diff --git a/index.html b/index.html
index 5227fd52..8d3d0416 100644
--- a/index.html
+++ b/index.html
@@ -5,9 +5,9 @@
   <title>Web Neural Network API</title>
   <meta content="w3c/ED" name="w3c-status">
   <link href="https://www.w3.org/StyleSheets/TR/2021/W3C-ED" rel="stylesheet">
-  <meta content="Bikeshed version d29f71adb, updated Wed Jul 19 11:08:56 2023 -0700" name="generator">
+  <meta content="Bikeshed version 6edc88947, updated Thu Aug 17 11:18:09 2023 -0700" name="generator">
   <link href="https://www.w3.org/TR/webnn/" rel="canonical">
-  <meta content="8e63e0709754b09ec4ed84e0ba7b2a8526137f0d" name="document-revision">
+  <meta content="34a4d8028e2cd2b7abe04009000c329e129795c7" name="document-revision">
 <style>
 /* Make <dl> blocks more distinct from their surroundings. */
 main dl:not(.switch) {
@@ -21,19 +21,19 @@
     margin: 1em 0;
 }
 
-/* Box for Valid Usage requirements. */
-div.validusage {
-    padding: .5em;
-    border: thin solid #88e !important;
-    border-radius: .5em;
-}
-
 /*
  * Stylistic labels, for clarity of presentation of these blocks.
  *
  * NOTE: This text is non-accessible and non-selectable; surrounding
  * text must also explain the context.
  */
+
+/* Box for Valid Usage requirements. */
+div.validusage {
+    padding: .5em;
+    border: thin solid #88e !important;
+    border-radius: .5em;
+}
 .validusage {
     position: relative;
 }
@@ -51,6 +51,90 @@
     content: "Valid Usage";
 }
 
+details {
+    padding: .5em;
+    border: thin solid #88e !important;
+    border-radius: .5em;
+}
+
+summary {
+    font-weight: bold;
+    margin: -0.5em -0.5em 0;
+    padding: 0.5em;
+}
+
+/* Box for algorithm steps. */
+
+div.algorithm-steps {
+    padding: .5em;
+    background-color: ghostwhite;
+}
+
+.algorithm-steps {
+    position: relative;
+    overflow: hidden;
+}
+.algorithm-steps::after {
+    font-weight: bold;
+    font-style: italic;
+    font-size: 130%;
+    color: rgba(0, 0, 0, 0.15);
+    color: var(--watermark-text);
+    position: absolute;
+    right: .3em;
+    bottom: .1em;
+}
+.algorithm-steps::after {
+    content: "Algorithm";
+}
+
+/* Informal steps */
+div.informalsteps {
+    padding: .5em;
+    border: thin solid #88e !important;
+    border-radius: .5em;
+    background-color: ghostwhite;
+}
+
+.informalsteps {
+    position: relative;
+}
+.informalsteps::after {
+    font-weight: bold;
+    font-style: italic;
+    font-size: 130%;
+    color: rgba(0, 0, 0, 0.15);
+    color: var(--watermark-text);
+    position: absolute;
+    right: .3em;
+    bottom: .1em;
+}
+.informalsteps::after {
+    content: "Non-normative";
+}
+
+/* Internal slots */
+div.internal-slots {
+    padding: .5em;
+    border: thin solid #88e !important;
+    border-radius: .5em;
+    background-color: aliceblue;
+}
+
+.internal-slots {
+    position: relative;
+}
+.internal-slots::after {
+    font-weight: bold;
+    font-style: italic;
+    font-size: 130%;
+    color: rgba(0, 0, 0, 0.15);
+    color: var(--watermark-text);
+    position: absolute;
+    right: .3em;
+    bottom: .1em;
+}
+
 /*
  * Ensure that argumentdef blocks don’t overflow algorithm section borders. This is made far harder
  * than it needs to be because the top-level W3C stylesheet has several @media + min-width variants
@@ -149,6 +233,76 @@
     }
 }
 
+
+/* Floating button for collapse/expand all details elements */
+
+.collapse-expand-button {
+  position: fixed;
+  bottom: 40px;
+  right: 40px;
+  width: 40px;
+  height: 40px;
+  border: none;
+  border-radius: 50%;
+  background-color: green;
+  color: ghostwhite;
+  font-size: 32px;
+  text-align: center;
+  align-items:center;
+  justify-content:center;
+  cursor: pointer;
+}
+
+.collapse-expand-button:hover {
+  background-color: green;
+}
+
+.collapse-expand-button.expand {
+  background-color: red;
+}
+
+.collapse-expand-button.expand::before {
+  content: "+";
+}
+
+.collapse-expand-button.collapse {
+  background-color: green;
+}
+
+.collapse-expand-button.collapse::before {
+  content: "-";
+}
+
+.collapse-expand-button .tooltiptext {
+  visibility: hidden;
+  bottom: 20px;
+  right: 20px;
+  width: 120px;
+  background-color: ghostwhite;
+  color: black;
+  font-size: 18px;
+  text-align: center;
+  align-items:center;
+  justify-content:center;
+  padding: 5px 0;
+  border-radius: 5px;
+
+  /* position */
+  position: absolute;
+  z-index: 1;
+  bottom: 100%;
+  left: 50%;
+  margin-left: -60px;
+      /* Use half of the width (120/2 = 60), to center the tooltip */
+}
+
+.collapse-expand-button:hover .tooltiptext {
+  visibility: visible;
+  opacity: 0.75;
+}
+
+/* end of floating collapse/expand button */
+
 </style>
 <style>/* Boilerplate: style-autolinks */
 .css.css, .property.property, .descriptor.descriptor {
@@ -811,7 +965,7 @@
   <div class="head">
    <p data-fill-with="logo"><a class="logo" href="https://www.w3.org/"> <img alt="W3C" height="48" src="https://www.w3.org/StyleSheets/TR/2021/logos/W3C" width="72"> </a> </p>
    <h1 class="p-name no-ref" id="title">Web Neural Network API</h1>
-   <p id="w3c-state"><a href="https://www.w3.org/standards/types#ED">Editor’s Draft</a>, <time class="dt-updated" datetime="2023-08-10">10 August 2023</time></p>
+   <p id="w3c-state"><a href="https://www.w3.org/standards/types#ED">Editor’s Draft</a>, <time class="dt-updated" datetime="2023-08-30">30 August 2023</time></p>
    <details open>
     <summary>More details about this document</summary>
     <div data-fill-with="spec-metadata">
@@ -945,7 +1099,11 @@ <h2 class="no-num no-toc no-ref" id="contents">Table of Contents</h2>
          <ol class="toc">
           <li><a href="#api-mloperand-create"><span class="secno">7.3.2.1</span> <span class="content">Creating <code class="idl"><span>MLOperand</span></code></span></a>
          </ol>
-        <li><a href="#api-mlactivation"><span class="secno">7.3.3</span> <span class="content">The MLActivation interface</span></a>
+        <li>
+         <a href="#api-mlactivation"><span class="secno">7.3.3</span> <span class="content">The MLActivation interface</span></a>
+         <ol class="toc">
+          <li><a href="#api-mlactivation-create"><span class="secno">7.3.3.1</span> <span class="content">Creating <code class="idl"><span>MLActivation</span></code></span></a>
+         </ol>
        </ol>
       <li>
        <a href="#api-mlcontext"><span class="secno">7.4</span> <span class="content">The MLContext interface</span></a>
@@ -977,46 +1135,112 @@ <h2 class="no-num no-toc no-ref" id="contents">Table of Contents</h2>
         <li><a href="#api-mlgraphbuilder-constructor"><span class="secno">7.6.1</span> <span class="content">The <code class="idl"><span>MLGraphBuilder</span></code> constructor</span></a>
         <li><a href="#api-mlgraphbuilder-input"><span class="secno">7.6.2</span> <span class="content">The <code class="idl"><span>input()</span></code> method</span></a>
         <li>
-         <a href="#api-mlgraphbuilder-constant-method"><span class="secno">7.6.3</span> <span class="content">The constant() method</span></a>
+         <a href="#api-mlgraphbuilder-build"><span class="secno">7.6.3</span> <span class="content">The build() method</span></a>
+         <ol class="toc">
+          <li><a href="#api-mlgraphbuilder-build-outputs"><span class="secno">7.6.3.1</span> <span class="content">The <code class="idl"><span>build(outputs)</span></code> method</span></a>
+          <li><a href="#api-mlgraphbuilder-buildsync-outputs"><span class="secno">7.6.3.2</span> <span class="content">The <code class="idl"><span>buildSync(outputs)</span></code> method</span></a>
+         </ol>
+        <li>
+         <a href="#api-mlgraphbuilder-constant-method"><span class="secno">7.6.4</span> <span class="content">The constant() method</span></a>
+         <ol class="toc">
+          <li><a href="#api-mlgraphbuilder-constant"><span class="secno">7.6.4.1</span> <span class="content">The <code class="idl"><span>constant(descriptor, bufferView)</span></code> method</span></a>
+          <li><a href="#api-mlgraphbuilder-constant-value-type"><span class="secno">7.6.4.2</span> <span class="content">The <code class="idl"><span>constant(value, type)</span></code> method</span></a>
+         </ol>
+        <li><a href="#api-mlgraphbuilder-batchnorm"><span class="secno">7.6.5</span> <span class="content">The batchNormalization() method</span></a>
+        <li>
+         <a href="#api-mlgraphbuilder-clamp"><span class="secno">7.6.6</span> <span class="content">The clamp() method</span></a>
+         <ol class="toc">
+          <li><a href="#api-mlgraphbuilder-clamp-operand-options"><span class="secno">7.6.6.1</span> <span class="content">The <code class="idl"><span>clamp(operand, options)</span></code> method</span></a>
+          <li><a href="#api-mlgraphbuilder-clamp-options"><span class="secno">7.6.6.2</span> <span class="content">The <code class="idl"><span>clamp(options)</span></code> method</span></a>
+         </ol>
+        <li><a href="#api-mlgraphbuilder-concat"><span class="secno">7.6.7</span> <span class="content">The concat() method</span></a>
+        <li><a href="#api-mlgraphbuilder-conv2d"><span class="secno">7.6.8</span> <span class="content">The conv2d() method</span></a>
+        <li><a href="#api-mlgraphbuilder-convtranspose2d"><span class="secno">7.6.9</span> <span class="content">The convTranspose2d() method</span></a>
+        <li><a href="#api-mlgraphbuilder-binary"><span class="secno">7.6.10</span> <span class="content">Element-wise binary operations</span></a>
+        <li><a href="#api-mlgraphbuilder-unary"><span class="secno">7.6.11</span> <span class="content">Element-wise unary operations</span></a>
+        <li>
+         <a href="#api-mlgraphbuilder-elu"><span class="secno">7.6.12</span> <span class="content">The elu() method</span></a>
+         <ol class="toc">
+          <li><a href="#api-mlgraphbuilder-elu-input-options"><span class="secno">7.6.12.1</span> <span class="content">The <code class="idl"><span>elu(input, options)</span></code> method</span></a>
+          <li><a href="#api-mlgraphbuilder-elu-options"><span class="secno">7.6.12.2</span> <span class="content">The <code class="idl"><span>elu(options)</span></code> method</span></a>
+         </ol>
+        <li><a href="#api-mlgraphbuilder-gemm"><span class="secno">7.6.13</span> <span class="content">The gemm() method</span></a>
+        <li><a href="#api-mlgraphbuilder-gru"><span class="secno">7.6.14</span> <span class="content">The gru() method</span></a>
+        <li><a href="#api-mlgraphbuilder-grucell"><span class="secno">7.6.15</span> <span class="content">The gruCell() method</span></a>
+        <li>
+         <a href="#api-mlgraphbuilder-hard-sigmoid"><span class="secno">7.6.16</span> <span class="content">The hardSigmoid() method</span></a>
+         <ol class="toc">
+          <li><a href="#api-mlgraphbuilder-hardsigmoid-input-options"><span class="secno">7.6.16.1</span> <span class="content">The <code class="idl"><span>hardSigmoid(input, options)</span></code> method</span></a>
+          <li><a href="#api-mlgraphbuilder-hardsigmoid-options"><span class="secno">7.6.16.2</span> <span class="content">The <code class="idl"><span>hardSigmoid(options)</span></code> method</span></a>
+         </ol>
+        <li>
+         <a href="#api-mlgraphbuilder-hard-swish"><span class="secno">7.6.17</span> <span class="content">The hardSwish() method</span></a>
+         <ol class="toc">
+          <li><a href="#api-mlgraphbuilder-hardswish-input"><span class="secno">7.6.17.1</span> <span class="content">The <code class="idl"><span>hardSwish(input)</span></code> method</span></a>
+          <li><a href="#api-mlgraphbuilder-hardswish"><span class="secno">7.6.17.2</span> <span class="content">The <code class="idl"><span>hardSwish()</span></code> method</span></a>
+         </ol>
+        <li><a href="#api-mlgraphbuilder-instancenorm"><span class="secno">7.6.18</span> <span class="content">The instanceNormalization() method</span></a>
+        <li>
+         <a href="#api-mlgraphbuilder-leakyrelu"><span class="secno">7.6.19</span> <span class="content">The leakyRelu() method</span></a>
+         <ol class="toc">
+          <li><a href="#api-mlgraphbuilder-leaky-relu-input-options"><span class="secno">7.6.19.1</span> <span class="content">The <code class="idl"><span>leakyRelu(input, options)</span></code> method</span></a>
+          <li><a href="#api-mlgraphbuilder-leaky-relu-options"><span class="secno">7.6.19.2</span> <span class="content">The <code class="idl"><span>leakyRelu(options)</span></code> method</span></a>
+         </ol>
+        <li>
+         <a href="#api-mlgraphbuilder-linear"><span class="secno">7.6.20</span> <span class="content">The linear() method</span></a>
+         <ol class="toc">
+          <li><a href="#api-mlgraphbuilder-linear-input-options"><span class="secno">7.6.20.1</span> <span class="content">The <code class="idl"><span>linear(input, options)</span></code> method</span></a>
+          <li><a href="#api-mlgraphbuilder-linear-options"><span class="secno">7.6.20.2</span> <span class="content">The <code class="idl"><span>linear(options)</span></code> method</span></a>
+         </ol>
+        <li><a href="#api-mlgraphbuilder-lstm"><span class="secno">7.6.21</span> <span class="content">The lstm() method</span></a>
+        <li><a href="#api-mlgraphbuilder-lstmcell"><span class="secno">7.6.22</span> <span class="content">The lstmCell() method</span></a>
+        <li><a href="#api-mlgraphbuilder-matmul"><span class="secno">7.6.23</span> <span class="content">The matmul() method</span></a>
+        <li><a href="#api-mlgraphbuilder-pad"><span class="secno">7.6.24</span> <span class="content">The pad() method</span></a>
+        <li><a href="#api-mlgraphbuilder-pool2d"><span class="secno">7.6.25</span> <span class="content">Pooling operations</span></a>
+        <li><a href="#api-mlgraphbuilder-prelu"><span class="secno">7.6.26</span> <span class="content">The prelu() method</span></a>
+        <li><a href="#api-mlgraphbuilder-reduce"><span class="secno">7.6.27</span> <span class="content">Reduction operations</span></a>
+        <li>
+         <a href="#api-mlgraphbuilder-relu-method"><span class="secno">7.6.28</span> <span class="content">The relu() method</span></a>
+         <ol class="toc">
+          <li><a href="#api-mlgraphbuilder-relu-input"><span class="secno">7.6.28.1</span> <span class="content">The <code class="idl"><span>relu(input)</span></code> method</span></a>
+          <li><a href="#api-mlgraphbuilder-relu"><span class="secno">7.6.28.2</span> <span class="content">The <code class="idl"><span>relu()</span></code> method</span></a>
+         </ol>
+        <li><a href="#api-mlgraphbuilder-resample2d-method"><span class="secno">7.6.29</span> <span class="content">The resample2d() method</span></a>
+        <li><a href="#api-mlgraphbuilder-reshape-method"><span class="secno">7.6.30</span> <span class="content">The reshape() method</span></a>
+        <li>
+         <a href="#api-mlgraphbuilder-sigmoid-method"><span class="secno">7.6.31</span> <span class="content">The sigmoid() method</span></a>
+         <ol class="toc">
+          <li><a href="#api-mlgraphbuilder-sigmoid-input"><span class="secno">7.6.31.1</span> <span class="content">The <code class="idl"><span>sigmoid(input)</span></code> method</span></a>
+          <li><a href="#api-mlgraphbuilder-sigmoid"><span class="secno">7.6.31.2</span> <span class="content">The <code class="idl"><span>sigmoid()</span></code> method</span></a>
+         </ol>
+        <li><a href="#api-mlgraphbuilder-slice"><span class="secno">7.6.32</span> <span class="content">The slice() method</span></a>
+        <li>
+         <a href="#api-mlgraphbuilder-softmax-method"><span class="secno">7.6.33</span> <span class="content">The softmax() method</span></a>
+         <ol class="toc">
+          <li><a href="#api-mlgraphbuilder-softmax-input"><span class="secno">7.6.33.1</span> <span class="content">The <code class="idl"><span>softmax(input)</span></code> method</span></a>
+          <li><a href="#api-mlgraphbuilder-softmax"><span class="secno">7.6.33.2</span> <span class="content">The <code class="idl"><span>softmax()</span></code> method</span></a>
+         </ol>
+        <li>
+         <a href="#api-mlgraphbuilder-softplus-method"><span class="secno">7.6.34</span> <span class="content">The softplus() method</span></a>
+         <ol class="toc">
+          <li><a href="#api-mlgraphbuilder-softplus-input-options"><span class="secno">7.6.34.1</span> <span class="content">The <code class="idl"><span>softplus(input, options)</span></code> method</span></a>
+          <li><a href="#api-mlgraphbuilder-softplus-options"><span class="secno">7.6.34.2</span> <span class="content">The <code class="idl"><span>softplus(options)</span></code> method</span></a>
+         </ol>
+        <li>
+         <a href="#api-mlgraphbuilder-softsign-method"><span class="secno">7.6.35</span> <span class="content">The softsign() method</span></a>
+         <ol class="toc">
+          <li><a href="#api-mlgraphbuilder-softsign-input"><span class="secno">7.6.35.1</span> <span class="content">The <code class="idl"><span>softsign(input)</span></code> method</span></a>
+          <li><a href="#api-mlgraphbuilder-softsign"><span class="secno">7.6.35.2</span> <span class="content">The <code class="idl"><span>softsign()</span></code> method</span></a>
+         </ol>
+        <li><a href="#api-mlgraphbuilder-split"><span class="secno">7.6.36</span> <span class="content">The split() method</span></a>
+        <li><a href="#api-mlgraphbuilder-squeeze"><span class="secno">7.6.37</span> <span class="content">The squeeze() method</span></a>
+        <li>
+         <a href="#api-mlgraphbuilder-tanh-method"><span class="secno">7.6.38</span> <span class="content">The tanh() method</span></a>
          <ol class="toc">
-          <li><a href="#api-mlgraphbuilder-constant"><span class="secno">7.6.3.1</span> <span class="content">The <code class="idl"><span>constant(descriptor, bufferView)</span></code> method</span></a>
-          <li><a href="#api-mlgraphbuilder-constant-value-type"><span class="secno">7.6.3.2</span> <span class="content">The <code class="idl"><span>constant(value, type)</span></code> method</span></a>
+          <li><a href="#api-mlgraphbuilder-tanh-input"><span class="secno">7.6.38.1</span> <span class="content">The <code class="idl"><span>tanh(input)</span></code> method</span></a>
+          <li><a href="#api-mlgraphbuilder-tanh"><span class="secno">7.6.38.2</span> <span class="content">The <code class="idl"><span>tanh()</span></code> method</span></a>
          </ol>
-        <li><a href="#api-mlgraphbuilder-batchnorm"><span class="secno">7.6.4</span> <span class="content">The batchNormalization() method</span></a>
-        <li><a href="#api-mlgraphbuilder-clamp"><span class="secno">7.6.5</span> <span class="content">The clamp() method</span></a>
-        <li><a href="#api-mlgraphbuilder-concat"><span class="secno">7.6.6</span> <span class="content">The concat() method</span></a>
-        <li><a href="#api-mlgraphbuilder-conv2d"><span class="secno">7.6.7</span> <span class="content">The conv2d() method</span></a>
-        <li><a href="#api-mlgraphbuilder-convtranspose2d"><span class="secno">7.6.8</span> <span class="content">The convTranspose2d() method</span></a>
-        <li><a href="#api-mlgraphbuilder-binary"><span class="secno">7.6.9</span> <span class="content">Element-wise binary operations</span></a>
-        <li><a href="#api-mlgraphbuilder-unary"><span class="secno">7.6.10</span> <span class="content">Element-wise unary operations</span></a>
-        <li><a href="#api-mlgraphbuilder-elu"><span class="secno">7.6.11</span> <span class="content">The elu() method</span></a>
-        <li><a href="#api-mlgraphbuilder-gemm"><span class="secno">7.6.12</span> <span class="content">The gemm() method</span></a>
-        <li><a href="#api-mlgraphbuilder-gru"><span class="secno">7.6.13</span> <span class="content">The gru() method</span></a>
-        <li><a href="#api-mlgraphbuilder-grucell"><span class="secno">7.6.14</span> <span class="content">The gruCell() method</span></a>
-        <li><a href="#api-mlgraphbuilder-hard-sigmoid"><span class="secno">7.6.15</span> <span class="content">The hardSigmoid() method</span></a>
-        <li><a href="#api-mlgraphbuilder-hard-swish"><span class="secno">7.6.16</span> <span class="content">The hardSwish() method</span></a>
-        <li><a href="#api-mlgraphbuilder-instancenorm"><span class="secno">7.6.17</span> <span class="content">The instanceNormalization() method</span></a>
-        <li><a href="#api-mlgraphbuilder-leakyrelu"><span class="secno">7.6.18</span> <span class="content">The leakyRelu() method</span></a>
-        <li><a href="#api-mlgraphbuilder-linear"><span class="secno">7.6.19</span> <span class="content">The linear() method</span></a>
-        <li><a href="#api-mlgraphbuilder-lstm"><span class="secno">7.6.20</span> <span class="content">The lstm() method</span></a>
-        <li><a href="#api-mlgraphbuilder-lstmcell"><span class="secno">7.6.21</span> <span class="content">The lstmCell() method</span></a>
-        <li><a href="#api-mlgraphbuilder-matmul"><span class="secno">7.6.22</span> <span class="content">The matmul() method</span></a>
-        <li><a href="#api-mlgraphbuilder-pad"><span class="secno">7.6.23</span> <span class="content">The pad() method</span></a>
-        <li><a href="#api-mlgraphbuilder-pool2d"><span class="secno">7.6.24</span> <span class="content">Pooling operations</span></a>
-        <li><a href="#api-mlgraphbuilder-prelu"><span class="secno">7.6.25</span> <span class="content">The prelu() method</span></a>
-        <li><a href="#api-mlgraphbuilder-reduce"><span class="secno">7.6.26</span> <span class="content">Reduction operations</span></a>
-        <li><a href="#api-mlgraphbuilder-relu"><span class="secno">7.6.27</span> <span class="content">The relu() method</span></a>
-        <li><a href="#api-mlgraphbuilder-resample2d"><span class="secno">7.6.28</span> <span class="content">The resample2d() method</span></a>
-        <li><a href="#api-mlgraphbuilder-reshape"><span class="secno">7.6.29</span> <span class="content">The reshape() method</span></a>
-        <li><a href="#api-mlgraphbuilder-sigmoid"><span class="secno">7.6.30</span> <span class="content">The sigmoid() method</span></a>
-        <li><a href="#api-mlgraphbuilder-slice"><span class="secno">7.6.31</span> <span class="content">The slice() method</span></a>
-        <li><a href="#api-mlgraphbuilder-softmax"><span class="secno">7.6.32</span> <span class="content">The softmax() method</span></a>
-        <li><a href="#api-mlgraphbuilder-softplus"><span class="secno">7.6.33</span> <span class="content">The softplus() method</span></a>
-        <li><a href="#api-mlgraphbuilder-softsign"><span class="secno">7.6.34</span> <span class="content">The softsign() method</span></a>
-        <li><a href="#api-mlgraphbuilder-split"><span class="secno">7.6.35</span> <span class="content">The split() method</span></a>
-        <li><a href="#api-mlgraphbuilder-squeeze"><span class="secno">7.6.36</span> <span class="content">The squeeze() method</span></a>
-        <li><a href="#api-mlgraphbuilder-tanh"><span class="secno">7.6.37</span> <span class="content">The tanh() method</span></a>
-        <li><a href="#api-mlgraphbuilder-transpose"><span class="secno">7.6.38</span> <span class="content">The transpose() method</span></a>
+        <li><a href="#api-mlgraphbuilder-transpose"><span class="secno">7.6.39</span> <span class="content">The transpose() method</span></a>
        </ol>
      </ol>
     <li><a href="#examples"><span class="secno">8</span> <span class="content">Examples</span></a>
@@ -1049,6 +1273,40 @@ <h2 class="no-num no-toc no-ref" id="contents">Table of Contents</h2>
    </ol>
   </nav>
   <main>
+   <button class="collapse-expand-button" onclick="toggleCE()"> <span class="tooltiptext">Collapse all</span> </button>
+<script>
+    var ceButton = document.querySelector(".collapse-expand-button");
+    ceButton.classList.add("collapse");  // All details are expanded by default.
+    var scrollY = window.scrollY;
+    window.addEventListener('scroll', function() {
+        scrollY = window.scrollY;
+        ceButton.style.top = scrollY + window.innerHeight - 60 + 'px';
+    });
+
+    function toggleCE() {
+        var button = document.querySelector(".collapse-expand-button");
+        var tip = document.querySelector(".tooltiptext");
+        var allDetails = document.querySelectorAll('details');
+
+        Array.from(allDetails).forEach(function(detail, index) {
+            if (button.classList.contains("expand")) {
+                detail.open = true;
+            } else {
+                detail.removeAttribute('open');
+            }
+        });
+
+        if (button.classList.contains("expand")) {
+            button.classList.remove("expand");
+            button.classList.add("collapse");
+            tip.innerHTML = "Collapse all";
+        } else {
+            button.classList.remove("collapse");
+            button.classList.add("expand");
+            tip.innerHTML = "Expand all";
+        }
+    }
+</script>
    <h2 class="heading settled" data-level="1" id="intro"><span class="secno">1. </span><span class="content">Introduction</span><a class="self-link" href="#intro"></a></h2>
    <p>The Web Neural Network API defines a web-friendly hardware-agnostic abstraction layer that makes use of Machine Learning capabilities of operating systems and underlying hardware platforms without being tied to platform-specific capabilities. The abstraction layer addresses the requirements of key Machine Learning JavaScript frameworks and also allows web developers familiar with the ML domain to write custom code without the help of libraries. A complementary <a href="https://webmachinelearning.github.io/model-loader/">Model Loader API</a> defines a higher-level abstraction targeting primarily web developers.</p>
    <p>For an illustrated introduction, please see the <a href="https://github.com/webmachinelearning/webnn/blob/master/explainer.md">explainer</a>.</p>
@@ -1177,7 +1435,7 @@ <h4 class="heading settled" data-level="2.2.5" id="usecase-real-time-video-proce
    <h2 class="heading settled" data-level="3" id="security"><span class="secno">3. </span><span class="content">Security Considerations</span><a class="self-link" href="#security"></a></h2>
     This specification defines a low-level API for neural network inference hardware acceleration. This API is considered a powerful feature <a data-link-type="biblio" href="#biblio-powerful-features" title="Secure Contexts">[POWERFUL-FEATURES]</a> because it grants low-level access to a user’s computer. To meet the authentication and confidentiality expectations of a powerful feature and to prevent man-in-the-middle attacks, all interfaces defined by this specification are only available in a secure context. 
    <p>This API is disabled by default in all cross-origin frames using the <a href="#permissions-policy-integration">§ 7.2.1 Permissions Policy Integration</a>. This prevents third-party content from using this API unless the embedding page explicitly sets a policy that grants permission.</p>
-   <p>This API allows creation of an <code class="idl"><a data-link-type="idl" href="#mlcontext" id="ref-for-mlcontext">MLContext</a></code> from a <code class="idl"><a data-link-type="idl" href="https://gpuweb.github.io/gpuweb/#gpu-device" id="ref-for-gpu-device">GPUDevice</a></code> defined by WebGPU specification. See <a href="https://gpuweb.github.io/gpuweb/#security-considerations">WebGPU Security Considerations</a> for more information regarding security characteristics of this context.</p>
+   <p>This API allows creation of an <code class="idl"><a data-link-type="idl" href="#mlcontext" id="ref-for-mlcontext">MLContext</a></code> from a <code class="idl"><a data-link-type="idl" href="https://gpuweb.github.io/gpuweb/#gpudevice" id="ref-for-gpudevice">GPUDevice</a></code> defined by WebGPU specification. See <a href="https://gpuweb.github.io/gpuweb/#security-considerations">WebGPU Security Considerations</a> for more information regarding security characteristics of this context.</p>
    <p>Once the graph is fully constructed and compiled, the input shapes into each of the operations in the graph are inferred and finalized. The bounds checking occurs when the compute method is invoked that executes the graph against the actual data. No actual data is bound to the compiled graph before this stage. It is the implementation’s responsibility to make sure proper bounds checking occurs against the shapes of the data already inferred by that time.</p>
    <p class="issue" id="issue-9e2aaedc"><a class="self-link" href="#issue-9e2aaedc"></a> Document operations susceptible to out-of-bounds access as a guidance to implementers.</p>
    <p>As a future-proofing measure, the API design allows certain operations that can be generically emulated to be deprecated for security, performance, or other reasons without breaking compatibility. This is made possible by high-level functions that are defined in terms of smaller primitive operations defined in this specifications. This enables a native implementation of a high-level function to be replaced with a polyfill implementation.</p>
@@ -1220,8 +1478,8 @@ <h2 class="heading settled" data-level="4" id="privacy"><span class="secno">4. <
    <p>This API exposes the minimum amount of information necessary to address the identified <a href="#usecases">§ 2 Use cases</a> for the best performance and reliability of results.</p>
    <p>No information from the underlying platform is exposed directly. An execution time analysis may reveal indirectly the performance of the underlying platform’s neural network hardware acceleration capabilities relative to another underlying platform.</p>
    <p class="note" role="note"><span class="marker">Note:</span> The group is <a href="https://github.com/webmachinelearning/webnn/issues/85">soliciting further input</a> on the proposed execution time analysis fingerprinting vector and will augment this section with more information and mitigations to inform the implementers of this API.</p>
-   <p>Unlike WebGPU, this API does not intrinsically support custom shader authoring; and as a result is not prone to timing attacks that rely on shader caches, or other persistent data. The API builds upon pre-existing shaders and lower level primitives of the browser or the underlying OS. Web developers who interface with <code class="idl"><a data-link-type="idl" href="https://gpuweb.github.io/gpuweb/#gpu-device" id="ref-for-gpu-device①">GPUDevice</a></code> are expected to be aware of <a href="https://gpuweb.github.io/gpuweb/#privacy-user-agent-state">WebGPU compilation cache considerations</a>.</p>
-   <p>The WebGPU API identifies <a href="https://gpuweb.github.io/gpuweb/#privacy-machine-artifacts">machine-specific artifacts</a> as a privacy consideration. Given the WebNN API defines means to record an ML workload onto a WebGPU-compatible <code class="idl"><a data-link-type="idl" href="https://gpuweb.github.io/gpuweb/#command-buffers" id="ref-for-command-buffers">GPUCommandBuffer</a></code>, compute unit scheduling may under certain circumstances introduce a fingerprint. However, similarly to WebGPU, such fingerprints are identical across most or all of the devices of each vendor, mitigating the concern. Furthermore, software implementations can be used to further eliminate such artifacts.</p>
+   <p>Unlike WebGPU, this API does not intrinsically support custom shader authoring; and as a result is not prone to timing attacks that rely on shader caches, or other persistent data. The API builds upon pre-existing shaders and lower level primitives of the browser or the underlying OS. Web developers who interface with <code class="idl"><a data-link-type="idl" href="https://gpuweb.github.io/gpuweb/#gpudevice" id="ref-for-gpudevice①">GPUDevice</a></code> are expected to be aware of <a href="https://gpuweb.github.io/gpuweb/#privacy-user-agent-state">WebGPU compilation cache considerations</a>.</p>
+   <p>The WebGPU API identifies <a href="https://gpuweb.github.io/gpuweb/#privacy-machine-artifacts">machine-specific artifacts</a> as a privacy consideration. Given the WebNN API defines means to record an ML workload onto a WebGPU-compatible <code class="idl"><a data-link-type="idl" href="https://gpuweb.github.io/gpuweb/#gpucommandbuffer" id="ref-for-gpucommandbuffer">GPUCommandBuffer</a></code>, compute unit scheduling may under certain circumstances introduce a fingerprint. However, similarly to WebGPU, such fingerprints are identical across most or all of the devices of each vendor, mitigating the concern. Furthermore, software implementations can be used to further eliminate such artifacts.</p>
    <p>The WebNN API defines two developer-settable preferences to help inform <a href="#programming-model-device-selection">§ 6.2 Device Selection</a> and allow the implementation to better select the most appropriate underlying execution device for the workload. <a data-link-type="dfn" href="#device-type" id="ref-for-device-type">Device type</a> normatively indicates the kind of device and is either "cpu" or "gpu". If this type cannot be satisfied, an "<code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#operationerror" id="ref-for-operationerror">OperationError</a></code>" <code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#idl-DOMException" id="ref-for-idl-DOMException">DOMException</a></code> is thrown, thus this type can in some cases add two bits of entropy to the fingerprint. <a data-link-type="dfn" href="#power-preference" id="ref-for-power-preference">Power preference</a> indicates preference as related to the power consumption and is considered a hint only and as such does not increase entropy of the fingerprint.</p>
    <p>If a future version of this specification introduces support for new a <a data-link-type="dfn" href="#device-type" id="ref-for-device-type①">device type</a> that can only support a subset of <code class="idl"><a data-link-type="idl" href="#enumdef-mloperandtype" id="ref-for-enumdef-mloperandtype">MLOperandType</a></code>s, that may introduce a new fingerprint.</p>
    <p>In general, implementers of this API are expected to apply <a href="https://gpuweb.github.io/gpuweb/#privacy-considerations">WebGPU Privacy Considerations</a> to their implementations where applicable.</p>
@@ -1242,7 +1500,7 @@ <h3 class="heading settled" data-level="6.1" id="programming-model-overview"><sp
 during inference, as well as the output values of inference.
 At inference time, every <code class="idl"><a data-link-type="idl" href="#mloperand" id="ref-for-mloperand①">MLOperand</a></code> will be bound to a tensor (the actual data).</p>
    <p>The <code class="idl"><a data-link-type="idl" href="#mlgraphbuilder" id="ref-for-mlgraphbuilder②">MLGraphBuilder</a></code> interface enables the creation of <code class="idl"><a data-link-type="idl" href="#mloperand" id="ref-for-mloperand②">MLOperand</a></code>s.
-A key part of the <code class="idl"><a data-link-type="idl" href="#mlgraphbuilder" id="ref-for-mlgraphbuilder③">MLGraphBuilder</a></code> interface are the operations (such as <code class="idl"><a data-link-type="idl" href="#mlgraphbuilder" id="ref-for-mlgraphbuilder④">MLGraphBuilder</a></code>.<code class="idl"><a data-link-type="idl" href="#dom-mlgraphbuilder-gemm" id="ref-for-dom-mlgraphbuilder-gemm">gemm()</a></code> and <code class="idl"><a data-link-type="idl" href="#mlgraphbuilder" id="ref-for-mlgraphbuilder⑤">MLGraphBuilder</a></code>.<code class="idl"><a data-link-type="idl" href="#dom-mlgraphbuilder-softmax①" id="ref-for-dom-mlgraphbuilder-softmax①">softmax()</a></code>). The operations have a functional
+A key part of the <code class="idl"><a data-link-type="idl" href="#mlgraphbuilder" id="ref-for-mlgraphbuilder③">MLGraphBuilder</a></code> interface are the operations (such as <code class="idl"><a data-link-type="idl" href="#mlgraphbuilder" id="ref-for-mlgraphbuilder④">MLGraphBuilder</a></code>.<code class="idl"><a data-link-type="idl" href="#dom-mlgraphbuilder-gemm" id="ref-for-dom-mlgraphbuilder-gemm">gemm()</a></code> and <code class="idl"><a data-link-type="idl" href="#mlgraphbuilder" id="ref-for-mlgraphbuilder⑤">MLGraphBuilder</a></code>.<code class="idl"><a data-link-type="idl" href="#softmax-noargs" id="ref-for-softmax-noargs">softmax()</a></code>). The operations have a functional
 semantics, with no side effects.
 Each operation invocation conceptually returns a distinct new value, without
 changing the value of any other <code class="idl"><a data-link-type="idl" href="#mloperand" id="ref-for-mloperand③">MLOperand</a></code>.</p>
@@ -1274,11 +1532,11 @@ <h3 class="heading settled" data-level="6.1" id="programming-model-overview"><sp
    <p>The <code class="idl"><a data-link-type="idl" href="#mlcommandencoder" id="ref-for-mlcommandencoder①">MLCommandEncoder</a></code> interface created by the <code class="idl"><a data-link-type="idl" href="#mlcontext" id="ref-for-mlcontext⑦">MLContext</a></code>.<code class="idl"><a data-link-type="idl" href="#dom-mlcontext-createcommandencoder" id="ref-for-dom-mlcontext-createcommandencoder">createCommandEncoder()</a></code> method supports
 a graph execution method that provides the maximum flexibility to callers that also utilize WebGPU in their
 application. It does this by placing the workload required to initialize and compute the results of the
-operations in the graph onto a <code class="idl"><a data-link-type="idl" href="https://gpuweb.github.io/gpuweb/#command-buffers" id="ref-for-command-buffers①">GPUCommandBuffer</a></code>. The callers are responsible for the eventual submission
-of this workload on the <code class="idl"><a data-link-type="idl" href="https://gpuweb.github.io/gpuweb/#queues" id="ref-for-queues">GPUQueue</a></code> through the WebGPU queue submission mechanism. Once the submitted workload
+operations in the graph onto a <code class="idl"><a data-link-type="idl" href="https://gpuweb.github.io/gpuweb/#gpucommandbuffer" id="ref-for-gpucommandbuffer①">GPUCommandBuffer</a></code>. The callers are responsible for the eventual submission
+of this workload on the <code class="idl"><a data-link-type="idl" href="https://gpuweb.github.io/gpuweb/#gpuqueue" id="ref-for-gpuqueue">GPUQueue</a></code> through the WebGPU queue submission mechanism. Once the submitted workload
 is completely executed, the result is avaialble in the bound output buffers.</p>
    <h3 class="heading settled" data-level="6.2" id="programming-model-device-selection"><span class="secno">6.2. </span><span class="content">Device Selection</span><a class="self-link" href="#programming-model-device-selection"></a></h3>
-   <p>An <code class="idl"><a data-link-type="idl" href="#mlcontext" id="ref-for-mlcontext⑧">MLContext</a></code> interface represents a global state of neural network execution. One of the important context states is the underlying execution device that manages the resources and facilitates the compilation and the eventual execution of the neural network graph. In addition to the default method of creation with <code class="idl"><a data-link-type="idl" href="#dictdef-mlcontextoptions" id="ref-for-dictdef-mlcontextoptions">MLContextOptions</a></code>, an <code class="idl"><a data-link-type="idl" href="#mlcontext" id="ref-for-mlcontext⑨">MLContext</a></code> could also be created from a specific <code class="idl"><a data-link-type="idl" href="https://gpuweb.github.io/gpuweb/#gpu-device" id="ref-for-gpu-device②">GPUDevice</a></code> that is already in use by the application, in which case the corresponding <code class="idl"><a data-link-type="idl" href="https://gpuweb.github.io/gpuweb/#buffer-interface" id="ref-for-buffer-interface">GPUBuffer</a></code> resources used as graph constants, as well as the <code class="idl"><a data-link-type="idl" href="https://gpuweb.github.io/gpuweb/#texture-interface" id="ref-for-texture-interface">GPUTexture</a></code> as graph inputs must also be created from the same device. In a multi-adapter configuration, the device used for <code class="idl"><a data-link-type="idl" href="#mlcontext" id="ref-for-mlcontext①⓪">MLContext</a></code> must be created from the same adapter as the device used to allocate the resources referenced in the graph.</p>
+   <p>An <code class="idl"><a data-link-type="idl" href="#mlcontext" id="ref-for-mlcontext⑧">MLContext</a></code> interface represents a global state of neural network execution. One of the important context states is the underlying execution device that manages the resources and facilitates the compilation and the eventual execution of the neural network graph. In addition to the default method of creation with <code class="idl"><a data-link-type="idl" href="#dictdef-mlcontextoptions" id="ref-for-dictdef-mlcontextoptions">MLContextOptions</a></code>, an <code class="idl"><a data-link-type="idl" href="#mlcontext" id="ref-for-mlcontext⑨">MLContext</a></code> could also be created from a specific <code class="idl"><a data-link-type="idl" href="https://gpuweb.github.io/gpuweb/#gpudevice" id="ref-for-gpudevice②">GPUDevice</a></code> that is already in use by the application, in which case the corresponding <code class="idl"><a data-link-type="idl" href="https://gpuweb.github.io/gpuweb/#gpubuffer" id="ref-for-gpubuffer">GPUBuffer</a></code> resources used as graph constants, as well as the <code class="idl"><a data-link-type="idl" href="https://gpuweb.github.io/gpuweb/#gputexture" id="ref-for-gputexture">GPUTexture</a></code> as graph inputs must also be created from the same device. In a multi-adapter configuration, the device used for <code class="idl"><a data-link-type="idl" href="#mlcontext" id="ref-for-mlcontext①⓪">MLContext</a></code> must be created from the same adapter as the device used to allocate the resources referenced in the graph.</p>
    <p>In a situation when a GPU context executes a graph with a constant or an input in the system memory as an <code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#ArrayBufferView" id="ref-for-ArrayBufferView">ArrayBufferView</a></code>, the input content is automatically uploaded from the system memory to the GPU memory, and downloaded back to the system memory of an <code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#ArrayBufferView" id="ref-for-ArrayBufferView①">ArrayBufferView</a></code> output buffer at the end of the graph execution. This data upload and download cycles will only occur whenever the execution device requires the data to be copied out of and back into the system memory, such as in the case of the GPU. It doesn’t occur when the device is a CPU device. Additionally, the result of the graph execution is in a known layout format. While the execution may be optimized for a native memory access pattern in an intermediate result within the graph, the output of the last operation of the graph must convert the content back to a known layout format at the end of the graph in order to maintain the expected behavior from the caller’s perspective.</p>
    <p>When an <code class="idl"><a data-link-type="idl" href="#mlcontext" id="ref-for-mlcontext①①">MLContext</a></code> is created with <code class="idl"><a data-link-type="idl" href="#dictdef-mlcontextoptions" id="ref-for-dictdef-mlcontextoptions①">MLContextOptions</a></code>, the user agent selects and creates the underlying execution device by taking into account the application’s <a data-link-type="dfn" href="#power-preference" id="ref-for-power-preference①">power preference</a> and <a data-link-type="dfn" href="#device-type" id="ref-for-device-type②">device type</a> specified in the <code class="idl"><a data-link-type="idl" href="#enumdef-mlpowerpreference" id="ref-for-enumdef-mlpowerpreference">MLPowerPreference</a></code> and <code class="idl"><a data-link-type="idl" href="#enumdef-mldevicetype" id="ref-for-enumdef-mldevicetype">MLDeviceType</a></code> options.</p>
    <p>The following table summarizes the types of resource supported by the context created through different method of creation:</p>
@@ -1330,37 +1588,28 @@ <h3 class="heading settled" data-level="7.2" id="api-ml"><span class="secno">7.2
 
 [<a class="idl-code" data-link-type="extended-attribute" href="https://webidl.spec.whatwg.org/#SecureContext" id="ref-for-SecureContext①"><c- g>SecureContext</c-></a>, <a class="idl-code" data-link-type="extended-attribute" href="https://webidl.spec.whatwg.org/#Exposed" id="ref-for-Exposed"><c- g>Exposed</c-></a>=(<c- n>Window</c->, <c- n>DedicatedWorker</c->)]
 <c- b>interface</c-> <dfn class="dfn-paneled idl-code" data-dfn-type="interface" data-export id="ml"><code><c- g>ML</c-></code></dfn> {
-  <a class="idl-code" data-link-type="interface" href="https://webidl.spec.whatwg.org/#idl-promise" id="ref-for-idl-promise①"><c- b>Promise</c-></a>&lt;<a data-link-type="idl-name" href="#mlcontext" id="ref-for-mlcontext①②"><c- n>MLContext</c-></a>> <dfn class="dfn-paneled idl-code" data-dfn-for="ML" data-dfn-type="method" data-export data-lt="createContext(options)|createContext()" id="dom-ml-createcontext"><code><c- g>createContext</c-></code></dfn>(<c- b>optional</c-> <a data-link-type="idl-name" href="#dictdef-mlcontextoptions" id="ref-for-dictdef-mlcontextoptions②"><c- n>MLContextOptions</c-></a> <dfn class="idl-code" data-dfn-for="ML/createContext(options), ML/createContext()" data-dfn-type="argument" data-export id="dom-ml-createcontext-options-options"><code><c- g>options</c-></code><a class="self-link" href="#dom-ml-createcontext-options-options"></a></dfn> = {});
-  <a class="idl-code" data-link-type="interface" href="https://webidl.spec.whatwg.org/#idl-promise" id="ref-for-idl-promise②"><c- b>Promise</c-></a>&lt;<a data-link-type="idl-name" href="#mlcontext" id="ref-for-mlcontext①③"><c- n>MLContext</c-></a>> <dfn class="idl-code" data-dfn-for="ML" data-dfn-type="method" data-export data-lt="createContext(gpuDevice)" id="dom-ml-createcontext-gpudevice"><code><c- g>createContext</c-></code><a class="self-link" href="#dom-ml-createcontext-gpudevice"></a></dfn>(<a data-link-type="idl-name" href="https://gpuweb.github.io/gpuweb/#gpu-device" id="ref-for-gpu-device③"><c- n>GPUDevice</c-></a> <dfn class="idl-code" data-dfn-for="ML/createContext(gpuDevice)" data-dfn-type="argument" data-export id="dom-ml-createcontext-gpudevice-gpudevice"><code><c- g>gpuDevice</c-></code><a class="self-link" href="#dom-ml-createcontext-gpudevice-gpudevice"></a></dfn>);
+  <a class="idl-code" data-link-type="interface" href="https://webidl.spec.whatwg.org/#idl-promise" id="ref-for-idl-promise①"><c- b>Promise</c-></a>&lt;<a data-link-type="idl-name" href="#mlcontext" id="ref-for-mlcontext①②"><c- n>MLContext</c-></a>> <a class="idl-code" data-link-type="method" href="#dom-ml-createcontext" id="ref-for-dom-ml-createcontext"><c- g>createContext</c-></a>(<c- b>optional</c-> <a data-link-type="idl-name" href="#dictdef-mlcontextoptions" id="ref-for-dictdef-mlcontextoptions②"><c- n>MLContextOptions</c-></a> <dfn class="idl-code" data-dfn-for="ML/createContext(options), ML/createContext()" data-dfn-type="argument" data-export id="dom-ml-createcontext-options-options"><code><c- g>options</c-></code><a class="self-link" href="#dom-ml-createcontext-options-options"></a></dfn> = {});
+  <a class="idl-code" data-link-type="interface" href="https://webidl.spec.whatwg.org/#idl-promise" id="ref-for-idl-promise②"><c- b>Promise</c-></a>&lt;<a data-link-type="idl-name" href="#mlcontext" id="ref-for-mlcontext①③"><c- n>MLContext</c-></a>> <a class="idl-code" data-link-type="method" href="#dom-ml-createcontext-gpudevice" id="ref-for-dom-ml-createcontext-gpudevice"><c- g>createContext</c-></a>(<a data-link-type="idl-name" href="https://gpuweb.github.io/gpuweb/#gpudevice" id="ref-for-gpudevice③"><c- n>GPUDevice</c-></a> <dfn class="idl-code" data-dfn-for="ML/createContext(gpuDevice)" data-dfn-type="argument" data-export id="dom-ml-createcontext-gpudevice-gpudevice"><code><c- g>gpuDevice</c-></code><a class="self-link" href="#dom-ml-createcontext-gpudevice-gpudevice"></a></dfn>);
 
   [<a class="idl-code" data-link-type="extended-attribute" href="https://webidl.spec.whatwg.org/#Exposed" id="ref-for-Exposed①"><c- g>Exposed</c-></a>=(<c- n>DedicatedWorker</c->)]
-  <a data-link-type="idl-name" href="#mlcontext" id="ref-for-mlcontext①④"><c- n>MLContext</c-></a> <dfn class="dfn-paneled idl-code" data-dfn-for="ML" data-dfn-type="method" data-export data-lt="createContextSync(options)|createContextSync()" id="dom-ml-createcontextsync"><code><c- g>createContextSync</c-></code></dfn>(<c- b>optional</c-> <a data-link-type="idl-name" href="#dictdef-mlcontextoptions" id="ref-for-dictdef-mlcontextoptions③"><c- n>MLContextOptions</c-></a> <dfn class="idl-code" data-dfn-for="ML/createContextSync(options), ML/createContextSync()" data-dfn-type="argument" data-export id="dom-ml-createcontextsync-options-options"><code><c- g>options</c-></code><a class="self-link" href="#dom-ml-createcontextsync-options-options"></a></dfn> = {});
+  <a data-link-type="idl-name" href="#mlcontext" id="ref-for-mlcontext①④"><c- n>MLContext</c-></a> <a class="idl-code" data-link-type="method" href="#dom-ml-createcontextsync" id="ref-for-dom-ml-createcontextsync"><c- g>createContextSync</c-></a>(<c- b>optional</c-> <a data-link-type="idl-name" href="#dictdef-mlcontextoptions" id="ref-for-dictdef-mlcontextoptions③"><c- n>MLContextOptions</c-></a> <dfn class="idl-code" data-dfn-for="ML/createContextSync(options), ML/createContextSync()" data-dfn-type="argument" data-export id="dom-ml-createcontextsync-options-options"><code><c- g>options</c-></code><a class="self-link" href="#dom-ml-createcontextsync-options-options"></a></dfn> = {});
   [<a class="idl-code" data-link-type="extended-attribute" href="https://webidl.spec.whatwg.org/#Exposed" id="ref-for-Exposed②"><c- g>Exposed</c-></a>=(<c- n>DedicatedWorker</c->)]
-  <a data-link-type="idl-name" href="#mlcontext" id="ref-for-mlcontext①⑤"><c- n>MLContext</c-></a> <dfn class="idl-code" data-dfn-for="ML" data-dfn-type="method" data-export data-lt="createContextSync(gpuDevice)" id="dom-ml-createcontextsync-gpudevice"><code><c- g>createContextSync</c-></code><a class="self-link" href="#dom-ml-createcontextsync-gpudevice"></a></dfn>(<a data-link-type="idl-name" href="https://gpuweb.github.io/gpuweb/#gpu-device" id="ref-for-gpu-device④"><c- n>GPUDevice</c-></a> <dfn class="idl-code" data-dfn-for="ML/createContextSync(gpuDevice)" data-dfn-type="argument" data-export id="dom-ml-createcontextsync-gpudevice-gpudevice"><code><c- g>gpuDevice</c-></code><a class="self-link" href="#dom-ml-createcontextsync-gpudevice-gpudevice"></a></dfn>);
+  <a data-link-type="idl-name" href="#mlcontext" id="ref-for-mlcontext①⑤"><c- n>MLContext</c-></a> <a class="idl-code" data-link-type="method" href="#dom-ml-createcontextsync-gpudevice" id="ref-for-dom-ml-createcontextsync-gpudevice"><c- g>createContextSync</c-></a>(<a data-link-type="idl-name" href="https://gpuweb.github.io/gpuweb/#gpudevice" id="ref-for-gpudevice④"><c- n>GPUDevice</c-></a> <dfn class="idl-code" data-dfn-for="ML/createContextSync(gpuDevice)" data-dfn-type="argument" data-export id="dom-ml-createcontextsync-gpudevice-gpudevice"><code><c- g>gpuDevice</c-></code><a class="self-link" href="#dom-ml-createcontextsync-gpudevice-gpudevice"></a></dfn>);
 };
 </pre>
    <h4 class="heading settled" data-level="7.2.1" id="permissions-policy-integration"><span class="secno">7.2.1. </span><span class="content">Permissions Policy Integration</span><a class="self-link" href="#permissions-policy-integration"></a></h4>
    <p>This specification defines a <a data-link-type="dfn" href="https://w3c.github.io/webappsec-permissions-policy/#policy-controlled-feature" id="ref-for-policy-controlled-feature">policy-controlled feature</a> identified by the
 string "<code><dfn class="dfn-paneled" data-dfn-type="dfn" data-lt="webnn-feature" data-noexport id="webnn-feature">webnn</dfn></code>".
 Its <a data-link-type="dfn" href="https://w3c.github.io/webappsec-permissions-policy/#policy-controlled-feature-default-allowlist" id="ref-for-policy-controlled-feature-default-allowlist">default allowlist</a> is <code>'self'</code>.</p>
-   <h4 class="heading settled" data-level="7.2.2" id="api-ml-createcontext"><span class="secno">7.2.2. </span><span class="content">The <code class="idl"><a data-link-type="idl" href="#dom-ml-createcontext" id="ref-for-dom-ml-createcontext">createContext()</a></code> method</span><a class="self-link" href="#api-ml-createcontext"></a></h4>
-    The <code class="idl"><a data-link-type="idl" href="#dom-ml-createcontext" id="ref-for-dom-ml-createcontext①">createContext()</a></code> method steps are: 
-   <ol>
-    <li data-md>
-     <p>If <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#this" id="ref-for-this">this</a>'s <a data-link-type="dfn" href="https://html.spec.whatwg.org/multipage/webappapis.html#concept-relevant-global" id="ref-for-concept-relevant-global">relevant global object</a>'s <a data-link-type="dfn" href="https://html.spec.whatwg.org/multipage/nav-history-apis.html#concept-document-window" id="ref-for-concept-document-window">associated Document</a> is not <a data-link-type="dfn" href="https://html.spec.whatwg.org/multipage/iframe-embed-object.html#allowed-to-use" id="ref-for-allowed-to-use">allowed to use</a> the <a data-link-type="dfn" href="#webnn-feature" id="ref-for-webnn-feature">webnn</a> feature, return <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#a-new-promise" id="ref-for-a-new-promise">a new promise</a> <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#reject" id="ref-for-reject">rejected</a> with a "<code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#securityerror" id="ref-for-securityerror">SecurityError</a></code>" <code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#idl-DOMException" id="ref-for-idl-DOMException①">DOMException</a></code> and abort these steps.</p>
-    <li data-md>
-     <p>Let <var>promise</var> be <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#a-new-promise" id="ref-for-a-new-promise①">a new promise</a>.</p>
-    <li data-md>
-     <p>Return <var>promise</var> and run the following steps <a data-link-type="dfn" href="https://html.spec.whatwg.org/multipage/infrastructure.html#in-parallel" id="ref-for-in-parallel">in parallel</a>.</p>
-    <li data-md>
-     <p>Let <var>options</var> be the first argument.</p>
-    <li data-md>
-     <p>Run the <dfn class="dfn-paneled" data-dfn-type="dfn" data-noexport id="create-context">create context</dfn> steps given <var>options</var>:</p>
+   <h4 class="heading settled" data-level="7.2.2" id="api-ml-createcontext"><span class="secno">7.2.2. </span><span class="content">The <code class="idl"><a data-link-type="idl" href="#dom-ml-createcontext" id="ref-for-dom-ml-createcontext①">createContext()</a></code> method</span><a class="self-link" href="#api-ml-createcontext"></a></h4>
+   <details class="algorithm" data-algorithm="create a context" open>
+    <summary> To <dfn class="dfn-paneled" data-dfn-type="dfn" data-noexport id="create-a-context">create a context</dfn> given <var>options</var>, run these steps: </summary>
+    <div class="algorithm-steps">
      <ol>
       <li data-md>
        <p>Let <var>context</var> be a new <code class="idl"><a data-link-type="idl" href="#mlcontext" id="ref-for-mlcontext①⑥">MLContext</a></code> object.</p>
       <li data-md>
-       <p>If <var>options</var> is a <code class="idl"><a data-link-type="idl" href="https://gpuweb.github.io/gpuweb/#gpu-device" id="ref-for-gpu-device⑤">GPUDevice</a></code> object,</p>
+       <p>If <var>options</var> is a <code class="idl"><a data-link-type="idl" href="https://gpuweb.github.io/gpuweb/#gpudevice" id="ref-for-gpudevice⑤">GPUDevice</a></code> object,</p>
        <ol>
         <li data-md>
          <p>Set <var>context</var>.<code class="idl"><a data-link-type="idl" href="#dom-mlcontext-contexttype-slot" id="ref-for-dom-mlcontext-contexttype-slot">[[contextType]]</a></code> to "<a data-link-type="dfn" href="#webgpu-context" id="ref-for-webgpu-context">webgpu</a>".</p>
@@ -1379,46 +1628,102 @@ <h4 class="heading settled" data-level="7.2.2" id="api-ml-createcontext"><span c
         <li data-md>
          <p>If <var>options</var>["<code class="idl"><a data-link-type="idl" href="#dom-mlcontextoptions-powerpreference" id="ref-for-dom-mlcontextoptions-powerpreference">powerPreference</a></code>"] <a data-link-type="dfn" href="https://infra.spec.whatwg.org/#map-exists" id="ref-for-map-exists①">exists</a>, then set <var>context</var>.<code class="idl"><a data-link-type="idl" href="#dom-mlcontext-powerpreference-slot" id="ref-for-dom-mlcontext-powerpreference-slot①">[[powerPreference]]</a></code> to <var>options</var>["<code class="idl"><a data-link-type="idl" href="#dom-mlcontextoptions-powerpreference" id="ref-for-dom-mlcontextoptions-powerpreference①">powerPreference</a></code>"]. Otherwise, set <var>context</var>.<code class="idl"><a data-link-type="idl" href="#dom-mlcontext-powerpreference-slot" id="ref-for-dom-mlcontext-powerpreference-slot②">[[powerPreference]]</a></code> to "<a data-link-type="dfn" href="#power-preference-default" id="ref-for-power-preference-default①">default</a>".</p>
        </ol>
+      <li data-md>
+       <p>Return <var>context</var>.</p>
      </ol>
-    <li data-md>
-     <p>If the <a data-link-type="dfn" href="#validate-mlcontext" id="ref-for-validate-mlcontext">validate MLContext</a> steps given <var>context</var> return <code>false</code>, <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#reject" id="ref-for-reject①">reject</a> <var>promise</var> with a "<code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#notsupportederror" id="ref-for-notsupportederror">NotSupportedError</a></code>" <code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#idl-DOMException" id="ref-for-idl-DOMException②">DOMException</a></code> and abort these steps.</p>
-    <li data-md>
-     <p><a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#resolve" id="ref-for-resolve">Resolve</a> <var>promise</var> with <var>context</var>.</p>
-   </ol>
-   <h4 class="heading settled" data-level="7.2.3" id="api-ml-createcontextsync"><span class="secno">7.2.3. </span><span class="content">The <code class="idl"><a data-link-type="idl" href="#dom-ml-createcontextsync" id="ref-for-dom-ml-createcontextsync">createContextSync()</a></code> method</span><a class="self-link" href="#api-ml-createcontextsync"></a></h4>
-    The <code class="idl"><a data-link-type="idl" href="#dom-ml-createcontextsync" id="ref-for-dom-ml-createcontextsync①">createContextSync()</a></code> method steps are: 
-   <ol>
-    <li data-md>
-     <p>If <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#this" id="ref-for-this①">this</a>'s <a data-link-type="dfn" href="https://html.spec.whatwg.org/multipage/webappapis.html#concept-relevant-global" id="ref-for-concept-relevant-global①">relevant global object</a>'s <a data-link-type="dfn" href="https://html.spec.whatwg.org/multipage/nav-history-apis.html#concept-document-window" id="ref-for-concept-document-window①">associated Document</a> is not <a data-link-type="dfn" href="https://html.spec.whatwg.org/multipage/iframe-embed-object.html#allowed-to-use" id="ref-for-allowed-to-use①">allowed to use</a> the <a data-link-type="dfn" href="#webnn-feature" id="ref-for-webnn-feature①">webnn</a> feature, throw a "<code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#securityerror" id="ref-for-securityerror①">SecurityError</a></code>" <code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#idl-DOMException" id="ref-for-idl-DOMException③">DOMException</a></code> and abort these steps.</p>
-    <li data-md>
-     <p>Let <var>options</var> be the first argument.</p>
-    <li data-md>
-     <p>Let <var>context</var> be the result of running the <a data-link-type="dfn" href="#create-context" id="ref-for-create-context">create context</a> steps given <var>options</var>.</p>
-    <li data-md>
-     <p>If the <a data-link-type="dfn" href="#validate-mlcontext" id="ref-for-validate-mlcontext①">validate MLContext</a> steps given <var>context</var> return <code>false</code>, throw a "<code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#notsupportederror" id="ref-for-notsupportederror①">NotSupportedError</a></code>" <code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#idl-DOMException" id="ref-for-idl-DOMException④">DOMException</a></code> and abort these steps.</p>
-    <li data-md>
-     <p>Return <var>context</var>.</p>
-   </ol>
+    </div>
+   </details>
+   <details class="algorithm" data-algorithm="createContext(options)" data-algorithm-for="ML" open>
+    <summary> The <dfn class="dfn-paneled idl-code" data-dfn-for="ML" data-dfn-type="method" data-export data-lt="createContext(options)|createContext()" id="dom-ml-createcontext"><code>createContext(<var>options</var>)</code></dfn> steps are: </summary>
+    <div class="algorithm-steps">
+     <ol>
+      <li data-md>
+       <p>If <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#this" id="ref-for-this">this</a>'s <a data-link-type="dfn" href="https://html.spec.whatwg.org/multipage/webappapis.html#concept-relevant-global" id="ref-for-concept-relevant-global">relevant global object</a>'s <a data-link-type="dfn" href="https://html.spec.whatwg.org/multipage/nav-history-apis.html#concept-document-window" id="ref-for-concept-document-window">associated Document</a> is not <a data-link-type="dfn" href="https://html.spec.whatwg.org/multipage/iframe-embed-object.html#allowed-to-use" id="ref-for-allowed-to-use">allowed to use</a> the <a data-link-type="dfn" href="#webnn-feature" id="ref-for-webnn-feature">webnn</a> feature, return <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#a-new-promise" id="ref-for-a-new-promise">a new promise</a> <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#reject" id="ref-for-reject">rejected</a> with a "<code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#securityerror" id="ref-for-securityerror">SecurityError</a></code>" <code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#idl-DOMException" id="ref-for-idl-DOMException①">DOMException</a></code>.</p>
+      <li data-md>
+       <p>Let <var>promise</var> be <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#a-new-promise" id="ref-for-a-new-promise①">a new promise</a>.</p>
+      <li data-md>
+       <p>Return <var>promise</var> and run the following steps <a data-link-type="dfn" href="https://html.spec.whatwg.org/multipage/infrastructure.html#in-parallel" id="ref-for-in-parallel">in parallel</a>.</p>
+      <li data-md>
+       <p>Let <var>context</var> be the result of <a data-link-type="dfn" href="#create-a-context" id="ref-for-create-a-context">creating a context</a> given <var>options</var>.</p>
+      <li data-md>
+       <p>If <a data-link-type="dfn" href="#validate-mlcontext" id="ref-for-validate-mlcontext">validating MLContext</a> given <var>context</var> returns false, <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#reject" id="ref-for-reject①">reject</a> <var>promise</var> with a "<code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#notsupportederror" id="ref-for-notsupportederror">NotSupportedError</a></code>" <code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#idl-DOMException" id="ref-for-idl-DOMException②">DOMException</a></code>.</p>
+      <li data-md>
+       <p><a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#resolve" id="ref-for-resolve">Resolve</a> <var>promise</var> with <var>context</var>.</p>
+     </ol>
+    </div>
+   </details>
+   <details class="algorithm" data-algorithm="createContext(gpuDevice)" data-algorithm-for="ML" open>
+    <summary> The <dfn class="dfn-paneled idl-code" data-dfn-for="ML" data-dfn-type="method" data-export id="dom-ml-createcontext-gpudevice"><code>createContext(<var>gpuDevice</var>)</code></dfn> method steps are: </summary>
+    <div class="algorithm-steps">
+     <ol>
+      <li data-md>
+       <p>If <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#this" id="ref-for-this①">this</a>'s <a data-link-type="dfn" href="https://html.spec.whatwg.org/multipage/webappapis.html#concept-relevant-global" id="ref-for-concept-relevant-global①">relevant global object</a>'s <a data-link-type="dfn" href="https://html.spec.whatwg.org/multipage/nav-history-apis.html#concept-document-window" id="ref-for-concept-document-window①">associated Document</a> is not <a data-link-type="dfn" href="https://html.spec.whatwg.org/multipage/iframe-embed-object.html#allowed-to-use" id="ref-for-allowed-to-use①">allowed to use</a> the <a data-link-type="dfn" href="#webnn-feature" id="ref-for-webnn-feature①">webnn</a> feature, return <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#a-new-promise" id="ref-for-a-new-promise②">a new promise</a> <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#reject" id="ref-for-reject②">rejected</a> with a "<code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#securityerror" id="ref-for-securityerror①">SecurityError</a></code>" <code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#idl-DOMException" id="ref-for-idl-DOMException③">DOMException</a></code>.</p>
+      <li data-md>
+       <p>Let <var>promise</var> be <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#a-new-promise" id="ref-for-a-new-promise③">a new promise</a>.</p>
+      <li data-md>
+       <p>Return <var>promise</var> and run the following steps <a data-link-type="dfn" href="https://html.spec.whatwg.org/multipage/infrastructure.html#in-parallel" id="ref-for-in-parallel①">in parallel</a>.</p>
+      <li data-md>
+       <p>Let <var>context</var> be the result of <a data-link-type="dfn" href="#create-a-context" id="ref-for-create-a-context①">creating a context</a> given <var>gpuDevice</var>.</p>
+      <li data-md>
+       <p>If <a data-link-type="dfn" href="#validate-mlcontext" id="ref-for-validate-mlcontext①">validating MLContext</a> given <var>context</var> returns false, <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#reject" id="ref-for-reject③">reject</a> <var>promise</var> with a "<code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#notsupportederror" id="ref-for-notsupportederror①">NotSupportedError</a></code>" <code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#idl-DOMException" id="ref-for-idl-DOMException④">DOMException</a></code>.</p>
+      <li data-md>
+       <p><a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#resolve" id="ref-for-resolve①">Resolve</a> <var>promise</var> with <var>context</var>.</p>
+     </ol>
+    </div>
+   </details>
+   <h4 class="heading settled" data-level="7.2.3" id="api-ml-createcontextsync"><span class="secno">7.2.3. </span><span class="content">The <code class="idl"><a data-link-type="idl" href="#dom-ml-createcontextsync" id="ref-for-dom-ml-createcontextsync①">createContextSync()</a></code> method</span><a class="self-link" href="#api-ml-createcontextsync"></a></h4>
+   <details class="algorithm" data-algorithm="createContextSync(options)" data-algorithm-for="ML" open>
+    <summary> The <dfn class="dfn-paneled idl-code" data-dfn-for="ML" data-dfn-type="method" data-export data-lt="createContextSync(options)|createContextSync()" id="dom-ml-createcontextsync"><code>createContextSync(<var>options</var>)</code></dfn> method steps are: </summary>
+    <div class="algorithm-steps">
+     <ol>
+      <li data-md>
+       <p>If <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#this" id="ref-for-this②">this</a>'s <a data-link-type="dfn" href="https://html.spec.whatwg.org/multipage/webappapis.html#concept-relevant-global" id="ref-for-concept-relevant-global②">relevant global object</a>'s <a data-link-type="dfn" href="https://html.spec.whatwg.org/multipage/nav-history-apis.html#concept-document-window" id="ref-for-concept-document-window②">associated Document</a> is not <a data-link-type="dfn" href="https://html.spec.whatwg.org/multipage/iframe-embed-object.html#allowed-to-use" id="ref-for-allowed-to-use②">allowed to use</a> the <a data-link-type="dfn" href="#webnn-feature" id="ref-for-webnn-feature②">webnn</a> feature, then <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#dfn-throw" id="ref-for-dfn-throw">throw</a> a "<code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#securityerror" id="ref-for-securityerror②">SecurityError</a></code>" <code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#idl-DOMException" id="ref-for-idl-DOMException⑤">DOMException</a></code>.</p>
+      <li data-md>
+       <p>Let <var>context</var> be the result <a data-link-type="dfn" href="#create-a-context" id="ref-for-create-a-context②">creating a context</a> <var>options</var>.</p>
+      <li data-md>
+       <p>If <a data-link-type="dfn" href="#validate-mlcontext" id="ref-for-validate-mlcontext②">validating MLContext</a> given <var>context</var> return false, then <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#dfn-throw" id="ref-for-dfn-throw①">throw</a> a "<code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#notsupportederror" id="ref-for-notsupportederror②">NotSupportedError</a></code>" <code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#idl-DOMException" id="ref-for-idl-DOMException⑥">DOMException</a></code>.</p>
+      <li data-md>
+       <p>Return <var>context</var>.</p>
+     </ol>
+    </div>
+   </details>
+   <details class="algorithm" data-algorithm="createContextSync(gpuDevice)" data-algorithm-for="ML" open>
+    <summary> The <dfn class="dfn-paneled idl-code" data-dfn-for="ML" data-dfn-type="method" data-export id="dom-ml-createcontextsync-gpudevice"><code>createContextSync(<var>gpuDevice</var>)</code></dfn> method steps are: </summary>
+    <div class="algorithm-steps">
+     <ol>
+      <li data-md>
+       <p>If <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#this" id="ref-for-this③">this</a>'s <a data-link-type="dfn" href="https://html.spec.whatwg.org/multipage/webappapis.html#concept-relevant-global" id="ref-for-concept-relevant-global③">relevant global object</a>'s <a data-link-type="dfn" href="https://html.spec.whatwg.org/multipage/nav-history-apis.html#concept-document-window" id="ref-for-concept-document-window③">associated Document</a> is not <a data-link-type="dfn" href="https://html.spec.whatwg.org/multipage/iframe-embed-object.html#allowed-to-use" id="ref-for-allowed-to-use③">allowed to use</a> the <a data-link-type="dfn" href="#webnn-feature" id="ref-for-webnn-feature③">webnn</a> feature, then <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#dfn-throw" id="ref-for-dfn-throw②">throw</a> a "<code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#securityerror" id="ref-for-securityerror③">SecurityError</a></code>" <code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#idl-DOMException" id="ref-for-idl-DOMException⑦">DOMException</a></code>.</p>
+      <li data-md>
+       <p>Let <var>context</var> be the result <a data-link-type="dfn" href="#create-a-context" id="ref-for-create-a-context③">creating a context</a> with <var>gpuDevice</var>.</p>
+      <li data-md>
+       <p>If <a data-link-type="dfn" href="#validate-mlcontext" id="ref-for-validate-mlcontext③">validating MLContext</a> given <var>context</var> return false, then <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#dfn-throw" id="ref-for-dfn-throw③">throw</a> a "<code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#notsupportederror" id="ref-for-notsupportederror③">NotSupportedError</a></code>" <code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#idl-DOMException" id="ref-for-idl-DOMException⑧">DOMException</a></code>.</p>
+      <li data-md>
+       <p>Return <var>context</var>.</p>
+     </ol>
+    </div>
+   </details>
    <h3 class="heading settled" data-level="7.3" id="api-mlgraph"><span class="secno">7.3. </span><span class="content">The MLGraph interface</span><a class="self-link" href="#api-mlgraph"></a></h3>
     The <code class="idl"><a data-link-type="idl" href="#mlgraph" id="ref-for-mlgraph⑥">MLGraph</a></code> interface represents a compiled computational graph. A compiled graph once constructed is immutable and cannot be subsequently changed. 
 <pre class="idl highlight def">[<a class="idl-code" data-link-type="extended-attribute" href="https://webidl.spec.whatwg.org/#SecureContext" id="ref-for-SecureContext②"><c- g>SecureContext</c-></a>, <a class="idl-code" data-link-type="extended-attribute" href="https://webidl.spec.whatwg.org/#Exposed" id="ref-for-Exposed③"><c- g>Exposed</c-></a>=(<c- n>Window</c->, <c- n>DedicatedWorker</c->)]
 <c- b>interface</c-> <dfn class="dfn-paneled idl-code" data-dfn-type="interface" data-export id="mlgraph"><code><c- g>MLGraph</c-></code></dfn> {};
 </pre>
-   <p><code class="idl"><a data-link-type="idl" href="#mlgraph" id="ref-for-mlgraph⑦">MLGraph</a></code> has the following internal slots:</p>
-   <dl>
-    <dt data-md><dfn class="dfn-paneled idl-code" data-dfn-for="MLGraph" data-dfn-type="attribute" data-export id="dom-mlgraph-context-slot"><code>[[context]]</code></dfn> of type <code class="idl"><a data-link-type="idl" href="#mlcontext" id="ref-for-mlcontext①⑦">MLContext</a></code>
-    <dd data-md>
-     <p>The context of type <code class="idl"><a data-link-type="idl" href="#mlcontext" id="ref-for-mlcontext①⑧">MLContext</a></code> associated with this <code class="idl"><a data-link-type="idl" href="#mlgraph" id="ref-for-mlgraph⑧">MLGraph</a></code>.</p>
-    <dt data-md><dfn class="dfn-paneled idl-code" data-dfn-for="MLGraph" data-dfn-type="attribute" data-export id="dom-mlgraph-inputdescriptors-slot"><code>[[inputDescriptors]]</code></dfn> of type <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#idl-record" id="ref-for-idl-record">record</a>&lt;<code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#idl-DOMString" id="ref-for-idl-DOMString">DOMString</a></code>, <code class="idl"><a data-link-type="idl" href="#dictdef-mloperanddescriptor" id="ref-for-dictdef-mloperanddescriptor">MLOperandDescriptor</a></code>>
-    <dd data-md>
-     <p>Maps the name of an input <code class="idl"><a data-link-type="idl" href="#mloperand" id="ref-for-mloperand⑦">MLOperand</a></code> to its <code class="idl"><a data-link-type="idl" href="#dictdef-mloperanddescriptor" id="ref-for-dictdef-mloperanddescriptor①">MLOperandDescriptor</a></code> for all input <code class="idl"><a data-link-type="idl" href="#mloperand" id="ref-for-mloperand⑧">MLOperand</a></code>s of this <code class="idl"><a data-link-type="idl" href="#mlgraph" id="ref-for-mlgraph⑨">MLGraph</a></code>.</p>
-    <dt data-md><dfn class="dfn-paneled idl-code" data-dfn-for="MLGraph" data-dfn-type="attribute" data-export id="dom-mlgraph-outputdescriptors-slot"><code>[[outputDescriptors]]</code></dfn> of type <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#idl-record" id="ref-for-idl-record①">record</a>&lt;<code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#idl-DOMString" id="ref-for-idl-DOMString①">DOMString</a></code>, <code class="idl"><a data-link-type="idl" href="#dictdef-mloperanddescriptor" id="ref-for-dictdef-mloperanddescriptor②">MLOperandDescriptor</a></code>>
-    <dd data-md>
-     <p>Maps the name of an output <code class="idl"><a data-link-type="idl" href="#mloperand" id="ref-for-mloperand⑨">MLOperand</a></code> to its <code class="idl"><a data-link-type="idl" href="#dictdef-mloperanddescriptor" id="ref-for-dictdef-mloperanddescriptor③">MLOperandDescriptor</a></code> for all output <code class="idl"><a data-link-type="idl" href="#mloperand" id="ref-for-mloperand①⓪">MLOperand</a></code>s of this <code class="idl"><a data-link-type="idl" href="#mlgraph" id="ref-for-mlgraph①⓪">MLGraph</a></code>.</p>
-    <dt data-md><dfn class="dfn-paneled idl-code" data-dfn-for="MLGraph" data-dfn-type="attribute" data-export id="dom-mlgraph-implementation-slot"><code>[[implementation]]</code></dfn>
-    <dd data-md>
-     <p>The underlying implementation provided by the User Agent.</p>
-   </dl>
+   <div class="internal-slots">
+     <code class="idl"><a data-link-type="idl" href="#mlgraph" id="ref-for-mlgraph⑦">MLGraph</a></code> has the following internal slots: 
+    <dl>
+     <dt data-md><dfn class="dfn-paneled idl-code" data-dfn-for="MLGraph" data-dfn-type="attribute" data-export id="dom-mlgraph-context-slot"><code>[[context]]</code></dfn> of type <code class="idl"><a data-link-type="idl" href="#mlcontext" id="ref-for-mlcontext①⑦">MLContext</a></code>
+     <dd data-md>
+      <p>The context of type <code class="idl"><a data-link-type="idl" href="#mlcontext" id="ref-for-mlcontext①⑧">MLContext</a></code> associated with this <code class="idl"><a data-link-type="idl" href="#mlgraph" id="ref-for-mlgraph⑧">MLGraph</a></code>.</p>
+     <dt data-md><dfn class="dfn-paneled idl-code" data-dfn-for="MLGraph" data-dfn-type="attribute" data-export id="dom-mlgraph-inputdescriptors-slot"><code>[[inputDescriptors]]</code></dfn> of type <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#idl-record" id="ref-for-idl-record">record</a>&lt;<code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#idl-DOMString" id="ref-for-idl-DOMString">DOMString</a></code>, <code class="idl"><a data-link-type="idl" href="#dictdef-mloperanddescriptor" id="ref-for-dictdef-mloperanddescriptor">MLOperandDescriptor</a></code>>
+     <dd data-md>
+      <p>Maps the name of an input <code class="idl"><a data-link-type="idl" href="#mloperand" id="ref-for-mloperand⑦">MLOperand</a></code> to its <code class="idl"><a data-link-type="idl" href="#dictdef-mloperanddescriptor" id="ref-for-dictdef-mloperanddescriptor①">MLOperandDescriptor</a></code> for all input <code class="idl"><a data-link-type="idl" href="#mloperand" id="ref-for-mloperand⑧">MLOperand</a></code>s of this <code class="idl"><a data-link-type="idl" href="#mlgraph" id="ref-for-mlgraph⑨">MLGraph</a></code>.</p>
+     <dt data-md><dfn class="dfn-paneled idl-code" data-dfn-for="MLGraph" data-dfn-type="attribute" data-export id="dom-mlgraph-outputdescriptors-slot"><code>[[outputDescriptors]]</code></dfn> of type <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#idl-record" id="ref-for-idl-record①">record</a>&lt;<code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#idl-DOMString" id="ref-for-idl-DOMString①">DOMString</a></code>, <code class="idl"><a data-link-type="idl" href="#dictdef-mloperanddescriptor" id="ref-for-dictdef-mloperanddescriptor②">MLOperandDescriptor</a></code>>
+     <dd data-md>
+      <p>Maps the name of an output <code class="idl"><a data-link-type="idl" href="#mloperand" id="ref-for-mloperand⑨">MLOperand</a></code> to its <code class="idl"><a data-link-type="idl" href="#dictdef-mloperanddescriptor" id="ref-for-dictdef-mloperanddescriptor③">MLOperandDescriptor</a></code> for all output <code class="idl"><a data-link-type="idl" href="#mloperand" id="ref-for-mloperand①⓪">MLOperand</a></code>s of this <code class="idl"><a data-link-type="idl" href="#mlgraph" id="ref-for-mlgraph①⓪">MLGraph</a></code>.</p>
+     <dt data-md><dfn class="dfn-paneled idl-code" data-dfn-for="MLGraph" data-dfn-type="attribute" data-export id="dom-mlgraph-implementation-slot"><code>[[implementation]]</code></dfn>
+     <dd data-md>
+      <p>The underlying implementation provided by the User Agent.</p>
+    </dl>
+   </div>
    <h4 class="heading settled" data-level="7.3.1" id="api-mloperanddescriptor"><span class="secno">7.3.1. </span><span class="content">The MLOperandDescriptor dictionary</span><a class="self-link" href="#api-mloperanddescriptor"></a></h4>
 <pre class="idl highlight def"><c- b>enum</c-> <dfn class="dfn-paneled idl-code" data-dfn-type="enum" data-export id="enumdef-mlinputoperandlayout"><code><c- g>MLInputOperandLayout</c-></code></dfn> {
   <dfn class="idl-code" data-dfn-for="MLInputOperandLayout" data-dfn-type="enum-value" data-export id="dom-mlinputoperandlayout-nchw"><code><c- s>"nchw"</c-></code><a class="self-link" href="#dom-mlinputoperandlayout-nchw"></a></dfn>,
@@ -1442,31 +1747,33 @@ <h4 class="heading settled" data-level="7.3.1" id="api-mloperanddescriptor"><spa
   <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#idl-sequence" id="ref-for-idl-sequence"><c- b>sequence</c-></a>&lt;<a class="idl-code" data-link-type="interface" href="https://webidl.spec.whatwg.org/#idl-unsigned-long" id="ref-for-idl-unsigned-long"><c- b>unsigned</c-> <c- b>long</c-></a>> <dfn class="dfn-paneled idl-code" data-dfn-for="MLOperandDescriptor" data-dfn-type="dict-member" data-export data-type="sequence<unsigned long> " id="dom-mloperanddescriptor-dimensions"><code><c- g>dimensions</c-></code></dfn>;
 };
 </pre>
-   <div class="algorithm" data-algorithm="byte length" data-algorithm-for="MLOperandDescriptor">
-     The <dfn class="dfn-paneled" data-dfn-for="MLOperandDescriptor" data-dfn-type="dfn" data-noexport id="mloperanddescriptor-byte-length">byte length</dfn> of an <code class="idl"><a data-link-type="idl" href="#dictdef-mloperanddescriptor" id="ref-for-dictdef-mloperanddescriptor④">MLOperandDescriptor</a></code> <var>desc</var> is the value returned by the following steps: 
-    <ol>
-     <li data-md>
-      <p>Let <var>elementLength</var> be 1.</p>
-     <li data-md>
-      <p>For each <var>dimension</var> of <var>desc</var>.<code class="idl"><a data-link-type="idl" href="#dom-mloperanddescriptor-dimensions" id="ref-for-dom-mloperanddescriptor-dimensions">dimensions</a></code>:</p>
-      <ol>
-       <li data-md>
-        <p>Set <var>elementLength</var> to <var>elementLength</var> × <var>dimension</var>.</p>
-      </ol>
-     <li data-md>
-      <p>Let <var>elementSize</var> be the <a data-link-type="dfn" href="https://tc39.es/ecma262/#table-the-typedarray-constructors" id="ref-for-table-the-typedarray-constructors">element size</a> of one of the <code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#ArrayBufferView" id="ref-for-ArrayBufferView②">ArrayBufferView</a></code> types that matches <var>desc</var>.<code class="idl"><a data-link-type="idl" href="#dom-mloperanddescriptor-type" id="ref-for-dom-mloperanddescriptor-type">type</a></code> according to <a href="#appendices-mloperandtype-arraybufferview-compatibility">this table</a>.</p>
-     <li data-md>
-      <p>Return <var>elementLength</var> × <var>elementSize</var>.</p>
-    </ol>
-   </div>
+   <details class="algorithm" data-algorithm="byte length" data-algorithm-for="MLOperandDescriptor" open>
+    <summary> The <dfn class="dfn-paneled" data-dfn-for="MLOperandDescriptor" data-dfn-type="dfn" data-noexport id="mloperanddescriptor-byte-length">byte length</dfn> of an <code class="idl"><a data-link-type="idl" href="#dictdef-mloperanddescriptor" id="ref-for-dictdef-mloperanddescriptor④">MLOperandDescriptor</a></code> <var>desc</var> is the value returned by the following steps: </summary>
+    <div class="algorithm-steps">
+     <ol>
+      <li data-md>
+       <p>Let <var>elementLength</var> be 1.</p>
+      <li data-md>
+       <p><a data-link-type="dfn" href="https://infra.spec.whatwg.org/#map-iterate" id="ref-for-map-iterate">For each</a> <var>dimension</var> of <var>desc</var>.<code class="idl"><a data-link-type="idl" href="#dom-mloperanddescriptor-dimensions" id="ref-for-dom-mloperanddescriptor-dimensions">dimensions</a></code>:</p>
+       <ol>
+        <li data-md>
+         <p>Set <var>elementLength</var> to <var>elementLength</var> × <var>dimension</var>.</p>
+       </ol>
+      <li data-md>
+       <p>Let <var>elementSize</var> be the <a data-link-type="dfn" href="https://tc39.es/ecma262/#table-the-typedarray-constructors" id="ref-for-table-the-typedarray-constructors">element size</a> of one of the <code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#ArrayBufferView" id="ref-for-ArrayBufferView②">ArrayBufferView</a></code> types that matches <var>desc</var>.<code class="idl"><a data-link-type="idl" href="#dom-mloperanddescriptor-type" id="ref-for-dom-mloperanddescriptor-type">type</a></code> according to <a href="#appendices-mloperandtype-arraybufferview-compatibility">this table</a>.</p>
+      <li data-md>
+       <p>Return <var>elementLength</var> × <var>elementSize</var>.</p>
+     </ol>
+    </div>
+   </details>
    <h4 class="heading settled" data-level="7.3.2" id="api-mloperand"><span class="secno">7.3.2. </span><span class="content">The MLOperand interface</span><a class="self-link" href="#api-mloperand"></a></h4>
    <p>An <code class="idl"><a data-link-type="idl" href="#mloperand" id="ref-for-mloperand①①">MLOperand</a></code> represents an intermediary graph being constructed as a result of compositing parts of an operation into a fully composed operation.</p>
    <p>For instance, an <code class="idl"><a data-link-type="idl" href="#mloperand" id="ref-for-mloperand①②">MLOperand</a></code> may represent a constant feeding to an operation or the result from combining multiple constants together into an operation. See also <a href="#programming-model">§ 6 Programming Model</a>.</p>
 <pre class="idl highlight def">[<a class="idl-code" data-link-type="extended-attribute" href="https://webidl.spec.whatwg.org/#SecureContext" id="ref-for-SecureContext③"><c- g>SecureContext</c-></a>, <a class="idl-code" data-link-type="extended-attribute" href="https://webidl.spec.whatwg.org/#Exposed" id="ref-for-Exposed④"><c- g>Exposed</c-></a>=(<c- n>Window</c->, <c- n>DedicatedWorker</c->)]
 <c- b>interface</c-> <dfn class="dfn-paneled idl-code" data-dfn-type="interface" data-export id="mloperand"><code><c- g>MLOperand</c-></code></dfn> {};
 </pre>
-   <p><code class="idl"><a data-link-type="idl" href="#mloperand" id="ref-for-mloperand①③">MLOperand</a></code> has the following internal slots:</p>
    <div class="internal-slots">
+     <code class="idl"><a data-link-type="idl" href="#mloperand" id="ref-for-mloperand①③">MLOperand</a></code> has the following internal slots: 
     <dl>
      <dt data-md><dfn class="dfn-paneled idl-code" data-dfn-for="MLOperand" data-dfn-type="attribute" data-export id="dom-mloperand-builder-slot"><code>[[builder]]</code></dfn> of type <code class="idl"><a data-link-type="idl" href="#mlgraphbuilder" id="ref-for-mlgraphbuilder⑧">MLGraphBuilder</a></code>
      <dd data-md>
@@ -1480,60 +1787,165 @@ <h4 class="heading settled" data-level="7.3.2" id="api-mloperand"><span class="s
      <dt data-md><dfn class="dfn-paneled idl-code" data-dfn-for="MLOperand" data-dfn-type="attribute" data-export id="dom-mloperand-operand-slot"><code>[[operand]]</code></dfn> of type <a data-link-type="dfn" href="https://w3c.github.io/FileAPI/#blob-url-entry-object" id="ref-for-blob-url-entry-object">object</a>
      <dd data-md>
       <p>Reference to <code class="idl"><a data-link-type="idl" href="#mloperand" id="ref-for-mloperand①⑦">MLOperand</a></code>'s corresponding <a data-link-type="dfn" href="https://infra.spec.whatwg.org/#implementation-defined" id="ref-for-implementation-defined">implementation-defined</a> platform operand object.</p>
-     <dt data-md><dfn class="idl-code" data-dfn-for="MLOperand" data-dfn-type="attribute" data-export id="dom-mloperand-operator-slot"><code>[[operator]]</code><a class="self-link" href="#dom-mloperand-operator-slot"></a></dfn> of type <a data-link-type="dfn" href="https://w3c.github.io/FileAPI/#blob-url-entry-object" id="ref-for-blob-url-entry-object①">object</a>
+     <dt data-md><dfn class="dfn-paneled idl-code" data-dfn-for="MLOperand" data-dfn-type="attribute" data-export id="dom-mloperand-operator-slot"><code>[[operator]]</code></dfn> of type <a data-link-type="dfn" href="https://w3c.github.io/FileAPI/#blob-url-entry-object" id="ref-for-blob-url-entry-object①">object</a>
      <dd data-md>
       <p>Reference to <code class="idl"><a data-link-type="idl" href="#mloperand" id="ref-for-mloperand①⑧">MLOperand</a></code>'s corresponding <a data-link-type="dfn" href="https://infra.spec.whatwg.org/#implementation-defined" id="ref-for-implementation-defined①">implementation-defined</a> platform operator object.</p>
     </dl>
    </div>
-   <p>Since the <code class="idl"><a data-link-type="idl" href="#dom-mloperand-builder-slot" id="ref-for-dom-mloperand-builder-slot">[[builder]]</a></code> object is bound by the <code class="idl"><a data-link-type="idl" href="#dom-mlgraphbuilder-mlgraphbuilder" id="ref-for-dom-mlgraphbuilder-mlgraphbuilder">MLGraphBuilder()</a></code> constructor to an <code class="idl"><a data-link-type="idl" href="#mlcontext" id="ref-for-mlcontext①⑨">MLContext</a></code> object, an <code class="idl"><a data-link-type="idl" href="#mloperand" id="ref-for-mloperand①⑨">MLOperand</a></code> is also always bound to the same <code class="idl"><a data-link-type="idl" href="#mlcontext" id="ref-for-mlcontext②⓪">MLContext</a></code> object.</p>
-   <h5 class="heading settled" data-level="7.3.2.1" id="api-mloperand-create"><span class="secno">7.3.2.1. </span><span class="content">Creating <code class="idl"><a data-link-type="idl" href="#mloperand" id="ref-for-mloperand②⓪">MLOperand</a></code></span><a class="self-link" href="#api-mloperand-create"></a></h5>
-    The <code class="idl"><a data-link-type="idl" href="#mloperand" id="ref-for-mloperand②①">MLOperand</a></code> objects are created by the methods of <code class="idl"><a data-link-type="idl" href="#mlgraphbuilder" id="ref-for-mlgraphbuilder⑨">MLGraphBuilder</a></code>, internally using the following algorithms. 
-   <p>To <dfn class="dfn-paneled" data-dfn-type="dfn" data-noexport id="create-mloperand">create MLOperand</dfn> given <var>builder</var> and <var>desc</var>, run the following steps:</p>
-   <div class="algorithm-steps">
-    <ol>
-     <li data-md>
-      <p>If <var>builder</var> is not an instance of <code class="idl"><a data-link-type="idl" href="#mlgraphbuilder" id="ref-for-mlgraphbuilder①⓪">MLGraphBuilder</a></code>, then throw a "<code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#exceptiondef-typeerror" id="ref-for-exceptiondef-typeerror">TypeError</a></code>" <code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#idl-DOMException" id="ref-for-idl-DOMException⑤">DOMException</a></code> and stop.</p>
-     <li data-md>
-      <p>If <var>desc</var> is not an <a data-link-type="dfn" href="https://w3c.github.io/FileAPI/#blob-url-entry-object" id="ref-for-blob-url-entry-object②">object</a> that <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#implements" id="ref-for-implements">implements</a> <code class="idl"><a data-link-type="idl" href="#dictdef-mloperanddescriptor" id="ref-for-dictdef-mloperanddescriptor⑥">MLOperandDescriptor</a></code>, then throw a "<code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#exceptiondef-typeerror" id="ref-for-exceptiondef-typeerror①">TypeError</a></code>" <code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#idl-DOMException" id="ref-for-idl-DOMException⑥">DOMException</a></code> and stop.</p>
-     <li data-md>
-      <p>Let <var>operand</var> be a new <a data-link-type="dfn" href="https://w3c.github.io/FileAPI/#blob-url-entry-object" id="ref-for-blob-url-entry-object③">object</a>.</p>
-     <li data-md>
-      <p>Set <var>operand</var>.<code class="idl"><a data-link-type="idl" href="#dom-mloperand-builder-slot" id="ref-for-dom-mloperand-builder-slot①">[[builder]]</a></code> to <var>builder</var>.</p>
-     <li data-md>
-      <p>Set <var>operand</var>.<code class="idl"><a data-link-type="idl" href="#dom-mloperand-descriptor-slot" id="ref-for-dom-mloperand-descriptor-slot">[[descriptor]]</a></code> to <var>desc</var>.</p>
-     <li data-md>
-      <p>Return <var>operand</var>.</p>
-    </ol>
-   </div>
-   <p>To <dfn class="dfn-paneled" data-dfn-type="dfn" data-noexport id="check-dimensions">check dimensions</dfn> given <var>dimensions</var> and <var>type</var>, run the following steps:</p>
-   <div class="algorithm-steps">
-    <ol>
-     <li data-md>
-      <p>If <var>dimensions</var> is not an array of positive numbers, return <code>false</code>;</p>
-     <li data-md>
-      <p>If <var>dimensions</var>.length is 0, return <code>false</code>.</p>
-     <li data-md>
-      <p>If <var>dimensions</var>.length is too large to be supported by the implementation, return <code>false</code>.</p>
-     <li data-md>
-      <p>If any element of <var>dimensions</var> is not a positive number, or it is too large to be supported by the implementation given <var>type</var>, return <code>false</code>.</p>
-     <li data-md>
-      <p>Return <code>true</code>.</p>
-    </ol>
+   <div class="algorithm" data-algorithm="rank" data-algorithm-for="MLOperand">
+     To get the <dfn class="dfn-paneled" data-dfn-for="MLOperand" data-dfn-type="dfn" data-noexport id="mloperand-rank">rank</dfn> of an <code class="idl"><a data-link-type="idl" href="#mloperand" id="ref-for-mloperand①⑨">MLOperand</a></code> <var>operand</var>, run the following steps: 
+    <div class="algorithm-steps">
+     <ol>
+      <li data-md>
+       <p>Return the <a data-link-type="dfn" href="https://infra.spec.whatwg.org/#list-size" id="ref-for-list-size">size</a> of <var>operand</var>.<code class="idl"><a data-link-type="idl" href="#dom-mloperand-descriptor-slot" id="ref-for-dom-mloperand-descriptor-slot">[[descriptor]]</a></code>.<code class="idl"><a data-link-type="idl" href="#dom-mloperanddescriptor-dimensions" id="ref-for-dom-mloperanddescriptor-dimensions①">dimensions</a></code>.</p>
+     </ol>
+    </div>
    </div>
-   <h4 class="heading settled" data-level="7.3.3" id="api-mlactivation"><span class="secno">7.3.3. </span><span class="content">The MLActivation interface</span><a class="self-link" href="#api-mlactivation"></a></h4>
-   <p>Objects implementing the <code class="idl"><a data-link-type="idl" href="#mlactivation" id="ref-for-mlactivation">MLActivation</a></code> interface represent activation function types.</p>
-<pre class="idl highlight def">[<a class="idl-code" data-link-type="extended-attribute" href="https://webidl.spec.whatwg.org/#SecureContext" id="ref-for-SecureContext④"><c- g>SecureContext</c-></a>, <a class="idl-code" data-link-type="extended-attribute" href="https://webidl.spec.whatwg.org/#Exposed" id="ref-for-Exposed⑤"><c- g>Exposed</c-></a>=(<c- n>Window</c->, <c- n>DedicatedWorker</c->)]
-<c- b>interface</c-> <dfn class="dfn-paneled idl-code" data-dfn-type="interface" data-export id="mlactivation"><code><c- g>MLActivation</c-></code></dfn> {};
-</pre>
-   <div class="note" role="note"> These activations function types are used to create other operations. One such use of this interface is for when an activation function is fused into another operation such as <a href="#api-mlgraphbuilder-conv2d">§ 7.6.7 The conv2d() method</a> or <a href="#api-mlgraphbuilder-batchnorm">§ 7.6.4 The batchNormalization() method</a> during a graph construction session. Such fused activation functions can provide a significant performance improvement when supported natively by the underlying implementation. This is intended as an optimization opportunity for implementers. </div>
-   <div class="note" role="note"> The implementation of the <code class="idl"><a data-link-type="idl" href="#mlactivation" id="ref-for-mlactivation①">MLActivation</a></code> interface can simply be a struct that holds a string type of the activation function along with other properties needed. The actual creation of the activation function e.g. a <a href="#api-mlgraphbuilder-sigmoid">§ 7.6.30 The sigmoid() method</a> or <a href="#api-mlgraphbuilder-relu">§ 7.6.27 The relu() method</a> can then be deferred until when the rest of the graph is ready to connect with it such as during the construction of <a href="#api-mlgraphbuilder-conv2d">§ 7.6.7 The conv2d() method</a> for example. </div>
-   <h3 class="heading settled" data-level="7.4" id="api-mlcontext"><span class="secno">7.4. </span><span class="content">The MLContext interface</span><a class="self-link" href="#api-mlcontext"></a></h3>
-    The <code class="idl"><a data-link-type="idl" href="#mlcontext" id="ref-for-mlcontext②①">MLContext</a></code> interface represents a global state of neural network compute workload and execution processes. Each <code class="idl"><a data-link-type="idl" href="#mlcontext" id="ref-for-mlcontext②②">MLContext</a></code> object has associated <a data-link-type="dfn" href="#context-type" id="ref-for-context-type">context type</a>, <a data-link-type="dfn" href="#device-type" id="ref-for-device-type③">device type</a> and <a data-link-type="dfn" href="#power-preference" id="ref-for-power-preference②">power preference</a>. 
-   <p>The <dfn class="dfn-paneled" data-dfn-type="dfn" data-noexport id="context-type">context type</dfn> is the type of the execution context that manages the resources and facilitates the compilation and execution of the neural network graph:</p>
-   <dl>
-    <dt>"<code><dfn class="dfn-paneled" data-dfn-type="dfn" data-lt="default-context" data-noexport id="default-context">default</dfn></code>"
-    <dd>Context created per user preference options.
-    <dt>"<code><dfn class="dfn-paneled" data-dfn-type="dfn" data-lt="webgpu-context" data-noexport id="webgpu-context">webgpu</dfn></code>"
+   <p>Since the <code class="idl"><a data-link-type="idl" href="#dom-mloperand-builder-slot" id="ref-for-dom-mloperand-builder-slot">[[builder]]</a></code> object is bound by the <code class="idl"><a data-link-type="idl" href="#dom-mlgraphbuilder-mlgraphbuilder" id="ref-for-dom-mlgraphbuilder-mlgraphbuilder">MLGraphBuilder()</a></code> constructor to an <code class="idl"><a data-link-type="idl" href="#mlcontext" id="ref-for-mlcontext①⑨">MLContext</a></code> object, an <code class="idl"><a data-link-type="idl" href="#mloperand" id="ref-for-mloperand②⓪">MLOperand</a></code> is also always bound to the same <code class="idl"><a data-link-type="idl" href="#mlcontext" id="ref-for-mlcontext②⓪">MLContext</a></code> object.</p>
+   <h5 class="heading settled" data-level="7.3.2.1" id="api-mloperand-create"><span class="secno">7.3.2.1. </span><span class="content">Creating <code class="idl"><a data-link-type="idl" href="#mloperand" id="ref-for-mloperand②①">MLOperand</a></code></span><a class="self-link" href="#api-mloperand-create"></a></h5>
+    The <code class="idl"><a data-link-type="idl" href="#mloperand" id="ref-for-mloperand②②">MLOperand</a></code> objects are created by the methods of <code class="idl"><a data-link-type="idl" href="#mlgraphbuilder" id="ref-for-mlgraphbuilder⑨">MLGraphBuilder</a></code>, internally using the following algorithms. 
+   <details class="algorithm" data-algorithm="create an MLOperand" open>
+    <summary> To <dfn class="dfn-paneled" data-dfn-type="dfn" data-noexport id="create-an-mloperand">create an MLOperand</dfn> given <var>builder</var> and <var>desc</var>, run the following steps: </summary>
+    <div class="algorithm-steps">
+     <ol>
+      <li data-md>
+       <p><a data-link-type="dfn" href="https://infra.spec.whatwg.org/#assert" id="ref-for-assert">Assert</a>: the type of <var>builder</var> is <code class="idl"><a data-link-type="idl" href="#mlgraphbuilder" id="ref-for-mlgraphbuilder①⓪">MLGraphBuilder</a></code>.</p>
+      <li data-md>
+       <p><a data-link-type="dfn" href="https://infra.spec.whatwg.org/#assert" id="ref-for-assert①">Assert</a>: the type of <var>desc</var> is <code class="idl"><a data-link-type="idl" href="#dictdef-mloperanddescriptor" id="ref-for-dictdef-mloperanddescriptor⑥">MLOperandDescriptor</a></code>.</p>
+      <li data-md>
+       <p>Let <var>operand</var> be a new <a data-link-type="dfn" href="https://w3c.github.io/FileAPI/#blob-url-entry-object" id="ref-for-blob-url-entry-object②">object</a>.</p>
+      <li data-md>
+       <p>Set <var>operand</var>.<code class="idl"><a data-link-type="idl" href="#dom-mloperand-builder-slot" id="ref-for-dom-mloperand-builder-slot①">[[builder]]</a></code> to <var>builder</var>.</p>
+      <li data-md>
+       <p>Set <var>operand</var>.<code class="idl"><a data-link-type="idl" href="#dom-mloperand-descriptor-slot" id="ref-for-dom-mloperand-descriptor-slot①">[[descriptor]]</a></code> to <var>desc</var>.</p>
+      <li data-md>
+       <p>Return <var>operand</var>.</p>
+     </ol>
+    </div>
+   </details>
+   <details class="algorithm" data-algorithm="copy an MLOperand" open>
+    <summary> To <dfn class="dfn-paneled" data-dfn-type="dfn" data-noexport id="copy-an-mloperand">copy an MLOperand</dfn> given <var>operand</var>, run the following steps: </summary>
+    <div class="algorithm-steps">
+     <ol>
+      <li data-md>
+       <p><a data-link-type="dfn" href="https://infra.spec.whatwg.org/#assert" id="ref-for-assert②">Assert</a>: the type of <var>operand</var> is <code class="idl"><a data-link-type="idl" href="#mloperand" id="ref-for-mloperand②③">MLOperand</a></code>.</p>
+      <li data-md>
+       <p>Let <var>result</var> be a new <a data-link-type="dfn" href="https://w3c.github.io/FileAPI/#blob-url-entry-object" id="ref-for-blob-url-entry-object③">object</a>.</p>
+      <li data-md>
+       <p>Set <var>result</var>.<code class="idl"><a data-link-type="idl" href="#dom-mloperand-builder-slot" id="ref-for-dom-mloperand-builder-slot②">[[builder]]</a></code> to <var>operand</var>.<code class="idl"><a data-link-type="idl" href="#dom-mloperand-builder-slot" id="ref-for-dom-mloperand-builder-slot③">[[builder]]</a></code>.</p>
+      <li data-md>
+       <p>Set <var>result</var>.<code class="idl"><a data-link-type="idl" href="#dom-mloperand-descriptor-slot" id="ref-for-dom-mloperand-descriptor-slot②">[[descriptor]]</a></code> to <var>operand</var>.<code class="idl"><a data-link-type="idl" href="#dom-mloperand-descriptor-slot" id="ref-for-dom-mloperand-descriptor-slot③">[[descriptor]]</a></code>.</p>
+      <li data-md>
+       <p>If <var>operand</var>.<code class="idl"><a data-link-type="idl" href="#dom-mloperand-name-slot" id="ref-for-dom-mloperand-name-slot">[[name]]</a></code> <a data-link-type="dfn" href="https://infra.spec.whatwg.org/#map-exists" id="ref-for-map-exists②">exists</a>, then set <var>result</var>.<code class="idl"><a data-link-type="idl" href="#dom-mloperand-name-slot" id="ref-for-dom-mloperand-name-slot①">[[name]]</a></code> to <var>operand</var>.<code class="idl"><a data-link-type="idl" href="#dom-mloperand-name-slot" id="ref-for-dom-mloperand-name-slot②">[[name]]</a></code>.</p>
+      <li data-md>
+       <p>Return <var>result</var>.</p>
+     </ol>
+    </div>
+   </details>
+   <details class="algorithm" data-algorithm="check dimensions" open>
+    <summary> To <dfn class="dfn-paneled" data-dfn-type="dfn" data-noexport id="check-dimensions">check dimensions</dfn> given <var>dimensions</var> and <var>type</var>, run the following steps: </summary>
+    <div class="algorithm-steps">
+     <ol>
+      <li data-md>
+       <p>If the <a data-link-type="dfn" href="https://infra.spec.whatwg.org/#list-size" id="ref-for-list-size①">size</a> of <var>dimensions</var> is 0, return false.</p>
+      <li data-md>
+       <p>If the <a data-link-type="dfn" href="https://infra.spec.whatwg.org/#list-size" id="ref-for-list-size②">size</a> of <var>dimensions</var> is too large to be supported by the implementation, return false.</p>
+      <li data-md>
+       <p>If any element of <var>dimensions</var> is not a positive number, or it is too large to be supported by the implementation given <var>type</var>, return false.</p>
+      <li data-md>
+       <p>Return true.</p>
+     </ol>
+    </div>
+   </details>
+   <details class="algorithm" data-algorithm="validate MLOperand" data-algorithm-for="MLOperand" open>
+    <summary> To <dfn class="dfn-paneled" data-dfn-for="MLOperand" data-dfn-type="dfn" data-noexport id="mloperand-validate-mloperand">validate MLOperand</dfn> given <var>operand</var> and <var>builder</var>, run the following steps: </summary>
+    <div class="algorithm-steps">
+     <ol>
+      <li data-md>
+       <p><a data-link-type="dfn" href="https://infra.spec.whatwg.org/#assert" id="ref-for-assert③">Assert</a>: the type of <var>operand</var>.<code class="idl"><a data-link-type="idl" href="#dom-mloperand-builder-slot" id="ref-for-dom-mloperand-builder-slot④">[[builder]]</a></code> is <code class="idl"><a data-link-type="idl" href="#mlgraphbuilder" id="ref-for-mlgraphbuilder①①">MLGraphBuilder</a></code>.</p>
+      <li data-md>
+       <p>If <var>builder</var> is not equal to <var>operand</var>.<code class="idl"><a data-link-type="idl" href="#dom-mloperand-builder-slot" id="ref-for-dom-mloperand-builder-slot⑤">[[builder]]</a></code>, return false.</p>
+      <li data-md>
+       <p>Let <var>desc</var> be <var>operand</var>.<code class="idl"><a data-link-type="idl" href="#dom-mloperand-descriptor-slot" id="ref-for-dom-mloperand-descriptor-slot④">[[descriptor]]</a></code>.</p>
+      <li data-md>
+       <p>If <var>desc</var>.<code class="idl"><a data-link-type="idl" href="#dom-mloperanddescriptor-dimensions" id="ref-for-dom-mloperanddescriptor-dimensions②">dimensions</a></code> <a data-link-type="dfn" href="https://infra.spec.whatwg.org/#map-exists" id="ref-for-map-exists③">exists</a> and invoking <a data-link-type="dfn" href="#check-dimensions" id="ref-for-check-dimensions">check dimensions</a> given <var>desc</var>.<code class="idl"><a data-link-type="idl" href="#dom-mloperanddescriptor-dimensions" id="ref-for-dom-mloperanddescriptor-dimensions③">dimensions</a></code> and <var>desc</var>.<code class="idl"><a data-link-type="idl" href="#dom-mloperanddescriptor-type" id="ref-for-dom-mloperanddescriptor-type①">type</a></code> returns false, then return false.</p>
+      <li data-md>
+       <p>Return true.</p>
+     </ol>
+    </div>
+   </details>
+   <h4 class="heading settled" data-level="7.3.3" id="api-mlactivation"><span class="secno">7.3.3. </span><span class="content">The MLActivation interface</span><a class="self-link" href="#api-mlactivation"></a></h4>
+   <p>Objects implementing the <code class="idl"><a data-link-type="idl" href="#mlactivation" id="ref-for-mlactivation">MLActivation</a></code> interface represent activation function types.</p>
+<pre class="idl highlight def">[<a class="idl-code" data-link-type="extended-attribute" href="https://webidl.spec.whatwg.org/#SecureContext" id="ref-for-SecureContext④"><c- g>SecureContext</c-></a>, <a class="idl-code" data-link-type="extended-attribute" href="https://webidl.spec.whatwg.org/#Exposed" id="ref-for-Exposed⑤"><c- g>Exposed</c-></a>=(<c- n>Window</c->, <c- n>DedicatedWorker</c->)]
+<c- b>interface</c-> <dfn class="dfn-paneled idl-code" data-dfn-type="interface" data-export id="mlactivation"><code><c- g>MLActivation</c-></code></dfn> {
+};
+</pre>
+   <div class="internal-slots">
+     <code class="idl"><a data-link-type="idl" href="#mlactivation" id="ref-for-mlactivation①">MLActivation</a></code> has the following internal slots: 
+    <dl>
+     <dt data-md><dfn class="dfn-paneled idl-code" data-dfn-for="MLActivation" data-dfn-type="attribute" data-export id="dom-mlactivation-name-slot"><code>[[name]]</code></dfn> of type <a data-link-type="dfn" href="https://infra.spec.whatwg.org/#string" id="ref-for-string①">string</a>
+     <dd data-md>
+      <p>The <code class="idl"><a data-link-type="idl" href="#mlactivation" id="ref-for-mlactivation②">MLActivation</a></code>'s name.</p>
+     <dt data-md><dfn class="dfn-paneled idl-code" data-dfn-for="MLActivation" data-dfn-type="attribute" data-export id="dom-mlactivation-builder-slot"><code>[[builder]]</code></dfn> of type <code class="idl"><a data-link-type="idl" href="#mlgraphbuilder" id="ref-for-mlgraphbuilder①②">MLGraphBuilder</a></code>
+     <dd data-md>
+      <p>The graph builder object this <code class="idl"><a data-link-type="idl" href="#mlactivation" id="ref-for-mlactivation③">MLActivation</a></code> belongs to.</p>
+     <dt data-md><dfn class="dfn-paneled idl-code" data-dfn-for="MLActivation" data-dfn-type="attribute" data-export id="dom-mlactivation-options-slot"><code>[[options]]</code></dfn> of type <a data-link-type="dfn" href="https://w3c.github.io/FileAPI/#blob-url-entry-object" id="ref-for-blob-url-entry-object④">object</a>
+     <dd data-md>
+      <p>A dictionary containing <code class="idl"><a data-link-type="idl" href="#mlactivation" id="ref-for-mlactivation④">MLActivation</a></code> options.</p>
+     <dt data-md><dfn class="dfn-paneled idl-code" data-dfn-for="MLActivation" data-dfn-type="attribute" data-export id="dom-mlactivation-operator-slot"><code>[[operator]]</code></dfn> of type <a data-link-type="dfn" href="https://w3c.github.io/FileAPI/#blob-url-entry-object" id="ref-for-blob-url-entry-object⑤">object</a>
+     <dd data-md>
+      <p>Reference to <code class="idl"><a data-link-type="idl" href="#mlactivation" id="ref-for-mlactivation⑤">MLActivation</a></code>'s corresponding <a data-link-type="dfn" href="https://infra.spec.whatwg.org/#implementation-defined" id="ref-for-implementation-defined②">implementation-defined</a> platform operator object.</p>
+    </dl>
+   </div>
+   <div class="note" role="note"> These activations function types are used to create other operations. One such use of this interface is for when an activation function is fused into another operation such as <a href="#api-mlgraphbuilder-conv2d">§ 7.6.8 The conv2d() method</a> or <a href="#api-mlgraphbuilder-batchnorm">§ 7.6.5 The batchNormalization() method</a> during a graph construction session. Such fused activation functions can provide a significant performance improvement when supported natively by the underlying implementation. This is intended as an optimization opportunity for implementers. </div>
+   <h5 class="heading settled" data-level="7.3.3.1" id="api-mlactivation-create"><span class="secno">7.3.3.1. </span><span class="content">Creating <code class="idl"><a data-link-type="idl" href="#mlactivation" id="ref-for-mlactivation⑥">MLActivation</a></code></span><a class="self-link" href="#api-mlactivation-create"></a></h5>
+   <div class="note" role="note"> The <code class="idl"><a data-link-type="idl" href="#mlactivation" id="ref-for-mlactivation⑦">MLActivation</a></code> objects (including the ones passed as input to methods) are created by the methods of <code class="idl"><a data-link-type="idl" href="#mlgraphbuilder" id="ref-for-mlgraphbuilder①③">MLGraphBuilder</a></code> and are identified by their name. The <var>options</var> dictionary is defined by those methods. The actual creation of the activation function e.g. a <a href="#api-mlgraphbuilder-sigmoid-method">§ 7.6.31 The sigmoid() method</a> or <a href="#api-mlgraphbuilder-relu-method">§ 7.6.28 The relu() method</a> can then be deferred until when the rest of the graph is ready to connect with it such as during the construction of <a href="#api-mlgraphbuilder-conv2d">§ 7.6.8 The conv2d() method</a> for example. </div>
+   <details class="algorithm" data-algorithm="create an MLActivation" open>
+    <summary> To <dfn class="dfn-paneled" data-dfn-type="dfn" data-noexport id="create-an-mlactivation">create an MLActivation</dfn> given <var>builder</var>, <var>name</var>, <var>options</var> and <var>init-steps</var>, run the following steps: </summary>
+    <div class="algorithm-steps">
+     <ol>
+      <li data-md>
+       <p><a data-link-type="dfn" href="https://infra.spec.whatwg.org/#assert" id="ref-for-assert④">Assert</a>: the type of <var>builder</var> is <code class="idl"><a data-link-type="idl" href="#mlgraphbuilder" id="ref-for-mlgraphbuilder①④">MLGraphBuilder</a></code>.</p>
+      <li data-md>
+       <p>If <var>name</var> is empty, then <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#dfn-throw" id="ref-for-dfn-throw④">throw</a> a "<code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#exceptiondef-typeerror" id="ref-for-exceptiondef-typeerror">TypeError</a></code>".</p>
+      <li data-md>
+       <p>Let <var>activation</var> be a new <a data-link-type="dfn" href="https://w3c.github.io/FileAPI/#blob-url-entry-object" id="ref-for-blob-url-entry-object⑥">object</a>.</p>
+      <li data-md>
+       <p>Set <var>activation</var>.<code class="idl"><a data-link-type="idl" href="#dom-mlactivation-builder-slot" id="ref-for-dom-mlactivation-builder-slot">[[builder]]</a></code> to <var>builder</var>.</p>
+      <li data-md>
+       <p>Set <var>activation</var>.<code class="idl"><a data-link-type="idl" href="#dom-mlactivation-name-slot" id="ref-for-dom-mlactivation-name-slot">[[name]]</a></code> to <var>name</var>.</p>
+      <li data-md>
+       <p>If <var>options</var> is an <a data-link-type="dfn" href="https://w3c.github.io/FileAPI/#blob-url-entry-object" id="ref-for-blob-url-entry-object⑦">object</a>, set <var>activation</var>.<code class="idl"><a data-link-type="idl" href="#dom-mlactivation-options-slot" id="ref-for-dom-mlactivation-options-slot">[[options]]</a></code> to <var>options</var>.</p>
+      <li data-md>
+       <p>If any of the following sub-steps fail, <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#dfn-throw" id="ref-for-dfn-throw⑤">throw</a> an "<code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#operationerror" id="ref-for-operationerror①">OperationError</a></code>" <code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#idl-DOMException" id="ref-for-idl-DOMException⑨">DOMException</a></code>.</p>
+       <ol>
+        <li data-md>
+         <p>Make a request to the underlying platform to:</p>
+         <ol>
+          <li data-md>
+           <p>Create an <a data-link-type="dfn" href="https://infra.spec.whatwg.org/#implementation-defined" id="ref-for-implementation-defined③">implementation-defined</a> platform operator <var>opImpl</var> for the given <var>name</var> operation.</p>
+          <li data-md>
+           <p>Store a reference of <var>opImpl</var> in <var>activation</var>.<code class="idl"><a data-link-type="idl" href="#dom-mlactivation-operator-slot" id="ref-for-dom-mlactivation-operator-slot">[[operator]]</a></code>.</p>
+         </ol>
+        <li data-md>
+         <p>If <var>init-steps</var> are defined, run <var>init-steps</var> with <var>options</var>.</p>
+         <ol>
+          <li data-md>
+           <p>Otherwise, initialize <var>activation</var>.<code class="idl"><a data-link-type="idl" href="#dom-mlactivation-operator-slot" id="ref-for-dom-mlactivation-operator-slot①">[[operator]]</a></code> given <var>options</var> in an <a data-link-type="dfn" href="https://infra.spec.whatwg.org/#implementation-defined" id="ref-for-implementation-defined④">implementation-defined</a> way for the given <var>name</var> operation.</p>
+         </ol>
+       </ol>
+      <li data-md>
+       <p>Return <var>activation</var>.</p>
+     </ol>
+    </div>
+   </details>
+   <h3 class="heading settled" data-level="7.4" id="api-mlcontext"><span class="secno">7.4. </span><span class="content">The MLContext interface</span><a class="self-link" href="#api-mlcontext"></a></h3>
+    The <code class="idl"><a data-link-type="idl" href="#mlcontext" id="ref-for-mlcontext②①">MLContext</a></code> interface represents a global state of neural network compute workload and execution processes. Each <code class="idl"><a data-link-type="idl" href="#mlcontext" id="ref-for-mlcontext②②">MLContext</a></code> object has associated <a data-link-type="dfn" href="#context-type" id="ref-for-context-type">context type</a>, <a data-link-type="dfn" href="#device-type" id="ref-for-device-type③">device type</a> and <a data-link-type="dfn" href="#power-preference" id="ref-for-power-preference②">power preference</a>. 
+   <p>The <dfn class="dfn-paneled" data-dfn-type="dfn" data-noexport id="context-type">context type</dfn> is the type of the execution context that manages the resources and facilitates the compilation and execution of the neural network graph:</p>
+   <dl>
+    <dt>"<code><dfn class="dfn-paneled" data-dfn-type="dfn" data-lt="default-context" data-noexport id="default-context">default</dfn></code>"
+    <dd>Context created per user preference options.
+    <dt>"<code><dfn class="dfn-paneled" data-dfn-type="dfn" data-lt="webgpu-context" data-noexport id="webgpu-context">webgpu</dfn></code>"
     <dd>Context created from WebGPU device.
    </dl>
    <p>The <dfn class="dfn-paneled" data-dfn-type="dfn" data-noexport id="device-type">device type</dfn> indicates the kind of device used for the context. It is one of the following:</p>
@@ -1557,43 +1969,49 @@ <h3 class="heading settled" data-level="7.4" id="api-mlcontext"><span class="sec
 [<a class="idl-code" data-link-type="extended-attribute" href="https://webidl.spec.whatwg.org/#SecureContext" id="ref-for-SecureContext⑤"><c- g>SecureContext</c-></a>, <a class="idl-code" data-link-type="extended-attribute" href="https://webidl.spec.whatwg.org/#Exposed" id="ref-for-Exposed⑥"><c- g>Exposed</c-></a>=(<c- n>Window</c->, <c- n>DedicatedWorker</c->)]
 <c- b>interface</c-> <dfn class="dfn-paneled idl-code" data-dfn-type="interface" data-export id="mlcontext"><code><c- g>MLContext</c-></code></dfn> {};
 </pre>
-   <p><code class="idl"><a data-link-type="idl" href="#mlcontext" id="ref-for-mlcontext②③">MLContext</a></code> has the following internal slots:</p>
-   <dl>
-    <dt data-md><dfn class="dfn-paneled idl-code" data-dfn-for="MLContext" data-dfn-type="attribute" data-export id="dom-mlcontext-contexttype-slot"><code>[[contextType]]</code></dfn> of type <a data-link-type="dfn" href="#context-type" id="ref-for-context-type①">context type</a>
-    <dd data-md>
-     <p>The <code class="idl"><a data-link-type="idl" href="#mlcontext" id="ref-for-mlcontext②④">MLContext</a></code>'s <a data-link-type="dfn" href="#context-type" id="ref-for-context-type②">context type</a>.</p>
-    <dt data-md><dfn class="dfn-paneled idl-code" data-dfn-for="MLContext" data-dfn-type="attribute" data-export id="dom-mlcontext-devicetype-slot"><code>[[deviceType]]</code></dfn> of type <a data-link-type="dfn" href="#device-type" id="ref-for-device-type④">device type</a>
-    <dd data-md>
-     <p>The <code class="idl"><a data-link-type="idl" href="#mlcontext" id="ref-for-mlcontext②⑤">MLContext</a></code>'s <a data-link-type="dfn" href="#device-type" id="ref-for-device-type⑤">device type</a>.</p>
-    <dt data-md><dfn class="dfn-paneled idl-code" data-dfn-for="MLContext" data-dfn-type="attribute" data-export id="dom-mlcontext-powerpreference-slot"><code>[[powerPreference]]</code></dfn> of type <a data-link-type="dfn" href="#power-preference" id="ref-for-power-preference③">power preference</a>
-    <dd data-md>
-     <p>The <code class="idl"><a data-link-type="idl" href="#mlcontext" id="ref-for-mlcontext②⑥">MLContext</a></code>'s <a data-link-type="dfn" href="#power-preference" id="ref-for-power-preference④">power preference</a>.</p>
-   </dl>
+   <div class="internal-slots">
+     <code class="idl"><a data-link-type="idl" href="#mlcontext" id="ref-for-mlcontext②③">MLContext</a></code> has the following internal slots: 
+    <dl>
+     <dt data-md><dfn class="dfn-paneled idl-code" data-dfn-for="MLContext" data-dfn-type="attribute" data-export id="dom-mlcontext-contexttype-slot"><code>[[contextType]]</code></dfn> of type <a data-link-type="dfn" href="#context-type" id="ref-for-context-type①">context type</a>
+     <dd data-md>
+      <p>The <code class="idl"><a data-link-type="idl" href="#mlcontext" id="ref-for-mlcontext②④">MLContext</a></code>'s <a data-link-type="dfn" href="#context-type" id="ref-for-context-type②">context type</a>.</p>
+     <dt data-md><dfn class="dfn-paneled idl-code" data-dfn-for="MLContext" data-dfn-type="attribute" data-export id="dom-mlcontext-devicetype-slot"><code>[[deviceType]]</code></dfn> of type <a data-link-type="dfn" href="#device-type" id="ref-for-device-type④">device type</a>
+     <dd data-md>
+      <p>The <code class="idl"><a data-link-type="idl" href="#mlcontext" id="ref-for-mlcontext②⑤">MLContext</a></code>'s <a data-link-type="dfn" href="#device-type" id="ref-for-device-type⑤">device type</a>.</p>
+     <dt data-md><dfn class="dfn-paneled idl-code" data-dfn-for="MLContext" data-dfn-type="attribute" data-export id="dom-mlcontext-powerpreference-slot"><code>[[powerPreference]]</code></dfn> of type <a data-link-type="dfn" href="#power-preference" id="ref-for-power-preference③">power preference</a>
+     <dd data-md>
+      <p>The <code class="idl"><a data-link-type="idl" href="#mlcontext" id="ref-for-mlcontext②⑥">MLContext</a></code>'s <a data-link-type="dfn" href="#power-preference" id="ref-for-power-preference④">power preference</a>.</p>
+    </dl>
+   </div>
    <div class="note" role="note"> When the <code class="idl"><a data-link-type="idl" href="#dom-mlcontext-contexttype-slot" id="ref-for-dom-mlcontext-contexttype-slot②">[[contextType]]</a></code> is set to <a data-link-type="dfn" href="#default-context" id="ref-for-default-context①">default</a> with the <code class="idl"><a data-link-type="idl" href="#dictdef-mlcontextoptions" id="ref-for-dictdef-mlcontextoptions④">MLContextOptions</a></code>.<code class="idl"><a data-link-type="idl" href="#dom-mlcontextoptions-devicetype" id="ref-for-dom-mlcontextoptions-devicetype②">deviceType</a></code> set to <a data-link-type="dfn" href="#device-type-gpu" id="ref-for-device-type-gpu①">gpu</a>, the user agent is responsible for creating an internal GPU device that operates within the context and is capable of ML workload submission on behalf of the calling application. In this setting however, only <code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#ArrayBufferView" id="ref-for-ArrayBufferView④">ArrayBufferView</a></code> inputs and outputs are allowed in and out of the graph execution since the application has no way to know what type of internal GPU device is being created on their behalf. In this case, the user agent is responsible for automatic uploads and downloads of the inputs and outputs to and from the GPU memory using this said internal device. </div>
    <h4 class="heading settled" data-level="7.4.1" id="api-mlcontext-validate"><span class="secno">7.4.1. </span><span class="content">The <code class="idl"><a data-link-type="idl" href="#mlcontext" id="ref-for-mlcontext②⑦">MLContext</a></code> validation algorithm</span><a class="self-link" href="#api-mlcontext-validate"></a></h4>
-    To <dfn class="dfn-paneled" data-dfn-type="dfn" data-noexport id="validate-mlcontext">validate <code class="idl"><a data-link-type="idl" href="#mlcontext" id="ref-for-mlcontext②⑧">MLContext</a></code></dfn>, given <var>context</var>, run these steps: 
-   <ol>
-    <li data-md>
-     <p>If <var>context</var>.<code class="idl"><a data-link-type="idl" href="#dom-mlcontext-contexttype-slot" id="ref-for-dom-mlcontext-contexttype-slot③">[[contextType]]</a></code> is not "<a data-link-type="dfn" href="#webgpu-context" id="ref-for-webgpu-context①">webgpu</a>" or "<a data-link-type="dfn" href="#default-context" id="ref-for-default-context②">default</a>, return <code>false</code>.</p>
-    <li data-md>
-     <p>If <var>context</var>.<code class="idl"><a data-link-type="idl" href="#dom-mlcontext-devicetype-slot" id="ref-for-dom-mlcontext-devicetype-slot③">[[deviceType]]</a></code> is not "<a data-link-type="dfn" href="#device-type-cpu" id="ref-for-device-type-cpu①">cpu</a>" or "<a data-link-type="dfn" href="#device-type-gpu" id="ref-for-device-type-gpu②">gpu</a>", return <code>false</code>.</p>
-    <li data-md>
-     <p>If <var>context</var>.<code class="idl"><a data-link-type="idl" href="#dom-mlcontext-powerpreference-slot" id="ref-for-dom-mlcontext-powerpreference-slot③">[[powerPreference]]</a></code> is not "<a data-link-type="dfn" href="#power-preference-default" id="ref-for-power-preference-default②">default</a>" or "<a data-link-type="dfn" href="#power-preference-high-performance" id="ref-for-power-preference-high-performance">high-performance</a>" or "<a data-link-type="dfn" href="#power-preference-low-power" id="ref-for-power-preference-low-power">low-power</a>", return <code>false</code>.</p>
-    <li data-md>
-     <p>If the user agent cannot support <var>context</var>.<code class="idl"><a data-link-type="idl" href="#dom-mlcontext-contexttype-slot" id="ref-for-dom-mlcontext-contexttype-slot④">[[contextType]]</a></code>, <var>context</var>.<code class="idl"><a data-link-type="idl" href="#dom-mlcontext-devicetype-slot" id="ref-for-dom-mlcontext-devicetype-slot④">[[deviceType]]</a></code> and <var>context</var>.<code class="idl"><a data-link-type="idl" href="#dom-mlcontext-powerpreference-slot" id="ref-for-dom-mlcontext-powerpreference-slot④">[[powerPreference]]</a></code>, return <code>false</code>.</p>
-    <li data-md>
-     <p>Return <code>true</code>;</p>
-   </ol>
+   <details class="algorithm" data-algorithm="validate MLContext" open>
+    <summary> To <dfn class="dfn-paneled" data-dfn-type="dfn" data-noexport id="validate-mlcontext">validate MLContext</dfn>, given <var>context</var>, run these steps: </summary>
+    <div class="algorithm-steps">
+     <ol>
+      <li data-md>
+       <p>If <var>context</var>.<code class="idl"><a data-link-type="idl" href="#dom-mlcontext-contexttype-slot" id="ref-for-dom-mlcontext-contexttype-slot③">[[contextType]]</a></code> is not "<a data-link-type="dfn" href="#webgpu-context" id="ref-for-webgpu-context①">webgpu</a>" or "<a data-link-type="dfn" href="#default-context" id="ref-for-default-context②">default</a>", return false.</p>
+      <li data-md>
+       <p>If <var>context</var>.<code class="idl"><a data-link-type="idl" href="#dom-mlcontext-devicetype-slot" id="ref-for-dom-mlcontext-devicetype-slot③">[[deviceType]]</a></code> is not "<a data-link-type="dfn" href="#device-type-cpu" id="ref-for-device-type-cpu①">cpu</a>" or "<a data-link-type="dfn" href="#device-type-gpu" id="ref-for-device-type-gpu②">gpu</a>", return false.</p>
+      <li data-md>
+       <p>If <var>context</var>.<code class="idl"><a data-link-type="idl" href="#dom-mlcontext-powerpreference-slot" id="ref-for-dom-mlcontext-powerpreference-slot③">[[powerPreference]]</a></code> is not "<a data-link-type="dfn" href="#power-preference-default" id="ref-for-power-preference-default②">default</a>" or "<a data-link-type="dfn" href="#power-preference-high-performance" id="ref-for-power-preference-high-performance">high-performance</a>" or "<a data-link-type="dfn" href="#power-preference-low-power" id="ref-for-power-preference-low-power">low-power</a>", return false.</p>
+      <li data-md>
+       <p>If the user agent cannot support <var>context</var>.<code class="idl"><a data-link-type="idl" href="#dom-mlcontext-contexttype-slot" id="ref-for-dom-mlcontext-contexttype-slot④">[[contextType]]</a></code>, <var>context</var>.<code class="idl"><a data-link-type="idl" href="#dom-mlcontext-devicetype-slot" id="ref-for-dom-mlcontext-devicetype-slot④">[[deviceType]]</a></code> and <var>context</var>.<code class="idl"><a data-link-type="idl" href="#dom-mlcontext-powerpreference-slot" id="ref-for-dom-mlcontext-powerpreference-slot④">[[powerPreference]]</a></code>, return false.</p>
+      <li data-md>
+       <p>Return true;</p>
+     </ol>
+    </div>
+   </details>
    <h4 class="heading settled" data-level="7.4.2" id="api-mlcontext-sync-execution"><span class="secno">7.4.2. </span><span class="content">Synchronous Execution</span><a class="self-link" href="#api-mlcontext-sync-execution"></a></h4>
-    Synchronously carries out the computational workload of a compiled graph <code class="idl"><a data-link-type="idl" href="#mlgraph" id="ref-for-mlgraph①①">MLGraph</a></code> on the calling thread, which must be a worker thread, to produce results as defined by the operations in the graph. This method of execution requires an <code class="idl"><a data-link-type="idl" href="#mlcontext" id="ref-for-mlcontext②⑨">MLContext</a></code> created with <code class="idl"><a data-link-type="idl" href="#dictdef-mlcontextoptions" id="ref-for-dictdef-mlcontextoptions⑤">MLContextOptions</a></code>. Otherwise, it throws an "<code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#operationerror" id="ref-for-operationerror①">OperationError</a></code>" <code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#idl-DOMException" id="ref-for-idl-DOMException⑦">DOMException</a></code>. 
-<pre class="idl highlight def"><c- b>partial</c-> <c- b>interface</c-> <a class="idl-code" data-link-type="interface" href="#mlcontext" id="ref-for-mlcontext③⓪"><c- g>MLContext</c-></a> {
+    Synchronously carries out the computational workload of a compiled graph <code class="idl"><a data-link-type="idl" href="#mlgraph" id="ref-for-mlgraph①①">MLGraph</a></code> on the calling thread, which must be a worker thread, to produce results as defined by the operations in the graph. This method of execution requires an <code class="idl"><a data-link-type="idl" href="#mlcontext" id="ref-for-mlcontext②⑧">MLContext</a></code> created with <code class="idl"><a data-link-type="idl" href="#dictdef-mlcontextoptions" id="ref-for-dictdef-mlcontextoptions⑤">MLContextOptions</a></code>. Otherwise, it<a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#dfn-throw" id="ref-for-dfn-throw⑥">throws</a> an "<code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#operationerror" id="ref-for-operationerror②">OperationError</a></code>" <code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#idl-DOMException" id="ref-for-idl-DOMException①⓪">DOMException</a></code>. 
+<pre class="idl highlight def"><c- b>partial</c-> <c- b>interface</c-> <a class="idl-code" data-link-type="interface" href="#mlcontext" id="ref-for-mlcontext②⑨"><c- g>MLContext</c-></a> {
   [<a class="idl-code" data-link-type="extended-attribute" href="https://webidl.spec.whatwg.org/#Exposed" id="ref-for-Exposed⑦"><c- g>Exposed</c-></a>=(<c- n>DedicatedWorker</c->)]
-  <a class="idl-code" data-link-type="interface" href="https://webidl.spec.whatwg.org/#idl-undefined" id="ref-for-idl-undefined"><c- b>undefined</c-></a> <dfn class="dfn-paneled idl-code" data-dfn-for="MLContext" data-dfn-type="method" data-export data-lt="computeSync(graph, inputs, outputs)" id="dom-mlcontext-computesync"><code><c- g>computeSync</c-></code></dfn>(
+  <a class="idl-code" data-link-type="interface" href="https://webidl.spec.whatwg.org/#idl-undefined" id="ref-for-idl-undefined"><c- b>undefined</c-></a> <a class="idl-code" data-link-type="method" href="#dom-mlcontext-computesync" id="ref-for-dom-mlcontext-computesync②"><c- g>computeSync</c-></a>(
       <a data-link-type="idl-name" href="#mlgraph" id="ref-for-mlgraph①②"><c- n>MLGraph</c-></a> <dfn class="idl-code" data-dfn-for="MLContext/computeSync(graph, inputs, outputs)" data-dfn-type="argument" data-export id="dom-mlcontext-computesync-graph-inputs-outputs-graph"><code><c- g>graph</c-></code><a class="self-link" href="#dom-mlcontext-computesync-graph-inputs-outputs-graph"></a></dfn>, <a data-link-type="idl-name" href="#typedefdef-mlnamedarraybufferviews" id="ref-for-typedefdef-mlnamedarraybufferviews②"><c- n>MLNamedArrayBufferViews</c-></a> <dfn class="idl-code" data-dfn-for="MLContext/computeSync(graph, inputs, outputs)" data-dfn-type="argument" data-export id="dom-mlcontext-computesync-graph-inputs-outputs-inputs"><code><c- g>inputs</c-></code><a class="self-link" href="#dom-mlcontext-computesync-graph-inputs-outputs-inputs"></a></dfn>, <a data-link-type="idl-name" href="#typedefdef-mlnamedarraybufferviews" id="ref-for-typedefdef-mlnamedarraybufferviews③"><c- n>MLNamedArrayBufferViews</c-></a> <dfn class="idl-code" data-dfn-for="MLContext/computeSync(graph, inputs, outputs)" data-dfn-type="argument" data-export id="dom-mlcontext-computesync-graph-inputs-outputs-outputs"><code><c- g>outputs</c-></code><a class="self-link" href="#dom-mlcontext-computesync-graph-inputs-outputs-outputs"></a></dfn>);
 };
 </pre>
-   <div class="algorithm" data-algorithm="mlcontext.computesync">
-    <p><strong>Arguments:</strong></p>
+   <div>
+     <strong>Arguments:</strong> 
     <ul>
      <li data-md>
       <p><em>graph</em>: an <code class="idl"><a data-link-type="idl" href="#mlgraph" id="ref-for-mlgraph①③">MLGraph</a></code>. The compiled graph to be executed.</p>
@@ -1603,104 +2021,118 @@ <h4 class="heading settled" data-level="7.4.2" id="api-mlcontext-sync-execution"
       <p><em>outputs</em>: an <code class="idl"><a data-link-type="idl" href="#typedefdef-mlnamedarraybufferviews" id="ref-for-typedefdef-mlnamedarraybufferviews⑤">MLNamedArrayBufferViews</a></code>. The pre-allocated resources of required outputs.</p>
     </ul>
     <p><strong>Returns:</strong> <code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#idl-undefined" id="ref-for-idl-undefined①">undefined</a></code>.</p>
-    <p>The <code class="idl"><a data-link-type="idl" href="#dom-mlcontext-computesync" id="ref-for-dom-mlcontext-computesync②">computeSync(graph, inputs, outputs)</a></code> method steps are:</p>
-    <ol>
-     <li data-md>
-      <p>If <var>graph</var>.<code class="idl"><a data-link-type="idl" href="#dom-mlgraph-context-slot" id="ref-for-dom-mlgraph-context-slot">[[context]]</a></code>.<code class="idl"><a data-link-type="idl" href="#dom-mlcontext-contexttype-slot" id="ref-for-dom-mlcontext-contexttype-slot⑤">[[contextType]]</a></code> is not "<a data-link-type="dfn" href="#default-context" id="ref-for-default-context③">default</a>, throw an "<code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#operationerror" id="ref-for-operationerror②">OperationError</a></code>" <code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#idl-DOMException" id="ref-for-idl-DOMException⑧">DOMException</a></code> and stop.</p>
-     <li data-md>
-      <p>If invoking the <a data-link-type="dfn" href="#validate-graph-resources" id="ref-for-validate-graph-resources">validate graph resources</a> algorithm given <var>inputs</var> and <var>graph</var>.<code class="idl"><a data-link-type="idl" href="#dom-mlgraph-inputdescriptors-slot" id="ref-for-dom-mlgraph-inputdescriptors-slot">[[inputDescriptors]]</a></code> returns <code>false</code>, then throw a "<code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#dataerror" id="ref-for-dataerror">DataError</a></code>" <code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#idl-DOMException" id="ref-for-idl-DOMException⑨">DOMException</a></code> and stop.</p>
-     <li data-md>
-      <p>If invoking the <a data-link-type="dfn" href="#validate-graph-resources" id="ref-for-validate-graph-resources①">validate graph resources</a> algorithm given <var>outputs</var> and <var>graph</var>.<code class="idl"><a data-link-type="idl" href="#dom-mlgraph-outputdescriptors-slot" id="ref-for-dom-mlgraph-outputdescriptors-slot">[[outputDescriptors]]</a></code> returns <code>false</code>, then throw a "<code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#dataerror" id="ref-for-dataerror①">DataError</a></code>" <code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#idl-DOMException" id="ref-for-idl-DOMException①⓪">DOMException</a></code> and stop.</p>
-     <li data-md>
-      <p>Invoke <a data-link-type="dfn" href="#execute-graph" id="ref-for-execute-graph">execute graph</a> given <var>graph</var>, <var>inputs</var> and <var>outputs</var>.</p>
-     <li data-md>
-      <p>If that throws an error, re-throw the error and stop.</p>
-     <li data-md>
-      <p>Return <code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#idl-undefined" id="ref-for-idl-undefined②">undefined</a></code>.</p>
-    </ol>
-    <div class="validusage">
-      To <dfn class="dfn-paneled" data-dfn-type="dfn" data-noexport id="validate-graph-resources">validate graph resources</dfn>, given <var>resources</var> and <var>descriptors</var>, run the following steps: 
+   </div>
+   <details class="algorithm" data-algorithm="computeSync(graph, inputs, outputs)" data-algorithm-for="MLContext" open>
+    <summary> The <dfn class="dfn-paneled idl-code" data-dfn-for="MLContext" data-dfn-type="method" data-export id="dom-mlcontext-computesync"><code>computeSync(<var>graph</var>, <var>inputs</var>, <var>outputs</var>)</code></dfn> method steps are: </summary>
+    <div class="algorithm-steps">
+     <ol>
+      <li data-md>
+       <p>If <var>graph</var>.<code class="idl"><a data-link-type="idl" href="#dom-mlgraph-context-slot" id="ref-for-dom-mlgraph-context-slot">[[context]]</a></code>.<code class="idl"><a data-link-type="idl" href="#dom-mlcontext-contexttype-slot" id="ref-for-dom-mlcontext-contexttype-slot⑤">[[contextType]]</a></code> is not "<a data-link-type="dfn" href="#default-context" id="ref-for-default-context③">default</a>", <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#dfn-throw" id="ref-for-dfn-throw⑦">throw</a> an "<code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#operationerror" id="ref-for-operationerror③">OperationError</a></code>" <code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#idl-DOMException" id="ref-for-idl-DOMException①①">DOMException</a></code>.</p>
+      <li data-md>
+       <p>If <a data-link-type="dfn" href="#validate-graph-resources" id="ref-for-validate-graph-resources">validating graph resources</a> given <var>inputs</var> and <var>graph</var>.<code class="idl"><a data-link-type="idl" href="#dom-mlgraph-inputdescriptors-slot" id="ref-for-dom-mlgraph-inputdescriptors-slot">[[inputDescriptors]]</a></code> returns false, then <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#dfn-throw" id="ref-for-dfn-throw⑧">throw</a> a "<code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#dataerror" id="ref-for-dataerror">DataError</a></code>" <code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#idl-DOMException" id="ref-for-idl-DOMException①②">DOMException</a></code>.</p>
+      <li data-md>
+       <p>If <a data-link-type="dfn" href="#validate-graph-resources" id="ref-for-validate-graph-resources①">validating graph resources</a> given <var>outputs</var> and <var>graph</var>.<code class="idl"><a data-link-type="idl" href="#dom-mlgraph-outputdescriptors-slot" id="ref-for-dom-mlgraph-outputdescriptors-slot">[[outputDescriptors]]</a></code> returns false, then <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#dfn-throw" id="ref-for-dfn-throw⑨">throw</a> a "<code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#dataerror" id="ref-for-dataerror①">DataError</a></code>" <code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#idl-DOMException" id="ref-for-idl-DOMException①③">DOMException</a></code>.</p>
+      <li data-md>
+       <p>Invoke <a data-link-type="dfn" href="#execute-graph" id="ref-for-execute-graph">execute graph</a> given <var>graph</var>, <var>inputs</var> and <var>outputs</var>.</p>
+      <li data-md>
+       <p>If that <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#dfn-throw" id="ref-for-dfn-throw①⓪">throws</a> an error, re-<a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#dfn-throw" id="ref-for-dfn-throw①①">throw</a> the error.</p>
+      <li data-md>
+       <p>Return <code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#idl-undefined" id="ref-for-idl-undefined②">undefined</a></code>.</p>
+     </ol>
+    </div>
+   </details>
+   <details class="algorithm" data-algorithm="validate graph resources" open>
+    <summary> To <dfn class="dfn-paneled" data-dfn-type="dfn" data-noexport id="validate-graph-resources">validate graph resources</dfn>, given <var>resources</var> and <var>descriptors</var>, run the following steps: </summary>
+    <div class="algorithm-steps">
      <ol>
       <li data-md>
-       <p><a data-link-type="dfn" href="https://infra.spec.whatwg.org/#assert" id="ref-for-assert">Assert</a>: the type of <var>resources</var> is <code class="idl"><a data-link-type="idl" href="#typedefdef-mlnamedarraybufferviews" id="ref-for-typedefdef-mlnamedarraybufferviews⑥">MLNamedArrayBufferViews</a></code>.</p>
+       <p><a data-link-type="dfn" href="https://infra.spec.whatwg.org/#assert" id="ref-for-assert⑤">Assert</a>: the type of <var>resources</var> is <code class="idl"><a data-link-type="idl" href="#typedefdef-mlnamedarraybufferviews" id="ref-for-typedefdef-mlnamedarraybufferviews⑥">MLNamedArrayBufferViews</a></code>.</p>
       <li data-md>
-       <p>For each <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#idl-record" id="ref-for-idl-record③">record</a> &lt;<var>key</var>, <var>value</var>> of <var>resources</var>:</p>
+       <p><a data-link-type="dfn" href="https://infra.spec.whatwg.org/#map-iterate" id="ref-for-map-iterate①">For each</a> <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#idl-record" id="ref-for-idl-record③">record</a> &lt;<var>key</var>, <var>value</var>> of <var>resources</var>:</p>
        <ol>
         <li data-md>
-         <p>If <var>descriptors</var>[<var>key</var>] does not <a data-link-type="dfn" href="https://infra.spec.whatwg.org/#map-exists" id="ref-for-map-exists②">exist</a>, return <code>false</code>.</p>
+         <p>If <var>descriptors</var>[<var>key</var>] does not <a data-link-type="dfn" href="https://infra.spec.whatwg.org/#map-exists" id="ref-for-map-exists④">exist</a>, return false.</p>
         <li data-md>
-         <p><a data-link-type="dfn" href="https://infra.spec.whatwg.org/#assert" id="ref-for-assert①">Assert</a>: the type of <var>value</var> is <code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#ArrayBufferView" id="ref-for-ArrayBufferView⑤">ArrayBufferView</a></code>.</p>
+         <p><a data-link-type="dfn" href="https://infra.spec.whatwg.org/#assert" id="ref-for-assert⑥">Assert</a>: the type of <var>value</var> is <code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#ArrayBufferView" id="ref-for-ArrayBufferView⑤">ArrayBufferView</a></code>.</p>
         <li data-md>
-         <p>If running the <a data-link-type="dfn" href="#validate-buffer-with-descriptor" id="ref-for-validate-buffer-with-descriptor">validate buffer with descriptor</a> given <var>value</var> and <var>descriptors</var>[<var>key</var>] return <code>false</code>, return <code>false</code>.</p>
+         <p>If <a data-link-type="dfn" href="#validate-buffer-with-descriptor" id="ref-for-validate-buffer-with-descriptor">validating buffer with descriptor</a> given <var>value</var> and <var>descriptors</var>[<var>key</var>] returns false, then return false.</p>
        </ol>
       <li data-md>
-       <p>Return <code>true</code>.</p>
+       <p>Return true.</p>
      </ol>
     </div>
-    <div class="validusage">
-      To <dfn class="dfn-paneled" data-dfn-type="dfn" data-noexport id="validate-buffer-with-descriptor">validate buffer with descriptor</dfn> given <var>bufferView</var> and <var>descriptor</var>, run the following steps: 
+   </details>
+   <details class="algorithm" data-algorithm="validate buffer with descriptor" open>
+    <summary> To <dfn class="dfn-paneled" data-dfn-type="dfn" data-noexport id="validate-buffer-with-descriptor">validate buffer with descriptor</dfn> given <var>bufferView</var> and <var>descriptor</var>, run the following steps: </summary>
+    <div class="algorithm-steps">
      <ol>
       <li data-md>
-       <p>If <var>bufferView</var> is not an <code class="idl"><a data-link-type="idl" href="#typedefdef-mlbufferview" id="ref-for-typedefdef-mlbufferview">MLBufferView</a></code>, return <code>false</code>.</p>
+       <p>If <var>bufferView</var> is not an <code class="idl"><a data-link-type="idl" href="#typedefdef-mlbufferview" id="ref-for-typedefdef-mlbufferview">MLBufferView</a></code>, return false.</p>
       <li data-md>
-       <p>If <var>bufferView</var>’s <a data-link-type="dfn" href="https://tc39.es/ecma262/#table-the-typedarray-constructors" id="ref-for-table-the-typedarray-constructors①">element type</a> does not match to <var>descriptor</var>.<code class="idl"><a data-link-type="idl" href="#dom-mloperanddescriptor-type" id="ref-for-dom-mloperanddescriptor-type①">type</a></code> according to <a href="#appendices-mloperandtype-arraybufferview-compatibility">this table</a>, return <code>false</code>.</p>
+       <p>If <var>bufferView</var>’s <a data-link-type="dfn" href="https://tc39.es/ecma262/#table-the-typedarray-constructors" id="ref-for-table-the-typedarray-constructors①">element type</a> does not match to <var>descriptor</var>.<code class="idl"><a data-link-type="idl" href="#dom-mloperanddescriptor-type" id="ref-for-dom-mloperanddescriptor-type②">type</a></code> according to <a href="#appendices-mloperandtype-arraybufferview-compatibility">this table</a>, return false.</p>
       <li data-md>
-       <p>If <var>bufferView</var>.[[ByteLength]] is not equal to the <a data-link-type="dfn" href="#mloperanddescriptor-byte-length" id="ref-for-mloperanddescriptor-byte-length">byte length</a> of <var>descriptor</var>, return <code>false</code>.</p>
+       <p>If <var>bufferView</var>.[[ByteLength]] is not equal to the <a data-link-type="dfn" href="#mloperanddescriptor-byte-length" id="ref-for-mloperanddescriptor-byte-length">byte length</a> of <var>descriptor</var>, return false.</p>
      </ol>
     </div>
-    <p>To <dfn class="dfn-paneled" data-dfn-type="dfn" data-noexport id="execute-graph">execute graph</dfn>, given <var>graph</var>, <var>inputs</var> and <var>outputs</var>, run the following steps:</p>
-    <ol>
-     <li data-md>
-      <p><a data-link-type="dfn" href="https://infra.spec.whatwg.org/#assert" id="ref-for-assert②">Assert</a>: the type of <var>inputs</var> is <code class="idl"><a data-link-type="idl" href="#typedefdef-mlnamedarraybufferviews" id="ref-for-typedefdef-mlnamedarraybufferviews⑦">MLNamedArrayBufferViews</a></code>.</p>
-     <li data-md>
-      <p>Let <var>inputResources</var> denote the input resources of <var>graph</var>.<code class="idl"><a data-link-type="idl" href="#dom-mlgraph-implementation-slot" id="ref-for-dom-mlgraph-implementation-slot">[[implementation]]</a></code>.</p>
-     <li data-md>
-      <p>For each &lt;<var>key</var>, <var>inputValue</var>> of <var>inputs</var>:</p>
-      <ol>
-       <li data-md>
-        <p>Let <var>inputDescriptor</var> be <var>graph</var>.<code class="idl"><a data-link-type="idl" href="#dom-mlgraph-inputdescriptors-slot" id="ref-for-dom-mlgraph-inputdescriptors-slot①">[[inputDescriptors]]</a></code>[<var>key</var>].</p>
-       <li data-md>
-        <p>Let <var>inputTensor</var> be a new tensor for <var>graph</var>.<code class="idl"><a data-link-type="idl" href="#dom-mlgraph-implementation-slot" id="ref-for-dom-mlgraph-implementation-slot①">[[implementation]]</a></code> as follows:</p>
-        <ol>
-         <li data-md>
-          <p>Set the data type of <var>inputTensor</var> to the one that matches the <a data-link-type="dfn" href="https://tc39.es/ecma262/#table-the-typedarray-constructors" id="ref-for-table-the-typedarray-constructors②">element type</a> of <var>inputValue</var>.</p>
-         <li data-md>
-          <p>Set the dimensions of <var>inputTensor</var> to <var>inputDescriptor</var>.<code class="idl"><a data-link-type="idl" href="#dom-mloperanddescriptor-dimensions" id="ref-for-dom-mloperanddescriptor-dimensions①">dimensions</a></code>.</p>
-         <li data-md>
-          <p>Set the values of elements in <var>inputTensor</var> to the values of elements in <var>inputValue</var>.</p>
-        </ol>
-       <li data-md>
-        <p>Request the underlying implementation of <var>graph</var> to bind <var>inputResources</var>[<var>key</var>] to <var>inputTensor</var>.</p>
-      </ol>
-     <li data-md>
-      <p><a data-link-type="dfn" href="https://infra.spec.whatwg.org/#assert" id="ref-for-assert③">Assert</a>: the type of <var>outputs</var> is <code class="idl"><a data-link-type="idl" href="#typedefdef-mlnamedarraybufferviews" id="ref-for-typedefdef-mlnamedarraybufferviews⑧">MLNamedArrayBufferViews</a></code>.</p>
-     <li data-md>
-      <p>For each &lt;<var>key</var>, <var>outputValue</var>> of <var>outputs</var>:</p>
-      <ol>
-       <li data-md>
-        <p>Issue a compute request to <var>graph</var>.<code class="idl"><a data-link-type="idl" href="#dom-mlgraph-implementation-slot" id="ref-for-dom-mlgraph-implementation-slot②">[[implementation]]</a></code> given <var>key</var> and <var>inputResources</var> and wait for completion.</p>
-        <ol>
-         <li data-md>
-          <p>If that returns an error, then throw an "<code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#operationerror" id="ref-for-operationerror③">OperationError</a></code>" <code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#idl-DOMException" id="ref-for-idl-DOMException①①">DOMException</a></code> and stop.</p>
-         <li data-md>
-          <p>Otherwise, store the result in <var>outputTensor</var>.</p>
-        </ol>
-       <li data-md>
-        <p>Let <var>outputDesc</var> be <var>graph</var>.<code class="idl"><a data-link-type="idl" href="#dom-mlgraph-outputdescriptors-slot" id="ref-for-dom-mlgraph-outputdescriptors-slot①">[[outputDescriptors]]</a></code>[<var>key</var>].</p>
-       <li data-md>
-        <p>If the byte length of <var>outputTensor</var> is not equal to the <a data-link-type="dfn" href="#mloperanddescriptor-byte-length" id="ref-for-mloperanddescriptor-byte-length①">byte length</a> of <var>outputDesc</var>, then throw a "<code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#dataerror" id="ref-for-dataerror②">DataError</a></code>" <code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#idl-DOMException" id="ref-for-idl-DOMException①②">DOMException</a></code> and stop.</p>
-       <li data-md>
-        <p>If the <a data-link-type="dfn" href="https://tc39.es/ecma262/#table-the-typedarray-constructors" id="ref-for-table-the-typedarray-constructors③">element type</a> of <var>outputTensor</var> doesn’t match the <a data-link-type="dfn" href="https://tc39.es/ecma262/#table-the-typedarray-constructors" id="ref-for-table-the-typedarray-constructors④">element type</a> of <var>outputValue</var>, then throw a "<code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#dataerror" id="ref-for-dataerror③">DataError</a></code>" <code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#idl-DOMException" id="ref-for-idl-DOMException①③">DOMException</a></code> and stop.</p>
-       <li data-md>
-        <p>Request the underlying implementation of <var>graph</var> to set the values of elements in <var>outputValue</var> to the values of elements in <var>outputTensor</var>.</p>
-      </ol>
-     <li data-md>
-      <p>Return <code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#idl-undefined" id="ref-for-idl-undefined③">undefined</a></code>.</p>
-    </ol>
-   </div>
+   </details>
+   <details class="algorithm" data-algorithm="execute graph" open>
+    <summary> To <dfn class="dfn-paneled" data-dfn-type="dfn" data-noexport id="execute-graph">execute graph</dfn>, given <var>graph</var>, <var>inputs</var> and <var>outputs</var>, run the following steps: </summary>
+    <div class="algorithm-steps">
+     <ol>
+      <li data-md>
+       <p><a data-link-type="dfn" href="https://infra.spec.whatwg.org/#assert" id="ref-for-assert⑦">Assert</a>: the type of <var>inputs</var> is <code class="idl"><a data-link-type="idl" href="#typedefdef-mlnamedarraybufferviews" id="ref-for-typedefdef-mlnamedarraybufferviews⑦">MLNamedArrayBufferViews</a></code>.</p>
+      <li data-md>
+       <p>Let <var>inputResources</var> denote the input resources of <var>graph</var>.<code class="idl"><a data-link-type="idl" href="#dom-mlgraph-implementation-slot" id="ref-for-dom-mlgraph-implementation-slot">[[implementation]]</a></code>.</p>
+      <li data-md>
+       <p><a data-link-type="dfn" href="https://infra.spec.whatwg.org/#map-iterate" id="ref-for-map-iterate②">For each</a> &lt;<var>key</var>, <var>inputValue</var>> of <var>inputs</var>:</p>
+       <ol>
+        <li data-md>
+         <p>Let <var>inputDescriptor</var> be <var>graph</var>.<code class="idl"><a data-link-type="idl" href="#dom-mlgraph-inputdescriptors-slot" id="ref-for-dom-mlgraph-inputdescriptors-slot①">[[inputDescriptors]]</a></code>[<var>key</var>].</p>
+        <li data-md>
+         <p>Let <var>inputTensor</var> be a new tensor for <var>graph</var>.<code class="idl"><a data-link-type="idl" href="#dom-mlgraph-implementation-slot" id="ref-for-dom-mlgraph-implementation-slot①">[[implementation]]</a></code> as follows:</p>
+         <ol>
+          <li data-md>
+           <p>Set the data type of <var>inputTensor</var> to the one that matches the <a data-link-type="dfn" href="https://tc39.es/ecma262/#table-the-typedarray-constructors" id="ref-for-table-the-typedarray-constructors②">element type</a> of <var>inputValue</var>.</p>
+          <li data-md>
+           <p>Set the dimensions of <var>inputTensor</var> to <var>inputDescriptor</var>.<code class="idl"><a data-link-type="idl" href="#dom-mloperanddescriptor-dimensions" id="ref-for-dom-mloperanddescriptor-dimensions④">dimensions</a></code>.</p>
+          <li data-md>
+           <p>Set the values of elements in <var>inputTensor</var> to the values of elements in <var>inputValue</var>.</p>
+         </ol>
+        <li data-md>
+         <p>Request the underlying implementation of <var>graph</var> to bind <var>inputResources</var>[<var>key</var>] to <var>inputTensor</var>.</p>
+       </ol>
+      <li data-md>
+       <p><a data-link-type="dfn" href="https://infra.spec.whatwg.org/#assert" id="ref-for-assert⑧">Assert</a>: the type of <var>outputs</var> is <code class="idl"><a data-link-type="idl" href="#typedefdef-mlnamedarraybufferviews" id="ref-for-typedefdef-mlnamedarraybufferviews⑧">MLNamedArrayBufferViews</a></code>.</p>
+      <li data-md>
+       <p><a data-link-type="dfn" href="https://infra.spec.whatwg.org/#map-iterate" id="ref-for-map-iterate③">For each</a> &lt;<var>key</var>, <var>outputValue</var>> of <var>outputs</var>:</p>
+       <ol>
+        <li data-md>
+         <p>Issue a compute request to <var>graph</var>.<code class="idl"><a data-link-type="idl" href="#dom-mlgraph-implementation-slot" id="ref-for-dom-mlgraph-implementation-slot②">[[implementation]]</a></code> given <var>key</var> and <var>inputResources</var> and wait for completion.</p>
+         <ol>
+          <li data-md>
+           <p>If that returns an error, then <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#dfn-throw" id="ref-for-dfn-throw①②">throw</a> an "<code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#operationerror" id="ref-for-operationerror④">OperationError</a></code>" <code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#idl-DOMException" id="ref-for-idl-DOMException①④">DOMException</a></code>.</p>
+          <li data-md>
+           <p>Otherwise, store the result in <var>outputTensor</var>.</p>
+         </ol>
+        <li data-md>
+         <p>Let <var>outputDesc</var> be <var>graph</var>.<code class="idl"><a data-link-type="idl" href="#dom-mlgraph-outputdescriptors-slot" id="ref-for-dom-mlgraph-outputdescriptors-slot①">[[outputDescriptors]]</a></code>[<var>key</var>].</p>
+        <li data-md>
+         <p>If the byte length of <var>outputTensor</var> is not equal to the <a data-link-type="dfn" href="#mloperanddescriptor-byte-length" id="ref-for-mloperanddescriptor-byte-length①">byte length</a> of <var>outputDesc</var>, then <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#dfn-throw" id="ref-for-dfn-throw①③">throw</a> a "<code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#dataerror" id="ref-for-dataerror②">DataError</a></code>" <code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#idl-DOMException" id="ref-for-idl-DOMException①⑤">DOMException</a></code>.</p>
+        <li data-md>
+         <p>If the <a data-link-type="dfn" href="https://tc39.es/ecma262/#table-the-typedarray-constructors" id="ref-for-table-the-typedarray-constructors③">element type</a> of <var>outputTensor</var> doesn’t match the <a data-link-type="dfn" href="https://tc39.es/ecma262/#table-the-typedarray-constructors" id="ref-for-table-the-typedarray-constructors④">element type</a> of <var>outputValue</var>, then <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#dfn-throw" id="ref-for-dfn-throw①④">throw</a> a "<code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#dataerror" id="ref-for-dataerror③">DataError</a></code>" <code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#idl-DOMException" id="ref-for-idl-DOMException①⑥">DOMException</a></code>.</p>
+        <li data-md>
+         <p>Request the underlying implementation of <var>graph</var> to set the values of elements in <var>outputValue</var> to the values of elements in <var>outputTensor</var>.</p>
+       </ol>
+      <li data-md>
+       <p>Return <code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#idl-undefined" id="ref-for-idl-undefined③">undefined</a></code>.</p>
+     </ol>
+    </div>
+   </details>
    <h5 class="heading settled" data-level="7.4.2.1" id="api-mlcontext-sync-execution-examples"><span class="secno">7.4.2.1. </span><span class="content">Examples</span><a class="self-link" href="#api-mlcontext-sync-execution-examples"></a></h5>
    <div class="example" id="example-007d60fb">
-    <a class="self-link" href="#example-007d60fb"></a> The following code showcases the synchronous computation with optional outputs in a worker. 
+    <a class="self-link" href="#example-007d60fb"></a> 
+    <details open>
+     <summary> The following code showcases the synchronous computation with optional outputs in a worker. </summary>
 <pre class="highlight"><c- a>const</c-> context <c- o>=</c-> navigator<c- p>.</c->ml<c- p>.</c->createContextSync<c- p>();</c->
 
 <c- c1>// Build a graph with two outputs.</c->
@@ -1730,43 +2162,48 @@ <h5 class="heading settled" data-level="7.4.2.1" id="api-mlcontext-sync-executio
 context<c- p>.</c->computeSync<c- p>(</c->graph<c- p>,</c-> inputs<c- p>,</c-> <c- p>{</c-><c- t>'e'</c-><c- o>:</c-> bufferE<c- p>});</c->
 console<c- p>.</c->log<c- p>(</c-><c- sb>`values: </c-><c- si>${</c->bufferE<c- si>}</c-><c- sb>`</c-><c- p>);</c->
 </pre>
+    </details>
    </div>
    <h4 class="heading settled" data-level="7.4.3" id="mlnamedarraybufferviews-transfer-alg"><span class="secno">7.4.3. </span><span class="content">The <code class="idl"><a data-link-type="idl" href="#typedefdef-mlnamedarraybufferviews" id="ref-for-typedefdef-mlnamedarraybufferviews⑨">MLNamedArrayBufferViews</a></code> transfer algorithm</span><a class="self-link" href="#mlnamedarraybufferviews-transfer-alg"></a></h4>
-    To <dfn class="dfn-paneled" data-dfn-for="MLNamedArrayBufferViews" data-dfn-type="dfn" data-noexport id="mlnamedarraybufferviews-transfer">transfer</dfn> an <code class="idl"><a data-link-type="idl" href="#typedefdef-mlnamedarraybufferviews" id="ref-for-typedefdef-mlnamedarraybufferviews①⓪">MLNamedArrayBufferViews</a></code> <var>views</var>: 
-   <ol>
-    <li data-md>
-     <p>Let <var>transferredViews</var> be a new <code class="idl"><a data-link-type="idl" href="#typedefdef-mlnamedarraybufferviews" id="ref-for-typedefdef-mlnamedarraybufferviews①①">MLNamedArrayBufferViews</a></code>.</p>
-    <li data-md>
-     <p>For each <var>key</var> -> <var>value</var> of <var>views</var>:</p>
+   <details class="algorithm" data-algorithm="transfer" data-algorithm-for="MLNamedArrayBufferViews" open>
+    <summary> To <dfn class="dfn-paneled" data-dfn-for="MLNamedArrayBufferViews" data-dfn-type="dfn" data-noexport id="mlnamedarraybufferviews-transfer">transfer</dfn> an <code class="idl"><a data-link-type="idl" href="#typedefdef-mlnamedarraybufferviews" id="ref-for-typedefdef-mlnamedarraybufferviews①⓪">MLNamedArrayBufferViews</a></code> <var>views</var>: </summary>
+    <div class="algorithm-steps">
      <ol>
       <li data-md>
-       <p>Let <var>transferredBuffer</var> be the result of <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#arraybuffer-transfer" id="ref-for-arraybuffer-transfer">transferring</a> the <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#buffersource-underlying-buffer" id="ref-for-buffersource-underlying-buffer">underlying buffer</a> of <var>value</var>.</p>
-      <li data-md>
-       <p>Let <var>constructor</var> be the appropriate <a data-link-type="dfn" href="https://tc39.es/ecma262/#table-the-typedarray-constructors" id="ref-for-table-the-typedarray-constructors⑤">view constructor</a> for the type of <code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#ArrayBufferView" id="ref-for-ArrayBufferView⑥">ArrayBufferView</a></code> <var>value</var>.</p>
+       <p>Let <var>transferredViews</var> be a new <code class="idl"><a data-link-type="idl" href="#typedefdef-mlnamedarraybufferviews" id="ref-for-typedefdef-mlnamedarraybufferviews①①">MLNamedArrayBufferViews</a></code>.</p>
       <li data-md>
-       <p>Let <var>elementsNumber</var> be the result of the <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#buffersource-byte-length" id="ref-for-buffersource-byte-length">byte length</a> of <var>value</var> ÷ <a data-link-type="dfn" href="https://tc39.es/ecma262/#table-the-typedarray-constructors" id="ref-for-table-the-typedarray-constructors⑥">element size</a> of <var>value</var>.</p>
-      <li data-md>
-       <p>Let <var>transferredView</var> be <a data-link-type="dfn" href="https://tc39.es/ecma262/#sec-construct" id="ref-for-sec-construct">Construct</a>(<var>constructor</var>, <var>transferredBuffer</var>, <var>value</var>.[[ByteOffset]], <var>elementsNumber</var>).</p>
+       <p><a data-link-type="dfn" href="https://infra.spec.whatwg.org/#map-iterate" id="ref-for-map-iterate④">For each</a> <var>key</var> → <var>value</var> of <var>views</var>:</p>
+       <ol>
+        <li data-md>
+         <p>Let <var>transferredBuffer</var> be the result of <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#arraybuffer-transfer" id="ref-for-arraybuffer-transfer">transferring</a> the <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#buffersource-underlying-buffer" id="ref-for-buffersource-underlying-buffer">underlying buffer</a> of <var>value</var>.</p>
+        <li data-md>
+         <p>Let <var>constructor</var> be the appropriate <a data-link-type="dfn" href="https://tc39.es/ecma262/#table-the-typedarray-constructors" id="ref-for-table-the-typedarray-constructors⑤">view constructor</a> for the type of <code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#ArrayBufferView" id="ref-for-ArrayBufferView⑥">ArrayBufferView</a></code> <var>value</var>.</p>
+        <li data-md>
+         <p>Let <var>elementsNumber</var> be the result of the <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#buffersource-byte-length" id="ref-for-buffersource-byte-length">byte length</a> of <var>value</var> ÷ <a data-link-type="dfn" href="https://tc39.es/ecma262/#table-the-typedarray-constructors" id="ref-for-table-the-typedarray-constructors⑥">element size</a> of <var>value</var>.</p>
+        <li data-md>
+         <p>Let <var>transferredView</var> be <a data-link-type="abstract-op" href="https://tc39.es/ecma262/multipage/abstract-operations.html#sec-construct" id="ref-for-sec-construct">Construct</a>(<var>constructor</var>, <var>transferredBuffer</var>, <var>value</var>.[[ByteOffset]], <var>elementsNumber</var>).</p>
+        <li data-md>
+         <p>Set <var>transferredViews</var>[<var>key</var>] to <var>transferredView</var>.</p>
+       </ol>
       <li data-md>
-       <p>Set <var>transferredViews</var>[<var>key</var>] to <var>transferredView</var>.</p>
+       <p>Return <var>transferredViews</var>.</p>
      </ol>
-    <li data-md>
-     <p>Return <var>transferredViews</var>.</p>
-   </ol>
+    </div>
+   </details>
    <h4 class="heading settled" data-level="7.4.4" id="api-mlcontext-async-execution"><span class="secno">7.4.4. </span><span class="content">Asynchronous Execution</span><a class="self-link" href="#api-mlcontext-async-execution"></a></h4>
-    Asynchronously carries out the computational workload of a compiled graph <code class="idl"><a data-link-type="idl" href="#mlgraph" id="ref-for-mlgraph①④">MLGraph</a></code> on a separate timeline, either on a worker thread for the CPU execution, or on a GPU timeline for the submission of GPU workload on the command queue. The asynchronous nature of this call avoids blocking the calling thread while the computation for result is ongoing. This method of execution requires an <code class="idl"><a data-link-type="idl" href="#mlcontext" id="ref-for-mlcontext③①">MLContext</a></code> created with <code class="idl"><a data-link-type="idl" href="#dictdef-mlcontextoptions" id="ref-for-dictdef-mlcontextoptions⑥">MLContextOptions</a></code>. Otherwise, it throws an "<code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#operationerror" id="ref-for-operationerror④">OperationError</a></code>" <code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#idl-DOMException" id="ref-for-idl-DOMException①④">DOMException</a></code>. 
+    Asynchronously carries out the computational workload of a compiled graph <code class="idl"><a data-link-type="idl" href="#mlgraph" id="ref-for-mlgraph①④">MLGraph</a></code> on a separate timeline, either on a worker thread for the CPU execution, or on a GPU timeline for the submission of GPU workload on the command queue. The asynchronous nature of this call avoids blocking the calling thread while the computation for result is ongoing. This method of execution requires an <code class="idl"><a data-link-type="idl" href="#mlcontext" id="ref-for-mlcontext③⓪">MLContext</a></code> created with <code class="idl"><a data-link-type="idl" href="#dictdef-mlcontextoptions" id="ref-for-dictdef-mlcontextoptions⑥">MLContextOptions</a></code>. Otherwise, it<a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#dfn-throw" id="ref-for-dfn-throw①⑤">throws</a> an "<code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#operationerror" id="ref-for-operationerror⑤">OperationError</a></code>" <code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#idl-DOMException" id="ref-for-idl-DOMException①⑦">DOMException</a></code>. 
    <div class="note" role="note"> In accordance with the <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#arraybufferview-write" id="ref-for-arraybufferview-write">Web IDL warning</a>, to prevent the calling thread from modifying the input and output resources while the computation is ongoing, this method <a data-link-type="dfn" href="#mlnamedarraybufferviews-transfer" id="ref-for-mlnamedarraybufferviews-transfer">transfers</a> the input and output <code class="idl"><a data-link-type="idl" href="#typedefdef-mlnamedarraybufferviews" id="ref-for-typedefdef-mlnamedarraybufferviews①②">MLNamedArrayBufferViews</a></code> to new views that share the same backing memory allocations. The transferred views are returned to the caller via the promise fulfillment with the computation result written into the backing memory of the output views. </div>
 <pre class="idl highlight def"><c- b>dictionary</c-> <dfn class="dfn-paneled idl-code" data-dfn-type="dictionary" data-export id="dictdef-mlcomputeresult"><code><c- g>MLComputeResult</c-></code></dfn> {
   <a data-link-type="idl-name" href="#typedefdef-mlnamedarraybufferviews" id="ref-for-typedefdef-mlnamedarraybufferviews①③"><c- n>MLNamedArrayBufferViews</c-></a> <dfn class="dfn-paneled idl-code" data-dfn-for="MLComputeResult" data-dfn-type="dict-member" data-export data-type="MLNamedArrayBufferViews " id="dom-mlcomputeresult-inputs"><code><c- g>inputs</c-></code></dfn>;
   <a data-link-type="idl-name" href="#typedefdef-mlnamedarraybufferviews" id="ref-for-typedefdef-mlnamedarraybufferviews①④"><c- n>MLNamedArrayBufferViews</c-></a> <dfn class="dfn-paneled idl-code" data-dfn-for="MLComputeResult" data-dfn-type="dict-member" data-export data-type="MLNamedArrayBufferViews " id="dom-mlcomputeresult-outputs"><code><c- g>outputs</c-></code></dfn>;
 };
 
-<c- b>partial</c-> <c- b>interface</c-> <a class="idl-code" data-link-type="interface" href="#mlcontext" id="ref-for-mlcontext③②"><c- g>MLContext</c-></a> {
-  <a class="idl-code" data-link-type="interface" href="https://webidl.spec.whatwg.org/#idl-promise" id="ref-for-idl-promise③"><c- b>Promise</c-></a>&lt;<a data-link-type="idl-name" href="#dictdef-mlcomputeresult" id="ref-for-dictdef-mlcomputeresult"><c- n>MLComputeResult</c-></a>> <dfn class="dfn-paneled idl-code" data-dfn-for="MLContext" data-dfn-type="method" data-export data-lt="compute(graph, inputs, outputs)" id="dom-mlcontext-compute"><code><c- g>compute</c-></code></dfn>(
+<c- b>partial</c-> <c- b>interface</c-> <a class="idl-code" data-link-type="interface" href="#mlcontext" id="ref-for-mlcontext③①"><c- g>MLContext</c-></a> {
+  <a class="idl-code" data-link-type="interface" href="https://webidl.spec.whatwg.org/#idl-promise" id="ref-for-idl-promise③"><c- b>Promise</c-></a>&lt;<a data-link-type="idl-name" href="#dictdef-mlcomputeresult" id="ref-for-dictdef-mlcomputeresult"><c- n>MLComputeResult</c-></a>> <a class="idl-code" data-link-type="method" href="#dom-mlcontext-compute" id="ref-for-dom-mlcontext-compute④"><c- g>compute</c-></a>(
       <a data-link-type="idl-name" href="#mlgraph" id="ref-for-mlgraph①⑤"><c- n>MLGraph</c-></a> <dfn class="idl-code" data-dfn-for="MLContext/compute(graph, inputs, outputs)" data-dfn-type="argument" data-export id="dom-mlcontext-compute-graph-inputs-outputs-graph"><code><c- g>graph</c-></code><a class="self-link" href="#dom-mlcontext-compute-graph-inputs-outputs-graph"></a></dfn>, <a data-link-type="idl-name" href="#typedefdef-mlnamedarraybufferviews" id="ref-for-typedefdef-mlnamedarraybufferviews①⑤"><c- n>MLNamedArrayBufferViews</c-></a> <dfn class="idl-code" data-dfn-for="MLContext/compute(graph, inputs, outputs)" data-dfn-type="argument" data-export id="dom-mlcontext-compute-graph-inputs-outputs-inputs"><code><c- g>inputs</c-></code><a class="self-link" href="#dom-mlcontext-compute-graph-inputs-outputs-inputs"></a></dfn>, <a data-link-type="idl-name" href="#typedefdef-mlnamedarraybufferviews" id="ref-for-typedefdef-mlnamedarraybufferviews①⑥"><c- n>MLNamedArrayBufferViews</c-></a> <dfn class="idl-code" data-dfn-for="MLContext/compute(graph, inputs, outputs)" data-dfn-type="argument" data-export id="dom-mlcontext-compute-graph-inputs-outputs-outputs"><code><c- g>outputs</c-></code><a class="self-link" href="#dom-mlcontext-compute-graph-inputs-outputs-outputs"></a></dfn>);
 };
 </pre>
-   <div class="algorithm" data-algorithm="mlcontext.compute">
+   <div>
      <strong>Arguments:</strong> 
     <ul>
      <li data-md>
@@ -1777,45 +2214,51 @@ <h4 class="heading settled" data-level="7.4.4" id="api-mlcontext-async-execution
       <p><em>outputs</em>: an <code class="idl"><a data-link-type="idl" href="#typedefdef-mlnamedarraybufferviews" id="ref-for-typedefdef-mlnamedarraybufferviews①⑧">MLNamedArrayBufferViews</a></code>. The pre-allocated resources of required outputs. Will be <a data-link-type="dfn" href="#mlnamedarraybufferviews-transfer" id="ref-for-mlnamedarraybufferviews-transfer②">transferred</a> if there are no validation errors.</p>
     </ul>
     <p><strong>Returns:</strong> Promise&lt;<code class="idl"><a data-link-type="idl" href="#dictdef-mlcomputeresult" id="ref-for-dictdef-mlcomputeresult①">MLComputeResult</a></code>>.</p>
-    <p>The <code class="idl"><a data-link-type="idl" href="#dom-mlcontext-compute" id="ref-for-dom-mlcontext-compute④">compute(graph, inputs, outputs)</a></code> method steps are:</p>
-    <ol>
-     <li data-md>
-      <p>Let <var>promise</var> be <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#a-new-promise" id="ref-for-a-new-promise②">a new promise</a>.</p>
-     <li data-md>
-      <p>Return <var>promise</var> and run the following steps <a data-link-type="dfn" href="https://html.spec.whatwg.org/multipage/infrastructure.html#in-parallel" id="ref-for-in-parallel①">in parallel</a>:</p>
-      <ol>
-       <li data-md>
-        <p>If <var>graph</var>.<code class="idl"><a data-link-type="idl" href="#dom-mlgraph-context-slot" id="ref-for-dom-mlgraph-context-slot①">[[context]]</a></code>.<code class="idl"><a data-link-type="idl" href="#dom-mlcontext-contexttype-slot" id="ref-for-dom-mlcontext-contexttype-slot⑥">[[contextType]]</a></code> is not "<a data-link-type="dfn" href="#default-context" id="ref-for-default-context④">default</a>, <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#reject" id="ref-for-reject②">reject</a> <var>promise</var> with an "<code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#operationerror" id="ref-for-operationerror⑤">OperationError</a></code>" <code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#idl-DOMException" id="ref-for-idl-DOMException①⑤">DOMException</a></code> and stop.</p>
-       <li data-md>
-        <p>If invoking the <a data-link-type="dfn" href="#validate-graph-resources" id="ref-for-validate-graph-resources②">validate graph resources</a> algorithm given <var>inputs</var> and <var>graph</var>.<code class="idl"><a data-link-type="idl" href="#dom-mlgraph-inputdescriptors-slot" id="ref-for-dom-mlgraph-inputdescriptors-slot②">[[inputDescriptors]]</a></code> returns <code>false</code>, then <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#reject" id="ref-for-reject③">reject</a> <var>promise</var> with a "<code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#dataerror" id="ref-for-dataerror④">DataError</a></code>" <code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#idl-DOMException" id="ref-for-idl-DOMException①⑥">DOMException</a></code> and stop.</p>
-       <li data-md>
-        <p>If invoking the <a data-link-type="dfn" href="#validate-graph-resources" id="ref-for-validate-graph-resources③">validate graph resources</a> algorithm given <var>outputs</var> and <var>graph</var>.<code class="idl"><a data-link-type="idl" href="#dom-mlgraph-outputdescriptors-slot" id="ref-for-dom-mlgraph-outputdescriptors-slot②">[[outputDescriptors]]</a></code> returns <code>false</code>, then <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#reject" id="ref-for-reject④">reject</a> <var>promise</var> with a "<code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#dataerror" id="ref-for-dataerror⑤">DataError</a></code>" <code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#idl-DOMException" id="ref-for-idl-DOMException①⑦">DOMException</a></code> and stop.</p>
-       <li data-md>
-        <p>Let <var>transferredInputs</var> be the result of <a data-link-type="dfn" href="#mlnamedarraybufferviews-transfer" id="ref-for-mlnamedarraybufferviews-transfer③">transferring</a> <code class="idl"><a data-link-type="idl" href="#typedefdef-mlnamedarraybufferviews" id="ref-for-typedefdef-mlnamedarraybufferviews①⑨">MLNamedArrayBufferViews</a></code> <var>inputs</var>.</p>
-       <li data-md>
-        <p>Let <var>transferredOutputs</var> be the result of <a data-link-type="dfn" href="#mlnamedarraybufferviews-transfer" id="ref-for-mlnamedarraybufferviews-transfer④">transferring</a> <code class="idl"><a data-link-type="idl" href="#typedefdef-mlnamedarraybufferviews" id="ref-for-typedefdef-mlnamedarraybufferviews②⓪">MLNamedArrayBufferViews</a></code> <var>outputs</var>.</p>
-       <li data-md>
-        <p>Invoke <a data-link-type="dfn" href="#execute-graph" id="ref-for-execute-graph①">execute graph</a> given <var>graph</var>, <var>transferredInputs</var> and <var>transferredOutputs</var>.</p>
-       <li data-md>
-        <p>If that throws an error, <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#reject" id="ref-for-reject⑤">reject</a> <var>promise</var> with the error and stop.</p>
-       <li data-md>
-        <p>Otherwise, when <a data-link-type="dfn" href="#execute-graph" id="ref-for-execute-graph②">execute graph</a> has completed:</p>
-        <ol>
-         <li data-md>
-          <p>Let <var>result</var> be a new <code class="idl"><a data-link-type="idl" href="#dictdef-mlcomputeresult" id="ref-for-dictdef-mlcomputeresult②">MLComputeResult</a></code>.</p>
-         <li data-md>
-          <p>Set <var>result</var>.<code class="idl"><a data-link-type="idl" href="#dom-mlcomputeresult-inputs" id="ref-for-dom-mlcomputeresult-inputs">inputs</a></code> to <var>transferredInputs</var>.</p>
-         <li data-md>
-          <p>Set <var>result</var>.<code class="idl"><a data-link-type="idl" href="#dom-mlcomputeresult-outputs" id="ref-for-dom-mlcomputeresult-outputs">outputs</a></code> to <var>transferredOutputs</var>.</p>
-         <li data-md>
-          <p><a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#resolve" id="ref-for-resolve①">Resolve</a> <var>promise</var> with <var>result</var> and stop.</p>
-        </ol>
-      </ol>
-    </ol>
    </div>
+   <details class="algorithm" data-algorithm="compute(graph, inputs, outputs)" data-algorithm-for="MLContext" open>
+    <summary> The <dfn class="dfn-paneled idl-code" data-dfn-for="MLContext" data-dfn-type="method" data-export id="dom-mlcontext-compute"><code>compute(<var>graph</var>, <var>inputs</var>, <var>outputs</var>)</code></dfn> method steps are: </summary>
+    <div class="algorithm-steps">
+     <ol>
+      <li data-md>
+       <p>Let <var>promise</var> be <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#a-new-promise" id="ref-for-a-new-promise④">a new promise</a>.</p>
+      <li data-md>
+       <p>Return <var>promise</var> and run the following steps <a data-link-type="dfn" href="https://html.spec.whatwg.org/multipage/infrastructure.html#in-parallel" id="ref-for-in-parallel②">in parallel</a>:</p>
+       <ol>
+        <li data-md>
+         <p>If <var>graph</var>.<code class="idl"><a data-link-type="idl" href="#dom-mlgraph-context-slot" id="ref-for-dom-mlgraph-context-slot①">[[context]]</a></code>.<code class="idl"><a data-link-type="idl" href="#dom-mlcontext-contexttype-slot" id="ref-for-dom-mlcontext-contexttype-slot⑥">[[contextType]]</a></code> is not "<a data-link-type="dfn" href="#default-context" id="ref-for-default-context④">default</a>", <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#reject" id="ref-for-reject④">reject</a> <var>promise</var> with an "<code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#operationerror" id="ref-for-operationerror⑥">OperationError</a></code>" <code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#idl-DOMException" id="ref-for-idl-DOMException①⑧">DOMException</a></code>.</p>
+        <li data-md>
+         <p>If <a data-link-type="dfn" href="#validate-graph-resources" id="ref-for-validate-graph-resources②">validating graph resources</a> given <var>inputs</var> and <var>graph</var>.<code class="idl"><a data-link-type="idl" href="#dom-mlgraph-inputdescriptors-slot" id="ref-for-dom-mlgraph-inputdescriptors-slot②">[[inputDescriptors]]</a></code> returns false, then <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#reject" id="ref-for-reject⑤">reject</a> <var>promise</var> with a "<code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#dataerror" id="ref-for-dataerror④">DataError</a></code>" <code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#idl-DOMException" id="ref-for-idl-DOMException①⑨">DOMException</a></code>.</p>
+        <li data-md>
+         <p>If <a data-link-type="dfn" href="#validate-graph-resources" id="ref-for-validate-graph-resources③">validating graph resources</a> given <var>outputs</var> and <var>graph</var>.<code class="idl"><a data-link-type="idl" href="#dom-mlgraph-outputdescriptors-slot" id="ref-for-dom-mlgraph-outputdescriptors-slot②">[[outputDescriptors]]</a></code> returns false, then <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#reject" id="ref-for-reject⑥">reject</a> <var>promise</var> with a "<code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#dataerror" id="ref-for-dataerror⑤">DataError</a></code>" <code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#idl-DOMException" id="ref-for-idl-DOMException②⓪">DOMException</a></code>.</p>
+        <li data-md>
+         <p>Let <var>transferredInputs</var> be the result of <a data-link-type="dfn" href="#mlnamedarraybufferviews-transfer" id="ref-for-mlnamedarraybufferviews-transfer③">transferring</a> <code class="idl"><a data-link-type="idl" href="#typedefdef-mlnamedarraybufferviews" id="ref-for-typedefdef-mlnamedarraybufferviews①⑨">MLNamedArrayBufferViews</a></code> <var>inputs</var>.</p>
+        <li data-md>
+         <p>Let <var>transferredOutputs</var> be the result of <a data-link-type="dfn" href="#mlnamedarraybufferviews-transfer" id="ref-for-mlnamedarraybufferviews-transfer④">transferring</a> <code class="idl"><a data-link-type="idl" href="#typedefdef-mlnamedarraybufferviews" id="ref-for-typedefdef-mlnamedarraybufferviews②⓪">MLNamedArrayBufferViews</a></code> <var>outputs</var>.</p>
+        <li data-md>
+         <p>Invoke <a data-link-type="dfn" href="#execute-graph" id="ref-for-execute-graph①">execute graph</a> given <var>graph</var>, <var>transferredInputs</var> and <var>transferredOutputs</var>.</p>
+        <li data-md>
+         <p>If that <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#dfn-throw" id="ref-for-dfn-throw①⑥">throws</a> an error, <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#reject" id="ref-for-reject⑦">reject</a> <var>promise</var> with the error.</p>
+        <li data-md>
+         <p>Otherwise, when <a data-link-type="dfn" href="#execute-graph" id="ref-for-execute-graph②">execute graph</a> has completed:</p>
+         <ol>
+          <li data-md>
+           <p>Let <var>result</var> be a new <code class="idl"><a data-link-type="idl" href="#dictdef-mlcomputeresult" id="ref-for-dictdef-mlcomputeresult②">MLComputeResult</a></code>.</p>
+          <li data-md>
+           <p>Set <var>result</var>.<code class="idl"><a data-link-type="idl" href="#dom-mlcomputeresult-inputs" id="ref-for-dom-mlcomputeresult-inputs">inputs</a></code> to <var>transferredInputs</var>.</p>
+          <li data-md>
+           <p>Set <var>result</var>.<code class="idl"><a data-link-type="idl" href="#dom-mlcomputeresult-outputs" id="ref-for-dom-mlcomputeresult-outputs">outputs</a></code> to <var>transferredOutputs</var>.</p>
+          <li data-md>
+           <p><a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#resolve" id="ref-for-resolve②">Resolve</a> <var>promise</var> with <var>result</var>.</p>
+         </ol>
+       </ol>
+     </ol>
+    </div>
+   </details>
    <h5 class="heading settled" data-level="7.4.4.1" id="api-mlcontext-async-execution-examples"><span class="secno">7.4.4.1. </span><span class="content">Examples</span><a class="self-link" href="#api-mlcontext-async-execution-examples"></a></h5>
    <div class="example" id="example-cfa89792">
-    <a class="self-link" href="#example-cfa89792"></a> The following code showcases the asynchronous computation. 
+    <a class="self-link" href="#example-cfa89792"></a> 
+    <details open>
+     <summary> The following code showcases the asynchronous computation. </summary>
 <pre class="highlight"><c- a>const</c-> operandType <c- o>=</c-> <c- p>{</c->type<c- o>:</c-> <c- t>'float32'</c-><c- p>,</c-> dimensions<c- o>:</c-> <c- p>[</c-><c- mf>2</c-><c- p>,</c-> <c- mf>2</c-><c- p>]};</c->
 <c- a>const</c-> context <c- o>=</c-> <c- k>await</c-> navigator<c- p>.</c->ml<c- p>.</c->createContext<c- p>();</c->
 <c- a>const</c-> builder <c- o>=</c-> <c- ow>new</c-> MLGraphBuilder<c- p>(</c->context<c- p>);</c->
@@ -1839,39 +2282,42 @@ <h5 class="heading settled" data-level="7.4.4.1" id="api-mlcontext-async-executi
 <c- c1>// Note: the result.outputs.C buffer is different from the bufferC, but it</c->
 <c- c1>// shares the same backing memory allocation.</c->
 </pre>
+    </details>
    </div>
    <h4 class="heading settled" data-level="7.4.5" id="api-mlcontext-webgpu-interop"><span class="secno">7.4.5. </span><span class="content">WebGPU Interoperability</span><a class="self-link" href="#api-mlcontext-webgpu-interop"></a></h4>
-    Create <code class="idl"><a data-link-type="idl" href="#mlcommandencoder" id="ref-for-mlcommandencoder②">MLCommandEncoder</a></code> interface used to record the ML workload onto a WebGPU-compatible <code class="idl"><a data-link-type="idl" href="https://gpuweb.github.io/gpuweb/#command-buffers" id="ref-for-command-buffers②">GPUCommandBuffer</a></code> to allow mixing of ML workload with other GPU workload in an application that leverages WebGPU. This method only succeeds on an <code class="idl"><a data-link-type="idl" href="#mlcontext" id="ref-for-mlcontext③③">MLContext</a></code> created with <code class="idl"><a data-link-type="idl" href="https://gpuweb.github.io/gpuweb/#gpu-device" id="ref-for-gpu-device⑥">GPUDevice</a></code>. Otherwise, it throws an "<code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#operationerror" id="ref-for-operationerror⑥">OperationError</a></code>" <code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#idl-DOMException" id="ref-for-idl-DOMException①⑧">DOMException</a></code>. 
-<pre class="idl highlight def"><c- b>partial</c-> <c- b>interface</c-> <a class="idl-code" data-link-type="interface" href="#mlcontext" id="ref-for-mlcontext③④"><c- g>MLContext</c-></a> {
+    Create <code class="idl"><a data-link-type="idl" href="#mlcommandencoder" id="ref-for-mlcommandencoder②">MLCommandEncoder</a></code> interface used to record the ML workload onto a WebGPU-compatible <code class="idl"><a data-link-type="idl" href="https://gpuweb.github.io/gpuweb/#gpucommandbuffer" id="ref-for-gpucommandbuffer②">GPUCommandBuffer</a></code> to allow mixing of ML workload with other GPU workload in an application that leverages WebGPU. This method only succeeds on an <code class="idl"><a data-link-type="idl" href="#mlcontext" id="ref-for-mlcontext③②">MLContext</a></code> created with <code class="idl"><a data-link-type="idl" href="https://gpuweb.github.io/gpuweb/#gpudevice" id="ref-for-gpudevice⑥">GPUDevice</a></code>. Otherwise, it<a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#dfn-throw" id="ref-for-dfn-throw①⑦">throws</a> an "<code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#operationerror" id="ref-for-operationerror⑦">OperationError</a></code>" <code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#idl-DOMException" id="ref-for-idl-DOMException②①">DOMException</a></code>. 
+<pre class="idl highlight def"><c- b>partial</c-> <c- b>interface</c-> <a class="idl-code" data-link-type="interface" href="#mlcontext" id="ref-for-mlcontext③③"><c- g>MLContext</c-></a> {
   <a data-link-type="idl-name" href="#mlcommandencoder" id="ref-for-mlcommandencoder③"><c- n>MLCommandEncoder</c-></a> <dfn class="dfn-paneled idl-code" data-dfn-for="MLContext" data-dfn-type="method" data-export data-lt="createCommandEncoder()" id="dom-mlcontext-createcommandencoder"><code><c- g>createCommandEncoder</c-></code></dfn>();
 };
 </pre>
    <div class="algorithm" data-algorithm="mlcontext.createcommandencoder"> <strong>Returns:</strong> <code class="idl"><a data-link-type="idl" href="#mlcommandencoder" id="ref-for-mlcommandencoder④">MLCommandEncoder</a></code>. The command encoder used to record ML workload on the GPU. </div>
    <h3 class="heading settled" data-level="7.5" id="api-mlcommandencoder"><span class="secno">7.5. </span><span class="content">The MLCommandEncoder interface</span><a class="self-link" href="#api-mlcommandencoder"></a></h3>
-    The <code class="idl"><a data-link-type="idl" href="#mlcommandencoder" id="ref-for-mlcommandencoder⑤">MLCommandEncoder</a></code> interface represents a method of execution that synchronously records the computational workload of a compiled <code class="idl"><a data-link-type="idl" href="#mlgraph" id="ref-for-mlgraph①⑦">MLGraph</a></code> to a <code class="idl"><a data-link-type="idl" href="https://gpuweb.github.io/gpuweb/#command-buffers" id="ref-for-command-buffers③">GPUCommandBuffer</a></code> on the calling thread. Since the workload is not immediately executed, just recorded, this method allows more flexibility for the caller to determine how and when the recorded commands will be submitted for execution on the GPU relative to other GPU workload on the same or different queue. 
-<pre class="idl highlight def"><c- b>typedef</c-> (<a data-link-type="idl-name" href="https://gpuweb.github.io/gpuweb/#buffer-interface" id="ref-for-buffer-interface①"><c- n>GPUBuffer</c-></a> <c- b>or</c-> <a data-link-type="idl-name" href="https://gpuweb.github.io/gpuweb/#texture-interface" id="ref-for-texture-interface①"><c- n>GPUTexture</c-></a>) <dfn class="dfn-paneled idl-code" data-dfn-type="typedef" data-export id="typedefdef-mlgpuresource"><code><c- g>MLGPUResource</c-></code></dfn>;
+    The <code class="idl"><a data-link-type="idl" href="#mlcommandencoder" id="ref-for-mlcommandencoder⑤">MLCommandEncoder</a></code> interface represents a method of execution that synchronously records the computational workload of a compiled <code class="idl"><a data-link-type="idl" href="#mlgraph" id="ref-for-mlgraph①⑦">MLGraph</a></code> to a <code class="idl"><a data-link-type="idl" href="https://gpuweb.github.io/gpuweb/#gpucommandbuffer" id="ref-for-gpucommandbuffer③">GPUCommandBuffer</a></code> on the calling thread. Since the workload is not immediately executed, just recorded, this method allows more flexibility for the caller to determine how and when the recorded commands will be submitted for execution on the GPU relative to other GPU workload on the same or different queue. 
+<pre class="idl highlight def"><c- b>typedef</c-> (<a data-link-type="idl-name" href="https://gpuweb.github.io/gpuweb/#gpubuffer" id="ref-for-gpubuffer①"><c- n>GPUBuffer</c-></a> <c- b>or</c-> <a data-link-type="idl-name" href="https://gpuweb.github.io/gpuweb/#gputexture" id="ref-for-gputexture①"><c- n>GPUTexture</c-></a>) <dfn class="dfn-paneled idl-code" data-dfn-type="typedef" data-export id="typedefdef-mlgpuresource"><code><c- g>MLGPUResource</c-></code></dfn>;
 
 <c- b>typedef</c-> <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#idl-record" id="ref-for-idl-record④"><c- b>record</c-></a>&lt;<a class="idl-code" data-link-type="interface" href="https://webidl.spec.whatwg.org/#idl-DOMString" id="ref-for-idl-DOMString③"><c- b>DOMString</c-></a>, <a data-link-type="idl-name" href="#typedefdef-mlgpuresource" id="ref-for-typedefdef-mlgpuresource"><c- n>MLGPUResource</c-></a>> <dfn class="dfn-paneled idl-code" data-dfn-type="typedef" data-export id="typedefdef-mlnamedgpuresources"><code><c- g>MLNamedGPUResources</c-></code></dfn>;
 
 [<a class="idl-code" data-link-type="extended-attribute" href="https://webidl.spec.whatwg.org/#SecureContext" id="ref-for-SecureContext⑥"><c- g>SecureContext</c-></a>, <a class="idl-code" data-link-type="extended-attribute" href="https://webidl.spec.whatwg.org/#Exposed" id="ref-for-Exposed⑧"><c- g>Exposed</c-></a>=(<c- n>Window</c->, <c- n>DedicatedWorker</c->)]
 <c- b>interface</c-> <dfn class="dfn-paneled idl-code" data-dfn-type="interface" data-export id="mlcommandencoder"><code><c- g>MLCommandEncoder</c-></code></dfn> {};
 </pre>
-   <p><code class="idl"><a data-link-type="idl" href="#mlcommandencoder" id="ref-for-mlcommandencoder⑥">MLCommandEncoder</a></code> has the following internal slots:</p>
-   <dl>
-    <dt data-md><dfn class="idl-code" data-dfn-for="MLCommandEncoder" data-dfn-type="attribute" data-export id="dom-mlcommandencoder-context-slot"><code>[[context]]</code><a class="self-link" href="#dom-mlcommandencoder-context-slot"></a></dfn> of type <code class="idl"><a data-link-type="idl" href="#mlcontext" id="ref-for-mlcontext③⑤">MLContext</a></code>
-    <dd data-md>
-     <p>The context of type <code class="idl"><a data-link-type="idl" href="#mlcontext" id="ref-for-mlcontext③⑥">MLContext</a></code> associated with this <code class="idl"><a data-link-type="idl" href="#mlcommandencoder" id="ref-for-mlcommandencoder⑦">MLCommandEncoder</a></code>.</p>
-    <dt data-md><dfn class="idl-code" data-dfn-for="MLCommandEncoder" data-dfn-type="attribute" data-export id="dom-mlcommandencoder-implementation-slot"><code>[[implementation]]</code><a class="self-link" href="#dom-mlcommandencoder-implementation-slot"></a></dfn>
-    <dd data-md>
-     <p>The underlying implementation provided by the User Agent.</p>
-   </dl>
+   <div class="internal-slots">
+     <code class="idl"><a data-link-type="idl" href="#mlcommandencoder" id="ref-for-mlcommandencoder⑥">MLCommandEncoder</a></code> has the following internal slots: 
+    <dl>
+     <dt data-md><dfn class="idl-code" data-dfn-for="MLCommandEncoder" data-dfn-type="attribute" data-export id="dom-mlcommandencoder-context-slot"><code>[[context]]</code><a class="self-link" href="#dom-mlcommandencoder-context-slot"></a></dfn> of type <code class="idl"><a data-link-type="idl" href="#mlcontext" id="ref-for-mlcontext③④">MLContext</a></code>
+     <dd data-md>
+      <p>The context of type <code class="idl"><a data-link-type="idl" href="#mlcontext" id="ref-for-mlcontext③⑤">MLContext</a></code> associated with this <code class="idl"><a data-link-type="idl" href="#mlcommandencoder" id="ref-for-mlcommandencoder⑦">MLCommandEncoder</a></code>.</p>
+     <dt data-md><dfn class="idl-code" data-dfn-for="MLCommandEncoder" data-dfn-type="attribute" data-export id="dom-mlcommandencoder-implementation-slot"><code>[[implementation]]</code><a class="self-link" href="#dom-mlcommandencoder-implementation-slot"></a></dfn>
+     <dd data-md>
+      <p>The underlying implementation provided by the User Agent.</p>
+    </dl>
+   </div>
    <h4 class="heading settled" data-level="7.5.1" id="api-mlcommandencoder-graph-initialization"><span class="secno">7.5.1. </span><span class="content">Graph Initialization</span><a class="self-link" href="#api-mlcommandencoder-graph-initialization"></a></h4>
     Record the initialization of the <code class="idl"><a data-link-type="idl" href="#mlgraph" id="ref-for-mlgraph①⑧">MLGraph</a></code>. This is a necessary step for optimal performance during graph execution as it gives the platform an opportunity to prepare and optimize constant input data for the subsequent execution of the graph. This method should only be called once per graph. 
 <pre class="idl highlight def"><c- b>partial</c-> <c- b>interface</c-> <a class="idl-code" data-link-type="interface" href="#mlcommandencoder" id="ref-for-mlcommandencoder⑧"><c- g>MLCommandEncoder</c-></a> {
-  <a class="idl-code" data-link-type="interface" href="https://webidl.spec.whatwg.org/#idl-undefined" id="ref-for-idl-undefined④"><c- b>undefined</c-></a> <dfn class="idl-code" data-dfn-for="MLCommandEncoder" data-dfn-type="method" data-export data-lt="initializeGraph(graph)" id="dom-mlcommandencoder-initializegraph"><code><c- g>initializeGraph</c-></code><a class="self-link" href="#dom-mlcommandencoder-initializegraph"></a></dfn>(<a data-link-type="idl-name" href="#mlgraph" id="ref-for-mlgraph①⑨"><c- n>MLGraph</c-></a> <dfn class="idl-code" data-dfn-for="MLCommandEncoder/initializeGraph(graph)" data-dfn-type="argument" data-export id="dom-mlcommandencoder-initializegraph-graph-graph"><code><c- g>graph</c-></code><a class="self-link" href="#dom-mlcommandencoder-initializegraph-graph-graph"></a></dfn>);
+  <a class="idl-code" data-link-type="interface" href="https://webidl.spec.whatwg.org/#idl-undefined" id="ref-for-idl-undefined④"><c- b>undefined</c-></a> <a class="idl-code" data-link-type="method" href="#dom-mlcommandencoder-initializegraph" id="ref-for-dom-mlcommandencoder-initializegraph"><c- g>initializeGraph</c-></a>(<a data-link-type="idl-name" href="#mlgraph" id="ref-for-mlgraph①⑨"><c- n>MLGraph</c-></a> <dfn class="idl-code" data-dfn-for="MLCommandEncoder/initializeGraph(graph)" data-dfn-type="argument" data-export id="dom-mlcommandencoder-initializegraph-graph-graph"><code><c- g>graph</c-></code><a class="self-link" href="#dom-mlcommandencoder-initializegraph-graph-graph"></a></dfn>);
 };
 </pre>
-   <div class="algorithm" data-algorithm="mlcommandencoder.initializegraph">
+   <div>
      <strong>Arguments:</strong> 
     <ul>
      <li data-md>
@@ -1879,14 +2325,19 @@ <h4 class="heading settled" data-level="7.5.1" id="api-mlcommandencoder-graph-in
     </ul>
     <p><strong>Returns:</strong> <code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#idl-undefined" id="ref-for-idl-undefined⑤">undefined</a></code>.</p>
    </div>
-   <div class="note" role="note"> Graph initialization stage typically involves a process known as "weight preprocessing" where all the constant inputs to the graph are preprocessed and cached at the operating system level for subsequent graph execution calls. The initializing inputs are typically the constant weight data specified through the <code class="idl"><a data-link-type="idl" href="#dom-mlgraphbuilder-constant" id="ref-for-dom-mlgraphbuilder-constant">MLGraphBuilder/constant()</a></code> method as constant operands during graph construction time. </div>
+   <details class="algorithm" data-algorithm="initializeGraph(graph)" data-algorithm-for="MLCommandEncoder" open>
+    <summary> The <dfn class="dfn-paneled idl-code" data-dfn-for="MLCommandEncoder" data-dfn-type="method" data-export id="dom-mlcommandencoder-initializegraph"><code>initializeGraph(<var>graph</var>)</code></dfn> method steps are: </summary>
+    <div>
+     <div class="note" role="note"> Graph initialization stage typically involves a process known as "weight preprocessing" where all the constant inputs to the graph are preprocessed and cached at the operating system level for subsequent graph execution calls. The initializing inputs are typically the constant weight data specified through the <code class="idl"><a data-link-type="idl" href="#dom-mlgraphbuilder-constant" id="ref-for-dom-mlgraphbuilder-constant">MLGraphBuilder/constant(value, type)</a></code> method as constant operands during graph construction time. </div>
+    </div>
+   </details>
    <h4 class="heading settled" data-level="7.5.2" id="api-mlcommandencoder-dispatch-commands"><span class="secno">7.5.2. </span><span class="content">Dispatch Execution Commands</span><a class="self-link" href="#api-mlcommandencoder-dispatch-commands"></a></h4>
     Record the <code class="idl"><a data-link-type="idl" href="#mlgraph" id="ref-for-mlgraph②①">MLGraph</a></code> execution with the inputs <code class="idl"><a data-link-type="idl" href="#typedefdef-mlnamedgpuresources" id="ref-for-typedefdef-mlnamedgpuresources">MLNamedGPUResources</a></code> and outputs <code class="idl"><a data-link-type="idl" href="#typedefdef-mlnamedgpuresources" id="ref-for-typedefdef-mlnamedgpuresources①">MLNamedGPUResources</a></code>. 
 <pre class="idl highlight def"><c- b>partial</c-> <c- b>interface</c-> <a class="idl-code" data-link-type="interface" href="#mlcommandencoder" id="ref-for-mlcommandencoder⑨"><c- g>MLCommandEncoder</c-></a> {
-  <a class="idl-code" data-link-type="interface" href="https://webidl.spec.whatwg.org/#idl-undefined" id="ref-for-idl-undefined⑥"><c- b>undefined</c-></a> <dfn class="idl-code" data-dfn-for="MLCommandEncoder" data-dfn-type="method" data-export data-lt="dispatch(graph, inputs, outputs)" id="dom-mlcommandencoder-dispatch"><code><c- g>dispatch</c-></code><a class="self-link" href="#dom-mlcommandencoder-dispatch"></a></dfn>(<a data-link-type="idl-name" href="#mlgraph" id="ref-for-mlgraph②②"><c- n>MLGraph</c-></a> <dfn class="idl-code" data-dfn-for="MLCommandEncoder/dispatch(graph, inputs, outputs)" data-dfn-type="argument" data-export id="dom-mlcommandencoder-dispatch-graph-inputs-outputs-graph"><code><c- g>graph</c-></code><a class="self-link" href="#dom-mlcommandencoder-dispatch-graph-inputs-outputs-graph"></a></dfn>, <a data-link-type="idl-name" href="#typedefdef-mlnamedgpuresources" id="ref-for-typedefdef-mlnamedgpuresources②"><c- n>MLNamedGPUResources</c-></a> <dfn class="idl-code" data-dfn-for="MLCommandEncoder/dispatch(graph, inputs, outputs)" data-dfn-type="argument" data-export id="dom-mlcommandencoder-dispatch-graph-inputs-outputs-inputs"><code><c- g>inputs</c-></code><a class="self-link" href="#dom-mlcommandencoder-dispatch-graph-inputs-outputs-inputs"></a></dfn>, <a data-link-type="idl-name" href="#typedefdef-mlnamedgpuresources" id="ref-for-typedefdef-mlnamedgpuresources③"><c- n>MLNamedGPUResources</c-></a> <dfn class="idl-code" data-dfn-for="MLCommandEncoder/dispatch(graph, inputs, outputs)" data-dfn-type="argument" data-export id="dom-mlcommandencoder-dispatch-graph-inputs-outputs-outputs"><code><c- g>outputs</c-></code><a class="self-link" href="#dom-mlcommandencoder-dispatch-graph-inputs-outputs-outputs"></a></dfn>);
+  <a class="idl-code" data-link-type="interface" href="https://webidl.spec.whatwg.org/#idl-undefined" id="ref-for-idl-undefined⑥"><c- b>undefined</c-></a> <a class="idl-code" data-link-type="method" href="#dom-mlcommandencoder-dispatch" id="ref-for-dom-mlcommandencoder-dispatch"><c- g>dispatch</c-></a>(<a data-link-type="idl-name" href="#mlgraph" id="ref-for-mlgraph②②"><c- n>MLGraph</c-></a> <dfn class="idl-code" data-dfn-for="MLCommandEncoder/dispatch(graph, inputs, outputs)" data-dfn-type="argument" data-export id="dom-mlcommandencoder-dispatch-graph-inputs-outputs-graph"><code><c- g>graph</c-></code><a class="self-link" href="#dom-mlcommandencoder-dispatch-graph-inputs-outputs-graph"></a></dfn>, <a data-link-type="idl-name" href="#typedefdef-mlnamedgpuresources" id="ref-for-typedefdef-mlnamedgpuresources②"><c- n>MLNamedGPUResources</c-></a> <dfn class="idl-code" data-dfn-for="MLCommandEncoder/dispatch(graph, inputs, outputs)" data-dfn-type="argument" data-export id="dom-mlcommandencoder-dispatch-graph-inputs-outputs-inputs"><code><c- g>inputs</c-></code><a class="self-link" href="#dom-mlcommandencoder-dispatch-graph-inputs-outputs-inputs"></a></dfn>, <a data-link-type="idl-name" href="#typedefdef-mlnamedgpuresources" id="ref-for-typedefdef-mlnamedgpuresources③"><c- n>MLNamedGPUResources</c-></a> <dfn class="idl-code" data-dfn-for="MLCommandEncoder/dispatch(graph, inputs, outputs)" data-dfn-type="argument" data-export id="dom-mlcommandencoder-dispatch-graph-inputs-outputs-outputs"><code><c- g>outputs</c-></code><a class="self-link" href="#dom-mlcommandencoder-dispatch-graph-inputs-outputs-outputs"></a></dfn>);
 };
 </pre>
-   <div class="algorithm" data-algorithm="mlcommandencoder.dispatch">
+   <div>
      <strong>Arguments:</strong> 
     <ul>
      <li data-md>
@@ -1897,87 +2348,108 @@ <h4 class="heading settled" data-level="7.5.2" id="api-mlcommandencoder-dispatch
       <p><em>outputs</em>: an <code class="idl"><a data-link-type="idl" href="#typedefdef-mlnamedgpuresources" id="ref-for-typedefdef-mlnamedgpuresources⑤">MLNamedGPUResources</a></code>. The pre-allocated resources of required outputs.</p>
     </ul>
     <p><strong>Returns:</strong> <code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#idl-undefined" id="ref-for-idl-undefined⑦">undefined</a></code>.</p>
-    <ol>
-     <li data-md>
-      <p>If any of the following requirements are unmet, then throw a "<code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#dataerror" id="ref-for-dataerror⑥">DataError</a></code>" <code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#idl-DOMException" id="ref-for-idl-DOMException①⑨">DOMException</a></code> and stop.</p>
-      <div class="validusage">
-       <ol>
+   </div>
+   <details class="algorithm" data-algorithm="dispatch(graph, inputs, outputs)" data-algorithm-for="MLCommandEncoder" open>
+    <summary> The <dfn class="dfn-paneled idl-code" data-dfn-for="MLCommandEncoder" data-dfn-type="method" data-export id="dom-mlcommandencoder-dispatch"><code>dispatch(<var>graph</var>, <var>inputs</var>, <var>outputs</var>)</code></dfn> method steps are: </summary>
+    <div class="algorithm-steps">
+     <ol>
+      <li data-md>
+       <p>If any of the following requirements are unmet, then <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#dfn-throw" id="ref-for-dfn-throw①⑧">throw</a> a "<code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#dataerror" id="ref-for-dataerror⑥">DataError</a></code>" <code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#idl-DOMException" id="ref-for-idl-DOMException②②">DOMException</a></code>.</p>
+       <div class="validusage">
+        <ol>
+         <li data-md>
+          <p><a data-link-type="dfn" href="https://infra.spec.whatwg.org/#map-iterate" id="ref-for-map-iterate⑤">For each</a> <var>key</var> → <var>value</var> of <var>inputs</var>:</p>
+          <ol>
+           <li data-md>
+            <p><var>graph</var>.<code class="idl"><a data-link-type="idl" href="#dom-mlgraph-inputdescriptors-slot" id="ref-for-dom-mlgraph-inputdescriptors-slot③">[[inputDescriptors]]</a></code>[<var>key</var>] must <a data-link-type="dfn" href="https://infra.spec.whatwg.org/#map-exists" id="ref-for-map-exists⑤">exist</a>.</p>
+           <li data-md>
+            <p>Let <var>inputDesc</var> be <var>graph</var>.<code class="idl"><a data-link-type="idl" href="#dom-mlgraph-inputdescriptors-slot" id="ref-for-dom-mlgraph-inputdescriptors-slot④">[[inputDescriptors]]</a></code>[<var>key</var>].</p>
+           <li data-md>
+            <p>If <var>value</var> is a <code class="idl"><a data-link-type="idl" href="https://gpuweb.github.io/gpuweb/#gpubuffer" id="ref-for-gpubuffer②">GPUBuffer</a></code>, then:</p>
+            <ol>
+             <li data-md>
+              <p><var>value</var>.<code class="idl"><a data-link-type="idl" href="https://gpuweb.github.io/gpuweb/#dom-gpubuffer-size" id="ref-for-dom-gpubuffer-size">size</a></code> must equal to <a data-link-type="dfn" href="#mloperanddescriptor-byte-length" id="ref-for-mloperanddescriptor-byte-length②">byte length</a> of <var>inputDesc</var>.</p>
+            </ol>
+          </ol>
+         <li data-md>
+          <p><a data-link-type="dfn" href="https://infra.spec.whatwg.org/#map-iterate" id="ref-for-map-iterate⑥">For each</a> <var>key</var> → <var>value</var> of <var>outputs</var>:</p>
+          <ol>
+           <li data-md>
+            <p><var>graph</var>.<code class="idl"><a data-link-type="idl" href="#dom-mlgraph-outputdescriptors-slot" id="ref-for-dom-mlgraph-outputdescriptors-slot③">[[outputDescriptors]]</a></code>[<var>key</var>] must <a data-link-type="dfn" href="https://infra.spec.whatwg.org/#map-exists" id="ref-for-map-exists⑥">exist</a>.</p>
+           <li data-md>
+            <p>Let <var>outputDesc</var> be <var>graph</var>.<code class="idl"><a data-link-type="idl" href="#dom-mlgraph-outputdescriptors-slot" id="ref-for-dom-mlgraph-outputdescriptors-slot④">[[outputDescriptors]]</a></code>[<var>key</var>].</p>
+           <li data-md>
+            <p>If <var>value</var> is a <code class="idl"><a data-link-type="idl" href="https://gpuweb.github.io/gpuweb/#gpubuffer" id="ref-for-gpubuffer③">GPUBuffer</a></code>, then:</p>
+            <ol>
+             <li data-md>
+              <p><var>value</var>.<code class="idl"><a data-link-type="idl" href="https://gpuweb.github.io/gpuweb/#dom-gpubuffer-size" id="ref-for-dom-gpubuffer-size①">size</a></code> must equal to <a data-link-type="dfn" href="#mloperanddescriptor-byte-length" id="ref-for-mloperanddescriptor-byte-length③">byte length</a> of <var>outputDesc</var>.</p>
+            </ol>
+          </ol>
+        </ol>
+       </div>
+      <li data-md>
+       <p><a data-link-type="dfn" href="https://infra.spec.whatwg.org/#map-iterate" id="ref-for-map-iterate⑦">For each</a> <var>key</var> → <var>value</var> of <var>inputs</var>:</p>
+       <ol>
         <li data-md>
-         <p>For each <var>key</var> -> <var>value</var> of <var>inputs</var>:</p>
-         <ol>
-          <li data-md>
-           <p><var>graph</var>.<code class="idl"><a data-link-type="idl" href="#dom-mlgraph-inputdescriptors-slot" id="ref-for-dom-mlgraph-inputdescriptors-slot③">[[inputDescriptors]]</a></code>[<var>key</var>] must <a data-link-type="dfn" href="https://infra.spec.whatwg.org/#map-exists" id="ref-for-map-exists③">exist</a>.</p>
-          <li data-md>
-           <p>Let <var>inputDesc</var> be <var>graph</var>.<code class="idl"><a data-link-type="idl" href="#dom-mlgraph-inputdescriptors-slot" id="ref-for-dom-mlgraph-inputdescriptors-slot④">[[inputDescriptors]]</a></code>[<var>key</var>].</p>
-          <li data-md>
-           <p>If <var>value</var> is a <code class="idl"><a data-link-type="idl" href="https://gpuweb.github.io/gpuweb/#buffer-interface" id="ref-for-buffer-interface②">GPUBuffer</a></code>, then:</p>
-           <ol>
-            <li data-md>
-             <p><var>value</var>.<code class="idl"><a data-link-type="idl" href="https://gpuweb.github.io/gpuweb/#dom-gpubuffer-size" id="ref-for-dom-gpubuffer-size">size</a></code> must equal to <a data-link-type="dfn" href="#mloperanddescriptor-byte-length" id="ref-for-mloperanddescriptor-byte-length②">byte length</a> of <var>inputDesc</var>.</p>
-           </ol>
-         </ol>
+         <p>Set the input of <var>graph</var>.<code class="idl"><a data-link-type="idl" href="#dom-mlgraph-implementation-slot" id="ref-for-dom-mlgraph-implementation-slot③">[[implementation]]</a></code> that is associated with <var>key</var> to <var>value</var>.</p>
+       </ol>
+      <li data-md>
+       <p><a data-link-type="dfn" href="https://infra.spec.whatwg.org/#map-iterate" id="ref-for-map-iterate⑧">For each</a> <var>key</var> → <var>value</var> of <var>outputs</var>:</p>
+       <ol>
         <li data-md>
-         <p>For each <var>key</var> -> <var>value</var> of <var>outputs</var>:</p>
-         <ol>
-          <li data-md>
-           <p><var>graph</var>.<code class="idl"><a data-link-type="idl" href="#dom-mlgraph-outputdescriptors-slot" id="ref-for-dom-mlgraph-outputdescriptors-slot③">[[outputDescriptors]]</a></code>[<var>key</var>] must <a data-link-type="dfn" href="https://infra.spec.whatwg.org/#map-exists" id="ref-for-map-exists④">exist</a>.</p>
-          <li data-md>
-           <p>Let <var>outputDesc</var> be <var>graph</var>.<code class="idl"><a data-link-type="idl" href="#dom-mlgraph-outputdescriptors-slot" id="ref-for-dom-mlgraph-outputdescriptors-slot④">[[outputDescriptors]]</a></code>[<var>key</var>].</p>
-          <li data-md>
-           <p>If <var>value</var> is a <code class="idl"><a data-link-type="idl" href="https://gpuweb.github.io/gpuweb/#buffer-interface" id="ref-for-buffer-interface③">GPUBuffer</a></code>, then:</p>
-           <ol>
-            <li data-md>
-             <p><var>value</var>.<code class="idl"><a data-link-type="idl" href="https://gpuweb.github.io/gpuweb/#dom-gpubuffer-size" id="ref-for-dom-gpubuffer-size①">size</a></code> must equal to <a data-link-type="dfn" href="#mloperanddescriptor-byte-length" id="ref-for-mloperanddescriptor-byte-length③">byte length</a> of <var>outputDesc</var>.</p>
-           </ol>
-         </ol>
+         <p>Set the output of <var>graph</var>.<code class="idl"><a data-link-type="idl" href="#dom-mlgraph-implementation-slot" id="ref-for-dom-mlgraph-implementation-slot④">[[implementation]]</a></code> that is associated with <var>key</var> to <var>value</var>.</p>
        </ol>
-      </div>
-     <li data-md>
-      <p>For each <var>key</var> -> <var>value</var> of <var>inputs</var>:</p>
-      <ol>
-       <li data-md>
-        <p>Set the input of <var>graph</var>.<code class="idl"><a data-link-type="idl" href="#dom-mlgraph-implementation-slot" id="ref-for-dom-mlgraph-implementation-slot③">[[implementation]]</a></code> that is associated with <var>key</var> to <var>value</var>.</p>
-      </ol>
-     <li data-md>
-      <p>For each <var>key</var> -> <var>value</var> of <var>outputs</var>:</p>
-      <ol>
-       <li data-md>
-        <p>Set the output of <var>graph</var>.<code class="idl"><a data-link-type="idl" href="#dom-mlgraph-implementation-slot" id="ref-for-dom-mlgraph-implementation-slot④">[[implementation]]</a></code> that is associated with <var>key</var> to <var>value</var>.</p>
-      </ol>
-     <li data-md>
-      <p>Issue a compute request of <var>graph</var>.<code class="idl"><a data-link-type="idl" href="#dom-mlgraph-implementation-slot" id="ref-for-dom-mlgraph-implementation-slot⑤">[[implementation]]</a></code>.</p>
-     <li data-md>
-      <p>If there is an error returned by <var>graph</var>.<code class="idl"><a data-link-type="idl" href="#dom-mlgraph-implementation-slot" id="ref-for-dom-mlgraph-implementation-slot⑥">[[implementation]]</a></code>, then:</p>
-      <ol>
-       <li data-md>
-        <p>Throw an "<code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#operationerror" id="ref-for-operationerror⑦">OperationError</a></code>" <code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#idl-DOMException" id="ref-for-idl-DOMException②⓪">DOMException</a></code> and stop.</p>
-      </ol>
-     <li data-md>
-      <p>Return <code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#idl-undefined" id="ref-for-idl-undefined⑧">undefined</a></code>.</p>
-    </ol>
-   </div>
+      <li data-md>
+       <p>Issue a compute request of <var>graph</var>.<code class="idl"><a data-link-type="idl" href="#dom-mlgraph-implementation-slot" id="ref-for-dom-mlgraph-implementation-slot⑤">[[implementation]]</a></code>.</p>
+      <li data-md>
+       <p>If there is an error returned by <var>graph</var>.<code class="idl"><a data-link-type="idl" href="#dom-mlgraph-implementation-slot" id="ref-for-dom-mlgraph-implementation-slot⑥">[[implementation]]</a></code>, then:</p>
+       <ol>
+        <li data-md>
+         <p>Throw an "<code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#operationerror" id="ref-for-operationerror⑧">OperationError</a></code>" <code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#idl-DOMException" id="ref-for-idl-DOMException②③">DOMException</a></code>.</p>
+       </ol>
+      <li data-md>
+       <p>Return <code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#idl-undefined" id="ref-for-idl-undefined⑧">undefined</a></code>.</p>
+     </ol>
+    </div>
+   </details>
    <h4 class="heading settled" data-level="7.5.3" id="api-mlcommandencoder-generate-gpu-command-buffer"><span class="secno">7.5.3. </span><span class="content">Generate GPU Command Buffer</span><a class="self-link" href="#api-mlcommandencoder-generate-gpu-command-buffer"></a></h4>
-    Complete the recording of ML workload and return a WebGPU-compatible <code class="idl"><a data-link-type="idl" href="https://gpuweb.github.io/gpuweb/#command-buffers" id="ref-for-command-buffers④">GPUCommandBuffer</a></code> containing the recorded workload. 
+    Complete the recording of ML workload and return a WebGPU-compatible <code class="idl"><a data-link-type="idl" href="https://gpuweb.github.io/gpuweb/#gpucommandbuffer" id="ref-for-gpucommandbuffer④">GPUCommandBuffer</a></code> containing the recorded workload. 
 <pre class="idl highlight def"><c- b>partial</c-> <c- b>interface</c-> <a class="idl-code" data-link-type="interface" href="#mlcommandencoder" id="ref-for-mlcommandencoder①⓪"><c- g>MLCommandEncoder</c-></a> {
-  <a data-link-type="idl-name" href="https://gpuweb.github.io/gpuweb/#command-buffers" id="ref-for-command-buffers⑤"><c- n>GPUCommandBuffer</c-></a> <dfn class="idl-code" data-dfn-for="MLCommandEncoder" data-dfn-type="method" data-export data-lt="finish(descriptor)|finish()" id="dom-mlcommandencoder-finish"><code><c- g>finish</c-></code><a class="self-link" href="#dom-mlcommandencoder-finish"></a></dfn>(<c- b>optional</c-> <a data-link-type="idl-name" href="https://gpuweb.github.io/gpuweb/#dictdef-gpucommandbufferdescriptor" id="ref-for-dictdef-gpucommandbufferdescriptor"><c- n>GPUCommandBufferDescriptor</c-></a> <dfn class="idl-code" data-dfn-for="MLCommandEncoder/finish(descriptor), MLCommandEncoder/finish()" data-dfn-type="argument" data-export id="dom-mlcommandencoder-finish-descriptor-descriptor"><code><c- g>descriptor</c-></code><a class="self-link" href="#dom-mlcommandencoder-finish-descriptor-descriptor"></a></dfn> = {});
+  <a data-link-type="idl-name" href="https://gpuweb.github.io/gpuweb/#gpucommandbuffer" id="ref-for-gpucommandbuffer⑤"><c- n>GPUCommandBuffer</c-></a> <a class="idl-code" data-link-type="method" href="#dom-mlcommandencoder-finish" id="ref-for-dom-mlcommandencoder-finish"><c- g>finish</c-></a>(<c- b>optional</c-> <a data-link-type="idl-name" href="https://gpuweb.github.io/gpuweb/#dictdef-gpucommandbufferdescriptor" id="ref-for-dictdef-gpucommandbufferdescriptor"><c- n>GPUCommandBufferDescriptor</c-></a> <dfn class="idl-code" data-dfn-for="MLCommandEncoder/finish(descriptor), MLCommandEncoder/finish()" data-dfn-type="argument" data-export id="dom-mlcommandencoder-finish-descriptor-descriptor"><code><c- g>descriptor</c-></code><a class="self-link" href="#dom-mlcommandencoder-finish-descriptor-descriptor"></a></dfn> = {});
 };
 </pre>
-   <div class="algorithm" data-algorithm="mlcommandencoder.finish">
+   <div>
      <strong>Arguments:</strong> 
     <ul>
      <li data-md>
       <p><em>descriptor</em>: an optional <code class="idl"><a data-link-type="idl" href="https://gpuweb.github.io/gpuweb/#dictdef-gpucommandbufferdescriptor" id="ref-for-dictdef-gpucommandbufferdescriptor①">GPUCommandBufferDescriptor</a></code>. Descriptor of the command buffer.</p>
     </ul>
-    <p><strong>Returns:</strong> <code class="idl"><a data-link-type="idl" href="https://gpuweb.github.io/gpuweb/#command-buffers" id="ref-for-command-buffers⑥">GPUCommandBuffer</a></code>.</p>
+    <p><strong>Returns:</strong> <code class="idl"><a data-link-type="idl" href="https://gpuweb.github.io/gpuweb/#gpucommandbuffer" id="ref-for-gpucommandbuffer⑥">GPUCommandBuffer</a></code>.</p>
    </div>
+   <details class="algorithm" data-algorithm="finish(descriptor)" data-algorithm-for="MLCommandEncoder" open>
+    <summary> The <dfn class="dfn-paneled idl-code" data-dfn-for="MLCommandEncoder" data-dfn-type="method" data-export data-lt="finish(descriptor)|finish()" id="dom-mlcommandencoder-finish"><code>finish(<var>descriptor</var>)</code></dfn> method steps are: </summary>
+    <div class="algorithm-steps">
+     <ol>
+      <li data-md>
+       <p>If any of the following sub-steps fail, <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#dfn-throw" id="ref-for-dfn-throw①⑨">throw</a> an "<code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#operationerror" id="ref-for-operationerror⑨">OperationError</a></code>" <code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#idl-DOMException" id="ref-for-idl-DOMException②④">DOMException</a></code>.</p>
+       <ol>
+        <li data-md>
+         <p>Make a request to the underlying platform to complete the recording of the ML workload, given <var>descriptor</var>.</p>
+         <div class="note" role="note"> See the related <a href="https://www.w3.org/TR/webgpu/#dom-gpucommandencoder-finish">WebGPU steps</a>. </div>
+       </ol>
+      <li data-md>
+       <p>Return a <code class="idl"><a data-link-type="idl" href="https://gpuweb.github.io/gpuweb/#gpucommandbuffer" id="ref-for-gpucommandbuffer⑦">GPUCommandBuffer</a></code> containing the recorded workload.</p>
+     </ol>
+    </div>
+   </details>
    <h3 class="heading settled" data-level="7.6" id="api-mlgraphbuilder"><span class="secno">7.6. </span><span class="content">The MLGraphBuilder interface</span><a class="self-link" href="#api-mlgraphbuilder"></a></h3>
-   <p>The <code class="idl"><a data-link-type="idl" href="#mlgraphbuilder" id="ref-for-mlgraphbuilder①①">MLGraphBuilder</a></code> interface defines a set of operations as identified by the <a href="#usecases">§ 2 Use cases</a> that can be composed into a computational graph. It also represents the intermediate state of a graph building session.</p>
-<pre class="idl highlight def"><c- b>typedef</c-> <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#idl-record" id="ref-for-idl-record⑤"><c- b>record</c-></a>&lt;<a class="idl-code" data-link-type="interface" href="https://webidl.spec.whatwg.org/#idl-DOMString" id="ref-for-idl-DOMString④"><c- b>DOMString</c-></a>, <a data-link-type="idl-name" href="#mloperand" id="ref-for-mloperand②②"><c- n>MLOperand</c-></a>> <dfn class="dfn-paneled idl-code" data-dfn-type="typedef" data-export id="typedefdef-mlnamedoperands"><code><c- g>MLNamedOperands</c-></code></dfn>;
+   <p>The <code class="idl"><a data-link-type="idl" href="#mlgraphbuilder" id="ref-for-mlgraphbuilder①⑤">MLGraphBuilder</a></code> interface defines a set of operations as identified by the <a href="#usecases">§ 2 Use cases</a> that can be composed into a computational graph. It also represents the intermediate state of a graph building session.</p>
+<pre class="idl highlight def"><c- b>typedef</c-> <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#idl-record" id="ref-for-idl-record⑤"><c- b>record</c-></a>&lt;<a class="idl-code" data-link-type="interface" href="https://webidl.spec.whatwg.org/#idl-DOMString" id="ref-for-idl-DOMString④"><c- b>DOMString</c-></a>, <a data-link-type="idl-name" href="#mloperand" id="ref-for-mloperand②④"><c- n>MLOperand</c-></a>> <dfn class="dfn-paneled idl-code" data-dfn-type="typedef" data-export id="typedefdef-mlnamedoperands"><code><c- g>MLNamedOperands</c-></code></dfn>;
 
 <c- b>dictionary</c-> <dfn class="dfn-paneled idl-code" data-dfn-type="dictionary" data-export id="dictdef-mlbufferresourceview"><code><c- g>MLBufferResourceView</c-></code></dfn> {
-  <c- b>required</c-> <a data-link-type="idl-name" href="https://gpuweb.github.io/gpuweb/#buffer-interface" id="ref-for-buffer-interface④"><c- n>GPUBuffer</c-></a> <dfn class="idl-code" data-dfn-for="MLBufferResourceView" data-dfn-type="dict-member" data-export data-type="GPUBuffer " id="dom-mlbufferresourceview-resource"><code><c- g>resource</c-></code><a class="self-link" href="#dom-mlbufferresourceview-resource"></a></dfn>;
-  <a class="idl-code" data-link-type="interface" href="https://webidl.spec.whatwg.org/#idl-unsigned-long-long" id="ref-for-idl-unsigned-long-long"><c- b>unsigned</c-> <c- b>long</c-> <c- b>long</c-></a> <dfn class="idl-code" data-default="0" data-dfn-for="MLBufferResourceView" data-dfn-type="dict-member" data-export data-type="unsigned long long " id="dom-mlbufferresourceview-offset"><code><c- g>offset</c-></code><a class="self-link" href="#dom-mlbufferresourceview-offset"></a></dfn> = 0;
-  <a class="idl-code" data-link-type="interface" href="https://webidl.spec.whatwg.org/#idl-unsigned-long-long" id="ref-for-idl-unsigned-long-long①"><c- b>unsigned</c-> <c- b>long</c-> <c- b>long</c-></a> <dfn class="idl-code" data-dfn-for="MLBufferResourceView" data-dfn-type="dict-member" data-export data-type="unsigned long long " id="dom-mlbufferresourceview-size"><code><c- g>size</c-></code><a class="self-link" href="#dom-mlbufferresourceview-size"></a></dfn>;
+  <c- b>required</c-> <a data-link-type="idl-name" href="https://gpuweb.github.io/gpuweb/#gpubuffer" id="ref-for-gpubuffer④"><c- n>GPUBuffer</c-></a> <a class="idl-code" data-link-type="dict-member" data-type="GPUBuffer " href="#dom-mlbufferresourceview-resource" id="ref-for-dom-mlbufferresourceview-resource"><c- g>resource</c-></a>;
+  <a class="idl-code" data-link-type="interface" href="https://webidl.spec.whatwg.org/#idl-unsigned-long-long" id="ref-for-idl-unsigned-long-long"><c- b>unsigned</c-> <c- b>long</c-> <c- b>long</c-></a> <a class="idl-code" data-default="0" data-link-type="dict-member" data-type="unsigned long long " href="#dom-mlbufferresourceview-offset" id="ref-for-dom-mlbufferresourceview-offset"><c- g>offset</c-></a> = 0;
+  <a class="idl-code" data-link-type="interface" href="https://webidl.spec.whatwg.org/#idl-unsigned-long-long" id="ref-for-idl-unsigned-long-long①"><c- b>unsigned</c-> <c- b>long</c-> <c- b>long</c-></a> <a class="idl-code" data-link-type="dict-member" data-type="unsigned long long " href="#dom-mlbufferresourceview-size" id="ref-for-dom-mlbufferresourceview-size"><c- g>size</c-></a>;
 };
 
 <c- b>typedef</c-> (<a data-link-type="idl-name" href="https://webidl.spec.whatwg.org/#ArrayBufferView" id="ref-for-ArrayBufferView⑦"><c- n>ArrayBufferView</c-></a> <c- b>or</c-> <a data-link-type="idl-name" href="#dictdef-mlbufferresourceview" id="ref-for-dictdef-mlbufferresourceview"><c- n>MLBufferResourceView</c-></a>) <dfn class="dfn-paneled idl-code" data-dfn-type="typedef" data-export id="typedefdef-mlbufferview"><code><c- g>MLBufferView</c-></code></dfn>;
@@ -1985,237 +2457,404 @@ <h3 class="heading settled" data-level="7.6" id="api-mlgraphbuilder"><span class
 [<a class="idl-code" data-link-type="extended-attribute" href="https://webidl.spec.whatwg.org/#SecureContext" id="ref-for-SecureContext⑦"><c- g>SecureContext</c-></a>, <a class="idl-code" data-link-type="extended-attribute" href="https://webidl.spec.whatwg.org/#Exposed" id="ref-for-Exposed⑨"><c- g>Exposed</c-></a>=(<c- n>Window</c->, <c- n>DedicatedWorker</c->)]
 <c- b>interface</c-> <dfn class="dfn-paneled idl-code" data-dfn-type="interface" data-export id="mlgraphbuilder"><code><c- g>MLGraphBuilder</c-></code></dfn> {
   // Construct the graph builder from the context.
-  <dfn class="dfn-paneled idl-code" data-dfn-for="MLGraphBuilder" data-dfn-type="constructor" data-export data-lt="MLGraphBuilder(context)|constructor(context)" id="dom-mlgraphbuilder-mlgraphbuilder"><code><c- g>constructor</c-></code></dfn>(<a data-link-type="idl-name" href="#mlcontext" id="ref-for-mlcontext③⑦"><c- n>MLContext</c-></a> <dfn class="idl-code" data-dfn-for="MLGraphBuilder/MLGraphBuilder(context), MLGraphBuilder/constructor(context)" data-dfn-type="argument" data-export id="dom-mlgraphbuilder-mlgraphbuilder-context-context"><code><c- g>context</c-></code><a class="self-link" href="#dom-mlgraphbuilder-mlgraphbuilder-context-context"></a></dfn>);
+  <a class="idl-code" data-link-type="constructor" href="#dom-mlgraphbuilder-mlgraphbuilder" id="ref-for-dom-mlgraphbuilder-mlgraphbuilder①"><c- g>constructor</c-></a>(<a data-link-type="idl-name" href="#mlcontext" id="ref-for-mlcontext③⑥"><c- n>MLContext</c-></a> <dfn class="idl-code" data-dfn-for="MLGraphBuilder/MLGraphBuilder(context), MLGraphBuilder/constructor(context)" data-dfn-type="argument" data-export id="dom-mlgraphbuilder-mlgraphbuilder-context-context"><code><c- g>context</c-></code><a class="self-link" href="#dom-mlgraphbuilder-mlgraphbuilder-context-context"></a></dfn>);
 
   // Create an operand for a graph input.
-  <a data-link-type="idl-name" href="#mloperand" id="ref-for-mloperand②③"><c- n>MLOperand</c-></a> <dfn class="dfn-paneled idl-code" data-dfn-for="MLGraphBuilder" data-dfn-type="method" data-export data-lt="input(name, descriptor)" id="dom-mlgraphbuilder-input"><code><c- g>input</c-></code></dfn>(<a class="idl-code" data-link-type="interface" href="https://webidl.spec.whatwg.org/#idl-DOMString" id="ref-for-idl-DOMString⑤"><c- b>DOMString</c-></a> <dfn class="idl-code" data-dfn-for="MLGraphBuilder/input(name, descriptor)" data-dfn-type="argument" data-export id="dom-mlgraphbuilder-input-name-descriptor-name"><code><c- g>name</c-></code><a class="self-link" href="#dom-mlgraphbuilder-input-name-descriptor-name"></a></dfn>, <a data-link-type="idl-name" href="#dictdef-mloperanddescriptor" id="ref-for-dictdef-mloperanddescriptor⑦"><c- n>MLOperandDescriptor</c-></a> <dfn class="idl-code" data-dfn-for="MLGraphBuilder/input(name, descriptor)" data-dfn-type="argument" data-export id="dom-mlgraphbuilder-input-name-descriptor-descriptor"><code><c- g>descriptor</c-></code><a class="self-link" href="#dom-mlgraphbuilder-input-name-descriptor-descriptor"></a></dfn>);
+  <a data-link-type="idl-name" href="#mloperand" id="ref-for-mloperand②⑤"><c- n>MLOperand</c-></a> <a class="idl-code" data-link-type="method" href="#dom-mlgraphbuilder-input" id="ref-for-dom-mlgraphbuilder-input"><c- g>input</c-></a>(<a class="idl-code" data-link-type="interface" href="https://webidl.spec.whatwg.org/#idl-DOMString" id="ref-for-idl-DOMString⑤"><c- b>DOMString</c-></a> <dfn class="idl-code" data-dfn-for="MLGraphBuilder/input(name, descriptor)" data-dfn-type="argument" data-export id="dom-mlgraphbuilder-input-name-descriptor-name"><code><c- g>name</c-></code><a class="self-link" href="#dom-mlgraphbuilder-input-name-descriptor-name"></a></dfn>, <a data-link-type="idl-name" href="#dictdef-mloperanddescriptor" id="ref-for-dictdef-mloperanddescriptor⑦"><c- n>MLOperandDescriptor</c-></a> <dfn class="idl-code" data-dfn-for="MLGraphBuilder/input(name, descriptor)" data-dfn-type="argument" data-export id="dom-mlgraphbuilder-input-name-descriptor-descriptor"><code><c- g>descriptor</c-></code><a class="self-link" href="#dom-mlgraphbuilder-input-name-descriptor-descriptor"></a></dfn>);
 
   // Create an operand for a graph constant.
-  <a data-link-type="idl-name" href="#mloperand" id="ref-for-mloperand②④"><c- n>MLOperand</c-></a> <dfn class="dfn-paneled idl-code" data-dfn-for="MLGraphBuilder" data-dfn-type="method" data-export data-lt="constant(descriptor, bufferView)" id="dom-mlgraphbuilder-constant"><code><c- g>constant</c-></code></dfn>(<a data-link-type="idl-name" href="#dictdef-mloperanddescriptor" id="ref-for-dictdef-mloperanddescriptor⑧"><c- n>MLOperandDescriptor</c-></a> <dfn class="idl-code" data-dfn-for="MLGraphBuilder/constant(descriptor, bufferView)" data-dfn-type="argument" data-export id="dom-mlgraphbuilder-constant-descriptor-bufferview-descriptor"><code><c- g>descriptor</c-></code><a class="self-link" href="#dom-mlgraphbuilder-constant-descriptor-bufferview-descriptor"></a></dfn>, <a data-link-type="idl-name" href="#typedefdef-mlbufferview" id="ref-for-typedefdef-mlbufferview①"><c- n>MLBufferView</c-></a> <dfn class="idl-code" data-dfn-for="MLGraphBuilder/constant(descriptor, bufferView)" data-dfn-type="argument" data-export id="dom-mlgraphbuilder-constant-descriptor-bufferview-bufferview"><code><c- g>bufferView</c-></code><a class="self-link" href="#dom-mlgraphbuilder-constant-descriptor-bufferview-bufferview"></a></dfn>);
+  <a data-link-type="idl-name" href="#mloperand" id="ref-for-mloperand②⑥"><c- n>MLOperand</c-></a> <a class="idl-code" data-link-type="method" href="#dom-mlgraphbuilder-constant" id="ref-for-dom-mlgraphbuilder-constant①"><c- g>constant</c-></a>(<a data-link-type="idl-name" href="#dictdef-mloperanddescriptor" id="ref-for-dictdef-mloperanddescriptor⑧"><c- n>MLOperandDescriptor</c-></a> <dfn class="idl-code" data-dfn-for="MLGraphBuilder/constant(descriptor, bufferView)" data-dfn-type="argument" data-export id="dom-mlgraphbuilder-constant-descriptor-bufferview-descriptor"><code><c- g>descriptor</c-></code><a class="self-link" href="#dom-mlgraphbuilder-constant-descriptor-bufferview-descriptor"></a></dfn>, <a data-link-type="idl-name" href="#typedefdef-mlbufferview" id="ref-for-typedefdef-mlbufferview①"><c- n>MLBufferView</c-></a> <dfn class="idl-code" data-dfn-for="MLGraphBuilder/constant(descriptor, bufferView)" data-dfn-type="argument" data-export id="dom-mlgraphbuilder-constant-descriptor-bufferview-bufferview"><code><c- g>bufferView</c-></code><a class="self-link" href="#dom-mlgraphbuilder-constant-descriptor-bufferview-bufferview"></a></dfn>);
 
   // Create a single-value operand from the specified number of the specified type.
-  <a data-link-type="idl-name" href="#mloperand" id="ref-for-mloperand②⑤"><c- n>MLOperand</c-></a> <dfn class="dfn-paneled idl-code" data-dfn-for="MLGraphBuilder" data-dfn-type="method" data-export data-lt="constant(value, type)|constant(value)" id="dom-mlgraphbuilder-constant-value-type"><code><c- g>constant</c-></code></dfn>(<a class="idl-code" data-link-type="interface" href="https://webidl.spec.whatwg.org/#idl-double" id="ref-for-idl-double"><c- b>double</c-></a> <dfn class="idl-code" data-dfn-for="MLGraphBuilder/constant(value, type), MLGraphBuilder/constant(value)" data-dfn-type="argument" data-export id="dom-mlgraphbuilder-constant-value-type-value"><code><c- g>value</c-></code><a class="self-link" href="#dom-mlgraphbuilder-constant-value-type-value"></a></dfn>, <c- b>optional</c-> <a data-link-type="idl-name" href="#enumdef-mloperandtype" id="ref-for-enumdef-mloperandtype②"><c- n>MLOperandType</c-></a> <dfn class="idl-code" data-dfn-for="MLGraphBuilder/constant(value, type), MLGraphBuilder/constant(value)" data-dfn-type="argument" data-export id="dom-mlgraphbuilder-constant-value-type-type"><code><c- g>type</c-></code><a class="self-link" href="#dom-mlgraphbuilder-constant-value-type-type"></a></dfn> = "float32");
+  <a data-link-type="idl-name" href="#mloperand" id="ref-for-mloperand②⑦"><c- n>MLOperand</c-></a> <a class="idl-code" data-link-type="method" href="#dom-mlgraphbuilder-constant-value-type" id="ref-for-dom-mlgraphbuilder-constant-value-type"><c- g>constant</c-></a>(<a class="idl-code" data-link-type="interface" href="https://webidl.spec.whatwg.org/#idl-double" id="ref-for-idl-double"><c- b>double</c-></a> <dfn class="idl-code" data-dfn-for="MLGraphBuilder/constant(value, type), MLGraphBuilder/constant(value)" data-dfn-type="argument" data-export id="dom-mlgraphbuilder-constant-value-type-value"><code><c- g>value</c-></code><a class="self-link" href="#dom-mlgraphbuilder-constant-value-type-value"></a></dfn>, <c- b>optional</c-> <a data-link-type="idl-name" href="#enumdef-mloperandtype" id="ref-for-enumdef-mloperandtype②"><c- n>MLOperandType</c-></a> <dfn class="idl-code" data-dfn-for="MLGraphBuilder/constant(value, type), MLGraphBuilder/constant(value)" data-dfn-type="argument" data-export id="dom-mlgraphbuilder-constant-value-type-type"><code><c- g>type</c-></code><a class="self-link" href="#dom-mlgraphbuilder-constant-value-type-type"></a></dfn> = "float32");
 
   // Compile the graph up to the specified output operands asynchronously.
-  <a class="idl-code" data-link-type="interface" href="https://webidl.spec.whatwg.org/#idl-promise" id="ref-for-idl-promise④"><c- b>Promise</c-></a>&lt;<a data-link-type="idl-name" href="#mlgraph" id="ref-for-mlgraph②④"><c- n>MLGraph</c-></a>> <dfn class="dfn-paneled idl-code" data-dfn-for="MLGraphBuilder" data-dfn-type="method" data-export data-lt="build(outputs)" id="dom-mlgraphbuilder-build"><code><c- g>build</c-></code></dfn>(<a data-link-type="idl-name" href="#typedefdef-mlnamedoperands" id="ref-for-typedefdef-mlnamedoperands"><c- n>MLNamedOperands</c-></a> <dfn class="idl-code" data-dfn-for="MLGraphBuilder/build(outputs)" data-dfn-type="argument" data-export id="dom-mlgraphbuilder-build-outputs-outputs"><code><c- g>outputs</c-></code><a class="self-link" href="#dom-mlgraphbuilder-build-outputs-outputs"></a></dfn>);
+  <a class="idl-code" data-link-type="interface" href="https://webidl.spec.whatwg.org/#idl-promise" id="ref-for-idl-promise④"><c- b>Promise</c-></a>&lt;<a data-link-type="idl-name" href="#mlgraph" id="ref-for-mlgraph②④"><c- n>MLGraph</c-></a>> <a class="idl-code" data-link-type="method" href="#dom-mlgraphbuilder-build" id="ref-for-dom-mlgraphbuilder-build①"><c- g>build</c-></a>(<a data-link-type="idl-name" href="#typedefdef-mlnamedoperands" id="ref-for-typedefdef-mlnamedoperands"><c- n>MLNamedOperands</c-></a> <dfn class="idl-code" data-dfn-for="MLGraphBuilder/build(outputs)" data-dfn-type="argument" data-export id="dom-mlgraphbuilder-build-outputs-outputs"><code><c- g>outputs</c-></code><a class="self-link" href="#dom-mlgraphbuilder-build-outputs-outputs"></a></dfn>);
 
   // Compile the graph up to the specified output operands synchronously.
   [<a class="idl-code" data-link-type="extended-attribute" href="https://webidl.spec.whatwg.org/#Exposed" id="ref-for-Exposed①⓪"><c- g>Exposed</c-></a>=(<c- n>DedicatedWorker</c->)]
-  <a data-link-type="idl-name" href="#mlgraph" id="ref-for-mlgraph②⑤"><c- n>MLGraph</c-></a> <dfn class="dfn-paneled idl-code" data-dfn-for="MLGraphBuilder" data-dfn-type="method" data-export data-lt="buildSync(outputs)" id="dom-mlgraphbuilder-buildsync"><code><c- g>buildSync</c-></code></dfn>(<a data-link-type="idl-name" href="#typedefdef-mlnamedoperands" id="ref-for-typedefdef-mlnamedoperands①"><c- n>MLNamedOperands</c-></a> <dfn class="idl-code" data-dfn-for="MLGraphBuilder/buildSync(outputs)" data-dfn-type="argument" data-export id="dom-mlgraphbuilder-buildsync-outputs-outputs"><code><c- g>outputs</c-></code><a class="self-link" href="#dom-mlgraphbuilder-buildsync-outputs-outputs"></a></dfn>);
+  <a data-link-type="idl-name" href="#mlgraph" id="ref-for-mlgraph②⑤"><c- n>MLGraph</c-></a> <a class="idl-code" data-link-type="method" href="#dom-mlgraphbuilder-buildsync" id="ref-for-dom-mlgraphbuilder-buildsync①"><c- g>buildSync</c-></a>(<a data-link-type="idl-name" href="#typedefdef-mlnamedoperands" id="ref-for-typedefdef-mlnamedoperands①"><c- n>MLNamedOperands</c-></a> <dfn class="idl-code" data-dfn-for="MLGraphBuilder/buildSync(outputs)" data-dfn-type="argument" data-export id="dom-mlgraphbuilder-buildsync-outputs-outputs"><code><c- g>outputs</c-></code><a class="self-link" href="#dom-mlgraphbuilder-buildsync-outputs-outputs"></a></dfn>);
 };
 </pre>
-   <div class="note" role="note"> Both <code class="idl"><a data-link-type="idl" href="#mlgraphbuilder" id="ref-for-mlgraphbuilder①②">MLGraphBuilder</a></code>.<code class="idl"><a data-link-type="idl" href="#dom-mlgraphbuilder-build" id="ref-for-dom-mlgraphbuilder-build①">build()</a></code> and <code class="idl"><a data-link-type="idl" href="#mlgraphbuilder" id="ref-for-mlgraphbuilder①③">MLGraphBuilder</a></code>.<code class="idl"><a data-link-type="idl" href="#dom-mlgraphbuilder-buildsync" id="ref-for-dom-mlgraphbuilder-buildsync①">buildSync()</a></code> methods compile the graph builder state up to the specified output operands into a compiled graph according to the type of <code class="idl"><a data-link-type="idl" href="#mlcontext" id="ref-for-mlcontext③⑧">MLContext</a></code> that creates it. Since this operation can be costly in some machine configurations, the calling thread of the <code class="idl"><a data-link-type="idl" href="#mlgraphbuilder" id="ref-for-mlgraphbuilder①④">MLGraphBuilder</a></code>.<code class="idl"><a data-link-type="idl" href="#dom-mlgraphbuilder-buildsync" id="ref-for-dom-mlgraphbuilder-buildsync②">buildSync()</a></code> method must only be a worker thread to avoid potential disruption of the user experience. When the <code class="idl"><a data-link-type="idl" href="#dom-mlcontext-contexttype-slot" id="ref-for-dom-mlcontext-contexttype-slot⑦">[[contextType]]</a></code> of the <code class="idl"><a data-link-type="idl" href="#mlcontext" id="ref-for-mlcontext③⑨">MLContext</a></code> is set to <a data-link-type="dfn" href="#default-context" id="ref-for-default-context⑤">default</a>, the compiled graph is initialized right before the <code class="idl"><a data-link-type="idl" href="#mlgraph" id="ref-for-mlgraph②⑥">MLGraph</a></code> is returned. This graph initialization stage is important for optimal performance of the subsequent graph executions. See <a href="#api-mlcommandencoder-graph-initialization">§ 7.5.1 Graph Initialization</a> for more detail. </div>
-   <p><code class="idl"><a data-link-type="idl" href="#mlgraphbuilder" id="ref-for-mlgraphbuilder①⑤">MLGraphBuilder</a></code> has the following internal slots:</p>
+   <div class="note" role="note"> Both <code class="idl"><a data-link-type="idl" href="#mlgraphbuilder" id="ref-for-mlgraphbuilder①⑥">MLGraphBuilder</a></code>.<code class="idl"><a data-link-type="idl" href="#dom-mlgraphbuilder-build" id="ref-for-dom-mlgraphbuilder-build②">build()</a></code> and <code class="idl"><a data-link-type="idl" href="#mlgraphbuilder" id="ref-for-mlgraphbuilder①⑦">MLGraphBuilder</a></code>.<code class="idl"><a data-link-type="idl" href="#dom-mlgraphbuilder-buildsync" id="ref-for-dom-mlgraphbuilder-buildsync②">buildSync()</a></code> methods compile the graph builder state up to the specified output operands into a compiled graph according to the type of <code class="idl"><a data-link-type="idl" href="#mlcontext" id="ref-for-mlcontext③⑦">MLContext</a></code> that creates it. Since this operation can be costly in some machine configurations, the calling thread of the <code class="idl"><a data-link-type="idl" href="#mlgraphbuilder" id="ref-for-mlgraphbuilder①⑧">MLGraphBuilder</a></code>.<code class="idl"><a data-link-type="idl" href="#dom-mlgraphbuilder-buildsync" id="ref-for-dom-mlgraphbuilder-buildsync③">buildSync()</a></code> method must only be a worker thread to avoid potential disruption of the user experience. When the <code class="idl"><a data-link-type="idl" href="#dom-mlcontext-contexttype-slot" id="ref-for-dom-mlcontext-contexttype-slot⑦">[[contextType]]</a></code> of the <code class="idl"><a data-link-type="idl" href="#mlcontext" id="ref-for-mlcontext③⑧">MLContext</a></code> is set to "<a data-link-type="dfn" href="#default-context" id="ref-for-default-context⑤">default</a>", the compiled graph is initialized right before the <code class="idl"><a data-link-type="idl" href="#mlgraph" id="ref-for-mlgraph②⑥">MLGraph</a></code> is returned. This graph initialization stage is important for optimal performance of the subsequent graph executions. See <a href="#api-mlcommandencoder-graph-initialization">§ 7.5.1 Graph Initialization</a> for more detail. </div>
+   <p><code class="idl"><a data-link-type="idl" href="#dictdef-mlbufferresourceview" id="ref-for-dictdef-mlbufferresourceview①">MLBufferResourceView</a></code> has the following members:</p>
    <dl>
-    <dt data-md><dfn class="dfn-paneled idl-code" data-dfn-for="MLGraphBuilder" data-dfn-type="attribute" data-export id="dom-mlgraphbuilder-context-slot"><code>[[context]]</code></dfn> of type <code class="idl"><a data-link-type="idl" href="#mlcontext" id="ref-for-mlcontext④⓪">MLContext</a></code>
+    <dt data-md><dfn class="dfn-paneled idl-code" data-dfn-for="MLBufferResourceView" data-dfn-type="dict-member" data-export id="dom-mlbufferresourceview-resource"><code>resource</code></dfn>, <span> of type <a data-link-type="idl-name" href="https://gpuweb.github.io/gpuweb/#gpubuffer" id="ref-for-gpubuffer⑤">GPUBuffer</a></span>
+    <dd data-md>
+     <p>A <code class="idl"><a data-link-type="idl" href="https://gpuweb.github.io/gpuweb/#gpubuffer" id="ref-for-gpubuffer⑥">GPUBuffer</a></code> object. Specifies the GPU buffer source.</p>
+    <dt data-md><dfn class="dfn-paneled idl-code" data-dfn-for="MLBufferResourceView" data-dfn-type="dict-member" data-export id="dom-mlbufferresourceview-offset"><code>offset</code></dfn>, <span> of type <a data-link-type="idl-name" href="https://webidl.spec.whatwg.org/#idl-unsigned-long-long" id="ref-for-idl-unsigned-long-long②">unsigned long long</a>, defaulting to <code>0</code></span>
     <dd data-md>
-     <p>The context of type <code class="idl"><a data-link-type="idl" href="#mlcontext" id="ref-for-mlcontext④①">MLContext</a></code> associated with this <code class="idl"><a data-link-type="idl" href="#mlgraphbuilder" id="ref-for-mlgraphbuilder①⑥">MLGraphBuilder</a></code>.</p>
+     <p>Specifies an <code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#idl-unsigned-long-long" id="ref-for-idl-unsigned-long-long③">unsigned long long</a></code> offset in the buffer source.</p>
+    <dt data-md><dfn class="dfn-paneled idl-code" data-dfn-for="MLBufferResourceView" data-dfn-type="dict-member" data-export id="dom-mlbufferresourceview-size"><code>size</code></dfn>, <span> of type <a data-link-type="idl-name" href="https://webidl.spec.whatwg.org/#idl-unsigned-long-long" id="ref-for-idl-unsigned-long-long④">unsigned long long</a></span>
+    <dd data-md>
+     <p>Specifies the <code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#idl-unsigned-long-long" id="ref-for-idl-unsigned-long-long⑤">unsigned long long</a></code> size of the buffer view.</p>
    </dl>
-   <h4 class="heading settled" data-level="7.6.1" id="api-mlgraphbuilder-constructor"><span class="secno">7.6.1. </span><span class="content">The <code class="idl"><a data-link-type="idl" href="#mlgraphbuilder" id="ref-for-mlgraphbuilder①⑦">MLGraphBuilder</a></code> constructor</span><a class="self-link" href="#api-mlgraphbuilder-constructor"></a></h4>
-    The <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#new" id="ref-for-new">new</a> <code class="idl"><a data-link-type="idl" href="#mlgraphbuilder" id="ref-for-mlgraphbuilder①⑧">MLGraphBuilder</a></code> constructor steps are: 
-   <ol>
-    <li data-md>
-     <p>If <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#this" id="ref-for-this②">this</a>'s <a data-link-type="dfn" href="https://html.spec.whatwg.org/multipage/webappapis.html#concept-relevant-global" id="ref-for-concept-relevant-global②">relevant global object</a>'s <a data-link-type="dfn" href="https://html.spec.whatwg.org/multipage/nav-history-apis.html#concept-document-window" id="ref-for-concept-document-window②">associated Document</a> is not <a data-link-type="dfn" href="https://html.spec.whatwg.org/multipage/iframe-embed-object.html#allowed-to-use" id="ref-for-allowed-to-use②">allowed to use</a> the <a data-link-type="dfn" href="#webnn-feature" id="ref-for-webnn-feature②">webnn</a> feature, throw a "<code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#securityerror" id="ref-for-securityerror②">SecurityError</a></code>" <code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#idl-DOMException" id="ref-for-idl-DOMException②①">DOMException</a></code> and abort these steps.</p>
-    <li data-md>
-     <p>Let <var>context</var> be the first argument.</p>
-    <li data-md>
-     <p>If the <a data-link-type="dfn" href="#validate-mlcontext" id="ref-for-validate-mlcontext②">validate MLContext</a> steps given <var>context</var> return <code>false</code>, throw a "<code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#exceptiondef-typeerror" id="ref-for-exceptiondef-typeerror②">TypeError</a></code>" and abort these steps.</p>
-    <li data-md>
-     <p>Set <code class="idl"><a data-link-type="idl" href="#dom-mlgraphbuilder-context-slot" id="ref-for-dom-mlgraphbuilder-context-slot">[[context]]</a></code> to <var>context</var>.</p>
-   </ol>
-   <h4 class="heading settled" data-level="7.6.2" id="api-mlgraphbuilder-input"><span class="secno">7.6.2. </span><span class="content">The <code class="idl"><a data-link-type="idl" href="#dom-mlgraphbuilder-input" id="ref-for-dom-mlgraphbuilder-input">input()</a></code> method</span><a class="self-link" href="#api-mlgraphbuilder-input"></a></h4>
-    Create a named <code class="idl"><a data-link-type="idl" href="#mloperand" id="ref-for-mloperand②⑥">MLOperand</a></code> based on a descriptor, that can be used as an input. 
-   <div class="note" role="note">
+   <div class="internal-slots">
+     <code class="idl"><a data-link-type="idl" href="#mlgraphbuilder" id="ref-for-mlgraphbuilder①⑨">MLGraphBuilder</a></code> has the following internal slots: 
+    <dl>
+     <dt data-md><dfn class="dfn-paneled idl-code" data-dfn-for="MLGraphBuilder" data-dfn-type="attribute" data-export id="dom-mlgraphbuilder-context-slot"><code>[[context]]</code></dfn> of type <code class="idl"><a data-link-type="idl" href="#mlcontext" id="ref-for-mlcontext③⑨">MLContext</a></code>
+     <dd data-md>
+      <p>The context of type <code class="idl"><a data-link-type="idl" href="#mlcontext" id="ref-for-mlcontext④⓪">MLContext</a></code> associated with this <code class="idl"><a data-link-type="idl" href="#mlgraphbuilder" id="ref-for-mlgraphbuilder②⓪">MLGraphBuilder</a></code>.</p>
+    </dl>
+   </div>
+   <h4 class="heading settled" data-level="7.6.1" id="api-mlgraphbuilder-constructor"><span class="secno">7.6.1. </span><span class="content">The <code class="idl"><a data-link-type="idl" href="#mlgraphbuilder" id="ref-for-mlgraphbuilder②①">MLGraphBuilder</a></code> constructor</span><a class="self-link" href="#api-mlgraphbuilder-constructor"></a></h4>
+   <details class="algorithm" data-algorithm="MLGraphBuilder(context)" data-algorithm-for="MLGraphBuilder" open>
+    <summary> The <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#new" id="ref-for-new">new</a> <dfn class="dfn-paneled idl-code" data-dfn-for="MLGraphBuilder" data-dfn-type="constructor" data-export data-lt="MLGraphBuilder(context)|constructor(context)" id="dom-mlgraphbuilder-mlgraphbuilder"><code>MLGraphBuilder(context)</code></dfn> constructor steps are: </summary>
+    <div class="algorithm-steps">
+     <ol>
+      <li data-md>
+       <p>If <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#this" id="ref-for-this④">this</a>'s <a data-link-type="dfn" href="https://html.spec.whatwg.org/multipage/webappapis.html#concept-relevant-global" id="ref-for-concept-relevant-global④">relevant global object</a>'s <a data-link-type="dfn" href="https://html.spec.whatwg.org/multipage/nav-history-apis.html#concept-document-window" id="ref-for-concept-document-window④">associated Document</a> is not <a data-link-type="dfn" href="https://html.spec.whatwg.org/multipage/iframe-embed-object.html#allowed-to-use" id="ref-for-allowed-to-use④">allowed to use</a> the <a data-link-type="dfn" href="#webnn-feature" id="ref-for-webnn-feature④">webnn</a> feature, then <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#dfn-throw" id="ref-for-dfn-throw②⓪">throw</a> a "<code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#securityerror" id="ref-for-securityerror④">SecurityError</a></code>" <code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#idl-DOMException" id="ref-for-idl-DOMException②⑤">DOMException</a></code>.</p>
+      <li data-md>
+       <p>If <a data-link-type="dfn" href="#validate-mlcontext" id="ref-for-validate-mlcontext④">validating MLContext</a> given <var>context</var> returns false, then <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#dfn-throw" id="ref-for-dfn-throw②①">throw</a> a "<code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#exceptiondef-typeerror" id="ref-for-exceptiondef-typeerror①">TypeError</a></code>".</p>
+      <li data-md>
+       <p>Set <code class="idl"><a data-link-type="idl" href="#dom-mlgraphbuilder-context-slot" id="ref-for-dom-mlgraphbuilder-context-slot">[[context]]</a></code> to <var>context</var>.</p>
+     </ol>
+    </div>
+   </details>
+   <h4 class="heading settled" data-level="7.6.2" id="api-mlgraphbuilder-input"><span class="secno">7.6.2. </span><span class="content">The <code class="idl"><a data-link-type="idl" href="#dom-mlgraphbuilder-input" id="ref-for-dom-mlgraphbuilder-input①">input()</a></code> method</span><a class="self-link" href="#api-mlgraphbuilder-input"></a></h4>
+    Create a named <code class="idl"><a data-link-type="idl" href="#mloperand" id="ref-for-mloperand②⑧">MLOperand</a></code> based on a descriptor, that can be used as an input. 
+   <div>
      <strong>Arguments:</strong> 
     <ul>
      <li data-md>
-      <p><em>name</em>: a <a data-link-type="dfn" href="https://infra.spec.whatwg.org/#string" id="ref-for-string①">string</a> name of the input.</p>
+      <p><em>name</em>: a <a data-link-type="dfn" href="https://infra.spec.whatwg.org/#string" id="ref-for-string②">string</a> name of the input.</p>
      <li data-md>
       <p><em>descriptor</em>: an <code class="idl"><a data-link-type="idl" href="#dictdef-mloperanddescriptor" id="ref-for-dictdef-mloperanddescriptor⑨">MLOperandDescriptor</a></code> object.</p>
     </ul>
-     <strong>Returns:</strong>: an <code class="idl"><a data-link-type="idl" href="#mloperand" id="ref-for-mloperand②⑦">MLOperand</a></code> object. 
+     <strong>Returns:</strong>: an <code class="idl"><a data-link-type="idl" href="#mloperand" id="ref-for-mloperand②⑨">MLOperand</a></code> object. 
    </div>
-   <div class="algorithm" data-algorithm="input">
-     The <code class="idl"><a data-link-type="idl" href="#dom-mlgraphbuilder-input" id="ref-for-dom-mlgraphbuilder-input①">input(name, descriptor)</a></code> steps are: 
-    <div class="note" role="note"> The permissions and context validity have been checked by <a href="#api-mlgraphbuilder-constructor">§ 7.6.1 The MLGraphBuilder constructor</a> steps. </div>
-    <ol>
-     <li data-md>
-      <p>Let <var>name</var> be the first argument.</p>
-      <ol>
-       <li data-md>
-        <p>If <var>name</var> is <code>undefined</code> or an empty <a data-link-type="dfn" href="https://infra.spec.whatwg.org/#string" id="ref-for-string②">string</a>, then throw a "<code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#exceptiondef-typeerror" id="ref-for-exceptiondef-typeerror③">TypeError</a></code>" <code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#idl-DOMException" id="ref-for-idl-DOMException②②">DOMException</a></code> and stop.</p>
-      </ol>
-     <li data-md>
-      <p>Let <var>descriptor</var> be the second argument.</p>
-      <ol>
-       <li data-md>
-        <p>If <var>descriptor</var> is not an an object that <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#implements" id="ref-for-implements①">implements</a> <code class="idl"><a data-link-type="idl" href="#dictdef-mloperanddescriptor" id="ref-for-dictdef-mloperanddescriptor①⓪">MLOperandDescriptor</a></code>, then throw a "<code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#exceptiondef-typeerror" id="ref-for-exceptiondef-typeerror④">TypeError</a></code>" <code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#idl-DOMException" id="ref-for-idl-DOMException②③">DOMException</a></code> and stop.</p>
-       <li data-md>
-        <p><a data-link-type="dfn" href="https://infra.spec.whatwg.org/#assert" id="ref-for-assert④">Assert</a>: If <var>descriptor</var>.<code class="idl"><a data-link-type="idl" href="#dom-mloperanddescriptor-dimensions" id="ref-for-dom-mloperanddescriptor-dimensions②">dimensions</a></code> does not <a data-link-type="dfn" href="https://infra.spec.whatwg.org/#map-exists" id="ref-for-map-exists⑤">exist</a>, then <var>descriptor</var> defines a scalar input.</p>
-       <li data-md>
-        <p>If <var>descriptor</var>.<code class="idl"><a data-link-type="idl" href="#dom-mloperanddescriptor-dimensions" id="ref-for-dom-mloperanddescriptor-dimensions③">dimensions</a></code> <a data-link-type="dfn" href="https://infra.spec.whatwg.org/#map-exists" id="ref-for-map-exists⑥">exists</a>:</p>
-        <ol>
-         <li data-md>
-          <p>If the <a data-link-type="dfn" href="#check-dimensions" id="ref-for-check-dimensions">check dimensions</a> steps given <var>descriptor</var>.<code class="idl"><a data-link-type="idl" href="#dom-mloperanddescriptor-type" id="ref-for-dom-mloperanddescriptor-type②">type</a></code> and <var>descriptor</var>.<code class="idl"><a data-link-type="idl" href="#dom-mloperanddescriptor-dimensions" id="ref-for-dom-mloperanddescriptor-dimensions④">dimensions</a></code> return <code>false</code>, throw a "<code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#dataerror" id="ref-for-dataerror⑦">DataError</a></code>" <code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#idl-DOMException" id="ref-for-idl-DOMException②④">DOMException</a></code> and stop.</p>
-         <li data-md>
-          <p>If the <a data-link-type="dfn" href="#mloperanddescriptor-byte-length" id="ref-for-mloperanddescriptor-byte-length④">byte length</a> of <var>descriptor</var> is not supported by the underlying platform, then throw a "<code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#dataerror" id="ref-for-dataerror⑧">DataError</a></code>" <code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#idl-DOMException" id="ref-for-idl-DOMException②⑤">DOMException</a></code> and stop.</p>
-        </ol>
-      </ol>
-     <li data-md>
-      <p>Let <var>operand</var> be the result of invoking the <a data-link-type="dfn" href="#create-mloperand" id="ref-for-create-mloperand">create MLOperand</a> steps with <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#this" id="ref-for-this③">this</a> and <var>descriptor</var>.</p>
-      <ol>
-       <li data-md>
-        <p>If that throws, re-throw the exception and stop.</p>
-      </ol>
-     <li data-md>
-      <p>Set <var>operand</var>.<code class="idl"><a data-link-type="idl" href="#dom-mloperand-name-slot" id="ref-for-dom-mloperand-name-slot">[[name]]</a></code> to <var>name</var>.</p>
-     <li data-md>
-      <p>Make a request to the underlying platform to register <var>operand</var> as an input and store a reference to the corresponding <a data-link-type="dfn" href="https://infra.spec.whatwg.org/#implementation-defined" id="ref-for-implementation-defined②">implementation-defined</a> platform object in <var>operand</var>.<code class="idl"><a data-link-type="idl" href="#dom-mloperand-operand-slot" id="ref-for-dom-mloperand-operand-slot">[[operand]]</a></code>.</p>
-      <ol>
-       <li data-md>
-        <p>If that fails, throw an "<code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#operationerror" id="ref-for-operationerror⑧">OperationError</a></code>" <code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#idl-DOMException" id="ref-for-idl-DOMException②⑥">DOMException</a></code> and abort these steps.</p>
-      </ol>
-    </ol>
-    <h4 class="heading settled" data-level="7.6.3" id="api-mlgraphbuilder-constant-method"><span class="secno">7.6.3. </span><span class="content">The constant() method</span><a class="self-link" href="#api-mlgraphbuilder-constant-method"></a></h4>
-     Create a constant <code class="idl"><a data-link-type="idl" href="#mloperand" id="ref-for-mloperand②⑧">MLOperand</a></code> that can be used in <code class="idl"><a data-link-type="idl" href="#mlgraphbuilder" id="ref-for-mlgraphbuilder①⑨">MLGraphBuilder</a></code> methods. 
-    <h5 class="heading settled" data-level="7.6.3.1" id="api-mlgraphbuilder-constant"><span class="secno">7.6.3.1. </span><span class="content">The <code class="idl"><a data-link-type="idl" href="#dom-mlgraphbuilder-constant" id="ref-for-dom-mlgraphbuilder-constant①">constant(descriptor, bufferView)</a></code> method</span><a class="self-link" href="#api-mlgraphbuilder-constant"></a></h5>
-    <div class="note" role="note">
-      <strong>Arguments:</strong> 
-     <ul>
-      <li data-md>
-       <p><em>descriptor</em>: an <code class="idl"><a data-link-type="idl" href="#dictdef-mloperanddescriptor" id="ref-for-dictdef-mloperanddescriptor①①">MLOperandDescriptor</a></code> object</p>
-      <li data-md>
-       <p><em>bufferView</em>: an <code class="idl"><a data-link-type="idl" href="#typedefdef-mlbufferview" id="ref-for-typedefdef-mlbufferview②">MLBufferView</a></code></p>
-     </ul>
-      <strong>Returns:</strong>: an <code class="idl"><a data-link-type="idl" href="#mloperand" id="ref-for-mloperand②⑨">MLOperand</a></code> object. 
-    </div>
-    <div class="algorithm-steps algorithm" data-algorithm="constant">
-      The <code class="idl"><a data-link-type="idl" href="#dom-mlgraphbuilder-constant" id="ref-for-dom-mlgraphbuilder-constant②">constant(descriptor, bufferView)</a></code> steps are: 
+   <details class="algorithm" data-algorithm="input(name, descriptor)" data-algorithm-for="MLGraphBuilder" open>
+    <summary> The <dfn class="dfn-paneled idl-code" data-dfn-for="MLGraphBuilder" data-dfn-type="method" data-export id="dom-mlgraphbuilder-input"><code>input(<var>name</var>, <var>descriptor</var>)</code></dfn> method steps are: </summary>
+    <div class="algorithm-steps">
      <div class="note" role="note"> The permissions and context validity have been checked by <a href="#api-mlgraphbuilder-constructor">§ 7.6.1 The MLGraphBuilder constructor</a> steps. </div>
      <ol>
       <li data-md>
-       <p>Let <var>descriptor</var> be the first argument.</p>
+       <p>If <var>name</var> is empty, then <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#dfn-throw" id="ref-for-dfn-throw②②">throw</a> a <code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#exceptiondef-typeerror" id="ref-for-exceptiondef-typeerror②">TypeError</a></code>.</p>
+      <li data-md>
+       <p><a data-link-type="dfn" href="https://infra.spec.whatwg.org/#assert" id="ref-for-assert⑨">Assert</a>: the type of <var>descriptor</var> is <code class="idl"><a data-link-type="idl" href="#dictdef-mloperanddescriptor" id="ref-for-dictdef-mloperanddescriptor①⓪">MLOperandDescriptor</a></code>.</p>
       <li data-md>
-       <p>If <var>descriptor</var> is not an an object that <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#implements" id="ref-for-implements②">implements</a> <code class="idl"><a data-link-type="idl" href="#dictdef-mloperanddescriptor" id="ref-for-dictdef-mloperanddescriptor①②">MLOperandDescriptor</a></code>, then throw a "<code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#exceptiondef-typeerror" id="ref-for-exceptiondef-typeerror⑤">TypeError</a></code>" <code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#idl-DOMException" id="ref-for-idl-DOMException②⑦">DOMException</a></code> and stop.</p>
+       <p><a data-link-type="dfn" href="https://infra.spec.whatwg.org/#assert" id="ref-for-assert①⓪">Assert</a>: If <var>descriptor</var>.<code class="idl"><a data-link-type="idl" href="#dom-mloperanddescriptor-dimensions" id="ref-for-dom-mloperanddescriptor-dimensions⑤">dimensions</a></code> does not <a data-link-type="dfn" href="https://infra.spec.whatwg.org/#map-exists" id="ref-for-map-exists⑦">exist</a>, then <var>descriptor</var> defines a scalar input.</p>
+      <li data-md>
+       <p>If <var>descriptor</var>.<code class="idl"><a data-link-type="idl" href="#dom-mloperanddescriptor-dimensions" id="ref-for-dom-mloperanddescriptor-dimensions⑥">dimensions</a></code> <a data-link-type="dfn" href="https://infra.spec.whatwg.org/#map-exists" id="ref-for-map-exists⑧">exists</a>:</p>
        <ol>
         <li data-md>
-         <p>If the <a data-link-type="dfn" href="#mloperanddescriptor-byte-length" id="ref-for-mloperanddescriptor-byte-length⑤">byte length</a> of <var>descriptor</var> is not supported by the underlying platform, then throw a "<code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#dataerror" id="ref-for-dataerror⑨">DataError</a></code>" <code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#idl-DOMException" id="ref-for-idl-DOMException②⑧">DOMException</a></code> and stop.</p>
+         <p>If the <a data-link-type="dfn" href="#check-dimensions" id="ref-for-check-dimensions①">check dimensions</a> steps given <var>descriptor</var>.<code class="idl"><a data-link-type="idl" href="#dom-mloperanddescriptor-type" id="ref-for-dom-mloperanddescriptor-type③">type</a></code> and <var>descriptor</var>.<code class="idl"><a data-link-type="idl" href="#dom-mloperanddescriptor-dimensions" id="ref-for-dom-mloperanddescriptor-dimensions⑦">dimensions</a></code> return false, then <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#dfn-throw" id="ref-for-dfn-throw②③">throw</a> a "<code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#dataerror" id="ref-for-dataerror⑦">DataError</a></code>" <code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#idl-DOMException" id="ref-for-idl-DOMException②⑥">DOMException</a></code>.</p>
         <li data-md>
-         <p>If the <a data-link-type="dfn" href="#check-dimensions" id="ref-for-check-dimensions①">check dimensions</a> steps given <var>descriptor</var>.<code class="idl"><a data-link-type="idl" href="#dom-mloperanddescriptor-type" id="ref-for-dom-mloperanddescriptor-type③">type</a></code> and <var>descriptor</var>.<code class="idl"><a data-link-type="idl" href="#dom-mloperanddescriptor-dimensions" id="ref-for-dom-mloperanddescriptor-dimensions⑤">dimensions</a></code> return <code>false</code>, throw a "<code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#dataerror" id="ref-for-dataerror①⓪">DataError</a></code>" <code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#idl-DOMException" id="ref-for-idl-DOMException②⑨">DOMException</a></code> and stop.</p>
+         <p>If the <a data-link-type="dfn" href="#mloperanddescriptor-byte-length" id="ref-for-mloperanddescriptor-byte-length④">byte length</a> of <var>descriptor</var> is not supported by the underlying platform, then <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#dfn-throw" id="ref-for-dfn-throw②④">throw</a> a "<code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#dataerror" id="ref-for-dataerror⑧">DataError</a></code>" <code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#idl-DOMException" id="ref-for-idl-DOMException②⑦">DOMException</a></code>.</p>
        </ol>
       <li data-md>
-       <p>Let <var>bufferView</var> be the second argument.</p>
+       <p>If any of the following sub-steps fail, <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#dfn-throw" id="ref-for-dfn-throw②⑤">throw</a> an "<code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#operationerror" id="ref-for-operationerror①⓪">OperationError</a></code>" <code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#idl-DOMException" id="ref-for-idl-DOMException②⑧">DOMException</a></code>.</p>
        <ol>
         <li data-md>
-         <p>If invoking <a data-link-type="dfn" href="#validate-buffer-with-descriptor" id="ref-for-validate-buffer-with-descriptor①">validate buffer with descriptor</a> given <var>bufferView</var> and <var>descriptor</var> return <code>false</code>, then throw a "<code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#exceptiondef-typeerror" id="ref-for-exceptiondef-typeerror⑥">TypeError</a></code>" <code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#idl-DOMException" id="ref-for-idl-DOMException③⓪">DOMException</a></code> and stop.</p>
+         <p>Let <var>operand</var> be the result of <a data-link-type="dfn" href="#create-an-mloperand" id="ref-for-create-an-mloperand">creating an MLOperand</a> given <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#this" id="ref-for-this⑤">this</a> and <var>descriptor</var>.</p>
+        <li data-md>
+         <p>Set <var>operand</var>.<code class="idl"><a data-link-type="idl" href="#dom-mloperand-name-slot" id="ref-for-dom-mloperand-name-slot③">[[name]]</a></code> to <var>name</var>.</p>
+        <li data-md>
+         <p>Make a request to the underlying platform to:</p>
+         <ol>
+          <li data-md>
+           <p>Create an <a data-link-type="dfn" href="https://infra.spec.whatwg.org/#implementation-defined" id="ref-for-implementation-defined⑤">implementation-defined</a> platform input operand <var>operandImpl</var> given <var>descriptor</var>.</p>
+          <li data-md>
+           <p>Store a reference of <var>operandImpl</var> in <var>operand</var>.<code class="idl"><a data-link-type="idl" href="#dom-mloperand-operand-slot" id="ref-for-dom-mloperand-operand-slot">[[operand]]</a></code>.</p>
+          <li data-md>
+           <p>Register <var>operand</var> as an input.</p>
+         </ol>
        </ol>
       <li data-md>
-       <p>Let <var>operand</var> be the result of invoking the <a data-link-type="dfn" href="#create-mloperand" id="ref-for-create-mloperand①">create MLOperand</a> steps with <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#this" id="ref-for-this④">this</a> and <var>descriptor</var>.</p>
+       <p>Return <var>operand</var>.</p>
+     </ol>
+    </div>
+   </details>
+   <h4 class="heading settled" data-level="7.6.3" id="api-mlgraphbuilder-build"><span class="secno">7.6.3. </span><span class="content">The build() method</span><a class="self-link" href="#api-mlgraphbuilder-build"></a></h4>
+    Build a composed graph up to a given output operand into a computational graph, asynchronously or synchronously. 
+   <h5 class="heading settled" data-level="7.6.3.1" id="api-mlgraphbuilder-build-outputs"><span class="secno">7.6.3.1. </span><span class="content">The <code class="idl"><a data-link-type="idl" href="#dom-mlgraphbuilder-build" id="ref-for-dom-mlgraphbuilder-build③">build(outputs)</a></code> method</span><a class="self-link" href="#api-mlgraphbuilder-build-outputs"></a></h5>
+   <details class="algorithm" data-algorithm="build(outputs)" data-algorithm-for="MLGraphBuilder" open>
+    <summary> The <dfn class="dfn-paneled idl-code" data-dfn-for="MLGraphBuilder" data-dfn-type="method" data-export id="dom-mlgraphbuilder-build"><code>build(<var>outputs</var>)</code></dfn> method steps are: </summary>
+    <div class="algorithm-steps">
+     <div class="note" role="note"> The permissions and context validity have been checked by <a href="#api-mlgraphbuilder-constructor">§ 7.6.1 The MLGraphBuilder constructor</a> steps. </div>
+     <ol>
+      <li data-md>
+       <p>Let <var>promise</var> be <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#a-new-promise" id="ref-for-a-new-promise⑤">a new promise</a>.</p>
+      <li data-md>
+       <p>Return <var>promise</var> and run the following steps <a data-link-type="dfn" href="https://html.spec.whatwg.org/multipage/infrastructure.html#in-parallel" id="ref-for-in-parallel③">in parallel</a>.</p>
+      <li data-md>
+       <p>Return the result of invoking <code class="idl"><a data-link-type="idl" href="#dom-mlgraphbuilder-buildsync" id="ref-for-dom-mlgraphbuilder-buildsync④">buildSync(outputs)</a></code> given <var>outputs</var>.</p>
        <ol>
         <li data-md>
-         <p>If that throws, re-throw the exception and stop.</p>
+         <p>If that <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#dfn-throw" id="ref-for-dfn-throw②⑥">throws</a>, re-<a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#dfn-throw" id="ref-for-dfn-throw②⑦">throw</a> the error.</p>
        </ol>
+     </ol>
+    </div>
+   </details>
+   <h5 class="heading settled" data-level="7.6.3.2" id="api-mlgraphbuilder-buildsync-outputs"><span class="secno">7.6.3.2. </span><span class="content">The <code class="idl"><a data-link-type="idl" href="#dom-mlgraphbuilder-buildsync" id="ref-for-dom-mlgraphbuilder-buildsync⑤">buildSync(outputs)</a></code> method</span><a class="self-link" href="#api-mlgraphbuilder-buildsync-outputs"></a></h5>
+   <details class="algorithm" data-algorithm="buildSync(outputs)" data-algorithm-for="MLGraphBuilder" open>
+    <summary> The <dfn class="dfn-paneled idl-code" data-dfn-for="MLGraphBuilder" data-dfn-type="method" data-export id="dom-mlgraphbuilder-buildsync"><code>buildSync(<var>outputs</var>)</code></dfn> method steps are: </summary>
+    <div class="algorithm-steps">
+     <div class="note" role="note"> The permissions and context validity have been checked by <a href="#api-mlgraphbuilder-constructor">§ 7.6.1 The MLGraphBuilder constructor</a> steps. </div>
+     <ol>
       <li data-md>
-       <p>Let <var>bytes</var> be the result of invoking [<a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#dfn-get-buffer-source-copy" id="ref-for-dfn-get-buffer-source-copy">get a copy of the bytes held by the buffer source</a>] given <var>bufferView</var>.</p>
+       <p><a data-link-type="dfn" href="https://infra.spec.whatwg.org/#assert" id="ref-for-assert①①">Assert</a>: the type of <var>outputs</var> is <code class="idl"><a data-link-type="idl" href="#typedefdef-mlnamedoperands" id="ref-for-typedefdef-mlnamedoperands②">MLNamedOperands</a></code>.</p>
       <li data-md>
-       <p>Make a request to the underlying platform to register <var>operand</var> as a tensor constant with <var>bytes</var> as value and store a reference to the corresponding <a data-link-type="dfn" href="https://infra.spec.whatwg.org/#implementation-defined" id="ref-for-implementation-defined③">implementation-defined</a> object to <var>operand</var>.<code class="idl"><a data-link-type="idl" href="#dom-mloperand-operand-slot" id="ref-for-dom-mloperand-operand-slot①">[[operand]]</a></code>.</p>
+       <p>If <var>outputs</var> is empty, then <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#dfn-throw" id="ref-for-dfn-throw②⑧">throw</a> a <code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#exceptiondef-typeerror" id="ref-for-exceptiondef-typeerror③">TypeError</a></code>.</p>
+      <li data-md>
+       <p><a data-link-type="dfn" href="https://infra.spec.whatwg.org/#map-iterate" id="ref-for-map-iterate⑨">For each</a> <var>element</var> in <var>outputs</var>:</p>
        <ol>
         <li data-md>
-         <p>If that fails, throw an "<code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#operationerror" id="ref-for-operationerror⑨">OperationError</a></code>" <code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#idl-DOMException" id="ref-for-idl-DOMException③①">DOMException</a></code> and stop.</p>
+         <p><a data-link-type="dfn" href="https://infra.spec.whatwg.org/#assert" id="ref-for-assert①②">Assert</a>: the type of <var>element</var>.key is <a data-link-type="dfn" href="https://infra.spec.whatwg.org/#string" id="ref-for-string③">string</a>.</p>
+        <li data-md>
+         <p>If <var>element</var>.key is empty, then <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#dfn-throw" id="ref-for-dfn-throw②⑨">throw</a> a <code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#exceptiondef-typeerror" id="ref-for-exceptiondef-typeerror④">TypeError</a></code>.</p>
+        <li data-md>
+         <p><a data-link-type="dfn" href="https://infra.spec.whatwg.org/#assert" id="ref-for-assert①③">Assert</a>: the type of <var>element</var>.value is <code class="idl"><a data-link-type="idl" href="#mloperand" id="ref-for-mloperand③⓪">MLOperand</a></code>.</p>
        </ol>
       <li data-md>
-       <p>Return <var>operand</var>.</p>
-     </ol>
-    </div>
-    <h5 class="heading settled" data-level="7.6.3.2" id="api-mlgraphbuilder-constant-value-type"><span class="secno">7.6.3.2. </span><span class="content">The <code class="idl"><a data-link-type="idl" href="#dom-mlgraphbuilder-constant-value-type" id="ref-for-dom-mlgraphbuilder-constant-value-type">constant(value, type)</a></code> method</span><a class="self-link" href="#api-mlgraphbuilder-constant-value-type"></a></h5>
-    <div class="note" role="note">
-      <strong>Arguments:</strong> 
-     <ul>
-      <li data-md>
-       <p><em>value</em>: a number</p>
+       <p>If any of the following sub-steps fail, <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#dfn-throw" id="ref-for-dfn-throw③⓪">throw</a> an "<code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#operationerror" id="ref-for-operationerror①①">OperationError</a></code>" <code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#idl-DOMException" id="ref-for-idl-DOMException②⑨">DOMException</a></code>.</p>
+       <ol>
+        <li data-md>
+         <p>Let <var>graph</var> be a new <code class="idl"><a data-link-type="idl" href="#mlgraph" id="ref-for-mlgraph②⑦">MLGraph</a></code>:</p>
+         <ol>
+          <li data-md>
+           <p>Set <var>graph</var>.<code class="idl"><a data-link-type="idl" href="#dom-mlgraph-context-slot" id="ref-for-dom-mlgraph-context-slot②">[[context]]</a></code> to <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#this" id="ref-for-this⑥">this</a>.<code class="idl"><a data-link-type="idl" href="#dom-mlgraphbuilder-context-slot" id="ref-for-dom-mlgraphbuilder-context-slot①">[[context]]</a></code>.</p>
+         </ol>
+        <li data-md>
+         <p>Make a request to the underlying platform to:</p>
+         <ol>
+          <li data-md>
+           <p>Connect <var>graph</var> to a new <a data-link-type="dfn" href="https://infra.spec.whatwg.org/#implementation-defined" id="ref-for-implementation-defined⑥">implementation-defined</a> graph implementation <var>graphImpl</var> given <var>graph</var>.</p>
+          <li data-md>
+           <p>Store a reference to <var>graphImpl</var> in <var>graph</var>.<code class="idl"><a data-link-type="idl" href="#dom-mlgraph-implementation-slot" id="ref-for-dom-mlgraph-implementation-slot⑦">[[implementation]]</a></code>.</p>
+         </ol>
+        <li data-md>
+         <p>Make a request to the underlying platform to initialize the graph:</p>
+         <ol>
+          <li data-md>
+           <p><a data-link-type="dfn" href="https://infra.spec.whatwg.org/#map-iterate" id="ref-for-map-iterate①⓪">For each</a> <var>operand</var> in <var>outputs</var>:</p>
+           <ol>
+            <li data-md>
+             <p>If <a data-link-type="dfn" href="#mloperand-validate-mloperand" id="ref-for-mloperand-validate-mloperand">validating MLOperand</a> given <var>operand</var> and <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#this" id="ref-for-this⑦">this</a> returns false, then <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#dfn-throw" id="ref-for-dfn-throw③①">throw</a> a <code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#exceptiondef-typeerror" id="ref-for-exceptiondef-typeerror⑤">TypeError</a></code>.</p>
+            <li data-md>
+             <p>If <var>operand</var> was created as an input by the underlying platform:</p>
+             <ol>
+              <li data-md>
+               <p>If <var>operand</var>.<code class="idl"><a data-link-type="idl" href="#dom-mloperand-name-slot" id="ref-for-dom-mloperand-name-slot④">[[name]]</a></code>] is not unique for <var>graphImpl</var>, then <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#dfn-throw" id="ref-for-dfn-throw③②">throw</a> a <code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#exceptiondef-typeerror" id="ref-for-exceptiondef-typeerror⑥">TypeError</a></code>.</p>
+              <li data-md>
+               <p>Add <var>operand</var>.<code class="idl"><a data-link-type="idl" href="#dom-mloperand-descriptor-slot" id="ref-for-dom-mloperand-descriptor-slot⑤">[[descriptor]]</a></code> to <var>graph</var>.<code class="idl"><a data-link-type="idl" href="#dom-mlgraph-inputdescriptors-slot" id="ref-for-dom-mlgraph-inputdescriptors-slot⑤">[[inputDescriptors]]</a></code>[<var>operand</var>.<code class="idl"><a data-link-type="idl" href="#dom-mloperand-name-slot" id="ref-for-dom-mloperand-name-slot⑤">[[name]]</a></code>].</p>
+             </ol>
+            <li data-md>
+             <p>If <var>operand</var> was created as a constant by the underlying platform:</p>
+             <ol>
+              <li data-md>
+               <p>Implementations MAY preprocess and optimize the tensor data of <var>operand</var> for the underlying platform.</p>
+             </ol>
+            <li data-md>
+             <p>Register <var>operand</var>.<code class="idl"><a data-link-type="idl" href="#dom-mloperand-operand-slot" id="ref-for-dom-mloperand-operand-slot①">[[operand]]</a></code> in <var>graphImpl</var> as graph output.</p>
+            <li data-md>
+             <p>Register <var>operand</var>.<code class="idl"><a data-link-type="idl" href="#dom-mloperand-operator-slot" id="ref-for-dom-mloperand-operator-slot">[[operator]]</a></code> to <var>graphImpl</var>.</p>
+           </ol>
+         </ol>
+       </ol>
       <li data-md>
-       <p><em>type</em>: an optional <code class="idl"><a data-link-type="idl" href="#enumdef-mloperandtype" id="ref-for-enumdef-mloperandtype③">MLOperandType</a></code>, by default <em>"float32"</em>.</p>
-     </ul>
-      <strong>Returns:</strong>: an <code class="idl"><a data-link-type="idl" href="#mloperand" id="ref-for-mloperand③⓪">MLOperand</a></code> object. 
+       <p>Return <var>graph</var>.</p>
+     </ol>
     </div>
-    <div class="algorithm-steps algorithm" data-algorithm="constant-value-type">
-      The <code class="idl"><a data-link-type="idl" href="#dom-mlgraphbuilder-constant-value-type" id="ref-for-dom-mlgraphbuilder-constant-value-type①">constant(value, type)</a></code> steps are: 
+   </details>
+   <h4 class="heading settled" data-level="7.6.4" id="api-mlgraphbuilder-constant-method"><span class="secno">7.6.4. </span><span class="content">The constant() method</span><a class="self-link" href="#api-mlgraphbuilder-constant-method"></a></h4>
+    Create a constant <code class="idl"><a data-link-type="idl" href="#mloperand" id="ref-for-mloperand③①">MLOperand</a></code> that can be used in <code class="idl"><a data-link-type="idl" href="#mlgraphbuilder" id="ref-for-mlgraphbuilder②②">MLGraphBuilder</a></code> methods. 
+   <h5 class="heading settled" data-level="7.6.4.1" id="api-mlgraphbuilder-constant"><span class="secno">7.6.4.1. </span><span class="content">The <code class="idl"><a data-link-type="idl" href="#dom-mlgraphbuilder-constant" id="ref-for-dom-mlgraphbuilder-constant②">constant(descriptor, bufferView)</a></code> method</span><a class="self-link" href="#api-mlgraphbuilder-constant"></a></h5>
+   <div>
+     <strong>Arguments:</strong> 
+    <ul>
+     <li data-md>
+      <p><em>descriptor</em>: an <code class="idl"><a data-link-type="idl" href="#dictdef-mloperanddescriptor" id="ref-for-dictdef-mloperanddescriptor①①">MLOperandDescriptor</a></code> object</p>
+     <li data-md>
+      <p><em>bufferView</em>: an <code class="idl"><a data-link-type="idl" href="#typedefdef-mlbufferview" id="ref-for-typedefdef-mlbufferview②">MLBufferView</a></code></p>
+    </ul>
+     <strong>Returns:</strong>: an <code class="idl"><a data-link-type="idl" href="#mloperand" id="ref-for-mloperand③②">MLOperand</a></code> object. 
+   </div>
+   <details class="algorithm" data-algorithm="constant(descriptor, bufferView)" data-algorithm-for="MLGraphBuilder" open>
+    <summary> The <dfn class="dfn-paneled idl-code" data-dfn-for="MLGraphBuilder" data-dfn-type="method" data-export id="dom-mlgraphbuilder-constant"><code>constant(<var>descriptor</var>, <var>bufferView</var>)</code></dfn> method steps are: </summary>
+    <div class="algorithm-steps">
      <div class="note" role="note"> The permissions and context validity have been checked by <a href="#api-mlgraphbuilder-constructor">§ 7.6.1 The MLGraphBuilder constructor</a> steps. </div>
      <ol>
       <li data-md>
-       <p>If <var>value</var> is not a <a data-link-type="dfn" href="https://drafts.csswg.org/css-values-4/#number" id="ref-for-number">number</a>, then then throw a "<code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#exceptiondef-typeerror" id="ref-for-exceptiondef-typeerror⑦">TypeError</a></code>" <code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#idl-DOMException" id="ref-for-idl-DOMException③②">DOMException</a></code> and stop.</p>
+       <p><a data-link-type="dfn" href="https://infra.spec.whatwg.org/#assert" id="ref-for-assert①④">Assert</a>: the type of <var>descriptor</var> is <code class="idl"><a data-link-type="idl" href="#dictdef-mloperanddescriptor" id="ref-for-dictdef-mloperanddescriptor①②">MLOperandDescriptor</a></code>.</p>
       <li data-md>
-       <p>If <var>type</var> is <code>undefined</code>, let <var>type</var> be <code>"float32"</code>.</p>
+       <p>If the <a data-link-type="dfn" href="#mloperanddescriptor-byte-length" id="ref-for-mloperanddescriptor-byte-length⑤">byte length</a> of <var>descriptor</var> is not supported by the underlying platform, then <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#dfn-throw" id="ref-for-dfn-throw③③">throw</a> a "<code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#dataerror" id="ref-for-dataerror⑨">DataError</a></code>" <code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#idl-DOMException" id="ref-for-idl-DOMException③⓪">DOMException</a></code>.</p>
       <li data-md>
-       <p>Otherwise, if <var>type</var> is not one of <code class="idl"><a data-link-type="idl" href="#enumdef-mloperandtype" id="ref-for-enumdef-mloperandtype④">MLOperandType</a></code>, then  throw a "<code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#exceptiondef-typeerror" id="ref-for-exceptiondef-typeerror⑧">TypeError</a></code>" <code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#idl-DOMException" id="ref-for-idl-DOMException③③">DOMException</a></code> and stop.</p>
+       <p>If the <a data-link-type="dfn" href="#check-dimensions" id="ref-for-check-dimensions②">check dimensions</a> steps given <var>descriptor</var>.<code class="idl"><a data-link-type="idl" href="#dom-mloperanddescriptor-type" id="ref-for-dom-mloperanddescriptor-type④">type</a></code> and <var>descriptor</var>.<code class="idl"><a data-link-type="idl" href="#dom-mloperanddescriptor-dimensions" id="ref-for-dom-mloperanddescriptor-dimensions⑧">dimensions</a></code> return false, then <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#dfn-throw" id="ref-for-dfn-throw③④">throw</a> a "<code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#dataerror" id="ref-for-dataerror①⓪">DataError</a></code>" <code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#idl-DOMException" id="ref-for-idl-DOMException③①">DOMException</a></code>.</p>
       <li data-md>
-       <p>Let <var>descriptor</var> be a new <code class="idl"><a data-link-type="idl" href="#dictdef-mloperanddescriptor" id="ref-for-dictdef-mloperanddescriptor①③">MLOperandDescriptor</a></code>.</p>
+       <p>If <a data-link-type="dfn" href="#validate-buffer-with-descriptor" id="ref-for-validate-buffer-with-descriptor①">validating buffer with descriptor</a> given <var>bufferView</var> and <var>descriptor</var> returns false, then <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#dfn-throw" id="ref-for-dfn-throw③⑤">throw</a> a <code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#exceptiondef-typeerror" id="ref-for-exceptiondef-typeerror⑦">TypeError</a></code>.</p>
+      <li data-md>
+       <p>If any of the following sub-steps fail, <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#dfn-throw" id="ref-for-dfn-throw③⑥">throw</a> an "<code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#operationerror" id="ref-for-operationerror①②">OperationError</a></code>" <code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#idl-DOMException" id="ref-for-idl-DOMException③②">DOMException</a></code>.</p>
        <ol>
         <li data-md>
-         <p>Set <var>descriptor</var>.<code class="idl"><a data-link-type="idl" href="#dom-mloperanddescriptor-type" id="ref-for-dom-mloperanddescriptor-type④">type</a></code> to <var>type</var>.</p>
+         <p>Let <var>operand</var> be the result of <a data-link-type="dfn" href="#create-an-mloperand" id="ref-for-create-an-mloperand①">creating an MLOperand</a> given <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#this" id="ref-for-this⑧">this</a> and <var>descriptor</var>.</p>
+        <li data-md>
+         <p>Let <var>bytes</var> be the result of invoking the <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#dfn-get-buffer-source-copy" id="ref-for-dfn-get-buffer-source-copy">get a copy of the bytes held by the buffer source</a> steps given <var>bufferView</var>.</p>
         <li data-md>
-         <p>Set <var>descriptor</var>.<code class="idl"><a data-link-type="idl" href="#dom-mloperanddescriptor-dimensions" id="ref-for-dom-mloperanddescriptor-dimensions⑥">dimensions</a></code> to <code>undefined</code>.</p>
-         <div class="note" role="note"> In the case of a scalar constant, <var>descriptor</var>.<code class="idl"><a data-link-type="idl" href="#dom-mloperanddescriptor-dimensions" id="ref-for-dom-mloperanddescriptor-dimensions⑦">dimensions</a></code> is ignored. </div>
+         <p>Make a request to the underlying platform to:</p>
+         <ol>
+          <li data-md>
+           <p>Create an <a data-link-type="dfn" href="https://infra.spec.whatwg.org/#implementation-defined" id="ref-for-implementation-defined⑦">implementation-defined</a> platform operand <var>constantImpl</var> to represent a constant, given <var>descriptor</var>.</p>
+          <li data-md>
+           <p>Store a reference of <var>constantImpl</var> in <var>operand</var>.<code class="idl"><a data-link-type="idl" href="#dom-mloperand-operand-slot" id="ref-for-dom-mloperand-operand-slot②">[[operand]]</a></code>.</p>
+          <li data-md>
+           <p>Register <var>operand</var> as a tensor constant with <var>bytes</var> as value.</p>
+         </ol>
        </ol>
       <li data-md>
-       <p>Let <var>operand</var> be the result of invoking the <a data-link-type="dfn" href="#create-mloperand" id="ref-for-create-mloperand②">create MLOperand</a> steps with <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#this" id="ref-for-this⑤">this</a> and <var>descriptor</var>.</p>
+       <p>Return <var>operand</var>.</p>
+     </ol>
+    </div>
+   </details>
+   <h5 class="heading settled" data-level="7.6.4.2" id="api-mlgraphbuilder-constant-value-type"><span class="secno">7.6.4.2. </span><span class="content">The <code class="idl"><a data-link-type="idl" href="#dom-mlgraphbuilder-constant-value-type" id="ref-for-dom-mlgraphbuilder-constant-value-type①">constant(value, type)</a></code> method</span><a class="self-link" href="#api-mlgraphbuilder-constant-value-type"></a></h5>
+   <div>
+     <strong>Arguments:</strong> 
+    <ul>
+     <li data-md>
+      <p><em>value</em>: a number</p>
+     <li data-md>
+      <p><em>type</em>: an optional <code class="idl"><a data-link-type="idl" href="#enumdef-mloperandtype" id="ref-for-enumdef-mloperandtype③">MLOperandType</a></code>, by default <em>"float32"</em>.</p>
+    </ul>
+     <strong>Returns:</strong>: an <code class="idl"><a data-link-type="idl" href="#mloperand" id="ref-for-mloperand③③">MLOperand</a></code> object. 
+   </div>
+   <details class="algorithm" data-algorithm="constant(value, type)" data-algorithm-for="MLGraphBuilder" open>
+    <summary> The <dfn class="dfn-paneled idl-code" data-dfn-for="MLGraphBuilder" data-dfn-type="method" data-export data-lt="constant(value, type)|constant(value)" id="dom-mlgraphbuilder-constant-value-type"><code>constant(<var>value</var>, <var>type</var>)</code></dfn> method steps are: </summary>
+    <div class="algorithm-steps">
+     <div class="note" role="note"> The permissions and context validity have been checked by <a href="#api-mlgraphbuilder-constructor">§ 7.6.1 The MLGraphBuilder constructor</a> steps. </div>
+     <ol>
+      <li data-md>
+       <p>Let <var>descriptor</var> be a new <code class="idl"><a data-link-type="idl" href="#dictdef-mloperanddescriptor" id="ref-for-dictdef-mloperanddescriptor①③">MLOperandDescriptor</a></code>.</p>
        <ol>
         <li data-md>
-         <p>If that throws, re-throw the exception and stop.</p>
+         <p>Set <var>descriptor</var>.<code class="idl"><a data-link-type="idl" href="#dom-mloperanddescriptor-type" id="ref-for-dom-mloperanddescriptor-type⑤">type</a></code> to <var>type</var>.</p>
+        <li data-md>
+         <p>Set <var>descriptor</var>.<code class="idl"><a data-link-type="idl" href="#dom-mloperanddescriptor-dimensions" id="ref-for-dom-mloperanddescriptor-dimensions⑨">dimensions</a></code> to <code>undefined</code>.</p>
+         <div class="note" role="note"> In the case of a scalar constant, <var>descriptor</var>.<code class="idl"><a data-link-type="idl" href="#dom-mloperanddescriptor-dimensions" id="ref-for-dom-mloperanddescriptor-dimensions①⓪">dimensions</a></code> is ignored. </div>
        </ol>
       <li data-md>
-       <p>Make a request to the underlying platform to register <var>operand</var> as a scalar constant with <var>value</var> as value and store a reference of the <a data-link-type="dfn" href="https://infra.spec.whatwg.org/#implementation-defined" id="ref-for-implementation-defined④">implementation-defined</a> platform object for the corresponding (scalar or tensor constant) operand to <var>operand</var>.<code class="idl"><a data-link-type="idl" href="#dom-mloperand-operand-slot" id="ref-for-dom-mloperand-operand-slot②">[[operand]]</a></code>.</p>
+       <p>If any of the following sub-steps fail, <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#dfn-throw" id="ref-for-dfn-throw③⑦">throw</a> an "<code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#operationerror" id="ref-for-operationerror①③">OperationError</a></code>" <code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#idl-DOMException" id="ref-for-idl-DOMException③③">DOMException</a></code>.</p>
        <ol>
         <li data-md>
-         <p>If that throws, re-throw the error and stop.</p>
+         <p>Let <var>operand</var> be the result of <a data-link-type="dfn" href="#create-an-mloperand" id="ref-for-create-an-mloperand②">creating an MLOperand</a> given <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#this" id="ref-for-this⑨">this</a> and <var>descriptor</var>.</p>
+        <li data-md>
+         <p>Make a request to the underlying platform to:</p>
+         <ol>
+          <li data-md>
+           <p>Create an <a data-link-type="dfn" href="https://infra.spec.whatwg.org/#implementation-defined" id="ref-for-implementation-defined⑧">implementation-defined</a> platform operand <var>constantImpl</var> to represent a constant, given <var>descriptor</var>.</p>
+          <li data-md>
+           <p>Store a reference of <var>constantImpl</var> in <var>operand</var>.<code class="idl"><a data-link-type="idl" href="#dom-mloperand-operand-slot" id="ref-for-dom-mloperand-operand-slot③">[[operand]]</a></code>.</p>
+          <li data-md>
+           <p>Register <var>operand</var> as a scalar constant with <var>value</var> as value.</p>
+         </ol>
        </ol>
       <li data-md>
        <p>Return <var>operand</var>.</p>
      </ol>
     </div>
-    <h4 class="heading settled" data-level="7.6.4" id="api-mlgraphbuilder-batchnorm"><span class="secno">7.6.4. </span><span class="content">The batchNormalization() method</span><a class="self-link" href="#api-mlgraphbuilder-batchnorm"></a></h4>
-     Normalize the tensor values of input features across the batch dimension using <a data-link-type="biblio" href="#biblio-batch-normalization" title="Batch Normalization: Accelerating Deep Network Training by Reducing Internal Covariate Shift">[Batch-Normalization]</a>. For each input feature, the mean and variance values of that feature supplied in this calculation as parameters are previously computed across the batch dimension of the input during the model training phase of this operation. 
+   </details>
+   <h4 class="heading settled" data-level="7.6.5" id="api-mlgraphbuilder-batchnorm"><span class="secno">7.6.5. </span><span class="content">The batchNormalization() method</span><a class="self-link" href="#api-mlgraphbuilder-batchnorm"></a></h4>
+    Normalize the tensor values of input features across the batch dimension using <a data-link-type="biblio" href="#biblio-batch-normalization" title="Batch Normalization: Accelerating Deep Network Training by Reducing Internal Covariate Shift">[Batch-Normalization]</a>. For each input feature, the mean and variance values of that feature supplied in this calculation as parameters are previously computed across the batch dimension of the input during the model training phase of this operation. 
 <pre class="idl highlight def"><c- b>dictionary</c-> <dfn class="dfn-paneled idl-code" data-dfn-type="dictionary" data-export id="dictdef-mlbatchnormalizationoptions"><code><c- g>MLBatchNormalizationOptions</c-></code></dfn> {
-  <a data-link-type="idl-name" href="#mloperand" id="ref-for-mloperand③①"><c- n>MLOperand</c-></a> <dfn class="idl-code" data-dfn-for="MLBatchNormalizationOptions" data-dfn-type="dict-member" data-export data-type="MLOperand " id="dom-mlbatchnormalizationoptions-scale"><code><c- g>scale</c-></code><a class="self-link" href="#dom-mlbatchnormalizationoptions-scale"></a></dfn>;
-  <a data-link-type="idl-name" href="#mloperand" id="ref-for-mloperand③②"><c- n>MLOperand</c-></a> <dfn class="idl-code" data-dfn-for="MLBatchNormalizationOptions" data-dfn-type="dict-member" data-export data-type="MLOperand " id="dom-mlbatchnormalizationoptions-bias"><code><c- g>bias</c-></code><a class="self-link" href="#dom-mlbatchnormalizationoptions-bias"></a></dfn>;
-  <a class="idl-code" data-link-type="interface" href="https://webidl.spec.whatwg.org/#idl-unsigned-long" id="ref-for-idl-unsigned-long①"><c- b>unsigned</c-> <c- b>long</c-></a> <dfn class="idl-code" data-default="1" data-dfn-for="MLBatchNormalizationOptions" data-dfn-type="dict-member" data-export data-type="unsigned long " id="dom-mlbatchnormalizationoptions-axis"><code><c- g>axis</c-></code><a class="self-link" href="#dom-mlbatchnormalizationoptions-axis"></a></dfn> = 1;
-  <a class="idl-code" data-link-type="interface" href="https://webidl.spec.whatwg.org/#idl-float" id="ref-for-idl-float"><c- b>float</c-></a> <dfn class="idl-code" data-default="1e-5" data-dfn-for="MLBatchNormalizationOptions" data-dfn-type="dict-member" data-export data-type="float " id="dom-mlbatchnormalizationoptions-epsilon"><code><c- g>epsilon</c-></code><a class="self-link" href="#dom-mlbatchnormalizationoptions-epsilon"></a></dfn> = 1e-5;
-  <a data-link-type="idl-name" href="#mlactivation" id="ref-for-mlactivation②"><c- n>MLActivation</c-></a> <dfn class="idl-code" data-dfn-for="MLBatchNormalizationOptions" data-dfn-type="dict-member" data-export data-type="MLActivation " id="dom-mlbatchnormalizationoptions-activation"><code><c- g>activation</c-></code><a class="self-link" href="#dom-mlbatchnormalizationoptions-activation"></a></dfn>;
+  <a data-link-type="idl-name" href="#mloperand" id="ref-for-mloperand③④"><c- n>MLOperand</c-></a> <a class="idl-code" data-link-type="dict-member" data-type="MLOperand " href="#dom-mlbatchnormalizationoptions-scale" id="ref-for-dom-mlbatchnormalizationoptions-scale"><c- g>scale</c-></a>;
+  <a data-link-type="idl-name" href="#mloperand" id="ref-for-mloperand③⑤"><c- n>MLOperand</c-></a> <a class="idl-code" data-link-type="dict-member" data-type="MLOperand " href="#dom-mlbatchnormalizationoptions-bias" id="ref-for-dom-mlbatchnormalizationoptions-bias"><c- g>bias</c-></a>;
+  <a class="idl-code" data-link-type="interface" href="https://webidl.spec.whatwg.org/#idl-unsigned-long" id="ref-for-idl-unsigned-long①"><c- b>unsigned</c-> <c- b>long</c-></a> <a class="idl-code" data-default="1" data-link-type="dict-member" data-type="unsigned long " href="#dom-mlbatchnormalizationoptions-axis" id="ref-for-dom-mlbatchnormalizationoptions-axis"><c- g>axis</c-></a> = 1;
+  <a class="idl-code" data-link-type="interface" href="https://webidl.spec.whatwg.org/#idl-float" id="ref-for-idl-float"><c- b>float</c-></a> <a class="idl-code" data-default="1e-5" data-link-type="dict-member" data-type="float " href="#dom-mlbatchnormalizationoptions-epsilon" id="ref-for-dom-mlbatchnormalizationoptions-epsilon"><c- g>epsilon</c-></a> = 1e-5;
+  <a data-link-type="idl-name" href="#mlactivation" id="ref-for-mlactivation⑧"><c- n>MLActivation</c-></a> <a class="idl-code" data-link-type="dict-member" data-type="MLActivation " href="#dom-mlbatchnormalizationoptions-activation" id="ref-for-dom-mlbatchnormalizationoptions-activation"><c- g>activation</c-></a>;
 };
 
-<c- b>partial</c-> <c- b>interface</c-> <a class="idl-code" data-link-type="interface" href="#mlgraphbuilder" id="ref-for-mlgraphbuilder②⓪"><c- g>MLGraphBuilder</c-></a> {
-  <a data-link-type="idl-name" href="#mloperand" id="ref-for-mloperand③③"><c- n>MLOperand</c-></a> <dfn class="idl-code" data-dfn-for="MLGraphBuilder" data-dfn-type="method" data-export data-lt="batchNormalization(input, mean, variance, options)|batchNormalization(input, mean, variance)" id="dom-mlgraphbuilder-batchnormalization"><code><c- g>batchNormalization</c-></code><a class="self-link" href="#dom-mlgraphbuilder-batchnormalization"></a></dfn>(<a data-link-type="idl-name" href="#mloperand" id="ref-for-mloperand③④"><c- n>MLOperand</c-></a> <dfn class="idl-code" data-dfn-for="MLGraphBuilder/batchNormalization(input, mean, variance, options), MLGraphBuilder/batchNormalization(input, mean, variance)" data-dfn-type="argument" data-export id="dom-mlgraphbuilder-batchnormalization-input-mean-variance-options-input"><code><c- g>input</c-></code><a class="self-link" href="#dom-mlgraphbuilder-batchnormalization-input-mean-variance-options-input"></a></dfn>, <a data-link-type="idl-name" href="#mloperand" id="ref-for-mloperand③⑤"><c- n>MLOperand</c-></a> <dfn class="idl-code" data-dfn-for="MLGraphBuilder/batchNormalization(input, mean, variance, options), MLGraphBuilder/batchNormalization(input, mean, variance)" data-dfn-type="argument" data-export id="dom-mlgraphbuilder-batchnormalization-input-mean-variance-options-mean"><code><c- g>mean</c-></code><a class="self-link" href="#dom-mlgraphbuilder-batchnormalization-input-mean-variance-options-mean"></a></dfn>, <a data-link-type="idl-name" href="#mloperand" id="ref-for-mloperand③⑥"><c- n>MLOperand</c-></a> <dfn class="idl-code" data-dfn-for="MLGraphBuilder/batchNormalization(input, mean, variance, options), MLGraphBuilder/batchNormalization(input, mean, variance)" data-dfn-type="argument" data-export id="dom-mlgraphbuilder-batchnormalization-input-mean-variance-options-variance"><code><c- g>variance</c-></code><a class="self-link" href="#dom-mlgraphbuilder-batchnormalization-input-mean-variance-options-variance"></a></dfn>,
+<c- b>partial</c-> <c- b>interface</c-> <a class="idl-code" data-link-type="interface" href="#mlgraphbuilder" id="ref-for-mlgraphbuilder②③"><c- g>MLGraphBuilder</c-></a> {
+  <a data-link-type="idl-name" href="#mloperand" id="ref-for-mloperand③⑥"><c- n>MLOperand</c-></a> <a class="idl-code" data-link-type="method" href="#dom-mlgraphbuilder-batchnormalization" id="ref-for-dom-mlgraphbuilder-batchnormalization"><c- g>batchNormalization</c-></a>(<a data-link-type="idl-name" href="#mloperand" id="ref-for-mloperand③⑦"><c- n>MLOperand</c-></a> <dfn class="idl-code" data-dfn-for="MLGraphBuilder/batchNormalization(input, mean, variance, options), MLGraphBuilder/batchNormalization(input, mean, variance)" data-dfn-type="argument" data-export id="dom-mlgraphbuilder-batchnormalization-input-mean-variance-options-input"><code><c- g>input</c-></code><a class="self-link" href="#dom-mlgraphbuilder-batchnormalization-input-mean-variance-options-input"></a></dfn>, <a data-link-type="idl-name" href="#mloperand" id="ref-for-mloperand③⑧"><c- n>MLOperand</c-></a> <dfn class="idl-code" data-dfn-for="MLGraphBuilder/batchNormalization(input, mean, variance, options), MLGraphBuilder/batchNormalization(input, mean, variance)" data-dfn-type="argument" data-export id="dom-mlgraphbuilder-batchnormalization-input-mean-variance-options-mean"><code><c- g>mean</c-></code><a class="self-link" href="#dom-mlgraphbuilder-batchnormalization-input-mean-variance-options-mean"></a></dfn>, <a data-link-type="idl-name" href="#mloperand" id="ref-for-mloperand③⑨"><c- n>MLOperand</c-></a> <dfn class="idl-code" data-dfn-for="MLGraphBuilder/batchNormalization(input, mean, variance, options), MLGraphBuilder/batchNormalization(input, mean, variance)" data-dfn-type="argument" data-export id="dom-mlgraphbuilder-batchnormalization-input-mean-variance-options-variance"><code><c- g>variance</c-></code><a class="self-link" href="#dom-mlgraphbuilder-batchnormalization-input-mean-variance-options-variance"></a></dfn>,
                              <c- b>optional</c-> <a data-link-type="idl-name" href="#dictdef-mlbatchnormalizationoptions" id="ref-for-dictdef-mlbatchnormalizationoptions"><c- n>MLBatchNormalizationOptions</c-></a> <dfn class="idl-code" data-dfn-for="MLGraphBuilder/batchNormalization(input, mean, variance, options), MLGraphBuilder/batchNormalization(input, mean, variance)" data-dfn-type="argument" data-export id="dom-mlgraphbuilder-batchnormalization-input-mean-variance-options-options"><code><c- g>options</c-></code><a class="self-link" href="#dom-mlgraphbuilder-batchnormalization-input-mean-variance-options-options"></a></dfn> = {});
 };
 </pre>
-    <div class="algorithm" data-algorithm="batchnorm">
-      <strong>Arguments:</strong> 
-     <ul>
+   <p><code class="idl"><a data-link-type="idl" href="#dictdef-mlbatchnormalizationoptions" id="ref-for-dictdef-mlbatchnormalizationoptions①">MLBatchNormalizationOptions</a></code> has the following members:</p>
+   <dl>
+    <dt data-md><dfn class="dfn-paneled idl-code" data-dfn-for="MLBatchNormalizationOptions" data-dfn-type="dict-member" data-export id="dom-mlbatchnormalizationoptions-scale"><code>scale</code></dfn>, <span> of type <a data-link-type="idl-name" href="#mloperand" id="ref-for-mloperand④⓪">MLOperand</a></span>
+    <dd data-md>
+     <p>An <code class="idl"><a data-link-type="idl" href="#mloperand" id="ref-for-mloperand④①">MLOperand</a></code>. Specifies the 1-D tensor of the scaling values whose is equal to the size of the input dimension denoted by <code class="idl"><a data-link-type="idl" href="#dom-mlbatchnormalizationoptions-axis" id="ref-for-dom-mlbatchnormalizationoptions-axis①">axis</a></code>.</p>
+    <dt data-md><dfn class="dfn-paneled idl-code" data-dfn-for="MLBatchNormalizationOptions" data-dfn-type="dict-member" data-export id="dom-mlbatchnormalizationoptions-bias"><code>bias</code></dfn>, <span> of type <a data-link-type="idl-name" href="#mloperand" id="ref-for-mloperand④②">MLOperand</a></span>
+    <dd data-md>
+     <p>An <code class="idl"><a data-link-type="idl" href="#mloperand" id="ref-for-mloperand④③">MLOperand</a></code>. Specifies the 1-D tensor of the bias values whose <a data-link-type="dfn" href="https://infra.spec.whatwg.org/#list-size" id="ref-for-list-size③">size</a> is equal to the size of the input dimension denoted by <code class="idl"><a data-link-type="idl" href="#dom-mlbatchnormalizationoptions-axis" id="ref-for-dom-mlbatchnormalizationoptions-axis②">axis</a></code>.</p>
+    <dt data-md><dfn class="dfn-paneled idl-code" data-dfn-for="MLBatchNormalizationOptions" data-dfn-type="dict-member" data-export id="dom-mlbatchnormalizationoptions-axis"><code>axis</code></dfn>, <span> of type <a data-link-type="idl-name" href="https://webidl.spec.whatwg.org/#idl-unsigned-long" id="ref-for-idl-unsigned-long②">unsigned long</a>, defaulting to <code>1</code></span>
+    <dd data-md>
+     <p>A <code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#idl-long" id="ref-for-idl-long">long</a></code> scalar. Specifies the index to the feature count dimension of the input shape for which the mean and variance values are. Its value must be in the range [0, N-1] where N is the <a data-link-type="dfn" href="#mloperand-rank" id="ref-for-mloperand-rank">rank</a> of the input tensor. The default value is 1, corresponding to the channel (<em>"c"</em>) dimension in the <em>"nchw"</em> data layout.</p>
+    <dt data-md><dfn class="dfn-paneled idl-code" data-dfn-for="MLBatchNormalizationOptions" data-dfn-type="dict-member" data-export id="dom-mlbatchnormalizationoptions-epsilon"><code>epsilon</code></dfn>, <span> of type <a data-link-type="idl-name" href="https://webidl.spec.whatwg.org/#idl-float" id="ref-for-idl-float①">float</a>, defaulting to <code>1e-5</code></span>
+    <dd data-md>
+     <p>A <code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#idl-float" id="ref-for-idl-float②">float</a></code> scalar. Specifies A small value to prevent computational error due to divide-by-zero.</p>
+    <dt data-md><dfn class="dfn-paneled idl-code" data-dfn-for="MLBatchNormalizationOptions" data-dfn-type="dict-member" data-export id="dom-mlbatchnormalizationoptions-activation"><code>activation</code></dfn>, <span> of type <a data-link-type="idl-name" href="#mlactivation" id="ref-for-mlactivation⑨">MLActivation</a></span>
+    <dd data-md>
+     <p>An <code class="idl"><a data-link-type="idl" href="#mlactivation" id="ref-for-mlactivation①⓪">MLActivation</a></code> object. Specifies the optional activation function that immediately follows the normalization operation.</p>
+   </dl>
+   <div>
+     <strong>Arguments:</strong> 
+    <ul>
+     <li data-md>
+      <p><em>input</em>: an <code class="idl"><a data-link-type="idl" href="#mloperand" id="ref-for-mloperand④④">MLOperand</a></code>. The input N-D tensor.</p>
+     <li data-md>
+      <p><em>mean</em>: an <code class="idl"><a data-link-type="idl" href="#mloperand" id="ref-for-mloperand④⑤">MLOperand</a></code>. Specifies the 1-D tensor of the mean values of the input features across the batch. Its <a data-link-type="dfn" href="https://infra.spec.whatwg.org/#list-size" id="ref-for-list-size④">size</a> is equal to the size of the input dimension denoted by <code class="idl"><a data-link-type="idl" href="#dom-mlbatchnormalizationoptions-axis" id="ref-for-dom-mlbatchnormalizationoptions-axis③">axis</a></code>.</p>
+     <li data-md>
+      <p><em>variance</em>: an <code class="idl"><a data-link-type="idl" href="#mloperand" id="ref-for-mloperand④⑥">MLOperand</a></code>. The 1-D tensor of the variance values of the input features across the batch whose <a data-link-type="dfn" href="https://infra.spec.whatwg.org/#list-size" id="ref-for-list-size⑤">size</a> is equal to the size of the input dimension denoted by <code class="idl"><a data-link-type="idl" href="#dom-mlbatchnormalizationoptions-axis" id="ref-for-dom-mlbatchnormalizationoptions-axis④">axis</a></code>.</p>
+     <li data-md>
+      <p><em>options</em>: an optional <code class="idl"><a data-link-type="idl" href="#dictdef-mlbatchnormalizationoptions" id="ref-for-dictdef-mlbatchnormalizationoptions②">MLBatchNormalizationOptions</a></code>. Specifies the optional parameters of the operation.</p>
+    </ul>
+    <p><strong>Returns:</strong> an <code class="idl"><a data-link-type="idl" href="#mloperand" id="ref-for-mloperand④⑦">MLOperand</a></code>. The batch-normalized N-D tensor of the same shape as <em>input</em>.</p>
+   </div>
+   <details class="algorithm" data-algorithm="batchNormalization(input, mean, variance, options)" data-algorithm-for="MLGraphBuilder" open>
+    <summary> The <dfn class="dfn-paneled idl-code" data-dfn-for="MLGraphBuilder" data-dfn-type="method" data-export data-lt="batchNormalization(input, mean, variance, options)|batchNormalization(input, mean, variance)" id="dom-mlgraphbuilder-batchnormalization"><code>batchNormalization(<var>input</var>, <var>mean</var>, <var>variance</var>, <var>options</var>)</code></dfn> method steps are: </summary>
+    <div class="algorithm-steps">
+     <ol>
       <li data-md>
-       <p><em>input</em>: an <code class="idl"><a data-link-type="idl" href="#mloperand" id="ref-for-mloperand③⑦">MLOperand</a></code>. The input N-D tensor.</p>
+       <p><a data-link-type="dfn" href="https://infra.spec.whatwg.org/#assert" id="ref-for-assert①⑤">Assert</a>: the type of <var>input</var>, <var>mean</var> and <var>variance</var> is <code class="idl"><a data-link-type="idl" href="#mloperand" id="ref-for-mloperand④⑧">MLOperand</a></code>.</p>
       <li data-md>
-       <p><em>mean</em>: an <code class="idl"><a data-link-type="idl" href="#mloperand" id="ref-for-mloperand③⑧">MLOperand</a></code>. The 1-D tensor of the mean values of the input features across the batch whose length is equal to the size of the input dimension denoted by <em>options.axis</em>.</p>
+       <p>If <var>options</var>.axis is not in <a data-link-type="dfn" href="https://infra.spec.whatwg.org/#the-range" id="ref-for-the-range">the range</a> 0 to the <a data-link-type="dfn" href="#mloperand-rank" id="ref-for-mloperand-rank①">rank</a> of <var>input</var>, exclusive, then <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#dfn-throw" id="ref-for-dfn-throw③⑧">throw</a> a <code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#exceptiondef-typeerror" id="ref-for-exceptiondef-typeerror⑧">TypeError</a></code>.</p>
       <li data-md>
-       <p><em>variance</em>: an <code class="idl"><a data-link-type="idl" href="#mloperand" id="ref-for-mloperand③⑨">MLOperand</a></code>. The 1-D tensor of the variance values of the input features across the batch whose length is equal to the size of the input dimension denoted by <em>options.axis</em>.</p>
+       <p>If the <a data-link-type="dfn" href="https://infra.spec.whatwg.org/#list-size" id="ref-for-list-size⑥">size</a> of <var>mean</var>.<code class="idl"><a data-link-type="idl" href="#dom-mloperand-descriptor-slot" id="ref-for-dom-mloperand-descriptor-slot⑥">[[descriptor]]</a></code>.<code class="idl"><a data-link-type="idl" href="#dom-mloperanddescriptor-dimensions" id="ref-for-dom-mloperanddescriptor-dimensions①①">dimensions</a></code> is not 1, then <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#dfn-throw" id="ref-for-dfn-throw③⑨">throw</a> a <code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#exceptiondef-typeerror" id="ref-for-exceptiondef-typeerror⑨">TypeError</a></code>.</p>
       <li data-md>
-       <p><em>options</em>: an optional <code class="idl"><a data-link-type="idl" href="#dictdef-mlbatchnormalizationoptions" id="ref-for-dictdef-mlbatchnormalizationoptions①">MLBatchNormalizationOptions</a></code>. The optional parameters of the operation.</p>
-       <ul>
+       <p>If <var>mean</var>.<code class="idl"><a data-link-type="idl" href="#dom-mloperand-descriptor-slot" id="ref-for-dom-mloperand-descriptor-slot⑦">[[descriptor]]</a></code>.<code class="idl"><a data-link-type="idl" href="#dom-mloperanddescriptor-dimensions" id="ref-for-dom-mloperanddescriptor-dimensions①②">dimensions</a></code>[0] is not equal to <var>input</var>.<code class="idl"><a data-link-type="idl" href="#dom-mloperand-descriptor-slot" id="ref-for-dom-mloperand-descriptor-slot⑧">[[descriptor]]</a></code>.<code class="idl"><a data-link-type="idl" href="#dom-mloperanddescriptor-dimensions" id="ref-for-dom-mloperanddescriptor-dimensions①③">dimensions</a></code>[<var>options</var>.<code class="idl"><a data-link-type="idl" href="#dom-mlbatchnormalizationoptions-axis" id="ref-for-dom-mlbatchnormalizationoptions-axis⑤">axis</a></code>], then <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#dfn-throw" id="ref-for-dfn-throw④⓪">throw</a> a <code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#exceptiondef-typeerror" id="ref-for-exceptiondef-typeerror①⓪">TypeError</a></code>.</p>
+      <li data-md>
+       <p>If the <a data-link-type="dfn" href="https://infra.spec.whatwg.org/#list-size" id="ref-for-list-size⑦">size</a> of <var>variance</var>.<code class="idl"><a data-link-type="idl" href="#dom-mloperand-descriptor-slot" id="ref-for-dom-mloperand-descriptor-slot⑨">[[descriptor]]</a></code>.<code class="idl"><a data-link-type="idl" href="#dom-mloperanddescriptor-dimensions" id="ref-for-dom-mloperanddescriptor-dimensions①④">dimensions</a></code> is not 1, then <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#dfn-throw" id="ref-for-dfn-throw④①">throw</a> a <code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#exceptiondef-typeerror" id="ref-for-exceptiondef-typeerror①①">TypeError</a></code>.</p>
+      <li data-md>
+       <p>If <var>variance</var>.<code class="idl"><a data-link-type="idl" href="#dom-mloperand-descriptor-slot" id="ref-for-dom-mloperand-descriptor-slot①⓪">[[descriptor]]</a></code>.<code class="idl"><a data-link-type="idl" href="#dom-mloperanddescriptor-dimensions" id="ref-for-dom-mloperanddescriptor-dimensions①⑤">dimensions</a></code>[0] is not equal to <var>input</var>.<code class="idl"><a data-link-type="idl" href="#dom-mloperand-descriptor-slot" id="ref-for-dom-mloperand-descriptor-slot①①">[[descriptor]]</a></code>.<code class="idl"><a data-link-type="idl" href="#dom-mloperanddescriptor-dimensions" id="ref-for-dom-mloperanddescriptor-dimensions①⑥">dimensions</a></code>[<var>options</var>.<code class="idl"><a data-link-type="idl" href="#dom-mlbatchnormalizationoptions-axis" id="ref-for-dom-mlbatchnormalizationoptions-axis⑥">axis</a></code>], then <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#dfn-throw" id="ref-for-dfn-throw④②">throw</a> a <code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#exceptiondef-typeerror" id="ref-for-exceptiondef-typeerror①②">TypeError</a></code>.</p>
+      <li data-md>
+       <p>If <var>options</var>.<code class="idl"><a data-link-type="idl" href="#dom-mlbatchnormalizationoptions-scale" id="ref-for-dom-mlbatchnormalizationoptions-scale①">scale</a></code> <a data-link-type="dfn" href="https://infra.spec.whatwg.org/#map-exists" id="ref-for-map-exists⑨">exists</a>:</p>
+       <ol>
         <li data-md>
-         <p><em>scale</em>: an <code class="idl"><a data-link-type="idl" href="#mloperand" id="ref-for-mloperand④⓪">MLOperand</a></code>. The 1-D tensor of the scaling values whose length is equal to the size of the input dimension denoted by <em>options.axis</em>.</p>
+         <p>If its <a data-link-type="dfn" href="https://infra.spec.whatwg.org/#list-size" id="ref-for-list-size⑧">size</a> is not 1, then <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#dfn-throw" id="ref-for-dfn-throw④③">throw</a> a <code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#exceptiondef-typeerror" id="ref-for-exceptiondef-typeerror①③">TypeError</a></code>.</p>
         <li data-md>
-         <p><em>bias</em>: an <code class="idl"><a data-link-type="idl" href="#mloperand" id="ref-for-mloperand④①">MLOperand</a></code>. The 1-D tensor of the bias values whose length is equal to the size of the input dimension denoted by <em>options.axis</em>.</p>
+         <p>If <var>options</var>.<code class="idl"><a data-link-type="idl" href="#dom-mlbatchnormalizationoptions-scale" id="ref-for-dom-mlbatchnormalizationoptions-scale②">scale</a></code>.<code class="idl"><a data-link-type="idl" href="#dom-mloperand-descriptor-slot" id="ref-for-dom-mloperand-descriptor-slot①②">[[descriptor]]</a></code>.<code class="idl"><a data-link-type="idl" href="#dom-mloperanddescriptor-dimensions" id="ref-for-dom-mloperanddescriptor-dimensions①⑦">dimensions</a></code>[0] is not equal to <var>input</var>.<code class="idl"><a data-link-type="idl" href="#dom-mloperand-descriptor-slot" id="ref-for-dom-mloperand-descriptor-slot①③">[[descriptor]]</a></code>.<code class="idl"><a data-link-type="idl" href="#dom-mloperanddescriptor-dimensions" id="ref-for-dom-mloperanddescriptor-dimensions①⑧">dimensions</a></code>[<var>options</var>.<code class="idl"><a data-link-type="idl" href="#dom-mlbatchnormalizationoptions-axis" id="ref-for-dom-mlbatchnormalizationoptions-axis⑦">axis</a></code>], then <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#dfn-throw" id="ref-for-dfn-throw④④">throw</a> a <code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#exceptiondef-typeerror" id="ref-for-exceptiondef-typeerror①④">TypeError</a></code>.</p>
+       </ol>
+      <li data-md>
+       <p>If <var>options</var>.<code class="idl"><a data-link-type="idl" href="#dom-mlbatchnormalizationoptions-bias" id="ref-for-dom-mlbatchnormalizationoptions-bias①">bias</a></code> <a data-link-type="dfn" href="https://infra.spec.whatwg.org/#map-exists" id="ref-for-map-exists①⓪">exists</a>:</p>
+       <ol>
         <li data-md>
-         <p><em>axis</em>: an <code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#idl-unsigned-long" id="ref-for-idl-unsigned-long②">unsigned long</a></code> scalar. The index to the feature count dimension of the input shape for which the mean and variance values are. Its value must be in the range [0, N-1] where N is the rank of input tensor. When it’s not specified, the default value is 1.</p>
+         <p>If its <a data-link-type="dfn" href="https://infra.spec.whatwg.org/#list-size" id="ref-for-list-size⑨">size</a> is not 1, then <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#dfn-throw" id="ref-for-dfn-throw④⑤">throw</a> a <code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#exceptiondef-typeerror" id="ref-for-exceptiondef-typeerror①⑤">TypeError</a></code>.</p>
         <li data-md>
-         <p><em>epsilon</em>: a <code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#idl-float" id="ref-for-idl-float①">float</a></code> scalar. A small value to prevent computational error due to divide-by-zero. The default value is 0.00001 when not specified.</p>
+         <p>If <var>options</var>.<code class="idl"><a data-link-type="idl" href="#dom-mlbatchnormalizationoptions-bias" id="ref-for-dom-mlbatchnormalizationoptions-bias②">bias</a></code>.<code class="idl"><a data-link-type="idl" href="#dom-mloperand-descriptor-slot" id="ref-for-dom-mloperand-descriptor-slot①④">[[descriptor]]</a></code>.<code class="idl"><a data-link-type="idl" href="#dom-mloperanddescriptor-dimensions" id="ref-for-dom-mloperanddescriptor-dimensions①⑨">dimensions</a></code>[0] is not equal to <var>input</var>.<code class="idl"><a data-link-type="idl" href="#dom-mloperand-descriptor-slot" id="ref-for-dom-mloperand-descriptor-slot①⑤">[[descriptor]]</a></code>.<code class="idl"><a data-link-type="idl" href="#dom-mloperanddescriptor-dimensions" id="ref-for-dom-mloperanddescriptor-dimensions②⓪">dimensions</a></code>[<var>options</var>.<code class="idl"><a data-link-type="idl" href="#dom-mlbatchnormalizationoptions-axis" id="ref-for-dom-mlbatchnormalizationoptions-axis⑧">axis</a></code>], then <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#dfn-throw" id="ref-for-dfn-throw④⑥">throw</a> a <code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#exceptiondef-typeerror" id="ref-for-exceptiondef-typeerror①⑥">TypeError</a></code>.</p>
+       </ol>
+      <li data-md>
+       <p>If any of the following sub-steps fail, <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#dfn-throw" id="ref-for-dfn-throw④⑦">throw</a> an "<code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#operationerror" id="ref-for-operationerror①④">OperationError</a></code>" <code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#idl-DOMException" id="ref-for-idl-DOMException③④">DOMException</a></code>.</p>
+       <ol>
         <li data-md>
-         <p><em>activation</em>: an <code class="idl"><a data-link-type="idl" href="#mlactivation" id="ref-for-mlactivation③">MLActivation</a></code>. The optional activation function that immediately follows the normalization operation.</p>
-       </ul>
-     </ul>
-     <p><strong>Returns:</strong> an <code class="idl"><a data-link-type="idl" href="#mloperand" id="ref-for-mloperand④②">MLOperand</a></code>. The batch-normalized N-D tensor of the same shape as the input tensor.</p>
-     <p>When <em>input</em> is a 4-D tensor of the <em>"nchw"</em> or <em>"nhwc"</em> layout, <em>options.axis</em> should be set to 1 or 3 respectively. The axis value designates the feature or channel count dimension of the input tensor.</p>
-     <div class="note" role="note">
-       The behavior of this operation when the input tensor is 4-D of the <em>"nchw"</em> layout and the activation is of operator type <em>relu</em> can be generically emulated from the usage of other operations as follow. However, user agents typically have a more efficient implementation for it, therefore its usage is encouraged from the performance standpoint. 
+         <p>Let <var>output</var> be the result of <a data-link-type="dfn" href="#create-an-mloperand" id="ref-for-create-an-mloperand③">creating an MLOperand</a> given <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#this" id="ref-for-this①⓪">this</a> and <var>input</var>.<code class="idl"><a data-link-type="idl" href="#dom-mloperand-descriptor-slot" id="ref-for-dom-mloperand-descriptor-slot①⑥">[[descriptor]]</a></code>, that may use the same underlying data as <var>input</var>.</p>
+        <li data-md>
+         <p>Make a request to the underlying platform to initialize the batch normalization:</p>
+         <ol>
+          <li data-md>
+           <p>Create an <a data-link-type="dfn" href="https://infra.spec.whatwg.org/#implementation-defined" id="ref-for-implementation-defined⑨">implementation-defined</a> platform operator <var>batchNormImpl</var> for this method, given <var>input</var>, <var>mean</var>, <var>variance</var> and <var>options</var>.</p>
+          <li data-md>
+           <p>If <var>options</var>.activation <a data-link-type="dfn" href="https://infra.spec.whatwg.org/#map-exists" id="ref-for-map-exists①①">exists</a>,register it as activation to <var>batchNormImpl</var>.</p>
+          <li data-md>
+           <p>Connect <var>output</var> as output to <var>batchNormImpl</var>.</p>
+         </ol>
+       </ol>
+      <li data-md>
+       <p>Return <var>output</var>.</p>
+     </ol>
+    </div>
+   </details>
+   <div class="note" role="note">
+    <details open>
+     <summary> The behavior of this operation when the input tensor is 4-D of the <em>"nchw"</em> layout and the activation is of operator type <em>relu</em> can be generically emulated from the usage of other operations as follow. However, user agents typically have a more efficient implementation for it, therefore its usage is encouraged from the performance standpoint. </summary>
 <pre class="highlight"><c- a>const</c-> shape <c- o>=</c-> <c- p>[</c-><c- mf>1</c-><c- p>,</c-><c- kc>null</c-><c- p>,</c-><c- mf>1</c-><c- p>,</c-><c- mf>1</c-><c- p>];</c->
 <c- k>return</c-> builder<c- p>.</c->relu<c- p>(</c->
   builder<c- p>.</c->add<c- p>(</c->
@@ -2229,86 +2868,238 @@ <h4 class="heading settled" data-level="7.6.4" id="api-mlgraphbuilder-batchnorm"
         <c- p>)),</c->
     builder<c- p>.</c->reshape<c- p>(</c->options<c- p>.</c->bias<c- p>,</c-> shape<c- p>)));</c->
 </pre>
-     </div>
-    </div>
-    <h4 class="heading settled" data-level="7.6.5" id="api-mlgraphbuilder-clamp"><span class="secno">7.6.5. </span><span class="content">The clamp() method</span><a class="self-link" href="#api-mlgraphbuilder-clamp"></a></h4>
-     Clamp the input tensor element-wise within a range specified by the minimum and maximum values. 
+    </details>
+   </div>
+   <h4 class="heading settled" data-level="7.6.6" id="api-mlgraphbuilder-clamp"><span class="secno">7.6.6. </span><span class="content">The clamp() method</span><a class="self-link" href="#api-mlgraphbuilder-clamp"></a></h4>
+    Clamp the input tensor element-wise within a range specified by the minimum and maximum values. 
 <pre class="idl highlight def"><c- b>dictionary</c-> <dfn class="dfn-paneled idl-code" data-dfn-type="dictionary" data-export id="dictdef-mlclampoptions"><code><c- g>MLClampOptions</c-></code></dfn> {
-  <a class="idl-code" data-link-type="interface" href="https://webidl.spec.whatwg.org/#idl-float" id="ref-for-idl-float②"><c- b>float</c-></a> <dfn class="idl-code" data-dfn-for="MLClampOptions" data-dfn-type="dict-member" data-export data-type="float " id="dom-mlclampoptions-minvalue"><code><c- g>minValue</c-></code><a class="self-link" href="#dom-mlclampoptions-minvalue"></a></dfn>;
-  <a class="idl-code" data-link-type="interface" href="https://webidl.spec.whatwg.org/#idl-float" id="ref-for-idl-float③"><c- b>float</c-></a> <dfn class="idl-code" data-dfn-for="MLClampOptions" data-dfn-type="dict-member" data-export data-type="float " id="dom-mlclampoptions-maxvalue"><code><c- g>maxValue</c-></code><a class="self-link" href="#dom-mlclampoptions-maxvalue"></a></dfn>;
+  <a class="idl-code" data-link-type="interface" href="https://webidl.spec.whatwg.org/#idl-float" id="ref-for-idl-float③"><c- b>float</c-></a> <dfn class="dfn-paneled idl-code" data-dfn-for="MLClampOptions" data-dfn-type="dict-member" data-export data-type="float " id="dom-mlclampoptions-minvalue"><code><c- g>minValue</c-></code></dfn>;
+  <a class="idl-code" data-link-type="interface" href="https://webidl.spec.whatwg.org/#idl-float" id="ref-for-idl-float④"><c- b>float</c-></a> <dfn class="dfn-paneled idl-code" data-dfn-for="MLClampOptions" data-dfn-type="dict-member" data-export data-type="float " id="dom-mlclampoptions-maxvalue"><code><c- g>maxValue</c-></code></dfn>;
 };
 
-<c- b>partial</c-> <c- b>interface</c-> <a class="idl-code" data-link-type="interface" href="#mlgraphbuilder" id="ref-for-mlgraphbuilder②①"><c- g>MLGraphBuilder</c-></a> {
-  <a data-link-type="idl-name" href="#mloperand" id="ref-for-mloperand④③"><c- n>MLOperand</c-></a> <dfn class="idl-code" data-dfn-for="MLGraphBuilder" data-dfn-type="method" data-export data-lt="clamp(x, options)|clamp(x)" id="dom-mlgraphbuilder-clamp"><code><c- g>clamp</c-></code><a class="self-link" href="#dom-mlgraphbuilder-clamp"></a></dfn>(<a data-link-type="idl-name" href="#mloperand" id="ref-for-mloperand④④"><c- n>MLOperand</c-></a> <dfn class="idl-code" data-dfn-for="MLGraphBuilder/clamp(x, options), MLGraphBuilder/clamp(x)" data-dfn-type="argument" data-export id="dom-mlgraphbuilder-clamp-x-options-x"><code><c- g>x</c-></code><a class="self-link" href="#dom-mlgraphbuilder-clamp-x-options-x"></a></dfn>, <c- b>optional</c-> <a data-link-type="idl-name" href="#dictdef-mlclampoptions" id="ref-for-dictdef-mlclampoptions"><c- n>MLClampOptions</c-></a> <dfn class="idl-code" data-dfn-for="MLGraphBuilder/clamp(x, options), MLGraphBuilder/clamp(x)" data-dfn-type="argument" data-export id="dom-mlgraphbuilder-clamp-x-options-options"><code><c- g>options</c-></code><a class="self-link" href="#dom-mlgraphbuilder-clamp-x-options-options"></a></dfn> = {});
-  <a data-link-type="idl-name" href="#mlactivation" id="ref-for-mlactivation④"><c- n>MLActivation</c-></a> <dfn class="idl-code" data-dfn-for="MLGraphBuilder" data-dfn-type="method" data-export data-lt="clamp(options)|clamp()" id="dom-mlgraphbuilder-clamp-options"><code><c- g>clamp</c-></code><a class="self-link" href="#dom-mlgraphbuilder-clamp-options"></a></dfn>(<c- b>optional</c-> <a data-link-type="idl-name" href="#dictdef-mlclampoptions" id="ref-for-dictdef-mlclampoptions①"><c- n>MLClampOptions</c-></a> <dfn class="idl-code" data-dfn-for="MLGraphBuilder/clamp(options), MLGraphBuilder/clamp()" data-dfn-type="argument" data-export id="dom-mlgraphbuilder-clamp-options-options"><code><c- g>options</c-></code><a class="self-link" href="#dom-mlgraphbuilder-clamp-options-options"></a></dfn> = {});
+<c- b>partial</c-> <c- b>interface</c-> <a class="idl-code" data-link-type="interface" href="#mlgraphbuilder" id="ref-for-mlgraphbuilder②④"><c- g>MLGraphBuilder</c-></a> {
+  <a data-link-type="idl-name" href="#mloperand" id="ref-for-mloperand④⑨"><c- n>MLOperand</c-></a> <a class="idl-code" data-link-type="method" href="#dom-mlgraphbuilder-clamp" id="ref-for-dom-mlgraphbuilder-clamp"><c- g>clamp</c-></a>(<a data-link-type="idl-name" href="#mloperand" id="ref-for-mloperand⑤⓪"><c- n>MLOperand</c-></a> <dfn class="idl-code" data-dfn-for="MLGraphBuilder/clamp(operand, options), MLGraphBuilder/clamp(operand)" data-dfn-type="argument" data-export id="dom-mlgraphbuilder-clamp-operand-options-operand"><code><c- g>operand</c-></code><a class="self-link" href="#dom-mlgraphbuilder-clamp-operand-options-operand"></a></dfn>, <c- b>optional</c-> <a data-link-type="idl-name" href="#dictdef-mlclampoptions" id="ref-for-dictdef-mlclampoptions"><c- n>MLClampOptions</c-></a> <dfn class="idl-code" data-dfn-for="MLGraphBuilder/clamp(operand, options), MLGraphBuilder/clamp(operand)" data-dfn-type="argument" data-export id="dom-mlgraphbuilder-clamp-operand-options-options"><code><c- g>options</c-></code><a class="self-link" href="#dom-mlgraphbuilder-clamp-operand-options-options"></a></dfn> = {});
+  <a data-link-type="idl-name" href="#mlactivation" id="ref-for-mlactivation①①"><c- n>MLActivation</c-></a> <a class="idl-code" data-link-type="method" href="#dom-mlgraphbuilder-clamp-options" id="ref-for-dom-mlgraphbuilder-clamp-options"><c- g>clamp</c-></a>(<c- b>optional</c-> <a data-link-type="idl-name" href="#dictdef-mlclampoptions" id="ref-for-dictdef-mlclampoptions①"><c- n>MLClampOptions</c-></a> <dfn class="idl-code" data-dfn-for="MLGraphBuilder/clamp(options), MLGraphBuilder/clamp()" data-dfn-type="argument" data-export id="dom-mlgraphbuilder-clamp-options-options"><code><c- g>options</c-></code><a class="self-link" href="#dom-mlgraphbuilder-clamp-options-options"></a></dfn> = {});
 };
 </pre>
-    <div class="algorithm" data-algorithm="clamp">
-      <strong>Arguments:</strong> 
-     <ul>
-      <li data-md>
-       <p><em>x</em>: an <code class="idl"><a data-link-type="idl" href="#mloperand" id="ref-for-mloperand④⑤">MLOperand</a></code>. The input tensor.</p>
-      <li data-md>
-       <p><em>options</em>: an optional <code class="idl"><a data-link-type="idl" href="#dictdef-mlclampoptions" id="ref-for-dictdef-mlclampoptions②">MLClampOptions</a></code>. The optional parameters of the operation.</p>
-       <ul>
-        <li data-md>
-         <p><em>minValue</em>: a <code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#idl-float" id="ref-for-idl-float④">float</a></code> scalar. Specifies the minimum value of the range. When it is not specified, the clamping is not performed on the lower limit of the range.</p>
-        <li data-md>
-         <p><em>maxValue</em>: a <code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#idl-float" id="ref-for-idl-float⑤">float</a></code> scalar. Specifies the maximum value of the range. When it is not specified, the clamping is not performed on the upper limit of the range.</p>
-       </ul>
-     </ul>
-     <p><strong>Returns:</strong></p>
-     <ul>
-      <li data-md>
-       <p>an <code class="idl"><a data-link-type="idl" href="#mloperand" id="ref-for-mloperand④⑥">MLOperand</a></code>. The output tensor of the same shape as <em>x</em>.</p>
-      <li data-md>
-       <p>an <code class="idl"><a data-link-type="idl" href="#mlactivation" id="ref-for-mlactivation⑤">MLActivation</a></code>. The activation function representing the clamp operation.</p>
-     </ul>
-     <div class="note" role="note">
-       The behavior of this operation can be generically emulated from the usage of
+   <div class="note" role="note">
+    <details open>
+     <summary> The behavior of this operation can be generically emulated from the usage of
     other operations as follow. However, user agents typically have a more
     efficient implementation for it, therefore its usage is encouraged from the
-    performance standpoint. 
+    performance standpoint. </summary>
 <pre class="highlight"><c- k>if</c-> <c- p>(</c->options<c- p>.</c->minValue <c- o>===</c-> <c- kc>undefined</c-><c- p>)</c-> <c- p>{</c->
   <c- k>if</c-> <c- p>(</c->options<c- p>.</c->maxValue <c- o>===</c-> <c- kc>undefined</c-><c- p>)</c-> <c- p>{</c->
-    <c- k>return</c-> x<c- p>;</c->
+    <c- k>return</c-> operand<c- p>;</c->
   <c- p>}</c-> <c- k>else</c-> <c- p>{</c->
-    <c- k>return</c-> builder<c- p>.</c->min<c- p>(</c->x<c- p>,</c-> builder<c- p>.</c->constant<c- p>(</c->options<c- p>.</c->maxValue<c- p>));</c->
+    <c- k>return</c-> builder<c- p>.</c->min<c- p>(</c->operand<c- p>,</c-> builder<c- p>.</c->constant<c- p>(</c->options<c- p>.</c->maxValue<c- p>));</c->
   <c- p>}</c->
 <c- p>}</c-> <c- k>else</c-> <c- p>{</c->
   <c- k>if</c-> <c- p>(</c->options<c- p>.</c->maxValue <c- o>===</c-> <c- kc>undefined</c-><c- p>)</c-> <c- p>{</c->
-    <c- k>return</c-> builder<c- p>.</c->max<c- p>(</c->x<c- p>,</c-> builder<c- p>.</c->constant<c- p>(</c->options<c- p>.</c->minValue<c- p>));</c->
+    <c- k>return</c-> builder<c- p>.</c->max<c- p>(</c->operand<c- p>,</c-> builder<c- p>.</c->constant<c- p>(</c->options<c- p>.</c->minValue<c- p>));</c->
   <c- p>}</c-> <c- k>else</c-> <c- p>{</c->
     <c- k>return</c-> builder<c- p>.</c->min<c- p>(</c->
-        builder<c- p>.</c->max<c- p>(</c->x<c- p>,</c-> builder<c- p>.</c->constant<c- p>(</c->options<c- p>.</c->minValue<c- p>)),</c->
+        builder<c- p>.</c->max<c- p>(</c->operand<c- p>,</c-> builder<c- p>.</c->constant<c- p>(</c->options<c- p>.</c->minValue<c- p>)),</c->
         builder<c- p>.</c->constant<c- p>(</c->options<c- p>.</c->maxValue<c- p>));</c->
   <c- p>}</c->
 <c- p>}</c->
 </pre>
-     </div>
+    </details>
+   </div>
+   <details class="algorithm" data-algorithm="check clamp options" open>
+    <summary> To <dfn class="dfn-paneled" data-dfn-type="dfn" data-noexport id="check-clamp-options">check clamp options</dfn> given <var>options</var>, run the following steps: </summary>
+    <div class="algorithm-steps">
+     <ol>
+      <li data-md>
+       <p>If <var>options</var>.<code class="idl"><a data-link-type="idl" href="#dom-mlclampoptions-minvalue" id="ref-for-dom-mlclampoptions-minvalue">minValue</a></code> is greater than <var>options</var>.<code class="idl"><a data-link-type="idl" href="#dom-mlclampoptions-maxvalue" id="ref-for-dom-mlclampoptions-maxvalue">maxValue</a></code>, then return false.</p>
+      <li data-md>
+       <p>Return true.</p>
+     </ol>
+    </div>
+   </details>
+   <h5 class="heading settled" data-level="7.6.6.1" id="api-mlgraphbuilder-clamp-operand-options"><span class="secno">7.6.6.1. </span><span class="content">The <code class="idl"><a data-link-type="idl" href="#dom-mlgraphbuilder-clamp" id="ref-for-dom-mlgraphbuilder-clamp①">clamp(operand, options)</a></code> method</span><a class="self-link" href="#api-mlgraphbuilder-clamp-operand-options"></a></h5>
+   <div>
+     <strong>Arguments:</strong> 
+    <ul>
+     <li data-md>
+      <p><em>operand</em>: an <code class="idl"><a data-link-type="idl" href="#mloperand" id="ref-for-mloperand⑤①">MLOperand</a></code>. The input tensor.</p>
+     <li data-md>
+      <p><em>options</em>: an optional <code class="idl"><a data-link-type="idl" href="#dictdef-mlclampoptions" id="ref-for-dictdef-mlclampoptions②">MLClampOptions</a></code>. The optional parameters of the operation.</p>
+      <ul>
+       <li data-md>
+        <p><em>minValue</em>: a <code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#idl-float" id="ref-for-idl-float⑤">float</a></code> scalar. Specifies the minimum value of the range. When it is not specified, the clamping is not performed on the lower limit of the range.</p>
+       <li data-md>
+        <p><em>maxValue</em>: a <code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#idl-float" id="ref-for-idl-float⑥">float</a></code> scalar. Specifies the maximum value of the range. When it is not specified, the clamping is not performed on the upper limit of the range.</p>
+      </ul>
+    </ul>
+     <strong>Returns:</strong> 
+    <ul>
+     <li data-md>
+      <p>an <code class="idl"><a data-link-type="idl" href="#mloperand" id="ref-for-mloperand⑤②">MLOperand</a></code>. The output tensor of the same shape as <em>operand</em>.</p>
+    </ul>
+   </div>
+   <details class="algorithm" data-algorithm="clamp(operand, options)" data-algorithm-for="MLGraphBuilder" open>
+    <summary> The <dfn class="dfn-paneled idl-code" data-dfn-for="MLGraphBuilder" data-dfn-type="method" data-export data-lt="clamp(operand, options)|clamp(operand)" id="dom-mlgraphbuilder-clamp"><code>clamp(<var>operand</var>, <var>options</var>)</code></dfn> method steps are: </summary>
+    <div class="algorithm-steps">
+     <ol>
+      <li data-md>
+       <p><a data-link-type="dfn" href="https://infra.spec.whatwg.org/#assert" id="ref-for-assert①⑥">Assert</a>: the type of <var>operand</var> is <code class="idl"><a data-link-type="idl" href="#mloperand" id="ref-for-mloperand⑤③">MLOperand</a></code>.</p>
+      <li data-md>
+       <p>If running the <a data-link-type="dfn" href="#check-clamp-options" id="ref-for-check-clamp-options">check clamp options</a> steps given <var>options</var> returns false, then <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#dfn-throw" id="ref-for-dfn-throw④⑧">throw</a> a <code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#exceptiondef-typeerror" id="ref-for-exceptiondef-typeerror①⑦">TypeError</a></code>.</p>
+      <li data-md>
+       <p>If any of the following sub-steps fail, <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#dfn-throw" id="ref-for-dfn-throw④⑨">throw</a> an "<code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#operationerror" id="ref-for-operationerror①⑤">OperationError</a></code>" <code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#idl-DOMException" id="ref-for-idl-DOMException③⑤">DOMException</a></code>.</p>
+       <ol>
+        <li data-md>
+         <p>Let <var>output</var> be the result of <a data-link-type="dfn" href="#copy-an-mloperand" id="ref-for-copy-an-mloperand">copying an MLOperand</a> given <var>operand</var>.</p>
+        <li data-md>
+         <p>Make a request to the underlying platform to:</p>
+         <ol>
+          <li data-md>
+           <p>Create an <a data-link-type="dfn" href="https://infra.spec.whatwg.org/#implementation-defined" id="ref-for-implementation-defined①⓪">implementation-defined</a> platform operator <var>clampImpl</var> for this method, given <var>options</var>.<code class="idl"><a data-link-type="idl" href="#dom-mlclampoptions-minvalue" id="ref-for-dom-mlclampoptions-minvalue①">minValue</a></code> and <var>options</var>.<code class="idl"><a data-link-type="idl" href="#dom-mlclampoptions-maxvalue" id="ref-for-dom-mlclampoptions-maxvalue①">maxValue</a></code>.</p>
+          <li data-md>
+           <p>Store a reference of <var>clampImpl</var> in <var>output</var>.<code class="idl"><a data-link-type="idl" href="#dom-mloperand-operator-slot" id="ref-for-dom-mloperand-operator-slot①">[[operator]]</a></code>.</p>
+          <li data-md>
+           <p>Create an <a data-link-type="dfn" href="https://infra.spec.whatwg.org/#implementation-defined" id="ref-for-implementation-defined①①">implementation-defined</a> platform operand <var>outputImpl</var> to represent clamp output, given <var>output</var> and <var>clampImpl</var>.</p>
+          <li data-md>
+           <p>Store a reference to <var>outputImpl</var> in <var>output</var>.<code class="idl"><a data-link-type="idl" href="#dom-mloperand-operand-slot" id="ref-for-dom-mloperand-operand-slot④">[[operand]]</a></code>.</p>
+         </ol>
+        <li data-md>
+         <p>Connect <var>operand</var>.<code class="idl"><a data-link-type="idl" href="#dom-mloperand-operand-slot" id="ref-for-dom-mloperand-operand-slot⑤">[[operand]]</a></code> as input to <var>clampImpl</var>.</p>
+        <li data-md>
+         <p>Connect <var>output</var>.<code class="idl"><a data-link-type="idl" href="#dom-mloperand-operand-slot" id="ref-for-dom-mloperand-operand-slot⑥">[[operand]]</a></code> as output to <var>clampImpl</var>.</p>
+       </ol>
+      <li data-md>
+       <p>Return <var>output</var>.</p>
+     </ol>
+    </div>
+   </details>
+   <h5 class="heading settled" data-level="7.6.6.2" id="api-mlgraphbuilder-clamp-options"><span class="secno">7.6.6.2. </span><span class="content">The <code class="idl"><a data-link-type="idl" href="#dom-mlgraphbuilder-clamp-options" id="ref-for-dom-mlgraphbuilder-clamp-options①">clamp(options)</a></code> method</span><a class="self-link" href="#api-mlgraphbuilder-clamp-options"></a></h5>
+   <div>
+     <strong>Arguments:</strong> 
+    <ul>
+     <li data-md>
+      <p><em>options</em>: an optional <code class="idl"><a data-link-type="idl" href="#dictdef-mlclampoptions" id="ref-for-dictdef-mlclampoptions③">MLClampOptions</a></code>. The optional parameters of the operation.</p>
+      <ul>
+       <li data-md>
+        <p><em>minValue</em>: a <code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#idl-float" id="ref-for-idl-float⑦">float</a></code> scalar. Specifies the minimum value of the range. When it is not specified, the clamping is not performed on the lower limit of the range.</p>
+       <li data-md>
+        <p><em>maxValue</em>: a <code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#idl-float" id="ref-for-idl-float⑧">float</a></code> scalar. Specifies the maximum value of the range. When it is not specified, the clamping is not performed on the upper limit of the range.</p>
+      </ul>
+    </ul>
+     <strong>Returns:</strong> 
+    <ul>
+     <li data-md>
+      <p>an <code class="idl"><a data-link-type="idl" href="#mlactivation" id="ref-for-mlactivation①②">MLActivation</a></code>. The operator representing the clamp operation.</p>
+    </ul>
+   </div>
+   <details class="algorithm" data-algorithm="clamp(options)" data-algorithm-for="MLGraphBuilder" open>
+    <summary> The <dfn class="dfn-paneled idl-code" data-dfn-for="MLGraphBuilder" data-dfn-type="method" data-export data-lt="clamp(options)|clamp()" id="dom-mlgraphbuilder-clamp-options"><code>clamp(<var>options</var>)</code></dfn> method steps are: </summary>
+    <div class="algorithm-steps">
+     <ol>
+      <li data-md>
+       <p>If running the <a data-link-type="dfn" href="#check-clamp-options" id="ref-for-check-clamp-options①">check clamp options</a> steps given <var>options</var> returns false, then <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#dfn-throw" id="ref-for-dfn-throw⑤⓪">throw</a> a <code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#exceptiondef-typeerror" id="ref-for-exceptiondef-typeerror①⑧">TypeError</a></code>.</p>
+      <li data-md>
+       <p>Let <var>op</var> be the result of <a data-link-type="dfn" href="#create-an-mlactivation" id="ref-for-create-an-mlactivation">creating an MLActivation</a> given <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#this" id="ref-for-this①①">this</a>, <code>"clamp"</code> and <var>options</var>.</p>
+       <ol>
+        <li data-md>
+         <p>If that <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#dfn-throw" id="ref-for-dfn-throw⑤①">throws</a> an error, re-<a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#dfn-throw" id="ref-for-dfn-throw⑤②">throw</a> the error.</p>
+       </ol>
+      <li data-md>
+       <p>Return <var>op</var>.</p>
+     </ol>
     </div>
-    <h4 class="heading settled" data-level="7.6.6" id="api-mlgraphbuilder-concat"><span class="secno">7.6.6. </span><span class="content">The concat() method</span><a class="self-link" href="#api-mlgraphbuilder-concat"></a></h4>
-     Concatenates the input tensors along a given axis. 
-<pre class="idl highlight def"><c- b>partial</c-> <c- b>interface</c-> <a class="idl-code" data-link-type="interface" href="#mlgraphbuilder" id="ref-for-mlgraphbuilder②②"><c- g>MLGraphBuilder</c-></a> {
-  <a data-link-type="idl-name" href="#mloperand" id="ref-for-mloperand④⑦"><c- n>MLOperand</c-></a> <dfn class="idl-code" data-dfn-for="MLGraphBuilder" data-dfn-type="method" data-export data-lt="concat(inputs, axis)" id="dom-mlgraphbuilder-concat"><code><c- g>concat</c-></code><a class="self-link" href="#dom-mlgraphbuilder-concat"></a></dfn>(<a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#idl-sequence" id="ref-for-idl-sequence①"><c- b>sequence</c-></a>&lt;<a data-link-type="idl-name" href="#mloperand" id="ref-for-mloperand④⑧"><c- n>MLOperand</c-></a>> <dfn class="idl-code" data-dfn-for="MLGraphBuilder/concat(inputs, axis)" data-dfn-type="argument" data-export id="dom-mlgraphbuilder-concat-inputs-axis-inputs"><code><c- g>inputs</c-></code><a class="self-link" href="#dom-mlgraphbuilder-concat-inputs-axis-inputs"></a></dfn>, <a class="idl-code" data-link-type="interface" href="https://webidl.spec.whatwg.org/#idl-unsigned-long" id="ref-for-idl-unsigned-long③"><c- b>unsigned</c-> <c- b>long</c-></a> <dfn class="idl-code" data-dfn-for="MLGraphBuilder/concat(inputs, axis)" data-dfn-type="argument" data-export id="dom-mlgraphbuilder-concat-inputs-axis-axis"><code><c- g>axis</c-></code><a class="self-link" href="#dom-mlgraphbuilder-concat-inputs-axis-axis"></a></dfn>);
+   </details>
+   <h4 class="heading settled" data-level="7.6.7" id="api-mlgraphbuilder-concat"><span class="secno">7.6.7. </span><span class="content">The concat() method</span><a class="self-link" href="#api-mlgraphbuilder-concat"></a></h4>
+    Concatenates the input tensors along a given axis. 
+<pre class="idl highlight def"><c- b>partial</c-> <c- b>interface</c-> <a class="idl-code" data-link-type="interface" href="#mlgraphbuilder" id="ref-for-mlgraphbuilder②⑤"><c- g>MLGraphBuilder</c-></a> {
+  <a data-link-type="idl-name" href="#mloperand" id="ref-for-mloperand⑤④"><c- n>MLOperand</c-></a> <a class="idl-code" data-link-type="method" href="#dom-mlgraphbuilder-concat" id="ref-for-dom-mlgraphbuilder-concat"><c- g>concat</c-></a>(<a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#idl-sequence" id="ref-for-idl-sequence①"><c- b>sequence</c-></a>&lt;<a data-link-type="idl-name" href="#mloperand" id="ref-for-mloperand⑤⑤"><c- n>MLOperand</c-></a>> <dfn class="idl-code" data-dfn-for="MLGraphBuilder/concat(inputs, axis)" data-dfn-type="argument" data-export id="dom-mlgraphbuilder-concat-inputs-axis-inputs"><code><c- g>inputs</c-></code><a class="self-link" href="#dom-mlgraphbuilder-concat-inputs-axis-inputs"></a></dfn>, <a class="idl-code" data-link-type="interface" href="https://webidl.spec.whatwg.org/#idl-unsigned-long" id="ref-for-idl-unsigned-long③"><c- b>unsigned</c-> <c- b>long</c-></a> <dfn class="idl-code" data-dfn-for="MLGraphBuilder/concat(inputs, axis)" data-dfn-type="argument" data-export id="dom-mlgraphbuilder-concat-inputs-axis-axis"><code><c- g>axis</c-></code><a class="self-link" href="#dom-mlgraphbuilder-concat-inputs-axis-axis"></a></dfn>);
 };
 </pre>
-    <div class="algorithm" data-algorithm="concat">
-      <strong>Arguments:</strong> 
-     <ul>
-      <li data-md>
-       <p><em>inputs</em>: a sequence of <code class="idl"><a data-link-type="idl" href="#mloperand" id="ref-for-mloperand④⑨">MLOperand</a></code>. All input tensors must have the
+   <div>
+     <strong>Arguments:</strong> 
+    <ul>
+     <li data-md>
+      <p><em>inputs</em>: a sequence of <code class="idl"><a data-link-type="idl" href="#mloperand" id="ref-for-mloperand⑤⑥">MLOperand</a></code>. All input tensors must have the
 same shape, except for the size of the dimension to concatenate on.</p>
-      <li data-md>
-       <p><em>axis</em>: an <code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#idl-unsigned-long" id="ref-for-idl-unsigned-long④">unsigned long</a></code> scalar. The axis that the inputs concatenate along. Its value must be in the range [0, N-1] where N is the rank of input tensors.</p>
-     </ul>
-     <p><strong>Returns:</strong> an <code class="idl"><a data-link-type="idl" href="#mloperand" id="ref-for-mloperand⑤⓪">MLOperand</a></code>. The concatenated tensor of all the inputs along
+     <li data-md>
+      <p><em>axis</em>: an <code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#idl-unsigned-long" id="ref-for-idl-unsigned-long④">unsigned long</a></code> scalar. The axis that the inputs concatenate along. Its value must be in the range [0, N-1] where N is the <a data-link-type="dfn" href="#mloperand-rank" id="ref-for-mloperand-rank②">rank</a> of the input tensors.</p>
+    </ul>
+    <p><strong>Returns:</strong> an <code class="idl"><a data-link-type="idl" href="#mloperand" id="ref-for-mloperand⑤⑦">MLOperand</a></code>. The concatenated tensor of all the inputs along
     the <em>axis</em>. The output tensor has the same shape except on the dimension
     that all the inputs concatenated along. The size of that dimension is
     computed as the sum of all the input sizes of the same dimension.</p>
+   </div>
+   <details class="algorithm" data-algorithm="concat(inputs, axis)" data-algorithm-for="MLGraphBuilder" open>
+    <summary> The <dfn class="dfn-paneled idl-code" data-dfn-for="MLGraphBuilder" data-dfn-type="method" data-export id="dom-mlgraphbuilder-concat"><code>concat(<var>inputs</var>, <var>axis</var>)</code></dfn> method steps are: </summary>
+    <div class="algorithm-steps">
+     <div class="note" role="note"> The permissions and context validity have been checked by <a href="#api-mlgraphbuilder-constructor">§ 7.6.1 The MLGraphBuilder constructor</a> steps. </div>
+     <ol>
+      <li data-md>
+       <p><a data-link-type="dfn" href="https://infra.spec.whatwg.org/#assert" id="ref-for-assert①⑦">Assert</a>: the type of <var>inputs</var> is sequence of <code class="idl"><a data-link-type="idl" href="#mloperand" id="ref-for-mloperand⑤⑧">MLOperand</a></code> objects.</p>
+      <li data-md>
+       <p><a data-link-type="dfn" href="https://infra.spec.whatwg.org/#assert" id="ref-for-assert①⑧">Assert</a>: the type of <var>axis</var> is <code>unsigned long</code>.</p>
+      <li data-md>
+       <p><a data-link-type="dfn" href="https://infra.spec.whatwg.org/#assert" id="ref-for-assert①⑨">Assert</a>: the shape, i.e. <code class="idl"><a data-link-type="idl" href="#dom-mloperanddescriptor-dimensions" id="ref-for-dom-mloperanddescriptor-dimensions②①">dimensions</a></code> of each operand in <var>inputs</var> is the same, except on the dimension given by <var>axis</var> on which they are concatenated.</p>
+      <li data-md>
+       <p><a data-link-type="dfn" href="https://infra.spec.whatwg.org/#assert" id="ref-for-assert②⓪">Assert</a>: the <code class="idl"><a data-link-type="idl" href="#dom-mloperanddescriptor-type" id="ref-for-dom-mloperanddescriptor-type⑥">type</a></code> of each operand in <var>inputs</var> is the same.</p>
+      <li data-md>
+       <p>If any of the following steps fail, then <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#dfn-throw" id="ref-for-dfn-throw⑤③">throw</a> a "<code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#dataerror" id="ref-for-dataerror①①">DataError</a></code>" <code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#idl-DOMException" id="ref-for-idl-DOMException③⑥">DOMException</a></code>.</p>
+       <ol>
+        <li data-md>
+         <p>Let <var>desc</var> be <var>inputs</var>[0].<code class="idl"><a data-link-type="idl" href="#dom-mloperand-descriptor-slot" id="ref-for-dom-mloperand-descriptor-slot①⑦">[[descriptor]]</a></code>.</p>
+        <li data-md>
+         <p>If <var>axis</var> is greater than or equal to the <a data-link-type="dfn" href="#mloperand-rank" id="ref-for-mloperand-rank③">rank</a> of <var>desc</var>, then <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#dfn-throw" id="ref-for-dfn-throw⑤④">throw</a> a "<code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#dataerror" id="ref-for-dataerror①②">DataError</a></code>" <code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#idl-DOMException" id="ref-for-idl-DOMException③⑦">DOMException</a></code>.</p>
+        <li data-md>
+         <p>Let <var>desc</var>.<code class="idl"><a data-link-type="idl" href="#dom-mloperanddescriptor-dimensions" id="ref-for-dom-mloperanddescriptor-dimensions②②">dimensions</a></code>[<var>axis</var>] be 0.</p>
+        <li data-md>
+         <p><a data-link-type="dfn" href="https://infra.spec.whatwg.org/#map-iterate" id="ref-for-map-iterate①①">For each</a> <var>index</var> in <a data-link-type="dfn" href="https://infra.spec.whatwg.org/#the-range" id="ref-for-the-range①">the range</a> 0 to the <a data-link-type="dfn" href="#mloperand-rank" id="ref-for-mloperand-rank④">rank</a> of <var>inputs</var>, exclusive:</p>
+         <ol>
+          <li data-md>
+           <p>Let <var>input</var> be <var>inputs</var>[<var>index</var>].</p>
+          <li data-md>
+           <p>If <a data-link-type="dfn" href="#mloperand-validate-mloperand" id="ref-for-mloperand-validate-mloperand①">validating MLOperand</a> given <var>input</var> and <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#this" id="ref-for-this①②">this</a> returns false, then <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#dfn-throw" id="ref-for-dfn-throw⑤⑤">throw</a> a "<code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#dataerror" id="ref-for-dataerror①③">DataError</a></code>" <code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#idl-DOMException" id="ref-for-idl-DOMException③⑧">DOMException</a></code>.</p>
+          <li data-md>
+           <p>If <var>input</var>.<code class="idl"><a data-link-type="idl" href="#dom-mloperand-descriptor-slot" id="ref-for-dom-mloperand-descriptor-slot①⑧">[[descriptor]]</a></code>.<code class="idl"><a data-link-type="idl" href="#dom-mloperanddescriptor-type" id="ref-for-dom-mloperanddescriptor-type⑦">type</a></code> is not equal to <var>inputs</var>[0].<code class="idl"><a data-link-type="idl" href="#dom-mloperand-descriptor-slot" id="ref-for-dom-mloperand-descriptor-slot①⑨">[[descriptor]]</a></code>.<code class="idl"><a data-link-type="idl" href="#dom-mloperanddescriptor-type" id="ref-for-dom-mloperanddescriptor-type⑧">type</a></code>, then <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#dfn-throw" id="ref-for-dfn-throw⑤⑥">throw</a> a "<code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#dataerror" id="ref-for-dataerror①④">DataError</a></code>" <code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#idl-DOMException" id="ref-for-idl-DOMException③⑨">DOMException</a></code>.</p>
+          <li data-md>
+           <p><a data-link-type="dfn" href="https://infra.spec.whatwg.org/#map-iterate" id="ref-for-map-iterate①②">For each</a> <var>dim</var> in <a data-link-type="dfn" href="https://infra.spec.whatwg.org/#the-range" id="ref-for-the-range②">the range</a> 0 to the <a data-link-type="dfn" href="#mloperand-rank" id="ref-for-mloperand-rank⑤">rank</a> of <var>input</var>.<code class="idl"><a data-link-type="idl" href="#dom-mloperand-descriptor-slot" id="ref-for-dom-mloperand-descriptor-slot②⓪">[[descriptor]]</a></code>.<code class="idl"><a data-link-type="idl" href="#dom-mloperanddescriptor-dimensions" id="ref-for-dom-mloperanddescriptor-dimensions②③">dimensions</a></code>, exclusive:</p>
+           <div class="note" role="note"> If the shape of each corresponding dimension and type of the operands, except for those of the dimension given by <var>axis</var>, is not the same, fail. </div>
+           <ol>
+            <li data-md>
+             <p>If <var>dim</var> is not equal to <var>axis</var> and if <var>input</var>.<code class="idl"><a data-link-type="idl" href="#dom-mloperanddescriptor-dimensions" id="ref-for-dom-mloperanddescriptor-dimensions②④">dimensions</a></code>[<var>dim</var>] is not equal to <var>inputs</var>[0].<code class="idl"><a data-link-type="idl" href="#dom-mloperanddescriptor-dimensions" id="ref-for-dom-mloperanddescriptor-dimensions②⑤">dimensions</a></code>[<var>dim</var>], then <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#dfn-throw" id="ref-for-dfn-throw⑤⑦">throw</a> a "<code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#dataerror" id="ref-for-dataerror①⑤">DataError</a></code>" <code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#idl-DOMException" id="ref-for-idl-DOMException④⓪">DOMException</a></code>.</p>
+            <li data-md>
+             <p>If <var>dim</var> is equal to <var>axis</var>, add to <var>desc</var>.<code class="idl"><a data-link-type="idl" href="#dom-mloperanddescriptor-dimensions" id="ref-for-dom-mloperanddescriptor-dimensions②⑥">dimensions</a></code>[<var>axis</var>] the value of <var>input</var>.<code class="idl"><a data-link-type="idl" href="#dom-mloperanddescriptor-dimensions" id="ref-for-dom-mloperanddescriptor-dimensions②⑦">dimensions</a></code>[<var>dim</var>].</p>
+           </ol>
+         </ol>
+       </ol>
+      <li data-md>
+       <p>If any of the following sub-steps fail, <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#dfn-throw" id="ref-for-dfn-throw⑤⑧">throw</a> an "<code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#operationerror" id="ref-for-operationerror①⑥">OperationError</a></code>" <code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#idl-DOMException" id="ref-for-idl-DOMException④①">DOMException</a></code>.</p>
+       <ol>
+        <li data-md>
+         <p>Let <var>output</var> be the result of <a data-link-type="dfn" href="#create-an-mloperand" id="ref-for-create-an-mloperand④">creating an MLOperand</a> given <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#this" id="ref-for-this①③">this</a> and <var>desc</var>.</p>
+        <li data-md>
+         <p>Make a request to the underlying platform to:</p>
+         <ol>
+          <li data-md>
+           <p>Create an <a data-link-type="dfn" href="https://infra.spec.whatwg.org/#implementation-defined" id="ref-for-implementation-defined①②">implementation-defined</a> platform operator <var>concatImpl</var> for this method, given <var>inputs</var> and <var>axis</var>.</p>
+          <li data-md>
+           <p>Store a reference of <var>concatImpl</var> in <var>output</var>.<code class="idl"><a data-link-type="idl" href="#dom-mloperand-operator-slot" id="ref-for-dom-mloperand-operator-slot②">[[operator]]</a></code>.</p>
+          <li data-md>
+           <p>Create an <a data-link-type="dfn" href="https://infra.spec.whatwg.org/#implementation-defined" id="ref-for-implementation-defined①③">implementation-defined</a> platform operand <var>outputImpl</var> to represent output,given <var>output</var> and <var>concatImpl</var>.</p>
+          <li data-md>
+           <p>Store a reference to <var>outputImpl</var> in <var>output</var>.<code class="idl"><a data-link-type="idl" href="#dom-mloperand-operand-slot" id="ref-for-dom-mloperand-operand-slot⑦">[[operand]]</a></code>.</p>
+         </ol>
+        <li data-md>
+         <p>Connect <var>inputs</var> as input to <var>concatImpl</var>.</p>
+        <li data-md>
+         <p>Connect <var>output</var>.<code class="idl"><a data-link-type="idl" href="#dom-mloperand-operand-slot" id="ref-for-dom-mloperand-operand-slot⑧">[[operand]]</a></code> as output to <var>concatImpl</var>.</p>
+       </ol>
+      <li data-md>
+       <p>Return <var>output</var>.</p>
+     </ol>
     </div>
-    <h4 class="heading settled" data-level="7.6.7" id="api-mlgraphbuilder-conv2d"><span class="secno">7.6.7. </span><span class="content">The conv2d() method</span><a class="self-link" href="#api-mlgraphbuilder-conv2d"></a></h4>
-     Compute a 2-D convolution given 4-D input and filter tensors 
+   </details>
+   <h4 class="heading settled" data-level="7.6.8" id="api-mlgraphbuilder-conv2d"><span class="secno">7.6.8. </span><span class="content">The conv2d() method</span><a class="self-link" href="#api-mlgraphbuilder-conv2d"></a></h4>
+    Compute a 2-D convolution given 4-D input and filter tensors 
 <pre class="idl highlight def"><c- b>enum</c-> <dfn class="dfn-paneled idl-code" data-dfn-type="enum" data-export id="enumdef-mlconv2dfilteroperandlayout"><code><c- g>MLConv2dFilterOperandLayout</c-></code></dfn> {
   <dfn class="idl-code" data-dfn-for="MLConv2dFilterOperandLayout" data-dfn-type="enum-value" data-export id="dom-mlconv2dfilteroperandlayout-oihw"><code><c- s>"oihw"</c-></code><a class="self-link" href="#dom-mlconv2dfilteroperandlayout-oihw"></a></dfn>,
   <dfn class="idl-code" data-dfn-for="MLConv2dFilterOperandLayout" data-dfn-type="enum-value" data-export id="dom-mlconv2dfilteroperandlayout-hwio"><code><c- s>"hwio"</c-></code><a class="self-link" href="#dom-mlconv2dfilteroperandlayout-hwio"></a></dfn>,
@@ -2323,94 +3114,211 @@ <h4 class="heading settled" data-level="7.6.7" id="api-mlgraphbuilder-conv2d"><s
 };
 
 <c- b>dictionary</c-> <dfn class="dfn-paneled idl-code" data-dfn-type="dictionary" data-export id="dictdef-mlconv2doptions"><code><c- g>MLConv2dOptions</c-></code></dfn> {
-  <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#idl-sequence" id="ref-for-idl-sequence②"><c- b>sequence</c-></a>&lt;<a class="idl-code" data-link-type="interface" href="https://webidl.spec.whatwg.org/#idl-unsigned-long" id="ref-for-idl-unsigned-long⑤"><c- b>unsigned</c-> <c- b>long</c-></a>> <dfn class="idl-code" data-dfn-for="MLConv2dOptions" data-dfn-type="dict-member" data-export data-type="sequence<unsigned long> " id="dom-mlconv2doptions-padding"><code><c- g>padding</c-></code><a class="self-link" href="#dom-mlconv2doptions-padding"></a></dfn>;
-  <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#idl-sequence" id="ref-for-idl-sequence③"><c- b>sequence</c-></a>&lt;<a class="idl-code" data-link-type="interface" href="https://webidl.spec.whatwg.org/#idl-unsigned-long" id="ref-for-idl-unsigned-long⑥"><c- b>unsigned</c-> <c- b>long</c-></a>> <dfn class="idl-code" data-dfn-for="MLConv2dOptions" data-dfn-type="dict-member" data-export data-type="sequence<unsigned long> " id="dom-mlconv2doptions-strides"><code><c- g>strides</c-></code><a class="self-link" href="#dom-mlconv2doptions-strides"></a></dfn>;
-  <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#idl-sequence" id="ref-for-idl-sequence④"><c- b>sequence</c-></a>&lt;<a class="idl-code" data-link-type="interface" href="https://webidl.spec.whatwg.org/#idl-unsigned-long" id="ref-for-idl-unsigned-long⑦"><c- b>unsigned</c-> <c- b>long</c-></a>> <dfn class="idl-code" data-dfn-for="MLConv2dOptions" data-dfn-type="dict-member" data-export data-type="sequence<unsigned long> " id="dom-mlconv2doptions-dilations"><code><c- g>dilations</c-></code><a class="self-link" href="#dom-mlconv2doptions-dilations"></a></dfn>;
-  <a data-link-type="idl-name" href="#enumdef-mlautopad" id="ref-for-enumdef-mlautopad"><c- n>MLAutoPad</c-></a> <dfn class="idl-code" data-default="&quot;explicit&quot;" data-dfn-for="MLConv2dOptions" data-dfn-type="dict-member" data-export data-type="MLAutoPad " id="dom-mlconv2doptions-autopad"><code><c- g>autoPad</c-></code><a class="self-link" href="#dom-mlconv2doptions-autopad"></a></dfn> = "explicit";
-  <a class="idl-code" data-link-type="interface" href="https://webidl.spec.whatwg.org/#idl-unsigned-long" id="ref-for-idl-unsigned-long⑧"><c- b>unsigned</c-> <c- b>long</c-></a> <dfn class="idl-code" data-default="1" data-dfn-for="MLConv2dOptions" data-dfn-type="dict-member" data-export data-type="unsigned long " id="dom-mlconv2doptions-groups"><code><c- g>groups</c-></code><a class="self-link" href="#dom-mlconv2doptions-groups"></a></dfn> = 1;
-  <a data-link-type="idl-name" href="#enumdef-mlinputoperandlayout" id="ref-for-enumdef-mlinputoperandlayout"><c- n>MLInputOperandLayout</c-></a> <dfn class="idl-code" data-default="&quot;nchw&quot;" data-dfn-for="MLConv2dOptions" data-dfn-type="dict-member" data-export data-type="MLInputOperandLayout " id="dom-mlconv2doptions-inputlayout"><code><c- g>inputLayout</c-></code><a class="self-link" href="#dom-mlconv2doptions-inputlayout"></a></dfn> = "nchw";
-  <a data-link-type="idl-name" href="#enumdef-mlconv2dfilteroperandlayout" id="ref-for-enumdef-mlconv2dfilteroperandlayout"><c- n>MLConv2dFilterOperandLayout</c-></a> <dfn class="idl-code" data-default="&quot;oihw&quot;" data-dfn-for="MLConv2dOptions" data-dfn-type="dict-member" data-export data-type="MLConv2dFilterOperandLayout " id="dom-mlconv2doptions-filterlayout"><code><c- g>filterLayout</c-></code><a class="self-link" href="#dom-mlconv2doptions-filterlayout"></a></dfn> = "oihw";
-  <a data-link-type="idl-name" href="#mloperand" id="ref-for-mloperand⑤①"><c- n>MLOperand</c-></a> <dfn class="idl-code" data-dfn-for="MLConv2dOptions" data-dfn-type="dict-member" data-export data-type="MLOperand " id="dom-mlconv2doptions-bias"><code><c- g>bias</c-></code><a class="self-link" href="#dom-mlconv2doptions-bias"></a></dfn>;
-  <a data-link-type="idl-name" href="#mlactivation" id="ref-for-mlactivation⑥"><c- n>MLActivation</c-></a> <dfn class="idl-code" data-dfn-for="MLConv2dOptions" data-dfn-type="dict-member" data-export data-type="MLActivation " id="dom-mlconv2doptions-activation"><code><c- g>activation</c-></code><a class="self-link" href="#dom-mlconv2doptions-activation"></a></dfn>;
+  <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#idl-sequence" id="ref-for-idl-sequence②"><c- b>sequence</c-></a>&lt;<a class="idl-code" data-link-type="interface" href="https://webidl.spec.whatwg.org/#idl-unsigned-long" id="ref-for-idl-unsigned-long⑤"><c- b>unsigned</c-> <c- b>long</c-></a>> <a class="idl-code" data-link-type="dict-member" data-type="sequence<unsigned long> " href="#dom-mlconv2doptions-padding" id="ref-for-dom-mlconv2doptions-padding"><c- g>padding</c-></a>;
+  <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#idl-sequence" id="ref-for-idl-sequence③"><c- b>sequence</c-></a>&lt;<a class="idl-code" data-link-type="interface" href="https://webidl.spec.whatwg.org/#idl-unsigned-long" id="ref-for-idl-unsigned-long⑥"><c- b>unsigned</c-> <c- b>long</c-></a>> <a class="idl-code" data-link-type="dict-member" data-type="sequence<unsigned long> " href="#dom-mlconv2doptions-strides" id="ref-for-dom-mlconv2doptions-strides"><c- g>strides</c-></a>;
+  <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#idl-sequence" id="ref-for-idl-sequence④"><c- b>sequence</c-></a>&lt;<a class="idl-code" data-link-type="interface" href="https://webidl.spec.whatwg.org/#idl-unsigned-long" id="ref-for-idl-unsigned-long⑦"><c- b>unsigned</c-> <c- b>long</c-></a>> <a class="idl-code" data-link-type="dict-member" data-type="sequence<unsigned long> " href="#dom-mlconv2doptions-dilations" id="ref-for-dom-mlconv2doptions-dilations"><c- g>dilations</c-></a>;
+  <a data-link-type="idl-name" href="#enumdef-mlautopad" id="ref-for-enumdef-mlautopad"><c- n>MLAutoPad</c-></a> <a class="idl-code" data-default="&quot;explicit&quot;" data-link-type="dict-member" data-type="MLAutoPad " href="#dom-mlconv2doptions-autopad" id="ref-for-dom-mlconv2doptions-autopad"><c- g>autoPad</c-></a> = "explicit";
+  <a class="idl-code" data-link-type="interface" href="https://webidl.spec.whatwg.org/#idl-unsigned-long" id="ref-for-idl-unsigned-long⑧"><c- b>unsigned</c-> <c- b>long</c-></a> <a class="idl-code" data-default="1" data-link-type="dict-member" data-type="unsigned long " href="#dom-mlconv2doptions-groups" id="ref-for-dom-mlconv2doptions-groups"><c- g>groups</c-></a> = 1;
+  <a data-link-type="idl-name" href="#enumdef-mlinputoperandlayout" id="ref-for-enumdef-mlinputoperandlayout"><c- n>MLInputOperandLayout</c-></a> <a class="idl-code" data-default="&quot;nchw&quot;" data-link-type="dict-member" data-type="MLInputOperandLayout " href="#dom-mlconv2doptions-inputlayout" id="ref-for-dom-mlconv2doptions-inputlayout"><c- g>inputLayout</c-></a> = "nchw";
+  <a data-link-type="idl-name" href="#enumdef-mlconv2dfilteroperandlayout" id="ref-for-enumdef-mlconv2dfilteroperandlayout"><c- n>MLConv2dFilterOperandLayout</c-></a> <a class="idl-code" data-default="&quot;oihw&quot;" data-link-type="dict-member" data-type="MLConv2dFilterOperandLayout " href="#dom-mlconv2doptions-filterlayout" id="ref-for-dom-mlconv2doptions-filterlayout"><c- g>filterLayout</c-></a> = "oihw";
+  <a data-link-type="idl-name" href="#mloperand" id="ref-for-mloperand⑤⑨"><c- n>MLOperand</c-></a> <a class="idl-code" data-link-type="dict-member" data-type="MLOperand " href="#dom-mlconv2doptions-bias" id="ref-for-dom-mlconv2doptions-bias"><c- g>bias</c-></a>;
+  <a data-link-type="idl-name" href="#mlactivation" id="ref-for-mlactivation①③"><c- n>MLActivation</c-></a> <a class="idl-code" data-link-type="dict-member" data-type="MLActivation " href="#dom-mlconv2doptions-activation" id="ref-for-dom-mlconv2doptions-activation"><c- g>activation</c-></a>;
 };
 
-<c- b>partial</c-> <c- b>interface</c-> <a class="idl-code" data-link-type="interface" href="#mlgraphbuilder" id="ref-for-mlgraphbuilder②③"><c- g>MLGraphBuilder</c-></a> {
-  <a data-link-type="idl-name" href="#mloperand" id="ref-for-mloperand⑤②"><c- n>MLOperand</c-></a> <dfn class="idl-code" data-dfn-for="MLGraphBuilder" data-dfn-type="method" data-export data-lt="conv2d(input, filter, options)|conv2d(input, filter)" id="dom-mlgraphbuilder-conv2d"><code><c- g>conv2d</c-></code><a class="self-link" href="#dom-mlgraphbuilder-conv2d"></a></dfn>(<a data-link-type="idl-name" href="#mloperand" id="ref-for-mloperand⑤③"><c- n>MLOperand</c-></a> <dfn class="idl-code" data-dfn-for="MLGraphBuilder/conv2d(input, filter, options), MLGraphBuilder/conv2d(input, filter)" data-dfn-type="argument" data-export id="dom-mlgraphbuilder-conv2d-input-filter-options-input"><code><c- g>input</c-></code><a class="self-link" href="#dom-mlgraphbuilder-conv2d-input-filter-options-input"></a></dfn>, <a data-link-type="idl-name" href="#mloperand" id="ref-for-mloperand⑤④"><c- n>MLOperand</c-></a> <dfn class="idl-code" data-dfn-for="MLGraphBuilder/conv2d(input, filter, options), MLGraphBuilder/conv2d(input, filter)" data-dfn-type="argument" data-export id="dom-mlgraphbuilder-conv2d-input-filter-options-filter"><code><c- g>filter</c-></code><a class="self-link" href="#dom-mlgraphbuilder-conv2d-input-filter-options-filter"></a></dfn>, <c- b>optional</c-> <a data-link-type="idl-name" href="#dictdef-mlconv2doptions" id="ref-for-dictdef-mlconv2doptions"><c- n>MLConv2dOptions</c-></a> <dfn class="idl-code" data-dfn-for="MLGraphBuilder/conv2d(input, filter, options), MLGraphBuilder/conv2d(input, filter)" data-dfn-type="argument" data-export id="dom-mlgraphbuilder-conv2d-input-filter-options-options"><code><c- g>options</c-></code><a class="self-link" href="#dom-mlgraphbuilder-conv2d-input-filter-options-options"></a></dfn> = {});
+<c- b>partial</c-> <c- b>interface</c-> <a class="idl-code" data-link-type="interface" href="#mlgraphbuilder" id="ref-for-mlgraphbuilder②⑥"><c- g>MLGraphBuilder</c-></a> {
+  <a data-link-type="idl-name" href="#mloperand" id="ref-for-mloperand⑥⓪"><c- n>MLOperand</c-></a> <a class="idl-code" data-link-type="method" href="#dom-mlgraphbuilder-conv2d" id="ref-for-dom-mlgraphbuilder-conv2d"><c- g>conv2d</c-></a>(<a data-link-type="idl-name" href="#mloperand" id="ref-for-mloperand⑥①"><c- n>MLOperand</c-></a> <dfn class="idl-code" data-dfn-for="MLGraphBuilder/conv2d(input, filter, options), MLGraphBuilder/conv2d(input, filter)" data-dfn-type="argument" data-export id="dom-mlgraphbuilder-conv2d-input-filter-options-input"><code><c- g>input</c-></code><a class="self-link" href="#dom-mlgraphbuilder-conv2d-input-filter-options-input"></a></dfn>, <a data-link-type="idl-name" href="#mloperand" id="ref-for-mloperand⑥②"><c- n>MLOperand</c-></a> <dfn class="idl-code" data-dfn-for="MLGraphBuilder/conv2d(input, filter, options), MLGraphBuilder/conv2d(input, filter)" data-dfn-type="argument" data-export id="dom-mlgraphbuilder-conv2d-input-filter-options-filter"><code><c- g>filter</c-></code><a class="self-link" href="#dom-mlgraphbuilder-conv2d-input-filter-options-filter"></a></dfn>, <c- b>optional</c-> <a data-link-type="idl-name" href="#dictdef-mlconv2doptions" id="ref-for-dictdef-mlconv2doptions"><c- n>MLConv2dOptions</c-></a> <dfn class="idl-code" data-dfn-for="MLGraphBuilder/conv2d(input, filter, options), MLGraphBuilder/conv2d(input, filter)" data-dfn-type="argument" data-export id="dom-mlgraphbuilder-conv2d-input-filter-options-options"><code><c- g>options</c-></code><a class="self-link" href="#dom-mlgraphbuilder-conv2d-input-filter-options-options"></a></dfn> = {});
 };
 </pre>
-    <div class="algorithm" data-algorithm="conv2d">
-      <strong>Arguments:</strong> 
+   <p><code class="idl"><a data-link-type="idl" href="#dictdef-mlconv2doptions" id="ref-for-dictdef-mlconv2doptions①">MLConv2dOptions</a></code> has the following members:</p>
+   <dl>
+    <dt data-md><dfn class="dfn-paneled idl-code" data-dfn-for="MLConv2dOptions" data-dfn-type="dict-member" data-export id="dom-mlconv2doptions-padding"><code>padding</code></dfn>, <span> of type <code class="idl-code">sequence&lt;unsigned long></code></span>
+    <dd data-md>
+     <p>A sequence of <code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#idl-unsigned-long" id="ref-for-idl-unsigned-long⑨">unsigned long</a></code> of length 4: [beginningHeight, endingHeight, beginningWidth, endingWidth].
+Specifies the additional rows and columns added to the beginning and ending of each spatial dimension of the convolution input.
+The default value is [0, 0, 0, 0].</p>
+    <dt data-md><dfn class="dfn-paneled idl-code" data-dfn-for="MLConv2dOptions" data-dfn-type="dict-member" data-export id="dom-mlconv2doptions-strides"><code>strides</code></dfn>, <span> of type <code class="idl-code">sequence&lt;unsigned long></code></span>
+    <dd data-md>
+     <p>A sequence of <code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#idl-unsigned-long" id="ref-for-idl-unsigned-long①⓪">unsigned long</a></code> of length 2: [strideHeight, strideWidth].
+Specifies the stride of the sliding window for each spatial dimension of the convolution input.
+The default value is [1, 1].</p>
+    <dt data-md><dfn class="dfn-paneled idl-code" data-dfn-for="MLConv2dOptions" data-dfn-type="dict-member" data-export id="dom-mlconv2doptions-dilations"><code>dilations</code></dfn>, <span> of type <code class="idl-code">sequence&lt;unsigned long></code></span>
+    <dd data-md>
+     <p>A sequence of <code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#idl-unsigned-long" id="ref-for-idl-unsigned-long①①">unsigned long</a></code> of length 2: [dilationHeight, dilationWidth]. Specifies the dilation factor for each spatial dimension applied on the convolution filter (kernel).
+The default value is [1, 1].</p>
+    <dt data-md><dfn class="dfn-paneled idl-code" data-dfn-for="MLConv2dOptions" data-dfn-type="dict-member" data-export id="dom-mlconv2doptions-autopad"><code>autoPad</code></dfn>, <span> of type <a data-link-type="idl-name" href="#enumdef-mlautopad" id="ref-for-enumdef-mlautopad①">MLAutoPad</a>, defaulting to <code>"explicit"</code></span>
+    <dd data-md>
+     <p>An <code class="idl"><a data-link-type="idl" href="#enumdef-mlautopad" id="ref-for-enumdef-mlautopad②">MLAutoPad</a></code> <a data-link-type="dfn" href="https://infra.spec.whatwg.org/#string" id="ref-for-string④">string</a>.
+Specifies the automatic input padding options.
+The default value is <em>"explicit"</em>, which means that the values in the <code class="idl"><a data-link-type="idl" href="#dom-mlconv2doptions-padding" id="ref-for-dom-mlconv2doptions-padding①">padding</a></code> array should be used for input padding.
+When the option is set other than <em>"explicit"</em>, the values in the <code class="idl"><a data-link-type="idl" href="#dom-mlconv2doptions-padding" id="ref-for-dom-mlconv2doptions-padding②">padding</a></code> array are ignored.</p>
+     <p>With the <em>"same-upper"</em> option, the padding values are automatically computed such that the additional ending padding of the spatial input dimensions would allow all of the input values in the corresponding dimension to be filtered.</p>
+     <p>The <em>"same-lower"</em> option is similar but padding is applied to the beginning padding of the spatial input dimensions instead of the ending one.</p>
+    <dt data-md><dfn class="dfn-paneled idl-code" data-dfn-for="MLConv2dOptions" data-dfn-type="dict-member" data-export id="dom-mlconv2doptions-groups"><code>groups</code></dfn>, <span> of type <a data-link-type="idl-name" href="https://webidl.spec.whatwg.org/#idl-unsigned-long" id="ref-for-idl-unsigned-long①②">unsigned long</a>, defaulting to <code>1</code></span>
+    <dd data-md>
+     <p>An <code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#idl-unsigned-long" id="ref-for-idl-unsigned-long①③">unsigned long</a></code> scalar.
+Specifies the number of groups that input channels and output channels are divided into.
+The default value is 1.</p>
+    <dt data-md><dfn class="dfn-paneled idl-code" data-dfn-for="MLConv2dOptions" data-dfn-type="dict-member" data-export id="dom-mlconv2doptions-inputlayout"><code>inputLayout</code></dfn>, <span> of type <a data-link-type="idl-name" href="#enumdef-mlinputoperandlayout" id="ref-for-enumdef-mlinputoperandlayout①">MLInputOperandLayout</a>, defaulting to <code>"nchw"</code></span>
+    <dd data-md>
+     <p>An <code class="idl"><a data-link-type="idl" href="#enumdef-mlinputoperandlayout" id="ref-for-enumdef-mlinputoperandlayout②">MLInputOperandLayout</a></code> <a data-link-type="dfn" href="https://infra.spec.whatwg.org/#string" id="ref-for-string⑤">string</a>.
+Specifies the layout format of the input and output tensor as follows:</p>
      <ul>
       <li data-md>
-       <p><em>input</em>: an <code class="idl"><a data-link-type="idl" href="#mloperand" id="ref-for-mloperand⑤⑤">MLOperand</a></code>. The input 4-D tensor. The logical shape
-is interpreted according to the value of <em>options.inputLayout</em>.</p>
-      <li data-md>
-       <p><em>filter</em>: an <code class="idl"><a data-link-type="idl" href="#mloperand" id="ref-for-mloperand⑤⑥">MLOperand</a></code>. The filter 4-D tensor. The logical shape is
-interpreted according to the value of <em>options.filterLayout</em> and <em>options.groups</em>.</p>
-      <li data-md>
-       <p><em>options</em>: an optional <code class="idl"><a data-link-type="idl" href="#dictdef-mlconv2doptions" id="ref-for-dictdef-mlconv2doptions①">MLConv2dOptions</a></code>. The optional parameters of the operation.</p>
+       <p><strong>"nchw"</strong></p>
        <ul>
         <li data-md>
-         <p><em>padding</em>: a sequence of <code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#idl-unsigned-long" id="ref-for-idl-unsigned-long⑨">unsigned long</a></code> of length 4. The additional rows and columns added to the beginning and ending of each spatial dimension of <em>input</em>, [beginning_height, ending_height, beginning_width, ending_width]. If not present, the values are assumed to be [0,0,0,0].</p>
+         <p>input tensor: <em>[batches, inputChannels, height, width]</em></p>
+        <li data-md>
+         <p>output tensor: <em>[batches, outputChannels, height, width]</em></p>
+       </ul>
+      <li data-md>
+       <p><strong>"nhwc"</strong>:</p>
+       <ul>
         <li data-md>
-         <p><em>strides</em>: a sequence of <code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#idl-unsigned-long" id="ref-for-idl-unsigned-long①⓪">unsigned long</a></code> of length 2. The stride of the sliding window for each spatial dimension of <em>input</em>, [stride_height, stride_width]. If not present, the values are assumed to be [1,1].</p>
+         <p>input tensor: <em>[batches, height, width, inputChannels]</em></p>
         <li data-md>
-         <p><em>dilations</em>: a sequence of <code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#idl-unsigned-long" id="ref-for-idl-unsigned-long①①">unsigned long</a></code> of length 2. The dilation factor for each spatial dimension of <em>input</em>, [dilation_height, dilation_width]. If not present, the values are assumed to be [1,1].</p>
+         <p>output tensor: <em>[batches, height, width, outputChannels]</em></p>
+       </ul>
+     </ul>
+     The default value is <em>"nchw"</em>. 
+    <dt data-md><dfn class="dfn-paneled idl-code" data-dfn-for="MLConv2dOptions" data-dfn-type="dict-member" data-export id="dom-mlconv2doptions-filterlayout"><code>filterLayout</code></dfn>, <span> of type <a data-link-type="idl-name" href="#enumdef-mlconv2dfilteroperandlayout" id="ref-for-enumdef-mlconv2dfilteroperandlayout①">MLConv2dFilterOperandLayout</a>, defaulting to <code>"oihw"</code></span>
+    <dd data-md>
+     <p>An <code class="idl"><a data-link-type="idl" href="#enumdef-mlconv2dfilteroperandlayout" id="ref-for-enumdef-mlconv2dfilteroperandlayout②">MLConv2dFilterOperandLayout</a></code> <a data-link-type="dfn" href="https://infra.spec.whatwg.org/#string" id="ref-for-string⑥">string</a>.
+  Specifies the layout format of the filter tensor as follow:</p>
+     <ul>
+      <li data-md>
+       <p><strong>"oihw"</strong>: <em>[outputChannels, inputChannels/groups, height, width]</em></p>
+      <li data-md>
+       <p><strong>"hwio"</strong>: <em>[height, width, inputChannels/groups, outputChannels]</em></p>
+      <li data-md>
+       <p><strong>"ohwi"</strong>: <em>[outputChannels, height, width, inputChannels/groups]</em></p>
+      <li data-md>
+       <p><strong>"ihwo"</strong>: <em>[inputChannels/groups, height, width, outputChannels]</em></p>
+     </ul>
+      The default value is <em>"oihw"</em>. 
+    <dt data-md><dfn class="dfn-paneled idl-code" data-dfn-for="MLConv2dOptions" data-dfn-type="dict-member" data-export id="dom-mlconv2doptions-bias"><code>bias</code></dfn>, <span> of type <a data-link-type="idl-name" href="#mloperand" id="ref-for-mloperand⑥③">MLOperand</a></span>
+    <dd data-md>
+     <p>An <code class="idl"><a data-link-type="idl" href="#mloperand" id="ref-for-mloperand⑥④">MLOperand</a></code> object.
+  Specifies the additional 1-D tensor with the shape of <em>[outputChannels]</em> whose values are to be added to the convolution result.</p>
+    <dt data-md><dfn class="dfn-paneled idl-code" data-dfn-for="MLConv2dOptions" data-dfn-type="dict-member" data-export id="dom-mlconv2doptions-activation"><code>activation</code></dfn>, <span> of type <a data-link-type="idl-name" href="#mlactivation" id="ref-for-mlactivation①④">MLActivation</a></span>
+    <dd data-md>
+     <p>An <code class="idl"><a data-link-type="idl" href="#mlactivation" id="ref-for-mlactivation①⑤">MLActivation</a></code> object.
+    Specifies the optional activation function that immediately follows the convolution operation.</p>
+   </dl>
+   <div>
+     <strong>Arguments:</strong> 
+    <ul>
+     <li data-md>
+      <p><em>input</em>: an <code class="idl"><a data-link-type="idl" href="#mloperand" id="ref-for-mloperand⑥⑤">MLOperand</a></code>. The input 4-D tensor. The logical shape
+is interpreted according to the value of <em>options</em>.<code class="idl"><a data-link-type="idl" href="#dom-mlconv2doptions-inputlayout" id="ref-for-dom-mlconv2doptions-inputlayout①">inputLayout</a></code>.</p>
+     <li data-md>
+      <p><em>filter</em>: an <code class="idl"><a data-link-type="idl" href="#mloperand" id="ref-for-mloperand⑥⑥">MLOperand</a></code>. The filter 4-D tensor. The logical shape is
+interpreted according to the value of <em>options</em>.<code class="idl"><a data-link-type="idl" href="#dom-mlconv2doptions-filterlayout" id="ref-for-dom-mlconv2doptions-filterlayout①">filterLayout</a></code> and <em>options</em>.<code class="idl"><a data-link-type="idl" href="#dom-mlconv2doptions-groups" id="ref-for-dom-mlconv2doptions-groups①">groups</a></code>.</p>
+     <li data-md>
+      <p><em>options</em>: an <code class="idl"><a data-link-type="idl" href="#dictdef-mlconv2doptions" id="ref-for-dictdef-mlconv2doptions②">MLConv2dOptions</a></code>. The optional parameters of the operation.</p>
+    </ul>
+    <p><strong>Returns:</strong> an <code class="idl"><a data-link-type="idl" href="#mloperand" id="ref-for-mloperand⑥⑦">MLOperand</a></code>. The output 4-D tensor that contains the convolution result. The output shape is interpreted according to the <em>options</em>.<code class="idl"><a data-link-type="idl" href="#dom-mlconv2doptions-inputlayout" id="ref-for-dom-mlconv2doptions-inputlayout②">inputLayout</a></code> value. More specifically, the spatial dimensions or the sizes of the last two dimensions of the output tensor for the <em>nchw</em> input layout can be calculated as follow:</p>
+    <p><em>outputSize = 1 + (inputSize - (filterSize - 1) *</em> <em>dilation - 1 + beginningPadding + endingPadding) / stride</em></p>
+   </div>
+   <div class="note" role="note"> A <em>depthwise</em> conv2d operation is a variant of grouped convolution, used in models like the MobileNet, where the <em>options.groups</em> = inputChannels = outputChannels and the shape of filter tensor is [options.groups, 1, height, width]
+    for <em>"oihw"</em> layout, [height, width, 1, options.groups] for <em>"hwio"</em> layout, [options.groups, height, width, 1] for <em>"ohwi"</em> layout and [1, height, width, options.groups] for <em>"ihwo"</em> layout. </div>
+   <details class="algorithm" data-algorithm="conv2d(input, filter, options)" data-algorithm-for="MLGraphBuilder" open>
+    <summary> The <dfn class="dfn-paneled idl-code" data-dfn-for="MLGraphBuilder" data-dfn-type="method" data-export data-lt="conv2d(input, filter, options)|conv2d(input, filter)" id="dom-mlgraphbuilder-conv2d"><code>conv2d(<var>input</var>, <var>filter</var>, <var>options</var>)</code></dfn> method steps are: </summary>
+    <div class="algorithm-steps">
+     <ol>
+      <li data-md>
+       <p><a data-link-type="dfn" href="https://infra.spec.whatwg.org/#assert" id="ref-for-assert②①">Assert</a>: the type of <var>input</var> and <var>filter</var> is <code class="idl"><a data-link-type="idl" href="#mloperand" id="ref-for-mloperand⑥⑧">MLOperand</a></code>.</p>
+      <li data-md>
+       <p>Let <var>inputSize</var> be the <a data-link-type="dfn" href="https://infra.spec.whatwg.org/#list-size" id="ref-for-list-size①⓪">size</a> of <var>input</var>.<code class="idl"><a data-link-type="idl" href="#dom-mloperand-descriptor-slot" id="ref-for-dom-mloperand-descriptor-slot②①">[[descriptor]]</a></code>.<code class="idl"><a data-link-type="idl" href="#dom-mloperanddescriptor-dimensions" id="ref-for-dom-mloperanddescriptor-dimensions②⑧">dimensions</a></code>.</p>
+      <li data-md>
+       <p>Let <var>filterSize</var> be the <a data-link-type="dfn" href="https://infra.spec.whatwg.org/#list-size" id="ref-for-list-size①①">size</a> of <var>filter</var>.<code class="idl"><a data-link-type="idl" href="#dom-mloperand-descriptor-slot" id="ref-for-dom-mloperand-descriptor-slot②②">[[descriptor]]</a></code>.<code class="idl"><a data-link-type="idl" href="#dom-mloperanddescriptor-dimensions" id="ref-for-dom-mloperanddescriptor-dimensions②⑨">dimensions</a></code>.</p>
+      <li data-md>
+       <p>If <var>inputSize</var> is not 4, then <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#dfn-throw" id="ref-for-dfn-throw⑤⑨">throw</a> a "<code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#dataerror" id="ref-for-dataerror①⑥">DataError</a></code>" <code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#idl-DOMException" id="ref-for-idl-DOMException④②">DOMException</a></code>.</p>
+      <li data-md>
+       <p>If <var>filterSize</var> is not 4, then <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#dfn-throw" id="ref-for-dfn-throw⑥⓪">throw</a> a "<code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#dataerror" id="ref-for-dataerror①⑦">DataError</a></code>" <code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#idl-DOMException" id="ref-for-idl-DOMException④③">DOMException</a></code>.</p>
+      <li data-md>
+       <p>If <var>input</var>.<code class="idl"><a data-link-type="idl" href="#dom-mloperand-descriptor-slot" id="ref-for-dom-mloperand-descriptor-slot②③">[[descriptor]]</a></code>.<code class="idl"><a data-link-type="idl" href="#dom-mloperanddescriptor-type" id="ref-for-dom-mloperanddescriptor-type⑨">type</a></code> is not the same as <var>filter</var>.<code class="idl"><a data-link-type="idl" href="#dom-mloperand-descriptor-slot" id="ref-for-dom-mloperand-descriptor-slot②④">[[descriptor]]</a></code>.<code class="idl"><a data-link-type="idl" href="#dom-mloperanddescriptor-type" id="ref-for-dom-mloperanddescriptor-type①⓪">type</a></code>, then <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#dfn-throw" id="ref-for-dfn-throw⑥①">throw</a> a <code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#exceptiondef-typeerror" id="ref-for-exceptiondef-typeerror①⑨">TypeError</a></code>.</p>
+      <li data-md>
+       <p>If <var>options</var>.<code class="idl"><a data-link-type="idl" href="#dom-mlconv2doptions-padding" id="ref-for-dom-mlconv2doptions-padding③">padding</a></code> does not <a data-link-type="dfn" href="https://infra.spec.whatwg.org/#map-exists" id="ref-for-map-exists①②">exist</a>, set it to <code>« 0, 0, 0, 0 »</code>.</p>
+      <li data-md>
+       <p>Else if the <a data-link-type="dfn" href="https://infra.spec.whatwg.org/#list-size" id="ref-for-list-size①②">size</a> of <var>options</var>.<code class="idl"><a data-link-type="idl" href="#dom-mlconv2doptions-padding" id="ref-for-dom-mlconv2doptions-padding④">padding</a></code> is not 4, then <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#dfn-throw" id="ref-for-dfn-throw⑥②">throw</a> a "<code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#dataerror" id="ref-for-dataerror①⑧">DataError</a></code>" <code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#idl-DOMException" id="ref-for-idl-DOMException④④">DOMException</a></code>.</p>
+      <li data-md>
+       <p>If <var>options</var>.<code class="idl"><a data-link-type="idl" href="#dom-mlconv2doptions-strides" id="ref-for-dom-mlconv2doptions-strides①">strides</a></code> does not <a data-link-type="dfn" href="https://infra.spec.whatwg.org/#map-exists" id="ref-for-map-exists①③">exist</a>, set it to <code>« 1, 1 »</code>.</p>
+      <li data-md>
+       <p>Else if the <a data-link-type="dfn" href="https://infra.spec.whatwg.org/#list-size" id="ref-for-list-size①③">size</a> of <var>options</var>.<code class="idl"><a data-link-type="idl" href="#dom-mlconv2doptions-strides" id="ref-for-dom-mlconv2doptions-strides②">strides</a></code> is not 2, then <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#dfn-throw" id="ref-for-dfn-throw⑥③">throw</a> a <code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#exceptiondef-typeerror" id="ref-for-exceptiondef-typeerror②⓪">TypeError</a></code>.</p>
+      <li data-md>
+       <p>If any element in <var>options</var>.<code class="idl"><a data-link-type="idl" href="#dom-mlconv2doptions-strides" id="ref-for-dom-mlconv2doptions-strides③">strides</a></code> is equal to 0, then <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#dfn-throw" id="ref-for-dfn-throw⑥④">throw</a> a <code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#exceptiondef-typeerror" id="ref-for-exceptiondef-typeerror②①">TypeError</a></code>.</p>
+      <li data-md>
+       <p>If <var>options</var>.<code class="idl"><a data-link-type="idl" href="#dom-mlconv2doptions-dilations" id="ref-for-dom-mlconv2doptions-dilations①">dilations</a></code> does not <a data-link-type="dfn" href="https://infra.spec.whatwg.org/#map-exists" id="ref-for-map-exists①④">exist</a>, set it to <code>« 1, 1 »</code>.</p>
+      <li data-md>
+       <p>Else if the <a data-link-type="dfn" href="https://infra.spec.whatwg.org/#list-size" id="ref-for-list-size①④">size</a> of <var>options</var>.<code class="idl"><a data-link-type="idl" href="#dom-mlconv2doptions-dilations" id="ref-for-dom-mlconv2doptions-dilations②">dilations</a></code> is not 2, then <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#dfn-throw" id="ref-for-dfn-throw⑥⑤">throw</a> a <code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#exceptiondef-typeerror" id="ref-for-exceptiondef-typeerror②②">TypeError</a></code>.</p>
+      <li data-md>
+       <p>If <var>options</var>.<code class="idl"><a data-link-type="idl" href="#dom-mlconv2doptions-autopad" id="ref-for-dom-mlconv2doptions-autopad①">autoPad</a></code> does not <a data-link-type="dfn" href="https://infra.spec.whatwg.org/#map-exists" id="ref-for-map-exists①⑤">exist</a>, set it to <code>"explicit"</code>.</p>
+      <li data-md>
+       <p>If <var>options</var>.<code class="idl"><a data-link-type="idl" href="#dom-mlconv2doptions-groups" id="ref-for-dom-mlconv2doptions-groups②">groups</a></code> is 0, then <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#dfn-throw" id="ref-for-dfn-throw⑥⑥">throw</a> a "<code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#dataerror" id="ref-for-dataerror①⑨">DataError</a></code>" <code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#idl-DOMException" id="ref-for-idl-DOMException④⑤">DOMException</a></code>.</p>
+      <li data-md>
+       <p>If <var>inputSize</var> / <var>options</var>.<code class="idl"><a data-link-type="idl" href="#dom-mlconv2doptions-groups" id="ref-for-dom-mlconv2doptions-groups③">groups</a></code> is not equal to <var>filterSize</var>, then <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#dfn-throw" id="ref-for-dfn-throw⑥⑦">throw</a> a "<code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#dataerror" id="ref-for-dataerror②⓪">DataError</a></code>" <code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#idl-DOMException" id="ref-for-idl-DOMException④⑥">DOMException</a></code>.</p>
+      <li data-md>
+       <p>Else if <var>inputSize</var> % <var>options</var>.<code class="idl"><a data-link-type="idl" href="#dom-mlconv2doptions-groups" id="ref-for-dom-mlconv2doptions-groups④">groups</a></code> is not 0, then <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#dfn-throw" id="ref-for-dfn-throw⑥⑧">throw</a> a "<code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#dataerror" id="ref-for-dataerror②①">DataError</a></code>" <code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#idl-DOMException" id="ref-for-idl-DOMException④⑦">DOMException</a></code>.</p>
+      <li data-md>
+       <p>If <var>options</var>.<code class="idl"><a data-link-type="idl" href="#dom-mlconv2doptions-bias" id="ref-for-dom-mlconv2doptions-bias①">bias</a></code> <a data-link-type="dfn" href="https://infra.spec.whatwg.org/#map-exists" id="ref-for-map-exists①⑥">exists</a>:</p>
+       <ol>
         <li data-md>
-         <p><em>autoPad</em>: an <code class="idl"><a data-link-type="idl" href="#enumdef-mlautopad" id="ref-for-enumdef-mlautopad①">MLAutoPad</a></code>. The automatic input padding options. By default, this argument is set to <em>"explicit"</em>, which means that the values in the <em>options.padding</em> array should be used for input padding. When the option is set other than <em>"explicit"</em>, the values in the <em>options.padding</em> array are ignored. With the <em>"same-upper"</em> option, the padding values are automatically computed such that the additional ending padding of the spatial input dimensions would allow all of the input values in the corresponding dimension to be filtered. The <em>"same-lower"</em> option is similar but padding is applied to the beginning padding of the spatial input dimensions instead of the ending one.</p>
+         <p><a data-link-type="dfn" href="https://infra.spec.whatwg.org/#assert" id="ref-for-assert②②">Assert</a>: the type of <var>options</var>.<code class="idl"><a data-link-type="idl" href="#dom-mlconv2doptions-bias" id="ref-for-dom-mlconv2doptions-bias②">bias</a></code> is <code class="idl"><a data-link-type="idl" href="#mloperand" id="ref-for-mloperand⑥⑨">MLOperand</a></code>.</p>
+       </ol>
+      <li data-md>
+       <p>If the <a data-link-type="dfn" href="https://infra.spec.whatwg.org/#list-size" id="ref-for-list-size①⑤">size</a> of <var>options</var>.<code class="idl"><a data-link-type="idl" href="#dom-mlconv2doptions-bias" id="ref-for-dom-mlconv2doptions-bias③">bias</a></code>.<code class="idl"><a data-link-type="idl" href="#dom-mloperand-descriptor-slot" id="ref-for-dom-mloperand-descriptor-slot②⑤">[[descriptor]]</a></code>.<code class="idl"><a data-link-type="idl" href="#dom-mloperanddescriptor-dimensions" id="ref-for-dom-mloperanddescriptor-dimensions③⓪">dimensions</a></code> is not 1, then <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#dfn-throw" id="ref-for-dfn-throw⑥⑨">throw</a> a <code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#exceptiondef-typeerror" id="ref-for-exceptiondef-typeerror②③">TypeError</a></code>.</p>
+      <li data-md>
+       <p>If <var>options</var>.<code class="idl"><a data-link-type="idl" href="#dom-mlconv2doptions-bias" id="ref-for-dom-mlconv2doptions-bias④">bias</a></code>.<code class="idl"><a data-link-type="idl" href="#dom-mloperand-descriptor-slot" id="ref-for-dom-mloperand-descriptor-slot②⑥">[[descriptor]]</a></code>.<code class="idl"><a data-link-type="idl" href="#dom-mloperanddescriptor-type" id="ref-for-dom-mloperanddescriptor-type①①">type</a></code> is not the same as <var>input</var>.<code class="idl"><a data-link-type="idl" href="#dom-mloperand-descriptor-slot" id="ref-for-dom-mloperand-descriptor-slot②⑦">[[descriptor]]</a></code>.<code class="idl"><a data-link-type="idl" href="#dom-mloperanddescriptor-type" id="ref-for-dom-mloperanddescriptor-type①②">type</a></code>, then <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#dfn-throw" id="ref-for-dfn-throw⑦⓪">throw</a> a <code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#exceptiondef-typeerror" id="ref-for-exceptiondef-typeerror②④">TypeError</a></code>.</p>
+      <li data-md>
+       <p>If <var>options</var>.<code class="idl"><a data-link-type="idl" href="#dom-mlconv2doptions-activation" id="ref-for-dom-mlconv2doptions-activation①">activation</a></code> <a data-link-type="dfn" href="https://infra.spec.whatwg.org/#map-exists" id="ref-for-map-exists①⑦">exists</a>:</p>
+       <ol>
         <li data-md>
-         <p><em>groups</em>: an <code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#idl-unsigned-long" id="ref-for-idl-unsigned-long①②">unsigned long</a></code> scalar. The number of groups that input channels and output channels are divided into, default to 1.</p>
+         <p><a data-link-type="dfn" href="https://infra.spec.whatwg.org/#assert" id="ref-for-assert②③">Assert</a>: the type of <var>options</var>.<code class="idl"><a data-link-type="idl" href="#dom-mlconv2doptions-activation" id="ref-for-dom-mlconv2doptions-activation②">activation</a></code> is <code class="idl"><a data-link-type="idl" href="#mlactivation" id="ref-for-mlactivation①⑥">MLActivation</a></code>.</p>
+       </ol>
+      <li data-md>
+       <p>Let <var>outputShape</var> be the result of invoking the underlying implementation for calculating output dimensions, given <var>options</var>.</p>
+      <li data-md>
+       <p>If <var>outputShape</var> is not the same as the shape of <var>options</var>.<code class="idl"><a data-link-type="idl" href="#dom-mlconv2doptions-bias" id="ref-for-dom-mlconv2doptions-bias⑤">bias</a></code>.<code class="idl"><a data-link-type="idl" href="#dom-mloperand-descriptor-slot" id="ref-for-dom-mloperand-descriptor-slot②⑧">[[descriptor]]</a></code>.<code class="idl"><a data-link-type="idl" href="#dom-mloperanddescriptor-dimensions" id="ref-for-dom-mloperanddescriptor-dimensions③①">dimensions</a></code>, then <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#dfn-throw" id="ref-for-dfn-throw⑦①">throw</a> a "<code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#dataerror" id="ref-for-dataerror②②">DataError</a></code>" <code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#idl-DOMException" id="ref-for-idl-DOMException④⑧">DOMException</a></code>.</p>
+      <li data-md>
+       <p>Let <var>desc</var> a new <code class="idl"><a data-link-type="idl" href="#dictdef-mloperanddescriptor" id="ref-for-dictdef-mloperanddescriptor①④">MLOperandDescriptor</a></code>.</p>
+      <li data-md>
+       <p>Set <var>desc</var>.<code class="idl"><a data-link-type="idl" href="#dom-mloperanddescriptor-type" id="ref-for-dom-mloperanddescriptor-type①③">type</a></code> to <var>input</var>.<code class="idl"><a data-link-type="idl" href="#dom-mloperand-descriptor-slot" id="ref-for-dom-mloperand-descriptor-slot②⑨">[[descriptor]]</a></code>.<code class="idl"><a data-link-type="idl" href="#dom-mloperanddescriptor-type" id="ref-for-dom-mloperanddescriptor-type①④">type</a></code>.</p>
+      <li data-md>
+       <p>Set <var>desc</var>.<code class="idl"><a data-link-type="idl" href="#dom-mloperanddescriptor-dimensions" id="ref-for-dom-mloperanddescriptor-dimensions③②">dimensions</a></code> to <var>outputShape</var>.</p>
+      <li data-md>
+       <p>If any of the following sub-steps fail, <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#dfn-throw" id="ref-for-dfn-throw⑦②">throw</a> an "<code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#operationerror" id="ref-for-operationerror①⑦">OperationError</a></code>" <code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#idl-DOMException" id="ref-for-idl-DOMException④⑨">DOMException</a></code>.</p>
+       <ol>
         <li data-md>
-         <p><em>inputLayout</em>: an <code class="idl"><a data-link-type="idl" href="#enumdef-mlinputoperandlayout" id="ref-for-enumdef-mlinputoperandlayout①">MLInputOperandLayout</a></code>. The default value is <em>"nchw"</em>. This option specifies the layout format of the input and output tensor as follow:</p>
-         <p>"nchw":</p>
-         <ul>
-          <li data-md>
-           <p>input tensor: [batches, input_channels, height, width]</p>
-          <li data-md>
-           <p>output tensor: [batches, output_channels, height, width]</p>
-         </ul>
-         <p>"nhwc":</p>
-         <ul>
-          <li data-md>
-           <p>input tensor: [batches, height, width, input_channels]</p>
-          <li data-md>
-           <p>output tensor: [batches, height, width, output_channels]</p>
-         </ul>
+         <p>Let <var>output</var> be the result of <a data-link-type="dfn" href="#create-an-mloperand" id="ref-for-create-an-mloperand⑤">creating an MLOperand</a> given <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#this" id="ref-for-this①④">this</a> and <var>desc</var>.</p>
         <li data-md>
-         <p><em>filterLayout</em>: an <code class="idl"><a data-link-type="idl" href="#enumdef-mlconv2dfilteroperandlayout" id="ref-for-enumdef-mlconv2dfilteroperandlayout①">MLConv2dFilterOperandLayout</a></code>. The default value is <em>"oihw"</em>. This option specifies the layout format of the filter tensor as follow:</p>
-         <p>"oihw":</p>
-         <ul>
+         <p>Make a request to the underlying platform to:</p>
+         <ol>
           <li data-md>
-           <p>[output_channels, input_channels/groups, height, width]</p>
-         </ul>
-         <p>"hwio":</p>
-         <ul>
+           <p>Create an <a data-link-type="dfn" href="https://infra.spec.whatwg.org/#implementation-defined" id="ref-for-implementation-defined①④">implementation-defined</a> platform operator <var>conv2dImpl</var> for this method, given <var>options</var> and <var>filter</var>.</p>
+           <ol>
+            <li data-md>
+             <p>If <var>options</var>.<code class="idl"><a data-link-type="idl" href="#dom-mlconv2doptions-activation" id="ref-for-dom-mlconv2doptions-activation③">activation</a></code> <a data-link-type="dfn" href="https://infra.spec.whatwg.org/#map-exists" id="ref-for-map-exists①⑧">exists</a>,register it as activation to <var>conv2dImpl</var>.</p>
+           </ol>
           <li data-md>
-           <p>[height, width, input_channels/groups, output_channels]</p>
-         </ul>
-         <p>"ohwi":</p>
-         <ul>
+           <p>Store a reference of <var>conv2dImpl</var> in <var>output</var>.<code class="idl"><a data-link-type="idl" href="#dom-mloperand-operator-slot" id="ref-for-dom-mloperand-operator-slot③">[[operator]]</a></code>.</p>
           <li data-md>
-           <p>[output_channels, height, width, input_channels/groups]</p>
-         </ul>
-         <p>"ihwo":</p>
-         <ul>
+           <p>Create an <a data-link-type="dfn" href="https://infra.spec.whatwg.org/#implementation-defined" id="ref-for-implementation-defined①⑤">implementation-defined</a> platform operand <var>outputImpl</var> to represent the output, given <var>output</var> and <var>conv2dImpl</var>.</p>
           <li data-md>
-           <p>[input_channels/groups, height, width, output_channels]</p>
-         </ul>
+           <p>Store a reference to <var>outputImpl</var> in <var>output</var>.<code class="idl"><a data-link-type="idl" href="#dom-mloperand-operand-slot" id="ref-for-dom-mloperand-operand-slot⑨">[[operand]]</a></code>.</p>
+         </ol>
         <li data-md>
-         <p><em>bias</em>: an <code class="idl"><a data-link-type="idl" href="#mloperand" id="ref-for-mloperand⑤⑦">MLOperand</a></code>. The additional 1-D tensor with the shape of [output_channels] whose values are to be added to the convolution result.</p>
+         <p>Connect <var>input</var>.<code class="idl"><a data-link-type="idl" href="#dom-mloperand-operand-slot" id="ref-for-dom-mloperand-operand-slot①⓪">[[operand]]</a></code> as input to <var>conv2dImpl</var>.</p>
         <li data-md>
-         <p><em>activation</em>: an <code class="idl"><a data-link-type="idl" href="#mlactivation" id="ref-for-mlactivation⑦">MLActivation</a></code>. The optional activation function that immediately follows the convolution operation.</p>
-       </ul>
-     </ul>
-     <p><strong>Returns:</strong> an <code class="idl"><a data-link-type="idl" href="#mloperand" id="ref-for-mloperand⑤⑧">MLOperand</a></code>. The output 4-D tensor that contains the convolution result. The output shape is interpreted according to the <em>options.inputLayout</em> value. More specifically, the spatial dimensions or the sizes of the last two dimensions of the output tensor for the <em>nchw</em> input layout can be calculated as follow:</p>
-     <p><em>output size = 1 + (input size - (filter size - 1) *</em> <em>dilation - 1 + beginning padding + ending padding) / stride</em></p>
-     <div class="note" role="note"> A <em>depthwise</em> conv2d operation is a variant of grouped convolution, used in models like the MobileNet, where the <em>options.groups</em> = input_channels = output_channels and the shape of filter tensor is [options.groups, 1, height, width]
-    for <em>"oihw"</em> layout, [height, width, 1, options.groups] for <em>"hwio"</em> layout, [options.groups, height, width, 1] for <em>"ohwi"</em> layout and [1, height, width, options.groups] for <em>"ihwo"</em> layout. </div>
+         <p>Connect <var>output</var>.<code class="idl"><a data-link-type="idl" href="#dom-mloperand-operand-slot" id="ref-for-dom-mloperand-operand-slot①①">[[operand]]</a></code> as output to <var>conv2dImpl</var>.</p>
+       </ol>
+      <li data-md>
+       <p>Return <var>output</var>.</p>
+     </ol>
     </div>
-    <h4 class="heading settled" data-level="7.6.8" id="api-mlgraphbuilder-convtranspose2d"><span class="secno">7.6.8. </span><span class="content">The convTranspose2d() method</span><a class="self-link" href="#api-mlgraphbuilder-convtranspose2d"></a></h4>
-     Compute a 2-D transposed convolution given 4-D input and filter tensors 
+   </details>
+   <h4 class="heading settled" data-level="7.6.9" id="api-mlgraphbuilder-convtranspose2d"><span class="secno">7.6.9. </span><span class="content">The convTranspose2d() method</span><a class="self-link" href="#api-mlgraphbuilder-convtranspose2d"></a></h4>
+    Compute a 2-D transposed convolution given 4-D input and filter tensors 
 <pre class="idl highlight def"><c- b>enum</c-> <dfn class="dfn-paneled idl-code" data-dfn-type="enum" data-export id="enumdef-mlconvtranspose2dfilteroperandlayout"><code><c- g>MLConvTranspose2dFilterOperandLayout</c-></code></dfn> {
   <dfn class="idl-code" data-dfn-for="MLConvTranspose2dFilterOperandLayout" data-dfn-type="enum-value" data-export id="dom-mlconvtranspose2dfilteroperandlayout-iohw"><code><c- s>"iohw"</c-></code><a class="self-link" href="#dom-mlconvtranspose2dfilteroperandlayout-iohw"></a></dfn>,
   <dfn class="idl-code" data-dfn-for="MLConvTranspose2dFilterOperandLayout" data-dfn-type="enum-value" data-export id="dom-mlconvtranspose2dfilteroperandlayout-hwoi"><code><c- s>"hwoi"</c-></code><a class="self-link" href="#dom-mlconvtranspose2dfilteroperandlayout-hwoi"></a></dfn>,
@@ -2418,216 +3326,653 @@ <h4 class="heading settled" data-level="7.6.8" id="api-mlgraphbuilder-convtransp
 };
 
 <c- b>dictionary</c-> <dfn class="dfn-paneled idl-code" data-dfn-type="dictionary" data-export id="dictdef-mlconvtranspose2doptions"><code><c- g>MLConvTranspose2dOptions</c-></code></dfn> {
-  <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#idl-sequence" id="ref-for-idl-sequence⑤"><c- b>sequence</c-></a>&lt;<a class="idl-code" data-link-type="interface" href="https://webidl.spec.whatwg.org/#idl-unsigned-long" id="ref-for-idl-unsigned-long①③"><c- b>unsigned</c-> <c- b>long</c-></a>> <dfn class="idl-code" data-dfn-for="MLConvTranspose2dOptions" data-dfn-type="dict-member" data-export data-type="sequence<unsigned long> " id="dom-mlconvtranspose2doptions-padding"><code><c- g>padding</c-></code><a class="self-link" href="#dom-mlconvtranspose2doptions-padding"></a></dfn>;
-  <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#idl-sequence" id="ref-for-idl-sequence⑥"><c- b>sequence</c-></a>&lt;<a class="idl-code" data-link-type="interface" href="https://webidl.spec.whatwg.org/#idl-unsigned-long" id="ref-for-idl-unsigned-long①④"><c- b>unsigned</c-> <c- b>long</c-></a>> <dfn class="idl-code" data-dfn-for="MLConvTranspose2dOptions" data-dfn-type="dict-member" data-export data-type="sequence<unsigned long> " id="dom-mlconvtranspose2doptions-strides"><code><c- g>strides</c-></code><a class="self-link" href="#dom-mlconvtranspose2doptions-strides"></a></dfn>;
-  <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#idl-sequence" id="ref-for-idl-sequence⑦"><c- b>sequence</c-></a>&lt;<a class="idl-code" data-link-type="interface" href="https://webidl.spec.whatwg.org/#idl-unsigned-long" id="ref-for-idl-unsigned-long①⑤"><c- b>unsigned</c-> <c- b>long</c-></a>> <dfn class="idl-code" data-dfn-for="MLConvTranspose2dOptions" data-dfn-type="dict-member" data-export data-type="sequence<unsigned long> " id="dom-mlconvtranspose2doptions-dilations"><code><c- g>dilations</c-></code><a class="self-link" href="#dom-mlconvtranspose2doptions-dilations"></a></dfn>;
-  <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#idl-sequence" id="ref-for-idl-sequence⑧"><c- b>sequence</c-></a>&lt;<a class="idl-code" data-link-type="interface" href="https://webidl.spec.whatwg.org/#idl-unsigned-long" id="ref-for-idl-unsigned-long①⑥"><c- b>unsigned</c-> <c- b>long</c-></a>> <dfn class="idl-code" data-dfn-for="MLConvTranspose2dOptions" data-dfn-type="dict-member" data-export data-type="sequence<unsigned long> " id="dom-mlconvtranspose2doptions-outputpadding"><code><c- g>outputPadding</c-></code><a class="self-link" href="#dom-mlconvtranspose2doptions-outputpadding"></a></dfn>;
-  <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#idl-sequence" id="ref-for-idl-sequence⑨"><c- b>sequence</c-></a>&lt;<a class="idl-code" data-link-type="interface" href="https://webidl.spec.whatwg.org/#idl-unsigned-long" id="ref-for-idl-unsigned-long①⑦"><c- b>unsigned</c-> <c- b>long</c-></a>> <dfn class="idl-code" data-dfn-for="MLConvTranspose2dOptions" data-dfn-type="dict-member" data-export data-type="sequence<unsigned long> " id="dom-mlconvtranspose2doptions-outputsizes"><code><c- g>outputSizes</c-></code><a class="self-link" href="#dom-mlconvtranspose2doptions-outputsizes"></a></dfn>;
-  <a data-link-type="idl-name" href="#enumdef-mlautopad" id="ref-for-enumdef-mlautopad②"><c- n>MLAutoPad</c-></a> <dfn class="idl-code" data-default="&quot;explicit&quot;" data-dfn-for="MLConvTranspose2dOptions" data-dfn-type="dict-member" data-export data-type="MLAutoPad " id="dom-mlconvtranspose2doptions-autopad"><code><c- g>autoPad</c-></code><a class="self-link" href="#dom-mlconvtranspose2doptions-autopad"></a></dfn> = "explicit";
-  <a class="idl-code" data-link-type="interface" href="https://webidl.spec.whatwg.org/#idl-unsigned-long" id="ref-for-idl-unsigned-long①⑧"><c- b>unsigned</c-> <c- b>long</c-></a> <dfn class="idl-code" data-default="1" data-dfn-for="MLConvTranspose2dOptions" data-dfn-type="dict-member" data-export data-type="unsigned long " id="dom-mlconvtranspose2doptions-groups"><code><c- g>groups</c-></code><a class="self-link" href="#dom-mlconvtranspose2doptions-groups"></a></dfn> = 1;
-  <a data-link-type="idl-name" href="#enumdef-mlinputoperandlayout" id="ref-for-enumdef-mlinputoperandlayout②"><c- n>MLInputOperandLayout</c-></a> <dfn class="idl-code" data-default="&quot;nchw&quot;" data-dfn-for="MLConvTranspose2dOptions" data-dfn-type="dict-member" data-export data-type="MLInputOperandLayout " id="dom-mlconvtranspose2doptions-inputlayout"><code><c- g>inputLayout</c-></code><a class="self-link" href="#dom-mlconvtranspose2doptions-inputlayout"></a></dfn> = "nchw";
-  <a data-link-type="idl-name" href="#enumdef-mlconvtranspose2dfilteroperandlayout" id="ref-for-enumdef-mlconvtranspose2dfilteroperandlayout"><c- n>MLConvTranspose2dFilterOperandLayout</c-></a> <dfn class="idl-code" data-default="&quot;iohw&quot;" data-dfn-for="MLConvTranspose2dOptions" data-dfn-type="dict-member" data-export data-type="MLConvTranspose2dFilterOperandLayout " id="dom-mlconvtranspose2doptions-filterlayout"><code><c- g>filterLayout</c-></code><a class="self-link" href="#dom-mlconvtranspose2doptions-filterlayout"></a></dfn> = "iohw";
-  <a data-link-type="idl-name" href="#mloperand" id="ref-for-mloperand⑤⑨"><c- n>MLOperand</c-></a> <dfn class="idl-code" data-dfn-for="MLConvTranspose2dOptions" data-dfn-type="dict-member" data-export data-type="MLOperand " id="dom-mlconvtranspose2doptions-bias"><code><c- g>bias</c-></code><a class="self-link" href="#dom-mlconvtranspose2doptions-bias"></a></dfn>;
-  <a data-link-type="idl-name" href="#mlactivation" id="ref-for-mlactivation⑧"><c- n>MLActivation</c-></a> <dfn class="idl-code" data-dfn-for="MLConvTranspose2dOptions" data-dfn-type="dict-member" data-export data-type="MLActivation " id="dom-mlconvtranspose2doptions-activation"><code><c- g>activation</c-></code><a class="self-link" href="#dom-mlconvtranspose2doptions-activation"></a></dfn>;
+  <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#idl-sequence" id="ref-for-idl-sequence⑤"><c- b>sequence</c-></a>&lt;<a class="idl-code" data-link-type="interface" href="https://webidl.spec.whatwg.org/#idl-unsigned-long" id="ref-for-idl-unsigned-long①④"><c- b>unsigned</c-> <c- b>long</c-></a>> <a class="idl-code" data-link-type="dict-member" data-type="sequence<unsigned long> " href="#dom-mlconvtranspose2doptions-padding" id="ref-for-dom-mlconvtranspose2doptions-padding"><c- g>padding</c-></a>;
+  <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#idl-sequence" id="ref-for-idl-sequence⑥"><c- b>sequence</c-></a>&lt;<a class="idl-code" data-link-type="interface" href="https://webidl.spec.whatwg.org/#idl-unsigned-long" id="ref-for-idl-unsigned-long①⑤"><c- b>unsigned</c-> <c- b>long</c-></a>> <a class="idl-code" data-link-type="dict-member" data-type="sequence<unsigned long> " href="#dom-mlconvtranspose2doptions-strides" id="ref-for-dom-mlconvtranspose2doptions-strides"><c- g>strides</c-></a>;
+  <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#idl-sequence" id="ref-for-idl-sequence⑦"><c- b>sequence</c-></a>&lt;<a class="idl-code" data-link-type="interface" href="https://webidl.spec.whatwg.org/#idl-unsigned-long" id="ref-for-idl-unsigned-long①⑥"><c- b>unsigned</c-> <c- b>long</c-></a>> <a class="idl-code" data-link-type="dict-member" data-type="sequence<unsigned long> " href="#dom-mlconvtranspose2doptions-dilations" id="ref-for-dom-mlconvtranspose2doptions-dilations"><c- g>dilations</c-></a>;
+  <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#idl-sequence" id="ref-for-idl-sequence⑧"><c- b>sequence</c-></a>&lt;<a class="idl-code" data-link-type="interface" href="https://webidl.spec.whatwg.org/#idl-unsigned-long" id="ref-for-idl-unsigned-long①⑦"><c- b>unsigned</c-> <c- b>long</c-></a>> <a class="idl-code" data-link-type="dict-member" data-type="sequence<unsigned long> " href="#dom-mlconvtranspose2doptions-outputpadding" id="ref-for-dom-mlconvtranspose2doptions-outputpadding"><c- g>outputPadding</c-></a>;
+  <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#idl-sequence" id="ref-for-idl-sequence⑨"><c- b>sequence</c-></a>&lt;<a class="idl-code" data-link-type="interface" href="https://webidl.spec.whatwg.org/#idl-unsigned-long" id="ref-for-idl-unsigned-long①⑧"><c- b>unsigned</c-> <c- b>long</c-></a>> <a class="idl-code" data-link-type="dict-member" data-type="sequence<unsigned long> " href="#dom-mlconvtranspose2doptions-outputsizes" id="ref-for-dom-mlconvtranspose2doptions-outputsizes"><c- g>outputSizes</c-></a>;
+  <a data-link-type="idl-name" href="#enumdef-mlautopad" id="ref-for-enumdef-mlautopad③"><c- n>MLAutoPad</c-></a> <a class="idl-code" data-default="&quot;explicit&quot;" data-link-type="dict-member" data-type="MLAutoPad " href="#dom-mlconvtranspose2doptions-autopad" id="ref-for-dom-mlconvtranspose2doptions-autopad"><c- g>autoPad</c-></a> = "explicit";
+  <a class="idl-code" data-link-type="interface" href="https://webidl.spec.whatwg.org/#idl-unsigned-long" id="ref-for-idl-unsigned-long①⑨"><c- b>unsigned</c-> <c- b>long</c-></a> <a class="idl-code" data-default="1" data-link-type="dict-member" data-type="unsigned long " href="#dom-mlconvtranspose2doptions-groups" id="ref-for-dom-mlconvtranspose2doptions-groups"><c- g>groups</c-></a> = 1;
+  <a data-link-type="idl-name" href="#enumdef-mlinputoperandlayout" id="ref-for-enumdef-mlinputoperandlayout③"><c- n>MLInputOperandLayout</c-></a> <a class="idl-code" data-default="&quot;nchw&quot;" data-link-type="dict-member" data-type="MLInputOperandLayout " href="#dom-mlconvtranspose2doptions-inputlayout" id="ref-for-dom-mlconvtranspose2doptions-inputlayout"><c- g>inputLayout</c-></a> = "nchw";
+  <a data-link-type="idl-name" href="#enumdef-mlconvtranspose2dfilteroperandlayout" id="ref-for-enumdef-mlconvtranspose2dfilteroperandlayout"><c- n>MLConvTranspose2dFilterOperandLayout</c-></a> <a class="idl-code" data-default="&quot;iohw&quot;" data-link-type="dict-member" data-type="MLConvTranspose2dFilterOperandLayout " href="#dom-mlconvtranspose2doptions-filterlayout" id="ref-for-dom-mlconvtranspose2doptions-filterlayout"><c- g>filterLayout</c-></a> = "iohw";
+  <a data-link-type="idl-name" href="#mloperand" id="ref-for-mloperand⑦⓪"><c- n>MLOperand</c-></a> <a class="idl-code" data-link-type="dict-member" data-type="MLOperand " href="#dom-mlconvtranspose2doptions-bias" id="ref-for-dom-mlconvtranspose2doptions-bias"><c- g>bias</c-></a>;
+  <a data-link-type="idl-name" href="#mlactivation" id="ref-for-mlactivation①⑦"><c- n>MLActivation</c-></a> <a class="idl-code" data-link-type="dict-member" data-type="MLActivation " href="#dom-mlconvtranspose2doptions-activation" id="ref-for-dom-mlconvtranspose2doptions-activation"><c- g>activation</c-></a>;
 };
 
-<c- b>partial</c-> <c- b>interface</c-> <a class="idl-code" data-link-type="interface" href="#mlgraphbuilder" id="ref-for-mlgraphbuilder②④"><c- g>MLGraphBuilder</c-></a> {
-  <a data-link-type="idl-name" href="#mloperand" id="ref-for-mloperand⑥⓪"><c- n>MLOperand</c-></a> <dfn class="idl-code" data-dfn-for="MLGraphBuilder" data-dfn-type="method" data-export data-lt="convTranspose2d(input, filter, options)|convTranspose2d(input, filter)" id="dom-mlgraphbuilder-convtranspose2d"><code><c- g>convTranspose2d</c-></code><a class="self-link" href="#dom-mlgraphbuilder-convtranspose2d"></a></dfn>(<a data-link-type="idl-name" href="#mloperand" id="ref-for-mloperand⑥①"><c- n>MLOperand</c-></a> <dfn class="idl-code" data-dfn-for="MLGraphBuilder/convTranspose2d(input, filter, options), MLGraphBuilder/convTranspose2d(input, filter)" data-dfn-type="argument" data-export id="dom-mlgraphbuilder-convtranspose2d-input-filter-options-input"><code><c- g>input</c-></code><a class="self-link" href="#dom-mlgraphbuilder-convtranspose2d-input-filter-options-input"></a></dfn>, <a data-link-type="idl-name" href="#mloperand" id="ref-for-mloperand⑥②"><c- n>MLOperand</c-></a> <dfn class="idl-code" data-dfn-for="MLGraphBuilder/convTranspose2d(input, filter, options), MLGraphBuilder/convTranspose2d(input, filter)" data-dfn-type="argument" data-export id="dom-mlgraphbuilder-convtranspose2d-input-filter-options-filter"><code><c- g>filter</c-></code><a class="self-link" href="#dom-mlgraphbuilder-convtranspose2d-input-filter-options-filter"></a></dfn>,
+<c- b>partial</c-> <c- b>interface</c-> <a class="idl-code" data-link-type="interface" href="#mlgraphbuilder" id="ref-for-mlgraphbuilder②⑦"><c- g>MLGraphBuilder</c-></a> {
+  <a data-link-type="idl-name" href="#mloperand" id="ref-for-mloperand⑦①"><c- n>MLOperand</c-></a> <a class="idl-code" data-link-type="method" href="#dom-mlgraphbuilder-convtranspose2d" id="ref-for-dom-mlgraphbuilder-convtranspose2d"><c- g>convTranspose2d</c-></a>(<a data-link-type="idl-name" href="#mloperand" id="ref-for-mloperand⑦②"><c- n>MLOperand</c-></a> <dfn class="idl-code" data-dfn-for="MLGraphBuilder/convTranspose2d(input, filter, options), MLGraphBuilder/convTranspose2d(input, filter)" data-dfn-type="argument" data-export id="dom-mlgraphbuilder-convtranspose2d-input-filter-options-input"><code><c- g>input</c-></code><a class="self-link" href="#dom-mlgraphbuilder-convtranspose2d-input-filter-options-input"></a></dfn>, <a data-link-type="idl-name" href="#mloperand" id="ref-for-mloperand⑦③"><c- n>MLOperand</c-></a> <dfn class="idl-code" data-dfn-for="MLGraphBuilder/convTranspose2d(input, filter, options), MLGraphBuilder/convTranspose2d(input, filter)" data-dfn-type="argument" data-export id="dom-mlgraphbuilder-convtranspose2d-input-filter-options-filter"><code><c- g>filter</c-></code><a class="self-link" href="#dom-mlgraphbuilder-convtranspose2d-input-filter-options-filter"></a></dfn>,
                             <c- b>optional</c-> <a data-link-type="idl-name" href="#dictdef-mlconvtranspose2doptions" id="ref-for-dictdef-mlconvtranspose2doptions"><c- n>MLConvTranspose2dOptions</c-></a> <dfn class="idl-code" data-dfn-for="MLGraphBuilder/convTranspose2d(input, filter, options), MLGraphBuilder/convTranspose2d(input, filter)" data-dfn-type="argument" data-export id="dom-mlgraphbuilder-convtranspose2d-input-filter-options-options"><code><c- g>options</c-></code><a class="self-link" href="#dom-mlgraphbuilder-convtranspose2d-input-filter-options-options"></a></dfn> = {});
 };
 </pre>
-    <div class="algorithm" data-algorithm="convtranspose2d">
-      <strong>Arguments:</strong> 
+   <p><code class="idl"><a data-link-type="idl" href="#dictdef-mlconvtranspose2doptions" id="ref-for-dictdef-mlconvtranspose2doptions①">MLConvTranspose2dOptions</a></code> has the following members:</p>
+   <dl>
+    <dt data-md><dfn class="dfn-paneled idl-code" data-dfn-for="MLConvTranspose2dOptions" data-dfn-type="dict-member" data-export id="dom-mlconvtranspose2doptions-padding"><code>padding</code></dfn>, <span> of type <code class="idl-code">sequence&lt;unsigned long></code></span>
+    <dd data-md>
+     <p>A sequence of <code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#idl-unsigned-long" id="ref-for-idl-unsigned-long②⓪">unsigned long</a></code> of length 4: [beginningHeight, endingHeight, beginningWidth, endingWidth].
+Specifies the additional rows and columns added to the beginning and ending of each spatial dimension of the convolution input.
+The default value is [0, 0, 0, 0].</p>
+    <dt data-md><dfn class="dfn-paneled idl-code" data-dfn-for="MLConvTranspose2dOptions" data-dfn-type="dict-member" data-export id="dom-mlconvtranspose2doptions-strides"><code>strides</code></dfn>, <span> of type <code class="idl-code">sequence&lt;unsigned long></code></span>
+    <dd data-md>
+     <p>A sequence of <code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#idl-unsigned-long" id="ref-for-idl-unsigned-long②①">unsigned long</a></code> of length 2: [strideHeight, strideWidth].
+Specifies the stride of the sliding window for each spatial dimension of the convolution input.
+The default value is [1, 1].</p>
+    <dt data-md><dfn class="dfn-paneled idl-code" data-dfn-for="MLConvTranspose2dOptions" data-dfn-type="dict-member" data-export id="dom-mlconvtranspose2doptions-dilations"><code>dilations</code></dfn>, <span> of type <code class="idl-code">sequence&lt;unsigned long></code></span>
+    <dd data-md>
+     <p>A sequence of <code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#idl-unsigned-long" id="ref-for-idl-unsigned-long②②">unsigned long</a></code> of length 2: [dilationHeight, dilationWidth]. Specifies the dilation factor for each spatial dimension applied on the convolution filter (kernel).
+The default value is [1, 1].</p>
+    <dt data-md><dfn class="dfn-paneled idl-code" data-dfn-for="MLConvTranspose2dOptions" data-dfn-type="dict-member" data-export id="dom-mlconvtranspose2doptions-outputpadding"><code>outputPadding</code></dfn>, <span> of type <code class="idl-code">sequence&lt;unsigned long></code></span>
+    <dd data-md>
+     <p>A sequence of <code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#idl-unsigned-long" id="ref-for-idl-unsigned-long②③">unsigned long</a></code> of length 2.
+Specifies the padding values applied to each spatial dimension of the output tensor. The explicit padding values are needed to disambiguate the output tensor shape for transposed convolution when the value of the <em>options</em>.<code class="idl"><a data-link-type="idl" href="#dom-mlconvtranspose2doptions-strides" id="ref-for-dom-mlconvtranspose2doptions-strides①">strides</a></code> is greater than 1.</p>
+     <p>Note that these values are only used to disambiguate output shape when needed; it does not necessarily cause any padding value to be written to the output tensor.</p>
+     <p>The default values is [0, 0].</p>
+    <dt data-md><dfn class="dfn-paneled idl-code" data-dfn-for="MLConvTranspose2dOptions" data-dfn-type="dict-member" data-export id="dom-mlconvtranspose2doptions-outputsizes"><code>outputSizes</code></dfn>, <span> of type <code class="idl-code">sequence&lt;unsigned long></code></span>
+    <dd data-md>
+     <p>A sequence of <code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#idl-unsigned-long" id="ref-for-idl-unsigned-long②④">unsigned long</a></code> of length 2.
+Specifies the sizes of the last two dimensions of the output tensor. When the output sizes are explicitly specified, the output padding values in <code class="idl"><a data-link-type="idl" href="#dom-mlconvtranspose2doptions-outputpadding" id="ref-for-dom-mlconvtranspose2doptions-outputpadding①">outputPadding</a></code> are ignored.</p>
+     <p>If not specified, the output sizes are automatically computed.</p>
+    <dt data-md><dfn class="dfn-paneled idl-code" data-dfn-for="MLConvTranspose2dOptions" data-dfn-type="dict-member" data-export id="dom-mlconvtranspose2doptions-autopad"><code>autoPad</code></dfn>, <span> of type <a data-link-type="idl-name" href="#enumdef-mlautopad" id="ref-for-enumdef-mlautopad④">MLAutoPad</a>, defaulting to <code>"explicit"</code></span>
+    <dd data-md>
+     <p>An <code class="idl"><a data-link-type="idl" href="#enumdef-mlautopad" id="ref-for-enumdef-mlautopad⑤">MLAutoPad</a></code> <a data-link-type="dfn" href="https://infra.spec.whatwg.org/#string" id="ref-for-string⑦">string</a>.
+Specifies the automatic input padding options.
+The default value is <em>"explicit"</em>, which means that the values in the <code class="idl"><a data-link-type="idl" href="#dom-mlconvtranspose2doptions-padding" id="ref-for-dom-mlconvtranspose2doptions-padding①">padding</a></code> array should be used for input padding.</p>
+     <p>When the option is set other than <em>"explicit"</em>, the values in the <code class="idl"><a data-link-type="idl" href="#dom-mlconvtranspose2doptions-padding" id="ref-for-dom-mlconvtranspose2doptions-padding②">padding</a></code> array are ignored.</p>
+     <p>With the <em>"same-upper"</em> option, the padding values are automatically computed such that the additional ending padding of the spatial input dimensions would allow all of the input values in the corresponding dimension to be filtered.</p>
+     <p>The <em>"same-lower"</em> option is similar but padding is applied to the beginning padding of the spatial input dimensions instead of the ending one.</p>
+    <dt data-md><dfn class="dfn-paneled idl-code" data-dfn-for="MLConvTranspose2dOptions" data-dfn-type="dict-member" data-export id="dom-mlconvtranspose2doptions-groups"><code>groups</code></dfn>, <span> of type <a data-link-type="idl-name" href="https://webidl.spec.whatwg.org/#idl-unsigned-long" id="ref-for-idl-unsigned-long②⑤">unsigned long</a>, defaulting to <code>1</code></span>
+    <dd data-md>
+     <p>An <code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#idl-unsigned-long" id="ref-for-idl-unsigned-long②⑥">unsigned long</a></code> scalar.
+Specifies the number of groups that input channels and output channels are divided into.
+The default value is 1.</p>
+    <dt data-md><dfn class="dfn-paneled idl-code" data-dfn-for="MLConvTranspose2dOptions" data-dfn-type="dict-member" data-export id="dom-mlconvtranspose2doptions-inputlayout"><code>inputLayout</code></dfn>, <span> of type <a data-link-type="idl-name" href="#enumdef-mlinputoperandlayout" id="ref-for-enumdef-mlinputoperandlayout④">MLInputOperandLayout</a>, defaulting to <code>"nchw"</code></span>
+    <dd data-md>
+     <p>An <code class="idl"><a data-link-type="idl" href="#enumdef-mlinputoperandlayout" id="ref-for-enumdef-mlinputoperandlayout⑤">MLInputOperandLayout</a></code> <a data-link-type="dfn" href="https://infra.spec.whatwg.org/#string" id="ref-for-string⑧">string</a>.
+Specifies the layout format of the input and output tensor as follows:</p>
      <ul>
       <li data-md>
-       <p><em>input</em>: an <code class="idl"><a data-link-type="idl" href="#mloperand" id="ref-for-mloperand⑥③">MLOperand</a></code>. The input 4-D tensor. The logical shape
-is interpreted according to the value of <em>options.inputLayout</em>.</p>
-      <li data-md>
-       <p><em>filter</em>: an <code class="idl"><a data-link-type="idl" href="#mloperand" id="ref-for-mloperand⑥④">MLOperand</a></code>. The filter 4-D tensor. The logical shape is
-interpreted according to the value of <em>options.filterLayout</em> and <em>options.groups</em>.</p>
-      <li data-md>
-       <p><em>options</em>: an optional <code class="idl"><a data-link-type="idl" href="#dictdef-mlconvtranspose2doptions" id="ref-for-dictdef-mlconvtranspose2doptions①">MLConvTranspose2dOptions</a></code>. The optional parameters of the operation.</p>
+       <p><strong>"nchw"</strong></p>
        <ul>
         <li data-md>
-         <p><em>padding</em>: a sequence of <code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#idl-unsigned-long" id="ref-for-idl-unsigned-long①⑨">unsigned long</a></code> of length 4. The additional rows and columns added to the beginning and ending of each spatial dimension of <em>input</em>, [beginning_height, ending_height, beginning_width, ending_width]. If not present, the values are assumed to be [0,0,0,0].</p>
-        <li data-md>
-         <p><em>strides</em>: a sequence of <code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#idl-unsigned-long" id="ref-for-idl-unsigned-long②⓪">unsigned long</a></code> of length 2. The stride of the sliding window for each spatial dimension of <em>input</em>, [stride_height, stride_width]. If not present, the values are assumed to be [1,1].</p>
-        <li data-md>
-         <p><em>dilations</em>: a sequence of <code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#idl-unsigned-long" id="ref-for-idl-unsigned-long②①">unsigned long</a></code> of length 2. The dilation factor for each spatial dimension of <em>input</em>, [dilation_height, dilation_width]. If not present, the values are assumed to be [1,1].</p>
-        <li data-md>
-         <p><em>outputPadding</em>: a sequence of <code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#idl-unsigned-long" id="ref-for-idl-unsigned-long②②">unsigned long</a></code> of length 2. The padding values applied to each spatial dimension of the output tensor. This explicit padding values are needed to disambiguate the output tensor shape for transposed convolution when the value of the <em>options.strides</em> is greater than 1. Note that these values are only used to disambiguate output shape when needed; it does not necessarily cause any padding value to be written to the output tensor. If not specified, the values are assumed to be [0,0].</p>
-        <li data-md>
-         <p><em>outputSizes</em>: a sequence of <code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#idl-unsigned-long" id="ref-for-idl-unsigned-long②③">unsigned long</a></code> of length 2. The sizes of the last two dimensions of the output tensor. When the output sizes are explicitly specified, the output padding values in <em>options.outputPadding</em> are ignored. If not specified, the output sizes are automatically computed.</p>
-        <li data-md>
-         <p><em>autoPad</em>: an <code class="idl"><a data-link-type="idl" href="#enumdef-mlautopad" id="ref-for-enumdef-mlautopad③">MLAutoPad</a></code>. The automatic input padding options. By default, this argument is set to <em>"explicit"</em>, which means that the values in the <em>options.padding</em> array should be used for input padding. When the option is set other than <em>"explicit"</em>, the values in the <em>options.padding</em> array are ignored. With the <em>"same-upper"</em> option, the padding values are automatically computed such that the additional ending padding of the spatial input dimensions would allow all of the input values in the corresponding dimension to be filtered. The <em>"same-lower"</em> option is similar but padding is applied to the beginning padding of the spatial input dimensions instead of the ending one.</p>
-        <li data-md>
-         <p><em>groups</em>: an <code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#idl-unsigned-long" id="ref-for-idl-unsigned-long②④">unsigned long</a></code> scalar. The number of groups that input channels and output channels are divided into, default to 1.</p>
-        <li data-md>
-         <p><em>inputLayout</em>: an <code class="idl"><a data-link-type="idl" href="#enumdef-mlinputoperandlayout" id="ref-for-enumdef-mlinputoperandlayout③">MLInputOperandLayout</a></code>. The default value is <em>"nchw"</em>. This option specifies the layout format of the input and output tensor as follow:</p>
-         <p>"nchw":</p>
-         <ul>
-          <li data-md>
-           <p>input tensor: [batches, input_channels, height, width]</p>
-          <li data-md>
-           <p>output tensor: [batches, output_channels, height, width]</p>
-         </ul>
-         <p>"nhwc":</p>
-         <ul>
-          <li data-md>
-           <p>input tensor: [batches, height, width, input_channels]</p>
-          <li data-md>
-           <p>output tensor: [batches, height, width, output_channels]</p>
-         </ul>
+         <p>input tensor: <em>[batches, inputChannels, height, width]</em></p>
         <li data-md>
-         <p><em>filterLayout</em>: an <code class="idl"><a data-link-type="idl" href="#enumdef-mlconvtranspose2dfilteroperandlayout" id="ref-for-enumdef-mlconvtranspose2dfilteroperandlayout①">MLConvTranspose2dFilterOperandLayout</a></code>. The default value is <em>"iohw"</em>. This option specifies the layout format of the filter tensor as follow:</p>
-         <p>"iohw":</p>
-         <ul>
-          <li data-md>
-           <p>[input_channels, output_channels/groups, height, width]</p>
-         </ul>
-         <p>"hwoi":</p>
-         <ul>
-          <li data-md>
-           <p>[height, width, output_channels/groups, input_channels]</p>
-         </ul>
-         <p>"ohwi":</p>
-         <ul>
-          <li data-md>
-           <p>[output_channels/groups, height, width, input_channels]</p>
-         </ul>
+         <p>output tensor: <em>[batches, outputChannels, height, width]</em></p>
+       </ul>
+      <li data-md>
+       <p><strong>"nhwc"</strong>:</p>
+       <ul>
         <li data-md>
-         <p><em>bias</em>: an <code class="idl"><a data-link-type="idl" href="#mloperand" id="ref-for-mloperand⑥⑤">MLOperand</a></code>. The additional 1-D tensor with the shape of [output_channels] whose values are to be added to the transposed convolution result.</p>
+         <p>input tensor: <em>[batches, height, width, inputChannels]</em></p>
         <li data-md>
-         <p><em>activation</em>: an <code class="idl"><a data-link-type="idl" href="#mlactivation" id="ref-for-mlactivation⑨">MLActivation</a></code>. The optional activation function that immediately follows the transposed convolution operation.</p>
+         <p>output tensor: <em>[batches, height, width, outputChannels]</em></p>
        </ul>
      </ul>
-     <p><strong>Returns:</strong> an <code class="idl"><a data-link-type="idl" href="#mloperand" id="ref-for-mloperand⑥⑥">MLOperand</a></code>. The output 4-D tensor that contains the transposed convolution result. The output shape is interpreted according to the <em>options.inputLayout</em> value. More specifically, unless the <em>options.outputSizes</em> values are explicitly specified, the <em>options.outputPadding</em> may be needed to compute the spatial dimension values of the output tensor as follow:</p>
-     <p><em>output size = (input size - 1) *</em> <em>stride + (filter size - 1) *</em> <em>dilation + 1 - beginning padding - ending padding + output padding</em></p>
-    </div>
-    <h4 class="heading settled" data-level="7.6.9" id="api-mlgraphbuilder-binary"><span class="secno">7.6.9. </span><span class="content">Element-wise binary operations</span><a class="self-link" href="#api-mlgraphbuilder-binary"></a></h4>
-     Compute the element-wise binary addition, subtraction, multiplication, division,
-maximum and minimum of the two input tensors. 
-<pre class="idl highlight def"><c- b>partial</c-> <c- b>interface</c-> <a class="idl-code" data-link-type="interface" href="#mlgraphbuilder" id="ref-for-mlgraphbuilder②⑤"><c- g>MLGraphBuilder</c-></a> {
-  <a data-link-type="idl-name" href="#mloperand" id="ref-for-mloperand⑥⑦"><c- n>MLOperand</c-></a> <dfn class="idl-code" data-dfn-for="MLGraphBuilder" data-dfn-type="method" data-export data-lt="add(a, b)" id="dom-mlgraphbuilder-add"><code><c- g>add</c-></code><a class="self-link" href="#dom-mlgraphbuilder-add"></a></dfn>(<a data-link-type="idl-name" href="#mloperand" id="ref-for-mloperand⑥⑧"><c- n>MLOperand</c-></a> <dfn class="idl-code" data-dfn-for="MLGraphBuilder/add(a, b)" data-dfn-type="argument" data-export id="dom-mlgraphbuilder-add-a-b-a"><code><c- g>a</c-></code><a class="self-link" href="#dom-mlgraphbuilder-add-a-b-a"></a></dfn>, <a data-link-type="idl-name" href="#mloperand" id="ref-for-mloperand⑥⑨"><c- n>MLOperand</c-></a> <dfn class="idl-code" data-dfn-for="MLGraphBuilder/add(a, b)" data-dfn-type="argument" data-export id="dom-mlgraphbuilder-add-a-b-b"><code><c- g>b</c-></code><a class="self-link" href="#dom-mlgraphbuilder-add-a-b-b"></a></dfn>);
-  <a data-link-type="idl-name" href="#mloperand" id="ref-for-mloperand⑦⓪"><c- n>MLOperand</c-></a> <dfn class="idl-code" data-dfn-for="MLGraphBuilder" data-dfn-type="method" data-export data-lt="sub(a, b)" id="dom-mlgraphbuilder-sub"><code><c- g>sub</c-></code><a class="self-link" href="#dom-mlgraphbuilder-sub"></a></dfn>(<a data-link-type="idl-name" href="#mloperand" id="ref-for-mloperand⑦①"><c- n>MLOperand</c-></a> <dfn class="idl-code" data-dfn-for="MLGraphBuilder/sub(a, b)" data-dfn-type="argument" data-export id="dom-mlgraphbuilder-sub-a-b-a"><code><c- g>a</c-></code><a class="self-link" href="#dom-mlgraphbuilder-sub-a-b-a"></a></dfn>, <a data-link-type="idl-name" href="#mloperand" id="ref-for-mloperand⑦②"><c- n>MLOperand</c-></a> <dfn class="idl-code" data-dfn-for="MLGraphBuilder/sub(a, b)" data-dfn-type="argument" data-export id="dom-mlgraphbuilder-sub-a-b-b"><code><c- g>b</c-></code><a class="self-link" href="#dom-mlgraphbuilder-sub-a-b-b"></a></dfn>);
-  <a data-link-type="idl-name" href="#mloperand" id="ref-for-mloperand⑦③"><c- n>MLOperand</c-></a> <dfn class="idl-code" data-dfn-for="MLGraphBuilder" data-dfn-type="method" data-export data-lt="mul(a, b)" id="dom-mlgraphbuilder-mul"><code><c- g>mul</c-></code><a class="self-link" href="#dom-mlgraphbuilder-mul"></a></dfn>(<a data-link-type="idl-name" href="#mloperand" id="ref-for-mloperand⑦④"><c- n>MLOperand</c-></a> <dfn class="idl-code" data-dfn-for="MLGraphBuilder/mul(a, b)" data-dfn-type="argument" data-export id="dom-mlgraphbuilder-mul-a-b-a"><code><c- g>a</c-></code><a class="self-link" href="#dom-mlgraphbuilder-mul-a-b-a"></a></dfn>, <a data-link-type="idl-name" href="#mloperand" id="ref-for-mloperand⑦⑤"><c- n>MLOperand</c-></a> <dfn class="idl-code" data-dfn-for="MLGraphBuilder/mul(a, b)" data-dfn-type="argument" data-export id="dom-mlgraphbuilder-mul-a-b-b"><code><c- g>b</c-></code><a class="self-link" href="#dom-mlgraphbuilder-mul-a-b-b"></a></dfn>);
-  <a data-link-type="idl-name" href="#mloperand" id="ref-for-mloperand⑦⑥"><c- n>MLOperand</c-></a> <dfn class="idl-code" data-dfn-for="MLGraphBuilder" data-dfn-type="method" data-export data-lt="div(a, b)" id="dom-mlgraphbuilder-div"><code><c- g>div</c-></code><a class="self-link" href="#dom-mlgraphbuilder-div"></a></dfn>(<a data-link-type="idl-name" href="#mloperand" id="ref-for-mloperand⑦⑦"><c- n>MLOperand</c-></a> <dfn class="idl-code" data-dfn-for="MLGraphBuilder/div(a, b)" data-dfn-type="argument" data-export id="dom-mlgraphbuilder-div-a-b-a"><code><c- g>a</c-></code><a class="self-link" href="#dom-mlgraphbuilder-div-a-b-a"></a></dfn>, <a data-link-type="idl-name" href="#mloperand" id="ref-for-mloperand⑦⑧"><c- n>MLOperand</c-></a> <dfn class="idl-code" data-dfn-for="MLGraphBuilder/div(a, b)" data-dfn-type="argument" data-export id="dom-mlgraphbuilder-div-a-b-b"><code><c- g>b</c-></code><a class="self-link" href="#dom-mlgraphbuilder-div-a-b-b"></a></dfn>);
-  <a data-link-type="idl-name" href="#mloperand" id="ref-for-mloperand⑦⑨"><c- n>MLOperand</c-></a> <dfn class="idl-code" data-dfn-for="MLGraphBuilder" data-dfn-type="method" data-export data-lt="max(a, b)" id="dom-mlgraphbuilder-max"><code><c- g>max</c-></code><a class="self-link" href="#dom-mlgraphbuilder-max"></a></dfn>(<a data-link-type="idl-name" href="#mloperand" id="ref-for-mloperand⑧⓪"><c- n>MLOperand</c-></a> <dfn class="idl-code" data-dfn-for="MLGraphBuilder/max(a, b)" data-dfn-type="argument" data-export id="dom-mlgraphbuilder-max-a-b-a"><code><c- g>a</c-></code><a class="self-link" href="#dom-mlgraphbuilder-max-a-b-a"></a></dfn>, <a data-link-type="idl-name" href="#mloperand" id="ref-for-mloperand⑧①"><c- n>MLOperand</c-></a> <dfn class="idl-code" data-dfn-for="MLGraphBuilder/max(a, b)" data-dfn-type="argument" data-export id="dom-mlgraphbuilder-max-a-b-b"><code><c- g>b</c-></code><a class="self-link" href="#dom-mlgraphbuilder-max-a-b-b"></a></dfn>);
-  <a data-link-type="idl-name" href="#mloperand" id="ref-for-mloperand⑧②"><c- n>MLOperand</c-></a> <dfn class="idl-code" data-dfn-for="MLGraphBuilder" data-dfn-type="method" data-export data-lt="min(a, b)" id="dom-mlgraphbuilder-min"><code><c- g>min</c-></code><a class="self-link" href="#dom-mlgraphbuilder-min"></a></dfn>(<a data-link-type="idl-name" href="#mloperand" id="ref-for-mloperand⑧③"><c- n>MLOperand</c-></a> <dfn class="idl-code" data-dfn-for="MLGraphBuilder/min(a, b)" data-dfn-type="argument" data-export id="dom-mlgraphbuilder-min-a-b-a"><code><c- g>a</c-></code><a class="self-link" href="#dom-mlgraphbuilder-min-a-b-a"></a></dfn>, <a data-link-type="idl-name" href="#mloperand" id="ref-for-mloperand⑧④"><c- n>MLOperand</c-></a> <dfn class="idl-code" data-dfn-for="MLGraphBuilder/min(a, b)" data-dfn-type="argument" data-export id="dom-mlgraphbuilder-min-a-b-b"><code><c- g>b</c-></code><a class="self-link" href="#dom-mlgraphbuilder-min-a-b-b"></a></dfn>);
-  <a data-link-type="idl-name" href="#mloperand" id="ref-for-mloperand⑧⑤"><c- n>MLOperand</c-></a> <dfn class="idl-code" data-dfn-for="MLGraphBuilder" data-dfn-type="method" data-export data-lt="pow(a, b)" id="dom-mlgraphbuilder-pow"><code><c- g>pow</c-></code><a class="self-link" href="#dom-mlgraphbuilder-pow"></a></dfn>(<a data-link-type="idl-name" href="#mloperand" id="ref-for-mloperand⑧⑥"><c- n>MLOperand</c-></a> <dfn class="idl-code" data-dfn-for="MLGraphBuilder/pow(a, b)" data-dfn-type="argument" data-export id="dom-mlgraphbuilder-pow-a-b-a"><code><c- g>a</c-></code><a class="self-link" href="#dom-mlgraphbuilder-pow-a-b-a"></a></dfn>, <a data-link-type="idl-name" href="#mloperand" id="ref-for-mloperand⑧⑦"><c- n>MLOperand</c-></a> <dfn class="idl-code" data-dfn-for="MLGraphBuilder/pow(a, b)" data-dfn-type="argument" data-export id="dom-mlgraphbuilder-pow-a-b-b"><code><c- g>b</c-></code><a class="self-link" href="#dom-mlgraphbuilder-pow-a-b-b"></a></dfn>);
-};
-</pre>
-    <div class="algorithm" data-algorithm="binary">
-      <strong>Arguments:</strong> 
+     The default value is <em>"nchw"</em>. 
+    <dt data-md><dfn class="dfn-paneled idl-code" data-dfn-for="MLConvTranspose2dOptions" data-dfn-type="dict-member" data-export id="dom-mlconvtranspose2doptions-filterlayout"><code>filterLayout</code></dfn>, <span> of type <a data-link-type="idl-name" href="#enumdef-mlconvtranspose2dfilteroperandlayout" id="ref-for-enumdef-mlconvtranspose2dfilteroperandlayout①">MLConvTranspose2dFilterOperandLayout</a>, defaulting to <code>"iohw"</code></span>
+    <dd data-md>
+     <p>An <code class="idl"><a data-link-type="idl" href="#enumdef-mlconvtranspose2dfilteroperandlayout" id="ref-for-enumdef-mlconvtranspose2dfilteroperandlayout②">MLConvTranspose2dFilterOperandLayout</a></code> <a data-link-type="dfn" href="https://infra.spec.whatwg.org/#string" id="ref-for-string⑨">string</a>.
+Specifies the layout format of the filter tensor as follow:</p>
      <ul>
       <li data-md>
-       <p><em>a</em>: an <code class="idl"><a data-link-type="idl" href="#mloperand" id="ref-for-mloperand⑧⑧">MLOperand</a></code>. The first input tensor.</p>
+       <p><strong>"iohw"</strong>: [inputChannels, outputChannels/groups, height, width]</p>
       <li data-md>
-       <p><em>b</em>: an <code class="idl"><a data-link-type="idl" href="#mloperand" id="ref-for-mloperand⑧⑨">MLOperand</a></code>. The second input tensor.</p>
-     </ul>
-     <p><strong>Returns:</strong> an <code class="idl"><a data-link-type="idl" href="#mloperand" id="ref-for-mloperand⑨⓪">MLOperand</a></code>. The output tensor that contains the result of
-    element-wise binary operation of the two input tensors.</p>
-     <p>The element-wise binary operation will be broadcasted according to <a data-link-type="biblio" href="#biblio-numpy-broadcasting-rule" title="General Broadcasting Rules of NumPy">[numpy-broadcasting-rule]</a>. The rank of the output tensor is the maximum
-    rank of the input tensors. For each dimension of the output tensor, its size
-    is the maximum size along that dimension of the input tensors.</p>
-     <p><strong>Operation types:</strong></p>
-     <ul>
+       <p><strong>"hwoi"</strong>: [height, width, outputChannels/groups, inputChannels]</p>
       <li data-md>
-       <p><em>add</em>: Add the values of the two input tensors, element-wise.</p>
+       <p><strong>"ohwi"</strong>: [outputChannels/groups, height, width, inputChannels]</p>
+     </ul>
+     The default value is <em>"iohw"</em>. 
+    <dt data-md><dfn class="dfn-paneled idl-code" data-dfn-for="MLConvTranspose2dOptions" data-dfn-type="dict-member" data-export id="dom-mlconvtranspose2doptions-bias"><code>bias</code></dfn>, <span> of type <a data-link-type="idl-name" href="#mloperand" id="ref-for-mloperand⑦④">MLOperand</a></span>
+    <dd data-md>
+     <p>An <code class="idl"><a data-link-type="idl" href="#mloperand" id="ref-for-mloperand⑦⑤">MLOperand</a></code> object.
+Specifies the additional 1-D tensor with the shape of <em>[outputChannels]</em> whose values are to be added to the convolution result.</p>
+    <dt data-md><dfn class="dfn-paneled idl-code" data-dfn-for="MLConvTranspose2dOptions" data-dfn-type="dict-member" data-export id="dom-mlconvtranspose2doptions-activation"><code>activation</code></dfn>, <span> of type <a data-link-type="idl-name" href="#mlactivation" id="ref-for-mlactivation①⑧">MLActivation</a></span>
+    <dd data-md>
+     <p>An <code class="idl"><a data-link-type="idl" href="#mlactivation" id="ref-for-mlactivation①⑨">MLActivation</a></code> object.
+Specifies the optional activation function that immediately follows the convolution operation.</p>
+   </dl>
+   <div>
+     <strong>Arguments:</strong> 
+    <ul>
+     <li data-md>
+      <p><em>input</em>: an <code class="idl"><a data-link-type="idl" href="#mloperand" id="ref-for-mloperand⑦⑥">MLOperand</a></code>. The input 4-D tensor. The logical shape
+is interpreted according to the value of <em>options</em>.<code class="idl"><a data-link-type="idl" href="#dom-mlconvtranspose2doptions-inputlayout" id="ref-for-dom-mlconvtranspose2doptions-inputlayout①">inputLayout</a></code>.</p>
+     <li data-md>
+      <p><em>filter</em>: an <code class="idl"><a data-link-type="idl" href="#mloperand" id="ref-for-mloperand⑦⑦">MLOperand</a></code>. The filter 4-D tensor. The logical shape is
+interpreted according to the value of <em>options</em>.<code class="idl"><a data-link-type="idl" href="#dom-mlconvtranspose2doptions-filterlayout" id="ref-for-dom-mlconvtranspose2doptions-filterlayout①">filterLayout</a></code> and <code class="idl"><a data-link-type="idl" href="#dom-mlconvtranspose2doptions-groups" id="ref-for-dom-mlconvtranspose2doptions-groups①">groups</a></code>.</p>
+     <li data-md>
+      <p><em>options</em>: an optional <code class="idl"><a data-link-type="idl" href="#dictdef-mlconvtranspose2doptions" id="ref-for-dictdef-mlconvtranspose2doptions②">MLConvTranspose2dOptions</a></code>.</p>
+    </ul>
+    <p><strong>Returns:</strong> an <code class="idl"><a data-link-type="idl" href="#mloperand" id="ref-for-mloperand⑦⑧">MLOperand</a></code>. The output 4-D tensor that contains the transposed convolution result. The output shape is interpreted according to the <em>options</em>.<code class="idl"><a data-link-type="idl" href="#dom-mlconvtranspose2doptions-inputlayout" id="ref-for-dom-mlconvtranspose2doptions-inputlayout②">inputLayout</a></code> value. More specifically, unless the <em>options</em>.<code class="idl"><a data-link-type="idl" href="#dom-mlconvtranspose2doptions-outputsizes" id="ref-for-dom-mlconvtranspose2doptions-outputsizes①">outputSizes</a></code> values are explicitly specified, the <em>options</em>.<code class="idl"><a data-link-type="idl" href="#dom-mlconvtranspose2doptions-outputpadding" id="ref-for-dom-mlconvtranspose2doptions-outputpadding②">outputPadding</a></code> may be needed to compute the spatial dimension values of the output tensor as follow:</p>
+    <p><em>outputSize = (inputSize - 1) *</em> <em>stride + (filterSize - 1) *</em> <em>dilation + 1 - beginningPadding - endingPadding + outputPadding</em></p>
+   </div>
+   <details class="algorithm" data-algorithm="convTranspose2d(input, filter, options)" data-algorithm-for="MLGraphBuilder" open>
+    <summary> The <dfn class="dfn-paneled idl-code" data-dfn-for="MLGraphBuilder" data-dfn-type="method" data-export data-lt="convTranspose2d(input, filter, options)|convTranspose2d(input, filter)" id="dom-mlgraphbuilder-convtranspose2d"><code>convTranspose2d(<var>input</var>, <var>filter</var>, <var>options</var>)</code></dfn> method steps are: </summary>
+    <div class="algorithm-steps">
+     <ol>
       <li data-md>
-       <p><em>sub</em>: Subtract the values of the second input tensor from the values of the first input tensor, element-wise.</p>
+       <p><a data-link-type="dfn" href="https://infra.spec.whatwg.org/#assert" id="ref-for-assert②④">Assert</a>: the type of <var>input</var> and <var>filter</var> is <code class="idl"><a data-link-type="idl" href="#mloperand" id="ref-for-mloperand⑦⑨">MLOperand</a></code>.</p>
       <li data-md>
-       <p><em>mul</em>: Multiply the values of the two input tensors, element-wise.</p>
+       <p>Let <var>inputSize</var> be the <a data-link-type="dfn" href="https://infra.spec.whatwg.org/#list-size" id="ref-for-list-size①⑥">size</a> of <var>input</var>.<code class="idl"><a data-link-type="idl" href="#dom-mloperand-descriptor-slot" id="ref-for-dom-mloperand-descriptor-slot③⓪">[[descriptor]]</a></code>.<code class="idl"><a data-link-type="idl" href="#dom-mloperanddescriptor-dimensions" id="ref-for-dom-mloperanddescriptor-dimensions③③">dimensions</a></code>.</p>
       <li data-md>
-       <p><em>div</em>: Divide the values of the first input tensor with the values of the second tensor, element-wise.</p>
+       <p>Let <var>filterSize</var> be the <a data-link-type="dfn" href="https://infra.spec.whatwg.org/#list-size" id="ref-for-list-size①⑦">size</a> of <var>filter</var>.<code class="idl"><a data-link-type="idl" href="#dom-mloperand-descriptor-slot" id="ref-for-dom-mloperand-descriptor-slot③①">[[descriptor]]</a></code>.<code class="idl"><a data-link-type="idl" href="#dom-mloperanddescriptor-dimensions" id="ref-for-dom-mloperanddescriptor-dimensions③④">dimensions</a></code>.</p>
       <li data-md>
-       <p><em>max</em>: Select the greater values of the two input tensors, element-wise.</p>
+       <p>If <var>inputSize</var> is not 4, then <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#dfn-throw" id="ref-for-dfn-throw⑦③">throw</a> a "<code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#dataerror" id="ref-for-dataerror②③">DataError</a></code>" <code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#idl-DOMException" id="ref-for-idl-DOMException⑤⓪">DOMException</a></code>.</p>
       <li data-md>
-       <p><em>min</em>: Select the lesser values of the two input tensors, element-wise.</p>
+       <p>If <var>filterSize</var> is not 4, then <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#dfn-throw" id="ref-for-dfn-throw⑦④">throw</a> a "<code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#dataerror" id="ref-for-dataerror②④">DataError</a></code>" <code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#idl-DOMException" id="ref-for-idl-DOMException⑤①">DOMException</a></code>.</p>
       <li data-md>
-       <p><em>pow</em>: Compute the values of the values of the first input tensor to the power of the values of the second input tensor, element-wise.</p>
-     </ul>
-    </div>
-    <h4 class="heading settled" data-level="7.6.10" id="api-mlgraphbuilder-unary"><span class="secno">7.6.10. </span><span class="content">Element-wise unary operations</span><a class="self-link" href="#api-mlgraphbuilder-unary"></a></h4>
-     Compute the element-wise unary operation for input tensor. 
-<pre class="idl highlight def"><c- b>partial</c-> <c- b>interface</c-> <a class="idl-code" data-link-type="interface" href="#mlgraphbuilder" id="ref-for-mlgraphbuilder②⑥"><c- g>MLGraphBuilder</c-></a> {
-  <a data-link-type="idl-name" href="#mloperand" id="ref-for-mloperand⑨①"><c- n>MLOperand</c-></a> <dfn class="idl-code" data-dfn-for="MLGraphBuilder" data-dfn-type="method" data-export data-lt="abs(x)" id="dom-mlgraphbuilder-abs"><code><c- g>abs</c-></code><a class="self-link" href="#dom-mlgraphbuilder-abs"></a></dfn>(<a data-link-type="idl-name" href="#mloperand" id="ref-for-mloperand⑨②"><c- n>MLOperand</c-></a> <dfn class="idl-code" data-dfn-for="MLGraphBuilder/abs(x)" data-dfn-type="argument" data-export id="dom-mlgraphbuilder-abs-x-x"><code><c- g>x</c-></code><a class="self-link" href="#dom-mlgraphbuilder-abs-x-x"></a></dfn>);
-  <a data-link-type="idl-name" href="#mloperand" id="ref-for-mloperand⑨③"><c- n>MLOperand</c-></a> <dfn class="idl-code" data-dfn-for="MLGraphBuilder" data-dfn-type="method" data-export data-lt="ceil(x)" id="dom-mlgraphbuilder-ceil"><code><c- g>ceil</c-></code><a class="self-link" href="#dom-mlgraphbuilder-ceil"></a></dfn>(<a data-link-type="idl-name" href="#mloperand" id="ref-for-mloperand⑨④"><c- n>MLOperand</c-></a> <dfn class="idl-code" data-dfn-for="MLGraphBuilder/ceil(x)" data-dfn-type="argument" data-export id="dom-mlgraphbuilder-ceil-x-x"><code><c- g>x</c-></code><a class="self-link" href="#dom-mlgraphbuilder-ceil-x-x"></a></dfn>);
-  <a data-link-type="idl-name" href="#mloperand" id="ref-for-mloperand⑨⑤"><c- n>MLOperand</c-></a> <dfn class="idl-code" data-dfn-for="MLGraphBuilder" data-dfn-type="method" data-export data-lt="cos(x)" id="dom-mlgraphbuilder-cos"><code><c- g>cos</c-></code><a class="self-link" href="#dom-mlgraphbuilder-cos"></a></dfn>(<a data-link-type="idl-name" href="#mloperand" id="ref-for-mloperand⑨⑥"><c- n>MLOperand</c-></a> <dfn class="idl-code" data-dfn-for="MLGraphBuilder/cos(x)" data-dfn-type="argument" data-export id="dom-mlgraphbuilder-cos-x-x"><code><c- g>x</c-></code><a class="self-link" href="#dom-mlgraphbuilder-cos-x-x"></a></dfn>);
-  <a data-link-type="idl-name" href="#mloperand" id="ref-for-mloperand⑨⑦"><c- n>MLOperand</c-></a> <dfn class="idl-code" data-dfn-for="MLGraphBuilder" data-dfn-type="method" data-export data-lt="exp(x)" id="dom-mlgraphbuilder-exp"><code><c- g>exp</c-></code><a class="self-link" href="#dom-mlgraphbuilder-exp"></a></dfn>(<a data-link-type="idl-name" href="#mloperand" id="ref-for-mloperand⑨⑧"><c- n>MLOperand</c-></a> <dfn class="idl-code" data-dfn-for="MLGraphBuilder/exp(x)" data-dfn-type="argument" data-export id="dom-mlgraphbuilder-exp-x-x"><code><c- g>x</c-></code><a class="self-link" href="#dom-mlgraphbuilder-exp-x-x"></a></dfn>);
-  <a data-link-type="idl-name" href="#mloperand" id="ref-for-mloperand⑨⑨"><c- n>MLOperand</c-></a> <dfn class="idl-code" data-dfn-for="MLGraphBuilder" data-dfn-type="method" data-export data-lt="floor(x)" id="dom-mlgraphbuilder-floor"><code><c- g>floor</c-></code><a class="self-link" href="#dom-mlgraphbuilder-floor"></a></dfn>(<a data-link-type="idl-name" href="#mloperand" id="ref-for-mloperand①⓪⓪"><c- n>MLOperand</c-></a> <dfn class="idl-code" data-dfn-for="MLGraphBuilder/floor(x)" data-dfn-type="argument" data-export id="dom-mlgraphbuilder-floor-x-x"><code><c- g>x</c-></code><a class="self-link" href="#dom-mlgraphbuilder-floor-x-x"></a></dfn>);
-  <a data-link-type="idl-name" href="#mloperand" id="ref-for-mloperand①⓪①"><c- n>MLOperand</c-></a> <dfn class="idl-code" data-dfn-for="MLGraphBuilder" data-dfn-type="method" data-export data-lt="log(x)" id="dom-mlgraphbuilder-log"><code><c- g>log</c-></code><a class="self-link" href="#dom-mlgraphbuilder-log"></a></dfn>(<a data-link-type="idl-name" href="#mloperand" id="ref-for-mloperand①⓪②"><c- n>MLOperand</c-></a> <dfn class="idl-code" data-dfn-for="MLGraphBuilder/log(x)" data-dfn-type="argument" data-export id="dom-mlgraphbuilder-log-x-x"><code><c- g>x</c-></code><a class="self-link" href="#dom-mlgraphbuilder-log-x-x"></a></dfn>);
-  <a data-link-type="idl-name" href="#mloperand" id="ref-for-mloperand①⓪③"><c- n>MLOperand</c-></a> <dfn class="idl-code" data-dfn-for="MLGraphBuilder" data-dfn-type="method" data-export data-lt="neg(x)" id="dom-mlgraphbuilder-neg"><code><c- g>neg</c-></code><a class="self-link" href="#dom-mlgraphbuilder-neg"></a></dfn>(<a data-link-type="idl-name" href="#mloperand" id="ref-for-mloperand①⓪④"><c- n>MLOperand</c-></a> <dfn class="idl-code" data-dfn-for="MLGraphBuilder/neg(x)" data-dfn-type="argument" data-export id="dom-mlgraphbuilder-neg-x-x"><code><c- g>x</c-></code><a class="self-link" href="#dom-mlgraphbuilder-neg-x-x"></a></dfn>);
-  <a data-link-type="idl-name" href="#mloperand" id="ref-for-mloperand①⓪⑤"><c- n>MLOperand</c-></a> <dfn class="idl-code" data-dfn-for="MLGraphBuilder" data-dfn-type="method" data-export data-lt="sin(x)" id="dom-mlgraphbuilder-sin"><code><c- g>sin</c-></code><a class="self-link" href="#dom-mlgraphbuilder-sin"></a></dfn>(<a data-link-type="idl-name" href="#mloperand" id="ref-for-mloperand①⓪⑥"><c- n>MLOperand</c-></a> <dfn class="idl-code" data-dfn-for="MLGraphBuilder/sin(x)" data-dfn-type="argument" data-export id="dom-mlgraphbuilder-sin-x-x"><code><c- g>x</c-></code><a class="self-link" href="#dom-mlgraphbuilder-sin-x-x"></a></dfn>);
-  <a data-link-type="idl-name" href="#mloperand" id="ref-for-mloperand①⓪⑦"><c- n>MLOperand</c-></a> <dfn class="idl-code" data-dfn-for="MLGraphBuilder" data-dfn-type="method" data-export data-lt="tan(x)" id="dom-mlgraphbuilder-tan"><code><c- g>tan</c-></code><a class="self-link" href="#dom-mlgraphbuilder-tan"></a></dfn>(<a data-link-type="idl-name" href="#mloperand" id="ref-for-mloperand①⓪⑧"><c- n>MLOperand</c-></a> <dfn class="idl-code" data-dfn-for="MLGraphBuilder/tan(x)" data-dfn-type="argument" data-export id="dom-mlgraphbuilder-tan-x-x"><code><c- g>x</c-></code><a class="self-link" href="#dom-mlgraphbuilder-tan-x-x"></a></dfn>);
-};
-</pre>
-    <div class="algorithm" data-algorithm="unary">
-      <strong>Arguments:</strong> 
-     <ul>
+       <p>If <var>input</var>.<code class="idl"><a data-link-type="idl" href="#dom-mloperand-descriptor-slot" id="ref-for-dom-mloperand-descriptor-slot③②">[[descriptor]]</a></code>.<code class="idl"><a data-link-type="idl" href="#dom-mloperanddescriptor-type" id="ref-for-dom-mloperanddescriptor-type①⑤">type</a></code> is not the same as <code class="idl"><a data-link-type="idl" href="#dom-mloperand-descriptor-slot" id="ref-for-dom-mloperand-descriptor-slot③③">[[descriptor]]</a></code>.<code class="idl"><a data-link-type="idl" href="#dom-mloperanddescriptor-type" id="ref-for-dom-mloperanddescriptor-type①⑥">type</a></code>, then <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#dfn-throw" id="ref-for-dfn-throw⑦⑤">throw</a> a <code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#exceptiondef-typeerror" id="ref-for-exceptiondef-typeerror②⑤">TypeError</a></code>.</p>
       <li data-md>
-       <p><em>x</em>: an <code class="idl"><a data-link-type="idl" href="#mloperand" id="ref-for-mloperand①⓪⑨">MLOperand</a></code>. The input tensor.</p>
-     </ul>
-     <p><strong>Returns:</strong> an <code class="idl"><a data-link-type="idl" href="#mloperand" id="ref-for-mloperand①①⓪">MLOperand</a></code>. The output tensor that contains the result of
-    element-wise unary operation of the input tensor. The shape of the output
-    tensor is the same as the shape of input tensor.</p>
-     <p><strong>Operation types:</strong></p>
-     <ul>
+       <p>If <var>options</var>.<code class="idl"><a data-link-type="idl" href="#dom-mlconvtranspose2doptions-padding" id="ref-for-dom-mlconvtranspose2doptions-padding③">padding</a></code> does not <a data-link-type="dfn" href="https://infra.spec.whatwg.org/#map-exists" id="ref-for-map-exists①⑨">exist</a>, set it to <code>« 0, 0, 0, 0 »</code>.</p>
       <li data-md>
-       <p><em>abs</em>: Compute the absolute value of the input tensor, element-wise.</p>
+       <p>Else if the <a data-link-type="dfn" href="https://infra.spec.whatwg.org/#list-size" id="ref-for-list-size①⑧">size</a> of <var>options</var>.<code class="idl"><a data-link-type="idl" href="#dom-mlconvtranspose2doptions-padding" id="ref-for-dom-mlconvtranspose2doptions-padding④">padding</a></code> is not 4, then <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#dfn-throw" id="ref-for-dfn-throw⑦⑥">throw</a> a "<code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#dataerror" id="ref-for-dataerror②⑤">DataError</a></code>" <code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#idl-DOMException" id="ref-for-idl-DOMException⑤②">DOMException</a></code>.</p>
       <li data-md>
-       <p><em>ceil</em>: Compute the ceiling of the input tensor, element-wise.</p>
+       <p>If <var>options</var>.<code class="idl"><a data-link-type="idl" href="#dom-mlconvtranspose2doptions-strides" id="ref-for-dom-mlconvtranspose2doptions-strides②">strides</a></code> does not <a data-link-type="dfn" href="https://infra.spec.whatwg.org/#map-exists" id="ref-for-map-exists②⓪">exist</a>, set it to <code>« 1, 1 »</code>.</p>
       <li data-md>
-       <p><em>cos</em>: Compute the cosine of the input tensor, element-wise.</p>
+       <p>Else if the <a data-link-type="dfn" href="https://infra.spec.whatwg.org/#list-size" id="ref-for-list-size①⑨">size</a> of <var>options</var>.<code class="idl"><a data-link-type="idl" href="#dom-mlconvtranspose2doptions-strides" id="ref-for-dom-mlconvtranspose2doptions-strides③">strides</a></code> is not 2, then <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#dfn-throw" id="ref-for-dfn-throw⑦⑦">throw</a> a <code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#exceptiondef-typeerror" id="ref-for-exceptiondef-typeerror②⑥">TypeError</a></code>.</p>
       <li data-md>
-       <p><em>exp</em>: Compute the exponential of the input tensor, element-wise.</p>
+       <p>If any element in <var>options</var>.<code class="idl"><a data-link-type="idl" href="#dom-mlconv2doptions-strides" id="ref-for-dom-mlconv2doptions-strides④">strides</a></code> is equal to 0, then <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#dfn-throw" id="ref-for-dfn-throw⑦⑧">throw</a> a <code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#exceptiondef-typeerror" id="ref-for-exceptiondef-typeerror②⑦">TypeError</a></code>.</p>
       <li data-md>
-       <p><em>floor</em>: Compute the floor of the input tensor, element-wise.</p>
+       <p>If <var>options</var>.<code class="idl"><a data-link-type="idl" href="#dom-mlconvtranspose2doptions-dilations" id="ref-for-dom-mlconvtranspose2doptions-dilations①">dilations</a></code> does not <a data-link-type="dfn" href="https://infra.spec.whatwg.org/#map-exists" id="ref-for-map-exists②①">exist</a>, set it to <code>« 1, 1 »</code>.</p>
       <li data-md>
-       <p><em>log</em>: Compute the natural logarithm of the input tensor, element-wise.</p>
+       <p>Else if the <a data-link-type="dfn" href="https://infra.spec.whatwg.org/#list-size" id="ref-for-list-size②⓪">size</a> of <var>options</var>.<code class="idl"><a data-link-type="idl" href="#dom-mlconvtranspose2doptions-dilations" id="ref-for-dom-mlconvtranspose2doptions-dilations②">dilations</a></code> is not 2, then <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#dfn-throw" id="ref-for-dfn-throw⑦⑨">throw</a> a <code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#exceptiondef-typeerror" id="ref-for-exceptiondef-typeerror②⑧">TypeError</a></code>.</p>
       <li data-md>
-       <p><em>neg</em>: Compute the numerical negative value of the input tensor, element-wise.</p>
+       <p>If <var>options</var>.<code class="idl"><a data-link-type="idl" href="#dom-mlconvtranspose2doptions-outputpadding" id="ref-for-dom-mlconvtranspose2doptions-outputpadding③">outputPadding</a></code> does not <a data-link-type="dfn" href="https://infra.spec.whatwg.org/#map-exists" id="ref-for-map-exists②②">exist</a>, set it to <code>« 0, 0 »</code>.</p>
       <li data-md>
-       <p><em>sin</em>: Compute the sine of the input tensor, element-wise.</p>
+       <p>Else if the <a data-link-type="dfn" href="https://infra.spec.whatwg.org/#list-size" id="ref-for-list-size②①">size</a> of <var>options</var>.<code class="idl"><a data-link-type="idl" href="#dom-mlconvtranspose2doptions-outputpadding" id="ref-for-dom-mlconvtranspose2doptions-outputpadding④">outputPadding</a></code> is not 2, then <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#dfn-throw" id="ref-for-dfn-throw⑧⓪">throw</a> a <code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#exceptiondef-typeerror" id="ref-for-exceptiondef-typeerror②⑨">TypeError</a></code>.</p>
       <li data-md>
-       <p><em>tan</em>: Compute the tangent of the input tensor, element-wise.</p>
-     </ul>
+       <p>If <var>options</var>.<code class="idl"><a data-link-type="idl" href="#dom-mlconvtranspose2doptions-outputsizes" id="ref-for-dom-mlconvtranspose2doptions-outputsizes②">outputSizes</a></code> <a data-link-type="dfn" href="https://infra.spec.whatwg.org/#map-exists" id="ref-for-map-exists②③">exists</a>:</p>
+       <ol>
+        <li data-md>
+         <p>If the <a data-link-type="dfn" href="https://infra.spec.whatwg.org/#list-size" id="ref-for-list-size②②">size</a> of <var>options</var>.<code class="idl"><a data-link-type="idl" href="#dom-mlconvtranspose2doptions-outputsizes" id="ref-for-dom-mlconvtranspose2doptions-outputsizes③">outputSizes</a></code> is not 2, then <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#dfn-throw" id="ref-for-dfn-throw⑧①">throw</a> a <code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#exceptiondef-typeerror" id="ref-for-exceptiondef-typeerror③⓪">TypeError</a></code>.</p>
+        <li data-md>
+         <p>If the elements of <var>options</var>.<code class="idl"><a data-link-type="idl" href="#dom-mlconvtranspose2doptions-outputsizes" id="ref-for-dom-mlconvtranspose2doptions-outputsizes④">outputSizes</a></code> are not smaller than the elements at the same dimension (index) for <var>options</var>.<code class="idl"><a data-link-type="idl" href="#dom-mlconvtranspose2doptions-strides" id="ref-for-dom-mlconvtranspose2doptions-strides④">strides</a></code>, then <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#dfn-throw" id="ref-for-dfn-throw⑧②">throw</a> a "<code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#dataerror" id="ref-for-dataerror②⑥">DataError</a></code>" <code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#idl-DOMException" id="ref-for-idl-DOMException⑤③">DOMException</a></code>.</p>
+       </ol>
+      <li data-md>
+       <p>If <var>inputSize</var> / <var>options</var>.<code class="idl"><a data-link-type="idl" href="#dom-mlconvtranspose2doptions-groups" id="ref-for-dom-mlconvtranspose2doptions-groups②">groups</a></code> is not equal to <var>filterSize</var>, then <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#dfn-throw" id="ref-for-dfn-throw⑧③">throw</a> a "<code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#dataerror" id="ref-for-dataerror②⑦">DataError</a></code>" <code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#idl-DOMException" id="ref-for-idl-DOMException⑤④">DOMException</a></code>.</p>
+      <li data-md>
+       <p>Else if <var>inputSize</var> % <var>options</var>.<code class="idl"><a data-link-type="idl" href="#dom-mlconvtranspose2doptions-groups" id="ref-for-dom-mlconvtranspose2doptions-groups③">groups</a></code> is not 0, then <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#dfn-throw" id="ref-for-dfn-throw⑧④">throw</a> a "<code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#dataerror" id="ref-for-dataerror②⑧">DataError</a></code>" <code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#idl-DOMException" id="ref-for-idl-DOMException⑤⑤">DOMException</a></code>.</p>
+      <li data-md>
+       <p>If <var>options</var>.<code class="idl"><a data-link-type="idl" href="#dom-mlconvtranspose2doptions-bias" id="ref-for-dom-mlconvtranspose2doptions-bias①">bias</a></code> <a data-link-type="dfn" href="https://infra.spec.whatwg.org/#map-exists" id="ref-for-map-exists②④">exists</a>:</p>
+       <ol>
+        <li data-md>
+         <p><a data-link-type="dfn" href="https://infra.spec.whatwg.org/#assert" id="ref-for-assert②⑤">Assert</a>: the type of <var>options</var>.<code class="idl"><a data-link-type="idl" href="#dom-mlconvtranspose2doptions-bias" id="ref-for-dom-mlconvtranspose2doptions-bias②">bias</a></code> is <code class="idl"><a data-link-type="idl" href="#mloperand" id="ref-for-mloperand⑧⓪">MLOperand</a></code>.</p>
+       </ol>
+      <li data-md>
+       <p>If the <a data-link-type="dfn" href="https://infra.spec.whatwg.org/#list-size" id="ref-for-list-size②③">size</a> of <var>options</var>.<code class="idl"><a data-link-type="idl" href="#dom-mlconvtranspose2doptions-bias" id="ref-for-dom-mlconvtranspose2doptions-bias③">bias</a></code>.<code class="idl"><a data-link-type="idl" href="#dom-mloperand-descriptor-slot" id="ref-for-dom-mloperand-descriptor-slot③④">[[descriptor]]</a></code>.<code class="idl"><a data-link-type="idl" href="#dom-mloperanddescriptor-dimensions" id="ref-for-dom-mloperanddescriptor-dimensions③⑤">dimensions</a></code> is not 1, then <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#dfn-throw" id="ref-for-dfn-throw⑧⑤">throw</a> a <code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#exceptiondef-typeerror" id="ref-for-exceptiondef-typeerror③①">TypeError</a></code>.</p>
+      <li data-md>
+       <p>If <var>options</var>.<code class="idl"><a data-link-type="idl" href="#dom-mlconvtranspose2doptions-bias" id="ref-for-dom-mlconvtranspose2doptions-bias④">bias</a></code>.<code class="idl"><a data-link-type="idl" href="#dom-mloperand-descriptor-slot" id="ref-for-dom-mloperand-descriptor-slot③⑤">[[descriptor]]</a></code>.<code class="idl"><a data-link-type="idl" href="#dom-mloperanddescriptor-type" id="ref-for-dom-mloperanddescriptor-type①⑦">type</a></code> is not the same as <var>input</var>.<code class="idl"><a data-link-type="idl" href="#dom-mloperand-descriptor-slot" id="ref-for-dom-mloperand-descriptor-slot③⑥">[[descriptor]]</a></code>.<code class="idl"><a data-link-type="idl" href="#dom-mloperanddescriptor-type" id="ref-for-dom-mloperanddescriptor-type①⑧">type</a></code>, then <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#dfn-throw" id="ref-for-dfn-throw⑧⑥">throw</a> a <code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#exceptiondef-typeerror" id="ref-for-exceptiondef-typeerror③②">TypeError</a></code>.</p>
+      <li data-md>
+       <p>If <var>options</var>.<code class="idl"><a data-link-type="idl" href="#dom-mlconvtranspose2doptions-activation" id="ref-for-dom-mlconvtranspose2doptions-activation①">activation</a></code> <a data-link-type="dfn" href="https://infra.spec.whatwg.org/#map-exists" id="ref-for-map-exists②⑤">exists</a>:</p>
+       <ol>
+        <li data-md>
+         <p><a data-link-type="dfn" href="https://infra.spec.whatwg.org/#assert" id="ref-for-assert②⑥">Assert</a>: the type of <var>options</var>.<code class="idl"><a data-link-type="idl" href="#dom-mlconvtranspose2doptions-activation" id="ref-for-dom-mlconvtranspose2doptions-activation②">activation</a></code> is <code class="idl"><a data-link-type="idl" href="#mlactivation" id="ref-for-mlactivation②⓪">MLActivation</a></code>.</p>
+       </ol>
+      <li data-md>
+       <p>Let <var>outputShape</var> be the result of invoking the underlying implementation for calculating output dimensions, given <var>options</var>.</p>
+      <li data-md>
+       <p>If <var>outputShape</var> is not the same as the shape of <var>options</var>.<code class="idl"><a data-link-type="idl" href="#dom-mlconvtranspose2doptions-bias" id="ref-for-dom-mlconvtranspose2doptions-bias⑤">bias</a></code>.<code class="idl"><a data-link-type="idl" href="#dom-mloperand-descriptor-slot" id="ref-for-dom-mloperand-descriptor-slot③⑦">[[descriptor]]</a></code>.<code class="idl"><a data-link-type="idl" href="#dom-mloperanddescriptor-dimensions" id="ref-for-dom-mloperanddescriptor-dimensions③⑥">dimensions</a></code>, then <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#dfn-throw" id="ref-for-dfn-throw⑧⑦">throw</a> a "<code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#dataerror" id="ref-for-dataerror②⑨">DataError</a></code>" <code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#idl-DOMException" id="ref-for-idl-DOMException⑤⑥">DOMException</a></code>.</p>
+      <li data-md>
+       <p>Let <var>desc</var> a new <code class="idl"><a data-link-type="idl" href="#dictdef-mloperanddescriptor" id="ref-for-dictdef-mloperanddescriptor①⑤">MLOperandDescriptor</a></code>.</p>
+      <li data-md>
+       <p>Set <var>desc</var>.<code class="idl"><a data-link-type="idl" href="#dom-mloperanddescriptor-type" id="ref-for-dom-mloperanddescriptor-type①⑨">type</a></code> to <var>input</var>.<code class="idl"><a data-link-type="idl" href="#dom-mloperand-descriptor-slot" id="ref-for-dom-mloperand-descriptor-slot③⑧">[[descriptor]]</a></code>.<code class="idl"><a data-link-type="idl" href="#dom-mloperanddescriptor-type" id="ref-for-dom-mloperanddescriptor-type②⓪">type</a></code>.</p>
+      <li data-md>
+       <p>Set <var>desc</var>.<code class="idl"><a data-link-type="idl" href="#dom-mloperanddescriptor-dimensions" id="ref-for-dom-mloperanddescriptor-dimensions③⑦">dimensions</a></code> to <var>outputShape</var>.</p>
+      <li data-md>
+       <p>If any of the following sub-steps fail, <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#dfn-throw" id="ref-for-dfn-throw⑧⑧">throw</a> an "<code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#operationerror" id="ref-for-operationerror①⑧">OperationError</a></code>" <code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#idl-DOMException" id="ref-for-idl-DOMException⑤⑦">DOMException</a></code>.</p>
+       <ol>
+        <li data-md>
+         <p>Let <var>output</var> be the result of <a data-link-type="dfn" href="#create-an-mloperand" id="ref-for-create-an-mloperand⑥">creating an MLOperand</a> given <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#this" id="ref-for-this①⑤">this</a> and <var>desc</var>.</p>
+        <li data-md>
+         <p>Make a request to the underlying platform to:</p>
+         <ol>
+          <li data-md>
+           <p>Create an <a data-link-type="dfn" href="https://infra.spec.whatwg.org/#implementation-defined" id="ref-for-implementation-defined①⑥">implementation-defined</a> platform operator <var>convTranspose2dImpl</var> for this method, given <var>options</var> and <var>filter</var>.</p>
+           <ol>
+            <li data-md>
+             <p>If <var>options</var>.<code class="idl"><a data-link-type="idl" href="#dom-mlconvtranspose2doptions-activation" id="ref-for-dom-mlconvtranspose2doptions-activation③">activation</a></code> <a data-link-type="dfn" href="https://infra.spec.whatwg.org/#map-exists" id="ref-for-map-exists②⑥">exists</a>,register it as activation to <var>convTranspose2dImpl</var>.</p>
+           </ol>
+          <li data-md>
+           <p>Store a reference of <var>convTranspose2dImpl</var> in <var>output</var>.<code class="idl"><a data-link-type="idl" href="#dom-mloperand-operator-slot" id="ref-for-dom-mloperand-operator-slot④">[[operator]]</a></code>.</p>
+          <li data-md>
+           <p>Create an <a data-link-type="dfn" href="https://infra.spec.whatwg.org/#implementation-defined" id="ref-for-implementation-defined①⑦">implementation-defined</a> platform operand <var>outputImpl</var> to represent the output, given <var>output</var> and <var>convTranspose2dImpl</var>.</p>
+          <li data-md>
+           <p>Store a reference to <var>outputImpl</var> in <var>output</var>.<code class="idl"><a data-link-type="idl" href="#dom-mloperand-operand-slot" id="ref-for-dom-mloperand-operand-slot①②">[[operand]]</a></code>.</p>
+         </ol>
+        <li data-md>
+         <p>Connect <var>input</var>.<code class="idl"><a data-link-type="idl" href="#dom-mloperand-operand-slot" id="ref-for-dom-mloperand-operand-slot①③">[[operand]]</a></code> as input to <var>convTranspose2dImpl</var>.</p>
+        <li data-md>
+         <p>Connect <var>output</var>.<code class="idl"><a data-link-type="idl" href="#dom-mloperand-operand-slot" id="ref-for-dom-mloperand-operand-slot①④">[[operand]]</a></code> as output to <var>convTranspose2dImpl</var>.</p>
+       </ol>
+      <li data-md>
+       <p>Return <var>output</var>.</p>
+     </ol>
     </div>
-    <h4 class="heading settled" data-level="7.6.11" id="api-mlgraphbuilder-elu"><span class="secno">7.6.11. </span><span class="content">The elu() method</span><a class="self-link" href="#api-mlgraphbuilder-elu"></a></h4>
-     Calculate the <a href="https://en.wikipedia.org/wiki/Rectifier_(neural_networks)#ELU"> exponential linear unit function</a> on the input tensor element-wise. The calculation follows the expression <code>max(0, x) + alpha * (exp(min(0, x)) - 1)</code>. 
-<pre class="idl highlight def"><c- b>dictionary</c-> <dfn class="dfn-paneled idl-code" data-dfn-type="dictionary" data-export id="dictdef-mleluoptions"><code><c- g>MLEluOptions</c-></code></dfn> {
-  <a class="idl-code" data-link-type="interface" href="https://webidl.spec.whatwg.org/#idl-float" id="ref-for-idl-float⑥"><c- b>float</c-></a> <dfn class="idl-code" data-default="1" data-dfn-for="MLEluOptions" data-dfn-type="dict-member" data-export data-type="float " id="dom-mleluoptions-alpha"><code><c- g>alpha</c-></code><a class="self-link" href="#dom-mleluoptions-alpha"></a></dfn> = 1;
-};
-
-<c- b>partial</c-> <c- b>interface</c-> <a class="idl-code" data-link-type="interface" href="#mlgraphbuilder" id="ref-for-mlgraphbuilder②⑦"><c- g>MLGraphBuilder</c-></a> {
-  <a data-link-type="idl-name" href="#mloperand" id="ref-for-mloperand①①①"><c- n>MLOperand</c-></a> <dfn class="idl-code" data-dfn-for="MLGraphBuilder" data-dfn-type="method" data-export data-lt="elu(x, options)|elu(x)" id="dom-mlgraphbuilder-elu"><code><c- g>elu</c-></code><a class="self-link" href="#dom-mlgraphbuilder-elu"></a></dfn>(<a data-link-type="idl-name" href="#mloperand" id="ref-for-mloperand①①②"><c- n>MLOperand</c-></a> <dfn class="idl-code" data-dfn-for="MLGraphBuilder/elu(x, options), MLGraphBuilder/elu(x)" data-dfn-type="argument" data-export id="dom-mlgraphbuilder-elu-x-options-x"><code><c- g>x</c-></code><a class="self-link" href="#dom-mlgraphbuilder-elu-x-options-x"></a></dfn>, <c- b>optional</c-> <a data-link-type="idl-name" href="#dictdef-mleluoptions" id="ref-for-dictdef-mleluoptions"><c- n>MLEluOptions</c-></a> <dfn class="idl-code" data-dfn-for="MLGraphBuilder/elu(x, options), MLGraphBuilder/elu(x)" data-dfn-type="argument" data-export id="dom-mlgraphbuilder-elu-x-options-options"><code><c- g>options</c-></code><a class="self-link" href="#dom-mlgraphbuilder-elu-x-options-options"></a></dfn> = {});
-  <a data-link-type="idl-name" href="#mlactivation" id="ref-for-mlactivation①⓪"><c- n>MLActivation</c-></a> <dfn class="idl-code" data-dfn-for="MLGraphBuilder" data-dfn-type="method" data-export data-lt="elu(options)|elu()" id="dom-mlgraphbuilder-elu-options"><code><c- g>elu</c-></code><a class="self-link" href="#dom-mlgraphbuilder-elu-options"></a></dfn>(<c- b>optional</c-> <a data-link-type="idl-name" href="#dictdef-mleluoptions" id="ref-for-dictdef-mleluoptions①"><c- n>MLEluOptions</c-></a> <dfn class="idl-code" data-dfn-for="MLGraphBuilder/elu(options), MLGraphBuilder/elu()" data-dfn-type="argument" data-export id="dom-mlgraphbuilder-elu-options-options"><code><c- g>options</c-></code><a class="self-link" href="#dom-mlgraphbuilder-elu-options-options"></a></dfn> = {});
+   </details>
+   <h4 class="heading settled" data-level="7.6.10" id="api-mlgraphbuilder-binary"><span class="secno">7.6.10. </span><span class="content">Element-wise binary operations</span><a class="self-link" href="#api-mlgraphbuilder-binary"></a></h4>
+    Compute the element-wise binary addition, subtraction, multiplication, division, power, maximum and minimum of the two input tensors. 
+   <p>The element-wise binary operations will be broadcasted according to <a data-link-type="biblio" href="#biblio-numpy-broadcasting-rule" title="General Broadcasting Rules of NumPy">[numpy-broadcasting-rule]</a>. The <a data-link-type="dfn" href="#mloperand-rank" id="ref-for-mloperand-rank⑥">rank</a> of the output tensor is the maximum <a data-link-type="dfn" href="#mloperand-rank" id="ref-for-mloperand-rank⑦">rank</a> of the input tensors. For each dimension of the output tensor, its size
+is the maximum size along that dimension of the input tensors.</p>
+<pre class="idl highlight def"><c- b>partial</c-> <c- b>interface</c-> <a class="idl-code" data-link-type="interface" href="#mlgraphbuilder" id="ref-for-mlgraphbuilder②⑧"><c- g>MLGraphBuilder</c-></a> {
+  <a data-link-type="idl-name" href="#mloperand" id="ref-for-mloperand⑧①"><c- n>MLOperand</c-></a> <a class="idl-code" data-link-type="method" href="#dom-mlgraphbuilder-add" id="ref-for-dom-mlgraphbuilder-add"><c- g>add</c-></a>(<a data-link-type="idl-name" href="#mloperand" id="ref-for-mloperand⑧②"><c- n>MLOperand</c-></a> <dfn class="idl-code" data-dfn-for="MLGraphBuilder/add(a, b)" data-dfn-type="argument" data-export id="dom-mlgraphbuilder-add-a-b-a"><code><c- g>a</c-></code><a class="self-link" href="#dom-mlgraphbuilder-add-a-b-a"></a></dfn>, <a data-link-type="idl-name" href="#mloperand" id="ref-for-mloperand⑧③"><c- n>MLOperand</c-></a> <dfn class="idl-code" data-dfn-for="MLGraphBuilder/add(a, b)" data-dfn-type="argument" data-export id="dom-mlgraphbuilder-add-a-b-b"><code><c- g>b</c-></code><a class="self-link" href="#dom-mlgraphbuilder-add-a-b-b"></a></dfn>);
+  <a data-link-type="idl-name" href="#mloperand" id="ref-for-mloperand⑧④"><c- n>MLOperand</c-></a> <a class="idl-code" data-link-type="method" href="#dom-mlgraphbuilder-sub" id="ref-for-dom-mlgraphbuilder-sub"><c- g>sub</c-></a>(<a data-link-type="idl-name" href="#mloperand" id="ref-for-mloperand⑧⑤"><c- n>MLOperand</c-></a> <dfn class="idl-code" data-dfn-for="MLGraphBuilder/sub(a, b)" data-dfn-type="argument" data-export id="dom-mlgraphbuilder-sub-a-b-a"><code><c- g>a</c-></code><a class="self-link" href="#dom-mlgraphbuilder-sub-a-b-a"></a></dfn>, <a data-link-type="idl-name" href="#mloperand" id="ref-for-mloperand⑧⑥"><c- n>MLOperand</c-></a> <dfn class="idl-code" data-dfn-for="MLGraphBuilder/sub(a, b)" data-dfn-type="argument" data-export id="dom-mlgraphbuilder-sub-a-b-b"><code><c- g>b</c-></code><a class="self-link" href="#dom-mlgraphbuilder-sub-a-b-b"></a></dfn>);
+  <a data-link-type="idl-name" href="#mloperand" id="ref-for-mloperand⑧⑦"><c- n>MLOperand</c-></a> <a class="idl-code" data-link-type="method" href="#dom-mlgraphbuilder-mul" id="ref-for-dom-mlgraphbuilder-mul"><c- g>mul</c-></a>(<a data-link-type="idl-name" href="#mloperand" id="ref-for-mloperand⑧⑧"><c- n>MLOperand</c-></a> <dfn class="idl-code" data-dfn-for="MLGraphBuilder/mul(a, b)" data-dfn-type="argument" data-export id="dom-mlgraphbuilder-mul-a-b-a"><code><c- g>a</c-></code><a class="self-link" href="#dom-mlgraphbuilder-mul-a-b-a"></a></dfn>, <a data-link-type="idl-name" href="#mloperand" id="ref-for-mloperand⑧⑨"><c- n>MLOperand</c-></a> <dfn class="idl-code" data-dfn-for="MLGraphBuilder/mul(a, b)" data-dfn-type="argument" data-export id="dom-mlgraphbuilder-mul-a-b-b"><code><c- g>b</c-></code><a class="self-link" href="#dom-mlgraphbuilder-mul-a-b-b"></a></dfn>);
+  <a data-link-type="idl-name" href="#mloperand" id="ref-for-mloperand⑨⓪"><c- n>MLOperand</c-></a> <a class="idl-code" data-link-type="method" href="#dom-mlgraphbuilder-div" id="ref-for-dom-mlgraphbuilder-div"><c- g>div</c-></a>(<a data-link-type="idl-name" href="#mloperand" id="ref-for-mloperand⑨①"><c- n>MLOperand</c-></a> <dfn class="idl-code" data-dfn-for="MLGraphBuilder/div(a, b)" data-dfn-type="argument" data-export id="dom-mlgraphbuilder-div-a-b-a"><code><c- g>a</c-></code><a class="self-link" href="#dom-mlgraphbuilder-div-a-b-a"></a></dfn>, <a data-link-type="idl-name" href="#mloperand" id="ref-for-mloperand⑨②"><c- n>MLOperand</c-></a> <dfn class="idl-code" data-dfn-for="MLGraphBuilder/div(a, b)" data-dfn-type="argument" data-export id="dom-mlgraphbuilder-div-a-b-b"><code><c- g>b</c-></code><a class="self-link" href="#dom-mlgraphbuilder-div-a-b-b"></a></dfn>);
+  <a data-link-type="idl-name" href="#mloperand" id="ref-for-mloperand⑨③"><c- n>MLOperand</c-></a> <a class="idl-code" data-link-type="method" href="#dom-mlgraphbuilder-max" id="ref-for-dom-mlgraphbuilder-max"><c- g>max</c-></a>(<a data-link-type="idl-name" href="#mloperand" id="ref-for-mloperand⑨④"><c- n>MLOperand</c-></a> <dfn class="idl-code" data-dfn-for="MLGraphBuilder/max(a, b)" data-dfn-type="argument" data-export id="dom-mlgraphbuilder-max-a-b-a"><code><c- g>a</c-></code><a class="self-link" href="#dom-mlgraphbuilder-max-a-b-a"></a></dfn>, <a data-link-type="idl-name" href="#mloperand" id="ref-for-mloperand⑨⑤"><c- n>MLOperand</c-></a> <dfn class="idl-code" data-dfn-for="MLGraphBuilder/max(a, b)" data-dfn-type="argument" data-export id="dom-mlgraphbuilder-max-a-b-b"><code><c- g>b</c-></code><a class="self-link" href="#dom-mlgraphbuilder-max-a-b-b"></a></dfn>);
+  <a data-link-type="idl-name" href="#mloperand" id="ref-for-mloperand⑨⑥"><c- n>MLOperand</c-></a> <a class="idl-code" data-link-type="method" href="#dom-mlgraphbuilder-min" id="ref-for-dom-mlgraphbuilder-min"><c- g>min</c-></a>(<a data-link-type="idl-name" href="#mloperand" id="ref-for-mloperand⑨⑦"><c- n>MLOperand</c-></a> <dfn class="idl-code" data-dfn-for="MLGraphBuilder/min(a, b)" data-dfn-type="argument" data-export id="dom-mlgraphbuilder-min-a-b-a"><code><c- g>a</c-></code><a class="self-link" href="#dom-mlgraphbuilder-min-a-b-a"></a></dfn>, <a data-link-type="idl-name" href="#mloperand" id="ref-for-mloperand⑨⑧"><c- n>MLOperand</c-></a> <dfn class="idl-code" data-dfn-for="MLGraphBuilder/min(a, b)" data-dfn-type="argument" data-export id="dom-mlgraphbuilder-min-a-b-b"><code><c- g>b</c-></code><a class="self-link" href="#dom-mlgraphbuilder-min-a-b-b"></a></dfn>);
+  <a data-link-type="idl-name" href="#mloperand" id="ref-for-mloperand⑨⑨"><c- n>MLOperand</c-></a> <a class="idl-code" data-link-type="method" href="#dom-mlgraphbuilder-pow" id="ref-for-dom-mlgraphbuilder-pow"><c- g>pow</c-></a>(<a data-link-type="idl-name" href="#mloperand" id="ref-for-mloperand①⓪⓪"><c- n>MLOperand</c-></a> <dfn class="idl-code" data-dfn-for="MLGraphBuilder/pow(a, b)" data-dfn-type="argument" data-export id="dom-mlgraphbuilder-pow-a-b-a"><code><c- g>a</c-></code><a class="self-link" href="#dom-mlgraphbuilder-pow-a-b-a"></a></dfn>, <a data-link-type="idl-name" href="#mloperand" id="ref-for-mloperand①⓪①"><c- n>MLOperand</c-></a> <dfn class="idl-code" data-dfn-for="MLGraphBuilder/pow(a, b)" data-dfn-type="argument" data-export id="dom-mlgraphbuilder-pow-a-b-b"><code><c- g>b</c-></code><a class="self-link" href="#dom-mlgraphbuilder-pow-a-b-b"></a></dfn>);
 };
 </pre>
-    <div class="algorithm" data-algorithm="elu">
-      <strong>Arguments:</strong> 
-     <ul>
+   <div>
+     <strong>Arguments:</strong> 
+    <ul>
+     <li data-md>
+      <p><em>a</em>: an <code class="idl"><a data-link-type="idl" href="#mloperand" id="ref-for-mloperand①⓪②">MLOperand</a></code>. The first input tensor.</p>
+     <li data-md>
+      <p><em>b</em>: an <code class="idl"><a data-link-type="idl" href="#mloperand" id="ref-for-mloperand①⓪③">MLOperand</a></code>. The second input tensor.</p>
+    </ul>
+    <p><strong>Returns:</strong> an <code class="idl"><a data-link-type="idl" href="#mloperand" id="ref-for-mloperand①⓪④">MLOperand</a></code>. The output tensor that contains the result of
+    element-wise binary operation of the two input tensors.</p>
+   </div>
+   <div>
+     <strong>Operation types:</strong> 
+    <ul>
+     <li data-md>
+      <p><em>add</em>: Add the values of the two input tensors, element-wise.</p>
+     <li data-md>
+      <p><em>sub</em>: Subtract the values of the second input tensor from the values of the first input tensor, element-wise.</p>
+     <li data-md>
+      <p><em>mul</em>: Multiply the values of the two input tensors, element-wise.</p>
+     <li data-md>
+      <p><em>div</em>: Divide the values of the first input tensor with the values of the second tensor, element-wise.</p>
+     <li data-md>
+      <p><em>max</em>: Select the greater values of the two input tensors, element-wise.</p>
+     <li data-md>
+      <p><em>min</em>: Select the lesser values of the two input tensors, element-wise.</p>
+     <li data-md>
+      <p><em>pow</em>: Compute the values of the values of the first input tensor to the power of the values of the second input tensor, element-wise.</p>
+    </ul>
+   </div>
+   <details class="algorithm" data-algorithm="element-wise-binary-op" data-algorithm-for="MLGraphBuilder" open>
+    <summary> To <dfn class="dfn-paneled" data-dfn-for="MLGraphBuilder" data-dfn-type="dfn" data-lt="element-wise-binary-op" data-noexport id="mlgraphbuilder-element-wise-binary-op">create element-wise binary operation</dfn> given <var>op</var>, <var>a</var> and <var>b</var>, run the following steps: </summary>
+    <div class="algorithm-steps">
+     <ol>
       <li data-md>
-       <p><em>x</em>: an <code class="idl"><a data-link-type="idl" href="#mloperand" id="ref-for-mloperand①①③">MLOperand</a></code>. The input tensor.</p>
+       <p><a data-link-type="dfn" href="https://infra.spec.whatwg.org/#assert" id="ref-for-assert②⑦">Assert</a>: <var>op</var> is one of "add", "sub", "mul", "div", "max", "min", "pow".</p>
       <li data-md>
-       <p><em>options</em>: an optional <code class="idl"><a data-link-type="idl" href="#dictdef-mleluoptions" id="ref-for-dictdef-mleluoptions②">MLEluOptions</a></code>. The optional parameters of the operation.</p>
-       <ul>
+       <p><a data-link-type="dfn" href="https://infra.spec.whatwg.org/#assert" id="ref-for-assert②⑧">Assert</a>: the type of <var>a</var> and <var>b</var> is <code class="idl"><a data-link-type="idl" href="#mloperand" id="ref-for-mloperand①⓪⑤">MLOperand</a></code>.</p>
+      <li data-md>
+       <p>If <var>a</var>.<code class="idl"><a data-link-type="idl" href="#dom-mloperand-descriptor-slot" id="ref-for-dom-mloperand-descriptor-slot③⑨">[[descriptor]]</a></code>.<code class="idl"><a data-link-type="idl" href="#dom-mloperanddescriptor-type" id="ref-for-dom-mloperanddescriptor-type②①">type</a></code> is not equal to <var>b</var>.<code class="idl"><a data-link-type="idl" href="#dom-mloperand-descriptor-slot" id="ref-for-dom-mloperand-descriptor-slot④⓪">[[descriptor]]</a></code>.<code class="idl"><a data-link-type="idl" href="#dom-mloperanddescriptor-type" id="ref-for-dom-mloperanddescriptor-type②②">type</a></code>, then <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#dfn-throw" id="ref-for-dfn-throw⑧⑨">throw</a> a "<code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#dataerror" id="ref-for-dataerror③⓪">DataError</a></code>" <code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#idl-DOMException" id="ref-for-idl-DOMException⑤⑧">DOMException</a></code>.</p>
+      <li data-md>
+       <p>Let <var>descriptor</var> be a new <code class="idl"><a data-link-type="idl" href="#dictdef-mloperanddescriptor" id="ref-for-dictdef-mloperanddescriptor①⑥">MLOperandDescriptor</a></code>.</p>
+      <li data-md>
+       <p>Set <var>descriptor</var>.<code class="idl"><a data-link-type="idl" href="#dom-mloperanddescriptor-type" id="ref-for-dom-mloperanddescriptor-type②③">type</a></code> to <var>a</var>.<code class="idl"><a data-link-type="idl" href="#dom-mloperand-descriptor-slot" id="ref-for-dom-mloperand-descriptor-slot④①">[[descriptor]]</a></code>.<code class="idl"><a data-link-type="idl" href="#dom-mloperanddescriptor-type" id="ref-for-dom-mloperanddescriptor-type②④">type</a></code>.</p>
+      <li data-md>
+       <p>Let <var>descriptor</var>.<code class="idl"><a data-link-type="idl" href="#dom-mloperanddescriptor-dimensions" id="ref-for-dom-mloperanddescriptor-dimensions③⑧">dimensions</a></code> be the result of running the <a data-link-type="dfn" href="#mlgraphbuilder-broadcast-shapes" id="ref-for-mlgraphbuilder-broadcast-shapes">broadcast-shapes</a> steps given <var>a</var>.<code class="idl"><a data-link-type="idl" href="#dom-mloperand-descriptor-slot" id="ref-for-dom-mloperand-descriptor-slot④②">[[descriptor]]</a></code>.<code class="idl"><a data-link-type="idl" href="#dom-mloperanddescriptor-dimensions" id="ref-for-dom-mloperanddescriptor-dimensions③⑨">dimensions</a></code> and <var>b</var>.<code class="idl"><a data-link-type="idl" href="#dom-mloperand-descriptor-slot" id="ref-for-dom-mloperand-descriptor-slot④③">[[descriptor]]</a></code>.<code class="idl"><a data-link-type="idl" href="#dom-mloperanddescriptor-dimensions" id="ref-for-dom-mloperanddescriptor-dimensions④⓪">dimensions</a></code>.</p>
+       <ol>
         <li data-md>
-         <p><em>alpha</em>: a <code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#idl-float" id="ref-for-idl-float⑦">float</a></code> scalar multiplier, default to 1.</p>
-       </ul>
-     </ul>
-     <p><strong>Returns:</strong></p>
-     <ul>
+         <p>If that <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#dfn-throw" id="ref-for-dfn-throw⑨⓪">throws</a> an error, re-<a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#dfn-throw" id="ref-for-dfn-throw⑨①">throw</a> the error.</p>
+       </ol>
       <li data-md>
-       <p>an <code class="idl"><a data-link-type="idl" href="#mloperand" id="ref-for-mloperand①①④">MLOperand</a></code>. The output tensor of the same shape as <em>x</em>.</p>
+       <p>If any of the following sub-steps fail, <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#dfn-throw" id="ref-for-dfn-throw⑨②">throw</a> an "<code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#operationerror" id="ref-for-operationerror①⑨">OperationError</a></code>" <code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#idl-DOMException" id="ref-for-idl-DOMException⑤⑨">DOMException</a></code>.</p>
+       <ol>
+        <li data-md>
+         <p>Let <var>output</var> be the result of <a data-link-type="dfn" href="#create-an-mloperand" id="ref-for-create-an-mloperand⑦">creating an MLOperand</a> given <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#this" id="ref-for-this①⑥">this</a> and <var>descriptor</var>.</p>
+        <li data-md>
+         <p>Make a request to the underlying platform to:</p>
+         <ol>
+          <li data-md>
+           <p>Let <var>opImpl</var> be an <a data-link-type="dfn" href="https://infra.spec.whatwg.org/#implementation-defined" id="ref-for-implementation-defined①⑧">implementation-defined</a> platform operator for the binary operation <var>op</var>, given <var>a</var> and <var>b</var>.</p>
+          <li data-md>
+           <p>Store a reference of <var>opImpl</var> in <var>output</var>.<code class="idl"><a data-link-type="idl" href="#dom-mloperand-operator-slot" id="ref-for-dom-mloperand-operator-slot⑤">[[operator]]</a></code>.</p>
+          <li data-md>
+           <p>Create an <a data-link-type="dfn" href="https://infra.spec.whatwg.org/#implementation-defined" id="ref-for-implementation-defined①⑨">implementation-defined</a> platform operand <var>outputImpl</var> to represent the output, given <var>output</var> and <var>opImpl</var>.</p>
+          <li data-md>
+           <p>Store a reference to <var>outputImpl</var> in <var>output</var>.<code class="idl"><a data-link-type="idl" href="#dom-mloperand-operand-slot" id="ref-for-dom-mloperand-operand-slot①⑤">[[operand]]</a></code>.</p>
+         </ol>
+        <li data-md>
+         <p>Connect <var>a</var>.<code class="idl"><a data-link-type="idl" href="#dom-mloperand-operand-slot" id="ref-for-dom-mloperand-operand-slot①⑥">[[operand]]</a></code> and <var>b</var>.<code class="idl"><a data-link-type="idl" href="#dom-mloperand-operand-slot" id="ref-for-dom-mloperand-operand-slot①⑦">[[operand]]</a></code> as inputs to <var>opImpl</var>.</p>
+        <li data-md>
+         <p>Connect <var>output</var>.<code class="idl"><a data-link-type="idl" href="#dom-mloperand-operand-slot" id="ref-for-dom-mloperand-operand-slot①⑧">[[operand]]</a></code> as output to <var>opImpl</var>.</p>
+       </ol>
       <li data-md>
-       <p>an <code class="idl"><a data-link-type="idl" href="#mlactivation" id="ref-for-mlactivation①①">MLActivation</a></code>. The activation function representing the elu operation.</p>
-     </ul>
-     <div class="note" role="note">
-       The behavior of this operation can be generically emulated from the usage of
+       <p>Return <var>output</var>.</p>
+     </ol>
+    </div>
+   </details>
+   <details class="algorithm" data-algorithm="broadcast-shapes" data-algorithm-for="MLGraphBuilder" open>
+    <summary> To <dfn class="dfn-paneled" data-dfn-for="MLGraphBuilder" data-dfn-type="dfn" data-noexport id="mlgraphbuilder-broadcast-shapes">broadcast-shapes</dfn> given <var>shape1</var> and <var>shape2</var>, run the following steps: </summary>
+    <div class="algorithm-steps">
+     <ol>
+      <li data-md>
+       <p><a data-link-type="dfn" href="https://infra.spec.whatwg.org/#assert" id="ref-for-assert②⑨">Assert</a>: The type of <var>shape1</var> and <var>shape2</var> is <code>sequence of unsigned long</code>.</p>
+      <li data-md>
+       <p>Let <var>output</var> be the result of invoking the <a data-link-type="dfn" href="https://infra.spec.whatwg.org/#implementation-defined" id="ref-for-implementation-defined②⓪">implementation-defined</a> shape broadcast on <var>shape1</var> and <var>shape2</var>.</p>
+       <ol>
+        <li data-md>
+         <p>If that fails, then <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#dfn-throw" id="ref-for-dfn-throw⑨③">throw</a> a "<code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#dataerror" id="ref-for-dataerror③①">DataError</a></code>" <code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#idl-DOMException" id="ref-for-idl-DOMException⑥⓪">DOMException</a></code>.</p>
+       </ol>
+      <li data-md>
+       <p>Return <var>output</var>.</p>
+       <div class="note" role="note"> The most common implementation is that two shapes are compatible, when each of their corresponding dimensions are equal, or one of them is 1. The output shape consists of the maximum of the corresponding dimensions. </div>
+     </ol>
+    </div>
+   </details>
+   <details open>
+    <summary> The element-wise binary operation algorithms invoke the <a data-link-type="dfn" href="#mlgraphbuilder-element-wise-binary-op" id="ref-for-mlgraphbuilder-element-wise-binary-op"> create element-wise binary operation</a> steps as follows. </summary>
+    <div class="algorithm-steps">
+     <div class="algorithm" data-algorithm="add(a, b)" data-algorithm-for="MLGraphBuilder">
+       The <dfn class="dfn-paneled idl-code" data-dfn-for="MLGraphBuilder" data-dfn-type="method" data-export id="dom-mlgraphbuilder-add"><code>add(<var>a</var>, <var>b</var>)</code></dfn> method steps are: 
+      <ol>
+       <li data-md>
+        <p>Let <var>output</var> be the result of running the <a data-link-type="dfn" href="#mlgraphbuilder-element-wise-binary-op" id="ref-for-mlgraphbuilder-element-wise-binary-op①"> create element-wise binary operation</a> given "add", <var>a</var> and <var>b</var>.</p>
+        <ol>
+         <li data-md>
+          <p>If that <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#dfn-throw" id="ref-for-dfn-throw⑨④">throws</a> an error, then re-<a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#dfn-throw" id="ref-for-dfn-throw⑨⑤">throw</a> the error.</p>
+        </ol>
+       <li data-md>
+        <p>Return <var>output</var>.</p>
+      </ol>
+     </div>
+     <div class="algorithm" data-algorithm="sub(a, b)" data-algorithm-for="MLGraphBuilder">
+       The <dfn class="dfn-paneled idl-code" data-dfn-for="MLGraphBuilder" data-dfn-type="method" data-export id="dom-mlgraphbuilder-sub"><code>sub(<var>a</var>, <var>b</var>)</code></dfn> method steps are: 
+      <ol>
+       <li data-md>
+        <p>Let <var>output</var> be the result of running the <a data-link-type="dfn" href="#mlgraphbuilder-element-wise-binary-op" id="ref-for-mlgraphbuilder-element-wise-binary-op②"> create element-wise binary operation</a> given "sub", <var>a</var> and <var>b</var>.</p>
+        <ol>
+         <li data-md>
+          <p>If that <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#dfn-throw" id="ref-for-dfn-throw⑨⑥">throws</a> an error, then re-<a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#dfn-throw" id="ref-for-dfn-throw⑨⑦">throw</a> the error.</p>
+        </ol>
+       <li data-md>
+        <p>Return <var>output</var>.</p>
+      </ol>
+     </div>
+     <div class="algorithm" data-algorithm="mul(a, b)" data-algorithm-for="MLGraphBuilder">
+       The <dfn class="dfn-paneled idl-code" data-dfn-for="MLGraphBuilder" data-dfn-type="method" data-export id="dom-mlgraphbuilder-mul"><code>mul(<var>a</var>, <var>b</var>)</code></dfn> method steps are: 
+      <ol>
+       <li data-md>
+        <p>Let <var>output</var> be the result of running the <a data-link-type="dfn" href="#mlgraphbuilder-element-wise-binary-op" id="ref-for-mlgraphbuilder-element-wise-binary-op③"> create element-wise binary operation</a> given "mul", <var>a</var> and <var>b</var>.</p>
+        <ol>
+         <li data-md>
+          <p>If that <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#dfn-throw" id="ref-for-dfn-throw⑨⑧">throws</a> an error, then re-<a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#dfn-throw" id="ref-for-dfn-throw⑨⑨">throw</a> the error.</p>
+        </ol>
+       <li data-md>
+        <p>Return <var>output</var>.</p>
+      </ol>
+     </div>
+     <div class="algorithm" data-algorithm="div(a, b)" data-algorithm-for="MLGraphBuilder">
+       The <dfn class="dfn-paneled idl-code" data-dfn-for="MLGraphBuilder" data-dfn-type="method" data-export id="dom-mlgraphbuilder-div"><code>div(<var>a</var>, <var>b</var>)</code></dfn> method steps are: 
+      <ol>
+       <li data-md>
+        <p>Let <var>output</var> be the result of running the <a data-link-type="dfn" href="#mlgraphbuilder-element-wise-binary-op" id="ref-for-mlgraphbuilder-element-wise-binary-op④"> create element-wise binary operation</a> given "div", <var>a</var> and <var>b</var>.</p>
+        <ol>
+         <li data-md>
+          <p>If that <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#dfn-throw" id="ref-for-dfn-throw①⓪⓪">throws</a> an error, then re-<a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#dfn-throw" id="ref-for-dfn-throw①⓪①">throw</a> the error.</p>
+        </ol>
+       <li data-md>
+        <p>Return <var>output</var>.</p>
+      </ol>
+     </div>
+     <div class="algorithm" data-algorithm="max(a, b)" data-algorithm-for="MLGraphBuilder">
+       The <dfn class="dfn-paneled idl-code" data-dfn-for="MLGraphBuilder" data-dfn-type="method" data-export id="dom-mlgraphbuilder-max"><code>max(<var>a</var>, <var>b</var>)</code></dfn> method steps are: 
+      <ol>
+       <li data-md>
+        <p>Let <var>output</var> be the result of running the <a data-link-type="dfn" href="#mlgraphbuilder-element-wise-binary-op" id="ref-for-mlgraphbuilder-element-wise-binary-op⑤"> create element-wise binary operation</a> given "max", <var>a</var> and <var>b</var>.</p>
+        <ol>
+         <li data-md>
+          <p>If that <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#dfn-throw" id="ref-for-dfn-throw①⓪②">throws</a> an error, then re-<a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#dfn-throw" id="ref-for-dfn-throw①⓪③">throw</a> the error.</p>
+        </ol>
+       <li data-md>
+        <p>Return <var>output</var>.</p>
+      </ol>
+     </div>
+     <div class="algorithm" data-algorithm="min(a, b)" data-algorithm-for="MLGraphBuilder">
+       The <dfn class="dfn-paneled idl-code" data-dfn-for="MLGraphBuilder" data-dfn-type="method" data-export id="dom-mlgraphbuilder-min"><code>min(<var>a</var>, <var>b</var>)</code></dfn> method steps are: 
+      <ol>
+       <li data-md>
+        <p>Let <var>output</var> be the result of running the <a data-link-type="dfn" href="#mlgraphbuilder-element-wise-binary-op" id="ref-for-mlgraphbuilder-element-wise-binary-op⑥"> create element-wise binary operation</a> given "min", <var>a</var> and <var>b</var>.</p>
+        <ol>
+         <li data-md>
+          <p>If that <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#dfn-throw" id="ref-for-dfn-throw①⓪④">throws</a> an error, then re-<a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#dfn-throw" id="ref-for-dfn-throw①⓪⑤">throw</a> the error.</p>
+        </ol>
+       <li data-md>
+        <p>Return <var>output</var>.</p>
+      </ol>
+     </div>
+     <div class="algorithm" data-algorithm="pow(a, b)" data-algorithm-for="MLGraphBuilder">
+       The <dfn class="dfn-paneled idl-code" data-dfn-for="MLGraphBuilder" data-dfn-type="method" data-export id="dom-mlgraphbuilder-pow"><code>pow(<var>a</var>, <var>b</var>)</code></dfn> method steps are: 
+      <ol>
+       <li data-md>
+        <p>Let <var>output</var> be the result of running the <a data-link-type="dfn" href="#mlgraphbuilder-element-wise-binary-op" id="ref-for-mlgraphbuilder-element-wise-binary-op⑦"> create element-wise binary operation</a> given "pow", <var>a</var> and <var>b</var>.</p>
+        <ol>
+         <li data-md>
+          <p>If that <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#dfn-throw" id="ref-for-dfn-throw①⓪⑥">throws</a> an error, then re-<a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#dfn-throw" id="ref-for-dfn-throw①⓪⑦">throw</a> the error.</p>
+        </ol>
+       <li data-md>
+        <p>Return <var>output</var>.</p>
+      </ol>
+     </div>
+    </div>
+   </details>
+   <h4 class="heading settled" data-level="7.6.11" id="api-mlgraphbuilder-unary"><span class="secno">7.6.11. </span><span class="content">Element-wise unary operations</span><a class="self-link" href="#api-mlgraphbuilder-unary"></a></h4>
+    Compute the element-wise unary operation for input tensor. 
+<pre class="idl highlight def"><c- b>partial</c-> <c- b>interface</c-> <a class="idl-code" data-link-type="interface" href="#mlgraphbuilder" id="ref-for-mlgraphbuilder②⑨"><c- g>MLGraphBuilder</c-></a> {
+  <a data-link-type="idl-name" href="#mloperand" id="ref-for-mloperand①⓪⑥"><c- n>MLOperand</c-></a> <a class="idl-code" data-link-type="method" href="#dom-mlgraphbuilder-abs" id="ref-for-dom-mlgraphbuilder-abs"><c- g>abs</c-></a>(<a data-link-type="idl-name" href="#mloperand" id="ref-for-mloperand①⓪⑦"><c- n>MLOperand</c-></a> <dfn class="idl-code" data-dfn-for="MLGraphBuilder/abs(input)" data-dfn-type="argument" data-export id="dom-mlgraphbuilder-abs-input-input"><code><c- g>input</c-></code><a class="self-link" href="#dom-mlgraphbuilder-abs-input-input"></a></dfn>);
+  <a data-link-type="idl-name" href="#mloperand" id="ref-for-mloperand①⓪⑧"><c- n>MLOperand</c-></a> <a class="idl-code" data-link-type="method" href="#dom-mlgraphbuilder-ceil" id="ref-for-dom-mlgraphbuilder-ceil"><c- g>ceil</c-></a>(<a data-link-type="idl-name" href="#mloperand" id="ref-for-mloperand①⓪⑨"><c- n>MLOperand</c-></a> <dfn class="idl-code" data-dfn-for="MLGraphBuilder/ceil(input)" data-dfn-type="argument" data-export id="dom-mlgraphbuilder-ceil-input-input"><code><c- g>input</c-></code><a class="self-link" href="#dom-mlgraphbuilder-ceil-input-input"></a></dfn>);
+  <a data-link-type="idl-name" href="#mloperand" id="ref-for-mloperand①①⓪"><c- n>MLOperand</c-></a> <a class="idl-code" data-link-type="method" href="#dom-mlgraphbuilder-cos" id="ref-for-dom-mlgraphbuilder-cos"><c- g>cos</c-></a>(<a data-link-type="idl-name" href="#mloperand" id="ref-for-mloperand①①①"><c- n>MLOperand</c-></a> <dfn class="idl-code" data-dfn-for="MLGraphBuilder/cos(input)" data-dfn-type="argument" data-export id="dom-mlgraphbuilder-cos-input-input"><code><c- g>input</c-></code><a class="self-link" href="#dom-mlgraphbuilder-cos-input-input"></a></dfn>);
+  <a data-link-type="idl-name" href="#mloperand" id="ref-for-mloperand①①②"><c- n>MLOperand</c-></a> <a class="idl-code" data-link-type="method" href="#dom-mlgraphbuilder-exp" id="ref-for-dom-mlgraphbuilder-exp"><c- g>exp</c-></a>(<a data-link-type="idl-name" href="#mloperand" id="ref-for-mloperand①①③"><c- n>MLOperand</c-></a> <dfn class="idl-code" data-dfn-for="MLGraphBuilder/exp(input)" data-dfn-type="argument" data-export id="dom-mlgraphbuilder-exp-input-input"><code><c- g>input</c-></code><a class="self-link" href="#dom-mlgraphbuilder-exp-input-input"></a></dfn>);
+  <a data-link-type="idl-name" href="#mloperand" id="ref-for-mloperand①①④"><c- n>MLOperand</c-></a> <a class="idl-code" data-link-type="method" href="#dom-mlgraphbuilder-floor" id="ref-for-dom-mlgraphbuilder-floor"><c- g>floor</c-></a>(<a data-link-type="idl-name" href="#mloperand" id="ref-for-mloperand①①⑤"><c- n>MLOperand</c-></a> <dfn class="idl-code" data-dfn-for="MLGraphBuilder/floor(input)" data-dfn-type="argument" data-export id="dom-mlgraphbuilder-floor-input-input"><code><c- g>input</c-></code><a class="self-link" href="#dom-mlgraphbuilder-floor-input-input"></a></dfn>);
+  <a data-link-type="idl-name" href="#mloperand" id="ref-for-mloperand①①⑥"><c- n>MLOperand</c-></a> <a class="idl-code" data-link-type="method" href="#dom-mlgraphbuilder-log" id="ref-for-dom-mlgraphbuilder-log"><c- g>log</c-></a>(<a data-link-type="idl-name" href="#mloperand" id="ref-for-mloperand①①⑦"><c- n>MLOperand</c-></a> <dfn class="idl-code" data-dfn-for="MLGraphBuilder/log(input)" data-dfn-type="argument" data-export id="dom-mlgraphbuilder-log-input-input"><code><c- g>input</c-></code><a class="self-link" href="#dom-mlgraphbuilder-log-input-input"></a></dfn>);
+  <a data-link-type="idl-name" href="#mloperand" id="ref-for-mloperand①①⑧"><c- n>MLOperand</c-></a> <a class="idl-code" data-link-type="method" href="#dom-mlgraphbuilder-neg" id="ref-for-dom-mlgraphbuilder-neg"><c- g>neg</c-></a>(<a data-link-type="idl-name" href="#mloperand" id="ref-for-mloperand①①⑨"><c- n>MLOperand</c-></a> <dfn class="idl-code" data-dfn-for="MLGraphBuilder/neg(input)" data-dfn-type="argument" data-export id="dom-mlgraphbuilder-neg-input-input"><code><c- g>input</c-></code><a class="self-link" href="#dom-mlgraphbuilder-neg-input-input"></a></dfn>);
+  <a data-link-type="idl-name" href="#mloperand" id="ref-for-mloperand①②⓪"><c- n>MLOperand</c-></a> <a class="idl-code" data-link-type="method" href="#dom-mlgraphbuilder-sin" id="ref-for-dom-mlgraphbuilder-sin"><c- g>sin</c-></a>(<a data-link-type="idl-name" href="#mloperand" id="ref-for-mloperand①②①"><c- n>MLOperand</c-></a> <dfn class="idl-code" data-dfn-for="MLGraphBuilder/sin(input)" data-dfn-type="argument" data-export id="dom-mlgraphbuilder-sin-input-input"><code><c- g>input</c-></code><a class="self-link" href="#dom-mlgraphbuilder-sin-input-input"></a></dfn>);
+  <a data-link-type="idl-name" href="#mloperand" id="ref-for-mloperand①②②"><c- n>MLOperand</c-></a> <a class="idl-code" data-link-type="method" href="#dom-mlgraphbuilder-tan" id="ref-for-dom-mlgraphbuilder-tan"><c- g>tan</c-></a>(<a data-link-type="idl-name" href="#mloperand" id="ref-for-mloperand①②③"><c- n>MLOperand</c-></a> <dfn class="idl-code" data-dfn-for="MLGraphBuilder/tan(input)" data-dfn-type="argument" data-export id="dom-mlgraphbuilder-tan-input-input"><code><c- g>input</c-></code><a class="self-link" href="#dom-mlgraphbuilder-tan-input-input"></a></dfn>);
+};
+</pre>
+   <div>
+     <strong>Arguments:</strong> 
+    <ul>
+     <li data-md>
+      <p><em>input</em>: an <code class="idl"><a data-link-type="idl" href="#mloperand" id="ref-for-mloperand①②④">MLOperand</a></code>. The input tensor.</p>
+    </ul>
+    <p><strong>Returns:</strong> an <code class="idl"><a data-link-type="idl" href="#mloperand" id="ref-for-mloperand①②⑤">MLOperand</a></code>. The output tensor that contains the result of
+    element-wise unary operation of the input tensor. The shape of the output
+    tensor is the same as the shape of input tensor.</p>
+   </div>
+   <div>
+     <strong>Operation types:</strong> 
+    <ul>
+     <li data-md>
+      <p><em>abs</em>: Compute the absolute value of the input tensor, element-wise.</p>
+     <li data-md>
+      <p><em>ceil</em>: Compute the ceiling of the input tensor, element-wise.</p>
+     <li data-md>
+      <p><em>cos</em>: Compute the cosine of the input tensor, element-wise.</p>
+     <li data-md>
+      <p><em>exp</em>: Compute the exponential of the input tensor, element-wise.</p>
+     <li data-md>
+      <p><em>floor</em>: Compute the floor of the input tensor, element-wise.</p>
+     <li data-md>
+      <p><em>log</em>: Compute the natural logarithm of the input tensor, element-wise.</p>
+     <li data-md>
+      <p><em>neg</em>: Compute the numerical negative value of the input tensor, element-wise.</p>
+     <li data-md>
+      <p><em>sin</em>: Compute the sine of the input tensor, element-wise.</p>
+     <li data-md>
+      <p><em>tan</em>: Compute the tangent of the input tensor, element-wise.</p>
+    </ul>
+   </div>
+   <details class="algorithm" data-algorithm="element-wise-unary-op" data-algorithm-for="MLGraphBuilder" open>
+    <summary> To <dfn class="dfn-paneled" data-dfn-for="MLGraphBuilder" data-dfn-type="dfn" data-lt="element-wise-unary-op" data-noexport id="mlgraphbuilder-element-wise-unary-op">create element-wise unary operation</dfn> given <var>op</var> and <var>input</var>, run the following steps: </summary>
+    <div class="algorithm-steps">
+     <ol>
+      <li data-md>
+       <p><a data-link-type="dfn" href="https://infra.spec.whatwg.org/#assert" id="ref-for-assert③⓪">Assert</a>: <var>op</var> is one of "abs", "ceil", "cos", "exp", "floor", "log", "neg", "sin", "tan".</p>
+      <li data-md>
+       <p><a data-link-type="dfn" href="https://infra.spec.whatwg.org/#assert" id="ref-for-assert③①">Assert</a>: the type of <var>input</var> is <code class="idl"><a data-link-type="idl" href="#mloperand" id="ref-for-mloperand①②⑥">MLOperand</a></code>.</p>
+      <li data-md>
+       <p>If any of the following sub-steps fail, <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#dfn-throw" id="ref-for-dfn-throw①⓪⑧">throw</a> an "<code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#operationerror" id="ref-for-operationerror②⓪">OperationError</a></code>" <code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#idl-DOMException" id="ref-for-idl-DOMException⑥①">DOMException</a></code>.</p>
+       <ol>
+        <li data-md>
+         <p>Let <var>output</var> be the result of <a data-link-type="dfn" href="#copy-an-mloperand" id="ref-for-copy-an-mloperand①">copying an MLOperand</a> given <var>input</var>.</p>
+        <li data-md>
+         <p>Make a request to the underlying platform to:</p>
+         <ol>
+          <li data-md>
+           <p>Let <var>opImpl</var> be an <a data-link-type="dfn" href="https://infra.spec.whatwg.org/#implementation-defined" id="ref-for-implementation-defined②①">implementation-defined</a> platform operator for the unary operation <var>op</var>.</p>
+          <li data-md>
+           <p>Store a reference of <var>opImpl</var> in <var>output</var>.<code class="idl"><a data-link-type="idl" href="#dom-mloperand-operator-slot" id="ref-for-dom-mloperand-operator-slot⑥">[[operator]]</a></code>.</p>
+          <li data-md>
+           <p>Create an <a data-link-type="dfn" href="https://infra.spec.whatwg.org/#implementation-defined" id="ref-for-implementation-defined②②">implementation-defined</a> platform operand <var>outputImpl</var> to represent the output, given <var>output</var> and <var>opImpl</var>.</p>
+          <li data-md>
+           <p>Store a reference to <var>outputImpl</var> in <var>output</var>.<code class="idl"><a data-link-type="idl" href="#dom-mloperand-operand-slot" id="ref-for-dom-mloperand-operand-slot①⑨">[[operand]]</a></code>.</p>
+         </ol>
+        <li data-md>
+         <p>Connect <var>input</var>.<code class="idl"><a data-link-type="idl" href="#dom-mloperand-operand-slot" id="ref-for-dom-mloperand-operand-slot②⓪">[[operand]]</a></code> as input to <var>opImpl</var>.</p>
+        <li data-md>
+         <p>Connect <var>output</var>.<code class="idl"><a data-link-type="idl" href="#dom-mloperand-operand-slot" id="ref-for-dom-mloperand-operand-slot②①">[[operand]]</a></code> as output to <var>opImpl</var>.</p>
+       </ol>
+      <li data-md>
+       <p>Return <var>output</var>.</p>
+     </ol>
+    </div>
+   </details>
+   <details open>
+    <summary> The element-wise unary operation algorithms invoke the <a data-link-type="dfn" href="#mlgraphbuilder-element-wise-unary-op" id="ref-for-mlgraphbuilder-element-wise-unary-op"> create element-wise unary operation</a> steps as follows. </summary>
+    <div class="algorithm-steps">
+     <div class="algorithm" data-algorithm="abs(input)" data-algorithm-for="MLGraphBuilder">
+       The <dfn class="dfn-paneled idl-code" data-dfn-for="MLGraphBuilder" data-dfn-type="method" data-export id="dom-mlgraphbuilder-abs"><code>abs(<var>input</var>)</code></dfn> method steps are: 
+      <ol>
+       <li data-md>
+        <p>Let <var>output</var> be the result of running the <a data-link-type="dfn" href="#mlgraphbuilder-element-wise-unary-op" id="ref-for-mlgraphbuilder-element-wise-unary-op①"> create element-wise unary operation</a> given "abs" and <var>input</var>.</p>
+        <ol>
+         <li data-md>
+          <p>If that <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#dfn-throw" id="ref-for-dfn-throw①⓪⑨">throws</a> an error, then re-<a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#dfn-throw" id="ref-for-dfn-throw①①⓪">throw</a> the error.</p>
+        </ol>
+       <li data-md>
+        <p>Return <var>output</var>.</p>
+      </ol>
+     </div>
+     <div class="algorithm" data-algorithm="ceil(input)" data-algorithm-for="MLGraphBuilder">
+       The <dfn class="dfn-paneled idl-code" data-dfn-for="MLGraphBuilder" data-dfn-type="method" data-export id="dom-mlgraphbuilder-ceil"><code>ceil(<var>input</var>)</code></dfn> method steps are: 
+      <ol>
+       <li data-md>
+        <p>Let <var>output</var> be the result of running the <a data-link-type="dfn" href="#mlgraphbuilder-element-wise-unary-op" id="ref-for-mlgraphbuilder-element-wise-unary-op②"> create element-wise unary operation</a> given "ceil" and <var>input</var>.</p>
+        <ol>
+         <li data-md>
+          <p>If that <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#dfn-throw" id="ref-for-dfn-throw①①①">throws</a> an error, then re-<a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#dfn-throw" id="ref-for-dfn-throw①①②">throw</a> the error.</p>
+        </ol>
+       <li data-md>
+        <p>Return <var>output</var>.</p>
+      </ol>
+     </div>
+     <div class="algorithm" data-algorithm="cos(input)" data-algorithm-for="MLGraphBuilder">
+       The <dfn class="dfn-paneled idl-code" data-dfn-for="MLGraphBuilder" data-dfn-type="method" data-export id="dom-mlgraphbuilder-cos"><code>cos(<var>input</var>)</code></dfn> method steps are: 
+      <ol>
+       <li data-md>
+        <p>Let <var>output</var> be the result of running the <a data-link-type="dfn" href="#mlgraphbuilder-element-wise-unary-op" id="ref-for-mlgraphbuilder-element-wise-unary-op③"> create element-wise unary operation</a> given "cos" and <var>input</var>.</p>
+        <ol>
+         <li data-md>
+          <p>If that <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#dfn-throw" id="ref-for-dfn-throw①①③">throws</a> an error, then re-<a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#dfn-throw" id="ref-for-dfn-throw①①④">throw</a> the error.</p>
+        </ol>
+       <li data-md>
+        <p>Return <var>output</var>.</p>
+      </ol>
+     </div>
+     <div class="algorithm" data-algorithm="exp(input)" data-algorithm-for="MLGraphBuilder">
+       The <dfn class="dfn-paneled idl-code" data-dfn-for="MLGraphBuilder" data-dfn-type="method" data-export id="dom-mlgraphbuilder-exp"><code>exp(<var>input</var>)</code></dfn> method steps are: 
+      <ol>
+       <li data-md>
+        <p>Let <var>output</var> be the result of running the <a data-link-type="dfn" href="#mlgraphbuilder-element-wise-unary-op" id="ref-for-mlgraphbuilder-element-wise-unary-op④"> create element-wise unary operation</a> given "exp" and <var>input</var>.</p>
+        <ol>
+         <li data-md>
+          <p>If that <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#dfn-throw" id="ref-for-dfn-throw①①⑤">throws</a> an error, then re-<a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#dfn-throw" id="ref-for-dfn-throw①①⑥">throw</a> the error.</p>
+        </ol>
+       <li data-md>
+        <p>Return <var>output</var>.</p>
+      </ol>
+     </div>
+     <div class="algorithm" data-algorithm="floor(input)" data-algorithm-for="MLGraphBuilder">
+       The <dfn class="dfn-paneled idl-code" data-dfn-for="MLGraphBuilder" data-dfn-type="method" data-export id="dom-mlgraphbuilder-floor"><code>floor(<var>input</var>)</code></dfn> method steps are: 
+      <ol>
+       <li data-md>
+        <p>Let <var>output</var> be the result of running the <a data-link-type="dfn" href="#mlgraphbuilder-element-wise-unary-op" id="ref-for-mlgraphbuilder-element-wise-unary-op⑤"> create element-wise unary operation</a> given "floor" and <var>input</var>.</p>
+        <ol>
+         <li data-md>
+          <p>If that <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#dfn-throw" id="ref-for-dfn-throw①①⑦">throws</a> an error, then re-<a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#dfn-throw" id="ref-for-dfn-throw①①⑧">throw</a> the error.</p>
+        </ol>
+       <li data-md>
+        <p>Return <var>output</var>.</p>
+      </ol>
+     </div>
+     <div class="algorithm" data-algorithm="log(input)" data-algorithm-for="MLGraphBuilder">
+       The <dfn class="dfn-paneled idl-code" data-dfn-for="MLGraphBuilder" data-dfn-type="method" data-export id="dom-mlgraphbuilder-log"><code>log(<var>input</var>)</code></dfn> method steps are: 
+      <ol>
+       <li data-md>
+        <p>Let <var>output</var> be the result of running the <a data-link-type="dfn" href="#mlgraphbuilder-element-wise-unary-op" id="ref-for-mlgraphbuilder-element-wise-unary-op⑥"> create element-wise unary operation</a> given "log" and <var>input</var>.</p>
+        <ol>
+         <li data-md>
+          <p>If that <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#dfn-throw" id="ref-for-dfn-throw①①⑨">throws</a> an error, then re-<a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#dfn-throw" id="ref-for-dfn-throw①②⓪">throw</a> the error.</p>
+        </ol>
+       <li data-md>
+        <p>Return <var>output</var>.</p>
+      </ol>
+     </div>
+     <div class="algorithm" data-algorithm="neg(input)" data-algorithm-for="MLGraphBuilder">
+       The <dfn class="dfn-paneled idl-code" data-dfn-for="MLGraphBuilder" data-dfn-type="method" data-export id="dom-mlgraphbuilder-neg"><code>neg(<var>input</var>)</code></dfn> method steps are: 
+      <ol>
+       <li data-md>
+        <p>Let <var>output</var> be the result of running the <a data-link-type="dfn" href="#mlgraphbuilder-element-wise-unary-op" id="ref-for-mlgraphbuilder-element-wise-unary-op⑦"> create element-wise unary operation</a> given "neg" and <var>input</var>.</p>
+        <ol>
+         <li data-md>
+          <p>If that <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#dfn-throw" id="ref-for-dfn-throw①②①">throws</a> an error, then re-<a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#dfn-throw" id="ref-for-dfn-throw①②②">throw</a> the error.</p>
+        </ol>
+       <li data-md>
+        <p>Return <var>output</var>.</p>
+      </ol>
+     </div>
+     <div class="algorithm" data-algorithm="sin(input)" data-algorithm-for="MLGraphBuilder">
+       The <dfn class="dfn-paneled idl-code" data-dfn-for="MLGraphBuilder" data-dfn-type="method" data-export id="dom-mlgraphbuilder-sin"><code>sin(<var>input</var>)</code></dfn> method steps are: 
+      <ol>
+       <li data-md>
+        <p>Let <var>output</var> be the result of running the <a data-link-type="dfn" href="#mlgraphbuilder-element-wise-unary-op" id="ref-for-mlgraphbuilder-element-wise-unary-op⑧"> create element-wise unary operation</a> given "sin" and <var>input</var>.</p>
+        <ol>
+         <li data-md>
+          <p>If that <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#dfn-throw" id="ref-for-dfn-throw①②③">throws</a> an error, then re-<a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#dfn-throw" id="ref-for-dfn-throw①②④">throw</a> the error.</p>
+        </ol>
+       <li data-md>
+        <p>Return <var>output</var>.</p>
+      </ol>
+     </div>
+     <div class="algorithm" data-algorithm="tan(input)" data-algorithm-for="MLGraphBuilder">
+       The <dfn class="dfn-paneled idl-code" data-dfn-for="MLGraphBuilder" data-dfn-type="method" data-export id="dom-mlgraphbuilder-tan"><code>tan(<var>input</var>)</code></dfn> method steps are: 
+      <ol>
+       <li data-md>
+        <p>Let <var>output</var> be the result of running the <a data-link-type="dfn" href="#mlgraphbuilder-element-wise-unary-op" id="ref-for-mlgraphbuilder-element-wise-unary-op⑨"> create element-wise unary operation</a> given "tan" and <var>input</var>.</p>
+        <ol>
+         <li data-md>
+          <p>If that <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#dfn-throw" id="ref-for-dfn-throw①②⑤">throws</a> an error, then re-<a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#dfn-throw" id="ref-for-dfn-throw①②⑥">throw</a> the error.</p>
+        </ol>
+       <li data-md>
+        <p>Return <var>output</var>.</p>
+      </ol>
+     </div>
+    </div>
+   </details>
+   <h4 class="heading settled" data-level="7.6.12" id="api-mlgraphbuilder-elu"><span class="secno">7.6.12. </span><span class="content">The elu() method</span><a class="self-link" href="#api-mlgraphbuilder-elu"></a></h4>
+    Calculate the <a href="https://en.wikipedia.org/wiki/Rectifier_(neural_networks)#ELU"> exponential linear unit function</a> (ELU) on the input tensor element-wise. The calculation follows the expression <code>max(0, x) + alpha * (exp(min(0, x)) - 1)</code>. 
+<pre class="idl highlight def"><c- b>dictionary</c-> <dfn class="dfn-paneled idl-code" data-dfn-type="dictionary" data-export id="dictdef-mleluoptions"><code><c- g>MLEluOptions</c-></code></dfn> {
+  <a class="idl-code" data-link-type="interface" href="https://webidl.spec.whatwg.org/#idl-float" id="ref-for-idl-float⑨"><c- b>float</c-></a> <dfn class="idl-code" data-default="1" data-dfn-for="MLEluOptions" data-dfn-type="dict-member" data-export data-type="float " id="dom-mleluoptions-alpha"><code><c- g>alpha</c-></code><a class="self-link" href="#dom-mleluoptions-alpha"></a></dfn> = 1;
+};
+
+<c- b>partial</c-> <c- b>interface</c-> <a class="idl-code" data-link-type="interface" href="#mlgraphbuilder" id="ref-for-mlgraphbuilder③⓪"><c- g>MLGraphBuilder</c-></a> {
+  <a data-link-type="idl-name" href="#mloperand" id="ref-for-mloperand①②⑦"><c- n>MLOperand</c-></a> <a class="idl-code" data-link-type="method" href="#dom-mlgraphbuilder-elu" id="ref-for-dom-mlgraphbuilder-elu"><c- g>elu</c-></a>(<a data-link-type="idl-name" href="#mloperand" id="ref-for-mloperand①②⑧"><c- n>MLOperand</c-></a> <dfn class="idl-code" data-dfn-for="MLGraphBuilder/elu(input, options), MLGraphBuilder/elu(input)" data-dfn-type="argument" data-export id="dom-mlgraphbuilder-elu-input-options-input"><code><c- g>input</c-></code><a class="self-link" href="#dom-mlgraphbuilder-elu-input-options-input"></a></dfn>, <c- b>optional</c-> <a data-link-type="idl-name" href="#dictdef-mleluoptions" id="ref-for-dictdef-mleluoptions"><c- n>MLEluOptions</c-></a> <dfn class="idl-code" data-dfn-for="MLGraphBuilder/elu(input, options), MLGraphBuilder/elu(input)" data-dfn-type="argument" data-export id="dom-mlgraphbuilder-elu-input-options-options"><code><c- g>options</c-></code><a class="self-link" href="#dom-mlgraphbuilder-elu-input-options-options"></a></dfn> = {});
+  <a data-link-type="idl-name" href="#mlactivation" id="ref-for-mlactivation②①"><c- n>MLActivation</c-></a> <a class="idl-code" data-link-type="method" href="#dom-mlgraphbuilder-elu-options" id="ref-for-dom-mlgraphbuilder-elu-options"><c- g>elu</c-></a>(<c- b>optional</c-> <a data-link-type="idl-name" href="#dictdef-mleluoptions" id="ref-for-dictdef-mleluoptions①"><c- n>MLEluOptions</c-></a> <dfn class="idl-code" data-dfn-for="MLGraphBuilder/elu(options), MLGraphBuilder/elu()" data-dfn-type="argument" data-export id="dom-mlgraphbuilder-elu-options-options"><code><c- g>options</c-></code><a class="self-link" href="#dom-mlgraphbuilder-elu-options-options"></a></dfn> = {});
+};
+</pre>
+   <div class="note" role="note">
+    <details open>
+     <summary> The behavior of this operation can be generically emulated from the usage of
     other operations as follow. However, user agents typically have a more
     efficient implementation for it, therefore its usage is encouraged from the
-    performance standpoint. 
+    performance standpoint. </summary>
 <pre class="highlight"><c- k>return</c-> builder<c- p>.</c->add<c- p>(</c->
           builder<c- p>.</c->max<c- p>(</c->builder<c- p>.</c->constant<c- p>(</c-><c- mf>0</c-><c- p>),</c-> x<c- p>),</c->
           builder<c- p>.</c->mul<c- p>(</c->
@@ -2636,47 +3981,187 @@ <h4 class="heading settled" data-level="7.6.11" id="api-mlgraphbuilder-elu"><spa
               builder<c- p>.</c->exp<c- p>(</c->builder<c- p>.</c->min<c- p>(</c->builder<c- p>.</c->constant<c- p>(</c-><c- mf>0</c-><c- p>),</c-> x<c- p>)),</c->
               builder<c- p>.</c->constant<c- p>(</c-><c- mf>1</c-><c- p>))));</c->
 </pre>
-     </div>
+    </details>
+   </div>
+   <h5 class="heading settled" data-level="7.6.12.1" id="api-mlgraphbuilder-elu-input-options"><span class="secno">7.6.12.1. </span><span class="content">The <code class="idl"><a data-link-type="idl" href="#dom-mlgraphbuilder-elu" id="ref-for-dom-mlgraphbuilder-elu①">elu(input, options)</a></code> method</span><a class="self-link" href="#api-mlgraphbuilder-elu-input-options"></a></h5>
+   <div>
+     <strong>Arguments:</strong> 
+    <ul>
+     <li data-md>
+      <p><em>input</em>: an <code class="idl"><a data-link-type="idl" href="#mloperand" id="ref-for-mloperand①②⑨">MLOperand</a></code>. The input tensor.</p>
+     <li data-md>
+      <p><em>options</em>: an optional <code class="idl"><a data-link-type="idl" href="#dictdef-mleluoptions" id="ref-for-dictdef-mleluoptions②">MLEluOptions</a></code>. The optional parameters of the operation.</p>
+      <ul>
+       <li data-md>
+        <p><em>alpha</em>: a <code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#idl-float" id="ref-for-idl-float①⓪">float</a></code> scalar multiplier, default to 1.</p>
+      </ul>
+    </ul>
+    <p><strong>Returns:</strong></p>
+    <ul>
+     <li data-md>
+      <p>an <code class="idl"><a data-link-type="idl" href="#mloperand" id="ref-for-mloperand①③⓪">MLOperand</a></code>. The output tensor of the same shape as <em>input</em>.</p>
+    </ul>
+   </div>
+   <details class="algorithm" data-algorithm="elu(input, options)" data-algorithm-for="MLGraphBuilder" open>
+    <summary> The <dfn class="dfn-paneled idl-code" data-dfn-for="MLGraphBuilder" data-dfn-type="method" data-export data-lt="elu(input, options)|elu(input)" id="dom-mlgraphbuilder-elu"><code>elu(<var>input</var>, <var>options</var>)</code></dfn> method steps are: </summary>
+    <div class="algorithm-steps">
+     <ol>
+      <li data-md>
+       <p>If any of the following sub-steps fail, <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#dfn-throw" id="ref-for-dfn-throw①②⑦">throw</a> an "<code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#operationerror" id="ref-for-operationerror②①">OperationError</a></code>" <code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#idl-DOMException" id="ref-for-idl-DOMException⑥②">DOMException</a></code>.</p>
+       <ol>
+        <li data-md>
+         <p>Let <var>output</var> be the result of <a data-link-type="dfn" href="#copy-an-mloperand" id="ref-for-copy-an-mloperand②">copying an MLOperand</a> given <var>input</var>.</p>
+        <li data-md>
+         <p>Make a request to the underlying platform to:</p>
+         <ol>
+          <li data-md>
+           <p>Let <var>opImpl</var> be an <a data-link-type="dfn" href="https://infra.spec.whatwg.org/#implementation-defined" id="ref-for-implementation-defined②③">implementation-defined</a> platform operator for the ELU operation, given <var>options</var>.</p>
+          <li data-md>
+           <p>Store a reference of <var>opImpl</var> in <var>output</var>.<code class="idl"><a data-link-type="idl" href="#dom-mloperand-operator-slot" id="ref-for-dom-mloperand-operator-slot⑦">[[operator]]</a></code>.</p>
+          <li data-md>
+           <p>Create an <a data-link-type="dfn" href="https://infra.spec.whatwg.org/#implementation-defined" id="ref-for-implementation-defined②④">implementation-defined</a> platform operand <var>outputImpl</var> to represent the output, given <var>output</var> and <var>opImpl</var>.</p>
+          <li data-md>
+           <p>Store a reference to <var>outputImpl</var> in <var>output</var>.<code class="idl"><a data-link-type="idl" href="#dom-mloperand-operand-slot" id="ref-for-dom-mloperand-operand-slot②②">[[operand]]</a></code>.</p>
+         </ol>
+        <li data-md>
+         <p>Connect <var>input</var>.<code class="idl"><a data-link-type="idl" href="#dom-mloperand-operand-slot" id="ref-for-dom-mloperand-operand-slot②③">[[operand]]</a></code> as input to <var>opImpl</var>.</p>
+        <li data-md>
+         <p>Connect <var>output</var>.<code class="idl"><a data-link-type="idl" href="#dom-mloperand-operand-slot" id="ref-for-dom-mloperand-operand-slot②④">[[operand]]</a></code> as output to <var>opImpl</var>.</p>
+       </ol>
+      <li data-md>
+       <p>Return <var>output</var>.</p>
+     </ol>
+    </div>
+   </details>
+   <h5 class="heading settled" data-level="7.6.12.2" id="api-mlgraphbuilder-elu-options"><span class="secno">7.6.12.2. </span><span class="content">The <code class="idl"><a data-link-type="idl" href="#dom-mlgraphbuilder-elu-options" id="ref-for-dom-mlgraphbuilder-elu-options①">elu(options)</a></code> method</span><a class="self-link" href="#api-mlgraphbuilder-elu-options"></a></h5>
+   <div>
+     <strong>Arguments:</strong> 
+    <ul>
+     <li data-md>
+      <p><em>options</em>: an optional <code class="idl"><a data-link-type="idl" href="#dictdef-mleluoptions" id="ref-for-dictdef-mleluoptions③">MLEluOptions</a></code>. The optional parameters of the operation.</p>
+      <ul>
+       <li data-md>
+        <p><em>alpha</em>: a <code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#idl-float" id="ref-for-idl-float①①">float</a></code> scalar multiplier, default to 1.</p>
+      </ul>
+    </ul>
+    <p><strong>Returns:</strong></p>
+    <ul>
+     <li data-md>
+      <p>an <code class="idl"><a data-link-type="idl" href="#mlactivation" id="ref-for-mlactivation②②">MLActivation</a></code>. The activation function representing the elu operation.</p>
+    </ul>
+   </div>
+   <details class="algorithm" data-algorithm="elu(options)" data-algorithm-for="MLGraphBuilder" open>
+    <summary> The <dfn class="dfn-paneled idl-code" data-dfn-for="MLGraphBuilder" data-dfn-type="method" data-export data-lt="elu(options)|elu()" id="dom-mlgraphbuilder-elu-options"><code>elu(<var>options</var>)</code></dfn> method steps are: </summary>
+    <div class="algorithm-steps">
+     <ol>
+      <li data-md>
+       <p>Let <var>op</var> be the result of <a data-link-type="dfn" href="#create-an-mlactivation" id="ref-for-create-an-mlactivation①">creating an MLActivation</a> given <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#this" id="ref-for-this①⑦">this</a>, <code>"elu"</code> and <var>options</var>.</p>
+      <li data-md>
+       <p>Return <var>op</var>.</p>
+     </ol>
     </div>
-    <h4 class="heading settled" data-level="7.6.12" id="api-mlgraphbuilder-gemm"><span class="secno">7.6.12. </span><span class="content">The gemm() method</span><a class="self-link" href="#api-mlgraphbuilder-gemm"></a></h4>
-     Calculate the <a href="https://en.wikipedia.org/wiki/Basic_Linear_Algebra_Subprograms#Level_3">general matrix multiplication of the Basic Linear Algebra Subprograms</a>. The calculation follows the expression <code>alpha * A * B + beta * C</code>, where <code>A</code> is a 2-D tensor with shape [M, K] or [K, M], <code>B</code> is a 2-D tensor with shape [K, N] or [N, K], and <code>C</code> is broadcastable to the shape [M, N]. <code>A</code> and <code>B</code> may optionally be transposed prior to the calculation. 
+   </details>
+   <h4 class="heading settled" data-level="7.6.13" id="api-mlgraphbuilder-gemm"><span class="secno">7.6.13. </span><span class="content">The gemm() method</span><a class="self-link" href="#api-mlgraphbuilder-gemm"></a></h4>
+    Calculate the <a href="https://en.wikipedia.org/wiki/Basic_Linear_Algebra_Subprograms#Level_3">general matrix multiplication of the Basic Linear Algebra Subprograms</a>. The calculation follows the expression <code>alpha * A * B + beta * C</code>, where <code>A</code> is a 2-D tensor with shape [M, K] or [K, M], <code>B</code> is a 2-D tensor with shape [K, N] or [N, K], and <code>C</code> is broadcastable to the shape [M, N]. <code>A</code> and <code>B</code> may optionally be transposed prior to the calculation. 
 <pre class="idl highlight def"><c- b>dictionary</c-> <dfn class="dfn-paneled idl-code" data-dfn-type="dictionary" data-export id="dictdef-mlgemmoptions"><code><c- g>MLGemmOptions</c-></code></dfn> {
-  <a data-link-type="idl-name" href="#mloperand" id="ref-for-mloperand①①⑤"><c- n>MLOperand</c-></a> <dfn class="idl-code" data-dfn-for="MLGemmOptions" data-dfn-type="dict-member" data-export data-type="MLOperand " id="dom-mlgemmoptions-c"><code><c- g>c</c-></code><a class="self-link" href="#dom-mlgemmoptions-c"></a></dfn>;
-  <a class="idl-code" data-link-type="interface" href="https://webidl.spec.whatwg.org/#idl-float" id="ref-for-idl-float⑧"><c- b>float</c-></a> <dfn class="idl-code" data-default="1.0" data-dfn-for="MLGemmOptions" data-dfn-type="dict-member" data-export data-type="float " id="dom-mlgemmoptions-alpha"><code><c- g>alpha</c-></code><a class="self-link" href="#dom-mlgemmoptions-alpha"></a></dfn> = 1.0;
-  <a class="idl-code" data-link-type="interface" href="https://webidl.spec.whatwg.org/#idl-float" id="ref-for-idl-float⑨"><c- b>float</c-></a> <dfn class="idl-code" data-default="1.0" data-dfn-for="MLGemmOptions" data-dfn-type="dict-member" data-export data-type="float " id="dom-mlgemmoptions-beta"><code><c- g>beta</c-></code><a class="self-link" href="#dom-mlgemmoptions-beta"></a></dfn> = 1.0;
-  <a class="idl-code" data-link-type="interface" href="https://webidl.spec.whatwg.org/#idl-boolean" id="ref-for-idl-boolean"><c- b>boolean</c-></a> <dfn class="idl-code" data-default="false" data-dfn-for="MLGemmOptions" data-dfn-type="dict-member" data-export data-type="boolean " id="dom-mlgemmoptions-atranspose"><code><c- g>aTranspose</c-></code><a class="self-link" href="#dom-mlgemmoptions-atranspose"></a></dfn> = <c- b>false</c->;
-  <a class="idl-code" data-link-type="interface" href="https://webidl.spec.whatwg.org/#idl-boolean" id="ref-for-idl-boolean①"><c- b>boolean</c-></a> <dfn class="idl-code" data-default="false" data-dfn-for="MLGemmOptions" data-dfn-type="dict-member" data-export data-type="boolean " id="dom-mlgemmoptions-btranspose"><code><c- g>bTranspose</c-></code><a class="self-link" href="#dom-mlgemmoptions-btranspose"></a></dfn> = <c- b>false</c->;
+  <a data-link-type="idl-name" href="#mloperand" id="ref-for-mloperand①③①"><c- n>MLOperand</c-></a> <a class="idl-code" data-link-type="dict-member" data-type="MLOperand " href="#dom-mlgemmoptions-c" id="ref-for-dom-mlgemmoptions-c"><c- g>c</c-></a>;
+  <a class="idl-code" data-link-type="interface" href="https://webidl.spec.whatwg.org/#idl-float" id="ref-for-idl-float①②"><c- b>float</c-></a> <a class="idl-code" data-default="1.0" data-link-type="dict-member" data-type="float " href="#dom-mlgemmoptions-alpha" id="ref-for-dom-mlgemmoptions-alpha"><c- g>alpha</c-></a> = 1.0;
+  <a class="idl-code" data-link-type="interface" href="https://webidl.spec.whatwg.org/#idl-float" id="ref-for-idl-float①③"><c- b>float</c-></a> <a class="idl-code" data-default="1.0" data-link-type="dict-member" data-type="float " href="#dom-mlgemmoptions-beta" id="ref-for-dom-mlgemmoptions-beta"><c- g>beta</c-></a> = 1.0;
+  <a class="idl-code" data-link-type="interface" href="https://webidl.spec.whatwg.org/#idl-boolean" id="ref-for-idl-boolean"><c- b>boolean</c-></a> <a class="idl-code" data-default="false" data-link-type="dict-member" data-type="boolean " href="#dom-mlgemmoptions-atranspose" id="ref-for-dom-mlgemmoptions-atranspose"><c- g>aTranspose</c-></a> = <c- b>false</c->;
+  <a class="idl-code" data-link-type="interface" href="https://webidl.spec.whatwg.org/#idl-boolean" id="ref-for-idl-boolean①"><c- b>boolean</c-></a> <a class="idl-code" data-default="false" data-link-type="dict-member" data-type="boolean " href="#dom-mlgemmoptions-btranspose" id="ref-for-dom-mlgemmoptions-btranspose"><c- g>bTranspose</c-></a> = <c- b>false</c->;
 };
 
-<c- b>partial</c-> <c- b>interface</c-> <a class="idl-code" data-link-type="interface" href="#mlgraphbuilder" id="ref-for-mlgraphbuilder②⑧"><c- g>MLGraphBuilder</c-></a> {
-  <a data-link-type="idl-name" href="#mloperand" id="ref-for-mloperand①①⑥"><c- n>MLOperand</c-></a> <dfn class="dfn-paneled idl-code" data-dfn-for="MLGraphBuilder" data-dfn-type="method" data-export data-lt="gemm(a, b, options)|gemm(a, b)" id="dom-mlgraphbuilder-gemm"><code><c- g>gemm</c-></code></dfn>(<a data-link-type="idl-name" href="#mloperand" id="ref-for-mloperand①①⑦"><c- n>MLOperand</c-></a> <dfn class="idl-code" data-dfn-for="MLGraphBuilder/gemm(a, b, options), MLGraphBuilder/gemm(a, b)" data-dfn-type="argument" data-export id="dom-mlgraphbuilder-gemm-a-b-options-a"><code><c- g>a</c-></code><a class="self-link" href="#dom-mlgraphbuilder-gemm-a-b-options-a"></a></dfn>, <a data-link-type="idl-name" href="#mloperand" id="ref-for-mloperand①①⑧"><c- n>MLOperand</c-></a> <dfn class="idl-code" data-dfn-for="MLGraphBuilder/gemm(a, b, options), MLGraphBuilder/gemm(a, b)" data-dfn-type="argument" data-export id="dom-mlgraphbuilder-gemm-a-b-options-b"><code><c- g>b</c-></code><a class="self-link" href="#dom-mlgraphbuilder-gemm-a-b-options-b"></a></dfn>, <c- b>optional</c-> <a data-link-type="idl-name" href="#dictdef-mlgemmoptions" id="ref-for-dictdef-mlgemmoptions"><c- n>MLGemmOptions</c-></a> <dfn class="idl-code" data-dfn-for="MLGraphBuilder/gemm(a, b, options), MLGraphBuilder/gemm(a, b)" data-dfn-type="argument" data-export id="dom-mlgraphbuilder-gemm-a-b-options-options"><code><c- g>options</c-></code><a class="self-link" href="#dom-mlgraphbuilder-gemm-a-b-options-options"></a></dfn> = {});
+<c- b>partial</c-> <c- b>interface</c-> <a class="idl-code" data-link-type="interface" href="#mlgraphbuilder" id="ref-for-mlgraphbuilder③①"><c- g>MLGraphBuilder</c-></a> {
+  <a data-link-type="idl-name" href="#mloperand" id="ref-for-mloperand①③②"><c- n>MLOperand</c-></a> <a class="idl-code" data-link-type="method" href="#dom-mlgraphbuilder-gemm" id="ref-for-dom-mlgraphbuilder-gemm①"><c- g>gemm</c-></a>(<a data-link-type="idl-name" href="#mloperand" id="ref-for-mloperand①③③"><c- n>MLOperand</c-></a> <dfn class="idl-code" data-dfn-for="MLGraphBuilder/gemm(a, b, options), MLGraphBuilder/gemm(a, b)" data-dfn-type="argument" data-export id="dom-mlgraphbuilder-gemm-a-b-options-a"><code><c- g>a</c-></code><a class="self-link" href="#dom-mlgraphbuilder-gemm-a-b-options-a"></a></dfn>, <a data-link-type="idl-name" href="#mloperand" id="ref-for-mloperand①③④"><c- n>MLOperand</c-></a> <dfn class="idl-code" data-dfn-for="MLGraphBuilder/gemm(a, b, options), MLGraphBuilder/gemm(a, b)" data-dfn-type="argument" data-export id="dom-mlgraphbuilder-gemm-a-b-options-b"><code><c- g>b</c-></code><a class="self-link" href="#dom-mlgraphbuilder-gemm-a-b-options-b"></a></dfn>, <c- b>optional</c-> <a data-link-type="idl-name" href="#dictdef-mlgemmoptions" id="ref-for-dictdef-mlgemmoptions"><c- n>MLGemmOptions</c-></a> <dfn class="idl-code" data-dfn-for="MLGraphBuilder/gemm(a, b, options), MLGraphBuilder/gemm(a, b)" data-dfn-type="argument" data-export id="dom-mlgraphbuilder-gemm-a-b-options-options"><code><c- g>options</c-></code><a class="self-link" href="#dom-mlgraphbuilder-gemm-a-b-options-options"></a></dfn> = {});
 };
 </pre>
-    <div class="algorithm" data-algorithm="gemm">
-      <strong>Arguments:</strong> 
-     <ul>
+   <p><code class="idl"><a data-link-type="idl" href="#dictdef-mlgemmoptions" id="ref-for-dictdef-mlgemmoptions①">MLGemmOptions</a></code> has the following members:</p>
+   <dl>
+    <dt data-md><dfn class="dfn-paneled idl-code" data-dfn-for="MLGemmOptions" data-dfn-type="dict-member" data-export id="dom-mlgemmoptions-c"><code>c</code></dfn>, <span> of type <a data-link-type="idl-name" href="#mloperand" id="ref-for-mloperand①③⑤">MLOperand</a></span>
+    <dd data-md>
+     <p>An <code class="idl"><a data-link-type="idl" href="#mloperand" id="ref-for-mloperand①③⑥">MLOperand</a></code>. Specifies the third input tensor. It is either a scalar, or of the shape that is unidirectionally broadcastable to the shape [M, N] according to <a data-link-type="biblio" href="#biblio-numpy-broadcasting-rule" title="General Broadcasting Rules of NumPy">[numpy-broadcasting-rule]</a>. When it is not specified, the computation is done as if <em>c</em> is a scalar 0.0.</p>
+    <dt data-md><dfn class="dfn-paneled idl-code" data-dfn-for="MLGemmOptions" data-dfn-type="dict-member" data-export id="dom-mlgemmoptions-alpha"><code>alpha</code></dfn>, <span> of type <a data-link-type="idl-name" href="https://webidl.spec.whatwg.org/#idl-float" id="ref-for-idl-float①④">float</a>, defaulting to <code>1.0</code></span>
+    <dd data-md>
+     <p>A <code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#idl-float" id="ref-for-idl-float①⑤">float</a></code> scalar multiplier for the first input.</p>
+    <dt data-md><dfn class="dfn-paneled idl-code" data-dfn-for="MLGemmOptions" data-dfn-type="dict-member" data-export id="dom-mlgemmoptions-beta"><code>beta</code></dfn>, <span> of type <a data-link-type="idl-name" href="https://webidl.spec.whatwg.org/#idl-float" id="ref-for-idl-float①⑥">float</a>, defaulting to <code>1.0</code></span>
+    <dd data-md>
+     <p>A <code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#idl-float" id="ref-for-idl-float①⑦">float</a></code> scalar multiplier for the third input <code class="idl"><a data-link-type="idl" href="#dom-mlgemmoptions-c" id="ref-for-dom-mlgemmoptions-c①">c</a></code>.</p>
+    <dt data-md><dfn class="dfn-paneled idl-code" data-dfn-for="MLGemmOptions" data-dfn-type="dict-member" data-export id="dom-mlgemmoptions-atranspose"><code>aTranspose</code></dfn>, <span> of type <a data-link-type="idl-name" href="https://webidl.spec.whatwg.org/#idl-boolean" id="ref-for-idl-boolean②">boolean</a>, defaulting to <code>false</code></span>
+    <dd data-md>
+     <p>A <code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#idl-boolean" id="ref-for-idl-boolean③">boolean</a></code> indicating if the first input should be transposed prior to calculating the output.</p>
+    <dt data-md><dfn class="dfn-paneled idl-code" data-dfn-for="MLGemmOptions" data-dfn-type="dict-member" data-export id="dom-mlgemmoptions-btranspose"><code>bTranspose</code></dfn>, <span> of type <a data-link-type="idl-name" href="https://webidl.spec.whatwg.org/#idl-boolean" id="ref-for-idl-boolean④">boolean</a>, defaulting to <code>false</code></span>
+    <dd data-md>
+     <p>A <code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#idl-boolean" id="ref-for-idl-boolean⑤">boolean</a></code> indicating if the second input should be transposed prior to calculating the output.</p>
+   </dl>
+   <div>
+     <strong>Arguments:</strong> 
+    <ul>
+     <li data-md>
+      <p><em>a</em>: an <code class="idl"><a data-link-type="idl" href="#mloperand" id="ref-for-mloperand①③⑦">MLOperand</a></code>. The first input 2-D tensor with shape [M, K] if <em>aTranspose</em> is false, or [K, M] if <em>aTranspose</em> is true.</p>
+     <li data-md>
+      <p><em>b</em>: an <code class="idl"><a data-link-type="idl" href="#mloperand" id="ref-for-mloperand①③⑧">MLOperand</a></code>. The second input 2-D tensor with shape [K, N] if <em>bTranspose</em> is false, or [N, K] if <em>bTranspose</em> is true.</p>
+     <li data-md>
+      <p><em>options</em>: an optional <code class="idl"><a data-link-type="idl" href="#dictdef-mlgemmoptions" id="ref-for-dictdef-mlgemmoptions②">MLGemmOptions</a></code>. The optional parameters of the operation.</p>
+    </ul>
+    <p><strong>Returns:</strong> an <code class="idl"><a data-link-type="idl" href="#mloperand" id="ref-for-mloperand①③⑨">MLOperand</a></code>. The output 2-D tensor of shape [M, N] that contains the calculated product of all the inputs.</p>
+   </div>
+   <details class="algorithm" data-algorithm="gemm(a, b, options)" data-algorithm-for="MLGraphBuilder" open>
+    <summary> The <dfn class="dfn-paneled idl-code" data-dfn-for="MLGraphBuilder" data-dfn-type="method" data-export data-lt="gemm(a, b, options)|gemm(a, b)" id="dom-mlgraphbuilder-gemm"><code>gemm(<var>a</var>, <var>b</var>, <var>options</var>)</code></dfn> method steps are: </summary>
+    <div class="algorithm-steps">
+     <ol>
+      <li data-md>
+       <p><a data-link-type="dfn" href="https://infra.spec.whatwg.org/#assert" id="ref-for-assert③②">Assert</a>: the type of <var>a</var> and <var>b</var> is <code class="idl"><a data-link-type="idl" href="#mloperand" id="ref-for-mloperand①④⓪">MLOperand</a></code>.</p>
+      <li data-md>
+       <p>Let <var>shapeA</var> be <var>a</var>.<code class="idl"><a data-link-type="idl" href="#dom-mloperand-descriptor-slot" id="ref-for-dom-mloperand-descriptor-slot④④">[[descriptor]]</a></code>.<code class="idl"><a data-link-type="idl" href="#dom-mloperanddescriptor-dimensions" id="ref-for-dom-mloperanddescriptor-dimensions④①">dimensions</a></code> and <var>sizeA</var> the <a data-link-type="dfn" href="https://infra.spec.whatwg.org/#list-size" id="ref-for-list-size②④">size</a> of <var>shapeA</var>.</p>
+      <li data-md>
+       <p>Let <var>shapeB</var> be <var>b</var>.<code class="idl"><a data-link-type="idl" href="#dom-mloperand-descriptor-slot" id="ref-for-dom-mloperand-descriptor-slot④⑤">[[descriptor]]</a></code>.<code class="idl"><a data-link-type="idl" href="#dom-mloperanddescriptor-dimensions" id="ref-for-dom-mloperanddescriptor-dimensions④②">dimensions</a></code> and <var>sizeB</var> the <a data-link-type="dfn" href="https://infra.spec.whatwg.org/#list-size" id="ref-for-list-size②⑤">size</a> of <var>shapeB</var>.</p>
+      <li data-md>
+       <p>If <var>sizeA</var> is not 2 or <var>sizeB</var> is not 2, then <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#dfn-throw" id="ref-for-dfn-throw①②⑧">throw</a> a "<code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#dataerror" id="ref-for-dataerror③②">DataError</a></code>" <code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#idl-DOMException" id="ref-for-idl-DOMException⑥③">DOMException</a></code>.</p>
+      <li data-md>
+       <p>If <var>options</var>.<code class="idl"><a data-link-type="idl" href="#dom-mlgemmoptions-atranspose" id="ref-for-dom-mlgemmoptions-atranspose①">aTranspose</a></code> is true, then let <var>shapeA</var> be the reverse array of <var>shapeA</var>.</p>
+      <li data-md>
+       <p>If <var>options</var>.<code class="idl"><a data-link-type="idl" href="#dom-mlgemmoptions-btranspose" id="ref-for-dom-mlgemmoptions-btranspose①">bTranspose</a></code> is true, then let <var>shapeB</var> be the reverse array of <var>shapeB</var>.</p>
       <li data-md>
-       <p><em>a</em>: an <code class="idl"><a data-link-type="idl" href="#mloperand" id="ref-for-mloperand①①⑨">MLOperand</a></code>. The first input 2-D tensor with shape [M, K] if <em>aTranspose</em> is false, or [K, M] if <em>aTranspose</em> is true.</p>
+       <p>If <var>shapeA</var>[1] is not equal to <var>shapeB</var>[0], then <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#dfn-throw" id="ref-for-dfn-throw①②⑨">throw</a> a "<code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#dataerror" id="ref-for-dataerror③③">DataError</a></code>" <code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#idl-DOMException" id="ref-for-idl-DOMException⑥④">DOMException</a></code>.</p>
       <li data-md>
-       <p><em>b</em>: an <code class="idl"><a data-link-type="idl" href="#mloperand" id="ref-for-mloperand①②⓪">MLOperand</a></code>. The second input 2-D tensor with shape [K, N] if <em>bTranspose</em> is false, or [N, K] if <em>bTranspose</em> is true.</p>
+       <p>If <var>options</var>.<code class="idl"><a data-link-type="idl" href="#dom-mlgemmoptions-c" id="ref-for-dom-mlgemmoptions-c②">c</a></code> <a data-link-type="dfn" href="https://infra.spec.whatwg.org/#map-exists" id="ref-for-map-exists②⑦">exists</a> and is not unidirectionally broadcastable to the shape [<var>shapeA</var>[0], <var>shapeB</var>[1]] according to the <a data-link-type="biblio" href="#biblio-numpy-broadcasting-rule" title="General Broadcasting Rules of NumPy">[numpy-broadcasting-rule]</a>, then <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#dfn-throw" id="ref-for-dfn-throw①③⓪">throw</a> a "<code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#dataerror" id="ref-for-dataerror③④">DataError</a></code>" <code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#idl-DOMException" id="ref-for-idl-DOMException⑥⑤">DOMException</a></code>.</p>
+       <div class="note" role="note"> Type compatibility between <var>a</var>, <var>b</var> and <var>options</var>.<code class="idl"><a data-link-type="idl" href="#dom-mlgemmoptions-c" id="ref-for-dom-mlgemmoptions-c③">c</a></code> can be also checked. </div>
       <li data-md>
-       <p><em>options</em>: an optional <code class="idl"><a data-link-type="idl" href="#dictdef-mlgemmoptions" id="ref-for-dictdef-mlgemmoptions①">MLGemmOptions</a></code>. The optional parameters of the operation.</p>
-       <ul>
-        <li data-md>
-         <p><em>c</em>: an <code class="idl"><a data-link-type="idl" href="#mloperand" id="ref-for-mloperand①②①">MLOperand</a></code>. The third input tensor. It is either a scalar, or of the shape that is unidirectionally broadcastable to the shape [M, N] according to <a data-link-type="biblio" href="#biblio-numpy-broadcasting-rule" title="General Broadcasting Rules of NumPy">[numpy-broadcasting-rule]</a>. When it is not specified, the computation is done as if <em>c</em> is a scalar 0.0.</p>
+       <p>Let <var>desc</var> a new <code class="idl"><a data-link-type="idl" href="#dictdef-mloperanddescriptor" id="ref-for-dictdef-mloperanddescriptor①⑦">MLOperandDescriptor</a></code>.</p>
+      <li data-md>
+       <p>Set <var>desc</var>.<code class="idl"><a data-link-type="idl" href="#dom-mloperanddescriptor-dimensions" id="ref-for-dom-mloperanddescriptor-dimensions④③">dimensions</a></code> to [<var>shapeA</var>[0], <var>shapeB</var>[1]].</p>
+      <li data-md>
+       <p>Set <var>desc</var>.<code class="idl"><a data-link-type="idl" href="#dom-mloperanddescriptor-type" id="ref-for-dom-mloperanddescriptor-type②⑤">type</a></code> to <var>a</var>.<code class="idl"><a data-link-type="idl" href="#dom-mloperand-descriptor-slot" id="ref-for-dom-mloperand-descriptor-slot④⑥">[[descriptor]]</a></code>.<code class="idl"><a data-link-type="idl" href="#dom-mloperanddescriptor-type" id="ref-for-dom-mloperanddescriptor-type②⑥">type</a></code>.</p>
+      <li data-md>
+       <p>If any of the following sub-steps fail, <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#dfn-throw" id="ref-for-dfn-throw①③①">throw</a> an "<code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#operationerror" id="ref-for-operationerror②②">OperationError</a></code>" <code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#idl-DOMException" id="ref-for-idl-DOMException⑥⑥">DOMException</a></code>.</p>
+       <ol>
         <li data-md>
-         <p><em>alpha</em>: a <code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#idl-float" id="ref-for-idl-float①⓪">float</a></code> scalar multiplier for the first input, default to 1.0.</p>
+         <p>Let <var>output</var> be the result of <a data-link-type="dfn" href="#create-an-mloperand" id="ref-for-create-an-mloperand⑧">creating an MLOperand</a> given <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#this" id="ref-for-this①⑧">this</a> and <var>desc</var>.</p>
         <li data-md>
-         <p><em>beta</em>: a <code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#idl-float" id="ref-for-idl-float①①">float</a></code> scalar multiplier for the third input, default to 1.0.</p>
+         <p>Make a request to the underlying platform to:</p>
+         <ol>
+          <li data-md>
+           <p>Let <var>opImpl</var> be an <a data-link-type="dfn" href="https://infra.spec.whatwg.org/#implementation-defined" id="ref-for-implementation-defined②⑤">implementation-defined</a> platform operator for the GEMM operation, given <var>options</var>.</p>
+          <li data-md>
+           <p>Store a reference of <var>opImpl</var> in <var>output</var>.<code class="idl"><a data-link-type="idl" href="#dom-mloperand-operator-slot" id="ref-for-dom-mloperand-operator-slot⑧">[[operator]]</a></code>.</p>
+          <li data-md>
+           <p>Create an <a data-link-type="dfn" href="https://infra.spec.whatwg.org/#implementation-defined" id="ref-for-implementation-defined②⑥">implementation-defined</a> platform operand <var>outputImpl</var> to represent the output, given <var>output</var> and <var>opImpl</var>.</p>
+          <li data-md>
+           <p>Store a reference to <var>outputImpl</var> in <var>output</var>.<code class="idl"><a data-link-type="idl" href="#dom-mloperand-operand-slot" id="ref-for-dom-mloperand-operand-slot②⑤">[[operand]]</a></code>.</p>
+         </ol>
         <li data-md>
-         <p><em>aTranspose</em>: a <code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#idl-boolean" id="ref-for-idl-boolean②">boolean</a></code> indicating if the first input should be transposed prior to calculating the output, default to false.</p>
+         <p>Connect <var>a</var>.<code class="idl"><a data-link-type="idl" href="#dom-mloperand-operand-slot" id="ref-for-dom-mloperand-operand-slot②⑥">[[operand]]</a></code> and <var>b</var>.<code class="idl"><a data-link-type="idl" href="#dom-mloperand-operand-slot" id="ref-for-dom-mloperand-operand-slot②⑦">[[operand]]</a></code> as inputs to <var>opImpl</var>.</p>
         <li data-md>
-         <p><em>bTranspose</em>: a <code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#idl-boolean" id="ref-for-idl-boolean③">boolean</a></code> indicating if the second input should be transposed prior to calculating the output, default to false.</p>
-       </ul>
-     </ul>
-     <p><strong>Returns:</strong> an <code class="idl"><a data-link-type="idl" href="#mloperand" id="ref-for-mloperand①②②">MLOperand</a></code>. The output 2-D tensor of shape [M, N] that contains the calculated product of all the inputs.</p>
-     <div class="note" role="note">
-       The behavior of this operation can be generically emulated from the usage of other operations as follow. However, user agents typically have a more efficient implementation for it, therefore its usage is encouraged from the performance standpoint. 
+         <p>Connect <var>output</var>.<code class="idl"><a data-link-type="idl" href="#dom-mloperand-operand-slot" id="ref-for-dom-mloperand-operand-slot②⑧">[[operand]]</a></code> as output to <var>opImpl</var>.</p>
+       </ol>
+      <li data-md>
+       <p>Return <var>output</var>.</p>
+     </ol>
+    </div>
+   </details>
+   <div class="note" role="note">
+    <details open>
+     <summary> The behavior of this operation can be generically emulated from the usage of other operations as follow. However, user agents typically have a more efficient implementation for it, therefore its usage is encouraged from the performance standpoint. </summary>
 <pre class="highlight"><c- k>if</c-> <c- p>(</c->options<c- p>.</c->aTranspose<c- p>)</c->
   a <c- o>=</c-> builder<c- p>.</c->transpose<c- p>(</c->a<c- p>);</c->
 
@@ -2686,10 +4171,10 @@ <h4 class="heading settled" data-level="7.6.12" id="api-mlgraphbuilder-gemm"><sp
 <c- a>let</c-> ab <c- o>=</c-> builder<c- p>.</c->matmul<c- p>(</c->builder<c- p>.</c->mul<c- p>(</c->builder<c- p>.</c->constant<c- p>(</c->options<c- p>.</c->alpha<c- p>),</c-> a<c- p>),</c-> b<c- p>);</c->
 <c- k>return</c-> <c- p>(</c->c <c- o>?</c-> builder<c- p>.</c->add<c- p>(</c->ab<c- p>,</c-> builder<c- p>.</c->mul<c- p>(</c->builder<c- p>.</c->constant<c- p>(</c->options<c- p>.</c->beta<c- p>),</c-> c<c- p>))</c-> <c- o>:</c-> ab<c- p>);</c->
 </pre>
-     </div>
-    </div>
-    <h4 class="heading settled" data-level="7.6.13" id="api-mlgraphbuilder-gru"><span class="secno">7.6.13. </span><span class="content">The gru() method</span><a class="self-link" href="#api-mlgraphbuilder-gru"></a></h4>
-     Gated Recurrent Unit <a data-link-type="biblio" href="#biblio-gru" title="Learning Phrase Representations using RNN Encoder–Decoder for Statistical Machine Translation">[GRU]</a> recurrent network uses an update, reset, and new gate to compute the output state that rolls into the output across the temporal sequence of the network. 
+    </details>
+   </div>
+   <h4 class="heading settled" data-level="7.6.14" id="api-mlgraphbuilder-gru"><span class="secno">7.6.14. </span><span class="content">The gru() method</span><a class="self-link" href="#api-mlgraphbuilder-gru"></a></h4>
+    Gated Recurrent Unit <a data-link-type="biblio" href="#biblio-gru" title="Learning Phrase Representations using RNN Encoder–Decoder for Statistical Machine Translation">[GRU]</a> recurrent network uses an update, reset, and new gate to compute the output state that rolls into the output across the temporal sequence of the network. 
 <pre class="idl highlight def"><c- b>enum</c-> <dfn class="dfn-paneled idl-code" data-dfn-type="enum" data-export id="enumdef-mlgruweightlayout"><code><c- g>MLGruWeightLayout</c-></code></dfn> {
   <dfn class="idl-code" data-dfn-for="MLGruWeightLayout" data-dfn-type="enum-value" data-export id="dom-mlgruweightlayout-zrn"><code><c- s>"zrn"</c-></code><a class="self-link" href="#dom-mlgruweightlayout-zrn"></a></dfn>,  // update-reset-new gate ordering
   <dfn class="idl-code" data-dfn-for="MLGruWeightLayout" data-dfn-type="enum-value" data-export id="dom-mlgruweightlayout-rzn"><code><c- s>"rzn"</c-></code><a class="self-link" href="#dom-mlgruweightlayout-rzn"></a></dfn>   // reset-update-new gate ordering
@@ -2702,59 +4187,129 @@ <h4 class="heading settled" data-level="7.6.13" id="api-mlgraphbuilder-gru"><spa
 };
 
 <c- b>dictionary</c-> <dfn class="dfn-paneled idl-code" data-dfn-type="dictionary" data-export id="dictdef-mlgruoptions"><code><c- g>MLGruOptions</c-></code></dfn> {
-  <a data-link-type="idl-name" href="#mloperand" id="ref-for-mloperand①②③"><c- n>MLOperand</c-></a> <dfn class="idl-code" data-dfn-for="MLGruOptions" data-dfn-type="dict-member" data-export data-type="MLOperand " id="dom-mlgruoptions-bias"><code><c- g>bias</c-></code><a class="self-link" href="#dom-mlgruoptions-bias"></a></dfn>;
-  <a data-link-type="idl-name" href="#mloperand" id="ref-for-mloperand①②④"><c- n>MLOperand</c-></a> <dfn class="idl-code" data-dfn-for="MLGruOptions" data-dfn-type="dict-member" data-export data-type="MLOperand " id="dom-mlgruoptions-recurrentbias"><code><c- g>recurrentBias</c-></code><a class="self-link" href="#dom-mlgruoptions-recurrentbias"></a></dfn>;
-  <a data-link-type="idl-name" href="#mloperand" id="ref-for-mloperand①②⑤"><c- n>MLOperand</c-></a> <dfn class="idl-code" data-dfn-for="MLGruOptions" data-dfn-type="dict-member" data-export data-type="MLOperand " id="dom-mlgruoptions-initialhiddenstate"><code><c- g>initialHiddenState</c-></code><a class="self-link" href="#dom-mlgruoptions-initialhiddenstate"></a></dfn>;
-  <a class="idl-code" data-link-type="interface" href="https://webidl.spec.whatwg.org/#idl-boolean" id="ref-for-idl-boolean④"><c- b>boolean</c-></a> <dfn class="idl-code" data-default="true" data-dfn-for="MLGruOptions" data-dfn-type="dict-member" data-export data-type="boolean " id="dom-mlgruoptions-resetafter"><code><c- g>resetAfter</c-></code><a class="self-link" href="#dom-mlgruoptions-resetafter"></a></dfn> = <c- b>true</c->;
-  <a class="idl-code" data-link-type="interface" href="https://webidl.spec.whatwg.org/#idl-boolean" id="ref-for-idl-boolean⑤"><c- b>boolean</c-></a> <dfn class="idl-code" data-default="false" data-dfn-for="MLGruOptions" data-dfn-type="dict-member" data-export data-type="boolean " id="dom-mlgruoptions-returnsequence"><code><c- g>returnSequence</c-></code><a class="self-link" href="#dom-mlgruoptions-returnsequence"></a></dfn> = <c- b>false</c->;
-  <a data-link-type="idl-name" href="#enumdef-mlrecurrentnetworkdirection" id="ref-for-enumdef-mlrecurrentnetworkdirection"><c- n>MLRecurrentNetworkDirection</c-></a> <dfn class="idl-code" data-default="&quot;forward&quot;" data-dfn-for="MLGruOptions" data-dfn-type="dict-member" data-export data-type="MLRecurrentNetworkDirection " id="dom-mlgruoptions-direction"><code><c- g>direction</c-></code><a class="self-link" href="#dom-mlgruoptions-direction"></a></dfn> = "forward";
-  <a data-link-type="idl-name" href="#enumdef-mlgruweightlayout" id="ref-for-enumdef-mlgruweightlayout"><c- n>MLGruWeightLayout</c-></a> <dfn class="idl-code" data-default="&quot;zrn&quot;" data-dfn-for="MLGruOptions" data-dfn-type="dict-member" data-export data-type="MLGruWeightLayout " id="dom-mlgruoptions-layout"><code><c- g>layout</c-></code><a class="self-link" href="#dom-mlgruoptions-layout"></a></dfn> = "zrn";
-  <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#idl-sequence" id="ref-for-idl-sequence①⓪"><c- b>sequence</c-></a>&lt;<a data-link-type="idl-name" href="#mlactivation" id="ref-for-mlactivation①②"><c- n>MLActivation</c-></a>> <dfn class="idl-code" data-dfn-for="MLGruOptions" data-dfn-type="dict-member" data-export data-type="sequence<MLActivation> " id="dom-mlgruoptions-activations"><code><c- g>activations</c-></code><a class="self-link" href="#dom-mlgruoptions-activations"></a></dfn>;
-};
-
-<c- b>partial</c-> <c- b>interface</c-> <a class="idl-code" data-link-type="interface" href="#mlgraphbuilder" id="ref-for-mlgraphbuilder②⑨"><c- g>MLGraphBuilder</c-></a> {
-  <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#idl-sequence" id="ref-for-idl-sequence①①"><c- b>sequence</c-></a>&lt;<a data-link-type="idl-name" href="#mloperand" id="ref-for-mloperand①②⑥"><c- n>MLOperand</c-></a>> <dfn class="idl-code" data-dfn-for="MLGraphBuilder" data-dfn-type="method" data-export data-lt="gru(input, weight, recurrentWeight, steps, hiddenSize, options)|gru(input, weight, recurrentWeight, steps, hiddenSize)" id="dom-mlgraphbuilder-gru"><code><c- g>gru</c-></code><a class="self-link" href="#dom-mlgraphbuilder-gru"></a></dfn>(<a data-link-type="idl-name" href="#mloperand" id="ref-for-mloperand①②⑦"><c- n>MLOperand</c-></a> <dfn class="idl-code" data-dfn-for="MLGraphBuilder/gru(input, weight, recurrentWeight, steps, hiddenSize, options), MLGraphBuilder/gru(input, weight, recurrentWeight, steps, hiddenSize)" data-dfn-type="argument" data-export id="dom-mlgraphbuilder-gru-input-weight-recurrentweight-steps-hiddensize-options-input"><code><c- g>input</c-></code><a class="self-link" href="#dom-mlgraphbuilder-gru-input-weight-recurrentweight-steps-hiddensize-options-input"></a></dfn>, <a data-link-type="idl-name" href="#mloperand" id="ref-for-mloperand①②⑧"><c- n>MLOperand</c-></a> <dfn class="idl-code" data-dfn-for="MLGraphBuilder/gru(input, weight, recurrentWeight, steps, hiddenSize, options), MLGraphBuilder/gru(input, weight, recurrentWeight, steps, hiddenSize)" data-dfn-type="argument" data-export id="dom-mlgraphbuilder-gru-input-weight-recurrentweight-steps-hiddensize-options-weight"><code><c- g>weight</c-></code><a class="self-link" href="#dom-mlgraphbuilder-gru-input-weight-recurrentweight-steps-hiddensize-options-weight"></a></dfn>, <a data-link-type="idl-name" href="#mloperand" id="ref-for-mloperand①②⑨"><c- n>MLOperand</c-></a> <dfn class="idl-code" data-dfn-for="MLGraphBuilder/gru(input, weight, recurrentWeight, steps, hiddenSize, options), MLGraphBuilder/gru(input, weight, recurrentWeight, steps, hiddenSize)" data-dfn-type="argument" data-export id="dom-mlgraphbuilder-gru-input-weight-recurrentweight-steps-hiddensize-options-recurrentweight"><code><c- g>recurrentWeight</c-></code><a class="self-link" href="#dom-mlgraphbuilder-gru-input-weight-recurrentweight-steps-hiddensize-options-recurrentweight"></a></dfn>,
-                          <a class="idl-code" data-link-type="interface" href="https://webidl.spec.whatwg.org/#idl-unsigned-long" id="ref-for-idl-unsigned-long②⑤"><c- b>unsigned</c-> <c- b>long</c-></a> <dfn class="idl-code" data-dfn-for="MLGraphBuilder/gru(input, weight, recurrentWeight, steps, hiddenSize, options), MLGraphBuilder/gru(input, weight, recurrentWeight, steps, hiddenSize)" data-dfn-type="argument" data-export id="dom-mlgraphbuilder-gru-input-weight-recurrentweight-steps-hiddensize-options-steps"><code><c- g>steps</c-></code><a class="self-link" href="#dom-mlgraphbuilder-gru-input-weight-recurrentweight-steps-hiddensize-options-steps"></a></dfn>, <a class="idl-code" data-link-type="interface" href="https://webidl.spec.whatwg.org/#idl-unsigned-long" id="ref-for-idl-unsigned-long②⑥"><c- b>unsigned</c-> <c- b>long</c-></a> <dfn class="idl-code" data-dfn-for="MLGraphBuilder/gru(input, weight, recurrentWeight, steps, hiddenSize, options), MLGraphBuilder/gru(input, weight, recurrentWeight, steps, hiddenSize)" data-dfn-type="argument" data-export id="dom-mlgraphbuilder-gru-input-weight-recurrentweight-steps-hiddensize-options-hiddensize"><code><c- g>hiddenSize</c-></code><a class="self-link" href="#dom-mlgraphbuilder-gru-input-weight-recurrentweight-steps-hiddensize-options-hiddensize"></a></dfn>,
+  <a data-link-type="idl-name" href="#mloperand" id="ref-for-mloperand①④①"><c- n>MLOperand</c-></a> <a class="idl-code" data-link-type="dict-member" data-type="MLOperand " href="#dom-mlgruoptions-bias" id="ref-for-dom-mlgruoptions-bias"><c- g>bias</c-></a>;
+  <a data-link-type="idl-name" href="#mloperand" id="ref-for-mloperand①④②"><c- n>MLOperand</c-></a> <a class="idl-code" data-link-type="dict-member" data-type="MLOperand " href="#dom-mlgruoptions-recurrentbias" id="ref-for-dom-mlgruoptions-recurrentbias"><c- g>recurrentBias</c-></a>;
+  <a data-link-type="idl-name" href="#mloperand" id="ref-for-mloperand①④③"><c- n>MLOperand</c-></a> <a class="idl-code" data-link-type="dict-member" data-type="MLOperand " href="#dom-mlgruoptions-initialhiddenstate" id="ref-for-dom-mlgruoptions-initialhiddenstate"><c- g>initialHiddenState</c-></a>;
+  <a class="idl-code" data-link-type="interface" href="https://webidl.spec.whatwg.org/#idl-boolean" id="ref-for-idl-boolean⑥"><c- b>boolean</c-></a> <a class="idl-code" data-default="true" data-link-type="dict-member" data-type="boolean " href="#dom-mlgruoptions-resetafter" id="ref-for-dom-mlgruoptions-resetafter"><c- g>resetAfter</c-></a> = <c- b>true</c->;
+  <a class="idl-code" data-link-type="interface" href="https://webidl.spec.whatwg.org/#idl-boolean" id="ref-for-idl-boolean⑦"><c- b>boolean</c-></a> <a class="idl-code" data-default="false" data-link-type="dict-member" data-type="boolean " href="#dom-mlgruoptions-returnsequence" id="ref-for-dom-mlgruoptions-returnsequence"><c- g>returnSequence</c-></a> = <c- b>false</c->;
+  <a data-link-type="idl-name" href="#enumdef-mlrecurrentnetworkdirection" id="ref-for-enumdef-mlrecurrentnetworkdirection"><c- n>MLRecurrentNetworkDirection</c-></a> <a class="idl-code" data-default="&quot;forward&quot;" data-link-type="dict-member" data-type="MLRecurrentNetworkDirection " href="#dom-mlgruoptions-direction" id="ref-for-dom-mlgruoptions-direction"><c- g>direction</c-></a> = "forward";
+  <a data-link-type="idl-name" href="#enumdef-mlgruweightlayout" id="ref-for-enumdef-mlgruweightlayout"><c- n>MLGruWeightLayout</c-></a> <a class="idl-code" data-default="&quot;zrn&quot;" data-link-type="dict-member" data-type="MLGruWeightLayout " href="#dom-mlgruoptions-layout" id="ref-for-dom-mlgruoptions-layout"><c- g>layout</c-></a> = "zrn";
+  <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#idl-sequence" id="ref-for-idl-sequence①⓪"><c- b>sequence</c-></a>&lt;<a data-link-type="idl-name" href="#mlactivation" id="ref-for-mlactivation②③"><c- n>MLActivation</c-></a>> <a class="idl-code" data-link-type="dict-member" data-type="sequence<MLActivation> " href="#dom-mlgruoptions-activations" id="ref-for-dom-mlgruoptions-activations"><c- g>activations</c-></a>;
+};
+
+<c- b>partial</c-> <c- b>interface</c-> <a class="idl-code" data-link-type="interface" href="#mlgraphbuilder" id="ref-for-mlgraphbuilder③②"><c- g>MLGraphBuilder</c-></a> {
+  <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#idl-sequence" id="ref-for-idl-sequence①①"><c- b>sequence</c-></a>&lt;<a data-link-type="idl-name" href="#mloperand" id="ref-for-mloperand①④④"><c- n>MLOperand</c-></a>> <a class="idl-code" data-link-type="method" href="#dom-mlgraphbuilder-gru" id="ref-for-dom-mlgraphbuilder-gru"><c- g>gru</c-></a>(<a data-link-type="idl-name" href="#mloperand" id="ref-for-mloperand①④⑤"><c- n>MLOperand</c-></a> <dfn class="idl-code" data-dfn-for="MLGraphBuilder/gru(input, weight, recurrentWeight, steps, hiddenSize, options), MLGraphBuilder/gru(input, weight, recurrentWeight, steps, hiddenSize)" data-dfn-type="argument" data-export id="dom-mlgraphbuilder-gru-input-weight-recurrentweight-steps-hiddensize-options-input"><code><c- g>input</c-></code><a class="self-link" href="#dom-mlgraphbuilder-gru-input-weight-recurrentweight-steps-hiddensize-options-input"></a></dfn>, <a data-link-type="idl-name" href="#mloperand" id="ref-for-mloperand①④⑥"><c- n>MLOperand</c-></a> <dfn class="idl-code" data-dfn-for="MLGraphBuilder/gru(input, weight, recurrentWeight, steps, hiddenSize, options), MLGraphBuilder/gru(input, weight, recurrentWeight, steps, hiddenSize)" data-dfn-type="argument" data-export id="dom-mlgraphbuilder-gru-input-weight-recurrentweight-steps-hiddensize-options-weight"><code><c- g>weight</c-></code><a class="self-link" href="#dom-mlgraphbuilder-gru-input-weight-recurrentweight-steps-hiddensize-options-weight"></a></dfn>, <a data-link-type="idl-name" href="#mloperand" id="ref-for-mloperand①④⑦"><c- n>MLOperand</c-></a> <dfn class="idl-code" data-dfn-for="MLGraphBuilder/gru(input, weight, recurrentWeight, steps, hiddenSize, options), MLGraphBuilder/gru(input, weight, recurrentWeight, steps, hiddenSize)" data-dfn-type="argument" data-export id="dom-mlgraphbuilder-gru-input-weight-recurrentweight-steps-hiddensize-options-recurrentweight"><code><c- g>recurrentWeight</c-></code><a class="self-link" href="#dom-mlgraphbuilder-gru-input-weight-recurrentweight-steps-hiddensize-options-recurrentweight"></a></dfn>,
+                          <a class="idl-code" data-link-type="interface" href="https://webidl.spec.whatwg.org/#idl-unsigned-long" id="ref-for-idl-unsigned-long②⑦"><c- b>unsigned</c-> <c- b>long</c-></a> <dfn class="idl-code" data-dfn-for="MLGraphBuilder/gru(input, weight, recurrentWeight, steps, hiddenSize, options), MLGraphBuilder/gru(input, weight, recurrentWeight, steps, hiddenSize)" data-dfn-type="argument" data-export id="dom-mlgraphbuilder-gru-input-weight-recurrentweight-steps-hiddensize-options-steps"><code><c- g>steps</c-></code><a class="self-link" href="#dom-mlgraphbuilder-gru-input-weight-recurrentweight-steps-hiddensize-options-steps"></a></dfn>, <a class="idl-code" data-link-type="interface" href="https://webidl.spec.whatwg.org/#idl-unsigned-long" id="ref-for-idl-unsigned-long②⑧"><c- b>unsigned</c-> <c- b>long</c-></a> <dfn class="idl-code" data-dfn-for="MLGraphBuilder/gru(input, weight, recurrentWeight, steps, hiddenSize, options), MLGraphBuilder/gru(input, weight, recurrentWeight, steps, hiddenSize)" data-dfn-type="argument" data-export id="dom-mlgraphbuilder-gru-input-weight-recurrentweight-steps-hiddensize-options-hiddensize"><code><c- g>hiddenSize</c-></code><a class="self-link" href="#dom-mlgraphbuilder-gru-input-weight-recurrentweight-steps-hiddensize-options-hiddensize"></a></dfn>,
                           <c- b>optional</c-> <a data-link-type="idl-name" href="#dictdef-mlgruoptions" id="ref-for-dictdef-mlgruoptions"><c- n>MLGruOptions</c-></a> <dfn class="idl-code" data-dfn-for="MLGraphBuilder/gru(input, weight, recurrentWeight, steps, hiddenSize, options), MLGraphBuilder/gru(input, weight, recurrentWeight, steps, hiddenSize)" data-dfn-type="argument" data-export id="dom-mlgraphbuilder-gru-input-weight-recurrentweight-steps-hiddensize-options-options"><code><c- g>options</c-></code><a class="self-link" href="#dom-mlgraphbuilder-gru-input-weight-recurrentweight-steps-hiddensize-options-options"></a></dfn> = {});
 };
 </pre>
-    <div class="algorithm" data-algorithm="gru">
-      <strong>Arguments:</strong> 
-     <ul>
-      <li data-md>
-       <p><em>input</em>: an <code class="idl"><a data-link-type="idl" href="#mloperand" id="ref-for-mloperand①③⓪">MLOperand</a></code>. The input 3-D tensor of shape [steps, batch_size, input_size].</p>
-      <li data-md>
-       <p><em>weight</em>: an <code class="idl"><a data-link-type="idl" href="#mloperand" id="ref-for-mloperand①③①">MLOperand</a></code>. The 3-D input weight tensor of shape [num_directions, 3 * hidden_size, input_size]. The ordering of the weight vectors in the second dimension of the tensor shape is specified according to the <em>options.layout</em> argument.</p>
-      <li data-md>
-       <p><em>recurrentWeight</em>: an <code class="idl"><a data-link-type="idl" href="#mloperand" id="ref-for-mloperand①③②">MLOperand</a></code>. The 3-D recurrent weight tensor of shape [num_directions, 3 * hidden_size, hidden_size]. The ordering of the weight vectors in the second dimension of the tensor shape is specified according to the <em>options.layout</em> argument.</p>
+   <p><code class="idl"><a data-link-type="idl" href="#dictdef-mlgruoptions" id="ref-for-dictdef-mlgruoptions①">MLGruOptions</a></code> has the following members:</p>
+   <dl>
+    <dt data-md><dfn class="dfn-paneled idl-code" data-dfn-for="MLGruOptions" data-dfn-type="dict-member" data-export id="dom-mlgruoptions-bias"><code>bias</code></dfn>, <span> of type <a data-link-type="idl-name" href="#mloperand" id="ref-for-mloperand①④⑧">MLOperand</a></span>
+    <dd data-md>
+     <p>An <code class="idl"><a data-link-type="idl" href="#mloperand" id="ref-for-mloperand①④⑨">MLOperand</a></code>. Specifies the 2-D input bias tensor of shape [numDirections, 3 * hiddenSize]. The ordering of the bias vectors in the second dimension of the tensor shape is specified according to the <code class="idl"><a data-link-type="idl" href="#dom-mlgruoptions-layout" id="ref-for-dom-mlgruoptions-layout①">layout</a></code> argument.</p>
+    <dt data-md><dfn class="dfn-paneled idl-code" data-dfn-for="MLGruOptions" data-dfn-type="dict-member" data-export id="dom-mlgruoptions-recurrentbias"><code>recurrentBias</code></dfn>, <span> of type <a data-link-type="idl-name" href="#mloperand" id="ref-for-mloperand①⑤⓪">MLOperand</a></span>
+    <dd data-md>
+     <p>An <code class="idl"><a data-link-type="idl" href="#mloperand" id="ref-for-mloperand①⑤①">MLOperand</a></code>. Specifies the 2-D recurrent bias tensor of shape [numDirections, 3 * hiddenSize]. The ordering of the bias vectors in the second dimension of the tensor shape is specified according to the <code class="idl"><a data-link-type="idl" href="#dom-mlgruoptions-layout" id="ref-for-dom-mlgruoptions-layout②">layout</a></code> argument.</p>
+    <dt data-md><dfn class="dfn-paneled idl-code" data-dfn-for="MLGruOptions" data-dfn-type="dict-member" data-export id="dom-mlgruoptions-initialhiddenstate"><code>initialHiddenState</code></dfn>, <span> of type <a data-link-type="idl-name" href="#mloperand" id="ref-for-mloperand①⑤②">MLOperand</a></span>
+    <dd data-md>
+     <p>An <code class="idl"><a data-link-type="idl" href="#mloperand" id="ref-for-mloperand①⑤③">MLOperand</a></code>. The 3-D initial hidden state tensor of shape [numDirections, batchSize, hiddenSize].
+When not specified, implementations SHOULD use a tensor filled with zero.</p>
+    <dt data-md><dfn class="dfn-paneled idl-code" data-dfn-for="MLGruOptions" data-dfn-type="dict-member" data-export id="dom-mlgruoptions-resetafter"><code>resetAfter</code></dfn>, <span> of type <a data-link-type="idl-name" href="https://webidl.spec.whatwg.org/#idl-boolean" id="ref-for-idl-boolean⑧">boolean</a>, defaulting to <code>true</code></span>
+    <dd data-md>
+     <p>A <code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#idl-boolean" id="ref-for-idl-boolean⑨">boolean</a></code> indicating whether to apply the reset gate after or before matrix multiplication. The default value is true.</p>
+    <dt data-md><dfn class="dfn-paneled idl-code" data-dfn-for="MLGruOptions" data-dfn-type="dict-member" data-export id="dom-mlgruoptions-returnsequence"><code>returnSequence</code></dfn>, <span> of type <a data-link-type="idl-name" href="https://webidl.spec.whatwg.org/#idl-boolean" id="ref-for-idl-boolean①⓪">boolean</a>, defaulting to <code>false</code></span>
+    <dd data-md>
+     <p>A <code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#idl-boolean" id="ref-for-idl-boolean①①">boolean</a></code> indicating whether to also return the entire sequence with every output from each time step in it in addition to the output of the last time step.
+The default value is false.</p>
+    <dt data-md><dfn class="dfn-paneled idl-code" data-dfn-for="MLGruOptions" data-dfn-type="dict-member" data-export id="dom-mlgruoptions-direction"><code>direction</code></dfn>, <span> of type <a data-link-type="idl-name" href="#enumdef-mlrecurrentnetworkdirection" id="ref-for-enumdef-mlrecurrentnetworkdirection①">MLRecurrentNetworkDirection</a>, defaulting to <code>"forward"</code></span>
+    <dd data-md>
+     <p>An <code class="idl"><a data-link-type="idl" href="#enumdef-mlrecurrentnetworkdirection" id="ref-for-enumdef-mlrecurrentnetworkdirection②">MLRecurrentNetworkDirection</a></code>. Specifies the processing direction of the input sequence. When set to <code>"both"</code>, the size of the first dimension of the weight and the bias tensor shapes must be 2, and the input is processed in both directions.</p>
+    <dt data-md><dfn class="dfn-paneled idl-code" data-dfn-for="MLGruOptions" data-dfn-type="dict-member" data-export id="dom-mlgruoptions-layout"><code>layout</code></dfn>, <span> of type <a data-link-type="idl-name" href="#enumdef-mlgruweightlayout" id="ref-for-enumdef-mlgruweightlayout①">MLGruWeightLayout</a>, defaulting to <code>"zrn"</code></span>
+    <dd data-md>
+     <p>An <code class="idl"><a data-link-type="idl" href="#enumdef-mlgruweightlayout" id="ref-for-enumdef-mlgruweightlayout②">MLGruWeightLayout</a></code>. The ordering of the weight and bias vectors for the internal gates of GRU, specifically the <code>update (z)</code>, <code>reset (r)</code>, and <code>new (n)</code> gate, as indicated in the second dimension of the weight and bias tensor shape. When not specified, the default layout is <code>"zrn"</code>.bias</p>
+    <dt data-md><dfn class="dfn-paneled idl-code" data-dfn-for="MLGruOptions" data-dfn-type="dict-member" data-export id="dom-mlgruoptions-activations"><code>activations</code></dfn>, <span> of type sequence&lt;<a data-link-type="idl-name" href="#mlactivation" id="ref-for-mlactivation②④">MLActivation</a>></span>
+    <dd data-md>
+     <p>A sequence of <code class="idl"><a data-link-type="idl" href="#mlactivation" id="ref-for-mlactivation②⑤">MLActivation</a></code>. Specifies a pair of activation functions with the first function used for the update and reset gate, and the second used for the new gate. When not specified, implementations SHOULD use the the pair of sigmoid (<code>"sigmoid"</code>) and the hyperbolic tangent (<code>"tanh"</code>) functions, respectively.</p>
+   </dl>
+   <div>
+     <strong>Arguments:</strong> 
+    <ul>
+     <li data-md>
+      <p><em>input</em>: an <code class="idl"><a data-link-type="idl" href="#mloperand" id="ref-for-mloperand①⑤④">MLOperand</a></code>. The input 3-D tensor of shape [steps, batchSize, inputSize].</p>
+     <li data-md>
+      <p><em>weight</em>: an <code class="idl"><a data-link-type="idl" href="#mloperand" id="ref-for-mloperand①⑤⑤">MLOperand</a></code>. The 3-D input weight tensor of shape [numDirections, 3 * hiddenSize, inputSize]. The ordering of the weight vectors in the second dimension of the tensor shape is specified according to the <var>options</var>.<code class="idl"><a data-link-type="idl" href="#dom-mlgruoptions-layout" id="ref-for-dom-mlgruoptions-layout③">layout</a></code> argument.</p>
+     <li data-md>
+      <p><em>recurrentWeight</em>: an <code class="idl"><a data-link-type="idl" href="#mloperand" id="ref-for-mloperand①⑤⑥">MLOperand</a></code>. The 3-D recurrent weight tensor of shape [numDirections, 3 * hiddenSize, hiddenSize]. The ordering of the weight vectors in the second dimension of the tensor shape is specified according to the <var>options</var>.<code class="idl"><a data-link-type="idl" href="#dom-mlgruoptions-layout" id="ref-for-dom-mlgruoptions-layout④">layout</a></code> argument.</p>
+     <li data-md>
+      <p><em>steps</em>: an <code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#idl-unsigned-long" id="ref-for-idl-unsigned-long②⑨">unsigned long</a></code> scalar. The number of time steps in the recurrent network. The value must be greater than 0.</p>
+     <li data-md>
+      <p><em>hiddenSize</em>: an <code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#idl-unsigned-long" id="ref-for-idl-unsigned-long③⓪">unsigned long</a></code> scalar. The value of the third dimension of the cell output tensor shape. It indicates the number of features in the hidden state.</p>
+     <li data-md>
+      <p><em>options</em>: an optional <code class="idl"><a data-link-type="idl" href="#dictdef-mlgruoptions" id="ref-for-dictdef-mlgruoptions②">MLGruOptions</a></code>. The optional parameters of the operation.</p>
+    </ul>
+    <p><strong>Returns:</strong> a sequence of <code class="idl"><a data-link-type="idl" href="#mloperand" id="ref-for-mloperand①⑤⑦">MLOperand</a></code>. The first element of the sequence is a 3-D tensor of shape [numDirections, batchSize, hiddenSize], the cell output from the last time step of the network. Additionally, if <var>options</var>.<code class="idl"><a data-link-type="idl" href="#dom-mlgruoptions-returnsequence" id="ref-for-dom-mlgruoptions-returnsequence①">returnSequence</a></code> is set to true, the second element is the 4-D output tensor of shape [steps, numDirections, batchSize, hiddenSize] containing every cell outputs from each time step in the temporal sequence.</p>
+   </div>
+   <details class="algorithm" data-algorithm="gru(input, weight, recurrentWeight, steps, hiddenSize, options)" data-algorithm-for="MLGraphBuilder" open>
+    <summary> The <dfn class="dfn-paneled idl-code" data-dfn-for="MLGraphBuilder" data-dfn-type="method" data-export data-lt="gru(input, weight, recurrentWeight, steps, hiddenSize, options)|gru(input, weight, recurrentWeight, steps, hiddenSize)" id="dom-mlgraphbuilder-gru"><code>gru(<var>input</var>, <var>weight</var>, <var>recurrentWeight</var>, <var>steps</var>, <var>hiddenSize</var>, <var>options</var>)</code></dfn> method steps are: </summary>
+    <div class="algorithm-steps">
+     <ol>
       <li data-md>
-       <p><em>steps</em>: an <code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#idl-unsigned-long" id="ref-for-idl-unsigned-long②⑦">unsigned long</a></code> scalar. The number of time steps in the recurrent network. The value must be greater than 0.</p>
+       <p><a data-link-type="dfn" href="https://infra.spec.whatwg.org/#assert" id="ref-for-assert③③">Assert</a>: the type of <var>input</var>, <var>weight</var> and <var>recurrentWeight</var> is <code class="idl"><a data-link-type="idl" href="#mloperand" id="ref-for-mloperand①⑤⑧">MLOperand</a></code>.</p>
       <li data-md>
-       <p><em>hiddenSize</em>: an <code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#idl-unsigned-long" id="ref-for-idl-unsigned-long②⑧">unsigned long</a></code> scalar. The value of the third dimension of the cell output tensor shape. It indicates the number of features in the hidden state.</p>
+       <p>If the <a data-link-type="dfn" href="#mloperand-rank" id="ref-for-mloperand-rank⑧">rank</a> of <var>input</var> or <var>weight</var> or <var>recurrentWeight</var> is not 3, then <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#dfn-throw" id="ref-for-dfn-throw①③②">throw</a> a "<code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#dataerror" id="ref-for-dataerror③⑤">DataError</a></code>" <code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#idl-DOMException" id="ref-for-idl-DOMException⑥⑦">DOMException</a></code>.</p>
       <li data-md>
-       <p><em>options</em>: an optional <code class="idl"><a data-link-type="idl" href="#dictdef-mlgruoptions" id="ref-for-dictdef-mlgruoptions①">MLGruOptions</a></code>. The optional parameters of the operation.</p>
-       <ul>
+       <p>If <var>options</var>.<code class="idl"><a data-link-type="idl" href="#dom-mlgruoptions-bias" id="ref-for-dom-mlgruoptions-bias①">bias</a></code> <a data-link-type="dfn" href="https://infra.spec.whatwg.org/#map-exists" id="ref-for-map-exists②⑧">exists</a>.</p>
+       <ol>
+        <li data-md>
+         <p><a data-link-type="dfn" href="https://infra.spec.whatwg.org/#assert" id="ref-for-assert③④">Assert</a>: its type is <code class="idl"><a data-link-type="idl" href="#mloperand" id="ref-for-mloperand①⑤⑨">MLOperand</a></code>.</p>
         <li data-md>
-         <p><em>bias</em>: an <code class="idl"><a data-link-type="idl" href="#mloperand" id="ref-for-mloperand①③③">MLOperand</a></code>. The 2-D input bias tensor of shape [num_directions, 3 * hidden_size]. The ordering of the bias vectors in the second dimension of the tensor shape is specified according to the <em>options.layout</em> argument.</p>
+         <p>If <var>options</var>.<code class="idl"><a data-link-type="idl" href="#dom-mlgruoptions-bias" id="ref-for-dom-mlgruoptions-bias②">bias</a></code>.<code class="idl"><a data-link-type="idl" href="#dom-mloperand-descriptor-slot" id="ref-for-dom-mloperand-descriptor-slot④⑦">[[descriptor]]</a></code>.<code class="idl"><a data-link-type="idl" href="#dom-mloperanddescriptor-dimensions" id="ref-for-dom-mloperanddescriptor-dimensions④④">dimensions</a></code>[1] is not equal to 3 * <var>hiddenSize</var>, then <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#dfn-throw" id="ref-for-dfn-throw①③③">throw</a> a "<code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#dataerror" id="ref-for-dataerror③⑥">DataError</a></code>" <code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#idl-DOMException" id="ref-for-idl-DOMException⑥⑧">DOMException</a></code>.</p>
+       </ol>
+      <li data-md>
+       <p>If <var>options</var>.<code class="idl"><a data-link-type="idl" href="#dom-mlgruoptions-recurrentbias" id="ref-for-dom-mlgruoptions-recurrentbias①">recurrentBias</a></code> <a data-link-type="dfn" href="https://infra.spec.whatwg.org/#map-exists" id="ref-for-map-exists②⑨">exists</a>.</p>
+       <ol>
         <li data-md>
-         <p><em>recurrentBias</em>: an <code class="idl"><a data-link-type="idl" href="#mloperand" id="ref-for-mloperand①③④">MLOperand</a></code>. The 2-D recurrent bias tensor of shape [num_directions, 3 * hidden_size]. The ordering of the bias vectors in the second dimension of the tensor shape is specified according to the <em>options.layout</em> argument.</p>
+         <p><a data-link-type="dfn" href="https://infra.spec.whatwg.org/#assert" id="ref-for-assert③⑤">Assert</a>: its type is <code class="idl"><a data-link-type="idl" href="#mloperand" id="ref-for-mloperand①⑥⓪">MLOperand</a></code>.</p>
         <li data-md>
-         <p><em>initialHiddenState</em>: an <code class="idl"><a data-link-type="idl" href="#mloperand" id="ref-for-mloperand①③⑤">MLOperand</a></code>. The 3-D initial hidden state tensor of shape [num_directions, batch_size, hidden_size]. When not specified, it’s assumed to be a tensor filled with zero.</p>
+         <p>If <var>options</var>.<code class="idl"><a data-link-type="idl" href="#dom-mlgruoptions-recurrentbias" id="ref-for-dom-mlgruoptions-recurrentbias②">recurrentBias</a></code>.<code class="idl"><a data-link-type="idl" href="#dom-mloperand-descriptor-slot" id="ref-for-dom-mloperand-descriptor-slot④⑧">[[descriptor]]</a></code>.<code class="idl"><a data-link-type="idl" href="#dom-mloperanddescriptor-dimensions" id="ref-for-dom-mloperanddescriptor-dimensions④⑤">dimensions</a></code>[1] is not equal to 3 * <var>hiddenSize</var>, then <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#dfn-throw" id="ref-for-dfn-throw①③④">throw</a> a "<code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#dataerror" id="ref-for-dataerror③⑦">DataError</a></code>" <code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#idl-DOMException" id="ref-for-idl-DOMException⑥⑨">DOMException</a></code>.</p>
+       </ol>
+      <li data-md>
+       <p>If <var>options</var>.<code class="idl"><a data-link-type="idl" href="#dom-mlgruoptions-initialhiddenstate" id="ref-for-dom-mlgruoptions-initialhiddenstate①">initialHiddenState</a></code> <a data-link-type="dfn" href="https://infra.spec.whatwg.org/#map-exists" id="ref-for-map-exists③⓪">exists</a>.</p>
+       <ol>
         <li data-md>
-         <p><em>resetAfter</em>: a <code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#idl-boolean" id="ref-for-idl-boolean⑥">boolean</a></code> indicating whether to apply the reset gate after or before matrix multiplication. Default to true.</p>
+         <p><a data-link-type="dfn" href="https://infra.spec.whatwg.org/#assert" id="ref-for-assert③⑥">Assert</a>: its type is <code class="idl"><a data-link-type="idl" href="#mloperand" id="ref-for-mloperand①⑥①">MLOperand</a></code>.</p>
         <li data-md>
-         <p><em>returnSequence</em>: a <code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#idl-boolean" id="ref-for-idl-boolean⑦">boolean</a></code> indicating whether to also return the entire sequence with every output from each time step in it in addition to the output of the last time step. Default to false.</p>
+         <p>If its rank is not 3, then <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#dfn-throw" id="ref-for-dfn-throw①③⑤">throw</a> a "<code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#dataerror" id="ref-for-dataerror③⑧">DataError</a></code>" <code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#idl-DOMException" id="ref-for-idl-DOMException⑦⓪">DOMException</a></code>.</p>
+       </ol>
+      <li data-md>
+       <p>If <var>options</var>.<code class="idl"><a data-link-type="idl" href="#dom-mlgruoptions-activations" id="ref-for-dom-mlgruoptions-activations①">activations</a></code> <a data-link-type="dfn" href="https://infra.spec.whatwg.org/#map-exists" id="ref-for-map-exists③①">exists</a> and its <a data-link-type="dfn" href="https://infra.spec.whatwg.org/#list-size" id="ref-for-list-size②⑥">size</a> is not 2, then <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#dfn-throw" id="ref-for-dfn-throw①③⑥">throw</a> a <code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#exceptiondef-typeerror" id="ref-for-exceptiondef-typeerror③③">TypeError</a></code>.</p>
+      <li data-md>
+       <p>If <var>steps</var> is not equal to <var>input</var>.<code class="idl"><a data-link-type="idl" href="#dom-mloperand-descriptor-slot" id="ref-for-dom-mloperand-descriptor-slot④⑨">[[descriptor]]</a></code>.<code class="idl"><a data-link-type="idl" href="#dom-mloperanddescriptor-dimensions" id="ref-for-dom-mloperanddescriptor-dimensions④⑥">dimensions</a></code>[0], then <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#dfn-throw" id="ref-for-dfn-throw①③⑦">throw</a> a <code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#exceptiondef-typeerror" id="ref-for-exceptiondef-typeerror③④">TypeError</a></code>.</p>
+      <li data-md>
+       <p>Let <var>output</var> be an empty sequence of <code class="idl"><a data-link-type="idl" href="#mloperand" id="ref-for-mloperand①⑥②">MLOperand</a></code> objects.</p>
+      <li data-md>
+       <p>If any of the following sub-steps fail, <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#dfn-throw" id="ref-for-dfn-throw①③⑧">throw</a> an "<code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#operationerror" id="ref-for-operationerror②③">OperationError</a></code>" <code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#idl-DOMException" id="ref-for-idl-DOMException⑦①">DOMException</a></code>.</p>
+       <ol>
         <li data-md>
-         <p><em>direction</em>: an <code class="idl"><a data-link-type="idl" href="#enumdef-mlrecurrentnetworkdirection" id="ref-for-enumdef-mlrecurrentnetworkdirection①">MLRecurrentNetworkDirection</a></code>. The processing direction of the input sequence. When set to <em>"both"</em>, the size of the first dimension of the weight and the bias tensor shapes must be 2, and the input is processed in both directions.</p>
+         <p>Make a request to the underlying platform to:</p>
+         <ol>
+          <li data-md>
+           <p>Let <var>opImpl</var> be an <a data-link-type="dfn" href="https://infra.spec.whatwg.org/#implementation-defined" id="ref-for-implementation-defined②⑦">implementation-defined</a> platform operator for <code>"gru"</code>, given <var>weight</var>, <var>recurrentWeight</var>, <var>steps</var>, <var>hiddenSize</var> and <var>options</var> as parameters.</p>
+         </ol>
         <li data-md>
-         <p><em>layout</em>: an <code class="idl"><a data-link-type="idl" href="#enumdef-mlgruweightlayout" id="ref-for-enumdef-mlgruweightlayout①">MLGruWeightLayout</a></code>. The ordering of the weight and bias vectors for the internal gates of GRU, specifically the <em>update (z)</em>, <em>reset (r)</em>, and <em>new (n)</em> gate, as indicated in the second dimension of the weight and bias tensor shape. When not specified, the default layout is <em>"zrn"</em>.</p>
+         <p>Connect <var>input</var>.<code class="idl"><a data-link-type="idl" href="#dom-mloperand-operand-slot" id="ref-for-dom-mloperand-operand-slot②⑨">[[operand]]</a></code> as input to <var>opImpl</var>.</p>
         <li data-md>
-         <p><em>activations</em>: a sequence of <code class="idl"><a data-link-type="idl" href="#mlactivation" id="ref-for-mlactivation①③">MLActivation</a></code>. A pair of activation functions with the first function used for the update and reset gate, and the second used for the new gate. When not specified, it’s assumed to be the sigmoid (<em>"sigmoid"</em>) and the hyperbolic tangent (<em>"tanh"</em>) function respectively.</p>
-       </ul>
-     </ul>
-     <p><strong>Returns:</strong> a sequence of <code class="idl"><a data-link-type="idl" href="#mloperand" id="ref-for-mloperand①③⑥">MLOperand</a></code>. The first element of the sequence is a 3-D tensor of shape [num_directions, batch_size, hidden_size], the cell output from the last time step of the network. Additionally, if <em>options.returnSequence</em> is set to true, the second element is the 4-D output tensor of shape [steps, num_directions, batch_size, hidden_size] containing every cell outputs from each time step in the temporal sequence.</p>
-     <div class="note" role="note">
-       The behavior of this operation can be generically emulated from the usage of other operations as follows. However, user agents typically have a more efficient implementation for it, therefore its usage is encouraged from the performance standpoint. 
+         <p>Connect <var>output</var> as output to <var>opImpl</var>.</p>
+       </ol>
+      <li data-md>
+       <p>Return <var>output</var>.</p>
+     </ol>
+    </div>
+   </details>
+   <div class="note" role="note">
+    <details open>
+     <summary> The behavior of this operation can be generically emulated from the usage of other operations as follows. However, user agents typically have a more efficient implementation for it, therefore its usage is encouraged from the performance standpoint. </summary>
 <pre class="highlight"><c- a>const</c-> numDirections <c- o>=</c-> <c- p>(</c->options<c- p>.</c->direction <c- o>==</c-> <c- u>"both"</c-> <c- o>?</c-> <c- mf>2</c-> <c- o>:</c-> <c- mf>1</c-><c- p>);</c->
 <c- a>let</c-> hiddenState <c- o>=</c-> options<c- p>.</c->initialHiddenState<c- p>;</c->
 
@@ -2771,11 +4326,11 @@ <h4 class="heading settled" data-level="7.6.13" id="api-mlgraphbuilder-gru"><spa
 <c- a>let</c-> currentRecurrentBias <c- o>=</c-> <c- p>[];</c->
 
 <c- k>for</c-> <c- p>(</c-><c- a>let</c-> dir <c- o>=</c-> <c- mf>0</c-><c- p>;</c-> dir <c- o>&lt;</c-> numDirections<c- p>;</c-> <c- o>++</c->dir<c- p>)</c-> <c- p>{</c->
-  currentWeight<c- p>.</c->push<c- p>(</c->builder<c- p>.</c->squeeze<c- p>(</c->builder<c- p>.</c->slice<c- p>(</c->weight<c- p>,</c-> <c- p>[</c->dir<c- p>,</c-> <c- mf>0</c-><c- p>,</c-> <c- mf>0</c-><c- p>],</c-> <c- p>[</c-><c- mf>1</c-><c- p>,</c-> <c- mf>3</c-> <c- o>*</c-> hidden_size<c- p>,</c-> input_size<c- p>]),</c-> <c- p>{</c-> axes<c- o>:</c-> <c- p>[</c-><c- mf>0</c-><c- p>]</c-> <c- p>}));</c->
-  currentRecurrentWeight<c- p>.</c->push<c- p>(</c->builder<c- p>.</c->squeeze<c- p>(</c->builder<c- p>.</c->slice<c- p>(</c->recurrentWeight<c- p>,</c-> <c- p>[</c->dir<c- p>,</c-> <c- mf>0</c-><c- p>,</c-> <c- mf>0</c-><c- p>],</c-> <c- p>[</c-><c- mf>1</c-><c- p>,</c-> <c- mf>3</c-> <c- o>*</c-> hidden_size<c- p>,</c-> hidden_size<c- p>]),</c-> <c- p>{</c-> axes<c- o>:</c-> <c- p>[</c-><c- mf>0</c-><c- p>]</c-> <c- p>}));</c->
-  currentBias<c- p>.</c->push<c- p>(</c->options<c- p>.</c->bias <c- o>?</c-> <c- p>(</c->builder<c- p>.</c->squeeze<c- p>(</c->builder<c- p>.</c->slice<c- p>(</c->options<c- p>.</c->bias<c- p>,</c-> <c- p>[</c->dir<c- p>,</c-> <c- mf>0</c-><c- p>],</c-> <c- p>[</c-><c- mf>1</c-><c- p>,</c-> <c- mf>3</c-> <c- o>*</c-> hidden_size<c- p>]),</c-> <c- p>{</c-> axes<c- o>:</c-> <c- p>[</c-><c- mf>0</c-><c- p>]</c-> <c- p>}))</c-> <c- o>:</c-> <c- kc>null</c-><c- p>);</c->
+  currentWeight<c- p>.</c->push<c- p>(</c->builder<c- p>.</c->squeeze<c- p>(</c->builder<c- p>.</c->slice<c- p>(</c->weight<c- p>,</c-> <c- p>[</c->dir<c- p>,</c-> <c- mf>0</c-><c- p>,</c-> <c- mf>0</c-><c- p>],</c-> <c- p>[</c-><c- mf>1</c-><c- p>,</c-> <c- mf>3</c-> <c- o>*</c-> hiddenSize<c- p>,</c-> inputSize<c- p>]),</c-> <c- p>{</c-> axes<c- o>:</c-> <c- p>[</c-><c- mf>0</c-><c- p>]</c-> <c- p>}));</c->
+  currentRecurrentWeight<c- p>.</c->push<c- p>(</c->builder<c- p>.</c->squeeze<c- p>(</c->builder<c- p>.</c->slice<c- p>(</c->recurrentWeight<c- p>,</c-> <c- p>[</c->dir<c- p>,</c-> <c- mf>0</c-><c- p>,</c-> <c- mf>0</c-><c- p>],</c-> <c- p>[</c-><c- mf>1</c-><c- p>,</c-> <c- mf>3</c-> <c- o>*</c-> hiddenSize<c- p>,</c-> hiddenSize<c- p>]),</c-> <c- p>{</c-> axes<c- o>:</c-> <c- p>[</c-><c- mf>0</c-><c- p>]</c-> <c- p>}));</c->
+  currentBias<c- p>.</c->push<c- p>(</c->options<c- p>.</c->bias <c- o>?</c-> <c- p>(</c->builder<c- p>.</c->squeeze<c- p>(</c->builder<c- p>.</c->slice<c- p>(</c->options<c- p>.</c->bias<c- p>,</c-> <c- p>[</c->dir<c- p>,</c-> <c- mf>0</c-><c- p>],</c-> <c- p>[</c-><c- mf>1</c-><c- p>,</c-> <c- mf>3</c-> <c- o>*</c-> hiddenSize<c- p>]),</c-> <c- p>{</c-> axes<c- o>:</c-> <c- p>[</c-><c- mf>0</c-><c- p>]</c-> <c- p>}))</c-> <c- o>:</c-> <c- kc>null</c-><c- p>);</c->
   currentRecurrentBias<c- p>.</c->push<c- p>(</c->options<c- p>.</c->recurrentBias <c- o>?</c->
-    <c- p>(</c->builder<c- p>.</c->squeeze<c- p>(</c->builder<c- p>.</c->slice<c- p>(</c->options<c- p>.</c->recurrentBias<c- p>,</c-> <c- p>[</c->dir<c- p>,</c-> <c- mf>0</c-><c- p>],</c-> <c- p>[</c-><c- mf>1</c-><c- p>,</c-> <c- mf>3</c-> <c- o>*</c-> hidden_size<c- p>]),</c-> <c- p>{</c-> axes<c- o>:</c-> <c- p>[</c-><c- mf>0</c-><c- p>]</c-> <c- p>}))</c-> <c- o>:</c-> <c- kc>null</c-><c- p>);</c->
+    <c- p>(</c->builder<c- p>.</c->squeeze<c- p>(</c->builder<c- p>.</c->slice<c- p>(</c->options<c- p>.</c->recurrentBias<c- p>,</c-> <c- p>[</c->dir<c- p>,</c-> <c- mf>0</c-><c- p>],</c-> <c- p>[</c-><c- mf>1</c-><c- p>,</c-> <c- mf>3</c-> <c- o>*</c-> hiddenSize<c- p>]),</c-> <c- p>{</c-> axes<c- o>:</c-> <c- p>[</c-><c- mf>0</c-><c- p>]</c-> <c- p>}))</c-> <c- o>:</c-> <c- kc>null</c-><c- p>);</c->
 <c- p>}</c->
 
 <c- k>for</c-> <c- p>(</c-><c- a>let</c-> step <c- o>=</c-> <c- mf>0</c-><c- p>;</c-> step <c- o>&lt;</c-> steps<c- p>;</c-> <c- o>++</c->step<c- p>)</c-> <c- p>{</c->
@@ -2783,12 +4338,12 @@ <h4 class="heading settled" data-level="7.6.13" id="api-mlgraphbuilder-gru"><spa
   <c- a>let</c-> currentOutput <c- o>=</c-> <c- kc>null</c-><c- p>;</c->
 
   <c- k>for</c-> <c- p>(</c-><c- a>let</c-> dir <c- o>=</c-> <c- mf>0</c-><c- p>;</c-> dir <c- o>&lt;</c-> numDirections<c- p>;</c-> <c- o>++</c->dir<c- p>)</c-> <c- p>{</c->
-    currentHidden<c- p>.</c->push<c- p>(</c->builder<c- p>.</c->squeeze<c- p>(</c->builder<c- p>.</c->slice<c- p>(</c->hiddenState<c- p>,</c-> <c- p>[</c->dir<c- p>,</c-> <c- mf>0</c-><c- p>,</c-> <c- mf>0</c-><c- p>],</c-> <c- p>[</c-><c- mf>1</c-><c- p>,</c-> batch_size<c- p>,</c-> hidden_size<c- p>]),</c-> <c- p>{</c-> axes<c- o>:</c-> <c- p>[</c-><c- mf>0</c-><c- p>]</c-> <c- p>}));</c->
+    currentHidden<c- p>.</c->push<c- p>(</c->builder<c- p>.</c->squeeze<c- p>(</c->builder<c- p>.</c->slice<c- p>(</c->hiddenState<c- p>,</c-> <c- p>[</c->dir<c- p>,</c-> <c- mf>0</c-><c- p>,</c-> <c- mf>0</c-><c- p>],</c-> <c- p>[</c-><c- mf>1</c-><c- p>,</c-> batchSize<c- p>,</c-> hiddenSize<c- p>]),</c-> <c- p>{</c-> axes<c- o>:</c-> <c- p>[</c-><c- mf>0</c-><c- p>]</c-> <c- p>}));</c->
   <c- p>}</c->
 
   <c- k>for</c-> <c- p>(</c-><c- a>let</c-> dir <c- o>=</c-> <c- mf>0</c-><c- p>;</c-> dir <c- o>&lt;</c-> numDirections<c- p>;</c-> <c- o>++</c->dir<c- p>)</c-> <c- p>{</c->
     <c- a>let</c-> slice <c- o>=</c-> <c- p>(</c->dir <c- o>==</c-> <c- mf>1</c-> <c- o>||</c-> options<c- p>.</c->direction <c- o>==</c-> <c- u>"backward"</c-> <c- o>?</c-> steps <c- o>-</c-> step <c- o>-</c-> <c- mf>1</c-> <c- o>:</c-> step<c- p>);</c->
-    <c- a>let</c-> currentInput <c- o>=</c-> builder<c- p>.</c->squeeze<c- p>(</c->builder<c- p>.</c->slice<c- p>(</c->input<c- p>,</c-> <c- p>[</c->slice<c- p>,</c-> <c- mf>0</c-><c- p>,</c-> <c- mf>0</c-><c- p>],</c-> <c- p>[</c-><c- mf>1</c-><c- p>,</c-> batch_size<c- p>,</c-> input_size<c- p>]),</c-> <c- p>{</c-> axes<c- o>:</c-> <c- p>[</c-><c- mf>0</c-><c- p>]</c-> <c- p>});</c->
+    <c- a>let</c-> currentInput <c- o>=</c-> builder<c- p>.</c->squeeze<c- p>(</c->builder<c- p>.</c->slice<c- p>(</c->input<c- p>,</c-> <c- p>[</c->slice<c- p>,</c-> <c- mf>0</c-><c- p>,</c-> <c- mf>0</c-><c- p>],</c-> <c- p>[</c-><c- mf>1</c-><c- p>,</c-> batchSize<c- p>,</c-> inputSize<c- p>]),</c-> <c- p>{</c-> axes<c- o>:</c-> <c- p>[</c-><c- mf>0</c-><c- p>]</c-> <c- p>});</c->
 
     <c- a>let</c-> result <c- o>=</c-> builder<c- p>.</c->reshape<c- p>(</c->
       builder<c- p>.</c->gruCell<c- p>(</c->
@@ -2811,55 +4366,126 @@ <h4 class="heading settled" data-level="7.6.13" id="api-mlgraphbuilder-gru"><spa
 
 <c- k>return</c-> <c- p>(</c->sequence <c- o>?</c-> <c- p>[</c->hiddenState<c- p>,</c-> sequence<c- p>]</c-> <c- o>:</c-> <c- p>[</c->hiddenState<c- p>]);</c->
 </pre>
-     </div>
-    </div>
-    <h4 class="heading settled" data-level="7.6.14" id="api-mlgraphbuilder-grucell"><span class="secno">7.6.14. </span><span class="content">The gruCell() method</span><a class="self-link" href="#api-mlgraphbuilder-grucell"></a></h4>
-     A single time step of the Gated Recurrent Unit <a data-link-type="biblio" href="#biblio-gru" title="Learning Phrase Representations using RNN Encoder–Decoder for Statistical Machine Translation">[GRU]</a> recurrent network using an update gate and a reset gate to compute the hidden state that rolls into the output across the temporal sequence of a recurrent network. 
+    </details>
+   </div>
+   <h4 class="heading settled" data-level="7.6.15" id="api-mlgraphbuilder-grucell"><span class="secno">7.6.15. </span><span class="content">The gruCell() method</span><a class="self-link" href="#api-mlgraphbuilder-grucell"></a></h4>
+    A single time step of the Gated Recurrent Unit <a data-link-type="biblio" href="#biblio-gru" title="Learning Phrase Representations using RNN Encoder–Decoder for Statistical Machine Translation">[GRU]</a> recurrent network using an update gate and a reset gate to compute the hidden state that rolls into the output across the temporal sequence of a recurrent network. 
 <pre class="idl highlight def"><c- b>dictionary</c-> <dfn class="dfn-paneled idl-code" data-dfn-type="dictionary" data-export id="dictdef-mlgrucelloptions"><code><c- g>MLGruCellOptions</c-></code></dfn> {
-  <a data-link-type="idl-name" href="#mloperand" id="ref-for-mloperand①③⑦"><c- n>MLOperand</c-></a> <dfn class="idl-code" data-dfn-for="MLGruCellOptions" data-dfn-type="dict-member" data-export data-type="MLOperand " id="dom-mlgrucelloptions-bias"><code><c- g>bias</c-></code><a class="self-link" href="#dom-mlgrucelloptions-bias"></a></dfn>;
-  <a data-link-type="idl-name" href="#mloperand" id="ref-for-mloperand①③⑧"><c- n>MLOperand</c-></a> <dfn class="idl-code" data-dfn-for="MLGruCellOptions" data-dfn-type="dict-member" data-export data-type="MLOperand " id="dom-mlgrucelloptions-recurrentbias"><code><c- g>recurrentBias</c-></code><a class="self-link" href="#dom-mlgrucelloptions-recurrentbias"></a></dfn>;
-  <a class="idl-code" data-link-type="interface" href="https://webidl.spec.whatwg.org/#idl-boolean" id="ref-for-idl-boolean⑧"><c- b>boolean</c-></a> <dfn class="idl-code" data-default="true" data-dfn-for="MLGruCellOptions" data-dfn-type="dict-member" data-export data-type="boolean " id="dom-mlgrucelloptions-resetafter"><code><c- g>resetAfter</c-></code><a class="self-link" href="#dom-mlgrucelloptions-resetafter"></a></dfn> = <c- b>true</c->;
-  <a data-link-type="idl-name" href="#enumdef-mlgruweightlayout" id="ref-for-enumdef-mlgruweightlayout②"><c- n>MLGruWeightLayout</c-></a> <dfn class="idl-code" data-default="&quot;zrn&quot;" data-dfn-for="MLGruCellOptions" data-dfn-type="dict-member" data-export data-type="MLGruWeightLayout " id="dom-mlgrucelloptions-layout"><code><c- g>layout</c-></code><a class="self-link" href="#dom-mlgrucelloptions-layout"></a></dfn> = "zrn";
-  <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#idl-sequence" id="ref-for-idl-sequence①②"><c- b>sequence</c-></a>&lt;<a data-link-type="idl-name" href="#mlactivation" id="ref-for-mlactivation①④"><c- n>MLActivation</c-></a>> <dfn class="idl-code" data-dfn-for="MLGruCellOptions" data-dfn-type="dict-member" data-export data-type="sequence<MLActivation> " id="dom-mlgrucelloptions-activations"><code><c- g>activations</c-></code><a class="self-link" href="#dom-mlgrucelloptions-activations"></a></dfn>;
+  <a data-link-type="idl-name" href="#mloperand" id="ref-for-mloperand①⑥③"><c- n>MLOperand</c-></a> <a class="idl-code" data-link-type="dict-member" data-type="MLOperand " href="#dom-mlgrucelloptions-bias" id="ref-for-dom-mlgrucelloptions-bias"><c- g>bias</c-></a>;
+  <a data-link-type="idl-name" href="#mloperand" id="ref-for-mloperand①⑥④"><c- n>MLOperand</c-></a> <a class="idl-code" data-link-type="dict-member" data-type="MLOperand " href="#dom-mlgrucelloptions-recurrentbias" id="ref-for-dom-mlgrucelloptions-recurrentbias"><c- g>recurrentBias</c-></a>;
+  <a class="idl-code" data-link-type="interface" href="https://webidl.spec.whatwg.org/#idl-boolean" id="ref-for-idl-boolean①②"><c- b>boolean</c-></a> <a class="idl-code" data-default="true" data-link-type="dict-member" data-type="boolean " href="#dom-mlgrucelloptions-resetafter" id="ref-for-dom-mlgrucelloptions-resetafter"><c- g>resetAfter</c-></a> = <c- b>true</c->;
+  <a data-link-type="idl-name" href="#enumdef-mlgruweightlayout" id="ref-for-enumdef-mlgruweightlayout③"><c- n>MLGruWeightLayout</c-></a> <a class="idl-code" data-default="&quot;zrn&quot;" data-link-type="dict-member" data-type="MLGruWeightLayout " href="#dom-mlgrucelloptions-layout" id="ref-for-dom-mlgrucelloptions-layout"><c- g>layout</c-></a> = "zrn";
+  <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#idl-sequence" id="ref-for-idl-sequence①②"><c- b>sequence</c-></a>&lt;<a data-link-type="idl-name" href="#mlactivation" id="ref-for-mlactivation②⑥"><c- n>MLActivation</c-></a>> <a class="idl-code" data-link-type="dict-member" data-type="sequence<MLActivation> " href="#dom-mlgrucelloptions-activations" id="ref-for-dom-mlgrucelloptions-activations"><c- g>activations</c-></a>;
 };
 
-<c- b>partial</c-> <c- b>interface</c-> <a class="idl-code" data-link-type="interface" href="#mlgraphbuilder" id="ref-for-mlgraphbuilder③⓪"><c- g>MLGraphBuilder</c-></a> {
-  <a data-link-type="idl-name" href="#mloperand" id="ref-for-mloperand①③⑨"><c- n>MLOperand</c-></a> <dfn class="idl-code" data-dfn-for="MLGraphBuilder" data-dfn-type="method" data-export data-lt="gruCell(input, weight, recurrentWeight, hiddenState, hiddenSize, options)|gruCell(input, weight, recurrentWeight, hiddenState, hiddenSize)" id="dom-mlgraphbuilder-grucell"><code><c- g>gruCell</c-></code><a class="self-link" href="#dom-mlgraphbuilder-grucell"></a></dfn>(<a data-link-type="idl-name" href="#mloperand" id="ref-for-mloperand①④⓪"><c- n>MLOperand</c-></a> <dfn class="idl-code" data-dfn-for="MLGraphBuilder/gruCell(input, weight, recurrentWeight, hiddenState, hiddenSize, options), MLGraphBuilder/gruCell(input, weight, recurrentWeight, hiddenState, hiddenSize)" data-dfn-type="argument" data-export id="dom-mlgraphbuilder-grucell-input-weight-recurrentweight-hiddenstate-hiddensize-options-input"><code><c- g>input</c-></code><a class="self-link" href="#dom-mlgraphbuilder-grucell-input-weight-recurrentweight-hiddenstate-hiddensize-options-input"></a></dfn>, <a data-link-type="idl-name" href="#mloperand" id="ref-for-mloperand①④①"><c- n>MLOperand</c-></a> <dfn class="idl-code" data-dfn-for="MLGraphBuilder/gruCell(input, weight, recurrentWeight, hiddenState, hiddenSize, options), MLGraphBuilder/gruCell(input, weight, recurrentWeight, hiddenState, hiddenSize)" data-dfn-type="argument" data-export id="dom-mlgraphbuilder-grucell-input-weight-recurrentweight-hiddenstate-hiddensize-options-weight"><code><c- g>weight</c-></code><a class="self-link" href="#dom-mlgraphbuilder-grucell-input-weight-recurrentweight-hiddenstate-hiddensize-options-weight"></a></dfn>, <a data-link-type="idl-name" href="#mloperand" id="ref-for-mloperand①④②"><c- n>MLOperand</c-></a> <dfn class="idl-code" data-dfn-for="MLGraphBuilder/gruCell(input, weight, recurrentWeight, hiddenState, hiddenSize, options), MLGraphBuilder/gruCell(input, weight, recurrentWeight, hiddenState, hiddenSize)" data-dfn-type="argument" data-export id="dom-mlgraphbuilder-grucell-input-weight-recurrentweight-hiddenstate-hiddensize-options-recurrentweight"><code><c- g>recurrentWeight</c-></code><a class="self-link" href="#dom-mlgraphbuilder-grucell-input-weight-recurrentweight-hiddenstate-hiddensize-options-recurrentweight"></a></dfn>,
-                    <a data-link-type="idl-name" href="#mloperand" id="ref-for-mloperand①④③"><c- n>MLOperand</c-></a> <dfn class="idl-code" data-dfn-for="MLGraphBuilder/gruCell(input, weight, recurrentWeight, hiddenState, hiddenSize, options), MLGraphBuilder/gruCell(input, weight, recurrentWeight, hiddenState, hiddenSize)" data-dfn-type="argument" data-export id="dom-mlgraphbuilder-grucell-input-weight-recurrentweight-hiddenstate-hiddensize-options-hiddenstate"><code><c- g>hiddenState</c-></code><a class="self-link" href="#dom-mlgraphbuilder-grucell-input-weight-recurrentweight-hiddenstate-hiddensize-options-hiddenstate"></a></dfn>, <a class="idl-code" data-link-type="interface" href="https://webidl.spec.whatwg.org/#idl-unsigned-long" id="ref-for-idl-unsigned-long②⑨"><c- b>unsigned</c-> <c- b>long</c-></a> <dfn class="idl-code" data-dfn-for="MLGraphBuilder/gruCell(input, weight, recurrentWeight, hiddenState, hiddenSize, options), MLGraphBuilder/gruCell(input, weight, recurrentWeight, hiddenState, hiddenSize)" data-dfn-type="argument" data-export id="dom-mlgraphbuilder-grucell-input-weight-recurrentweight-hiddenstate-hiddensize-options-hiddensize"><code><c- g>hiddenSize</c-></code><a class="self-link" href="#dom-mlgraphbuilder-grucell-input-weight-recurrentweight-hiddenstate-hiddensize-options-hiddensize"></a></dfn>,
+<c- b>partial</c-> <c- b>interface</c-> <a class="idl-code" data-link-type="interface" href="#mlgraphbuilder" id="ref-for-mlgraphbuilder③③"><c- g>MLGraphBuilder</c-></a> {
+  <a data-link-type="idl-name" href="#mloperand" id="ref-for-mloperand①⑥⑤"><c- n>MLOperand</c-></a> <a class="idl-code" data-link-type="method" href="#dom-mlgraphbuilder-grucell" id="ref-for-dom-mlgraphbuilder-grucell"><c- g>gruCell</c-></a>(<a data-link-type="idl-name" href="#mloperand" id="ref-for-mloperand①⑥⑥"><c- n>MLOperand</c-></a> <dfn class="idl-code" data-dfn-for="MLGraphBuilder/gruCell(input, weight, recurrentWeight, hiddenState, hiddenSize, options), MLGraphBuilder/gruCell(input, weight, recurrentWeight, hiddenState, hiddenSize)" data-dfn-type="argument" data-export id="dom-mlgraphbuilder-grucell-input-weight-recurrentweight-hiddenstate-hiddensize-options-input"><code><c- g>input</c-></code><a class="self-link" href="#dom-mlgraphbuilder-grucell-input-weight-recurrentweight-hiddenstate-hiddensize-options-input"></a></dfn>, <a data-link-type="idl-name" href="#mloperand" id="ref-for-mloperand①⑥⑦"><c- n>MLOperand</c-></a> <dfn class="idl-code" data-dfn-for="MLGraphBuilder/gruCell(input, weight, recurrentWeight, hiddenState, hiddenSize, options), MLGraphBuilder/gruCell(input, weight, recurrentWeight, hiddenState, hiddenSize)" data-dfn-type="argument" data-export id="dom-mlgraphbuilder-grucell-input-weight-recurrentweight-hiddenstate-hiddensize-options-weight"><code><c- g>weight</c-></code><a class="self-link" href="#dom-mlgraphbuilder-grucell-input-weight-recurrentweight-hiddenstate-hiddensize-options-weight"></a></dfn>, <a data-link-type="idl-name" href="#mloperand" id="ref-for-mloperand①⑥⑧"><c- n>MLOperand</c-></a> <dfn class="idl-code" data-dfn-for="MLGraphBuilder/gruCell(input, weight, recurrentWeight, hiddenState, hiddenSize, options), MLGraphBuilder/gruCell(input, weight, recurrentWeight, hiddenState, hiddenSize)" data-dfn-type="argument" data-export id="dom-mlgraphbuilder-grucell-input-weight-recurrentweight-hiddenstate-hiddensize-options-recurrentweight"><code><c- g>recurrentWeight</c-></code><a class="self-link" href="#dom-mlgraphbuilder-grucell-input-weight-recurrentweight-hiddenstate-hiddensize-options-recurrentweight"></a></dfn>,
+                    <a data-link-type="idl-name" href="#mloperand" id="ref-for-mloperand①⑥⑨"><c- n>MLOperand</c-></a> <dfn class="idl-code" data-dfn-for="MLGraphBuilder/gruCell(input, weight, recurrentWeight, hiddenState, hiddenSize, options), MLGraphBuilder/gruCell(input, weight, recurrentWeight, hiddenState, hiddenSize)" data-dfn-type="argument" data-export id="dom-mlgraphbuilder-grucell-input-weight-recurrentweight-hiddenstate-hiddensize-options-hiddenstate"><code><c- g>hiddenState</c-></code><a class="self-link" href="#dom-mlgraphbuilder-grucell-input-weight-recurrentweight-hiddenstate-hiddensize-options-hiddenstate"></a></dfn>, <a class="idl-code" data-link-type="interface" href="https://webidl.spec.whatwg.org/#idl-unsigned-long" id="ref-for-idl-unsigned-long③①"><c- b>unsigned</c-> <c- b>long</c-></a> <dfn class="idl-code" data-dfn-for="MLGraphBuilder/gruCell(input, weight, recurrentWeight, hiddenState, hiddenSize, options), MLGraphBuilder/gruCell(input, weight, recurrentWeight, hiddenState, hiddenSize)" data-dfn-type="argument" data-export id="dom-mlgraphbuilder-grucell-input-weight-recurrentweight-hiddenstate-hiddensize-options-hiddensize"><code><c- g>hiddenSize</c-></code><a class="self-link" href="#dom-mlgraphbuilder-grucell-input-weight-recurrentweight-hiddenstate-hiddensize-options-hiddensize"></a></dfn>,
                     <c- b>optional</c-> <a data-link-type="idl-name" href="#dictdef-mlgrucelloptions" id="ref-for-dictdef-mlgrucelloptions"><c- n>MLGruCellOptions</c-></a> <dfn class="idl-code" data-dfn-for="MLGraphBuilder/gruCell(input, weight, recurrentWeight, hiddenState, hiddenSize, options), MLGraphBuilder/gruCell(input, weight, recurrentWeight, hiddenState, hiddenSize)" data-dfn-type="argument" data-export id="dom-mlgraphbuilder-grucell-input-weight-recurrentweight-hiddenstate-hiddensize-options-options"><code><c- g>options</c-></code><a class="self-link" href="#dom-mlgraphbuilder-grucell-input-weight-recurrentweight-hiddenstate-hiddensize-options-options"></a></dfn> = {});
 };
 </pre>
-    <div class="algorithm" data-algorithm="grucell">
-      <strong>Arguments:</strong> 
-     <ul>
+   <p><code class="idl"><a data-link-type="idl" href="#dictdef-mlgrucelloptions" id="ref-for-dictdef-mlgrucelloptions①">MLGruCellOptions</a></code> has the following members:</p>
+   <dl>
+    <dt data-md><dfn class="dfn-paneled idl-code" data-dfn-for="MLGruCellOptions" data-dfn-type="dict-member" data-export id="dom-mlgrucelloptions-bias"><code>bias</code></dfn>, <span> of type <a data-link-type="idl-name" href="#mloperand" id="ref-for-mloperand①⑦⓪">MLOperand</a></span>
+    <dd data-md>
+     <p>An <code class="idl"><a data-link-type="idl" href="#mloperand" id="ref-for-mloperand①⑦①">MLOperand</a></code>. Specifies the 1-D input bias tensor of shape [3 * hiddenSize]. The ordering of the bias vectors in the second dimension of the tensor shape is specified according to the <code class="idl"><a data-link-type="idl" href="#dom-mlgruoptions-layout" id="ref-for-dom-mlgruoptions-layout⑤">layout</a></code> argument.</p>
+    <dt data-md><dfn class="dfn-paneled idl-code" data-dfn-for="MLGruCellOptions" data-dfn-type="dict-member" data-export id="dom-mlgrucelloptions-recurrentbias"><code>recurrentBias</code></dfn>, <span> of type <a data-link-type="idl-name" href="#mloperand" id="ref-for-mloperand①⑦②">MLOperand</a></span>
+    <dd data-md>
+     <p>An <code class="idl"><a data-link-type="idl" href="#mloperand" id="ref-for-mloperand①⑦③">MLOperand</a></code>. Specifies the 1-D recurrent bias tensor of shape [3 * hiddenSize]. The ordering of the bias vectors in the second dimension of the tensor shape is specified according to the <code class="idl"><a data-link-type="idl" href="#dom-mlgruoptions-layout" id="ref-for-dom-mlgruoptions-layout⑥">layout</a></code> argument.</p>
+    <dt data-md><dfn class="dfn-paneled idl-code" data-dfn-for="MLGruCellOptions" data-dfn-type="dict-member" data-export id="dom-mlgrucelloptions-resetafter"><code>resetAfter</code></dfn>, <span> of type <a data-link-type="idl-name" href="https://webidl.spec.whatwg.org/#idl-boolean" id="ref-for-idl-boolean①③">boolean</a>, defaulting to <code>true</code></span>
+    <dd data-md>
+     <p>A <code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#idl-boolean" id="ref-for-idl-boolean①④">boolean</a></code> indicating whether to apply the reset gate after or before matrix multiplication. The default value is true.</p>
+    <dt data-md><dfn class="dfn-paneled idl-code" data-dfn-for="MLGruCellOptions" data-dfn-type="dict-member" data-export id="dom-mlgrucelloptions-layout"><code>layout</code></dfn>, <span> of type <a data-link-type="idl-name" href="#enumdef-mlgruweightlayout" id="ref-for-enumdef-mlgruweightlayout④">MLGruWeightLayout</a>, defaulting to <code>"zrn"</code></span>
+    <dd data-md>
+     <p>An <code class="idl"><a data-link-type="idl" href="#enumdef-mlgruweightlayout" id="ref-for-enumdef-mlgruweightlayout⑤">MLGruWeightLayout</a></code>. The ordering of the weight and bias vectors for the internal gates of GRU, specifically the <code>update (z)</code>, <code>reset (r)</code>, and <code>new (n)</code> gate, as indicated in the second dimension of the weight and bias tensor shape. When not specified, the default layout is <code>"zrn"</code>.</p>
+    <dt data-md><dfn class="dfn-paneled idl-code" data-dfn-for="MLGruCellOptions" data-dfn-type="dict-member" data-export id="dom-mlgrucelloptions-activations"><code>activations</code></dfn>, <span> of type sequence&lt;<a data-link-type="idl-name" href="#mlactivation" id="ref-for-mlactivation②⑦">MLActivation</a>></span>
+    <dd data-md>
+     <p>A sequence of <code class="idl"><a data-link-type="idl" href="#mlactivation" id="ref-for-mlactivation②⑧">MLActivation</a></code>. Specifies a pair of activation functions with the first function used for the update and reset gate, and the second used for the new gate. When not specified, implementations SHOULD use the the pair of sigmoid (<code>"sigmoid"</code>) and the hyperbolic tangent (<code>"tanh"</code>) functions, respectively.</p>
+   </dl>
+   <div>
+     <strong>Arguments:</strong> 
+    <ul>
+     <li data-md>
+      <p><em>input</em>: an <code class="idl"><a data-link-type="idl" href="#mloperand" id="ref-for-mloperand①⑦④">MLOperand</a></code>. The input 2-D tensor of shape [batchSize, inputSize].</p>
+     <li data-md>
+      <p><em>weight</em>: an <code class="idl"><a data-link-type="idl" href="#mloperand" id="ref-for-mloperand①⑦⑤">MLOperand</a></code>. The 2-D input weight tensor of shape [3 * hiddenSize, inputSize]. The ordering of the weight vectors in the first dimension of the tensor shape is specified according to the <em>options.layout</em> argument.</p>
+     <li data-md>
+      <p><em>recurrentWeight</em>: an <code class="idl"><a data-link-type="idl" href="#mloperand" id="ref-for-mloperand①⑦⑥">MLOperand</a></code>. The 2-D recurrent weight tensor of shape [3 * hiddenSize, hiddenSize]. The ordering of the weight vectors in the first dimension of the tensor shape is specified according to the <em>options.layout</em> argument.</p>
+     <li data-md>
+      <p><em>hiddenState</em>: an <code class="idl"><a data-link-type="idl" href="#mloperand" id="ref-for-mloperand①⑦⑦">MLOperand</a></code>. The 2-D input hidden state tensor of shape [batchSize, hiddenSize].</p>
+     <li data-md>
+      <p><em>hiddenSize</em>: an <code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#idl-unsigned-long" id="ref-for-idl-unsigned-long③②">unsigned long</a></code> scalar. The value of the second dimension of the output tensor shape. It indicates the number of features in the hidden state.</p>
+     <li data-md>
+      <p><em>options</em>: an optional <code class="idl"><a data-link-type="idl" href="#dictdef-mlgrucelloptions" id="ref-for-dictdef-mlgrucelloptions②">MLGruCellOptions</a></code>. The optional parameters of the operation.</p>
+    </ul>
+    <p><strong>Returns:</strong> an <code class="idl"><a data-link-type="idl" href="#mloperand" id="ref-for-mloperand①⑦⑧">MLOperand</a></code>. The 2-D tensor of shape [batchSize, hiddenSize], the cell output hidden state of a single time step of the recurrent network.</p>
+   </div>
+   <details class="algorithm" data-algorithm="gruCell(input, weight, recurrentWeight, hiddenState, hiddenSize, options)" data-algorithm-for="MLGraphBuilder" open>
+    <summary> The <dfn class="dfn-paneled idl-code" data-dfn-for="MLGraphBuilder" data-dfn-type="method" data-export data-lt="gruCell(input, weight, recurrentWeight, hiddenState, hiddenSize, options)|gruCell(input, weight, recurrentWeight, hiddenState, hiddenSize)" id="dom-mlgraphbuilder-grucell"><code>gruCell(<var>input</var>, <var>weight</var>, <var>recurrentWeight</var>, <var>hiddenState</var>, <var>hiddenSize</var>, <var>options</var>)</code></dfn> method steps are: </summary>
+    <div class="algorithm-steps">
+     <ol>
       <li data-md>
-       <p><em>input</em>: an <code class="idl"><a data-link-type="idl" href="#mloperand" id="ref-for-mloperand①④④">MLOperand</a></code>. The input 2-D tensor of shape [batch_size, input_size].</p>
+       <p><a data-link-type="dfn" href="https://infra.spec.whatwg.org/#assert" id="ref-for-assert③⑦">Assert</a>: the type of <var>input</var>, <var>weight</var> and <var>recurrentWeight</var> is <code class="idl"><a data-link-type="idl" href="#mloperand" id="ref-for-mloperand①⑦⑨">MLOperand</a></code>.</p>
       <li data-md>
-       <p><em>weight</em>: an <code class="idl"><a data-link-type="idl" href="#mloperand" id="ref-for-mloperand①④⑤">MLOperand</a></code>. The 2-D input weight tensor of shape [3 * hidden_size, input_size]. The ordering of the weight vectors in the first dimension of the tensor shape is specified according to the <em>options.layout</em> argument.</p>
+       <p>If the <a data-link-type="dfn" href="#mloperand-rank" id="ref-for-mloperand-rank⑨">rank</a> of <var>input</var> or <var>weight</var> or <var>recurrentWeight</var> or <var>hiddenState</var> is not 2, then <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#dfn-throw" id="ref-for-dfn-throw①③⑨">throw</a> a "<code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#dataerror" id="ref-for-dataerror③⑨">DataError</a></code>" <code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#idl-DOMException" id="ref-for-idl-DOMException⑦②">DOMException</a></code>.</p>
       <li data-md>
-       <p><em>recurrentWeight</em>: an <code class="idl"><a data-link-type="idl" href="#mloperand" id="ref-for-mloperand①④⑥">MLOperand</a></code>. The 2-D recurrent weight tensor of shape [3 * hidden_size, hidden_size]. The ordering of the weight vectors in the first dimension of the tensor shape is specified according to the <em>options.layout</em> argument.</p>
+       <p>If <var>weight</var>.<code class="idl"><a data-link-type="idl" href="#dom-mloperand-descriptor-slot" id="ref-for-dom-mloperand-descriptor-slot⑤⓪">[[descriptor]]</a></code>.<code class="idl"><a data-link-type="idl" href="#dom-mloperanddescriptor-dimensions" id="ref-for-dom-mloperanddescriptor-dimensions④⑦">dimensions</a></code>[0] is not equal to 3 * <var>hiddenSize</var>, then <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#dfn-throw" id="ref-for-dfn-throw①④⓪">throw</a> a "<code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#dataerror" id="ref-for-dataerror④⓪">DataError</a></code>" <code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#idl-DOMException" id="ref-for-idl-DOMException⑦③">DOMException</a></code>.</p>
       <li data-md>
-       <p><em>hiddenState</em>: an <code class="idl"><a data-link-type="idl" href="#mloperand" id="ref-for-mloperand①④⑦">MLOperand</a></code>. The 2-D input hidden state tensor of shape [batch_size, hidden_size].</p>
+       <p>If <var>recurrentWeight</var>.<code class="idl"><a data-link-type="idl" href="#dom-mloperand-descriptor-slot" id="ref-for-dom-mloperand-descriptor-slot⑤①">[[descriptor]]</a></code>.<code class="idl"><a data-link-type="idl" href="#dom-mloperanddescriptor-dimensions" id="ref-for-dom-mloperanddescriptor-dimensions④⑧">dimensions</a></code>[0] is not equal to 3 * <var>hiddenSize</var>, then <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#dfn-throw" id="ref-for-dfn-throw①④①">throw</a> a "<code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#dataerror" id="ref-for-dataerror④①">DataError</a></code>" <code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#idl-DOMException" id="ref-for-idl-DOMException⑦④">DOMException</a></code>.</p>
       <li data-md>
-       <p><em>hiddenSize</em>: an <code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#idl-unsigned-long" id="ref-for-idl-unsigned-long③⓪">unsigned long</a></code> scalar. The value of the second dimension of the output tensor shape. It indicates the number of features in the hidden state.</p>
+       <p>If <var>options</var>.<code class="idl"><a data-link-type="idl" href="#dom-mlgruoptions-bias" id="ref-for-dom-mlgruoptions-bias③">bias</a></code> <a data-link-type="dfn" href="https://infra.spec.whatwg.org/#map-exists" id="ref-for-map-exists③②">exists</a>:</p>
+       <ol>
+        <li data-md>
+         <p><a data-link-type="dfn" href="https://infra.spec.whatwg.org/#assert" id="ref-for-assert③⑧">Assert</a>: its type is <code class="idl"><a data-link-type="idl" href="#mloperand" id="ref-for-mloperand①⑧⓪">MLOperand</a></code>.</p>
+        <li data-md>
+         <p>If its rank is not equal to 3 * <var>hiddenSize</var>, then <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#dfn-throw" id="ref-for-dfn-throw①④②">throw</a> a "<code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#dataerror" id="ref-for-dataerror④②">DataError</a></code>" <code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#idl-DOMException" id="ref-for-idl-DOMException⑦⑤">DOMException</a></code>.</p>
+       </ol>
       <li data-md>
-       <p><em>options</em>: an optional <code class="idl"><a data-link-type="idl" href="#dictdef-mlgrucelloptions" id="ref-for-dictdef-mlgrucelloptions①">MLGruCellOptions</a></code>. The optional parameters of the operation.</p>
-       <ul>
+       <p>If <var>options</var>.<code class="idl"><a data-link-type="idl" href="#dom-mlgruoptions-recurrentbias" id="ref-for-dom-mlgruoptions-recurrentbias③">recurrentBias</a></code> <a data-link-type="dfn" href="https://infra.spec.whatwg.org/#map-exists" id="ref-for-map-exists③③">exists</a>:</p>
+       <ol>
+        <li data-md>
+         <p><a data-link-type="dfn" href="https://infra.spec.whatwg.org/#assert" id="ref-for-assert③⑨">Assert</a>: its type is <code class="idl"><a data-link-type="idl" href="#mloperand" id="ref-for-mloperand①⑧①">MLOperand</a></code>.</p>
         <li data-md>
-         <p><em>bias</em>: an <code class="idl"><a data-link-type="idl" href="#mloperand" id="ref-for-mloperand①④⑧">MLOperand</a></code>. The 1-D input bias tensor of shape [3 * hidden_size]. The ordering of the bias vectors in the first dimension of the tensor shape is specified according to the <em>options.layout</em> argument.</p>
+         <p>If its rank is not equal to 3 * <var>hiddenSize</var>, then <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#dfn-throw" id="ref-for-dfn-throw①④③">throw</a> a "<code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#dataerror" id="ref-for-dataerror④③">DataError</a></code>" <code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#idl-DOMException" id="ref-for-idl-DOMException⑦⑥">DOMException</a></code>.</p>
+       </ol>
+      <li data-md>
+       <p>If <var>options</var>.<code class="idl"><a data-link-type="idl" href="#dom-mlgruoptions-activations" id="ref-for-dom-mlgruoptions-activations②">activations</a></code> <a data-link-type="dfn" href="https://infra.spec.whatwg.org/#map-exists" id="ref-for-map-exists③④">exists</a> and its <a data-link-type="dfn" href="https://infra.spec.whatwg.org/#list-size" id="ref-for-list-size②⑦">size</a> is not 2, then <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#dfn-throw" id="ref-for-dfn-throw①④④">throw</a> a <code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#exceptiondef-typeerror" id="ref-for-exceptiondef-typeerror③⑤">TypeError</a></code>.</p>
+      <li data-md>
+       <p>Let <var>desc</var> a new <code class="idl"><a data-link-type="idl" href="#dictdef-mloperanddescriptor" id="ref-for-dictdef-mloperanddescriptor①⑧">MLOperandDescriptor</a></code>.</p>
+      <li data-md>
+       <p>Set <var>desc</var>.<code class="idl"><a data-link-type="idl" href="#dom-mloperanddescriptor-dimensions" id="ref-for-dom-mloperanddescriptor-dimensions④⑨">dimensions</a></code> to [ <var>input</var>.<code class="idl"><a data-link-type="idl" href="#dom-mloperanddescriptor-dimensions" id="ref-for-dom-mloperanddescriptor-dimensions⑤⓪">dimensions</a></code>[0], <var>hiddenSize</var> ].</p>
+      <li data-md>
+       <p>Set <var>desc</var>.<code class="idl"><a data-link-type="idl" href="#dom-mloperanddescriptor-type" id="ref-for-dom-mloperanddescriptor-type②⑦">type</a></code> to <var>input</var>.<code class="idl"><a data-link-type="idl" href="#dom-mloperand-descriptor-slot" id="ref-for-dom-mloperand-descriptor-slot⑤②">[[descriptor]]</a></code>.<code class="idl"><a data-link-type="idl" href="#dom-mloperanddescriptor-type" id="ref-for-dom-mloperanddescriptor-type②⑧">type</a></code>.</p>
+      <li data-md>
+       <p>If any of the following sub-steps fail, <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#dfn-throw" id="ref-for-dfn-throw①④⑤">throw</a> an "<code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#operationerror" id="ref-for-operationerror②④">OperationError</a></code>" <code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#idl-DOMException" id="ref-for-idl-DOMException⑦⑦">DOMException</a></code>.</p>
+       <ol>
         <li data-md>
-         <p><em>recurrentBias</em>: an <code class="idl"><a data-link-type="idl" href="#mloperand" id="ref-for-mloperand①④⑨">MLOperand</a></code>. The 1-D recurrent bias tensor of shape [3 * hidden_size]. The ordering of the bias vectors in the first dimension of the tensor shape is specified according to the <em>options.layout</em> argument.</p>
+         <p>Let <var>output</var> be the result of <a data-link-type="dfn" href="#create-an-mloperand" id="ref-for-create-an-mloperand⑨">creating an MLOperand</a> given <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#this" id="ref-for-this①⑨">this</a> and <var>desc</var>.</p>
         <li data-md>
-         <p><em>resetAfter</em>: a <code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#idl-boolean" id="ref-for-idl-boolean⑨">boolean</a></code> indicating whether to apply the reset gate after or before matrix multiplication. Default to true.</p>
+         <p>Make a request to the underlying platform to:</p>
+         <ol>
+          <li data-md>
+           <p>Let <var>opImpl</var> be an <a data-link-type="dfn" href="https://infra.spec.whatwg.org/#implementation-defined" id="ref-for-implementation-defined②⑧">implementation-defined</a> platform operator for <code>"gruCell"</code>, given <var>weight</var>, <var>recurrentWeight</var>, <var>hiddenState</var>, <var>hiddenSize</var> and <var>options</var> as parameters.</p>
+          <li data-md>
+           <p>Store a reference of <var>opImpl</var> in <var>output</var>.<code class="idl"><a data-link-type="idl" href="#dom-mloperand-operator-slot" id="ref-for-dom-mloperand-operator-slot⑨">[[operator]]</a></code>.</p>
+          <li data-md>
+           <p>Create an <a data-link-type="dfn" href="https://infra.spec.whatwg.org/#implementation-defined" id="ref-for-implementation-defined②⑨">implementation-defined</a> platform operand <var>outputImpl</var> to represent the output, given <var>output</var> and <var>opImpl</var>.</p>
+          <li data-md>
+           <p>Store a reference to <var>outputImpl</var> in <var>output</var>.<code class="idl"><a data-link-type="idl" href="#dom-mloperand-operand-slot" id="ref-for-dom-mloperand-operand-slot③⓪">[[operand]]</a></code>.</p>
+         </ol>
         <li data-md>
-         <p><em>layout</em>: an <code class="idl"><a data-link-type="idl" href="#enumdef-mlgruweightlayout" id="ref-for-enumdef-mlgruweightlayout③">MLGruWeightLayout</a></code>. The ordering of the weight and bias vectors for the internal gates of GRU, specifically the <em>update (z)</em>, <em>reset (r)</em>, and <em>new (n)</em> gate, as indicated in the first dimension of the weight and bias tensor shapes. When not specified, the default layout is <em>"zrn"</em>.</p>
+         <p>Connect <var>input</var>.<code class="idl"><a data-link-type="idl" href="#dom-mloperand-operand-slot" id="ref-for-dom-mloperand-operand-slot③①">[[operand]]</a></code> as input to <var>opImpl</var>.</p>
         <li data-md>
-         <p><em>activations</em>: a sequence of <code class="idl"><a data-link-type="idl" href="#mlactivation" id="ref-for-mlactivation①⑤">MLActivation</a></code>. A pair of activation functions with the first function used for the <em>update (z)</em> and <em>reset (r)</em> gate, and the second used for the <em>new (n)</em> gate. When not specified, it’s default to the sigmoid (<em>"sigmoid"</em>) and the hyperbolic tangent (<em>"tanh"</em>) function respectively.</p>
-       </ul>
-     </ul>
-     <p><strong>Returns:</strong> an <code class="idl"><a data-link-type="idl" href="#mloperand" id="ref-for-mloperand①⑤⓪">MLOperand</a></code>. The 2-D tensor of shape [batch_size, hidden_size], the cell output hidden state of a single time step of the recurrent network.</p>
-     <div class="note" role="note">
-       The behavior of this operation can be generically emulated via other operations as shown below, when the weight layout is the default <em>"zrn"</em> layout, and the activation functions of the update/reset gate and new gate are of the operator types <em>sigmoid</em> and <em>tanh</em> respectively. 
+         <p>Connect <var>output</var>.<code class="idl"><a data-link-type="idl" href="#dom-mloperand-operand-slot" id="ref-for-dom-mloperand-operand-slot③②">[[operand]]</a></code> as output to <var>opImpl</var>.</p>
+       </ol>
+      <li data-md>
+       <p>Return <var>output</var>.</p>
+     </ol>
+    </div>
+   </details>
+   <div class="note" role="note">
+    <details open>
+     <summary> The behavior of this operation can be generically emulated via other operations as shown below, when the weight layout is the default <em>"zrn"</em> layout, and the activation functions of the update/reset gate and new gate are of the operator types <em>sigmoid</em> and <em>tanh</em> respectively. </summary>
 <pre class="highlight"><c- a>const</c-> one <c- o>=</c-> builder<c- p>.</c->constant<c- p>(</c-><c- mf>1</c-><c- p>);</c->
 <c- a>const</c-> zero <c- o>=</c-> builder<c- p>.</c->constant<c- p>(</c-><c- mf>0</c-><c- p>);</c->
 
@@ -2873,11 +4499,11 @@ <h4 class="heading settled" data-level="7.6.14" id="api-mlgraphbuilder-grucell">
     builder<c- p>.</c->add<c- p>(</c->
       builder<c- p>.</c->matmul<c- p>(</c->
         input<c- p>,</c->
-        builder<c- p>.</c->transpose<c- p>(</c->builder<c- p>.</c->slice<c- p>(</c->weight<c- p>,</c-> <c- p>[</c-><c- mf>0</c-><c- p>,</c-> <c- mf>0</c-><c- p>],</c-> <c- p>[</c->hiddenSize<c- p>,</c-> input_size<c- p>]))</c->
+        builder<c- p>.</c->transpose<c- p>(</c->builder<c- p>.</c->slice<c- p>(</c->weight<c- p>,</c-> <c- p>[</c-><c- mf>0</c-><c- p>,</c-> <c- mf>0</c-><c- p>],</c-> <c- p>[</c->hiddenSize<c- p>,</c-> inputSize<c- p>]))</c->
         <c- p>),</c->
       builder<c- p>.</c->matmul<c- p>(</c->
         hiddenState<c- p>,</c->
-        builder<c- p>.</c->transpose<c- p>(</c->builder<c- p>.</c->slice<c- p>(</c->recurrentWeight<c- p>,</c-> <c- p>[</c-><c- mf>0</c-><c- p>,</c-> <c- mf>0</c-><c- p>],</c-> <c- p>[</c->hiddenSize<c- p>,</c-> hidden_size<c- p>]))</c->
+        builder<c- p>.</c->transpose<c- p>(</c->builder<c- p>.</c->slice<c- p>(</c->recurrentWeight<c- p>,</c-> <c- p>[</c-><c- mf>0</c-><c- p>,</c-> <c- mf>0</c-><c- p>],</c-> <c- p>[</c->hiddenSize<c- p>,</c-> hiddenSize<c- p>]))</c->
         <c- p>)</c->
       <c- p>)</c->
     <c- p>)</c->
@@ -2893,11 +4519,11 @@ <h4 class="heading settled" data-level="7.6.14" id="api-mlgraphbuilder-grucell">
     builder<c- p>.</c->add<c- p>(</c->
       builder<c- p>.</c->matmul<c- p>(</c->
         input<c- p>,</c->
-        builder<c- p>.</c->transpose<c- p>(</c->builder<c- p>.</c->slice<c- p>(</c->weight<c- p>,</c-> <c- p>[</c->hiddenSize<c- p>,</c-> <c- mf>0</c-><c- p>],</c-> <c- p>[</c->hiddenSize<c- p>,</c-> input_size<c- p>]))</c->
+        builder<c- p>.</c->transpose<c- p>(</c->builder<c- p>.</c->slice<c- p>(</c->weight<c- p>,</c-> <c- p>[</c->hiddenSize<c- p>,</c-> <c- mf>0</c-><c- p>],</c-> <c- p>[</c->hiddenSize<c- p>,</c-> inputSize<c- p>]))</c->
         <c- p>),</c->
       builder<c- p>.</c->matmul<c- p>(</c->
         hiddenState<c- p>,</c->
-        builder<c- p>.</c->transpose<c- p>(</c->builder<c- p>.</c->slice<c- p>(</c->recurrentWeight<c- p>,</c-> <c- p>[</c->hiddenSize<c- p>,</c-> <c- mf>0</c-><c- p>],</c-> <c- p>[</c->hiddenSize<c- p>,</c-> hidden_size<c- p>]))</c->
+        builder<c- p>.</c->transpose<c- p>(</c->builder<c- p>.</c->slice<c- p>(</c->recurrentWeight<c- p>,</c-> <c- p>[</c->hiddenSize<c- p>,</c-> <c- mf>0</c-><c- p>],</c-> <c- p>[</c->hiddenSize<c- p>,</c-> hiddenSize<c- p>]))</c->
         <c- p>)</c->
       <c- p>)</c->
     <c- p>)</c->
@@ -2912,7 +4538,7 @@ <h4 class="heading settled" data-level="7.6.14" id="api-mlgraphbuilder-grucell">
       builder<c- p>.</c->add<c- p>(</c->
         builder<c- p>.</c->matmul<c- p>(</c->
           input<c- p>,</c->
-          builder<c- p>.</c->transpose<c- p>(</c->builder<c- p>.</c->slice<c- p>(</c->weight<c- p>,</c-> <c- p>[</c-><c- mf>2</c-> <c- o>*</c-> hiddenSize<c- p>,</c-> <c- mf>0</c-><c- p>],</c-> <c- p>[</c->hiddenSize<c- p>,</c-> input_size<c- p>]))</c->
+          builder<c- p>.</c->transpose<c- p>(</c->builder<c- p>.</c->slice<c- p>(</c->weight<c- p>,</c-> <c- p>[</c-><c- mf>2</c-> <c- o>*</c-> hiddenSize<c- p>,</c-> <c- mf>0</c-><c- p>],</c-> <c- p>[</c->hiddenSize<c- p>,</c-> inputSize<c- p>]))</c->
           <c- p>),</c->
         builder<c- p>.</c->mul<c- p>(</c->
           r<c- p>,</c->
@@ -2920,7 +4546,7 @@ <h4 class="heading settled" data-level="7.6.14" id="api-mlgraphbuilder-grucell">
             <c- p>(</c->options<c- p>.</c->recurrentBias <c- o>?</c-> builder<c- p>.</c->slice<c- p>(</c->options<c- p>.</c->recurrentBias<c- p>,</c-> <c- p>[</c-><c- mf>2</c-> <c- o>*</c-> hiddenSize<c- p>],</c-> <c- p>[</c->hiddenSize<c- p>])</c-> <c- o>:</c-> zero<c- p>),</c->
             builder<c- p>.</c->matmul<c- p>(</c->
               hiddenState<c- p>,</c->
-              builder<c- p>.</c->transpose<c- p>(</c->builder<c- p>.</c->slice<c- p>(</c->recurrentWeight<c- p>,</c-> <c- p>[</c-><c- mf>2</c-> <c- o>*</c-> hiddenSize<c- p>,</c-> <c- mf>0</c-><c- p>],</c-> <c- p>[</c->hiddenSize<c- p>,</c-> hidden_size<c- p>]))</c->
+              builder<c- p>.</c->transpose<c- p>(</c->builder<c- p>.</c->slice<c- p>(</c->recurrentWeight<c- p>,</c-> <c- p>[</c-><c- mf>2</c-> <c- o>*</c-> hiddenSize<c- p>,</c-> <c- mf>0</c-><c- p>],</c-> <c- p>[</c->hiddenSize<c- p>,</c-> hiddenSize<c- p>]))</c->
               <c- p>)</c->
             <c- p>)</c->
           <c- p>)</c->
@@ -2938,11 +4564,11 @@ <h4 class="heading settled" data-level="7.6.14" id="api-mlgraphbuilder-grucell">
       builder<c- p>.</c->add<c- p>(</c->
         builder<c- p>.</c->matmul<c- p>(</c->
           input<c- p>,</c->
-          builder<c- p>.</c->transpose<c- p>(</c->builder<c- p>.</c->slice<c- p>(</c->weight<c- p>,</c-> <c- p>[</c-><c- mf>2</c-> <c- o>*</c-> hiddenSize<c- p>,</c-> <c- mf>0</c-><c- p>],</c-> <c- p>[</c->hiddenSize<c- p>,</c-> input_size<c- p>]))</c->
+          builder<c- p>.</c->transpose<c- p>(</c->builder<c- p>.</c->slice<c- p>(</c->weight<c- p>,</c-> <c- p>[</c-><c- mf>2</c-> <c- o>*</c-> hiddenSize<c- p>,</c-> <c- mf>0</c-><c- p>],</c-> <c- p>[</c->hiddenSize<c- p>,</c-> inputSize<c- p>]))</c->
           <c- p>),</c->
         builder<c- p>.</c->matmul<c- p>(</c->
           builder<c- p>.</c->mul<c- p>(</c->r<c- p>,</c-> hiddenState<c- p>),</c->
-          builder<c- p>.</c->transpose<c- p>(</c->builder<c- p>.</c->slice<c- p>(</c->recurrentWeight<c- p>,</c-> <c- p>[</c-><c- mf>2</c-> <c- o>*</c-> hiddenSize<c- p>,</c-> <c- mf>0</c-><c- p>],</c-> <c- p>[</c->hiddenSize<c- p>,</c-> hidden_size<c- p>]))</c->
+          builder<c- p>.</c->transpose<c- p>(</c->builder<c- p>.</c->slice<c- p>(</c->recurrentWeight<c- p>,</c-> <c- p>[</c-><c- mf>2</c-> <c- o>*</c-> hiddenSize<c- p>,</c-> <c- mf>0</c-><c- p>],</c-> <c- p>[</c->hiddenSize<c- p>,</c-> hiddenSize<c- p>]))</c->
           <c- p>)</c->
         <c- p>)</c->
       <c- p>)</c->
@@ -2952,46 +4578,26 @@ <h4 class="heading settled" data-level="7.6.14" id="api-mlgraphbuilder-grucell">
 <c- c1>// compute the new hidden state</c->
 <c- k>return</c-> builder<c- p>.</c->add<c- p>(</c->builder<c- p>.</c->mul<c- p>(</c->z<c- p>,</c-> hiddenState<c- p>),</c-> builder<c- p>.</c->mul<c- p>(</c->n<c- p>,</c-> builder<c- p>.</c->sub<c- p>(</c->one<c- p>,</c-> z<c- p>)));</c->
 </pre>
-     </div>
-    </div>
-    <h4 class="heading settled" data-level="7.6.15" id="api-mlgraphbuilder-hard-sigmoid"><span class="secno">7.6.15. </span><span class="content">The hardSigmoid() method</span><a class="self-link" href="#api-mlgraphbuilder-hard-sigmoid"></a></h4>
-     Calculate the <a href="https://en.wikipedia.org/wiki/Hard_sigmoid"> non-smooth function</a> used in place of a sigmoid function on the input tensor. 
+    </details>
+   </div>
+   <h4 class="heading settled" data-level="7.6.16" id="api-mlgraphbuilder-hard-sigmoid"><span class="secno">7.6.16. </span><span class="content">The hardSigmoid() method</span><a class="self-link" href="#api-mlgraphbuilder-hard-sigmoid"></a></h4>
+    Calculate the non-smooth <a href="https://en.wikipedia.org/wiki/Hard_sigmoid">hard sigmoid function</a> on the input tensor, used instead of the sigmoid function for faster computation. 
 <pre class="idl highlight def"><c- b>dictionary</c-> <dfn class="dfn-paneled idl-code" data-dfn-type="dictionary" data-export id="dictdef-mlhardsigmoidoptions"><code><c- g>MLHardSigmoidOptions</c-></code></dfn> {
-  <a class="idl-code" data-link-type="interface" href="https://webidl.spec.whatwg.org/#idl-float" id="ref-for-idl-float①②"><c- b>float</c-></a> <dfn class="idl-code" data-default="0.2" data-dfn-for="MLHardSigmoidOptions" data-dfn-type="dict-member" data-export data-type="float " id="dom-mlhardsigmoidoptions-alpha"><code><c- g>alpha</c-></code><a class="self-link" href="#dom-mlhardsigmoidoptions-alpha"></a></dfn> = 0.2;
-  <a class="idl-code" data-link-type="interface" href="https://webidl.spec.whatwg.org/#idl-float" id="ref-for-idl-float①③"><c- b>float</c-></a> <dfn class="idl-code" data-default="0.5" data-dfn-for="MLHardSigmoidOptions" data-dfn-type="dict-member" data-export data-type="float " id="dom-mlhardsigmoidoptions-beta"><code><c- g>beta</c-></code><a class="self-link" href="#dom-mlhardsigmoidoptions-beta"></a></dfn> = 0.5;
+  <a class="idl-code" data-link-type="interface" href="https://webidl.spec.whatwg.org/#idl-float" id="ref-for-idl-float①⑧"><c- b>float</c-></a> <a class="idl-code" data-default="0.2" data-link-type="dict-member" data-type="float " href="#dom-mlhardsigmoidoptions-alpha" id="ref-for-dom-mlhardsigmoidoptions-alpha"><c- g>alpha</c-></a> = 0.2;
+  <a class="idl-code" data-link-type="interface" href="https://webidl.spec.whatwg.org/#idl-float" id="ref-for-idl-float①⑨"><c- b>float</c-></a> <a class="idl-code" data-default="0.5" data-link-type="dict-member" data-type="float " href="#dom-mlhardsigmoidoptions-beta" id="ref-for-dom-mlhardsigmoidoptions-beta"><c- g>beta</c-></a> = 0.5;
 };
 
-<c- b>partial</c-> <c- b>interface</c-> <a class="idl-code" data-link-type="interface" href="#mlgraphbuilder" id="ref-for-mlgraphbuilder③①"><c- g>MLGraphBuilder</c-></a> {
-  <a data-link-type="idl-name" href="#mloperand" id="ref-for-mloperand①⑤①"><c- n>MLOperand</c-></a> <dfn class="idl-code" data-dfn-for="MLGraphBuilder" data-dfn-type="method" data-export data-lt="hardSigmoid(x, options)|hardSigmoid(x)" id="dom-mlgraphbuilder-hardsigmoid"><code><c- g>hardSigmoid</c-></code><a class="self-link" href="#dom-mlgraphbuilder-hardsigmoid"></a></dfn>(<a data-link-type="idl-name" href="#mloperand" id="ref-for-mloperand①⑤②"><c- n>MLOperand</c-></a> <dfn class="idl-code" data-dfn-for="MLGraphBuilder/hardSigmoid(x, options), MLGraphBuilder/hardSigmoid(x)" data-dfn-type="argument" data-export id="dom-mlgraphbuilder-hardsigmoid-x-options-x"><code><c- g>x</c-></code><a class="self-link" href="#dom-mlgraphbuilder-hardsigmoid-x-options-x"></a></dfn>, <c- b>optional</c-> <a data-link-type="idl-name" href="#dictdef-mlhardsigmoidoptions" id="ref-for-dictdef-mlhardsigmoidoptions"><c- n>MLHardSigmoidOptions</c-></a> <dfn class="idl-code" data-dfn-for="MLGraphBuilder/hardSigmoid(x, options), MLGraphBuilder/hardSigmoid(x)" data-dfn-type="argument" data-export id="dom-mlgraphbuilder-hardsigmoid-x-options-options"><code><c- g>options</c-></code><a class="self-link" href="#dom-mlgraphbuilder-hardsigmoid-x-options-options"></a></dfn> = {});
-  <a data-link-type="idl-name" href="#mlactivation" id="ref-for-mlactivation①⑥"><c- n>MLActivation</c-></a> <dfn class="idl-code" data-dfn-for="MLGraphBuilder" data-dfn-type="method" data-export data-lt="hardSigmoid(options)|hardSigmoid()" id="dom-mlgraphbuilder-hardsigmoid-options"><code><c- g>hardSigmoid</c-></code><a class="self-link" href="#dom-mlgraphbuilder-hardsigmoid-options"></a></dfn>(<c- b>optional</c-> <a data-link-type="idl-name" href="#dictdef-mlhardsigmoidoptions" id="ref-for-dictdef-mlhardsigmoidoptions①"><c- n>MLHardSigmoidOptions</c-></a> <dfn class="idl-code" data-dfn-for="MLGraphBuilder/hardSigmoid(options), MLGraphBuilder/hardSigmoid()" data-dfn-type="argument" data-export id="dom-mlgraphbuilder-hardsigmoid-options-options"><code><c- g>options</c-></code><a class="self-link" href="#dom-mlgraphbuilder-hardsigmoid-options-options"></a></dfn> = {});
+<c- b>partial</c-> <c- b>interface</c-> <a class="idl-code" data-link-type="interface" href="#mlgraphbuilder" id="ref-for-mlgraphbuilder③④"><c- g>MLGraphBuilder</c-></a> {
+  <a data-link-type="idl-name" href="#mloperand" id="ref-for-mloperand①⑧②"><c- n>MLOperand</c-></a> <a class="idl-code" data-link-type="method" href="#dom-mlgraphbuilder-hardsigmoid" id="ref-for-dom-mlgraphbuilder-hardsigmoid"><c- g>hardSigmoid</c-></a>(<a data-link-type="idl-name" href="#mloperand" id="ref-for-mloperand①⑧③"><c- n>MLOperand</c-></a> <dfn class="idl-code" data-dfn-for="MLGraphBuilder/hardSigmoid(input, options), MLGraphBuilder/hardSigmoid(input)" data-dfn-type="argument" data-export id="dom-mlgraphbuilder-hardsigmoid-input-options-input"><code><c- g>input</c-></code><a class="self-link" href="#dom-mlgraphbuilder-hardsigmoid-input-options-input"></a></dfn>, <c- b>optional</c-> <a data-link-type="idl-name" href="#dictdef-mlhardsigmoidoptions" id="ref-for-dictdef-mlhardsigmoidoptions"><c- n>MLHardSigmoidOptions</c-></a> <dfn class="idl-code" data-dfn-for="MLGraphBuilder/hardSigmoid(input, options), MLGraphBuilder/hardSigmoid(input)" data-dfn-type="argument" data-export id="dom-mlgraphbuilder-hardsigmoid-input-options-options"><code><c- g>options</c-></code><a class="self-link" href="#dom-mlgraphbuilder-hardsigmoid-input-options-options"></a></dfn> = {});
+  <a data-link-type="idl-name" href="#mlactivation" id="ref-for-mlactivation②⑨"><c- n>MLActivation</c-></a> <a class="idl-code" data-link-type="method" href="#dom-mlgraphbuilder-hardsigmoid-options" id="ref-for-dom-mlgraphbuilder-hardsigmoid-options"><c- g>hardSigmoid</c-></a>(<c- b>optional</c-> <a data-link-type="idl-name" href="#dictdef-mlhardsigmoidoptions" id="ref-for-dictdef-mlhardsigmoidoptions①"><c- n>MLHardSigmoidOptions</c-></a> <dfn class="idl-code" data-dfn-for="MLGraphBuilder/hardSigmoid(options), MLGraphBuilder/hardSigmoid()" data-dfn-type="argument" data-export id="dom-mlgraphbuilder-hardsigmoid-options-options"><code><c- g>options</c-></code><a class="self-link" href="#dom-mlgraphbuilder-hardsigmoid-options-options"></a></dfn> = {});
 };
 </pre>
-    <div class="algorithm" data-algorithm="hard-sigmoid">
-      <strong>Arguments:</strong> 
-     <ul>
-      <li data-md>
-       <p><em>x</em>: an <code class="idl"><a data-link-type="idl" href="#mloperand" id="ref-for-mloperand①⑤③">MLOperand</a></code>. The input tensor.</p>
-      <li data-md>
-       <p><em>options</em>: an optional <code class="idl"><a data-link-type="idl" href="#dictdef-mlhardsigmoidoptions" id="ref-for-dictdef-mlhardsigmoidoptions②">MLHardSigmoidOptions</a></code>. The optional parameters of the operation.</p>
-       <ul>
-        <li data-md>
-         <p><em>alpha</em>: a <code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#idl-float" id="ref-for-idl-float①④">float</a></code> scalar multiplier, default to 0.2.</p>
-        <li data-md>
-         <p><em>beta</em>: a <code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#idl-float" id="ref-for-idl-float①⑤">float</a></code> scalar addition, default to 0.5.</p>
-       </ul>
-     </ul>
-     <p><strong>Returns:</strong></p>
-     <ul>
-      <li data-md>
-       <p>an <code class="idl"><a data-link-type="idl" href="#mloperand" id="ref-for-mloperand①⑤④">MLOperand</a></code>. The output tensor of the same shape as <em>x</em>.</p>
-      <li data-md>
-       <p>an <code class="idl"><a data-link-type="idl" href="#mlactivation" id="ref-for-mlactivation①⑦">MLActivation</a></code>. The activation function representing the hard sigmoid operation.</p>
-     </ul>
-     <div class="note" role="note">
-       The behavior of this operation can be generically emulated from the usage of
+   <div class="note" role="note">
+    <details open>
+     <summary> The behavior of this operation can be generically emulated from the usage of
     other operations as follow. However, user agents typically have a more
     efficient implementation for it, therefore its usage is encouraged from the
-    performance standpoint. 
+    performance standpoint. </summary>
 <pre class="highlight"><c- k>return</c-> builder<c- p>.</c->max<c- p>(</c->
            builder<c- p>.</c->min<c- p>(</c->
                builder<c- p>.</c->add<c- p>(</c->
@@ -3000,33 +4606,108 @@ <h4 class="heading settled" data-level="7.6.15" id="api-mlgraphbuilder-hard-sigm
                builder<c- p>.</c->constant<c- p>(</c-><c- mf>1</c-><c- p>)),</c->
            builder<c- p>.</c->constant<c- p>(</c-><c- mf>0</c-><c- p>));</c->
 </pre>
-     </div>
-    </div>
-    <h4 class="heading settled" data-level="7.6.16" id="api-mlgraphbuilder-hard-swish"><span class="secno">7.6.16. </span><span class="content">The hardSwish() method</span><a class="self-link" href="#api-mlgraphbuilder-hard-swish"></a></h4>
-     Computes the nonlinear function <code>y = x * max(0, min(6, (x + 3))) / 6</code> that is introduced by <a data-link-type="biblio" href="#biblio-mobilenetv3" title="Searching for MobileNetV3">[MobileNetV3]</a> on the input tensor element-wise. 
-<pre class="idl highlight def"><c- b>partial</c-> <c- b>interface</c-> <a class="idl-code" data-link-type="interface" href="#mlgraphbuilder" id="ref-for-mlgraphbuilder③②"><c- g>MLGraphBuilder</c-></a> {
-  <a data-link-type="idl-name" href="#mloperand" id="ref-for-mloperand①⑤⑤"><c- n>MLOperand</c-></a> <dfn class="idl-code" data-dfn-for="MLGraphBuilder" data-dfn-type="method" data-export data-lt="hardSwish(x)" id="dom-mlgraphbuilder-hardswish"><code><c- g>hardSwish</c-></code><a class="self-link" href="#dom-mlgraphbuilder-hardswish"></a></dfn>(<a data-link-type="idl-name" href="#mloperand" id="ref-for-mloperand①⑤⑥"><c- n>MLOperand</c-></a> <dfn class="idl-code" data-dfn-for="MLGraphBuilder/hardSwish(x)" data-dfn-type="argument" data-export id="dom-mlgraphbuilder-hardswish-x-x"><code><c- g>x</c-></code><a class="self-link" href="#dom-mlgraphbuilder-hardswish-x-x"></a></dfn>);
-  <a data-link-type="idl-name" href="#mlactivation" id="ref-for-mlactivation①⑧"><c- n>MLActivation</c-></a> <dfn class="idl-code" data-dfn-for="MLGraphBuilder" data-dfn-type="method" data-export data-lt="hardSwish()" id="dom-mlgraphbuilder-hardswish①"><code><c- g>hardSwish</c-></code><a class="self-link" href="#dom-mlgraphbuilder-hardswish①"></a></dfn>();
-};
-</pre>
-    <div class="algorithm" data-algorithm="hard-swish">
-      <strong>Arguments:</strong> 
-     <ul>
+    </details>
+   </div>
+   <p><code class="idl"><a data-link-type="idl" href="#dictdef-mlhardsigmoidoptions" id="ref-for-dictdef-mlhardsigmoidoptions②">MLHardSigmoidOptions</a></code> has the following members:</p>
+   <dl>
+    <dt data-md><dfn class="dfn-paneled idl-code" data-dfn-for="MLHardSigmoidOptions" data-dfn-type="dict-member" data-export id="dom-mlhardsigmoidoptions-alpha"><code>alpha</code></dfn>, <span> of type <a data-link-type="idl-name" href="https://webidl.spec.whatwg.org/#idl-float" id="ref-for-idl-float②⓪">float</a>, defaulting to <code>0.2</code></span>
+    <dd data-md>
+     <p>A <code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#idl-float" id="ref-for-idl-float②①">float</a></code> scalar multiplier.
+ The default value is 0.2.</p>
+    <dt data-md><dfn class="dfn-paneled idl-code" data-dfn-for="MLHardSigmoidOptions" data-dfn-type="dict-member" data-export id="dom-mlhardsigmoidoptions-beta"><code>beta</code></dfn>, <span> of type <a data-link-type="idl-name" href="https://webidl.spec.whatwg.org/#idl-float" id="ref-for-idl-float②②">float</a>, defaulting to <code>0.5</code></span>
+    <dd data-md>
+     <p>A <code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#idl-float" id="ref-for-idl-float②③">float</a></code> scalar addition.
+ The default value is 0.5.</p>
+   </dl>
+   <h5 class="heading settled" data-level="7.6.16.1" id="api-mlgraphbuilder-hardsigmoid-input-options"><span class="secno">7.6.16.1. </span><span class="content">The <code class="idl"><a data-link-type="idl" href="#dom-mlgraphbuilder-hardsigmoid" id="ref-for-dom-mlgraphbuilder-hardsigmoid①">hardSigmoid(input, options)</a></code> method</span><a class="self-link" href="#api-mlgraphbuilder-hardsigmoid-input-options"></a></h5>
+   <div>
+     <strong>Arguments:</strong> 
+    <ul>
+     <li data-md>
+      <p><em>input</em>: an <code class="idl"><a data-link-type="idl" href="#mloperand" id="ref-for-mloperand①⑧④">MLOperand</a></code>. The input tensor.</p>
+     <li data-md>
+      <p><em>options</em>: an optional <code class="idl"><a data-link-type="idl" href="#dictdef-mlhardsigmoidoptions" id="ref-for-dictdef-mlhardsigmoidoptions③">MLHardSigmoidOptions</a></code>. The optional parameters of the operation.</p>
+    </ul>
+    <p><strong>Returns:</strong></p>
+    <ul>
+     <li data-md>
+      <p>an <code class="idl"><a data-link-type="idl" href="#mloperand" id="ref-for-mloperand①⑧⑤">MLOperand</a></code>. The output tensor of the same shape as <em>input</em>.</p>
+    </ul>
+   </div>
+   <details class="algorithm" data-algorithm="hardSigmoid(input, options)" data-algorithm-for="MLGraphBuilder" open>
+    <summary> The <dfn class="dfn-paneled idl-code" data-dfn-for="MLGraphBuilder" data-dfn-type="method" data-export data-lt="hardSigmoid(input, options)|hardSigmoid(input)" id="dom-mlgraphbuilder-hardsigmoid"><code>hardSigmoid(<var>input</var>, <var>options</var>)</code></dfn> method steps are: </summary>
+    <div class="algorithm-steps">
+     <ol>
       <li data-md>
-       <p><em>x</em>: an <code class="idl"><a data-link-type="idl" href="#mloperand" id="ref-for-mloperand①⑤⑦">MLOperand</a></code>. The input tensor.</p>
-     </ul>
-     <p><strong>Returns:</strong></p>
-     <ul>
+       <p><a data-link-type="dfn" href="https://infra.spec.whatwg.org/#assert" id="ref-for-assert④⓪">Assert</a>: the type of <var>input</var> is <code class="idl"><a data-link-type="idl" href="#mloperand" id="ref-for-mloperand①⑧⑥">MLOperand</a></code>.</p>
       <li data-md>
-       <p>an <code class="idl"><a data-link-type="idl" href="#mloperand" id="ref-for-mloperand①⑤⑧">MLOperand</a></code>. The output tensor of the same shape as <em>x</em>.</p>
+       <p>If any of the following sub-steps fail, <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#dfn-throw" id="ref-for-dfn-throw①④⑥">throw</a> an "<code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#operationerror" id="ref-for-operationerror②⑤">OperationError</a></code>" <code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#idl-DOMException" id="ref-for-idl-DOMException⑦⑧">DOMException</a></code>.</p>
+       <ol>
+        <li data-md>
+         <p>Let <var>output</var> be the result of <a data-link-type="dfn" href="#copy-an-mloperand" id="ref-for-copy-an-mloperand③">copying an MLOperand</a> given <var>input</var>.</p>
+        <li data-md>
+         <p>Make a request to the underlying platform to:</p>
+         <ol>
+          <li data-md>
+           <p>Let <var>opImpl</var> be an <a data-link-type="dfn" href="https://infra.spec.whatwg.org/#implementation-defined" id="ref-for-implementation-defined③⓪">implementation-defined</a> platform operator for the hard sigmoid operation, given <var>options</var>.</p>
+          <li data-md>
+           <p>Store a reference of <var>opImpl</var> in <var>output</var>.<code class="idl"><a data-link-type="idl" href="#dom-mloperand-operator-slot" id="ref-for-dom-mloperand-operator-slot①⓪">[[operator]]</a></code>.</p>
+          <li data-md>
+           <p>Create an <a data-link-type="dfn" href="https://infra.spec.whatwg.org/#implementation-defined" id="ref-for-implementation-defined③①">implementation-defined</a> platform operand <var>outputImpl</var> to represent the output, given <var>output</var> and <var>opImpl</var>.</p>
+          <li data-md>
+           <p>Store a reference to <var>outputImpl</var> in <var>output</var>.<code class="idl"><a data-link-type="idl" href="#dom-mloperand-operand-slot" id="ref-for-dom-mloperand-operand-slot③③">[[operand]]</a></code>.</p>
+         </ol>
+        <li data-md>
+         <p>Connect <var>input</var>.<code class="idl"><a data-link-type="idl" href="#dom-mloperand-operand-slot" id="ref-for-dom-mloperand-operand-slot③④">[[operand]]</a></code> as input to <var>opImpl</var>.</p>
+        <li data-md>
+         <p>Connect <var>output</var>.<code class="idl"><a data-link-type="idl" href="#dom-mloperand-operand-slot" id="ref-for-dom-mloperand-operand-slot③⑤">[[operand]]</a></code> as output to <var>opImpl</var>.</p>
+       </ol>
       <li data-md>
-       <p>an <code class="idl"><a data-link-type="idl" href="#mlactivation" id="ref-for-mlactivation①⑨">MLActivation</a></code>. The activation function representing the hard-swish operation.</p>
-     </ul>
-     <div class="note" role="note">
-       The behavior of this operation can be generically emulated from the usage of
+       <p>Return <var>output</var>.</p>
+     </ol>
+    </div>
+   </details>
+   <h5 class="heading settled" data-level="7.6.16.2" id="api-mlgraphbuilder-hardsigmoid-options"><span class="secno">7.6.16.2. </span><span class="content">The <code class="idl"><a data-link-type="idl" href="#dom-mlgraphbuilder-hardsigmoid-options" id="ref-for-dom-mlgraphbuilder-hardsigmoid-options①">hardSigmoid(options)</a></code> method</span><a class="self-link" href="#api-mlgraphbuilder-hardsigmoid-options"></a></h5>
+   <div>
+     <strong>Arguments:</strong> 
+    <ul>
+     <li data-md>
+      <p><em>options</em>: an optional <code class="idl"><a data-link-type="idl" href="#dictdef-mlhardsigmoidoptions" id="ref-for-dictdef-mlhardsigmoidoptions④">MLHardSigmoidOptions</a></code>. The optional parameters of the operation.</p>
+    </ul>
+    <p><strong>Returns:</strong></p>
+    <ul>
+     <li data-md>
+      <p>an <code class="idl"><a data-link-type="idl" href="#mlactivation" id="ref-for-mlactivation③⓪">MLActivation</a></code>. The activation function representing the hard sigmoid operation.</p>
+    </ul>
+   </div>
+   <details class="algorithm" data-algorithm="hardSigmoid(options)" data-algorithm-for="MLGraphBuilder" open>
+    <summary> The <dfn class="dfn-paneled idl-code" data-dfn-for="MLGraphBuilder" data-dfn-type="method" data-export data-lt="hardSigmoid(options)|hardSigmoid()" id="dom-mlgraphbuilder-hardsigmoid-options"><code>hardSigmoid(<var>options</var>)</code></dfn> method steps are: </summary>
+    <div class="algorithm-steps">
+     <ol>
+      <li data-md>
+       <p>Let <var>op</var> be the result of <a data-link-type="dfn" href="#create-an-mlactivation" id="ref-for-create-an-mlactivation②">creating an MLActivation</a> given <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#this" id="ref-for-this②⓪">this</a>, <code>"hardSigmoid"</code> and <var>options</var>.</p>
+       <ol>
+        <li data-md>
+         <p>If that <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#dfn-throw" id="ref-for-dfn-throw①④⑦">throws</a> an error, re-<a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#dfn-throw" id="ref-for-dfn-throw①④⑧">throw</a> the error.</p>
+       </ol>
+      <li data-md>
+       <p>Return <var>op</var>.</p>
+     </ol>
+    </div>
+   </details>
+   <h4 class="heading settled" data-level="7.6.17" id="api-mlgraphbuilder-hard-swish"><span class="secno">7.6.17. </span><span class="content">The hardSwish() method</span><a class="self-link" href="#api-mlgraphbuilder-hard-swish"></a></h4>
+    Computes the nonlinear function <code>y = x * max(0, min(6, (x + 3))) / 6</code> that is introduced by <a data-link-type="biblio" href="#biblio-mobilenetv3" title="Searching for MobileNetV3">[MobileNetV3]</a> on the input tensor element-wise. 
+<pre class="idl highlight def"><c- b>partial</c-> <c- b>interface</c-> <a class="idl-code" data-link-type="interface" href="#mlgraphbuilder" id="ref-for-mlgraphbuilder③⑤"><c- g>MLGraphBuilder</c-></a> {
+  <a data-link-type="idl-name" href="#mloperand" id="ref-for-mloperand①⑧⑦"><c- n>MLOperand</c-></a> <a class="idl-code" data-link-type="method" href="#dom-mlgraphbuilder-hardswish" id="ref-for-dom-mlgraphbuilder-hardswish"><c- g>hardSwish</c-></a>(<a data-link-type="idl-name" href="#mloperand" id="ref-for-mloperand①⑧⑧"><c- n>MLOperand</c-></a> <dfn class="idl-code" data-dfn-for="MLGraphBuilder/hardSwish(input)" data-dfn-type="argument" data-export id="dom-mlgraphbuilder-hardswish-input-input"><code><c- g>input</c-></code><a class="self-link" href="#dom-mlgraphbuilder-hardswish-input-input"></a></dfn>);
+  <a data-link-type="idl-name" href="#mlactivation" id="ref-for-mlactivation③①"><c- n>MLActivation</c-></a> <a class="idl-code" data-link-type="method" href="#hardswish-noargs" id="ref-for-hardswish-noargs"><c- g>hardSwish</c-></a>();
+};
+</pre>
+   <div class="note" role="note">
+    <details open>
+     <summary> The behavior of this operation can be generically emulated from the usage of
     other operations as follow. However, user agents typically have a more
     efficient implementation for it, therefore its usage is encouraged from the
-    performance standpoint. 
+    performance standpoint. </summary>
 <pre class="highlight"><c- k>return</c-> builder<c- p>.</c->div<c- p>(</c->
            builder<c- p>.</c->mul<c- p>(</c->
                x<c- p>,</c->
@@ -3037,45 +4718,167 @@ <h4 class="heading settled" data-level="7.6.16" id="api-mlgraphbuilder-hard-swis
                        builder<c- p>.</c->add<c- p>(</c->x<c- p>,</c-> builder<c- p>.</c->constant<c- p>(</c-><c- mf>3</c-><c- p>))))),</c->
            builder<c- p>.</c->constant<c- p>(</c-><c- mf>6</c-><c- p>));</c->
 </pre>
-     </div>
+    </details>
+   </div>
+   <h5 class="heading settled" data-level="7.6.17.1" id="api-mlgraphbuilder-hardswish-input"><span class="secno">7.6.17.1. </span><span class="content">The <code class="idl"><a data-link-type="idl" href="#dom-mlgraphbuilder-hardswish" id="ref-for-dom-mlgraphbuilder-hardswish①">hardSwish(input)</a></code> method</span><a class="self-link" href="#api-mlgraphbuilder-hardswish-input"></a></h5>
+   <div>
+     <strong>Arguments:</strong> 
+    <ul>
+     <li data-md>
+      <p><em>input</em>: an <code class="idl"><a data-link-type="idl" href="#mloperand" id="ref-for-mloperand①⑧⑨">MLOperand</a></code>. The input tensor.</p>
+    </ul>
+    <p><strong>Returns:</strong></p>
+    <ul>
+     <li data-md>
+      <p>an <code class="idl"><a data-link-type="idl" href="#mloperand" id="ref-for-mloperand①⑨⓪">MLOperand</a></code>. The output tensor of the same shape as <em>input</em>.</p>
+    </ul>
+   </div>
+   <details class="algorithm" data-algorithm="hardSwish(input)" data-algorithm-for="MLGraphBuilder" open>
+    <summary> The <dfn class="dfn-paneled idl-code" data-dfn-for="MLGraphBuilder" data-dfn-type="method" data-export id="dom-mlgraphbuilder-hardswish"><code>hardSwish(<var>input</var>)</code></dfn> method steps are: </summary>
+    <div class="algorithm-steps">
+     <ol>
+      <li data-md>
+       <p>If any of the following sub-steps fail, <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#dfn-throw" id="ref-for-dfn-throw①④⑨">throw</a> an "<code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#operationerror" id="ref-for-operationerror②⑥">OperationError</a></code>" <code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#idl-DOMException" id="ref-for-idl-DOMException⑦⑨">DOMException</a></code>.</p>
+       <ol>
+        <li data-md>
+         <p>Let <var>output</var> be the result of <a data-link-type="dfn" href="#copy-an-mloperand" id="ref-for-copy-an-mloperand④">copying an MLOperand</a> given <var>input</var>.</p>
+        <li data-md>
+         <p>Make a request to the underlying platform to:</p>
+         <ol>
+          <li data-md>
+           <p>Let <var>opImpl</var> be an <a data-link-type="dfn" href="https://infra.spec.whatwg.org/#implementation-defined" id="ref-for-implementation-defined③②">implementation-defined</a> platform operator for the hard-swish operation.</p>
+          <li data-md>
+           <p>Store a reference of <var>opImpl</var> in <var>output</var>.<code class="idl"><a data-link-type="idl" href="#dom-mloperand-operator-slot" id="ref-for-dom-mloperand-operator-slot①①">[[operator]]</a></code>.</p>
+          <li data-md>
+           <p>Create an <a data-link-type="dfn" href="https://infra.spec.whatwg.org/#implementation-defined" id="ref-for-implementation-defined③③">implementation-defined</a> platform operand <var>outputImpl</var> to represent the output, given <var>output</var> and <var>opImpl</var>.</p>
+          <li data-md>
+           <p>Store a reference to <var>outputImpl</var> in <var>output</var>.<code class="idl"><a data-link-type="idl" href="#dom-mloperand-operand-slot" id="ref-for-dom-mloperand-operand-slot③⑥">[[operand]]</a></code>.</p>
+         </ol>
+        <li data-md>
+         <p>Connect <var>input</var>.<code class="idl"><a data-link-type="idl" href="#dom-mloperand-operand-slot" id="ref-for-dom-mloperand-operand-slot③⑦">[[operand]]</a></code> as input to <var>opImpl</var>.</p>
+        <li data-md>
+         <p>Connect <var>output</var>.<code class="idl"><a data-link-type="idl" href="#dom-mloperand-operand-slot" id="ref-for-dom-mloperand-operand-slot③⑧">[[operand]]</a></code> as output to <var>opImpl</var>.</p>
+       </ol>
+      <li data-md>
+       <p>Return <var>output</var>.</p>
+     </ol>
+    </div>
+   </details>
+   <h5 class="heading settled" data-level="7.6.17.2" id="api-mlgraphbuilder-hardswish"><span class="secno">7.6.17.2. </span><span class="content">The <code class="idl"><a data-link-type="idl" href="#hardswish-noargs" id="ref-for-hardswish-noargs①">hardSwish()</a></code> method</span><a class="self-link" href="#api-mlgraphbuilder-hardswish"></a></h5>
+   <div>
+     <strong>Arguments:</strong> 
+    <ul>
+     <li data-md>
+      <p>None.</p>
+    </ul>
+    <p><strong>Returns:</strong></p>
+    <ul>
+     <li data-md>
+      <p>an <code class="idl"><a data-link-type="idl" href="#mlactivation" id="ref-for-mlactivation③②">MLActivation</a></code>. The activation function representing the hard-swish operation.</p>
+    </ul>
+   </div>
+   <details class="algorithm" data-algorithm="hardSwish()" data-algorithm-for="MLGraphBuilder" open>
+    <summary> The <dfn class="dfn-paneled idl-code" data-dfn-for="MLGraphBuilder" data-dfn-type="method" data-export id="hardswish-noargs"><code>hardSwish()</code></dfn> method steps are: </summary>
+    <div class="algorithm-steps">
+     <ol>
+      <li data-md>
+       <p>Let <var>op</var> be the result of <a data-link-type="dfn" href="#create-an-mlactivation" id="ref-for-create-an-mlactivation③">creating an MLActivation</a> given <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#this" id="ref-for-this②①">this</a> and <code>"hardSwish"</code>.</p>
+       <ol>
+        <li data-md>
+         <p>If that <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#dfn-throw" id="ref-for-dfn-throw①⑤⓪">throws</a> an error, re-<a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#dfn-throw" id="ref-for-dfn-throw①⑤①">throw</a> the error.</p>
+       </ol>
+      <li data-md>
+       <p>Return <var>op</var>.</p>
+     </ol>
     </div>
-    <h4 class="heading settled" data-level="7.6.17" id="api-mlgraphbuilder-instancenorm"><span class="secno">7.6.17. </span><span class="content">The instanceNormalization() method</span><a class="self-link" href="#api-mlgraphbuilder-instancenorm"></a></h4>
-     Normalize the input features using <a data-link-type="biblio" href="#biblio-instance-normalization" title="Instance Normalization: The Missing Ingredient for Fast Stylization">[Instance-Normalization]</a>. Unlike <a href="#api-mlgraphbuilder-batchnorm">§ 7.6.4 The batchNormalization() method</a> where the mean and variance values used in the calculation are previously computed across the batch dimension during the model training phase, the mean and variance values used in the calculation of an instance normalization are computed internally on the fly per input feature. 
+   </details>
+   <h4 class="heading settled" data-level="7.6.18" id="api-mlgraphbuilder-instancenorm"><span class="secno">7.6.18. </span><span class="content">The instanceNormalization() method</span><a class="self-link" href="#api-mlgraphbuilder-instancenorm"></a></h4>
+    Normalize the input features using <a data-link-type="biblio" href="#biblio-instance-normalization" title="Instance Normalization: The Missing Ingredient for Fast Stylization">[Instance-Normalization]</a>. Unlike <a href="#api-mlgraphbuilder-batchnorm">§ 7.6.5 The batchNormalization() method</a> where the mean and variance values used in the calculation are previously computed across the batch dimension during the model training phase, the mean and variance values used in the calculation of an instance normalization are computed internally on the fly per input feature. 
 <pre class="idl highlight def"><c- b>dictionary</c-> <dfn class="dfn-paneled idl-code" data-dfn-type="dictionary" data-export id="dictdef-mlinstancenormalizationoptions"><code><c- g>MLInstanceNormalizationOptions</c-></code></dfn> {
-  <a data-link-type="idl-name" href="#mloperand" id="ref-for-mloperand①⑤⑨"><c- n>MLOperand</c-></a> <dfn class="idl-code" data-dfn-for="MLInstanceNormalizationOptions" data-dfn-type="dict-member" data-export data-type="MLOperand " id="dom-mlinstancenormalizationoptions-scale"><code><c- g>scale</c-></code><a class="self-link" href="#dom-mlinstancenormalizationoptions-scale"></a></dfn>;
-  <a data-link-type="idl-name" href="#mloperand" id="ref-for-mloperand①⑥⓪"><c- n>MLOperand</c-></a> <dfn class="idl-code" data-dfn-for="MLInstanceNormalizationOptions" data-dfn-type="dict-member" data-export data-type="MLOperand " id="dom-mlinstancenormalizationoptions-bias"><code><c- g>bias</c-></code><a class="self-link" href="#dom-mlinstancenormalizationoptions-bias"></a></dfn>;
-  <a class="idl-code" data-link-type="interface" href="https://webidl.spec.whatwg.org/#idl-float" id="ref-for-idl-float①⑥"><c- b>float</c-></a> <dfn class="idl-code" data-default="1e-5" data-dfn-for="MLInstanceNormalizationOptions" data-dfn-type="dict-member" data-export data-type="float " id="dom-mlinstancenormalizationoptions-epsilon"><code><c- g>epsilon</c-></code><a class="self-link" href="#dom-mlinstancenormalizationoptions-epsilon"></a></dfn> = 1e-5;
-  <a data-link-type="idl-name" href="#enumdef-mlinputoperandlayout" id="ref-for-enumdef-mlinputoperandlayout④"><c- n>MLInputOperandLayout</c-></a> <dfn class="idl-code" data-default="&quot;nchw&quot;" data-dfn-for="MLInstanceNormalizationOptions" data-dfn-type="dict-member" data-export data-type="MLInputOperandLayout " id="dom-mlinstancenormalizationoptions-layout"><code><c- g>layout</c-></code><a class="self-link" href="#dom-mlinstancenormalizationoptions-layout"></a></dfn> = "nchw";
+  <a data-link-type="idl-name" href="#mloperand" id="ref-for-mloperand①⑨①"><c- n>MLOperand</c-></a> <a class="idl-code" data-link-type="dict-member" data-type="MLOperand " href="#dom-mlinstancenormalizationoptions-scale" id="ref-for-dom-mlinstancenormalizationoptions-scale"><c- g>scale</c-></a>;
+  <a data-link-type="idl-name" href="#mloperand" id="ref-for-mloperand①⑨②"><c- n>MLOperand</c-></a> <a class="idl-code" data-link-type="dict-member" data-type="MLOperand " href="#dom-mlinstancenormalizationoptions-bias" id="ref-for-dom-mlinstancenormalizationoptions-bias"><c- g>bias</c-></a>;
+  <a class="idl-code" data-link-type="interface" href="https://webidl.spec.whatwg.org/#idl-float" id="ref-for-idl-float②④"><c- b>float</c-></a> <a class="idl-code" data-default="1e-5" data-link-type="dict-member" data-type="float " href="#dom-mlinstancenormalizationoptions-epsilon" id="ref-for-dom-mlinstancenormalizationoptions-epsilon"><c- g>epsilon</c-></a> = 1e-5;
+  <a data-link-type="idl-name" href="#enumdef-mlinputoperandlayout" id="ref-for-enumdef-mlinputoperandlayout⑥"><c- n>MLInputOperandLayout</c-></a> <a class="idl-code" data-default="&quot;nchw&quot;" data-link-type="dict-member" data-type="MLInputOperandLayout " href="#dom-mlinstancenormalizationoptions-layout" id="ref-for-dom-mlinstancenormalizationoptions-layout"><c- g>layout</c-></a> = "nchw";
 };
 
-<c- b>partial</c-> <c- b>interface</c-> <a class="idl-code" data-link-type="interface" href="#mlgraphbuilder" id="ref-for-mlgraphbuilder③③"><c- g>MLGraphBuilder</c-></a> {
-  <a data-link-type="idl-name" href="#mloperand" id="ref-for-mloperand①⑥①"><c- n>MLOperand</c-></a> <dfn class="idl-code" data-dfn-for="MLGraphBuilder" data-dfn-type="method" data-export data-lt="instanceNormalization(input, options)|instanceNormalization(input)" id="dom-mlgraphbuilder-instancenormalization"><code><c- g>instanceNormalization</c-></code><a class="self-link" href="#dom-mlgraphbuilder-instancenormalization"></a></dfn>(<a data-link-type="idl-name" href="#mloperand" id="ref-for-mloperand①⑥②"><c- n>MLOperand</c-></a> <dfn class="idl-code" data-dfn-for="MLGraphBuilder/instanceNormalization(input, options), MLGraphBuilder/instanceNormalization(input)" data-dfn-type="argument" data-export id="dom-mlgraphbuilder-instancenormalization-input-options-input"><code><c- g>input</c-></code><a class="self-link" href="#dom-mlgraphbuilder-instancenormalization-input-options-input"></a></dfn>,
+<c- b>partial</c-> <c- b>interface</c-> <a class="idl-code" data-link-type="interface" href="#mlgraphbuilder" id="ref-for-mlgraphbuilder③⑥"><c- g>MLGraphBuilder</c-></a> {
+  <a data-link-type="idl-name" href="#mloperand" id="ref-for-mloperand①⑨③"><c- n>MLOperand</c-></a> <a class="idl-code" data-link-type="method" href="#dom-mlgraphbuilder-instancenormalization" id="ref-for-dom-mlgraphbuilder-instancenormalization"><c- g>instanceNormalization</c-></a>(<a data-link-type="idl-name" href="#mloperand" id="ref-for-mloperand①⑨④"><c- n>MLOperand</c-></a> <dfn class="idl-code" data-dfn-for="MLGraphBuilder/instanceNormalization(input, options), MLGraphBuilder/instanceNormalization(input)" data-dfn-type="argument" data-export id="dom-mlgraphbuilder-instancenormalization-input-options-input"><code><c- g>input</c-></code><a class="self-link" href="#dom-mlgraphbuilder-instancenormalization-input-options-input"></a></dfn>,
                                 <c- b>optional</c-> <a data-link-type="idl-name" href="#dictdef-mlinstancenormalizationoptions" id="ref-for-dictdef-mlinstancenormalizationoptions"><c- n>MLInstanceNormalizationOptions</c-></a> <dfn class="idl-code" data-dfn-for="MLGraphBuilder/instanceNormalization(input, options), MLGraphBuilder/instanceNormalization(input)" data-dfn-type="argument" data-export id="dom-mlgraphbuilder-instancenormalization-input-options-options"><code><c- g>options</c-></code><a class="self-link" href="#dom-mlgraphbuilder-instancenormalization-input-options-options"></a></dfn> = {});
 };
 </pre>
-    <div class="algorithm" data-algorithm="instancenorm">
-      <strong>Arguments:</strong> 
-     <ul>
+   <p>The <code class="idl"><a data-link-type="idl" href="#dictdef-mlinstancenormalizationoptions" id="ref-for-dictdef-mlinstancenormalizationoptions①">MLInstanceNormalizationOptions</a></code> members are:</p>
+   <dl>
+    <dt data-md><dfn class="dfn-paneled idl-code" data-dfn-for="MLInstanceNormalizationOptions" data-dfn-type="dict-member" data-export id="dom-mlinstancenormalizationoptions-scale"><code>scale</code></dfn>, <span> of type <a data-link-type="idl-name" href="#mloperand" id="ref-for-mloperand①⑨⑤">MLOperand</a></span>
+    <dd data-md>
+     <p>An <code class="idl"><a data-link-type="idl" href="#mloperand" id="ref-for-mloperand①⑨⑥">MLOperand</a></code>. Specifies the 1-D tensor of the scaling values whose <a data-link-type="dfn" href="https://infra.spec.whatwg.org/#list-size" id="ref-for-list-size②⑧">size</a> is equal to the number of channels, i.e. the size of the feature dimension of the input. For example, for an <var>input</var> tensor with <code>nchw</code> layout, the <a data-link-type="dfn" href="https://infra.spec.whatwg.org/#list-size" id="ref-for-list-size②⑨">size</a> is the value of <var>input</var>.<code class="idl"><a data-link-type="idl" href="#dom-mloperand-descriptor-slot" id="ref-for-dom-mloperand-descriptor-slot⑤③">[[descriptor]]</a></code>.<code class="idl"><a data-link-type="idl" href="#dom-mloperanddescriptor-dimensions" id="ref-for-dom-mloperanddescriptor-dimensions⑤①">dimensions</a></code>[1].</p>
+    <dt data-md><dfn class="dfn-paneled idl-code" data-dfn-for="MLInstanceNormalizationOptions" data-dfn-type="dict-member" data-export id="dom-mlinstancenormalizationoptions-bias"><code>bias</code></dfn>, <span> of type <a data-link-type="idl-name" href="#mloperand" id="ref-for-mloperand①⑨⑦">MLOperand</a></span>
+    <dd data-md>
+     <p>An <code class="idl"><a data-link-type="idl" href="#mloperand" id="ref-for-mloperand①⑨⑧">MLOperand</a></code>. Specifies the 1-D tensor of the bias values whose <a data-link-type="dfn" href="https://infra.spec.whatwg.org/#list-size" id="ref-for-list-size③⓪">size</a> is equal to the size of the feature dimension of the input. For example, for an <var>input</var> tensor with <code>nchw</code> layout, the len<a data-link-type="dfn" href="https://infra.spec.whatwg.org/#list-size" id="ref-for-list-size③①">size</a>gth is the value of <var>input</var>.<code class="idl"><a data-link-type="idl" href="#dom-mloperand-descriptor-slot" id="ref-for-dom-mloperand-descriptor-slot⑤④">[[descriptor]]</a></code>.<code class="idl"><a data-link-type="idl" href="#dom-mloperanddescriptor-dimensions" id="ref-for-dom-mloperanddescriptor-dimensions⑤②">dimensions</a></code>[1].</p>
+    <dt data-md><dfn class="dfn-paneled idl-code" data-dfn-for="MLInstanceNormalizationOptions" data-dfn-type="dict-member" data-export id="dom-mlinstancenormalizationoptions-epsilon"><code>epsilon</code></dfn>, <span> of type <a data-link-type="idl-name" href="https://webidl.spec.whatwg.org/#idl-float" id="ref-for-idl-float②⑤">float</a>, defaulting to <code>1e-5</code></span>
+    <dd data-md>
+     <p>A <code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#idl-float" id="ref-for-idl-float②⑥">float</a></code> scalar. Specifies a small value to prevent computational error due to divide-by-zero.</p>
+    <dt data-md><dfn class="dfn-paneled idl-code" data-dfn-for="MLInstanceNormalizationOptions" data-dfn-type="dict-member" data-export id="dom-mlinstancenormalizationoptions-layout"><code>layout</code></dfn>, <span> of type <a data-link-type="idl-name" href="#enumdef-mlinputoperandlayout" id="ref-for-enumdef-mlinputoperandlayout⑦">MLInputOperandLayout</a>, defaulting to <code>"nchw"</code></span>
+    <dd data-md>
+     <p>An <code class="idl"><a data-link-type="idl" href="#enumdef-mlinputoperandlayout" id="ref-for-enumdef-mlinputoperandlayout⑧">MLInputOperandLayout</a></code>. Specifies the layout format of the input.</p>
+   </dl>
+   <div>
+     <strong>Arguments:</strong> 
+    <ul>
+     <li data-md>
+      <p><em>input</em>: an <code class="idl"><a data-link-type="idl" href="#mloperand" id="ref-for-mloperand①⑨⑨">MLOperand</a></code>. The input 4-D tensor.</p>
+     <li data-md>
+      <p><em>options</em>: an optional <code class="idl"><a data-link-type="idl" href="#dictdef-mlinstancenormalizationoptions" id="ref-for-dictdef-mlinstancenormalizationoptions②">MLInstanceNormalizationOptions</a></code>. The optional parameters of the operation.</p>
+    </ul>
+    <p><strong>Returns:</strong> an <code class="idl"><a data-link-type="idl" href="#mloperand" id="ref-for-mloperand②⓪⓪">MLOperand</a></code>. The instance-normalized 4-D tensor of the same shape as <em>input</em>.</p>
+   </div>
+   <details class="algorithm" data-algorithm="instanceNormalization(input, options)" data-algorithm-for="MLGraphBuilder" open>
+    <summary> The <dfn class="dfn-paneled idl-code" data-dfn-for="MLGraphBuilder" data-dfn-type="method" data-export data-lt="instanceNormalization(input, options)|instanceNormalization(input)" id="dom-mlgraphbuilder-instancenormalization"><code>instanceNormalization(<var>input</var>, <var>options</var>)</code></dfn> method steps are: </summary>
+    <div class="algorithm-steps">
+     <ol>
       <li data-md>
-       <p><em>input</em>: an <code class="idl"><a data-link-type="idl" href="#mloperand" id="ref-for-mloperand①⑥③">MLOperand</a></code>. The input 4-D tensor.</p>
+       <p><a data-link-type="dfn" href="https://infra.spec.whatwg.org/#assert" id="ref-for-assert④①">Assert</a>: the type of <var>input</var> is <code class="idl"><a data-link-type="idl" href="#mloperand" id="ref-for-mloperand②⓪①">MLOperand</a></code>.</p>
       <li data-md>
-       <p><em>options</em>: an optional <code class="idl"><a data-link-type="idl" href="#dictdef-mlinstancenormalizationoptions" id="ref-for-dictdef-mlinstancenormalizationoptions①">MLInstanceNormalizationOptions</a></code>. The optional parameters of the operation.</p>
-       <ul>
+       <p>If the <a data-link-type="dfn" href="#mloperand-rank" id="ref-for-mloperand-rank①⓪">rank</a> of <var>input</var> is not 4, then <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#dfn-throw" id="ref-for-dfn-throw①⑤②">throw</a> a "<code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#dataerror" id="ref-for-dataerror④④">DataError</a></code>" <code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#idl-DOMException" id="ref-for-idl-DOMException⑧⓪">DOMException</a></code>.</p>
+      <li data-md>
+       <p><a data-link-type="dfn" href="https://infra.spec.whatwg.org/#assert" id="ref-for-assert④②">Assert</a>:  the type of <var>options</var>.<code class="idl"><a data-link-type="idl" href="#dom-mlinstancenormalizationoptions-scale" id="ref-for-dom-mlinstancenormalizationoptions-scale①">scale</a></code> is <code class="idl"><a data-link-type="idl" href="#mloperand" id="ref-for-mloperand②⓪②">MLOperand</a></code>.</p>
+      <li data-md>
+       <p>If the <a data-link-type="dfn" href="#mloperand-rank" id="ref-for-mloperand-rank①①">rank</a> of <var>options</var>.<code class="idl"><a data-link-type="idl" href="#dom-mlinstancenormalizationoptions-scale" id="ref-for-dom-mlinstancenormalizationoptions-scale②">scale</a></code> is not equal to the <a data-link-type="dfn" href="https://infra.spec.whatwg.org/#list-size" id="ref-for-list-size③②">size</a> of the channel dimension of <var>input</var>, then <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#dfn-throw" id="ref-for-dfn-throw①⑤③">throw</a> a "<code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#dataerror" id="ref-for-dataerror④⑤">DataError</a></code>" <code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#idl-DOMException" id="ref-for-idl-DOMException⑧①">DOMException</a></code>.</p>
+      <li data-md>
+       <p><a data-link-type="dfn" href="https://infra.spec.whatwg.org/#assert" id="ref-for-assert④③">Assert</a>: the type of <var>options</var>.<code class="idl"><a data-link-type="idl" href="#dom-mlinstancenormalizationoptions-bias" id="ref-for-dom-mlinstancenormalizationoptions-bias①">bias</a></code> is <code class="idl"><a data-link-type="idl" href="#mloperand" id="ref-for-mloperand②⓪③">MLOperand</a></code>.</p>
+      <li data-md>
+       <p>If the <a data-link-type="dfn" href="#mloperand-rank" id="ref-for-mloperand-rank①②">rank</a> of <var>options</var>.<code class="idl"><a data-link-type="idl" href="#dom-mlinstancenormalizationoptions-bias" id="ref-for-dom-mlinstancenormalizationoptions-bias②">bias</a></code> is not equal to the <a data-link-type="dfn" href="https://infra.spec.whatwg.org/#list-size" id="ref-for-list-size③③">size</a> of the channel dimension of <var>input</var>, then <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#dfn-throw" id="ref-for-dfn-throw①⑤④">throw</a> a "<code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#dataerror" id="ref-for-dataerror④⑥">DataError</a></code>" <code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#idl-DOMException" id="ref-for-idl-DOMException⑧②">DOMException</a></code>.</p>
+      <li data-md>
+       <p>If any of the following sub-steps fail, <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#dfn-throw" id="ref-for-dfn-throw①⑤⑤">throw</a> an "<code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#operationerror" id="ref-for-operationerror②⑦">OperationError</a></code>" <code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#idl-DOMException" id="ref-for-idl-DOMException⑧③">DOMException</a></code>.</p>
+       <ol>
         <li data-md>
-         <p><em>scale</em>: an <code class="idl"><a data-link-type="idl" href="#mloperand" id="ref-for-mloperand①⑥④">MLOperand</a></code>. The 1-D tensor of the scaling values whose length is equal to the size of the feature dimension of the input e.g. for the input tensor with <em>nchw</em> layout, the feature dimension is 1.</p>
+         <p>Let <var>output</var> be the result of <a data-link-type="dfn" href="#copy-an-mloperand" id="ref-for-copy-an-mloperand⑤">copying an MLOperand</a> given <var>input</var>.</p>
         <li data-md>
-         <p><em>bias</em>: an <code class="idl"><a data-link-type="idl" href="#mloperand" id="ref-for-mloperand①⑥⑤">MLOperand</a></code>. The 1-D tensor of the bias values whose length is equal to the size of the feature dimension of the input e.g. for the input tensor with <em>nchw</em> layout, the feature dimension is 1.</p>
+         <p>Make a request to the underlying platform to:</p>
+         <ol>
+          <li data-md>
+           <p>Let <var>opImpl</var> be an <a data-link-type="dfn" href="https://infra.spec.whatwg.org/#implementation-defined" id="ref-for-implementation-defined③④">implementation-defined</a> platform operator for the instance normalization operation, given <var>options</var>.</p>
+          <li data-md>
+           <p>Store a reference of <var>opImpl</var> in <var>output</var>.<code class="idl"><a data-link-type="idl" href="#dom-mloperand-operator-slot" id="ref-for-dom-mloperand-operator-slot①②">[[operator]]</a></code>.</p>
+          <li data-md>
+           <p>Create an <a data-link-type="dfn" href="https://infra.spec.whatwg.org/#implementation-defined" id="ref-for-implementation-defined③⑤">implementation-defined</a> platform operand <var>outputImpl</var> to represent the output, given <var>output</var> and <var>opImpl</var>.</p>
+          <li data-md>
+           <p>Store a reference to <var>outputImpl</var> in <var>output</var>.<code class="idl"><a data-link-type="idl" href="#dom-mloperand-operand-slot" id="ref-for-dom-mloperand-operand-slot③⑨">[[operand]]</a></code>.</p>
+         </ol>
         <li data-md>
-         <p><em>epsilon</em>: a <code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#idl-float" id="ref-for-idl-float①⑦">float</a></code> scalar. A small value to prevent computational error due to divide-by-zero. The default value is 0.00001 when not specified.</p>
+         <p>Connect <var>input</var>.<code class="idl"><a data-link-type="idl" href="#dom-mloperand-operand-slot" id="ref-for-dom-mloperand-operand-slot④⓪">[[operand]]</a></code> as input to <var>opImpl</var>.</p>
         <li data-md>
-         <p><em>layout</em>: an <code class="idl"><a data-link-type="idl" href="#enumdef-mlinputoperandlayout" id="ref-for-enumdef-mlinputoperandlayout⑤">MLInputOperandLayout</a></code>. This option specifies the layout format of the input. The default value is <em>"nchw"</em>.</p>
-       </ul>
-     </ul>
-     <p><strong>Returns:</strong> an <code class="idl"><a data-link-type="idl" href="#mloperand" id="ref-for-mloperand①⑥⑥">MLOperand</a></code>. The instance-normalized 4-D tensor of the same shape as the input tensor.</p>
-     <div class="note" role="note">
-       The behavior of this operation when the input tensor is 4-D of the <em>"nchw"</em> layout can be generically emulated from
+         <p>Connect <var>output</var>.<code class="idl"><a data-link-type="idl" href="#dom-mloperand-operand-slot" id="ref-for-dom-mloperand-operand-slot④①">[[operand]]</a></code> as output to <var>opImpl</var>.</p>
+       </ol>
+      <li data-md>
+       <p>Return <var>output</var>.</p>
+     </ol>
+    </div>
+   </details>
+   <div class="note" role="note">
+    <details open>
+     <summary> The behavior of this operation when the input tensor is 4-D of the <em>"nchw"</em> layout can be generically emulated from
     the usage of other operations as follow. However, user agents typically have a more efficient implementation for it,
-    therefore its usage is encouraged from the performance standpoint. 
+    therefore its usage is encouraged from the performance standpoint. </summary>
 <pre class="highlight"><c- c1>// The mean reductions happen over the spatial dimensions of the input</c->
 <c- c1>// e.g. axis 2 and 3 of the input tensor.</c->
 <c- a>const</c-> reduceOptions <c- o>=</c-> <c- p>{</c-> axes<c- o>:</c-> <c- p>[</c-><c- mf>2</c-><c- p>,</c-><c- mf>3</c-><c- p>],</c-> keepDimensions<c- o>:</c-> <c- kc>true</c-> <c- p>};</c->
@@ -3103,156 +4906,428 @@ <h4 class="heading settled" data-level="7.6.17" id="api-mlgraphbuilder-instancen
   builder<c- p>.</c->reshape<c- p>(</c->options<c- p>.</c->bias<c- p>,</c-> shape<c- p>)</c->
   <c- p>);</c->
 </pre>
-     </div>
-    </div>
-    <h4 class="heading settled" data-level="7.6.18" id="api-mlgraphbuilder-leakyrelu"><span class="secno">7.6.18. </span><span class="content">The leakyRelu() method</span><a class="self-link" href="#api-mlgraphbuilder-leakyrelu"></a></h4>
-     Calculate the <a href="https://en.wikipedia.org/wiki/Rectifier_(neural_networks)#Leaky_ReLU"> leaky version of rectified linear function</a> on the input tensor element-wise. The calculation follows the expression <code>max(0, x) + alpha ∗ min(0, x)</code>. 
+    </details>
+   </div>
+   <h4 class="heading settled" data-level="7.6.19" id="api-mlgraphbuilder-leakyrelu"><span class="secno">7.6.19. </span><span class="content">The leakyRelu() method</span><a class="self-link" href="#api-mlgraphbuilder-leakyrelu"></a></h4>
+    Calculate the <a href="https://en.wikipedia.org/wiki/Rectifier_(neural_networks)#Leaky_ReLU"> leaky version of rectified linear function</a> on the input tensor element-wise. The calculation follows the expression <code>max(0, x) + alpha ∗ min(0, x)</code>. 
 <pre class="idl highlight def"><c- b>dictionary</c-> <dfn class="dfn-paneled idl-code" data-dfn-type="dictionary" data-export id="dictdef-mlleakyreluoptions"><code><c- g>MLLeakyReluOptions</c-></code></dfn> {
-  <a class="idl-code" data-link-type="interface" href="https://webidl.spec.whatwg.org/#idl-float" id="ref-for-idl-float①⑧"><c- b>float</c-></a> <dfn class="idl-code" data-default="0.01" data-dfn-for="MLLeakyReluOptions" data-dfn-type="dict-member" data-export data-type="float " id="dom-mlleakyreluoptions-alpha"><code><c- g>alpha</c-></code><a class="self-link" href="#dom-mlleakyreluoptions-alpha"></a></dfn> = 0.01;
+  <a class="idl-code" data-link-type="interface" href="https://webidl.spec.whatwg.org/#idl-float" id="ref-for-idl-float②⑦"><c- b>float</c-></a> <a class="idl-code" data-default="0.01" data-link-type="dict-member" data-type="float " href="#dom-mlleakyreluoptions-alpha" id="ref-for-dom-mlleakyreluoptions-alpha"><c- g>alpha</c-></a> = 0.01;
 };
 
-<c- b>partial</c-> <c- b>interface</c-> <a class="idl-code" data-link-type="interface" href="#mlgraphbuilder" id="ref-for-mlgraphbuilder③④"><c- g>MLGraphBuilder</c-></a> {
-  <a data-link-type="idl-name" href="#mloperand" id="ref-for-mloperand①⑥⑦"><c- n>MLOperand</c-></a> <dfn class="idl-code" data-dfn-for="MLGraphBuilder" data-dfn-type="method" data-export data-lt="leakyRelu(x, options)|leakyRelu(x)" id="dom-mlgraphbuilder-leakyrelu"><code><c- g>leakyRelu</c-></code><a class="self-link" href="#dom-mlgraphbuilder-leakyrelu"></a></dfn>(<a data-link-type="idl-name" href="#mloperand" id="ref-for-mloperand①⑥⑧"><c- n>MLOperand</c-></a> <dfn class="idl-code" data-dfn-for="MLGraphBuilder/leakyRelu(x, options), MLGraphBuilder/leakyRelu(x)" data-dfn-type="argument" data-export id="dom-mlgraphbuilder-leakyrelu-x-options-x"><code><c- g>x</c-></code><a class="self-link" href="#dom-mlgraphbuilder-leakyrelu-x-options-x"></a></dfn>, <c- b>optional</c-> <a data-link-type="idl-name" href="#dictdef-mlleakyreluoptions" id="ref-for-dictdef-mlleakyreluoptions"><c- n>MLLeakyReluOptions</c-></a> <dfn class="idl-code" data-dfn-for="MLGraphBuilder/leakyRelu(x, options), MLGraphBuilder/leakyRelu(x)" data-dfn-type="argument" data-export id="dom-mlgraphbuilder-leakyrelu-x-options-options"><code><c- g>options</c-></code><a class="self-link" href="#dom-mlgraphbuilder-leakyrelu-x-options-options"></a></dfn> = {});
-  <a data-link-type="idl-name" href="#mlactivation" id="ref-for-mlactivation②⓪"><c- n>MLActivation</c-></a> <dfn class="idl-code" data-dfn-for="MLGraphBuilder" data-dfn-type="method" data-export data-lt="leakyRelu(options)|leakyRelu()" id="dom-mlgraphbuilder-leakyrelu-options"><code><c- g>leakyRelu</c-></code><a class="self-link" href="#dom-mlgraphbuilder-leakyrelu-options"></a></dfn>(<c- b>optional</c-> <a data-link-type="idl-name" href="#dictdef-mlleakyreluoptions" id="ref-for-dictdef-mlleakyreluoptions①"><c- n>MLLeakyReluOptions</c-></a> <dfn class="idl-code" data-dfn-for="MLGraphBuilder/leakyRelu(options), MLGraphBuilder/leakyRelu()" data-dfn-type="argument" data-export id="dom-mlgraphbuilder-leakyrelu-options-options"><code><c- g>options</c-></code><a class="self-link" href="#dom-mlgraphbuilder-leakyrelu-options-options"></a></dfn> = {});
+<c- b>partial</c-> <c- b>interface</c-> <a class="idl-code" data-link-type="interface" href="#mlgraphbuilder" id="ref-for-mlgraphbuilder③⑦"><c- g>MLGraphBuilder</c-></a> {
+  <a data-link-type="idl-name" href="#mloperand" id="ref-for-mloperand②⓪④"><c- n>MLOperand</c-></a> <a class="idl-code" data-link-type="method" href="#dom-mlgraphbuilder-leakyrelu" id="ref-for-dom-mlgraphbuilder-leakyrelu"><c- g>leakyRelu</c-></a>(<a data-link-type="idl-name" href="#mloperand" id="ref-for-mloperand②⓪⑤"><c- n>MLOperand</c-></a> <dfn class="idl-code" data-dfn-for="MLGraphBuilder/leakyRelu(input, options), MLGraphBuilder/leakyRelu(input)" data-dfn-type="argument" data-export id="dom-mlgraphbuilder-leakyrelu-input-options-input"><code><c- g>input</c-></code><a class="self-link" href="#dom-mlgraphbuilder-leakyrelu-input-options-input"></a></dfn>, <c- b>optional</c-> <a data-link-type="idl-name" href="#dictdef-mlleakyreluoptions" id="ref-for-dictdef-mlleakyreluoptions"><c- n>MLLeakyReluOptions</c-></a> <dfn class="idl-code" data-dfn-for="MLGraphBuilder/leakyRelu(input, options), MLGraphBuilder/leakyRelu(input)" data-dfn-type="argument" data-export id="dom-mlgraphbuilder-leakyrelu-input-options-options"><code><c- g>options</c-></code><a class="self-link" href="#dom-mlgraphbuilder-leakyrelu-input-options-options"></a></dfn> = {});
+  <a data-link-type="idl-name" href="#mlactivation" id="ref-for-mlactivation③③"><c- n>MLActivation</c-></a> <a class="idl-code" data-link-type="method" href="#dom-mlgraphbuilder-leakyrelu-options" id="ref-for-dom-mlgraphbuilder-leakyrelu-options"><c- g>leakyRelu</c-></a>(<c- b>optional</c-> <a data-link-type="idl-name" href="#dictdef-mlleakyreluoptions" id="ref-for-dictdef-mlleakyreluoptions①"><c- n>MLLeakyReluOptions</c-></a> <dfn class="idl-code" data-dfn-for="MLGraphBuilder/leakyRelu(options), MLGraphBuilder/leakyRelu()" data-dfn-type="argument" data-export id="dom-mlgraphbuilder-leakyrelu-options-options"><code><c- g>options</c-></code><a class="self-link" href="#dom-mlgraphbuilder-leakyrelu-options-options"></a></dfn> = {});
 };
 </pre>
-    <div class="algorithm" data-algorithm="leakyrelu">
-      <strong>Arguments:</strong> 
-     <ul>
-      <li data-md>
-       <p><em>x</em>: an <code class="idl"><a data-link-type="idl" href="#mloperand" id="ref-for-mloperand①⑥⑨">MLOperand</a></code>. The input tensor.</p>
-      <li data-md>
-       <p><em>options</em>: an optional <code class="idl"><a data-link-type="idl" href="#dictdef-mlleakyreluoptions" id="ref-for-dictdef-mlleakyreluoptions②">MLLeakyReluOptions</a></code>. The optional parameters of the operation.</p>
-       <ul>
-        <li data-md>
-         <p><em>alpha</em>: a <code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#idl-float" id="ref-for-idl-float①⑨">float</a></code> scalar multiplier, default to 0.01.</p>
-       </ul>
-     </ul>
-     <p><strong>Returns:</strong></p>
-     <ul>
-      <li data-md>
-       <p>an <code class="idl"><a data-link-type="idl" href="#mloperand" id="ref-for-mloperand①⑦⓪">MLOperand</a></code>. The output tensor of the same shape as <em>x</em>.</p>
-      <li data-md>
-       <p>an <code class="idl"><a data-link-type="idl" href="#mlactivation" id="ref-for-mlactivation②①">MLActivation</a></code>. The activation function representing the leaky relu operation.</p>
-     </ul>
-     <div class="note" role="note">
-       The behavior of this operation can be generically emulated from the usage of
+   <div class="note" role="note">
+    <details open>
+     <summary> The behavior of this operation can be generically emulated from the usage of
     other operations as follow. However, user agents typically have a more
     efficient implementation for it, therefore its usage is encouraged from the
-    performance standpoint. 
+    performance standpoint. </summary>
 <pre class="highlight"><c- k>return</c-> builder<c- p>.</c->add<c- p>(</c->builder<c- p>.</c->max<c- p>(</c->builder<c- p>.</c->constant<c- p>(</c-><c- mf>0</c-><c- p>),</c-> x<c- p>),</c->
           builder<c- p>.</c->mul<c- p>(</c->builder<c- p>.</c->constant<c- p>(</c->options<c- p>.</c->alpha<c- p>),</c-> builder<c- p>.</c->min<c- p>(</c->builder<c- p>.</c->constant<c- p>(</c-><c- mf>0</c-><c- p>),</c-> x<c- p>)));</c->
 </pre>
-     </div>
+    </details>
+   </div>
+   <p><code class="idl"><a data-link-type="idl" href="#dictdef-mlleakyreluoptions" id="ref-for-dictdef-mlleakyreluoptions②">MLLeakyReluOptions</a></code> has the following members:</p>
+   <dl>
+    <dt data-md><dfn class="dfn-paneled idl-code" data-dfn-for="MLLeakyReluOptions" data-dfn-type="dict-member" data-export id="dom-mlleakyreluoptions-alpha"><code>alpha</code></dfn>, <span> of type <a data-link-type="idl-name" href="https://webidl.spec.whatwg.org/#idl-float" id="ref-for-idl-float②⑧">float</a>, defaulting to <code>0.01</code></span>
+    <dd data-md>
+     <p>A <code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#idl-float" id="ref-for-idl-float②⑨">float</a></code> scalar multiplier.
+ The default value is 0.01.</p>
+   </dl>
+   <h5 class="heading settled" data-level="7.6.19.1" id="api-mlgraphbuilder-leaky-relu-input-options"><span class="secno">7.6.19.1. </span><span class="content">The <code class="idl"><a data-link-type="idl" href="#dom-mlgraphbuilder-leakyrelu" id="ref-for-dom-mlgraphbuilder-leakyrelu①">leakyRelu(input, options)</a></code> method</span><a class="self-link" href="#api-mlgraphbuilder-leaky-relu-input-options"></a></h5>
+   <div>
+     <strong>Arguments:</strong> 
+    <ul>
+     <li data-md>
+      <p><em>input</em>: an <code class="idl"><a data-link-type="idl" href="#mloperand" id="ref-for-mloperand②⓪⑥">MLOperand</a></code>. The input tensor.</p>
+     <li data-md>
+      <p><em>options</em>: an optional <code class="idl"><a data-link-type="idl" href="#dictdef-mlleakyreluoptions" id="ref-for-dictdef-mlleakyreluoptions③">MLLeakyReluOptions</a></code>. The optional parameters of the operation.</p>
+    </ul>
+    <p><strong>Returns:</strong></p>
+    <ul>
+     <li data-md>
+      <p>an <code class="idl"><a data-link-type="idl" href="#mloperand" id="ref-for-mloperand②⓪⑦">MLOperand</a></code>. The output tensor of the same shape as <em>input</em>.</p>
+    </ul>
+   </div>
+   <details class="algorithm" data-algorithm="leakyRelu(input, options)" data-algorithm-for="MLGraphBuilder" open>
+    <summary> The <dfn class="dfn-paneled idl-code" data-dfn-for="MLGraphBuilder" data-dfn-type="method" data-export data-lt="leakyRelu(input, options)|leakyRelu(input)" id="dom-mlgraphbuilder-leakyrelu"><code>leakyRelu(<var>input</var>, <var>options</var>)</code></dfn> method steps are: </summary>
+    <div class="algorithm-steps">
+     <ol>
+      <li data-md>
+       <p>If any of the following sub-steps fail, <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#dfn-throw" id="ref-for-dfn-throw①⑤⑥">throw</a> an "<code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#operationerror" id="ref-for-operationerror②⑧">OperationError</a></code>" <code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#idl-DOMException" id="ref-for-idl-DOMException⑧④">DOMException</a></code>.</p>
+       <ol>
+        <li data-md>
+         <p>Let <var>output</var> be the result of <a data-link-type="dfn" href="#copy-an-mloperand" id="ref-for-copy-an-mloperand⑥">copying an MLOperand</a> given <var>input</var>.</p>
+        <li data-md>
+         <p>Make a request to the underlying platform to:</p>
+         <ol>
+          <li data-md>
+           <p>Let <var>opImpl</var> be an <a data-link-type="dfn" href="https://infra.spec.whatwg.org/#implementation-defined" id="ref-for-implementation-defined③⑥">implementation-defined</a> platform operator for the Leaky RELU operation, given <var>options</var>.</p>
+          <li data-md>
+           <p>Store a reference of <var>opImpl</var> in <var>output</var>.<code class="idl"><a data-link-type="idl" href="#dom-mloperand-operator-slot" id="ref-for-dom-mloperand-operator-slot①③">[[operator]]</a></code>.</p>
+          <li data-md>
+           <p>Create an <a data-link-type="dfn" href="https://infra.spec.whatwg.org/#implementation-defined" id="ref-for-implementation-defined③⑦">implementation-defined</a> platform operand <var>outputImpl</var> to represent the output, given <var>output</var> and <var>opImpl</var>.</p>
+          <li data-md>
+           <p>Store a reference to <var>outputImpl</var> in <var>output</var>.<code class="idl"><a data-link-type="idl" href="#dom-mloperand-operand-slot" id="ref-for-dom-mloperand-operand-slot④②">[[operand]]</a></code>.</p>
+         </ol>
+        <li data-md>
+         <p>Connect <var>input</var>.<code class="idl"><a data-link-type="idl" href="#dom-mloperand-operand-slot" id="ref-for-dom-mloperand-operand-slot④③">[[operand]]</a></code> as input to <var>opImpl</var>.</p>
+        <li data-md>
+         <p>Connect <var>output</var>.<code class="idl"><a data-link-type="idl" href="#dom-mloperand-operand-slot" id="ref-for-dom-mloperand-operand-slot④④">[[operand]]</a></code> as output to <var>opImpl</var>.</p>
+       </ol>
+      <li data-md>
+       <p>Return <var>output</var>.</p>
+     </ol>
+    </div>
+   </details>
+   <h5 class="heading settled" data-level="7.6.19.2" id="api-mlgraphbuilder-leaky-relu-options"><span class="secno">7.6.19.2. </span><span class="content">The <code class="idl"><a data-link-type="idl" href="#dom-mlgraphbuilder-leakyrelu-options" id="ref-for-dom-mlgraphbuilder-leakyrelu-options①">leakyRelu(options)</a></code> method</span><a class="self-link" href="#api-mlgraphbuilder-leaky-relu-options"></a></h5>
+   <div>
+     <strong>Arguments:</strong> 
+    <ul>
+     <li data-md>
+      <p><em>options</em>: an optional <code class="idl"><a data-link-type="idl" href="#dictdef-mlleakyreluoptions" id="ref-for-dictdef-mlleakyreluoptions④">MLLeakyReluOptions</a></code>. The optional parameters of the operation.</p>
+    </ul>
+    <p><strong>Returns:</strong></p>
+    <ul>
+     <li data-md>
+      <p>an <code class="idl"><a data-link-type="idl" href="#mlactivation" id="ref-for-mlactivation③④">MLActivation</a></code>. The activation function representing the leaky relu operation.</p>
+    </ul>
+   </div>
+   <details class="algorithm" data-algorithm="leakyRelu(options)" data-algorithm-for="MLGraphBuilder" open>
+    <summary> The <dfn class="dfn-paneled idl-code" data-dfn-for="MLGraphBuilder" data-dfn-type="method" data-export data-lt="leakyRelu(options)|leakyRelu()" id="dom-mlgraphbuilder-leakyrelu-options"><code>leakyRelu(<var>options</var>)</code></dfn> method steps are: </summary>
+    <div class="algorithm-steps">
+     <ol>
+      <li data-md>
+       <p>Let <var>op</var> be the result of <a data-link-type="dfn" href="#create-an-mlactivation" id="ref-for-create-an-mlactivation④">creating an MLActivation</a> given <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#this" id="ref-for-this②②">this</a>, <code>"leakyRelu"</code> and <var>options</var>.</p>
+       <ol>
+        <li data-md>
+         <p>If that <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#dfn-throw" id="ref-for-dfn-throw①⑤⑦">throws</a> an error, re-<a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#dfn-throw" id="ref-for-dfn-throw①⑤⑧">throw</a> the error.</p>
+       </ol>
+      <li data-md>
+       <p>Return <var>op</var>.</p>
+     </ol>
     </div>
-    <h4 class="heading settled" data-level="7.6.19" id="api-mlgraphbuilder-linear"><span class="secno">7.6.19. </span><span class="content">The linear() method</span><a class="self-link" href="#api-mlgraphbuilder-linear"></a></h4>
-     Calculate a linear function <code>y = alpha * x + beta</code> on the input tensor. 
+   </details>
+   <h4 class="heading settled" data-level="7.6.20" id="api-mlgraphbuilder-linear"><span class="secno">7.6.20. </span><span class="content">The linear() method</span><a class="self-link" href="#api-mlgraphbuilder-linear"></a></h4>
+    Calculate a linear function <code>y = alpha * x + beta</code> on the input tensor. 
 <pre class="idl highlight def"><c- b>dictionary</c-> <dfn class="dfn-paneled idl-code" data-dfn-type="dictionary" data-export id="dictdef-mllinearoptions"><code><c- g>MLLinearOptions</c-></code></dfn> {
-  <a class="idl-code" data-link-type="interface" href="https://webidl.spec.whatwg.org/#idl-float" id="ref-for-idl-float②⓪"><c- b>float</c-></a> <dfn class="idl-code" data-default="1" data-dfn-for="MLLinearOptions" data-dfn-type="dict-member" data-export data-type="float " id="dom-mllinearoptions-alpha"><code><c- g>alpha</c-></code><a class="self-link" href="#dom-mllinearoptions-alpha"></a></dfn> = 1;
-  <a class="idl-code" data-link-type="interface" href="https://webidl.spec.whatwg.org/#idl-float" id="ref-for-idl-float②①"><c- b>float</c-></a> <dfn class="idl-code" data-default="0" data-dfn-for="MLLinearOptions" data-dfn-type="dict-member" data-export data-type="float " id="dom-mllinearoptions-beta"><code><c- g>beta</c-></code><a class="self-link" href="#dom-mllinearoptions-beta"></a></dfn> = 0;
+  <a class="idl-code" data-link-type="interface" href="https://webidl.spec.whatwg.org/#idl-float" id="ref-for-idl-float③⓪"><c- b>float</c-></a> <a class="idl-code" data-default="1" data-link-type="dict-member" data-type="float " href="#dom-mllinearoptions-alpha" id="ref-for-dom-mllinearoptions-alpha"><c- g>alpha</c-></a> = 1;
+  <a class="idl-code" data-link-type="interface" href="https://webidl.spec.whatwg.org/#idl-float" id="ref-for-idl-float③①"><c- b>float</c-></a> <a class="idl-code" data-default="0" data-link-type="dict-member" data-type="float " href="#dom-mllinearoptions-beta" id="ref-for-dom-mllinearoptions-beta"><c- g>beta</c-></a> = 0;
 };
 
-<c- b>partial</c-> <c- b>interface</c-> <a class="idl-code" data-link-type="interface" href="#mlgraphbuilder" id="ref-for-mlgraphbuilder③⑤"><c- g>MLGraphBuilder</c-></a> {
-  <a data-link-type="idl-name" href="#mloperand" id="ref-for-mloperand①⑦①"><c- n>MLOperand</c-></a> <dfn class="idl-code" data-dfn-for="MLGraphBuilder" data-dfn-type="method" data-export data-lt="linear(x, options)|linear(x)" id="dom-mlgraphbuilder-linear"><code><c- g>linear</c-></code><a class="self-link" href="#dom-mlgraphbuilder-linear"></a></dfn>(<a data-link-type="idl-name" href="#mloperand" id="ref-for-mloperand①⑦②"><c- n>MLOperand</c-></a> <dfn class="idl-code" data-dfn-for="MLGraphBuilder/linear(x, options), MLGraphBuilder/linear(x)" data-dfn-type="argument" data-export id="dom-mlgraphbuilder-linear-x-options-x"><code><c- g>x</c-></code><a class="self-link" href="#dom-mlgraphbuilder-linear-x-options-x"></a></dfn>, <c- b>optional</c-> <a data-link-type="idl-name" href="#dictdef-mllinearoptions" id="ref-for-dictdef-mllinearoptions"><c- n>MLLinearOptions</c-></a> <dfn class="idl-code" data-dfn-for="MLGraphBuilder/linear(x, options), MLGraphBuilder/linear(x)" data-dfn-type="argument" data-export id="dom-mlgraphbuilder-linear-x-options-options"><code><c- g>options</c-></code><a class="self-link" href="#dom-mlgraphbuilder-linear-x-options-options"></a></dfn> = {});
-  <a data-link-type="idl-name" href="#mlactivation" id="ref-for-mlactivation②②"><c- n>MLActivation</c-></a> <dfn class="idl-code" data-dfn-for="MLGraphBuilder" data-dfn-type="method" data-export data-lt="linear(options)|linear()" id="dom-mlgraphbuilder-linear-options"><code><c- g>linear</c-></code><a class="self-link" href="#dom-mlgraphbuilder-linear-options"></a></dfn>(<c- b>optional</c-> <a data-link-type="idl-name" href="#dictdef-mllinearoptions" id="ref-for-dictdef-mllinearoptions①"><c- n>MLLinearOptions</c-></a> <dfn class="idl-code" data-dfn-for="MLGraphBuilder/linear(options), MLGraphBuilder/linear()" data-dfn-type="argument" data-export id="dom-mlgraphbuilder-linear-options-options"><code><c- g>options</c-></code><a class="self-link" href="#dom-mlgraphbuilder-linear-options-options"></a></dfn> = {});
+<c- b>partial</c-> <c- b>interface</c-> <a class="idl-code" data-link-type="interface" href="#mlgraphbuilder" id="ref-for-mlgraphbuilder③⑧"><c- g>MLGraphBuilder</c-></a> {
+  <a data-link-type="idl-name" href="#mloperand" id="ref-for-mloperand②⓪⑧"><c- n>MLOperand</c-></a> <a class="idl-code" data-link-type="method" href="#dom-mlgraphbuilder-linear" id="ref-for-dom-mlgraphbuilder-linear"><c- g>linear</c-></a>(<a data-link-type="idl-name" href="#mloperand" id="ref-for-mloperand②⓪⑨"><c- n>MLOperand</c-></a> <dfn class="idl-code" data-dfn-for="MLGraphBuilder/linear(input, options), MLGraphBuilder/linear(input)" data-dfn-type="argument" data-export id="dom-mlgraphbuilder-linear-input-options-input"><code><c- g>input</c-></code><a class="self-link" href="#dom-mlgraphbuilder-linear-input-options-input"></a></dfn>, <c- b>optional</c-> <a data-link-type="idl-name" href="#dictdef-mllinearoptions" id="ref-for-dictdef-mllinearoptions"><c- n>MLLinearOptions</c-></a> <dfn class="idl-code" data-dfn-for="MLGraphBuilder/linear(input, options), MLGraphBuilder/linear(input)" data-dfn-type="argument" data-export id="dom-mlgraphbuilder-linear-input-options-options"><code><c- g>options</c-></code><a class="self-link" href="#dom-mlgraphbuilder-linear-input-options-options"></a></dfn> = {});
+  <a data-link-type="idl-name" href="#mlactivation" id="ref-for-mlactivation③⑤"><c- n>MLActivation</c-></a> <a class="idl-code" data-link-type="method" href="#dom-mlgraphbuilder-linear-options" id="ref-for-dom-mlgraphbuilder-linear-options"><c- g>linear</c-></a>(<c- b>optional</c-> <a data-link-type="idl-name" href="#dictdef-mllinearoptions" id="ref-for-dictdef-mllinearoptions①"><c- n>MLLinearOptions</c-></a> <dfn class="idl-code" data-dfn-for="MLGraphBuilder/linear(options), MLGraphBuilder/linear()" data-dfn-type="argument" data-export id="dom-mlgraphbuilder-linear-options-options"><code><c- g>options</c-></code><a class="self-link" href="#dom-mlgraphbuilder-linear-options-options"></a></dfn> = {});
 };
 </pre>
-    <div class="algorithm" data-algorithm="linear">
-      <strong>Arguments:</strong> 
-     <ul>
-      <li data-md>
-       <p><em>x</em>: an <code class="idl"><a data-link-type="idl" href="#mloperand" id="ref-for-mloperand①⑦③">MLOperand</a></code>. The input tensor.</p>
-      <li data-md>
-       <p><em>options</em>: an optional <code class="idl"><a data-link-type="idl" href="#dictdef-mllinearoptions" id="ref-for-dictdef-mllinearoptions②">MLLinearOptions</a></code>. The optional parameters of the operation.</p>
-       <ul>
-        <li data-md>
-         <p><em>alpha</em>: a <code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#idl-float" id="ref-for-idl-float②②">float</a></code> scalar multiplier, default to 1.</p>
-        <li data-md>
-         <p><em>beta</em>: a <code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#idl-float" id="ref-for-idl-float②③">float</a></code> scalar addition, default to 0.</p>
-       </ul>
-     </ul>
-     <p><strong>Returns:</strong></p>
-     <ul>
-      <li data-md>
-       <p>an <code class="idl"><a data-link-type="idl" href="#mloperand" id="ref-for-mloperand①⑦④">MLOperand</a></code>. The output tensor of the same shape as <em>x</em>.</p>
-      <li data-md>
-       <p>an <code class="idl"><a data-link-type="idl" href="#mlactivation" id="ref-for-mlactivation②③">MLActivation</a></code>. The activation function representing the linear operation.</p>
-     </ul>
-     <div class="note" role="note">
-       The behavior of this operation can be generically emulated from the usage of
+   <div class="note" role="note">
+    <details open>
+     <summary> The behavior of this operation can be generically emulated from the usage of
     other operations as follow. However, user agents typically have a more
     efficient implementation for it, therefore its usage is encouraged from the
-    performance standpoint. 
+    performance standpoint. </summary>
 <pre class="highlight"><c- k>return</c-> builder<c- p>.</c->add<c- p>(</c->
           builder<c- p>.</c->mul<c- p>(</c->x<c- p>,</c-> builder<c- p>.</c->constant<c- p>(</c->options<c- p>.</c->alpha<c- p>)),</c->
           builder<c- p>.</c->constant<c- p>(</c->options<c- p>.</c->beta<c- p>));</c->
 </pre>
-     </div>
+    </details>
+   </div>
+   <p><code class="idl"><a data-link-type="idl" href="#dictdef-mllinearoptions" id="ref-for-dictdef-mllinearoptions②">MLLinearOptions</a></code> has the following members:</p>
+   <dl>
+    <dt data-md><dfn class="dfn-paneled idl-code" data-dfn-for="MLLinearOptions" data-dfn-type="dict-member" data-export id="dom-mllinearoptions-alpha"><code>alpha</code></dfn>, <span> of type <a data-link-type="idl-name" href="https://webidl.spec.whatwg.org/#idl-float" id="ref-for-idl-float③②">float</a>, defaulting to <code>1</code></span>
+    <dd data-md>
+     <p>A <code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#idl-float" id="ref-for-idl-float③③">float</a></code> scalar multiplier.
+ The default value is 1.</p>
+    <dt data-md><dfn class="dfn-paneled idl-code" data-dfn-for="MLLinearOptions" data-dfn-type="dict-member" data-export id="dom-mllinearoptions-beta"><code>beta</code></dfn>, <span> of type <a data-link-type="idl-name" href="https://webidl.spec.whatwg.org/#idl-float" id="ref-for-idl-float③④">float</a>, defaulting to <code>0</code></span>
+    <dd data-md>
+     <p>A <code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#idl-float" id="ref-for-idl-float③⑤">float</a></code> scalar addition.
+ The default value is 0.</p>
+   </dl>
+   <h5 class="heading settled" data-level="7.6.20.1" id="api-mlgraphbuilder-linear-input-options"><span class="secno">7.6.20.1. </span><span class="content">The <code class="idl"><a data-link-type="idl" href="#dom-mlgraphbuilder-linear" id="ref-for-dom-mlgraphbuilder-linear①">linear(input, options)</a></code> method</span><a class="self-link" href="#api-mlgraphbuilder-linear-input-options"></a></h5>
+   <div>
+     <strong>Arguments:</strong> 
+    <ul>
+     <li data-md>
+      <p><em>input</em>: an <code class="idl"><a data-link-type="idl" href="#mloperand" id="ref-for-mloperand②①⓪">MLOperand</a></code>. The input tensor.</p>
+     <li data-md>
+      <p><em>options</em>: an optional <code class="idl"><a data-link-type="idl" href="#dictdef-mllinearoptions" id="ref-for-dictdef-mllinearoptions③">MLLinearOptions</a></code>. The optional parameters of the operation.</p>
+    </ul>
+    <p><strong>Returns:</strong></p>
+    <ul>
+     <li data-md>
+      <p>an <code class="idl"><a data-link-type="idl" href="#mloperand" id="ref-for-mloperand②①①">MLOperand</a></code>. The output tensor of the same shape as <em>input</em>.</p>
+    </ul>
+   </div>
+   <details class="algorithm" data-algorithm="linear(input, options)" data-algorithm-for="MLGraphBuilder" open>
+    <summary> The <dfn class="dfn-paneled idl-code" data-dfn-for="MLGraphBuilder" data-dfn-type="method" data-export data-lt="linear(input, options)|linear(input)" id="dom-mlgraphbuilder-linear"><code>linear(<var>input</var>, <var>options</var>)</code></dfn> method steps are: </summary>
+    <div class="algorithm-steps">
+     <ol>
+      <li data-md>
+       <p>If any of the following sub-steps fail, <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#dfn-throw" id="ref-for-dfn-throw①⑤⑨">throw</a> an "<code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#operationerror" id="ref-for-operationerror②⑨">OperationError</a></code>" <code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#idl-DOMException" id="ref-for-idl-DOMException⑧⑤">DOMException</a></code>.</p>
+       <ol>
+        <li data-md>
+         <p>Let <var>output</var> be the result of <a data-link-type="dfn" href="#copy-an-mloperand" id="ref-for-copy-an-mloperand⑦">copying an MLOperand</a> given <var>input</var>.</p>
+        <li data-md>
+         <p>Make a request to the underlying platform to:</p>
+         <ol>
+          <li data-md>
+           <p>Let <var>opImpl</var> be an <a data-link-type="dfn" href="https://infra.spec.whatwg.org/#implementation-defined" id="ref-for-implementation-defined③⑧">implementation-defined</a> platform operator for the linear operation, given <var>options</var>.</p>
+          <li data-md>
+           <p>Store a reference of <var>opImpl</var> in <var>output</var>.<code class="idl"><a data-link-type="idl" href="#dom-mloperand-operator-slot" id="ref-for-dom-mloperand-operator-slot①④">[[operator]]</a></code>.</p>
+          <li data-md>
+           <p>Create an <a data-link-type="dfn" href="https://infra.spec.whatwg.org/#implementation-defined" id="ref-for-implementation-defined③⑨">implementation-defined</a> platform operand <var>outputImpl</var> to represent the output, given <var>output</var> and <var>opImpl</var>.</p>
+          <li data-md>
+           <p>Store a reference to <var>outputImpl</var> in <var>output</var>.<code class="idl"><a data-link-type="idl" href="#dom-mloperand-operand-slot" id="ref-for-dom-mloperand-operand-slot④⑤">[[operand]]</a></code>.</p>
+         </ol>
+        <li data-md>
+         <p>Connect <var>input</var>.<code class="idl"><a data-link-type="idl" href="#dom-mloperand-operand-slot" id="ref-for-dom-mloperand-operand-slot④⑥">[[operand]]</a></code> as input to <var>opImpl</var>.</p>
+        <li data-md>
+         <p>Connect <var>output</var>.<code class="idl"><a data-link-type="idl" href="#dom-mloperand-operand-slot" id="ref-for-dom-mloperand-operand-slot④⑦">[[operand]]</a></code> as output to <var>opImpl</var>.</p>
+       </ol>
+      <li data-md>
+       <p>Return <var>output</var>.</p>
+     </ol>
+    </div>
+   </details>
+   <h5 class="heading settled" data-level="7.6.20.2" id="api-mlgraphbuilder-linear-options"><span class="secno">7.6.20.2. </span><span class="content">The <code class="idl"><a data-link-type="idl" href="#dom-mlgraphbuilder-linear-options" id="ref-for-dom-mlgraphbuilder-linear-options①">linear(options)</a></code> method</span><a class="self-link" href="#api-mlgraphbuilder-linear-options"></a></h5>
+   <div>
+     <strong>Arguments:</strong> 
+    <ul>
+     <li data-md>
+      <p><em>options</em>: an optional <code class="idl"><a data-link-type="idl" href="#dictdef-mllinearoptions" id="ref-for-dictdef-mllinearoptions④">MLLinearOptions</a></code>. The optional parameters of the operation.</p>
+    </ul>
+    <p><strong>Returns:</strong></p>
+    <ul>
+     <li data-md>
+      <p>an <code class="idl"><a data-link-type="idl" href="#mlactivation" id="ref-for-mlactivation③⑥">MLActivation</a></code>. The activation function representing the linear operation.</p>
+    </ul>
+   </div>
+   <details class="algorithm" data-algorithm="linear(options)" data-algorithm-for="MLGraphBuilder" open>
+    <summary> The <dfn class="dfn-paneled idl-code" data-dfn-for="MLGraphBuilder" data-dfn-type="method" data-export data-lt="linear(options)|linear()" id="dom-mlgraphbuilder-linear-options"><code>linear(<var>options</var>)</code></dfn> method steps are: </summary>
+    <div class="algorithm-steps">
+     <ol>
+      <li data-md>
+       <p>Let <var>op</var> be the result of <a data-link-type="dfn" href="#create-an-mlactivation" id="ref-for-create-an-mlactivation⑤">creating an MLActivation</a> given <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#this" id="ref-for-this②③">this</a>, <code>"linear"</code> and <var>options</var>.</p>
+       <ol>
+        <li data-md>
+         <p>If that <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#dfn-throw" id="ref-for-dfn-throw①⑥⓪">throws</a> an error, re-<a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#dfn-throw" id="ref-for-dfn-throw①⑥①">throw</a> the error.</p>
+       </ol>
+      <li data-md>
+       <p>Return <var>op</var>.</p>
+     </ol>
     </div>
-    <h4 class="heading settled" data-level="7.6.20" id="api-mlgraphbuilder-lstm"><span class="secno">7.6.20. </span><span class="content">The lstm() method</span><a class="self-link" href="#api-mlgraphbuilder-lstm"></a></h4>
-     Long Short-Term Memory <a data-link-type="biblio" href="#biblio-lstm" title="Long Short-Term Memory">[LSTM]</a> recurrent network uses an input, output, forget, and cell gate to compute the output state that rolls into the output across the temporal sequence of the network. 
+   </details>
+   <h4 class="heading settled" data-level="7.6.21" id="api-mlgraphbuilder-lstm"><span class="secno">7.6.21. </span><span class="content">The lstm() method</span><a class="self-link" href="#api-mlgraphbuilder-lstm"></a></h4>
+    Long Short-Term Memory <a data-link-type="biblio" href="#biblio-lstm" title="Long Short-Term Memory">[LSTM]</a> recurrent network uses an input, output, forget, and cell gate to compute the output state that rolls into the output across the temporal sequence of the network. 
 <pre class="idl highlight def"><c- b>enum</c-> <dfn class="dfn-paneled idl-code" data-dfn-type="enum" data-export id="enumdef-mllstmweightlayout"><code><c- g>MLLstmWeightLayout</c-></code></dfn> {
   <dfn class="idl-code" data-dfn-for="MLLstmWeightLayout" data-dfn-type="enum-value" data-export id="dom-mllstmweightlayout-iofg"><code><c- s>"iofg"</c-></code><a class="self-link" href="#dom-mllstmweightlayout-iofg"></a></dfn>, // input-output-forget-cell gate ordering
   <dfn class="idl-code" data-dfn-for="MLLstmWeightLayout" data-dfn-type="enum-value" data-export id="dom-mllstmweightlayout-ifgo"><code><c- s>"ifgo"</c-></code><a class="self-link" href="#dom-mllstmweightlayout-ifgo"></a></dfn>  // input-forget-cell-output gate ordering
 };
 
 <c- b>dictionary</c-> <dfn class="dfn-paneled idl-code" data-dfn-type="dictionary" data-export id="dictdef-mllstmoptions"><code><c- g>MLLstmOptions</c-></code></dfn> {
-  <a data-link-type="idl-name" href="#mloperand" id="ref-for-mloperand①⑦⑤"><c- n>MLOperand</c-></a> <dfn class="idl-code" data-dfn-for="MLLstmOptions" data-dfn-type="dict-member" data-export data-type="MLOperand " id="dom-mllstmoptions-bias"><code><c- g>bias</c-></code><a class="self-link" href="#dom-mllstmoptions-bias"></a></dfn>;
-  <a data-link-type="idl-name" href="#mloperand" id="ref-for-mloperand①⑦⑥"><c- n>MLOperand</c-></a> <dfn class="idl-code" data-dfn-for="MLLstmOptions" data-dfn-type="dict-member" data-export data-type="MLOperand " id="dom-mllstmoptions-recurrentbias"><code><c- g>recurrentBias</c-></code><a class="self-link" href="#dom-mllstmoptions-recurrentbias"></a></dfn>;
-  <a data-link-type="idl-name" href="#mloperand" id="ref-for-mloperand①⑦⑦"><c- n>MLOperand</c-></a> <dfn class="idl-code" data-dfn-for="MLLstmOptions" data-dfn-type="dict-member" data-export data-type="MLOperand " id="dom-mllstmoptions-peepholeweight"><code><c- g>peepholeWeight</c-></code><a class="self-link" href="#dom-mllstmoptions-peepholeweight"></a></dfn>;
-  <a data-link-type="idl-name" href="#mloperand" id="ref-for-mloperand①⑦⑧"><c- n>MLOperand</c-></a> <dfn class="idl-code" data-dfn-for="MLLstmOptions" data-dfn-type="dict-member" data-export data-type="MLOperand " id="dom-mllstmoptions-initialhiddenstate"><code><c- g>initialHiddenState</c-></code><a class="self-link" href="#dom-mllstmoptions-initialhiddenstate"></a></dfn>;
-  <a data-link-type="idl-name" href="#mloperand" id="ref-for-mloperand①⑦⑨"><c- n>MLOperand</c-></a> <dfn class="idl-code" data-dfn-for="MLLstmOptions" data-dfn-type="dict-member" data-export data-type="MLOperand " id="dom-mllstmoptions-initialcellstate"><code><c- g>initialCellState</c-></code><a class="self-link" href="#dom-mllstmoptions-initialcellstate"></a></dfn>;
-  <a class="idl-code" data-link-type="interface" href="https://webidl.spec.whatwg.org/#idl-boolean" id="ref-for-idl-boolean①⓪"><c- b>boolean</c-></a> <dfn class="idl-code" data-default="false" data-dfn-for="MLLstmOptions" data-dfn-type="dict-member" data-export data-type="boolean " id="dom-mllstmoptions-returnsequence"><code><c- g>returnSequence</c-></code><a class="self-link" href="#dom-mllstmoptions-returnsequence"></a></dfn> = <c- b>false</c->;
-  <a data-link-type="idl-name" href="#enumdef-mlrecurrentnetworkdirection" id="ref-for-enumdef-mlrecurrentnetworkdirection②"><c- n>MLRecurrentNetworkDirection</c-></a> <dfn class="idl-code" data-default="&quot;forward&quot;" data-dfn-for="MLLstmOptions" data-dfn-type="dict-member" data-export data-type="MLRecurrentNetworkDirection " id="dom-mllstmoptions-direction"><code><c- g>direction</c-></code><a class="self-link" href="#dom-mllstmoptions-direction"></a></dfn> = "forward";
-  <a data-link-type="idl-name" href="#enumdef-mllstmweightlayout" id="ref-for-enumdef-mllstmweightlayout"><c- n>MLLstmWeightLayout</c-></a> <dfn class="idl-code" data-default="&quot;iofg&quot;" data-dfn-for="MLLstmOptions" data-dfn-type="dict-member" data-export data-type="MLLstmWeightLayout " id="dom-mllstmoptions-layout"><code><c- g>layout</c-></code><a class="self-link" href="#dom-mllstmoptions-layout"></a></dfn> = "iofg";
-  <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#idl-sequence" id="ref-for-idl-sequence①③"><c- b>sequence</c-></a>&lt;<a data-link-type="idl-name" href="#mlactivation" id="ref-for-mlactivation②④"><c- n>MLActivation</c-></a>> <dfn class="idl-code" data-dfn-for="MLLstmOptions" data-dfn-type="dict-member" data-export data-type="sequence<MLActivation> " id="dom-mllstmoptions-activations"><code><c- g>activations</c-></code><a class="self-link" href="#dom-mllstmoptions-activations"></a></dfn>;
+  <a data-link-type="idl-name" href="#mloperand" id="ref-for-mloperand②①②"><c- n>MLOperand</c-></a> <a class="idl-code" data-link-type="dict-member" data-type="MLOperand " href="#dom-mllstmoptions-bias" id="ref-for-dom-mllstmoptions-bias"><c- g>bias</c-></a>;
+  <a data-link-type="idl-name" href="#mloperand" id="ref-for-mloperand②①③"><c- n>MLOperand</c-></a> <a class="idl-code" data-link-type="dict-member" data-type="MLOperand " href="#dom-mllstmoptions-recurrentbias" id="ref-for-dom-mllstmoptions-recurrentbias"><c- g>recurrentBias</c-></a>;
+  <a data-link-type="idl-name" href="#mloperand" id="ref-for-mloperand②①④"><c- n>MLOperand</c-></a> <a class="idl-code" data-link-type="dict-member" data-type="MLOperand " href="#dom-mllstmoptions-peepholeweight" id="ref-for-dom-mllstmoptions-peepholeweight"><c- g>peepholeWeight</c-></a>;
+  <a data-link-type="idl-name" href="#mloperand" id="ref-for-mloperand②①⑤"><c- n>MLOperand</c-></a> <a class="idl-code" data-link-type="dict-member" data-type="MLOperand " href="#dom-mllstmoptions-initialhiddenstate" id="ref-for-dom-mllstmoptions-initialhiddenstate"><c- g>initialHiddenState</c-></a>;
+  <a data-link-type="idl-name" href="#mloperand" id="ref-for-mloperand②①⑥"><c- n>MLOperand</c-></a> <a class="idl-code" data-link-type="dict-member" data-type="MLOperand " href="#dom-mllstmoptions-initialcellstate" id="ref-for-dom-mllstmoptions-initialcellstate"><c- g>initialCellState</c-></a>;
+  <a class="idl-code" data-link-type="interface" href="https://webidl.spec.whatwg.org/#idl-boolean" id="ref-for-idl-boolean①⑤"><c- b>boolean</c-></a> <a class="idl-code" data-default="false" data-link-type="dict-member" data-type="boolean " href="#dom-mllstmoptions-returnsequence" id="ref-for-dom-mllstmoptions-returnsequence"><c- g>returnSequence</c-></a> = <c- b>false</c->;
+  <a data-link-type="idl-name" href="#enumdef-mlrecurrentnetworkdirection" id="ref-for-enumdef-mlrecurrentnetworkdirection③"><c- n>MLRecurrentNetworkDirection</c-></a> <a class="idl-code" data-default="&quot;forward&quot;" data-link-type="dict-member" data-type="MLRecurrentNetworkDirection " href="#dom-mllstmoptions-direction" id="ref-for-dom-mllstmoptions-direction"><c- g>direction</c-></a> = "forward";
+  <a data-link-type="idl-name" href="#enumdef-mllstmweightlayout" id="ref-for-enumdef-mllstmweightlayout"><c- n>MLLstmWeightLayout</c-></a> <a class="idl-code" data-default="&quot;iofg&quot;" data-link-type="dict-member" data-type="MLLstmWeightLayout " href="#dom-mllstmoptions-layout" id="ref-for-dom-mllstmoptions-layout"><c- g>layout</c-></a> = "iofg";
+  <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#idl-sequence" id="ref-for-idl-sequence①③"><c- b>sequence</c-></a>&lt;<a data-link-type="idl-name" href="#mlactivation" id="ref-for-mlactivation③⑦"><c- n>MLActivation</c-></a>> <a class="idl-code" data-link-type="dict-member" data-type="sequence<MLActivation> " href="#dom-mllstmoptions-activations" id="ref-for-dom-mllstmoptions-activations"><c- g>activations</c-></a>;
 };
 
-<c- b>partial</c-> <c- b>interface</c-> <a class="idl-code" data-link-type="interface" href="#mlgraphbuilder" id="ref-for-mlgraphbuilder③⑥"><c- g>MLGraphBuilder</c-></a> {
-  <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#idl-sequence" id="ref-for-idl-sequence①④"><c- b>sequence</c-></a>&lt;<a data-link-type="idl-name" href="#mloperand" id="ref-for-mloperand①⑧⓪"><c- n>MLOperand</c-></a>> <dfn class="idl-code" data-dfn-for="MLGraphBuilder" data-dfn-type="method" data-export data-lt="lstm(input, weight, recurrentWeight, steps, hiddenSize, options)|lstm(input, weight, recurrentWeight, steps, hiddenSize)" id="dom-mlgraphbuilder-lstm"><code><c- g>lstm</c-></code><a class="self-link" href="#dom-mlgraphbuilder-lstm"></a></dfn>(<a data-link-type="idl-name" href="#mloperand" id="ref-for-mloperand①⑧①"><c- n>MLOperand</c-></a> <dfn class="idl-code" data-dfn-for="MLGraphBuilder/lstm(input, weight, recurrentWeight, steps, hiddenSize, options), MLGraphBuilder/lstm(input, weight, recurrentWeight, steps, hiddenSize)" data-dfn-type="argument" data-export id="dom-mlgraphbuilder-lstm-input-weight-recurrentweight-steps-hiddensize-options-input"><code><c- g>input</c-></code><a class="self-link" href="#dom-mlgraphbuilder-lstm-input-weight-recurrentweight-steps-hiddensize-options-input"></a></dfn>, <a data-link-type="idl-name" href="#mloperand" id="ref-for-mloperand①⑧②"><c- n>MLOperand</c-></a> <dfn class="idl-code" data-dfn-for="MLGraphBuilder/lstm(input, weight, recurrentWeight, steps, hiddenSize, options), MLGraphBuilder/lstm(input, weight, recurrentWeight, steps, hiddenSize)" data-dfn-type="argument" data-export id="dom-mlgraphbuilder-lstm-input-weight-recurrentweight-steps-hiddensize-options-weight"><code><c- g>weight</c-></code><a class="self-link" href="#dom-mlgraphbuilder-lstm-input-weight-recurrentweight-steps-hiddensize-options-weight"></a></dfn>, <a data-link-type="idl-name" href="#mloperand" id="ref-for-mloperand①⑧③"><c- n>MLOperand</c-></a> <dfn class="idl-code" data-dfn-for="MLGraphBuilder/lstm(input, weight, recurrentWeight, steps, hiddenSize, options), MLGraphBuilder/lstm(input, weight, recurrentWeight, steps, hiddenSize)" data-dfn-type="argument" data-export id="dom-mlgraphbuilder-lstm-input-weight-recurrentweight-steps-hiddensize-options-recurrentweight"><code><c- g>recurrentWeight</c-></code><a class="self-link" href="#dom-mlgraphbuilder-lstm-input-weight-recurrentweight-steps-hiddensize-options-recurrentweight"></a></dfn>,
-                           <a class="idl-code" data-link-type="interface" href="https://webidl.spec.whatwg.org/#idl-unsigned-long" id="ref-for-idl-unsigned-long③①"><c- b>unsigned</c-> <c- b>long</c-></a> <dfn class="idl-code" data-dfn-for="MLGraphBuilder/lstm(input, weight, recurrentWeight, steps, hiddenSize, options), MLGraphBuilder/lstm(input, weight, recurrentWeight, steps, hiddenSize)" data-dfn-type="argument" data-export id="dom-mlgraphbuilder-lstm-input-weight-recurrentweight-steps-hiddensize-options-steps"><code><c- g>steps</c-></code><a class="self-link" href="#dom-mlgraphbuilder-lstm-input-weight-recurrentweight-steps-hiddensize-options-steps"></a></dfn>, <a class="idl-code" data-link-type="interface" href="https://webidl.spec.whatwg.org/#idl-unsigned-long" id="ref-for-idl-unsigned-long③②"><c- b>unsigned</c-> <c- b>long</c-></a> <dfn class="idl-code" data-dfn-for="MLGraphBuilder/lstm(input, weight, recurrentWeight, steps, hiddenSize, options), MLGraphBuilder/lstm(input, weight, recurrentWeight, steps, hiddenSize)" data-dfn-type="argument" data-export id="dom-mlgraphbuilder-lstm-input-weight-recurrentweight-steps-hiddensize-options-hiddensize"><code><c- g>hiddenSize</c-></code><a class="self-link" href="#dom-mlgraphbuilder-lstm-input-weight-recurrentweight-steps-hiddensize-options-hiddensize"></a></dfn>,
+<c- b>partial</c-> <c- b>interface</c-> <a class="idl-code" data-link-type="interface" href="#mlgraphbuilder" id="ref-for-mlgraphbuilder③⑨"><c- g>MLGraphBuilder</c-></a> {
+  <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#idl-sequence" id="ref-for-idl-sequence①④"><c- b>sequence</c-></a>&lt;<a data-link-type="idl-name" href="#mloperand" id="ref-for-mloperand②①⑦"><c- n>MLOperand</c-></a>> <a class="idl-code" data-link-type="method" href="#dom-mlgraphbuilder-lstm" id="ref-for-dom-mlgraphbuilder-lstm"><c- g>lstm</c-></a>(<a data-link-type="idl-name" href="#mloperand" id="ref-for-mloperand②①⑧"><c- n>MLOperand</c-></a> <dfn class="idl-code" data-dfn-for="MLGraphBuilder/lstm(input, weight, recurrentWeight, steps, hiddenSize, options), MLGraphBuilder/lstm(input, weight, recurrentWeight, steps, hiddenSize)" data-dfn-type="argument" data-export id="dom-mlgraphbuilder-lstm-input-weight-recurrentweight-steps-hiddensize-options-input"><code><c- g>input</c-></code><a class="self-link" href="#dom-mlgraphbuilder-lstm-input-weight-recurrentweight-steps-hiddensize-options-input"></a></dfn>, <a data-link-type="idl-name" href="#mloperand" id="ref-for-mloperand②①⑨"><c- n>MLOperand</c-></a> <dfn class="idl-code" data-dfn-for="MLGraphBuilder/lstm(input, weight, recurrentWeight, steps, hiddenSize, options), MLGraphBuilder/lstm(input, weight, recurrentWeight, steps, hiddenSize)" data-dfn-type="argument" data-export id="dom-mlgraphbuilder-lstm-input-weight-recurrentweight-steps-hiddensize-options-weight"><code><c- g>weight</c-></code><a class="self-link" href="#dom-mlgraphbuilder-lstm-input-weight-recurrentweight-steps-hiddensize-options-weight"></a></dfn>, <a data-link-type="idl-name" href="#mloperand" id="ref-for-mloperand②②⓪"><c- n>MLOperand</c-></a> <dfn class="idl-code" data-dfn-for="MLGraphBuilder/lstm(input, weight, recurrentWeight, steps, hiddenSize, options), MLGraphBuilder/lstm(input, weight, recurrentWeight, steps, hiddenSize)" data-dfn-type="argument" data-export id="dom-mlgraphbuilder-lstm-input-weight-recurrentweight-steps-hiddensize-options-recurrentweight"><code><c- g>recurrentWeight</c-></code><a class="self-link" href="#dom-mlgraphbuilder-lstm-input-weight-recurrentweight-steps-hiddensize-options-recurrentweight"></a></dfn>,
+                           <a class="idl-code" data-link-type="interface" href="https://webidl.spec.whatwg.org/#idl-unsigned-long" id="ref-for-idl-unsigned-long③③"><c- b>unsigned</c-> <c- b>long</c-></a> <dfn class="idl-code" data-dfn-for="MLGraphBuilder/lstm(input, weight, recurrentWeight, steps, hiddenSize, options), MLGraphBuilder/lstm(input, weight, recurrentWeight, steps, hiddenSize)" data-dfn-type="argument" data-export id="dom-mlgraphbuilder-lstm-input-weight-recurrentweight-steps-hiddensize-options-steps"><code><c- g>steps</c-></code><a class="self-link" href="#dom-mlgraphbuilder-lstm-input-weight-recurrentweight-steps-hiddensize-options-steps"></a></dfn>, <a class="idl-code" data-link-type="interface" href="https://webidl.spec.whatwg.org/#idl-unsigned-long" id="ref-for-idl-unsigned-long③④"><c- b>unsigned</c-> <c- b>long</c-></a> <dfn class="idl-code" data-dfn-for="MLGraphBuilder/lstm(input, weight, recurrentWeight, steps, hiddenSize, options), MLGraphBuilder/lstm(input, weight, recurrentWeight, steps, hiddenSize)" data-dfn-type="argument" data-export id="dom-mlgraphbuilder-lstm-input-weight-recurrentweight-steps-hiddensize-options-hiddensize"><code><c- g>hiddenSize</c-></code><a class="self-link" href="#dom-mlgraphbuilder-lstm-input-weight-recurrentweight-steps-hiddensize-options-hiddensize"></a></dfn>,
                            <c- b>optional</c-> <a data-link-type="idl-name" href="#dictdef-mllstmoptions" id="ref-for-dictdef-mllstmoptions"><c- n>MLLstmOptions</c-></a> <dfn class="idl-code" data-dfn-for="MLGraphBuilder/lstm(input, weight, recurrentWeight, steps, hiddenSize, options), MLGraphBuilder/lstm(input, weight, recurrentWeight, steps, hiddenSize)" data-dfn-type="argument" data-export id="dom-mlgraphbuilder-lstm-input-weight-recurrentweight-steps-hiddensize-options-options"><code><c- g>options</c-></code><a class="self-link" href="#dom-mlgraphbuilder-lstm-input-weight-recurrentweight-steps-hiddensize-options-options"></a></dfn> = {});
 };
 </pre>
-    <div class="algorithm" data-algorithm="lstm">
-      <strong>Arguments:</strong> 
-     <ul>
+   <p><code class="idl"><a data-link-type="idl" href="#dictdef-mllstmoptions" id="ref-for-dictdef-mllstmoptions①">MLLstmOptions</a></code> has the following members:</p>
+   <dl>
+    <dt data-md><dfn class="dfn-paneled idl-code" data-dfn-for="MLLstmOptions" data-dfn-type="dict-member" data-export id="dom-mllstmoptions-bias"><code>bias</code></dfn>, <span> of type <a data-link-type="idl-name" href="#mloperand" id="ref-for-mloperand②②①">MLOperand</a></span>
+    <dd data-md>
+     <p>An <code class="idl"><a data-link-type="idl" href="#mloperand" id="ref-for-mloperand②②②">MLOperand</a></code>. Specifies the 2-D input bias tensor of shape [numDirections, 4 * hiddenSize]. The ordering of the bias vectors in the second dimension of the tensor shape is specified according to <code class="idl"><a data-link-type="idl" href="#dom-mllstmoptions-layout" id="ref-for-dom-mllstmoptions-layout①">layout</a></code>.</p>
+    <dt data-md><dfn class="dfn-paneled idl-code" data-dfn-for="MLLstmOptions" data-dfn-type="dict-member" data-export id="dom-mllstmoptions-recurrentbias"><code>recurrentBias</code></dfn>, <span> of type <a data-link-type="idl-name" href="#mloperand" id="ref-for-mloperand②②③">MLOperand</a></span>
+    <dd data-md>
+     <p>An <code class="idl"><a data-link-type="idl" href="#mloperand" id="ref-for-mloperand②②④">MLOperand</a></code>. Specifies the 2-D recurrent bias tensor of shape [numDirections, 4 * hiddenSize]. The ordering of the bias vectors in the first dimension of the tensor shape is specified according to <code class="idl"><a data-link-type="idl" href="#dom-mllstmoptions-layout" id="ref-for-dom-mllstmoptions-layout②">layout</a></code>.</p>
+    <dt data-md><dfn class="dfn-paneled idl-code" data-dfn-for="MLLstmOptions" data-dfn-type="dict-member" data-export id="dom-mllstmoptions-peepholeweight"><code>peepholeWeight</code></dfn>, <span> of type <a data-link-type="idl-name" href="#mloperand" id="ref-for-mloperand②②⑤">MLOperand</a></span>
+    <dd data-md>
+     <p>An <code class="idl"><a data-link-type="idl" href="#mloperand" id="ref-for-mloperand②②⑥">MLOperand</a></code>. Specifies the 2-D weight tensor for peepholes of shape [numDirections, 3 * hiddenSize]. The pack ordering of the weight vectors is for the <code>input (i)</code>, <code>output (o)</code>, and <code>forget (f)</code> gate, respectively.</p>
+    <dt data-md><dfn class="dfn-paneled idl-code" data-dfn-for="MLLstmOptions" data-dfn-type="dict-member" data-export id="dom-mllstmoptions-initialhiddenstate"><code>initialHiddenState</code></dfn>, <span> of type <a data-link-type="idl-name" href="#mloperand" id="ref-for-mloperand②②⑦">MLOperand</a></span>
+    <dd data-md>
+     <p>An <code class="idl"><a data-link-type="idl" href="#mloperand" id="ref-for-mloperand②②⑧">MLOperand</a></code>. Specifies the 3-D initial hidden state tensor of shape [numDirections, batchSize, hiddenSize]. When not specified, implementations SHOULD use a tensor filled with zero.</p>
+    <dt data-md><dfn class="dfn-paneled idl-code" data-dfn-for="MLLstmOptions" data-dfn-type="dict-member" data-export id="dom-mllstmoptions-initialcellstate"><code>initialCellState</code></dfn>, <span> of type <a data-link-type="idl-name" href="#mloperand" id="ref-for-mloperand②②⑨">MLOperand</a></span>
+    <dd data-md>
+     <p>An <code class="idl"><a data-link-type="idl" href="#mloperand" id="ref-for-mloperand②③⓪">MLOperand</a></code>. Specifies the 3-D initial hidden state tensor of shape [numDirections, batchSize, hiddenSize]. When not specified, implementations SHOULD use a tensor filled with zero.</p>
+    <dt data-md><dfn class="dfn-paneled idl-code" data-dfn-for="MLLstmOptions" data-dfn-type="dict-member" data-export id="dom-mllstmoptions-returnsequence"><code>returnSequence</code></dfn>, <span> of type <a data-link-type="idl-name" href="https://webidl.spec.whatwg.org/#idl-boolean" id="ref-for-idl-boolean①⑥">boolean</a>, defaulting to <code>false</code></span>
+    <dd data-md>
+     <p>A <code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#idl-boolean" id="ref-for-idl-boolean①⑦">boolean</a></code> indicating whether to also return the entire sequence with every output from each time step in it in addition to the output of the last time step.</p>
+    <dt data-md><dfn class="dfn-paneled idl-code" data-dfn-for="MLLstmOptions" data-dfn-type="dict-member" data-export id="dom-mllstmoptions-direction"><code>direction</code></dfn>, <span> of type <a data-link-type="idl-name" href="#enumdef-mlrecurrentnetworkdirection" id="ref-for-enumdef-mlrecurrentnetworkdirection④">MLRecurrentNetworkDirection</a>, defaulting to <code>"forward"</code></span>
+    <dd data-md>
+     <p>An <code class="idl"><a data-link-type="idl" href="#enumdef-mlrecurrentnetworkdirection" id="ref-for-enumdef-mlrecurrentnetworkdirection⑤">MLRecurrentNetworkDirection</a></code>. Specifies the processing direction of the input sequence. When set to <code>"both"</code>, the size of the first dimension of the weight and the bias tensor shapes must be 2, and the input is processed in both directions.</p>
+    <dt data-md><dfn class="dfn-paneled idl-code" data-dfn-for="MLLstmOptions" data-dfn-type="dict-member" data-export id="dom-mllstmoptions-layout"><code>layout</code></dfn>, <span> of type <a data-link-type="idl-name" href="#enumdef-mllstmweightlayout" id="ref-for-enumdef-mllstmweightlayout①">MLLstmWeightLayout</a>, defaulting to <code>"iofg"</code></span>
+    <dd data-md>
+     <p>An <code class="idl"><a data-link-type="idl" href="#enumdef-mllstmweightlayout" id="ref-for-enumdef-mllstmweightlayout②">MLLstmWeightLayout</a></code>. The ordering of the weight and bias vectors for the internal gates of LSTM, specifically the <code>input (i)</code>, <code>output (o)</code>, <code>forget (f)</code>, and <code>cell (g)</code> gate, as indicated in the first dimension of the weight and bias tensor shapes. When not specified, the default layout is <code>"iofg"</code>.</p>
+    <dt data-md><dfn class="dfn-paneled idl-code" data-dfn-for="MLLstmOptions" data-dfn-type="dict-member" data-export id="dom-mllstmoptions-activations"><code>activations</code></dfn>, <span> of type sequence&lt;<a data-link-type="idl-name" href="#mlactivation" id="ref-for-mlactivation③⑧">MLActivation</a>></span>
+    <dd data-md>
+     <p>A sequence of <code class="idl"><a data-link-type="idl" href="#mlactivation" id="ref-for-mlactivation③⑨">MLActivation</a></code>. A sequence of three activation functions, the first one is used for the <code>input (i)</code>, <code>forget (f)</code>, and <code>output (o)</code> gate, the second one is used for the <code>cell (g)</code> gate, and the last used for filtering the output cell state before combining it with the result of the output gate to form the output hidden state. When not specified, implementations SHOULD use the sequence of the sigmoid function (<code>"sigmoid"</code>) followed by two hyperbolic tangent functions (<code>"tanh"</code>) respectively.</p>
+   </dl>
+   <div>
+     <strong>Arguments:</strong> 
+    <ul>
+     <li data-md>
+      <p><em>input</em>: an <code class="idl"><a data-link-type="idl" href="#mloperand" id="ref-for-mloperand②③①">MLOperand</a></code>. The input 3-D tensor of shape [steps, batchSize, inputSize].</p>
+     <li data-md>
+      <p><em>weight</em>: an <code class="idl"><a data-link-type="idl" href="#mloperand" id="ref-for-mloperand②③②">MLOperand</a></code>. The 3-D input weight tensor of shape [numDirections, 4 * hiddenSize, inputSize]. The ordering of the weight vectors in the second dimension of the tensor shape is specified according to the <var>options</var>.<code class="idl"><a data-link-type="idl" href="#dom-mllstmoptions-layout" id="ref-for-dom-mllstmoptions-layout③">layout</a></code>.</p>
+     <li data-md>
+      <p><em>recurrentWeight</em>: an <code class="idl"><a data-link-type="idl" href="#mloperand" id="ref-for-mloperand②③③">MLOperand</a></code>. The 3-D recurrent weight tensor of shape [numDirections, 4 * hiddenSize, hiddenSize]. The ordering of the weight vectors in the second dimension of the tensor shape is specified according to the <var>options</var>.<code class="idl"><a data-link-type="idl" href="#dom-mllstmoptions-layout" id="ref-for-dom-mllstmoptions-layout④">layout</a></code> argument.</p>
+     <li data-md>
+      <p><em>steps</em>: an <code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#idl-unsigned-long" id="ref-for-idl-unsigned-long③⑤">unsigned long</a></code> scalar. The number of time steps in the recurrent network. The value must be greater than 0.</p>
+     <li data-md>
+      <p><em>hiddenSize</em>: an <code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#idl-unsigned-long" id="ref-for-idl-unsigned-long③⑥">unsigned long</a></code> scalar. The value of the third dimension of the cell output tensor shape. It indicates the number of features in the hidden state.</p>
+     <li data-md>
+      <p><em>options</em>: an optional <code class="idl"><a data-link-type="idl" href="#dictdef-mllstmoptions" id="ref-for-dictdef-mllstmoptions②">MLLstmOptions</a></code>. The optional parameters of the operation.</p>
+    </ul>
+    <p><strong>Returns:</strong> a sequence of <code class="idl"><a data-link-type="idl" href="#mloperand" id="ref-for-mloperand②③④">MLOperand</a></code>. The first element of the sequence is a 3-D tensor of shape [numDirections, batchSize, hiddenSize], the output hidden state from the last time step of the network. The second element is a 3-D tensor of shape [numDirections, batchSize, hiddenSize], the output cell state from the last time step of the network. Additionally, if <var>options</var>.<code class="idl"><a data-link-type="idl" href="#dom-mllstmoptions-returnsequence" id="ref-for-dom-mllstmoptions-returnsequence①">returnSequence</a></code> is set to true, the third element is the 4-D output tensor of shape [steps, numDirections, batchSize, hiddenSize] containing every output from each time step in the temporal sequence.</p>
+   </div>
+   <details class="algorithm" data-algorithm="lstm(input, weight, recurrentWeight, steps, hiddenSize, options)" data-algorithm-for="MLGraphBuilder" open>
+    <summary> The <dfn class="dfn-paneled idl-code" data-dfn-for="MLGraphBuilder" data-dfn-type="method" data-export data-lt="lstm(input, weight, recurrentWeight, steps, hiddenSize, options)|lstm(input, weight, recurrentWeight, steps, hiddenSize)" id="dom-mlgraphbuilder-lstm"><code>lstm(<var>input</var>, <var>weight</var>, <var>recurrentWeight</var>, <var>steps</var>, <var>hiddenSize</var>, <var>options</var>)</code></dfn> method steps are: </summary>
+    <div class="algorithm-steps">
+     <ol>
       <li data-md>
-       <p><em>input</em>: an <code class="idl"><a data-link-type="idl" href="#mloperand" id="ref-for-mloperand①⑧④">MLOperand</a></code>. The input 3-D tensor of shape [steps, batch_size, input_size].</p>
+       <p>Let <var>numDirections</var> be 1 if <var>options</var>.<code class="idl"><a data-link-type="idl" href="#dom-mllstmoptions-direction" id="ref-for-dom-mllstmoptions-direction①">direction</a></code> is <code>"forward"</code>, or otherwise let it be 2.</p>
       <li data-md>
-       <p><em>weight</em>: an <code class="idl"><a data-link-type="idl" href="#mloperand" id="ref-for-mloperand①⑧⑤">MLOperand</a></code>. The 3-D input weight tensor of shape [num_directions, 4 * hidden_size, input_size]. The ordering of the weight vectors in the second dimension of the tensor shape is specified according to the <em>options.layout</em> argument.</p>
+       <p><a data-link-type="dfn" href="https://infra.spec.whatwg.org/#assert" id="ref-for-assert④④">Assert</a>: the type of <var>input</var>, <var>weight</var> and <var>recurrentWeight</var> is <code class="idl"><a data-link-type="idl" href="#mloperand" id="ref-for-mloperand②③⑤">MLOperand</a></code>.</p>
+       <div class="note" role="note"> The shape of <var>input</var>, <var>weight</var> or <var>recurrentWeight</var> could be also checked here. </div>
       <li data-md>
-       <p><em>recurrentWeight</em>: an <code class="idl"><a data-link-type="idl" href="#mloperand" id="ref-for-mloperand①⑧⑥">MLOperand</a></code>. The 3-D recurrent weight tensor of shape [num_directions, 4 * hidden_size, hidden_size]. The ordering of the weight vectors in the second dimension of the tensor shape is specified according to the <em>options.layout</em> argument.</p>
+       <p>If <var>input</var>.<code class="idl"><a data-link-type="idl" href="#dom-mloperand-descriptor-slot" id="ref-for-dom-mloperand-descriptor-slot⑤⑤">[[descriptor]]</a></code>.<code class="idl"><a data-link-type="idl" href="#dom-mloperanddescriptor-dimensions" id="ref-for-dom-mloperanddescriptor-dimensions⑤③">dimensions</a></code>[0] is not equal to <var>steps</var>, then <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#dfn-throw" id="ref-for-dfn-throw①⑥②">throw</a> a "<code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#dataerror" id="ref-for-dataerror④⑦">DataError</a></code>" <code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#idl-DOMException" id="ref-for-idl-DOMException⑧⑥">DOMException</a></code>.</p>
       <li data-md>
-       <p><em>steps</em>: an <code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#idl-unsigned-long" id="ref-for-idl-unsigned-long③③">unsigned long</a></code> scalar. The number of time steps in the recurrent network. The value must be greater than 0.</p>
+       <p>Let <var>batchSize</var> be <var>input</var>.<code class="idl"><a data-link-type="idl" href="#dom-mloperand-descriptor-slot" id="ref-for-dom-mloperand-descriptor-slot⑤⑥">[[descriptor]]</a></code>.<code class="idl"><a data-link-type="idl" href="#dom-mloperanddescriptor-dimensions" id="ref-for-dom-mloperanddescriptor-dimensions⑤④">dimensions</a></code>[1].</p>
       <li data-md>
-       <p><em>hiddenSize</em>: an <code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#idl-unsigned-long" id="ref-for-idl-unsigned-long③④">unsigned long</a></code> scalar. The value of the third dimension of the cell output tensor shape. It indicates the number of features in the hidden state.</p>
+       <p>If <var>options</var>.<code class="idl"><a data-link-type="idl" href="#dom-mllstmoptions-bias" id="ref-for-dom-mllstmoptions-bias①">bias</a></code> <a data-link-type="dfn" href="https://infra.spec.whatwg.org/#map-exists" id="ref-for-map-exists③⑤">exists</a>:</p>
+       <ol>
+        <li data-md>
+         <p><a data-link-type="dfn" href="https://infra.spec.whatwg.org/#assert" id="ref-for-assert④⑤">Assert</a>: its type is <code class="idl"><a data-link-type="idl" href="#mloperand" id="ref-for-mloperand②③⑥">MLOperand</a></code>.</p>
+        <li data-md>
+         <p>If its rank is not 2, then <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#dfn-throw" id="ref-for-dfn-throw①⑥③">throw</a> a "<code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#dataerror" id="ref-for-dataerror④⑧">DataError</a></code>" <code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#idl-DOMException" id="ref-for-idl-DOMException⑧⑦">DOMException</a></code>.</p>
+        <li data-md>
+         <p>If <var>options</var>.<code class="idl"><a data-link-type="idl" href="#dom-mllstmoptions-bias" id="ref-for-dom-mllstmoptions-bias②">bias</a></code>.<code class="idl"><a data-link-type="idl" href="#dom-mloperand-descriptor-slot" id="ref-for-dom-mloperand-descriptor-slot⑤⑦">[[descriptor]]</a></code>.<code class="idl"><a data-link-type="idl" href="#dom-mloperanddescriptor-dimensions" id="ref-for-dom-mloperanddescriptor-dimensions⑤⑤">dimensions</a></code>[0] is not <var>numDirections</var>, then <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#dfn-throw" id="ref-for-dfn-throw①⑥④">throw</a> a "<code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#dataerror" id="ref-for-dataerror④⑨">DataError</a></code>" <code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#idl-DOMException" id="ref-for-idl-DOMException⑧⑧">DOMException</a></code>.</p>
+        <li data-md>
+         <p>If <var>options</var>.<code class="idl"><a data-link-type="idl" href="#dom-mllstmoptions-bias" id="ref-for-dom-mllstmoptions-bias③">bias</a></code>.<code class="idl"><a data-link-type="idl" href="#dom-mloperand-descriptor-slot" id="ref-for-dom-mloperand-descriptor-slot⑤⑧">[[descriptor]]</a></code>.<code class="idl"><a data-link-type="idl" href="#dom-mloperanddescriptor-dimensions" id="ref-for-dom-mloperanddescriptor-dimensions⑤⑥">dimensions</a></code>[1] is not 4 * <var>hiddenSize</var>, then <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#dfn-throw" id="ref-for-dfn-throw①⑥⑤">throw</a> a "<code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#dataerror" id="ref-for-dataerror⑤⓪">DataError</a></code>" <code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#idl-DOMException" id="ref-for-idl-DOMException⑧⑨">DOMException</a></code>.</p>
+       </ol>
       <li data-md>
-       <p><em>options</em>: an optional <code class="idl"><a data-link-type="idl" href="#dictdef-mlgruoptions" id="ref-for-dictdef-mlgruoptions②">MLGruOptions</a></code>. The optional parameters of the operation.</p>
-       <ul>
+       <p>If <var>options</var>.<code class="idl"><a data-link-type="idl" href="#dom-mllstmoptions-recurrentbias" id="ref-for-dom-mllstmoptions-recurrentbias①">recurrentBias</a></code> <a data-link-type="dfn" href="https://infra.spec.whatwg.org/#map-exists" id="ref-for-map-exists③⑥">exists</a>:</p>
+       <ol>
+        <li data-md>
+         <p><a data-link-type="dfn" href="https://infra.spec.whatwg.org/#assert" id="ref-for-assert④⑥">Assert</a>: its type is <code class="idl"><a data-link-type="idl" href="#mloperand" id="ref-for-mloperand②③⑦">MLOperand</a></code>.</p>
+        <li data-md>
+         <p>If its rank is not 2, then <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#dfn-throw" id="ref-for-dfn-throw①⑥⑥">throw</a> a "<code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#dataerror" id="ref-for-dataerror⑤①">DataError</a></code>" <code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#idl-DOMException" id="ref-for-idl-DOMException⑨⓪">DOMException</a></code>.</p>
+        <li data-md>
+         <p>If <var>options</var>.<code class="idl"><a data-link-type="idl" href="#dom-mllstmoptions-recurrentbias" id="ref-for-dom-mllstmoptions-recurrentbias②">recurrentBias</a></code>.<code class="idl"><a data-link-type="idl" href="#dom-mloperand-descriptor-slot" id="ref-for-dom-mloperand-descriptor-slot⑤⑨">[[descriptor]]</a></code>.<code class="idl"><a data-link-type="idl" href="#dom-mloperanddescriptor-dimensions" id="ref-for-dom-mloperanddescriptor-dimensions⑤⑦">dimensions</a></code>[0] is not <var>numDirections</var>, then <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#dfn-throw" id="ref-for-dfn-throw①⑥⑦">throw</a> a "<code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#dataerror" id="ref-for-dataerror⑤②">DataError</a></code>" <code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#idl-DOMException" id="ref-for-idl-DOMException⑨①">DOMException</a></code>.</p>
+        <li data-md>
+         <p>If <var>options</var>.<code class="idl"><a data-link-type="idl" href="#dom-mllstmoptions-recurrentbias" id="ref-for-dom-mllstmoptions-recurrentbias③">recurrentBias</a></code>.<code class="idl"><a data-link-type="idl" href="#dom-mloperand-descriptor-slot" id="ref-for-dom-mloperand-descriptor-slot⑥⓪">[[descriptor]]</a></code>.<code class="idl"><a data-link-type="idl" href="#dom-mloperanddescriptor-dimensions" id="ref-for-dom-mloperanddescriptor-dimensions⑤⑧">dimensions</a></code>[1] is not 4 * <var>hiddenSize</var>, then <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#dfn-throw" id="ref-for-dfn-throw①⑥⑧">throw</a> a "<code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#dataerror" id="ref-for-dataerror⑤③">DataError</a></code>" <code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#idl-DOMException" id="ref-for-idl-DOMException⑨②">DOMException</a></code>.</p>
+       </ol>
+      <li data-md>
+       <p>If <var>options</var>.<code class="idl"><a data-link-type="idl" href="#dom-mllstmoptions-peepholeweight" id="ref-for-dom-mllstmoptions-peepholeweight①">peepholeWeight</a></code> <a data-link-type="dfn" href="https://infra.spec.whatwg.org/#map-exists" id="ref-for-map-exists③⑦">exists</a>:</p>
+       <ol>
         <li data-md>
-         <p><em>bias</em>: an <code class="idl"><a data-link-type="idl" href="#mloperand" id="ref-for-mloperand①⑧⑦">MLOperand</a></code>. The 2-D input bias tensor of shape [num_directions, 4 * hidden_size]. The ordering of the bias vectors in the second dimension of the tensor shape is specified according to the <em>options.layout</em> argument.</p>
+         <p><a data-link-type="dfn" href="https://infra.spec.whatwg.org/#assert" id="ref-for-assert④⑦">Assert</a>: its type is <code class="idl"><a data-link-type="idl" href="#mloperand" id="ref-for-mloperand②③⑧">MLOperand</a></code>.</p>
         <li data-md>
-         <p><em>recurrentBias</em>: an <code class="idl"><a data-link-type="idl" href="#mloperand" id="ref-for-mloperand①⑧⑧">MLOperand</a></code>. The 2-D recurrent bias tensor of shape [num_directions, 4 * hidden_size]. The ordering of the bias vectors in the second dimension of the tensor shape is specified according to the <em>options.layout</em> argument.</p>
+         <p>If its rank is not 2, then <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#dfn-throw" id="ref-for-dfn-throw①⑥⑨">throw</a> a "<code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#dataerror" id="ref-for-dataerror⑤④">DataError</a></code>" <code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#idl-DOMException" id="ref-for-idl-DOMException⑨③">DOMException</a></code>.</p>
         <li data-md>
-         <p><em>peepholeWeight</em>: an <code class="idl"><a data-link-type="idl" href="#mloperand" id="ref-for-mloperand①⑧⑨">MLOperand</a></code>. The 2-D weight tensor for peepholes of shape [num_directions, 3 * hidden_size]. The pack ordering of the weight vectors is for the <em>input (i)</em>, <em>output (o)</em>, and <em>forget (f)</em> gate respectively.</p>
+         <p>If <var>options</var>.<code class="idl"><a data-link-type="idl" href="#dom-mllstmoptions-peepholeweight" id="ref-for-dom-mllstmoptions-peepholeweight②">peepholeWeight</a></code>.<code class="idl"><a data-link-type="idl" href="#dom-mloperand-descriptor-slot" id="ref-for-dom-mloperand-descriptor-slot⑥①">[[descriptor]]</a></code>.<code class="idl"><a data-link-type="idl" href="#dom-mloperanddescriptor-dimensions" id="ref-for-dom-mloperanddescriptor-dimensions⑤⑨">dimensions</a></code>[0] is not <var>numDirections</var>, then <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#dfn-throw" id="ref-for-dfn-throw①⑦⓪">throw</a> a "<code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#dataerror" id="ref-for-dataerror⑤⑤">DataError</a></code>" <code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#idl-DOMException" id="ref-for-idl-DOMException⑨④">DOMException</a></code>.</p>
         <li data-md>
-         <p><em>initialHiddenState</em>: an <code class="idl"><a data-link-type="idl" href="#mloperand" id="ref-for-mloperand①⑨⓪">MLOperand</a></code>. The 3-D initial hidden state tensor of shape [num_directions, batch_size, hidden_size]. When not specified, it’s assumed to be a tensor filled with zero.</p>
+         <p>If <var>options</var>.<code class="idl"><a data-link-type="idl" href="#dom-mllstmoptions-peepholeweight" id="ref-for-dom-mllstmoptions-peepholeweight③">peepholeWeight</a></code>.<code class="idl"><a data-link-type="idl" href="#dom-mloperand-descriptor-slot" id="ref-for-dom-mloperand-descriptor-slot⑥②">[[descriptor]]</a></code>.<code class="idl"><a data-link-type="idl" href="#dom-mloperanddescriptor-dimensions" id="ref-for-dom-mloperanddescriptor-dimensions⑥⓪">dimensions</a></code>[1] is not 4 * <var>hiddenSize</var>, then <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#dfn-throw" id="ref-for-dfn-throw①⑦①">throw</a> a "<code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#dataerror" id="ref-for-dataerror⑤⑥">DataError</a></code>" <code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#idl-DOMException" id="ref-for-idl-DOMException⑨⑤">DOMException</a></code>.</p>
+       </ol>
+      <li data-md>
+       <p>If <var>options</var>.<code class="idl"><a data-link-type="idl" href="#dom-mllstmoptions-initialhiddenstate" id="ref-for-dom-mllstmoptions-initialhiddenstate①">initialHiddenState</a></code> <a data-link-type="dfn" href="https://infra.spec.whatwg.org/#map-exists" id="ref-for-map-exists③⑧">exists</a>:</p>
+       <ol>
         <li data-md>
-         <p><em>initialCellState</em>: an <code class="idl"><a data-link-type="idl" href="#mloperand" id="ref-for-mloperand①⑨①">MLOperand</a></code>. The 3-D initial hidden state tensor of shape [num_directions, batch_size, hidden_size]. When not specified, it’s assumed to be a tensor filled with zero.</p>
+         <p><a data-link-type="dfn" href="https://infra.spec.whatwg.org/#assert" id="ref-for-assert④⑧">Assert</a>: its type is <code class="idl"><a data-link-type="idl" href="#mloperand" id="ref-for-mloperand②③⑨">MLOperand</a></code>.</p>
         <li data-md>
-         <p><em>returnSequence</em>: a <code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#idl-boolean" id="ref-for-idl-boolean①①">boolean</a></code> indicating whether to also return the entire sequence with every output from each time step in it in addition to the output of the last time step. Default to false.</p>
+         <p>If its rank is not 3, then <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#dfn-throw" id="ref-for-dfn-throw①⑦②">throw</a> a "<code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#dataerror" id="ref-for-dataerror⑤⑦">DataError</a></code>" <code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#idl-DOMException" id="ref-for-idl-DOMException⑨⑥">DOMException</a></code>.</p>
         <li data-md>
-         <p><em>direction</em>: an <code class="idl"><a data-link-type="idl" href="#enumdef-mlrecurrentnetworkdirection" id="ref-for-enumdef-mlrecurrentnetworkdirection③">MLRecurrentNetworkDirection</a></code>. The processing direction of the input sequence. When set to <em>"both"</em>, the size of the first dimension of the weight and the bias tensor shapes must be 2, and the input is processed in both directions.</p>
+         <p>If <var>options</var>.<code class="idl"><a data-link-type="idl" href="#dom-mllstmoptions-initialhiddenstate" id="ref-for-dom-mllstmoptions-initialhiddenstate②">initialHiddenState</a></code>.<code class="idl"><a data-link-type="idl" href="#dom-mloperand-descriptor-slot" id="ref-for-dom-mloperand-descriptor-slot⑥③">[[descriptor]]</a></code>.<code class="idl"><a data-link-type="idl" href="#dom-mloperanddescriptor-dimensions" id="ref-for-dom-mloperanddescriptor-dimensions⑥①">dimensions</a></code>[0] is not <var>numDirections</var>, then <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#dfn-throw" id="ref-for-dfn-throw①⑦③">throw</a> a "<code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#dataerror" id="ref-for-dataerror⑤⑧">DataError</a></code>" <code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#idl-DOMException" id="ref-for-idl-DOMException⑨⑦">DOMException</a></code>.</p>
         <li data-md>
-         <p><em>layout</em>: an <code class="idl"><a data-link-type="idl" href="#enumdef-mllstmweightlayout" id="ref-for-enumdef-mllstmweightlayout①">MLLstmWeightLayout</a></code>. The ordering of the weight and bias vectors for the internal gates of LSTM, specifically the <em>input (i)</em>, <em>output (o)</em>, <em>forget (f)</em>, and <em>cell (g)</em> gate, as indicated in the second dimension of the weight and bias tensor shapes. When not specified, the default layout is <em>"iofg"</em>.</p>
+         <p>If <var>options</var>.<code class="idl"><a data-link-type="idl" href="#dom-mllstmoptions-initialhiddenstate" id="ref-for-dom-mllstmoptions-initialhiddenstate③">initialHiddenState</a></code>.<code class="idl"><a data-link-type="idl" href="#dom-mloperand-descriptor-slot" id="ref-for-dom-mloperand-descriptor-slot⑥④">[[descriptor]]</a></code>.<code class="idl"><a data-link-type="idl" href="#dom-mloperanddescriptor-dimensions" id="ref-for-dom-mloperanddescriptor-dimensions⑥②">dimensions</a></code>[1] is not equal to <var>batchSize</var>, then <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#dfn-throw" id="ref-for-dfn-throw①⑦④">throw</a> a "<code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#dataerror" id="ref-for-dataerror⑤⑨">DataError</a></code>" <code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#idl-DOMException" id="ref-for-idl-DOMException⑨⑧">DOMException</a></code>.</p>
         <li data-md>
-         <p><em>activations</em>: a sequence of <code class="idl"><a data-link-type="idl" href="#mlactivation" id="ref-for-mlactivation②⑤">MLActivation</a></code>. A sequence of three activation functions, the first one is used for the <em>input (i)</em>, <em>forget (f)</em>, and <em>output (o)</em> gate, the second one is used for the <em>cell (g)</em> gate, and the last used for filtering the output cell state before combining it with the result of the output gate to form the output hidden state. When not specified, they are assumed to be of the sigmoid function (<em>"sigmoid"</em>) followed by two hyperbolic tangent functions (<em>"tanh"</em>) respectively.</p>
-       </ul>
-     </ul>
-     <p><strong>Returns:</strong> a sequence of <code class="idl"><a data-link-type="idl" href="#mloperand" id="ref-for-mloperand①⑨②">MLOperand</a></code>. The first element of the sequence is a 3-D tensor of shape [num_directions, batch_size, hidden_size], the output hidden state from the last time step of the network. The second element is a 3-D tensor of shape [num_directions, batch_size, hidden_size], the output cell state from the last time step of the network. Additionally, if <em>options.returnSequence</em> is set to true, the third element is the 4-D output tensor of shape [steps, num_directions, batch_size, hidden_size] containing every output from each time step in the temporal sequence.</p>
-     <div class="note" role="note">
-       The behavior of this operation can be generically emulated from the usage of other operations as follow. However, user agents typically have a more efficient implementation for it, therefore its usage is encouraged from the performance standpoint. 
+         <p>If <var>options</var>.<code class="idl"><a data-link-type="idl" href="#dom-mllstmoptions-initialhiddenstate" id="ref-for-dom-mllstmoptions-initialhiddenstate④">initialHiddenState</a></code>.<code class="idl"><a data-link-type="idl" href="#dom-mloperand-descriptor-slot" id="ref-for-dom-mloperand-descriptor-slot⑥⑤">[[descriptor]]</a></code>.<code class="idl"><a data-link-type="idl" href="#dom-mloperanddescriptor-dimensions" id="ref-for-dom-mloperanddescriptor-dimensions⑥③">dimensions</a></code>[2] is not <var>hiddenSize</var>, then <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#dfn-throw" id="ref-for-dfn-throw①⑦⑤">throw</a> a "<code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#dataerror" id="ref-for-dataerror⑥⓪">DataError</a></code>" <code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#idl-DOMException" id="ref-for-idl-DOMException⑨⑨">DOMException</a></code>.</p>
+       </ol>
+      <li data-md>
+       <p>If <var>options</var>.<code class="idl"><a data-link-type="idl" href="#dom-mllstmoptions-initialcellstate" id="ref-for-dom-mllstmoptions-initialcellstate①">initialCellState</a></code> <a data-link-type="dfn" href="https://infra.spec.whatwg.org/#map-exists" id="ref-for-map-exists③⑨">exists</a>:</p>
+       <ol>
+        <li data-md>
+         <p><a data-link-type="dfn" href="https://infra.spec.whatwg.org/#assert" id="ref-for-assert④⑨">Assert</a>: its type is <code class="idl"><a data-link-type="idl" href="#mloperand" id="ref-for-mloperand②④⓪">MLOperand</a></code>.</p>
+        <li data-md>
+         <p>If its rank is not 3, then <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#dfn-throw" id="ref-for-dfn-throw①⑦⑥">throw</a> a "<code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#dataerror" id="ref-for-dataerror⑥①">DataError</a></code>" <code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#idl-DOMException" id="ref-for-idl-DOMException①⓪⓪">DOMException</a></code>.</p>
+        <li data-md>
+         <p>If <var>options</var>.<code class="idl"><a data-link-type="idl" href="#dom-mllstmoptions-initialcellstate" id="ref-for-dom-mllstmoptions-initialcellstate②">initialCellState</a></code>.<code class="idl"><a data-link-type="idl" href="#dom-mloperand-descriptor-slot" id="ref-for-dom-mloperand-descriptor-slot⑥⑥">[[descriptor]]</a></code>.<code class="idl"><a data-link-type="idl" href="#dom-mloperanddescriptor-dimensions" id="ref-for-dom-mloperanddescriptor-dimensions⑥④">dimensions</a></code>[0] is not <var>numDirections</var>, then <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#dfn-throw" id="ref-for-dfn-throw①⑦⑦">throw</a> a "<code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#dataerror" id="ref-for-dataerror⑥②">DataError</a></code>" <code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#idl-DOMException" id="ref-for-idl-DOMException①⓪①">DOMException</a></code>.</p>
+        <li data-md>
+         <p>If <var>options</var>.<code class="idl"><a data-link-type="idl" href="#dom-mllstmoptions-initialcellstate" id="ref-for-dom-mllstmoptions-initialcellstate③">initialCellState</a></code>.<code class="idl"><a data-link-type="idl" href="#dom-mloperand-descriptor-slot" id="ref-for-dom-mloperand-descriptor-slot⑥⑦">[[descriptor]]</a></code>.<code class="idl"><a data-link-type="idl" href="#dom-mloperanddescriptor-dimensions" id="ref-for-dom-mloperanddescriptor-dimensions⑥⑤">dimensions</a></code>[1] is not equal to <var>batchSize</var>, then <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#dfn-throw" id="ref-for-dfn-throw①⑦⑧">throw</a> a "<code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#dataerror" id="ref-for-dataerror⑥③">DataError</a></code>" <code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#idl-DOMException" id="ref-for-idl-DOMException①⓪②">DOMException</a></code>.</p>
+        <li data-md>
+         <p>If <var>options</var>.<code class="idl"><a data-link-type="idl" href="#dom-mllstmoptions-initialcellstate" id="ref-for-dom-mllstmoptions-initialcellstate④">initialCellState</a></code>.<code class="idl"><a data-link-type="idl" href="#dom-mloperand-descriptor-slot" id="ref-for-dom-mloperand-descriptor-slot⑥⑧">[[descriptor]]</a></code>.<code class="idl"><a data-link-type="idl" href="#dom-mloperanddescriptor-dimensions" id="ref-for-dom-mloperanddescriptor-dimensions⑥⑥">dimensions</a></code>[2] is not <var>hiddenSize</var>, then <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#dfn-throw" id="ref-for-dfn-throw①⑦⑨">throw</a> a "<code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#dataerror" id="ref-for-dataerror⑥④">DataError</a></code>" <code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#idl-DOMException" id="ref-for-idl-DOMException①⓪③">DOMException</a></code>.</p>
+       </ol>
+      <li data-md>
+       <p>If <var>options</var>.<code class="idl"><a data-link-type="idl" href="#dom-mllstmoptions-activations" id="ref-for-dom-mllstmoptions-activations①">activations</a></code> <a data-link-type="dfn" href="https://infra.spec.whatwg.org/#map-exists" id="ref-for-map-exists④⓪">exists</a>:</p>
+       <ol>
+        <li data-md>
+         <p>If its <a data-link-type="dfn" href="https://infra.spec.whatwg.org/#list-size" id="ref-for-list-size③④">size</a> is not 3, then <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#dfn-throw" id="ref-for-dfn-throw①⑧⓪">throw</a> a <code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#exceptiondef-typeerror" id="ref-for-exceptiondef-typeerror③⑥">TypeError</a></code>.</p>
+        <li data-md>
+         <p><a data-link-type="dfn" href="https://infra.spec.whatwg.org/#assert" id="ref-for-assert⑤⓪">Assert</a>: the type of its elements is <code class="idl"><a data-link-type="idl" href="#mlactivation" id="ref-for-mlactivation④⓪">MLActivation</a></code>.</p>
+       </ol>
+      <li data-md>
+       <p>If any of the following sub-steps fail, <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#dfn-throw" id="ref-for-dfn-throw①⑧①">throw</a> an "<code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#operationerror" id="ref-for-operationerror③⓪">OperationError</a></code>" <code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#idl-DOMException" id="ref-for-idl-DOMException①⓪④">DOMException</a></code>.</p>
+       <ol>
+        <li data-md>
+         <p>Let <var>desc</var> a new <code class="idl"><a data-link-type="idl" href="#dictdef-mloperanddescriptor" id="ref-for-dictdef-mloperanddescriptor①⑨">MLOperandDescriptor</a></code>.</p>
+        <li data-md>
+         <p>Set <var>desc</var>.<code class="idl"><a data-link-type="idl" href="#dom-mloperanddescriptor-dimensions" id="ref-for-dom-mloperanddescriptor-dimensions⑥⑦">dimensions</a></code> to [ <var>numDirections</var>, <var>batchSize</var>, <var>hiddenSize</var> ].</p>
+        <li data-md>
+         <p>Set <var>desc</var>.<code class="idl"><a data-link-type="idl" href="#dom-mloperanddescriptor-type" id="ref-for-dom-mloperanddescriptor-type②⑨">type</a></code> to <var>input</var>.<code class="idl"><a data-link-type="idl" href="#dom-mloperand-descriptor-slot" id="ref-for-dom-mloperand-descriptor-slot⑥⑨">[[descriptor]]</a></code>.<code class="idl"><a data-link-type="idl" href="#dom-mloperanddescriptor-type" id="ref-for-dom-mloperanddescriptor-type③⓪">type</a></code>.</p>
+        <li data-md>
+         <p>Let <var>output0</var> be the result of <a data-link-type="dfn" href="#create-an-mloperand" id="ref-for-create-an-mloperand①⓪">creating an MLOperand</a> given <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#this" id="ref-for-this②④">this</a> and <var>desc</var>.</p>
+        <li data-md>
+         <p>Let <var>output1</var> be the result of <a data-link-type="dfn" href="#create-an-mloperand" id="ref-for-create-an-mloperand①①">creating an MLOperand</a> given <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#this" id="ref-for-this②⑤">this</a> and <var>desc</var>.</p>
+        <li data-md>
+         <p>Set <var>desc</var>.<code class="idl"><a data-link-type="idl" href="#dom-mloperanddescriptor-dimensions" id="ref-for-dom-mloperanddescriptor-dimensions⑥⑧">dimensions</a></code> to [ <var>steps</var>, <var>numDirections</var>, <var>batchSize</var>, <var>hiddenSize</var> ].</p>
+        <li data-md>
+         <p>If <var>options</var>.<code class="idl"><a data-link-type="idl" href="#dom-mllstmoptions-returnsequence" id="ref-for-dom-mllstmoptions-returnsequence②">returnSequence</a></code> is set to true:</p>
+         <ol>
+          <li data-md>
+           <p>Let <var>output2</var> be the result of <a data-link-type="dfn" href="#create-an-mloperand" id="ref-for-create-an-mloperand①②">creating an MLOperand</a> given <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#this" id="ref-for-this②⑥">this</a> and <var>desc</var>.</p>
+          <li data-md>
+           <p>Let <var>output</var> be the array [ <var>output0</var>, <var>output1</var>, <var>output2</var> ].</p>
+         </ol>
+        <li data-md>
+         <p>Otherwise, Let <var>output</var> be the array [ <var>output0</var>, <var>output1</var> ].</p>
+        <li data-md>
+         <p>Make a request to the underlying platform to:</p>
+         <ol>
+          <li data-md>
+           <p>Let <var>opImpl</var> be an <a data-link-type="dfn" href="https://infra.spec.whatwg.org/#implementation-defined" id="ref-for-implementation-defined④⓪">implementation-defined</a> platform operator for the LSTM operation, given <var>weight</var>, <var>recurrentWeight</var>, <var>steps</var>, <var>hiddenSize</var> and <var>options</var>.</p>
+          <li data-md>
+           <p>Store a reference of <var>opImpl</var> in <var>output0</var>.<code class="idl"><a data-link-type="idl" href="#dom-mloperand-operator-slot" id="ref-for-dom-mloperand-operator-slot①⑤">[[operator]]</a></code>, <var>output1</var>.<code class="idl"><a data-link-type="idl" href="#dom-mloperand-operator-slot" id="ref-for-dom-mloperand-operator-slot①⑥">[[operator]]</a></code> and <var>output2</var>.<code class="idl"><a data-link-type="idl" href="#dom-mloperand-operator-slot" id="ref-for-dom-mloperand-operator-slot①⑦">[[operator]]</a></code>.</p>
+          <li data-md>
+           <p>Create an <a data-link-type="dfn" href="https://infra.spec.whatwg.org/#implementation-defined" id="ref-for-implementation-defined④①">implementation-defined</a> platform operand <var>outputImpl</var> to represent the output, given <var>output</var> and <var>opImpl</var>.</p>
+          <li data-md>
+           <p>Store a reference to <var>outputImpl</var> in <var>output0</var>.<code class="idl"><a data-link-type="idl" href="#dom-mloperand-operand-slot" id="ref-for-dom-mloperand-operand-slot④⑧">[[operand]]</a></code>, <var>output1</var>.<code class="idl"><a data-link-type="idl" href="#dom-mloperand-operand-slot" id="ref-for-dom-mloperand-operand-slot④⑨">[[operand]]</a></code> and <var>output2</var>.<code class="idl"><a data-link-type="idl" href="#dom-mloperand-operand-slot" id="ref-for-dom-mloperand-operand-slot⑤⓪">[[operand]]</a></code>.</p>
+         </ol>
+        <li data-md>
+         <p>Connect <var>input</var>.<code class="idl"><a data-link-type="idl" href="#dom-mloperand-operand-slot" id="ref-for-dom-mloperand-operand-slot⑤①">[[operand]]</a></code> as input to <var>opImpl</var>.</p>
+        <li data-md>
+         <p>Connect <var>output</var> as output to <var>opImpl</var>.</p>
+       </ol>
+      <li data-md>
+       <p>Return <var>output</var>.</p>
+     </ol>
+    </div>
+   </details>
+   <div class="note" role="note">
+    <details open>
+     <summary> The behavior of this operation can be generically emulated from the usage of other operations as follow. However, user agents typically have a more efficient implementation for it, therefore its usage is encouraged from the performance standpoint. </summary>
 <pre class="highlight"><c- a>const</c-> numDirections <c- o>=</c-> <c- p>(</c->options<c- p>.</c->direction <c- o>==</c-> <c- u>"both"</c-> <c- o>?</c-> <c- mf>2</c-> <c- o>:</c-> <c- mf>1</c-><c- p>);</c->
 <c- a>let</c-> hiddenState <c- o>=</c-> options<c- p>.</c->initialHiddenState<c- p>;</c->
 <c- a>let</c-> cellState <c- o>=</c-> options<c- p>.</c->initialCellState<c- p>;</c->
@@ -3277,13 +5352,13 @@ <h4 class="heading settled" data-level="7.6.20" id="api-mlgraphbuilder-lstm"><sp
 <c- a>let</c-> currentPeepholeWeight <c- o>=</c-> <c- p>[];</c->
 
 <c- k>for</c-> <c- p>(</c-><c- a>let</c-> dir <c- o>=</c-> <c- mf>0</c-><c- p>;</c-> dir <c- o>&lt;</c-> numDirections<c- p>;</c-> <c- o>++</c->dir<c- p>)</c-> <c- p>{</c->
-  currentWeight<c- p>.</c->push<c- p>(</c->builder<c- p>.</c->squeeze<c- p>(</c->builder<c- p>.</c->slice<c- p>(</c->weight<c- p>,</c-> <c- p>[</c->dir<c- p>,</c-> <c- mf>0</c-><c- p>,</c-> <c- mf>0</c-><c- p>],</c-> <c- p>[</c-><c- mf>1</c-><c- p>,</c-> <c- mf>4</c-> <c- o>*</c-> hidden_size<c- p>,</c-> input_size<c- p>]),</c-> <c- p>{</c-> axes<c- o>:</c-> <c- p>[</c-><c- mf>0</c-><c- p>]</c-> <c- p>}));</c->
-  currentRecurrentWeight<c- p>.</c->push<c- p>(</c->builder<c- p>.</c->squeeze<c- p>(</c->builder<c- p>.</c->slice<c- p>(</c->recurrentWeight<c- p>,</c-> <c- p>[</c->dir<c- p>,</c-> <c- mf>0</c-><c- p>,</c-> <c- mf>0</c-><c- p>],</c-> <c- p>[</c-><c- mf>1</c-><c- p>,</c-> <c- mf>4</c-> <c- o>*</c-> hidden_size<c- p>,</c-> hidden_size<c- p>]),</c-> <c- p>{</c-> axes<c- o>:</c-> <c- p>[</c-><c- mf>0</c-><c- p>]</c-> <c- p>}));</c->
-  currentBias<c- p>.</c->push<c- p>(</c->options<c- p>.</c->bias <c- o>?</c-> <c- p>(</c->builder<c- p>.</c->squeeze<c- p>(</c->builder<c- p>.</c->slice<c- p>(</c->options<c- p>.</c->bias<c- p>,</c-> <c- p>[</c->dir<c- p>,</c-> <c- mf>0</c-><c- p>],</c-> <c- p>[</c-><c- mf>1</c-><c- p>,</c-> <c- mf>4</c-> <c- o>*</c-> hidden_size<c- p>]),</c-> <c- p>{</c-> axes<c- o>:</c-> <c- p>[</c-><c- mf>0</c-><c- p>]</c-> <c- p>}))</c-> <c- o>:</c-> <c- kc>null</c-><c- p>);</c->
+  currentWeight<c- p>.</c->push<c- p>(</c->builder<c- p>.</c->squeeze<c- p>(</c->builder<c- p>.</c->slice<c- p>(</c->weight<c- p>,</c-> <c- p>[</c->dir<c- p>,</c-> <c- mf>0</c-><c- p>,</c-> <c- mf>0</c-><c- p>],</c-> <c- p>[</c-><c- mf>1</c-><c- p>,</c-> <c- mf>4</c-> <c- o>*</c-> hiddenSize<c- p>,</c-> inputSize<c- p>]),</c-> <c- p>{</c-> axes<c- o>:</c-> <c- p>[</c-><c- mf>0</c-><c- p>]</c-> <c- p>}));</c->
+  currentRecurrentWeight<c- p>.</c->push<c- p>(</c->builder<c- p>.</c->squeeze<c- p>(</c->builder<c- p>.</c->slice<c- p>(</c->recurrentWeight<c- p>,</c-> <c- p>[</c->dir<c- p>,</c-> <c- mf>0</c-><c- p>,</c-> <c- mf>0</c-><c- p>],</c-> <c- p>[</c-><c- mf>1</c-><c- p>,</c-> <c- mf>4</c-> <c- o>*</c-> hiddenSize<c- p>,</c-> hiddenSize<c- p>]),</c-> <c- p>{</c-> axes<c- o>:</c-> <c- p>[</c-><c- mf>0</c-><c- p>]</c-> <c- p>}));</c->
+  currentBias<c- p>.</c->push<c- p>(</c->options<c- p>.</c->bias <c- o>?</c-> <c- p>(</c->builder<c- p>.</c->squeeze<c- p>(</c->builder<c- p>.</c->slice<c- p>(</c->options<c- p>.</c->bias<c- p>,</c-> <c- p>[</c->dir<c- p>,</c-> <c- mf>0</c-><c- p>],</c-> <c- p>[</c-><c- mf>1</c-><c- p>,</c-> <c- mf>4</c-> <c- o>*</c-> hiddenSize<c- p>]),</c-> <c- p>{</c-> axes<c- o>:</c-> <c- p>[</c-><c- mf>0</c-><c- p>]</c-> <c- p>}))</c-> <c- o>:</c-> <c- kc>null</c-><c- p>);</c->
   currentRecurrentBias<c- p>.</c->push<c- p>(</c->options<c- p>.</c->recurrentBias <c- o>?</c->
-    <c- p>(</c->builder<c- p>.</c->squeeze<c- p>(</c->builder<c- p>.</c->slice<c- p>(</c->options<c- p>.</c->recurrentBias<c- p>,</c-> <c- p>[</c->dir<c- p>,</c-> <c- mf>0</c-><c- p>],</c-> <c- p>[</c-><c- mf>1</c-><c- p>,</c-> <c- mf>4</c-> <c- o>*</c-> hidden_size<c- p>]),</c-> <c- p>{</c-> axes<c- o>:</c-> <c- p>[</c-><c- mf>0</c-><c- p>]</c-> <c- p>}))</c-> <c- o>:</c-> <c- kc>null</c-><c- p>);</c->
+    <c- p>(</c->builder<c- p>.</c->squeeze<c- p>(</c->builder<c- p>.</c->slice<c- p>(</c->options<c- p>.</c->recurrentBias<c- p>,</c-> <c- p>[</c->dir<c- p>,</c-> <c- mf>0</c-><c- p>],</c-> <c- p>[</c-><c- mf>1</c-><c- p>,</c-> <c- mf>4</c-> <c- o>*</c-> hiddenSize<c- p>]),</c-> <c- p>{</c-> axes<c- o>:</c-> <c- p>[</c-><c- mf>0</c-><c- p>]</c-> <c- p>}))</c-> <c- o>:</c-> <c- kc>null</c-><c- p>);</c->
   currentPeepholeWeight<c- p>.</c->push<c- p>(</c->options<c- p>.</c->peepholeWeight <c- o>?</c->
-    <c- p>(</c->builder<c- p>.</c->squeeze<c- p>(</c->builder<c- p>.</c->slice<c- p>(</c->options<c- p>.</c->peepholeWeight<c- p>,</c-> <c- p>[</c->dir<c- p>,</c-> <c- mf>0</c-><c- p>],</c-> <c- p>[</c-><c- mf>1</c-><c- p>,</c-> <c- mf>3</c-> <c- o>*</c-> hidden_size<c- p>]),</c-> <c- p>{</c-> axes<c- o>:</c-> <c- p>[</c-><c- mf>0</c-><c- p>]</c-> <c- p>}))</c-> <c- o>:</c-> <c- kc>null</c-><c- p>);</c->
+    <c- p>(</c->builder<c- p>.</c->squeeze<c- p>(</c->builder<c- p>.</c->slice<c- p>(</c->options<c- p>.</c->peepholeWeight<c- p>,</c-> <c- p>[</c->dir<c- p>,</c-> <c- mf>0</c-><c- p>],</c-> <c- p>[</c-><c- mf>1</c-><c- p>,</c-> <c- mf>3</c-> <c- o>*</c-> hiddenSize<c- p>]),</c-> <c- p>{</c-> axes<c- o>:</c-> <c- p>[</c-><c- mf>0</c-><c- p>]</c-> <c- p>}))</c-> <c- o>:</c-> <c- kc>null</c-><c- p>);</c->
 <c- p>}</c->
 
 <c- k>for</c-> <c- p>(</c-><c- a>let</c-> step <c- o>=</c-> <c- mf>0</c-><c- p>;</c-> step <c- o>&lt;</c-> steps<c- p>;</c-> <c- o>++</c->step<c- p>)</c-> <c- p>{</c->
@@ -3293,13 +5368,13 @@ <h4 class="heading settled" data-level="7.6.20" id="api-mlgraphbuilder-lstm"><sp
   <c- a>let</c-> nextCell <c- o>=</c-> <c- kc>null</c-><c- p>;</c->
 
   <c- k>for</c-> <c- p>(</c-><c- a>let</c-> dir <c- o>=</c-> <c- mf>0</c-><c- p>;</c-> dir <c- o>&lt;</c-> numDirections<c- p>;</c-> <c- o>++</c->dir<c- p>)</c-> <c- p>{</c->
-    currentHidden<c- p>.</c->push<c- p>(</c->builder<c- p>.</c->squeeze<c- p>(</c->builder<c- p>.</c->slice<c- p>(</c->hiddenState<c- p>,</c-> <c- p>[</c->dir<c- p>,</c-> <c- mf>0</c-><c- p>,</c-> <c- mf>0</c-><c- p>],</c-> <c- p>[</c-><c- mf>1</c-><c- p>,</c-> batch_size<c- p>,</c-> hidden_size<c- p>]),</c-> <c- p>{</c-> axes<c- o>:</c-> <c- p>[</c-><c- mf>0</c-><c- p>]</c-> <c- p>}));</c->
-    currentCell<c- p>.</c->push<c- p>(</c->builder<c- p>.</c->squeeze<c- p>(</c->builder<c- p>.</c->slice<c- p>(</c->cellState<c- p>,</c-> <c- p>[</c->dir<c- p>,</c-> <c- mf>0</c-><c- p>,</c-> <c- mf>0</c-><c- p>],</c-> <c- p>[</c-><c- mf>1</c-><c- p>,</c-> batch_size<c- p>,</c-> hidden_size<c- p>]),</c-> <c- p>{</c-> axes<c- o>:</c-> <c- p>[</c-><c- mf>0</c-><c- p>]</c-> <c- p>}));</c->
+    currentHidden<c- p>.</c->push<c- p>(</c->builder<c- p>.</c->squeeze<c- p>(</c->builder<c- p>.</c->slice<c- p>(</c->hiddenState<c- p>,</c-> <c- p>[</c->dir<c- p>,</c-> <c- mf>0</c-><c- p>,</c-> <c- mf>0</c-><c- p>],</c-> <c- p>[</c-><c- mf>1</c-><c- p>,</c-> batchSize<c- p>,</c-> hiddenSize<c- p>]),</c-> <c- p>{</c-> axes<c- o>:</c-> <c- p>[</c-><c- mf>0</c-><c- p>]</c-> <c- p>}));</c->
+    currentCell<c- p>.</c->push<c- p>(</c->builder<c- p>.</c->squeeze<c- p>(</c->builder<c- p>.</c->slice<c- p>(</c->cellState<c- p>,</c-> <c- p>[</c->dir<c- p>,</c-> <c- mf>0</c-><c- p>,</c-> <c- mf>0</c-><c- p>],</c-> <c- p>[</c-><c- mf>1</c-><c- p>,</c-> batchSize<c- p>,</c-> hiddenSize<c- p>]),</c-> <c- p>{</c-> axes<c- o>:</c-> <c- p>[</c-><c- mf>0</c-><c- p>]</c-> <c- p>}));</c->
   <c- p>}</c->
 
   <c- k>for</c-> <c- p>(</c-><c- a>let</c-> dir <c- o>=</c-> <c- mf>0</c-><c- p>;</c-> dir <c- o>&lt;</c-> numDirections<c- p>;</c-> <c- o>++</c->dir<c- p>)</c-> <c- p>{</c->
     <c- a>let</c-> slice <c- o>=</c-> <c- p>(</c->dir <c- o>==</c-> <c- mf>1</c-> <c- o>||</c-> options<c- p>.</c->direction <c- o>==</c-> <c- u>"backward"</c-> <c- o>?</c-> steps <c- o>-</c-> step <c- o>-</c-> <c- mf>1</c-> <c- o>:</c-> step<c- p>);</c->
-    <c- a>let</c-> currentInput <c- o>=</c-> builder<c- p>.</c->squeeze<c- p>(</c->builder<c- p>.</c->slice<c- p>(</c->input<c- p>,</c-> <c- p>[</c->slice<c- p>,</c-> <c- mf>0</c-><c- p>,</c-> <c- mf>0</c-><c- p>],</c-> <c- p>[</c-><c- mf>1</c-><c- p>,</c-> batch_size<c- p>,</c-> input_size<c- p>]),</c-> <c- p>{</c-> axes<c- o>:</c-> <c- p>[</c-><c- mf>0</c-><c- p>]</c-> <c- p>});</c->
+    <c- a>let</c-> currentInput <c- o>=</c-> builder<c- p>.</c->squeeze<c- p>(</c->builder<c- p>.</c->slice<c- p>(</c->input<c- p>,</c-> <c- p>[</c->slice<c- p>,</c-> <c- mf>0</c-><c- p>,</c-> <c- mf>0</c-><c- p>],</c-> <c- p>[</c-><c- mf>1</c-><c- p>,</c-> batchSize<c- p>,</c-> inputSize<c- p>]),</c-> <c- p>{</c-> axes<c- o>:</c-> <c- p>[</c-><c- mf>0</c-><c- p>]</c-> <c- p>});</c->
 
     <c- a>let</c-> results <c- o>=</c-> builder<c- p>.</c->lstmCell<c- p>(</c->
       currentInput<c- p>,</c-> currentWeight<c- p>[</c->dir<c- p>],</c-> currentRecurrentWeight<c- p>[</c->dir<c- p>],</c->
@@ -3325,57 +5400,150 @@ <h4 class="heading settled" data-level="7.6.20" id="api-mlgraphbuilder-lstm"><sp
 
 <c- k>return</c-> <c- p>(</c->sequence <c- o>?</c-> <c- p>[</c->hiddenState<c- p>,</c-> cellState<c- p>,</c-> sequence<c- p>]</c-> <c- o>:</c-> <c- p>[</c->hiddenState<c- p>,</c-> cellState<c- p>]);</c->
 </pre>
-     </div>
-    </div>
-    <h4 class="heading settled" data-level="7.6.21" id="api-mlgraphbuilder-lstmcell"><span class="secno">7.6.21. </span><span class="content">The lstmCell() method</span><a class="self-link" href="#api-mlgraphbuilder-lstmcell"></a></h4>
-     A single time step of the Long Short-Term Memory <a data-link-type="biblio" href="#biblio-lstm" title="Long Short-Term Memory">[LSTM]</a> recurrent network using a cell state, an input, output, and forget gate to compute the cell state and the hidden state of the next time step that rolls into the output across the temporal sequence of the network. 
+    </details>
+   </div>
+   <h4 class="heading settled" data-level="7.6.22" id="api-mlgraphbuilder-lstmcell"><span class="secno">7.6.22. </span><span class="content">The lstmCell() method</span><a class="self-link" href="#api-mlgraphbuilder-lstmcell"></a></h4>
+    A single time step of the Long Short-Term Memory <a data-link-type="biblio" href="#biblio-lstm" title="Long Short-Term Memory">[LSTM]</a> recurrent network using a cell state, an input, output, and forget gate to compute the cell state and the hidden state of the next time step that rolls into the output across the temporal sequence of the network. 
 <pre class="idl highlight def"><c- b>dictionary</c-> <dfn class="dfn-paneled idl-code" data-dfn-type="dictionary" data-export id="dictdef-mllstmcelloptions"><code><c- g>MLLstmCellOptions</c-></code></dfn> {
-  <a data-link-type="idl-name" href="#mloperand" id="ref-for-mloperand①⑨③"><c- n>MLOperand</c-></a> <dfn class="idl-code" data-dfn-for="MLLstmCellOptions" data-dfn-type="dict-member" data-export data-type="MLOperand " id="dom-mllstmcelloptions-bias"><code><c- g>bias</c-></code><a class="self-link" href="#dom-mllstmcelloptions-bias"></a></dfn>;
-  <a data-link-type="idl-name" href="#mloperand" id="ref-for-mloperand①⑨④"><c- n>MLOperand</c-></a> <dfn class="idl-code" data-dfn-for="MLLstmCellOptions" data-dfn-type="dict-member" data-export data-type="MLOperand " id="dom-mllstmcelloptions-recurrentbias"><code><c- g>recurrentBias</c-></code><a class="self-link" href="#dom-mllstmcelloptions-recurrentbias"></a></dfn>;
-  <a data-link-type="idl-name" href="#mloperand" id="ref-for-mloperand①⑨⑤"><c- n>MLOperand</c-></a> <dfn class="idl-code" data-dfn-for="MLLstmCellOptions" data-dfn-type="dict-member" data-export data-type="MLOperand " id="dom-mllstmcelloptions-peepholeweight"><code><c- g>peepholeWeight</c-></code><a class="self-link" href="#dom-mllstmcelloptions-peepholeweight"></a></dfn>;
-  <a data-link-type="idl-name" href="#enumdef-mllstmweightlayout" id="ref-for-enumdef-mllstmweightlayout②"><c- n>MLLstmWeightLayout</c-></a> <dfn class="idl-code" data-default="&quot;iofg&quot;" data-dfn-for="MLLstmCellOptions" data-dfn-type="dict-member" data-export data-type="MLLstmWeightLayout " id="dom-mllstmcelloptions-layout"><code><c- g>layout</c-></code><a class="self-link" href="#dom-mllstmcelloptions-layout"></a></dfn> = "iofg";
-  <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#idl-sequence" id="ref-for-idl-sequence①⑤"><c- b>sequence</c-></a>&lt;<a data-link-type="idl-name" href="#mlactivation" id="ref-for-mlactivation②⑥"><c- n>MLActivation</c-></a>> <dfn class="idl-code" data-dfn-for="MLLstmCellOptions" data-dfn-type="dict-member" data-export data-type="sequence<MLActivation> " id="dom-mllstmcelloptions-activations"><code><c- g>activations</c-></code><a class="self-link" href="#dom-mllstmcelloptions-activations"></a></dfn>;
+  <a data-link-type="idl-name" href="#mloperand" id="ref-for-mloperand②④①"><c- n>MLOperand</c-></a> <a class="idl-code" data-link-type="dict-member" data-type="MLOperand " href="#dom-mllstmcelloptions-bias" id="ref-for-dom-mllstmcelloptions-bias"><c- g>bias</c-></a>;
+  <a data-link-type="idl-name" href="#mloperand" id="ref-for-mloperand②④②"><c- n>MLOperand</c-></a> <a class="idl-code" data-link-type="dict-member" data-type="MLOperand " href="#dom-mllstmcelloptions-recurrentbias" id="ref-for-dom-mllstmcelloptions-recurrentbias"><c- g>recurrentBias</c-></a>;
+  <a data-link-type="idl-name" href="#mloperand" id="ref-for-mloperand②④③"><c- n>MLOperand</c-></a> <a class="idl-code" data-link-type="dict-member" data-type="MLOperand " href="#dom-mllstmcelloptions-peepholeweight" id="ref-for-dom-mllstmcelloptions-peepholeweight"><c- g>peepholeWeight</c-></a>;
+  <a data-link-type="idl-name" href="#enumdef-mllstmweightlayout" id="ref-for-enumdef-mllstmweightlayout③"><c- n>MLLstmWeightLayout</c-></a> <a class="idl-code" data-default="&quot;iofg&quot;" data-link-type="dict-member" data-type="MLLstmWeightLayout " href="#dom-mllstmcelloptions-layout" id="ref-for-dom-mllstmcelloptions-layout"><c- g>layout</c-></a> = "iofg";
+  <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#idl-sequence" id="ref-for-idl-sequence①⑤"><c- b>sequence</c-></a>&lt;<a data-link-type="idl-name" href="#mlactivation" id="ref-for-mlactivation④①"><c- n>MLActivation</c-></a>> <a class="idl-code" data-link-type="dict-member" data-type="sequence<MLActivation> " href="#dom-mllstmcelloptions-activations" id="ref-for-dom-mllstmcelloptions-activations"><c- g>activations</c-></a>;
 };
 
-<c- b>partial</c-> <c- b>interface</c-> <a class="idl-code" data-link-type="interface" href="#mlgraphbuilder" id="ref-for-mlgraphbuilder③⑦"><c- g>MLGraphBuilder</c-></a> {
-  <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#idl-sequence" id="ref-for-idl-sequence①⑥"><c- b>sequence</c-></a>&lt;<a data-link-type="idl-name" href="#mloperand" id="ref-for-mloperand①⑨⑥"><c- n>MLOperand</c-></a>> <dfn class="idl-code" data-dfn-for="MLGraphBuilder" data-dfn-type="method" data-export data-lt="lstmCell(input, weight, recurrentWeight, hiddenState, cellState, hiddenSize, options)|lstmCell(input, weight, recurrentWeight, hiddenState, cellState, hiddenSize)" id="dom-mlgraphbuilder-lstmcell"><code><c- g>lstmCell</c-></code><a class="self-link" href="#dom-mlgraphbuilder-lstmcell"></a></dfn>(<a data-link-type="idl-name" href="#mloperand" id="ref-for-mloperand①⑨⑦"><c- n>MLOperand</c-></a> <dfn class="idl-code" data-dfn-for="MLGraphBuilder/lstmCell(input, weight, recurrentWeight, hiddenState, cellState, hiddenSize, options), MLGraphBuilder/lstmCell(input, weight, recurrentWeight, hiddenState, cellState, hiddenSize)" data-dfn-type="argument" data-export id="dom-mlgraphbuilder-lstmcell-input-weight-recurrentweight-hiddenstate-cellstate-hiddensize-options-input"><code><c- g>input</c-></code><a class="self-link" href="#dom-mlgraphbuilder-lstmcell-input-weight-recurrentweight-hiddenstate-cellstate-hiddensize-options-input"></a></dfn>, <a data-link-type="idl-name" href="#mloperand" id="ref-for-mloperand①⑨⑧"><c- n>MLOperand</c-></a> <dfn class="idl-code" data-dfn-for="MLGraphBuilder/lstmCell(input, weight, recurrentWeight, hiddenState, cellState, hiddenSize, options), MLGraphBuilder/lstmCell(input, weight, recurrentWeight, hiddenState, cellState, hiddenSize)" data-dfn-type="argument" data-export id="dom-mlgraphbuilder-lstmcell-input-weight-recurrentweight-hiddenstate-cellstate-hiddensize-options-weight"><code><c- g>weight</c-></code><a class="self-link" href="#dom-mlgraphbuilder-lstmcell-input-weight-recurrentweight-hiddenstate-cellstate-hiddensize-options-weight"></a></dfn>, <a data-link-type="idl-name" href="#mloperand" id="ref-for-mloperand①⑨⑨"><c- n>MLOperand</c-></a> <dfn class="idl-code" data-dfn-for="MLGraphBuilder/lstmCell(input, weight, recurrentWeight, hiddenState, cellState, hiddenSize, options), MLGraphBuilder/lstmCell(input, weight, recurrentWeight, hiddenState, cellState, hiddenSize)" data-dfn-type="argument" data-export id="dom-mlgraphbuilder-lstmcell-input-weight-recurrentweight-hiddenstate-cellstate-hiddensize-options-recurrentweight"><code><c- g>recurrentWeight</c-></code><a class="self-link" href="#dom-mlgraphbuilder-lstmcell-input-weight-recurrentweight-hiddenstate-cellstate-hiddensize-options-recurrentweight"></a></dfn>,
-                               <a data-link-type="idl-name" href="#mloperand" id="ref-for-mloperand②⓪⓪"><c- n>MLOperand</c-></a> <dfn class="idl-code" data-dfn-for="MLGraphBuilder/lstmCell(input, weight, recurrentWeight, hiddenState, cellState, hiddenSize, options), MLGraphBuilder/lstmCell(input, weight, recurrentWeight, hiddenState, cellState, hiddenSize)" data-dfn-type="argument" data-export id="dom-mlgraphbuilder-lstmcell-input-weight-recurrentweight-hiddenstate-cellstate-hiddensize-options-hiddenstate"><code><c- g>hiddenState</c-></code><a class="self-link" href="#dom-mlgraphbuilder-lstmcell-input-weight-recurrentweight-hiddenstate-cellstate-hiddensize-options-hiddenstate"></a></dfn>, <a data-link-type="idl-name" href="#mloperand" id="ref-for-mloperand②⓪①"><c- n>MLOperand</c-></a> <dfn class="idl-code" data-dfn-for="MLGraphBuilder/lstmCell(input, weight, recurrentWeight, hiddenState, cellState, hiddenSize, options), MLGraphBuilder/lstmCell(input, weight, recurrentWeight, hiddenState, cellState, hiddenSize)" data-dfn-type="argument" data-export id="dom-mlgraphbuilder-lstmcell-input-weight-recurrentweight-hiddenstate-cellstate-hiddensize-options-cellstate"><code><c- g>cellState</c-></code><a class="self-link" href="#dom-mlgraphbuilder-lstmcell-input-weight-recurrentweight-hiddenstate-cellstate-hiddensize-options-cellstate"></a></dfn>, <a class="idl-code" data-link-type="interface" href="https://webidl.spec.whatwg.org/#idl-unsigned-long" id="ref-for-idl-unsigned-long③⑤"><c- b>unsigned</c-> <c- b>long</c-></a> <dfn class="idl-code" data-dfn-for="MLGraphBuilder/lstmCell(input, weight, recurrentWeight, hiddenState, cellState, hiddenSize, options), MLGraphBuilder/lstmCell(input, weight, recurrentWeight, hiddenState, cellState, hiddenSize)" data-dfn-type="argument" data-export id="dom-mlgraphbuilder-lstmcell-input-weight-recurrentweight-hiddenstate-cellstate-hiddensize-options-hiddensize"><code><c- g>hiddenSize</c-></code><a class="self-link" href="#dom-mlgraphbuilder-lstmcell-input-weight-recurrentweight-hiddenstate-cellstate-hiddensize-options-hiddensize"></a></dfn>,
+<c- b>partial</c-> <c- b>interface</c-> <a class="idl-code" data-link-type="interface" href="#mlgraphbuilder" id="ref-for-mlgraphbuilder④⓪"><c- g>MLGraphBuilder</c-></a> {
+  <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#idl-sequence" id="ref-for-idl-sequence①⑥"><c- b>sequence</c-></a>&lt;<a data-link-type="idl-name" href="#mloperand" id="ref-for-mloperand②④④"><c- n>MLOperand</c-></a>> <a class="idl-code" data-link-type="method" href="#dom-mlgraphbuilder-lstmcell" id="ref-for-dom-mlgraphbuilder-lstmcell"><c- g>lstmCell</c-></a>(<a data-link-type="idl-name" href="#mloperand" id="ref-for-mloperand②④⑤"><c- n>MLOperand</c-></a> <dfn class="idl-code" data-dfn-for="MLGraphBuilder/lstmCell(input, weight, recurrentWeight, hiddenState, cellState, hiddenSize, options), MLGraphBuilder/lstmCell(input, weight, recurrentWeight, hiddenState, cellState, hiddenSize)" data-dfn-type="argument" data-export id="dom-mlgraphbuilder-lstmcell-input-weight-recurrentweight-hiddenstate-cellstate-hiddensize-options-input"><code><c- g>input</c-></code><a class="self-link" href="#dom-mlgraphbuilder-lstmcell-input-weight-recurrentweight-hiddenstate-cellstate-hiddensize-options-input"></a></dfn>, <a data-link-type="idl-name" href="#mloperand" id="ref-for-mloperand②④⑥"><c- n>MLOperand</c-></a> <dfn class="idl-code" data-dfn-for="MLGraphBuilder/lstmCell(input, weight, recurrentWeight, hiddenState, cellState, hiddenSize, options), MLGraphBuilder/lstmCell(input, weight, recurrentWeight, hiddenState, cellState, hiddenSize)" data-dfn-type="argument" data-export id="dom-mlgraphbuilder-lstmcell-input-weight-recurrentweight-hiddenstate-cellstate-hiddensize-options-weight"><code><c- g>weight</c-></code><a class="self-link" href="#dom-mlgraphbuilder-lstmcell-input-weight-recurrentweight-hiddenstate-cellstate-hiddensize-options-weight"></a></dfn>, <a data-link-type="idl-name" href="#mloperand" id="ref-for-mloperand②④⑦"><c- n>MLOperand</c-></a> <dfn class="idl-code" data-dfn-for="MLGraphBuilder/lstmCell(input, weight, recurrentWeight, hiddenState, cellState, hiddenSize, options), MLGraphBuilder/lstmCell(input, weight, recurrentWeight, hiddenState, cellState, hiddenSize)" data-dfn-type="argument" data-export id="dom-mlgraphbuilder-lstmcell-input-weight-recurrentweight-hiddenstate-cellstate-hiddensize-options-recurrentweight"><code><c- g>recurrentWeight</c-></code><a class="self-link" href="#dom-mlgraphbuilder-lstmcell-input-weight-recurrentweight-hiddenstate-cellstate-hiddensize-options-recurrentweight"></a></dfn>,
+                               <a data-link-type="idl-name" href="#mloperand" id="ref-for-mloperand②④⑧"><c- n>MLOperand</c-></a> <dfn class="idl-code" data-dfn-for="MLGraphBuilder/lstmCell(input, weight, recurrentWeight, hiddenState, cellState, hiddenSize, options), MLGraphBuilder/lstmCell(input, weight, recurrentWeight, hiddenState, cellState, hiddenSize)" data-dfn-type="argument" data-export id="dom-mlgraphbuilder-lstmcell-input-weight-recurrentweight-hiddenstate-cellstate-hiddensize-options-hiddenstate"><code><c- g>hiddenState</c-></code><a class="self-link" href="#dom-mlgraphbuilder-lstmcell-input-weight-recurrentweight-hiddenstate-cellstate-hiddensize-options-hiddenstate"></a></dfn>, <a data-link-type="idl-name" href="#mloperand" id="ref-for-mloperand②④⑨"><c- n>MLOperand</c-></a> <dfn class="idl-code" data-dfn-for="MLGraphBuilder/lstmCell(input, weight, recurrentWeight, hiddenState, cellState, hiddenSize, options), MLGraphBuilder/lstmCell(input, weight, recurrentWeight, hiddenState, cellState, hiddenSize)" data-dfn-type="argument" data-export id="dom-mlgraphbuilder-lstmcell-input-weight-recurrentweight-hiddenstate-cellstate-hiddensize-options-cellstate"><code><c- g>cellState</c-></code><a class="self-link" href="#dom-mlgraphbuilder-lstmcell-input-weight-recurrentweight-hiddenstate-cellstate-hiddensize-options-cellstate"></a></dfn>, <a class="idl-code" data-link-type="interface" href="https://webidl.spec.whatwg.org/#idl-unsigned-long" id="ref-for-idl-unsigned-long③⑦"><c- b>unsigned</c-> <c- b>long</c-></a> <dfn class="idl-code" data-dfn-for="MLGraphBuilder/lstmCell(input, weight, recurrentWeight, hiddenState, cellState, hiddenSize, options), MLGraphBuilder/lstmCell(input, weight, recurrentWeight, hiddenState, cellState, hiddenSize)" data-dfn-type="argument" data-export id="dom-mlgraphbuilder-lstmcell-input-weight-recurrentweight-hiddenstate-cellstate-hiddensize-options-hiddensize"><code><c- g>hiddenSize</c-></code><a class="self-link" href="#dom-mlgraphbuilder-lstmcell-input-weight-recurrentweight-hiddenstate-cellstate-hiddensize-options-hiddensize"></a></dfn>,
                                <c- b>optional</c-> <a data-link-type="idl-name" href="#dictdef-mllstmcelloptions" id="ref-for-dictdef-mllstmcelloptions"><c- n>MLLstmCellOptions</c-></a> <dfn class="idl-code" data-dfn-for="MLGraphBuilder/lstmCell(input, weight, recurrentWeight, hiddenState, cellState, hiddenSize, options), MLGraphBuilder/lstmCell(input, weight, recurrentWeight, hiddenState, cellState, hiddenSize)" data-dfn-type="argument" data-export id="dom-mlgraphbuilder-lstmcell-input-weight-recurrentweight-hiddenstate-cellstate-hiddensize-options-options"><code><c- g>options</c-></code><a class="self-link" href="#dom-mlgraphbuilder-lstmcell-input-weight-recurrentweight-hiddenstate-cellstate-hiddensize-options-options"></a></dfn> = {});
 };
 </pre>
-    <div class="algorithm" data-algorithm="lstmcell">
-      <strong>Arguments:</strong> 
-     <ul>
+   <p><code class="idl"><a data-link-type="idl" href="#dictdef-mllstmcelloptions" id="ref-for-dictdef-mllstmcelloptions①">MLLstmCellOptions</a></code> has the following members:</p>
+   <dl>
+    <dt data-md><dfn class="dfn-paneled idl-code" data-dfn-for="MLLstmCellOptions" data-dfn-type="dict-member" data-export id="dom-mllstmcelloptions-bias"><code>bias</code></dfn>, <span> of type <a data-link-type="idl-name" href="#mloperand" id="ref-for-mloperand②⑤⓪">MLOperand</a></span>
+    <dd data-md>
+     <p>An <code class="idl"><a data-link-type="idl" href="#mloperand" id="ref-for-mloperand②⑤①">MLOperand</a></code>. The 1-D input bias tensor of shape [4 * hiddenSize]. The ordering of the bias vectors in the first dimension of the tensor shape is specified according to the <code class="idl"><a data-link-type="idl" href="#dom-mllstmcelloptions-layout" id="ref-for-dom-mllstmcelloptions-layout①">layout</a></code> argument.</p>
+    <dt data-md><dfn class="dfn-paneled idl-code" data-dfn-for="MLLstmCellOptions" data-dfn-type="dict-member" data-export id="dom-mllstmcelloptions-recurrentbias"><code>recurrentBias</code></dfn>, <span> of type <a data-link-type="idl-name" href="#mloperand" id="ref-for-mloperand②⑤②">MLOperand</a></span>
+    <dd data-md>
+     <p>An <code class="idl"><a data-link-type="idl" href="#mloperand" id="ref-for-mloperand②⑤③">MLOperand</a></code>. The 1-D recurrent bias tensor of shape [4 * hiddenSize]. The ordering of the bias vectors in the first dimension of the tensor shape is specified according to the <code class="idl"><a data-link-type="idl" href="#dom-mllstmcelloptions-layout" id="ref-for-dom-mllstmcelloptions-layout②">layout</a></code> argument.</p>
+    <dt data-md><dfn class="dfn-paneled idl-code" data-dfn-for="MLLstmCellOptions" data-dfn-type="dict-member" data-export id="dom-mllstmcelloptions-peepholeweight"><code>peepholeWeight</code></dfn>, <span> of type <a data-link-type="idl-name" href="#mloperand" id="ref-for-mloperand②⑤④">MLOperand</a></span>
+    <dd data-md>
+     <p>An <code class="idl"><a data-link-type="idl" href="#mloperand" id="ref-for-mloperand②⑤⑤">MLOperand</a></code>. The 1-D weight tensor for peepholes of shape [3 * hiddenSize]. The pack ordering of the weight vectors is for the <code>input (i)</code>, <code>output (o)</code>, and <code>forget (f)</code> gate, respectively.</p>
+    <dt data-md><dfn class="dfn-paneled idl-code" data-dfn-for="MLLstmCellOptions" data-dfn-type="dict-member" data-export id="dom-mllstmcelloptions-layout"><code>layout</code></dfn>, <span> of type <a data-link-type="idl-name" href="#enumdef-mllstmweightlayout" id="ref-for-enumdef-mllstmweightlayout④">MLLstmWeightLayout</a>, defaulting to <code>"iofg"</code></span>
+    <dd data-md>
+     <p>An <code class="idl"><a data-link-type="idl" href="#enumdef-mllstmweightlayout" id="ref-for-enumdef-mllstmweightlayout⑤">MLLstmWeightLayout</a></code>. The ordering of the weight and bias vectors for the internal gates of LSTM, specifically the <code>input (i)</code>, <code>output (o)</code>, <code>forget (f)</code>, and <code>cell (g)</code> gate, as indicated in the first dimension of the weight and bias tensor shapes. When not specified, the default layout is <code>"iofg"</code>.</p>
+    <dt data-md><dfn class="dfn-paneled idl-code" data-dfn-for="MLLstmCellOptions" data-dfn-type="dict-member" data-export id="dom-mllstmcelloptions-activations"><code>activations</code></dfn>, <span> of type sequence&lt;<a data-link-type="idl-name" href="#mlactivation" id="ref-for-mlactivation④②">MLActivation</a>></span>
+    <dd data-md>
+     <p>A sequence of <code class="idl"><a data-link-type="idl" href="#mlactivation" id="ref-for-mlactivation④③">MLActivation</a></code>. A sequence of three activation functions, the first one is used for the <code>input (i)</code>, <code>forget (f)</code>, and <code>output (o)</code> gate, the second one is used for the <code>cell (g)</code> gate, and the last used for filtering the output cell state before combining it with the result of the output gate to form the output hidden state. When not specified, they are assumed to be of the sigmoid function (<code>"sigmoid"</code>) followed by two hyperbolic tangent functions (<code>"tanh"</code>) respectively.</p>
+   </dl>
+   <div>
+     <strong>Arguments:</strong> 
+    <ul>
+     <li data-md>
+      <p><em>input</em>: an <code class="idl"><a data-link-type="idl" href="#mloperand" id="ref-for-mloperand②⑤⑥">MLOperand</a></code>. The input 2-D tensor of shape [batchSize, inputSize].</p>
+     <li data-md>
+      <p><em>weight</em>: an <code class="idl"><a data-link-type="idl" href="#mloperand" id="ref-for-mloperand②⑤⑦">MLOperand</a></code>. The 2-D input weight tensor of shape [4 * hiddenSize, inputSize]. The ordering of the weight vectors in the first dimension of the tensor shape is specified according to the <em>options.layout</em> argument.</p>
+     <li data-md>
+      <p><em>recurrentWeight</em>: an <code class="idl"><a data-link-type="idl" href="#mloperand" id="ref-for-mloperand②⑤⑧">MLOperand</a></code>. The 2-D recurrent weight tensor of shape [4 * hiddenSize, hiddenSize]. The ordering of the weight vectors in the first dimension of the tensor shape is specified according to the <em>options.layout</em> argument.</p>
+     <li data-md>
+      <p><em>hiddenState</em>: an <code class="idl"><a data-link-type="idl" href="#mloperand" id="ref-for-mloperand②⑤⑨">MLOperand</a></code>. The 2-D input hidden state tensor of shape [batchSize, hiddenSize].</p>
+     <li data-md>
+      <p><em>cellState</em>: an <code class="idl"><a data-link-type="idl" href="#mloperand" id="ref-for-mloperand②⑥⓪">MLOperand</a></code>. The 2-D input cell state tensor of shape [batchSize, hiddenSize].</p>
+     <li data-md>
+      <p><em>hiddenSize</em>: an <code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#idl-unsigned-long" id="ref-for-idl-unsigned-long③⑧">unsigned long</a></code> scalar. The value of the second dimension of the output tensor shape. It indicates the number of features in the hidden state.</p>
+     <li data-md>
+      <p><em>options</em>: an optional <code class="idl"><a data-link-type="idl" href="#dictdef-mllstmcelloptions" id="ref-for-dictdef-mllstmcelloptions②">MLLstmCellOptions</a></code>. The optional parameters of the operation.</p>
+    </ul>
+    <p><strong>Returns:</strong> a sequence of <code class="idl"><a data-link-type="idl" href="#mloperand" id="ref-for-mloperand②⑥①">MLOperand</a></code>. The first element of the sequence is the output hidden state of the current time step of the recurrent network. The following element is the output cell state. Both elements are 2-D tensors of shape [batchSize, hiddenSize].</p>
+   </div>
+   <details class="algorithm" data-algorithm="lstmCell(input, weight, recurrentWeight, hiddenState, cellState, hiddenSize, options)" data-algorithm-for="MLGraphBuilder" open>
+    <summary> The <dfn class="dfn-paneled idl-code" data-dfn-for="MLGraphBuilder" data-dfn-type="method" data-export data-lt="lstmCell(input, weight, recurrentWeight, hiddenState, cellState, hiddenSize, options)|lstmCell(input, weight, recurrentWeight, hiddenState, cellState, hiddenSize)" id="dom-mlgraphbuilder-lstmcell"><code>lstmCell(<var>input</var>, <var>weight</var>, <var>recurrentWeight</var>, <var>hiddenState</var>, <var>cellState</var>, <var>hiddenSize</var>, <var>options</var>)</code></dfn> method steps are: </summary>
+    <div class="algorithm-steps">
+     <ol>
       <li data-md>
-       <p><em>input</em>: an <code class="idl"><a data-link-type="idl" href="#mloperand" id="ref-for-mloperand②⓪②">MLOperand</a></code>. The input 2-D tensor of shape [batch_size, input_size].</p>
+       <p><a data-link-type="dfn" href="https://infra.spec.whatwg.org/#assert" id="ref-for-assert⑤①">Assert</a>: the type of <var>input</var>, <var>weight</var>, <var>recurrentWeight</var>, <var>hiddenState</var> and <var>cellState</var> is <code class="idl"><a data-link-type="idl" href="#mloperand" id="ref-for-mloperand②⑥②">MLOperand</a></code>.</p>
       <li data-md>
-       <p><em>weight</em>: an <code class="idl"><a data-link-type="idl" href="#mloperand" id="ref-for-mloperand②⓪③">MLOperand</a></code>. The 2-D input weight tensor of shape [4 * hidden_size, input_size]. The ordering of the weight vectors in the first dimension of the tensor shape is specified according to the <em>options.layout</em> argument.</p>
+       <p>If the <a data-link-type="dfn" href="#mloperand-rank" id="ref-for-mloperand-rank①③">rank</a> of <var>input</var>, <var>weight</var>, <var>recurrentWeight</var>, <var>hiddenState</var> or <var>cellState</var> is not 2, then <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#dfn-throw" id="ref-for-dfn-throw①⑧②">throw</a> a "<code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#dataerror" id="ref-for-dataerror⑥⑤">DataError</a></code>" <code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#idl-DOMException" id="ref-for-idl-DOMException①⓪⑤">DOMException</a></code>.</p>
       <li data-md>
-       <p><em>recurrentWeight</em>: an <code class="idl"><a data-link-type="idl" href="#mloperand" id="ref-for-mloperand②⓪④">MLOperand</a></code>. The 2-D recurrent weight tensor of shape [4 * hidden_size, hidden_size]. The ordering of the weight vectors in the first dimension of the tensor shape is specified according to the <em>options.layout</em> argument.</p>
+       <p>Let <var>batchSize</var> be <var>input</var>.<code class="idl"><a data-link-type="idl" href="#dom-mloperand-descriptor-slot" id="ref-for-dom-mloperand-descriptor-slot⑦⓪">[[descriptor]]</a></code>.<code class="idl"><a data-link-type="idl" href="#dom-mloperanddescriptor-dimensions" id="ref-for-dom-mloperanddescriptor-dimensions⑥⑨">dimensions</a></code>[0].</p>
       <li data-md>
-       <p><em>hiddenState</em>: an <code class="idl"><a data-link-type="idl" href="#mloperand" id="ref-for-mloperand②⓪⑤">MLOperand</a></code>. The 2-D input hidden state tensor of shape [batch_size, hidden_size].</p>
+       <p>If <var>options</var>.<code class="idl"><a data-link-type="idl" href="#dom-mllstmcelloptions-bias" id="ref-for-dom-mllstmcelloptions-bias①">bias</a></code> <a data-link-type="dfn" href="https://infra.spec.whatwg.org/#map-exists" id="ref-for-map-exists④①">exists</a>:</p>
+       <ol>
+        <li data-md>
+         <p><a data-link-type="dfn" href="https://infra.spec.whatwg.org/#assert" id="ref-for-assert⑤②">Assert</a>: its type is <code class="idl"><a data-link-type="idl" href="#mloperand" id="ref-for-mloperand②⑥③">MLOperand</a></code>.</p>
+        <li data-md>
+         <p>If its rank is not 1, then <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#dfn-throw" id="ref-for-dfn-throw①⑧③">throw</a> a "<code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#dataerror" id="ref-for-dataerror⑥⑥">DataError</a></code>" <code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#idl-DOMException" id="ref-for-idl-DOMException①⓪⑥">DOMException</a></code>.</p>
+        <li data-md>
+         <p>If <var>options</var>.<code class="idl"><a data-link-type="idl" href="#dom-mllstmcelloptions-bias" id="ref-for-dom-mllstmcelloptions-bias②">bias</a></code>.<code class="idl"><a data-link-type="idl" href="#dom-mloperand-descriptor-slot" id="ref-for-dom-mloperand-descriptor-slot⑦①">[[descriptor]]</a></code>.<code class="idl"><a data-link-type="idl" href="#dom-mloperanddescriptor-dimensions" id="ref-for-dom-mloperanddescriptor-dimensions⑦⓪">dimensions</a></code>[0] is not 4 * <var>hiddenSize</var>, then <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#dfn-throw" id="ref-for-dfn-throw①⑧④">throw</a> a "<code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#dataerror" id="ref-for-dataerror⑥⑦">DataError</a></code>" <code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#idl-DOMException" id="ref-for-idl-DOMException①⓪⑦">DOMException</a></code>.</p>
+       </ol>
       <li data-md>
-       <p><em>cellState</em>: an <code class="idl"><a data-link-type="idl" href="#mloperand" id="ref-for-mloperand②⓪⑥">MLOperand</a></code>. The 2-D input cell state tensor of shape [batch_size, hidden_size].</p>
+       <p>If <var>options</var>.<code class="idl"><a data-link-type="idl" href="#dom-mllstmcelloptions-recurrentbias" id="ref-for-dom-mllstmcelloptions-recurrentbias①">recurrentBias</a></code> <a data-link-type="dfn" href="https://infra.spec.whatwg.org/#map-exists" id="ref-for-map-exists④②">exists</a>:</p>
+       <ol>
+        <li data-md>
+         <p><a data-link-type="dfn" href="https://infra.spec.whatwg.org/#assert" id="ref-for-assert⑤③">Assert</a>: its type is <code class="idl"><a data-link-type="idl" href="#mloperand" id="ref-for-mloperand②⑥④">MLOperand</a></code>.</p>
+        <li data-md>
+         <p>If its rank is not 1, then <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#dfn-throw" id="ref-for-dfn-throw①⑧⑤">throw</a> a "<code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#dataerror" id="ref-for-dataerror⑥⑧">DataError</a></code>" <code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#idl-DOMException" id="ref-for-idl-DOMException①⓪⑧">DOMException</a></code>.</p>
+        <li data-md>
+         <p>If <var>options</var>.<code class="idl"><a data-link-type="idl" href="#dom-mllstmcelloptions-recurrentbias" id="ref-for-dom-mllstmcelloptions-recurrentbias②">recurrentBias</a></code>.<code class="idl"><a data-link-type="idl" href="#dom-mloperand-descriptor-slot" id="ref-for-dom-mloperand-descriptor-slot⑦②">[[descriptor]]</a></code>.<code class="idl"><a data-link-type="idl" href="#dom-mloperanddescriptor-dimensions" id="ref-for-dom-mloperanddescriptor-dimensions⑦①">dimensions</a></code>[0] is not 4 * <var>hiddenSize</var>, then <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#dfn-throw" id="ref-for-dfn-throw①⑧⑥">throw</a> a "<code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#dataerror" id="ref-for-dataerror⑥⑨">DataError</a></code>" <code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#idl-DOMException" id="ref-for-idl-DOMException①⓪⑨">DOMException</a></code>.</p>
+       </ol>
       <li data-md>
-       <p><em>hiddenSize</em>: an <code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#idl-unsigned-long" id="ref-for-idl-unsigned-long③⑥">unsigned long</a></code> scalar. The value of the second dimension of the output tensor shape. It indicates the number of features in the hidden state.</p>
+       <p>If <var>options</var>.<code class="idl"><a data-link-type="idl" href="#dom-mllstmcelloptions-peepholeweight" id="ref-for-dom-mllstmcelloptions-peepholeweight①">peepholeWeight</a></code> <a data-link-type="dfn" href="https://infra.spec.whatwg.org/#map-exists" id="ref-for-map-exists④③">exists</a>:</p>
+       <ol>
+        <li data-md>
+         <p><a data-link-type="dfn" href="https://infra.spec.whatwg.org/#assert" id="ref-for-assert⑤④">Assert</a>: its type is <code class="idl"><a data-link-type="idl" href="#mloperand" id="ref-for-mloperand②⑥⑤">MLOperand</a></code>.</p>
+        <li data-md>
+         <p>If its rank is not 1, then <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#dfn-throw" id="ref-for-dfn-throw①⑧⑦">throw</a> a "<code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#dataerror" id="ref-for-dataerror⑦⓪">DataError</a></code>" <code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#idl-DOMException" id="ref-for-idl-DOMException①①⓪">DOMException</a></code>.</p>
+        <li data-md>
+         <p>If <var>options</var>.<code class="idl"><a data-link-type="idl" href="#dom-mllstmcelloptions-peepholeweight" id="ref-for-dom-mllstmcelloptions-peepholeweight②">peepholeWeight</a></code>.<code class="idl"><a data-link-type="idl" href="#dom-mloperand-descriptor-slot" id="ref-for-dom-mloperand-descriptor-slot⑦③">[[descriptor]]</a></code>.<code class="idl"><a data-link-type="idl" href="#dom-mloperanddescriptor-dimensions" id="ref-for-dom-mloperanddescriptor-dimensions⑦②">dimensions</a></code>[0] is not 3 * <var>hiddenSize</var>, then <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#dfn-throw" id="ref-for-dfn-throw①⑧⑧">throw</a> a "<code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#dataerror" id="ref-for-dataerror⑦①">DataError</a></code>" <code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#idl-DOMException" id="ref-for-idl-DOMException①①①">DOMException</a></code>.</p>
+       </ol>
       <li data-md>
-       <p><em>options</em>: an optional <code class="idl"><a data-link-type="idl" href="#dictdef-mllstmcelloptions" id="ref-for-dictdef-mllstmcelloptions①">MLLstmCellOptions</a></code>. The optional parameters of the operation.</p>
-       <ul>
+       <p>If <var>options</var>.<code class="idl"><a data-link-type="idl" href="#dom-mllstmcelloptions-activations" id="ref-for-dom-mllstmcelloptions-activations①">activations</a></code> <a data-link-type="dfn" href="https://infra.spec.whatwg.org/#map-exists" id="ref-for-map-exists④④">exists</a>:</p>
+       <ol>
+        <li data-md>
+         <p>If its <a data-link-type="dfn" href="https://infra.spec.whatwg.org/#list-size" id="ref-for-list-size③⑤">size</a> is not 3, then <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#dfn-throw" id="ref-for-dfn-throw①⑧⑨">throw</a> a <code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#exceptiondef-typeerror" id="ref-for-exceptiondef-typeerror③⑦">TypeError</a></code>.</p>
+        <li data-md>
+         <p><a data-link-type="dfn" href="https://infra.spec.whatwg.org/#assert" id="ref-for-assert⑤⑤">Assert</a>: the type of its elements is <code class="idl"><a data-link-type="idl" href="#mlactivation" id="ref-for-mlactivation④④">MLActivation</a></code>.</p>
+       </ol>
+      <li data-md>
+       <p>Let <var>desc</var> a new <code class="idl"><a data-link-type="idl" href="#dictdef-mloperanddescriptor" id="ref-for-dictdef-mloperanddescriptor②⓪">MLOperandDescriptor</a></code>.</p>
+      <li data-md>
+       <p>Set <var>desc</var>.<code class="idl"><a data-link-type="idl" href="#dom-mloperanddescriptor-dimensions" id="ref-for-dom-mloperanddescriptor-dimensions⑦③">dimensions</a></code> to [ <var>batchSize</var>, <var>hiddenSize</var> ].</p>
+      <li data-md>
+       <p>Set <var>desc</var>.<code class="idl"><a data-link-type="idl" href="#dom-mloperanddescriptor-type" id="ref-for-dom-mloperanddescriptor-type③①">type</a></code> to <var>input</var>.<code class="idl"><a data-link-type="idl" href="#dom-mloperand-descriptor-slot" id="ref-for-dom-mloperand-descriptor-slot⑦④">[[descriptor]]</a></code>.<code class="idl"><a data-link-type="idl" href="#dom-mloperanddescriptor-type" id="ref-for-dom-mloperanddescriptor-type③②">type</a></code>.</p>
+      <li data-md>
+       <p>If any of the following sub-steps fail, <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#dfn-throw" id="ref-for-dfn-throw①⑨⓪">throw</a> an "<code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#operationerror" id="ref-for-operationerror③①">OperationError</a></code>" <code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#idl-DOMException" id="ref-for-idl-DOMException①①②">DOMException</a></code>.</p>
+       <ol>
+        <li data-md>
+         <p>Let <var>output0</var> be the result of <a data-link-type="dfn" href="#create-an-mloperand" id="ref-for-create-an-mloperand①③">creating an MLOperand</a> given <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#this" id="ref-for-this②⑦">this</a> and <var>desc</var>.</p>
         <li data-md>
-         <p><em>bias</em>: an <code class="idl"><a data-link-type="idl" href="#mloperand" id="ref-for-mloperand②⓪⑦">MLOperand</a></code>. The 1-D input bias tensor of shape [4 * hidden_size]. The ordering of the bias vectors in the first dimension of the tensor shape is specified according to the <em>options.layout</em> argument.</p>
+         <p>Let <var>output1</var> be the result of <a data-link-type="dfn" href="#create-an-mloperand" id="ref-for-create-an-mloperand①④">creating an MLOperand</a> given <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#this" id="ref-for-this②⑧">this</a> and <var>desc</var>.</p>
         <li data-md>
-         <p><em>recurrentBias</em>: an <code class="idl"><a data-link-type="idl" href="#mloperand" id="ref-for-mloperand②⓪⑧">MLOperand</a></code>. The 1-D recurrent bias tensor of shape [4 * hidden_size]. The ordering of the bias vectors in the first dimension of the tensor shape is specified according to the <em>options.layout</em> argument.</p>
+         <p>Let <var>output</var> be the array [ <var>output0</var>, <var>output1</var> ].</p>
         <li data-md>
-         <p><em>peepholeWeight</em>: an <code class="idl"><a data-link-type="idl" href="#mloperand" id="ref-for-mloperand②⓪⑨">MLOperand</a></code>. The 1-D weight tensor for peepholes of shape [3 * hidden_size]. The pack ordering of the weight vectors is for the <em>input (i)</em>, <em>output (o)</em>, and <em>forget (f)</em> gate respectively.</p>
+         <p>Make a request to the underlying platform to:</p>
+         <ol>
+          <li data-md>
+           <p>Let <var>opImpl</var> be an <a data-link-type="dfn" href="https://infra.spec.whatwg.org/#implementation-defined" id="ref-for-implementation-defined④②">implementation-defined</a> platform operator for the LSTM cell operation, given <var>weight</var>, <var>recurrentWeight</var>, <var>hiddenState</var>, <var>cellState</var>, <var>hiddenSize</var> and <var>options</var>.</p>
+          <li data-md>
+           <p>Store a reference of <var>opImpl</var> in <var>output0</var>.<code class="idl"><a data-link-type="idl" href="#dom-mloperand-operator-slot" id="ref-for-dom-mloperand-operator-slot①⑧">[[operator]]</a></code> and <var>output1</var>.<code class="idl"><a data-link-type="idl" href="#dom-mloperand-operator-slot" id="ref-for-dom-mloperand-operator-slot①⑨">[[operator]]</a></code>.</p>
+          <li data-md>
+           <p>Create an <a data-link-type="dfn" href="https://infra.spec.whatwg.org/#implementation-defined" id="ref-for-implementation-defined④③">implementation-defined</a> platform operand <var>outputImpl</var> to represent the output, given <var>output</var> and <var>opImpl</var>.</p>
+          <li data-md>
+           <p>Store a reference to <var>outputImpl</var> in <var>output0</var>.<code class="idl"><a data-link-type="idl" href="#dom-mloperand-operand-slot" id="ref-for-dom-mloperand-operand-slot⑤②">[[operand]]</a></code> and <var>output1</var>.<code class="idl"><a data-link-type="idl" href="#dom-mloperand-operand-slot" id="ref-for-dom-mloperand-operand-slot⑤③">[[operand]]</a></code>.</p>
+         </ol>
         <li data-md>
-         <p><em>layout</em>: an <code class="idl"><a data-link-type="idl" href="#enumdef-mllstmweightlayout" id="ref-for-enumdef-mllstmweightlayout③">MLLstmWeightLayout</a></code>. The ordering of the weight and bias vectors for the internal gates of LSTM, specifically the <em>input (i)</em>, <em>output (o)</em>, <em>forget (f)</em>, and <em>cell (g)</em> gate, as indicated in the first dimension of the weight and bias tensor shapes. When not specified, the default layout is <em>"iofg"</em>.</p>
+         <p>Connect <var>input</var>.<code class="idl"><a data-link-type="idl" href="#dom-mloperand-operand-slot" id="ref-for-dom-mloperand-operand-slot⑤④">[[operand]]</a></code> as input to <var>opImpl</var>.</p>
         <li data-md>
-         <p><em>activations</em>: a sequence of <code class="idl"><a data-link-type="idl" href="#mlactivation" id="ref-for-mlactivation②⑦">MLActivation</a></code>. A sequence of three activation functions, the first one is used for the <em>input (i)</em>, <em>forget (f)</em>, and <em>output (o)</em> gate, the second one is used for the <em>cell (g)</em> gate, and the last used for filtering the output cell state before combining it with the result of the output gate to form the output hidden state. When not specified, they are assumed to be of the sigmoid function (<em>"sigmoid"</em>) followed by two hyperbolic tangent functions (<em>"tanh"</em>) respectively.</p>
-       </ul>
-     </ul>
-     <p><strong>Returns:</strong> a sequence of <code class="idl"><a data-link-type="idl" href="#mloperand" id="ref-for-mloperand②①⓪">MLOperand</a></code>. The first element of the sequence is the output hidden state of the current time step of the recurrent network. The following element is the output cell state. Both elements are 2-D tensors of shape [batch_size, hidden_size].</p>
-     <div class="note" role="note">
-       The behavior of this operation can be generically emulated via other operations as shown below, when the weight layout is the default <em>"iofg"</em> layout, and the activation functions of the input/forget/output gate and the cell gate/the cell state’s filter for the output hidden state are of the operator types <em>sigmoid</em> and <em>tanh</em> respectively. 
+         <p>Connect <var>output</var> as output to <var>opImpl</var>.</p>
+       </ol>
+      <li data-md>
+       <p>Return <var>output</var>.</p>
+     </ol>
+    </div>
+   </details>
+   <div class="note" role="note">
+    <details open>
+     <summary> The behavior of this operation can be generically emulated via other operations as shown below, when the weight layout is the default <em>"iofg"</em> layout, and the activation functions of the input/forget/output gate and the cell gate/the cell state’s filter for the output hidden state are of the operator types <em>sigmoid</em> and <em>tanh</em> respectively. </summary>
 <pre class="highlight"><c- a>const</c-> zero <c- o>=</c-> builder<c- p>.</c->constant<c- p>(</c-><c- mf>0</c-><c- p>);</c->
 
 <c- c1>// input gate (i)</c->
@@ -3393,11 +5561,11 @@ <h4 class="heading settled" data-level="7.6.21" id="api-mlgraphbuilder-lstmcell"
       builder<c- p>.</c->add<c- p>(</c->
         builder<c- p>.</c->matmul<c- p>(</c->
           input<c- p>,</c->
-          builder<c- p>.</c->transpose<c- p>(</c->builder<c- p>.</c->slice<c- p>(</c->weight<c- p>,</c-> <c- p>[</c-><c- mf>0</c-><c- p>,</c-> <c- mf>0</c-><c- p>],</c-> <c- p>[</c->hiddenSize<c- p>,</c-> input_size<c- p>]))</c->
+          builder<c- p>.</c->transpose<c- p>(</c->builder<c- p>.</c->slice<c- p>(</c->weight<c- p>,</c-> <c- p>[</c-><c- mf>0</c-><c- p>,</c-> <c- mf>0</c-><c- p>],</c-> <c- p>[</c->hiddenSize<c- p>,</c-> inputSize<c- p>]))</c->
         <c- p>),</c->
         builder<c- p>.</c->matmul<c- p>(</c->
           hiddenState<c- p>,</c->
-          builder<c- p>.</c->transpose<c- p>(</c->builder<c- p>.</c->slice<c- p>(</c->recurrentWeight<c- p>,</c-> <c- p>[</c-><c- mf>0</c-><c- p>,</c-> <c- mf>0</c-><c- p>],</c-> <c- p>[</c->hiddenSize<c- p>,</c-> hidden_size<c- p>]))</c->
+          builder<c- p>.</c->transpose<c- p>(</c->builder<c- p>.</c->slice<c- p>(</c->recurrentWeight<c- p>,</c-> <c- p>[</c-><c- mf>0</c-><c- p>,</c-> <c- mf>0</c-><c- p>],</c-> <c- p>[</c->hiddenSize<c- p>,</c-> hiddenSize<c- p>]))</c->
         <c- p>)</c->
       <c- p>)</c->
     <c- p>)</c->
@@ -3419,11 +5587,11 @@ <h4 class="heading settled" data-level="7.6.21" id="api-mlgraphbuilder-lstmcell"
       builder<c- p>.</c->add<c- p>(</c->
         builder<c- p>.</c->matmul<c- p>(</c->
           input<c- p>,</c->
-          builder<c- p>.</c->transpose<c- p>(</c->builder<c- p>.</c->slice<c- p>(</c->weight<c- p>,</c-> <c- p>[</c-><c- mf>2</c-> <c- o>*</c-> hiddenSize<c- p>,</c-> <c- mf>0</c-><c- p>],</c-> <c- p>[</c->hiddenSize<c- p>,</c-> input_size<c- p>]))</c->
+          builder<c- p>.</c->transpose<c- p>(</c->builder<c- p>.</c->slice<c- p>(</c->weight<c- p>,</c-> <c- p>[</c-><c- mf>2</c-> <c- o>*</c-> hiddenSize<c- p>,</c-> <c- mf>0</c-><c- p>],</c-> <c- p>[</c->hiddenSize<c- p>,</c-> inputSize<c- p>]))</c->
         <c- p>),</c->
         builder<c- p>.</c->matmul<c- p>(</c->
           hiddenState<c- p>,</c->
-          builder<c- p>.</c->transpose<c- p>(</c->builder<c- p>.</c->slice<c- p>(</c->recurrentWeight<c- p>,</c-> <c- p>[</c-><c- mf>2</c-> <c- o>*</c-> hiddenSize<c- p>,</c-> <c- mf>0</c-><c- p>],</c-> <c- p>[</c->hiddenSize<c- p>,</c-> hidden_size<c- p>]))</c->
+          builder<c- p>.</c->transpose<c- p>(</c->builder<c- p>.</c->slice<c- p>(</c->recurrentWeight<c- p>,</c-> <c- p>[</c-><c- mf>2</c-> <c- o>*</c-> hiddenSize<c- p>,</c-> <c- mf>0</c-><c- p>],</c-> <c- p>[</c->hiddenSize<c- p>,</c-> hiddenSize<c- p>]))</c->
         <c- p>)</c->
       <c- p>)</c->
     <c- p>)</c->
@@ -3440,11 +5608,11 @@ <h4 class="heading settled" data-level="7.6.21" id="api-mlgraphbuilder-lstmcell"
     builder<c- p>.</c->add<c- p>(</c->
       builder<c- p>.</c->matmul<c- p>(</c->
         input<c- p>,</c->
-        builder<c- p>.</c->transpose<c- p>(</c->builder<c- p>.</c->slice<c- p>(</c->weight<c- p>,</c-> <c- p>[</c-><c- mf>3</c-> <c- o>*</c-> hiddenSize<c- p>,</c-> <c- mf>0</c-><c- p>],</c-> <c- p>[</c->hiddenSize<c- p>,</c-> input_size<c- p>]))</c->
+        builder<c- p>.</c->transpose<c- p>(</c->builder<c- p>.</c->slice<c- p>(</c->weight<c- p>,</c-> <c- p>[</c-><c- mf>3</c-> <c- o>*</c-> hiddenSize<c- p>,</c-> <c- mf>0</c-><c- p>],</c-> <c- p>[</c->hiddenSize<c- p>,</c-> inputSize<c- p>]))</c->
       <c- p>),</c->
       builder<c- p>.</c->matmul<c- p>(</c->
         hiddenState<c- p>,</c->
-        builder<c- p>.</c->transpose<c- p>(</c->builder<c- p>.</c->slice<c- p>(</c->recurrentWeight<c- p>,</c-> <c- p>[</c-><c- mf>3</c-> <c- o>*</c-> hiddenSize<c- p>,</c-> <c- mf>0</c-><c- p>],</c-> <c- p>[</c->hiddenSize<c- p>,</c-> hidden_size<c- p>]))</c->
+        builder<c- p>.</c->transpose<c- p>(</c->builder<c- p>.</c->slice<c- p>(</c->recurrentWeight<c- p>,</c-> <c- p>[</c-><c- mf>3</c-> <c- o>*</c-> hiddenSize<c- p>,</c-> <c- mf>0</c-><c- p>],</c-> <c- p>[</c->hiddenSize<c- p>,</c-> hiddenSize<c- p>]))</c->
       <c- p>)</c->
     <c- p>)</c->
   <c- p>)</c->
@@ -3465,11 +5633,11 @@ <h4 class="heading settled" data-level="7.6.21" id="api-mlgraphbuilder-lstmcell"
       builder<c- p>.</c->add<c- p>(</c->
         builder<c- p>.</c->matmul<c- p>(</c->
           input<c- p>,</c->
-          builder<c- p>.</c->transpose<c- p>(</c->builder<c- p>.</c->slice<c- p>(</c->weight<c- p>,</c-> <c- p>[</c->hiddenSize<c- p>,</c-> <c- mf>0</c-><c- p>],</c-> <c- p>[</c->hiddenSize<c- p>,</c-> input_size<c- p>]))</c->
+          builder<c- p>.</c->transpose<c- p>(</c->builder<c- p>.</c->slice<c- p>(</c->weight<c- p>,</c-> <c- p>[</c->hiddenSize<c- p>,</c-> <c- mf>0</c-><c- p>],</c-> <c- p>[</c->hiddenSize<c- p>,</c-> inputSize<c- p>]))</c->
         <c- p>),</c->
         builder<c- p>.</c->matmul<c- p>(</c->
           hiddenState<c- p>,</c->
-          builder<c- p>.</c->transpose<c- p>(</c->builder<c- p>.</c->slice<c- p>(</c->recurrentWeight<c- p>,</c-> <c- p>[</c->hiddenSize<c- p>,</c-> <c- mf>0</c-><c- p>],</c-> <c- p>[</c->hiddenSize<c- p>,</c-> hidden_size<c- p>]))</c->
+          builder<c- p>.</c->transpose<c- p>(</c->builder<c- p>.</c->slice<c- p>(</c->recurrentWeight<c- p>,</c-> <c- p>[</c->hiddenSize<c- p>,</c-> <c- mf>0</c-><c- p>],</c-> <c- p>[</c->hiddenSize<c- p>,</c-> hiddenSize<c- p>]))</c->
         <c- p>)</c->
       <c- p>)</c->
     <c- p>)</c->
@@ -3484,49 +5652,113 @@ <h4 class="heading settled" data-level="7.6.21" id="api-mlgraphbuilder-lstmcell"
 
 <c- k>return</c-> <c- p>[</c->ht<c- p>,</c-> ct<c- p>];</c->
 </pre>
-     </div>
-    </div>
-    <h4 class="heading settled" data-level="7.6.22" id="api-mlgraphbuilder-matmul"><span class="secno">7.6.22. </span><span class="content">The matmul() method</span><a class="self-link" href="#api-mlgraphbuilder-matmul"></a></h4>
-     Compute the matrix product of two input tensors. 
-<pre class="idl highlight def"><c- b>partial</c-> <c- b>interface</c-> <a class="idl-code" data-link-type="interface" href="#mlgraphbuilder" id="ref-for-mlgraphbuilder③⑧"><c- g>MLGraphBuilder</c-></a> {
-  <a data-link-type="idl-name" href="#mloperand" id="ref-for-mloperand②①①"><c- n>MLOperand</c-></a> <dfn class="idl-code" data-dfn-for="MLGraphBuilder" data-dfn-type="method" data-export data-lt="matmul(a, b)" id="dom-mlgraphbuilder-matmul"><code><c- g>matmul</c-></code><a class="self-link" href="#dom-mlgraphbuilder-matmul"></a></dfn>(<a data-link-type="idl-name" href="#mloperand" id="ref-for-mloperand②①②"><c- n>MLOperand</c-></a> <dfn class="idl-code" data-dfn-for="MLGraphBuilder/matmul(a, b)" data-dfn-type="argument" data-export id="dom-mlgraphbuilder-matmul-a-b-a"><code><c- g>a</c-></code><a class="self-link" href="#dom-mlgraphbuilder-matmul-a-b-a"></a></dfn>, <a data-link-type="idl-name" href="#mloperand" id="ref-for-mloperand②①③"><c- n>MLOperand</c-></a> <dfn class="idl-code" data-dfn-for="MLGraphBuilder/matmul(a, b)" data-dfn-type="argument" data-export id="dom-mlgraphbuilder-matmul-a-b-b"><code><c- g>b</c-></code><a class="self-link" href="#dom-mlgraphbuilder-matmul-a-b-b"></a></dfn>);
+    </details>
+   </div>
+   <h4 class="heading settled" data-level="7.6.23" id="api-mlgraphbuilder-matmul"><span class="secno">7.6.23. </span><span class="content">The matmul() method</span><a class="self-link" href="#api-mlgraphbuilder-matmul"></a></h4>
+    Compute the matrix product of two input tensors. 
+<pre class="idl highlight def"><c- b>partial</c-> <c- b>interface</c-> <a class="idl-code" data-link-type="interface" href="#mlgraphbuilder" id="ref-for-mlgraphbuilder④①"><c- g>MLGraphBuilder</c-></a> {
+  <a data-link-type="idl-name" href="#mloperand" id="ref-for-mloperand②⑥⑥"><c- n>MLOperand</c-></a> <a class="idl-code" data-link-type="method" href="#dom-mlgraphbuilder-matmul" id="ref-for-dom-mlgraphbuilder-matmul"><c- g>matmul</c-></a>(<a data-link-type="idl-name" href="#mloperand" id="ref-for-mloperand②⑥⑦"><c- n>MLOperand</c-></a> <dfn class="idl-code" data-dfn-for="MLGraphBuilder/matmul(a, b)" data-dfn-type="argument" data-export id="dom-mlgraphbuilder-matmul-a-b-a"><code><c- g>a</c-></code><a class="self-link" href="#dom-mlgraphbuilder-matmul-a-b-a"></a></dfn>, <a data-link-type="idl-name" href="#mloperand" id="ref-for-mloperand②⑥⑧"><c- n>MLOperand</c-></a> <dfn class="idl-code" data-dfn-for="MLGraphBuilder/matmul(a, b)" data-dfn-type="argument" data-export id="dom-mlgraphbuilder-matmul-a-b-b"><code><c- g>b</c-></code><a class="self-link" href="#dom-mlgraphbuilder-matmul-a-b-b"></a></dfn>);
 };
 </pre>
-    <div class="algorithm" data-algorithm="matmul">
-      <strong>Arguments:</strong> 
-     <ul>
+   <div>
+     <strong>Arguments:</strong> 
+    <ul>
+     <li data-md>
+      <p><em>a</em>: an <code class="idl"><a data-link-type="idl" href="#mloperand" id="ref-for-mloperand②⑥⑨">MLOperand</a></code>. The first N-dimensional input tensor.</p>
+     <li data-md>
+      <p><em>b</em>: an <code class="idl"><a data-link-type="idl" href="#mloperand" id="ref-for-mloperand②⑦⓪">MLOperand</a></code>. The second N-dimensional input tensor.</p>
+    </ul>
+    <p><strong>Returns:</strong> an <code class="idl"><a data-link-type="idl" href="#mloperand" id="ref-for-mloperand②⑦①">MLOperand</a></code>. The output tensor that contains the matrix
+    product of two input tensors.</p>
+   </div>
+   <div>
+     Computes the matrix product of two input tensors as follows: 
+    <ul>
+     <li data-md>
+      <p>If both <em>a</em> and <em>b</em> are 2-dimensional, they are multiplied like conventional
+matrices and produce a 2-dimensional tensor as the output.</p>
+     <li data-md>
+      <p>If either <em>a</em> or <em>b</em> is <code>N</code>-dimensional where <code>N > 2</code>, it is treated as a stack of matrices with dimensions corresponding to the last two indices. The matrix multiplication will be broadcasted accordingly by following the <a data-link-type="biblio" href="#biblio-numpy-broadcasting-rule" title="General Broadcasting Rules of NumPy">[numpy-broadcasting-rule]</a>. The output is a <code>N</code>-dimensional tensor whose rank is the maximum <a data-link-type="dfn" href="#mloperand-rank" id="ref-for-mloperand-rank①④">rank</a> of the input tensors. For each dimension, except the last two, of the output tensor, its size is the maximum size along that dimension of the input tensors.</p>
+     <li data-md>
+      <p>If <em>a</em> is 1-dimensional, it is converted to a 2-dimensional tensor by prepending a 1 to its dimensions.</p>
+     <li data-md>
+      <p>If <em>b</em> is 1-dimensional, it is converted to a 2-dimensional tensor by by appending a 1 to its dimensions.</p>
+     <li data-md>
+      <p>If both <em>a</em> and <em>b</em> are 1-dimensional, the operation is a vector dot-product, which produces a scalar output.</p>
+    </ul>
+   </div>
+   <details class="algorithm" data-algorithm="calculate matmul output sizes" data-algorithm-for="MLGraphBuilder" open>
+    <summary> To <dfn class="dfn-paneled" data-dfn-for="MLGraphBuilder" data-dfn-type="dfn" data-noexport id="mlgraphbuilder-calculate-matmul-output-sizes">calculate matmul output sizes</dfn>, given <var>a</var> and <var>b</var> run the following steps: </summary>
+    <div class="algorithm-steps">
+     <ol>
       <li data-md>
-       <p><em>a</em>: an <code class="idl"><a data-link-type="idl" href="#mloperand" id="ref-for-mloperand②①④">MLOperand</a></code>. The first input N-D tensor.</p>
+       <p>Let <var>shapeA</var> be <var>a</var>.<code class="idl"><a data-link-type="idl" href="#dom-mloperand-descriptor-slot" id="ref-for-dom-mloperand-descriptor-slot⑦⑤">[[descriptor]]</a></code>.<code class="idl"><a data-link-type="idl" href="#dom-mloperanddescriptor-dimensions" id="ref-for-dom-mloperanddescriptor-dimensions⑦④">dimensions</a></code> and <var>sizeA</var> the <a data-link-type="dfn" href="https://infra.spec.whatwg.org/#list-size" id="ref-for-list-size③⑥">size</a> of <var>shapeA</var>.</p>
       <li data-md>
-       <p><em>b</em>: an <code class="idl"><a data-link-type="idl" href="#mloperand" id="ref-for-mloperand②①⑤">MLOperand</a></code>. The second input N-D tensor.</p>
-     </ul>
-     <p><strong>Returns:</strong> an <code class="idl"><a data-link-type="idl" href="#mloperand" id="ref-for-mloperand②①⑥">MLOperand</a></code>. The output N-D tensor that contains the matrix
-    product of two input tensors.</p>
-     <p>Compute the matrix product of two input tensors. It behaves as following:</p>
-     <ul>
+       <p>Let <var>shapeB</var> be <var>b</var>.<code class="idl"><a data-link-type="idl" href="#dom-mloperand-descriptor-slot" id="ref-for-dom-mloperand-descriptor-slot⑦⑥">[[descriptor]]</a></code>.<code class="idl"><a data-link-type="idl" href="#dom-mloperanddescriptor-dimensions" id="ref-for-dom-mloperanddescriptor-dimensions⑦⑤">dimensions</a></code> and <var>sizeB</var> the <a data-link-type="dfn" href="https://infra.spec.whatwg.org/#list-size" id="ref-for-list-size③⑦">size</a> of <var>shapeB</var>.</p>
       <li data-md>
-       <p>If both <em>a</em> and <em>b</em> are 2-D, they are multiplied like conventional
-matrices and produce a 2-D tensor as the output.</p>
+       <p>If <var>sizeA</var> and <var>sizeB</var> is 1, return <code>«  1  »</code>.</p>
       <li data-md>
-       <p>If either <em>a</em> or <em>b</em> is N-D, N > 2, it is treated as a stack of
-matrices with dimensions corresponding to the last two indices. The
-matrix multiplication will be broadcasted accordingly by following <a data-link-type="biblio" href="#biblio-numpy-broadcasting-rule" title="General Broadcasting Rules of NumPy">[numpy-broadcasting-rule]</a>. The output is a N-D tensor whose rank
-is the maximum rank of the input tensors. For each dimension, except
-the last two, of the output tensor, its size is the maximum size
-along that dimension of the input tensors.</p>
+       <p>If <var>sizeA</var> is 1 and <var>sizeB</var> is not, then insert 1 in the front of <var>shapeA</var> to become [ 1 | <var>shapeA</var> ] and let <var>sizeA</var> be 2.</p>
       <li data-md>
-       <p>If <em>a</em> is 1-D, it is converted to a 2-D tensor by prepending a 1 to
-its dimensions.</p>
+       <p>If <var>shapeA</var>[0] is not equal to <var>shapeB</var>[<var>sizeB</var> - 2], then <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#dfn-throw" id="ref-for-dfn-throw①⑨①">throw</a> an "<code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#operationerror" id="ref-for-operationerror③②">OperationError</a></code>" <code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#idl-DOMException" id="ref-for-idl-DOMException①①③">DOMException</a></code>.</p>
       <li data-md>
-       <p>If <em>b</em> is 1-D, it is converted to a 2-D tensor by by appending a 1 to
-its dimensions.</p>
+       <p>If <var>sizeB</var> is 1 and <var>sizeA</var> is not, then append 1 to <var>shapeB</var> to become [ <var>shapeB</var> | 1 ] and let <var>sizeB</var> be 2.</p>
       <li data-md>
-       <p>If both <em>a</em> and <em>b</em> are 1-D, the operation is a vector dot-product,
-which produces a scalar output.</p>
-     </ul>
+       <p>If <var>shapeA</var>[<var>sizeA</var> - 1] is not equal to <var>shapeB</var>[0], then <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#dfn-throw" id="ref-for-dfn-throw①⑨②">throw</a> an "<code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#operationerror" id="ref-for-operationerror③③">OperationError</a></code>" <code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#idl-DOMException" id="ref-for-idl-DOMException①①④">DOMException</a></code>.</p>
+      <li data-md>
+       <p>Let <var>shape</var> be an array whose size <var>size</var> is the maximum of <var>sizeA</var> and <var>sizeB</var>.</p>
+      <li data-md>
+       <p><a data-link-type="dfn" href="https://infra.spec.whatwg.org/#map-iterate" id="ref-for-map-iterate①③">For each</a> <var>index</var> in <a data-link-type="dfn" href="https://infra.spec.whatwg.org/#the-range" id="ref-for-the-range③">the range</a> 0 to <var>size</var>, exclusive:</p>
+      <li data-md>
+       <p>Set <var>shape</var>[<var>index</var>] to the maximum of <var>shapeA</var>[<var>index</var>] and <var>shapeB</var>[<var>index</var>].</p>
+      <li data-md>
+       <p>Return <var>shape</var>.</p>
+     </ol>
+    </div>
+   </details>
+   <details class="algorithm" data-algorithm="matmul(a, b)" data-algorithm-for="MLGraphBuilder" open>
+    <summary> The <dfn class="dfn-paneled idl-code" data-dfn-for="MLGraphBuilder" data-dfn-type="method" data-export id="dom-mlgraphbuilder-matmul"><code>matmul(<var>a</var>, <var>b</var>)</code></dfn> method steps are: </summary>
+    <div class="algorithm-steps">
+     <ol>
+      <li data-md>
+       <p><a data-link-type="dfn" href="https://infra.spec.whatwg.org/#assert" id="ref-for-assert⑤⑥">Assert</a>: the type of <var>a</var> and <var>b</var> is <code class="idl"><a data-link-type="idl" href="#mloperand" id="ref-for-mloperand②⑦②">MLOperand</a></code>.</p>
+      <li data-md>
+       <p>Let <var>desc</var> a new <code class="idl"><a data-link-type="idl" href="#dictdef-mloperanddescriptor" id="ref-for-dictdef-mloperanddescriptor②①">MLOperandDescriptor</a></code>.</p>
+      <li data-md>
+       <p>Set <var>desc</var>.<code class="idl"><a data-link-type="idl" href="#dom-mloperanddescriptor-dimensions" id="ref-for-dom-mloperanddescriptor-dimensions⑦⑥">dimensions</a></code> to the result of invoking the <a data-link-type="dfn" href="#mlgraphbuilder-calculate-matmul-output-sizes" id="ref-for-mlgraphbuilder-calculate-matmul-output-sizes">calculate matmul output sizes</a> steps given <var>a</var> and <var>b</var>.</p>
+      <li data-md>
+       <p>If that throws an error, re-<a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#dfn-throw" id="ref-for-dfn-throw①⑨③">throw</a> the error.</p>
+      <li data-md>
+       <p>Set <var>desc</var>.<code class="idl"><a data-link-type="idl" href="#dom-mloperanddescriptor-type" id="ref-for-dom-mloperanddescriptor-type③③">type</a></code> to <var>a</var>.<code class="idl"><a data-link-type="idl" href="#dom-mloperand-descriptor-slot" id="ref-for-dom-mloperand-descriptor-slot⑦⑦">[[descriptor]]</a></code>.<code class="idl"><a data-link-type="idl" href="#dom-mloperanddescriptor-type" id="ref-for-dom-mloperanddescriptor-type③④">type</a></code>.</p>
+      <li data-md>
+       <p>If any of the following sub-steps fail, <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#dfn-throw" id="ref-for-dfn-throw①⑨④">throw</a> an "<code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#operationerror" id="ref-for-operationerror③④">OperationError</a></code>" <code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#idl-DOMException" id="ref-for-idl-DOMException①①⑤">DOMException</a></code>.</p>
+       <ol>
+        <li data-md>
+         <p>Let <var>output</var> be the result of <a data-link-type="dfn" href="#create-an-mloperand" id="ref-for-create-an-mloperand①⑤">creating an MLOperand</a> given <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#this" id="ref-for-this②⑨">this</a> and <var>desc</var>.</p>
+        <li data-md>
+         <p>Make a request to the underlying platform to:</p>
+         <ol>
+          <li data-md>
+           <p>Let <var>opImpl</var> be an <a data-link-type="dfn" href="https://infra.spec.whatwg.org/#implementation-defined" id="ref-for-implementation-defined④④">implementation-defined</a> platform operator for the matrix multiplication operation.</p>
+          <li data-md>
+           <p>Store a reference of <var>opImpl</var> in <var>output</var>.<code class="idl"><a data-link-type="idl" href="#dom-mloperand-operator-slot" id="ref-for-dom-mloperand-operator-slot②⓪">[[operator]]</a></code>.</p>
+          <li data-md>
+           <p>Create an <a data-link-type="dfn" href="https://infra.spec.whatwg.org/#implementation-defined" id="ref-for-implementation-defined④⑤">implementation-defined</a> platform operand <var>outputImpl</var> to represent the output, given <var>output</var> and <var>opImpl</var>.</p>
+          <li data-md>
+           <p>Store a reference to <var>outputImpl</var> in <var>output</var>.<code class="idl"><a data-link-type="idl" href="#dom-mloperand-operand-slot" id="ref-for-dom-mloperand-operand-slot⑤⑤">[[operand]]</a></code>.</p>
+         </ol>
+        <li data-md>
+         <p>Connect <var>a</var>.<code class="idl"><a data-link-type="idl" href="#dom-mloperand-operand-slot" id="ref-for-dom-mloperand-operand-slot⑤⑥">[[operand]]</a></code> and <var>b</var>.<code class="idl"><a data-link-type="idl" href="#dom-mloperand-operand-slot" id="ref-for-dom-mloperand-operand-slot⑤⑦">[[operand]]</a></code> as inputs to <var>opImpl</var>.</p>
+        <li data-md>
+         <p>Connect <var>output</var>.<code class="idl"><a data-link-type="idl" href="#dom-mloperand-operand-slot" id="ref-for-dom-mloperand-operand-slot⑤⑧">[[operand]]</a></code> as output to <var>opImpl</var>.</p>
+       </ol>
+      <li data-md>
+       <p>Return <var>output</var>.</p>
+     </ol>
     </div>
-    <h4 class="heading settled" data-level="7.6.23" id="api-mlgraphbuilder-pad"><span class="secno">7.6.23. </span><span class="content">The pad() method</span><a class="self-link" href="#api-mlgraphbuilder-pad"></a></h4>
-     Inflate the tensor with constant or mirrored values on the edges. 
+   </details>
+   <h4 class="heading settled" data-level="7.6.24" id="api-mlgraphbuilder-pad"><span class="secno">7.6.24. </span><span class="content">The pad() method</span><a class="self-link" href="#api-mlgraphbuilder-pad"></a></h4>
+    Inflate the tensor with constant or mirrored values on the edges. 
 <pre class="idl highlight def"><c- b>enum</c-> <dfn class="dfn-paneled idl-code" data-dfn-type="enum" data-export id="enumdef-mlpaddingmode"><code><c- g>MLPaddingMode</c-></code></dfn> {
   <dfn class="idl-code" data-dfn-for="MLPaddingMode" data-dfn-type="enum-value" data-export id="dom-mlpaddingmode-constant"><code><c- s>"constant"</c-></code><a class="self-link" href="#dom-mlpaddingmode-constant"></a></dfn>,
   <dfn class="idl-code" data-dfn-for="MLPaddingMode" data-dfn-type="enum-value" data-export id="dom-mlpaddingmode-edge"><code><c- s>"edge"</c-></code><a class="self-link" href="#dom-mlpaddingmode-edge"></a></dfn>,
@@ -3535,39 +5767,107 @@ <h4 class="heading settled" data-level="7.6.23" id="api-mlgraphbuilder-pad"><spa
 };
 
 <c- b>dictionary</c-> <dfn class="dfn-paneled idl-code" data-dfn-type="dictionary" data-export id="dictdef-mlpadoptions"><code><c- g>MLPadOptions</c-></code></dfn> {
-  <a data-link-type="idl-name" href="#enumdef-mlpaddingmode" id="ref-for-enumdef-mlpaddingmode"><c- n>MLPaddingMode</c-></a> <dfn class="idl-code" data-default="&quot;constant&quot;" data-dfn-for="MLPadOptions" data-dfn-type="dict-member" data-export data-type="MLPaddingMode " id="dom-mlpadoptions-mode"><code><c- g>mode</c-></code><a class="self-link" href="#dom-mlpadoptions-mode"></a></dfn> = "constant";
-  <a class="idl-code" data-link-type="interface" href="https://webidl.spec.whatwg.org/#idl-float" id="ref-for-idl-float②④"><c- b>float</c-></a> <dfn class="idl-code" data-default="0" data-dfn-for="MLPadOptions" data-dfn-type="dict-member" data-export data-type="float " id="dom-mlpadoptions-value"><code><c- g>value</c-></code><a class="self-link" href="#dom-mlpadoptions-value"></a></dfn> = 0;
+  <a data-link-type="idl-name" href="#enumdef-mlpaddingmode" id="ref-for-enumdef-mlpaddingmode"><c- n>MLPaddingMode</c-></a> <a class="idl-code" data-default="&quot;constant&quot;" data-link-type="dict-member" data-type="MLPaddingMode " href="#dom-mlpadoptions-mode" id="ref-for-dom-mlpadoptions-mode"><c- g>mode</c-></a> = "constant";
+  <a class="idl-code" data-link-type="interface" href="https://webidl.spec.whatwg.org/#idl-float" id="ref-for-idl-float③⑥"><c- b>float</c-></a> <a class="idl-code" data-default="0" data-link-type="dict-member" data-type="float " href="#dom-mlpadoptions-value" id="ref-for-dom-mlpadoptions-value"><c- g>value</c-></a> = 0;
 };
 
-<c- b>partial</c-> <c- b>interface</c-> <a class="idl-code" data-link-type="interface" href="#mlgraphbuilder" id="ref-for-mlgraphbuilder③⑨"><c- g>MLGraphBuilder</c-></a> {
-  <a data-link-type="idl-name" href="#mloperand" id="ref-for-mloperand②①⑦"><c- n>MLOperand</c-></a> <dfn class="idl-code" data-dfn-for="MLGraphBuilder" data-dfn-type="method" data-export data-lt="pad(input, beginningPadding, endingPadding, options)|pad(input, beginningPadding, endingPadding)" id="dom-mlgraphbuilder-pad"><code><c- g>pad</c-></code><a class="self-link" href="#dom-mlgraphbuilder-pad"></a></dfn>(<a data-link-type="idl-name" href="#mloperand" id="ref-for-mloperand②①⑧"><c- n>MLOperand</c-></a> <dfn class="idl-code" data-dfn-for="MLGraphBuilder/pad(input, beginningPadding, endingPadding, options), MLGraphBuilder/pad(input, beginningPadding, endingPadding)" data-dfn-type="argument" data-export id="dom-mlgraphbuilder-pad-input-beginningpadding-endingpadding-options-input"><code><c- g>input</c-></code><a class="self-link" href="#dom-mlgraphbuilder-pad-input-beginningpadding-endingpadding-options-input"></a></dfn>,
-                <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#idl-sequence" id="ref-for-idl-sequence①⑦"><c- b>sequence</c-></a>&lt;<a class="idl-code" data-link-type="interface" href="https://webidl.spec.whatwg.org/#idl-unsigned-long" id="ref-for-idl-unsigned-long③⑦"><c- b>unsigned</c-> <c- b>long</c-></a>> <dfn class="idl-code" data-dfn-for="MLGraphBuilder/pad(input, beginningPadding, endingPadding, options), MLGraphBuilder/pad(input, beginningPadding, endingPadding)" data-dfn-type="argument" data-export id="dom-mlgraphbuilder-pad-input-beginningpadding-endingpadding-options-beginningpadding"><code><c- g>beginningPadding</c-></code><a class="self-link" href="#dom-mlgraphbuilder-pad-input-beginningpadding-endingpadding-options-beginningpadding"></a></dfn>,
-                <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#idl-sequence" id="ref-for-idl-sequence①⑧"><c- b>sequence</c-></a>&lt;<a class="idl-code" data-link-type="interface" href="https://webidl.spec.whatwg.org/#idl-unsigned-long" id="ref-for-idl-unsigned-long③⑧"><c- b>unsigned</c-> <c- b>long</c-></a>> <dfn class="idl-code" data-dfn-for="MLGraphBuilder/pad(input, beginningPadding, endingPadding, options), MLGraphBuilder/pad(input, beginningPadding, endingPadding)" data-dfn-type="argument" data-export id="dom-mlgraphbuilder-pad-input-beginningpadding-endingpadding-options-endingpadding"><code><c- g>endingPadding</c-></code><a class="self-link" href="#dom-mlgraphbuilder-pad-input-beginningpadding-endingpadding-options-endingpadding"></a></dfn>,
+<c- b>partial</c-> <c- b>interface</c-> <a class="idl-code" data-link-type="interface" href="#mlgraphbuilder" id="ref-for-mlgraphbuilder④②"><c- g>MLGraphBuilder</c-></a> {
+  <a data-link-type="idl-name" href="#mloperand" id="ref-for-mloperand②⑦③"><c- n>MLOperand</c-></a> <a class="idl-code" data-link-type="method" href="#dom-mlgraphbuilder-pad" id="ref-for-dom-mlgraphbuilder-pad"><c- g>pad</c-></a>(<a data-link-type="idl-name" href="#mloperand" id="ref-for-mloperand②⑦④"><c- n>MLOperand</c-></a> <dfn class="idl-code" data-dfn-for="MLGraphBuilder/pad(input, beginningPadding, endingPadding, options), MLGraphBuilder/pad(input, beginningPadding, endingPadding)" data-dfn-type="argument" data-export id="dom-mlgraphbuilder-pad-input-beginningpadding-endingpadding-options-input"><code><c- g>input</c-></code><a class="self-link" href="#dom-mlgraphbuilder-pad-input-beginningpadding-endingpadding-options-input"></a></dfn>,
+                <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#idl-sequence" id="ref-for-idl-sequence①⑦"><c- b>sequence</c-></a>&lt;<a class="idl-code" data-link-type="interface" href="https://webidl.spec.whatwg.org/#idl-unsigned-long" id="ref-for-idl-unsigned-long③⑨"><c- b>unsigned</c-> <c- b>long</c-></a>> <dfn class="idl-code" data-dfn-for="MLGraphBuilder/pad(input, beginningPadding, endingPadding, options), MLGraphBuilder/pad(input, beginningPadding, endingPadding)" data-dfn-type="argument" data-export id="dom-mlgraphbuilder-pad-input-beginningpadding-endingpadding-options-beginningpadding"><code><c- g>beginningPadding</c-></code><a class="self-link" href="#dom-mlgraphbuilder-pad-input-beginningpadding-endingpadding-options-beginningpadding"></a></dfn>,
+                <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#idl-sequence" id="ref-for-idl-sequence①⑧"><c- b>sequence</c-></a>&lt;<a class="idl-code" data-link-type="interface" href="https://webidl.spec.whatwg.org/#idl-unsigned-long" id="ref-for-idl-unsigned-long④⓪"><c- b>unsigned</c-> <c- b>long</c-></a>> <dfn class="idl-code" data-dfn-for="MLGraphBuilder/pad(input, beginningPadding, endingPadding, options), MLGraphBuilder/pad(input, beginningPadding, endingPadding)" data-dfn-type="argument" data-export id="dom-mlgraphbuilder-pad-input-beginningpadding-endingpadding-options-endingpadding"><code><c- g>endingPadding</c-></code><a class="self-link" href="#dom-mlgraphbuilder-pad-input-beginningpadding-endingpadding-options-endingpadding"></a></dfn>,
                 <c- b>optional</c-> <a data-link-type="idl-name" href="#dictdef-mlpadoptions" id="ref-for-dictdef-mlpadoptions"><c- n>MLPadOptions</c-></a> <dfn class="idl-code" data-dfn-for="MLGraphBuilder/pad(input, beginningPadding, endingPadding, options), MLGraphBuilder/pad(input, beginningPadding, endingPadding)" data-dfn-type="argument" data-export id="dom-mlgraphbuilder-pad-input-beginningpadding-endingpadding-options-options"><code><c- g>options</c-></code><a class="self-link" href="#dom-mlgraphbuilder-pad-input-beginningpadding-endingpadding-options-options"></a></dfn> = {});
 };
 </pre>
-    <div class="algorithm" data-algorithm="pad">
-      <strong>Arguments:</strong> 
-     <ul>
+   <p><code class="idl"><a data-link-type="idl" href="#dictdef-mlpadoptions" id="ref-for-dictdef-mlpadoptions①">MLPadOptions</a></code> has the following members:</p>
+   <dl>
+    <dt data-md><dfn class="dfn-paneled idl-code" data-dfn-for="MLPadOptions" data-dfn-type="dict-member" data-export id="dom-mlpadoptions-mode"><code>mode</code></dfn>, <span> of type <a data-link-type="idl-name" href="#enumdef-mlpaddingmode" id="ref-for-enumdef-mlpaddingmode①">MLPaddingMode</a>, defaulting to <code>"constant"</code></span>
+    <dd data-md>
+     <p>An <code class="idl"><a data-link-type="idl" href="#enumdef-mlpaddingmode" id="ref-for-enumdef-mlpaddingmode②">MLPaddingMode</a></code> <a data-link-type="dfn" href="https://infra.spec.whatwg.org/#string" id="ref-for-string①⓪">string</a>.
+Specifies the different ways to pad the tensor.
+The default value is <code>"constant"</code>.</p>
+    <dt data-md><dfn class="dfn-paneled idl-code" data-dfn-for="MLPadOptions" data-dfn-type="dict-member" data-export id="dom-mlpadoptions-value"><code>value</code></dfn>, <span> of type <a data-link-type="idl-name" href="https://webidl.spec.whatwg.org/#idl-float" id="ref-for-idl-float③⑦">float</a>, defaulting to <code>0</code></span>
+    <dd data-md>
+     <p>A <code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#idl-float" id="ref-for-idl-float③⑧">float</a></code>.
+Specifies the padding value when <code class="idl"><a data-link-type="idl" href="#dom-mlpadoptions-mode" id="ref-for-dom-mlpadoptions-mode①">mode</a></code> is set to <code>"constant"</code>.
+The default value is 0.</p>
+   </dl>
+   <div>
+     <strong>Arguments:</strong> 
+    <ul>
+     <li data-md>
+      <p><em>input</em>: an <code class="idl"><a data-link-type="idl" href="#mloperand" id="ref-for-mloperand②⑦⑤">MLOperand</a></code>. The input tensor.</p>
+     <li data-md>
+      <p><em>beginningPadding</em>: a sequence of <code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#idl-unsigned-long" id="ref-for-idl-unsigned-long④①">unsigned long</a></code>. The sequence of unsigned integer values indicating the number of padding values to add at the beginning of each input dimension, of length <em>N</em> where <em>N</em> is the <a data-link-type="dfn" href="#mloperand-rank" id="ref-for-mloperand-rank①⑤">rank</a> of the input tensor. For each dimension <em>d</em> of <em>input</em>, <em>beginningPadding[d]</em> indicates how many values to add before the content in that dimension.</p>
+     <li data-md>
+      <p><em>endingPadding</em>: a sequence of <code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#idl-unsigned-long" id="ref-for-idl-unsigned-long④②">unsigned long</a></code>. The sequence of unsigned integer values indicating the number of padding values to add at the ending of each input dimension, of length <em>N</em> where <em>N</em> is the <a data-link-type="dfn" href="#mloperand-rank" id="ref-for-mloperand-rank①⑥">rank</a> of the input tensor. For each dimension <em>d</em> of <em>input</em>, <em>endingPadding[d]</em> indicates how many values to add after the content in that dimension.</p>
+     <li data-md>
+      <p><em>options</em>: an optional <code class="idl"><a data-link-type="idl" href="#dictdef-mlpadoptions" id="ref-for-dictdef-mlpadoptions②">MLPadOptions</a></code>. The optional parameters of the operation.</p>
+    </ul>
+    <p><strong>Returns:</strong> an <code class="idl"><a data-link-type="idl" href="#mloperand" id="ref-for-mloperand②⑦⑥">MLOperand</a></code>. The padded output tensor. Each dimension of the output tensor can be calculated as follow:</p>
+    <p><em>output size = beginning padding + input size + ending padding</em></p>
+   </div>
+   <details class="algorithm" data-algorithm="calculate padding output sizes" data-algorithm-for="MLGraphBuilder" open>
+    <summary> To <dfn class="dfn-paneled" data-dfn-for="MLGraphBuilder" data-dfn-type="dfn" data-noexport id="mlgraphbuilder-calculate-padding-output-sizes">calculate padding output sizes</dfn>, given <var>input</var>, <var>beginningPadding</var> and <var>endingPadding</var>, run the following steps: </summary>
+    <div class="algorithm-steps">
+     <ol>
+      <li data-md>
+       <p>Let <var>shape</var> be a copy of <var>input</var>.<code class="idl"><a data-link-type="idl" href="#dom-mloperand-descriptor-slot" id="ref-for-dom-mloperand-descriptor-slot⑦⑧">[[descriptor]]</a></code>.<code class="idl"><a data-link-type="idl" href="#dom-mloperanddescriptor-dimensions" id="ref-for-dom-mloperanddescriptor-dimensions⑦⑦">dimensions</a></code>.</p>
+      <li data-md>
+       <p>For <var>index</var> in <a data-link-type="dfn" href="https://infra.spec.whatwg.org/#the-range" id="ref-for-the-range④">the range</a> 0 to the <a data-link-type="dfn" href="#mloperand-rank" id="ref-for-mloperand-rank①⑦">rank</a> of <var>shape</var>, exclusive:</p>
+       <ol>
+        <li data-md>
+         <p>Add to <var>shape</var>[<var>index</var>] the value of <var>beginningPadding</var>[<var>index</var>].</p>
+        <li data-md>
+         <p>Add to <var>shape</var>[<var>index</var>] the value of <var>endingPadding</var>[<var>index</var>].</p>
+       </ol>
       <li data-md>
-       <p><em>input</em>: an <code class="idl"><a data-link-type="idl" href="#mloperand" id="ref-for-mloperand②①⑨">MLOperand</a></code>. The input tensor.</p>
+       <p>Return <var>shape</var>.</p>
+     </ol>
+    </div>
+   </details>
+   <details class="algorithm" data-algorithm="pad(input, beginningPadding, endingPadding, options)" data-algorithm-for="MLGraphBuilder" open>
+    <summary> The <dfn class="dfn-paneled idl-code" data-dfn-for="MLGraphBuilder" data-dfn-type="method" data-export data-lt="pad(input, beginningPadding, endingPadding, options)|pad(input, beginningPadding, endingPadding)" id="dom-mlgraphbuilder-pad"><code>pad(<var>input</var>, <var>beginningPadding</var>, <var>endingPadding</var>, <var>options</var>)</code></dfn> method steps are: </summary>
+    <div class="algorithm-steps">
+     <ol>
       <li data-md>
-       <p><em>beginningPadding</em>: a sequence of <code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#idl-unsigned-long" id="ref-for-idl-unsigned-long③⑨">unsigned long</a></code>. The sequence of unsigned integer values indicating the number of padding values to add at the beginning of each input dimension, of length <em>N</em> where <em>N</em> is the rank of the input tensor. For each dimension <em>d</em> of <em>input</em>, <em>beginningPadding[d]</em> indicates how many values to add before the content in that dimension.</p>
+       <p><a data-link-type="dfn" href="https://infra.spec.whatwg.org/#assert" id="ref-for-assert⑤⑦">Assert</a>: the type of <var>input</var> is <code class="idl"><a data-link-type="idl" href="#mloperand" id="ref-for-mloperand②⑦⑦">MLOperand</a></code>.</p>
       <li data-md>
-       <p><em>endingPadding</em>: a sequence of <code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#idl-unsigned-long" id="ref-for-idl-unsigned-long④⓪">unsigned long</a></code>. The sequence of unsigned integer values indicating the number of padding values to add at the ending of each input dimension, of length <em>N</em> where <em>N</em> is the rank of the input tensor. For each dimension <em>d</em> of <em>input</em>, <em>endingPadding[d]</em> indicates how many values to add after the content in that dimension.</p>
+       <p>If the <a data-link-type="dfn" href="https://infra.spec.whatwg.org/#list-size" id="ref-for-list-size③⑧">size</a> of <var>beginningPadding</var> and <var>endingPadding</var> is not equal to the <a data-link-type="dfn" href="#mloperand-rank" id="ref-for-mloperand-rank①⑧">rank</a> of <var>input</var>, then then <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#dfn-throw" id="ref-for-dfn-throw①⑨⑤">throw</a> a "<code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#exceptiondef-typeerror" id="ref-for-exceptiondef-typeerror③⑧">TypeError</a></code>".</p>
       <li data-md>
-       <p><em>options</em>: an optional <code class="idl"><a data-link-type="idl" href="#dictdef-mlpadoptions" id="ref-for-dictdef-mlpadoptions①">MLPadOptions</a></code>. The optional parameters of the operation.</p>
-       <ul>
+       <p>Let <var>desc</var> be a copy of <var>input</var>.<code class="idl"><a data-link-type="idl" href="#dom-mloperand-descriptor-slot" id="ref-for-dom-mloperand-descriptor-slot⑦⑨">[[descriptor]]</a></code>.</p>
+      <li data-md>
+       <p>Set <var>desc</var>.<code class="idl"><a data-link-type="idl" href="#dom-mloperanddescriptor-dimensions" id="ref-for-dom-mloperanddescriptor-dimensions⑦⑧">dimensions</a></code> to the result of invoking the <a data-link-type="dfn" href="#mlgraphbuilder-calculate-padding-output-sizes" id="ref-for-mlgraphbuilder-calculate-padding-output-sizes">calculate padding output sizes</a> steps given <var>input</var>, <var>beginningPadding</var> and <var>endingPadding</var>.</p>
+      <li data-md>
+       <p>If any of the following sub-steps fail, <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#dfn-throw" id="ref-for-dfn-throw①⑨⑥">throw</a> an "<code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#operationerror" id="ref-for-operationerror③⑤">OperationError</a></code>" <code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#idl-DOMException" id="ref-for-idl-DOMException①①⑥">DOMException</a></code>.</p>
+       <ol>
         <li data-md>
-         <p><em>mode</em>: an <code class="idl"><a data-link-type="idl" href="#enumdef-mlpaddingmode" id="ref-for-enumdef-mlpaddingmode①">MLPaddingMode</a></code>. The different ways to pad the tensor. When not set, it’s assumed to be "constant".</p>
+         <p>Let <var>output</var> be the result of <a data-link-type="dfn" href="#create-an-mloperand" id="ref-for-create-an-mloperand①⑥">creating an MLOperand</a> given <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#this" id="ref-for-this③⓪">this</a> and <var>desc</var>.</p>
         <li data-md>
-         <p><em>value</em>: a <code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#idl-float" id="ref-for-idl-float②⑤">float</a></code>. The pad value when the <em>options.mode</em> is set to <em>"constant"</em>. When not set, it’s assumed to be 0.</p>
-       </ul>
-     </ul>
-     <p><strong>Returns:</strong> an <code class="idl"><a data-link-type="idl" href="#mloperand" id="ref-for-mloperand②②⓪">MLOperand</a></code>. The padded output tensor. Each dimension of the output tensor can be calculated as follow:</p>
-     <p><em>output size = beginning padding + input size + ending padding</em></p>
-     <div class="example" id="example-69d64c7d">
-      <a class="self-link" href="#example-69d64c7d"></a> 
+         <p>Make a request to the underlying platform to:</p>
+         <ol>
+          <li data-md>
+           <p>Let <var>opImpl</var> be an <a data-link-type="dfn" href="https://infra.spec.whatwg.org/#implementation-defined" id="ref-for-implementation-defined④⑥">implementation-defined</a> platform operator for the padding operation, given <var>beginningPadding</var>, <var>endingPadding</var> and <var>options</var>.</p>
+          <li data-md>
+           <p>Store a reference of <var>opImpl</var> in <var>output</var>.<code class="idl"><a data-link-type="idl" href="#dom-mloperand-operator-slot" id="ref-for-dom-mloperand-operator-slot②①">[[operator]]</a></code>.</p>
+          <li data-md>
+           <p>Create an <a data-link-type="dfn" href="https://infra.spec.whatwg.org/#implementation-defined" id="ref-for-implementation-defined④⑦">implementation-defined</a> platform operand <var>outputImpl</var> to represent the output, given <var>output</var> and <var>opImpl</var>.</p>
+          <li data-md>
+           <p>Store a reference to <var>outputImpl</var> in <var>output</var>.<code class="idl"><a data-link-type="idl" href="#dom-mloperand-operand-slot" id="ref-for-dom-mloperand-operand-slot⑤⑨">[[operand]]</a></code>.</p>
+         </ol>
+        <li data-md>
+         <p>Connect <var>input</var>.<code class="idl"><a data-link-type="idl" href="#dom-mloperand-operand-slot" id="ref-for-dom-mloperand-operand-slot⑥⓪">[[operand]]</a></code> as input to <var>opImpl</var>.</p>
+        <li data-md>
+         <p>Connect <var>output</var>.<code class="idl"><a data-link-type="idl" href="#dom-mloperand-operand-slot" id="ref-for-dom-mloperand-operand-slot⑥①">[[operand]]</a></code> as output to <var>opImpl</var>.</p>
+       </ol>
+      <li data-md>
+       <p>Return <var>output</var>.</p>
+     </ol>
+    </div>
+   </details>
+   <div class="example" id="example-6180d419">
+    <a class="self-link" href="#example-6180d419"></a> 
+    <details open>
+     <summary> Examples for constant, edge, reflection and symmetric padding: </summary>
 <pre class="highlight"><c- c1>// input: [[1,2,3], [4,5,6]]</c->
 <c- a>const</c-> input <c- o>=</c-> builder<c- p>.</c->constant<c- p>(</c->
   <c- p>{</c-> type<c- o>:</c-> <c- t>'float32'</c-><c- p>,</c-> dimensions<c- o>:</c-> <c- p>[</c-><c- mf>2</c-><c- p>,</c-><c- mf>3</c-><c- p>]</c-> <c- p>},</c-> <c- ow>new</c-> Float32Array<c- p>([</c-><c- mf>1</c-><c- p>,</c-><c- mf>2</c-><c- p>,</c-><c- mf>3</c-><c- p>,</c-><c- mf>4</c-><c- p>,</c-><c- mf>5</c-><c- p>,</c-><c- mf>6</c-><c- p>]));</c->
@@ -3603,350 +5903,1003 @@ <h4 class="heading settled" data-level="7.6.23" id="api-mlgraphbuilder-pad"><spa
 <c- c1>//     [5,4,4,5,6,6,5]]</c->
 builder<c- p>.</c->pad<c- p>(</c->input<c- p>,</c-> beginningPadding<c- p>,</c-> endingPadding<c- p>,</c-> <c- p>{</c-> mode<c- o>:</c-> <c- u>"symmetric"</c-> <c- p>});</c->
 </pre>
-     </div>
-    </div>
-    <h4 class="heading settled" data-level="7.6.24" id="api-mlgraphbuilder-pool2d"><span class="secno">7.6.24. </span><span class="content">Pooling operations</span><a class="self-link" href="#api-mlgraphbuilder-pool2d"></a></h4>
-     Compute a <em>mean</em>, <em>L2 norm</em>, or <em>max</em> reduction operation across all the elements within the moving window over the input tensor. See the description of each type of reduction in <a href="#api-mlgraphbuilder-reduce">§ 7.6.26 Reduction operations</a>. 
+    </details>
+   </div>
+   <h4 class="heading settled" data-level="7.6.25" id="api-mlgraphbuilder-pool2d"><span class="secno">7.6.25. </span><span class="content">Pooling operations</span><a class="self-link" href="#api-mlgraphbuilder-pool2d"></a></h4>
+    Compute a <em>mean</em>, <em>L2 norm</em>, or <em>max</em> reduction operation across all the elements within the moving window over the input tensor. See the description of each type of reduction in <a href="#api-mlgraphbuilder-reduce">§ 7.6.27 Reduction operations</a>. 
 <pre class="idl highlight def"><c- b>enum</c-> <dfn class="dfn-paneled idl-code" data-dfn-type="enum" data-export id="enumdef-mlroundingtype"><code><c- g>MLRoundingType</c-></code></dfn> {
   <dfn class="idl-code" data-dfn-for="MLRoundingType" data-dfn-type="enum-value" data-export id="dom-mlroundingtype-floor"><code><c- s>"floor"</c-></code><a class="self-link" href="#dom-mlroundingtype-floor"></a></dfn>,
   <dfn class="idl-code" data-dfn-for="MLRoundingType" data-dfn-type="enum-value" data-export id="dom-mlroundingtype-ceil"><code><c- s>"ceil"</c-></code><a class="self-link" href="#dom-mlroundingtype-ceil"></a></dfn>
 };
 
 <c- b>dictionary</c-> <dfn class="dfn-paneled idl-code" data-dfn-type="dictionary" data-export id="dictdef-mlpool2doptions"><code><c- g>MLPool2dOptions</c-></code></dfn> {
-  <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#idl-sequence" id="ref-for-idl-sequence①⑨"><c- b>sequence</c-></a>&lt;<a class="idl-code" data-link-type="interface" href="https://webidl.spec.whatwg.org/#idl-unsigned-long" id="ref-for-idl-unsigned-long④①"><c- b>unsigned</c-> <c- b>long</c-></a>> <dfn class="idl-code" data-dfn-for="MLPool2dOptions" data-dfn-type="dict-member" data-export data-type="sequence<unsigned long> " id="dom-mlpool2doptions-windowdimensions"><code><c- g>windowDimensions</c-></code><a class="self-link" href="#dom-mlpool2doptions-windowdimensions"></a></dfn>;
-  <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#idl-sequence" id="ref-for-idl-sequence②⓪"><c- b>sequence</c-></a>&lt;<a class="idl-code" data-link-type="interface" href="https://webidl.spec.whatwg.org/#idl-unsigned-long" id="ref-for-idl-unsigned-long④②"><c- b>unsigned</c-> <c- b>long</c-></a>> <dfn class="idl-code" data-dfn-for="MLPool2dOptions" data-dfn-type="dict-member" data-export data-type="sequence<unsigned long> " id="dom-mlpool2doptions-padding"><code><c- g>padding</c-></code><a class="self-link" href="#dom-mlpool2doptions-padding"></a></dfn>;
-  <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#idl-sequence" id="ref-for-idl-sequence②①"><c- b>sequence</c-></a>&lt;<a class="idl-code" data-link-type="interface" href="https://webidl.spec.whatwg.org/#idl-unsigned-long" id="ref-for-idl-unsigned-long④③"><c- b>unsigned</c-> <c- b>long</c-></a>> <dfn class="idl-code" data-dfn-for="MLPool2dOptions" data-dfn-type="dict-member" data-export data-type="sequence<unsigned long> " id="dom-mlpool2doptions-strides"><code><c- g>strides</c-></code><a class="self-link" href="#dom-mlpool2doptions-strides"></a></dfn>;
-  <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#idl-sequence" id="ref-for-idl-sequence②②"><c- b>sequence</c-></a>&lt;<a class="idl-code" data-link-type="interface" href="https://webidl.spec.whatwg.org/#idl-unsigned-long" id="ref-for-idl-unsigned-long④④"><c- b>unsigned</c-> <c- b>long</c-></a>> <dfn class="idl-code" data-dfn-for="MLPool2dOptions" data-dfn-type="dict-member" data-export data-type="sequence<unsigned long> " id="dom-mlpool2doptions-dilations"><code><c- g>dilations</c-></code><a class="self-link" href="#dom-mlpool2doptions-dilations"></a></dfn>;
-  <a data-link-type="idl-name" href="#enumdef-mlautopad" id="ref-for-enumdef-mlautopad④"><c- n>MLAutoPad</c-></a> <dfn class="idl-code" data-default="&quot;explicit&quot;" data-dfn-for="MLPool2dOptions" data-dfn-type="dict-member" data-export data-type="MLAutoPad " id="dom-mlpool2doptions-autopad"><code><c- g>autoPad</c-></code><a class="self-link" href="#dom-mlpool2doptions-autopad"></a></dfn> = "explicit";
-  <a data-link-type="idl-name" href="#enumdef-mlinputoperandlayout" id="ref-for-enumdef-mlinputoperandlayout⑥"><c- n>MLInputOperandLayout</c-></a> <dfn class="idl-code" data-default="&quot;nchw&quot;" data-dfn-for="MLPool2dOptions" data-dfn-type="dict-member" data-export data-type="MLInputOperandLayout " id="dom-mlpool2doptions-layout"><code><c- g>layout</c-></code><a class="self-link" href="#dom-mlpool2doptions-layout"></a></dfn> = "nchw";
-  <a data-link-type="idl-name" href="#enumdef-mlroundingtype" id="ref-for-enumdef-mlroundingtype"><c- n>MLRoundingType</c-></a> <dfn class="idl-code" data-default="&quot;floor&quot;" data-dfn-for="MLPool2dOptions" data-dfn-type="dict-member" data-export data-type="MLRoundingType " id="dom-mlpool2doptions-roundingtype"><code><c- g>roundingType</c-></code><a class="self-link" href="#dom-mlpool2doptions-roundingtype"></a></dfn> = "floor";
-  <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#idl-sequence" id="ref-for-idl-sequence②③"><c- b>sequence</c-></a>&lt;<a class="idl-code" data-link-type="interface" href="https://webidl.spec.whatwg.org/#idl-unsigned-long" id="ref-for-idl-unsigned-long④⑤"><c- b>unsigned</c-> <c- b>long</c-></a>> <dfn class="idl-code" data-dfn-for="MLPool2dOptions" data-dfn-type="dict-member" data-export data-type="sequence<unsigned long> " id="dom-mlpool2doptions-outputsizes"><code><c- g>outputSizes</c-></code><a class="self-link" href="#dom-mlpool2doptions-outputsizes"></a></dfn>;
+  <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#idl-sequence" id="ref-for-idl-sequence①⑨"><c- b>sequence</c-></a>&lt;<a class="idl-code" data-link-type="interface" href="https://webidl.spec.whatwg.org/#idl-unsigned-long" id="ref-for-idl-unsigned-long④③"><c- b>unsigned</c-> <c- b>long</c-></a>> <a class="idl-code" data-link-type="dict-member" data-type="sequence<unsigned long> " href="#dom-mlpool2doptions-windowdimensions" id="ref-for-dom-mlpool2doptions-windowdimensions"><c- g>windowDimensions</c-></a>;
+  <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#idl-sequence" id="ref-for-idl-sequence②⓪"><c- b>sequence</c-></a>&lt;<a class="idl-code" data-link-type="interface" href="https://webidl.spec.whatwg.org/#idl-unsigned-long" id="ref-for-idl-unsigned-long④④"><c- b>unsigned</c-> <c- b>long</c-></a>> <a class="idl-code" data-link-type="dict-member" data-type="sequence<unsigned long> " href="#dom-mlpool2doptions-padding" id="ref-for-dom-mlpool2doptions-padding"><c- g>padding</c-></a>;
+  <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#idl-sequence" id="ref-for-idl-sequence②①"><c- b>sequence</c-></a>&lt;<a class="idl-code" data-link-type="interface" href="https://webidl.spec.whatwg.org/#idl-unsigned-long" id="ref-for-idl-unsigned-long④⑤"><c- b>unsigned</c-> <c- b>long</c-></a>> <a class="idl-code" data-link-type="dict-member" data-type="sequence<unsigned long> " href="#dom-mlpool2doptions-strides" id="ref-for-dom-mlpool2doptions-strides"><c- g>strides</c-></a>;
+  <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#idl-sequence" id="ref-for-idl-sequence②②"><c- b>sequence</c-></a>&lt;<a class="idl-code" data-link-type="interface" href="https://webidl.spec.whatwg.org/#idl-unsigned-long" id="ref-for-idl-unsigned-long④⑥"><c- b>unsigned</c-> <c- b>long</c-></a>> <a class="idl-code" data-link-type="dict-member" data-type="sequence<unsigned long> " href="#dom-mlpool2doptions-dilations" id="ref-for-dom-mlpool2doptions-dilations"><c- g>dilations</c-></a>;
+  <a data-link-type="idl-name" href="#enumdef-mlautopad" id="ref-for-enumdef-mlautopad⑥"><c- n>MLAutoPad</c-></a> <a class="idl-code" data-default="&quot;explicit&quot;" data-link-type="dict-member" data-type="MLAutoPad " href="#dom-mlpool2doptions-autopad" id="ref-for-dom-mlpool2doptions-autopad"><c- g>autoPad</c-></a> = "explicit";
+  <a data-link-type="idl-name" href="#enumdef-mlinputoperandlayout" id="ref-for-enumdef-mlinputoperandlayout⑨"><c- n>MLInputOperandLayout</c-></a> <a class="idl-code" data-default="&quot;nchw&quot;" data-link-type="dict-member" data-type="MLInputOperandLayout " href="#dom-mlpool2doptions-layout" id="ref-for-dom-mlpool2doptions-layout"><c- g>layout</c-></a> = "nchw";
+  <a data-link-type="idl-name" href="#enumdef-mlroundingtype" id="ref-for-enumdef-mlroundingtype"><c- n>MLRoundingType</c-></a> <a class="idl-code" data-default="&quot;floor&quot;" data-link-type="dict-member" data-type="MLRoundingType " href="#dom-mlpool2doptions-roundingtype" id="ref-for-dom-mlpool2doptions-roundingtype"><c- g>roundingType</c-></a> = "floor";
+  <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#idl-sequence" id="ref-for-idl-sequence②③"><c- b>sequence</c-></a>&lt;<a class="idl-code" data-link-type="interface" href="https://webidl.spec.whatwg.org/#idl-unsigned-long" id="ref-for-idl-unsigned-long④⑦"><c- b>unsigned</c-> <c- b>long</c-></a>> <a class="idl-code" data-link-type="dict-member" data-type="sequence<unsigned long> " href="#dom-mlpool2doptions-outputsizes" id="ref-for-dom-mlpool2doptions-outputsizes"><c- g>outputSizes</c-></a>;
 };
 
-<c- b>partial</c-> <c- b>interface</c-> <a class="idl-code" data-link-type="interface" href="#mlgraphbuilder" id="ref-for-mlgraphbuilder④⓪"><c- g>MLGraphBuilder</c-></a> {
-  <a data-link-type="idl-name" href="#mloperand" id="ref-for-mloperand②②①"><c- n>MLOperand</c-></a> <dfn class="idl-code" data-dfn-for="MLGraphBuilder" data-dfn-type="method" data-export data-lt="averagePool2d(input, options)|averagePool2d(input)" id="dom-mlgraphbuilder-averagepool2d"><code><c- g>averagePool2d</c-></code><a class="self-link" href="#dom-mlgraphbuilder-averagepool2d"></a></dfn>(<a data-link-type="idl-name" href="#mloperand" id="ref-for-mloperand②②②"><c- n>MLOperand</c-></a> <dfn class="idl-code" data-dfn-for="MLGraphBuilder/averagePool2d(input, options), MLGraphBuilder/averagePool2d(input)" data-dfn-type="argument" data-export id="dom-mlgraphbuilder-averagepool2d-input-options-input"><code><c- g>input</c-></code><a class="self-link" href="#dom-mlgraphbuilder-averagepool2d-input-options-input"></a></dfn>, <c- b>optional</c-> <a data-link-type="idl-name" href="#dictdef-mlpool2doptions" id="ref-for-dictdef-mlpool2doptions"><c- n>MLPool2dOptions</c-></a> <dfn class="idl-code" data-dfn-for="MLGraphBuilder/averagePool2d(input, options), MLGraphBuilder/averagePool2d(input)" data-dfn-type="argument" data-export id="dom-mlgraphbuilder-averagepool2d-input-options-options"><code><c- g>options</c-></code><a class="self-link" href="#dom-mlgraphbuilder-averagepool2d-input-options-options"></a></dfn> = {});
-  <a data-link-type="idl-name" href="#mloperand" id="ref-for-mloperand②②③"><c- n>MLOperand</c-></a> <dfn class="idl-code" data-dfn-for="MLGraphBuilder" data-dfn-type="method" data-export data-lt="l2Pool2d(input, options)|l2Pool2d(input)" id="dom-mlgraphbuilder-l2pool2d"><code><c- g>l2Pool2d</c-></code><a class="self-link" href="#dom-mlgraphbuilder-l2pool2d"></a></dfn>(<a data-link-type="idl-name" href="#mloperand" id="ref-for-mloperand②②④"><c- n>MLOperand</c-></a> <dfn class="idl-code" data-dfn-for="MLGraphBuilder/l2Pool2d(input, options), MLGraphBuilder/l2Pool2d(input)" data-dfn-type="argument" data-export id="dom-mlgraphbuilder-l2pool2d-input-options-input"><code><c- g>input</c-></code><a class="self-link" href="#dom-mlgraphbuilder-l2pool2d-input-options-input"></a></dfn>, <c- b>optional</c-> <a data-link-type="idl-name" href="#dictdef-mlpool2doptions" id="ref-for-dictdef-mlpool2doptions①"><c- n>MLPool2dOptions</c-></a> <dfn class="idl-code" data-dfn-for="MLGraphBuilder/l2Pool2d(input, options), MLGraphBuilder/l2Pool2d(input)" data-dfn-type="argument" data-export id="dom-mlgraphbuilder-l2pool2d-input-options-options"><code><c- g>options</c-></code><a class="self-link" href="#dom-mlgraphbuilder-l2pool2d-input-options-options"></a></dfn> = {});
-  <a data-link-type="idl-name" href="#mloperand" id="ref-for-mloperand②②⑤"><c- n>MLOperand</c-></a> <dfn class="idl-code" data-dfn-for="MLGraphBuilder" data-dfn-type="method" data-export data-lt="maxPool2d(input, options)|maxPool2d(input)" id="dom-mlgraphbuilder-maxpool2d"><code><c- g>maxPool2d</c-></code><a class="self-link" href="#dom-mlgraphbuilder-maxpool2d"></a></dfn>(<a data-link-type="idl-name" href="#mloperand" id="ref-for-mloperand②②⑥"><c- n>MLOperand</c-></a> <dfn class="idl-code" data-dfn-for="MLGraphBuilder/maxPool2d(input, options), MLGraphBuilder/maxPool2d(input)" data-dfn-type="argument" data-export id="dom-mlgraphbuilder-maxpool2d-input-options-input"><code><c- g>input</c-></code><a class="self-link" href="#dom-mlgraphbuilder-maxpool2d-input-options-input"></a></dfn>, <c- b>optional</c-> <a data-link-type="idl-name" href="#dictdef-mlpool2doptions" id="ref-for-dictdef-mlpool2doptions②"><c- n>MLPool2dOptions</c-></a> <dfn class="idl-code" data-dfn-for="MLGraphBuilder/maxPool2d(input, options), MLGraphBuilder/maxPool2d(input)" data-dfn-type="argument" data-export id="dom-mlgraphbuilder-maxpool2d-input-options-options"><code><c- g>options</c-></code><a class="self-link" href="#dom-mlgraphbuilder-maxpool2d-input-options-options"></a></dfn> = {});
+<c- b>partial</c-> <c- b>interface</c-> <a class="idl-code" data-link-type="interface" href="#mlgraphbuilder" id="ref-for-mlgraphbuilder④③"><c- g>MLGraphBuilder</c-></a> {
+  <a data-link-type="idl-name" href="#mloperand" id="ref-for-mloperand②⑦⑧"><c- n>MLOperand</c-></a> <a class="idl-code" data-link-type="method" href="#dom-mlgraphbuilder-averagepool2d" id="ref-for-dom-mlgraphbuilder-averagepool2d"><c- g>averagePool2d</c-></a>(<a data-link-type="idl-name" href="#mloperand" id="ref-for-mloperand②⑦⑨"><c- n>MLOperand</c-></a> <dfn class="idl-code" data-dfn-for="MLGraphBuilder/averagePool2d(input, options), MLGraphBuilder/averagePool2d(input)" data-dfn-type="argument" data-export id="dom-mlgraphbuilder-averagepool2d-input-options-input"><code><c- g>input</c-></code><a class="self-link" href="#dom-mlgraphbuilder-averagepool2d-input-options-input"></a></dfn>, <c- b>optional</c-> <a data-link-type="idl-name" href="#dictdef-mlpool2doptions" id="ref-for-dictdef-mlpool2doptions"><c- n>MLPool2dOptions</c-></a> <dfn class="idl-code" data-dfn-for="MLGraphBuilder/averagePool2d(input, options), MLGraphBuilder/averagePool2d(input)" data-dfn-type="argument" data-export id="dom-mlgraphbuilder-averagepool2d-input-options-options"><code><c- g>options</c-></code><a class="self-link" href="#dom-mlgraphbuilder-averagepool2d-input-options-options"></a></dfn> = {});
+  <a data-link-type="idl-name" href="#mloperand" id="ref-for-mloperand②⑧⓪"><c- n>MLOperand</c-></a> <a class="idl-code" data-link-type="method" href="#dom-mlgraphbuilder-l2pool2d" id="ref-for-dom-mlgraphbuilder-l2pool2d"><c- g>l2Pool2d</c-></a>(<a data-link-type="idl-name" href="#mloperand" id="ref-for-mloperand②⑧①"><c- n>MLOperand</c-></a> <dfn class="idl-code" data-dfn-for="MLGraphBuilder/l2Pool2d(input, options), MLGraphBuilder/l2Pool2d(input)" data-dfn-type="argument" data-export id="dom-mlgraphbuilder-l2pool2d-input-options-input"><code><c- g>input</c-></code><a class="self-link" href="#dom-mlgraphbuilder-l2pool2d-input-options-input"></a></dfn>, <c- b>optional</c-> <a data-link-type="idl-name" href="#dictdef-mlpool2doptions" id="ref-for-dictdef-mlpool2doptions①"><c- n>MLPool2dOptions</c-></a> <dfn class="idl-code" data-dfn-for="MLGraphBuilder/l2Pool2d(input, options), MLGraphBuilder/l2Pool2d(input)" data-dfn-type="argument" data-export id="dom-mlgraphbuilder-l2pool2d-input-options-options"><code><c- g>options</c-></code><a class="self-link" href="#dom-mlgraphbuilder-l2pool2d-input-options-options"></a></dfn> = {});
+  <a data-link-type="idl-name" href="#mloperand" id="ref-for-mloperand②⑧②"><c- n>MLOperand</c-></a> <a class="idl-code" data-link-type="method" href="#dom-mlgraphbuilder-maxpool2d" id="ref-for-dom-mlgraphbuilder-maxpool2d"><c- g>maxPool2d</c-></a>(<a data-link-type="idl-name" href="#mloperand" id="ref-for-mloperand②⑧③"><c- n>MLOperand</c-></a> <dfn class="idl-code" data-dfn-for="MLGraphBuilder/maxPool2d(input, options), MLGraphBuilder/maxPool2d(input)" data-dfn-type="argument" data-export id="dom-mlgraphbuilder-maxpool2d-input-options-input"><code><c- g>input</c-></code><a class="self-link" href="#dom-mlgraphbuilder-maxpool2d-input-options-input"></a></dfn>, <c- b>optional</c-> <a data-link-type="idl-name" href="#dictdef-mlpool2doptions" id="ref-for-dictdef-mlpool2doptions②"><c- n>MLPool2dOptions</c-></a> <dfn class="idl-code" data-dfn-for="MLGraphBuilder/maxPool2d(input, options), MLGraphBuilder/maxPool2d(input)" data-dfn-type="argument" data-export id="dom-mlgraphbuilder-maxpool2d-input-options-options"><code><c- g>options</c-></code><a class="self-link" href="#dom-mlgraphbuilder-maxpool2d-input-options-options"></a></dfn> = {});
+};
+</pre>
+   <div>
+     <strong>Arguments:</strong> 
+    <ul>
+     <li data-md>
+      <p><em>input</em>: an <code class="idl"><a data-link-type="idl" href="#mloperand" id="ref-for-mloperand②⑧④">MLOperand</a></code>. The input 4-D tensor. The logical shape
+is interpreted according to the value of <em>options.layout</em>.</p>
+     <li data-md>
+      <p><em>options</em>: an optional <code class="idl"><a data-link-type="idl" href="#dictdef-mlpool2doptions" id="ref-for-dictdef-mlpool2doptions③">MLPool2dOptions</a></code>. The optional parameters of the operation.</p>
+    </ul>
+    <p><strong>Returns:</strong> an <code class="idl"><a data-link-type="idl" href="#mloperand" id="ref-for-mloperand②⑧⑤">MLOperand</a></code>. The output 4-D tensor that contains the
+    result of the reduction. The logical shape is interpreted according to the
+    value of <em>layout</em>. More specifically, if the <em>options.roundingType</em> is <em>"floor"</em>, the spatial dimensions of the output tensor can be calculated as follow:</p>
+    <p><em>output size = floor(1 + (input size - filter size + beginning padding + ending padding) / stride)</em></p>
+    <p>or if <em>options.roundingType</em> is <em>"ceil"</em>:</p>
+    <p><em>output size = ceil(1 + (input size - filter size + beginning padding + ending padding) / stride)</em></p>
+   </div>
+   <div class="note" role="note">
+     A <em>global</em> pooling operation such as one for the max pooling operation is a variant of pooling where the window dimensions is the spatial dimensions (last two dimensions) of the input shape, as follow. 
+<pre class="highlight"><c- c1>// 'global' max pooling</c->
+builder<c- p>.</c->maxPool2d<c- p>(</c->input<c- p>);</c->
+</pre>
+   </div>
+   <p><code class="idl"><a data-link-type="idl" href="#dictdef-mlpool2doptions" id="ref-for-dictdef-mlpool2doptions④">MLPool2dOptions</a></code> has the following members:</p>
+   <dl>
+    <dt data-md><dfn class="dfn-paneled idl-code" data-dfn-for="MLPool2dOptions" data-dfn-type="dict-member" data-export id="dom-mlpool2doptions-windowdimensions"><code>windowDimensions</code></dfn>, <span> of type <code class="idl-code">sequence&lt;unsigned long></code></span>
+    <dd data-md>
+     <p>A sequence of <code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#idl-unsigned-long" id="ref-for-idl-unsigned-long④⑧">unsigned long</a></code> of length 2: [windowHeight, windowWidth].
+ Specifies the dimensions of the sliding window.
+ The default value for the window dimensions are the height and width dimensions of the input shape.</p>
+    <dt data-md><dfn class="dfn-paneled idl-code" data-dfn-for="MLPool2dOptions" data-dfn-type="dict-member" data-export id="dom-mlpool2doptions-padding"><code>padding</code></dfn>, <span> of type <code class="idl-code">sequence&lt;unsigned long></code></span>
+    <dd data-md>
+     <p>A sequence of <code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#idl-unsigned-long" id="ref-for-idl-unsigned-long④⑨">unsigned long</a></code> of length 4: [beginningHeight, endingHeight, beginningWidth, endingWidth].
+Specifies the additional rows and columns added to the beginning and ending of each spatial dimension of the convolution input.
+The default value is [0,0,0,0].</p>
+    <dt data-md><dfn class="dfn-paneled idl-code" data-dfn-for="MLPool2dOptions" data-dfn-type="dict-member" data-export id="dom-mlpool2doptions-strides"><code>strides</code></dfn>, <span> of type <code class="idl-code">sequence&lt;unsigned long></code></span>
+    <dd data-md>
+     <p>A sequence of <code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#idl-unsigned-long" id="ref-for-idl-unsigned-long⑤⓪">unsigned long</a></code> of length 2: [strideHeight, strideWidth].
+Specifies the stride of the sliding window for each spatial dimension of the convolution input.
+The default value is [1,1].</p>
+    <dt data-md><dfn class="dfn-paneled idl-code" data-dfn-for="MLPool2dOptions" data-dfn-type="dict-member" data-export id="dom-mlpool2doptions-dilations"><code>dilations</code></dfn>, <span> of type <code class="idl-code">sequence&lt;unsigned long></code></span>
+    <dd data-md>
+     <p>A sequence of <code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#idl-unsigned-long" id="ref-for-idl-unsigned-long⑤①">unsigned long</a></code> of length 2: [dilationHeight, dilationWidth]. Specifies the dilation factor for each spatial dimension applied on the convolution filter (kernel).
+The default value is [1,1].</p>
+    <dt data-md><dfn class="dfn-paneled idl-code" data-dfn-for="MLPool2dOptions" data-dfn-type="dict-member" data-export id="dom-mlpool2doptions-autopad"><code>autoPad</code></dfn>, <span> of type <a data-link-type="idl-name" href="#enumdef-mlautopad" id="ref-for-enumdef-mlautopad⑦">MLAutoPad</a>, defaulting to <code>"explicit"</code></span>
+    <dd data-md>
+     <p>An <code class="idl"><a data-link-type="idl" href="#enumdef-mlautopad" id="ref-for-enumdef-mlautopad⑧">MLAutoPad</a></code> <a data-link-type="dfn" href="https://infra.spec.whatwg.org/#string" id="ref-for-string①①">string</a>.
+Specifies the automatic input padding options.
+The default value is <em>"explicit"</em>, which means that the values in the <code class="idl"><a data-link-type="idl" href="#dom-mlpool2doptions-padding" id="ref-for-dom-mlpool2doptions-padding①">padding</a></code> array should be used for input padding.
+When the option is set other than <em>"explicit"</em>, the values in the <code class="idl"><a data-link-type="idl" href="#dom-mlpool2doptions-padding" id="ref-for-dom-mlpool2doptions-padding②">padding</a></code> array are ignored.</p>
+     <p>With the <em>"same-upper"</em> option, the padding values are automatically computed such that the additional ending padding of the spatial input dimensions would allow all of the input values in the corresponding dimension to be filtered.</p>
+     <p>The <em>"same-lower"</em> option is similar but padding is applied to the beginning padding of the spatial input dimensions instead of the ending one.</p>
+    <dt data-md><dfn class="dfn-paneled idl-code" data-dfn-for="MLPool2dOptions" data-dfn-type="dict-member" data-export id="dom-mlpool2doptions-layout"><code>layout</code></dfn>, <span> of type <a data-link-type="idl-name" href="#enumdef-mlinputoperandlayout" id="ref-for-enumdef-mlinputoperandlayout①⓪">MLInputOperandLayout</a>, defaulting to <code>"nchw"</code></span>
+    <dd data-md>
+     <p>An <code class="idl"><a data-link-type="idl" href="#enumdef-mlinputoperandlayout" id="ref-for-enumdef-mlinputoperandlayout①①">MLInputOperandLayout</a></code> <a data-link-type="dfn" href="https://infra.spec.whatwg.org/#string" id="ref-for-string①②">string</a>.
+Specifies the layout format of the input and output tensor as follows:</p>
+     <ul>
+      <li data-md>
+       <p><strong>"nchw"</strong></p>
+       <ul>
+        <li data-md>
+         <p>input tensor: <em>[batches, inputChannels, height, width]</em></p>
+        <li data-md>
+         <p>output tensor: <em>[batches, outputChannels, height, width]</em></p>
+       </ul>
+      <li data-md>
+       <p><strong>"nhwc"</strong>:</p>
+       <ul>
+        <li data-md>
+         <p>input tensor: <em>[batches, height, width, inputChannels]</em></p>
+        <li data-md>
+         <p>output tensor: <em>[batches, height, width, outputChannels]</em></p>
+       </ul>
+     </ul>
+     The default value is <em>"nchw"</em>. 
+    <dt data-md><dfn class="dfn-paneled idl-code" data-dfn-for="MLPool2dOptions" data-dfn-type="dict-member" data-export id="dom-mlpool2doptions-roundingtype"><code>roundingType</code></dfn>, <span> of type <a data-link-type="idl-name" href="#enumdef-mlroundingtype" id="ref-for-enumdef-mlroundingtype①">MLRoundingType</a>, defaulting to <code>"floor"</code></span>
+    <dd data-md>
+     <p>An <code class="idl"><a data-link-type="idl" href="#enumdef-mlroundingtype" id="ref-for-enumdef-mlroundingtype②">MLRoundingType</a></code> <a data-link-type="dfn" href="https://infra.spec.whatwg.org/#string" id="ref-for-string①③">string</a>.
+Specifies the rounding function used to compute the output shape.</p>
+    <dt data-md><dfn class="dfn-paneled idl-code" data-dfn-for="MLPool2dOptions" data-dfn-type="dict-member" data-export id="dom-mlpool2doptions-outputsizes"><code>outputSizes</code></dfn>, <span> of type <code class="idl-code">sequence&lt;unsigned long></code></span>
+    <dd data-md>
+     <p>A sequence of <code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#idl-unsigned-long" id="ref-for-idl-unsigned-long⑤②">unsigned long</a></code> of length 2.
+Specifies the sizes of the two spacial dimensions of the output tensor. When the output sizes are explicitly specified, the <code class="idl"><a data-link-type="idl" href="#dom-mlpool2doptions-roundingtype" id="ref-for-dom-mlpool2doptions-roundingtype①">roundingType</a></code> is ignored.</p>
+     <p>If not specified, the output sizes are automatically computed.</p>
+   </dl>
+   <details class="algorithm" data-algorithm="pooling-op" data-algorithm-for="MLGraphBuilder" open>
+    <summary> To <dfn class="dfn-paneled" data-dfn-for="MLGraphBuilder" data-dfn-type="dfn" data-lt="pooling-op" data-noexport id="mlgraphbuilder-pooling-op">create pooling operation</dfn> given <var>op</var>, <var>input</var> and <var>options</var>, run the following steps: </summary>
+    <div class="algorithm-steps">
+     <ol>
+      <li data-md>
+       <p><a data-link-type="dfn" href="https://infra.spec.whatwg.org/#assert" id="ref-for-assert⑤⑧">Assert</a>: <var>op</var> is one of "averagePool2d", "l2Pool2d", "maxPool2d".</p>
+      <li data-md>
+       <p><a data-link-type="dfn" href="https://infra.spec.whatwg.org/#assert" id="ref-for-assert⑤⑨">Assert</a>: the type of <var>input</var> is <code class="idl"><a data-link-type="idl" href="#mloperand" id="ref-for-mloperand②⑧⑥">MLOperand</a></code>.</p>
+      <li data-md>
+       <p>If the <a data-link-type="dfn" href="https://infra.spec.whatwg.org/#list-size" id="ref-for-list-size③⑨">size</a> of <var>input</var>.<code class="idl"><a data-link-type="idl" href="#dom-mloperand-descriptor-slot" id="ref-for-dom-mloperand-descriptor-slot⑧⓪">[[descriptor]]</a></code>.<code class="idl"><a data-link-type="idl" href="#dom-mloperanddescriptor-dimensions" id="ref-for-dom-mloperanddescriptor-dimensions⑦⑨">dimensions</a></code> is not 4, then <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#dfn-throw" id="ref-for-dfn-throw①⑨⑦">throw</a> a "<code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#dataerror" id="ref-for-dataerror⑦②">DataError</a></code>" <code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#idl-DOMException" id="ref-for-idl-DOMException①①⑦">DOMException</a></code>.</p>
+      <li data-md>
+       <p>If <var>options</var>.<code class="idl"><a data-link-type="idl" href="#dom-mlpool2doptions-windowdimensions" id="ref-for-dom-mlpool2doptions-windowdimensions①">windowDimensions</a></code> <a data-link-type="dfn" href="https://infra.spec.whatwg.org/#map-exists" id="ref-for-map-exists④⑤">exists</a> and its <a data-link-type="dfn" href="https://infra.spec.whatwg.org/#list-size" id="ref-for-list-size④⓪">size</a> is not 2, then <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#dfn-throw" id="ref-for-dfn-throw①⑨⑧">throw</a> a "<code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#dataerror" id="ref-for-dataerror⑦③">DataError</a></code>" <code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#idl-DOMException" id="ref-for-idl-DOMException①①⑧">DOMException</a></code>.</p>
+      <li data-md>
+       <p>Otherwise, set <var>options</var>.<code class="idl"><a data-link-type="idl" href="#dom-mlpool2doptions-windowdimensions" id="ref-for-dom-mlpool2doptions-windowdimensions②">windowDimensions</a></code> to the height and width dimensions of the shape of <var>input</var>.</p>
+      <li data-md>
+       <p>If <var>options</var>.<code class="idl"><a data-link-type="idl" href="#dom-mlpool2doptions-outputsizes" id="ref-for-dom-mlpool2doptions-outputsizes①">outputSizes</a></code> <a data-link-type="dfn" href="https://infra.spec.whatwg.org/#map-exists" id="ref-for-map-exists④⑥">exists</a>, or if <var>options</var>.<code class="idl"><a data-link-type="idl" href="#dom-mlpool2doptions-padding" id="ref-for-dom-mlpool2doptions-padding③">padding</a></code> does not <a data-link-type="dfn" href="https://infra.spec.whatwg.org/#map-exists" id="ref-for-map-exists④⑦">exist</a>, set <var>options</var>.<code class="idl"><a data-link-type="idl" href="#dom-mlpool2doptions-padding" id="ref-for-dom-mlpool2doptions-padding④">padding</a></code> to <code>« 0, 0, 0, 0 »</code>.</p>
+      <li data-md>
+       <p>If the len<a data-link-type="dfn" href="https://infra.spec.whatwg.org/#list-size" id="ref-for-list-size④①">size</a>gth of <var>options</var>.<code class="idl"><a data-link-type="idl" href="#dom-mlpool2doptions-padding" id="ref-for-dom-mlpool2doptions-padding⑤">padding</a></code> is not 4, then <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#dfn-throw" id="ref-for-dfn-throw①⑨⑨">throw</a> a "<code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#dataerror" id="ref-for-dataerror⑦④">DataError</a></code>" <code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#idl-DOMException" id="ref-for-idl-DOMException①①⑨">DOMException</a></code>.</p>
+      <li data-md>
+       <p>If <var>options</var>.<code class="idl"><a data-link-type="idl" href="#dom-mlpool2doptions-strides" id="ref-for-dom-mlpool2doptions-strides①">strides</a></code> does not <a data-link-type="dfn" href="https://infra.spec.whatwg.org/#map-exists" id="ref-for-map-exists④⑧">exist</a>, set <var>options</var>.<code class="idl"><a data-link-type="idl" href="#dom-mlpool2doptions-strides" id="ref-for-dom-mlpool2doptions-strides②">strides</a></code> to <code>« 1, 1 »</code>.</p>
+      <li data-md>
+       <p>If the <a data-link-type="dfn" href="https://infra.spec.whatwg.org/#list-size" id="ref-for-list-size④②">size</a> of <var>options</var>.<code class="idl"><a data-link-type="idl" href="#dom-mlpool2doptions-strides" id="ref-for-dom-mlpool2doptions-strides③">strides</a></code> is not 2, then <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#dfn-throw" id="ref-for-dfn-throw②⓪⓪">throw</a> a "<code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#dataerror" id="ref-for-dataerror⑦⑤">DataError</a></code>" <code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#idl-DOMException" id="ref-for-idl-DOMException①②⓪">DOMException</a></code>.</p>
+      <li data-md>
+       <p>If any value in <var>options</var>.<code class="idl"><a data-link-type="idl" href="#dom-mlpool2doptions-strides" id="ref-for-dom-mlpool2doptions-strides④">strides</a></code> is not greater than 0, then <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#dfn-throw" id="ref-for-dfn-throw②⓪①">throw</a> a "<code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#dataerror" id="ref-for-dataerror⑦⑥">DataError</a></code>" <code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#idl-DOMException" id="ref-for-idl-DOMException①②①">DOMException</a></code>.</p>
+      <li data-md>
+       <p>If <var>options</var>.<code class="idl"><a data-link-type="idl" href="#dom-mlpool2doptions-outputsizes" id="ref-for-dom-mlpool2doptions-outputsizes②">outputSizes</a></code> <a data-link-type="dfn" href="https://infra.spec.whatwg.org/#map-exists" id="ref-for-map-exists④⑨">exists</a>:</p>
+       <ol>
+        <li data-md>
+         <p>If the <a data-link-type="dfn" href="https://infra.spec.whatwg.org/#list-size" id="ref-for-list-size④③">size</a> of <var>options</var>.<code class="idl"><a data-link-type="idl" href="#dom-mlpool2doptions-outputsizes" id="ref-for-dom-mlpool2doptions-outputsizes③">outputSizes</a></code> is not 2, then <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#dfn-throw" id="ref-for-dfn-throw②⓪②">throw</a> a <code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#exceptiondef-typeerror" id="ref-for-exceptiondef-typeerror③⑨">TypeError</a></code>.</p>
+        <li data-md>
+         <p>If the elements of <var>options</var>.<code class="idl"><a data-link-type="idl" href="#dom-mlpool2doptions-outputsizes" id="ref-for-dom-mlpool2doptions-outputsizes④">outputSizes</a></code> are not smaller than the elements at the same dimension (index) for <var>options</var>.<code class="idl"><a data-link-type="idl" href="#dom-mlpool2doptions-strides" id="ref-for-dom-mlpool2doptions-strides⑤">strides</a></code>, then <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#dfn-throw" id="ref-for-dfn-throw②⓪③">throw</a> a "<code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#dataerror" id="ref-for-dataerror⑦⑦">DataError</a></code>" <code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#idl-DOMException" id="ref-for-idl-DOMException①②②">DOMException</a></code>.</p>
+       </ol>
+      <li data-md>
+       <p>If <var>options</var>.<code class="idl"><a data-link-type="idl" href="#dom-mlpool2doptions-dilations" id="ref-for-dom-mlpool2doptions-dilations①">dilations</a></code> does not <a data-link-type="dfn" href="https://infra.spec.whatwg.org/#map-exists" id="ref-for-map-exists⑤⓪">exist</a>, set <var>options</var>.<code class="idl"><a data-link-type="idl" href="#dom-mlpool2doptions-dilations" id="ref-for-dom-mlpool2doptions-dilations②">dilations</a></code> to <code>« 1, 1 »</code>.</p>
+      <li data-md>
+       <p>If the <a data-link-type="dfn" href="https://infra.spec.whatwg.org/#list-size" id="ref-for-list-size④④">size</a> of <var>options</var>.<code class="idl"><a data-link-type="idl" href="#dom-mlpool2doptions-dilations" id="ref-for-dom-mlpool2doptions-dilations③">dilations</a></code> is not 2, then <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#dfn-throw" id="ref-for-dfn-throw②⓪④">throw</a> a "<code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#dataerror" id="ref-for-dataerror⑦⑧">DataError</a></code>" <code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#idl-DOMException" id="ref-for-idl-DOMException①②③">DOMException</a></code>.</p>
+      <li data-md>
+       <p>If any value in <var>options</var>.<code class="idl"><a data-link-type="idl" href="#dom-mlpool2doptions-dilations" id="ref-for-dom-mlpool2doptions-dilations④">dilations</a></code> is not greater than 0, then <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#dfn-throw" id="ref-for-dfn-throw②⓪⑤">throw</a> a "<code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#dataerror" id="ref-for-dataerror⑦⑨">DataError</a></code>" <code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#idl-DOMException" id="ref-for-idl-DOMException①②④">DOMException</a></code>.</p>
+      <li data-md>
+       <p>If <var>options</var>.<code class="idl"><a data-link-type="idl" href="#dom-mlpool2doptions-autopad" id="ref-for-dom-mlpool2doptions-autopad①">autoPad</a></code> is not <code>"explicit"</code>, set <var>options</var>.<code class="idl"><a data-link-type="idl" href="#dom-mlpool2doptions-padding" id="ref-for-dom-mlpool2doptions-padding⑥">padding</a></code> to <code>« 0, 0, 0, 0 »</code>.</p>
+      <li data-md>
+       <p>Let <var>desc</var> be a copy of <var>input</var>.<code class="idl"><a data-link-type="idl" href="#dom-mloperand-descriptor-slot" id="ref-for-dom-mloperand-descriptor-slot⑧①">[[descriptor]]</a></code>.</p>
+      <li data-md>
+       <p>If any of the following sub-steps fail, <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#dfn-throw" id="ref-for-dfn-throw②⓪⑥">throw</a> an "<code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#operationerror" id="ref-for-operationerror③⑥">OperationError</a></code>" <code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#idl-DOMException" id="ref-for-idl-DOMException①②⑤">DOMException</a></code>.</p>
+       <ol>
+        <li data-md>
+         <p>Make a request to the underlying platform to:</p>
+         <ol>
+          <li data-md>
+           <p>Calculate the output dimensions given <var>input</var> and <var>options</var>. Let <var>desc</var>.<code class="idl"><a data-link-type="idl" href="#dom-mloperanddescriptor-dimensions" id="ref-for-dom-mloperanddescriptor-dimensions⑧⓪">dimensions</a></code> be the result of that.</p>
+          <li data-md>
+           <p>Let <var>output</var> be the result of <a data-link-type="dfn" href="#create-an-mloperand" id="ref-for-create-an-mloperand①⑦">creating an MLOperand</a> given <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#this" id="ref-for-this③①">this</a> and <var>desc</var>.</p>
+          <li data-md>
+           <p>Let <var>opImpl</var> be an <a data-link-type="dfn" href="https://infra.spec.whatwg.org/#implementation-defined" id="ref-for-implementation-defined④⑧">implementation-defined</a> platform operator for the <var>op</var> pooling operation, given <var>options</var>.</p>
+          <li data-md>
+           <p>Store a reference of <var>opImpl</var> in <var>output</var>.<code class="idl"><a data-link-type="idl" href="#dom-mloperand-operator-slot" id="ref-for-dom-mloperand-operator-slot②②">[[operator]]</a></code>.</p>
+          <li data-md>
+           <p>Create an <a data-link-type="dfn" href="https://infra.spec.whatwg.org/#implementation-defined" id="ref-for-implementation-defined④⑨">implementation-defined</a> platform operand <var>outputImpl</var> to represent the output, given <var>output</var> and <var>opImpl</var>.</p>
+          <li data-md>
+           <p>Store a reference to <var>outputImpl</var> in <var>output</var>.<code class="idl"><a data-link-type="idl" href="#dom-mloperand-operand-slot" id="ref-for-dom-mloperand-operand-slot⑥②">[[operand]]</a></code>.</p>
+         </ol>
+        <li data-md>
+         <p>Connect <var>input</var>.<code class="idl"><a data-link-type="idl" href="#dom-mloperand-operand-slot" id="ref-for-dom-mloperand-operand-slot⑥③">[[operand]]</a></code> as input to <var>opImpl</var>.</p>
+        <li data-md>
+         <p>Connect <var>output</var>.<code class="idl"><a data-link-type="idl" href="#dom-mloperand-operand-slot" id="ref-for-dom-mloperand-operand-slot⑥④">[[operand]]</a></code> as output to <var>opImpl</var>.</p>
+       </ol>
+      <li data-md>
+       <p>Return <var>output</var>.</p>
+     </ol>
+    </div>
+   </details>
+   <details open>
+    <summary> The following pooling algorithms are supported. </summary>
+    <div class="algorithm-steps algorithm" data-algorithm="averagePool2d(input, options)" data-algorithm-for="MLGraphBuilder">
+      The <dfn class="dfn-paneled idl-code" data-dfn-for="MLGraphBuilder" data-dfn-type="method" data-export data-lt="averagePool2d(input, options)|averagePool2d(input)" id="dom-mlgraphbuilder-averagepool2d"><code>averagePool2d(<var>input</var>, <var>options</var>)</code></dfn> method steps are: 
+     <ol>
+      <li data-md>
+       <p>Let <var>output</var> be the result of running the <a data-link-type="dfn" href="#mlgraphbuilder-pooling-op" id="ref-for-mlgraphbuilder-pooling-op"> create pooling operation</a> given <code>"averagePool2d"</code>, <var>input</var> and <var>options</var>.</p>
+       <ol>
+        <li data-md>
+         <p>If that <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#dfn-throw" id="ref-for-dfn-throw②⓪⑦">throws</a> an error, then re-<a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#dfn-throw" id="ref-for-dfn-throw②⓪⑧">throw</a> the error.</p>
+       </ol>
+      <li data-md>
+       <p>Return <var>output</var>.</p>
+     </ol>
+    </div>
+    <div class="algorithm-steps algorithm" data-algorithm="l2Pool2d(input, options)" data-algorithm-for="MLGraphBuilder">
+      The <dfn class="dfn-paneled idl-code" data-dfn-for="MLGraphBuilder" data-dfn-type="method" data-export data-lt="l2Pool2d(input, options)|l2Pool2d(input)" id="dom-mlgraphbuilder-l2pool2d"><code>l2Pool2d(<var>input</var>, <var>options</var>)</code></dfn> method steps are: 
+     <ol>
+      <li data-md>
+       <p>Let <var>output</var> be the result of running the <a data-link-type="dfn" href="#mlgraphbuilder-pooling-op" id="ref-for-mlgraphbuilder-pooling-op①"> create pooling operation</a> given <code>"l2Pool2d"</code>, <var>input</var> and <var>options</var>.</p>
+       <ol>
+        <li data-md>
+         <p>If that <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#dfn-throw" id="ref-for-dfn-throw②⓪⑨">throws</a> an error, then re-<a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#dfn-throw" id="ref-for-dfn-throw②①⓪">throw</a> the error.</p>
+       </ol>
+      <li data-md>
+       <p>Return <var>output</var>.</p>
+     </ol>
+    </div>
+    <div class="algorithm-steps algorithm" data-algorithm="maxPool2d(input, options)" data-algorithm-for="MLGraphBuilder">
+      The <dfn class="dfn-paneled idl-code" data-dfn-for="MLGraphBuilder" data-dfn-type="method" data-export data-lt="maxPool2d(input, options)|maxPool2d(input)" id="dom-mlgraphbuilder-maxpool2d"><code>maxPool2d(<var>input</var>, <var>options</var>)</code></dfn> method steps are: 
+     <ol>
+      <li data-md>
+       <p>Let <var>output</var> be the result of running the <a data-link-type="dfn" href="#mlgraphbuilder-pooling-op" id="ref-for-mlgraphbuilder-pooling-op②"> create pooling operation</a> given <code>"maxPool2d"</code>, <var>input</var> and <var>options</var>.</p>
+       <ol>
+        <li data-md>
+         <p>If that <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#dfn-throw" id="ref-for-dfn-throw②①①">throws</a> an error, then re-<a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#dfn-throw" id="ref-for-dfn-throw②①②">throw</a> the error.</p>
+       </ol>
+      <li data-md>
+       <p>Return <var>output</var>.</p>
+     </ol>
+    </div>
+   </details>
+   <h4 class="heading settled" data-level="7.6.26" id="api-mlgraphbuilder-prelu"><span class="secno">7.6.26. </span><span class="content">The prelu() method</span><a class="self-link" href="#api-mlgraphbuilder-prelu"></a></h4>
+    Calculate the <a href="https://en.wikipedia.org/wiki/Rectifier_(neural_networks)#Parametric_ReLU">parametric version of rectified linear function (Parametric ReLU)</a> on the input tensor element-wise. Parametric ReLU is a type of leaky ReLU that, instead of having a scalar slope like 0.01, making the slope (coefficient of leakage) into a parameter that is learned during the model training phase of this operation. The calculation follows the expression <code>max(0, x) + slope ∗ min(0, x)</code>. 
+<pre class="idl highlight def"><c- b>partial</c-> <c- b>interface</c-> <a class="idl-code" data-link-type="interface" href="#mlgraphbuilder" id="ref-for-mlgraphbuilder④④"><c- g>MLGraphBuilder</c-></a> {
+  <a data-link-type="idl-name" href="#mloperand" id="ref-for-mloperand②⑧⑦"><c- n>MLOperand</c-></a> <a class="idl-code" data-link-type="method" href="#dom-mlgraphbuilder-prelu" id="ref-for-dom-mlgraphbuilder-prelu"><c- g>prelu</c-></a>(<a data-link-type="idl-name" href="#mloperand" id="ref-for-mloperand②⑧⑧"><c- n>MLOperand</c-></a> <dfn class="idl-code" data-dfn-for="MLGraphBuilder/prelu(input, slope)" data-dfn-type="argument" data-export id="dom-mlgraphbuilder-prelu-input-slope-input"><code><c- g>input</c-></code><a class="self-link" href="#dom-mlgraphbuilder-prelu-input-slope-input"></a></dfn>, <a data-link-type="idl-name" href="#mloperand" id="ref-for-mloperand②⑧⑨"><c- n>MLOperand</c-></a> <dfn class="idl-code" data-dfn-for="MLGraphBuilder/prelu(input, slope)" data-dfn-type="argument" data-export id="dom-mlgraphbuilder-prelu-input-slope-slope"><code><c- g>slope</c-></code><a class="self-link" href="#dom-mlgraphbuilder-prelu-input-slope-slope"></a></dfn>);
 };
 </pre>
-    <div class="algorithm" data-algorithm="pool2d">
-      <strong>Arguments:</strong> 
-     <ul>
+   <div>
+     <strong>Arguments:</strong> 
+    <ul>
+     <li data-md>
+      <p><em>input</em>: an <code class="idl"><a data-link-type="idl" href="#mloperand" id="ref-for-mloperand②⑨⓪">MLOperand</a></code>. The input tensor.</p>
+     <li data-md>
+      <p><em>slope</em>: an <code class="idl"><a data-link-type="idl" href="#mloperand" id="ref-for-mloperand②⑨①">MLOperand</a></code>. The slope tensor. Its shape is either the same as, or unidirectionally broadcastable to the shape of input tensor <em>input</em> according to <a data-link-type="biblio" href="#biblio-numpy-broadcasting-rule" title="General Broadcasting Rules of NumPy">[numpy-broadcasting-rule]</a>.</p>
+    </ul>
+    <p><strong>Returns:</strong></p>
+    <ul>
+     <li data-md>
+      <p>an <code class="idl"><a data-link-type="idl" href="#mloperand" id="ref-for-mloperand②⑨②">MLOperand</a></code>. The output tensor of the same shape as <em>input</em>.</p>
+    </ul>
+   </div>
+   <details class="algorithm" data-algorithm="prelu(input, slope)" data-algorithm-for="MLGraphBuilder" open>
+    <summary> The <dfn class="dfn-paneled idl-code" data-dfn-for="MLGraphBuilder" data-dfn-type="method" data-export id="dom-mlgraphbuilder-prelu"><code>prelu(<var>input</var>, <var>slope</var>)</code></dfn> method steps are: </summary>
+    <div class="algorithm-steps">
+     <ol>
       <li data-md>
-       <p><em>input</em>: an <code class="idl"><a data-link-type="idl" href="#mloperand" id="ref-for-mloperand②②⑦">MLOperand</a></code>. The input 4-D tensor. The logical shape
-is interpreted according to the value of <em>options.layout</em>.</p>
+       <p><a data-link-type="dfn" href="https://infra.spec.whatwg.org/#assert" id="ref-for-assert⑥⓪">Assert</a>: the type of <var>input</var> and <var>slope</var> is <code class="idl"><a data-link-type="idl" href="#mloperand" id="ref-for-mloperand②⑨③">MLOperand</a></code>.</p>
       <li data-md>
-       <p><em>options</em>: an optional <code class="idl"><a data-link-type="idl" href="#dictdef-mlpool2doptions" id="ref-for-dictdef-mlpool2doptions③">MLPool2dOptions</a></code>. The optional parameters of the operation.</p>
-       <ul>
-        <li data-md>
-         <p><em>windowDimensions</em>: a sequence of <code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#idl-unsigned-long" id="ref-for-idl-unsigned-long④⑥">unsigned long</a></code> of length 2. The dimensions of the sliding window,
-[window_height, window_width]. If not present, the window dimensions are assumed to be the height
-and width dimensions of the input shape.</p>
-        <li data-md>
-         <p><em>padding</em>: a sequence of <code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#idl-unsigned-long" id="ref-for-idl-unsigned-long④⑦">unsigned long</a></code> of length 4. The additional rows and columns added to the beginning and ending of each spatial dimension of <em>input</em>, [beginning_height, ending_height, beginning_width, ending_width]. If not present, the values are assumed to be [0,0,0,0].</p>
-        <li data-md>
-         <p><em>strides</em>: a sequence of <code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#idl-unsigned-long" id="ref-for-idl-unsigned-long④⑧">unsigned long</a></code> of length 2. The stride of the
-sliding window for each spatial dimension of <em>input</em>,
-[stride_height, stride_width]. If not present, the values are assumed to be [1,1].</p>
+       <p>Let <var>descriptor</var> be a new <code class="idl"><a data-link-type="idl" href="#dictdef-mloperanddescriptor" id="ref-for-dictdef-mloperanddescriptor②②">MLOperandDescriptor</a></code>.</p>
+      <li data-md>
+       <p>Set <var>descriptor</var>.<code class="idl"><a data-link-type="idl" href="#dom-mloperanddescriptor-type" id="ref-for-dom-mloperanddescriptor-type③⑤">type</a></code> to <var>input</var>.<code class="idl"><a data-link-type="idl" href="#dom-mloperand-descriptor-slot" id="ref-for-dom-mloperand-descriptor-slot⑧②">[[descriptor]]</a></code>.<code class="idl"><a data-link-type="idl" href="#dom-mloperanddescriptor-type" id="ref-for-dom-mloperanddescriptor-type③⑥">type</a></code>.</p>
+      <li data-md>
+       <p>Let <var>descriptor</var>.<code class="idl"><a data-link-type="idl" href="#dom-mloperanddescriptor-dimensions" id="ref-for-dom-mloperanddescriptor-dimensions⑧①">dimensions</a></code> be the result of running the <a data-link-type="dfn" href="#mlgraphbuilder-broadcast-shapes" id="ref-for-mlgraphbuilder-broadcast-shapes①">broadcast-shapes</a> steps given <var>input</var>.<code class="idl"><a data-link-type="idl" href="#dom-mloperand-descriptor-slot" id="ref-for-dom-mloperand-descriptor-slot⑧③">[[descriptor]]</a></code>.<code class="idl"><a data-link-type="idl" href="#dom-mloperanddescriptor-dimensions" id="ref-for-dom-mloperanddescriptor-dimensions⑧②">dimensions</a></code> and <var>slope</var>.<code class="idl"><a data-link-type="idl" href="#dom-mloperand-descriptor-slot" id="ref-for-dom-mloperand-descriptor-slot⑧④">[[descriptor]]</a></code>.<code class="idl"><a data-link-type="idl" href="#dom-mloperanddescriptor-dimensions" id="ref-for-dom-mloperanddescriptor-dimensions⑧③">dimensions</a></code>.</p>
+       <ol>
         <li data-md>
-         <p><em>dilations</em>: a sequence of <code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#idl-unsigned-long" id="ref-for-idl-unsigned-long④⑨">unsigned long</a></code> of length 2. The dilation factor
-for each spatial dimension of <em>input</em>, [dilation_height, dilation_width].
-If not present, the values are assumed to be [1,1].</p>
+         <p>If that <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#dfn-throw" id="ref-for-dfn-throw②①③">throws</a> an error, re-<a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#dfn-throw" id="ref-for-dfn-throw②①④">throw</a> the error.</p>
+       </ol>
+      <li data-md>
+       <p>If any of the following sub-steps fail, <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#dfn-throw" id="ref-for-dfn-throw②①⑤">throw</a> an "<code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#operationerror" id="ref-for-operationerror③⑦">OperationError</a></code>" <code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#idl-DOMException" id="ref-for-idl-DOMException①②⑥">DOMException</a></code>.</p>
+       <ol>
         <li data-md>
-         <p><em>autoPad</em>: an <code class="idl"><a data-link-type="idl" href="#enumdef-mlautopad" id="ref-for-enumdef-mlautopad⑤">MLAutoPad</a></code>. The automatic input padding options. By default, this argument is set to <em>"explicit"</em>, which means that the values in the <em>options.padding</em> array should be used for input padding. When the option is set other than <em>"explicit"</em>, the values in the <em>options.padding</em> array are ignored. With the <em>"same-upper"</em> option, the padding values are automatically computed such that the additional ending padding of the spatial input dimensions would allow all of the input values in the corresponding dimension to be filtered. The <em>"same-lower"</em> option is similar but padding is applied to the beginning padding of the spatial input dimensions instead of the ending one.</p>
+         <p>Let <var>output</var> be the result of <a data-link-type="dfn" href="#create-an-mloperand" id="ref-for-create-an-mloperand①⑧">creating an MLOperand</a> given <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#this" id="ref-for-this③②">this</a> and <var>descriptor</var>.</p>
         <li data-md>
-         <p><em>layout</em>: an <code class="idl"><a data-link-type="idl" href="#enumdef-mlinputoperandlayout" id="ref-for-enumdef-mlinputoperandlayout⑦">MLInputOperandLayout</a></code>. The default value is <em>"nchw"</em>. This option specifies the
-layout format of the input and output tensor as follow:</p>
-         <p>"nchw":</p>
-         <ul>
+         <p>Make a request to the underlying platform to:</p>
+         <ol>
           <li data-md>
-           <p>input tensor: [batches, channels, height, width]</p>
+           <p>Let <var>opImpl</var> be an <a data-link-type="dfn" href="https://infra.spec.whatwg.org/#implementation-defined" id="ref-for-implementation-defined⑤⓪">implementation-defined</a> platform operator for the PreLU operation, given <var>slope</var>.</p>
           <li data-md>
-           <p>output tensor: [batches, channels, height, width]</p>
-         </ul>
-         <p>"nhwc":</p>
-         <ul>
+           <p>Store a reference of <var>opImpl</var> in <var>output</var>.<code class="idl"><a data-link-type="idl" href="#dom-mloperand-operator-slot" id="ref-for-dom-mloperand-operator-slot②③">[[operator]]</a></code>.</p>
           <li data-md>
-           <p>input tensor: [batches, height, width, channels]</p>
+           <p>Create an <a data-link-type="dfn" href="https://infra.spec.whatwg.org/#implementation-defined" id="ref-for-implementation-defined⑤①">implementation-defined</a> platform operand <var>outputImpl</var> to represent the output, given <var>output</var> and <var>opImpl</var>.</p>
           <li data-md>
-           <p>output tensor: [batches, height, width, channels]</p>
-         </ul>
+           <p>Store a reference to <var>outputImpl</var> in <var>output</var>.<code class="idl"><a data-link-type="idl" href="#dom-mloperand-operand-slot" id="ref-for-dom-mloperand-operand-slot⑥⑤">[[operand]]</a></code>.</p>
+         </ol>
         <li data-md>
-         <p><em>roundingType</em>: an <code class="idl"><a data-link-type="idl" href="#enumdef-mlroundingtype" id="ref-for-enumdef-mlroundingtype①">MLRoundingType</a></code>. The option specifies the rounding function used to compute the output shape.</p>
+         <p>Connect <var>input</var>.<code class="idl"><a data-link-type="idl" href="#dom-mloperand-operand-slot" id="ref-for-dom-mloperand-operand-slot⑥⑥">[[operand]]</a></code> as input to <var>opImpl</var>.</p>
         <li data-md>
-         <p><em>outputSizes</em>: a sequence of <code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#idl-unsigned-long" id="ref-for-idl-unsigned-long⑤⓪">unsigned long</a></code> of length 2. The sizes of the two spacial dimensions of the output tensor. When the output sizes are explicitly specified, the options.roundingType is ignored. If not specified, the output sizes are automatically computed.</p>
-       </ul>
-     </ul>
-     <p><strong>Returns:</strong> an <code class="idl"><a data-link-type="idl" href="#mloperand" id="ref-for-mloperand②②⑧">MLOperand</a></code>. The output 4-D tensor that contains the
-    result of the reduction. The logical shape is interpreted according to the
-    value of <em>layout</em>. More specifically, if the <em>options.roundingType</em> is <em>"floor"</em>, the spatial dimensions of the output tensor can be calculated as follow:</p>
-     <p><em>output size = floor(1 + (input size - filter size + beginning padding + ending padding) / stride)</em></p>
-     <p>or if <em>options.roundingType</em> is <em>"ceil"</em>:</p>
-     <p><em>output size = ceil(1 + (input size - filter size + beginning padding + ending padding) / stride)</em></p>
-     <div class="note" role="note">
-       A <em>global</em> pooling operation such as one for the max pooling operation is a variant of pooling where the window dimensions is the spatial dimensions (last two dimensions) of the input shape, as follow. 
-<pre class="highlight"><c- c1>// 'global' max pooling</c->
-builder<c- p>.</c->maxPool2d<c- p>(</c->input<c- p>);</c->
-</pre>
-     </div>
-    </div>
-    <h4 class="heading settled" data-level="7.6.25" id="api-mlgraphbuilder-prelu"><span class="secno">7.6.25. </span><span class="content">The prelu() method</span><a class="self-link" href="#api-mlgraphbuilder-prelu"></a></h4>
-     Calculate the <a href="https://en.wikipedia.org/wiki/Rectifier_(neural_networks)#Parametric_ReLU">parametric version of rectified linear function (Parametric Relu)</a> on the input tensor element-wise. Parametric Relu is a type of leaky ReLU that, instead of having a scalar slope like 0.01, making the slope (coefficient of leakage) into a parameter that is learned during the model training phase of this operation. The calculation follows the expression <code>max(0, x) + slope ∗ min(0, x)</code>. 
-<pre class="idl highlight def"><c- b>partial</c-> <c- b>interface</c-> <a class="idl-code" data-link-type="interface" href="#mlgraphbuilder" id="ref-for-mlgraphbuilder④①"><c- g>MLGraphBuilder</c-></a> {
-  <a data-link-type="idl-name" href="#mloperand" id="ref-for-mloperand②②⑨"><c- n>MLOperand</c-></a> <dfn class="idl-code" data-dfn-for="MLGraphBuilder" data-dfn-type="method" data-export data-lt="prelu(x, slope)" id="dom-mlgraphbuilder-prelu"><code><c- g>prelu</c-></code><a class="self-link" href="#dom-mlgraphbuilder-prelu"></a></dfn>(<a data-link-type="idl-name" href="#mloperand" id="ref-for-mloperand②③⓪"><c- n>MLOperand</c-></a> <dfn class="idl-code" data-dfn-for="MLGraphBuilder/prelu(x, slope)" data-dfn-type="argument" data-export id="dom-mlgraphbuilder-prelu-x-slope-x"><code><c- g>x</c-></code><a class="self-link" href="#dom-mlgraphbuilder-prelu-x-slope-x"></a></dfn>, <a data-link-type="idl-name" href="#mloperand" id="ref-for-mloperand②③①"><c- n>MLOperand</c-></a> <dfn class="idl-code" data-dfn-for="MLGraphBuilder/prelu(x, slope)" data-dfn-type="argument" data-export id="dom-mlgraphbuilder-prelu-x-slope-slope"><code><c- g>slope</c-></code><a class="self-link" href="#dom-mlgraphbuilder-prelu-x-slope-slope"></a></dfn>);
-};
-</pre>
-    <div class="algorithm" data-algorithm="prelu">
-      <strong>Arguments:</strong> 
-     <ul>
-      <li data-md>
-       <p><em>x</em>: an <code class="idl"><a data-link-type="idl" href="#mloperand" id="ref-for-mloperand②③②">MLOperand</a></code>. The input tensor.</p>
-      <li data-md>
-       <p><em>slope</em>: an <code class="idl"><a data-link-type="idl" href="#mloperand" id="ref-for-mloperand②③③">MLOperand</a></code>. The slope tensor. Its shape is either the same as, or unidirectionally broadcastable to the shape of input tensor <em>x</em> according to <a data-link-type="biblio" href="#biblio-numpy-broadcasting-rule" title="General Broadcasting Rules of NumPy">[numpy-broadcasting-rule]</a>.</p>
-     </ul>
-     <p><strong>Returns:</strong></p>
-     <ul>
+         <p>Connect <var>output</var>.<code class="idl"><a data-link-type="idl" href="#dom-mloperand-operand-slot" id="ref-for-dom-mloperand-operand-slot⑥⑦">[[operand]]</a></code> as output to <var>opImpl</var>.</p>
+       </ol>
       <li data-md>
-       <p>an <code class="idl"><a data-link-type="idl" href="#mloperand" id="ref-for-mloperand②③④">MLOperand</a></code>. The output tensor of the same shape as <em>x</em>.</p>
-     </ul>
-     <div class="note" role="note">
-       The behavior of this operation can be generically emulated from the usage of
+       <p>Return <var>output</var>.</p>
+     </ol>
+    </div>
+   </details>
+   <div class="note" role="note">
+    <details open>
+     <summary> The behavior of this operation can be generically emulated from the usage of
     other operations as follow. However, user agents typically have a more
     efficient implementation for it, therefore its usage is encouraged from the
-    performance standpoint. 
+    performance standpoint. </summary>
 <pre class="highlight"><c- k>return</c-> builder<c- p>.</c->add<c- p>(</c->builder<c- p>.</c->max<c- p>(</c->builder<c- p>.</c->constant<c- p>(</c-><c- mf>0</c-><c- p>),</c-> x<c- p>),</c->
                    builder<c- p>.</c->mul<c- p>(</c->slope<c- p>,</c-> builder<c- p>.</c->min<c- p>(</c->builder<c- p>.</c->constant<c- p>(</c-><c- mf>0</c-><c- p>),</c-> x<c- p>)));</c->
 </pre>
-     </div>
-    </div>
-    <h4 class="heading settled" data-level="7.6.26" id="api-mlgraphbuilder-reduce"><span class="secno">7.6.26. </span><span class="content">Reduction operations</span><a class="self-link" href="#api-mlgraphbuilder-reduce"></a></h4>
-     Reduce the input along the dimensions given in <em>axes</em>. 
+    </details>
+   </div>
+   <h4 class="heading settled" data-level="7.6.27" id="api-mlgraphbuilder-reduce"><span class="secno">7.6.27. </span><span class="content">Reduction operations</span><a class="self-link" href="#api-mlgraphbuilder-reduce"></a></h4>
+    Reduce the input tensor along all dimensions, or along the axes specified in the <code class="idl"><a data-link-type="idl" href="#dom-mlreduceoptions-axes" id="ref-for-dom-mlreduceoptions-axes">axes</a></code> array parameter. For each specified axis, the dimension with that index is reduced, i.e. the resulting tensor will not contain it, unless the <code class="idl"><a data-link-type="idl" href="#dom-mlreduceoptions-keepdimensions" id="ref-for-dom-mlreduceoptions-keepdimensions">keepDimensions</a></code> option is specified. The values of the resulting tensor are calculated using the specified reduction function that takes as parameters all the values across the reduced dimension. 
 <pre class="idl highlight def"><c- b>dictionary</c-> <dfn class="dfn-paneled idl-code" data-dfn-type="dictionary" data-export id="dictdef-mlreduceoptions"><code><c- g>MLReduceOptions</c-></code></dfn> {
-  <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#idl-sequence" id="ref-for-idl-sequence②④"><c- b>sequence</c-></a>&lt;<a class="idl-code" data-link-type="interface" href="https://webidl.spec.whatwg.org/#idl-unsigned-long" id="ref-for-idl-unsigned-long⑤①"><c- b>unsigned</c-> <c- b>long</c-></a>> <dfn class="idl-code" data-default="null" data-dfn-for="MLReduceOptions" data-dfn-type="dict-member" data-export data-type="sequence<unsigned long> " id="dom-mlreduceoptions-axes"><code><c- g>axes</c-></code><a class="self-link" href="#dom-mlreduceoptions-axes"></a></dfn> = <c- b>null</c->;
-  <a class="idl-code" data-link-type="interface" href="https://webidl.spec.whatwg.org/#idl-boolean" id="ref-for-idl-boolean①②"><c- b>boolean</c-></a> <dfn class="idl-code" data-default="false" data-dfn-for="MLReduceOptions" data-dfn-type="dict-member" data-export data-type="boolean " id="dom-mlreduceoptions-keepdimensions"><code><c- g>keepDimensions</c-></code><a class="self-link" href="#dom-mlreduceoptions-keepdimensions"></a></dfn> = <c- b>false</c->;
-};
-
-<c- b>partial</c-> <c- b>interface</c-> <a class="idl-code" data-link-type="interface" href="#mlgraphbuilder" id="ref-for-mlgraphbuilder④②"><c- g>MLGraphBuilder</c-></a> {
-  <a data-link-type="idl-name" href="#mloperand" id="ref-for-mloperand②③⑤"><c- n>MLOperand</c-></a> <dfn class="idl-code" data-dfn-for="MLGraphBuilder" data-dfn-type="method" data-export data-lt="reduceL1(input, options)|reduceL1(input)" id="dom-mlgraphbuilder-reducel1"><code><c- g>reduceL1</c-></code><a class="self-link" href="#dom-mlgraphbuilder-reducel1"></a></dfn>(<a data-link-type="idl-name" href="#mloperand" id="ref-for-mloperand②③⑥"><c- n>MLOperand</c-></a> <dfn class="idl-code" data-dfn-for="MLGraphBuilder/reduceL1(input, options), MLGraphBuilder/reduceL1(input)" data-dfn-type="argument" data-export id="dom-mlgraphbuilder-reducel1-input-options-input"><code><c- g>input</c-></code><a class="self-link" href="#dom-mlgraphbuilder-reducel1-input-options-input"></a></dfn>, <c- b>optional</c-> <a data-link-type="idl-name" href="#dictdef-mlreduceoptions" id="ref-for-dictdef-mlreduceoptions"><c- n>MLReduceOptions</c-></a> <dfn class="idl-code" data-dfn-for="MLGraphBuilder/reduceL1(input, options), MLGraphBuilder/reduceL1(input)" data-dfn-type="argument" data-export id="dom-mlgraphbuilder-reducel1-input-options-options"><code><c- g>options</c-></code><a class="self-link" href="#dom-mlgraphbuilder-reducel1-input-options-options"></a></dfn> = {});
-  <a data-link-type="idl-name" href="#mloperand" id="ref-for-mloperand②③⑦"><c- n>MLOperand</c-></a> <dfn class="idl-code" data-dfn-for="MLGraphBuilder" data-dfn-type="method" data-export data-lt="reduceL2(input, options)|reduceL2(input)" id="dom-mlgraphbuilder-reducel2"><code><c- g>reduceL2</c-></code><a class="self-link" href="#dom-mlgraphbuilder-reducel2"></a></dfn>(<a data-link-type="idl-name" href="#mloperand" id="ref-for-mloperand②③⑧"><c- n>MLOperand</c-></a> <dfn class="idl-code" data-dfn-for="MLGraphBuilder/reduceL2(input, options), MLGraphBuilder/reduceL2(input)" data-dfn-type="argument" data-export id="dom-mlgraphbuilder-reducel2-input-options-input"><code><c- g>input</c-></code><a class="self-link" href="#dom-mlgraphbuilder-reducel2-input-options-input"></a></dfn>, <c- b>optional</c-> <a data-link-type="idl-name" href="#dictdef-mlreduceoptions" id="ref-for-dictdef-mlreduceoptions①"><c- n>MLReduceOptions</c-></a> <dfn class="idl-code" data-dfn-for="MLGraphBuilder/reduceL2(input, options), MLGraphBuilder/reduceL2(input)" data-dfn-type="argument" data-export id="dom-mlgraphbuilder-reducel2-input-options-options"><code><c- g>options</c-></code><a class="self-link" href="#dom-mlgraphbuilder-reducel2-input-options-options"></a></dfn> = {});
-  <a data-link-type="idl-name" href="#mloperand" id="ref-for-mloperand②③⑨"><c- n>MLOperand</c-></a> <dfn class="idl-code" data-dfn-for="MLGraphBuilder" data-dfn-type="method" data-export data-lt="reduceLogSum(input, options)|reduceLogSum(input)" id="dom-mlgraphbuilder-reducelogsum"><code><c- g>reduceLogSum</c-></code><a class="self-link" href="#dom-mlgraphbuilder-reducelogsum"></a></dfn>(<a data-link-type="idl-name" href="#mloperand" id="ref-for-mloperand②④⓪"><c- n>MLOperand</c-></a> <dfn class="idl-code" data-dfn-for="MLGraphBuilder/reduceLogSum(input, options), MLGraphBuilder/reduceLogSum(input)" data-dfn-type="argument" data-export id="dom-mlgraphbuilder-reducelogsum-input-options-input"><code><c- g>input</c-></code><a class="self-link" href="#dom-mlgraphbuilder-reducelogsum-input-options-input"></a></dfn>, <c- b>optional</c-> <a data-link-type="idl-name" href="#dictdef-mlreduceoptions" id="ref-for-dictdef-mlreduceoptions②"><c- n>MLReduceOptions</c-></a> <dfn class="idl-code" data-dfn-for="MLGraphBuilder/reduceLogSum(input, options), MLGraphBuilder/reduceLogSum(input)" data-dfn-type="argument" data-export id="dom-mlgraphbuilder-reducelogsum-input-options-options"><code><c- g>options</c-></code><a class="self-link" href="#dom-mlgraphbuilder-reducelogsum-input-options-options"></a></dfn> = {});
-  <a data-link-type="idl-name" href="#mloperand" id="ref-for-mloperand②④①"><c- n>MLOperand</c-></a> <dfn class="idl-code" data-dfn-for="MLGraphBuilder" data-dfn-type="method" data-export data-lt="reduceLogSumExp(input, options)|reduceLogSumExp(input)" id="dom-mlgraphbuilder-reducelogsumexp"><code><c- g>reduceLogSumExp</c-></code><a class="self-link" href="#dom-mlgraphbuilder-reducelogsumexp"></a></dfn>(<a data-link-type="idl-name" href="#mloperand" id="ref-for-mloperand②④②"><c- n>MLOperand</c-></a> <dfn class="idl-code" data-dfn-for="MLGraphBuilder/reduceLogSumExp(input, options), MLGraphBuilder/reduceLogSumExp(input)" data-dfn-type="argument" data-export id="dom-mlgraphbuilder-reducelogsumexp-input-options-input"><code><c- g>input</c-></code><a class="self-link" href="#dom-mlgraphbuilder-reducelogsumexp-input-options-input"></a></dfn>, <c- b>optional</c-> <a data-link-type="idl-name" href="#dictdef-mlreduceoptions" id="ref-for-dictdef-mlreduceoptions③"><c- n>MLReduceOptions</c-></a> <dfn class="idl-code" data-dfn-for="MLGraphBuilder/reduceLogSumExp(input, options), MLGraphBuilder/reduceLogSumExp(input)" data-dfn-type="argument" data-export id="dom-mlgraphbuilder-reducelogsumexp-input-options-options"><code><c- g>options</c-></code><a class="self-link" href="#dom-mlgraphbuilder-reducelogsumexp-input-options-options"></a></dfn> = {});
-  <a data-link-type="idl-name" href="#mloperand" id="ref-for-mloperand②④③"><c- n>MLOperand</c-></a> <dfn class="idl-code" data-dfn-for="MLGraphBuilder" data-dfn-type="method" data-export data-lt="reduceMax(input, options)|reduceMax(input)" id="dom-mlgraphbuilder-reducemax"><code><c- g>reduceMax</c-></code><a class="self-link" href="#dom-mlgraphbuilder-reducemax"></a></dfn>(<a data-link-type="idl-name" href="#mloperand" id="ref-for-mloperand②④④"><c- n>MLOperand</c-></a> <dfn class="idl-code" data-dfn-for="MLGraphBuilder/reduceMax(input, options), MLGraphBuilder/reduceMax(input)" data-dfn-type="argument" data-export id="dom-mlgraphbuilder-reducemax-input-options-input"><code><c- g>input</c-></code><a class="self-link" href="#dom-mlgraphbuilder-reducemax-input-options-input"></a></dfn>, <c- b>optional</c-> <a data-link-type="idl-name" href="#dictdef-mlreduceoptions" id="ref-for-dictdef-mlreduceoptions④"><c- n>MLReduceOptions</c-></a> <dfn class="idl-code" data-dfn-for="MLGraphBuilder/reduceMax(input, options), MLGraphBuilder/reduceMax(input)" data-dfn-type="argument" data-export id="dom-mlgraphbuilder-reducemax-input-options-options"><code><c- g>options</c-></code><a class="self-link" href="#dom-mlgraphbuilder-reducemax-input-options-options"></a></dfn> = {});
-  <a data-link-type="idl-name" href="#mloperand" id="ref-for-mloperand②④⑤"><c- n>MLOperand</c-></a> <dfn class="idl-code" data-dfn-for="MLGraphBuilder" data-dfn-type="method" data-export data-lt="reduceMean(input, options)|reduceMean(input)" id="dom-mlgraphbuilder-reducemean"><code><c- g>reduceMean</c-></code><a class="self-link" href="#dom-mlgraphbuilder-reducemean"></a></dfn>(<a data-link-type="idl-name" href="#mloperand" id="ref-for-mloperand②④⑥"><c- n>MLOperand</c-></a> <dfn class="idl-code" data-dfn-for="MLGraphBuilder/reduceMean(input, options), MLGraphBuilder/reduceMean(input)" data-dfn-type="argument" data-export id="dom-mlgraphbuilder-reducemean-input-options-input"><code><c- g>input</c-></code><a class="self-link" href="#dom-mlgraphbuilder-reducemean-input-options-input"></a></dfn>, <c- b>optional</c-> <a data-link-type="idl-name" href="#dictdef-mlreduceoptions" id="ref-for-dictdef-mlreduceoptions⑤"><c- n>MLReduceOptions</c-></a> <dfn class="idl-code" data-dfn-for="MLGraphBuilder/reduceMean(input, options), MLGraphBuilder/reduceMean(input)" data-dfn-type="argument" data-export id="dom-mlgraphbuilder-reducemean-input-options-options"><code><c- g>options</c-></code><a class="self-link" href="#dom-mlgraphbuilder-reducemean-input-options-options"></a></dfn> = {});
-  <a data-link-type="idl-name" href="#mloperand" id="ref-for-mloperand②④⑦"><c- n>MLOperand</c-></a> <dfn class="idl-code" data-dfn-for="MLGraphBuilder" data-dfn-type="method" data-export data-lt="reduceMin(input, options)|reduceMin(input)" id="dom-mlgraphbuilder-reducemin"><code><c- g>reduceMin</c-></code><a class="self-link" href="#dom-mlgraphbuilder-reducemin"></a></dfn>(<a data-link-type="idl-name" href="#mloperand" id="ref-for-mloperand②④⑧"><c- n>MLOperand</c-></a> <dfn class="idl-code" data-dfn-for="MLGraphBuilder/reduceMin(input, options), MLGraphBuilder/reduceMin(input)" data-dfn-type="argument" data-export id="dom-mlgraphbuilder-reducemin-input-options-input"><code><c- g>input</c-></code><a class="self-link" href="#dom-mlgraphbuilder-reducemin-input-options-input"></a></dfn>, <c- b>optional</c-> <a data-link-type="idl-name" href="#dictdef-mlreduceoptions" id="ref-for-dictdef-mlreduceoptions⑥"><c- n>MLReduceOptions</c-></a> <dfn class="idl-code" data-dfn-for="MLGraphBuilder/reduceMin(input, options), MLGraphBuilder/reduceMin(input)" data-dfn-type="argument" data-export id="dom-mlgraphbuilder-reducemin-input-options-options"><code><c- g>options</c-></code><a class="self-link" href="#dom-mlgraphbuilder-reducemin-input-options-options"></a></dfn> = {});
-  <a data-link-type="idl-name" href="#mloperand" id="ref-for-mloperand②④⑨"><c- n>MLOperand</c-></a> <dfn class="idl-code" data-dfn-for="MLGraphBuilder" data-dfn-type="method" data-export data-lt="reduceProduct(input, options)|reduceProduct(input)" id="dom-mlgraphbuilder-reduceproduct"><code><c- g>reduceProduct</c-></code><a class="self-link" href="#dom-mlgraphbuilder-reduceproduct"></a></dfn>(<a data-link-type="idl-name" href="#mloperand" id="ref-for-mloperand②⑤⓪"><c- n>MLOperand</c-></a> <dfn class="idl-code" data-dfn-for="MLGraphBuilder/reduceProduct(input, options), MLGraphBuilder/reduceProduct(input)" data-dfn-type="argument" data-export id="dom-mlgraphbuilder-reduceproduct-input-options-input"><code><c- g>input</c-></code><a class="self-link" href="#dom-mlgraphbuilder-reduceproduct-input-options-input"></a></dfn>, <c- b>optional</c-> <a data-link-type="idl-name" href="#dictdef-mlreduceoptions" id="ref-for-dictdef-mlreduceoptions⑦"><c- n>MLReduceOptions</c-></a> <dfn class="idl-code" data-dfn-for="MLGraphBuilder/reduceProduct(input, options), MLGraphBuilder/reduceProduct(input)" data-dfn-type="argument" data-export id="dom-mlgraphbuilder-reduceproduct-input-options-options"><code><c- g>options</c-></code><a class="self-link" href="#dom-mlgraphbuilder-reduceproduct-input-options-options"></a></dfn> = {});
-  <a data-link-type="idl-name" href="#mloperand" id="ref-for-mloperand②⑤①"><c- n>MLOperand</c-></a> <dfn class="idl-code" data-dfn-for="MLGraphBuilder" data-dfn-type="method" data-export data-lt="reduceSum(input, options)|reduceSum(input)" id="dom-mlgraphbuilder-reducesum"><code><c- g>reduceSum</c-></code><a class="self-link" href="#dom-mlgraphbuilder-reducesum"></a></dfn>(<a data-link-type="idl-name" href="#mloperand" id="ref-for-mloperand②⑤②"><c- n>MLOperand</c-></a> <dfn class="idl-code" data-dfn-for="MLGraphBuilder/reduceSum(input, options), MLGraphBuilder/reduceSum(input)" data-dfn-type="argument" data-export id="dom-mlgraphbuilder-reducesum-input-options-input"><code><c- g>input</c-></code><a class="self-link" href="#dom-mlgraphbuilder-reducesum-input-options-input"></a></dfn>, <c- b>optional</c-> <a data-link-type="idl-name" href="#dictdef-mlreduceoptions" id="ref-for-dictdef-mlreduceoptions⑧"><c- n>MLReduceOptions</c-></a> <dfn class="idl-code" data-dfn-for="MLGraphBuilder/reduceSum(input, options), MLGraphBuilder/reduceSum(input)" data-dfn-type="argument" data-export id="dom-mlgraphbuilder-reducesum-input-options-options"><code><c- g>options</c-></code><a class="self-link" href="#dom-mlgraphbuilder-reducesum-input-options-options"></a></dfn> = {});
-  <a data-link-type="idl-name" href="#mloperand" id="ref-for-mloperand②⑤③"><c- n>MLOperand</c-></a> <dfn class="idl-code" data-dfn-for="MLGraphBuilder" data-dfn-type="method" data-export data-lt="reduceSumSquare(input, options)|reduceSumSquare(input)" id="dom-mlgraphbuilder-reducesumsquare"><code><c- g>reduceSumSquare</c-></code><a class="self-link" href="#dom-mlgraphbuilder-reducesumsquare"></a></dfn>(<a data-link-type="idl-name" href="#mloperand" id="ref-for-mloperand②⑤④"><c- n>MLOperand</c-></a> <dfn class="idl-code" data-dfn-for="MLGraphBuilder/reduceSumSquare(input, options), MLGraphBuilder/reduceSumSquare(input)" data-dfn-type="argument" data-export id="dom-mlgraphbuilder-reducesumsquare-input-options-input"><code><c- g>input</c-></code><a class="self-link" href="#dom-mlgraphbuilder-reducesumsquare-input-options-input"></a></dfn>, <c- b>optional</c-> <a data-link-type="idl-name" href="#dictdef-mlreduceoptions" id="ref-for-dictdef-mlreduceoptions⑨"><c- n>MLReduceOptions</c-></a> <dfn class="idl-code" data-dfn-for="MLGraphBuilder/reduceSumSquare(input, options), MLGraphBuilder/reduceSumSquare(input)" data-dfn-type="argument" data-export id="dom-mlgraphbuilder-reducesumsquare-input-options-options"><code><c- g>options</c-></code><a class="self-link" href="#dom-mlgraphbuilder-reducesumsquare-input-options-options"></a></dfn> = {});
+  <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#idl-sequence" id="ref-for-idl-sequence②④"><c- b>sequence</c-></a>&lt;<a class="idl-code" data-link-type="interface" href="https://webidl.spec.whatwg.org/#idl-unsigned-long" id="ref-for-idl-unsigned-long⑤③"><c- b>unsigned</c-> <c- b>long</c-></a>> <dfn class="dfn-paneled idl-code" data-default="null" data-dfn-for="MLReduceOptions" data-dfn-type="dict-member" data-export data-type="sequence<unsigned long> " id="dom-mlreduceoptions-axes"><code><c- g>axes</c-></code></dfn> = <c- b>null</c->;
+  <a class="idl-code" data-link-type="interface" href="https://webidl.spec.whatwg.org/#idl-boolean" id="ref-for-idl-boolean①⑧"><c- b>boolean</c-></a> <dfn class="dfn-paneled idl-code" data-default="false" data-dfn-for="MLReduceOptions" data-dfn-type="dict-member" data-export data-type="boolean " id="dom-mlreduceoptions-keepdimensions"><code><c- g>keepDimensions</c-></code></dfn> = <c- b>false</c->;
+};
+
+<c- b>partial</c-> <c- b>interface</c-> <a class="idl-code" data-link-type="interface" href="#mlgraphbuilder" id="ref-for-mlgraphbuilder④⑤"><c- g>MLGraphBuilder</c-></a> {
+  <a data-link-type="idl-name" href="#mloperand" id="ref-for-mloperand②⑨④"><c- n>MLOperand</c-></a> <a class="idl-code" data-link-type="method" href="#dom-mlgraphbuilder-reducel1" id="ref-for-dom-mlgraphbuilder-reducel1"><c- g>reduceL1</c-></a>(<a data-link-type="idl-name" href="#mloperand" id="ref-for-mloperand②⑨⑤"><c- n>MLOperand</c-></a> <dfn class="idl-code" data-dfn-for="MLGraphBuilder/reduceL1(input, options), MLGraphBuilder/reduceL1(input)" data-dfn-type="argument" data-export id="dom-mlgraphbuilder-reducel1-input-options-input"><code><c- g>input</c-></code><a class="self-link" href="#dom-mlgraphbuilder-reducel1-input-options-input"></a></dfn>, <c- b>optional</c-> <a data-link-type="idl-name" href="#dictdef-mlreduceoptions" id="ref-for-dictdef-mlreduceoptions"><c- n>MLReduceOptions</c-></a> <dfn class="idl-code" data-dfn-for="MLGraphBuilder/reduceL1(input, options), MLGraphBuilder/reduceL1(input)" data-dfn-type="argument" data-export id="dom-mlgraphbuilder-reducel1-input-options-options"><code><c- g>options</c-></code><a class="self-link" href="#dom-mlgraphbuilder-reducel1-input-options-options"></a></dfn> = {});
+  <a data-link-type="idl-name" href="#mloperand" id="ref-for-mloperand②⑨⑥"><c- n>MLOperand</c-></a> <a class="idl-code" data-link-type="method" href="#dom-mlgraphbuilder-reducel2" id="ref-for-dom-mlgraphbuilder-reducel2"><c- g>reduceL2</c-></a>(<a data-link-type="idl-name" href="#mloperand" id="ref-for-mloperand②⑨⑦"><c- n>MLOperand</c-></a> <dfn class="idl-code" data-dfn-for="MLGraphBuilder/reduceL2(input, options), MLGraphBuilder/reduceL2(input)" data-dfn-type="argument" data-export id="dom-mlgraphbuilder-reducel2-input-options-input"><code><c- g>input</c-></code><a class="self-link" href="#dom-mlgraphbuilder-reducel2-input-options-input"></a></dfn>, <c- b>optional</c-> <a data-link-type="idl-name" href="#dictdef-mlreduceoptions" id="ref-for-dictdef-mlreduceoptions①"><c- n>MLReduceOptions</c-></a> <dfn class="idl-code" data-dfn-for="MLGraphBuilder/reduceL2(input, options), MLGraphBuilder/reduceL2(input)" data-dfn-type="argument" data-export id="dom-mlgraphbuilder-reducel2-input-options-options"><code><c- g>options</c-></code><a class="self-link" href="#dom-mlgraphbuilder-reducel2-input-options-options"></a></dfn> = {});
+  <a data-link-type="idl-name" href="#mloperand" id="ref-for-mloperand②⑨⑧"><c- n>MLOperand</c-></a> <a class="idl-code" data-link-type="method" href="#dom-mlgraphbuilder-reducelogsum" id="ref-for-dom-mlgraphbuilder-reducelogsum"><c- g>reduceLogSum</c-></a>(<a data-link-type="idl-name" href="#mloperand" id="ref-for-mloperand②⑨⑨"><c- n>MLOperand</c-></a> <dfn class="idl-code" data-dfn-for="MLGraphBuilder/reduceLogSum(input, options), MLGraphBuilder/reduceLogSum(input)" data-dfn-type="argument" data-export id="dom-mlgraphbuilder-reducelogsum-input-options-input"><code><c- g>input</c-></code><a class="self-link" href="#dom-mlgraphbuilder-reducelogsum-input-options-input"></a></dfn>, <c- b>optional</c-> <a data-link-type="idl-name" href="#dictdef-mlreduceoptions" id="ref-for-dictdef-mlreduceoptions②"><c- n>MLReduceOptions</c-></a> <dfn class="idl-code" data-dfn-for="MLGraphBuilder/reduceLogSum(input, options), MLGraphBuilder/reduceLogSum(input)" data-dfn-type="argument" data-export id="dom-mlgraphbuilder-reducelogsum-input-options-options"><code><c- g>options</c-></code><a class="self-link" href="#dom-mlgraphbuilder-reducelogsum-input-options-options"></a></dfn> = {});
+  <a data-link-type="idl-name" href="#mloperand" id="ref-for-mloperand③⓪⓪"><c- n>MLOperand</c-></a> <a class="idl-code" data-link-type="method" href="#dom-mlgraphbuilder-reducelogsumexp" id="ref-for-dom-mlgraphbuilder-reducelogsumexp"><c- g>reduceLogSumExp</c-></a>(<a data-link-type="idl-name" href="#mloperand" id="ref-for-mloperand③⓪①"><c- n>MLOperand</c-></a> <dfn class="idl-code" data-dfn-for="MLGraphBuilder/reduceLogSumExp(input, options), MLGraphBuilder/reduceLogSumExp(input)" data-dfn-type="argument" data-export id="dom-mlgraphbuilder-reducelogsumexp-input-options-input"><code><c- g>input</c-></code><a class="self-link" href="#dom-mlgraphbuilder-reducelogsumexp-input-options-input"></a></dfn>, <c- b>optional</c-> <a data-link-type="idl-name" href="#dictdef-mlreduceoptions" id="ref-for-dictdef-mlreduceoptions③"><c- n>MLReduceOptions</c-></a> <dfn class="idl-code" data-dfn-for="MLGraphBuilder/reduceLogSumExp(input, options), MLGraphBuilder/reduceLogSumExp(input)" data-dfn-type="argument" data-export id="dom-mlgraphbuilder-reducelogsumexp-input-options-options"><code><c- g>options</c-></code><a class="self-link" href="#dom-mlgraphbuilder-reducelogsumexp-input-options-options"></a></dfn> = {});
+  <a data-link-type="idl-name" href="#mloperand" id="ref-for-mloperand③⓪②"><c- n>MLOperand</c-></a> <a class="idl-code" data-link-type="method" href="#dom-mlgraphbuilder-reducemax" id="ref-for-dom-mlgraphbuilder-reducemax"><c- g>reduceMax</c-></a>(<a data-link-type="idl-name" href="#mloperand" id="ref-for-mloperand③⓪③"><c- n>MLOperand</c-></a> <dfn class="idl-code" data-dfn-for="MLGraphBuilder/reduceMax(input, options), MLGraphBuilder/reduceMax(input)" data-dfn-type="argument" data-export id="dom-mlgraphbuilder-reducemax-input-options-input"><code><c- g>input</c-></code><a class="self-link" href="#dom-mlgraphbuilder-reducemax-input-options-input"></a></dfn>, <c- b>optional</c-> <a data-link-type="idl-name" href="#dictdef-mlreduceoptions" id="ref-for-dictdef-mlreduceoptions④"><c- n>MLReduceOptions</c-></a> <dfn class="idl-code" data-dfn-for="MLGraphBuilder/reduceMax(input, options), MLGraphBuilder/reduceMax(input)" data-dfn-type="argument" data-export id="dom-mlgraphbuilder-reducemax-input-options-options"><code><c- g>options</c-></code><a class="self-link" href="#dom-mlgraphbuilder-reducemax-input-options-options"></a></dfn> = {});
+  <a data-link-type="idl-name" href="#mloperand" id="ref-for-mloperand③⓪④"><c- n>MLOperand</c-></a> <a class="idl-code" data-link-type="method" href="#dom-mlgraphbuilder-reducemean" id="ref-for-dom-mlgraphbuilder-reducemean"><c- g>reduceMean</c-></a>(<a data-link-type="idl-name" href="#mloperand" id="ref-for-mloperand③⓪⑤"><c- n>MLOperand</c-></a> <dfn class="idl-code" data-dfn-for="MLGraphBuilder/reduceMean(input, options), MLGraphBuilder/reduceMean(input)" data-dfn-type="argument" data-export id="dom-mlgraphbuilder-reducemean-input-options-input"><code><c- g>input</c-></code><a class="self-link" href="#dom-mlgraphbuilder-reducemean-input-options-input"></a></dfn>, <c- b>optional</c-> <a data-link-type="idl-name" href="#dictdef-mlreduceoptions" id="ref-for-dictdef-mlreduceoptions⑤"><c- n>MLReduceOptions</c-></a> <dfn class="idl-code" data-dfn-for="MLGraphBuilder/reduceMean(input, options), MLGraphBuilder/reduceMean(input)" data-dfn-type="argument" data-export id="dom-mlgraphbuilder-reducemean-input-options-options"><code><c- g>options</c-></code><a class="self-link" href="#dom-mlgraphbuilder-reducemean-input-options-options"></a></dfn> = {});
+  <a data-link-type="idl-name" href="#mloperand" id="ref-for-mloperand③⓪⑥"><c- n>MLOperand</c-></a> <a class="idl-code" data-link-type="method" href="#dom-mlgraphbuilder-reducemin" id="ref-for-dom-mlgraphbuilder-reducemin"><c- g>reduceMin</c-></a>(<a data-link-type="idl-name" href="#mloperand" id="ref-for-mloperand③⓪⑦"><c- n>MLOperand</c-></a> <dfn class="idl-code" data-dfn-for="MLGraphBuilder/reduceMin(input, options), MLGraphBuilder/reduceMin(input)" data-dfn-type="argument" data-export id="dom-mlgraphbuilder-reducemin-input-options-input"><code><c- g>input</c-></code><a class="self-link" href="#dom-mlgraphbuilder-reducemin-input-options-input"></a></dfn>, <c- b>optional</c-> <a data-link-type="idl-name" href="#dictdef-mlreduceoptions" id="ref-for-dictdef-mlreduceoptions⑥"><c- n>MLReduceOptions</c-></a> <dfn class="idl-code" data-dfn-for="MLGraphBuilder/reduceMin(input, options), MLGraphBuilder/reduceMin(input)" data-dfn-type="argument" data-export id="dom-mlgraphbuilder-reducemin-input-options-options"><code><c- g>options</c-></code><a class="self-link" href="#dom-mlgraphbuilder-reducemin-input-options-options"></a></dfn> = {});
+  <a data-link-type="idl-name" href="#mloperand" id="ref-for-mloperand③⓪⑧"><c- n>MLOperand</c-></a> <a class="idl-code" data-link-type="method" href="#dom-mlgraphbuilder-reduceproduct" id="ref-for-dom-mlgraphbuilder-reduceproduct"><c- g>reduceProduct</c-></a>(<a data-link-type="idl-name" href="#mloperand" id="ref-for-mloperand③⓪⑨"><c- n>MLOperand</c-></a> <dfn class="idl-code" data-dfn-for="MLGraphBuilder/reduceProduct(input, options), MLGraphBuilder/reduceProduct(input)" data-dfn-type="argument" data-export id="dom-mlgraphbuilder-reduceproduct-input-options-input"><code><c- g>input</c-></code><a class="self-link" href="#dom-mlgraphbuilder-reduceproduct-input-options-input"></a></dfn>, <c- b>optional</c-> <a data-link-type="idl-name" href="#dictdef-mlreduceoptions" id="ref-for-dictdef-mlreduceoptions⑦"><c- n>MLReduceOptions</c-></a> <dfn class="idl-code" data-dfn-for="MLGraphBuilder/reduceProduct(input, options), MLGraphBuilder/reduceProduct(input)" data-dfn-type="argument" data-export id="dom-mlgraphbuilder-reduceproduct-input-options-options"><code><c- g>options</c-></code><a class="self-link" href="#dom-mlgraphbuilder-reduceproduct-input-options-options"></a></dfn> = {});
+  <a data-link-type="idl-name" href="#mloperand" id="ref-for-mloperand③①⓪"><c- n>MLOperand</c-></a> <a class="idl-code" data-link-type="method" href="#dom-mlgraphbuilder-reducesum" id="ref-for-dom-mlgraphbuilder-reducesum"><c- g>reduceSum</c-></a>(<a data-link-type="idl-name" href="#mloperand" id="ref-for-mloperand③①①"><c- n>MLOperand</c-></a> <dfn class="idl-code" data-dfn-for="MLGraphBuilder/reduceSum(input, options), MLGraphBuilder/reduceSum(input)" data-dfn-type="argument" data-export id="dom-mlgraphbuilder-reducesum-input-options-input"><code><c- g>input</c-></code><a class="self-link" href="#dom-mlgraphbuilder-reducesum-input-options-input"></a></dfn>, <c- b>optional</c-> <a data-link-type="idl-name" href="#dictdef-mlreduceoptions" id="ref-for-dictdef-mlreduceoptions⑧"><c- n>MLReduceOptions</c-></a> <dfn class="idl-code" data-dfn-for="MLGraphBuilder/reduceSum(input, options), MLGraphBuilder/reduceSum(input)" data-dfn-type="argument" data-export id="dom-mlgraphbuilder-reducesum-input-options-options"><code><c- g>options</c-></code><a class="self-link" href="#dom-mlgraphbuilder-reducesum-input-options-options"></a></dfn> = {});
+  <a data-link-type="idl-name" href="#mloperand" id="ref-for-mloperand③①②"><c- n>MLOperand</c-></a> <a class="idl-code" data-link-type="method" href="#dom-mlgraphbuilder-reducesumsquare" id="ref-for-dom-mlgraphbuilder-reducesumsquare"><c- g>reduceSumSquare</c-></a>(<a data-link-type="idl-name" href="#mloperand" id="ref-for-mloperand③①③"><c- n>MLOperand</c-></a> <dfn class="idl-code" data-dfn-for="MLGraphBuilder/reduceSumSquare(input, options), MLGraphBuilder/reduceSumSquare(input)" data-dfn-type="argument" data-export id="dom-mlgraphbuilder-reducesumsquare-input-options-input"><code><c- g>input</c-></code><a class="self-link" href="#dom-mlgraphbuilder-reducesumsquare-input-options-input"></a></dfn>, <c- b>optional</c-> <a data-link-type="idl-name" href="#dictdef-mlreduceoptions" id="ref-for-dictdef-mlreduceoptions⑨"><c- n>MLReduceOptions</c-></a> <dfn class="idl-code" data-dfn-for="MLGraphBuilder/reduceSumSquare(input, options), MLGraphBuilder/reduceSumSquare(input)" data-dfn-type="argument" data-export id="dom-mlgraphbuilder-reducesumsquare-input-options-options"><code><c- g>options</c-></code><a class="self-link" href="#dom-mlgraphbuilder-reducesumsquare-input-options-options"></a></dfn> = {});
 };
 </pre>
-    <div class="algorithm" data-algorithm="reduce">
-      <strong>Arguments:</strong> 
-     <ul>
+   <div>
+     <strong>Arguments:</strong> 
+    <ul>
+     <li data-md>
+      <p><em>input</em>: an <code class="idl"><a data-link-type="idl" href="#mloperand" id="ref-for-mloperand③①④">MLOperand</a></code>. The input tensor.</p>
+     <li data-md>
+      <p><em>options</em>: an optional <code class="idl"><a data-link-type="idl" href="#dictdef-mlreduceoptions" id="ref-for-dictdef-mlreduceoptions①⓪">MLReduceOptions</a></code>. The optional parameters of the operation.</p>
+      <ul>
+       <li data-md>
+        <p><em>axes</em>: a sequence of <code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#idl-unsigned-long" id="ref-for-idl-unsigned-long⑤④">unsigned long</a></code>. The dimensions to reduce. The values in the sequence must be in the range [0, N-1] where N is the <a data-link-type="dfn" href="#mloperand-rank" id="ref-for-mloperand-rank①⑨">rank</a> of the input tensor.
+If not present, all dimensions are reduced.</p>
+       <li data-md>
+        <p><em>keepDimensions</em>: a <code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#idl-boolean" id="ref-for-idl-boolean①⑨">boolean</a></code>. If true, retains reduced dimensions with <a data-link-type="dfn" href="https://infra.spec.whatwg.org/#list-size" id="ref-for-list-size④⑤">size</a> 1.
+The default value is false.</p>
+      </ul>
+    </ul>
+    <p><strong>Returns:</strong> an <code class="idl"><a data-link-type="idl" href="#mloperand" id="ref-for-mloperand③①⑤">MLOperand</a></code>. The reduced output tensor.</p>
+   </div>
+   <div class="note" role="note">
+     <strong>Reduction types:</strong> 
+    <ul>
+     <li data-md>
+      <p><em>L1</em>: Compute the <a href="https://mathworld.wolfram.com/L1-Norm.html">L1 norm</a> of all the input values along the axes.</p>
+     <li data-md>
+      <p><em>L2</em>: Compute the <a href="https://mathworld.wolfram.com/L2-Norm.html">L2 norm</a> of all the input values along the axes.</p>
+     <li data-md>
+      <p><em>LogSum</em>: Compute the log value of the sum of all the input values along the axes.</p>
+     <li data-md>
+      <p><em>LogSumExp</em>: Compute the log value of the sum of the exponent of all the input values along the axes.</p>
+     <li data-md>
+      <p><em>Max</em>: Compute the maximum value of all the input values along the axes.</p>
+     <li data-md>
+      <p><em>Mean</em>: Compute the average value of all the input values along the axes.</p>
+     <li data-md>
+      <p><em>Min</em>: Compute the minimum value of all the input values along the axes.</p>
+     <li data-md>
+      <p><em>Product</em>: Compute the product of all the input values along the axes.</p>
+     <li data-md>
+      <p><em>Sum</em>: Compute the sum of all the input values along the axes.</p>
+     <li data-md>
+      <p><em>SumSquare</em>: Compute the sum of the square of all the input values along the axes.</p>
+    </ul>
+   </div>
+   <details class="algorithm" data-algorithm="reduce-op" data-algorithm-for="MLGraphBuilder" open>
+    <summary> To <dfn class="dfn-paneled" data-dfn-for="MLGraphBuilder" data-dfn-type="dfn" data-lt="reduce-op" data-noexport id="mlgraphbuilder-reduce-op">create reduce operation</dfn> given <var>op</var>, <var>input</var> and <var>options</var>, run the following steps: </summary>
+    <div class="algorithm-steps">
+     <ol>
       <li data-md>
-       <p><em>input</em>: an <code class="idl"><a data-link-type="idl" href="#mloperand" id="ref-for-mloperand②⑤⑤">MLOperand</a></code>. The input tensor.</p>
+       <p><a data-link-type="dfn" href="https://infra.spec.whatwg.org/#assert" id="ref-for-assert⑥①">Assert</a>: <var>op</var> is one of "reduceL1", "reduceL2", "reduceLogSum", "reduceLogSumExp", "reduceMax", "reduceMean", "reduceMin", "reduceProduct", "reduceSum", "reduceSumSquare".</p>
       <li data-md>
-       <p><em>options</em>: an optional <code class="idl"><a data-link-type="idl" href="#dictdef-mlreduceoptions" id="ref-for-dictdef-mlreduceoptions①⓪">MLReduceOptions</a></code>. The optional parameters of the operation.</p>
-       <ul>
+       <p><a data-link-type="dfn" href="https://infra.spec.whatwg.org/#assert" id="ref-for-assert⑥②">Assert</a>: the type of <var>input</var> is <code class="idl"><a data-link-type="idl" href="#mloperand" id="ref-for-mloperand③①⑥">MLOperand</a></code>.</p>
+      <li data-md>
+       <p>If <var>options</var>.<code class="idl"><a data-link-type="idl" href="#dom-mlreduceoptions-axes" id="ref-for-dom-mlreduceoptions-axes①">axes</a></code> <a data-link-type="dfn" href="https://infra.spec.whatwg.org/#map-exists" id="ref-for-map-exists⑤①">exists</a>, if any of its elements is not in <a data-link-type="dfn" href="https://infra.spec.whatwg.org/#the-range" id="ref-for-the-range⑤">the range</a> 0 to the <a data-link-type="dfn" href="#mloperand-rank" id="ref-for-mloperand-rank②⓪">rank</a> of <var>input</var>, exclusive, then <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#dfn-throw" id="ref-for-dfn-throw②①⑥">throw</a> a "<code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#dataerror" id="ref-for-dataerror⑧⓪">DataError</a></code>" <code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#idl-DOMException" id="ref-for-idl-DOMException①②⑦">DOMException</a></code>.</p>
+      <li data-md>
+       <p>If any of the following sub-steps fail, <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#dfn-throw" id="ref-for-dfn-throw②①⑦">throw</a> an "<code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#operationerror" id="ref-for-operationerror③⑧">OperationError</a></code>" <code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#idl-DOMException" id="ref-for-idl-DOMException①②⑧">DOMException</a></code>.</p>
+       <ol>
         <li data-md>
-         <p><em>axes</em>: a sequence of <code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#idl-unsigned-long" id="ref-for-idl-unsigned-long⑤②">unsigned long</a></code>. The dimensions to reduce. The values in the sequence must be in the range [0, N-1] where N is the rank of input tensor.
-If not present, all dimensions are reduced.</p>
+         <p>Let <var>output</var> be the result of <a data-link-type="dfn" href="#copy-an-mloperand" id="ref-for-copy-an-mloperand⑧">copying an MLOperand</a> given <var>input</var>.</p>
         <li data-md>
-         <p><em>keepDimensions</em>: a <code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#idl-boolean" id="ref-for-idl-boolean①③">boolean</a></code>. If true, retains reduced dimensions with size of 1.
-The default value is false.</p>
-       </ul>
-     </ul>
-     <p><strong>Returns:</strong> an <code class="idl"><a data-link-type="idl" href="#mloperand" id="ref-for-mloperand②⑤⑥">MLOperand</a></code>. The reduced output tensor.</p>
-     <p><strong>Reduction types:</strong></p>
-     <ul>
+         <p>Make a request to the underlying platform to:</p>
+         <ol>
+          <li data-md>
+           <p>Let <var>opImpl</var> be an <a data-link-type="dfn" href="https://infra.spec.whatwg.org/#implementation-defined" id="ref-for-implementation-defined⑤②">implementation-defined</a> platform operator for the <var>op</var> reduce operation, given <var>options</var>.</p>
+          <li data-md>
+           <p>Store a reference of <var>opImpl</var> in <var>output</var>.<code class="idl"><a data-link-type="idl" href="#dom-mloperand-operator-slot" id="ref-for-dom-mloperand-operator-slot②④">[[operator]]</a></code>.</p>
+          <li data-md>
+           <p>Create an <a data-link-type="dfn" href="https://infra.spec.whatwg.org/#implementation-defined" id="ref-for-implementation-defined⑤③">implementation-defined</a> platform operand <var>outputImpl</var> to represent the output, given <var>output</var> and <var>opImpl</var>.</p>
+          <li data-md>
+           <p>Store a reference to <var>outputImpl</var> in <var>output</var>.<code class="idl"><a data-link-type="idl" href="#dom-mloperand-operand-slot" id="ref-for-dom-mloperand-operand-slot⑥⑧">[[operand]]</a></code>.</p>
+         </ol>
+        <li data-md>
+         <p>Connect <var>input</var>.<code class="idl"><a data-link-type="idl" href="#dom-mloperand-operand-slot" id="ref-for-dom-mloperand-operand-slot⑥⑨">[[operand]]</a></code> as input to <var>opImpl</var>.</p>
+        <li data-md>
+         <p>Connect <var>output</var>.<code class="idl"><a data-link-type="idl" href="#dom-mloperand-operand-slot" id="ref-for-dom-mloperand-operand-slot⑦⓪">[[operand]]</a></code> as output to <var>opImpl</var>.</p>
+       </ol>
+      <li data-md>
+       <p>Return <var>output</var>.</p>
+     </ol>
+    </div>
+   </details>
+   <details open>
+    <summary> The following reduce algorithms are supported. </summary>
+    <div class="algorithm" data-algorithm="reduceL1(input, options)" data-algorithm-for="MLGraphBuilder">
+      The <dfn class="dfn-paneled idl-code" data-dfn-for="MLGraphBuilder" data-dfn-type="method" data-export data-lt="reduceL1(input, options)|reduceL1(input)" id="dom-mlgraphbuilder-reducel1"><code>reduceL1(<var>input</var>, <var>options</var>)</code></dfn> method steps are: 
+     <ol>
+      <li data-md>
+       <p>Let <var>output</var> be the result of running the <a data-link-type="dfn" href="#mlgraphbuilder-reduce-op" id="ref-for-mlgraphbuilder-reduce-op"> create reduce operation</a> given "reduceL1", <var>input</var> and <var>options</var>.</p>
+       <ol>
+        <li data-md>
+         <p>If that <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#dfn-throw" id="ref-for-dfn-throw②①⑧">throws</a> an error, then re-<a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#dfn-throw" id="ref-for-dfn-throw②①⑨">throw</a> the error.</p>
+       </ol>
+      <li data-md>
+       <p>Return <var>output</var>.</p>
+     </ol>
+    </div>
+    <div class="algorithm" data-algorithm="reduceL2(input, options)" data-algorithm-for="MLGraphBuilder">
+      The <dfn class="dfn-paneled idl-code" data-dfn-for="MLGraphBuilder" data-dfn-type="method" data-export data-lt="reduceL2(input, options)|reduceL2(input)" id="dom-mlgraphbuilder-reducel2"><code>reduceL2(<var>input</var>, <var>options</var>)</code></dfn> method steps are: 
+     <ol>
+      <li data-md>
+       <p>Let <var>output</var> be the result of running the <a data-link-type="dfn" href="#mlgraphbuilder-reduce-op" id="ref-for-mlgraphbuilder-reduce-op①"> create reduce operation</a> given "reduceL2", <var>input</var> and <var>options</var>.</p>
+       <ol>
+        <li data-md>
+         <p>If that <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#dfn-throw" id="ref-for-dfn-throw②②⓪">throws</a> an error, then re-<a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#dfn-throw" id="ref-for-dfn-throw②②①">throw</a> the error.</p>
+       </ol>
+      <li data-md>
+       <p>Return <var>output</var>.</p>
+     </ol>
+    </div>
+    <div class="algorithm" data-algorithm="reduceLogSum(input, options)" data-algorithm-for="MLGraphBuilder">
+      The <dfn class="dfn-paneled idl-code" data-dfn-for="MLGraphBuilder" data-dfn-type="method" data-export data-lt="reduceLogSum(input, options)|reduceLogSum(input)" id="dom-mlgraphbuilder-reducelogsum"><code>reduceLogSum(<var>input</var>, <var>options</var>)</code></dfn> method steps are: 
+     <ol>
+      <li data-md>
+       <p>Let <var>output</var> be the result of running the <a data-link-type="dfn" href="#mlgraphbuilder-reduce-op" id="ref-for-mlgraphbuilder-reduce-op②"> create reduce operation</a> given "reduceLogSum", <var>input</var> and <var>options</var>.</p>
+       <ol>
+        <li data-md>
+         <p>If that <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#dfn-throw" id="ref-for-dfn-throw②②②">throws</a> an error, then re-<a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#dfn-throw" id="ref-for-dfn-throw②②③">throw</a> the error.</p>
+       </ol>
+      <li data-md>
+       <p>Return <var>output</var>.</p>
+     </ol>
+    </div>
+    <div class="algorithm" data-algorithm="reduceLogSumExp(input, options)" data-algorithm-for="MLGraphBuilder">
+      The <dfn class="dfn-paneled idl-code" data-dfn-for="MLGraphBuilder" data-dfn-type="method" data-export data-lt="reduceLogSumExp(input, options)|reduceLogSumExp(input)" id="dom-mlgraphbuilder-reducelogsumexp"><code>reduceLogSumExp(<var>input</var>, <var>options</var>)</code></dfn> method steps are: 
+     <ol>
       <li data-md>
-       <p><em>L1</em>: Compute the <a href="https://mathworld.wolfram.com/L1-Norm.html">L1 norm</a> of all the input values along the axes.</p>
+       <p>Let <var>output</var> be the result of running the <a data-link-type="dfn" href="#mlgraphbuilder-reduce-op" id="ref-for-mlgraphbuilder-reduce-op③"> create reduce operation</a> given "reduceLogSumExp", <var>input</var> and <var>options</var>.</p>
+       <ol>
+        <li data-md>
+         <p>If that <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#dfn-throw" id="ref-for-dfn-throw②②④">throws</a> an error, then re-<a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#dfn-throw" id="ref-for-dfn-throw②②⑤">throw</a> the error.</p>
+       </ol>
       <li data-md>
-       <p><em>L2</em>: Compute the <a href="https://mathworld.wolfram.com/L2-Norm.html">L2 norm</a> of all the input values along the axes.</p>
+       <p>Return <var>output</var>.</p>
+     </ol>
+    </div>
+    <div class="algorithm" data-algorithm="reduceMax(input, options)" data-algorithm-for="MLGraphBuilder">
+      The <dfn class="dfn-paneled idl-code" data-dfn-for="MLGraphBuilder" data-dfn-type="method" data-export data-lt="reduceMax(input, options)|reduceMax(input)" id="dom-mlgraphbuilder-reducemax"><code>reduceMax(<var>input</var>, <var>options</var>)</code></dfn> method steps are: 
+     <ol>
       <li data-md>
-       <p><em>LogSum</em>: Compute the log value of the sum of all the input values along the axes.</p>
+       <p>Let <var>output</var> be the result of running the <a data-link-type="dfn" href="#mlgraphbuilder-reduce-op" id="ref-for-mlgraphbuilder-reduce-op④"> create reduce operation</a> given "reduceMax", <var>input</var> and <var>options</var>.</p>
+       <ol>
+        <li data-md>
+         <p>If that <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#dfn-throw" id="ref-for-dfn-throw②②⑥">throws</a> an error, then re-<a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#dfn-throw" id="ref-for-dfn-throw②②⑦">throw</a> the error.</p>
+       </ol>
       <li data-md>
-       <p><em>LogSumExp</em>: Compute the log value of the sum of the exponent of all the input values along the axes.</p>
+       <p>Return <var>output</var>.</p>
+     </ol>
+    </div>
+    <div class="algorithm" data-algorithm="reduceMean(input, options)" data-algorithm-for="MLGraphBuilder">
+      The <dfn class="dfn-paneled idl-code" data-dfn-for="MLGraphBuilder" data-dfn-type="method" data-export data-lt="reduceMean(input, options)|reduceMean(input)" id="dom-mlgraphbuilder-reducemean"><code>reduceMean(<var>input</var>, <var>options</var>)</code></dfn> method steps are: 
+     <ol>
       <li data-md>
-       <p><em>Max</em>: Compute the maximum value of all the input values along the axes.</p>
+       <p>Let <var>output</var> be the result of running the <a data-link-type="dfn" href="#mlgraphbuilder-reduce-op" id="ref-for-mlgraphbuilder-reduce-op⑤"> create reduce operation</a> given "reduceMean", <var>input</var> and <var>options</var>.</p>
+       <ol>
+        <li data-md>
+         <p>If that <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#dfn-throw" id="ref-for-dfn-throw②②⑧">throws</a> an error, then re-<a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#dfn-throw" id="ref-for-dfn-throw②②⑨">throw</a> the error.</p>
+       </ol>
       <li data-md>
-       <p><em>Mean</em>: Compute the average value of all the input values along the axes.</p>
+       <p>Return <var>output</var>.</p>
+     </ol>
+    </div>
+    <div class="algorithm" data-algorithm="reduceMin(input, options)" data-algorithm-for="MLGraphBuilder">
+      The <dfn class="dfn-paneled idl-code" data-dfn-for="MLGraphBuilder" data-dfn-type="method" data-export data-lt="reduceMin(input, options)|reduceMin(input)" id="dom-mlgraphbuilder-reducemin"><code>reduceMin(<var>input</var>, <var>options</var>)</code></dfn> method steps are: 
+     <ol>
       <li data-md>
-       <p><em>Min</em>: Compute the minimum value of all the input values along the axes.</p>
+       <p>Let <var>output</var> be the result of running the <a data-link-type="dfn" href="#mlgraphbuilder-reduce-op" id="ref-for-mlgraphbuilder-reduce-op⑥"> create reduce operation</a> given "reduceMin", <var>input</var> and <var>options</var>.</p>
+       <ol>
+        <li data-md>
+         <p>If that <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#dfn-throw" id="ref-for-dfn-throw②③⓪">throws</a> an error, then re-<a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#dfn-throw" id="ref-for-dfn-throw②③①">throw</a> the error.</p>
+       </ol>
       <li data-md>
-       <p><em>Product</em>: Compute the product of all the input values along the axes.</p>
+       <p>Return <var>output</var>.</p>
+     </ol>
+    </div>
+    <div class="algorithm" data-algorithm="reduceProduct(input, options)" data-algorithm-for="MLGraphBuilder">
+      The <dfn class="dfn-paneled idl-code" data-dfn-for="MLGraphBuilder" data-dfn-type="method" data-export data-lt="reduceProduct(input, options)|reduceProduct(input)" id="dom-mlgraphbuilder-reduceproduct"><code>reduceProduct(<var>input</var>, <var>options</var>)</code></dfn> method steps are: 
+     <ol>
       <li data-md>
-       <p><em>Sum</em>: Compute the sum of all the input values along the axes.</p>
+       <p>Let <var>output</var> be the result of running the <a data-link-type="dfn" href="#mlgraphbuilder-reduce-op" id="ref-for-mlgraphbuilder-reduce-op⑦"> create reduce operation</a> given "reduceProduct", <var>input</var> and <var>options</var>.</p>
+       <ol>
+        <li data-md>
+         <p>If that <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#dfn-throw" id="ref-for-dfn-throw②③②">throws</a> an error, then re-<a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#dfn-throw" id="ref-for-dfn-throw②③③">throw</a> the error.</p>
+       </ol>
       <li data-md>
-       <p><em>SumSquare</em>: Compute the sum of the square of all the input values along the axes.</p>
-     </ul>
+       <p>Return <var>output</var>.</p>
+     </ol>
     </div>
-    <h4 class="heading settled" data-level="7.6.27" id="api-mlgraphbuilder-relu"><span class="secno">7.6.27. </span><span class="content">The relu() method</span><a class="self-link" href="#api-mlgraphbuilder-relu"></a></h4>
-     Compute the <a href="https://en.wikipedia.org/wiki/Rectifier_(neural_networks)">rectified linear function</a> of the input tensor. 
-<pre class="idl highlight def"><c- b>partial</c-> <c- b>interface</c-> <a class="idl-code" data-link-type="interface" href="#mlgraphbuilder" id="ref-for-mlgraphbuilder④③"><c- g>MLGraphBuilder</c-></a> {
-  <a data-link-type="idl-name" href="#mloperand" id="ref-for-mloperand②⑤⑦"><c- n>MLOperand</c-></a> <dfn class="idl-code" data-dfn-for="MLGraphBuilder" data-dfn-type="method" data-export data-lt="relu(x)" id="dom-mlgraphbuilder-relu"><code><c- g>relu</c-></code><a class="self-link" href="#dom-mlgraphbuilder-relu"></a></dfn>(<a data-link-type="idl-name" href="#mloperand" id="ref-for-mloperand②⑤⑧"><c- n>MLOperand</c-></a> <dfn class="idl-code" data-dfn-for="MLGraphBuilder/relu(x)" data-dfn-type="argument" data-export id="dom-mlgraphbuilder-relu-x-x"><code><c- g>x</c-></code><a class="self-link" href="#dom-mlgraphbuilder-relu-x-x"></a></dfn>);
-  <a data-link-type="idl-name" href="#mlactivation" id="ref-for-mlactivation②⑧"><c- n>MLActivation</c-></a> <dfn class="idl-code" data-dfn-for="MLGraphBuilder" data-dfn-type="method" data-export data-lt="relu()" id="dom-mlgraphbuilder-relu①"><code><c- g>relu</c-></code><a class="self-link" href="#dom-mlgraphbuilder-relu①"></a></dfn>();
-};
-</pre>
-    <div class="algorithm" data-algorithm="relu">
-      <strong>Arguments:</strong> 
-     <ul>
+    <div class="algorithm" data-algorithm="reduceSum(input, options)" data-algorithm-for="MLGraphBuilder">
+      The <dfn class="dfn-paneled idl-code" data-dfn-for="MLGraphBuilder" data-dfn-type="method" data-export data-lt="reduceSum(input, options)|reduceSum(input)" id="dom-mlgraphbuilder-reducesum"><code>reduceSum(<var>input</var>, <var>options</var>)</code></dfn> method steps are: 
+     <ol>
       <li data-md>
-       <p><em>x</em>: an <code class="idl"><a data-link-type="idl" href="#mloperand" id="ref-for-mloperand②⑤⑨">MLOperand</a></code>. The input tensor.</p>
-     </ul>
-     <p><strong>Returns:</strong></p>
-     <ul>
+       <p>Let <var>output</var> be the result of running the <a data-link-type="dfn" href="#mlgraphbuilder-reduce-op" id="ref-for-mlgraphbuilder-reduce-op⑧"> create reduce operation</a> given "reduceSum", <var>input</var> and <var>options</var>.</p>
+       <ol>
+        <li data-md>
+         <p>If that <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#dfn-throw" id="ref-for-dfn-throw②③④">throws</a> an error, then re-<a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#dfn-throw" id="ref-for-dfn-throw②③⑤">throw</a> the error.</p>
+       </ol>
+      <li data-md>
+       <p>Return <var>output</var>.</p>
+     </ol>
+    </div>
+    <div class="algorithm" data-algorithm="reduceSumSquare(input, options)" data-algorithm-for="MLGraphBuilder">
+      The <dfn class="dfn-paneled idl-code" data-dfn-for="MLGraphBuilder" data-dfn-type="method" data-export data-lt="reduceSumSquare(input, options)|reduceSumSquare(input)" id="dom-mlgraphbuilder-reducesumsquare"><code>reduceSumSquare(<var>input</var>, <var>options</var>)</code></dfn> method steps are: 
+     <ol>
       <li data-md>
-       <p>an <code class="idl"><a data-link-type="idl" href="#mloperand" id="ref-for-mloperand②⑥⓪">MLOperand</a></code>. The output tensor of the same shape as <em>x</em>.</p>
+       <p>Let <var>output</var> be the result of running the <a data-link-type="dfn" href="#mlgraphbuilder-reduce-op" id="ref-for-mlgraphbuilder-reduce-op⑨"> create reduce operation</a> given "reduceSumSquare", <var>input</var> and <var>options</var>.</p>
+       <ol>
+        <li data-md>
+         <p>If that <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#dfn-throw" id="ref-for-dfn-throw②③⑥">throws</a> an error, then re-<a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#dfn-throw" id="ref-for-dfn-throw②③⑦">throw</a> the error.</p>
+       </ol>
       <li data-md>
-       <p>an <code class="idl"><a data-link-type="idl" href="#mlactivation" id="ref-for-mlactivation②⑨">MLActivation</a></code>. The activation function representing the relu operation.</p>
-     </ul>
-     <div class="note" role="note">
-       The behavior of this operation can be generically emulated from the usage of
+       <p>Return <var>output</var>.</p>
+     </ol>
+    </div>
+   </details>
+   <h4 class="heading settled" data-level="7.6.28" id="api-mlgraphbuilder-relu-method"><span class="secno">7.6.28. </span><span class="content">The relu() method</span><a class="self-link" href="#api-mlgraphbuilder-relu-method"></a></h4>
+    Compute the <a href="https://en.wikipedia.org/wiki/Rectifier_(neural_networks)">rectified linear function</a> of the input tensor. 
+<pre class="idl highlight def"><c- b>partial</c-> <c- b>interface</c-> <a class="idl-code" data-link-type="interface" href="#mlgraphbuilder" id="ref-for-mlgraphbuilder④⑥"><c- g>MLGraphBuilder</c-></a> {
+  <a data-link-type="idl-name" href="#mloperand" id="ref-for-mloperand③①⑦"><c- n>MLOperand</c-></a> <a class="idl-code" data-link-type="method" href="#dom-mlgraphbuilder-relu" id="ref-for-dom-mlgraphbuilder-relu"><c- g>relu</c-></a>(<a data-link-type="idl-name" href="#mloperand" id="ref-for-mloperand③①⑧"><c- n>MLOperand</c-></a> <dfn class="idl-code" data-dfn-for="MLGraphBuilder/relu(input)" data-dfn-type="argument" data-export id="dom-mlgraphbuilder-relu-input-input"><code><c- g>input</c-></code><a class="self-link" href="#dom-mlgraphbuilder-relu-input-input"></a></dfn>);
+  <a data-link-type="idl-name" href="#mlactivation" id="ref-for-mlactivation④⑤"><c- n>MLActivation</c-></a> <a class="idl-code" data-link-type="method" href="#relu-noargs" id="ref-for-relu-noargs"><c- g>relu</c-></a>();
+};
+</pre>
+   <div class="note" role="note">
+    <details open>
+     <summary> The behavior of this operation can be generically emulated from the usage of
     other operations as follow. However, user agents typically have a more
     efficient implementation for it, therefore its usage is encouraged from the
-    performance standpoint. 
+    performance standpoint. </summary>
 <pre class="highlight"><c- k>return</c-> builder<c- p>.</c->max<c- p>(</c->builder<c- p>.</c->constant<c- p>(</c-><c- mf>0</c-><c- p>),</c-> x<c- p>);</c->
 </pre>
-     </div>
+    </details>
+   </div>
+   <h5 class="heading settled" data-level="7.6.28.1" id="api-mlgraphbuilder-relu-input"><span class="secno">7.6.28.1. </span><span class="content">The <code class="idl"><a data-link-type="idl" href="#dom-mlgraphbuilder-relu" id="ref-for-dom-mlgraphbuilder-relu①">relu(input)</a></code> method</span><a class="self-link" href="#api-mlgraphbuilder-relu-input"></a></h5>
+   <div>
+     <strong>Arguments:</strong> 
+    <ul>
+     <li data-md>
+      <p><em>input</em>: an <code class="idl"><a data-link-type="idl" href="#mloperand" id="ref-for-mloperand③①⑨">MLOperand</a></code>. The input tensor.</p>
+    </ul>
+    <p><strong>Returns:</strong></p>
+    <ul>
+     <li data-md>
+      <p>an <code class="idl"><a data-link-type="idl" href="#mloperand" id="ref-for-mloperand③②⓪">MLOperand</a></code>. The output tensor of the same shape as <em>input</em>.</p>
+    </ul>
+   </div>
+   <details class="algorithm" data-algorithm="relu(input)" data-algorithm-for="MLGraphBuilder" open>
+    <summary> The <dfn class="dfn-paneled idl-code" data-dfn-for="MLGraphBuilder" data-dfn-type="method" data-export id="dom-mlgraphbuilder-relu"><code>relu(<var>input</var>)</code></dfn> method steps are: </summary>
+    <div class="algorithm-steps">
+     <ol>
+      <li data-md>
+       <p><a data-link-type="dfn" href="https://infra.spec.whatwg.org/#assert" id="ref-for-assert⑥③">Assert</a>: the type of <var>input</var> is <code class="idl"><a data-link-type="idl" href="#mloperand" id="ref-for-mloperand③②①">MLOperand</a></code>.</p>
+      <li data-md>
+       <p>If any of the following sub-steps fail, <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#dfn-throw" id="ref-for-dfn-throw②③⑧">throw</a> an "<code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#operationerror" id="ref-for-operationerror③⑨">OperationError</a></code>" <code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#idl-DOMException" id="ref-for-idl-DOMException①②⑨">DOMException</a></code>.</p>
+       <ol>
+        <li data-md>
+         <p>Let <var>output</var> be the result of <a data-link-type="dfn" href="#copy-an-mloperand" id="ref-for-copy-an-mloperand⑨">copying an MLOperand</a> given <var>input</var>.</p>
+        <li data-md>
+         <p>Make a request to the underlying platform to:</p>
+         <ol>
+          <li data-md>
+           <p>Let <var>opImpl</var> be an <a data-link-type="dfn" href="https://infra.spec.whatwg.org/#implementation-defined" id="ref-for-implementation-defined⑤④">implementation-defined</a> platform operator for the ReLU operation.</p>
+          <li data-md>
+           <p>Store a reference of <var>opImpl</var> in <var>output</var>.<code class="idl"><a data-link-type="idl" href="#dom-mloperand-operator-slot" id="ref-for-dom-mloperand-operator-slot②⑤">[[operator]]</a></code>.</p>
+          <li data-md>
+           <p>Create an <a data-link-type="dfn" href="https://infra.spec.whatwg.org/#implementation-defined" id="ref-for-implementation-defined⑤⑤">implementation-defined</a> platform operand <var>outputImpl</var> to represent the output, given <var>output</var> and <var>opImpl</var>.</p>
+          <li data-md>
+           <p>Store a reference to <var>outputImpl</var> in <var>output</var>.<code class="idl"><a data-link-type="idl" href="#dom-mloperand-operand-slot" id="ref-for-dom-mloperand-operand-slot⑦①">[[operand]]</a></code>.</p>
+         </ol>
+        <li data-md>
+         <p>Connect <var>input</var>.<code class="idl"><a data-link-type="idl" href="#dom-mloperand-operand-slot" id="ref-for-dom-mloperand-operand-slot⑦②">[[operand]]</a></code> as input to <var>opImpl</var>.</p>
+        <li data-md>
+         <p>Connect <var>output</var>.<code class="idl"><a data-link-type="idl" href="#dom-mloperand-operand-slot" id="ref-for-dom-mloperand-operand-slot⑦③">[[operand]]</a></code> as output to <var>opImpl</var>.</p>
+       </ol>
+      <li data-md>
+       <p>Return <var>output</var>.</p>
+     </ol>
+    </div>
+   </details>
+   <h5 class="heading settled" data-level="7.6.28.2" id="api-mlgraphbuilder-relu"><span class="secno">7.6.28.2. </span><span class="content">The <code class="idl"><a data-link-type="idl" href="#relu-noargs" id="ref-for-relu-noargs①">relu()</a></code> method</span><a class="self-link" href="#api-mlgraphbuilder-relu"></a></h5>
+   <div>
+     <strong>Arguments:</strong> 
+    <ul>
+     <li data-md>
+      <p>None.</p>
+    </ul>
+    <p><strong>Returns:</strong></p>
+    <ul>
+     <li data-md>
+      <p>an <code class="idl"><a data-link-type="idl" href="#mlactivation" id="ref-for-mlactivation④⑥">MLActivation</a></code>. The activation function representing the relu operation.</p>
+    </ul>
+   </div>
+   <details class="algorithm" data-algorithm="relu()" data-algorithm-for="MLGraphBuilder" open>
+    <summary> The <dfn class="dfn-paneled idl-code" data-dfn-for="MLGraphBuilder" data-dfn-type="method" data-export id="relu-noargs"><code>relu()</code></dfn> method steps are: </summary>
+    <div class="algorithm-steps">
+     <ol>
+      <li data-md>
+       <p>Let <var>op</var> be the result of <a data-link-type="dfn" href="#create-an-mlactivation" id="ref-for-create-an-mlactivation⑥">creating an MLActivation</a> given <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#this" id="ref-for-this③③">this</a> and <code>"relu"</code>.</p>
+       <ol>
+        <li data-md>
+         <p>If that <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#dfn-throw" id="ref-for-dfn-throw②③⑨">throws</a> an error, re-<a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#dfn-throw" id="ref-for-dfn-throw②④⓪">throw</a> the error.</p>
+       </ol>
+      <li data-md>
+       <p>Return <var>op</var>.</p>
+     </ol>
     </div>
-    <h4 class="heading settled" data-level="7.6.28" id="api-mlgraphbuilder-resample2d"><span class="secno">7.6.28. </span><span class="content">The resample2d() method</span><a class="self-link" href="#api-mlgraphbuilder-resample2d"></a></h4>
-     Resample the tensor values from the source to the destination spatial dimensions according to the scaling factors. 
+   </details>
+   <h4 class="heading settled" data-level="7.6.29" id="api-mlgraphbuilder-resample2d-method"><span class="secno">7.6.29. </span><span class="content">The resample2d() method</span><a class="self-link" href="#api-mlgraphbuilder-resample2d-method"></a></h4>
+    Resample the tensor values from the source to the destination spatial dimensions according to the scaling factors. 
 <pre class="idl highlight def"><c- b>enum</c-> <dfn class="dfn-paneled idl-code" data-dfn-type="enum" data-export id="enumdef-mlinterpolationmode"><code><c- g>MLInterpolationMode</c-></code></dfn> {
   <dfn class="idl-code" data-dfn-for="MLInterpolationMode" data-dfn-type="enum-value" data-export id="dom-mlinterpolationmode-nearest-neighbor"><code><c- s>"nearest-neighbor"</c-></code><a class="self-link" href="#dom-mlinterpolationmode-nearest-neighbor"></a></dfn>,
   <dfn class="idl-code" data-dfn-for="MLInterpolationMode" data-dfn-type="enum-value" data-export id="dom-mlinterpolationmode-linear"><code><c- s>"linear"</c-></code><a class="self-link" href="#dom-mlinterpolationmode-linear"></a></dfn>
 };
 
 <c- b>dictionary</c-> <dfn class="dfn-paneled idl-code" data-dfn-type="dictionary" data-export id="dictdef-mlresample2doptions"><code><c- g>MLResample2dOptions</c-></code></dfn> {
-  <a data-link-type="idl-name" href="#enumdef-mlinterpolationmode" id="ref-for-enumdef-mlinterpolationmode"><c- n>MLInterpolationMode</c-></a> <dfn class="idl-code" data-default="&quot;nearest-neighbor&quot;" data-dfn-for="MLResample2dOptions" data-dfn-type="dict-member" data-export data-type="MLInterpolationMode " id="dom-mlresample2doptions-mode"><code><c- g>mode</c-></code><a class="self-link" href="#dom-mlresample2doptions-mode"></a></dfn> = "nearest-neighbor";
-  <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#idl-sequence" id="ref-for-idl-sequence②⑤"><c- b>sequence</c-></a>&lt;<a class="idl-code" data-link-type="interface" href="https://webidl.spec.whatwg.org/#idl-float" id="ref-for-idl-float②⑥"><c- b>float</c-></a>> <dfn class="idl-code" data-dfn-for="MLResample2dOptions" data-dfn-type="dict-member" data-export data-type="sequence<float> " id="dom-mlresample2doptions-scales"><code><c- g>scales</c-></code><a class="self-link" href="#dom-mlresample2doptions-scales"></a></dfn>;
-  <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#idl-sequence" id="ref-for-idl-sequence②⑥"><c- b>sequence</c-></a>&lt;<a class="idl-code" data-link-type="interface" href="https://webidl.spec.whatwg.org/#idl-unsigned-long" id="ref-for-idl-unsigned-long⑤③"><c- b>unsigned</c-> <c- b>long</c-></a>> <dfn class="idl-code" data-dfn-for="MLResample2dOptions" data-dfn-type="dict-member" data-export data-type="sequence<unsigned long> " id="dom-mlresample2doptions-sizes"><code><c- g>sizes</c-></code><a class="self-link" href="#dom-mlresample2doptions-sizes"></a></dfn>;
-  <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#idl-sequence" id="ref-for-idl-sequence②⑦"><c- b>sequence</c-></a>&lt;<a class="idl-code" data-link-type="interface" href="https://webidl.spec.whatwg.org/#idl-unsigned-long" id="ref-for-idl-unsigned-long⑤④"><c- b>unsigned</c-> <c- b>long</c-></a>> <dfn class="idl-code" data-dfn-for="MLResample2dOptions" data-dfn-type="dict-member" data-export data-type="sequence<unsigned long> " id="dom-mlresample2doptions-axes"><code><c- g>axes</c-></code><a class="self-link" href="#dom-mlresample2doptions-axes"></a></dfn>;
+  <a data-link-type="idl-name" href="#enumdef-mlinterpolationmode" id="ref-for-enumdef-mlinterpolationmode"><c- n>MLInterpolationMode</c-></a> <a class="idl-code" data-default="&quot;nearest-neighbor&quot;" data-link-type="dict-member" data-type="MLInterpolationMode " href="#dom-mlresample2doptions-mode" id="ref-for-dom-mlresample2doptions-mode"><c- g>mode</c-></a> = "nearest-neighbor";
+  <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#idl-sequence" id="ref-for-idl-sequence②⑤"><c- b>sequence</c-></a>&lt;<a class="idl-code" data-link-type="interface" href="https://webidl.spec.whatwg.org/#idl-float" id="ref-for-idl-float③⑨"><c- b>float</c-></a>> <a class="idl-code" data-link-type="dict-member" data-type="sequence<float> " href="#dom-mlresample2doptions-scales" id="ref-for-dom-mlresample2doptions-scales"><c- g>scales</c-></a>;
+  <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#idl-sequence" id="ref-for-idl-sequence②⑥"><c- b>sequence</c-></a>&lt;<a class="idl-code" data-link-type="interface" href="https://webidl.spec.whatwg.org/#idl-unsigned-long" id="ref-for-idl-unsigned-long⑤⑤"><c- b>unsigned</c-> <c- b>long</c-></a>> <a class="idl-code" data-link-type="dict-member" data-type="sequence<unsigned long> " href="#dom-mlresample2doptions-sizes" id="ref-for-dom-mlresample2doptions-sizes"><c- g>sizes</c-></a>;
+  <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#idl-sequence" id="ref-for-idl-sequence②⑦"><c- b>sequence</c-></a>&lt;<a class="idl-code" data-link-type="interface" href="https://webidl.spec.whatwg.org/#idl-unsigned-long" id="ref-for-idl-unsigned-long⑤⑥"><c- b>unsigned</c-> <c- b>long</c-></a>> <a class="idl-code" data-link-type="dict-member" data-type="sequence<unsigned long> " href="#dom-mlresample2doptions-axes" id="ref-for-dom-mlresample2doptions-axes"><c- g>axes</c-></a>;
 };
 
-<c- b>partial</c-> <c- b>interface</c-> <a class="idl-code" data-link-type="interface" href="#mlgraphbuilder" id="ref-for-mlgraphbuilder④④"><c- g>MLGraphBuilder</c-></a> {
-  <a data-link-type="idl-name" href="#mloperand" id="ref-for-mloperand②⑥①"><c- n>MLOperand</c-></a> <dfn class="idl-code" data-dfn-for="MLGraphBuilder" data-dfn-type="method" data-export data-lt="resample2d(input, options)|resample2d(input)" id="dom-mlgraphbuilder-resample2d"><code><c- g>resample2d</c-></code><a class="self-link" href="#dom-mlgraphbuilder-resample2d"></a></dfn>(<a data-link-type="idl-name" href="#mloperand" id="ref-for-mloperand②⑥②"><c- n>MLOperand</c-></a> <dfn class="idl-code" data-dfn-for="MLGraphBuilder/resample2d(input, options), MLGraphBuilder/resample2d(input)" data-dfn-type="argument" data-export id="dom-mlgraphbuilder-resample2d-input-options-input"><code><c- g>input</c-></code><a class="self-link" href="#dom-mlgraphbuilder-resample2d-input-options-input"></a></dfn>, <c- b>optional</c-> <a data-link-type="idl-name" href="#dictdef-mlresample2doptions" id="ref-for-dictdef-mlresample2doptions"><c- n>MLResample2dOptions</c-></a> <dfn class="idl-code" data-dfn-for="MLGraphBuilder/resample2d(input, options), MLGraphBuilder/resample2d(input)" data-dfn-type="argument" data-export id="dom-mlgraphbuilder-resample2d-input-options-options"><code><c- g>options</c-></code><a class="self-link" href="#dom-mlgraphbuilder-resample2d-input-options-options"></a></dfn> = {});
+<c- b>partial</c-> <c- b>interface</c-> <a class="idl-code" data-link-type="interface" href="#mlgraphbuilder" id="ref-for-mlgraphbuilder④⑦"><c- g>MLGraphBuilder</c-></a> {
+  <a data-link-type="idl-name" href="#mloperand" id="ref-for-mloperand③②②"><c- n>MLOperand</c-></a> <a class="idl-code" data-link-type="method" href="#dom-mlgraphbuilder-resample2d" id="ref-for-dom-mlgraphbuilder-resample2d"><c- g>resample2d</c-></a>(<a data-link-type="idl-name" href="#mloperand" id="ref-for-mloperand③②③"><c- n>MLOperand</c-></a> <dfn class="idl-code" data-dfn-for="MLGraphBuilder/resample2d(input, options), MLGraphBuilder/resample2d(input)" data-dfn-type="argument" data-export id="dom-mlgraphbuilder-resample2d-input-options-input"><code><c- g>input</c-></code><a class="self-link" href="#dom-mlgraphbuilder-resample2d-input-options-input"></a></dfn>, <c- b>optional</c-> <a data-link-type="idl-name" href="#dictdef-mlresample2doptions" id="ref-for-dictdef-mlresample2doptions"><c- n>MLResample2dOptions</c-></a> <dfn class="idl-code" data-dfn-for="MLGraphBuilder/resample2d(input, options), MLGraphBuilder/resample2d(input)" data-dfn-type="argument" data-export id="dom-mlgraphbuilder-resample2d-input-options-options"><code><c- g>options</c-></code><a class="self-link" href="#dom-mlgraphbuilder-resample2d-input-options-options"></a></dfn> = {});
 };
 </pre>
-    <div class="algorithm" data-algorithm="resample2d">
-      <strong>Arguments:</strong> 
-     <ul>
+   <div>
+     <strong>Arguments:</strong> 
+    <ul>
+     <li data-md>
+      <p><em>input</em>: an <code class="idl"><a data-link-type="idl" href="#mloperand" id="ref-for-mloperand③②④">MLOperand</a></code>. The input 4-D tensor.</p>
+     <li data-md>
+      <p><em>options</em>: an optional <code class="idl"><a data-link-type="idl" href="#dictdef-mlresample2doptions" id="ref-for-dictdef-mlresample2doptions①">MLResample2dOptions</a></code>. The optional parameters of the operation.</p>
+    </ul>
+    <p><strong>Returns:</strong> an <code class="idl"><a data-link-type="idl" href="#mloperand" id="ref-for-mloperand③②⑤">MLOperand</a></code>. The output 4-D tensor.</p>
+   </div>
+   <p><code class="idl"><a data-link-type="idl" href="#dictdef-mlresample2doptions" id="ref-for-dictdef-mlresample2doptions②">MLResample2dOptions</a></code> has the following members:</p>
+   <dl>
+    <dt data-md><dfn class="dfn-paneled idl-code" data-dfn-for="MLResample2dOptions" data-dfn-type="dict-member" data-export id="dom-mlresample2doptions-mode"><code>mode</code></dfn>, <span> of type <a data-link-type="idl-name" href="#enumdef-mlinterpolationmode" id="ref-for-enumdef-mlinterpolationmode①">MLInterpolationMode</a>, defaulting to <code>"nearest-neighbor"</code></span>
+    <dd data-md>
+     <p>An <code class="idl"><a data-link-type="idl" href="#enumdef-mlinterpolationmode" id="ref-for-enumdef-mlinterpolationmode②">MLInterpolationMode</a></code> <a data-link-type="dfn" href="https://infra.spec.whatwg.org/#string" id="ref-for-string①④">string</a>.
+Specifies the interpolation algorithm used to fill the output tensor values.
+The default value is <code>"nearest-neighbor"</code>, standing for <em>Nearest Neighbor</em> interpolation.</p>
+    <dt data-md><dfn class="dfn-paneled idl-code" data-dfn-for="MLResample2dOptions" data-dfn-type="dict-member" data-export id="dom-mlresample2doptions-scales"><code>scales</code></dfn>, <span> of type sequence&lt;<a data-link-type="idl-name" href="https://webidl.spec.whatwg.org/#idl-float" id="ref-for-idl-float④⓪">float</a>></span>
+    <dd data-md>
+     <p>A sequence of <code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#idl-float" id="ref-for-idl-float④①">float</a></code> of length 2.
+Specifies the scaling factor in each spatial dimensions of the input: [scaleHeight, scaleWidth].
+The default value is [1.0, 1.0].</p>
+    <dt data-md><dfn class="dfn-paneled idl-code" data-dfn-for="MLResample2dOptions" data-dfn-type="dict-member" data-export id="dom-mlresample2doptions-sizes"><code>sizes</code></dfn>, <span> of type <code class="idl-code">sequence&lt;unsigned long></code></span>
+    <dd data-md>
+     <p>A sequence of <code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#idl-unsigned-long" id="ref-for-idl-unsigned-long⑤⑦">unsigned long</a></code> of length 2.
+Specifies the target sizes for each spatial dimensions of the input: [sizeHeight, sizeWidth]. When the target sizes are specified, the <code class="idl"><a data-link-type="idl" href="#dom-mlresample2doptions-scales" id="ref-for-dom-mlresample2doptions-scales①">scales</a></code> argument is ignored, since the scaling factor values are derived from the target sizes of each spatial dimension of the input.</p>
+    <dt data-md><dfn class="dfn-paneled idl-code" data-dfn-for="MLResample2dOptions" data-dfn-type="dict-member" data-export id="dom-mlresample2doptions-axes"><code>axes</code></dfn>, <span> of type <code class="idl-code">sequence&lt;unsigned long></code></span>
+    <dd data-md>
+     <p>A sequence of <code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#idl-unsigned-long" id="ref-for-idl-unsigned-long⑤⑧">unsigned long</a></code> of length 2.
+Specifies the two consecutive dimensions of the input tensor to which the interpolation algorithm applies. The valid values in the sequence are [0, 1], [1, 2] or [2, 3].
+The default value is [2, 3].</p>
+   </dl>
+   <details class="algorithm" data-algorithm="check resample options" data-algorithm-for="MLGraphBuilder" open>
+    <summary> To <dfn class="dfn-paneled" data-dfn-for="MLGraphBuilder" data-dfn-type="dfn" data-noexport id="mlgraphbuilder-check-resample-options">check resample options</dfn> given <var>options</var>, run the following steps: </summary>
+    <div class="algorithm-steps">
+     <ol>
+      <li data-md>
+       <p>If <var>options</var>.<code class="idl"><a data-link-type="idl" href="#dom-mlresample2doptions-scales" id="ref-for-dom-mlresample2doptions-scales②">scales</a></code> does not <a data-link-type="dfn" href="https://infra.spec.whatwg.org/#map-exists" id="ref-for-map-exists⑤②">exist</a>, set it to to <code>« 1.0, 1.0 »</code>.</p>
+      <li data-md>
+       <p>Otherwise, if any of its values is not greater than 0, or if its <a data-link-type="dfn" href="https://infra.spec.whatwg.org/#list-size" id="ref-for-list-size④⑥">size</a> is not 2, return false.</p>
+      <li data-md>
+       <p>If <var>options</var>.<code class="idl"><a data-link-type="idl" href="#dom-mlresample2doptions-sizes" id="ref-for-dom-mlresample2doptions-sizes①">sizes</a></code> <a data-link-type="dfn" href="https://infra.spec.whatwg.org/#map-exists" id="ref-for-map-exists⑤③">exists</a>, and if its size is not 2, or if any of its values is not greater than 0, return false.</p>
+      <li data-md>
+       <p>If <var>options</var>.<code class="idl"><a data-link-type="idl" href="#dom-mlresample2doptions-axes" id="ref-for-dom-mlresample2doptions-axes①">axes</a></code> does not <a data-link-type="dfn" href="https://infra.spec.whatwg.org/#map-exists" id="ref-for-map-exists⑤④">exists</a>, set it to <code>« 2, 3 »</code>.</p>
+      <li data-md>
+       <p>Otherwise, if its value is not one of <code>« 0, 1», « 1, 2», « 2, 3 »</code>, return false.</p>
+      <li data-md>
+       <p>Return true.</p>
+     </ol>
+    </div>
+   </details>
+   <details class="algorithm" data-algorithm="resample output sizes" data-algorithm-for="MLGraphBuilder" open>
+    <summary> To <dfn class="dfn-paneled" data-dfn-for="MLGraphBuilder" data-dfn-type="dfn" data-noexport id="mlgraphbuilder-resample-output-sizes">resample output sizes</dfn> given <var>input</var> and <var>options</var>, run the following steps: </summary>
+    <div class="algorithm-steps">
+     <ol>
+      <li data-md>
+       <p>Let <var>desc</var> be an <code class="idl"><a data-link-type="idl" href="#dictdef-mloperanddescriptor" id="ref-for-dictdef-mloperanddescriptor②③">MLOperandDescriptor</a></code> initialized to <var>input</var>.<code class="idl"><a data-link-type="idl" href="#dom-mloperand-descriptor-slot" id="ref-for-dom-mloperand-descriptor-slot⑧⑤">[[descriptor]]</a></code>.</p>
+      <li data-md>
+       <p>For <var>index</var> in <a data-link-type="dfn" href="https://infra.spec.whatwg.org/#the-range" id="ref-for-the-range⑥">the range</a> 0 to the <a data-link-type="dfn" href="https://infra.spec.whatwg.org/#list-size" id="ref-for-list-size④⑦">size</a> of <var>options</var>.<code class="idl"><a data-link-type="idl" href="#dom-mlresample2doptions-axes" id="ref-for-dom-mlresample2doptions-axes②">axes</a></code>, exclusive:</p>
+       <ol>
+        <li data-md>
+         <p>If <var>options</var>.<code class="idl"><a data-link-type="idl" href="#dom-mlresample2doptions-sizes" id="ref-for-dom-mlresample2doptions-sizes②">sizes</a></code> <a data-link-type="dfn" href="https://infra.spec.whatwg.org/#map-exists" id="ref-for-map-exists⑤⑤">exists</a>, set <var>desc</var>.<code class="idl"><a data-link-type="idl" href="#dom-mloperand-descriptor-slot" id="ref-for-dom-mloperand-descriptor-slot⑧⑥">[[descriptor]]</a></code>.<code class="idl"><a data-link-type="idl" href="#dom-mloperanddescriptor-dimensions" id="ref-for-dom-mloperanddescriptor-dimensions⑧④">dimensions</a></code>[<var>options</var>.<code class="idl"><a data-link-type="idl" href="#dom-mlresample2doptions-axes" id="ref-for-dom-mlresample2doptions-axes③">axes</a></code>[<var>index</var>]] to <var>options</var>.<code class="idl"><a data-link-type="idl" href="#dom-mlresample2doptions-sizes" id="ref-for-dom-mlresample2doptions-sizes③">sizes</a></code>[<var>index</var>] and return <var>desc</var>.</p>
+        <li data-md>
+         <p>Otherwise, set <var>desc</var>.<code class="idl"><a data-link-type="idl" href="#dom-mloperand-descriptor-slot" id="ref-for-dom-mloperand-descriptor-slot⑧⑦">[[descriptor]]</a></code>.<code class="idl"><a data-link-type="idl" href="#dom-mloperanddescriptor-dimensions" id="ref-for-dom-mloperanddescriptor-dimensions⑧⑤">dimensions</a></code>[<var>options</var>.<code class="idl"><a data-link-type="idl" href="#dom-mlresample2doptions-axes" id="ref-for-dom-mlresample2doptions-axes④">axes</a></code>[<var>index</var>]] to <var>input</var>.<code class="idl"><a data-link-type="idl" href="#dom-mloperand-descriptor-slot" id="ref-for-dom-mloperand-descriptor-slot⑧⑧">[[descriptor]]</a></code>.<code class="idl"><a data-link-type="idl" href="#dom-mloperanddescriptor-dimensions" id="ref-for-dom-mloperanddescriptor-dimensions⑧⑥">dimensions</a></code>[<var>index</var>] multiplied by <var>options</var>.<code class="idl"><a data-link-type="idl" href="#dom-mlresample2doptions-scales" id="ref-for-dom-mlresample2doptions-scales③">scales</a></code>.</p>
+       </ol>
+      <li data-md>
+       <p>Return <var>desc</var>.</p>
+     </ol>
+    </div>
+   </details>
+   <details class="algorithm" data-algorithm="resample2d(input, options)" data-algorithm-for="MLGraphBuilder" open>
+    <summary> The <dfn class="dfn-paneled idl-code" data-dfn-for="MLGraphBuilder" data-dfn-type="method" data-export data-lt="resample2d(input, options)|resample2d(input)" id="dom-mlgraphbuilder-resample2d"><code>resample2d(<var>input</var>, <var>options</var>)</code></dfn> method steps are: </summary>
+    <div class="algorithm-steps">
+     <ol>
       <li data-md>
-       <p><em>input</em>: an <code class="idl"><a data-link-type="idl" href="#mloperand" id="ref-for-mloperand②⑥③">MLOperand</a></code>. The input 4-D tensor.</p>
+       <p>If the <a data-link-type="dfn" href="https://infra.spec.whatwg.org/#list-size" id="ref-for-list-size④⑧">size</a> of <var>input</var>.<code class="idl"><a data-link-type="idl" href="#dom-mloperand-descriptor-slot" id="ref-for-dom-mloperand-descriptor-slot⑧⑨">[[descriptor]]</a></code>.<code class="idl"><a data-link-type="idl" href="#dom-mloperanddescriptor-dimensions" id="ref-for-dom-mloperanddescriptor-dimensions⑧⑦">dimensions</a></code> is not 4, then <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#dfn-throw" id="ref-for-dfn-throw②④①">throw</a> a "<code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#dataerror" id="ref-for-dataerror⑧①">DataError</a></code>" <code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#idl-DOMException" id="ref-for-idl-DOMException①③⓪">DOMException</a></code>.</p>
       <li data-md>
-       <p><em>options</em>: an optional <code class="idl"><a data-link-type="idl" href="#dictdef-mlresample2doptions" id="ref-for-dictdef-mlresample2doptions①">MLResample2dOptions</a></code>. The optional parameters of the operation.</p>
-       <ul>
+       <p>If running the <a data-link-type="dfn" href="#mlgraphbuilder-check-resample-options" id="ref-for-mlgraphbuilder-check-resample-options">check resample options</a> steps given <var>options</var> returns false, then <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#dfn-throw" id="ref-for-dfn-throw②④②">throw</a> a "<code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#dataerror" id="ref-for-dataerror⑧②">DataError</a></code>" <code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#idl-DOMException" id="ref-for-idl-DOMException①③①">DOMException</a></code>.</p>
+      <li data-md>
+       <p>Let <var>desc</var> be the result of running the <a data-link-type="dfn" href="#mlgraphbuilder-resample-output-sizes" id="ref-for-mlgraphbuilder-resample-output-sizes">resample output sizes</a> steps given <var>options</var>.</p>
+       <ol>
+        <li data-md>
+         <p>If that <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#dfn-throw" id="ref-for-dfn-throw②④③">throws</a> an error, re-<a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#dfn-throw" id="ref-for-dfn-throw②④④">throw</a> the error.</p>
+       </ol>
+      <li data-md>
+       <p>If any of the following sub-steps fail, <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#dfn-throw" id="ref-for-dfn-throw②④⑤">throw</a> an "<code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#operationerror" id="ref-for-operationerror④⓪">OperationError</a></code>" <code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#idl-DOMException" id="ref-for-idl-DOMException①③②">DOMException</a></code>.</p>
+       <ol>
         <li data-md>
-         <p><em>mode</em>: an <code class="idl"><a data-link-type="idl" href="#enumdef-mlinterpolationmode" id="ref-for-enumdef-mlinterpolationmode①">MLInterpolationMode</a></code>. The interpolation algorithm used to fill the output tensor values.
-If not set, it is assumed to be the <em>Nearest Neighbor</em> interpolation.</p>
+         <p>Let <var>output</var> be the result of <a data-link-type="dfn" href="#create-an-mloperand" id="ref-for-create-an-mloperand①⑨">creating an MLOperand</a> given <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#this" id="ref-for-this③④">this</a> and <var>desc</var>.</p>
         <li data-md>
-         <p><em>scales</em>: a sequence of <code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#idl-float" id="ref-for-idl-float②⑦">float</a></code> of length 2. Each value represents the scaling factor used to scale in each spatial dimensions of input, [scale_height, scale_width]. If not set, the values are assumed to be [1.0, 1.0].</p>
+         <p>Make a request to the underlying platform to:</p>
+         <ol>
+          <li data-md>
+           <p>Let <var>opImpl</var> be an <a data-link-type="dfn" href="https://infra.spec.whatwg.org/#implementation-defined" id="ref-for-implementation-defined⑤⑥">implementation-defined</a> platform operator for the resample 2D operation, given <var>options</var>.</p>
+          <li data-md>
+           <p>Store a reference of <var>opImpl</var> in <var>output</var>.<code class="idl"><a data-link-type="idl" href="#dom-mloperand-operator-slot" id="ref-for-dom-mloperand-operator-slot②⑥">[[operator]]</a></code>.</p>
+          <li data-md>
+           <p>Create an <a data-link-type="dfn" href="https://infra.spec.whatwg.org/#implementation-defined" id="ref-for-implementation-defined⑤⑦">implementation-defined</a> platform operand <var>outputImpl</var> to represent the output, given <var>output</var> and <var>opImpl</var>.</p>
+          <li data-md>
+           <p>Store a reference to <var>outputImpl</var> in <var>output</var>.<code class="idl"><a data-link-type="idl" href="#dom-mloperand-operand-slot" id="ref-for-dom-mloperand-operand-slot⑦④">[[operand]]</a></code>.</p>
+         </ol>
         <li data-md>
-         <p><em>sizes</em>: a sequence of <code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#idl-unsigned-long" id="ref-for-idl-unsigned-long⑤⑤">unsigned long</a></code> of length 2. The target sizes for each spatial dimensions of input, [size_height, size_width]. When the target sizes are specified, the <em>options.scales</em> argument is ignored as the scaling factor values are derived from the target sizes of each spatial dimension of input.</p>
+         <p>Connect <var>input</var>.<code class="idl"><a data-link-type="idl" href="#dom-mloperand-operand-slot" id="ref-for-dom-mloperand-operand-slot⑦⑤">[[operand]]</a></code> as input to <var>opImpl</var>.</p>
         <li data-md>
-         <p><em>axes</em>: a sequence of <code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#idl-unsigned-long" id="ref-for-idl-unsigned-long⑤⑥">unsigned long</a></code> of length 2. The two consecutive dimensions of the input tensor to which the interpolation algorithm applies. The valid values in the sequence are [0, 1], [1, 2] or [2, 3]. When not specified, the sequence is assumed to be [2, 3].</p>
-       </ul>
-     </ul>
-     <p><strong>Returns:</strong> an <code class="idl"><a data-link-type="idl" href="#mloperand" id="ref-for-mloperand②⑥④">MLOperand</a></code>. The output 4-D tensor.</p>
+         <p>Connect <var>output</var>.<code class="idl"><a data-link-type="idl" href="#dom-mloperand-operand-slot" id="ref-for-dom-mloperand-operand-slot⑦⑥">[[operand]]</a></code> as output to <var>opImpl</var>.</p>
+       </ol>
+      <li data-md>
+       <p>Return <var>output</var>.</p>
+     </ol>
     </div>
-    <h4 class="heading settled" data-level="7.6.29" id="api-mlgraphbuilder-reshape"><span class="secno">7.6.29. </span><span class="content">The reshape() method</span><a class="self-link" href="#api-mlgraphbuilder-reshape"></a></h4>
-     Alter the shape of a tensor to a new shape. Reshape does not copy or change the content of the tensor. It just changes the tensor’s logical dimensions for the subsequent operations. 
-<pre class="idl highlight def"><c- b>partial</c-> <c- b>interface</c-> <a class="idl-code" data-link-type="interface" href="#mlgraphbuilder" id="ref-for-mlgraphbuilder④⑤"><c- g>MLGraphBuilder</c-></a> {
-  <a data-link-type="idl-name" href="#mloperand" id="ref-for-mloperand②⑥⑤"><c- n>MLOperand</c-></a> <dfn class="idl-code" data-dfn-for="MLGraphBuilder" data-dfn-type="method" data-export data-lt="reshape(input, newShape)" id="dom-mlgraphbuilder-reshape"><code><c- g>reshape</c-></code><a class="self-link" href="#dom-mlgraphbuilder-reshape"></a></dfn>(<a data-link-type="idl-name" href="#mloperand" id="ref-for-mloperand②⑥⑥"><c- n>MLOperand</c-></a> <dfn class="idl-code" data-dfn-for="MLGraphBuilder/reshape(input, newShape)" data-dfn-type="argument" data-export id="dom-mlgraphbuilder-reshape-input-newshape-input"><code><c- g>input</c-></code><a class="self-link" href="#dom-mlgraphbuilder-reshape-input-newshape-input"></a></dfn>, <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#idl-sequence" id="ref-for-idl-sequence②⑧"><c- b>sequence</c-></a>&lt;<a class="idl-code" data-link-type="interface" href="https://webidl.spec.whatwg.org/#idl-unsigned-long" id="ref-for-idl-unsigned-long⑤⑦"><c- b>unsigned</c-> <c- b>long</c-></a>?> <dfn class="idl-code" data-dfn-for="MLGraphBuilder/reshape(input, newShape)" data-dfn-type="argument" data-export id="dom-mlgraphbuilder-reshape-input-newshape-newshape"><code><c- g>newShape</c-></code><a class="self-link" href="#dom-mlgraphbuilder-reshape-input-newshape-newshape"></a></dfn>);
+   </details>
+   <h4 class="heading settled" data-level="7.6.30" id="api-mlgraphbuilder-reshape-method"><span class="secno">7.6.30. </span><span class="content">The reshape() method</span><a class="self-link" href="#api-mlgraphbuilder-reshape-method"></a></h4>
+    Alter the shape of a tensor to a new shape. Reshape does not copy or change the content of the tensor. It just changes the tensor’s logical dimensions for the subsequent operations. 
+<pre class="idl highlight def"><c- b>partial</c-> <c- b>interface</c-> <a class="idl-code" data-link-type="interface" href="#mlgraphbuilder" id="ref-for-mlgraphbuilder④⑧"><c- g>MLGraphBuilder</c-></a> {
+  <a data-link-type="idl-name" href="#mloperand" id="ref-for-mloperand③②⑥"><c- n>MLOperand</c-></a> <a class="idl-code" data-link-type="method" href="#dom-mlgraphbuilder-reshape" id="ref-for-dom-mlgraphbuilder-reshape"><c- g>reshape</c-></a>(<a data-link-type="idl-name" href="#mloperand" id="ref-for-mloperand③②⑦"><c- n>MLOperand</c-></a> <dfn class="idl-code" data-dfn-for="MLGraphBuilder/reshape(input, newShape)" data-dfn-type="argument" data-export id="dom-mlgraphbuilder-reshape-input-newshape-input"><code><c- g>input</c-></code><a class="self-link" href="#dom-mlgraphbuilder-reshape-input-newshape-input"></a></dfn>, <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#idl-sequence" id="ref-for-idl-sequence②⑧"><c- b>sequence</c-></a>&lt;<a class="idl-code" data-link-type="interface" href="https://webidl.spec.whatwg.org/#idl-unsigned-long" id="ref-for-idl-unsigned-long⑤⑨"><c- b>unsigned</c-> <c- b>long</c-></a>?> <dfn class="idl-code" data-dfn-for="MLGraphBuilder/reshape(input, newShape)" data-dfn-type="argument" data-export id="dom-mlgraphbuilder-reshape-input-newshape-newshape"><code><c- g>newShape</c-></code><a class="self-link" href="#dom-mlgraphbuilder-reshape-input-newshape-newshape"></a></dfn>);
 };
 </pre>
-    <div class="algorithm" data-algorithm="reshape">
-      <strong>Arguments:</strong> 
-     <ul>
-      <li data-md>
-       <p><em>input</em>: an <code class="idl"><a data-link-type="idl" href="#mloperand" id="ref-for-mloperand②⑥⑦">MLOperand</a></code>. The input tensor.</p>
-      <li data-md>
-       <p><em>newShape</em>: a sequence of <code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#idl-nullable-type" id="ref-for-idl-nullable-type">nullable</a></code> <code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#idl-unsigned-long" id="ref-for-idl-unsigned-long⑤⑧">unsigned long</a></code>. The shape of the output tensor.
+   <div>
+     <strong>Arguments:</strong> 
+    <ul>
+     <li data-md>
+      <p><em>input</em>: an <code class="idl"><a data-link-type="idl" href="#mloperand" id="ref-for-mloperand③②⑧">MLOperand</a></code>. The input tensor.</p>
+     <li data-md>
+      <p><em>newShape</em>: a sequence of <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#dfn-nullable-type" id="ref-for-dfn-nullable-type">nullable</a> <code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#idl-unsigned-long" id="ref-for-idl-unsigned-long⑥⓪">unsigned long</a></code>. The shape of the output tensor.
 The number of elements implied by <em>newShape</em> must be the same as the
 number of elements in the input tensor. Only one component of <em>newShape</em> can be the special value of <code>null</code>. The size of the dimension
 with the value <code>null</code> is computed so that the total size remains
 constant.</p>
-     </ul>
-     <p><strong>Returns:</strong> an <code class="idl"><a data-link-type="idl" href="#mloperand" id="ref-for-mloperand②⑥⑧">MLOperand</a></code>. The output tensor. The values of the output
+    </ul>
+    <p><strong>Returns:</strong> an <code class="idl"><a data-link-type="idl" href="#mloperand" id="ref-for-mloperand③②⑨">MLOperand</a></code>. The output tensor. The values of the output
     tensor are the same as values of the input tensor. The shape of the output
     tensor is specified by the <em>newShape</em> argument.</p>
-    </div>
-    <h4 class="heading settled" data-level="7.6.30" id="api-mlgraphbuilder-sigmoid"><span class="secno">7.6.30. </span><span class="content">The sigmoid() method</span><a class="self-link" href="#api-mlgraphbuilder-sigmoid"></a></h4>
-     Compute the <a href="https://en.wikipedia.org/wiki/Sigmoid_function">sigmoid function</a> of the input tensor. The calculation follows the expression <code>1 / (exp(-x) + 1)</code>. 
-<pre class="idl highlight def"><c- b>partial</c-> <c- b>interface</c-> <a class="idl-code" data-link-type="interface" href="#mlgraphbuilder" id="ref-for-mlgraphbuilder④⑥"><c- g>MLGraphBuilder</c-></a> {
-  <a data-link-type="idl-name" href="#mloperand" id="ref-for-mloperand②⑥⑨"><c- n>MLOperand</c-></a> <dfn class="idl-code" data-dfn-for="MLGraphBuilder" data-dfn-type="method" data-export data-lt="sigmoid(x)" id="dom-mlgraphbuilder-sigmoid"><code><c- g>sigmoid</c-></code><a class="self-link" href="#dom-mlgraphbuilder-sigmoid"></a></dfn>(<a data-link-type="idl-name" href="#mloperand" id="ref-for-mloperand②⑦⓪"><c- n>MLOperand</c-></a> <dfn class="idl-code" data-dfn-for="MLGraphBuilder/sigmoid(x)" data-dfn-type="argument" data-export id="dom-mlgraphbuilder-sigmoid-x-x"><code><c- g>x</c-></code><a class="self-link" href="#dom-mlgraphbuilder-sigmoid-x-x"></a></dfn>);
-  <a data-link-type="idl-name" href="#mlactivation" id="ref-for-mlactivation③⓪"><c- n>MLActivation</c-></a> <dfn class="idl-code" data-dfn-for="MLGraphBuilder" data-dfn-type="method" data-export data-lt="sigmoid()" id="dom-mlgraphbuilder-sigmoid①"><code><c- g>sigmoid</c-></code><a class="self-link" href="#dom-mlgraphbuilder-sigmoid①"></a></dfn>();
-};
-</pre>
-    <div class="algorithm" data-algorithm="sigmoid">
-      <strong>Arguments:</strong> 
-     <ul>
+   </div>
+   <details class="algorithm" data-algorithm="reshape(input, newShape)" data-algorithm-for="MLGraphBuilder" open>
+    <summary> The <dfn class="dfn-paneled idl-code" data-dfn-for="MLGraphBuilder" data-dfn-type="method" data-export id="dom-mlgraphbuilder-reshape"><code>reshape(<var>input</var>, <var>newShape</var>)</code></dfn> method steps are: </summary>
+    <div class="algorithm-steps">
+     <ol>
       <li data-md>
-       <p><em>x</em>: an <code class="idl"><a data-link-type="idl" href="#mloperand" id="ref-for-mloperand②⑦①">MLOperand</a></code>. The input tensor.</p>
-     </ul>
-     <p><strong>Returns:</strong></p>
-     <ul>
+       <p><a data-link-type="dfn" href="https://infra.spec.whatwg.org/#assert" id="ref-for-assert⑥④">Assert</a>: the type of <var>input</var> is <code class="idl"><a data-link-type="idl" href="#mloperand" id="ref-for-mloperand③③⓪">MLOperand</a></code>.</p>
       <li data-md>
-       <p>an <code class="idl"><a data-link-type="idl" href="#mloperand" id="ref-for-mloperand②⑦②">MLOperand</a></code>. The output tensor of the same shape as <em>x</em>.</p>
+       <p>Let <var>outputShape</var> be an empty array of <code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#idl-unsigned-long" id="ref-for-idl-unsigned-long⑥①">unsigned long</a></code>.</p>
       <li data-md>
-       <p>an <code class="idl"><a data-link-type="idl" href="#mlactivation" id="ref-for-mlactivation③①">MLActivation</a></code>. The activation function representing the sigmoid operation.</p>
-     </ul>
-     <div class="note" role="note">
-       The behavior of this operation can be generically emulated from the usage of
+       <p>If <var>newShape</var> is a scalar <a data-link-type="dfn" href="https://drafts.csswg.org/css-values-4/#number" id="ref-for-number">number</a>, set <var>outputShape</var> to <code>«  1  »</code>.</p>
+      <li data-md>
+       <p>Otherwise, if <var>newShape</var> is an array of <code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#idl-unsigned-long" id="ref-for-idl-unsigned-long⑥②">unsigned long</a></code>:</p>
+       <ol>
+        <li data-md>
+         <p>If the <a data-link-type="dfn" href="https://infra.spec.whatwg.org/#list-size" id="ref-for-list-size④⑨">size</a> of <var>newShape</var> is 0, set <var>outputShape</var> to <code>«  1  »</code> (reshaping to scalar).</p>
+        <li data-md>
+         <p>If <var>newShape</var> contains more than one <code>null</code> value, then <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#dfn-throw" id="ref-for-dfn-throw②④⑥">throw</a> a "<code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#dataerror" id="ref-for-dataerror⑧③">DataError</a></code>" <code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#idl-DOMException" id="ref-for-idl-DOMException①③③">DOMException</a></code>.</p>
+        <li data-md>
+         <p>If any value in <var>newShape</var> is 0, then <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#dfn-throw" id="ref-for-dfn-throw②④⑦">throw</a> a "<code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#dataerror" id="ref-for-dataerror⑧④">DataError</a></code>" <code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#idl-DOMException" id="ref-for-idl-DOMException①③④">DOMException</a></code>.</p>
+        <li data-md>
+         <p>Let <var>inputElementCount</var> be the product of all elements in <var>input</var>.<code class="idl"><a data-link-type="idl" href="#dom-mloperand-descriptor-slot" id="ref-for-dom-mloperand-descriptor-slot⑨⓪">[[descriptor]]</a></code>.<code class="idl"><a data-link-type="idl" href="#dom-mloperanddescriptor-dimensions" id="ref-for-dom-mloperanddescriptor-dimensions⑧⑧">dimensions</a></code>.</p>
+        <li data-md>
+         <p>If <var>newShape</var> contains a <code>null</code> value, set that value to <var>inputElementCount</var> divided by the product of all other values in <var>newShape</var>.</p>
+         <ol>
+          <li data-md>
+           <p>If that value is too large for <code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#idl-unsigned-long" id="ref-for-idl-unsigned-long⑥③">unsigned long</a></code>, then <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#dfn-throw" id="ref-for-dfn-throw②④⑧">throw</a> a "<code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#dataerror" id="ref-for-dataerror⑧⑤">DataError</a></code>" <code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#idl-DOMException" id="ref-for-idl-DOMException①③⑤">DOMException</a></code>.</p>
+         </ol>
+        <li data-md>
+         <p>If product of all values in <var>newShape</var> is not equal to <var>inputElementCount</var>, then <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#dfn-throw" id="ref-for-dfn-throw②④⑨">throw</a> a "<code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#dataerror" id="ref-for-dataerror⑧⑥">DataError</a></code>" <code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#idl-DOMException" id="ref-for-idl-DOMException①③⑥">DOMException</a></code>.</p>
+       </ol>
+      <li data-md>
+       <p>Let <var>desc</var> be a copy of <var>input</var>.<code class="idl"><a data-link-type="idl" href="#dom-mloperand-descriptor-slot" id="ref-for-dom-mloperand-descriptor-slot⑨①">[[descriptor]]</a></code>.</p>
+      <li data-md>
+       <p>Set <var>desc</var>.<code class="idl"><a data-link-type="idl" href="#dom-mloperanddescriptor-dimensions" id="ref-for-dom-mloperanddescriptor-dimensions⑧⑨">dimensions</a></code> to <var>newShape</var>.</p>
+      <li data-md>
+       <p>If any of the following sub-steps fail, <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#dfn-throw" id="ref-for-dfn-throw②⑤⓪">throw</a> an "<code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#operationerror" id="ref-for-operationerror④①">OperationError</a></code>" <code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#idl-DOMException" id="ref-for-idl-DOMException①③⑦">DOMException</a></code>.</p>
+       <ol>
+        <li data-md>
+         <p>Let <var>output</var> be the result of <a data-link-type="dfn" href="#create-an-mloperand" id="ref-for-create-an-mloperand②⓪">creating an MLOperand</a> given <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#this" id="ref-for-this③⑤">this</a> and <var>desc</var>.</p>
+        <li data-md>
+         <p>Make a request to the underlying platform to:</p>
+         <ol>
+          <li data-md>
+           <p>Let <var>opImpl</var> be an <a data-link-type="dfn" href="https://infra.spec.whatwg.org/#implementation-defined" id="ref-for-implementation-defined⑤⑧">implementation-defined</a> platform operator for the reshape operation.</p>
+          <li data-md>
+           <p>Store a reference of <var>opImpl</var> in <var>output</var>.<code class="idl"><a data-link-type="idl" href="#dom-mloperand-operator-slot" id="ref-for-dom-mloperand-operator-slot②⑦">[[operator]]</a></code>.</p>
+          <li data-md>
+           <p>Create an <a data-link-type="dfn" href="https://infra.spec.whatwg.org/#implementation-defined" id="ref-for-implementation-defined⑤⑨">implementation-defined</a> platform operand <var>outputImpl</var> to represent the output, given <var>output</var> and <var>opImpl</var>.</p>
+          <li data-md>
+           <p>Store a reference to <var>outputImpl</var> in <var>output</var>.<code class="idl"><a data-link-type="idl" href="#dom-mloperand-operand-slot" id="ref-for-dom-mloperand-operand-slot⑦⑦">[[operand]]</a></code>.</p>
+         </ol>
+        <li data-md>
+         <p>Connect <var>input</var>.<code class="idl"><a data-link-type="idl" href="#dom-mloperand-operand-slot" id="ref-for-dom-mloperand-operand-slot⑦⑧">[[operand]]</a></code> as input to <var>opImpl</var>.</p>
+        <li data-md>
+         <p>Connect <var>output</var>.<code class="idl"><a data-link-type="idl" href="#dom-mloperand-operand-slot" id="ref-for-dom-mloperand-operand-slot⑦⑨">[[operand]]</a></code> as output to <var>opImpl</var>.</p>
+       </ol>
+      <li data-md>
+       <p>Return <var>output</var>.</p>
+     </ol>
+    </div>
+   </details>
+   <h4 class="heading settled" data-level="7.6.31" id="api-mlgraphbuilder-sigmoid-method"><span class="secno">7.6.31. </span><span class="content">The sigmoid() method</span><a class="self-link" href="#api-mlgraphbuilder-sigmoid-method"></a></h4>
+    Compute the <a href="https://en.wikipedia.org/wiki/Sigmoid_function">sigmoid function</a> of the input tensor. The calculation follows the expression <code>1 / (exp(-x) + 1)</code>. 
+<pre class="idl highlight def"><c- b>partial</c-> <c- b>interface</c-> <a class="idl-code" data-link-type="interface" href="#mlgraphbuilder" id="ref-for-mlgraphbuilder④⑨"><c- g>MLGraphBuilder</c-></a> {
+  <a data-link-type="idl-name" href="#mloperand" id="ref-for-mloperand③③①"><c- n>MLOperand</c-></a> <a class="idl-code" data-link-type="method" href="#dom-mlgraphbuilder-sigmoid" id="ref-for-dom-mlgraphbuilder-sigmoid"><c- g>sigmoid</c-></a>(<a data-link-type="idl-name" href="#mloperand" id="ref-for-mloperand③③②"><c- n>MLOperand</c-></a> <dfn class="idl-code" data-dfn-for="MLGraphBuilder/sigmoid(input)" data-dfn-type="argument" data-export id="dom-mlgraphbuilder-sigmoid-input-input"><code><c- g>input</c-></code><a class="self-link" href="#dom-mlgraphbuilder-sigmoid-input-input"></a></dfn>);
+  <a data-link-type="idl-name" href="#mlactivation" id="ref-for-mlactivation④⑦"><c- n>MLActivation</c-></a> <a class="idl-code" data-link-type="method" href="#sigmoid-noargs" id="ref-for-sigmoid-noargs"><c- g>sigmoid</c-></a>();
+};
+</pre>
+   <div class="note" role="note">
+    <details open>
+     <summary> The behavior of this operation can be generically emulated from the usage of
     other operations as follow. However, user agents typically have a more
     efficient implementation for it, therefore its usage is encouraged from the
-    performance standpoint. 
+    performance standpoint. </summary>
 <pre class="highlight"><c- k>return</c-> builder<c- p>.</c->div<c- p>(</c->
           builder<c- p>.</c->constant<c- p>(</c-><c- mf>1</c-><c- p>),</c->
           builder<c- p>.</c->add<c- p>(</c->
             builder<c- p>.</c->exp<c- p>(</c->builder<c- p>.</c->neg<c- p>(</c->x<c- p>)),</c->
             builder<c- p>.</c->constant<c- p>(</c-><c- mf>1</c-><c- p>)));</c->
 </pre>
-     </div>
+    </details>
+   </div>
+   <h5 class="heading settled" data-level="7.6.31.1" id="api-mlgraphbuilder-sigmoid-input"><span class="secno">7.6.31.1. </span><span class="content">The <code class="idl"><a data-link-type="idl" href="#dom-mlgraphbuilder-sigmoid" id="ref-for-dom-mlgraphbuilder-sigmoid①">sigmoid(input)</a></code> method</span><a class="self-link" href="#api-mlgraphbuilder-sigmoid-input"></a></h5>
+   <div>
+     <strong>Arguments:</strong> 
+    <ul>
+     <li data-md>
+      <p><em>input</em>: an <code class="idl"><a data-link-type="idl" href="#mloperand" id="ref-for-mloperand③③③">MLOperand</a></code>. The input tensor.</p>
+    </ul>
+    <p><strong>Returns:</strong></p>
+    <ul>
+     <li data-md>
+      <p>an <code class="idl"><a data-link-type="idl" href="#mloperand" id="ref-for-mloperand③③④">MLOperand</a></code>. The output tensor of the same shape as <em>input</em>.</p>
+    </ul>
+   </div>
+   <details class="algorithm" data-algorithm="sigmoid(input)" data-algorithm-for="MLGraphBuilder" open>
+    <summary> The <dfn class="dfn-paneled idl-code" data-dfn-for="MLGraphBuilder" data-dfn-type="method" data-export id="dom-mlgraphbuilder-sigmoid"><code>sigmoid(<var>input</var>)</code></dfn> method steps are: </summary>
+    <div class="algorithm-steps">
+     <ol>
+      <li data-md>
+       <p><a data-link-type="dfn" href="https://infra.spec.whatwg.org/#assert" id="ref-for-assert⑥⑤">Assert</a>: the type of <var>input</var> is <code class="idl"><a data-link-type="idl" href="#mloperand" id="ref-for-mloperand③③⑤">MLOperand</a></code>.</p>
+      <li data-md>
+       <p>If any of the following sub-steps fail, <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#dfn-throw" id="ref-for-dfn-throw②⑤①">throw</a> an "<code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#operationerror" id="ref-for-operationerror④②">OperationError</a></code>" <code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#idl-DOMException" id="ref-for-idl-DOMException①③⑧">DOMException</a></code>.</p>
+       <ol>
+        <li data-md>
+         <p>Let <var>output</var> be the result of <a data-link-type="dfn" href="#copy-an-mloperand" id="ref-for-copy-an-mloperand①⓪">copying an MLOperand</a> given <var>input</var>.</p>
+        <li data-md>
+         <p>Make a request to the underlying platform to:</p>
+         <ol>
+          <li data-md>
+           <p>Let <var>opImpl</var> be an <a data-link-type="dfn" href="https://infra.spec.whatwg.org/#implementation-defined" id="ref-for-implementation-defined⑥⓪">implementation-defined</a> platform operator for the sigmoid operation.</p>
+          <li data-md>
+           <p>Store a reference of <var>opImpl</var> in <var>output</var>.<code class="idl"><a data-link-type="idl" href="#dom-mloperand-operator-slot" id="ref-for-dom-mloperand-operator-slot②⑧">[[operator]]</a></code>.</p>
+          <li data-md>
+           <p>Create an <a data-link-type="dfn" href="https://infra.spec.whatwg.org/#implementation-defined" id="ref-for-implementation-defined⑥①">implementation-defined</a> platform operand <var>outputImpl</var> to represent the output, given <var>output</var> and <var>opImpl</var>.</p>
+          <li data-md>
+           <p>Store a reference to <var>outputImpl</var> in <var>output</var>.<code class="idl"><a data-link-type="idl" href="#dom-mloperand-operand-slot" id="ref-for-dom-mloperand-operand-slot⑧⓪">[[operand]]</a></code>.</p>
+         </ol>
+        <li data-md>
+         <p>Connect <var>input</var>.<code class="idl"><a data-link-type="idl" href="#dom-mloperand-operand-slot" id="ref-for-dom-mloperand-operand-slot⑧①">[[operand]]</a></code> as input to <var>opImpl</var>.</p>
+        <li data-md>
+         <p>Connect <var>output</var>.<code class="idl"><a data-link-type="idl" href="#dom-mloperand-operand-slot" id="ref-for-dom-mloperand-operand-slot⑧②">[[operand]]</a></code> as output to <var>opImpl</var>.</p>
+       </ol>
+      <li data-md>
+       <p>Return <var>output</var>.</p>
+     </ol>
+    </div>
+   </details>
+   <h5 class="heading settled" data-level="7.6.31.2" id="api-mlgraphbuilder-sigmoid"><span class="secno">7.6.31.2. </span><span class="content">The <code class="idl"><a data-link-type="idl" href="#sigmoid-noargs" id="ref-for-sigmoid-noargs①">sigmoid()</a></code> method</span><a class="self-link" href="#api-mlgraphbuilder-sigmoid"></a></h5>
+   <div>
+     <strong>Arguments:</strong> 
+    <ul>
+     <li data-md>
+      <p>None.</p>
+    </ul>
+    <p><strong>Returns:</strong></p>
+    <ul>
+     <li data-md>
+      <p>an <code class="idl"><a data-link-type="idl" href="#mlactivation" id="ref-for-mlactivation④⑧">MLActivation</a></code>. The activation function representing the sigmoid operation.</p>
+    </ul>
+   </div>
+   <details class="algorithm" data-algorithm="sigmoid()" data-algorithm-for="MLGraphBuilder" open>
+    <summary> The <dfn class="dfn-paneled idl-code" data-dfn-for="MLGraphBuilder" data-dfn-type="method" data-export id="sigmoid-noargs"><code>sigmoid()</code></dfn> method steps are: </summary>
+    <div class="algorithm-steps">
+     <ol>
+      <li data-md>
+       <p>Let <var>op</var> be the result of <a data-link-type="dfn" href="#create-an-mlactivation" id="ref-for-create-an-mlactivation⑦">creating an MLActivation</a> given <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#this" id="ref-for-this③⑥">this</a> and <code>"sigmoid"</code>.</p>
+       <ol>
+        <li data-md>
+         <p>If that <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#dfn-throw" id="ref-for-dfn-throw②⑤②">throws</a> an error, re-<a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#dfn-throw" id="ref-for-dfn-throw②⑤③">throw</a> the error.</p>
+       </ol>
+      <li data-md>
+       <p>Return <var>op</var>.</p>
+     </ol>
     </div>
-    <h4 class="heading settled" data-level="7.6.31" id="api-mlgraphbuilder-slice"><span class="secno">7.6.31. </span><span class="content">The slice() method</span><a class="self-link" href="#api-mlgraphbuilder-slice"></a></h4>
-     Produce a slice of the input tensor. 
-<pre class="idl highlight def"><c- b>partial</c-> <c- b>interface</c-> <a class="idl-code" data-link-type="interface" href="#mlgraphbuilder" id="ref-for-mlgraphbuilder④⑦"><c- g>MLGraphBuilder</c-></a> {
-  <a data-link-type="idl-name" href="#mloperand" id="ref-for-mloperand②⑦③"><c- n>MLOperand</c-></a> <dfn class="idl-code" data-dfn-for="MLGraphBuilder" data-dfn-type="method" data-export data-lt="slice(input, starts, sizes)" id="dom-mlgraphbuilder-slice"><code><c- g>slice</c-></code><a class="self-link" href="#dom-mlgraphbuilder-slice"></a></dfn>(<a data-link-type="idl-name" href="#mloperand" id="ref-for-mloperand②⑦④"><c- n>MLOperand</c-></a> <dfn class="idl-code" data-dfn-for="MLGraphBuilder/slice(input, starts, sizes)" data-dfn-type="argument" data-export id="dom-mlgraphbuilder-slice-input-starts-sizes-input"><code><c- g>input</c-></code><a class="self-link" href="#dom-mlgraphbuilder-slice-input-starts-sizes-input"></a></dfn>, <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#idl-sequence" id="ref-for-idl-sequence②⑨"><c- b>sequence</c-></a>&lt;<a class="idl-code" data-link-type="interface" href="https://webidl.spec.whatwg.org/#idl-unsigned-long" id="ref-for-idl-unsigned-long⑤⑨"><c- b>unsigned</c-> <c- b>long</c-></a>> <dfn class="idl-code" data-dfn-for="MLGraphBuilder/slice(input, starts, sizes)" data-dfn-type="argument" data-export id="dom-mlgraphbuilder-slice-input-starts-sizes-starts"><code><c- g>starts</c-></code><a class="self-link" href="#dom-mlgraphbuilder-slice-input-starts-sizes-starts"></a></dfn>, <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#idl-sequence" id="ref-for-idl-sequence③⓪"><c- b>sequence</c-></a>&lt;<a class="idl-code" data-link-type="interface" href="https://webidl.spec.whatwg.org/#idl-unsigned-long" id="ref-for-idl-unsigned-long⑥⓪"><c- b>unsigned</c-> <c- b>long</c-></a>> <dfn class="idl-code" data-dfn-for="MLGraphBuilder/slice(input, starts, sizes)" data-dfn-type="argument" data-export id="dom-mlgraphbuilder-slice-input-starts-sizes-sizes"><code><c- g>sizes</c-></code><a class="self-link" href="#dom-mlgraphbuilder-slice-input-starts-sizes-sizes"></a></dfn>);
+   </details>
+   <h4 class="heading settled" data-level="7.6.32" id="api-mlgraphbuilder-slice"><span class="secno">7.6.32. </span><span class="content">The slice() method</span><a class="self-link" href="#api-mlgraphbuilder-slice"></a></h4>
+    Produce a slice of the input tensor. 
+<pre class="idl highlight def"><c- b>partial</c-> <c- b>interface</c-> <a class="idl-code" data-link-type="interface" href="#mlgraphbuilder" id="ref-for-mlgraphbuilder⑤⓪"><c- g>MLGraphBuilder</c-></a> {
+  <a data-link-type="idl-name" href="#mloperand" id="ref-for-mloperand③③⑥"><c- n>MLOperand</c-></a> <a class="idl-code" data-link-type="method" href="#dom-mlgraphbuilder-slice" id="ref-for-dom-mlgraphbuilder-slice"><c- g>slice</c-></a>(<a data-link-type="idl-name" href="#mloperand" id="ref-for-mloperand③③⑦"><c- n>MLOperand</c-></a> <dfn class="idl-code" data-dfn-for="MLGraphBuilder/slice(input, starts, sizes)" data-dfn-type="argument" data-export id="dom-mlgraphbuilder-slice-input-starts-sizes-input"><code><c- g>input</c-></code><a class="self-link" href="#dom-mlgraphbuilder-slice-input-starts-sizes-input"></a></dfn>, <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#idl-sequence" id="ref-for-idl-sequence②⑨"><c- b>sequence</c-></a>&lt;<a class="idl-code" data-link-type="interface" href="https://webidl.spec.whatwg.org/#idl-unsigned-long" id="ref-for-idl-unsigned-long⑥④"><c- b>unsigned</c-> <c- b>long</c-></a>> <dfn class="idl-code" data-dfn-for="MLGraphBuilder/slice(input, starts, sizes)" data-dfn-type="argument" data-export id="dom-mlgraphbuilder-slice-input-starts-sizes-starts"><code><c- g>starts</c-></code><a class="self-link" href="#dom-mlgraphbuilder-slice-input-starts-sizes-starts"></a></dfn>, <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#idl-sequence" id="ref-for-idl-sequence③⓪"><c- b>sequence</c-></a>&lt;<a class="idl-code" data-link-type="interface" href="https://webidl.spec.whatwg.org/#idl-unsigned-long" id="ref-for-idl-unsigned-long⑥⑤"><c- b>unsigned</c-> <c- b>long</c-></a>> <dfn class="idl-code" data-dfn-for="MLGraphBuilder/slice(input, starts, sizes)" data-dfn-type="argument" data-export id="dom-mlgraphbuilder-slice-input-starts-sizes-sizes"><code><c- g>sizes</c-></code><a class="self-link" href="#dom-mlgraphbuilder-slice-input-starts-sizes-sizes"></a></dfn>);
 };
 </pre>
-    <div class="algorithm" data-algorithm="slice">
-      <strong>Arguments:</strong> 
-     <ul>
+   <div>
+     <strong>Arguments:</strong> 
+    <ul>
+     <li data-md>
+      <p><em>input</em>: an <code class="idl"><a data-link-type="idl" href="#mloperand" id="ref-for-mloperand③③⑧">MLOperand</a></code>. The input tensor.</p>
+     <li data-md>
+      <p><em>starts</em>: a sequence of <code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#idl-unsigned-long" id="ref-for-idl-unsigned-long⑥⑥">unsigned long</a></code>. The sequence of unsigned integer values indicating the starting index to slice of each input dimension, of length N where N is the <a data-link-type="dfn" href="#mloperand-rank" id="ref-for-mloperand-rank②①">rank</a> of the input tensor. For each dimension <em>d</em> of <em>input</em>, <em>starts[d]</em> indicates the starting index to slice in that dimension. The starting index must be in the range [0, input size - 1] in that dimension.</p>
+     <li data-md>
+      <p><em>sizes</em>: a sequence of <code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#idl-unsigned-long" id="ref-for-idl-unsigned-long⑥⑦">unsigned long</a></code>. The sequence of unsigned integer values indicating the number of elements to slice of each input dimension, of length N where N is the <a data-link-type="dfn" href="#mloperand-rank" id="ref-for-mloperand-rank②②">rank</a> of the input tensor. For each dimension <em>d</em> of <em>input</em>, <em>sizes[d]</em> indicates the number of elements to slice in that dimension. The size must not be 0 and must satisfy the constraint <em>starting index + size &lt;= input size</em> in that dimension.</p>
+    </ul>
+    <p><strong>Returns:</strong> an <code class="idl"><a data-link-type="idl" href="#mloperand" id="ref-for-mloperand③③⑨">MLOperand</a></code>. The output tensor of the same rank as the input tensor with tensor values stripped to the specified starting and ending indices in each dimension.</p>
+   </div>
+   <details class="algorithm" data-algorithm="slice(input, starts, sizes)" data-algorithm-for="MLGraphBuilder" open>
+    <summary> The <dfn class="dfn-paneled idl-code" data-dfn-for="MLGraphBuilder" data-dfn-type="method" data-export id="dom-mlgraphbuilder-slice"><code>slice(<var>input</var>, <var>starts</var>, <var>sizes</var>)</code></dfn> method steps are: </summary>
+    <div class="algorithm-steps">
+     <ol>
       <li data-md>
-       <p><em>input</em>: an <code class="idl"><a data-link-type="idl" href="#mloperand" id="ref-for-mloperand②⑦⑤">MLOperand</a></code>. The input tensor.</p>
+       <p><a data-link-type="dfn" href="https://infra.spec.whatwg.org/#assert" id="ref-for-assert⑥⑥">Assert</a>: the type of <var>input</var> is <code class="idl"><a data-link-type="idl" href="#mloperand" id="ref-for-mloperand③④⓪">MLOperand</a></code>.</p>
       <li data-md>
-       <p><em>starts</em>: a sequence of <code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#idl-unsigned-long" id="ref-for-idl-unsigned-long⑥①">unsigned long</a></code>. The sequence of unsigned integer values indicating the starting index to slice of each input dimension, of length N where N is the rank of the input tensor. For each dimension <em>d</em> of <em>input</em>, <em>starts[d]</em> indicates the starting index to slice in that dimension. The starting index must be in the range [0, input size - 1] in that dimension.</p>
+       <p>If <var>sizes</var>.size is 0, then <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#dfn-throw" id="ref-for-dfn-throw②⑤④">throw</a> a <code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#exceptiondef-typeerror" id="ref-for-exceptiondef-typeerror④⓪">TypeError</a></code>.</p>
       <li data-md>
-       <p><em>sizes</em>: a sequence of <code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#idl-unsigned-long" id="ref-for-idl-unsigned-long⑥②">unsigned long</a></code>. The sequence of unsigned integer values indicating the number of elements to slice of each input dimension, of length N where N is the rank of the input tensor. For each dimension <em>d</em> of <em>input</em>, <em>sizes[d]</em> indicates the number of elements to slice in that dimension. The size must not be 0 and must satisfy the constraint <em>starting index + size &lt;= input size</em> in that dimension.</p>
-     </ul>
-     <p><strong>Returns:</strong> an <code class="idl"><a data-link-type="idl" href="#mloperand" id="ref-for-mloperand②⑦⑥">MLOperand</a></code>. The output tensor of the same rank as the input tensor with tensor values stripped to the specified starting and ending indices in each dimension.</p>
+       <p>If the <a data-link-type="dfn" href="https://infra.spec.whatwg.org/#list-size" id="ref-for-list-size⑤⓪">size</a> of <var>starts</var> and <var>sizes</var> is not equal to the <a data-link-type="dfn" href="#mloperand-rank" id="ref-for-mloperand-rank②③">rank</a> of <var>input</var>, then <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#dfn-throw" id="ref-for-dfn-throw②⑤⑤">throw</a> a <code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#exceptiondef-typeerror" id="ref-for-exceptiondef-typeerror④①">TypeError</a></code>.</p>
+      <li data-md>
+       <p>If any of the following sub-steps fail, <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#dfn-throw" id="ref-for-dfn-throw②⑤⑥">throw</a> an "<code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#operationerror" id="ref-for-operationerror④③">OperationError</a></code>" <code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#idl-DOMException" id="ref-for-idl-DOMException①③⑨">DOMException</a></code>.</p>
+       <ol>
+        <li data-md>
+         <p>Let <var>output</var> be the result of <a data-link-type="dfn" href="#copy-an-mloperand" id="ref-for-copy-an-mloperand①①">copying an MLOperand</a> given <var>input</var>.</p>
+        <li data-md>
+         <p>Make a request to the underlying platform to:</p>
+         <ol>
+          <li data-md>
+           <p>Let <var>opImpl</var> be an <a data-link-type="dfn" href="https://infra.spec.whatwg.org/#implementation-defined" id="ref-for-implementation-defined⑥②">implementation-defined</a> platform operator for the slice operation, given <var>starts</var> and <var>sizes</var>.</p>
+          <li data-md>
+           <p>Store a reference of <var>opImpl</var> in <var>output</var>.<code class="idl"><a data-link-type="idl" href="#dom-mloperand-operator-slot" id="ref-for-dom-mloperand-operator-slot②⑨">[[operator]]</a></code>.</p>
+          <li data-md>
+           <p>Create an <a data-link-type="dfn" href="https://infra.spec.whatwg.org/#implementation-defined" id="ref-for-implementation-defined⑥③">implementation-defined</a> platform operand <var>outputImpl</var> to represent the output, given <var>output</var> and <var>opImpl</var>.</p>
+          <li data-md>
+           <p>Store a reference to <var>outputImpl</var> in <var>output</var>.<code class="idl"><a data-link-type="idl" href="#dom-mloperand-operand-slot" id="ref-for-dom-mloperand-operand-slot⑧③">[[operand]]</a></code>.</p>
+         </ol>
+        <li data-md>
+         <p>Connect <var>input</var>.<code class="idl"><a data-link-type="idl" href="#dom-mloperand-operand-slot" id="ref-for-dom-mloperand-operand-slot⑧④">[[operand]]</a></code> as input to <var>opImpl</var>.</p>
+        <li data-md>
+         <p>Connect <var>output</var>.<code class="idl"><a data-link-type="idl" href="#dom-mloperand-operand-slot" id="ref-for-dom-mloperand-operand-slot⑧⑤">[[operand]]</a></code> as output to <var>opImpl</var>.</p>
+       </ol>
+      <li data-md>
+       <p>Return <var>output</var>.</p>
+     </ol>
     </div>
-    <h4 class="heading settled" data-level="7.6.32" id="api-mlgraphbuilder-softmax"><span class="secno">7.6.32. </span><span class="content">The softmax() method</span><a class="self-link" href="#api-mlgraphbuilder-softmax"></a></h4>
-     Compute the <a href="https://en.wikipedia.org/wiki/Softmax_function">softmax</a> values of
+   </details>
+   <h4 class="heading settled" data-level="7.6.33" id="api-mlgraphbuilder-softmax-method"><span class="secno">7.6.33. </span><span class="content">The softmax() method</span><a class="self-link" href="#api-mlgraphbuilder-softmax-method"></a></h4>
+    Compute the <a href="https://en.wikipedia.org/wiki/Softmax_function">softmax</a> values of
 the 2-D input tensor along axis 1. 
-<pre class="idl highlight def"><c- b>partial</c-> <c- b>interface</c-> <a class="idl-code" data-link-type="interface" href="#mlgraphbuilder" id="ref-for-mlgraphbuilder④⑧"><c- g>MLGraphBuilder</c-></a> {
-  <a data-link-type="idl-name" href="#mloperand" id="ref-for-mloperand②⑦⑦"><c- n>MLOperand</c-></a> <dfn class="idl-code" data-dfn-for="MLGraphBuilder" data-dfn-type="method" data-export data-lt="softmax(x)" id="dom-mlgraphbuilder-softmax"><code><c- g>softmax</c-></code><a class="self-link" href="#dom-mlgraphbuilder-softmax"></a></dfn>(<a data-link-type="idl-name" href="#mloperand" id="ref-for-mloperand②⑦⑧"><c- n>MLOperand</c-></a> <dfn class="idl-code" data-dfn-for="MLGraphBuilder/softmax(x)" data-dfn-type="argument" data-export id="dom-mlgraphbuilder-softmax-x-x"><code><c- g>x</c-></code><a class="self-link" href="#dom-mlgraphbuilder-softmax-x-x"></a></dfn>);
-  <a data-link-type="idl-name" href="#mlactivation" id="ref-for-mlactivation③②"><c- n>MLActivation</c-></a> <dfn class="dfn-paneled idl-code" data-dfn-for="MLGraphBuilder" data-dfn-type="method" data-export data-lt="softmax()" id="dom-mlgraphbuilder-softmax①"><code><c- g>softmax</c-></code></dfn>();
+<pre class="idl highlight def"><c- b>partial</c-> <c- b>interface</c-> <a class="idl-code" data-link-type="interface" href="#mlgraphbuilder" id="ref-for-mlgraphbuilder⑤①"><c- g>MLGraphBuilder</c-></a> {
+  <a data-link-type="idl-name" href="#mloperand" id="ref-for-mloperand③④①"><c- n>MLOperand</c-></a> <a class="idl-code" data-link-type="method" href="#dom-mlgraphbuilder-softmax" id="ref-for-dom-mlgraphbuilder-softmax"><c- g>softmax</c-></a>(<a data-link-type="idl-name" href="#mloperand" id="ref-for-mloperand③④②"><c- n>MLOperand</c-></a> <dfn class="idl-code" data-dfn-for="MLGraphBuilder/softmax(input)" data-dfn-type="argument" data-export id="dom-mlgraphbuilder-softmax-input-input"><code><c- g>input</c-></code><a class="self-link" href="#dom-mlgraphbuilder-softmax-input-input"></a></dfn>);
+  <a data-link-type="idl-name" href="#mlactivation" id="ref-for-mlactivation④⑨"><c- n>MLActivation</c-></a> <a class="idl-code" data-link-type="method" href="#softmax-noargs" id="ref-for-softmax-noargs①"><c- g>softmax</c-></a>();
 };
 </pre>
-    <div class="algorithm" data-algorithm="softmax">
-      <strong>Arguments:</strong> 
-     <ul>
-      <li data-md>
-       <p><em>x</em>: an <code class="idl"><a data-link-type="idl" href="#mloperand" id="ref-for-mloperand②⑦⑨">MLOperand</a></code>. The input 2-D tensor.</p>
-     </ul>
-     <p><strong>Returns:</strong></p>
-     <ul>
-      <li data-md>
-       <p>an <code class="idl"><a data-link-type="idl" href="#mloperand" id="ref-for-mloperand②⑧⓪">MLOperand</a></code>. The output 2-D tensor that contains the softmax results, of the same shape as the input tensor.</p>
-      <li data-md>
-       <p>an <code class="idl"><a data-link-type="idl" href="#mlactivation" id="ref-for-mlactivation③③">MLActivation</a></code>. The activation function representing the softmax operation.</p>
-     </ul>
-     <div class="note" role="note">
-       The behavior of this operation can be generically emulated from the usage of
+   <div class="note" role="note">
+    <details open>
+     <summary> The behavior of this operation can be generically emulated from the usage of
     other operations as follow. However, user agents typically have a more
     efficient implementation for it, therefore its usage is encouraged from the
-    performance standpoint. 
+    performance standpoint. </summary>
 <pre class="highlight"><c- c1>// This sample deploys a well-known implementation trick [1] to compute the</c->
 <c- c1>// exponentials of the distances to the max value, instead of the exponentials</c->
 <c- c1>// of the input values itself, in order to increase the numerical stability of</c->
@@ -3956,37 +6909,101 @@ <h4 class="heading settled" data-level="7.6.32" id="api-mlgraphbuilder-softmax">
 <c- a>const</c-> exp_x <c- o>=</c-> builder<c- p>.</c->exp<c- p>(</c->builder<c- p>.</c->sub<c- p>(</c->x<c- p>,</c-> max_x<c- p>));</c->
 <c- k>return</c-> builder<c- p>.</c->div<c- p>(</c->exp_x<c- p>,</c-> builder<c- p>.</c->reduceSum<c- p>(</c->exp_x<c- p>,</c-> <c- p>{</c-> axes<c- o>:</c-> <c- p>[</c-><c- mf>1</c-><c- p>],</c-> keepDimensions<c- o>:</c-> <c- kc>true</c-> <c- p>}));</c->
 </pre>
-     </div>
+    </details>
+   </div>
+   <h5 class="heading settled" data-level="7.6.33.1" id="api-mlgraphbuilder-softmax-input"><span class="secno">7.6.33.1. </span><span class="content">The <code class="idl"><a data-link-type="idl" href="#dom-mlgraphbuilder-softmax" id="ref-for-dom-mlgraphbuilder-softmax①">softmax(input)</a></code> method</span><a class="self-link" href="#api-mlgraphbuilder-softmax-input"></a></h5>
+   <div>
+     <strong>Arguments:</strong> 
+    <ul>
+     <li data-md>
+      <p><em>input</em>: an <code class="idl"><a data-link-type="idl" href="#mloperand" id="ref-for-mloperand③④③">MLOperand</a></code>. The input 2-D tensor.</p>
+    </ul>
+    <p><strong>Returns:</strong></p>
+    <ul>
+     <li data-md>
+      <p>an <code class="idl"><a data-link-type="idl" href="#mloperand" id="ref-for-mloperand③④④">MLOperand</a></code>. The output 2-D tensor that contains the softmax results, of the same shape as <em>input</em>.</p>
+    </ul>
+   </div>
+   <details class="algorithm" data-algorithm="softmax(input)" data-algorithm-for="MLGraphBuilder" open>
+    <summary> The <dfn class="dfn-paneled idl-code" data-dfn-for="MLGraphBuilder" data-dfn-type="method" data-export id="dom-mlgraphbuilder-softmax"><code>softmax(<var>input</var>)</code></dfn> method steps are: </summary>
+    <div class="algorithm-steps">
+     <ol>
+      <li data-md>
+       <p><a data-link-type="dfn" href="https://infra.spec.whatwg.org/#assert" id="ref-for-assert⑥⑦">Assert</a>: the type of <var>input</var> is <code class="idl"><a data-link-type="idl" href="#mloperand" id="ref-for-mloperand③④⑤">MLOperand</a></code>.</p>
+      <li data-md>
+       <p>If the <a data-link-type="dfn" href="https://infra.spec.whatwg.org/#list-size" id="ref-for-list-size⑤①">size</a> of <var>input</var>.<code class="idl"><a data-link-type="idl" href="#dom-mloperand-descriptor-slot" id="ref-for-dom-mloperand-descriptor-slot⑨②">[[descriptor]]</a></code>.<code class="idl"><a data-link-type="idl" href="#dom-mloperanddescriptor-dimensions" id="ref-for-dom-mloperanddescriptor-dimensions⑨⓪">dimensions</a></code> is not 2, then <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#dfn-throw" id="ref-for-dfn-throw②⑤⑦">throw</a> a "<code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#dataerror" id="ref-for-dataerror⑧⑦">DataError</a></code>" <code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#idl-DOMException" id="ref-for-idl-DOMException①④⓪">DOMException</a></code>.</p>
+      <li data-md>
+       <p>If any of the following sub-steps fail, <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#dfn-throw" id="ref-for-dfn-throw②⑤⑧">throw</a> an "<code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#operationerror" id="ref-for-operationerror④④">OperationError</a></code>" <code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#idl-DOMException" id="ref-for-idl-DOMException①④①">DOMException</a></code>.</p>
+       <ol>
+        <li data-md>
+         <p>Let <var>output</var> be the result of <a data-link-type="dfn" href="#copy-an-mloperand" id="ref-for-copy-an-mloperand①②">copying an MLOperand</a> given <var>input</var>.</p>
+        <li data-md>
+         <p>Make a request to the underlying platform to:</p>
+         <ol>
+          <li data-md>
+           <p>Let <var>opImpl</var> be an <a data-link-type="dfn" href="https://infra.spec.whatwg.org/#implementation-defined" id="ref-for-implementation-defined⑥④">implementation-defined</a> platform operator for the softmax operation.</p>
+          <li data-md>
+           <p>Store a reference of <var>opImpl</var> in <var>output</var>.<code class="idl"><a data-link-type="idl" href="#dom-mloperand-operator-slot" id="ref-for-dom-mloperand-operator-slot③⓪">[[operator]]</a></code>.</p>
+          <li data-md>
+           <p>Create an <a data-link-type="dfn" href="https://infra.spec.whatwg.org/#implementation-defined" id="ref-for-implementation-defined⑥⑤">implementation-defined</a> platform operand <var>outputImpl</var> to represent the output, given <var>output</var> and <var>opImpl</var>.</p>
+          <li data-md>
+           <p>Store a reference to <var>outputImpl</var> in <var>output</var>.<code class="idl"><a data-link-type="idl" href="#dom-mloperand-operand-slot" id="ref-for-dom-mloperand-operand-slot⑧⑥">[[operand]]</a></code>.</p>
+         </ol>
+        <li data-md>
+         <p>Connect <var>input</var>.<code class="idl"><a data-link-type="idl" href="#dom-mloperand-operand-slot" id="ref-for-dom-mloperand-operand-slot⑧⑦">[[operand]]</a></code> as input to <var>opImpl</var>.</p>
+        <li data-md>
+         <p>Connect <var>output</var>.<code class="idl"><a data-link-type="idl" href="#dom-mloperand-operand-slot" id="ref-for-dom-mloperand-operand-slot⑧⑧">[[operand]]</a></code> as output to <var>opImpl</var>.</p>
+       </ol>
+      <li data-md>
+       <p>Return <var>output</var>.</p>
+     </ol>
+    </div>
+   </details>
+   <h5 class="heading settled" data-level="7.6.33.2" id="api-mlgraphbuilder-softmax"><span class="secno">7.6.33.2. </span><span class="content">The <code class="idl"><a data-link-type="idl" href="#softmax-noargs" id="ref-for-softmax-noargs②">softmax()</a></code> method</span><a class="self-link" href="#api-mlgraphbuilder-softmax"></a></h5>
+   <div>
+     <strong>Arguments:</strong> 
+    <ul>
+     <li data-md>
+      <p>None.</p>
+    </ul>
+    <p><strong>Returns:</strong></p>
+    <ul>
+     <li data-md>
+      <p>an <code class="idl"><a data-link-type="idl" href="#mlactivation" id="ref-for-mlactivation⑤⓪">MLActivation</a></code>. The activation function representing the softmax operation.</p>
+    </ul>
+   </div>
+   <details class="algorithm" data-algorithm="softmax()" data-algorithm-for="MLGraphBuilder" open>
+    <summary> The <dfn class="dfn-paneled idl-code" data-dfn-for="MLGraphBuilder" data-dfn-type="method" data-export id="softmax-noargs"><code>softmax()</code></dfn> method steps are: </summary>
+    <div class="algorithm-steps">
+     <ol>
+      <li data-md>
+       <p>Let <var>op</var> be the result of <a data-link-type="dfn" href="#create-an-mlactivation" id="ref-for-create-an-mlactivation⑧">creating an MLActivation</a> given  and <code>"softmax"</code>.</p>
+       <ol>
+        <li data-md>
+         <p>If that <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#dfn-throw" id="ref-for-dfn-throw②⑤⑨">throws</a> an error, re-<a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#dfn-throw" id="ref-for-dfn-throw②⑥⓪">throw</a> the error.</p>
+       </ol>
+      <li data-md>
+       <p>Return <var>op</var>.</p>
+     </ol>
     </div>
-    <h4 class="heading settled" data-level="7.6.33" id="api-mlgraphbuilder-softplus"><span class="secno">7.6.33. </span><span class="content">The softplus() method</span><a class="self-link" href="#api-mlgraphbuilder-softplus"></a></h4>
-     Compute the <a href="https://en.wikipedia.org/wiki/Rectifier_(neural_networks)#Softplus">softplus function</a> of the input tensor. The calculation follows the expression <code>ln(1 + exp(steepness * x)) / steepness</code>. 
+   </details>
+   <h4 class="heading settled" data-level="7.6.34" id="api-mlgraphbuilder-softplus-method"><span class="secno">7.6.34. </span><span class="content">The softplus() method</span><a class="self-link" href="#api-mlgraphbuilder-softplus-method"></a></h4>
+    Compute the <a href="https://en.wikipedia.org/wiki/Rectifier_(neural_networks)#Softplus">softplus function</a> of the input tensor. The calculation follows the expression <code>ln(1 + exp(steepness * x)) / steepness</code>. 
 <pre class="idl highlight def"><c- b>dictionary</c-> <dfn class="dfn-paneled idl-code" data-dfn-type="dictionary" data-export id="dictdef-mlsoftplusoptions"><code><c- g>MLSoftplusOptions</c-></code></dfn> {
-  <a class="idl-code" data-link-type="interface" href="https://webidl.spec.whatwg.org/#idl-float" id="ref-for-idl-float②⑧"><c- b>float</c-></a> <dfn class="idl-code" data-default="1" data-dfn-for="MLSoftplusOptions" data-dfn-type="dict-member" data-export data-type="float " id="dom-mlsoftplusoptions-steepness"><code><c- g>steepness</c-></code><a class="self-link" href="#dom-mlsoftplusoptions-steepness"></a></dfn> = 1;
+  <a class="idl-code" data-link-type="interface" href="https://webidl.spec.whatwg.org/#idl-float" id="ref-for-idl-float④②"><c- b>float</c-></a> <a class="idl-code" data-default="1" data-link-type="dict-member" data-type="float " href="#dom-mlsoftplusoptions-steepness" id="ref-for-dom-mlsoftplusoptions-steepness"><c- g>steepness</c-></a> = 1;
 };
 
-<c- b>partial</c-> <c- b>interface</c-> <a class="idl-code" data-link-type="interface" href="#mlgraphbuilder" id="ref-for-mlgraphbuilder④⑨"><c- g>MLGraphBuilder</c-></a> {
-  <a data-link-type="idl-name" href="#mloperand" id="ref-for-mloperand②⑧①"><c- n>MLOperand</c-></a> <dfn class="idl-code" data-dfn-for="MLGraphBuilder" data-dfn-type="method" data-export data-lt="softplus(x, options)|softplus(x)" id="dom-mlgraphbuilder-softplus"><code><c- g>softplus</c-></code><a class="self-link" href="#dom-mlgraphbuilder-softplus"></a></dfn>(<a data-link-type="idl-name" href="#mloperand" id="ref-for-mloperand②⑧②"><c- n>MLOperand</c-></a> <dfn class="idl-code" data-dfn-for="MLGraphBuilder/softplus(x, options), MLGraphBuilder/softplus(x)" data-dfn-type="argument" data-export id="dom-mlgraphbuilder-softplus-x-options-x"><code><c- g>x</c-></code><a class="self-link" href="#dom-mlgraphbuilder-softplus-x-options-x"></a></dfn>, <c- b>optional</c-> <a data-link-type="idl-name" href="#dictdef-mlsoftplusoptions" id="ref-for-dictdef-mlsoftplusoptions"><c- n>MLSoftplusOptions</c-></a> <dfn class="idl-code" data-dfn-for="MLGraphBuilder/softplus(x, options), MLGraphBuilder/softplus(x)" data-dfn-type="argument" data-export id="dom-mlgraphbuilder-softplus-x-options-options"><code><c- g>options</c-></code><a class="self-link" href="#dom-mlgraphbuilder-softplus-x-options-options"></a></dfn> = {});
-  <a data-link-type="idl-name" href="#mlactivation" id="ref-for-mlactivation③④"><c- n>MLActivation</c-></a> <dfn class="idl-code" data-dfn-for="MLGraphBuilder" data-dfn-type="method" data-export data-lt="softplus(options)|softplus()" id="dom-mlgraphbuilder-softplus-options"><code><c- g>softplus</c-></code><a class="self-link" href="#dom-mlgraphbuilder-softplus-options"></a></dfn>(<c- b>optional</c-> <a data-link-type="idl-name" href="#dictdef-mlsoftplusoptions" id="ref-for-dictdef-mlsoftplusoptions①"><c- n>MLSoftplusOptions</c-></a> <dfn class="idl-code" data-dfn-for="MLGraphBuilder/softplus(options), MLGraphBuilder/softplus()" data-dfn-type="argument" data-export id="dom-mlgraphbuilder-softplus-options-options"><code><c- g>options</c-></code><a class="self-link" href="#dom-mlgraphbuilder-softplus-options-options"></a></dfn> = {});
+<c- b>partial</c-> <c- b>interface</c-> <a class="idl-code" data-link-type="interface" href="#mlgraphbuilder" id="ref-for-mlgraphbuilder⑤②"><c- g>MLGraphBuilder</c-></a> {
+  <a data-link-type="idl-name" href="#mloperand" id="ref-for-mloperand③④⑥"><c- n>MLOperand</c-></a> <a class="idl-code" data-link-type="method" href="#dom-mlgraphbuilder-softplus" id="ref-for-dom-mlgraphbuilder-softplus"><c- g>softplus</c-></a>(<a data-link-type="idl-name" href="#mloperand" id="ref-for-mloperand③④⑦"><c- n>MLOperand</c-></a> <dfn class="idl-code" data-dfn-for="MLGraphBuilder/softplus(input, options), MLGraphBuilder/softplus(input)" data-dfn-type="argument" data-export id="dom-mlgraphbuilder-softplus-input-options-input"><code><c- g>input</c-></code><a class="self-link" href="#dom-mlgraphbuilder-softplus-input-options-input"></a></dfn>, <c- b>optional</c-> <a data-link-type="idl-name" href="#dictdef-mlsoftplusoptions" id="ref-for-dictdef-mlsoftplusoptions"><c- n>MLSoftplusOptions</c-></a> <dfn class="idl-code" data-dfn-for="MLGraphBuilder/softplus(input, options), MLGraphBuilder/softplus(input)" data-dfn-type="argument" data-export id="dom-mlgraphbuilder-softplus-input-options-options"><code><c- g>options</c-></code><a class="self-link" href="#dom-mlgraphbuilder-softplus-input-options-options"></a></dfn> = {});
+  <a data-link-type="idl-name" href="#mlactivation" id="ref-for-mlactivation⑤①"><c- n>MLActivation</c-></a> <a class="idl-code" data-link-type="method" href="#dom-mlgraphbuilder-softplus-options" id="ref-for-dom-mlgraphbuilder-softplus-options"><c- g>softplus</c-></a>(<c- b>optional</c-> <a data-link-type="idl-name" href="#dictdef-mlsoftplusoptions" id="ref-for-dictdef-mlsoftplusoptions①"><c- n>MLSoftplusOptions</c-></a> <dfn class="idl-code" data-dfn-for="MLGraphBuilder/softplus(options), MLGraphBuilder/softplus()" data-dfn-type="argument" data-export id="dom-mlgraphbuilder-softplus-options-options"><code><c- g>options</c-></code><a class="self-link" href="#dom-mlgraphbuilder-softplus-options-options"></a></dfn> = {});
 };
 </pre>
-    <div class="algorithm" data-algorithm="softplus">
-      <strong>Arguments:</strong> 
-     <ul>
-      <li data-md>
-       <p><em>x</em>: an <code class="idl"><a data-link-type="idl" href="#mloperand" id="ref-for-mloperand②⑧③">MLOperand</a></code>. The input tensor.</p>
-     </ul>
-     <p><strong>Returns:</strong></p>
-     <ul>
-      <li data-md>
-       <p>an <code class="idl"><a data-link-type="idl" href="#mloperand" id="ref-for-mloperand②⑧④">MLOperand</a></code>. The output tensor of the same shape as <em>x</em>.</p>
-      <li data-md>
-       <p>an <code class="idl"><a data-link-type="idl" href="#mlactivation" id="ref-for-mlactivation③⑤">MLActivation</a></code>. The activation function representing the softplus operation.</p>
-     </ul>
-     <div class="note" role="note">
-       The behavior of this operation can be generically emulated from the usage of
+   <div class="note" role="note">
+    <details open>
+     <summary> The behavior of this operation can be generically emulated from the usage of
     other operations as follow. However, user agents typically have a more
     efficient implementation for it, therefore its usage is encouraged from the
-    performance standpoint. 
+    performance standpoint. </summary>
 <pre class="highlight"><c- k>return</c-> builder<c- p>.</c->div<c- p>(</c->
           builder<c- p>.</c->log<c- p>(</c->
             builder<c- p>.</c->add<c- p>(</c->
@@ -3994,69 +7011,254 @@ <h4 class="heading settled" data-level="7.6.33" id="api-mlgraphbuilder-softplus"
               builder<c- p>.</c->constant<c- p>(</c-><c- mf>1</c-><c- p>))),</c->
           builder<c- p>.</c->constant<c- p>(</c->options<c- p>.</c->steepness<c- p>));</c->
 </pre>
-     </div>
+    </details>
+   </div>
+   <p><code class="idl"><a data-link-type="idl" href="#dictdef-mlsoftplusoptions" id="ref-for-dictdef-mlsoftplusoptions②">MLSoftplusOptions</a></code> has the following members:</p>
+   <dl>
+    <dt data-md><dfn class="dfn-paneled idl-code" data-dfn-for="MLSoftplusOptions" data-dfn-type="dict-member" data-export id="dom-mlsoftplusoptions-steepness"><code>steepness</code></dfn>, <span> of type <a data-link-type="idl-name" href="https://webidl.spec.whatwg.org/#idl-float" id="ref-for-idl-float④③">float</a>, defaulting to <code>1</code></span>
+    <dd data-md>
+     <p>A <code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#idl-float" id="ref-for-idl-float④④">float</a></code> scalar parameter.
+ The default value is 1.</p>
+   </dl>
+   <h5 class="heading settled" data-level="7.6.34.1" id="api-mlgraphbuilder-softplus-input-options"><span class="secno">7.6.34.1. </span><span class="content">The <code class="idl"><a data-link-type="idl" href="#dom-mlgraphbuilder-softplus" id="ref-for-dom-mlgraphbuilder-softplus①">softplus(input, options)</a></code> method</span><a class="self-link" href="#api-mlgraphbuilder-softplus-input-options"></a></h5>
+   <div>
+     <strong>Arguments:</strong> 
+    <ul>
+     <li data-md>
+      <p><em>input</em>: an <code class="idl"><a data-link-type="idl" href="#mloperand" id="ref-for-mloperand③④⑧">MLOperand</a></code>. The input tensor.</p>
+     <li data-md>
+      <p><em>options</em>: an optional <code class="idl"><a data-link-type="idl" href="#dictdef-mlsoftplusoptions" id="ref-for-dictdef-mlsoftplusoptions③">MLSoftplusOptions</a></code>. The optional parameters of the operation.</p>
+    </ul>
+    <p><strong>Returns:</strong></p>
+    <ul>
+     <li data-md>
+      <p>an <code class="idl"><a data-link-type="idl" href="#mloperand" id="ref-for-mloperand③④⑨">MLOperand</a></code>. The output tensor of the same shape as <em>input</em>.</p>
+    </ul>
+   </div>
+   <details class="algorithm" data-algorithm="softplus(input, options)" data-algorithm-for="MLGraphBuilder" open>
+    <summary> The <dfn class="dfn-paneled idl-code" data-dfn-for="MLGraphBuilder" data-dfn-type="method" data-export data-lt="softplus(input, options)|softplus(input)" id="dom-mlgraphbuilder-softplus"><code>softplus(<var>input</var>, <var>options</var>)</code></dfn> method steps are: </summary>
+    <div class="algorithm-steps">
+     <ol>
+      <li data-md>
+       <p>If any of the following sub-steps fail, <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#dfn-throw" id="ref-for-dfn-throw②⑥①">throw</a> an "<code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#operationerror" id="ref-for-operationerror④⑤">OperationError</a></code>" <code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#idl-DOMException" id="ref-for-idl-DOMException①④②">DOMException</a></code>.</p>
+       <ol>
+        <li data-md>
+         <p>Let <var>output</var> be the result of <a data-link-type="dfn" href="#copy-an-mloperand" id="ref-for-copy-an-mloperand①③">copying an MLOperand</a> given <var>input</var>.</p>
+        <li data-md>
+         <p>Make a request to the underlying platform to:</p>
+         <ol>
+          <li data-md>
+           <p>Let <var>opImpl</var> be an <a data-link-type="dfn" href="https://infra.spec.whatwg.org/#implementation-defined" id="ref-for-implementation-defined⑥⑥">implementation-defined</a> platform operator for the softplus operation, given <var>options</var>.</p>
+          <li data-md>
+           <p>Store a reference of <var>opImpl</var> in <var>output</var>.<code class="idl"><a data-link-type="idl" href="#dom-mloperand-operator-slot" id="ref-for-dom-mloperand-operator-slot③①">[[operator]]</a></code>.</p>
+          <li data-md>
+           <p>Create an <a data-link-type="dfn" href="https://infra.spec.whatwg.org/#implementation-defined" id="ref-for-implementation-defined⑥⑦">implementation-defined</a> platform operand <var>outputImpl</var> to represent the output, given <var>output</var> and <var>opImpl</var>.</p>
+          <li data-md>
+           <p>Store a reference to <var>outputImpl</var> in <var>output</var>.<code class="idl"><a data-link-type="idl" href="#dom-mloperand-operand-slot" id="ref-for-dom-mloperand-operand-slot⑧⑨">[[operand]]</a></code>.</p>
+         </ol>
+        <li data-md>
+         <p>Connect <var>input</var>.<code class="idl"><a data-link-type="idl" href="#dom-mloperand-operand-slot" id="ref-for-dom-mloperand-operand-slot⑨⓪">[[operand]]</a></code> as input to <var>opImpl</var>.</p>
+        <li data-md>
+         <p>Connect <var>output</var>.<code class="idl"><a data-link-type="idl" href="#dom-mloperand-operand-slot" id="ref-for-dom-mloperand-operand-slot⑨①">[[operand]]</a></code> as output to <var>opImpl</var>.</p>
+       </ol>
+      <li data-md>
+       <p>Return <var>output</var>.</p>
+     </ol>
+    </div>
+   </details>
+   <h5 class="heading settled" data-level="7.6.34.2" id="api-mlgraphbuilder-softplus-options"><span class="secno">7.6.34.2. </span><span class="content">The <code class="idl"><a data-link-type="idl" href="#dom-mlgraphbuilder-softplus-options" id="ref-for-dom-mlgraphbuilder-softplus-options①">softplus(options)</a></code> method</span><a class="self-link" href="#api-mlgraphbuilder-softplus-options"></a></h5>
+   <div>
+     <strong>Arguments:</strong> 
+    <ul>
+     <li data-md>
+      <p><em>options</em>: an optional <code class="idl"><a data-link-type="idl" href="#dictdef-mlsoftplusoptions" id="ref-for-dictdef-mlsoftplusoptions④">MLSoftplusOptions</a></code>. The optional parameters of the operation.</p>
+    </ul>
+    <p><strong>Returns:</strong></p>
+    <ul>
+     <li data-md>
+      <p>an <code class="idl"><a data-link-type="idl" href="#mlactivation" id="ref-for-mlactivation⑤②">MLActivation</a></code>. The activation function representing the softplus operation.</p>
+    </ul>
+   </div>
+   <details class="algorithm" data-algorithm="softplus(options)" data-algorithm-for="MLGraphBuilder" open>
+    <summary> The <dfn class="dfn-paneled idl-code" data-dfn-for="MLGraphBuilder" data-dfn-type="method" data-export data-lt="softplus(options)|softplus()" id="dom-mlgraphbuilder-softplus-options"><code>softplus(<var>options</var>)</code></dfn> method steps are: </summary>
+    <div class="algorithm-steps">
+     <ol>
+      <li data-md>
+       <p>Let <var>op</var> be the result of <a data-link-type="dfn" href="#create-an-mlactivation" id="ref-for-create-an-mlactivation⑨">creating an MLActivation</a> given <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#this" id="ref-for-this③⑦">this</a>, <code>"softplus"</code> and <var>options</var>.</p>
+       <ol>
+        <li data-md>
+         <p>If that <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#dfn-throw" id="ref-for-dfn-throw②⑥②">throws</a> an error, re-<a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#dfn-throw" id="ref-for-dfn-throw②⑥③">throw</a> the error.</p>
+       </ol>
+      <li data-md>
+       <p>Return <var>op</var>.</p>
+     </ol>
     </div>
-    <h4 class="heading settled" data-level="7.6.34" id="api-mlgraphbuilder-softsign"><span class="secno">7.6.34. </span><span class="content">The softsign() method</span><a class="self-link" href="#api-mlgraphbuilder-softsign"></a></h4>
-     Compute the <a href="https://pytorch.org/docs/stable/generated/torch.nn.Softsign.html">softsign function</a> of the input tensor. The calculation follows the expression <code>x / (1 + |x|)</code>. 
-<pre class="idl highlight def"><c- b>partial</c-> <c- b>interface</c-> <a class="idl-code" data-link-type="interface" href="#mlgraphbuilder" id="ref-for-mlgraphbuilder⑤⓪"><c- g>MLGraphBuilder</c-></a> {
-  <a data-link-type="idl-name" href="#mloperand" id="ref-for-mloperand②⑧⑤"><c- n>MLOperand</c-></a> <dfn class="idl-code" data-dfn-for="MLGraphBuilder" data-dfn-type="method" data-export data-lt="softsign(x)" id="dom-mlgraphbuilder-softsign"><code><c- g>softsign</c-></code><a class="self-link" href="#dom-mlgraphbuilder-softsign"></a></dfn>(<a data-link-type="idl-name" href="#mloperand" id="ref-for-mloperand②⑧⑥"><c- n>MLOperand</c-></a> <dfn class="idl-code" data-dfn-for="MLGraphBuilder/softsign(x)" data-dfn-type="argument" data-export id="dom-mlgraphbuilder-softsign-x-x"><code><c- g>x</c-></code><a class="self-link" href="#dom-mlgraphbuilder-softsign-x-x"></a></dfn>);
-  <a data-link-type="idl-name" href="#mlactivation" id="ref-for-mlactivation③⑥"><c- n>MLActivation</c-></a> <dfn class="idl-code" data-dfn-for="MLGraphBuilder" data-dfn-type="method" data-export data-lt="softsign()" id="dom-mlgraphbuilder-softsign①"><code><c- g>softsign</c-></code><a class="self-link" href="#dom-mlgraphbuilder-softsign①"></a></dfn>();
+   </details>
+   <h4 class="heading settled" data-level="7.6.35" id="api-mlgraphbuilder-softsign-method"><span class="secno">7.6.35. </span><span class="content">The softsign() method</span><a class="self-link" href="#api-mlgraphbuilder-softsign-method"></a></h4>
+    Compute the <a href="https://pytorch.org/docs/stable/generated/torch.nn.Softsign.html">softsign function</a> of the input tensor. The calculation follows the expression <code>x / (1 + |x|)</code>. 
+<pre class="idl highlight def"><c- b>partial</c-> <c- b>interface</c-> <a class="idl-code" data-link-type="interface" href="#mlgraphbuilder" id="ref-for-mlgraphbuilder⑤③"><c- g>MLGraphBuilder</c-></a> {
+  <a data-link-type="idl-name" href="#mloperand" id="ref-for-mloperand③⑤⓪"><c- n>MLOperand</c-></a> <a class="idl-code" data-link-type="method" href="#dom-mlgraphbuilder-softsign" id="ref-for-dom-mlgraphbuilder-softsign"><c- g>softsign</c-></a>(<a data-link-type="idl-name" href="#mloperand" id="ref-for-mloperand③⑤①"><c- n>MLOperand</c-></a> <dfn class="idl-code" data-dfn-for="MLGraphBuilder/softsign(input)" data-dfn-type="argument" data-export id="dom-mlgraphbuilder-softsign-input-input"><code><c- g>input</c-></code><a class="self-link" href="#dom-mlgraphbuilder-softsign-input-input"></a></dfn>);
+  <a data-link-type="idl-name" href="#mlactivation" id="ref-for-mlactivation⑤③"><c- n>MLActivation</c-></a> <a class="idl-code" data-link-type="method" href="#softsign-noargs" id="ref-for-softsign-noargs"><c- g>softsign</c-></a>();
 };
 </pre>
-    <div class="algorithm" data-algorithm="softsign">
-      <strong>Arguments:</strong> 
-     <ul>
-      <li data-md>
-       <p><em>x</em>: an <code class="idl"><a data-link-type="idl" href="#mloperand" id="ref-for-mloperand②⑧⑦">MLOperand</a></code>. The input tensor.</p>
-     </ul>
-     <p><strong>Returns:</strong></p>
-     <ul>
-      <li data-md>
-       <p>an <code class="idl"><a data-link-type="idl" href="#mloperand" id="ref-for-mloperand②⑧⑧">MLOperand</a></code>. The output tensor of the same shape as <em>x</em>.</p>
-      <li data-md>
-       <p>an <code class="idl"><a data-link-type="idl" href="#mlactivation" id="ref-for-mlactivation③⑦">MLActivation</a></code>. The activation function representing the softsign operation.</p>
-     </ul>
-     <div class="note" role="note">
-       The behavior of this operation can be generically emulated from the usage of
+   <div class="note" role="note">
+    <details open>
+     <summary> The behavior of this operation can be generically emulated from the usage of
     other operations as follow. However, user agents typically have a more
     efficient implementation for it, therefore its usage is encouraged from the
-    performance standpoint. 
+    performance standpoint. </summary>
 <pre class="highlight"><c- k>return</c-> builder<c- p>.</c->div<c- p>(</c->x<c- p>,</c-> builder<c- p>.</c->add<c- p>(</c->builder<c- p>.</c->constant<c- p>(</c-><c- mf>1</c-><c- p>),</c-> builder<c- p>.</c->abs<c- p>(</c->x<c- p>)));</c->
 </pre>
-     </div>
+    </details>
+   </div>
+   <h5 class="heading settled" data-level="7.6.35.1" id="api-mlgraphbuilder-softsign-input"><span class="secno">7.6.35.1. </span><span class="content">The <code class="idl"><a data-link-type="idl" href="#dom-mlgraphbuilder-softsign" id="ref-for-dom-mlgraphbuilder-softsign①">softsign(input)</a></code> method</span><a class="self-link" href="#api-mlgraphbuilder-softsign-input"></a></h5>
+   <div>
+     <strong>Arguments:</strong> 
+    <ul>
+     <li data-md>
+      <p><em>input</em>: an <code class="idl"><a data-link-type="idl" href="#mloperand" id="ref-for-mloperand③⑤②">MLOperand</a></code>. The input tensor.</p>
+    </ul>
+    <p><strong>Returns:</strong></p>
+    <ul>
+     <li data-md>
+      <p>an <code class="idl"><a data-link-type="idl" href="#mloperand" id="ref-for-mloperand③⑤③">MLOperand</a></code>. The output tensor of the same shape as <em>input</em>.</p>
+    </ul>
+   </div>
+   <details class="algorithm" data-algorithm="softsign(input)" data-algorithm-for="MLGraphBuilder" open>
+    <summary> The <dfn class="dfn-paneled idl-code" data-dfn-for="MLGraphBuilder" data-dfn-type="method" data-export id="dom-mlgraphbuilder-softsign"><code>softsign(<var>input</var>)</code></dfn> method steps are: </summary>
+    <div class="algorithm-steps">
+     <ol>
+      <li data-md>
+       <p><a data-link-type="dfn" href="https://infra.spec.whatwg.org/#assert" id="ref-for-assert⑥⑧">Assert</a>: the type of <var>input</var> is <code class="idl"><a data-link-type="idl" href="#mloperand" id="ref-for-mloperand③⑤④">MLOperand</a></code>.</p>
+      <li data-md>
+       <p>If any of the following sub-steps fail, <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#dfn-throw" id="ref-for-dfn-throw②⑥④">throw</a> an "<code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#operationerror" id="ref-for-operationerror④⑥">OperationError</a></code>" <code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#idl-DOMException" id="ref-for-idl-DOMException①④③">DOMException</a></code>.</p>
+       <ol>
+        <li data-md>
+         <p>Let <var>output</var> be the result of <a data-link-type="dfn" href="#copy-an-mloperand" id="ref-for-copy-an-mloperand①④">copying an MLOperand</a> given <var>input</var>.</p>
+        <li data-md>
+         <p>Make a request to the underlying platform to:</p>
+         <ol>
+          <li data-md>
+           <p>Let <var>opImpl</var> be an <a data-link-type="dfn" href="https://infra.spec.whatwg.org/#implementation-defined" id="ref-for-implementation-defined⑥⑧">implementation-defined</a> platform operator for the softsign operation.</p>
+          <li data-md>
+           <p>Store a reference of <var>opImpl</var> in <var>output</var>.<code class="idl"><a data-link-type="idl" href="#dom-mloperand-operator-slot" id="ref-for-dom-mloperand-operator-slot③②">[[operator]]</a></code>.</p>
+          <li data-md>
+           <p>Create an <a data-link-type="dfn" href="https://infra.spec.whatwg.org/#implementation-defined" id="ref-for-implementation-defined⑥⑨">implementation-defined</a> platform operand <var>outputImpl</var> to represent the output, given <var>output</var> and <var>opImpl</var>.</p>
+          <li data-md>
+           <p>Store a reference to <var>outputImpl</var> in <var>output</var>.<code class="idl"><a data-link-type="idl" href="#dom-mloperand-operand-slot" id="ref-for-dom-mloperand-operand-slot⑨②">[[operand]]</a></code>.</p>
+         </ol>
+        <li data-md>
+         <p>Connect <var>input</var>.<code class="idl"><a data-link-type="idl" href="#dom-mloperand-operand-slot" id="ref-for-dom-mloperand-operand-slot⑨③">[[operand]]</a></code> as input to <var>opImpl</var>.</p>
+        <li data-md>
+         <p>Connect <var>output</var>.<code class="idl"><a data-link-type="idl" href="#dom-mloperand-operand-slot" id="ref-for-dom-mloperand-operand-slot⑨④">[[operand]]</a></code> as output to <var>opImpl</var>.</p>
+       </ol>
+      <li data-md>
+       <p>Return <var>output</var>.</p>
+     </ol>
+    </div>
+   </details>
+   <h5 class="heading settled" data-level="7.6.35.2" id="api-mlgraphbuilder-softsign"><span class="secno">7.6.35.2. </span><span class="content">The <code class="idl"><a data-link-type="idl" href="#softsign-noargs" id="ref-for-softsign-noargs①">softsign()</a></code> method</span><a class="self-link" href="#api-mlgraphbuilder-softsign"></a></h5>
+   <div>
+     <strong>Arguments:</strong> 
+    <ul>
+     <li data-md>
+      <p>None.</p>
+    </ul>
+    <p><strong>Returns:</strong></p>
+    <ul>
+     <li data-md>
+      <p>an <code class="idl"><a data-link-type="idl" href="#mlactivation" id="ref-for-mlactivation⑤④">MLActivation</a></code>. The activation function representing the softsign operation.</p>
+    </ul>
+   </div>
+   <details class="algorithm" data-algorithm="softsign()" data-algorithm-for="MLGraphBuilder" open>
+    <summary> The <dfn class="dfn-paneled idl-code" data-dfn-for="MLGraphBuilder" data-dfn-type="method" data-export id="softsign-noargs"><code>softsign()</code></dfn> method steps are: </summary>
+    <div class="algorithm-steps">
+     <ol>
+      <li data-md>
+       <p>Let <var>op</var> be the result of <a data-link-type="dfn" href="#create-an-mlactivation" id="ref-for-create-an-mlactivation①⓪">creating an MLActivation</a> given <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#this" id="ref-for-this③⑧">this</a> and <code>"softsign"</code>.</p>
+       <ol>
+        <li data-md>
+         <p>If that <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#dfn-throw" id="ref-for-dfn-throw②⑥⑤">throws</a> an error, re-<a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#dfn-throw" id="ref-for-dfn-throw②⑥⑥">throw</a> the error.</p>
+       </ol>
+      <li data-md>
+       <p>Return <var>op</var>.</p>
+     </ol>
     </div>
-    <h4 class="heading settled" data-level="7.6.35" id="api-mlgraphbuilder-split"><span class="secno">7.6.35. </span><span class="content">The split() method</span><a class="self-link" href="#api-mlgraphbuilder-split"></a></h4>
-     Split the input tensor into a number of sub tensors along the given axis. 
+   </details>
+   <h4 class="heading settled" data-level="7.6.36" id="api-mlgraphbuilder-split"><span class="secno">7.6.36. </span><span class="content">The split() method</span><a class="self-link" href="#api-mlgraphbuilder-split"></a></h4>
+    Split the input tensor into a number of sub tensors along the given axis. 
 <pre class="idl highlight def"><c- b>dictionary</c-> <dfn class="dfn-paneled idl-code" data-dfn-type="dictionary" data-export id="dictdef-mlsplitoptions"><code><c- g>MLSplitOptions</c-></code></dfn> {
-  <a class="idl-code" data-link-type="interface" href="https://webidl.spec.whatwg.org/#idl-unsigned-long" id="ref-for-idl-unsigned-long⑥③"><c- b>unsigned</c-> <c- b>long</c-></a> <dfn class="idl-code" data-default="0" data-dfn-for="MLSplitOptions" data-dfn-type="dict-member" data-export data-type="unsigned long " id="dom-mlsplitoptions-axis"><code><c- g>axis</c-></code><a class="self-link" href="#dom-mlsplitoptions-axis"></a></dfn> = 0;
+  <a class="idl-code" data-link-type="interface" href="https://webidl.spec.whatwg.org/#idl-unsigned-long" id="ref-for-idl-unsigned-long⑥⑧"><c- b>unsigned</c-> <c- b>long</c-></a> <a class="idl-code" data-default="0" data-link-type="dict-member" data-type="unsigned long " href="#dom-mlsplitoptions-axis" id="ref-for-dom-mlsplitoptions-axis"><c- g>axis</c-></a> = 0;
 };
 
-<c- b>partial</c-> <c- b>interface</c-> <a class="idl-code" data-link-type="interface" href="#mlgraphbuilder" id="ref-for-mlgraphbuilder⑤①"><c- g>MLGraphBuilder</c-></a> {
-  <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#idl-sequence" id="ref-for-idl-sequence③①"><c- b>sequence</c-></a>&lt;<a data-link-type="idl-name" href="#mloperand" id="ref-for-mloperand②⑧⑨"><c- n>MLOperand</c-></a>> <dfn class="idl-code" data-dfn-for="MLGraphBuilder" data-dfn-type="method" data-export data-lt="split(input, splits, options)|split(input, splits)" id="dom-mlgraphbuilder-split"><code><c- g>split</c-></code><a class="self-link" href="#dom-mlgraphbuilder-split"></a></dfn>(<a data-link-type="idl-name" href="#mloperand" id="ref-for-mloperand②⑨⓪"><c- n>MLOperand</c-></a> <dfn class="idl-code" data-dfn-for="MLGraphBuilder/split(input, splits, options), MLGraphBuilder/split(input, splits)" data-dfn-type="argument" data-export id="dom-mlgraphbuilder-split-input-splits-options-input"><code><c- g>input</c-></code><a class="self-link" href="#dom-mlgraphbuilder-split-input-splits-options-input"></a></dfn>,
-                          (<a class="idl-code" data-link-type="interface" href="https://webidl.spec.whatwg.org/#idl-unsigned-long" id="ref-for-idl-unsigned-long⑥④"><c- b>unsigned</c-> <c- b>long</c-></a> <c- b>or</c-> <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#idl-sequence" id="ref-for-idl-sequence③②"><c- b>sequence</c-></a>&lt;<a class="idl-code" data-link-type="interface" href="https://webidl.spec.whatwg.org/#idl-unsigned-long" id="ref-for-idl-unsigned-long⑥⑤"><c- b>unsigned</c-> <c- b>long</c-></a>>) <dfn class="idl-code" data-dfn-for="MLGraphBuilder/split(input, splits, options), MLGraphBuilder/split(input, splits)" data-dfn-type="argument" data-export id="dom-mlgraphbuilder-split-input-splits-options-splits"><code><c- g>splits</c-></code><a class="self-link" href="#dom-mlgraphbuilder-split-input-splits-options-splits"></a></dfn>,
+<c- b>partial</c-> <c- b>interface</c-> <a class="idl-code" data-link-type="interface" href="#mlgraphbuilder" id="ref-for-mlgraphbuilder⑤④"><c- g>MLGraphBuilder</c-></a> {
+  <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#idl-sequence" id="ref-for-idl-sequence③①"><c- b>sequence</c-></a>&lt;<a data-link-type="idl-name" href="#mloperand" id="ref-for-mloperand③⑤⑤"><c- n>MLOperand</c-></a>> <a class="idl-code" data-link-type="method" href="#dom-mlgraphbuilder-split" id="ref-for-dom-mlgraphbuilder-split"><c- g>split</c-></a>(<a data-link-type="idl-name" href="#mloperand" id="ref-for-mloperand③⑤⑥"><c- n>MLOperand</c-></a> <dfn class="idl-code" data-dfn-for="MLGraphBuilder/split(input, splits, options), MLGraphBuilder/split(input, splits)" data-dfn-type="argument" data-export id="dom-mlgraphbuilder-split-input-splits-options-input"><code><c- g>input</c-></code><a class="self-link" href="#dom-mlgraphbuilder-split-input-splits-options-input"></a></dfn>,
+                          (<a class="idl-code" data-link-type="interface" href="https://webidl.spec.whatwg.org/#idl-unsigned-long" id="ref-for-idl-unsigned-long⑥⑨"><c- b>unsigned</c-> <c- b>long</c-></a> <c- b>or</c-> <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#idl-sequence" id="ref-for-idl-sequence③②"><c- b>sequence</c-></a>&lt;<a class="idl-code" data-link-type="interface" href="https://webidl.spec.whatwg.org/#idl-unsigned-long" id="ref-for-idl-unsigned-long⑦⓪"><c- b>unsigned</c-> <c- b>long</c-></a>>) <dfn class="idl-code" data-dfn-for="MLGraphBuilder/split(input, splits, options), MLGraphBuilder/split(input, splits)" data-dfn-type="argument" data-export id="dom-mlgraphbuilder-split-input-splits-options-splits"><code><c- g>splits</c-></code><a class="self-link" href="#dom-mlgraphbuilder-split-input-splits-options-splits"></a></dfn>,
                           <c- b>optional</c-> <a data-link-type="idl-name" href="#dictdef-mlsplitoptions" id="ref-for-dictdef-mlsplitoptions"><c- n>MLSplitOptions</c-></a> <dfn class="idl-code" data-dfn-for="MLGraphBuilder/split(input, splits, options), MLGraphBuilder/split(input, splits)" data-dfn-type="argument" data-export id="dom-mlgraphbuilder-split-input-splits-options-options"><code><c- g>options</c-></code><a class="self-link" href="#dom-mlgraphbuilder-split-input-splits-options-options"></a></dfn> = {});
 };
 </pre>
-    <div class="algorithm" data-algorithm="split">
-      <strong>Arguments:</strong> 
-     <ul>
+   <div>
+     <strong>Arguments:</strong> 
+    <ul>
+     <li data-md>
+      <p><em>input</em>: an <code class="idl"><a data-link-type="idl" href="#mloperand" id="ref-for-mloperand③⑤⑦">MLOperand</a></code>. The input tensor.</p>
+     <li data-md>
+      <p><em>splits</em>: an <code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#idl-unsigned-long" id="ref-for-idl-unsigned-long⑦①">unsigned long</a></code> or a sequence of <code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#idl-unsigned-long" id="ref-for-idl-unsigned-long⑦②">unsigned long</a></code>. If an <code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#idl-unsigned-long" id="ref-for-idl-unsigned-long⑦③">unsigned long</a></code>, it specifies the number of output tensors along the axis. The number must evenly divide the dimension size of <em>input</em> along <em>options.axis</em>. If a sequence of <code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#idl-unsigned-long" id="ref-for-idl-unsigned-long⑦④">unsigned long</a></code>, it specifies the sizes of each output tensor along the <em>options.axis</em>. The sum of sizes must equal to the dimension size of <em>input</em> along <em>options.axis</em>.</p>
+     <li data-md>
+      <p><em>options</em>: an optional <code class="idl"><a data-link-type="idl" href="#dictdef-mlsplitoptions" id="ref-for-dictdef-mlsplitoptions①">MLSplitOptions</a></code>. The optional parameters of the operation.</p>
+    </ul>
+    <p><strong>Returns:</strong> a sequence of <code class="idl"><a data-link-type="idl" href="#mloperand" id="ref-for-mloperand③⑤⑧">MLOperand</a></code>. The splitted output tensors. If <em>splits</em> is an <code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#idl-unsigned-long" id="ref-for-idl-unsigned-long⑦⑤">unsigned long</a></code>, the <a data-link-type="dfn" href="https://infra.spec.whatwg.org/#list-size" id="ref-for-list-size⑤②">size</a> of the output sequence equals to <em>splits</em>. The shape of each output tensor is the same as <em>input</em> except the dimension size of <em>axis</em> equals to the quotient of dividing the dimension size of <em>input</em> along <em>axis</em> by <em>splits</em>. If <em>splits</em> is a sequence of <code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#idl-unsigned-long" id="ref-for-idl-unsigned-long⑦⑥">unsigned long</a></code>, the <a data-link-type="dfn" href="https://infra.spec.whatwg.org/#list-size" id="ref-for-list-size⑤③">size</a> of the output sequence equals to the <a data-link-type="dfn" href="https://infra.spec.whatwg.org/#list-size" id="ref-for-list-size⑤④">size</a> of <em>splits</em>. The shape of the i-th output tensor is the same as as <em>input</em> except along <em>axis</em> where the dimension size is <em>splits[i]</em>.</p>
+   </div>
+   <p><code class="idl"><a data-link-type="idl" href="#dictdef-mlsplitoptions" id="ref-for-dictdef-mlsplitoptions②">MLSplitOptions</a></code> has the following members:</p>
+   <dl>
+    <dt data-md><dfn class="dfn-paneled idl-code" data-dfn-for="MLSplitOptions" data-dfn-type="dict-member" data-export id="dom-mlsplitoptions-axis"><code>axis</code></dfn>, <span> of type <a data-link-type="idl-name" href="https://webidl.spec.whatwg.org/#idl-unsigned-long" id="ref-for-idl-unsigned-long⑦⑦">unsigned long</a>, defaulting to <code>0</code></span>
+    <dd data-md>
+     <p>An <code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#idl-unsigned-long" id="ref-for-idl-unsigned-long⑦⑧">unsigned long</a></code> scalar. The dimension along which to split. Its value must be in the range [0, N-1] where N is the <a data-link-type="dfn" href="#mloperand-rank" id="ref-for-mloperand-rank②④">rank</a> of the input tensor.
+The default value is 0.</p>
+   </dl>
+   <details class="algorithm" data-algorithm="split(input, splits, options)" data-algorithm-for="MLGraphBuilder" open>
+    <summary> The <dfn class="dfn-paneled idl-code" data-dfn-for="MLGraphBuilder" data-dfn-type="method" data-export data-lt="split(input, splits, options)|split(input, splits)" id="dom-mlgraphbuilder-split"><code>split(<var>input</var>, <var>splits</var>, <var>options</var>)</code></dfn> method steps are: </summary>
+    <div class="algorithm-steps">
+     <ol>
       <li data-md>
-       <p><em>input</em>: an <code class="idl"><a data-link-type="idl" href="#mloperand" id="ref-for-mloperand②⑨①">MLOperand</a></code>. The input tensor.</p>
+       <p><a data-link-type="dfn" href="https://infra.spec.whatwg.org/#assert" id="ref-for-assert⑥⑨">Assert</a>: the type of <var>input</var> is <code class="idl"><a data-link-type="idl" href="#mloperand" id="ref-for-mloperand③⑤⑨">MLOperand</a></code>.</p>
       <li data-md>
-       <p><em>splits</em>: an <code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#idl-unsigned-long" id="ref-for-idl-unsigned-long⑥⑥">unsigned long</a></code> or a sequence of <code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#idl-unsigned-long" id="ref-for-idl-unsigned-long⑥⑦">unsigned long</a></code>. If an <code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#idl-unsigned-long" id="ref-for-idl-unsigned-long⑥⑧">unsigned long</a></code>, it specifies the number of output tensors along the axis. The number must evenly divide the dimension size of <em>input</em> along <em>options.axis</em>. If a sequence of <code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#idl-unsigned-long" id="ref-for-idl-unsigned-long⑥⑨">unsigned long</a></code>, it specifies the sizes of each output tensor along the <em>options.axis</em>. The sum of sizes must equal to the dimension size of <em>input</em> along <em>options.axis</em>.</p>
+       <p>If <var>splits</var> is an <code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#idl-unsigned-long" id="ref-for-idl-unsigned-long⑦⑨">unsigned long</a></code>, and <var>input</var>.<code class="idl"><a data-link-type="idl" href="#dom-mloperanddescriptor-dimensions" id="ref-for-dom-mloperanddescriptor-dimensions⑨①">dimensions</a></code>[<var>options</var>.<code class="idl"><a data-link-type="idl" href="#dom-mlsplitoptions-axis" id="ref-for-dom-mlsplitoptions-axis①">axis</a></code>] % <var>splits</var> is not 0, then <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#dfn-throw" id="ref-for-dfn-throw②⑥⑦">throw</a> a <code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#exceptiondef-typeerror" id="ref-for-exceptiondef-typeerror④②">TypeError</a></code>.</p>
       <li data-md>
-       <p><em>options</em>: an optional <code class="idl"><a data-link-type="idl" href="#dictdef-mlsplitoptions" id="ref-for-dictdef-mlsplitoptions①">MLSplitOptions</a></code>. The optional parameters of the operation.</p>
-       <ul>
+       <p>If <var>splits</var> is a sequence of <code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#idl-unsigned-long" id="ref-for-idl-unsigned-long⑧⓪">unsigned long</a></code>, and the sum of its elements is not equal to <var>input</var>.<code class="idl"><a data-link-type="idl" href="#dom-mloperanddescriptor-dimensions" id="ref-for-dom-mloperanddescriptor-dimensions⑨②">dimensions</a></code>[<var>options</var>.<code class="idl"><a data-link-type="idl" href="#dom-mlsplitoptions-axis" id="ref-for-dom-mlsplitoptions-axis②">axis</a></code>], then <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#dfn-throw" id="ref-for-dfn-throw②⑥⑧">throw</a> a <code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#exceptiondef-typeerror" id="ref-for-exceptiondef-typeerror④③">TypeError</a></code>.</p>
+      <li data-md>
+       <p>If any of the following sub-steps fail, <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#dfn-throw" id="ref-for-dfn-throw②⑥⑨">throw</a> an "<code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#operationerror" id="ref-for-operationerror④⑦">OperationError</a></code>" <code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#idl-DOMException" id="ref-for-idl-DOMException①④④">DOMException</a></code>.</p>
+       <ol>
         <li data-md>
-         <p><em>axis</em>: an <code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#idl-unsigned-long" id="ref-for-idl-unsigned-long⑦⓪">unsigned long</a></code> scalar. The dimension along which to split. Its value must be in the range [0, N-1] where N is the rank of input tensor. Default to 0.</p>
-       </ul>
-     </ul>
-     <p><strong>Returns:</strong> a sequence of <code class="idl"><a data-link-type="idl" href="#mloperand" id="ref-for-mloperand②⑨②">MLOperand</a></code>. The splitted output tensors. If <em>splits</em> is an <code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#idl-unsigned-long" id="ref-for-idl-unsigned-long⑦①">unsigned long</a></code>, the length of the output sequence equals to <em>splits</em>. The shape of each output tensor is the same as <em>input</em> except the dimension size of <em>axis</em> equals to the quotient of dividing the dimension size of <em>input</em> along <em>axis</em> by <em>splits</em>. If <em>splits</em> is a sequence of <code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#idl-unsigned-long" id="ref-for-idl-unsigned-long⑦②">unsigned long</a></code>, the length of the output sequence equals to the length of <em>splits</em>. The shape of the i-th output tensor is the same as as <em>input</em> except along <em>axis</em> where the dimension size is <em>splits[i]</em>.</p>
-     <div class="note" role="note">
-       The behavior of this operation can be generically emulated from the usage of
+         <p>Let <var>output</var> be the result of <a data-link-type="dfn" href="#copy-an-mloperand" id="ref-for-copy-an-mloperand①⑤">copying an MLOperand</a> given <var>input</var>.</p>
+        <li data-md>
+         <p>Make a request to the underlying platform to:</p>
+         <ol>
+          <li data-md>
+           <p>Let <var>opImpl</var> be an <a data-link-type="dfn" href="https://infra.spec.whatwg.org/#implementation-defined" id="ref-for-implementation-defined⑦⓪">implementation-defined</a> platform operator for the split operation, given <var>splits</var> and <var>options</var>.</p>
+          <li data-md>
+           <p>Store a reference of <var>opImpl</var> in <var>output</var>.<code class="idl"><a data-link-type="idl" href="#dom-mloperand-operator-slot" id="ref-for-dom-mloperand-operator-slot③③">[[operator]]</a></code>.</p>
+          <li data-md>
+           <p>Create an <a data-link-type="dfn" href="https://infra.spec.whatwg.org/#implementation-defined" id="ref-for-implementation-defined⑦①">implementation-defined</a> platform operand <var>outputImpl</var> to represent the output, given <var>output</var> and <var>opImpl</var>.</p>
+          <li data-md>
+           <p>Store a reference to <var>outputImpl</var> in <var>output</var>.<code class="idl"><a data-link-type="idl" href="#dom-mloperand-operand-slot" id="ref-for-dom-mloperand-operand-slot⑨⑤">[[operand]]</a></code>.</p>
+         </ol>
+        <li data-md>
+         <p>Connect <var>input</var>.<code class="idl"><a data-link-type="idl" href="#dom-mloperand-operand-slot" id="ref-for-dom-mloperand-operand-slot⑨⑥">[[operand]]</a></code> as input to <var>opImpl</var>.</p>
+        <li data-md>
+         <p>Connect <var>output</var>.<code class="idl"><a data-link-type="idl" href="#dom-mloperand-operand-slot" id="ref-for-dom-mloperand-operand-slot⑨⑦">[[operand]]</a></code> as output to <var>opImpl</var>.</p>
+       </ol>
+      <li data-md>
+       <p>Return <var>output</var>.</p>
+     </ol>
+    </div>
+   </details>
+   <div class="note" role="note">
+    <details open>
+     <summary> The behavior of this operation can be generically emulated from the usage of
     other operations as follow. However, user agents typically have a more
     efficient implementation for it, therefore its usage is encouraged from the
-    performance standpoint. 
+    performance standpoint. </summary>
 <pre class="highlight"><c- c1>// This sample shows the case that the splits parameter is an array.</c->
 <c- a>const</c-> outputs <c- o>=</c-> <c- p>[];</c->
 <c- a>let</c-> starts <c- o>=</c-> Array<c- p>(</c->input_rank<c- p>).</c->fill<c- p>(</c-><c- mf>0</c-><c- p>);</c->
@@ -4070,103 +7272,272 @@ <h4 class="heading settled" data-level="7.6.35" id="api-mlgraphbuilder-split"><s
 <c- p>}</c->
 <c- k>return</c-> outputs<c- p>;</c->
 </pre>
-     </div>
-    </div>
-    <h4 class="heading settled" data-level="7.6.36" id="api-mlgraphbuilder-squeeze"><span class="secno">7.6.36. </span><span class="content">The squeeze() method</span><a class="self-link" href="#api-mlgraphbuilder-squeeze"></a></h4>
-     Reduce the rank of a tensor by eliminating dimensions with size 1 of the tensor shape. Squeeze only affects the tensor’s logical dimensions. It does not copy or change the content in the tensor. 
+    </details>
+   </div>
+   <h4 class="heading settled" data-level="7.6.37" id="api-mlgraphbuilder-squeeze"><span class="secno">7.6.37. </span><span class="content">The squeeze() method</span><a class="self-link" href="#api-mlgraphbuilder-squeeze"></a></h4>
+    Reduce the <a data-link-type="dfn" href="#mloperand-rank" id="ref-for-mloperand-rank②⑤">rank</a> of a tensor by eliminating dimensions with size 1 of the tensor shape. Squeeze only affects the tensor’s logical dimensions. It does not copy or change the content in the tensor. 
 <pre class="idl highlight def"><c- b>dictionary</c-> <dfn class="dfn-paneled idl-code" data-dfn-type="dictionary" data-export id="dictdef-mlsqueezeoptions"><code><c- g>MLSqueezeOptions</c-></code></dfn> {
-  <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#idl-sequence" id="ref-for-idl-sequence③③"><c- b>sequence</c-></a>&lt;<a class="idl-code" data-link-type="interface" href="https://webidl.spec.whatwg.org/#idl-unsigned-long" id="ref-for-idl-unsigned-long⑦③"><c- b>unsigned</c-> <c- b>long</c-></a>> <dfn class="idl-code" data-dfn-for="MLSqueezeOptions" data-dfn-type="dict-member" data-export data-type="sequence<unsigned long> " id="dom-mlsqueezeoptions-axes"><code><c- g>axes</c-></code><a class="self-link" href="#dom-mlsqueezeoptions-axes"></a></dfn>;
+  <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#idl-sequence" id="ref-for-idl-sequence③③"><c- b>sequence</c-></a>&lt;<a class="idl-code" data-link-type="interface" href="https://webidl.spec.whatwg.org/#idl-unsigned-long" id="ref-for-idl-unsigned-long⑧①"><c- b>unsigned</c-> <c- b>long</c-></a>> <a class="idl-code" data-link-type="dict-member" data-type="sequence<unsigned long> " href="#dom-mlsqueezeoptions-axes" id="ref-for-dom-mlsqueezeoptions-axes"><c- g>axes</c-></a>;
 };
 
-<c- b>partial</c-> <c- b>interface</c-> <a class="idl-code" data-link-type="interface" href="#mlgraphbuilder" id="ref-for-mlgraphbuilder⑤②"><c- g>MLGraphBuilder</c-></a> {
-  <a data-link-type="idl-name" href="#mloperand" id="ref-for-mloperand②⑨③"><c- n>MLOperand</c-></a> <dfn class="idl-code" data-dfn-for="MLGraphBuilder" data-dfn-type="method" data-export data-lt="squeeze(input, options)|squeeze(input)" id="dom-mlgraphbuilder-squeeze"><code><c- g>squeeze</c-></code><a class="self-link" href="#dom-mlgraphbuilder-squeeze"></a></dfn>(<a data-link-type="idl-name" href="#mloperand" id="ref-for-mloperand②⑨④"><c- n>MLOperand</c-></a> <dfn class="idl-code" data-dfn-for="MLGraphBuilder/squeeze(input, options), MLGraphBuilder/squeeze(input)" data-dfn-type="argument" data-export id="dom-mlgraphbuilder-squeeze-input-options-input"><code><c- g>input</c-></code><a class="self-link" href="#dom-mlgraphbuilder-squeeze-input-options-input"></a></dfn>, <c- b>optional</c-> <a data-link-type="idl-name" href="#dictdef-mlsqueezeoptions" id="ref-for-dictdef-mlsqueezeoptions"><c- n>MLSqueezeOptions</c-></a> <dfn class="idl-code" data-dfn-for="MLGraphBuilder/squeeze(input, options), MLGraphBuilder/squeeze(input)" data-dfn-type="argument" data-export id="dom-mlgraphbuilder-squeeze-input-options-options"><code><c- g>options</c-></code><a class="self-link" href="#dom-mlgraphbuilder-squeeze-input-options-options"></a></dfn> = {});
+<c- b>partial</c-> <c- b>interface</c-> <a class="idl-code" data-link-type="interface" href="#mlgraphbuilder" id="ref-for-mlgraphbuilder⑤⑤"><c- g>MLGraphBuilder</c-></a> {
+  <a data-link-type="idl-name" href="#mloperand" id="ref-for-mloperand③⑥⓪"><c- n>MLOperand</c-></a> <a class="idl-code" data-link-type="method" href="#dom-mlgraphbuilder-squeeze" id="ref-for-dom-mlgraphbuilder-squeeze"><c- g>squeeze</c-></a>(<a data-link-type="idl-name" href="#mloperand" id="ref-for-mloperand③⑥①"><c- n>MLOperand</c-></a> <dfn class="idl-code" data-dfn-for="MLGraphBuilder/squeeze(input, options), MLGraphBuilder/squeeze(input)" data-dfn-type="argument" data-export id="dom-mlgraphbuilder-squeeze-input-options-input"><code><c- g>input</c-></code><a class="self-link" href="#dom-mlgraphbuilder-squeeze-input-options-input"></a></dfn>, <c- b>optional</c-> <a data-link-type="idl-name" href="#dictdef-mlsqueezeoptions" id="ref-for-dictdef-mlsqueezeoptions"><c- n>MLSqueezeOptions</c-></a> <dfn class="idl-code" data-dfn-for="MLGraphBuilder/squeeze(input, options), MLGraphBuilder/squeeze(input)" data-dfn-type="argument" data-export id="dom-mlgraphbuilder-squeeze-input-options-options"><code><c- g>options</c-></code><a class="self-link" href="#dom-mlgraphbuilder-squeeze-input-options-options"></a></dfn> = {});
 };
 </pre>
-    <div class="algorithm" data-algorithm="squeeze">
-      <strong>Arguments:</strong> 
-     <ul>
+   <div>
+     <strong>Arguments:</strong> 
+    <ul>
+     <li data-md>
+      <p><em>input</em>: an <code class="idl"><a data-link-type="idl" href="#mloperand" id="ref-for-mloperand③⑥②">MLOperand</a></code>. The input tensor.</p>
+     <li data-md>
+      <p><em>options</em>: an optional <code class="idl"><a data-link-type="idl" href="#dictdef-mlsqueezeoptions" id="ref-for-dictdef-mlsqueezeoptions①">MLSqueezeOptions</a></code>. The optional parameters of the operation.</p>
+    </ul>
+    <p><strong>Returns:</strong> an <code class="idl"><a data-link-type="idl" href="#mloperand" id="ref-for-mloperand③⑥③">MLOperand</a></code>. The output tensor of the same or reduced rank with the shape dimensions of size 1 eliminated.</p>
+   </div>
+   <p><code class="idl"><a data-link-type="idl" href="#dictdef-mlsqueezeoptions" id="ref-for-dictdef-mlsqueezeoptions②">MLSqueezeOptions</a></code> has the following members:</p>
+   <dl>
+    <dt data-md><dfn class="dfn-paneled idl-code" data-dfn-for="MLSqueezeOptions" data-dfn-type="dict-member" data-export id="dom-mlsqueezeoptions-axes"><code>axes</code></dfn>, <span> of type <code class="idl-code">sequence&lt;unsigned long></code></span>
+    <dd data-md>
+     <p>A sequence of <code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#idl-unsigned-long" id="ref-for-idl-unsigned-long⑧②">unsigned long</a></code>.
+Specifies the indices to the shape dimensions of size 1 to eliminate. The values in the sequence must be in the range [0, N-1] where N is the <a data-link-type="dfn" href="#mloperand-rank" id="ref-for-mloperand-rank②⑥">rank</a> of the input tensor.
+When not specified, every shape dimensions of size 1 in the tensor are eliminated.</p>
+   </dl>
+   <details class="algorithm" data-algorithm="squeeze(input, options)" data-algorithm-for="MLGraphBuilder" open>
+    <summary> The <dfn class="dfn-paneled idl-code" data-dfn-for="MLGraphBuilder" data-dfn-type="method" data-export data-lt="squeeze(input, options)|squeeze(input)" id="dom-mlgraphbuilder-squeeze"><code>squeeze(<var>input</var>, <var>options</var>)</code></dfn> method steps are: </summary>
+    <div class="algorithm-steps">
+     <ol>
       <li data-md>
-       <p><em>input</em>: an <code class="idl"><a data-link-type="idl" href="#mloperand" id="ref-for-mloperand②⑨⑤">MLOperand</a></code>. The input tensor.</p>
+       <p><a data-link-type="dfn" href="https://infra.spec.whatwg.org/#assert" id="ref-for-assert⑦⓪">Assert</a>: the type of <var>input</var> is <code class="idl"><a data-link-type="idl" href="#mloperand" id="ref-for-mloperand③⑥④">MLOperand</a></code>.</p>
       <li data-md>
-       <p><em>options</em>: an optional <code class="idl"><a data-link-type="idl" href="#dictdef-mlsqueezeoptions" id="ref-for-dictdef-mlsqueezeoptions①">MLSqueezeOptions</a></code>. The optional parameters of the operation.</p>
-       <ul>
+       <p>If <var>options</var>.<code class="idl"><a data-link-type="idl" href="#dom-mlsqueezeoptions-axes" id="ref-for-dom-mlsqueezeoptions-axes①">axes</a></code> <a data-link-type="dfn" href="https://infra.spec.whatwg.org/#map-exists" id="ref-for-map-exists⑤⑥">exists</a>, then:</p>
+       <ol>
         <li data-md>
-         <p><em>axes</em>: a sequence of <code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#idl-unsigned-long" id="ref-for-idl-unsigned-long⑦④">unsigned long</a></code>. Indices to the shape dimensions of size 1 to eliminate. The values in the sequence must be in the range [0, N-1] where N is the rank of input tensor. When not specified, every shape dimensions of size 1 in the tensor are eliminated.</p>
-       </ul>
-     </ul>
-     <p><strong>Returns:</strong> an <code class="idl"><a data-link-type="idl" href="#mloperand" id="ref-for-mloperand②⑨⑥">MLOperand</a></code>. The output tensor of the same or reduced rank with the shape dimensions of size 1 eliminated.</p>
+         <p>Let <var>dimensions</var> be <var>input</var>.<code class="idl"><a data-link-type="idl" href="#dom-mloperand-descriptor-slot" id="ref-for-dom-mloperand-descriptor-slot⑨③">[[descriptor]]</a></code>.<code class="idl"><a data-link-type="idl" href="#dom-mloperanddescriptor-dimensions" id="ref-for-dom-mloperanddescriptor-dimensions⑨③">dimensions</a></code>.</p>
+        <li data-md>
+         <p>Let <var>axesLength</var> be the <a data-link-type="dfn" href="https://infra.spec.whatwg.org/#list-size" id="ref-for-list-size⑤⑤">size</a> of <var>options</var>.<code class="idl"><a data-link-type="idl" href="#dom-mlsqueezeoptions-axes" id="ref-for-dom-mlsqueezeoptions-axes②">axes</a></code>.</p>
+        <li data-md>
+         <p>If <var>axesLength</var> is not smaller than the <a data-link-type="dfn" href="#mloperand-rank" id="ref-for-mloperand-rank②⑦">rank</a> of <var>dimensions</var>,</p>
+        <li data-md>
+         <p>For <var>index</var> in <a data-link-type="dfn" href="https://infra.spec.whatwg.org/#the-range" id="ref-for-the-range⑦">the range</a> 0 to <var>axesLength</var>, exclusive:</p>
+         <ol>
+          <li data-md>
+           <p>Let <var>oneDimIndex</var> be <var>options</var>.<code class="idl"><a data-link-type="idl" href="#dom-mlsqueezeoptions-axes" id="ref-for-dom-mlsqueezeoptions-axes③">axes</a></code>[<var>index</var>].</p>
+          <li data-md>
+           <p>If <var>dimensions</var>[<var>oneDimIndex</var>] is not 1, then <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#dfn-throw" id="ref-for-dfn-throw②⑦⓪">throw</a> a <code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#exceptiondef-typeerror" id="ref-for-exceptiondef-typeerror④④">TypeError</a></code>.</p>
+         </ol>
+       </ol>
+      <li data-md>
+       <p>If any of the following sub-steps fail, <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#dfn-throw" id="ref-for-dfn-throw②⑦①">throw</a> an "<code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#operationerror" id="ref-for-operationerror④⑧">OperationError</a></code>" <code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#idl-DOMException" id="ref-for-idl-DOMException①④⑤">DOMException</a></code>.</p>
+       <ol>
+        <li data-md>
+         <p>Let <var>output</var> be the result of <a data-link-type="dfn" href="#copy-an-mloperand" id="ref-for-copy-an-mloperand①⑥">copying an MLOperand</a> given <var>input</var>.</p>
+        <li data-md>
+         <p>Make a request to the underlying platform to:</p>
+         <ol>
+          <li data-md>
+           <p>Let <var>opImpl</var> be an <a data-link-type="dfn" href="https://infra.spec.whatwg.org/#implementation-defined" id="ref-for-implementation-defined⑦②">implementation-defined</a> platform operator for the squeeze operation, given <var>options</var>.</p>
+          <li data-md>
+           <p>Store a reference of <var>opImpl</var> in <var>output</var>.<code class="idl"><a data-link-type="idl" href="#dom-mloperand-operator-slot" id="ref-for-dom-mloperand-operator-slot③④">[[operator]]</a></code>.</p>
+          <li data-md>
+           <p>Create an <a data-link-type="dfn" href="https://infra.spec.whatwg.org/#implementation-defined" id="ref-for-implementation-defined⑦③">implementation-defined</a> platform operand <var>outputImpl</var> to represent the output, given <var>output</var> and <var>opImpl</var>.</p>
+          <li data-md>
+           <p>Store a reference to <var>outputImpl</var> in <var>output</var>.<code class="idl"><a data-link-type="idl" href="#dom-mloperand-operand-slot" id="ref-for-dom-mloperand-operand-slot⑨⑧">[[operand]]</a></code>.</p>
+         </ol>
+        <li data-md>
+         <p>Connect <var>input</var>.<code class="idl"><a data-link-type="idl" href="#dom-mloperand-operand-slot" id="ref-for-dom-mloperand-operand-slot⑨⑨">[[operand]]</a></code> as input to <var>opImpl</var>.</p>
+        <li data-md>
+         <p>Connect <var>output</var>.<code class="idl"><a data-link-type="idl" href="#dom-mloperand-operand-slot" id="ref-for-dom-mloperand-operand-slot①⓪⓪">[[operand]]</a></code> as output to <var>opImpl</var>.</p>
+       </ol>
+      <li data-md>
+       <p>Return <var>output</var>.</p>
+     </ol>
     </div>
-    <h4 class="heading settled" data-level="7.6.37" id="api-mlgraphbuilder-tanh"><span class="secno">7.6.37. </span><span class="content">The tanh() method</span><a class="self-link" href="#api-mlgraphbuilder-tanh"></a></h4>
-     Compute the <a href="https://en.wikipedia.org/wiki/Hyperbolic_functions">hyperbolic tangent function</a> of the input tensor. The calculation follows the expression <code>(exp(2 * x) - 1) / (exp(2 * x) + 1)</code>. 
-<pre class="idl highlight def"><c- b>partial</c-> <c- b>interface</c-> <a class="idl-code" data-link-type="interface" href="#mlgraphbuilder" id="ref-for-mlgraphbuilder⑤③"><c- g>MLGraphBuilder</c-></a> {
-  <a data-link-type="idl-name" href="#mloperand" id="ref-for-mloperand②⑨⑦"><c- n>MLOperand</c-></a> <dfn class="idl-code" data-dfn-for="MLGraphBuilder" data-dfn-type="method" data-export data-lt="tanh(x)" id="dom-mlgraphbuilder-tanh"><code><c- g>tanh</c-></code><a class="self-link" href="#dom-mlgraphbuilder-tanh"></a></dfn>(<a data-link-type="idl-name" href="#mloperand" id="ref-for-mloperand②⑨⑧"><c- n>MLOperand</c-></a> <dfn class="idl-code" data-dfn-for="MLGraphBuilder/tanh(x)" data-dfn-type="argument" data-export id="dom-mlgraphbuilder-tanh-x-x"><code><c- g>x</c-></code><a class="self-link" href="#dom-mlgraphbuilder-tanh-x-x"></a></dfn>);
-  <a data-link-type="idl-name" href="#mlactivation" id="ref-for-mlactivation③⑧"><c- n>MLActivation</c-></a> <dfn class="idl-code" data-dfn-for="MLGraphBuilder" data-dfn-type="method" data-export data-lt="tanh()" id="dom-mlgraphbuilder-tanh①"><code><c- g>tanh</c-></code><a class="self-link" href="#dom-mlgraphbuilder-tanh①"></a></dfn>();
+   </details>
+   <h4 class="heading settled" data-level="7.6.38" id="api-mlgraphbuilder-tanh-method"><span class="secno">7.6.38. </span><span class="content">The tanh() method</span><a class="self-link" href="#api-mlgraphbuilder-tanh-method"></a></h4>
+    Compute the <a href="https://en.wikipedia.org/wiki/Hyperbolic_functions">hyperbolic tangent function</a> of the input tensor. The calculation follows the expression <code>(exp(2 * x) - 1) / (exp(2 * x) + 1)</code>. 
+<pre class="idl highlight def"><c- b>partial</c-> <c- b>interface</c-> <a class="idl-code" data-link-type="interface" href="#mlgraphbuilder" id="ref-for-mlgraphbuilder⑤⑥"><c- g>MLGraphBuilder</c-></a> {
+  <a data-link-type="idl-name" href="#mloperand" id="ref-for-mloperand③⑥⑤"><c- n>MLOperand</c-></a> <a class="idl-code" data-link-type="method" href="#dom-mlgraphbuilder-tanh" id="ref-for-dom-mlgraphbuilder-tanh"><c- g>tanh</c-></a>(<a data-link-type="idl-name" href="#mloperand" id="ref-for-mloperand③⑥⑥"><c- n>MLOperand</c-></a> <dfn class="idl-code" data-dfn-for="MLGraphBuilder/tanh(input)" data-dfn-type="argument" data-export id="dom-mlgraphbuilder-tanh-input-input"><code><c- g>input</c-></code><a class="self-link" href="#dom-mlgraphbuilder-tanh-input-input"></a></dfn>);
+  <a data-link-type="idl-name" href="#mlactivation" id="ref-for-mlactivation⑤⑤"><c- n>MLActivation</c-></a> <a class="idl-code" data-link-type="method" href="#tanh-noargs" id="ref-for-tanh-noargs"><c- g>tanh</c-></a>();
 };
 </pre>
-    <div class="algorithm" data-algorithm="tanh">
-      <strong>Arguments:</strong> 
-     <ul>
-      <li data-md>
-       <p><em>x</em>: an <code class="idl"><a data-link-type="idl" href="#mloperand" id="ref-for-mloperand②⑨⑨">MLOperand</a></code>. The input tensor.</p>
-     </ul>
-     <p><strong>Returns:</strong></p>
-     <ul>
-      <li data-md>
-       <p>an <code class="idl"><a data-link-type="idl" href="#mloperand" id="ref-for-mloperand③⓪⓪">MLOperand</a></code>. The output tensor of the same shape as <em>x</em>.</p>
-      <li data-md>
-       <p>an <code class="idl"><a data-link-type="idl" href="#mlactivation" id="ref-for-mlactivation③⑨">MLActivation</a></code>. The activation function representing the tanh operation.</p>
-     </ul>
-     <div class="note" role="note">
-       The behavior of this operation can be generically emulated from the usage of
+   <div class="note" role="note">
+    <details open>
+     <summary> The behavior of this operation can be generically emulated from the usage of
     other operations as follow. However, user agents typically have a more
     efficient implementation for it, therefore its usage is encouraged from the
-    performance standpoint. 
+    performance standpoint. </summary>
 <pre class="highlight"><c- k>return</c-> builder<c- p>.</c->div<c- p>(</c->
           builder<c- p>.</c->sub<c- p>(</c->builder<c- p>.</c->exp<c- p>(</c->builder<c- p>.</c->mul<c- p>(</c->builder<c- p>.</c->constant<c- p>(</c-><c- mf>2</c-><c- p>),</c-> x<c- p>)),</c-> builder<c- p>.</c->constant<c- p>(</c-><c- mf>1</c-><c- p>)),</c->
           builder<c- p>.</c->add<c- p>(</c->builder<c- p>.</c->exp<c- p>(</c->builder<c- p>.</c->mul<c- p>(</c->builder<c- p>.</c->constant<c- p>(</c-><c- mf>2</c-><c- p>),</c-> x<c- p>)),</c-> builder<c- p>.</c->constant<c- p>(</c-><c- mf>1</c-><c- p>)));</c->
 </pre>
-     </div>
+    </details>
+   </div>
+   <h5 class="heading settled" data-level="7.6.38.1" id="api-mlgraphbuilder-tanh-input"><span class="secno">7.6.38.1. </span><span class="content">The <code class="idl"><a data-link-type="idl" href="#dom-mlgraphbuilder-tanh" id="ref-for-dom-mlgraphbuilder-tanh①">tanh(input)</a></code> method</span><a class="self-link" href="#api-mlgraphbuilder-tanh-input"></a></h5>
+   <div>
+     <strong>Arguments:</strong> 
+    <ul>
+     <li data-md>
+      <p><em>input</em>: an <code class="idl"><a data-link-type="idl" href="#mloperand" id="ref-for-mloperand③⑥⑦">MLOperand</a></code>. The input tensor.</p>
+    </ul>
+    <p><strong>Returns:</strong></p>
+    <ul>
+     <li data-md>
+      <p>an <code class="idl"><a data-link-type="idl" href="#mloperand" id="ref-for-mloperand③⑥⑧">MLOperand</a></code>. The output tensor of the same shape as <em>input</em>.</p>
+    </ul>
+   </div>
+   <details class="algorithm" data-algorithm="tanh(input)" data-algorithm-for="MLGraphBuilder" open>
+    <summary> The <dfn class="dfn-paneled idl-code" data-dfn-for="MLGraphBuilder" data-dfn-type="method" data-export id="dom-mlgraphbuilder-tanh"><code>tanh(<var>input</var>)</code></dfn> method steps are: </summary>
+    <div class="algorithm-steps">
+     <ol>
+      <li data-md>
+       <p><a data-link-type="dfn" href="https://infra.spec.whatwg.org/#assert" id="ref-for-assert⑦①">Assert</a>: the type of <var>input</var> is <code class="idl"><a data-link-type="idl" href="#mloperand" id="ref-for-mloperand③⑥⑨">MLOperand</a></code>.</p>
+      <li data-md>
+       <p>If any of the following sub-steps fail, <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#dfn-throw" id="ref-for-dfn-throw②⑦②">throw</a> an "<code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#operationerror" id="ref-for-operationerror④⑨">OperationError</a></code>" <code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#idl-DOMException" id="ref-for-idl-DOMException①④⑥">DOMException</a></code>.</p>
+       <ol>
+        <li data-md>
+         <p>Let <var>output</var> be the result of <a data-link-type="dfn" href="#copy-an-mloperand" id="ref-for-copy-an-mloperand①⑦">copying an MLOperand</a> given <var>input</var>.</p>
+        <li data-md>
+         <p>Make a request to the underlying platform to:</p>
+         <ol>
+          <li data-md>
+           <p>Let <var>opImpl</var> be an <a data-link-type="dfn" href="https://infra.spec.whatwg.org/#implementation-defined" id="ref-for-implementation-defined⑦④">implementation-defined</a> platform operator for the hyperbolic tangent operation.</p>
+          <li data-md>
+           <p>Store a reference of <var>opImpl</var> in <var>output</var>.<code class="idl"><a data-link-type="idl" href="#dom-mloperand-operator-slot" id="ref-for-dom-mloperand-operator-slot③⑤">[[operator]]</a></code>.</p>
+          <li data-md>
+           <p>Create an <a data-link-type="dfn" href="https://infra.spec.whatwg.org/#implementation-defined" id="ref-for-implementation-defined⑦⑤">implementation-defined</a> platform operand <var>outputImpl</var> to represent the output, given <var>output</var> and <var>opImpl</var>.</p>
+          <li data-md>
+           <p>Store a reference to <var>outputImpl</var> in <var>output</var>.<code class="idl"><a data-link-type="idl" href="#dom-mloperand-operand-slot" id="ref-for-dom-mloperand-operand-slot①⓪①">[[operand]]</a></code>.</p>
+         </ol>
+        <li data-md>
+         <p>Connect <var>input</var>.<code class="idl"><a data-link-type="idl" href="#dom-mloperand-operand-slot" id="ref-for-dom-mloperand-operand-slot①⓪②">[[operand]]</a></code> as input to <var>opImpl</var>.</p>
+        <li data-md>
+         <p>Connect <var>output</var>.<code class="idl"><a data-link-type="idl" href="#dom-mloperand-operand-slot" id="ref-for-dom-mloperand-operand-slot①⓪③">[[operand]]</a></code> as output to <var>opImpl</var>.</p>
+       </ol>
+      <li data-md>
+       <p>Return <var>output</var>.</p>
+     </ol>
+    </div>
+   </details>
+   <h5 class="heading settled" data-level="7.6.38.2" id="api-mlgraphbuilder-tanh"><span class="secno">7.6.38.2. </span><span class="content">The <code class="idl"><a data-link-type="idl" href="#tanh-noargs" id="ref-for-tanh-noargs①">tanh()</a></code> method</span><a class="self-link" href="#api-mlgraphbuilder-tanh"></a></h5>
+   <div>
+     <strong>Arguments:</strong> 
+    <ul>
+     <li data-md>
+      <p>None.</p>
+    </ul>
+    <p><strong>Returns:</strong></p>
+    <ul>
+     <li data-md>
+      <p>an <code class="idl"><a data-link-type="idl" href="#mlactivation" id="ref-for-mlactivation⑤⑥">MLActivation</a></code>. The activation function representing the tanh operation.</p>
+    </ul>
+   </div>
+   <details class="algorithm" data-algorithm="tanh()" data-algorithm-for="MLGraphBuilder" open>
+    <summary> The <dfn class="dfn-paneled idl-code" data-dfn-for="MLGraphBuilder" data-dfn-type="method" data-export id="tanh-noargs"><code>tanh()</code></dfn> method steps are: </summary>
+    <div class="algorithm-steps">
+     <ol>
+      <li data-md>
+       <p>Let <var>op</var> be the result of <a data-link-type="dfn" href="#create-an-mlactivation" id="ref-for-create-an-mlactivation①①">creating an MLActivation</a> given <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#this" id="ref-for-this③⑨">this</a> and <code>"tanh"</code>.</p>
+       <ol>
+        <li data-md>
+         <p>If that <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#dfn-throw" id="ref-for-dfn-throw②⑦③">throws</a> an error, re-<a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#dfn-throw" id="ref-for-dfn-throw②⑦④">throw</a> the error.</p>
+       </ol>
+      <li data-md>
+       <p>Return <var>op</var>.</p>
+     </ol>
     </div>
-    <h4 class="heading settled" data-level="7.6.38" id="api-mlgraphbuilder-transpose"><span class="secno">7.6.38. </span><span class="content">The transpose() method</span><a class="self-link" href="#api-mlgraphbuilder-transpose"></a></h4>
-     Permute the dimensions of the input tensor according to the <em>permutation</em> argument. 
+   </details>
+   <h4 class="heading settled" data-level="7.6.39" id="api-mlgraphbuilder-transpose"><span class="secno">7.6.39. </span><span class="content">The transpose() method</span><a class="self-link" href="#api-mlgraphbuilder-transpose"></a></h4>
+    Permute the dimensions of the input tensor according to the <em>permutation</em> argument. 
 <pre class="idl highlight def"><c- b>dictionary</c-> <dfn class="dfn-paneled idl-code" data-dfn-type="dictionary" data-export id="dictdef-mltransposeoptions"><code><c- g>MLTransposeOptions</c-></code></dfn> {
-  <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#idl-sequence" id="ref-for-idl-sequence③④"><c- b>sequence</c-></a>&lt;<a class="idl-code" data-link-type="interface" href="https://webidl.spec.whatwg.org/#idl-unsigned-long" id="ref-for-idl-unsigned-long⑦⑤"><c- b>unsigned</c-> <c- b>long</c-></a>> <dfn class="idl-code" data-dfn-for="MLTransposeOptions" data-dfn-type="dict-member" data-export data-type="sequence<unsigned long> " id="dom-mltransposeoptions-permutation"><code><c- g>permutation</c-></code><a class="self-link" href="#dom-mltransposeoptions-permutation"></a></dfn>;
+  <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#idl-sequence" id="ref-for-idl-sequence③④"><c- b>sequence</c-></a>&lt;<a class="idl-code" data-link-type="interface" href="https://webidl.spec.whatwg.org/#idl-unsigned-long" id="ref-for-idl-unsigned-long⑧③"><c- b>unsigned</c-> <c- b>long</c-></a>> <a class="idl-code" data-link-type="dict-member" data-type="sequence<unsigned long> " href="#dom-mltransposeoptions-permutation" id="ref-for-dom-mltransposeoptions-permutation"><c- g>permutation</c-></a>;
 };
 
-<c- b>partial</c-> <c- b>interface</c-> <a class="idl-code" data-link-type="interface" href="#mlgraphbuilder" id="ref-for-mlgraphbuilder⑤④"><c- g>MLGraphBuilder</c-></a> {
-  <a data-link-type="idl-name" href="#mloperand" id="ref-for-mloperand③⓪①"><c- n>MLOperand</c-></a> <dfn class="idl-code" data-dfn-for="MLGraphBuilder" data-dfn-type="method" data-export data-lt="transpose(input, options)|transpose(input)" id="dom-mlgraphbuilder-transpose"><code><c- g>transpose</c-></code><a class="self-link" href="#dom-mlgraphbuilder-transpose"></a></dfn>(<a data-link-type="idl-name" href="#mloperand" id="ref-for-mloperand③⓪②"><c- n>MLOperand</c-></a> <dfn class="idl-code" data-dfn-for="MLGraphBuilder/transpose(input, options), MLGraphBuilder/transpose(input)" data-dfn-type="argument" data-export id="dom-mlgraphbuilder-transpose-input-options-input"><code><c- g>input</c-></code><a class="self-link" href="#dom-mlgraphbuilder-transpose-input-options-input"></a></dfn>, <c- b>optional</c-> <a data-link-type="idl-name" href="#dictdef-mltransposeoptions" id="ref-for-dictdef-mltransposeoptions"><c- n>MLTransposeOptions</c-></a> <dfn class="idl-code" data-dfn-for="MLGraphBuilder/transpose(input, options), MLGraphBuilder/transpose(input)" data-dfn-type="argument" data-export id="dom-mlgraphbuilder-transpose-input-options-options"><code><c- g>options</c-></code><a class="self-link" href="#dom-mlgraphbuilder-transpose-input-options-options"></a></dfn> = {});
+<c- b>partial</c-> <c- b>interface</c-> <a class="idl-code" data-link-type="interface" href="#mlgraphbuilder" id="ref-for-mlgraphbuilder⑤⑦"><c- g>MLGraphBuilder</c-></a> {
+  <a data-link-type="idl-name" href="#mloperand" id="ref-for-mloperand③⑦⓪"><c- n>MLOperand</c-></a> <a class="idl-code" data-link-type="method" href="#dom-mlgraphbuilder-transpose" id="ref-for-dom-mlgraphbuilder-transpose"><c- g>transpose</c-></a>(<a data-link-type="idl-name" href="#mloperand" id="ref-for-mloperand③⑦①"><c- n>MLOperand</c-></a> <dfn class="idl-code" data-dfn-for="MLGraphBuilder/transpose(input, options), MLGraphBuilder/transpose(input)" data-dfn-type="argument" data-export id="dom-mlgraphbuilder-transpose-input-options-input"><code><c- g>input</c-></code><a class="self-link" href="#dom-mlgraphbuilder-transpose-input-options-input"></a></dfn>, <c- b>optional</c-> <a data-link-type="idl-name" href="#dictdef-mltransposeoptions" id="ref-for-dictdef-mltransposeoptions"><c- n>MLTransposeOptions</c-></a> <dfn class="idl-code" data-dfn-for="MLGraphBuilder/transpose(input, options), MLGraphBuilder/transpose(input)" data-dfn-type="argument" data-export id="dom-mlgraphbuilder-transpose-input-options-options"><code><c- g>options</c-></code><a class="self-link" href="#dom-mlgraphbuilder-transpose-input-options-options"></a></dfn> = {});
 };
 </pre>
-    <div class="algorithm" data-algorithm="transpose">
-      <strong>Arguments:</strong> 
-     <ul>
+   <div>
+     <strong>Arguments:</strong> 
+    <ul>
+     <li data-md>
+      <p><em>input</em>: an <code class="idl"><a data-link-type="idl" href="#mloperand" id="ref-for-mloperand③⑦②">MLOperand</a></code>. The input N-D tensor.</p>
+     <li data-md>
+      <p><em>options</em>: an optional <code class="idl"><a data-link-type="idl" href="#dictdef-mltransposeoptions" id="ref-for-dictdef-mltransposeoptions①">MLTransposeOptions</a></code>. The optional parameters of the operation.</p>
+    </ul>
+    <p><strong>Returns:</strong> an <code class="idl"><a data-link-type="idl" href="#mloperand" id="ref-for-mloperand③⑦③">MLOperand</a></code>. The permuted or transposed N-D tensor.</p>
+   </div>
+   <p><code class="idl"><a data-link-type="idl" href="#dictdef-mltransposeoptions" id="ref-for-dictdef-mltransposeoptions②">MLTransposeOptions</a></code> has the following members:</p>
+   <dl>
+    <dt data-md><dfn class="dfn-paneled idl-code" data-dfn-for="MLTransposeOptions" data-dfn-type="dict-member" data-export id="dom-mltransposeoptions-permutation"><code>permutation</code></dfn>, <span> of type <code class="idl-code">sequence&lt;unsigned long></code></span>
+    <dd data-md>
+     <p>A sequence of <code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#idl-unsigned-long" id="ref-for-idl-unsigned-long⑧④">unsigned long</a></code> values.
+Specifies the values used to permute the output shape.
+The default value is [N-1, ..., 0], where N is the <a data-link-type="dfn" href="#mloperand-rank" id="ref-for-mloperand-rank②⑧">rank</a> of the input tensor, e.g. [2,1,0] for a 3-D tensor.
+These default values cause the output to become a transposed tensor of the input. When specified, the number of values in the sequence must be the same as the <a data-link-type="dfn" href="#mloperand-rank" id="ref-for-mloperand-rank②⑨">rank</a> of the input tensor, and the values in the sequence must be within the range from 0 to N-1 with no two or more same values found in the sequence.</p>
+   </dl>
+   <details class="algorithm" data-algorithm="transpose(input, options)" data-algorithm-for="MLGraphBuilder" open>
+    <summary> The <dfn class="dfn-paneled idl-code" data-dfn-for="MLGraphBuilder" data-dfn-type="method" data-export data-lt="transpose(input, options)|transpose(input)" id="dom-mlgraphbuilder-transpose"><code>transpose(<var>input</var>, <var>options</var>)</code></dfn> method steps are: </summary>
+    <div class="algorithm-steps">
+     <ol>
       <li data-md>
-       <p><em>input</em>: an <code class="idl"><a data-link-type="idl" href="#mloperand" id="ref-for-mloperand③⓪③">MLOperand</a></code>. The input N-D tensor.</p>
+       <p><a data-link-type="dfn" href="https://infra.spec.whatwg.org/#assert" id="ref-for-assert⑦②">Assert</a>: the type of <var>input</var> is <code class="idl"><a data-link-type="idl" href="#mloperand" id="ref-for-mloperand③⑦④">MLOperand</a></code>.</p>
       <li data-md>
-       <p><em>options</em>: an optional <code class="idl"><a data-link-type="idl" href="#dictdef-mltransposeoptions" id="ref-for-dictdef-mltransposeoptions①">MLTransposeOptions</a></code>. The optional parameters of the operation.</p>
-       <ul>
+       <p>If <var>options</var>.<code class="idl"><a data-link-type="idl" href="#dom-mltransposeoptions-permutation" id="ref-for-dom-mltransposeoptions-permutation①">permutation</a></code> does not <a data-link-type="dfn" href="https://infra.spec.whatwg.org/#map-exists" id="ref-for-map-exists⑤⑦">exist</a>, let <var>options</var>.<code class="idl"><a data-link-type="idl" href="#dom-mltransposeoptions-permutation" id="ref-for-dom-mltransposeoptions-permutation②">permutation</a></code> be the reversed sequence of all indices for <var>input</var>.<code class="idl"><a data-link-type="idl" href="#dom-mloperand-descriptor-slot" id="ref-for-dom-mloperand-descriptor-slot⑨④">[[descriptor]]</a></code>.<code class="idl"><a data-link-type="idl" href="#dom-mloperanddescriptor-dimensions" id="ref-for-dom-mloperanddescriptor-dimensions⑨④">dimensions</a></code>.</p>
+      <li data-md>
+       <p>Otherwise if <var>options</var>.<code class="idl"><a data-link-type="idl" href="#dom-mltransposeoptions-permutation" id="ref-for-dom-mltransposeoptions-permutation③">permutation</a></code> <a data-link-type="dfn" href="https://infra.spec.whatwg.org/#map-exists" id="ref-for-map-exists⑤⑧">exists</a>:</p>
+       <ol>
         <li data-md>
-         <p><em>permutation</em>: a sequence of <code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#idl-unsigned-long" id="ref-for-idl-unsigned-long⑦⑥">unsigned long</a></code> values. The values used to permute the output shape. When it’s not specified, it’s set to [N-1, ..., 0], where N is the rank of the input tensor, e.g. [2,1,0] for a 3-D tensor. These default values cause the output to become a transposed tensor of the input. When specified, the number of values in the sequence must be the same as the rank of the input tensor, and the values in the sequence must be within the range from 0 to N-1 with no two or more same values found in the sequence.</p>
-       </ul>
-     </ul>
-     <p><strong>Returns:</strong> an <code class="idl"><a data-link-type="idl" href="#mloperand" id="ref-for-mloperand③⓪④">MLOperand</a></code>. The permuted or transposed N-D tensor.</p>
+         <p>If the <a data-link-type="dfn" href="#mloperand-rank" id="ref-for-mloperand-rank③⓪">rank</a> of <var>options</var>.<code class="idl"><a data-link-type="idl" href="#dom-mltransposeoptions-permutation" id="ref-for-dom-mltransposeoptions-permutation④">permutation</a></code> is not the same as the <a data-link-type="dfn" href="#mloperand-rank" id="ref-for-mloperand-rank③①">rank</a> of <var>input</var>.<code class="idl"><a data-link-type="idl" href="#dom-mloperand-descriptor-slot" id="ref-for-dom-mloperand-descriptor-slot⑨⑤">[[descriptor]]</a></code>.<code class="idl"><a data-link-type="idl" href="#dom-mloperanddescriptor-dimensions" id="ref-for-dom-mloperanddescriptor-dimensions⑨⑤">dimensions</a></code>, then <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#dfn-throw" id="ref-for-dfn-throw②⑦⑤">throw</a> a <code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#exceptiondef-typeerror" id="ref-for-exceptiondef-typeerror④⑤">TypeError</a></code>.</p>
+        <li data-md>
+         <p>If the values in <var>options</var>.<code class="idl"><a data-link-type="idl" href="#dom-mltransposeoptions-permutation" id="ref-for-dom-mltransposeoptions-permutation⑤">permutation</a></code> are not in <a data-link-type="dfn" href="https://infra.spec.whatwg.org/#the-range" id="ref-for-the-range⑧">the range</a> 0 and the <a data-link-type="dfn" href="#mloperand-rank" id="ref-for-mloperand-rank③②">rank</a> of <var>input</var>.<code class="idl"><a data-link-type="idl" href="#dom-mloperand-descriptor-slot" id="ref-for-dom-mloperand-descriptor-slot⑨⑥">[[descriptor]]</a></code>.<code class="idl"><a data-link-type="idl" href="#dom-mloperanddescriptor-dimensions" id="ref-for-dom-mloperanddescriptor-dimensions⑨⑥">dimensions</a></code> exclusive, then <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#dfn-throw" id="ref-for-dfn-throw②⑦⑥">throw</a> a <code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#exceptiondef-typeerror" id="ref-for-exceptiondef-typeerror④⑥">TypeError</a></code>.</p>
+        <li data-md>
+         <p>If the values in <var>options</var>.<code class="idl"><a data-link-type="idl" href="#dom-mltransposeoptions-permutation" id="ref-for-dom-mltransposeoptions-permutation⑥">permutation</a></code> contain duplicate value, then <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#dfn-throw" id="ref-for-dfn-throw②⑦⑦">throw</a> a <code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#exceptiondef-typeerror" id="ref-for-exceptiondef-typeerror④⑦">TypeError</a></code>.</p>
+       </ol>
+      <li data-md>
+       <p>If any of the following sub-steps fail, <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#dfn-throw" id="ref-for-dfn-throw②⑦⑧">throw</a> an "<code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#operationerror" id="ref-for-operationerror⑤⓪">OperationError</a></code>" <code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#idl-DOMException" id="ref-for-idl-DOMException①④⑦">DOMException</a></code>.</p>
+       <ol>
+        <li data-md>
+         <p>Let <var>output</var> be the result of <a data-link-type="dfn" href="#copy-an-mloperand" id="ref-for-copy-an-mloperand①⑧">copying an MLOperand</a> given <var>input</var>.</p>
+        <li data-md>
+         <p>Make a request to the underlying platform to:</p>
+         <ol>
+          <li data-md>
+           <p>Let <var>opImpl</var> be an <a data-link-type="dfn" href="https://infra.spec.whatwg.org/#implementation-defined" id="ref-for-implementation-defined⑦⑥">implementation-defined</a> platform operator for the transpose operation, given <var>options</var>.</p>
+          <li data-md>
+           <p>Store a reference of <var>opImpl</var> in <var>output</var>.<code class="idl"><a data-link-type="idl" href="#dom-mloperand-operator-slot" id="ref-for-dom-mloperand-operator-slot③⑥">[[operator]]</a></code>.</p>
+          <li data-md>
+           <p>Create an <a data-link-type="dfn" href="https://infra.spec.whatwg.org/#implementation-defined" id="ref-for-implementation-defined⑦⑦">implementation-defined</a> platform operand <var>outputImpl</var> to represent the output, given <var>output</var> and <var>opImpl</var>.</p>
+          <li data-md>
+           <p>Store a reference to <var>outputImpl</var> in <var>output</var>.<code class="idl"><a data-link-type="idl" href="#dom-mloperand-operand-slot" id="ref-for-dom-mloperand-operand-slot①⓪④">[[operand]]</a></code>.</p>
+         </ol>
+        <li data-md>
+         <p>Connect <var>input</var>.<code class="idl"><a data-link-type="idl" href="#dom-mloperand-operand-slot" id="ref-for-dom-mloperand-operand-slot①⓪⑤">[[operand]]</a></code> as input to <var>opImpl</var>.</p>
+        <li data-md>
+         <p>Connect <var>output</var>.<code class="idl"><a data-link-type="idl" href="#dom-mloperand-operand-slot" id="ref-for-dom-mloperand-operand-slot①⓪⑥">[[operand]]</a></code> as output to <var>opImpl</var>.</p>
+       </ol>
+      <li data-md>
+       <p>Return <var>output</var>.</p>
+     </ol>
     </div>
-    <h2 class="heading settled" data-level="8" id="examples"><span class="secno">8. </span><span class="content">Examples</span><a class="self-link" href="#examples"></a></h2>
-    <div class="example" id="example-99852ef0">
-     <a class="self-link" href="#example-99852ef0"></a> The following code gets the MLContext object. 
+   </details>
+   <h2 class="heading settled" data-level="8" id="examples"><span class="secno">8. </span><span class="content">Examples</span><a class="self-link" href="#examples"></a></h2>
+   <div class="example" id="example-99852ef0">
+    <a class="self-link" href="#example-99852ef0"></a> The following code gets the MLContext object. 
 <pre class="highlight"><c- a>const</c-> context <c- o>=</c-> <c- k>await</c-> navigator<c- p>.</c->ml<c- p>.</c->createContext<c- p>({</c->powerPreference<c- o>:</c-> <c- t>'low-power'</c-><c- p>});</c->
 </pre>
-    </div>
-    <div class="example" id="example-4863c0b5">
-     <a class="self-link" href="#example-4863c0b5"></a> The following code builds a graph as: 
+   </div>
+   <div class="example" id="example-0dff3e64">
+    <a class="self-link" href="#example-0dff3e64"></a> Given the following build graph: 
 <pre>constant1 ---+
-             +--- Add ---> intermediateOutput1 ---+
+            +--- Add ---> intermediateOutput1 ---+
 input1    ---+                                    |
-                                                  +--- Mul---> output
+                                                +--- Mul---> output
 constant2 ---+                                    |
-             +--- Add ---> intermediateOutput2 ---+
+            +--- Add ---> intermediateOutput2 ---+
 input2    ---+
 </pre>
+    <details open>
+     <summary> The following code implements the graph: </summary>
 <pre class="highlight"><c- c1>// Use tensors in 4 dimensions.</c->
 <c- a>const</c-> TENSOR_DIMS <c- o>=</c-> <c- p>[</c-><c- mf>1</c-><c- p>,</c-> <c- mf>2</c-><c- p>,</c-> <c- mf>2</c-><c- p>,</c-> <c- mf>2</c-><c- p>];</c->
 <c- a>const</c-> TENSOR_SIZE <c- o>=</c-> <c- mf>8</c-><c- p>;</c->
@@ -4199,15 +7570,18 @@ <h2 class="heading settled" data-level="8" id="examples"><span class="secno">8.
 <c- c1>// output is the output MLOperand of the Mul operation.</c->
 <c- a>const</c-> output <c- o>=</c-> builder<c- p>.</c->mul<c- p>(</c->intermediateOutput1<c- p>,</c-> intermediateOutput2<c- p>);</c->
 </pre>
-    </div>
-    <div class="example" id="example-215c9f06">
-     <a class="self-link" href="#example-215c9f06"></a> Compile the graph up to the output operand. 
+    </details>
+   </div>
+   <div class="example" id="example-215c9f06">
+    <a class="self-link" href="#example-215c9f06"></a> Compile the graph up to the output operand. 
 <pre class="highlight"><c- c1>// Compile the constructed graph.</c->
 <c- a>const</c-> graph <c- o>=</c-> <c- k>await</c-> builder<c- p>.</c->build<c- p>({</c-><c- t>'output'</c-><c- o>:</c-> output<c- p>});</c->
 </pre>
-    </div>
-    <div class="example" id="example-e1b87669">
-     <a class="self-link" href="#example-e1b87669"></a> The following code executes the compiled graph. 
+   </div>
+   <div class="example" id="example-e1b87669">
+    <a class="self-link" href="#example-e1b87669"></a> 
+    <details open>
+     <summary> The following code executes the compiled graph. </summary>
 <pre class="highlight"><c- c1>// Setup the input buffers with value 1.</c->
 <c- a>const</c-> inputBuffer1 <c- o>=</c-> <c- ow>new</c-> Float32Array<c- p>(</c->TENSOR_SIZE<c- p>).</c->fill<c- p>(</c-><c- mf>1</c-><c- p>);</c->
 <c- a>const</c-> inputBuffer2 <c- o>=</c-> <c- ow>new</c-> Float32Array<c- p>(</c->TENSOR_SIZE<c- p>).</c->fill<c- p>(</c-><c- mf>1</c-><c- p>);</c->
@@ -4215,8 +7589,8 @@ <h2 class="heading settled" data-level="8" id="examples"><span class="secno">8.
 
 <c- c1>// Execute the compiled graph with the specified inputs.</c->
 <c- a>const</c-> inputs <c- o>=</c-> <c- p>{</c->
-  <c- t>'input1'</c-><c- o>:</c-> inputBuffer1<c- p>,</c->
-  <c- t>'input2'</c-><c- o>:</c-> inputBuffer2<c- p>,</c->
+<c- t>'input1'</c-><c- o>:</c-> inputBuffer1<c- p>,</c->
+<c- t>'input2'</c-><c- o>:</c-> inputBuffer2<c- p>,</c->
 <c- p>};</c->
 <c- a>const</c-> outputs <c- o>=</c-> <c- p>{</c-><c- t>'output'</c-><c- o>:</c-> outputBuffer<c- p>};</c->
 <c- a>const</c-> result <c- o>=</c-> <c- k>await</c-> context<c- p>.</c->compute<c- p>(</c->graph<c- p>,</c-> inputs<c- p>,</c-> outputs<c- p>);</c->
@@ -4224,49 +7598,50 @@ <h2 class="heading settled" data-level="8" id="examples"><span class="secno">8.
 console<c- p>.</c->log<c- p>(</c-><c- t>'Output value: '</c-> <c- o>+</c-> result<c- p>.</c->outputs<c- p>.</c->output<c- p>);</c->
 <c- c1>// Output value: 2.25,2.25,2.25,2.25,2.25,2.25,2.25,2.25</c->
 </pre>
-    </div>
-    <h2 class="heading settled" data-level="9" id="appendices"><span class="secno">9. </span><span class="content">Appendices</span><a class="self-link" href="#appendices"></a></h2>
-    <h3 class="heading settled" data-level="9.1" id="appendices-mloperandtype-arraybufferview-compatibility"><span class="secno">9.1. </span><span class="content"><code class="idl"><a data-link-type="idl" href="#enumdef-mloperandtype" id="ref-for-enumdef-mloperandtype⑤">MLOperandType</a></code> and <code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#ArrayBufferView" id="ref-for-ArrayBufferView⑧">ArrayBufferView</a></code> compatibility</span><a class="self-link" href="#appendices-mloperandtype-arraybufferview-compatibility"></a></h3>
-    <table class="data">
-     <thead class="stickyheader">
-      <tr>
-       <th><code class="idl"><a data-link-type="idl" href="#enumdef-mloperandtype" id="ref-for-enumdef-mloperandtype⑥">MLOperandType</a></code> 
-       <th><code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#ArrayBufferView" id="ref-for-ArrayBufferView⑨">ArrayBufferView</a></code> 
-     <tbody>
-      <tr>
-       <td><code class="idl"><a data-link-type="idl" href="#dom-mloperandtype-float32" id="ref-for-dom-mloperandtype-float32">float32</a></code> 
-       <td><code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#idl-Float32Array" id="ref-for-idl-Float32Array">Float32Array</a></code> 
-      <tr>
-       <td><code class="idl"><a data-link-type="idl" href="#dom-mloperandtype-float16" id="ref-for-dom-mloperandtype-float16">float16</a></code> 
-       <td><code class="idl"><a data-link-type="idl" href="https://tc39.es/proposal-float16array/#sec-float16array" id="ref-for-sec-float16array">Float16Array</a></code> 
-      <tr>
-       <td><code class="idl"><a data-link-type="idl" href="#dom-mloperandtype-int32" id="ref-for-dom-mloperandtype-int32">int32</a></code> 
-       <td><code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#idl-Int32Array" id="ref-for-idl-Int32Array">Int32Array</a></code> 
-      <tr>
-       <td><code class="idl"><a data-link-type="idl" href="#dom-mloperandtype-uint32" id="ref-for-dom-mloperandtype-uint32">uint32</a></code> 
-       <td><code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#idl-Uint32Array" id="ref-for-idl-Uint32Array">Uint32Array</a></code> 
-      <tr>
-       <td><code class="idl"><a data-link-type="idl" href="#dom-mloperandtype-int8" id="ref-for-dom-mloperandtype-int8">int8</a></code> 
-       <td><code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#idl-Int8Array" id="ref-for-idl-Int8Array">Int8Array</a></code> 
-      <tr>
-       <td><code class="idl"><a data-link-type="idl" href="#dom-mloperandtype-uint8" id="ref-for-dom-mloperandtype-uint8">uint8</a></code> 
-       <td><code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#idl-Uint8Array" id="ref-for-idl-Uint8Array">Uint8Array</a></code> 
-    </table>
-    <p class="note" role="note"><code class="idl"><a data-link-type="idl" href="https://tc39.es/proposal-float16array/#sec-float16array" id="ref-for-sec-float16array①">Float16Array</a></code> is at <a href="https://tc39.es/process-document/">ECMA Stage 3</a> signaling its design is finished. Implementers wanting to enable this type ahead native implementations can emulate the type by passing raw bits via <code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#idl-Uint16Array" id="ref-for-idl-Uint16Array">Uint16Array</a></code>. <a href="https://github.com/webmachinelearning/webnn/issues/373">[Issue webnn#373]</a></p>
-    <h2 class="heading settled" data-level="10" id="acknowledgements"><span class="secno">10. </span><span class="content">Acknowledgements</span><a class="self-link" href="#acknowledgements"></a></h2>
-    <p>This specification follows the concepts of the Android Neural Networks API C
+    </details>
+   </div>
+   <h2 class="heading settled" data-level="9" id="appendices"><span class="secno">9. </span><span class="content">Appendices</span><a class="self-link" href="#appendices"></a></h2>
+   <h3 class="heading settled" data-level="9.1" id="appendices-mloperandtype-arraybufferview-compatibility"><span class="secno">9.1. </span><span class="content"><code class="idl"><a data-link-type="idl" href="#enumdef-mloperandtype" id="ref-for-enumdef-mloperandtype④">MLOperandType</a></code> and <code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#ArrayBufferView" id="ref-for-ArrayBufferView⑧">ArrayBufferView</a></code> compatibility</span><a class="self-link" href="#appendices-mloperandtype-arraybufferview-compatibility"></a></h3>
+   <table class="data">
+    <thead class="stickyheader">
+     <tr>
+      <th><code class="idl"><a data-link-type="idl" href="#enumdef-mloperandtype" id="ref-for-enumdef-mloperandtype⑤">MLOperandType</a></code> 
+      <th><code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#ArrayBufferView" id="ref-for-ArrayBufferView⑨">ArrayBufferView</a></code> 
+    <tbody>
+     <tr>
+      <td><code class="idl"><a data-link-type="idl" href="#dom-mloperandtype-float32" id="ref-for-dom-mloperandtype-float32">float32</a></code> 
+      <td><code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#idl-Float32Array" id="ref-for-idl-Float32Array">Float32Array</a></code> 
+     <tr>
+      <td><code class="idl"><a data-link-type="idl" href="#dom-mloperandtype-float16" id="ref-for-dom-mloperandtype-float16">float16</a></code> 
+      <td><code class="idl"><a data-link-type="idl" href="https://tc39.es/proposal-float16array/#sec-float16array" id="ref-for-sec-float16array">Float16Array</a></code> 
+     <tr>
+      <td><code class="idl"><a data-link-type="idl" href="#dom-mloperandtype-int32" id="ref-for-dom-mloperandtype-int32">int32</a></code> 
+      <td><code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#idl-Int32Array" id="ref-for-idl-Int32Array">Int32Array</a></code> 
+     <tr>
+      <td><code class="idl"><a data-link-type="idl" href="#dom-mloperandtype-uint32" id="ref-for-dom-mloperandtype-uint32">uint32</a></code> 
+      <td><code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#idl-Uint32Array" id="ref-for-idl-Uint32Array">Uint32Array</a></code> 
+     <tr>
+      <td><code class="idl"><a data-link-type="idl" href="#dom-mloperandtype-int8" id="ref-for-dom-mloperandtype-int8">int8</a></code> 
+      <td><code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#idl-Int8Array" id="ref-for-idl-Int8Array">Int8Array</a></code> 
+     <tr>
+      <td><code class="idl"><a data-link-type="idl" href="#dom-mloperandtype-uint8" id="ref-for-dom-mloperandtype-uint8">uint8</a></code> 
+      <td><code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#idl-Uint8Array" id="ref-for-idl-Uint8Array">Uint8Array</a></code> 
+   </table>
+   <p class="note" role="note"><code class="idl"><a data-link-type="idl" href="https://tc39.es/proposal-float16array/#sec-float16array" id="ref-for-sec-float16array①">Float16Array</a></code> is at <a href="https://tc39.es/process-document/">ECMA Stage 3</a> signaling its design is finished. Implementers wanting to enable this type ahead native implementations can emulate the type by passing raw bits via <code class="idl"><a data-link-type="idl" href="https://webidl.spec.whatwg.org/#idl-Uint16Array" id="ref-for-idl-Uint16Array">Uint16Array</a></code>. <a href="https://github.com/webmachinelearning/webnn/issues/373">[Issue webnn#373]</a></p>
+   <h2 class="heading settled" data-level="10" id="acknowledgements"><span class="secno">10. </span><span class="content">Acknowledgements</span><a class="self-link" href="#acknowledgements"></a></h2>
+   <p>This specification follows the concepts of the Android Neural Networks API C
 API.</p>
-    <p>Thanks to Tomoyuki Shimizu, Ningxin Hu, Zhiqiang Yu and Belem Zhang for the use
+   <p>Thanks to Tomoyuki Shimizu, Ningxin Hu, Zhiqiang Yu and Belem Zhang for the use
 cases.</p>
-    <p>Thanks to Nikhil Thorat, Daniel Smilkov, Ganesan Ramalingam, Rafael Cintron and
+   <p>Thanks to Nikhil Thorat, Daniel Smilkov, Ganesan Ramalingam, Rafael Cintron and
 Benjamin Poulain for their contributions to the API specification.</p>
-    <p>Thanks to Sangwhan Moon and the W3C Technical Architecture Group for review of this specification for web architecture fit, design consistency and developer ergonomics.</p>
-    <p>Thanks to W3C Privacy Interest Group for privacy and security review and feedback.</p>
-    <p>Thanks to Alex Gough and the Chrome Security team for security review and questions.</p>
-    <p>Thanks to Michal Karzynski for sharing practical guidelines and learnings from ONNX.</p>
-    <p>Thanks to Kaustubha Govind and Chrome privacy reviewers for feedback and privacy considerations.</p>
-    <p>Thanks to Jiewei Qian for Chromium implementation review and feedback.</p>
-   </div>
+   <p>Thanks to Sangwhan Moon and the W3C Technical Architecture Group for review of this specification for web architecture fit, design consistency and developer ergonomics.</p>
+   <p>Thanks to Zoltan Kis for adding algorithms and making navigating this specification a delightful experience. Thanks to Joshua Bell for aligning the specification with modern editorial conventions. Thanks to Ningxin Hu, Lisha Guo, Shiyi Zou, Mingming Xu, Junwei Fu, Bruce Dai and Bin Miao for careful review and comments.</p>
+   <p>Thanks to W3C Privacy Interest Group for privacy and security review and feedback.</p>
+   <p>Thanks to Alex Gough and the Chrome Security team for security review and questions.</p>
+   <p>Thanks to Michal Karzynski for sharing practical guidelines and learnings from ONNX.</p>
+   <p>Thanks to Kaustubha Govind and Chrome privacy reviewers for feedback and privacy considerations.</p>
+   <p>Thanks to Jiewei Qian for Chromium implementation review and feedback.</p>
   </main>
   <div data-fill-with="conformance">
    <h2 class="no-ref no-num heading settled" id="w3c-conformance"><span class="content">Conformance</span><a class="self-link" href="#w3c-conformance"></a></h2>
@@ -4315,99 +7690,109 @@ <h3 class="no-ref no-num heading settled" id="w3c-conformant-algorithms"><span c
   <h2 class="no-num no-ref heading settled" id="index"><span class="content">Index</span><a class="self-link" href="#index"></a></h2>
   <h3 class="no-num no-ref heading settled" id="index-defined-here"><span class="content">Terms defined by this specification</span><a class="self-link" href="#index-defined-here"></a></h3>
   <ul class="index">
-   <li><a href="#dom-mlgraphbuilder-abs">abs(x)</a><span>, in § 7.6.10</span>
+   <li><a href="#dom-mlgraphbuilder-abs">abs(input)</a><span>, in § 7.6.11</span>
    <li>
     activation
     <ul>
-     <li><a href="#dom-mlbatchnormalizationoptions-activation">dict-member for MLBatchNormalizationOptions</a><span>, in § 7.6.4</span>
-     <li><a href="#dom-mlconv2doptions-activation">dict-member for MLConv2dOptions</a><span>, in § 7.6.7</span>
-     <li><a href="#dom-mlconvtranspose2doptions-activation">dict-member for MLConvTranspose2dOptions</a><span>, in § 7.6.8</span>
+     <li><a href="#dom-mlbatchnormalizationoptions-activation">dict-member for MLBatchNormalizationOptions</a><span>, in § 7.6.5</span>
+     <li><a href="#dom-mlconv2doptions-activation">dict-member for MLConv2dOptions</a><span>, in § 7.6.8</span>
+     <li><a href="#dom-mlconvtranspose2doptions-activation">dict-member for MLConvTranspose2dOptions</a><span>, in § 7.6.9</span>
     </ul>
    <li>
     activations
     <ul>
-     <li><a href="#dom-mlgrucelloptions-activations">dict-member for MLGruCellOptions</a><span>, in § 7.6.14</span>
-     <li><a href="#dom-mlgruoptions-activations">dict-member for MLGruOptions</a><span>, in § 7.6.13</span>
-     <li><a href="#dom-mllstmcelloptions-activations">dict-member for MLLstmCellOptions</a><span>, in § 7.6.21</span>
-     <li><a href="#dom-mllstmoptions-activations">dict-member for MLLstmOptions</a><span>, in § 7.6.20</span>
+     <li><a href="#dom-mlgrucelloptions-activations">dict-member for MLGruCellOptions</a><span>, in § 7.6.15</span>
+     <li><a href="#dom-mlgruoptions-activations">dict-member for MLGruOptions</a><span>, in § 7.6.14</span>
+     <li><a href="#dom-mllstmcelloptions-activations">dict-member for MLLstmCellOptions</a><span>, in § 7.6.22</span>
+     <li><a href="#dom-mllstmoptions-activations">dict-member for MLLstmOptions</a><span>, in § 7.6.21</span>
     </ul>
-   <li><a href="#dom-mlgraphbuilder-add">add(a, b)</a><span>, in § 7.6.9</span>
+   <li><a href="#dom-mlgraphbuilder-add">add(a, b)</a><span>, in § 7.6.10</span>
    <li>
     alpha
     <ul>
-     <li><a href="#dom-mleluoptions-alpha">dict-member for MLEluOptions</a><span>, in § 7.6.11</span>
-     <li><a href="#dom-mlgemmoptions-alpha">dict-member for MLGemmOptions</a><span>, in § 7.6.12</span>
-     <li><a href="#dom-mlhardsigmoidoptions-alpha">dict-member for MLHardSigmoidOptions</a><span>, in § 7.6.15</span>
-     <li><a href="#dom-mlleakyreluoptions-alpha">dict-member for MLLeakyReluOptions</a><span>, in § 7.6.18</span>
-     <li><a href="#dom-mllinearoptions-alpha">dict-member for MLLinearOptions</a><span>, in § 7.6.19</span>
+     <li><a href="#dom-mleluoptions-alpha">dict-member for MLEluOptions</a><span>, in § 7.6.12</span>
+     <li><a href="#dom-mlgemmoptions-alpha">dict-member for MLGemmOptions</a><span>, in § 7.6.13</span>
+     <li><a href="#dom-mlhardsigmoidoptions-alpha">dict-member for MLHardSigmoidOptions</a><span>, in § 7.6.16</span>
+     <li><a href="#dom-mlleakyreluoptions-alpha">dict-member for MLLeakyReluOptions</a><span>, in § 7.6.19</span>
+     <li><a href="#dom-mllinearoptions-alpha">dict-member for MLLinearOptions</a><span>, in § 7.6.20</span>
     </ul>
-   <li><a href="#dom-mlgemmoptions-atranspose">aTranspose</a><span>, in § 7.6.12</span>
+   <li><a href="#dom-mlgemmoptions-atranspose">aTranspose</a><span>, in § 7.6.13</span>
    <li>
     autoPad
     <ul>
-     <li><a href="#dom-mlconv2doptions-autopad">dict-member for MLConv2dOptions</a><span>, in § 7.6.7</span>
-     <li><a href="#dom-mlconvtranspose2doptions-autopad">dict-member for MLConvTranspose2dOptions</a><span>, in § 7.6.8</span>
-     <li><a href="#dom-mlpool2doptions-autopad">dict-member for MLPool2dOptions</a><span>, in § 7.6.24</span>
+     <li><a href="#dom-mlconv2doptions-autopad">dict-member for MLConv2dOptions</a><span>, in § 7.6.8</span>
+     <li><a href="#dom-mlconvtranspose2doptions-autopad">dict-member for MLConvTranspose2dOptions</a><span>, in § 7.6.9</span>
+     <li><a href="#dom-mlpool2doptions-autopad">dict-member for MLPool2dOptions</a><span>, in § 7.6.25</span>
     </ul>
-   <li><a href="#dom-mlgraphbuilder-averagepool2d">averagePool2d(input)</a><span>, in § 7.6.24</span>
-   <li><a href="#dom-mlgraphbuilder-averagepool2d">averagePool2d(input, options)</a><span>, in § 7.6.24</span>
+   <li><a href="#dom-mlgraphbuilder-averagepool2d">averagePool2d(input)</a><span>, in § 7.6.25</span>
+   <li><a href="#dom-mlgraphbuilder-averagepool2d">averagePool2d(input, options)</a><span>, in § 7.6.25</span>
    <li>
     axes
     <ul>
-     <li><a href="#dom-mlreduceoptions-axes">dict-member for MLReduceOptions</a><span>, in § 7.6.26</span>
-     <li><a href="#dom-mlresample2doptions-axes">dict-member for MLResample2dOptions</a><span>, in § 7.6.28</span>
-     <li><a href="#dom-mlsqueezeoptions-axes">dict-member for MLSqueezeOptions</a><span>, in § 7.6.36</span>
+     <li><a href="#dom-mlreduceoptions-axes">dict-member for MLReduceOptions</a><span>, in § 7.6.27</span>
+     <li><a href="#dom-mlresample2doptions-axes">dict-member for MLResample2dOptions</a><span>, in § 7.6.29</span>
+     <li><a href="#dom-mlsqueezeoptions-axes">dict-member for MLSqueezeOptions</a><span>, in § 7.6.37</span>
     </ul>
    <li>
     axis
     <ul>
-     <li><a href="#dom-mlbatchnormalizationoptions-axis">dict-member for MLBatchNormalizationOptions</a><span>, in § 7.6.4</span>
-     <li><a href="#dom-mlsplitoptions-axis">dict-member for MLSplitOptions</a><span>, in § 7.6.35</span>
+     <li><a href="#dom-mlbatchnormalizationoptions-axis">dict-member for MLBatchNormalizationOptions</a><span>, in § 7.6.5</span>
+     <li><a href="#dom-mlsplitoptions-axis">dict-member for MLSplitOptions</a><span>, in § 7.6.36</span>
     </ul>
-   <li><a href="#dom-mlrecurrentnetworkdirection-backward">"backward"</a><span>, in § 7.6.13</span>
-   <li><a href="#dom-mlgraphbuilder-batchnormalization">batchNormalization(input, mean, variance)</a><span>, in § 7.6.4</span>
-   <li><a href="#dom-mlgraphbuilder-batchnormalization">batchNormalization(input, mean, variance, options)</a><span>, in § 7.6.4</span>
+   <li><a href="#dom-mlrecurrentnetworkdirection-backward">"backward"</a><span>, in § 7.6.14</span>
+   <li><a href="#dom-mlgraphbuilder-batchnormalization">batchNormalization(input, mean, variance)</a><span>, in § 7.6.5</span>
+   <li><a href="#dom-mlgraphbuilder-batchnormalization">batchNormalization(input, mean, variance, options)</a><span>, in § 7.6.5</span>
    <li>
     beta
     <ul>
-     <li><a href="#dom-mlgemmoptions-beta">dict-member for MLGemmOptions</a><span>, in § 7.6.12</span>
-     <li><a href="#dom-mlhardsigmoidoptions-beta">dict-member for MLHardSigmoidOptions</a><span>, in § 7.6.15</span>
-     <li><a href="#dom-mllinearoptions-beta">dict-member for MLLinearOptions</a><span>, in § 7.6.19</span>
+     <li><a href="#dom-mlgemmoptions-beta">dict-member for MLGemmOptions</a><span>, in § 7.6.13</span>
+     <li><a href="#dom-mlhardsigmoidoptions-beta">dict-member for MLHardSigmoidOptions</a><span>, in § 7.6.16</span>
+     <li><a href="#dom-mllinearoptions-beta">dict-member for MLLinearOptions</a><span>, in § 7.6.20</span>
     </ul>
    <li>
     bias
     <ul>
-     <li><a href="#dom-mlbatchnormalizationoptions-bias">dict-member for MLBatchNormalizationOptions</a><span>, in § 7.6.4</span>
-     <li><a href="#dom-mlconv2doptions-bias">dict-member for MLConv2dOptions</a><span>, in § 7.6.7</span>
-     <li><a href="#dom-mlconvtranspose2doptions-bias">dict-member for MLConvTranspose2dOptions</a><span>, in § 7.6.8</span>
-     <li><a href="#dom-mlgrucelloptions-bias">dict-member for MLGruCellOptions</a><span>, in § 7.6.14</span>
-     <li><a href="#dom-mlgruoptions-bias">dict-member for MLGruOptions</a><span>, in § 7.6.13</span>
-     <li><a href="#dom-mlinstancenormalizationoptions-bias">dict-member for MLInstanceNormalizationOptions</a><span>, in § 7.6.17</span>
-     <li><a href="#dom-mllstmcelloptions-bias">dict-member for MLLstmCellOptions</a><span>, in § 7.6.21</span>
-     <li><a href="#dom-mllstmoptions-bias">dict-member for MLLstmOptions</a><span>, in § 7.6.20</span>
+     <li><a href="#dom-mlbatchnormalizationoptions-bias">dict-member for MLBatchNormalizationOptions</a><span>, in § 7.6.5</span>
+     <li><a href="#dom-mlconv2doptions-bias">dict-member for MLConv2dOptions</a><span>, in § 7.6.8</span>
+     <li><a href="#dom-mlconvtranspose2doptions-bias">dict-member for MLConvTranspose2dOptions</a><span>, in § 7.6.9</span>
+     <li><a href="#dom-mlgrucelloptions-bias">dict-member for MLGruCellOptions</a><span>, in § 7.6.15</span>
+     <li><a href="#dom-mlgruoptions-bias">dict-member for MLGruOptions</a><span>, in § 7.6.14</span>
+     <li><a href="#dom-mlinstancenormalizationoptions-bias">dict-member for MLInstanceNormalizationOptions</a><span>, in § 7.6.18</span>
+     <li><a href="#dom-mllstmcelloptions-bias">dict-member for MLLstmCellOptions</a><span>, in § 7.6.22</span>
+     <li><a href="#dom-mllstmoptions-bias">dict-member for MLLstmOptions</a><span>, in § 7.6.21</span>
+    </ul>
+   <li><a href="#dom-mlrecurrentnetworkdirection-both">"both"</a><span>, in § 7.6.14</span>
+   <li><a href="#mlgraphbuilder-broadcast-shapes">broadcast-shapes</a><span>, in § 7.6.10</span>
+   <li><a href="#dom-mlgemmoptions-btranspose">bTranspose</a><span>, in § 7.6.13</span>
+   <li>
+    [[builder]]
+    <ul>
+     <li><a href="#dom-mlactivation-builder-slot">attribute for MLActivation</a><span>, in § 7.3.3</span>
+     <li><a href="#dom-mloperand-builder-slot">attribute for MLOperand</a><span>, in § 7.3.2</span>
     </ul>
-   <li><a href="#dom-mlrecurrentnetworkdirection-both">"both"</a><span>, in § 7.6.13</span>
-   <li><a href="#dom-mlgemmoptions-btranspose">bTranspose</a><span>, in § 7.6.12</span>
-   <li><a href="#dom-mloperand-builder-slot">[[builder]]</a><span>, in § 7.3.2</span>
-   <li><a href="#dom-mlgraphbuilder-build">build(outputs)</a><span>, in § 7.6</span>
-   <li><a href="#dom-mlgraphbuilder-buildsync">buildSync(outputs)</a><span>, in § 7.6</span>
+   <li><a href="#dom-mlgraphbuilder-build">build(outputs)</a><span>, in § 7.6.3.1</span>
+   <li><a href="#dom-mlgraphbuilder-buildsync">buildSync(outputs)</a><span>, in § 7.6.3.2</span>
    <li><a href="#mloperanddescriptor-byte-length">byte length</a><span>, in § 7.3.1</span>
-   <li><a href="#dom-mlgemmoptions-c">c</a><span>, in § 7.6.12</span>
-   <li><a href="#dom-mlroundingtype-ceil">"ceil"</a><span>, in § 7.6.24</span>
-   <li><a href="#dom-mlgraphbuilder-ceil">ceil(x)</a><span>, in § 7.6.10</span>
+   <li><a href="#dom-mlgemmoptions-c">c</a><span>, in § 7.6.13</span>
+   <li><a href="#mlgraphbuilder-calculate-matmul-output-sizes">calculate matmul output sizes</a><span>, in § 7.6.23</span>
+   <li><a href="#mlgraphbuilder-calculate-padding-output-sizes">calculate padding output sizes</a><span>, in § 7.6.24</span>
+   <li><a href="#dom-mlroundingtype-ceil">"ceil"</a><span>, in § 7.6.25</span>
+   <li><a href="#dom-mlgraphbuilder-ceil">ceil(input)</a><span>, in § 7.6.11</span>
+   <li><a href="#check-clamp-options">check clamp options</a><span>, in § 7.6.6</span>
    <li><a href="#check-dimensions">check dimensions</a><span>, in § 7.3.2.1</span>
-   <li><a href="#dom-mlgraphbuilder-clamp-options">clamp()</a><span>, in § 7.6.5</span>
-   <li><a href="#dom-mlgraphbuilder-clamp-options">clamp(options)</a><span>, in § 7.6.5</span>
-   <li><a href="#dom-mlgraphbuilder-clamp">clamp(x)</a><span>, in § 7.6.5</span>
-   <li><a href="#dom-mlgraphbuilder-clamp">clamp(x, options)</a><span>, in § 7.6.5</span>
+   <li><a href="#mlgraphbuilder-check-resample-options">check resample options</a><span>, in § 7.6.29</span>
+   <li><a href="#dom-mlgraphbuilder-clamp-options">clamp()</a><span>, in § 7.6.6.2</span>
+   <li><a href="#dom-mlgraphbuilder-clamp">clamp(operand)</a><span>, in § 7.6.6.1</span>
+   <li><a href="#dom-mlgraphbuilder-clamp">clamp(operand, options)</a><span>, in § 7.6.6.1</span>
+   <li><a href="#dom-mlgraphbuilder-clamp-options">clamp(options)</a><span>, in § 7.6.6.2</span>
    <li><a href="#dom-mlcontext-compute">compute(graph, inputs, outputs)</a><span>, in § 7.4.4</span>
    <li><a href="#dom-mlcontext-computesync">computeSync(graph, inputs, outputs)</a><span>, in § 7.4.2</span>
-   <li><a href="#dom-mlgraphbuilder-concat">concat(inputs, axis)</a><span>, in § 7.6.6</span>
-   <li><a href="#dom-mlpaddingmode-constant">"constant"</a><span>, in § 7.6.23</span>
-   <li><a href="#dom-mlgraphbuilder-constant">constant(descriptor, bufferView)</a><span>, in § 7.6</span>
-   <li><a href="#dom-mlgraphbuilder-constant-value-type">constant(value)</a><span>, in § 7.6</span>
-   <li><a href="#dom-mlgraphbuilder-constant-value-type">constant(value, type)</a><span>, in § 7.6</span>
-   <li><a href="#dom-mlgraphbuilder-mlgraphbuilder">constructor(context)</a><span>, in § 7.6</span>
+   <li><a href="#dom-mlgraphbuilder-concat">concat(inputs, axis)</a><span>, in § 7.6.7</span>
+   <li><a href="#dom-mlpaddingmode-constant">"constant"</a><span>, in § 7.6.24</span>
+   <li><a href="#dom-mlgraphbuilder-constant">constant(descriptor, bufferView)</a><span>, in § 7.6.4.1</span>
+   <li><a href="#dom-mlgraphbuilder-constant-value-type">constant(value)</a><span>, in § 7.6.4.2</span>
+   <li><a href="#dom-mlgraphbuilder-constant-value-type">constant(value, type)</a><span>, in § 7.6.4.2</span>
+   <li><a href="#dom-mlgraphbuilder-mlgraphbuilder">constructor(context)</a><span>, in § 7.6.1</span>
    <li>
     [[context]]
     <ul>
@@ -4417,21 +7802,23 @@ <h3 class="no-num no-ref heading settled" id="index-defined-here"><span class="c
     </ul>
    <li><a href="#dom-mlcontext-contexttype-slot">[[contextType]]</a><span>, in § 7.4</span>
    <li><a href="#context-type">context type</a><span>, in § 7.4</span>
-   <li><a href="#dom-mlgraphbuilder-conv2d">conv2d(input, filter)</a><span>, in § 7.6.7</span>
-   <li><a href="#dom-mlgraphbuilder-conv2d">conv2d(input, filter, options)</a><span>, in § 7.6.7</span>
-   <li><a href="#dom-mlgraphbuilder-convtranspose2d">convTranspose2d(input, filter)</a><span>, in § 7.6.8</span>
-   <li><a href="#dom-mlgraphbuilder-convtranspose2d">convTranspose2d(input, filter, options)</a><span>, in § 7.6.8</span>
-   <li><a href="#dom-mlgraphbuilder-cos">cos(x)</a><span>, in § 7.6.10</span>
+   <li><a href="#dom-mlgraphbuilder-conv2d">conv2d(input, filter)</a><span>, in § 7.6.8</span>
+   <li><a href="#dom-mlgraphbuilder-conv2d">conv2d(input, filter, options)</a><span>, in § 7.6.8</span>
+   <li><a href="#dom-mlgraphbuilder-convtranspose2d">convTranspose2d(input, filter)</a><span>, in § 7.6.9</span>
+   <li><a href="#dom-mlgraphbuilder-convtranspose2d">convTranspose2d(input, filter, options)</a><span>, in § 7.6.9</span>
+   <li><a href="#copy-an-mloperand">copy an MLOperand</a><span>, in § 7.3.2.1</span>
+   <li><a href="#dom-mlgraphbuilder-cos">cos(input)</a><span>, in § 7.6.11</span>
    <li><a href="#dom-mldevicetype-cpu">"cpu"</a><span>, in § 7.2</span>
+   <li><a href="#create-a-context">create a context</a><span>, in § 7.2.2</span>
+   <li><a href="#create-an-mlactivation">create an MLActivation</a><span>, in § 7.3.3.1</span>
+   <li><a href="#create-an-mloperand">create an MLOperand</a><span>, in § 7.3.2.1</span>
    <li><a href="#dom-mlcontext-createcommandencoder">createCommandEncoder()</a><span>, in § 7.4.5</span>
-   <li><a href="#create-context">create context</a><span>, in § 7.2.2</span>
-   <li><a href="#dom-ml-createcontext">createContext()</a><span>, in § 7.2</span>
-   <li><a href="#dom-ml-createcontext-gpudevice">createContext(gpuDevice)</a><span>, in § 7.2</span>
-   <li><a href="#dom-ml-createcontext">createContext(options)</a><span>, in § 7.2</span>
-   <li><a href="#dom-ml-createcontextsync">createContextSync()</a><span>, in § 7.2</span>
-   <li><a href="#dom-ml-createcontextsync-gpudevice">createContextSync(gpuDevice)</a><span>, in § 7.2</span>
-   <li><a href="#dom-ml-createcontextsync">createContextSync(options)</a><span>, in § 7.2</span>
-   <li><a href="#create-mloperand">create MLOperand</a><span>, in § 7.3.2.1</span>
+   <li><a href="#dom-ml-createcontext">createContext()</a><span>, in § 7.2.2</span>
+   <li><a href="#dom-ml-createcontext-gpudevice">createContext(gpuDevice)</a><span>, in § 7.2.2</span>
+   <li><a href="#dom-ml-createcontext">createContext(options)</a><span>, in § 7.2.2</span>
+   <li><a href="#dom-ml-createcontextsync">createContextSync()</a><span>, in § 7.2.3</span>
+   <li><a href="#dom-ml-createcontextsync-gpudevice">createContextSync(gpuDevice)</a><span>, in § 7.2.3</span>
+   <li><a href="#dom-ml-createcontextsync">createContextSync(options)</a><span>, in § 7.2.3</span>
    <li><a href="#dom-mlpowerpreference-default">"default"</a><span>, in § 7.2</span>
    <li><a href="#default-context">default-context</a><span>, in § 7.4</span>
    <li><a href="#dom-mloperand-descriptor-slot">[[descriptor]]</a><span>, in § 7.3.2</span>
@@ -4443,346 +7830,364 @@ <h3 class="no-num no-ref heading settled" id="index-defined-here"><span class="c
    <li>
     dilations
     <ul>
-     <li><a href="#dom-mlconv2doptions-dilations">dict-member for MLConv2dOptions</a><span>, in § 7.6.7</span>
-     <li><a href="#dom-mlconvtranspose2doptions-dilations">dict-member for MLConvTranspose2dOptions</a><span>, in § 7.6.8</span>
-     <li><a href="#dom-mlpool2doptions-dilations">dict-member for MLPool2dOptions</a><span>, in § 7.6.24</span>
+     <li><a href="#dom-mlconv2doptions-dilations">dict-member for MLConv2dOptions</a><span>, in § 7.6.8</span>
+     <li><a href="#dom-mlconvtranspose2doptions-dilations">dict-member for MLConvTranspose2dOptions</a><span>, in § 7.6.9</span>
+     <li><a href="#dom-mlpool2doptions-dilations">dict-member for MLPool2dOptions</a><span>, in § 7.6.25</span>
     </ul>
    <li><a href="#dom-mloperanddescriptor-dimensions">dimensions</a><span>, in § 7.3.1</span>
    <li>
     direction
     <ul>
-     <li><a href="#dom-mlgruoptions-direction">dict-member for MLGruOptions</a><span>, in § 7.6.13</span>
-     <li><a href="#dom-mllstmoptions-direction">dict-member for MLLstmOptions</a><span>, in § 7.6.20</span>
+     <li><a href="#dom-mlgruoptions-direction">dict-member for MLGruOptions</a><span>, in § 7.6.14</span>
+     <li><a href="#dom-mllstmoptions-direction">dict-member for MLLstmOptions</a><span>, in § 7.6.21</span>
     </ul>
    <li><a href="#dom-mlcommandencoder-dispatch">dispatch(graph, inputs, outputs)</a><span>, in § 7.5.2</span>
-   <li><a href="#dom-mlgraphbuilder-div">div(a, b)</a><span>, in § 7.6.9</span>
-   <li><a href="#dom-mlpaddingmode-edge">"edge"</a><span>, in § 7.6.23</span>
-   <li><a href="#dom-mlgraphbuilder-elu-options">elu()</a><span>, in § 7.6.11</span>
-   <li><a href="#dom-mlgraphbuilder-elu-options">elu(options)</a><span>, in § 7.6.11</span>
-   <li><a href="#dom-mlgraphbuilder-elu">elu(x)</a><span>, in § 7.6.11</span>
-   <li><a href="#dom-mlgraphbuilder-elu">elu(x, options)</a><span>, in § 7.6.11</span>
+   <li><a href="#dom-mlgraphbuilder-div">div(a, b)</a><span>, in § 7.6.10</span>
+   <li><a href="#dom-mlpaddingmode-edge">"edge"</a><span>, in § 7.6.24</span>
+   <li><a href="#mlgraphbuilder-element-wise-binary-op">element-wise-binary-op</a><span>, in § 7.6.10</span>
+   <li><a href="#mlgraphbuilder-element-wise-unary-op">element-wise-unary-op</a><span>, in § 7.6.11</span>
+   <li><a href="#dom-mlgraphbuilder-elu-options">elu()</a><span>, in § 7.6.12.2</span>
+   <li><a href="#dom-mlgraphbuilder-elu">elu(input)</a><span>, in § 7.6.12.1</span>
+   <li><a href="#dom-mlgraphbuilder-elu">elu(input, options)</a><span>, in § 7.6.12.1</span>
+   <li><a href="#dom-mlgraphbuilder-elu-options">elu(options)</a><span>, in § 7.6.12.2</span>
    <li>
     epsilon
     <ul>
-     <li><a href="#dom-mlbatchnormalizationoptions-epsilon">dict-member for MLBatchNormalizationOptions</a><span>, in § 7.6.4</span>
-     <li><a href="#dom-mlinstancenormalizationoptions-epsilon">dict-member for MLInstanceNormalizationOptions</a><span>, in § 7.6.17</span>
+     <li><a href="#dom-mlbatchnormalizationoptions-epsilon">dict-member for MLBatchNormalizationOptions</a><span>, in § 7.6.5</span>
+     <li><a href="#dom-mlinstancenormalizationoptions-epsilon">dict-member for MLInstanceNormalizationOptions</a><span>, in § 7.6.18</span>
     </ul>
    <li><a href="#execute-graph">execute graph</a><span>, in § 7.4.2</span>
-   <li><a href="#dom-mlautopad-explicit">"explicit"</a><span>, in § 7.6.7</span>
-   <li><a href="#dom-mlgraphbuilder-exp">exp(x)</a><span>, in § 7.6.10</span>
+   <li><a href="#dom-mlgraphbuilder-exp">exp(input)</a><span>, in § 7.6.11</span>
+   <li><a href="#dom-mlautopad-explicit">"explicit"</a><span>, in § 7.6.8</span>
    <li>
     filterLayout
     <ul>
-     <li><a href="#dom-mlconv2doptions-filterlayout">dict-member for MLConv2dOptions</a><span>, in § 7.6.7</span>
-     <li><a href="#dom-mlconvtranspose2doptions-filterlayout">dict-member for MLConvTranspose2dOptions</a><span>, in § 7.6.8</span>
+     <li><a href="#dom-mlconv2doptions-filterlayout">dict-member for MLConv2dOptions</a><span>, in § 7.6.8</span>
+     <li><a href="#dom-mlconvtranspose2doptions-filterlayout">dict-member for MLConvTranspose2dOptions</a><span>, in § 7.6.9</span>
     </ul>
    <li><a href="#dom-mlcommandencoder-finish">finish()</a><span>, in § 7.5.3</span>
    <li><a href="#dom-mlcommandencoder-finish">finish(descriptor)</a><span>, in § 7.5.3</span>
    <li><a href="#dom-mloperandtype-float16">"float16"</a><span>, in § 7.3.1</span>
    <li><a href="#dom-mloperandtype-float32">"float32"</a><span>, in § 7.3.1</span>
-   <li><a href="#dom-mlroundingtype-floor">"floor"</a><span>, in § 7.6.24</span>
-   <li><a href="#dom-mlgraphbuilder-floor">floor(x)</a><span>, in § 7.6.10</span>
-   <li><a href="#dom-mlrecurrentnetworkdirection-forward">"forward"</a><span>, in § 7.6.13</span>
-   <li><a href="#dom-mlgraphbuilder-gemm">gemm(a, b)</a><span>, in § 7.6.12</span>
-   <li><a href="#dom-mlgraphbuilder-gemm">gemm(a, b, options)</a><span>, in § 7.6.12</span>
+   <li><a href="#dom-mlroundingtype-floor">"floor"</a><span>, in § 7.6.25</span>
+   <li><a href="#dom-mlgraphbuilder-floor">floor(input)</a><span>, in § 7.6.11</span>
+   <li><a href="#dom-mlrecurrentnetworkdirection-forward">"forward"</a><span>, in § 7.6.14</span>
+   <li><a href="#dom-mlgraphbuilder-gemm">gemm(a, b)</a><span>, in § 7.6.13</span>
+   <li><a href="#dom-mlgraphbuilder-gemm">gemm(a, b, options)</a><span>, in § 7.6.13</span>
    <li><a href="#dom-mldevicetype-gpu">"gpu"</a><span>, in § 7.2</span>
    <li>
     groups
     <ul>
-     <li><a href="#dom-mlconv2doptions-groups">dict-member for MLConv2dOptions</a><span>, in § 7.6.7</span>
-     <li><a href="#dom-mlconvtranspose2doptions-groups">dict-member for MLConvTranspose2dOptions</a><span>, in § 7.6.8</span>
+     <li><a href="#dom-mlconv2doptions-groups">dict-member for MLConv2dOptions</a><span>, in § 7.6.8</span>
+     <li><a href="#dom-mlconvtranspose2doptions-groups">dict-member for MLConvTranspose2dOptions</a><span>, in § 7.6.9</span>
     </ul>
-   <li><a href="#dom-mlgraphbuilder-grucell">gruCell(input, weight, recurrentWeight, hiddenState, hiddenSize)</a><span>, in § 7.6.14</span>
-   <li><a href="#dom-mlgraphbuilder-grucell">gruCell(input, weight, recurrentWeight, hiddenState, hiddenSize, options)</a><span>, in § 7.6.14</span>
-   <li><a href="#dom-mlgraphbuilder-gru">gru(input, weight, recurrentWeight, steps, hiddenSize)</a><span>, in § 7.6.13</span>
-   <li><a href="#dom-mlgraphbuilder-gru">gru(input, weight, recurrentWeight, steps, hiddenSize, options)</a><span>, in § 7.6.13</span>
-   <li><a href="#dom-mlgraphbuilder-hardsigmoid-options">hardSigmoid()</a><span>, in § 7.6.15</span>
-   <li><a href="#dom-mlgraphbuilder-hardsigmoid-options">hardSigmoid(options)</a><span>, in § 7.6.15</span>
-   <li><a href="#dom-mlgraphbuilder-hardsigmoid">hardSigmoid(x)</a><span>, in § 7.6.15</span>
-   <li><a href="#dom-mlgraphbuilder-hardsigmoid">hardSigmoid(x, options)</a><span>, in § 7.6.15</span>
-   <li><a href="#dom-mlgraphbuilder-hardswish①">hardSwish()</a><span>, in § 7.6.16</span>
-   <li><a href="#dom-mlgraphbuilder-hardswish">hardSwish(x)</a><span>, in § 7.6.16</span>
+   <li><a href="#dom-mlgraphbuilder-grucell">gruCell(input, weight, recurrentWeight, hiddenState, hiddenSize)</a><span>, in § 7.6.15</span>
+   <li><a href="#dom-mlgraphbuilder-grucell">gruCell(input, weight, recurrentWeight, hiddenState, hiddenSize, options)</a><span>, in § 7.6.15</span>
+   <li><a href="#dom-mlgraphbuilder-gru">gru(input, weight, recurrentWeight, steps, hiddenSize)</a><span>, in § 7.6.14</span>
+   <li><a href="#dom-mlgraphbuilder-gru">gru(input, weight, recurrentWeight, steps, hiddenSize, options)</a><span>, in § 7.6.14</span>
+   <li><a href="#dom-mlgraphbuilder-hardsigmoid-options">hardSigmoid()</a><span>, in § 7.6.16.2</span>
+   <li><a href="#dom-mlgraphbuilder-hardsigmoid">hardSigmoid(input)</a><span>, in § 7.6.16.1</span>
+   <li><a href="#dom-mlgraphbuilder-hardsigmoid">hardSigmoid(input, options)</a><span>, in § 7.6.16.1</span>
+   <li><a href="#dom-mlgraphbuilder-hardsigmoid-options">hardSigmoid(options)</a><span>, in § 7.6.16.2</span>
+   <li><a href="#hardswish-noargs">hardSwish()</a><span>, in § 7.6.17.2</span>
+   <li><a href="#dom-mlgraphbuilder-hardswish">hardSwish(input)</a><span>, in § 7.6.17.1</span>
    <li><a href="#dom-mlpowerpreference-high-performance">"high-performance"</a><span>, in § 7.2</span>
-   <li><a href="#dom-mlconv2dfilteroperandlayout-hwio">"hwio"</a><span>, in § 7.6.7</span>
-   <li><a href="#dom-mlconvtranspose2dfilteroperandlayout-hwoi">"hwoi"</a><span>, in § 7.6.8</span>
-   <li><a href="#dom-mllstmweightlayout-ifgo">"ifgo"</a><span>, in § 7.6.20</span>
-   <li><a href="#dom-mlconv2dfilteroperandlayout-ihwo">"ihwo"</a><span>, in § 7.6.7</span>
+   <li><a href="#dom-mlconv2dfilteroperandlayout-hwio">"hwio"</a><span>, in § 7.6.8</span>
+   <li><a href="#dom-mlconvtranspose2dfilteroperandlayout-hwoi">"hwoi"</a><span>, in § 7.6.9</span>
+   <li><a href="#dom-mllstmweightlayout-ifgo">"ifgo"</a><span>, in § 7.6.21</span>
+   <li><a href="#dom-mlconv2dfilteroperandlayout-ihwo">"ihwo"</a><span>, in § 7.6.8</span>
    <li>
     [[implementation]]
     <ul>
      <li><a href="#dom-mlcommandencoder-implementation-slot">attribute for MLCommandEncoder</a><span>, in § 7.5</span>
      <li><a href="#dom-mlgraph-implementation-slot">attribute for MLGraph</a><span>, in § 7.3</span>
     </ul>
-   <li><a href="#dom-mllstmoptions-initialcellstate">initialCellState</a><span>, in § 7.6.20</span>
+   <li><a href="#dom-mllstmoptions-initialcellstate">initialCellState</a><span>, in § 7.6.21</span>
    <li>
     initialHiddenState
     <ul>
-     <li><a href="#dom-mlgruoptions-initialhiddenstate">dict-member for MLGruOptions</a><span>, in § 7.6.13</span>
-     <li><a href="#dom-mllstmoptions-initialhiddenstate">dict-member for MLLstmOptions</a><span>, in § 7.6.20</span>
+     <li><a href="#dom-mlgruoptions-initialhiddenstate">dict-member for MLGruOptions</a><span>, in § 7.6.14</span>
+     <li><a href="#dom-mllstmoptions-initialhiddenstate">dict-member for MLLstmOptions</a><span>, in § 7.6.21</span>
     </ul>
    <li><a href="#dom-mlcommandencoder-initializegraph">initializeGraph(graph)</a><span>, in § 7.5.1</span>
    <li><a href="#dom-mlgraph-inputdescriptors-slot">[[inputDescriptors]]</a><span>, in § 7.3</span>
    <li>
     inputLayout
     <ul>
-     <li><a href="#dom-mlconv2doptions-inputlayout">dict-member for MLConv2dOptions</a><span>, in § 7.6.7</span>
-     <li><a href="#dom-mlconvtranspose2doptions-inputlayout">dict-member for MLConvTranspose2dOptions</a><span>, in § 7.6.8</span>
+     <li><a href="#dom-mlconv2doptions-inputlayout">dict-member for MLConv2dOptions</a><span>, in § 7.6.8</span>
+     <li><a href="#dom-mlconvtranspose2doptions-inputlayout">dict-member for MLConvTranspose2dOptions</a><span>, in § 7.6.9</span>
     </ul>
-   <li><a href="#dom-mlgraphbuilder-input">input(name, descriptor)</a><span>, in § 7.6</span>
+   <li><a href="#dom-mlgraphbuilder-input">input(name, descriptor)</a><span>, in § 7.6.2</span>
    <li><a href="#dom-mlcomputeresult-inputs">inputs</a><span>, in § 7.4.4</span>
-   <li><a href="#dom-mlgraphbuilder-instancenormalization">instanceNormalization(input)</a><span>, in § 7.6.17</span>
-   <li><a href="#dom-mlgraphbuilder-instancenormalization">instanceNormalization(input, options)</a><span>, in § 7.6.17</span>
+   <li><a href="#dom-mlgraphbuilder-instancenormalization">instanceNormalization(input)</a><span>, in § 7.6.18</span>
+   <li><a href="#dom-mlgraphbuilder-instancenormalization">instanceNormalization(input, options)</a><span>, in § 7.6.18</span>
    <li><a href="#dom-mloperandtype-int32">"int32"</a><span>, in § 7.3.1</span>
    <li><a href="#dom-mloperandtype-int8">"int8"</a><span>, in § 7.3.1</span>
-   <li><a href="#dom-mllstmweightlayout-iofg">"iofg"</a><span>, in § 7.6.20</span>
-   <li><a href="#dom-mlconvtranspose2dfilteroperandlayout-iohw">"iohw"</a><span>, in § 7.6.8</span>
-   <li><a href="#dom-mlreduceoptions-keepdimensions">keepDimensions</a><span>, in § 7.6.26</span>
-   <li><a href="#dom-mlgraphbuilder-l2pool2d">l2Pool2d(input)</a><span>, in § 7.6.24</span>
-   <li><a href="#dom-mlgraphbuilder-l2pool2d">l2Pool2d(input, options)</a><span>, in § 7.6.24</span>
+   <li><a href="#dom-mllstmweightlayout-iofg">"iofg"</a><span>, in § 7.6.21</span>
+   <li><a href="#dom-mlconvtranspose2dfilteroperandlayout-iohw">"iohw"</a><span>, in § 7.6.9</span>
+   <li><a href="#dom-mlreduceoptions-keepdimensions">keepDimensions</a><span>, in § 7.6.27</span>
+   <li><a href="#dom-mlgraphbuilder-l2pool2d">l2Pool2d(input)</a><span>, in § 7.6.25</span>
+   <li><a href="#dom-mlgraphbuilder-l2pool2d">l2Pool2d(input, options)</a><span>, in § 7.6.25</span>
    <li>
     layout
     <ul>
-     <li><a href="#dom-mlgrucelloptions-layout">dict-member for MLGruCellOptions</a><span>, in § 7.6.14</span>
-     <li><a href="#dom-mlgruoptions-layout">dict-member for MLGruOptions</a><span>, in § 7.6.13</span>
-     <li><a href="#dom-mlinstancenormalizationoptions-layout">dict-member for MLInstanceNormalizationOptions</a><span>, in § 7.6.17</span>
-     <li><a href="#dom-mllstmcelloptions-layout">dict-member for MLLstmCellOptions</a><span>, in § 7.6.21</span>
-     <li><a href="#dom-mllstmoptions-layout">dict-member for MLLstmOptions</a><span>, in § 7.6.20</span>
-     <li><a href="#dom-mlpool2doptions-layout">dict-member for MLPool2dOptions</a><span>, in § 7.6.24</span>
+     <li><a href="#dom-mlgrucelloptions-layout">dict-member for MLGruCellOptions</a><span>, in § 7.6.15</span>
+     <li><a href="#dom-mlgruoptions-layout">dict-member for MLGruOptions</a><span>, in § 7.6.14</span>
+     <li><a href="#dom-mlinstancenormalizationoptions-layout">dict-member for MLInstanceNormalizationOptions</a><span>, in § 7.6.18</span>
+     <li><a href="#dom-mllstmcelloptions-layout">dict-member for MLLstmCellOptions</a><span>, in § 7.6.22</span>
+     <li><a href="#dom-mllstmoptions-layout">dict-member for MLLstmOptions</a><span>, in § 7.6.21</span>
+     <li><a href="#dom-mlpool2doptions-layout">dict-member for MLPool2dOptions</a><span>, in § 7.6.25</span>
     </ul>
-   <li><a href="#dom-mlgraphbuilder-leakyrelu-options">leakyRelu()</a><span>, in § 7.6.18</span>
-   <li><a href="#dom-mlgraphbuilder-leakyrelu-options">leakyRelu(options)</a><span>, in § 7.6.18</span>
-   <li><a href="#dom-mlgraphbuilder-leakyrelu">leakyRelu(x)</a><span>, in § 7.6.18</span>
-   <li><a href="#dom-mlgraphbuilder-leakyrelu">leakyRelu(x, options)</a><span>, in § 7.6.18</span>
-   <li><a href="#dom-mlinterpolationmode-linear">"linear"</a><span>, in § 7.6.28</span>
-   <li><a href="#dom-mlgraphbuilder-linear-options">linear()</a><span>, in § 7.6.19</span>
-   <li><a href="#dom-mlgraphbuilder-linear-options">linear(options)</a><span>, in § 7.6.19</span>
-   <li><a href="#dom-mlgraphbuilder-linear">linear(x)</a><span>, in § 7.6.19</span>
-   <li><a href="#dom-mlgraphbuilder-linear">linear(x, options)</a><span>, in § 7.6.19</span>
-   <li><a href="#dom-mlgraphbuilder-log">log(x)</a><span>, in § 7.6.10</span>
+   <li><a href="#dom-mlgraphbuilder-leakyrelu-options">leakyRelu()</a><span>, in § 7.6.19.2</span>
+   <li><a href="#dom-mlgraphbuilder-leakyrelu">leakyRelu(input)</a><span>, in § 7.6.19.1</span>
+   <li><a href="#dom-mlgraphbuilder-leakyrelu">leakyRelu(input, options)</a><span>, in § 7.6.19.1</span>
+   <li><a href="#dom-mlgraphbuilder-leakyrelu-options">leakyRelu(options)</a><span>, in § 7.6.19.2</span>
+   <li><a href="#dom-mlinterpolationmode-linear">"linear"</a><span>, in § 7.6.29</span>
+   <li><a href="#dom-mlgraphbuilder-linear-options">linear()</a><span>, in § 7.6.20.2</span>
+   <li><a href="#dom-mlgraphbuilder-linear">linear(input)</a><span>, in § 7.6.20.1</span>
+   <li><a href="#dom-mlgraphbuilder-linear">linear(input, options)</a><span>, in § 7.6.20.1</span>
+   <li><a href="#dom-mlgraphbuilder-linear-options">linear(options)</a><span>, in § 7.6.20.2</span>
+   <li><a href="#dom-mlgraphbuilder-log">log(input)</a><span>, in § 7.6.11</span>
    <li><a href="#dom-mlpowerpreference-low-power">"low-power"</a><span>, in § 7.2</span>
-   <li><a href="#dom-mlgraphbuilder-lstmcell">lstmCell(input, weight, recurrentWeight, hiddenState, cellState, hiddenSize)</a><span>, in § 7.6.21</span>
-   <li><a href="#dom-mlgraphbuilder-lstmcell">lstmCell(input, weight, recurrentWeight, hiddenState, cellState, hiddenSize, options)</a><span>, in § 7.6.21</span>
-   <li><a href="#dom-mlgraphbuilder-lstm">lstm(input, weight, recurrentWeight, steps, hiddenSize)</a><span>, in § 7.6.20</span>
-   <li><a href="#dom-mlgraphbuilder-lstm">lstm(input, weight, recurrentWeight, steps, hiddenSize, options)</a><span>, in § 7.6.20</span>
-   <li><a href="#dom-mlgraphbuilder-matmul">matmul(a, b)</a><span>, in § 7.6.22</span>
-   <li><a href="#dom-mlgraphbuilder-max">max(a, b)</a><span>, in § 7.6.9</span>
-   <li><a href="#dom-mlgraphbuilder-maxpool2d">maxPool2d(input)</a><span>, in § 7.6.24</span>
-   <li><a href="#dom-mlgraphbuilder-maxpool2d">maxPool2d(input, options)</a><span>, in § 7.6.24</span>
-   <li><a href="#dom-mlclampoptions-maxvalue">maxValue</a><span>, in § 7.6.5</span>
-   <li><a href="#dom-mlgraphbuilder-min">min(a, b)</a><span>, in § 7.6.9</span>
-   <li><a href="#dom-mlclampoptions-minvalue">minValue</a><span>, in § 7.6.5</span>
+   <li><a href="#dom-mlgraphbuilder-lstmcell">lstmCell(input, weight, recurrentWeight, hiddenState, cellState, hiddenSize)</a><span>, in § 7.6.22</span>
+   <li><a href="#dom-mlgraphbuilder-lstmcell">lstmCell(input, weight, recurrentWeight, hiddenState, cellState, hiddenSize, options)</a><span>, in § 7.6.22</span>
+   <li><a href="#dom-mlgraphbuilder-lstm">lstm(input, weight, recurrentWeight, steps, hiddenSize)</a><span>, in § 7.6.21</span>
+   <li><a href="#dom-mlgraphbuilder-lstm">lstm(input, weight, recurrentWeight, steps, hiddenSize, options)</a><span>, in § 7.6.21</span>
+   <li><a href="#dom-mlgraphbuilder-matmul">matmul(a, b)</a><span>, in § 7.6.23</span>
+   <li><a href="#dom-mlgraphbuilder-max">max(a, b)</a><span>, in § 7.6.10</span>
+   <li><a href="#dom-mlgraphbuilder-maxpool2d">maxPool2d(input)</a><span>, in § 7.6.25</span>
+   <li><a href="#dom-mlgraphbuilder-maxpool2d">maxPool2d(input, options)</a><span>, in § 7.6.25</span>
+   <li><a href="#dom-mlclampoptions-maxvalue">maxValue</a><span>, in § 7.6.6</span>
+   <li><a href="#dom-mlgraphbuilder-min">min(a, b)</a><span>, in § 7.6.10</span>
+   <li><a href="#dom-mlclampoptions-minvalue">minValue</a><span>, in § 7.6.6</span>
    <li><a href="#ml">ML</a><span>, in § 7.2</span>
    <li><a href="#dom-navigatorml-ml">ml</a><span>, in § 7.1</span>
    <li><a href="#mlactivation">MLActivation</a><span>, in § 7.3.3</span>
-   <li><a href="#enumdef-mlautopad">MLAutoPad</a><span>, in § 7.6.7</span>
-   <li><a href="#dictdef-mlbatchnormalizationoptions">MLBatchNormalizationOptions</a><span>, in § 7.6.4</span>
+   <li><a href="#enumdef-mlautopad">MLAutoPad</a><span>, in § 7.6.8</span>
+   <li><a href="#dictdef-mlbatchnormalizationoptions">MLBatchNormalizationOptions</a><span>, in § 7.6.5</span>
    <li><a href="#dictdef-mlbufferresourceview">MLBufferResourceView</a><span>, in § 7.6</span>
    <li><a href="#typedefdef-mlbufferview">MLBufferView</a><span>, in § 7.6</span>
-   <li><a href="#dictdef-mlclampoptions">MLClampOptions</a><span>, in § 7.6.5</span>
+   <li><a href="#dictdef-mlclampoptions">MLClampOptions</a><span>, in § 7.6.6</span>
    <li><a href="#mlcommandencoder">MLCommandEncoder</a><span>, in § 7.5</span>
    <li><a href="#dictdef-mlcomputeresult">MLComputeResult</a><span>, in § 7.4.4</span>
    <li><a href="#mlcontext">MLContext</a><span>, in § 7.4</span>
    <li><a href="#dictdef-mlcontextoptions">MLContextOptions</a><span>, in § 7.2</span>
-   <li><a href="#enumdef-mlconv2dfilteroperandlayout">MLConv2dFilterOperandLayout</a><span>, in § 7.6.7</span>
-   <li><a href="#dictdef-mlconv2doptions">MLConv2dOptions</a><span>, in § 7.6.7</span>
-   <li><a href="#enumdef-mlconvtranspose2dfilteroperandlayout">MLConvTranspose2dFilterOperandLayout</a><span>, in § 7.6.8</span>
-   <li><a href="#dictdef-mlconvtranspose2doptions">MLConvTranspose2dOptions</a><span>, in § 7.6.8</span>
+   <li><a href="#enumdef-mlconv2dfilteroperandlayout">MLConv2dFilterOperandLayout</a><span>, in § 7.6.8</span>
+   <li><a href="#dictdef-mlconv2doptions">MLConv2dOptions</a><span>, in § 7.6.8</span>
+   <li><a href="#enumdef-mlconvtranspose2dfilteroperandlayout">MLConvTranspose2dFilterOperandLayout</a><span>, in § 7.6.9</span>
+   <li><a href="#dictdef-mlconvtranspose2doptions">MLConvTranspose2dOptions</a><span>, in § 7.6.9</span>
    <li><a href="#enumdef-mldevicetype">MLDeviceType</a><span>, in § 7.2</span>
-   <li><a href="#dictdef-mleluoptions">MLEluOptions</a><span>, in § 7.6.11</span>
-   <li><a href="#dictdef-mlgemmoptions">MLGemmOptions</a><span>, in § 7.6.12</span>
+   <li><a href="#dictdef-mleluoptions">MLEluOptions</a><span>, in § 7.6.12</span>
+   <li><a href="#dictdef-mlgemmoptions">MLGemmOptions</a><span>, in § 7.6.13</span>
    <li><a href="#typedefdef-mlgpuresource">MLGPUResource</a><span>, in § 7.5</span>
    <li><a href="#mlgraph">MLGraph</a><span>, in § 7.3</span>
    <li><a href="#mlgraphbuilder">MLGraphBuilder</a><span>, in § 7.6</span>
-   <li><a href="#dom-mlgraphbuilder-mlgraphbuilder">MLGraphBuilder(context)</a><span>, in § 7.6</span>
-   <li><a href="#dictdef-mlgrucelloptions">MLGruCellOptions</a><span>, in § 7.6.14</span>
-   <li><a href="#dictdef-mlgruoptions">MLGruOptions</a><span>, in § 7.6.13</span>
-   <li><a href="#enumdef-mlgruweightlayout">MLGruWeightLayout</a><span>, in § 7.6.13</span>
-   <li><a href="#dictdef-mlhardsigmoidoptions">MLHardSigmoidOptions</a><span>, in § 7.6.15</span>
+   <li><a href="#dom-mlgraphbuilder-mlgraphbuilder">MLGraphBuilder(context)</a><span>, in § 7.6.1</span>
+   <li><a href="#dictdef-mlgrucelloptions">MLGruCellOptions</a><span>, in § 7.6.15</span>
+   <li><a href="#dictdef-mlgruoptions">MLGruOptions</a><span>, in § 7.6.14</span>
+   <li><a href="#enumdef-mlgruweightlayout">MLGruWeightLayout</a><span>, in § 7.6.14</span>
+   <li><a href="#dictdef-mlhardsigmoidoptions">MLHardSigmoidOptions</a><span>, in § 7.6.16</span>
    <li><a href="#enumdef-mlinputoperandlayout">MLInputOperandLayout</a><span>, in § 7.3.1</span>
-   <li><a href="#dictdef-mlinstancenormalizationoptions">MLInstanceNormalizationOptions</a><span>, in § 7.6.17</span>
-   <li><a href="#enumdef-mlinterpolationmode">MLInterpolationMode</a><span>, in § 7.6.28</span>
-   <li><a href="#dictdef-mlleakyreluoptions">MLLeakyReluOptions</a><span>, in § 7.6.18</span>
-   <li><a href="#dictdef-mllinearoptions">MLLinearOptions</a><span>, in § 7.6.19</span>
-   <li><a href="#dictdef-mllstmcelloptions">MLLstmCellOptions</a><span>, in § 7.6.21</span>
-   <li><a href="#dictdef-mllstmoptions">MLLstmOptions</a><span>, in § 7.6.20</span>
-   <li><a href="#enumdef-mllstmweightlayout">MLLstmWeightLayout</a><span>, in § 7.6.20</span>
+   <li><a href="#dictdef-mlinstancenormalizationoptions">MLInstanceNormalizationOptions</a><span>, in § 7.6.18</span>
+   <li><a href="#enumdef-mlinterpolationmode">MLInterpolationMode</a><span>, in § 7.6.29</span>
+   <li><a href="#dictdef-mlleakyreluoptions">MLLeakyReluOptions</a><span>, in § 7.6.19</span>
+   <li><a href="#dictdef-mllinearoptions">MLLinearOptions</a><span>, in § 7.6.20</span>
+   <li><a href="#dictdef-mllstmcelloptions">MLLstmCellOptions</a><span>, in § 7.6.22</span>
+   <li><a href="#dictdef-mllstmoptions">MLLstmOptions</a><span>, in § 7.6.21</span>
+   <li><a href="#enumdef-mllstmweightlayout">MLLstmWeightLayout</a><span>, in § 7.6.21</span>
    <li><a href="#typedefdef-mlnamedarraybufferviews">MLNamedArrayBufferViews</a><span>, in § 7.4</span>
    <li><a href="#typedefdef-mlnamedgpuresources">MLNamedGPUResources</a><span>, in § 7.5</span>
    <li><a href="#typedefdef-mlnamedoperands">MLNamedOperands</a><span>, in § 7.6</span>
    <li><a href="#mloperand">MLOperand</a><span>, in § 7.3.2</span>
    <li><a href="#dictdef-mloperanddescriptor">MLOperandDescriptor</a><span>, in § 7.3.1</span>
    <li><a href="#enumdef-mloperandtype">MLOperandType</a><span>, in § 7.3.1</span>
-   <li><a href="#enumdef-mlpaddingmode">MLPaddingMode</a><span>, in § 7.6.23</span>
-   <li><a href="#dictdef-mlpadoptions">MLPadOptions</a><span>, in § 7.6.23</span>
-   <li><a href="#dictdef-mlpool2doptions">MLPool2dOptions</a><span>, in § 7.6.24</span>
+   <li><a href="#enumdef-mlpaddingmode">MLPaddingMode</a><span>, in § 7.6.24</span>
+   <li><a href="#dictdef-mlpadoptions">MLPadOptions</a><span>, in § 7.6.24</span>
+   <li><a href="#dictdef-mlpool2doptions">MLPool2dOptions</a><span>, in § 7.6.25</span>
    <li><a href="#enumdef-mlpowerpreference">MLPowerPreference</a><span>, in § 7.2</span>
-   <li><a href="#enumdef-mlrecurrentnetworkdirection">MLRecurrentNetworkDirection</a><span>, in § 7.6.13</span>
-   <li><a href="#dictdef-mlreduceoptions">MLReduceOptions</a><span>, in § 7.6.26</span>
-   <li><a href="#dictdef-mlresample2doptions">MLResample2dOptions</a><span>, in § 7.6.28</span>
-   <li><a href="#enumdef-mlroundingtype">MLRoundingType</a><span>, in § 7.6.24</span>
-   <li><a href="#dictdef-mlsoftplusoptions">MLSoftplusOptions</a><span>, in § 7.6.33</span>
-   <li><a href="#dictdef-mlsplitoptions">MLSplitOptions</a><span>, in § 7.6.35</span>
-   <li><a href="#dictdef-mlsqueezeoptions">MLSqueezeOptions</a><span>, in § 7.6.36</span>
-   <li><a href="#dictdef-mltransposeoptions">MLTransposeOptions</a><span>, in § 7.6.38</span>
+   <li><a href="#enumdef-mlrecurrentnetworkdirection">MLRecurrentNetworkDirection</a><span>, in § 7.6.14</span>
+   <li><a href="#dictdef-mlreduceoptions">MLReduceOptions</a><span>, in § 7.6.27</span>
+   <li><a href="#dictdef-mlresample2doptions">MLResample2dOptions</a><span>, in § 7.6.29</span>
+   <li><a href="#enumdef-mlroundingtype">MLRoundingType</a><span>, in § 7.6.25</span>
+   <li><a href="#dictdef-mlsoftplusoptions">MLSoftplusOptions</a><span>, in § 7.6.34</span>
+   <li><a href="#dictdef-mlsplitoptions">MLSplitOptions</a><span>, in § 7.6.36</span>
+   <li><a href="#dictdef-mlsqueezeoptions">MLSqueezeOptions</a><span>, in § 7.6.37</span>
+   <li><a href="#dictdef-mltransposeoptions">MLTransposeOptions</a><span>, in § 7.6.39</span>
    <li>
     mode
     <ul>
-     <li><a href="#dom-mlpadoptions-mode">dict-member for MLPadOptions</a><span>, in § 7.6.23</span>
-     <li><a href="#dom-mlresample2doptions-mode">dict-member for MLResample2dOptions</a><span>, in § 7.6.28</span>
+     <li><a href="#dom-mlpadoptions-mode">dict-member for MLPadOptions</a><span>, in § 7.6.24</span>
+     <li><a href="#dom-mlresample2doptions-mode">dict-member for MLResample2dOptions</a><span>, in § 7.6.29</span>
+    </ul>
+   <li><a href="#dom-mlgraphbuilder-mul">mul(a, b)</a><span>, in § 7.6.10</span>
+   <li>
+    [[name]]
+    <ul>
+     <li><a href="#dom-mlactivation-name-slot">attribute for MLActivation</a><span>, in § 7.3.3</span>
+     <li><a href="#dom-mloperand-name-slot">attribute for MLOperand</a><span>, in § 7.3.2</span>
     </ul>
-   <li><a href="#dom-mlgraphbuilder-mul">mul(a, b)</a><span>, in § 7.6.9</span>
-   <li><a href="#dom-mloperand-name-slot">[[name]]</a><span>, in § 7.3.2</span>
    <li><a href="#navigatorml">NavigatorML</a><span>, in § 7.1</span>
    <li><a href="#dom-mlinputoperandlayout-nchw">"nchw"</a><span>, in § 7.3.1</span>
-   <li><a href="#dom-mlinterpolationmode-nearest-neighbor">"nearest-neighbor"</a><span>, in § 7.6.28</span>
-   <li><a href="#dom-mlgraphbuilder-neg">neg(x)</a><span>, in § 7.6.10</span>
+   <li><a href="#dom-mlinterpolationmode-nearest-neighbor">"nearest-neighbor"</a><span>, in § 7.6.29</span>
+   <li><a href="#dom-mlgraphbuilder-neg">neg(input)</a><span>, in § 7.6.11</span>
    <li><a href="#dom-mlinputoperandlayout-nhwc">"nhwc"</a><span>, in § 7.3.1</span>
    <li><a href="#dom-mlbufferresourceview-offset">offset</a><span>, in § 7.6</span>
    <li>
     "ohwi"
     <ul>
-     <li><a href="#dom-mlconv2dfilteroperandlayout-ohwi">enum-value for MLConv2dFilterOperandLayout</a><span>, in § 7.6.7</span>
-     <li><a href="#dom-mlconvtranspose2dfilteroperandlayout-ohwi">enum-value for MLConvTranspose2dFilterOperandLayout</a><span>, in § 7.6.8</span>
+     <li><a href="#dom-mlconv2dfilteroperandlayout-ohwi">enum-value for MLConv2dFilterOperandLayout</a><span>, in § 7.6.8</span>
+     <li><a href="#dom-mlconvtranspose2dfilteroperandlayout-ohwi">enum-value for MLConvTranspose2dFilterOperandLayout</a><span>, in § 7.6.9</span>
     </ul>
-   <li><a href="#dom-mlconv2dfilteroperandlayout-oihw">"oihw"</a><span>, in § 7.6.7</span>
+   <li><a href="#dom-mlconv2dfilteroperandlayout-oihw">"oihw"</a><span>, in § 7.6.8</span>
    <li><a href="#dom-mloperand-operand-slot">[[operand]]</a><span>, in § 7.3.2</span>
-   <li><a href="#dom-mloperand-operator-slot">[[operator]]</a><span>, in § 7.3.2</span>
+   <li>
+    [[operator]]
+    <ul>
+     <li><a href="#dom-mlactivation-operator-slot">attribute for MLActivation</a><span>, in § 7.3.3</span>
+     <li><a href="#dom-mloperand-operator-slot">attribute for MLOperand</a><span>, in § 7.3.2</span>
+    </ul>
+   <li><a href="#dom-mlactivation-options-slot">[[options]]</a><span>, in § 7.3.3</span>
    <li><a href="#dom-mlgraph-outputdescriptors-slot">[[outputDescriptors]]</a><span>, in § 7.3</span>
-   <li><a href="#dom-mlconvtranspose2doptions-outputpadding">outputPadding</a><span>, in § 7.6.8</span>
+   <li><a href="#dom-mlconvtranspose2doptions-outputpadding">outputPadding</a><span>, in § 7.6.9</span>
    <li><a href="#dom-mlcomputeresult-outputs">outputs</a><span>, in § 7.4.4</span>
    <li>
     outputSizes
     <ul>
-     <li><a href="#dom-mlconvtranspose2doptions-outputsizes">dict-member for MLConvTranspose2dOptions</a><span>, in § 7.6.8</span>
-     <li><a href="#dom-mlpool2doptions-outputsizes">dict-member for MLPool2dOptions</a><span>, in § 7.6.24</span>
+     <li><a href="#dom-mlconvtranspose2doptions-outputsizes">dict-member for MLConvTranspose2dOptions</a><span>, in § 7.6.9</span>
+     <li><a href="#dom-mlpool2doptions-outputsizes">dict-member for MLPool2dOptions</a><span>, in § 7.6.25</span>
     </ul>
    <li>
     padding
     <ul>
-     <li><a href="#dom-mlconv2doptions-padding">dict-member for MLConv2dOptions</a><span>, in § 7.6.7</span>
-     <li><a href="#dom-mlconvtranspose2doptions-padding">dict-member for MLConvTranspose2dOptions</a><span>, in § 7.6.8</span>
-     <li><a href="#dom-mlpool2doptions-padding">dict-member for MLPool2dOptions</a><span>, in § 7.6.24</span>
+     <li><a href="#dom-mlconv2doptions-padding">dict-member for MLConv2dOptions</a><span>, in § 7.6.8</span>
+     <li><a href="#dom-mlconvtranspose2doptions-padding">dict-member for MLConvTranspose2dOptions</a><span>, in § 7.6.9</span>
+     <li><a href="#dom-mlpool2doptions-padding">dict-member for MLPool2dOptions</a><span>, in § 7.6.25</span>
     </ul>
-   <li><a href="#dom-mlgraphbuilder-pad">pad(input, beginningPadding, endingPadding)</a><span>, in § 7.6.23</span>
-   <li><a href="#dom-mlgraphbuilder-pad">pad(input, beginningPadding, endingPadding, options)</a><span>, in § 7.6.23</span>
+   <li><a href="#dom-mlgraphbuilder-pad">pad(input, beginningPadding, endingPadding)</a><span>, in § 7.6.24</span>
+   <li><a href="#dom-mlgraphbuilder-pad">pad(input, beginningPadding, endingPadding, options)</a><span>, in § 7.6.24</span>
    <li>
     peepholeWeight
     <ul>
-     <li><a href="#dom-mllstmcelloptions-peepholeweight">dict-member for MLLstmCellOptions</a><span>, in § 7.6.21</span>
-     <li><a href="#dom-mllstmoptions-peepholeweight">dict-member for MLLstmOptions</a><span>, in § 7.6.20</span>
+     <li><a href="#dom-mllstmcelloptions-peepholeweight">dict-member for MLLstmCellOptions</a><span>, in § 7.6.22</span>
+     <li><a href="#dom-mllstmoptions-peepholeweight">dict-member for MLLstmOptions</a><span>, in § 7.6.21</span>
     </ul>
-   <li><a href="#dom-mltransposeoptions-permutation">permutation</a><span>, in § 7.6.38</span>
-   <li><a href="#dom-mlgraphbuilder-pow">pow(a, b)</a><span>, in § 7.6.9</span>
+   <li><a href="#dom-mltransposeoptions-permutation">permutation</a><span>, in § 7.6.39</span>
+   <li><a href="#mlgraphbuilder-pooling-op">pooling-op</a><span>, in § 7.6.25</span>
+   <li><a href="#dom-mlgraphbuilder-pow">pow(a, b)</a><span>, in § 7.6.10</span>
    <li><a href="#dom-mlcontext-powerpreference-slot">[[powerPreference]]</a><span>, in § 7.4</span>
    <li><a href="#power-preference">power preference</a><span>, in § 7.4</span>
    <li><a href="#dom-mlcontextoptions-powerpreference">powerPreference</a><span>, in § 7.2</span>
    <li><a href="#power-preference-default">power-preference-default</a><span>, in § 7.4</span>
    <li><a href="#power-preference-high-performance">power-preference-high-performance</a><span>, in § 7.4</span>
    <li><a href="#power-preference-low-power">power-preference-low-power</a><span>, in § 7.4</span>
-   <li><a href="#dom-mlgraphbuilder-prelu">prelu(x, slope)</a><span>, in § 7.6.25</span>
+   <li><a href="#dom-mlgraphbuilder-prelu">prelu(input, slope)</a><span>, in § 7.6.26</span>
+   <li><a href="#mloperand-rank">rank</a><span>, in § 7.3.2</span>
    <li>
     recurrentBias
     <ul>
-     <li><a href="#dom-mlgrucelloptions-recurrentbias">dict-member for MLGruCellOptions</a><span>, in § 7.6.14</span>
-     <li><a href="#dom-mlgruoptions-recurrentbias">dict-member for MLGruOptions</a><span>, in § 7.6.13</span>
-     <li><a href="#dom-mllstmcelloptions-recurrentbias">dict-member for MLLstmCellOptions</a><span>, in § 7.6.21</span>
-     <li><a href="#dom-mllstmoptions-recurrentbias">dict-member for MLLstmOptions</a><span>, in § 7.6.20</span>
+     <li><a href="#dom-mlgrucelloptions-recurrentbias">dict-member for MLGruCellOptions</a><span>, in § 7.6.15</span>
+     <li><a href="#dom-mlgruoptions-recurrentbias">dict-member for MLGruOptions</a><span>, in § 7.6.14</span>
+     <li><a href="#dom-mllstmcelloptions-recurrentbias">dict-member for MLLstmCellOptions</a><span>, in § 7.6.22</span>
+     <li><a href="#dom-mllstmoptions-recurrentbias">dict-member for MLLstmOptions</a><span>, in § 7.6.21</span>
     </ul>
-   <li><a href="#dom-mlgraphbuilder-reducel1">reduceL1(input)</a><span>, in § 7.6.26</span>
-   <li><a href="#dom-mlgraphbuilder-reducel1">reduceL1(input, options)</a><span>, in § 7.6.26</span>
-   <li><a href="#dom-mlgraphbuilder-reducel2">reduceL2(input)</a><span>, in § 7.6.26</span>
-   <li><a href="#dom-mlgraphbuilder-reducel2">reduceL2(input, options)</a><span>, in § 7.6.26</span>
-   <li><a href="#dom-mlgraphbuilder-reducelogsumexp">reduceLogSumExp(input)</a><span>, in § 7.6.26</span>
-   <li><a href="#dom-mlgraphbuilder-reducelogsumexp">reduceLogSumExp(input, options)</a><span>, in § 7.6.26</span>
-   <li><a href="#dom-mlgraphbuilder-reducelogsum">reduceLogSum(input)</a><span>, in § 7.6.26</span>
-   <li><a href="#dom-mlgraphbuilder-reducelogsum">reduceLogSum(input, options)</a><span>, in § 7.6.26</span>
-   <li><a href="#dom-mlgraphbuilder-reducemax">reduceMax(input)</a><span>, in § 7.6.26</span>
-   <li><a href="#dom-mlgraphbuilder-reducemax">reduceMax(input, options)</a><span>, in § 7.6.26</span>
-   <li><a href="#dom-mlgraphbuilder-reducemean">reduceMean(input)</a><span>, in § 7.6.26</span>
-   <li><a href="#dom-mlgraphbuilder-reducemean">reduceMean(input, options)</a><span>, in § 7.6.26</span>
-   <li><a href="#dom-mlgraphbuilder-reducemin">reduceMin(input)</a><span>, in § 7.6.26</span>
-   <li><a href="#dom-mlgraphbuilder-reducemin">reduceMin(input, options)</a><span>, in § 7.6.26</span>
-   <li><a href="#dom-mlgraphbuilder-reduceproduct">reduceProduct(input)</a><span>, in § 7.6.26</span>
-   <li><a href="#dom-mlgraphbuilder-reduceproduct">reduceProduct(input, options)</a><span>, in § 7.6.26</span>
-   <li><a href="#dom-mlgraphbuilder-reducesum">reduceSum(input)</a><span>, in § 7.6.26</span>
-   <li><a href="#dom-mlgraphbuilder-reducesum">reduceSum(input, options)</a><span>, in § 7.6.26</span>
-   <li><a href="#dom-mlgraphbuilder-reducesumsquare">reduceSumSquare(input)</a><span>, in § 7.6.26</span>
-   <li><a href="#dom-mlgraphbuilder-reducesumsquare">reduceSumSquare(input, options)</a><span>, in § 7.6.26</span>
-   <li><a href="#dom-mlpaddingmode-reflection">"reflection"</a><span>, in § 7.6.23</span>
-   <li><a href="#dom-mlgraphbuilder-relu①">relu()</a><span>, in § 7.6.27</span>
-   <li><a href="#dom-mlgraphbuilder-relu">relu(x)</a><span>, in § 7.6.27</span>
-   <li><a href="#dom-mlgraphbuilder-resample2d">resample2d(input)</a><span>, in § 7.6.28</span>
-   <li><a href="#dom-mlgraphbuilder-resample2d">resample2d(input, options)</a><span>, in § 7.6.28</span>
+   <li><a href="#dom-mlgraphbuilder-reducel1">reduceL1(input)</a><span>, in § 7.6.27</span>
+   <li><a href="#dom-mlgraphbuilder-reducel1">reduceL1(input, options)</a><span>, in § 7.6.27</span>
+   <li><a href="#dom-mlgraphbuilder-reducel2">reduceL2(input)</a><span>, in § 7.6.27</span>
+   <li><a href="#dom-mlgraphbuilder-reducel2">reduceL2(input, options)</a><span>, in § 7.6.27</span>
+   <li><a href="#dom-mlgraphbuilder-reducelogsumexp">reduceLogSumExp(input)</a><span>, in § 7.6.27</span>
+   <li><a href="#dom-mlgraphbuilder-reducelogsumexp">reduceLogSumExp(input, options)</a><span>, in § 7.6.27</span>
+   <li><a href="#dom-mlgraphbuilder-reducelogsum">reduceLogSum(input)</a><span>, in § 7.6.27</span>
+   <li><a href="#dom-mlgraphbuilder-reducelogsum">reduceLogSum(input, options)</a><span>, in § 7.6.27</span>
+   <li><a href="#dom-mlgraphbuilder-reducemax">reduceMax(input)</a><span>, in § 7.6.27</span>
+   <li><a href="#dom-mlgraphbuilder-reducemax">reduceMax(input, options)</a><span>, in § 7.6.27</span>
+   <li><a href="#dom-mlgraphbuilder-reducemean">reduceMean(input)</a><span>, in § 7.6.27</span>
+   <li><a href="#dom-mlgraphbuilder-reducemean">reduceMean(input, options)</a><span>, in § 7.6.27</span>
+   <li><a href="#dom-mlgraphbuilder-reducemin">reduceMin(input)</a><span>, in § 7.6.27</span>
+   <li><a href="#dom-mlgraphbuilder-reducemin">reduceMin(input, options)</a><span>, in § 7.6.27</span>
+   <li><a href="#mlgraphbuilder-reduce-op">reduce-op</a><span>, in § 7.6.27</span>
+   <li><a href="#dom-mlgraphbuilder-reduceproduct">reduceProduct(input)</a><span>, in § 7.6.27</span>
+   <li><a href="#dom-mlgraphbuilder-reduceproduct">reduceProduct(input, options)</a><span>, in § 7.6.27</span>
+   <li><a href="#dom-mlgraphbuilder-reducesum">reduceSum(input)</a><span>, in § 7.6.27</span>
+   <li><a href="#dom-mlgraphbuilder-reducesum">reduceSum(input, options)</a><span>, in § 7.6.27</span>
+   <li><a href="#dom-mlgraphbuilder-reducesumsquare">reduceSumSquare(input)</a><span>, in § 7.6.27</span>
+   <li><a href="#dom-mlgraphbuilder-reducesumsquare">reduceSumSquare(input, options)</a><span>, in § 7.6.27</span>
+   <li><a href="#dom-mlpaddingmode-reflection">"reflection"</a><span>, in § 7.6.24</span>
+   <li><a href="#relu-noargs">relu()</a><span>, in § 7.6.28.2</span>
+   <li><a href="#dom-mlgraphbuilder-relu">relu(input)</a><span>, in § 7.6.28.1</span>
+   <li><a href="#dom-mlgraphbuilder-resample2d">resample2d(input)</a><span>, in § 7.6.29</span>
+   <li><a href="#dom-mlgraphbuilder-resample2d">resample2d(input, options)</a><span>, in § 7.6.29</span>
+   <li><a href="#mlgraphbuilder-resample-output-sizes">resample output sizes</a><span>, in § 7.6.29</span>
    <li>
     resetAfter
     <ul>
-     <li><a href="#dom-mlgrucelloptions-resetafter">dict-member for MLGruCellOptions</a><span>, in § 7.6.14</span>
-     <li><a href="#dom-mlgruoptions-resetafter">dict-member for MLGruOptions</a><span>, in § 7.6.13</span>
+     <li><a href="#dom-mlgrucelloptions-resetafter">dict-member for MLGruCellOptions</a><span>, in § 7.6.15</span>
+     <li><a href="#dom-mlgruoptions-resetafter">dict-member for MLGruOptions</a><span>, in § 7.6.14</span>
     </ul>
-   <li><a href="#dom-mlgraphbuilder-reshape">reshape(input, newShape)</a><span>, in § 7.6.29</span>
+   <li><a href="#dom-mlgraphbuilder-reshape">reshape(input, newShape)</a><span>, in § 7.6.30</span>
    <li><a href="#dom-mlbufferresourceview-resource">resource</a><span>, in § 7.6</span>
    <li>
     returnSequence
     <ul>
-     <li><a href="#dom-mlgruoptions-returnsequence">dict-member for MLGruOptions</a><span>, in § 7.6.13</span>
-     <li><a href="#dom-mllstmoptions-returnsequence">dict-member for MLLstmOptions</a><span>, in § 7.6.20</span>
+     <li><a href="#dom-mlgruoptions-returnsequence">dict-member for MLGruOptions</a><span>, in § 7.6.14</span>
+     <li><a href="#dom-mllstmoptions-returnsequence">dict-member for MLLstmOptions</a><span>, in § 7.6.21</span>
     </ul>
-   <li><a href="#dom-mlpool2doptions-roundingtype">roundingType</a><span>, in § 7.6.24</span>
-   <li><a href="#dom-mlgruweightlayout-rzn">"rzn"</a><span>, in § 7.6.13</span>
-   <li><a href="#dom-mlautopad-same-lower">"same-lower"</a><span>, in § 7.6.7</span>
-   <li><a href="#dom-mlautopad-same-upper">"same-upper"</a><span>, in § 7.6.7</span>
+   <li><a href="#dom-mlpool2doptions-roundingtype">roundingType</a><span>, in § 7.6.25</span>
+   <li><a href="#dom-mlgruweightlayout-rzn">"rzn"</a><span>, in § 7.6.14</span>
+   <li><a href="#dom-mlautopad-same-lower">"same-lower"</a><span>, in § 7.6.8</span>
+   <li><a href="#dom-mlautopad-same-upper">"same-upper"</a><span>, in § 7.6.8</span>
    <li>
     scale
     <ul>
-     <li><a href="#dom-mlbatchnormalizationoptions-scale">dict-member for MLBatchNormalizationOptions</a><span>, in § 7.6.4</span>
-     <li><a href="#dom-mlinstancenormalizationoptions-scale">dict-member for MLInstanceNormalizationOptions</a><span>, in § 7.6.17</span>
+     <li><a href="#dom-mlbatchnormalizationoptions-scale">dict-member for MLBatchNormalizationOptions</a><span>, in § 7.6.5</span>
+     <li><a href="#dom-mlinstancenormalizationoptions-scale">dict-member for MLInstanceNormalizationOptions</a><span>, in § 7.6.18</span>
     </ul>
-   <li><a href="#dom-mlresample2doptions-scales">scales</a><span>, in § 7.6.28</span>
-   <li><a href="#dom-mlgraphbuilder-sigmoid①">sigmoid()</a><span>, in § 7.6.30</span>
-   <li><a href="#dom-mlgraphbuilder-sigmoid">sigmoid(x)</a><span>, in § 7.6.30</span>
-   <li><a href="#dom-mlgraphbuilder-sin">sin(x)</a><span>, in § 7.6.10</span>
+   <li><a href="#dom-mlresample2doptions-scales">scales</a><span>, in § 7.6.29</span>
+   <li><a href="#sigmoid-noargs">sigmoid()</a><span>, in § 7.6.31.2</span>
+   <li><a href="#dom-mlgraphbuilder-sigmoid">sigmoid(input)</a><span>, in § 7.6.31.1</span>
+   <li><a href="#dom-mlgraphbuilder-sin">sin(input)</a><span>, in § 7.6.11</span>
    <li><a href="#dom-mlbufferresourceview-size">size</a><span>, in § 7.6</span>
-   <li><a href="#dom-mlresample2doptions-sizes">sizes</a><span>, in § 7.6.28</span>
-   <li><a href="#dom-mlgraphbuilder-slice">slice(input, starts, sizes)</a><span>, in § 7.6.31</span>
-   <li><a href="#dom-mlgraphbuilder-softmax①">softmax()</a><span>, in § 7.6.32</span>
-   <li><a href="#dom-mlgraphbuilder-softmax">softmax(x)</a><span>, in § 7.6.32</span>
-   <li><a href="#dom-mlgraphbuilder-softplus-options">softplus()</a><span>, in § 7.6.33</span>
-   <li><a href="#dom-mlgraphbuilder-softplus-options">softplus(options)</a><span>, in § 7.6.33</span>
-   <li><a href="#dom-mlgraphbuilder-softplus">softplus(x)</a><span>, in § 7.6.33</span>
-   <li><a href="#dom-mlgraphbuilder-softplus">softplus(x, options)</a><span>, in § 7.6.33</span>
-   <li><a href="#dom-mlgraphbuilder-softsign①">softsign()</a><span>, in § 7.6.34</span>
-   <li><a href="#dom-mlgraphbuilder-softsign">softsign(x)</a><span>, in § 7.6.34</span>
-   <li><a href="#dom-mlgraphbuilder-split">split(input, splits)</a><span>, in § 7.6.35</span>
-   <li><a href="#dom-mlgraphbuilder-split">split(input, splits, options)</a><span>, in § 7.6.35</span>
-   <li><a href="#dom-mlgraphbuilder-squeeze">squeeze(input)</a><span>, in § 7.6.36</span>
-   <li><a href="#dom-mlgraphbuilder-squeeze">squeeze(input, options)</a><span>, in § 7.6.36</span>
-   <li><a href="#dom-mlsoftplusoptions-steepness">steepness</a><span>, in § 7.6.33</span>
+   <li><a href="#dom-mlresample2doptions-sizes">sizes</a><span>, in § 7.6.29</span>
+   <li><a href="#dom-mlgraphbuilder-slice">slice(input, starts, sizes)</a><span>, in § 7.6.32</span>
+   <li><a href="#softmax-noargs">softmax()</a><span>, in § 7.6.33.2</span>
+   <li><a href="#dom-mlgraphbuilder-softmax">softmax(input)</a><span>, in § 7.6.33.1</span>
+   <li><a href="#dom-mlgraphbuilder-softplus-options">softplus()</a><span>, in § 7.6.34.2</span>
+   <li><a href="#dom-mlgraphbuilder-softplus">softplus(input)</a><span>, in § 7.6.34.1</span>
+   <li><a href="#dom-mlgraphbuilder-softplus">softplus(input, options)</a><span>, in § 7.6.34.1</span>
+   <li><a href="#dom-mlgraphbuilder-softplus-options">softplus(options)</a><span>, in § 7.6.34.2</span>
+   <li><a href="#softsign-noargs">softsign()</a><span>, in § 7.6.35.2</span>
+   <li><a href="#dom-mlgraphbuilder-softsign">softsign(input)</a><span>, in § 7.6.35.1</span>
+   <li><a href="#dom-mlgraphbuilder-split">split(input, splits)</a><span>, in § 7.6.36</span>
+   <li><a href="#dom-mlgraphbuilder-split">split(input, splits, options)</a><span>, in § 7.6.36</span>
+   <li><a href="#dom-mlgraphbuilder-squeeze">squeeze(input)</a><span>, in § 7.6.37</span>
+   <li><a href="#dom-mlgraphbuilder-squeeze">squeeze(input, options)</a><span>, in § 7.6.37</span>
+   <li><a href="#dom-mlsoftplusoptions-steepness">steepness</a><span>, in § 7.6.34</span>
    <li>
     strides
     <ul>
-     <li><a href="#dom-mlconv2doptions-strides">dict-member for MLConv2dOptions</a><span>, in § 7.6.7</span>
-     <li><a href="#dom-mlconvtranspose2doptions-strides">dict-member for MLConvTranspose2dOptions</a><span>, in § 7.6.8</span>
-     <li><a href="#dom-mlpool2doptions-strides">dict-member for MLPool2dOptions</a><span>, in § 7.6.24</span>
+     <li><a href="#dom-mlconv2doptions-strides">dict-member for MLConv2dOptions</a><span>, in § 7.6.8</span>
+     <li><a href="#dom-mlconvtranspose2doptions-strides">dict-member for MLConvTranspose2dOptions</a><span>, in § 7.6.9</span>
+     <li><a href="#dom-mlpool2doptions-strides">dict-member for MLPool2dOptions</a><span>, in § 7.6.25</span>
     </ul>
-   <li><a href="#dom-mlgraphbuilder-sub">sub(a, b)</a><span>, in § 7.6.9</span>
-   <li><a href="#dom-mlpaddingmode-symmetric">"symmetric"</a><span>, in § 7.6.23</span>
-   <li><a href="#dom-mlgraphbuilder-tanh①">tanh()</a><span>, in § 7.6.37</span>
-   <li><a href="#dom-mlgraphbuilder-tanh">tanh(x)</a><span>, in § 7.6.37</span>
-   <li><a href="#dom-mlgraphbuilder-tan">tan(x)</a><span>, in § 7.6.10</span>
+   <li><a href="#dom-mlgraphbuilder-sub">sub(a, b)</a><span>, in § 7.6.10</span>
+   <li><a href="#dom-mlpaddingmode-symmetric">"symmetric"</a><span>, in § 7.6.24</span>
+   <li><a href="#tanh-noargs">tanh()</a><span>, in § 7.6.38.2</span>
+   <li><a href="#dom-mlgraphbuilder-tanh">tanh(input)</a><span>, in § 7.6.38.1</span>
+   <li><a href="#dom-mlgraphbuilder-tan">tan(input)</a><span>, in § 7.6.11</span>
    <li><a href="#mlnamedarraybufferviews-transfer">transfer</a><span>, in § 7.4.3</span>
-   <li><a href="#dom-mlgraphbuilder-transpose">transpose(input)</a><span>, in § 7.6.38</span>
-   <li><a href="#dom-mlgraphbuilder-transpose">transpose(input, options)</a><span>, in § 7.6.38</span>
+   <li><a href="#dom-mlgraphbuilder-transpose">transpose(input)</a><span>, in § 7.6.39</span>
+   <li><a href="#dom-mlgraphbuilder-transpose">transpose(input, options)</a><span>, in § 7.6.39</span>
    <li><a href="#dom-mloperanddescriptor-type">type</a><span>, in § 7.3.1</span>
    <li><a href="#dom-mloperandtype-uint32">"uint32"</a><span>, in § 7.3.1</span>
    <li><a href="#dom-mloperandtype-uint8">"uint8"</a><span>, in § 7.3.1</span>
    <li><a href="#validate-buffer-with-descriptor">validate buffer with descriptor</a><span>, in § 7.4.2</span>
    <li><a href="#validate-graph-resources">validate graph resources</a><span>, in § 7.4.2</span>
    <li><a href="#validate-mlcontext">validate MLContext</a><span>, in § 7.4.1</span>
-   <li><a href="#dom-mlpadoptions-value">value</a><span>, in § 7.6.23</span>
+   <li><a href="#mloperand-validate-mloperand">validate MLOperand</a><span>, in § 7.3.2.1</span>
+   <li><a href="#dom-mlpadoptions-value">value</a><span>, in § 7.6.24</span>
    <li><a href="#webgpu-context">webgpu-context</a><span>, in § 7.4</span>
    <li><a href="#webnn-feature">webnn-feature</a><span>, in § 7.2.1</span>
-   <li><a href="#dom-mlpool2doptions-windowdimensions">windowDimensions</a><span>, in § 7.6.24</span>
-   <li><a href="#dom-mlgruweightlayout-zrn">"zrn"</a><span>, in § 7.6.13</span>
+   <li><a href="#dom-mlpool2doptions-windowdimensions">windowDimensions</a><span>, in § 7.6.25</span>
+   <li><a href="#dom-mlgruweightlayout-zrn">"zrn"</a><span>, in § 7.6.14</span>
   </ul>
   <h3 class="no-num no-ref heading settled" id="index-defined-elsewhere"><span class="content">Terms defined by reference</span><a class="self-link" href="#index-defined-elsewhere"></a></h3>
   <ul class="index">
@@ -4794,7 +8199,7 @@ <h3 class="no-num no-ref heading settled" id="index-defined-elsewhere"><span cla
    <li>
     <a data-link-type="biblio">[ECMASCRIPT]</a> defines the following terms:
     <ul>
-     <li><span class="dfn-paneled" id="134d399c">construct</span>
+     <li><span class="dfn-paneled" id="944206e4">Construct</span>
      <li><span class="dfn-paneled" id="e63ec40a">element size</span>
      <li><span class="dfn-paneled" id="38d90164">element type</span>
      <li><span class="dfn-paneled" id="023f906d">view constructor</span>
@@ -4826,8 +8231,11 @@ <h3 class="no-num no-ref heading settled" id="index-defined-elsewhere"><span cla
     <ul>
      <li><span class="dfn-paneled" id="77b4c09a">assert</span>
      <li><span class="dfn-paneled" id="1243a891">exist</span>
+     <li><span class="dfn-paneled" id="45209803">for each</span>
      <li><span class="dfn-paneled" id="860300d4">implementation-defined</span>
+     <li><span class="dfn-paneled" id="0204d188">size</span>
      <li><span class="dfn-paneled" id="0698d556">string</span>
+     <li><span class="dfn-paneled" id="febebe0e">the range</span>
     </ul>
    <li>
     <a data-link-type="biblio">[PERMISSIONS-POLICY-1]</a> defines the following terms:
@@ -4838,12 +8246,12 @@ <h3 class="no-num no-ref heading settled" id="index-defined-elsewhere"><span cla
    <li>
     <a data-link-type="biblio">[WEBGPU]</a> defines the following terms:
     <ul>
-     <li><span class="dfn-paneled" id="4acc25f4">GPUBuffer</span>
-     <li><span class="dfn-paneled" id="12d473bb">GPUCommandBuffer</span>
+     <li><span class="dfn-paneled" id="be5908a3">GPUBuffer</span>
+     <li><span class="dfn-paneled" id="819f210c">GPUCommandBuffer</span>
      <li><span class="dfn-paneled" id="30267b02">GPUCommandBufferDescriptor</span>
-     <li><span class="dfn-paneled" id="09adced4">GPUDevice</span>
-     <li><span class="dfn-paneled" id="2e08cf49">GPUQueue</span>
-     <li><span class="dfn-paneled" id="4abf7301">GPUTexture</span>
+     <li><span class="dfn-paneled" id="189809f1">GPUDevice</span>
+     <li><span class="dfn-paneled" id="899c078f">GPUQueue</span>
+     <li><span class="dfn-paneled" id="4b3142af">GPUTexture</span>
      <li><span class="dfn-paneled" id="49f6fafc">size</span>
     </ul>
    <li>
@@ -4869,18 +8277,19 @@ <h3 class="no-num no-ref heading settled" id="index-defined-elsewhere"><span cla
      <li><span class="dfn-paneled" id="95d7775a">Uint8Array</span>
      <li><span class="dfn-paneled" id="dacde8b5">a new promise</span>
      <li><span class="dfn-paneled" id="5372cca8">boolean</span>
-     <li><span class="dfn-paneled" id="e9acf661">buffer byte length</span>
+     <li><span class="dfn-paneled" id="495737df">byte length</span>
      <li><span class="dfn-paneled" id="8c800cdf">double</span>
      <li><span class="dfn-paneled" id="eec1be2d">float</span>
      <li><span class="dfn-paneled" id="7b05fe24">get a copy of the bytes held by the buffer source</span>
-     <li><span class="dfn-paneled" id="a32c65d4">implements</span>
+     <li><span class="dfn-paneled" id="f8de33a3">long</span>
      <li><span class="dfn-paneled" id="56f81a8e">new</span>
-     <li><span class="dfn-paneled" id="aedfadcb">nullable</span>
+     <li><span class="dfn-paneled" id="51978e94">nullable type</span>
      <li><span class="dfn-paneled" id="7787d359">record</span>
      <li><span class="dfn-paneled" id="b262501e">reject</span>
      <li><span class="dfn-paneled" id="3b90bdcd">resolve</span>
      <li><span class="dfn-paneled" id="9cce47fd">sequence</span>
      <li><span class="dfn-paneled" id="4013a022">this</span>
+     <li><span class="dfn-paneled" id="b4cfa5ce">throw</span>
      <li><span class="dfn-paneled" id="69c80471">transfer</span>
      <li><span class="dfn-paneled" id="5f90bbfb">undefined</span>
      <li><span class="dfn-paneled" id="f32e046f">underlying buffer</span>
@@ -4909,7 +8318,7 @@ <h3 class="no-num no-ref heading settled" id="normative"><span class="content">N
    <dt id="biblio-rfc2119">[RFC2119]
    <dd>S. Bradner. <a href="https://datatracker.ietf.org/doc/html/rfc2119"><cite>Key words for use in RFCs to Indicate Requirement Levels</cite></a>. March 1997. Best Current Practice. URL: <a href="https://datatracker.ietf.org/doc/html/rfc2119">https://datatracker.ietf.org/doc/html/rfc2119</a>
    <dt id="biblio-webgpu">[WEBGPU]
-   <dd>Dzmitry Malyshau; Kai Ninomiya. <a href="https://gpuweb.github.io/gpuweb/"><cite>WebGPU</cite></a>. ED. URL: <a href="https://gpuweb.github.io/gpuweb/">https://gpuweb.github.io/gpuweb/</a>
+   <dd>Kai Ninomiya; Brandon Jones; Myles Maxfield. <a href="https://gpuweb.github.io/gpuweb/"><cite>WebGPU</cite></a>. URL: <a href="https://gpuweb.github.io/gpuweb/">https://gpuweb.github.io/gpuweb/</a>
    <dt id="biblio-webidl">[WEBIDL]
    <dd>Edgar Chen; Timothy Gu. <a href="https://webidl.spec.whatwg.org/"><cite>Web IDL Standard</cite></a>. Living Standard. URL: <a href="https://webidl.spec.whatwg.org/">https://webidl.spec.whatwg.org/</a>
   </dl>
@@ -4970,7 +8379,7 @@ <h3 class="no-num no-ref heading settled" id="informative"><span class="content"
    <dt id="biblio-video-summarization-with-lstm">[Video-Summarization-with-LSTM]
    <dd>Ke Zhang; et al. <a href="http://www-scf.usc.edu/~zhan355/ke_eccv2016.pdf"><cite>Video summarization with long short-term memory</cite></a>. October 2016. URL: <a href="http://www-scf.usc.edu/~zhan355/ke_eccv2016.pdf">http://www-scf.usc.edu/~zhan355/ke_eccv2016.pdf</a>
    <dt id="biblio-webmachinelearning-ethics">[WEBMACHINELEARNING-ETHICS]
-   <dd>James Fletcher; Anssi Kostiainen. <a href="https://webmachinelearning.github.io/webmachinelearning-ethics/"><cite>Ethical Principles for Web Machine Learning</cite></a>. URL: <a href="https://webmachinelearning.github.io/webmachinelearning-ethics/">https://webmachinelearning.github.io/webmachinelearning-ethics/</a>
+   <dd>Anssi Kostiainen. <a href="https://webmachinelearning.github.io/webmachinelearning-ethics/"><cite>Ethical Principles for Web Machine Learning</cite></a>. URL: <a href="https://webmachinelearning.github.io/webmachinelearning-ethics/">https://webmachinelearning.github.io/webmachinelearning-ethics/</a>
    <dt id="biblio-yolo">[YOLO]
    <dd>Joseph Redmon; et al. <a href="https://arxiv.org/abs/1506.02640"><cite>You Only Look Once: Unified, Real-Time Object Detection</cite></a>. May 2016. URL: <a href="https://arxiv.org/abs/1506.02640">https://arxiv.org/abs/1506.02640</a>
   </dl>
@@ -4999,13 +8408,13 @@ <h2 class="no-num no-ref heading settled" id="idl-index"><span class="content">I
 
 [<a class="idl-code" data-link-type="extended-attribute" href="https://webidl.spec.whatwg.org/#SecureContext"><c- g>SecureContext</c-></a>, <a class="idl-code" data-link-type="extended-attribute" href="https://webidl.spec.whatwg.org/#Exposed"><c- g>Exposed</c-></a>=(<c- n>Window</c->, <c- n>DedicatedWorker</c->)]
 <c- b>interface</c-> <a href="#ml"><code><c- g>ML</c-></code></a> {
-  <a class="idl-code" data-link-type="interface" href="https://webidl.spec.whatwg.org/#idl-promise"><c- b>Promise</c-></a>&lt;<a data-link-type="idl-name" href="#mlcontext"><c- n>MLContext</c-></a>> <a href="#dom-ml-createcontext"><code><c- g>createContext</c-></code></a>(<c- b>optional</c-> <a data-link-type="idl-name" href="#dictdef-mlcontextoptions"><c- n>MLContextOptions</c-></a> <a href="#dom-ml-createcontext-options-options"><code><c- g>options</c-></code></a> = {});
-  <a class="idl-code" data-link-type="interface" href="https://webidl.spec.whatwg.org/#idl-promise"><c- b>Promise</c-></a>&lt;<a data-link-type="idl-name" href="#mlcontext"><c- n>MLContext</c-></a>> <a href="#dom-ml-createcontext-gpudevice"><code><c- g>createContext</c-></code></a>(<a data-link-type="idl-name" href="https://gpuweb.github.io/gpuweb/#gpu-device"><c- n>GPUDevice</c-></a> <a href="#dom-ml-createcontext-gpudevice-gpudevice"><code><c- g>gpuDevice</c-></code></a>);
+  <a class="idl-code" data-link-type="interface" href="https://webidl.spec.whatwg.org/#idl-promise"><c- b>Promise</c-></a>&lt;<a data-link-type="idl-name" href="#mlcontext"><c- n>MLContext</c-></a>> <a class="idl-code" data-link-type="method" href="#dom-ml-createcontext"><c- g>createContext</c-></a>(<c- b>optional</c-> <a data-link-type="idl-name" href="#dictdef-mlcontextoptions"><c- n>MLContextOptions</c-></a> <a href="#dom-ml-createcontext-options-options"><code><c- g>options</c-></code></a> = {});
+  <a class="idl-code" data-link-type="interface" href="https://webidl.spec.whatwg.org/#idl-promise"><c- b>Promise</c-></a>&lt;<a data-link-type="idl-name" href="#mlcontext"><c- n>MLContext</c-></a>> <a class="idl-code" data-link-type="method" href="#dom-ml-createcontext-gpudevice"><c- g>createContext</c-></a>(<a data-link-type="idl-name" href="https://gpuweb.github.io/gpuweb/#gpudevice"><c- n>GPUDevice</c-></a> <a href="#dom-ml-createcontext-gpudevice-gpudevice"><code><c- g>gpuDevice</c-></code></a>);
 
   [<a class="idl-code" data-link-type="extended-attribute" href="https://webidl.spec.whatwg.org/#Exposed"><c- g>Exposed</c-></a>=(<c- n>DedicatedWorker</c->)]
-  <a data-link-type="idl-name" href="#mlcontext"><c- n>MLContext</c-></a> <a href="#dom-ml-createcontextsync"><code><c- g>createContextSync</c-></code></a>(<c- b>optional</c-> <a data-link-type="idl-name" href="#dictdef-mlcontextoptions"><c- n>MLContextOptions</c-></a> <a href="#dom-ml-createcontextsync-options-options"><code><c- g>options</c-></code></a> = {});
+  <a data-link-type="idl-name" href="#mlcontext"><c- n>MLContext</c-></a> <a class="idl-code" data-link-type="method" href="#dom-ml-createcontextsync"><c- g>createContextSync</c-></a>(<c- b>optional</c-> <a data-link-type="idl-name" href="#dictdef-mlcontextoptions"><c- n>MLContextOptions</c-></a> <a href="#dom-ml-createcontextsync-options-options"><code><c- g>options</c-></code></a> = {});
   [<a class="idl-code" data-link-type="extended-attribute" href="https://webidl.spec.whatwg.org/#Exposed"><c- g>Exposed</c-></a>=(<c- n>DedicatedWorker</c->)]
-  <a data-link-type="idl-name" href="#mlcontext"><c- n>MLContext</c-></a> <a href="#dom-ml-createcontextsync-gpudevice"><code><c- g>createContextSync</c-></code></a>(<a data-link-type="idl-name" href="https://gpuweb.github.io/gpuweb/#gpu-device"><c- n>GPUDevice</c-></a> <a href="#dom-ml-createcontextsync-gpudevice-gpudevice"><code><c- g>gpuDevice</c-></code></a>);
+  <a data-link-type="idl-name" href="#mlcontext"><c- n>MLContext</c-></a> <a class="idl-code" data-link-type="method" href="#dom-ml-createcontextsync-gpudevice"><c- g>createContextSync</c-></a>(<a data-link-type="idl-name" href="https://gpuweb.github.io/gpuweb/#gpudevice"><c- n>GPUDevice</c-></a> <a href="#dom-ml-createcontextsync-gpudevice-gpudevice"><code><c- g>gpuDevice</c-></code></a>);
 };
 
 [<a class="idl-code" data-link-type="extended-attribute" href="https://webidl.spec.whatwg.org/#SecureContext"><c- g>SecureContext</c-></a>, <a class="idl-code" data-link-type="extended-attribute" href="https://webidl.spec.whatwg.org/#Exposed"><c- g>Exposed</c-></a>=(<c- n>Window</c->, <c- n>DedicatedWorker</c->)]
@@ -5037,7 +8446,8 @@ <h2 class="no-num no-ref heading settled" id="idl-index"><span class="content">I
 <c- b>interface</c-> <a href="#mloperand"><code><c- g>MLOperand</c-></code></a> {};
 
 [<a class="idl-code" data-link-type="extended-attribute" href="https://webidl.spec.whatwg.org/#SecureContext"><c- g>SecureContext</c-></a>, <a class="idl-code" data-link-type="extended-attribute" href="https://webidl.spec.whatwg.org/#Exposed"><c- g>Exposed</c-></a>=(<c- n>Window</c->, <c- n>DedicatedWorker</c->)]
-<c- b>interface</c-> <a href="#mlactivation"><code><c- g>MLActivation</c-></code></a> {};
+<c- b>interface</c-> <a href="#mlactivation"><code><c- g>MLActivation</c-></code></a> {
+};
 
 <c- b>typedef</c-> <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#idl-record"><c- b>record</c-></a>&lt;<a class="idl-code" data-link-type="interface" href="https://webidl.spec.whatwg.org/#idl-DOMString"><c- b>DOMString</c-></a>, <a data-link-type="idl-name" href="https://webidl.spec.whatwg.org/#ArrayBufferView"><c- n>ArrayBufferView</c-></a>> <a href="#typedefdef-mlnamedarraybufferviews"><code><c- g>MLNamedArrayBufferViews</c-></code></a>;
 
@@ -5046,7 +8456,7 @@ <h2 class="no-num no-ref heading settled" id="idl-index"><span class="content">I
 
 <c- b>partial</c-> <c- b>interface</c-> <a class="idl-code" data-link-type="interface" href="#mlcontext"><c- g>MLContext</c-></a> {
   [<a class="idl-code" data-link-type="extended-attribute" href="https://webidl.spec.whatwg.org/#Exposed"><c- g>Exposed</c-></a>=(<c- n>DedicatedWorker</c->)]
-  <a class="idl-code" data-link-type="interface" href="https://webidl.spec.whatwg.org/#idl-undefined"><c- b>undefined</c-></a> <a href="#dom-mlcontext-computesync"><code><c- g>computeSync</c-></code></a>(
+  <a class="idl-code" data-link-type="interface" href="https://webidl.spec.whatwg.org/#idl-undefined"><c- b>undefined</c-></a> <a class="idl-code" data-link-type="method" href="#dom-mlcontext-computesync"><c- g>computeSync</c-></a>(
       <a data-link-type="idl-name" href="#mlgraph"><c- n>MLGraph</c-></a> <a href="#dom-mlcontext-computesync-graph-inputs-outputs-graph"><code><c- g>graph</c-></code></a>, <a data-link-type="idl-name" href="#typedefdef-mlnamedarraybufferviews"><c- n>MLNamedArrayBufferViews</c-></a> <a href="#dom-mlcontext-computesync-graph-inputs-outputs-inputs"><code><c- g>inputs</c-></code></a>, <a data-link-type="idl-name" href="#typedefdef-mlnamedarraybufferviews"><c- n>MLNamedArrayBufferViews</c-></a> <a href="#dom-mlcontext-computesync-graph-inputs-outputs-outputs"><code><c- g>outputs</c-></code></a>);
 };
 
@@ -5056,7 +8466,7 @@ <h2 class="no-num no-ref heading settled" id="idl-index"><span class="content">I
 };
 
 <c- b>partial</c-> <c- b>interface</c-> <a class="idl-code" data-link-type="interface" href="#mlcontext"><c- g>MLContext</c-></a> {
-  <a class="idl-code" data-link-type="interface" href="https://webidl.spec.whatwg.org/#idl-promise"><c- b>Promise</c-></a>&lt;<a data-link-type="idl-name" href="#dictdef-mlcomputeresult"><c- n>MLComputeResult</c-></a>> <a href="#dom-mlcontext-compute"><code><c- g>compute</c-></code></a>(
+  <a class="idl-code" data-link-type="interface" href="https://webidl.spec.whatwg.org/#idl-promise"><c- b>Promise</c-></a>&lt;<a data-link-type="idl-name" href="#dictdef-mlcomputeresult"><c- n>MLComputeResult</c-></a>> <a class="idl-code" data-link-type="method" href="#dom-mlcontext-compute"><c- g>compute</c-></a>(
       <a data-link-type="idl-name" href="#mlgraph"><c- n>MLGraph</c-></a> <a href="#dom-mlcontext-compute-graph-inputs-outputs-graph"><code><c- g>graph</c-></code></a>, <a data-link-type="idl-name" href="#typedefdef-mlnamedarraybufferviews"><c- n>MLNamedArrayBufferViews</c-></a> <a href="#dom-mlcontext-compute-graph-inputs-outputs-inputs"><code><c- g>inputs</c-></code></a>, <a data-link-type="idl-name" href="#typedefdef-mlnamedarraybufferviews"><c- n>MLNamedArrayBufferViews</c-></a> <a href="#dom-mlcontext-compute-graph-inputs-outputs-outputs"><code><c- g>outputs</c-></code></a>);
 };
 
@@ -5064,7 +8474,7 @@ <h2 class="no-num no-ref heading settled" id="idl-index"><span class="content">I
   <a data-link-type="idl-name" href="#mlcommandencoder"><c- n>MLCommandEncoder</c-></a> <a href="#dom-mlcontext-createcommandencoder"><code><c- g>createCommandEncoder</c-></code></a>();
 };
 
-<c- b>typedef</c-> (<a data-link-type="idl-name" href="https://gpuweb.github.io/gpuweb/#buffer-interface"><c- n>GPUBuffer</c-></a> <c- b>or</c-> <a data-link-type="idl-name" href="https://gpuweb.github.io/gpuweb/#texture-interface"><c- n>GPUTexture</c-></a>) <a href="#typedefdef-mlgpuresource"><code><c- g>MLGPUResource</c-></code></a>;
+<c- b>typedef</c-> (<a data-link-type="idl-name" href="https://gpuweb.github.io/gpuweb/#gpubuffer"><c- n>GPUBuffer</c-></a> <c- b>or</c-> <a data-link-type="idl-name" href="https://gpuweb.github.io/gpuweb/#gputexture"><c- n>GPUTexture</c-></a>) <a href="#typedefdef-mlgpuresource"><code><c- g>MLGPUResource</c-></code></a>;
 
 <c- b>typedef</c-> <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#idl-record"><c- b>record</c-></a>&lt;<a class="idl-code" data-link-type="interface" href="https://webidl.spec.whatwg.org/#idl-DOMString"><c- b>DOMString</c-></a>, <a data-link-type="idl-name" href="#typedefdef-mlgpuresource"><c- n>MLGPUResource</c-></a>> <a href="#typedefdef-mlnamedgpuresources"><code><c- g>MLNamedGPUResources</c-></code></a>;
 
@@ -5072,23 +8482,23 @@ <h2 class="no-num no-ref heading settled" id="idl-index"><span class="content">I
 <c- b>interface</c-> <a href="#mlcommandencoder"><code><c- g>MLCommandEncoder</c-></code></a> {};
 
 <c- b>partial</c-> <c- b>interface</c-> <a class="idl-code" data-link-type="interface" href="#mlcommandencoder"><c- g>MLCommandEncoder</c-></a> {
-  <a class="idl-code" data-link-type="interface" href="https://webidl.spec.whatwg.org/#idl-undefined"><c- b>undefined</c-></a> <a href="#dom-mlcommandencoder-initializegraph"><code><c- g>initializeGraph</c-></code></a>(<a data-link-type="idl-name" href="#mlgraph"><c- n>MLGraph</c-></a> <a href="#dom-mlcommandencoder-initializegraph-graph-graph"><code><c- g>graph</c-></code></a>);
+  <a class="idl-code" data-link-type="interface" href="https://webidl.spec.whatwg.org/#idl-undefined"><c- b>undefined</c-></a> <a class="idl-code" data-link-type="method" href="#dom-mlcommandencoder-initializegraph"><c- g>initializeGraph</c-></a>(<a data-link-type="idl-name" href="#mlgraph"><c- n>MLGraph</c-></a> <a href="#dom-mlcommandencoder-initializegraph-graph-graph"><code><c- g>graph</c-></code></a>);
 };
 
 <c- b>partial</c-> <c- b>interface</c-> <a class="idl-code" data-link-type="interface" href="#mlcommandencoder"><c- g>MLCommandEncoder</c-></a> {
-  <a class="idl-code" data-link-type="interface" href="https://webidl.spec.whatwg.org/#idl-undefined"><c- b>undefined</c-></a> <a href="#dom-mlcommandencoder-dispatch"><code><c- g>dispatch</c-></code></a>(<a data-link-type="idl-name" href="#mlgraph"><c- n>MLGraph</c-></a> <a href="#dom-mlcommandencoder-dispatch-graph-inputs-outputs-graph"><code><c- g>graph</c-></code></a>, <a data-link-type="idl-name" href="#typedefdef-mlnamedgpuresources"><c- n>MLNamedGPUResources</c-></a> <a href="#dom-mlcommandencoder-dispatch-graph-inputs-outputs-inputs"><code><c- g>inputs</c-></code></a>, <a data-link-type="idl-name" href="#typedefdef-mlnamedgpuresources"><c- n>MLNamedGPUResources</c-></a> <a href="#dom-mlcommandencoder-dispatch-graph-inputs-outputs-outputs"><code><c- g>outputs</c-></code></a>);
+  <a class="idl-code" data-link-type="interface" href="https://webidl.spec.whatwg.org/#idl-undefined"><c- b>undefined</c-></a> <a class="idl-code" data-link-type="method" href="#dom-mlcommandencoder-dispatch"><c- g>dispatch</c-></a>(<a data-link-type="idl-name" href="#mlgraph"><c- n>MLGraph</c-></a> <a href="#dom-mlcommandencoder-dispatch-graph-inputs-outputs-graph"><code><c- g>graph</c-></code></a>, <a data-link-type="idl-name" href="#typedefdef-mlnamedgpuresources"><c- n>MLNamedGPUResources</c-></a> <a href="#dom-mlcommandencoder-dispatch-graph-inputs-outputs-inputs"><code><c- g>inputs</c-></code></a>, <a data-link-type="idl-name" href="#typedefdef-mlnamedgpuresources"><c- n>MLNamedGPUResources</c-></a> <a href="#dom-mlcommandencoder-dispatch-graph-inputs-outputs-outputs"><code><c- g>outputs</c-></code></a>);
 };
 
 <c- b>partial</c-> <c- b>interface</c-> <a class="idl-code" data-link-type="interface" href="#mlcommandencoder"><c- g>MLCommandEncoder</c-></a> {
-  <a data-link-type="idl-name" href="https://gpuweb.github.io/gpuweb/#command-buffers"><c- n>GPUCommandBuffer</c-></a> <a href="#dom-mlcommandencoder-finish"><code><c- g>finish</c-></code></a>(<c- b>optional</c-> <a data-link-type="idl-name" href="https://gpuweb.github.io/gpuweb/#dictdef-gpucommandbufferdescriptor"><c- n>GPUCommandBufferDescriptor</c-></a> <a href="#dom-mlcommandencoder-finish-descriptor-descriptor"><code><c- g>descriptor</c-></code></a> = {});
+  <a data-link-type="idl-name" href="https://gpuweb.github.io/gpuweb/#gpucommandbuffer"><c- n>GPUCommandBuffer</c-></a> <a class="idl-code" data-link-type="method" href="#dom-mlcommandencoder-finish"><c- g>finish</c-></a>(<c- b>optional</c-> <a data-link-type="idl-name" href="https://gpuweb.github.io/gpuweb/#dictdef-gpucommandbufferdescriptor"><c- n>GPUCommandBufferDescriptor</c-></a> <a href="#dom-mlcommandencoder-finish-descriptor-descriptor"><code><c- g>descriptor</c-></code></a> = {});
 };
 
 <c- b>typedef</c-> <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#idl-record"><c- b>record</c-></a>&lt;<a class="idl-code" data-link-type="interface" href="https://webidl.spec.whatwg.org/#idl-DOMString"><c- b>DOMString</c-></a>, <a data-link-type="idl-name" href="#mloperand"><c- n>MLOperand</c-></a>> <a href="#typedefdef-mlnamedoperands"><code><c- g>MLNamedOperands</c-></code></a>;
 
 <c- b>dictionary</c-> <a href="#dictdef-mlbufferresourceview"><code><c- g>MLBufferResourceView</c-></code></a> {
-  <c- b>required</c-> <a data-link-type="idl-name" href="https://gpuweb.github.io/gpuweb/#buffer-interface"><c- n>GPUBuffer</c-></a> <a data-type="GPUBuffer " href="#dom-mlbufferresourceview-resource"><code><c- g>resource</c-></code></a>;
-  <a class="idl-code" data-link-type="interface" href="https://webidl.spec.whatwg.org/#idl-unsigned-long-long"><c- b>unsigned</c-> <c- b>long</c-> <c- b>long</c-></a> <a data-default="0" data-type="unsigned long long " href="#dom-mlbufferresourceview-offset"><code><c- g>offset</c-></code></a> = 0;
-  <a class="idl-code" data-link-type="interface" href="https://webidl.spec.whatwg.org/#idl-unsigned-long-long"><c- b>unsigned</c-> <c- b>long</c-> <c- b>long</c-></a> <a data-type="unsigned long long " href="#dom-mlbufferresourceview-size"><code><c- g>size</c-></code></a>;
+  <c- b>required</c-> <a data-link-type="idl-name" href="https://gpuweb.github.io/gpuweb/#gpubuffer"><c- n>GPUBuffer</c-></a> <a class="idl-code" data-link-type="dict-member" data-type="GPUBuffer " href="#dom-mlbufferresourceview-resource"><c- g>resource</c-></a>;
+  <a class="idl-code" data-link-type="interface" href="https://webidl.spec.whatwg.org/#idl-unsigned-long-long"><c- b>unsigned</c-> <c- b>long</c-> <c- b>long</c-></a> <a class="idl-code" data-default="0" data-link-type="dict-member" data-type="unsigned long long " href="#dom-mlbufferresourceview-offset"><c- g>offset</c-></a> = 0;
+  <a class="idl-code" data-link-type="interface" href="https://webidl.spec.whatwg.org/#idl-unsigned-long-long"><c- b>unsigned</c-> <c- b>long</c-> <c- b>long</c-></a> <a class="idl-code" data-link-type="dict-member" data-type="unsigned long long " href="#dom-mlbufferresourceview-size"><c- g>size</c-></a>;
 };
 
 <c- b>typedef</c-> (<a data-link-type="idl-name" href="https://webidl.spec.whatwg.org/#ArrayBufferView"><c- n>ArrayBufferView</c-></a> <c- b>or</c-> <a data-link-type="idl-name" href="#dictdef-mlbufferresourceview"><c- n>MLBufferResourceView</c-></a>) <a href="#typedefdef-mlbufferview"><code><c- g>MLBufferView</c-></code></a>;
@@ -5096,35 +8506,35 @@ <h2 class="no-num no-ref heading settled" id="idl-index"><span class="content">I
 [<a class="idl-code" data-link-type="extended-attribute" href="https://webidl.spec.whatwg.org/#SecureContext"><c- g>SecureContext</c-></a>, <a class="idl-code" data-link-type="extended-attribute" href="https://webidl.spec.whatwg.org/#Exposed"><c- g>Exposed</c-></a>=(<c- n>Window</c->, <c- n>DedicatedWorker</c->)]
 <c- b>interface</c-> <a href="#mlgraphbuilder"><code><c- g>MLGraphBuilder</c-></code></a> {
   // Construct the graph builder from the context.
-  <a href="#dom-mlgraphbuilder-mlgraphbuilder"><code><c- g>constructor</c-></code></a>(<a data-link-type="idl-name" href="#mlcontext"><c- n>MLContext</c-></a> <a href="#dom-mlgraphbuilder-mlgraphbuilder-context-context"><code><c- g>context</c-></code></a>);
+  <a class="idl-code" data-link-type="constructor" href="#dom-mlgraphbuilder-mlgraphbuilder"><c- g>constructor</c-></a>(<a data-link-type="idl-name" href="#mlcontext"><c- n>MLContext</c-></a> <a href="#dom-mlgraphbuilder-mlgraphbuilder-context-context"><code><c- g>context</c-></code></a>);
 
   // Create an operand for a graph input.
-  <a data-link-type="idl-name" href="#mloperand"><c- n>MLOperand</c-></a> <a href="#dom-mlgraphbuilder-input"><code><c- g>input</c-></code></a>(<a class="idl-code" data-link-type="interface" href="https://webidl.spec.whatwg.org/#idl-DOMString"><c- b>DOMString</c-></a> <a href="#dom-mlgraphbuilder-input-name-descriptor-name"><code><c- g>name</c-></code></a>, <a data-link-type="idl-name" href="#dictdef-mloperanddescriptor"><c- n>MLOperandDescriptor</c-></a> <a href="#dom-mlgraphbuilder-input-name-descriptor-descriptor"><code><c- g>descriptor</c-></code></a>);
+  <a data-link-type="idl-name" href="#mloperand"><c- n>MLOperand</c-></a> <a class="idl-code" data-link-type="method" href="#dom-mlgraphbuilder-input"><c- g>input</c-></a>(<a class="idl-code" data-link-type="interface" href="https://webidl.spec.whatwg.org/#idl-DOMString"><c- b>DOMString</c-></a> <a href="#dom-mlgraphbuilder-input-name-descriptor-name"><code><c- g>name</c-></code></a>, <a data-link-type="idl-name" href="#dictdef-mloperanddescriptor"><c- n>MLOperandDescriptor</c-></a> <a href="#dom-mlgraphbuilder-input-name-descriptor-descriptor"><code><c- g>descriptor</c-></code></a>);
 
   // Create an operand for a graph constant.
-  <a data-link-type="idl-name" href="#mloperand"><c- n>MLOperand</c-></a> <a href="#dom-mlgraphbuilder-constant"><code><c- g>constant</c-></code></a>(<a data-link-type="idl-name" href="#dictdef-mloperanddescriptor"><c- n>MLOperandDescriptor</c-></a> <a href="#dom-mlgraphbuilder-constant-descriptor-bufferview-descriptor"><code><c- g>descriptor</c-></code></a>, <a data-link-type="idl-name" href="#typedefdef-mlbufferview"><c- n>MLBufferView</c-></a> <a href="#dom-mlgraphbuilder-constant-descriptor-bufferview-bufferview"><code><c- g>bufferView</c-></code></a>);
+  <a data-link-type="idl-name" href="#mloperand"><c- n>MLOperand</c-></a> <a class="idl-code" data-link-type="method" href="#dom-mlgraphbuilder-constant"><c- g>constant</c-></a>(<a data-link-type="idl-name" href="#dictdef-mloperanddescriptor"><c- n>MLOperandDescriptor</c-></a> <a href="#dom-mlgraphbuilder-constant-descriptor-bufferview-descriptor"><code><c- g>descriptor</c-></code></a>, <a data-link-type="idl-name" href="#typedefdef-mlbufferview"><c- n>MLBufferView</c-></a> <a href="#dom-mlgraphbuilder-constant-descriptor-bufferview-bufferview"><code><c- g>bufferView</c-></code></a>);
 
   // Create a single-value operand from the specified number of the specified type.
-  <a data-link-type="idl-name" href="#mloperand"><c- n>MLOperand</c-></a> <a href="#dom-mlgraphbuilder-constant-value-type"><code><c- g>constant</c-></code></a>(<a class="idl-code" data-link-type="interface" href="https://webidl.spec.whatwg.org/#idl-double"><c- b>double</c-></a> <a href="#dom-mlgraphbuilder-constant-value-type-value"><code><c- g>value</c-></code></a>, <c- b>optional</c-> <a data-link-type="idl-name" href="#enumdef-mloperandtype"><c- n>MLOperandType</c-></a> <a href="#dom-mlgraphbuilder-constant-value-type-type"><code><c- g>type</c-></code></a> = "float32");
+  <a data-link-type="idl-name" href="#mloperand"><c- n>MLOperand</c-></a> <a class="idl-code" data-link-type="method" href="#dom-mlgraphbuilder-constant-value-type"><c- g>constant</c-></a>(<a class="idl-code" data-link-type="interface" href="https://webidl.spec.whatwg.org/#idl-double"><c- b>double</c-></a> <a href="#dom-mlgraphbuilder-constant-value-type-value"><code><c- g>value</c-></code></a>, <c- b>optional</c-> <a data-link-type="idl-name" href="#enumdef-mloperandtype"><c- n>MLOperandType</c-></a> <a href="#dom-mlgraphbuilder-constant-value-type-type"><code><c- g>type</c-></code></a> = "float32");
 
   // Compile the graph up to the specified output operands asynchronously.
-  <a class="idl-code" data-link-type="interface" href="https://webidl.spec.whatwg.org/#idl-promise"><c- b>Promise</c-></a>&lt;<a data-link-type="idl-name" href="#mlgraph"><c- n>MLGraph</c-></a>> <a href="#dom-mlgraphbuilder-build"><code><c- g>build</c-></code></a>(<a data-link-type="idl-name" href="#typedefdef-mlnamedoperands"><c- n>MLNamedOperands</c-></a> <a href="#dom-mlgraphbuilder-build-outputs-outputs"><code><c- g>outputs</c-></code></a>);
+  <a class="idl-code" data-link-type="interface" href="https://webidl.spec.whatwg.org/#idl-promise"><c- b>Promise</c-></a>&lt;<a data-link-type="idl-name" href="#mlgraph"><c- n>MLGraph</c-></a>> <a class="idl-code" data-link-type="method" href="#dom-mlgraphbuilder-build"><c- g>build</c-></a>(<a data-link-type="idl-name" href="#typedefdef-mlnamedoperands"><c- n>MLNamedOperands</c-></a> <a href="#dom-mlgraphbuilder-build-outputs-outputs"><code><c- g>outputs</c-></code></a>);
 
   // Compile the graph up to the specified output operands synchronously.
   [<a class="idl-code" data-link-type="extended-attribute" href="https://webidl.spec.whatwg.org/#Exposed"><c- g>Exposed</c-></a>=(<c- n>DedicatedWorker</c->)]
-  <a data-link-type="idl-name" href="#mlgraph"><c- n>MLGraph</c-></a> <a href="#dom-mlgraphbuilder-buildsync"><code><c- g>buildSync</c-></code></a>(<a data-link-type="idl-name" href="#typedefdef-mlnamedoperands"><c- n>MLNamedOperands</c-></a> <a href="#dom-mlgraphbuilder-buildsync-outputs-outputs"><code><c- g>outputs</c-></code></a>);
+  <a data-link-type="idl-name" href="#mlgraph"><c- n>MLGraph</c-></a> <a class="idl-code" data-link-type="method" href="#dom-mlgraphbuilder-buildsync"><c- g>buildSync</c-></a>(<a data-link-type="idl-name" href="#typedefdef-mlnamedoperands"><c- n>MLNamedOperands</c-></a> <a href="#dom-mlgraphbuilder-buildsync-outputs-outputs"><code><c- g>outputs</c-></code></a>);
 };
 
 <c- b>dictionary</c-> <a href="#dictdef-mlbatchnormalizationoptions"><code><c- g>MLBatchNormalizationOptions</c-></code></a> {
-  <a data-link-type="idl-name" href="#mloperand"><c- n>MLOperand</c-></a> <a data-type="MLOperand " href="#dom-mlbatchnormalizationoptions-scale"><code><c- g>scale</c-></code></a>;
-  <a data-link-type="idl-name" href="#mloperand"><c- n>MLOperand</c-></a> <a data-type="MLOperand " href="#dom-mlbatchnormalizationoptions-bias"><code><c- g>bias</c-></code></a>;
-  <a class="idl-code" data-link-type="interface" href="https://webidl.spec.whatwg.org/#idl-unsigned-long"><c- b>unsigned</c-> <c- b>long</c-></a> <a data-default="1" data-type="unsigned long " href="#dom-mlbatchnormalizationoptions-axis"><code><c- g>axis</c-></code></a> = 1;
-  <a class="idl-code" data-link-type="interface" href="https://webidl.spec.whatwg.org/#idl-float"><c- b>float</c-></a> <a data-default="1e-5" data-type="float " href="#dom-mlbatchnormalizationoptions-epsilon"><code><c- g>epsilon</c-></code></a> = 1e-5;
-  <a data-link-type="idl-name" href="#mlactivation"><c- n>MLActivation</c-></a> <a data-type="MLActivation " href="#dom-mlbatchnormalizationoptions-activation"><code><c- g>activation</c-></code></a>;
+  <a data-link-type="idl-name" href="#mloperand"><c- n>MLOperand</c-></a> <a class="idl-code" data-link-type="dict-member" data-type="MLOperand " href="#dom-mlbatchnormalizationoptions-scale"><c- g>scale</c-></a>;
+  <a data-link-type="idl-name" href="#mloperand"><c- n>MLOperand</c-></a> <a class="idl-code" data-link-type="dict-member" data-type="MLOperand " href="#dom-mlbatchnormalizationoptions-bias"><c- g>bias</c-></a>;
+  <a class="idl-code" data-link-type="interface" href="https://webidl.spec.whatwg.org/#idl-unsigned-long"><c- b>unsigned</c-> <c- b>long</c-></a> <a class="idl-code" data-default="1" data-link-type="dict-member" data-type="unsigned long " href="#dom-mlbatchnormalizationoptions-axis"><c- g>axis</c-></a> = 1;
+  <a class="idl-code" data-link-type="interface" href="https://webidl.spec.whatwg.org/#idl-float"><c- b>float</c-></a> <a class="idl-code" data-default="1e-5" data-link-type="dict-member" data-type="float " href="#dom-mlbatchnormalizationoptions-epsilon"><c- g>epsilon</c-></a> = 1e-5;
+  <a data-link-type="idl-name" href="#mlactivation"><c- n>MLActivation</c-></a> <a class="idl-code" data-link-type="dict-member" data-type="MLActivation " href="#dom-mlbatchnormalizationoptions-activation"><c- g>activation</c-></a>;
 };
 
 <c- b>partial</c-> <c- b>interface</c-> <a class="idl-code" data-link-type="interface" href="#mlgraphbuilder"><c- g>MLGraphBuilder</c-></a> {
-  <a data-link-type="idl-name" href="#mloperand"><c- n>MLOperand</c-></a> <a href="#dom-mlgraphbuilder-batchnormalization"><code><c- g>batchNormalization</c-></code></a>(<a data-link-type="idl-name" href="#mloperand"><c- n>MLOperand</c-></a> <a href="#dom-mlgraphbuilder-batchnormalization-input-mean-variance-options-input"><code><c- g>input</c-></code></a>, <a data-link-type="idl-name" href="#mloperand"><c- n>MLOperand</c-></a> <a href="#dom-mlgraphbuilder-batchnormalization-input-mean-variance-options-mean"><code><c- g>mean</c-></code></a>, <a data-link-type="idl-name" href="#mloperand"><c- n>MLOperand</c-></a> <a href="#dom-mlgraphbuilder-batchnormalization-input-mean-variance-options-variance"><code><c- g>variance</c-></code></a>,
+  <a data-link-type="idl-name" href="#mloperand"><c- n>MLOperand</c-></a> <a class="idl-code" data-link-type="method" href="#dom-mlgraphbuilder-batchnormalization"><c- g>batchNormalization</c-></a>(<a data-link-type="idl-name" href="#mloperand"><c- n>MLOperand</c-></a> <a href="#dom-mlgraphbuilder-batchnormalization-input-mean-variance-options-input"><code><c- g>input</c-></code></a>, <a data-link-type="idl-name" href="#mloperand"><c- n>MLOperand</c-></a> <a href="#dom-mlgraphbuilder-batchnormalization-input-mean-variance-options-mean"><code><c- g>mean</c-></code></a>, <a data-link-type="idl-name" href="#mloperand"><c- n>MLOperand</c-></a> <a href="#dom-mlgraphbuilder-batchnormalization-input-mean-variance-options-variance"><code><c- g>variance</c-></code></a>,
                              <c- b>optional</c-> <a data-link-type="idl-name" href="#dictdef-mlbatchnormalizationoptions"><c- n>MLBatchNormalizationOptions</c-></a> <a href="#dom-mlgraphbuilder-batchnormalization-input-mean-variance-options-options"><code><c- g>options</c-></code></a> = {});
 };
 
@@ -5134,12 +8544,12 @@ <h2 class="no-num no-ref heading settled" id="idl-index"><span class="content">I
 };
 
 <c- b>partial</c-> <c- b>interface</c-> <a class="idl-code" data-link-type="interface" href="#mlgraphbuilder"><c- g>MLGraphBuilder</c-></a> {
-  <a data-link-type="idl-name" href="#mloperand"><c- n>MLOperand</c-></a> <a href="#dom-mlgraphbuilder-clamp"><code><c- g>clamp</c-></code></a>(<a data-link-type="idl-name" href="#mloperand"><c- n>MLOperand</c-></a> <a href="#dom-mlgraphbuilder-clamp-x-options-x"><code><c- g>x</c-></code></a>, <c- b>optional</c-> <a data-link-type="idl-name" href="#dictdef-mlclampoptions"><c- n>MLClampOptions</c-></a> <a href="#dom-mlgraphbuilder-clamp-x-options-options"><code><c- g>options</c-></code></a> = {});
-  <a data-link-type="idl-name" href="#mlactivation"><c- n>MLActivation</c-></a> <a href="#dom-mlgraphbuilder-clamp-options"><code><c- g>clamp</c-></code></a>(<c- b>optional</c-> <a data-link-type="idl-name" href="#dictdef-mlclampoptions"><c- n>MLClampOptions</c-></a> <a href="#dom-mlgraphbuilder-clamp-options-options"><code><c- g>options</c-></code></a> = {});
+  <a data-link-type="idl-name" href="#mloperand"><c- n>MLOperand</c-></a> <a class="idl-code" data-link-type="method" href="#dom-mlgraphbuilder-clamp"><c- g>clamp</c-></a>(<a data-link-type="idl-name" href="#mloperand"><c- n>MLOperand</c-></a> <a href="#dom-mlgraphbuilder-clamp-operand-options-operand"><code><c- g>operand</c-></code></a>, <c- b>optional</c-> <a data-link-type="idl-name" href="#dictdef-mlclampoptions"><c- n>MLClampOptions</c-></a> <a href="#dom-mlgraphbuilder-clamp-operand-options-options"><code><c- g>options</c-></code></a> = {});
+  <a data-link-type="idl-name" href="#mlactivation"><c- n>MLActivation</c-></a> <a class="idl-code" data-link-type="method" href="#dom-mlgraphbuilder-clamp-options"><c- g>clamp</c-></a>(<c- b>optional</c-> <a data-link-type="idl-name" href="#dictdef-mlclampoptions"><c- n>MLClampOptions</c-></a> <a href="#dom-mlgraphbuilder-clamp-options-options"><code><c- g>options</c-></code></a> = {});
 };
 
 <c- b>partial</c-> <c- b>interface</c-> <a class="idl-code" data-link-type="interface" href="#mlgraphbuilder"><c- g>MLGraphBuilder</c-></a> {
-  <a data-link-type="idl-name" href="#mloperand"><c- n>MLOperand</c-></a> <a href="#dom-mlgraphbuilder-concat"><code><c- g>concat</c-></code></a>(<a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#idl-sequence"><c- b>sequence</c-></a>&lt;<a data-link-type="idl-name" href="#mloperand"><c- n>MLOperand</c-></a>> <a href="#dom-mlgraphbuilder-concat-inputs-axis-inputs"><code><c- g>inputs</c-></code></a>, <a class="idl-code" data-link-type="interface" href="https://webidl.spec.whatwg.org/#idl-unsigned-long"><c- b>unsigned</c-> <c- b>long</c-></a> <a href="#dom-mlgraphbuilder-concat-inputs-axis-axis"><code><c- g>axis</c-></code></a>);
+  <a data-link-type="idl-name" href="#mloperand"><c- n>MLOperand</c-></a> <a class="idl-code" data-link-type="method" href="#dom-mlgraphbuilder-concat"><c- g>concat</c-></a>(<a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#idl-sequence"><c- b>sequence</c-></a>&lt;<a data-link-type="idl-name" href="#mloperand"><c- n>MLOperand</c-></a>> <a href="#dom-mlgraphbuilder-concat-inputs-axis-inputs"><code><c- g>inputs</c-></code></a>, <a class="idl-code" data-link-type="interface" href="https://webidl.spec.whatwg.org/#idl-unsigned-long"><c- b>unsigned</c-> <c- b>long</c-></a> <a href="#dom-mlgraphbuilder-concat-inputs-axis-axis"><code><c- g>axis</c-></code></a>);
 };
 
 <c- b>enum</c-> <a href="#enumdef-mlconv2dfilteroperandlayout"><code><c- g>MLConv2dFilterOperandLayout</c-></code></a> {
@@ -5156,19 +8566,19 @@ <h2 class="no-num no-ref heading settled" id="idl-index"><span class="content">I
 };
 
 <c- b>dictionary</c-> <a href="#dictdef-mlconv2doptions"><code><c- g>MLConv2dOptions</c-></code></a> {
-  <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#idl-sequence"><c- b>sequence</c-></a>&lt;<a class="idl-code" data-link-type="interface" href="https://webidl.spec.whatwg.org/#idl-unsigned-long"><c- b>unsigned</c-> <c- b>long</c-></a>> <a data-type="sequence<unsigned long> " href="#dom-mlconv2doptions-padding"><code><c- g>padding</c-></code></a>;
-  <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#idl-sequence"><c- b>sequence</c-></a>&lt;<a class="idl-code" data-link-type="interface" href="https://webidl.spec.whatwg.org/#idl-unsigned-long"><c- b>unsigned</c-> <c- b>long</c-></a>> <a data-type="sequence<unsigned long> " href="#dom-mlconv2doptions-strides"><code><c- g>strides</c-></code></a>;
-  <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#idl-sequence"><c- b>sequence</c-></a>&lt;<a class="idl-code" data-link-type="interface" href="https://webidl.spec.whatwg.org/#idl-unsigned-long"><c- b>unsigned</c-> <c- b>long</c-></a>> <a data-type="sequence<unsigned long> " href="#dom-mlconv2doptions-dilations"><code><c- g>dilations</c-></code></a>;
-  <a data-link-type="idl-name" href="#enumdef-mlautopad"><c- n>MLAutoPad</c-></a> <a data-default="&quot;explicit&quot;" data-type="MLAutoPad " href="#dom-mlconv2doptions-autopad"><code><c- g>autoPad</c-></code></a> = "explicit";
-  <a class="idl-code" data-link-type="interface" href="https://webidl.spec.whatwg.org/#idl-unsigned-long"><c- b>unsigned</c-> <c- b>long</c-></a> <a data-default="1" data-type="unsigned long " href="#dom-mlconv2doptions-groups"><code><c- g>groups</c-></code></a> = 1;
-  <a data-link-type="idl-name" href="#enumdef-mlinputoperandlayout"><c- n>MLInputOperandLayout</c-></a> <a data-default="&quot;nchw&quot;" data-type="MLInputOperandLayout " href="#dom-mlconv2doptions-inputlayout"><code><c- g>inputLayout</c-></code></a> = "nchw";
-  <a data-link-type="idl-name" href="#enumdef-mlconv2dfilteroperandlayout"><c- n>MLConv2dFilterOperandLayout</c-></a> <a data-default="&quot;oihw&quot;" data-type="MLConv2dFilterOperandLayout " href="#dom-mlconv2doptions-filterlayout"><code><c- g>filterLayout</c-></code></a> = "oihw";
-  <a data-link-type="idl-name" href="#mloperand"><c- n>MLOperand</c-></a> <a data-type="MLOperand " href="#dom-mlconv2doptions-bias"><code><c- g>bias</c-></code></a>;
-  <a data-link-type="idl-name" href="#mlactivation"><c- n>MLActivation</c-></a> <a data-type="MLActivation " href="#dom-mlconv2doptions-activation"><code><c- g>activation</c-></code></a>;
+  <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#idl-sequence"><c- b>sequence</c-></a>&lt;<a class="idl-code" data-link-type="interface" href="https://webidl.spec.whatwg.org/#idl-unsigned-long"><c- b>unsigned</c-> <c- b>long</c-></a>> <a class="idl-code" data-link-type="dict-member" data-type="sequence<unsigned long> " href="#dom-mlconv2doptions-padding"><c- g>padding</c-></a>;
+  <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#idl-sequence"><c- b>sequence</c-></a>&lt;<a class="idl-code" data-link-type="interface" href="https://webidl.spec.whatwg.org/#idl-unsigned-long"><c- b>unsigned</c-> <c- b>long</c-></a>> <a class="idl-code" data-link-type="dict-member" data-type="sequence<unsigned long> " href="#dom-mlconv2doptions-strides"><c- g>strides</c-></a>;
+  <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#idl-sequence"><c- b>sequence</c-></a>&lt;<a class="idl-code" data-link-type="interface" href="https://webidl.spec.whatwg.org/#idl-unsigned-long"><c- b>unsigned</c-> <c- b>long</c-></a>> <a class="idl-code" data-link-type="dict-member" data-type="sequence<unsigned long> " href="#dom-mlconv2doptions-dilations"><c- g>dilations</c-></a>;
+  <a data-link-type="idl-name" href="#enumdef-mlautopad"><c- n>MLAutoPad</c-></a> <a class="idl-code" data-default="&quot;explicit&quot;" data-link-type="dict-member" data-type="MLAutoPad " href="#dom-mlconv2doptions-autopad"><c- g>autoPad</c-></a> = "explicit";
+  <a class="idl-code" data-link-type="interface" href="https://webidl.spec.whatwg.org/#idl-unsigned-long"><c- b>unsigned</c-> <c- b>long</c-></a> <a class="idl-code" data-default="1" data-link-type="dict-member" data-type="unsigned long " href="#dom-mlconv2doptions-groups"><c- g>groups</c-></a> = 1;
+  <a data-link-type="idl-name" href="#enumdef-mlinputoperandlayout"><c- n>MLInputOperandLayout</c-></a> <a class="idl-code" data-default="&quot;nchw&quot;" data-link-type="dict-member" data-type="MLInputOperandLayout " href="#dom-mlconv2doptions-inputlayout"><c- g>inputLayout</c-></a> = "nchw";
+  <a data-link-type="idl-name" href="#enumdef-mlconv2dfilteroperandlayout"><c- n>MLConv2dFilterOperandLayout</c-></a> <a class="idl-code" data-default="&quot;oihw&quot;" data-link-type="dict-member" data-type="MLConv2dFilterOperandLayout " href="#dom-mlconv2doptions-filterlayout"><c- g>filterLayout</c-></a> = "oihw";
+  <a data-link-type="idl-name" href="#mloperand"><c- n>MLOperand</c-></a> <a class="idl-code" data-link-type="dict-member" data-type="MLOperand " href="#dom-mlconv2doptions-bias"><c- g>bias</c-></a>;
+  <a data-link-type="idl-name" href="#mlactivation"><c- n>MLActivation</c-></a> <a class="idl-code" data-link-type="dict-member" data-type="MLActivation " href="#dom-mlconv2doptions-activation"><c- g>activation</c-></a>;
 };
 
 <c- b>partial</c-> <c- b>interface</c-> <a class="idl-code" data-link-type="interface" href="#mlgraphbuilder"><c- g>MLGraphBuilder</c-></a> {
-  <a data-link-type="idl-name" href="#mloperand"><c- n>MLOperand</c-></a> <a href="#dom-mlgraphbuilder-conv2d"><code><c- g>conv2d</c-></code></a>(<a data-link-type="idl-name" href="#mloperand"><c- n>MLOperand</c-></a> <a href="#dom-mlgraphbuilder-conv2d-input-filter-options-input"><code><c- g>input</c-></code></a>, <a data-link-type="idl-name" href="#mloperand"><c- n>MLOperand</c-></a> <a href="#dom-mlgraphbuilder-conv2d-input-filter-options-filter"><code><c- g>filter</c-></code></a>, <c- b>optional</c-> <a data-link-type="idl-name" href="#dictdef-mlconv2doptions"><c- n>MLConv2dOptions</c-></a> <a href="#dom-mlgraphbuilder-conv2d-input-filter-options-options"><code><c- g>options</c-></code></a> = {});
+  <a data-link-type="idl-name" href="#mloperand"><c- n>MLOperand</c-></a> <a class="idl-code" data-link-type="method" href="#dom-mlgraphbuilder-conv2d"><c- g>conv2d</c-></a>(<a data-link-type="idl-name" href="#mloperand"><c- n>MLOperand</c-></a> <a href="#dom-mlgraphbuilder-conv2d-input-filter-options-input"><code><c- g>input</c-></code></a>, <a data-link-type="idl-name" href="#mloperand"><c- n>MLOperand</c-></a> <a href="#dom-mlgraphbuilder-conv2d-input-filter-options-filter"><code><c- g>filter</c-></code></a>, <c- b>optional</c-> <a data-link-type="idl-name" href="#dictdef-mlconv2doptions"><c- n>MLConv2dOptions</c-></a> <a href="#dom-mlgraphbuilder-conv2d-input-filter-options-options"><code><c- g>options</c-></code></a> = {});
 };
 
 <c- b>enum</c-> <a href="#enumdef-mlconvtranspose2dfilteroperandlayout"><code><c- g>MLConvTranspose2dFilterOperandLayout</c-></code></a> {
@@ -5178,44 +8588,44 @@ <h2 class="no-num no-ref heading settled" id="idl-index"><span class="content">I
 };
 
 <c- b>dictionary</c-> <a href="#dictdef-mlconvtranspose2doptions"><code><c- g>MLConvTranspose2dOptions</c-></code></a> {
-  <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#idl-sequence"><c- b>sequence</c-></a>&lt;<a class="idl-code" data-link-type="interface" href="https://webidl.spec.whatwg.org/#idl-unsigned-long"><c- b>unsigned</c-> <c- b>long</c-></a>> <a data-type="sequence<unsigned long> " href="#dom-mlconvtranspose2doptions-padding"><code><c- g>padding</c-></code></a>;
-  <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#idl-sequence"><c- b>sequence</c-></a>&lt;<a class="idl-code" data-link-type="interface" href="https://webidl.spec.whatwg.org/#idl-unsigned-long"><c- b>unsigned</c-> <c- b>long</c-></a>> <a data-type="sequence<unsigned long> " href="#dom-mlconvtranspose2doptions-strides"><code><c- g>strides</c-></code></a>;
-  <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#idl-sequence"><c- b>sequence</c-></a>&lt;<a class="idl-code" data-link-type="interface" href="https://webidl.spec.whatwg.org/#idl-unsigned-long"><c- b>unsigned</c-> <c- b>long</c-></a>> <a data-type="sequence<unsigned long> " href="#dom-mlconvtranspose2doptions-dilations"><code><c- g>dilations</c-></code></a>;
-  <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#idl-sequence"><c- b>sequence</c-></a>&lt;<a class="idl-code" data-link-type="interface" href="https://webidl.spec.whatwg.org/#idl-unsigned-long"><c- b>unsigned</c-> <c- b>long</c-></a>> <a data-type="sequence<unsigned long> " href="#dom-mlconvtranspose2doptions-outputpadding"><code><c- g>outputPadding</c-></code></a>;
-  <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#idl-sequence"><c- b>sequence</c-></a>&lt;<a class="idl-code" data-link-type="interface" href="https://webidl.spec.whatwg.org/#idl-unsigned-long"><c- b>unsigned</c-> <c- b>long</c-></a>> <a data-type="sequence<unsigned long> " href="#dom-mlconvtranspose2doptions-outputsizes"><code><c- g>outputSizes</c-></code></a>;
-  <a data-link-type="idl-name" href="#enumdef-mlautopad"><c- n>MLAutoPad</c-></a> <a data-default="&quot;explicit&quot;" data-type="MLAutoPad " href="#dom-mlconvtranspose2doptions-autopad"><code><c- g>autoPad</c-></code></a> = "explicit";
-  <a class="idl-code" data-link-type="interface" href="https://webidl.spec.whatwg.org/#idl-unsigned-long"><c- b>unsigned</c-> <c- b>long</c-></a> <a data-default="1" data-type="unsigned long " href="#dom-mlconvtranspose2doptions-groups"><code><c- g>groups</c-></code></a> = 1;
-  <a data-link-type="idl-name" href="#enumdef-mlinputoperandlayout"><c- n>MLInputOperandLayout</c-></a> <a data-default="&quot;nchw&quot;" data-type="MLInputOperandLayout " href="#dom-mlconvtranspose2doptions-inputlayout"><code><c- g>inputLayout</c-></code></a> = "nchw";
-  <a data-link-type="idl-name" href="#enumdef-mlconvtranspose2dfilteroperandlayout"><c- n>MLConvTranspose2dFilterOperandLayout</c-></a> <a data-default="&quot;iohw&quot;" data-type="MLConvTranspose2dFilterOperandLayout " href="#dom-mlconvtranspose2doptions-filterlayout"><code><c- g>filterLayout</c-></code></a> = "iohw";
-  <a data-link-type="idl-name" href="#mloperand"><c- n>MLOperand</c-></a> <a data-type="MLOperand " href="#dom-mlconvtranspose2doptions-bias"><code><c- g>bias</c-></code></a>;
-  <a data-link-type="idl-name" href="#mlactivation"><c- n>MLActivation</c-></a> <a data-type="MLActivation " href="#dom-mlconvtranspose2doptions-activation"><code><c- g>activation</c-></code></a>;
+  <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#idl-sequence"><c- b>sequence</c-></a>&lt;<a class="idl-code" data-link-type="interface" href="https://webidl.spec.whatwg.org/#idl-unsigned-long"><c- b>unsigned</c-> <c- b>long</c-></a>> <a class="idl-code" data-link-type="dict-member" data-type="sequence<unsigned long> " href="#dom-mlconvtranspose2doptions-padding"><c- g>padding</c-></a>;
+  <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#idl-sequence"><c- b>sequence</c-></a>&lt;<a class="idl-code" data-link-type="interface" href="https://webidl.spec.whatwg.org/#idl-unsigned-long"><c- b>unsigned</c-> <c- b>long</c-></a>> <a class="idl-code" data-link-type="dict-member" data-type="sequence<unsigned long> " href="#dom-mlconvtranspose2doptions-strides"><c- g>strides</c-></a>;
+  <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#idl-sequence"><c- b>sequence</c-></a>&lt;<a class="idl-code" data-link-type="interface" href="https://webidl.spec.whatwg.org/#idl-unsigned-long"><c- b>unsigned</c-> <c- b>long</c-></a>> <a class="idl-code" data-link-type="dict-member" data-type="sequence<unsigned long> " href="#dom-mlconvtranspose2doptions-dilations"><c- g>dilations</c-></a>;
+  <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#idl-sequence"><c- b>sequence</c-></a>&lt;<a class="idl-code" data-link-type="interface" href="https://webidl.spec.whatwg.org/#idl-unsigned-long"><c- b>unsigned</c-> <c- b>long</c-></a>> <a class="idl-code" data-link-type="dict-member" data-type="sequence<unsigned long> " href="#dom-mlconvtranspose2doptions-outputpadding"><c- g>outputPadding</c-></a>;
+  <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#idl-sequence"><c- b>sequence</c-></a>&lt;<a class="idl-code" data-link-type="interface" href="https://webidl.spec.whatwg.org/#idl-unsigned-long"><c- b>unsigned</c-> <c- b>long</c-></a>> <a class="idl-code" data-link-type="dict-member" data-type="sequence<unsigned long> " href="#dom-mlconvtranspose2doptions-outputsizes"><c- g>outputSizes</c-></a>;
+  <a data-link-type="idl-name" href="#enumdef-mlautopad"><c- n>MLAutoPad</c-></a> <a class="idl-code" data-default="&quot;explicit&quot;" data-link-type="dict-member" data-type="MLAutoPad " href="#dom-mlconvtranspose2doptions-autopad"><c- g>autoPad</c-></a> = "explicit";
+  <a class="idl-code" data-link-type="interface" href="https://webidl.spec.whatwg.org/#idl-unsigned-long"><c- b>unsigned</c-> <c- b>long</c-></a> <a class="idl-code" data-default="1" data-link-type="dict-member" data-type="unsigned long " href="#dom-mlconvtranspose2doptions-groups"><c- g>groups</c-></a> = 1;
+  <a data-link-type="idl-name" href="#enumdef-mlinputoperandlayout"><c- n>MLInputOperandLayout</c-></a> <a class="idl-code" data-default="&quot;nchw&quot;" data-link-type="dict-member" data-type="MLInputOperandLayout " href="#dom-mlconvtranspose2doptions-inputlayout"><c- g>inputLayout</c-></a> = "nchw";
+  <a data-link-type="idl-name" href="#enumdef-mlconvtranspose2dfilteroperandlayout"><c- n>MLConvTranspose2dFilterOperandLayout</c-></a> <a class="idl-code" data-default="&quot;iohw&quot;" data-link-type="dict-member" data-type="MLConvTranspose2dFilterOperandLayout " href="#dom-mlconvtranspose2doptions-filterlayout"><c- g>filterLayout</c-></a> = "iohw";
+  <a data-link-type="idl-name" href="#mloperand"><c- n>MLOperand</c-></a> <a class="idl-code" data-link-type="dict-member" data-type="MLOperand " href="#dom-mlconvtranspose2doptions-bias"><c- g>bias</c-></a>;
+  <a data-link-type="idl-name" href="#mlactivation"><c- n>MLActivation</c-></a> <a class="idl-code" data-link-type="dict-member" data-type="MLActivation " href="#dom-mlconvtranspose2doptions-activation"><c- g>activation</c-></a>;
 };
 
 <c- b>partial</c-> <c- b>interface</c-> <a class="idl-code" data-link-type="interface" href="#mlgraphbuilder"><c- g>MLGraphBuilder</c-></a> {
-  <a data-link-type="idl-name" href="#mloperand"><c- n>MLOperand</c-></a> <a href="#dom-mlgraphbuilder-convtranspose2d"><code><c- g>convTranspose2d</c-></code></a>(<a data-link-type="idl-name" href="#mloperand"><c- n>MLOperand</c-></a> <a href="#dom-mlgraphbuilder-convtranspose2d-input-filter-options-input"><code><c- g>input</c-></code></a>, <a data-link-type="idl-name" href="#mloperand"><c- n>MLOperand</c-></a> <a href="#dom-mlgraphbuilder-convtranspose2d-input-filter-options-filter"><code><c- g>filter</c-></code></a>,
+  <a data-link-type="idl-name" href="#mloperand"><c- n>MLOperand</c-></a> <a class="idl-code" data-link-type="method" href="#dom-mlgraphbuilder-convtranspose2d"><c- g>convTranspose2d</c-></a>(<a data-link-type="idl-name" href="#mloperand"><c- n>MLOperand</c-></a> <a href="#dom-mlgraphbuilder-convtranspose2d-input-filter-options-input"><code><c- g>input</c-></code></a>, <a data-link-type="idl-name" href="#mloperand"><c- n>MLOperand</c-></a> <a href="#dom-mlgraphbuilder-convtranspose2d-input-filter-options-filter"><code><c- g>filter</c-></code></a>,
                             <c- b>optional</c-> <a data-link-type="idl-name" href="#dictdef-mlconvtranspose2doptions"><c- n>MLConvTranspose2dOptions</c-></a> <a href="#dom-mlgraphbuilder-convtranspose2d-input-filter-options-options"><code><c- g>options</c-></code></a> = {});
 };
 
 <c- b>partial</c-> <c- b>interface</c-> <a class="idl-code" data-link-type="interface" href="#mlgraphbuilder"><c- g>MLGraphBuilder</c-></a> {
-  <a data-link-type="idl-name" href="#mloperand"><c- n>MLOperand</c-></a> <a href="#dom-mlgraphbuilder-add"><code><c- g>add</c-></code></a>(<a data-link-type="idl-name" href="#mloperand"><c- n>MLOperand</c-></a> <a href="#dom-mlgraphbuilder-add-a-b-a"><code><c- g>a</c-></code></a>, <a data-link-type="idl-name" href="#mloperand"><c- n>MLOperand</c-></a> <a href="#dom-mlgraphbuilder-add-a-b-b"><code><c- g>b</c-></code></a>);
-  <a data-link-type="idl-name" href="#mloperand"><c- n>MLOperand</c-></a> <a href="#dom-mlgraphbuilder-sub"><code><c- g>sub</c-></code></a>(<a data-link-type="idl-name" href="#mloperand"><c- n>MLOperand</c-></a> <a href="#dom-mlgraphbuilder-sub-a-b-a"><code><c- g>a</c-></code></a>, <a data-link-type="idl-name" href="#mloperand"><c- n>MLOperand</c-></a> <a href="#dom-mlgraphbuilder-sub-a-b-b"><code><c- g>b</c-></code></a>);
-  <a data-link-type="idl-name" href="#mloperand"><c- n>MLOperand</c-></a> <a href="#dom-mlgraphbuilder-mul"><code><c- g>mul</c-></code></a>(<a data-link-type="idl-name" href="#mloperand"><c- n>MLOperand</c-></a> <a href="#dom-mlgraphbuilder-mul-a-b-a"><code><c- g>a</c-></code></a>, <a data-link-type="idl-name" href="#mloperand"><c- n>MLOperand</c-></a> <a href="#dom-mlgraphbuilder-mul-a-b-b"><code><c- g>b</c-></code></a>);
-  <a data-link-type="idl-name" href="#mloperand"><c- n>MLOperand</c-></a> <a href="#dom-mlgraphbuilder-div"><code><c- g>div</c-></code></a>(<a data-link-type="idl-name" href="#mloperand"><c- n>MLOperand</c-></a> <a href="#dom-mlgraphbuilder-div-a-b-a"><code><c- g>a</c-></code></a>, <a data-link-type="idl-name" href="#mloperand"><c- n>MLOperand</c-></a> <a href="#dom-mlgraphbuilder-div-a-b-b"><code><c- g>b</c-></code></a>);
-  <a data-link-type="idl-name" href="#mloperand"><c- n>MLOperand</c-></a> <a href="#dom-mlgraphbuilder-max"><code><c- g>max</c-></code></a>(<a data-link-type="idl-name" href="#mloperand"><c- n>MLOperand</c-></a> <a href="#dom-mlgraphbuilder-max-a-b-a"><code><c- g>a</c-></code></a>, <a data-link-type="idl-name" href="#mloperand"><c- n>MLOperand</c-></a> <a href="#dom-mlgraphbuilder-max-a-b-b"><code><c- g>b</c-></code></a>);
-  <a data-link-type="idl-name" href="#mloperand"><c- n>MLOperand</c-></a> <a href="#dom-mlgraphbuilder-min"><code><c- g>min</c-></code></a>(<a data-link-type="idl-name" href="#mloperand"><c- n>MLOperand</c-></a> <a href="#dom-mlgraphbuilder-min-a-b-a"><code><c- g>a</c-></code></a>, <a data-link-type="idl-name" href="#mloperand"><c- n>MLOperand</c-></a> <a href="#dom-mlgraphbuilder-min-a-b-b"><code><c- g>b</c-></code></a>);
-  <a data-link-type="idl-name" href="#mloperand"><c- n>MLOperand</c-></a> <a href="#dom-mlgraphbuilder-pow"><code><c- g>pow</c-></code></a>(<a data-link-type="idl-name" href="#mloperand"><c- n>MLOperand</c-></a> <a href="#dom-mlgraphbuilder-pow-a-b-a"><code><c- g>a</c-></code></a>, <a data-link-type="idl-name" href="#mloperand"><c- n>MLOperand</c-></a> <a href="#dom-mlgraphbuilder-pow-a-b-b"><code><c- g>b</c-></code></a>);
+  <a data-link-type="idl-name" href="#mloperand"><c- n>MLOperand</c-></a> <a class="idl-code" data-link-type="method" href="#dom-mlgraphbuilder-add"><c- g>add</c-></a>(<a data-link-type="idl-name" href="#mloperand"><c- n>MLOperand</c-></a> <a href="#dom-mlgraphbuilder-add-a-b-a"><code><c- g>a</c-></code></a>, <a data-link-type="idl-name" href="#mloperand"><c- n>MLOperand</c-></a> <a href="#dom-mlgraphbuilder-add-a-b-b"><code><c- g>b</c-></code></a>);
+  <a data-link-type="idl-name" href="#mloperand"><c- n>MLOperand</c-></a> <a class="idl-code" data-link-type="method" href="#dom-mlgraphbuilder-sub"><c- g>sub</c-></a>(<a data-link-type="idl-name" href="#mloperand"><c- n>MLOperand</c-></a> <a href="#dom-mlgraphbuilder-sub-a-b-a"><code><c- g>a</c-></code></a>, <a data-link-type="idl-name" href="#mloperand"><c- n>MLOperand</c-></a> <a href="#dom-mlgraphbuilder-sub-a-b-b"><code><c- g>b</c-></code></a>);
+  <a data-link-type="idl-name" href="#mloperand"><c- n>MLOperand</c-></a> <a class="idl-code" data-link-type="method" href="#dom-mlgraphbuilder-mul"><c- g>mul</c-></a>(<a data-link-type="idl-name" href="#mloperand"><c- n>MLOperand</c-></a> <a href="#dom-mlgraphbuilder-mul-a-b-a"><code><c- g>a</c-></code></a>, <a data-link-type="idl-name" href="#mloperand"><c- n>MLOperand</c-></a> <a href="#dom-mlgraphbuilder-mul-a-b-b"><code><c- g>b</c-></code></a>);
+  <a data-link-type="idl-name" href="#mloperand"><c- n>MLOperand</c-></a> <a class="idl-code" data-link-type="method" href="#dom-mlgraphbuilder-div"><c- g>div</c-></a>(<a data-link-type="idl-name" href="#mloperand"><c- n>MLOperand</c-></a> <a href="#dom-mlgraphbuilder-div-a-b-a"><code><c- g>a</c-></code></a>, <a data-link-type="idl-name" href="#mloperand"><c- n>MLOperand</c-></a> <a href="#dom-mlgraphbuilder-div-a-b-b"><code><c- g>b</c-></code></a>);
+  <a data-link-type="idl-name" href="#mloperand"><c- n>MLOperand</c-></a> <a class="idl-code" data-link-type="method" href="#dom-mlgraphbuilder-max"><c- g>max</c-></a>(<a data-link-type="idl-name" href="#mloperand"><c- n>MLOperand</c-></a> <a href="#dom-mlgraphbuilder-max-a-b-a"><code><c- g>a</c-></code></a>, <a data-link-type="idl-name" href="#mloperand"><c- n>MLOperand</c-></a> <a href="#dom-mlgraphbuilder-max-a-b-b"><code><c- g>b</c-></code></a>);
+  <a data-link-type="idl-name" href="#mloperand"><c- n>MLOperand</c-></a> <a class="idl-code" data-link-type="method" href="#dom-mlgraphbuilder-min"><c- g>min</c-></a>(<a data-link-type="idl-name" href="#mloperand"><c- n>MLOperand</c-></a> <a href="#dom-mlgraphbuilder-min-a-b-a"><code><c- g>a</c-></code></a>, <a data-link-type="idl-name" href="#mloperand"><c- n>MLOperand</c-></a> <a href="#dom-mlgraphbuilder-min-a-b-b"><code><c- g>b</c-></code></a>);
+  <a data-link-type="idl-name" href="#mloperand"><c- n>MLOperand</c-></a> <a class="idl-code" data-link-type="method" href="#dom-mlgraphbuilder-pow"><c- g>pow</c-></a>(<a data-link-type="idl-name" href="#mloperand"><c- n>MLOperand</c-></a> <a href="#dom-mlgraphbuilder-pow-a-b-a"><code><c- g>a</c-></code></a>, <a data-link-type="idl-name" href="#mloperand"><c- n>MLOperand</c-></a> <a href="#dom-mlgraphbuilder-pow-a-b-b"><code><c- g>b</c-></code></a>);
 };
 
 <c- b>partial</c-> <c- b>interface</c-> <a class="idl-code" data-link-type="interface" href="#mlgraphbuilder"><c- g>MLGraphBuilder</c-></a> {
-  <a data-link-type="idl-name" href="#mloperand"><c- n>MLOperand</c-></a> <a href="#dom-mlgraphbuilder-abs"><code><c- g>abs</c-></code></a>(<a data-link-type="idl-name" href="#mloperand"><c- n>MLOperand</c-></a> <a href="#dom-mlgraphbuilder-abs-x-x"><code><c- g>x</c-></code></a>);
-  <a data-link-type="idl-name" href="#mloperand"><c- n>MLOperand</c-></a> <a href="#dom-mlgraphbuilder-ceil"><code><c- g>ceil</c-></code></a>(<a data-link-type="idl-name" href="#mloperand"><c- n>MLOperand</c-></a> <a href="#dom-mlgraphbuilder-ceil-x-x"><code><c- g>x</c-></code></a>);
-  <a data-link-type="idl-name" href="#mloperand"><c- n>MLOperand</c-></a> <a href="#dom-mlgraphbuilder-cos"><code><c- g>cos</c-></code></a>(<a data-link-type="idl-name" href="#mloperand"><c- n>MLOperand</c-></a> <a href="#dom-mlgraphbuilder-cos-x-x"><code><c- g>x</c-></code></a>);
-  <a data-link-type="idl-name" href="#mloperand"><c- n>MLOperand</c-></a> <a href="#dom-mlgraphbuilder-exp"><code><c- g>exp</c-></code></a>(<a data-link-type="idl-name" href="#mloperand"><c- n>MLOperand</c-></a> <a href="#dom-mlgraphbuilder-exp-x-x"><code><c- g>x</c-></code></a>);
-  <a data-link-type="idl-name" href="#mloperand"><c- n>MLOperand</c-></a> <a href="#dom-mlgraphbuilder-floor"><code><c- g>floor</c-></code></a>(<a data-link-type="idl-name" href="#mloperand"><c- n>MLOperand</c-></a> <a href="#dom-mlgraphbuilder-floor-x-x"><code><c- g>x</c-></code></a>);
-  <a data-link-type="idl-name" href="#mloperand"><c- n>MLOperand</c-></a> <a href="#dom-mlgraphbuilder-log"><code><c- g>log</c-></code></a>(<a data-link-type="idl-name" href="#mloperand"><c- n>MLOperand</c-></a> <a href="#dom-mlgraphbuilder-log-x-x"><code><c- g>x</c-></code></a>);
-  <a data-link-type="idl-name" href="#mloperand"><c- n>MLOperand</c-></a> <a href="#dom-mlgraphbuilder-neg"><code><c- g>neg</c-></code></a>(<a data-link-type="idl-name" href="#mloperand"><c- n>MLOperand</c-></a> <a href="#dom-mlgraphbuilder-neg-x-x"><code><c- g>x</c-></code></a>);
-  <a data-link-type="idl-name" href="#mloperand"><c- n>MLOperand</c-></a> <a href="#dom-mlgraphbuilder-sin"><code><c- g>sin</c-></code></a>(<a data-link-type="idl-name" href="#mloperand"><c- n>MLOperand</c-></a> <a href="#dom-mlgraphbuilder-sin-x-x"><code><c- g>x</c-></code></a>);
-  <a data-link-type="idl-name" href="#mloperand"><c- n>MLOperand</c-></a> <a href="#dom-mlgraphbuilder-tan"><code><c- g>tan</c-></code></a>(<a data-link-type="idl-name" href="#mloperand"><c- n>MLOperand</c-></a> <a href="#dom-mlgraphbuilder-tan-x-x"><code><c- g>x</c-></code></a>);
+  <a data-link-type="idl-name" href="#mloperand"><c- n>MLOperand</c-></a> <a class="idl-code" data-link-type="method" href="#dom-mlgraphbuilder-abs"><c- g>abs</c-></a>(<a data-link-type="idl-name" href="#mloperand"><c- n>MLOperand</c-></a> <a href="#dom-mlgraphbuilder-abs-input-input"><code><c- g>input</c-></code></a>);
+  <a data-link-type="idl-name" href="#mloperand"><c- n>MLOperand</c-></a> <a class="idl-code" data-link-type="method" href="#dom-mlgraphbuilder-ceil"><c- g>ceil</c-></a>(<a data-link-type="idl-name" href="#mloperand"><c- n>MLOperand</c-></a> <a href="#dom-mlgraphbuilder-ceil-input-input"><code><c- g>input</c-></code></a>);
+  <a data-link-type="idl-name" href="#mloperand"><c- n>MLOperand</c-></a> <a class="idl-code" data-link-type="method" href="#dom-mlgraphbuilder-cos"><c- g>cos</c-></a>(<a data-link-type="idl-name" href="#mloperand"><c- n>MLOperand</c-></a> <a href="#dom-mlgraphbuilder-cos-input-input"><code><c- g>input</c-></code></a>);
+  <a data-link-type="idl-name" href="#mloperand"><c- n>MLOperand</c-></a> <a class="idl-code" data-link-type="method" href="#dom-mlgraphbuilder-exp"><c- g>exp</c-></a>(<a data-link-type="idl-name" href="#mloperand"><c- n>MLOperand</c-></a> <a href="#dom-mlgraphbuilder-exp-input-input"><code><c- g>input</c-></code></a>);
+  <a data-link-type="idl-name" href="#mloperand"><c- n>MLOperand</c-></a> <a class="idl-code" data-link-type="method" href="#dom-mlgraphbuilder-floor"><c- g>floor</c-></a>(<a data-link-type="idl-name" href="#mloperand"><c- n>MLOperand</c-></a> <a href="#dom-mlgraphbuilder-floor-input-input"><code><c- g>input</c-></code></a>);
+  <a data-link-type="idl-name" href="#mloperand"><c- n>MLOperand</c-></a> <a class="idl-code" data-link-type="method" href="#dom-mlgraphbuilder-log"><c- g>log</c-></a>(<a data-link-type="idl-name" href="#mloperand"><c- n>MLOperand</c-></a> <a href="#dom-mlgraphbuilder-log-input-input"><code><c- g>input</c-></code></a>);
+  <a data-link-type="idl-name" href="#mloperand"><c- n>MLOperand</c-></a> <a class="idl-code" data-link-type="method" href="#dom-mlgraphbuilder-neg"><c- g>neg</c-></a>(<a data-link-type="idl-name" href="#mloperand"><c- n>MLOperand</c-></a> <a href="#dom-mlgraphbuilder-neg-input-input"><code><c- g>input</c-></code></a>);
+  <a data-link-type="idl-name" href="#mloperand"><c- n>MLOperand</c-></a> <a class="idl-code" data-link-type="method" href="#dom-mlgraphbuilder-sin"><c- g>sin</c-></a>(<a data-link-type="idl-name" href="#mloperand"><c- n>MLOperand</c-></a> <a href="#dom-mlgraphbuilder-sin-input-input"><code><c- g>input</c-></code></a>);
+  <a data-link-type="idl-name" href="#mloperand"><c- n>MLOperand</c-></a> <a class="idl-code" data-link-type="method" href="#dom-mlgraphbuilder-tan"><c- g>tan</c-></a>(<a data-link-type="idl-name" href="#mloperand"><c- n>MLOperand</c-></a> <a href="#dom-mlgraphbuilder-tan-input-input"><code><c- g>input</c-></code></a>);
 };
 
 <c- b>dictionary</c-> <a href="#dictdef-mleluoptions"><code><c- g>MLEluOptions</c-></code></a> {
@@ -5223,20 +8633,20 @@ <h2 class="no-num no-ref heading settled" id="idl-index"><span class="content">I
 };
 
 <c- b>partial</c-> <c- b>interface</c-> <a class="idl-code" data-link-type="interface" href="#mlgraphbuilder"><c- g>MLGraphBuilder</c-></a> {
-  <a data-link-type="idl-name" href="#mloperand"><c- n>MLOperand</c-></a> <a href="#dom-mlgraphbuilder-elu"><code><c- g>elu</c-></code></a>(<a data-link-type="idl-name" href="#mloperand"><c- n>MLOperand</c-></a> <a href="#dom-mlgraphbuilder-elu-x-options-x"><code><c- g>x</c-></code></a>, <c- b>optional</c-> <a data-link-type="idl-name" href="#dictdef-mleluoptions"><c- n>MLEluOptions</c-></a> <a href="#dom-mlgraphbuilder-elu-x-options-options"><code><c- g>options</c-></code></a> = {});
-  <a data-link-type="idl-name" href="#mlactivation"><c- n>MLActivation</c-></a> <a href="#dom-mlgraphbuilder-elu-options"><code><c- g>elu</c-></code></a>(<c- b>optional</c-> <a data-link-type="idl-name" href="#dictdef-mleluoptions"><c- n>MLEluOptions</c-></a> <a href="#dom-mlgraphbuilder-elu-options-options"><code><c- g>options</c-></code></a> = {});
+  <a data-link-type="idl-name" href="#mloperand"><c- n>MLOperand</c-></a> <a class="idl-code" data-link-type="method" href="#dom-mlgraphbuilder-elu"><c- g>elu</c-></a>(<a data-link-type="idl-name" href="#mloperand"><c- n>MLOperand</c-></a> <a href="#dom-mlgraphbuilder-elu-input-options-input"><code><c- g>input</c-></code></a>, <c- b>optional</c-> <a data-link-type="idl-name" href="#dictdef-mleluoptions"><c- n>MLEluOptions</c-></a> <a href="#dom-mlgraphbuilder-elu-input-options-options"><code><c- g>options</c-></code></a> = {});
+  <a data-link-type="idl-name" href="#mlactivation"><c- n>MLActivation</c-></a> <a class="idl-code" data-link-type="method" href="#dom-mlgraphbuilder-elu-options"><c- g>elu</c-></a>(<c- b>optional</c-> <a data-link-type="idl-name" href="#dictdef-mleluoptions"><c- n>MLEluOptions</c-></a> <a href="#dom-mlgraphbuilder-elu-options-options"><code><c- g>options</c-></code></a> = {});
 };
 
 <c- b>dictionary</c-> <a href="#dictdef-mlgemmoptions"><code><c- g>MLGemmOptions</c-></code></a> {
-  <a data-link-type="idl-name" href="#mloperand"><c- n>MLOperand</c-></a> <a data-type="MLOperand " href="#dom-mlgemmoptions-c"><code><c- g>c</c-></code></a>;
-  <a class="idl-code" data-link-type="interface" href="https://webidl.spec.whatwg.org/#idl-float"><c- b>float</c-></a> <a data-default="1.0" data-type="float " href="#dom-mlgemmoptions-alpha"><code><c- g>alpha</c-></code></a> = 1.0;
-  <a class="idl-code" data-link-type="interface" href="https://webidl.spec.whatwg.org/#idl-float"><c- b>float</c-></a> <a data-default="1.0" data-type="float " href="#dom-mlgemmoptions-beta"><code><c- g>beta</c-></code></a> = 1.0;
-  <a class="idl-code" data-link-type="interface" href="https://webidl.spec.whatwg.org/#idl-boolean"><c- b>boolean</c-></a> <a data-default="false" data-type="boolean " href="#dom-mlgemmoptions-atranspose"><code><c- g>aTranspose</c-></code></a> = <c- b>false</c->;
-  <a class="idl-code" data-link-type="interface" href="https://webidl.spec.whatwg.org/#idl-boolean"><c- b>boolean</c-></a> <a data-default="false" data-type="boolean " href="#dom-mlgemmoptions-btranspose"><code><c- g>bTranspose</c-></code></a> = <c- b>false</c->;
+  <a data-link-type="idl-name" href="#mloperand"><c- n>MLOperand</c-></a> <a class="idl-code" data-link-type="dict-member" data-type="MLOperand " href="#dom-mlgemmoptions-c"><c- g>c</c-></a>;
+  <a class="idl-code" data-link-type="interface" href="https://webidl.spec.whatwg.org/#idl-float"><c- b>float</c-></a> <a class="idl-code" data-default="1.0" data-link-type="dict-member" data-type="float " href="#dom-mlgemmoptions-alpha"><c- g>alpha</c-></a> = 1.0;
+  <a class="idl-code" data-link-type="interface" href="https://webidl.spec.whatwg.org/#idl-float"><c- b>float</c-></a> <a class="idl-code" data-default="1.0" data-link-type="dict-member" data-type="float " href="#dom-mlgemmoptions-beta"><c- g>beta</c-></a> = 1.0;
+  <a class="idl-code" data-link-type="interface" href="https://webidl.spec.whatwg.org/#idl-boolean"><c- b>boolean</c-></a> <a class="idl-code" data-default="false" data-link-type="dict-member" data-type="boolean " href="#dom-mlgemmoptions-atranspose"><c- g>aTranspose</c-></a> = <c- b>false</c->;
+  <a class="idl-code" data-link-type="interface" href="https://webidl.spec.whatwg.org/#idl-boolean"><c- b>boolean</c-></a> <a class="idl-code" data-default="false" data-link-type="dict-member" data-type="boolean " href="#dom-mlgemmoptions-btranspose"><c- g>bTranspose</c-></a> = <c- b>false</c->;
 };
 
 <c- b>partial</c-> <c- b>interface</c-> <a class="idl-code" data-link-type="interface" href="#mlgraphbuilder"><c- g>MLGraphBuilder</c-></a> {
-  <a data-link-type="idl-name" href="#mloperand"><c- n>MLOperand</c-></a> <a href="#dom-mlgraphbuilder-gemm"><code><c- g>gemm</c-></code></a>(<a data-link-type="idl-name" href="#mloperand"><c- n>MLOperand</c-></a> <a href="#dom-mlgraphbuilder-gemm-a-b-options-a"><code><c- g>a</c-></code></a>, <a data-link-type="idl-name" href="#mloperand"><c- n>MLOperand</c-></a> <a href="#dom-mlgraphbuilder-gemm-a-b-options-b"><code><c- g>b</c-></code></a>, <c- b>optional</c-> <a data-link-type="idl-name" href="#dictdef-mlgemmoptions"><c- n>MLGemmOptions</c-></a> <a href="#dom-mlgraphbuilder-gemm-a-b-options-options"><code><c- g>options</c-></code></a> = {});
+  <a data-link-type="idl-name" href="#mloperand"><c- n>MLOperand</c-></a> <a class="idl-code" data-link-type="method" href="#dom-mlgraphbuilder-gemm"><c- g>gemm</c-></a>(<a data-link-type="idl-name" href="#mloperand"><c- n>MLOperand</c-></a> <a href="#dom-mlgraphbuilder-gemm-a-b-options-a"><code><c- g>a</c-></code></a>, <a data-link-type="idl-name" href="#mloperand"><c- n>MLOperand</c-></a> <a href="#dom-mlgraphbuilder-gemm-a-b-options-b"><code><c- g>b</c-></code></a>, <c- b>optional</c-> <a data-link-type="idl-name" href="#dictdef-mlgemmoptions"><c- n>MLGemmOptions</c-></a> <a href="#dom-mlgraphbuilder-gemm-a-b-options-options"><code><c- g>options</c-></code></a> = {});
 };
 
 <c- b>enum</c-> <a href="#enumdef-mlgruweightlayout"><code><c- g>MLGruWeightLayout</c-></code></a> {
@@ -5251,80 +8661,80 @@ <h2 class="no-num no-ref heading settled" id="idl-index"><span class="content">I
 };
 
 <c- b>dictionary</c-> <a href="#dictdef-mlgruoptions"><code><c- g>MLGruOptions</c-></code></a> {
-  <a data-link-type="idl-name" href="#mloperand"><c- n>MLOperand</c-></a> <a data-type="MLOperand " href="#dom-mlgruoptions-bias"><code><c- g>bias</c-></code></a>;
-  <a data-link-type="idl-name" href="#mloperand"><c- n>MLOperand</c-></a> <a data-type="MLOperand " href="#dom-mlgruoptions-recurrentbias"><code><c- g>recurrentBias</c-></code></a>;
-  <a data-link-type="idl-name" href="#mloperand"><c- n>MLOperand</c-></a> <a data-type="MLOperand " href="#dom-mlgruoptions-initialhiddenstate"><code><c- g>initialHiddenState</c-></code></a>;
-  <a class="idl-code" data-link-type="interface" href="https://webidl.spec.whatwg.org/#idl-boolean"><c- b>boolean</c-></a> <a data-default="true" data-type="boolean " href="#dom-mlgruoptions-resetafter"><code><c- g>resetAfter</c-></code></a> = <c- b>true</c->;
-  <a class="idl-code" data-link-type="interface" href="https://webidl.spec.whatwg.org/#idl-boolean"><c- b>boolean</c-></a> <a data-default="false" data-type="boolean " href="#dom-mlgruoptions-returnsequence"><code><c- g>returnSequence</c-></code></a> = <c- b>false</c->;
-  <a data-link-type="idl-name" href="#enumdef-mlrecurrentnetworkdirection"><c- n>MLRecurrentNetworkDirection</c-></a> <a data-default="&quot;forward&quot;" data-type="MLRecurrentNetworkDirection " href="#dom-mlgruoptions-direction"><code><c- g>direction</c-></code></a> = "forward";
-  <a data-link-type="idl-name" href="#enumdef-mlgruweightlayout"><c- n>MLGruWeightLayout</c-></a> <a data-default="&quot;zrn&quot;" data-type="MLGruWeightLayout " href="#dom-mlgruoptions-layout"><code><c- g>layout</c-></code></a> = "zrn";
-  <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#idl-sequence"><c- b>sequence</c-></a>&lt;<a data-link-type="idl-name" href="#mlactivation"><c- n>MLActivation</c-></a>> <a data-type="sequence<MLActivation> " href="#dom-mlgruoptions-activations"><code><c- g>activations</c-></code></a>;
+  <a data-link-type="idl-name" href="#mloperand"><c- n>MLOperand</c-></a> <a class="idl-code" data-link-type="dict-member" data-type="MLOperand " href="#dom-mlgruoptions-bias"><c- g>bias</c-></a>;
+  <a data-link-type="idl-name" href="#mloperand"><c- n>MLOperand</c-></a> <a class="idl-code" data-link-type="dict-member" data-type="MLOperand " href="#dom-mlgruoptions-recurrentbias"><c- g>recurrentBias</c-></a>;
+  <a data-link-type="idl-name" href="#mloperand"><c- n>MLOperand</c-></a> <a class="idl-code" data-link-type="dict-member" data-type="MLOperand " href="#dom-mlgruoptions-initialhiddenstate"><c- g>initialHiddenState</c-></a>;
+  <a class="idl-code" data-link-type="interface" href="https://webidl.spec.whatwg.org/#idl-boolean"><c- b>boolean</c-></a> <a class="idl-code" data-default="true" data-link-type="dict-member" data-type="boolean " href="#dom-mlgruoptions-resetafter"><c- g>resetAfter</c-></a> = <c- b>true</c->;
+  <a class="idl-code" data-link-type="interface" href="https://webidl.spec.whatwg.org/#idl-boolean"><c- b>boolean</c-></a> <a class="idl-code" data-default="false" data-link-type="dict-member" data-type="boolean " href="#dom-mlgruoptions-returnsequence"><c- g>returnSequence</c-></a> = <c- b>false</c->;
+  <a data-link-type="idl-name" href="#enumdef-mlrecurrentnetworkdirection"><c- n>MLRecurrentNetworkDirection</c-></a> <a class="idl-code" data-default="&quot;forward&quot;" data-link-type="dict-member" data-type="MLRecurrentNetworkDirection " href="#dom-mlgruoptions-direction"><c- g>direction</c-></a> = "forward";
+  <a data-link-type="idl-name" href="#enumdef-mlgruweightlayout"><c- n>MLGruWeightLayout</c-></a> <a class="idl-code" data-default="&quot;zrn&quot;" data-link-type="dict-member" data-type="MLGruWeightLayout " href="#dom-mlgruoptions-layout"><c- g>layout</c-></a> = "zrn";
+  <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#idl-sequence"><c- b>sequence</c-></a>&lt;<a data-link-type="idl-name" href="#mlactivation"><c- n>MLActivation</c-></a>> <a class="idl-code" data-link-type="dict-member" data-type="sequence<MLActivation> " href="#dom-mlgruoptions-activations"><c- g>activations</c-></a>;
 };
 
 <c- b>partial</c-> <c- b>interface</c-> <a class="idl-code" data-link-type="interface" href="#mlgraphbuilder"><c- g>MLGraphBuilder</c-></a> {
-  <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#idl-sequence"><c- b>sequence</c-></a>&lt;<a data-link-type="idl-name" href="#mloperand"><c- n>MLOperand</c-></a>> <a href="#dom-mlgraphbuilder-gru"><code><c- g>gru</c-></code></a>(<a data-link-type="idl-name" href="#mloperand"><c- n>MLOperand</c-></a> <a href="#dom-mlgraphbuilder-gru-input-weight-recurrentweight-steps-hiddensize-options-input"><code><c- g>input</c-></code></a>, <a data-link-type="idl-name" href="#mloperand"><c- n>MLOperand</c-></a> <a href="#dom-mlgraphbuilder-gru-input-weight-recurrentweight-steps-hiddensize-options-weight"><code><c- g>weight</c-></code></a>, <a data-link-type="idl-name" href="#mloperand"><c- n>MLOperand</c-></a> <a href="#dom-mlgraphbuilder-gru-input-weight-recurrentweight-steps-hiddensize-options-recurrentweight"><code><c- g>recurrentWeight</c-></code></a>,
+  <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#idl-sequence"><c- b>sequence</c-></a>&lt;<a data-link-type="idl-name" href="#mloperand"><c- n>MLOperand</c-></a>> <a class="idl-code" data-link-type="method" href="#dom-mlgraphbuilder-gru"><c- g>gru</c-></a>(<a data-link-type="idl-name" href="#mloperand"><c- n>MLOperand</c-></a> <a href="#dom-mlgraphbuilder-gru-input-weight-recurrentweight-steps-hiddensize-options-input"><code><c- g>input</c-></code></a>, <a data-link-type="idl-name" href="#mloperand"><c- n>MLOperand</c-></a> <a href="#dom-mlgraphbuilder-gru-input-weight-recurrentweight-steps-hiddensize-options-weight"><code><c- g>weight</c-></code></a>, <a data-link-type="idl-name" href="#mloperand"><c- n>MLOperand</c-></a> <a href="#dom-mlgraphbuilder-gru-input-weight-recurrentweight-steps-hiddensize-options-recurrentweight"><code><c- g>recurrentWeight</c-></code></a>,
                           <a class="idl-code" data-link-type="interface" href="https://webidl.spec.whatwg.org/#idl-unsigned-long"><c- b>unsigned</c-> <c- b>long</c-></a> <a href="#dom-mlgraphbuilder-gru-input-weight-recurrentweight-steps-hiddensize-options-steps"><code><c- g>steps</c-></code></a>, <a class="idl-code" data-link-type="interface" href="https://webidl.spec.whatwg.org/#idl-unsigned-long"><c- b>unsigned</c-> <c- b>long</c-></a> <a href="#dom-mlgraphbuilder-gru-input-weight-recurrentweight-steps-hiddensize-options-hiddensize"><code><c- g>hiddenSize</c-></code></a>,
                           <c- b>optional</c-> <a data-link-type="idl-name" href="#dictdef-mlgruoptions"><c- n>MLGruOptions</c-></a> <a href="#dom-mlgraphbuilder-gru-input-weight-recurrentweight-steps-hiddensize-options-options"><code><c- g>options</c-></code></a> = {});
 };
 
 <c- b>dictionary</c-> <a href="#dictdef-mlgrucelloptions"><code><c- g>MLGruCellOptions</c-></code></a> {
-  <a data-link-type="idl-name" href="#mloperand"><c- n>MLOperand</c-></a> <a data-type="MLOperand " href="#dom-mlgrucelloptions-bias"><code><c- g>bias</c-></code></a>;
-  <a data-link-type="idl-name" href="#mloperand"><c- n>MLOperand</c-></a> <a data-type="MLOperand " href="#dom-mlgrucelloptions-recurrentbias"><code><c- g>recurrentBias</c-></code></a>;
-  <a class="idl-code" data-link-type="interface" href="https://webidl.spec.whatwg.org/#idl-boolean"><c- b>boolean</c-></a> <a data-default="true" data-type="boolean " href="#dom-mlgrucelloptions-resetafter"><code><c- g>resetAfter</c-></code></a> = <c- b>true</c->;
-  <a data-link-type="idl-name" href="#enumdef-mlgruweightlayout"><c- n>MLGruWeightLayout</c-></a> <a data-default="&quot;zrn&quot;" data-type="MLGruWeightLayout " href="#dom-mlgrucelloptions-layout"><code><c- g>layout</c-></code></a> = "zrn";
-  <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#idl-sequence"><c- b>sequence</c-></a>&lt;<a data-link-type="idl-name" href="#mlactivation"><c- n>MLActivation</c-></a>> <a data-type="sequence<MLActivation> " href="#dom-mlgrucelloptions-activations"><code><c- g>activations</c-></code></a>;
+  <a data-link-type="idl-name" href="#mloperand"><c- n>MLOperand</c-></a> <a class="idl-code" data-link-type="dict-member" data-type="MLOperand " href="#dom-mlgrucelloptions-bias"><c- g>bias</c-></a>;
+  <a data-link-type="idl-name" href="#mloperand"><c- n>MLOperand</c-></a> <a class="idl-code" data-link-type="dict-member" data-type="MLOperand " href="#dom-mlgrucelloptions-recurrentbias"><c- g>recurrentBias</c-></a>;
+  <a class="idl-code" data-link-type="interface" href="https://webidl.spec.whatwg.org/#idl-boolean"><c- b>boolean</c-></a> <a class="idl-code" data-default="true" data-link-type="dict-member" data-type="boolean " href="#dom-mlgrucelloptions-resetafter"><c- g>resetAfter</c-></a> = <c- b>true</c->;
+  <a data-link-type="idl-name" href="#enumdef-mlgruweightlayout"><c- n>MLGruWeightLayout</c-></a> <a class="idl-code" data-default="&quot;zrn&quot;" data-link-type="dict-member" data-type="MLGruWeightLayout " href="#dom-mlgrucelloptions-layout"><c- g>layout</c-></a> = "zrn";
+  <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#idl-sequence"><c- b>sequence</c-></a>&lt;<a data-link-type="idl-name" href="#mlactivation"><c- n>MLActivation</c-></a>> <a class="idl-code" data-link-type="dict-member" data-type="sequence<MLActivation> " href="#dom-mlgrucelloptions-activations"><c- g>activations</c-></a>;
 };
 
 <c- b>partial</c-> <c- b>interface</c-> <a class="idl-code" data-link-type="interface" href="#mlgraphbuilder"><c- g>MLGraphBuilder</c-></a> {
-  <a data-link-type="idl-name" href="#mloperand"><c- n>MLOperand</c-></a> <a href="#dom-mlgraphbuilder-grucell"><code><c- g>gruCell</c-></code></a>(<a data-link-type="idl-name" href="#mloperand"><c- n>MLOperand</c-></a> <a href="#dom-mlgraphbuilder-grucell-input-weight-recurrentweight-hiddenstate-hiddensize-options-input"><code><c- g>input</c-></code></a>, <a data-link-type="idl-name" href="#mloperand"><c- n>MLOperand</c-></a> <a href="#dom-mlgraphbuilder-grucell-input-weight-recurrentweight-hiddenstate-hiddensize-options-weight"><code><c- g>weight</c-></code></a>, <a data-link-type="idl-name" href="#mloperand"><c- n>MLOperand</c-></a> <a href="#dom-mlgraphbuilder-grucell-input-weight-recurrentweight-hiddenstate-hiddensize-options-recurrentweight"><code><c- g>recurrentWeight</c-></code></a>,
+  <a data-link-type="idl-name" href="#mloperand"><c- n>MLOperand</c-></a> <a class="idl-code" data-link-type="method" href="#dom-mlgraphbuilder-grucell"><c- g>gruCell</c-></a>(<a data-link-type="idl-name" href="#mloperand"><c- n>MLOperand</c-></a> <a href="#dom-mlgraphbuilder-grucell-input-weight-recurrentweight-hiddenstate-hiddensize-options-input"><code><c- g>input</c-></code></a>, <a data-link-type="idl-name" href="#mloperand"><c- n>MLOperand</c-></a> <a href="#dom-mlgraphbuilder-grucell-input-weight-recurrentweight-hiddenstate-hiddensize-options-weight"><code><c- g>weight</c-></code></a>, <a data-link-type="idl-name" href="#mloperand"><c- n>MLOperand</c-></a> <a href="#dom-mlgraphbuilder-grucell-input-weight-recurrentweight-hiddenstate-hiddensize-options-recurrentweight"><code><c- g>recurrentWeight</c-></code></a>,
                     <a data-link-type="idl-name" href="#mloperand"><c- n>MLOperand</c-></a> <a href="#dom-mlgraphbuilder-grucell-input-weight-recurrentweight-hiddenstate-hiddensize-options-hiddenstate"><code><c- g>hiddenState</c-></code></a>, <a class="idl-code" data-link-type="interface" href="https://webidl.spec.whatwg.org/#idl-unsigned-long"><c- b>unsigned</c-> <c- b>long</c-></a> <a href="#dom-mlgraphbuilder-grucell-input-weight-recurrentweight-hiddenstate-hiddensize-options-hiddensize"><code><c- g>hiddenSize</c-></code></a>,
                     <c- b>optional</c-> <a data-link-type="idl-name" href="#dictdef-mlgrucelloptions"><c- n>MLGruCellOptions</c-></a> <a href="#dom-mlgraphbuilder-grucell-input-weight-recurrentweight-hiddenstate-hiddensize-options-options"><code><c- g>options</c-></code></a> = {});
 };
 
 <c- b>dictionary</c-> <a href="#dictdef-mlhardsigmoidoptions"><code><c- g>MLHardSigmoidOptions</c-></code></a> {
-  <a class="idl-code" data-link-type="interface" href="https://webidl.spec.whatwg.org/#idl-float"><c- b>float</c-></a> <a data-default="0.2" data-type="float " href="#dom-mlhardsigmoidoptions-alpha"><code><c- g>alpha</c-></code></a> = 0.2;
-  <a class="idl-code" data-link-type="interface" href="https://webidl.spec.whatwg.org/#idl-float"><c- b>float</c-></a> <a data-default="0.5" data-type="float " href="#dom-mlhardsigmoidoptions-beta"><code><c- g>beta</c-></code></a> = 0.5;
+  <a class="idl-code" data-link-type="interface" href="https://webidl.spec.whatwg.org/#idl-float"><c- b>float</c-></a> <a class="idl-code" data-default="0.2" data-link-type="dict-member" data-type="float " href="#dom-mlhardsigmoidoptions-alpha"><c- g>alpha</c-></a> = 0.2;
+  <a class="idl-code" data-link-type="interface" href="https://webidl.spec.whatwg.org/#idl-float"><c- b>float</c-></a> <a class="idl-code" data-default="0.5" data-link-type="dict-member" data-type="float " href="#dom-mlhardsigmoidoptions-beta"><c- g>beta</c-></a> = 0.5;
 };
 
 <c- b>partial</c-> <c- b>interface</c-> <a class="idl-code" data-link-type="interface" href="#mlgraphbuilder"><c- g>MLGraphBuilder</c-></a> {
-  <a data-link-type="idl-name" href="#mloperand"><c- n>MLOperand</c-></a> <a href="#dom-mlgraphbuilder-hardsigmoid"><code><c- g>hardSigmoid</c-></code></a>(<a data-link-type="idl-name" href="#mloperand"><c- n>MLOperand</c-></a> <a href="#dom-mlgraphbuilder-hardsigmoid-x-options-x"><code><c- g>x</c-></code></a>, <c- b>optional</c-> <a data-link-type="idl-name" href="#dictdef-mlhardsigmoidoptions"><c- n>MLHardSigmoidOptions</c-></a> <a href="#dom-mlgraphbuilder-hardsigmoid-x-options-options"><code><c- g>options</c-></code></a> = {});
-  <a data-link-type="idl-name" href="#mlactivation"><c- n>MLActivation</c-></a> <a href="#dom-mlgraphbuilder-hardsigmoid-options"><code><c- g>hardSigmoid</c-></code></a>(<c- b>optional</c-> <a data-link-type="idl-name" href="#dictdef-mlhardsigmoidoptions"><c- n>MLHardSigmoidOptions</c-></a> <a href="#dom-mlgraphbuilder-hardsigmoid-options-options"><code><c- g>options</c-></code></a> = {});
+  <a data-link-type="idl-name" href="#mloperand"><c- n>MLOperand</c-></a> <a class="idl-code" data-link-type="method" href="#dom-mlgraphbuilder-hardsigmoid"><c- g>hardSigmoid</c-></a>(<a data-link-type="idl-name" href="#mloperand"><c- n>MLOperand</c-></a> <a href="#dom-mlgraphbuilder-hardsigmoid-input-options-input"><code><c- g>input</c-></code></a>, <c- b>optional</c-> <a data-link-type="idl-name" href="#dictdef-mlhardsigmoidoptions"><c- n>MLHardSigmoidOptions</c-></a> <a href="#dom-mlgraphbuilder-hardsigmoid-input-options-options"><code><c- g>options</c-></code></a> = {});
+  <a data-link-type="idl-name" href="#mlactivation"><c- n>MLActivation</c-></a> <a class="idl-code" data-link-type="method" href="#dom-mlgraphbuilder-hardsigmoid-options"><c- g>hardSigmoid</c-></a>(<c- b>optional</c-> <a data-link-type="idl-name" href="#dictdef-mlhardsigmoidoptions"><c- n>MLHardSigmoidOptions</c-></a> <a href="#dom-mlgraphbuilder-hardsigmoid-options-options"><code><c- g>options</c-></code></a> = {});
 };
 
 <c- b>partial</c-> <c- b>interface</c-> <a class="idl-code" data-link-type="interface" href="#mlgraphbuilder"><c- g>MLGraphBuilder</c-></a> {
-  <a data-link-type="idl-name" href="#mloperand"><c- n>MLOperand</c-></a> <a href="#dom-mlgraphbuilder-hardswish"><code><c- g>hardSwish</c-></code></a>(<a data-link-type="idl-name" href="#mloperand"><c- n>MLOperand</c-></a> <a href="#dom-mlgraphbuilder-hardswish-x-x"><code><c- g>x</c-></code></a>);
-  <a data-link-type="idl-name" href="#mlactivation"><c- n>MLActivation</c-></a> <a href="#dom-mlgraphbuilder-hardswish①"><code><c- g>hardSwish</c-></code></a>();
+  <a data-link-type="idl-name" href="#mloperand"><c- n>MLOperand</c-></a> <a class="idl-code" data-link-type="method" href="#dom-mlgraphbuilder-hardswish"><c- g>hardSwish</c-></a>(<a data-link-type="idl-name" href="#mloperand"><c- n>MLOperand</c-></a> <a href="#dom-mlgraphbuilder-hardswish-input-input"><code><c- g>input</c-></code></a>);
+  <a data-link-type="idl-name" href="#mlactivation"><c- n>MLActivation</c-></a> <a class="idl-code" data-link-type="method" href="#hardswish-noargs"><c- g>hardSwish</c-></a>();
 };
 
 <c- b>dictionary</c-> <a href="#dictdef-mlinstancenormalizationoptions"><code><c- g>MLInstanceNormalizationOptions</c-></code></a> {
-  <a data-link-type="idl-name" href="#mloperand"><c- n>MLOperand</c-></a> <a data-type="MLOperand " href="#dom-mlinstancenormalizationoptions-scale"><code><c- g>scale</c-></code></a>;
-  <a data-link-type="idl-name" href="#mloperand"><c- n>MLOperand</c-></a> <a data-type="MLOperand " href="#dom-mlinstancenormalizationoptions-bias"><code><c- g>bias</c-></code></a>;
-  <a class="idl-code" data-link-type="interface" href="https://webidl.spec.whatwg.org/#idl-float"><c- b>float</c-></a> <a data-default="1e-5" data-type="float " href="#dom-mlinstancenormalizationoptions-epsilon"><code><c- g>epsilon</c-></code></a> = 1e-5;
-  <a data-link-type="idl-name" href="#enumdef-mlinputoperandlayout"><c- n>MLInputOperandLayout</c-></a> <a data-default="&quot;nchw&quot;" data-type="MLInputOperandLayout " href="#dom-mlinstancenormalizationoptions-layout"><code><c- g>layout</c-></code></a> = "nchw";
+  <a data-link-type="idl-name" href="#mloperand"><c- n>MLOperand</c-></a> <a class="idl-code" data-link-type="dict-member" data-type="MLOperand " href="#dom-mlinstancenormalizationoptions-scale"><c- g>scale</c-></a>;
+  <a data-link-type="idl-name" href="#mloperand"><c- n>MLOperand</c-></a> <a class="idl-code" data-link-type="dict-member" data-type="MLOperand " href="#dom-mlinstancenormalizationoptions-bias"><c- g>bias</c-></a>;
+  <a class="idl-code" data-link-type="interface" href="https://webidl.spec.whatwg.org/#idl-float"><c- b>float</c-></a> <a class="idl-code" data-default="1e-5" data-link-type="dict-member" data-type="float " href="#dom-mlinstancenormalizationoptions-epsilon"><c- g>epsilon</c-></a> = 1e-5;
+  <a data-link-type="idl-name" href="#enumdef-mlinputoperandlayout"><c- n>MLInputOperandLayout</c-></a> <a class="idl-code" data-default="&quot;nchw&quot;" data-link-type="dict-member" data-type="MLInputOperandLayout " href="#dom-mlinstancenormalizationoptions-layout"><c- g>layout</c-></a> = "nchw";
 };
 
 <c- b>partial</c-> <c- b>interface</c-> <a class="idl-code" data-link-type="interface" href="#mlgraphbuilder"><c- g>MLGraphBuilder</c-></a> {
-  <a data-link-type="idl-name" href="#mloperand"><c- n>MLOperand</c-></a> <a href="#dom-mlgraphbuilder-instancenormalization"><code><c- g>instanceNormalization</c-></code></a>(<a data-link-type="idl-name" href="#mloperand"><c- n>MLOperand</c-></a> <a href="#dom-mlgraphbuilder-instancenormalization-input-options-input"><code><c- g>input</c-></code></a>,
+  <a data-link-type="idl-name" href="#mloperand"><c- n>MLOperand</c-></a> <a class="idl-code" data-link-type="method" href="#dom-mlgraphbuilder-instancenormalization"><c- g>instanceNormalization</c-></a>(<a data-link-type="idl-name" href="#mloperand"><c- n>MLOperand</c-></a> <a href="#dom-mlgraphbuilder-instancenormalization-input-options-input"><code><c- g>input</c-></code></a>,
                                 <c- b>optional</c-> <a data-link-type="idl-name" href="#dictdef-mlinstancenormalizationoptions"><c- n>MLInstanceNormalizationOptions</c-></a> <a href="#dom-mlgraphbuilder-instancenormalization-input-options-options"><code><c- g>options</c-></code></a> = {});
 };
 
 <c- b>dictionary</c-> <a href="#dictdef-mlleakyreluoptions"><code><c- g>MLLeakyReluOptions</c-></code></a> {
-  <a class="idl-code" data-link-type="interface" href="https://webidl.spec.whatwg.org/#idl-float"><c- b>float</c-></a> <a data-default="0.01" data-type="float " href="#dom-mlleakyreluoptions-alpha"><code><c- g>alpha</c-></code></a> = 0.01;
+  <a class="idl-code" data-link-type="interface" href="https://webidl.spec.whatwg.org/#idl-float"><c- b>float</c-></a> <a class="idl-code" data-default="0.01" data-link-type="dict-member" data-type="float " href="#dom-mlleakyreluoptions-alpha"><c- g>alpha</c-></a> = 0.01;
 };
 
 <c- b>partial</c-> <c- b>interface</c-> <a class="idl-code" data-link-type="interface" href="#mlgraphbuilder"><c- g>MLGraphBuilder</c-></a> {
-  <a data-link-type="idl-name" href="#mloperand"><c- n>MLOperand</c-></a> <a href="#dom-mlgraphbuilder-leakyrelu"><code><c- g>leakyRelu</c-></code></a>(<a data-link-type="idl-name" href="#mloperand"><c- n>MLOperand</c-></a> <a href="#dom-mlgraphbuilder-leakyrelu-x-options-x"><code><c- g>x</c-></code></a>, <c- b>optional</c-> <a data-link-type="idl-name" href="#dictdef-mlleakyreluoptions"><c- n>MLLeakyReluOptions</c-></a> <a href="#dom-mlgraphbuilder-leakyrelu-x-options-options"><code><c- g>options</c-></code></a> = {});
-  <a data-link-type="idl-name" href="#mlactivation"><c- n>MLActivation</c-></a> <a href="#dom-mlgraphbuilder-leakyrelu-options"><code><c- g>leakyRelu</c-></code></a>(<c- b>optional</c-> <a data-link-type="idl-name" href="#dictdef-mlleakyreluoptions"><c- n>MLLeakyReluOptions</c-></a> <a href="#dom-mlgraphbuilder-leakyrelu-options-options"><code><c- g>options</c-></code></a> = {});
+  <a data-link-type="idl-name" href="#mloperand"><c- n>MLOperand</c-></a> <a class="idl-code" data-link-type="method" href="#dom-mlgraphbuilder-leakyrelu"><c- g>leakyRelu</c-></a>(<a data-link-type="idl-name" href="#mloperand"><c- n>MLOperand</c-></a> <a href="#dom-mlgraphbuilder-leakyrelu-input-options-input"><code><c- g>input</c-></code></a>, <c- b>optional</c-> <a data-link-type="idl-name" href="#dictdef-mlleakyreluoptions"><c- n>MLLeakyReluOptions</c-></a> <a href="#dom-mlgraphbuilder-leakyrelu-input-options-options"><code><c- g>options</c-></code></a> = {});
+  <a data-link-type="idl-name" href="#mlactivation"><c- n>MLActivation</c-></a> <a class="idl-code" data-link-type="method" href="#dom-mlgraphbuilder-leakyrelu-options"><c- g>leakyRelu</c-></a>(<c- b>optional</c-> <a data-link-type="idl-name" href="#dictdef-mlleakyreluoptions"><c- n>MLLeakyReluOptions</c-></a> <a href="#dom-mlgraphbuilder-leakyrelu-options-options"><code><c- g>options</c-></code></a> = {});
 };
 
 <c- b>dictionary</c-> <a href="#dictdef-mllinearoptions"><code><c- g>MLLinearOptions</c-></code></a> {
-  <a class="idl-code" data-link-type="interface" href="https://webidl.spec.whatwg.org/#idl-float"><c- b>float</c-></a> <a data-default="1" data-type="float " href="#dom-mllinearoptions-alpha"><code><c- g>alpha</c-></code></a> = 1;
-  <a class="idl-code" data-link-type="interface" href="https://webidl.spec.whatwg.org/#idl-float"><c- b>float</c-></a> <a data-default="0" data-type="float " href="#dom-mllinearoptions-beta"><code><c- g>beta</c-></code></a> = 0;
+  <a class="idl-code" data-link-type="interface" href="https://webidl.spec.whatwg.org/#idl-float"><c- b>float</c-></a> <a class="idl-code" data-default="1" data-link-type="dict-member" data-type="float " href="#dom-mllinearoptions-alpha"><c- g>alpha</c-></a> = 1;
+  <a class="idl-code" data-link-type="interface" href="https://webidl.spec.whatwg.org/#idl-float"><c- b>float</c-></a> <a class="idl-code" data-default="0" data-link-type="dict-member" data-type="float " href="#dom-mllinearoptions-beta"><c- g>beta</c-></a> = 0;
 };
 
 <c- b>partial</c-> <c- b>interface</c-> <a class="idl-code" data-link-type="interface" href="#mlgraphbuilder"><c- g>MLGraphBuilder</c-></a> {
-  <a data-link-type="idl-name" href="#mloperand"><c- n>MLOperand</c-></a> <a href="#dom-mlgraphbuilder-linear"><code><c- g>linear</c-></code></a>(<a data-link-type="idl-name" href="#mloperand"><c- n>MLOperand</c-></a> <a href="#dom-mlgraphbuilder-linear-x-options-x"><code><c- g>x</c-></code></a>, <c- b>optional</c-> <a data-link-type="idl-name" href="#dictdef-mllinearoptions"><c- n>MLLinearOptions</c-></a> <a href="#dom-mlgraphbuilder-linear-x-options-options"><code><c- g>options</c-></code></a> = {});
-  <a data-link-type="idl-name" href="#mlactivation"><c- n>MLActivation</c-></a> <a href="#dom-mlgraphbuilder-linear-options"><code><c- g>linear</c-></code></a>(<c- b>optional</c-> <a data-link-type="idl-name" href="#dictdef-mllinearoptions"><c- n>MLLinearOptions</c-></a> <a href="#dom-mlgraphbuilder-linear-options-options"><code><c- g>options</c-></code></a> = {});
+  <a data-link-type="idl-name" href="#mloperand"><c- n>MLOperand</c-></a> <a class="idl-code" data-link-type="method" href="#dom-mlgraphbuilder-linear"><c- g>linear</c-></a>(<a data-link-type="idl-name" href="#mloperand"><c- n>MLOperand</c-></a> <a href="#dom-mlgraphbuilder-linear-input-options-input"><code><c- g>input</c-></code></a>, <c- b>optional</c-> <a data-link-type="idl-name" href="#dictdef-mllinearoptions"><c- n>MLLinearOptions</c-></a> <a href="#dom-mlgraphbuilder-linear-input-options-options"><code><c- g>options</c-></code></a> = {});
+  <a data-link-type="idl-name" href="#mlactivation"><c- n>MLActivation</c-></a> <a class="idl-code" data-link-type="method" href="#dom-mlgraphbuilder-linear-options"><c- g>linear</c-></a>(<c- b>optional</c-> <a data-link-type="idl-name" href="#dictdef-mllinearoptions"><c- n>MLLinearOptions</c-></a> <a href="#dom-mlgraphbuilder-linear-options-options"><code><c- g>options</c-></code></a> = {});
 };
 
 <c- b>enum</c-> <a href="#enumdef-mllstmweightlayout"><code><c- g>MLLstmWeightLayout</c-></code></a> {
@@ -5333,39 +8743,39 @@ <h2 class="no-num no-ref heading settled" id="idl-index"><span class="content">I
 };
 
 <c- b>dictionary</c-> <a href="#dictdef-mllstmoptions"><code><c- g>MLLstmOptions</c-></code></a> {
-  <a data-link-type="idl-name" href="#mloperand"><c- n>MLOperand</c-></a> <a data-type="MLOperand " href="#dom-mllstmoptions-bias"><code><c- g>bias</c-></code></a>;
-  <a data-link-type="idl-name" href="#mloperand"><c- n>MLOperand</c-></a> <a data-type="MLOperand " href="#dom-mllstmoptions-recurrentbias"><code><c- g>recurrentBias</c-></code></a>;
-  <a data-link-type="idl-name" href="#mloperand"><c- n>MLOperand</c-></a> <a data-type="MLOperand " href="#dom-mllstmoptions-peepholeweight"><code><c- g>peepholeWeight</c-></code></a>;
-  <a data-link-type="idl-name" href="#mloperand"><c- n>MLOperand</c-></a> <a data-type="MLOperand " href="#dom-mllstmoptions-initialhiddenstate"><code><c- g>initialHiddenState</c-></code></a>;
-  <a data-link-type="idl-name" href="#mloperand"><c- n>MLOperand</c-></a> <a data-type="MLOperand " href="#dom-mllstmoptions-initialcellstate"><code><c- g>initialCellState</c-></code></a>;
-  <a class="idl-code" data-link-type="interface" href="https://webidl.spec.whatwg.org/#idl-boolean"><c- b>boolean</c-></a> <a data-default="false" data-type="boolean " href="#dom-mllstmoptions-returnsequence"><code><c- g>returnSequence</c-></code></a> = <c- b>false</c->;
-  <a data-link-type="idl-name" href="#enumdef-mlrecurrentnetworkdirection"><c- n>MLRecurrentNetworkDirection</c-></a> <a data-default="&quot;forward&quot;" data-type="MLRecurrentNetworkDirection " href="#dom-mllstmoptions-direction"><code><c- g>direction</c-></code></a> = "forward";
-  <a data-link-type="idl-name" href="#enumdef-mllstmweightlayout"><c- n>MLLstmWeightLayout</c-></a> <a data-default="&quot;iofg&quot;" data-type="MLLstmWeightLayout " href="#dom-mllstmoptions-layout"><code><c- g>layout</c-></code></a> = "iofg";
-  <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#idl-sequence"><c- b>sequence</c-></a>&lt;<a data-link-type="idl-name" href="#mlactivation"><c- n>MLActivation</c-></a>> <a data-type="sequence<MLActivation> " href="#dom-mllstmoptions-activations"><code><c- g>activations</c-></code></a>;
+  <a data-link-type="idl-name" href="#mloperand"><c- n>MLOperand</c-></a> <a class="idl-code" data-link-type="dict-member" data-type="MLOperand " href="#dom-mllstmoptions-bias"><c- g>bias</c-></a>;
+  <a data-link-type="idl-name" href="#mloperand"><c- n>MLOperand</c-></a> <a class="idl-code" data-link-type="dict-member" data-type="MLOperand " href="#dom-mllstmoptions-recurrentbias"><c- g>recurrentBias</c-></a>;
+  <a data-link-type="idl-name" href="#mloperand"><c- n>MLOperand</c-></a> <a class="idl-code" data-link-type="dict-member" data-type="MLOperand " href="#dom-mllstmoptions-peepholeweight"><c- g>peepholeWeight</c-></a>;
+  <a data-link-type="idl-name" href="#mloperand"><c- n>MLOperand</c-></a> <a class="idl-code" data-link-type="dict-member" data-type="MLOperand " href="#dom-mllstmoptions-initialhiddenstate"><c- g>initialHiddenState</c-></a>;
+  <a data-link-type="idl-name" href="#mloperand"><c- n>MLOperand</c-></a> <a class="idl-code" data-link-type="dict-member" data-type="MLOperand " href="#dom-mllstmoptions-initialcellstate"><c- g>initialCellState</c-></a>;
+  <a class="idl-code" data-link-type="interface" href="https://webidl.spec.whatwg.org/#idl-boolean"><c- b>boolean</c-></a> <a class="idl-code" data-default="false" data-link-type="dict-member" data-type="boolean " href="#dom-mllstmoptions-returnsequence"><c- g>returnSequence</c-></a> = <c- b>false</c->;
+  <a data-link-type="idl-name" href="#enumdef-mlrecurrentnetworkdirection"><c- n>MLRecurrentNetworkDirection</c-></a> <a class="idl-code" data-default="&quot;forward&quot;" data-link-type="dict-member" data-type="MLRecurrentNetworkDirection " href="#dom-mllstmoptions-direction"><c- g>direction</c-></a> = "forward";
+  <a data-link-type="idl-name" href="#enumdef-mllstmweightlayout"><c- n>MLLstmWeightLayout</c-></a> <a class="idl-code" data-default="&quot;iofg&quot;" data-link-type="dict-member" data-type="MLLstmWeightLayout " href="#dom-mllstmoptions-layout"><c- g>layout</c-></a> = "iofg";
+  <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#idl-sequence"><c- b>sequence</c-></a>&lt;<a data-link-type="idl-name" href="#mlactivation"><c- n>MLActivation</c-></a>> <a class="idl-code" data-link-type="dict-member" data-type="sequence<MLActivation> " href="#dom-mllstmoptions-activations"><c- g>activations</c-></a>;
 };
 
 <c- b>partial</c-> <c- b>interface</c-> <a class="idl-code" data-link-type="interface" href="#mlgraphbuilder"><c- g>MLGraphBuilder</c-></a> {
-  <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#idl-sequence"><c- b>sequence</c-></a>&lt;<a data-link-type="idl-name" href="#mloperand"><c- n>MLOperand</c-></a>> <a href="#dom-mlgraphbuilder-lstm"><code><c- g>lstm</c-></code></a>(<a data-link-type="idl-name" href="#mloperand"><c- n>MLOperand</c-></a> <a href="#dom-mlgraphbuilder-lstm-input-weight-recurrentweight-steps-hiddensize-options-input"><code><c- g>input</c-></code></a>, <a data-link-type="idl-name" href="#mloperand"><c- n>MLOperand</c-></a> <a href="#dom-mlgraphbuilder-lstm-input-weight-recurrentweight-steps-hiddensize-options-weight"><code><c- g>weight</c-></code></a>, <a data-link-type="idl-name" href="#mloperand"><c- n>MLOperand</c-></a> <a href="#dom-mlgraphbuilder-lstm-input-weight-recurrentweight-steps-hiddensize-options-recurrentweight"><code><c- g>recurrentWeight</c-></code></a>,
+  <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#idl-sequence"><c- b>sequence</c-></a>&lt;<a data-link-type="idl-name" href="#mloperand"><c- n>MLOperand</c-></a>> <a class="idl-code" data-link-type="method" href="#dom-mlgraphbuilder-lstm"><c- g>lstm</c-></a>(<a data-link-type="idl-name" href="#mloperand"><c- n>MLOperand</c-></a> <a href="#dom-mlgraphbuilder-lstm-input-weight-recurrentweight-steps-hiddensize-options-input"><code><c- g>input</c-></code></a>, <a data-link-type="idl-name" href="#mloperand"><c- n>MLOperand</c-></a> <a href="#dom-mlgraphbuilder-lstm-input-weight-recurrentweight-steps-hiddensize-options-weight"><code><c- g>weight</c-></code></a>, <a data-link-type="idl-name" href="#mloperand"><c- n>MLOperand</c-></a> <a href="#dom-mlgraphbuilder-lstm-input-weight-recurrentweight-steps-hiddensize-options-recurrentweight"><code><c- g>recurrentWeight</c-></code></a>,
                            <a class="idl-code" data-link-type="interface" href="https://webidl.spec.whatwg.org/#idl-unsigned-long"><c- b>unsigned</c-> <c- b>long</c-></a> <a href="#dom-mlgraphbuilder-lstm-input-weight-recurrentweight-steps-hiddensize-options-steps"><code><c- g>steps</c-></code></a>, <a class="idl-code" data-link-type="interface" href="https://webidl.spec.whatwg.org/#idl-unsigned-long"><c- b>unsigned</c-> <c- b>long</c-></a> <a href="#dom-mlgraphbuilder-lstm-input-weight-recurrentweight-steps-hiddensize-options-hiddensize"><code><c- g>hiddenSize</c-></code></a>,
                            <c- b>optional</c-> <a data-link-type="idl-name" href="#dictdef-mllstmoptions"><c- n>MLLstmOptions</c-></a> <a href="#dom-mlgraphbuilder-lstm-input-weight-recurrentweight-steps-hiddensize-options-options"><code><c- g>options</c-></code></a> = {});
 };
 
 <c- b>dictionary</c-> <a href="#dictdef-mllstmcelloptions"><code><c- g>MLLstmCellOptions</c-></code></a> {
-  <a data-link-type="idl-name" href="#mloperand"><c- n>MLOperand</c-></a> <a data-type="MLOperand " href="#dom-mllstmcelloptions-bias"><code><c- g>bias</c-></code></a>;
-  <a data-link-type="idl-name" href="#mloperand"><c- n>MLOperand</c-></a> <a data-type="MLOperand " href="#dom-mllstmcelloptions-recurrentbias"><code><c- g>recurrentBias</c-></code></a>;
-  <a data-link-type="idl-name" href="#mloperand"><c- n>MLOperand</c-></a> <a data-type="MLOperand " href="#dom-mllstmcelloptions-peepholeweight"><code><c- g>peepholeWeight</c-></code></a>;
-  <a data-link-type="idl-name" href="#enumdef-mllstmweightlayout"><c- n>MLLstmWeightLayout</c-></a> <a data-default="&quot;iofg&quot;" data-type="MLLstmWeightLayout " href="#dom-mllstmcelloptions-layout"><code><c- g>layout</c-></code></a> = "iofg";
-  <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#idl-sequence"><c- b>sequence</c-></a>&lt;<a data-link-type="idl-name" href="#mlactivation"><c- n>MLActivation</c-></a>> <a data-type="sequence<MLActivation> " href="#dom-mllstmcelloptions-activations"><code><c- g>activations</c-></code></a>;
+  <a data-link-type="idl-name" href="#mloperand"><c- n>MLOperand</c-></a> <a class="idl-code" data-link-type="dict-member" data-type="MLOperand " href="#dom-mllstmcelloptions-bias"><c- g>bias</c-></a>;
+  <a data-link-type="idl-name" href="#mloperand"><c- n>MLOperand</c-></a> <a class="idl-code" data-link-type="dict-member" data-type="MLOperand " href="#dom-mllstmcelloptions-recurrentbias"><c- g>recurrentBias</c-></a>;
+  <a data-link-type="idl-name" href="#mloperand"><c- n>MLOperand</c-></a> <a class="idl-code" data-link-type="dict-member" data-type="MLOperand " href="#dom-mllstmcelloptions-peepholeweight"><c- g>peepholeWeight</c-></a>;
+  <a data-link-type="idl-name" href="#enumdef-mllstmweightlayout"><c- n>MLLstmWeightLayout</c-></a> <a class="idl-code" data-default="&quot;iofg&quot;" data-link-type="dict-member" data-type="MLLstmWeightLayout " href="#dom-mllstmcelloptions-layout"><c- g>layout</c-></a> = "iofg";
+  <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#idl-sequence"><c- b>sequence</c-></a>&lt;<a data-link-type="idl-name" href="#mlactivation"><c- n>MLActivation</c-></a>> <a class="idl-code" data-link-type="dict-member" data-type="sequence<MLActivation> " href="#dom-mllstmcelloptions-activations"><c- g>activations</c-></a>;
 };
 
 <c- b>partial</c-> <c- b>interface</c-> <a class="idl-code" data-link-type="interface" href="#mlgraphbuilder"><c- g>MLGraphBuilder</c-></a> {
-  <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#idl-sequence"><c- b>sequence</c-></a>&lt;<a data-link-type="idl-name" href="#mloperand"><c- n>MLOperand</c-></a>> <a href="#dom-mlgraphbuilder-lstmcell"><code><c- g>lstmCell</c-></code></a>(<a data-link-type="idl-name" href="#mloperand"><c- n>MLOperand</c-></a> <a href="#dom-mlgraphbuilder-lstmcell-input-weight-recurrentweight-hiddenstate-cellstate-hiddensize-options-input"><code><c- g>input</c-></code></a>, <a data-link-type="idl-name" href="#mloperand"><c- n>MLOperand</c-></a> <a href="#dom-mlgraphbuilder-lstmcell-input-weight-recurrentweight-hiddenstate-cellstate-hiddensize-options-weight"><code><c- g>weight</c-></code></a>, <a data-link-type="idl-name" href="#mloperand"><c- n>MLOperand</c-></a> <a href="#dom-mlgraphbuilder-lstmcell-input-weight-recurrentweight-hiddenstate-cellstate-hiddensize-options-recurrentweight"><code><c- g>recurrentWeight</c-></code></a>,
+  <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#idl-sequence"><c- b>sequence</c-></a>&lt;<a data-link-type="idl-name" href="#mloperand"><c- n>MLOperand</c-></a>> <a class="idl-code" data-link-type="method" href="#dom-mlgraphbuilder-lstmcell"><c- g>lstmCell</c-></a>(<a data-link-type="idl-name" href="#mloperand"><c- n>MLOperand</c-></a> <a href="#dom-mlgraphbuilder-lstmcell-input-weight-recurrentweight-hiddenstate-cellstate-hiddensize-options-input"><code><c- g>input</c-></code></a>, <a data-link-type="idl-name" href="#mloperand"><c- n>MLOperand</c-></a> <a href="#dom-mlgraphbuilder-lstmcell-input-weight-recurrentweight-hiddenstate-cellstate-hiddensize-options-weight"><code><c- g>weight</c-></code></a>, <a data-link-type="idl-name" href="#mloperand"><c- n>MLOperand</c-></a> <a href="#dom-mlgraphbuilder-lstmcell-input-weight-recurrentweight-hiddenstate-cellstate-hiddensize-options-recurrentweight"><code><c- g>recurrentWeight</c-></code></a>,
                                <a data-link-type="idl-name" href="#mloperand"><c- n>MLOperand</c-></a> <a href="#dom-mlgraphbuilder-lstmcell-input-weight-recurrentweight-hiddenstate-cellstate-hiddensize-options-hiddenstate"><code><c- g>hiddenState</c-></code></a>, <a data-link-type="idl-name" href="#mloperand"><c- n>MLOperand</c-></a> <a href="#dom-mlgraphbuilder-lstmcell-input-weight-recurrentweight-hiddenstate-cellstate-hiddensize-options-cellstate"><code><c- g>cellState</c-></code></a>, <a class="idl-code" data-link-type="interface" href="https://webidl.spec.whatwg.org/#idl-unsigned-long"><c- b>unsigned</c-> <c- b>long</c-></a> <a href="#dom-mlgraphbuilder-lstmcell-input-weight-recurrentweight-hiddenstate-cellstate-hiddensize-options-hiddensize"><code><c- g>hiddenSize</c-></code></a>,
                                <c- b>optional</c-> <a data-link-type="idl-name" href="#dictdef-mllstmcelloptions"><c- n>MLLstmCellOptions</c-></a> <a href="#dom-mlgraphbuilder-lstmcell-input-weight-recurrentweight-hiddenstate-cellstate-hiddensize-options-options"><code><c- g>options</c-></code></a> = {});
 };
 
 <c- b>partial</c-> <c- b>interface</c-> <a class="idl-code" data-link-type="interface" href="#mlgraphbuilder"><c- g>MLGraphBuilder</c-></a> {
-  <a data-link-type="idl-name" href="#mloperand"><c- n>MLOperand</c-></a> <a href="#dom-mlgraphbuilder-matmul"><code><c- g>matmul</c-></code></a>(<a data-link-type="idl-name" href="#mloperand"><c- n>MLOperand</c-></a> <a href="#dom-mlgraphbuilder-matmul-a-b-a"><code><c- g>a</c-></code></a>, <a data-link-type="idl-name" href="#mloperand"><c- n>MLOperand</c-></a> <a href="#dom-mlgraphbuilder-matmul-a-b-b"><code><c- g>b</c-></code></a>);
+  <a data-link-type="idl-name" href="#mloperand"><c- n>MLOperand</c-></a> <a class="idl-code" data-link-type="method" href="#dom-mlgraphbuilder-matmul"><c- g>matmul</c-></a>(<a data-link-type="idl-name" href="#mloperand"><c- n>MLOperand</c-></a> <a href="#dom-mlgraphbuilder-matmul-a-b-a"><code><c- g>a</c-></code></a>, <a data-link-type="idl-name" href="#mloperand"><c- n>MLOperand</c-></a> <a href="#dom-mlgraphbuilder-matmul-a-b-b"><code><c- g>b</c-></code></a>);
 };
 
 <c- b>enum</c-> <a href="#enumdef-mlpaddingmode"><code><c- g>MLPaddingMode</c-></code></a> {
@@ -5376,12 +8786,12 @@ <h2 class="no-num no-ref heading settled" id="idl-index"><span class="content">I
 };
 
 <c- b>dictionary</c-> <a href="#dictdef-mlpadoptions"><code><c- g>MLPadOptions</c-></code></a> {
-  <a data-link-type="idl-name" href="#enumdef-mlpaddingmode"><c- n>MLPaddingMode</c-></a> <a data-default="&quot;constant&quot;" data-type="MLPaddingMode " href="#dom-mlpadoptions-mode"><code><c- g>mode</c-></code></a> = "constant";
-  <a class="idl-code" data-link-type="interface" href="https://webidl.spec.whatwg.org/#idl-float"><c- b>float</c-></a> <a data-default="0" data-type="float " href="#dom-mlpadoptions-value"><code><c- g>value</c-></code></a> = 0;
+  <a data-link-type="idl-name" href="#enumdef-mlpaddingmode"><c- n>MLPaddingMode</c-></a> <a class="idl-code" data-default="&quot;constant&quot;" data-link-type="dict-member" data-type="MLPaddingMode " href="#dom-mlpadoptions-mode"><c- g>mode</c-></a> = "constant";
+  <a class="idl-code" data-link-type="interface" href="https://webidl.spec.whatwg.org/#idl-float"><c- b>float</c-></a> <a class="idl-code" data-default="0" data-link-type="dict-member" data-type="float " href="#dom-mlpadoptions-value"><c- g>value</c-></a> = 0;
 };
 
 <c- b>partial</c-> <c- b>interface</c-> <a class="idl-code" data-link-type="interface" href="#mlgraphbuilder"><c- g>MLGraphBuilder</c-></a> {
-  <a data-link-type="idl-name" href="#mloperand"><c- n>MLOperand</c-></a> <a href="#dom-mlgraphbuilder-pad"><code><c- g>pad</c-></code></a>(<a data-link-type="idl-name" href="#mloperand"><c- n>MLOperand</c-></a> <a href="#dom-mlgraphbuilder-pad-input-beginningpadding-endingpadding-options-input"><code><c- g>input</c-></code></a>,
+  <a data-link-type="idl-name" href="#mloperand"><c- n>MLOperand</c-></a> <a class="idl-code" data-link-type="method" href="#dom-mlgraphbuilder-pad"><c- g>pad</c-></a>(<a data-link-type="idl-name" href="#mloperand"><c- n>MLOperand</c-></a> <a href="#dom-mlgraphbuilder-pad-input-beginningpadding-endingpadding-options-input"><code><c- g>input</c-></code></a>,
                 <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#idl-sequence"><c- b>sequence</c-></a>&lt;<a class="idl-code" data-link-type="interface" href="https://webidl.spec.whatwg.org/#idl-unsigned-long"><c- b>unsigned</c-> <c- b>long</c-></a>> <a href="#dom-mlgraphbuilder-pad-input-beginningpadding-endingpadding-options-beginningpadding"><code><c- g>beginningPadding</c-></code></a>,
                 <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#idl-sequence"><c- b>sequence</c-></a>&lt;<a class="idl-code" data-link-type="interface" href="https://webidl.spec.whatwg.org/#idl-unsigned-long"><c- b>unsigned</c-> <c- b>long</c-></a>> <a href="#dom-mlgraphbuilder-pad-input-beginningpadding-endingpadding-options-endingpadding"><code><c- g>endingPadding</c-></code></a>,
                 <c- b>optional</c-> <a data-link-type="idl-name" href="#dictdef-mlpadoptions"><c- n>MLPadOptions</c-></a> <a href="#dom-mlgraphbuilder-pad-input-beginningpadding-endingpadding-options-options"><code><c- g>options</c-></code></a> = {});
@@ -5393,24 +8803,24 @@ <h2 class="no-num no-ref heading settled" id="idl-index"><span class="content">I
 };
 
 <c- b>dictionary</c-> <a href="#dictdef-mlpool2doptions"><code><c- g>MLPool2dOptions</c-></code></a> {
-  <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#idl-sequence"><c- b>sequence</c-></a>&lt;<a class="idl-code" data-link-type="interface" href="https://webidl.spec.whatwg.org/#idl-unsigned-long"><c- b>unsigned</c-> <c- b>long</c-></a>> <a data-type="sequence<unsigned long> " href="#dom-mlpool2doptions-windowdimensions"><code><c- g>windowDimensions</c-></code></a>;
-  <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#idl-sequence"><c- b>sequence</c-></a>&lt;<a class="idl-code" data-link-type="interface" href="https://webidl.spec.whatwg.org/#idl-unsigned-long"><c- b>unsigned</c-> <c- b>long</c-></a>> <a data-type="sequence<unsigned long> " href="#dom-mlpool2doptions-padding"><code><c- g>padding</c-></code></a>;
-  <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#idl-sequence"><c- b>sequence</c-></a>&lt;<a class="idl-code" data-link-type="interface" href="https://webidl.spec.whatwg.org/#idl-unsigned-long"><c- b>unsigned</c-> <c- b>long</c-></a>> <a data-type="sequence<unsigned long> " href="#dom-mlpool2doptions-strides"><code><c- g>strides</c-></code></a>;
-  <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#idl-sequence"><c- b>sequence</c-></a>&lt;<a class="idl-code" data-link-type="interface" href="https://webidl.spec.whatwg.org/#idl-unsigned-long"><c- b>unsigned</c-> <c- b>long</c-></a>> <a data-type="sequence<unsigned long> " href="#dom-mlpool2doptions-dilations"><code><c- g>dilations</c-></code></a>;
-  <a data-link-type="idl-name" href="#enumdef-mlautopad"><c- n>MLAutoPad</c-></a> <a data-default="&quot;explicit&quot;" data-type="MLAutoPad " href="#dom-mlpool2doptions-autopad"><code><c- g>autoPad</c-></code></a> = "explicit";
-  <a data-link-type="idl-name" href="#enumdef-mlinputoperandlayout"><c- n>MLInputOperandLayout</c-></a> <a data-default="&quot;nchw&quot;" data-type="MLInputOperandLayout " href="#dom-mlpool2doptions-layout"><code><c- g>layout</c-></code></a> = "nchw";
-  <a data-link-type="idl-name" href="#enumdef-mlroundingtype"><c- n>MLRoundingType</c-></a> <a data-default="&quot;floor&quot;" data-type="MLRoundingType " href="#dom-mlpool2doptions-roundingtype"><code><c- g>roundingType</c-></code></a> = "floor";
-  <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#idl-sequence"><c- b>sequence</c-></a>&lt;<a class="idl-code" data-link-type="interface" href="https://webidl.spec.whatwg.org/#idl-unsigned-long"><c- b>unsigned</c-> <c- b>long</c-></a>> <a data-type="sequence<unsigned long> " href="#dom-mlpool2doptions-outputsizes"><code><c- g>outputSizes</c-></code></a>;
+  <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#idl-sequence"><c- b>sequence</c-></a>&lt;<a class="idl-code" data-link-type="interface" href="https://webidl.spec.whatwg.org/#idl-unsigned-long"><c- b>unsigned</c-> <c- b>long</c-></a>> <a class="idl-code" data-link-type="dict-member" data-type="sequence<unsigned long> " href="#dom-mlpool2doptions-windowdimensions"><c- g>windowDimensions</c-></a>;
+  <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#idl-sequence"><c- b>sequence</c-></a>&lt;<a class="idl-code" data-link-type="interface" href="https://webidl.spec.whatwg.org/#idl-unsigned-long"><c- b>unsigned</c-> <c- b>long</c-></a>> <a class="idl-code" data-link-type="dict-member" data-type="sequence<unsigned long> " href="#dom-mlpool2doptions-padding"><c- g>padding</c-></a>;
+  <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#idl-sequence"><c- b>sequence</c-></a>&lt;<a class="idl-code" data-link-type="interface" href="https://webidl.spec.whatwg.org/#idl-unsigned-long"><c- b>unsigned</c-> <c- b>long</c-></a>> <a class="idl-code" data-link-type="dict-member" data-type="sequence<unsigned long> " href="#dom-mlpool2doptions-strides"><c- g>strides</c-></a>;
+  <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#idl-sequence"><c- b>sequence</c-></a>&lt;<a class="idl-code" data-link-type="interface" href="https://webidl.spec.whatwg.org/#idl-unsigned-long"><c- b>unsigned</c-> <c- b>long</c-></a>> <a class="idl-code" data-link-type="dict-member" data-type="sequence<unsigned long> " href="#dom-mlpool2doptions-dilations"><c- g>dilations</c-></a>;
+  <a data-link-type="idl-name" href="#enumdef-mlautopad"><c- n>MLAutoPad</c-></a> <a class="idl-code" data-default="&quot;explicit&quot;" data-link-type="dict-member" data-type="MLAutoPad " href="#dom-mlpool2doptions-autopad"><c- g>autoPad</c-></a> = "explicit";
+  <a data-link-type="idl-name" href="#enumdef-mlinputoperandlayout"><c- n>MLInputOperandLayout</c-></a> <a class="idl-code" data-default="&quot;nchw&quot;" data-link-type="dict-member" data-type="MLInputOperandLayout " href="#dom-mlpool2doptions-layout"><c- g>layout</c-></a> = "nchw";
+  <a data-link-type="idl-name" href="#enumdef-mlroundingtype"><c- n>MLRoundingType</c-></a> <a class="idl-code" data-default="&quot;floor&quot;" data-link-type="dict-member" data-type="MLRoundingType " href="#dom-mlpool2doptions-roundingtype"><c- g>roundingType</c-></a> = "floor";
+  <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#idl-sequence"><c- b>sequence</c-></a>&lt;<a class="idl-code" data-link-type="interface" href="https://webidl.spec.whatwg.org/#idl-unsigned-long"><c- b>unsigned</c-> <c- b>long</c-></a>> <a class="idl-code" data-link-type="dict-member" data-type="sequence<unsigned long> " href="#dom-mlpool2doptions-outputsizes"><c- g>outputSizes</c-></a>;
 };
 
 <c- b>partial</c-> <c- b>interface</c-> <a class="idl-code" data-link-type="interface" href="#mlgraphbuilder"><c- g>MLGraphBuilder</c-></a> {
-  <a data-link-type="idl-name" href="#mloperand"><c- n>MLOperand</c-></a> <a href="#dom-mlgraphbuilder-averagepool2d"><code><c- g>averagePool2d</c-></code></a>(<a data-link-type="idl-name" href="#mloperand"><c- n>MLOperand</c-></a> <a href="#dom-mlgraphbuilder-averagepool2d-input-options-input"><code><c- g>input</c-></code></a>, <c- b>optional</c-> <a data-link-type="idl-name" href="#dictdef-mlpool2doptions"><c- n>MLPool2dOptions</c-></a> <a href="#dom-mlgraphbuilder-averagepool2d-input-options-options"><code><c- g>options</c-></code></a> = {});
-  <a data-link-type="idl-name" href="#mloperand"><c- n>MLOperand</c-></a> <a href="#dom-mlgraphbuilder-l2pool2d"><code><c- g>l2Pool2d</c-></code></a>(<a data-link-type="idl-name" href="#mloperand"><c- n>MLOperand</c-></a> <a href="#dom-mlgraphbuilder-l2pool2d-input-options-input"><code><c- g>input</c-></code></a>, <c- b>optional</c-> <a data-link-type="idl-name" href="#dictdef-mlpool2doptions"><c- n>MLPool2dOptions</c-></a> <a href="#dom-mlgraphbuilder-l2pool2d-input-options-options"><code><c- g>options</c-></code></a> = {});
-  <a data-link-type="idl-name" href="#mloperand"><c- n>MLOperand</c-></a> <a href="#dom-mlgraphbuilder-maxpool2d"><code><c- g>maxPool2d</c-></code></a>(<a data-link-type="idl-name" href="#mloperand"><c- n>MLOperand</c-></a> <a href="#dom-mlgraphbuilder-maxpool2d-input-options-input"><code><c- g>input</c-></code></a>, <c- b>optional</c-> <a data-link-type="idl-name" href="#dictdef-mlpool2doptions"><c- n>MLPool2dOptions</c-></a> <a href="#dom-mlgraphbuilder-maxpool2d-input-options-options"><code><c- g>options</c-></code></a> = {});
+  <a data-link-type="idl-name" href="#mloperand"><c- n>MLOperand</c-></a> <a class="idl-code" data-link-type="method" href="#dom-mlgraphbuilder-averagepool2d"><c- g>averagePool2d</c-></a>(<a data-link-type="idl-name" href="#mloperand"><c- n>MLOperand</c-></a> <a href="#dom-mlgraphbuilder-averagepool2d-input-options-input"><code><c- g>input</c-></code></a>, <c- b>optional</c-> <a data-link-type="idl-name" href="#dictdef-mlpool2doptions"><c- n>MLPool2dOptions</c-></a> <a href="#dom-mlgraphbuilder-averagepool2d-input-options-options"><code><c- g>options</c-></code></a> = {});
+  <a data-link-type="idl-name" href="#mloperand"><c- n>MLOperand</c-></a> <a class="idl-code" data-link-type="method" href="#dom-mlgraphbuilder-l2pool2d"><c- g>l2Pool2d</c-></a>(<a data-link-type="idl-name" href="#mloperand"><c- n>MLOperand</c-></a> <a href="#dom-mlgraphbuilder-l2pool2d-input-options-input"><code><c- g>input</c-></code></a>, <c- b>optional</c-> <a data-link-type="idl-name" href="#dictdef-mlpool2doptions"><c- n>MLPool2dOptions</c-></a> <a href="#dom-mlgraphbuilder-l2pool2d-input-options-options"><code><c- g>options</c-></code></a> = {});
+  <a data-link-type="idl-name" href="#mloperand"><c- n>MLOperand</c-></a> <a class="idl-code" data-link-type="method" href="#dom-mlgraphbuilder-maxpool2d"><c- g>maxPool2d</c-></a>(<a data-link-type="idl-name" href="#mloperand"><c- n>MLOperand</c-></a> <a href="#dom-mlgraphbuilder-maxpool2d-input-options-input"><code><c- g>input</c-></code></a>, <c- b>optional</c-> <a data-link-type="idl-name" href="#dictdef-mlpool2doptions"><c- n>MLPool2dOptions</c-></a> <a href="#dom-mlgraphbuilder-maxpool2d-input-options-options"><code><c- g>options</c-></code></a> = {});
 };
 
 <c- b>partial</c-> <c- b>interface</c-> <a class="idl-code" data-link-type="interface" href="#mlgraphbuilder"><c- g>MLGraphBuilder</c-></a> {
-  <a data-link-type="idl-name" href="#mloperand"><c- n>MLOperand</c-></a> <a href="#dom-mlgraphbuilder-prelu"><code><c- g>prelu</c-></code></a>(<a data-link-type="idl-name" href="#mloperand"><c- n>MLOperand</c-></a> <a href="#dom-mlgraphbuilder-prelu-x-slope-x"><code><c- g>x</c-></code></a>, <a data-link-type="idl-name" href="#mloperand"><c- n>MLOperand</c-></a> <a href="#dom-mlgraphbuilder-prelu-x-slope-slope"><code><c- g>slope</c-></code></a>);
+  <a data-link-type="idl-name" href="#mloperand"><c- n>MLOperand</c-></a> <a class="idl-code" data-link-type="method" href="#dom-mlgraphbuilder-prelu"><c- g>prelu</c-></a>(<a data-link-type="idl-name" href="#mloperand"><c- n>MLOperand</c-></a> <a href="#dom-mlgraphbuilder-prelu-input-slope-input"><code><c- g>input</c-></code></a>, <a data-link-type="idl-name" href="#mloperand"><c- n>MLOperand</c-></a> <a href="#dom-mlgraphbuilder-prelu-input-slope-slope"><code><c- g>slope</c-></code></a>);
 };
 
 <c- b>dictionary</c-> <a href="#dictdef-mlreduceoptions"><code><c- g>MLReduceOptions</c-></code></a> {
@@ -5419,21 +8829,21 @@ <h2 class="no-num no-ref heading settled" id="idl-index"><span class="content">I
 };
 
 <c- b>partial</c-> <c- b>interface</c-> <a class="idl-code" data-link-type="interface" href="#mlgraphbuilder"><c- g>MLGraphBuilder</c-></a> {
-  <a data-link-type="idl-name" href="#mloperand"><c- n>MLOperand</c-></a> <a href="#dom-mlgraphbuilder-reducel1"><code><c- g>reduceL1</c-></code></a>(<a data-link-type="idl-name" href="#mloperand"><c- n>MLOperand</c-></a> <a href="#dom-mlgraphbuilder-reducel1-input-options-input"><code><c- g>input</c-></code></a>, <c- b>optional</c-> <a data-link-type="idl-name" href="#dictdef-mlreduceoptions"><c- n>MLReduceOptions</c-></a> <a href="#dom-mlgraphbuilder-reducel1-input-options-options"><code><c- g>options</c-></code></a> = {});
-  <a data-link-type="idl-name" href="#mloperand"><c- n>MLOperand</c-></a> <a href="#dom-mlgraphbuilder-reducel2"><code><c- g>reduceL2</c-></code></a>(<a data-link-type="idl-name" href="#mloperand"><c- n>MLOperand</c-></a> <a href="#dom-mlgraphbuilder-reducel2-input-options-input"><code><c- g>input</c-></code></a>, <c- b>optional</c-> <a data-link-type="idl-name" href="#dictdef-mlreduceoptions"><c- n>MLReduceOptions</c-></a> <a href="#dom-mlgraphbuilder-reducel2-input-options-options"><code><c- g>options</c-></code></a> = {});
-  <a data-link-type="idl-name" href="#mloperand"><c- n>MLOperand</c-></a> <a href="#dom-mlgraphbuilder-reducelogsum"><code><c- g>reduceLogSum</c-></code></a>(<a data-link-type="idl-name" href="#mloperand"><c- n>MLOperand</c-></a> <a href="#dom-mlgraphbuilder-reducelogsum-input-options-input"><code><c- g>input</c-></code></a>, <c- b>optional</c-> <a data-link-type="idl-name" href="#dictdef-mlreduceoptions"><c- n>MLReduceOptions</c-></a> <a href="#dom-mlgraphbuilder-reducelogsum-input-options-options"><code><c- g>options</c-></code></a> = {});
-  <a data-link-type="idl-name" href="#mloperand"><c- n>MLOperand</c-></a> <a href="#dom-mlgraphbuilder-reducelogsumexp"><code><c- g>reduceLogSumExp</c-></code></a>(<a data-link-type="idl-name" href="#mloperand"><c- n>MLOperand</c-></a> <a href="#dom-mlgraphbuilder-reducelogsumexp-input-options-input"><code><c- g>input</c-></code></a>, <c- b>optional</c-> <a data-link-type="idl-name" href="#dictdef-mlreduceoptions"><c- n>MLReduceOptions</c-></a> <a href="#dom-mlgraphbuilder-reducelogsumexp-input-options-options"><code><c- g>options</c-></code></a> = {});
-  <a data-link-type="idl-name" href="#mloperand"><c- n>MLOperand</c-></a> <a href="#dom-mlgraphbuilder-reducemax"><code><c- g>reduceMax</c-></code></a>(<a data-link-type="idl-name" href="#mloperand"><c- n>MLOperand</c-></a> <a href="#dom-mlgraphbuilder-reducemax-input-options-input"><code><c- g>input</c-></code></a>, <c- b>optional</c-> <a data-link-type="idl-name" href="#dictdef-mlreduceoptions"><c- n>MLReduceOptions</c-></a> <a href="#dom-mlgraphbuilder-reducemax-input-options-options"><code><c- g>options</c-></code></a> = {});
-  <a data-link-type="idl-name" href="#mloperand"><c- n>MLOperand</c-></a> <a href="#dom-mlgraphbuilder-reducemean"><code><c- g>reduceMean</c-></code></a>(<a data-link-type="idl-name" href="#mloperand"><c- n>MLOperand</c-></a> <a href="#dom-mlgraphbuilder-reducemean-input-options-input"><code><c- g>input</c-></code></a>, <c- b>optional</c-> <a data-link-type="idl-name" href="#dictdef-mlreduceoptions"><c- n>MLReduceOptions</c-></a> <a href="#dom-mlgraphbuilder-reducemean-input-options-options"><code><c- g>options</c-></code></a> = {});
-  <a data-link-type="idl-name" href="#mloperand"><c- n>MLOperand</c-></a> <a href="#dom-mlgraphbuilder-reducemin"><code><c- g>reduceMin</c-></code></a>(<a data-link-type="idl-name" href="#mloperand"><c- n>MLOperand</c-></a> <a href="#dom-mlgraphbuilder-reducemin-input-options-input"><code><c- g>input</c-></code></a>, <c- b>optional</c-> <a data-link-type="idl-name" href="#dictdef-mlreduceoptions"><c- n>MLReduceOptions</c-></a> <a href="#dom-mlgraphbuilder-reducemin-input-options-options"><code><c- g>options</c-></code></a> = {});
-  <a data-link-type="idl-name" href="#mloperand"><c- n>MLOperand</c-></a> <a href="#dom-mlgraphbuilder-reduceproduct"><code><c- g>reduceProduct</c-></code></a>(<a data-link-type="idl-name" href="#mloperand"><c- n>MLOperand</c-></a> <a href="#dom-mlgraphbuilder-reduceproduct-input-options-input"><code><c- g>input</c-></code></a>, <c- b>optional</c-> <a data-link-type="idl-name" href="#dictdef-mlreduceoptions"><c- n>MLReduceOptions</c-></a> <a href="#dom-mlgraphbuilder-reduceproduct-input-options-options"><code><c- g>options</c-></code></a> = {});
-  <a data-link-type="idl-name" href="#mloperand"><c- n>MLOperand</c-></a> <a href="#dom-mlgraphbuilder-reducesum"><code><c- g>reduceSum</c-></code></a>(<a data-link-type="idl-name" href="#mloperand"><c- n>MLOperand</c-></a> <a href="#dom-mlgraphbuilder-reducesum-input-options-input"><code><c- g>input</c-></code></a>, <c- b>optional</c-> <a data-link-type="idl-name" href="#dictdef-mlreduceoptions"><c- n>MLReduceOptions</c-></a> <a href="#dom-mlgraphbuilder-reducesum-input-options-options"><code><c- g>options</c-></code></a> = {});
-  <a data-link-type="idl-name" href="#mloperand"><c- n>MLOperand</c-></a> <a href="#dom-mlgraphbuilder-reducesumsquare"><code><c- g>reduceSumSquare</c-></code></a>(<a data-link-type="idl-name" href="#mloperand"><c- n>MLOperand</c-></a> <a href="#dom-mlgraphbuilder-reducesumsquare-input-options-input"><code><c- g>input</c-></code></a>, <c- b>optional</c-> <a data-link-type="idl-name" href="#dictdef-mlreduceoptions"><c- n>MLReduceOptions</c-></a> <a href="#dom-mlgraphbuilder-reducesumsquare-input-options-options"><code><c- g>options</c-></code></a> = {});
+  <a data-link-type="idl-name" href="#mloperand"><c- n>MLOperand</c-></a> <a class="idl-code" data-link-type="method" href="#dom-mlgraphbuilder-reducel1"><c- g>reduceL1</c-></a>(<a data-link-type="idl-name" href="#mloperand"><c- n>MLOperand</c-></a> <a href="#dom-mlgraphbuilder-reducel1-input-options-input"><code><c- g>input</c-></code></a>, <c- b>optional</c-> <a data-link-type="idl-name" href="#dictdef-mlreduceoptions"><c- n>MLReduceOptions</c-></a> <a href="#dom-mlgraphbuilder-reducel1-input-options-options"><code><c- g>options</c-></code></a> = {});
+  <a data-link-type="idl-name" href="#mloperand"><c- n>MLOperand</c-></a> <a class="idl-code" data-link-type="method" href="#dom-mlgraphbuilder-reducel2"><c- g>reduceL2</c-></a>(<a data-link-type="idl-name" href="#mloperand"><c- n>MLOperand</c-></a> <a href="#dom-mlgraphbuilder-reducel2-input-options-input"><code><c- g>input</c-></code></a>, <c- b>optional</c-> <a data-link-type="idl-name" href="#dictdef-mlreduceoptions"><c- n>MLReduceOptions</c-></a> <a href="#dom-mlgraphbuilder-reducel2-input-options-options"><code><c- g>options</c-></code></a> = {});
+  <a data-link-type="idl-name" href="#mloperand"><c- n>MLOperand</c-></a> <a class="idl-code" data-link-type="method" href="#dom-mlgraphbuilder-reducelogsum"><c- g>reduceLogSum</c-></a>(<a data-link-type="idl-name" href="#mloperand"><c- n>MLOperand</c-></a> <a href="#dom-mlgraphbuilder-reducelogsum-input-options-input"><code><c- g>input</c-></code></a>, <c- b>optional</c-> <a data-link-type="idl-name" href="#dictdef-mlreduceoptions"><c- n>MLReduceOptions</c-></a> <a href="#dom-mlgraphbuilder-reducelogsum-input-options-options"><code><c- g>options</c-></code></a> = {});
+  <a data-link-type="idl-name" href="#mloperand"><c- n>MLOperand</c-></a> <a class="idl-code" data-link-type="method" href="#dom-mlgraphbuilder-reducelogsumexp"><c- g>reduceLogSumExp</c-></a>(<a data-link-type="idl-name" href="#mloperand"><c- n>MLOperand</c-></a> <a href="#dom-mlgraphbuilder-reducelogsumexp-input-options-input"><code><c- g>input</c-></code></a>, <c- b>optional</c-> <a data-link-type="idl-name" href="#dictdef-mlreduceoptions"><c- n>MLReduceOptions</c-></a> <a href="#dom-mlgraphbuilder-reducelogsumexp-input-options-options"><code><c- g>options</c-></code></a> = {});
+  <a data-link-type="idl-name" href="#mloperand"><c- n>MLOperand</c-></a> <a class="idl-code" data-link-type="method" href="#dom-mlgraphbuilder-reducemax"><c- g>reduceMax</c-></a>(<a data-link-type="idl-name" href="#mloperand"><c- n>MLOperand</c-></a> <a href="#dom-mlgraphbuilder-reducemax-input-options-input"><code><c- g>input</c-></code></a>, <c- b>optional</c-> <a data-link-type="idl-name" href="#dictdef-mlreduceoptions"><c- n>MLReduceOptions</c-></a> <a href="#dom-mlgraphbuilder-reducemax-input-options-options"><code><c- g>options</c-></code></a> = {});
+  <a data-link-type="idl-name" href="#mloperand"><c- n>MLOperand</c-></a> <a class="idl-code" data-link-type="method" href="#dom-mlgraphbuilder-reducemean"><c- g>reduceMean</c-></a>(<a data-link-type="idl-name" href="#mloperand"><c- n>MLOperand</c-></a> <a href="#dom-mlgraphbuilder-reducemean-input-options-input"><code><c- g>input</c-></code></a>, <c- b>optional</c-> <a data-link-type="idl-name" href="#dictdef-mlreduceoptions"><c- n>MLReduceOptions</c-></a> <a href="#dom-mlgraphbuilder-reducemean-input-options-options"><code><c- g>options</c-></code></a> = {});
+  <a data-link-type="idl-name" href="#mloperand"><c- n>MLOperand</c-></a> <a class="idl-code" data-link-type="method" href="#dom-mlgraphbuilder-reducemin"><c- g>reduceMin</c-></a>(<a data-link-type="idl-name" href="#mloperand"><c- n>MLOperand</c-></a> <a href="#dom-mlgraphbuilder-reducemin-input-options-input"><code><c- g>input</c-></code></a>, <c- b>optional</c-> <a data-link-type="idl-name" href="#dictdef-mlreduceoptions"><c- n>MLReduceOptions</c-></a> <a href="#dom-mlgraphbuilder-reducemin-input-options-options"><code><c- g>options</c-></code></a> = {});
+  <a data-link-type="idl-name" href="#mloperand"><c- n>MLOperand</c-></a> <a class="idl-code" data-link-type="method" href="#dom-mlgraphbuilder-reduceproduct"><c- g>reduceProduct</c-></a>(<a data-link-type="idl-name" href="#mloperand"><c- n>MLOperand</c-></a> <a href="#dom-mlgraphbuilder-reduceproduct-input-options-input"><code><c- g>input</c-></code></a>, <c- b>optional</c-> <a data-link-type="idl-name" href="#dictdef-mlreduceoptions"><c- n>MLReduceOptions</c-></a> <a href="#dom-mlgraphbuilder-reduceproduct-input-options-options"><code><c- g>options</c-></code></a> = {});
+  <a data-link-type="idl-name" href="#mloperand"><c- n>MLOperand</c-></a> <a class="idl-code" data-link-type="method" href="#dom-mlgraphbuilder-reducesum"><c- g>reduceSum</c-></a>(<a data-link-type="idl-name" href="#mloperand"><c- n>MLOperand</c-></a> <a href="#dom-mlgraphbuilder-reducesum-input-options-input"><code><c- g>input</c-></code></a>, <c- b>optional</c-> <a data-link-type="idl-name" href="#dictdef-mlreduceoptions"><c- n>MLReduceOptions</c-></a> <a href="#dom-mlgraphbuilder-reducesum-input-options-options"><code><c- g>options</c-></code></a> = {});
+  <a data-link-type="idl-name" href="#mloperand"><c- n>MLOperand</c-></a> <a class="idl-code" data-link-type="method" href="#dom-mlgraphbuilder-reducesumsquare"><c- g>reduceSumSquare</c-></a>(<a data-link-type="idl-name" href="#mloperand"><c- n>MLOperand</c-></a> <a href="#dom-mlgraphbuilder-reducesumsquare-input-options-input"><code><c- g>input</c-></code></a>, <c- b>optional</c-> <a data-link-type="idl-name" href="#dictdef-mlreduceoptions"><c- n>MLReduceOptions</c-></a> <a href="#dom-mlgraphbuilder-reducesumsquare-input-options-options"><code><c- g>options</c-></code></a> = {});
 };
 
 <c- b>partial</c-> <c- b>interface</c-> <a class="idl-code" data-link-type="interface" href="#mlgraphbuilder"><c- g>MLGraphBuilder</c-></a> {
-  <a data-link-type="idl-name" href="#mloperand"><c- n>MLOperand</c-></a> <a href="#dom-mlgraphbuilder-relu"><code><c- g>relu</c-></code></a>(<a data-link-type="idl-name" href="#mloperand"><c- n>MLOperand</c-></a> <a href="#dom-mlgraphbuilder-relu-x-x"><code><c- g>x</c-></code></a>);
-  <a data-link-type="idl-name" href="#mlactivation"><c- n>MLActivation</c-></a> <a href="#dom-mlgraphbuilder-relu①"><code><c- g>relu</c-></code></a>();
+  <a data-link-type="idl-name" href="#mloperand"><c- n>MLOperand</c-></a> <a class="idl-code" data-link-type="method" href="#dom-mlgraphbuilder-relu"><c- g>relu</c-></a>(<a data-link-type="idl-name" href="#mloperand"><c- n>MLOperand</c-></a> <a href="#dom-mlgraphbuilder-relu-input-input"><code><c- g>input</c-></code></a>);
+  <a data-link-type="idl-name" href="#mlactivation"><c- n>MLActivation</c-></a> <a class="idl-code" data-link-type="method" href="#relu-noargs"><c- g>relu</c-></a>();
 };
 
 <c- b>enum</c-> <a href="#enumdef-mlinterpolationmode"><code><c- g>MLInterpolationMode</c-></code></a> {
@@ -5442,77 +8852,77 @@ <h2 class="no-num no-ref heading settled" id="idl-index"><span class="content">I
 };
 
 <c- b>dictionary</c-> <a href="#dictdef-mlresample2doptions"><code><c- g>MLResample2dOptions</c-></code></a> {
-  <a data-link-type="idl-name" href="#enumdef-mlinterpolationmode"><c- n>MLInterpolationMode</c-></a> <a data-default="&quot;nearest-neighbor&quot;" data-type="MLInterpolationMode " href="#dom-mlresample2doptions-mode"><code><c- g>mode</c-></code></a> = "nearest-neighbor";
-  <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#idl-sequence"><c- b>sequence</c-></a>&lt;<a class="idl-code" data-link-type="interface" href="https://webidl.spec.whatwg.org/#idl-float"><c- b>float</c-></a>> <a data-type="sequence<float> " href="#dom-mlresample2doptions-scales"><code><c- g>scales</c-></code></a>;
-  <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#idl-sequence"><c- b>sequence</c-></a>&lt;<a class="idl-code" data-link-type="interface" href="https://webidl.spec.whatwg.org/#idl-unsigned-long"><c- b>unsigned</c-> <c- b>long</c-></a>> <a data-type="sequence<unsigned long> " href="#dom-mlresample2doptions-sizes"><code><c- g>sizes</c-></code></a>;
-  <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#idl-sequence"><c- b>sequence</c-></a>&lt;<a class="idl-code" data-link-type="interface" href="https://webidl.spec.whatwg.org/#idl-unsigned-long"><c- b>unsigned</c-> <c- b>long</c-></a>> <a data-type="sequence<unsigned long> " href="#dom-mlresample2doptions-axes"><code><c- g>axes</c-></code></a>;
+  <a data-link-type="idl-name" href="#enumdef-mlinterpolationmode"><c- n>MLInterpolationMode</c-></a> <a class="idl-code" data-default="&quot;nearest-neighbor&quot;" data-link-type="dict-member" data-type="MLInterpolationMode " href="#dom-mlresample2doptions-mode"><c- g>mode</c-></a> = "nearest-neighbor";
+  <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#idl-sequence"><c- b>sequence</c-></a>&lt;<a class="idl-code" data-link-type="interface" href="https://webidl.spec.whatwg.org/#idl-float"><c- b>float</c-></a>> <a class="idl-code" data-link-type="dict-member" data-type="sequence<float> " href="#dom-mlresample2doptions-scales"><c- g>scales</c-></a>;
+  <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#idl-sequence"><c- b>sequence</c-></a>&lt;<a class="idl-code" data-link-type="interface" href="https://webidl.spec.whatwg.org/#idl-unsigned-long"><c- b>unsigned</c-> <c- b>long</c-></a>> <a class="idl-code" data-link-type="dict-member" data-type="sequence<unsigned long> " href="#dom-mlresample2doptions-sizes"><c- g>sizes</c-></a>;
+  <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#idl-sequence"><c- b>sequence</c-></a>&lt;<a class="idl-code" data-link-type="interface" href="https://webidl.spec.whatwg.org/#idl-unsigned-long"><c- b>unsigned</c-> <c- b>long</c-></a>> <a class="idl-code" data-link-type="dict-member" data-type="sequence<unsigned long> " href="#dom-mlresample2doptions-axes"><c- g>axes</c-></a>;
 };
 
 <c- b>partial</c-> <c- b>interface</c-> <a class="idl-code" data-link-type="interface" href="#mlgraphbuilder"><c- g>MLGraphBuilder</c-></a> {
-  <a data-link-type="idl-name" href="#mloperand"><c- n>MLOperand</c-></a> <a href="#dom-mlgraphbuilder-resample2d"><code><c- g>resample2d</c-></code></a>(<a data-link-type="idl-name" href="#mloperand"><c- n>MLOperand</c-></a> <a href="#dom-mlgraphbuilder-resample2d-input-options-input"><code><c- g>input</c-></code></a>, <c- b>optional</c-> <a data-link-type="idl-name" href="#dictdef-mlresample2doptions"><c- n>MLResample2dOptions</c-></a> <a href="#dom-mlgraphbuilder-resample2d-input-options-options"><code><c- g>options</c-></code></a> = {});
+  <a data-link-type="idl-name" href="#mloperand"><c- n>MLOperand</c-></a> <a class="idl-code" data-link-type="method" href="#dom-mlgraphbuilder-resample2d"><c- g>resample2d</c-></a>(<a data-link-type="idl-name" href="#mloperand"><c- n>MLOperand</c-></a> <a href="#dom-mlgraphbuilder-resample2d-input-options-input"><code><c- g>input</c-></code></a>, <c- b>optional</c-> <a data-link-type="idl-name" href="#dictdef-mlresample2doptions"><c- n>MLResample2dOptions</c-></a> <a href="#dom-mlgraphbuilder-resample2d-input-options-options"><code><c- g>options</c-></code></a> = {});
 };
 
 <c- b>partial</c-> <c- b>interface</c-> <a class="idl-code" data-link-type="interface" href="#mlgraphbuilder"><c- g>MLGraphBuilder</c-></a> {
-  <a data-link-type="idl-name" href="#mloperand"><c- n>MLOperand</c-></a> <a href="#dom-mlgraphbuilder-reshape"><code><c- g>reshape</c-></code></a>(<a data-link-type="idl-name" href="#mloperand"><c- n>MLOperand</c-></a> <a href="#dom-mlgraphbuilder-reshape-input-newshape-input"><code><c- g>input</c-></code></a>, <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#idl-sequence"><c- b>sequence</c-></a>&lt;<a class="idl-code" data-link-type="interface" href="https://webidl.spec.whatwg.org/#idl-unsigned-long"><c- b>unsigned</c-> <c- b>long</c-></a>?> <a href="#dom-mlgraphbuilder-reshape-input-newshape-newshape"><code><c- g>newShape</c-></code></a>);
+  <a data-link-type="idl-name" href="#mloperand"><c- n>MLOperand</c-></a> <a class="idl-code" data-link-type="method" href="#dom-mlgraphbuilder-reshape"><c- g>reshape</c-></a>(<a data-link-type="idl-name" href="#mloperand"><c- n>MLOperand</c-></a> <a href="#dom-mlgraphbuilder-reshape-input-newshape-input"><code><c- g>input</c-></code></a>, <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#idl-sequence"><c- b>sequence</c-></a>&lt;<a class="idl-code" data-link-type="interface" href="https://webidl.spec.whatwg.org/#idl-unsigned-long"><c- b>unsigned</c-> <c- b>long</c-></a>?> <a href="#dom-mlgraphbuilder-reshape-input-newshape-newshape"><code><c- g>newShape</c-></code></a>);
 };
 
 <c- b>partial</c-> <c- b>interface</c-> <a class="idl-code" data-link-type="interface" href="#mlgraphbuilder"><c- g>MLGraphBuilder</c-></a> {
-  <a data-link-type="idl-name" href="#mloperand"><c- n>MLOperand</c-></a> <a href="#dom-mlgraphbuilder-sigmoid"><code><c- g>sigmoid</c-></code></a>(<a data-link-type="idl-name" href="#mloperand"><c- n>MLOperand</c-></a> <a href="#dom-mlgraphbuilder-sigmoid-x-x"><code><c- g>x</c-></code></a>);
-  <a data-link-type="idl-name" href="#mlactivation"><c- n>MLActivation</c-></a> <a href="#dom-mlgraphbuilder-sigmoid①"><code><c- g>sigmoid</c-></code></a>();
+  <a data-link-type="idl-name" href="#mloperand"><c- n>MLOperand</c-></a> <a class="idl-code" data-link-type="method" href="#dom-mlgraphbuilder-sigmoid"><c- g>sigmoid</c-></a>(<a data-link-type="idl-name" href="#mloperand"><c- n>MLOperand</c-></a> <a href="#dom-mlgraphbuilder-sigmoid-input-input"><code><c- g>input</c-></code></a>);
+  <a data-link-type="idl-name" href="#mlactivation"><c- n>MLActivation</c-></a> <a class="idl-code" data-link-type="method" href="#sigmoid-noargs"><c- g>sigmoid</c-></a>();
 };
 
 <c- b>partial</c-> <c- b>interface</c-> <a class="idl-code" data-link-type="interface" href="#mlgraphbuilder"><c- g>MLGraphBuilder</c-></a> {
-  <a data-link-type="idl-name" href="#mloperand"><c- n>MLOperand</c-></a> <a href="#dom-mlgraphbuilder-slice"><code><c- g>slice</c-></code></a>(<a data-link-type="idl-name" href="#mloperand"><c- n>MLOperand</c-></a> <a href="#dom-mlgraphbuilder-slice-input-starts-sizes-input"><code><c- g>input</c-></code></a>, <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#idl-sequence"><c- b>sequence</c-></a>&lt;<a class="idl-code" data-link-type="interface" href="https://webidl.spec.whatwg.org/#idl-unsigned-long"><c- b>unsigned</c-> <c- b>long</c-></a>> <a href="#dom-mlgraphbuilder-slice-input-starts-sizes-starts"><code><c- g>starts</c-></code></a>, <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#idl-sequence"><c- b>sequence</c-></a>&lt;<a class="idl-code" data-link-type="interface" href="https://webidl.spec.whatwg.org/#idl-unsigned-long"><c- b>unsigned</c-> <c- b>long</c-></a>> <a href="#dom-mlgraphbuilder-slice-input-starts-sizes-sizes"><code><c- g>sizes</c-></code></a>);
+  <a data-link-type="idl-name" href="#mloperand"><c- n>MLOperand</c-></a> <a class="idl-code" data-link-type="method" href="#dom-mlgraphbuilder-slice"><c- g>slice</c-></a>(<a data-link-type="idl-name" href="#mloperand"><c- n>MLOperand</c-></a> <a href="#dom-mlgraphbuilder-slice-input-starts-sizes-input"><code><c- g>input</c-></code></a>, <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#idl-sequence"><c- b>sequence</c-></a>&lt;<a class="idl-code" data-link-type="interface" href="https://webidl.spec.whatwg.org/#idl-unsigned-long"><c- b>unsigned</c-> <c- b>long</c-></a>> <a href="#dom-mlgraphbuilder-slice-input-starts-sizes-starts"><code><c- g>starts</c-></code></a>, <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#idl-sequence"><c- b>sequence</c-></a>&lt;<a class="idl-code" data-link-type="interface" href="https://webidl.spec.whatwg.org/#idl-unsigned-long"><c- b>unsigned</c-> <c- b>long</c-></a>> <a href="#dom-mlgraphbuilder-slice-input-starts-sizes-sizes"><code><c- g>sizes</c-></code></a>);
 };
 
 <c- b>partial</c-> <c- b>interface</c-> <a class="idl-code" data-link-type="interface" href="#mlgraphbuilder"><c- g>MLGraphBuilder</c-></a> {
-  <a data-link-type="idl-name" href="#mloperand"><c- n>MLOperand</c-></a> <a href="#dom-mlgraphbuilder-softmax"><code><c- g>softmax</c-></code></a>(<a data-link-type="idl-name" href="#mloperand"><c- n>MLOperand</c-></a> <a href="#dom-mlgraphbuilder-softmax-x-x"><code><c- g>x</c-></code></a>);
-  <a data-link-type="idl-name" href="#mlactivation"><c- n>MLActivation</c-></a> <a href="#dom-mlgraphbuilder-softmax①"><code><c- g>softmax</c-></code></a>();
+  <a data-link-type="idl-name" href="#mloperand"><c- n>MLOperand</c-></a> <a class="idl-code" data-link-type="method" href="#dom-mlgraphbuilder-softmax"><c- g>softmax</c-></a>(<a data-link-type="idl-name" href="#mloperand"><c- n>MLOperand</c-></a> <a href="#dom-mlgraphbuilder-softmax-input-input"><code><c- g>input</c-></code></a>);
+  <a data-link-type="idl-name" href="#mlactivation"><c- n>MLActivation</c-></a> <a class="idl-code" data-link-type="method" href="#softmax-noargs"><c- g>softmax</c-></a>();
 };
 
 <c- b>dictionary</c-> <a href="#dictdef-mlsoftplusoptions"><code><c- g>MLSoftplusOptions</c-></code></a> {
-  <a class="idl-code" data-link-type="interface" href="https://webidl.spec.whatwg.org/#idl-float"><c- b>float</c-></a> <a data-default="1" data-type="float " href="#dom-mlsoftplusoptions-steepness"><code><c- g>steepness</c-></code></a> = 1;
+  <a class="idl-code" data-link-type="interface" href="https://webidl.spec.whatwg.org/#idl-float"><c- b>float</c-></a> <a class="idl-code" data-default="1" data-link-type="dict-member" data-type="float " href="#dom-mlsoftplusoptions-steepness"><c- g>steepness</c-></a> = 1;
 };
 
 <c- b>partial</c-> <c- b>interface</c-> <a class="idl-code" data-link-type="interface" href="#mlgraphbuilder"><c- g>MLGraphBuilder</c-></a> {
-  <a data-link-type="idl-name" href="#mloperand"><c- n>MLOperand</c-></a> <a href="#dom-mlgraphbuilder-softplus"><code><c- g>softplus</c-></code></a>(<a data-link-type="idl-name" href="#mloperand"><c- n>MLOperand</c-></a> <a href="#dom-mlgraphbuilder-softplus-x-options-x"><code><c- g>x</c-></code></a>, <c- b>optional</c-> <a data-link-type="idl-name" href="#dictdef-mlsoftplusoptions"><c- n>MLSoftplusOptions</c-></a> <a href="#dom-mlgraphbuilder-softplus-x-options-options"><code><c- g>options</c-></code></a> = {});
-  <a data-link-type="idl-name" href="#mlactivation"><c- n>MLActivation</c-></a> <a href="#dom-mlgraphbuilder-softplus-options"><code><c- g>softplus</c-></code></a>(<c- b>optional</c-> <a data-link-type="idl-name" href="#dictdef-mlsoftplusoptions"><c- n>MLSoftplusOptions</c-></a> <a href="#dom-mlgraphbuilder-softplus-options-options"><code><c- g>options</c-></code></a> = {});
+  <a data-link-type="idl-name" href="#mloperand"><c- n>MLOperand</c-></a> <a class="idl-code" data-link-type="method" href="#dom-mlgraphbuilder-softplus"><c- g>softplus</c-></a>(<a data-link-type="idl-name" href="#mloperand"><c- n>MLOperand</c-></a> <a href="#dom-mlgraphbuilder-softplus-input-options-input"><code><c- g>input</c-></code></a>, <c- b>optional</c-> <a data-link-type="idl-name" href="#dictdef-mlsoftplusoptions"><c- n>MLSoftplusOptions</c-></a> <a href="#dom-mlgraphbuilder-softplus-input-options-options"><code><c- g>options</c-></code></a> = {});
+  <a data-link-type="idl-name" href="#mlactivation"><c- n>MLActivation</c-></a> <a class="idl-code" data-link-type="method" href="#dom-mlgraphbuilder-softplus-options"><c- g>softplus</c-></a>(<c- b>optional</c-> <a data-link-type="idl-name" href="#dictdef-mlsoftplusoptions"><c- n>MLSoftplusOptions</c-></a> <a href="#dom-mlgraphbuilder-softplus-options-options"><code><c- g>options</c-></code></a> = {});
 };
 
 <c- b>partial</c-> <c- b>interface</c-> <a class="idl-code" data-link-type="interface" href="#mlgraphbuilder"><c- g>MLGraphBuilder</c-></a> {
-  <a data-link-type="idl-name" href="#mloperand"><c- n>MLOperand</c-></a> <a href="#dom-mlgraphbuilder-softsign"><code><c- g>softsign</c-></code></a>(<a data-link-type="idl-name" href="#mloperand"><c- n>MLOperand</c-></a> <a href="#dom-mlgraphbuilder-softsign-x-x"><code><c- g>x</c-></code></a>);
-  <a data-link-type="idl-name" href="#mlactivation"><c- n>MLActivation</c-></a> <a href="#dom-mlgraphbuilder-softsign①"><code><c- g>softsign</c-></code></a>();
+  <a data-link-type="idl-name" href="#mloperand"><c- n>MLOperand</c-></a> <a class="idl-code" data-link-type="method" href="#dom-mlgraphbuilder-softsign"><c- g>softsign</c-></a>(<a data-link-type="idl-name" href="#mloperand"><c- n>MLOperand</c-></a> <a href="#dom-mlgraphbuilder-softsign-input-input"><code><c- g>input</c-></code></a>);
+  <a data-link-type="idl-name" href="#mlactivation"><c- n>MLActivation</c-></a> <a class="idl-code" data-link-type="method" href="#softsign-noargs"><c- g>softsign</c-></a>();
 };
 
 <c- b>dictionary</c-> <a href="#dictdef-mlsplitoptions"><code><c- g>MLSplitOptions</c-></code></a> {
-  <a class="idl-code" data-link-type="interface" href="https://webidl.spec.whatwg.org/#idl-unsigned-long"><c- b>unsigned</c-> <c- b>long</c-></a> <a data-default="0" data-type="unsigned long " href="#dom-mlsplitoptions-axis"><code><c- g>axis</c-></code></a> = 0;
+  <a class="idl-code" data-link-type="interface" href="https://webidl.spec.whatwg.org/#idl-unsigned-long"><c- b>unsigned</c-> <c- b>long</c-></a> <a class="idl-code" data-default="0" data-link-type="dict-member" data-type="unsigned long " href="#dom-mlsplitoptions-axis"><c- g>axis</c-></a> = 0;
 };
 
 <c- b>partial</c-> <c- b>interface</c-> <a class="idl-code" data-link-type="interface" href="#mlgraphbuilder"><c- g>MLGraphBuilder</c-></a> {
-  <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#idl-sequence"><c- b>sequence</c-></a>&lt;<a data-link-type="idl-name" href="#mloperand"><c- n>MLOperand</c-></a>> <a href="#dom-mlgraphbuilder-split"><code><c- g>split</c-></code></a>(<a data-link-type="idl-name" href="#mloperand"><c- n>MLOperand</c-></a> <a href="#dom-mlgraphbuilder-split-input-splits-options-input"><code><c- g>input</c-></code></a>,
+  <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#idl-sequence"><c- b>sequence</c-></a>&lt;<a data-link-type="idl-name" href="#mloperand"><c- n>MLOperand</c-></a>> <a class="idl-code" data-link-type="method" href="#dom-mlgraphbuilder-split"><c- g>split</c-></a>(<a data-link-type="idl-name" href="#mloperand"><c- n>MLOperand</c-></a> <a href="#dom-mlgraphbuilder-split-input-splits-options-input"><code><c- g>input</c-></code></a>,
                           (<a class="idl-code" data-link-type="interface" href="https://webidl.spec.whatwg.org/#idl-unsigned-long"><c- b>unsigned</c-> <c- b>long</c-></a> <c- b>or</c-> <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#idl-sequence"><c- b>sequence</c-></a>&lt;<a class="idl-code" data-link-type="interface" href="https://webidl.spec.whatwg.org/#idl-unsigned-long"><c- b>unsigned</c-> <c- b>long</c-></a>>) <a href="#dom-mlgraphbuilder-split-input-splits-options-splits"><code><c- g>splits</c-></code></a>,
                           <c- b>optional</c-> <a data-link-type="idl-name" href="#dictdef-mlsplitoptions"><c- n>MLSplitOptions</c-></a> <a href="#dom-mlgraphbuilder-split-input-splits-options-options"><code><c- g>options</c-></code></a> = {});
 };
 
 <c- b>dictionary</c-> <a href="#dictdef-mlsqueezeoptions"><code><c- g>MLSqueezeOptions</c-></code></a> {
-  <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#idl-sequence"><c- b>sequence</c-></a>&lt;<a class="idl-code" data-link-type="interface" href="https://webidl.spec.whatwg.org/#idl-unsigned-long"><c- b>unsigned</c-> <c- b>long</c-></a>> <a data-type="sequence<unsigned long> " href="#dom-mlsqueezeoptions-axes"><code><c- g>axes</c-></code></a>;
+  <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#idl-sequence"><c- b>sequence</c-></a>&lt;<a class="idl-code" data-link-type="interface" href="https://webidl.spec.whatwg.org/#idl-unsigned-long"><c- b>unsigned</c-> <c- b>long</c-></a>> <a class="idl-code" data-link-type="dict-member" data-type="sequence<unsigned long> " href="#dom-mlsqueezeoptions-axes"><c- g>axes</c-></a>;
 };
 
 <c- b>partial</c-> <c- b>interface</c-> <a class="idl-code" data-link-type="interface" href="#mlgraphbuilder"><c- g>MLGraphBuilder</c-></a> {
-  <a data-link-type="idl-name" href="#mloperand"><c- n>MLOperand</c-></a> <a href="#dom-mlgraphbuilder-squeeze"><code><c- g>squeeze</c-></code></a>(<a data-link-type="idl-name" href="#mloperand"><c- n>MLOperand</c-></a> <a href="#dom-mlgraphbuilder-squeeze-input-options-input"><code><c- g>input</c-></code></a>, <c- b>optional</c-> <a data-link-type="idl-name" href="#dictdef-mlsqueezeoptions"><c- n>MLSqueezeOptions</c-></a> <a href="#dom-mlgraphbuilder-squeeze-input-options-options"><code><c- g>options</c-></code></a> = {});
+  <a data-link-type="idl-name" href="#mloperand"><c- n>MLOperand</c-></a> <a class="idl-code" data-link-type="method" href="#dom-mlgraphbuilder-squeeze"><c- g>squeeze</c-></a>(<a data-link-type="idl-name" href="#mloperand"><c- n>MLOperand</c-></a> <a href="#dom-mlgraphbuilder-squeeze-input-options-input"><code><c- g>input</c-></code></a>, <c- b>optional</c-> <a data-link-type="idl-name" href="#dictdef-mlsqueezeoptions"><c- n>MLSqueezeOptions</c-></a> <a href="#dom-mlgraphbuilder-squeeze-input-options-options"><code><c- g>options</c-></code></a> = {});
 };
 
 <c- b>partial</c-> <c- b>interface</c-> <a class="idl-code" data-link-type="interface" href="#mlgraphbuilder"><c- g>MLGraphBuilder</c-></a> {
-  <a data-link-type="idl-name" href="#mloperand"><c- n>MLOperand</c-></a> <a href="#dom-mlgraphbuilder-tanh"><code><c- g>tanh</c-></code></a>(<a data-link-type="idl-name" href="#mloperand"><c- n>MLOperand</c-></a> <a href="#dom-mlgraphbuilder-tanh-x-x"><code><c- g>x</c-></code></a>);
-  <a data-link-type="idl-name" href="#mlactivation"><c- n>MLActivation</c-></a> <a href="#dom-mlgraphbuilder-tanh①"><code><c- g>tanh</c-></code></a>();
+  <a data-link-type="idl-name" href="#mloperand"><c- n>MLOperand</c-></a> <a class="idl-code" data-link-type="method" href="#dom-mlgraphbuilder-tanh"><c- g>tanh</c-></a>(<a data-link-type="idl-name" href="#mloperand"><c- n>MLOperand</c-></a> <a href="#dom-mlgraphbuilder-tanh-input-input"><code><c- g>input</c-></code></a>);
+  <a data-link-type="idl-name" href="#mlactivation"><c- n>MLActivation</c-></a> <a class="idl-code" data-link-type="method" href="#tanh-noargs"><c- g>tanh</c-></a>();
 };
 
 <c- b>dictionary</c-> <a href="#dictdef-mltransposeoptions"><code><c- g>MLTransposeOptions</c-></code></a> {
-  <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#idl-sequence"><c- b>sequence</c-></a>&lt;<a class="idl-code" data-link-type="interface" href="https://webidl.spec.whatwg.org/#idl-unsigned-long"><c- b>unsigned</c-> <c- b>long</c-></a>> <a data-type="sequence<unsigned long> " href="#dom-mltransposeoptions-permutation"><code><c- g>permutation</c-></code></a>;
+  <a data-link-type="dfn" href="https://webidl.spec.whatwg.org/#idl-sequence"><c- b>sequence</c-></a>&lt;<a class="idl-code" data-link-type="interface" href="https://webidl.spec.whatwg.org/#idl-unsigned-long"><c- b>unsigned</c-> <c- b>long</c-></a>> <a class="idl-code" data-link-type="dict-member" data-type="sequence<unsigned long> " href="#dom-mltransposeoptions-permutation"><c- g>permutation</c-></a>;
 };
 
 <c- b>partial</c-> <c- b>interface</c-> <a class="idl-code" data-link-type="interface" href="#mlgraphbuilder"><c- g>MLGraphBuilder</c-></a> {
-  <a data-link-type="idl-name" href="#mloperand"><c- n>MLOperand</c-></a> <a href="#dom-mlgraphbuilder-transpose"><code><c- g>transpose</c-></code></a>(<a data-link-type="idl-name" href="#mloperand"><c- n>MLOperand</c-></a> <a href="#dom-mlgraphbuilder-transpose-input-options-input"><code><c- g>input</c-></code></a>, <c- b>optional</c-> <a data-link-type="idl-name" href="#dictdef-mltransposeoptions"><c- n>MLTransposeOptions</c-></a> <a href="#dom-mlgraphbuilder-transpose-input-options-options"><code><c- g>options</c-></code></a> = {});
+  <a data-link-type="idl-name" href="#mloperand"><c- n>MLOperand</c-></a> <a class="idl-code" data-link-type="method" href="#dom-mlgraphbuilder-transpose"><c- g>transpose</c-></a>(<a data-link-type="idl-name" href="#mloperand"><c- n>MLOperand</c-></a> <a href="#dom-mlgraphbuilder-transpose-input-options-input"><code><c- g>input</c-></code></a>, <c- b>optional</c-> <a data-link-type="idl-name" href="#dictdef-mltransposeoptions"><c- n>MLTransposeOptions</c-></a> <a href="#dom-mlgraphbuilder-transpose-input-options-options"><code><c- g>options</c-></code></a> = {});
 };
 
 </pre>
@@ -5830,71 +9240,75 @@ <h2 class="no-num no-ref heading settled" id="issues-index"><span class="content
 </script>
 <script>/* Boilerplate: script-dfn-panel-json */
 window.dfnpanelData = {};
-window.dfnpanelData['f0ba04de'] = {"dfnID": "f0ba04de", "url": "https://drafts.csswg.org/css-values-4/#number", "dfnText": "number", "refSections": [{"refs": [{"id": "ref-for-number"}], "title": "7.6.3.2. The constant(value, type) method"}], "external": true};
-window.dfnpanelData['134d399c'] = {"dfnID": "134d399c", "url": "https://tc39.es/ecma262/#sec-construct", "dfnText": "construct", "refSections": [{"refs": [{"id": "ref-for-sec-construct"}], "title": "7.4.3. The MLNamedArrayBufferViews transfer algorithm"}], "external": true};
+window.dfnpanelData['f0ba04de'] = {"dfnID": "f0ba04de", "url": "https://drafts.csswg.org/css-values-4/#number", "dfnText": "number", "refSections": [{"refs": [{"id": "ref-for-number"}], "title": "7.6.30. The reshape() method"}], "external": true};
+window.dfnpanelData['944206e4'] = {"dfnID": "944206e4", "url": "https://tc39.es/ecma262/multipage/abstract-operations.html#sec-construct", "dfnText": "Construct", "refSections": [{"refs": [{"id": "ref-for-sec-construct"}], "title": "7.4.3. The MLNamedArrayBufferViews transfer algorithm"}], "external": true};
 window.dfnpanelData['e63ec40a'] = {"dfnID": "e63ec40a", "url": "https://tc39.es/ecma262/#table-the-typedarray-constructors", "dfnText": "element size", "refSections": [{"refs": [{"id": "ref-for-table-the-typedarray-constructors"}], "title": "7.3.1. The MLOperandDescriptor dictionary"}, {"refs": [{"id": "ref-for-table-the-typedarray-constructors\u2460"}, {"id": "ref-for-table-the-typedarray-constructors\u2461"}, {"id": "ref-for-table-the-typedarray-constructors\u2462"}, {"id": "ref-for-table-the-typedarray-constructors\u2463"}], "title": "7.4.2. Synchronous Execution"}, {"refs": [{"id": "ref-for-table-the-typedarray-constructors\u2464"}, {"id": "ref-for-table-the-typedarray-constructors\u2465"}], "title": "7.4.3. The MLNamedArrayBufferViews transfer algorithm"}], "external": true};
 window.dfnpanelData['38d90164'] = {"dfnID": "38d90164", "url": "https://tc39.es/ecma262/#table-the-typedarray-constructors", "dfnText": "element type", "refSections": [{"refs": [{"id": "ref-for-table-the-typedarray-constructors"}], "title": "7.3.1. The MLOperandDescriptor dictionary"}, {"refs": [{"id": "ref-for-table-the-typedarray-constructors\u2460"}, {"id": "ref-for-table-the-typedarray-constructors\u2461"}, {"id": "ref-for-table-the-typedarray-constructors\u2462"}, {"id": "ref-for-table-the-typedarray-constructors\u2463"}], "title": "7.4.2. Synchronous Execution"}, {"refs": [{"id": "ref-for-table-the-typedarray-constructors\u2464"}, {"id": "ref-for-table-the-typedarray-constructors\u2465"}], "title": "7.4.3. The MLNamedArrayBufferViews transfer algorithm"}], "external": true};
 window.dfnpanelData['023f906d'] = {"dfnID": "023f906d", "url": "https://tc39.es/ecma262/#table-the-typedarray-constructors", "dfnText": "view constructor", "refSections": [{"refs": [{"id": "ref-for-table-the-typedarray-constructors"}], "title": "7.3.1. The MLOperandDescriptor dictionary"}, {"refs": [{"id": "ref-for-table-the-typedarray-constructors\u2460"}, {"id": "ref-for-table-the-typedarray-constructors\u2461"}, {"id": "ref-for-table-the-typedarray-constructors\u2462"}, {"id": "ref-for-table-the-typedarray-constructors\u2463"}], "title": "7.4.2. Synchronous Execution"}, {"refs": [{"id": "ref-for-table-the-typedarray-constructors\u2464"}, {"id": "ref-for-table-the-typedarray-constructors\u2465"}], "title": "7.4.3. The MLNamedArrayBufferViews transfer algorithm"}], "external": true};
-window.dfnpanelData['ea2e2fcc'] = {"dfnID": "ea2e2fcc", "url": "https://w3c.github.io/FileAPI/#blob-url-entry-object", "dfnText": "object", "refSections": [{"refs": [{"id": "ref-for-blob-url-entry-object"}, {"id": "ref-for-blob-url-entry-object\u2460"}], "title": "7.3.2. The MLOperand interface"}, {"refs": [{"id": "ref-for-blob-url-entry-object\u2461"}, {"id": "ref-for-blob-url-entry-object\u2462"}], "title": "7.3.2.1. Creating MLOperand"}], "external": true};
+window.dfnpanelData['ea2e2fcc'] = {"dfnID": "ea2e2fcc", "url": "https://w3c.github.io/FileAPI/#blob-url-entry-object", "dfnText": "object", "refSections": [{"refs": [{"id": "ref-for-blob-url-entry-object"}, {"id": "ref-for-blob-url-entry-object\u2460"}], "title": "7.3.2. The MLOperand interface"}, {"refs": [{"id": "ref-for-blob-url-entry-object\u2461"}, {"id": "ref-for-blob-url-entry-object\u2462"}], "title": "7.3.2.1. Creating MLOperand"}, {"refs": [{"id": "ref-for-blob-url-entry-object\u2463"}, {"id": "ref-for-blob-url-entry-object\u2464"}], "title": "7.3.3. The MLActivation interface"}, {"refs": [{"id": "ref-for-blob-url-entry-object\u2465"}, {"id": "ref-for-blob-url-entry-object\u2466"}], "title": "7.3.3.1. Creating MLActivation"}], "external": true};
 window.dfnpanelData['7bb7bc47'] = {"dfnID": "7bb7bc47", "url": "https://tc39.es/proposal-float16array/#sec-float16array", "dfnText": "Float16Array", "refSections": [{"refs": [{"id": "ref-for-sec-float16array"}, {"id": "ref-for-sec-float16array\u2460"}], "title": "9.1. MLOperandType and ArrayBufferView compatibility"}], "external": true};
 window.dfnpanelData['b8308563'] = {"dfnID": "b8308563", "url": "https://html.spec.whatwg.org/multipage/workers.html#dedicatedworkerglobalscope", "dfnText": "DedicatedWorkerGlobalScope", "refSections": [{"refs": [{"id": "ref-for-dedicatedworkerglobalscope"}], "title": "7.1. The navigator.ml interface"}], "external": true};
 window.dfnpanelData['be0c27b2'] = {"dfnID": "be0c27b2", "url": "https://html.spec.whatwg.org/multipage/system-state.html#navigator", "dfnText": "Navigator", "refSections": [{"refs": [{"id": "ref-for-navigator"}, {"id": "ref-for-navigator\u2460"}], "title": "7.1. The navigator.ml interface"}], "external": true};
 window.dfnpanelData['5d7209e9'] = {"dfnID": "5d7209e9", "url": "https://html.spec.whatwg.org/multipage/nav-history-apis.html#window", "dfnText": "Window", "refSections": [{"refs": [{"id": "ref-for-window"}], "title": "7.1. The navigator.ml interface"}], "external": true};
 window.dfnpanelData['a4785085'] = {"dfnID": "a4785085", "url": "https://html.spec.whatwg.org/multipage/workers.html#workernavigator", "dfnText": "WorkerNavigator", "refSections": [{"refs": [{"id": "ref-for-workernavigator"}, {"id": "ref-for-workernavigator\u2460"}], "title": "7.1. The navigator.ml interface"}], "external": true};
-window.dfnpanelData['3610bd67'] = {"dfnID": "3610bd67", "url": "https://html.spec.whatwg.org/multipage/iframe-embed-object.html#allowed-to-use", "dfnText": "allowed to use", "refSections": [{"refs": [{"id": "ref-for-allowed-to-use"}], "title": "7.2.2. The createContext() method"}, {"refs": [{"id": "ref-for-allowed-to-use\u2460"}], "title": "7.2.3. The createContextSync() method"}, {"refs": [{"id": "ref-for-allowed-to-use\u2461"}], "title": "7.6.1. The MLGraphBuilder constructor"}], "external": true};
-window.dfnpanelData['3349d69f'] = {"dfnID": "3349d69f", "url": "https://html.spec.whatwg.org/multipage/nav-history-apis.html#concept-document-window", "dfnText": "associated document", "refSections": [{"refs": [{"id": "ref-for-concept-document-window"}], "title": "7.2.2. The createContext() method"}, {"refs": [{"id": "ref-for-concept-document-window\u2460"}], "title": "7.2.3. The createContextSync() method"}, {"refs": [{"id": "ref-for-concept-document-window\u2461"}], "title": "7.6.1. The MLGraphBuilder constructor"}], "external": true};
-window.dfnpanelData['a72449dd'] = {"dfnID": "a72449dd", "url": "https://html.spec.whatwg.org/multipage/infrastructure.html#in-parallel", "dfnText": "in parallel", "refSections": [{"refs": [{"id": "ref-for-in-parallel"}], "title": "7.2.2. The createContext() method"}, {"refs": [{"id": "ref-for-in-parallel\u2460"}], "title": "7.4.4. Asynchronous Execution"}], "external": true};
-window.dfnpanelData['e99bd18e'] = {"dfnID": "e99bd18e", "url": "https://html.spec.whatwg.org/multipage/webappapis.html#concept-relevant-global", "dfnText": "relevant global object", "refSections": [{"refs": [{"id": "ref-for-concept-relevant-global"}], "title": "7.2.2. The createContext() method"}, {"refs": [{"id": "ref-for-concept-relevant-global\u2460"}], "title": "7.2.3. The createContextSync() method"}, {"refs": [{"id": "ref-for-concept-relevant-global\u2461"}], "title": "7.6.1. The MLGraphBuilder constructor"}], "external": true};
-window.dfnpanelData['77b4c09a'] = {"dfnID": "77b4c09a", "url": "https://infra.spec.whatwg.org/#assert", "dfnText": "assert", "refSections": [{"refs": [{"id": "ref-for-assert"}, {"id": "ref-for-assert\u2460"}, {"id": "ref-for-assert\u2461"}, {"id": "ref-for-assert\u2462"}], "title": "7.4.2. Synchronous Execution"}, {"refs": [{"id": "ref-for-assert\u2463"}], "title": "7.6.2. The input() method"}], "external": true};
-window.dfnpanelData['1243a891'] = {"dfnID": "1243a891", "url": "https://infra.spec.whatwg.org/#map-exists", "dfnText": "exist", "refSections": [{"refs": [{"id": "ref-for-map-exists"}, {"id": "ref-for-map-exists\u2460"}], "title": "7.2.2. The createContext() method"}, {"refs": [{"id": "ref-for-map-exists\u2461"}], "title": "7.4.2. Synchronous Execution"}, {"refs": [{"id": "ref-for-map-exists\u2462"}, {"id": "ref-for-map-exists\u2463"}], "title": "7.5.2. Dispatch Execution Commands"}, {"refs": [{"id": "ref-for-map-exists\u2464"}, {"id": "ref-for-map-exists\u2465"}], "title": "7.6.2. The input() method"}], "external": true};
-window.dfnpanelData['860300d4'] = {"dfnID": "860300d4", "url": "https://infra.spec.whatwg.org/#implementation-defined", "dfnText": "implementation-defined", "refSections": [{"refs": [{"id": "ref-for-implementation-defined"}, {"id": "ref-for-implementation-defined\u2460"}], "title": "7.3.2. The MLOperand interface"}, {"refs": [{"id": "ref-for-implementation-defined\u2461"}], "title": "7.6.2. The input() method"}, {"refs": [{"id": "ref-for-implementation-defined\u2462"}], "title": "7.6.3.1. The constant(descriptor, bufferView) method"}, {"refs": [{"id": "ref-for-implementation-defined\u2463"}], "title": "7.6.3.2. The constant(value, type) method"}], "external": true};
-window.dfnpanelData['0698d556'] = {"dfnID": "0698d556", "url": "https://infra.spec.whatwg.org/#string", "dfnText": "string", "refSections": [{"refs": [{"id": "ref-for-string"}], "title": "7.3.2. The MLOperand interface"}, {"refs": [{"id": "ref-for-string\u2460"}, {"id": "ref-for-string\u2461"}], "title": "7.6.2. The input() method"}], "external": true};
+window.dfnpanelData['3610bd67'] = {"dfnID": "3610bd67", "url": "https://html.spec.whatwg.org/multipage/iframe-embed-object.html#allowed-to-use", "dfnText": "allowed to use", "refSections": [{"refs": [{"id": "ref-for-allowed-to-use"}, {"id": "ref-for-allowed-to-use\u2460"}], "title": "7.2.2. The createContext() method"}, {"refs": [{"id": "ref-for-allowed-to-use\u2461"}, {"id": "ref-for-allowed-to-use\u2462"}], "title": "7.2.3. The createContextSync() method"}, {"refs": [{"id": "ref-for-allowed-to-use\u2463"}], "title": "7.6.1. The MLGraphBuilder constructor"}], "external": true};
+window.dfnpanelData['3349d69f'] = {"dfnID": "3349d69f", "url": "https://html.spec.whatwg.org/multipage/nav-history-apis.html#concept-document-window", "dfnText": "associated document", "refSections": [{"refs": [{"id": "ref-for-concept-document-window"}, {"id": "ref-for-concept-document-window\u2460"}], "title": "7.2.2. The createContext() method"}, {"refs": [{"id": "ref-for-concept-document-window\u2461"}, {"id": "ref-for-concept-document-window\u2462"}], "title": "7.2.3. The createContextSync() method"}, {"refs": [{"id": "ref-for-concept-document-window\u2463"}], "title": "7.6.1. The MLGraphBuilder constructor"}], "external": true};
+window.dfnpanelData['a72449dd'] = {"dfnID": "a72449dd", "url": "https://html.spec.whatwg.org/multipage/infrastructure.html#in-parallel", "dfnText": "in parallel", "refSections": [{"refs": [{"id": "ref-for-in-parallel"}, {"id": "ref-for-in-parallel\u2460"}], "title": "7.2.2. The createContext() method"}, {"refs": [{"id": "ref-for-in-parallel\u2461"}], "title": "7.4.4. Asynchronous Execution"}, {"refs": [{"id": "ref-for-in-parallel\u2462"}], "title": "7.6.3.1. The build(outputs) method"}], "external": true};
+window.dfnpanelData['e99bd18e'] = {"dfnID": "e99bd18e", "url": "https://html.spec.whatwg.org/multipage/webappapis.html#concept-relevant-global", "dfnText": "relevant global object", "refSections": [{"refs": [{"id": "ref-for-concept-relevant-global"}, {"id": "ref-for-concept-relevant-global\u2460"}], "title": "7.2.2. The createContext() method"}, {"refs": [{"id": "ref-for-concept-relevant-global\u2461"}, {"id": "ref-for-concept-relevant-global\u2462"}], "title": "7.2.3. The createContextSync() method"}, {"refs": [{"id": "ref-for-concept-relevant-global\u2463"}], "title": "7.6.1. The MLGraphBuilder constructor"}], "external": true};
+window.dfnpanelData['77b4c09a'] = {"dfnID": "77b4c09a", "url": "https://infra.spec.whatwg.org/#assert", "dfnText": "assert", "refSections": [{"refs": [{"id": "ref-for-assert"}, {"id": "ref-for-assert\u2460"}, {"id": "ref-for-assert\u2461"}, {"id": "ref-for-assert\u2462"}], "title": "7.3.2.1. Creating MLOperand"}, {"refs": [{"id": "ref-for-assert\u2463"}], "title": "7.3.3.1. Creating MLActivation"}, {"refs": [{"id": "ref-for-assert\u2464"}, {"id": "ref-for-assert\u2465"}, {"id": "ref-for-assert\u2466"}, {"id": "ref-for-assert\u2467"}], "title": "7.4.2. Synchronous Execution"}, {"refs": [{"id": "ref-for-assert\u2468"}, {"id": "ref-for-assert\u2460\u24ea"}], "title": "7.6.2. The input() method"}, {"refs": [{"id": "ref-for-assert\u2460\u2460"}, {"id": "ref-for-assert\u2460\u2461"}, {"id": "ref-for-assert\u2460\u2462"}], "title": "7.6.3.2. The buildSync(outputs) method"}, {"refs": [{"id": "ref-for-assert\u2460\u2463"}], "title": "7.6.4.1. The constant(descriptor, bufferView) method"}, {"refs": [{"id": "ref-for-assert\u2460\u2464"}], "title": "7.6.5. The batchNormalization() method"}, {"refs": [{"id": "ref-for-assert\u2460\u2465"}], "title": "7.6.6.1. The clamp(operand, options) method"}, {"refs": [{"id": "ref-for-assert\u2460\u2466"}, {"id": "ref-for-assert\u2460\u2467"}, {"id": "ref-for-assert\u2460\u2468"}, {"id": "ref-for-assert\u2461\u24ea"}], "title": "7.6.7. The concat() method"}, {"refs": [{"id": "ref-for-assert\u2461\u2460"}, {"id": "ref-for-assert\u2461\u2461"}, {"id": "ref-for-assert\u2461\u2462"}], "title": "7.6.8. The conv2d() method"}, {"refs": [{"id": "ref-for-assert\u2461\u2463"}, {"id": "ref-for-assert\u2461\u2464"}, {"id": "ref-for-assert\u2461\u2465"}], "title": "7.6.9. The convTranspose2d() method"}, {"refs": [{"id": "ref-for-assert\u2461\u2466"}, {"id": "ref-for-assert\u2461\u2467"}, {"id": "ref-for-assert\u2461\u2468"}], "title": "7.6.10. Element-wise binary operations"}, {"refs": [{"id": "ref-for-assert\u2462\u24ea"}, {"id": "ref-for-assert\u2462\u2460"}], "title": "7.6.11. Element-wise unary operations"}, {"refs": [{"id": "ref-for-assert\u2462\u2461"}], "title": "7.6.13. The gemm() method"}, {"refs": [{"id": "ref-for-assert\u2462\u2462"}, {"id": "ref-for-assert\u2462\u2463"}, {"id": "ref-for-assert\u2462\u2464"}, {"id": "ref-for-assert\u2462\u2465"}], "title": "7.6.14. The gru() method"}, {"refs": [{"id": "ref-for-assert\u2462\u2466"}, {"id": "ref-for-assert\u2462\u2467"}, {"id": "ref-for-assert\u2462\u2468"}], "title": "7.6.15. The gruCell() method"}, {"refs": [{"id": "ref-for-assert\u2463\u24ea"}], "title": "7.6.16.1. The hardSigmoid(input, options) method"}, {"refs": [{"id": "ref-for-assert\u2463\u2460"}, {"id": "ref-for-assert\u2463\u2461"}, {"id": "ref-for-assert\u2463\u2462"}], "title": "7.6.18. The instanceNormalization() method"}, {"refs": [{"id": "ref-for-assert\u2463\u2463"}, {"id": "ref-for-assert\u2463\u2464"}, {"id": "ref-for-assert\u2463\u2465"}, {"id": "ref-for-assert\u2463\u2466"}, {"id": "ref-for-assert\u2463\u2467"}, {"id": "ref-for-assert\u2463\u2468"}, {"id": "ref-for-assert\u2464\u24ea"}], "title": "7.6.21. The lstm() method"}, {"refs": [{"id": "ref-for-assert\u2464\u2460"}, {"id": "ref-for-assert\u2464\u2461"}, {"id": "ref-for-assert\u2464\u2462"}, {"id": "ref-for-assert\u2464\u2463"}, {"id": "ref-for-assert\u2464\u2464"}], "title": "7.6.22. The lstmCell() method"}, {"refs": [{"id": "ref-for-assert\u2464\u2465"}], "title": "7.6.23. The matmul() method"}, {"refs": [{"id": "ref-for-assert\u2464\u2466"}], "title": "7.6.24. The pad() method"}, {"refs": [{"id": "ref-for-assert\u2464\u2467"}, {"id": "ref-for-assert\u2464\u2468"}], "title": "7.6.25. Pooling operations"}, {"refs": [{"id": "ref-for-assert\u2465\u24ea"}], "title": "7.6.26. The prelu() method"}, {"refs": [{"id": "ref-for-assert\u2465\u2460"}, {"id": "ref-for-assert\u2465\u2461"}], "title": "7.6.27. Reduction operations"}, {"refs": [{"id": "ref-for-assert\u2465\u2462"}], "title": "7.6.28.1. The relu(input) method"}, {"refs": [{"id": "ref-for-assert\u2465\u2463"}], "title": "7.6.30. The reshape() method"}, {"refs": [{"id": "ref-for-assert\u2465\u2464"}], "title": "7.6.31.1. The sigmoid(input) method"}, {"refs": [{"id": "ref-for-assert\u2465\u2465"}], "title": "7.6.32. The slice() method"}, {"refs": [{"id": "ref-for-assert\u2465\u2466"}], "title": "7.6.33.1. The softmax(input) method"}, {"refs": [{"id": "ref-for-assert\u2465\u2467"}], "title": "7.6.35.1. The softsign(input) method"}, {"refs": [{"id": "ref-for-assert\u2465\u2468"}], "title": "7.6.36. The split() method"}, {"refs": [{"id": "ref-for-assert\u2466\u24ea"}], "title": "7.6.37. The squeeze() method"}, {"refs": [{"id": "ref-for-assert\u2466\u2460"}], "title": "7.6.38.1. The tanh(input) method"}, {"refs": [{"id": "ref-for-assert\u2466\u2461"}], "title": "7.6.39. The transpose() method"}], "external": true};
+window.dfnpanelData['1243a891'] = {"dfnID": "1243a891", "url": "https://infra.spec.whatwg.org/#map-exists", "dfnText": "exist", "refSections": [{"refs": [{"id": "ref-for-map-exists"}, {"id": "ref-for-map-exists\u2460"}], "title": "7.2.2. The createContext() method"}, {"refs": [{"id": "ref-for-map-exists\u2461"}, {"id": "ref-for-map-exists\u2462"}], "title": "7.3.2.1. Creating MLOperand"}, {"refs": [{"id": "ref-for-map-exists\u2463"}], "title": "7.4.2. Synchronous Execution"}, {"refs": [{"id": "ref-for-map-exists\u2464"}, {"id": "ref-for-map-exists\u2465"}], "title": "7.5.2. Dispatch Execution Commands"}, {"refs": [{"id": "ref-for-map-exists\u2466"}, {"id": "ref-for-map-exists\u2467"}], "title": "7.6.2. The input() method"}, {"refs": [{"id": "ref-for-map-exists\u2468"}, {"id": "ref-for-map-exists\u2460\u24ea"}, {"id": "ref-for-map-exists\u2460\u2460"}], "title": "7.6.5. The batchNormalization() method"}, {"refs": [{"id": "ref-for-map-exists\u2460\u2461"}, {"id": "ref-for-map-exists\u2460\u2462"}, {"id": "ref-for-map-exists\u2460\u2463"}, {"id": "ref-for-map-exists\u2460\u2464"}, {"id": "ref-for-map-exists\u2460\u2465"}, {"id": "ref-for-map-exists\u2460\u2466"}, {"id": "ref-for-map-exists\u2460\u2467"}], "title": "7.6.8. The conv2d() method"}, {"refs": [{"id": "ref-for-map-exists\u2460\u2468"}, {"id": "ref-for-map-exists\u2461\u24ea"}, {"id": "ref-for-map-exists\u2461\u2460"}, {"id": "ref-for-map-exists\u2461\u2461"}, {"id": "ref-for-map-exists\u2461\u2462"}, {"id": "ref-for-map-exists\u2461\u2463"}, {"id": "ref-for-map-exists\u2461\u2464"}, {"id": "ref-for-map-exists\u2461\u2465"}], "title": "7.6.9. The convTranspose2d() method"}, {"refs": [{"id": "ref-for-map-exists\u2461\u2466"}], "title": "7.6.13. The gemm() method"}, {"refs": [{"id": "ref-for-map-exists\u2461\u2467"}, {"id": "ref-for-map-exists\u2461\u2468"}, {"id": "ref-for-map-exists\u2462\u24ea"}, {"id": "ref-for-map-exists\u2462\u2460"}], "title": "7.6.14. The gru() method"}, {"refs": [{"id": "ref-for-map-exists\u2462\u2461"}, {"id": "ref-for-map-exists\u2462\u2462"}, {"id": "ref-for-map-exists\u2462\u2463"}], "title": "7.6.15. The gruCell() method"}, {"refs": [{"id": "ref-for-map-exists\u2462\u2464"}, {"id": "ref-for-map-exists\u2462\u2465"}, {"id": "ref-for-map-exists\u2462\u2466"}, {"id": "ref-for-map-exists\u2462\u2467"}, {"id": "ref-for-map-exists\u2462\u2468"}, {"id": "ref-for-map-exists\u2463\u24ea"}], "title": "7.6.21. The lstm() method"}, {"refs": [{"id": "ref-for-map-exists\u2463\u2460"}, {"id": "ref-for-map-exists\u2463\u2461"}, {"id": "ref-for-map-exists\u2463\u2462"}, {"id": "ref-for-map-exists\u2463\u2463"}], "title": "7.6.22. The lstmCell() method"}, {"refs": [{"id": "ref-for-map-exists\u2463\u2464"}, {"id": "ref-for-map-exists\u2463\u2465"}, {"id": "ref-for-map-exists\u2463\u2466"}, {"id": "ref-for-map-exists\u2463\u2467"}, {"id": "ref-for-map-exists\u2463\u2468"}, {"id": "ref-for-map-exists\u2464\u24ea"}], "title": "7.6.25. Pooling operations"}, {"refs": [{"id": "ref-for-map-exists\u2464\u2460"}], "title": "7.6.27. Reduction operations"}, {"refs": [{"id": "ref-for-map-exists\u2464\u2461"}, {"id": "ref-for-map-exists\u2464\u2462"}, {"id": "ref-for-map-exists\u2464\u2463"}, {"id": "ref-for-map-exists\u2464\u2464"}], "title": "7.6.29. The resample2d() method"}, {"refs": [{"id": "ref-for-map-exists\u2464\u2465"}], "title": "7.6.37. The squeeze() method"}, {"refs": [{"id": "ref-for-map-exists\u2464\u2466"}, {"id": "ref-for-map-exists\u2464\u2467"}], "title": "7.6.39. The transpose() method"}], "external": true};
+window.dfnpanelData['45209803'] = {"dfnID": "45209803", "url": "https://infra.spec.whatwg.org/#map-iterate", "dfnText": "for each", "refSections": [{"refs": [{"id": "ref-for-map-iterate"}], "title": "7.3.1. The MLOperandDescriptor dictionary"}, {"refs": [{"id": "ref-for-map-iterate\u2460"}, {"id": "ref-for-map-iterate\u2461"}, {"id": "ref-for-map-iterate\u2462"}], "title": "7.4.2. Synchronous Execution"}, {"refs": [{"id": "ref-for-map-iterate\u2463"}], "title": "7.4.3. The MLNamedArrayBufferViews transfer algorithm"}, {"refs": [{"id": "ref-for-map-iterate\u2464"}, {"id": "ref-for-map-iterate\u2465"}, {"id": "ref-for-map-iterate\u2466"}, {"id": "ref-for-map-iterate\u2467"}], "title": "7.5.2. Dispatch Execution Commands"}, {"refs": [{"id": "ref-for-map-iterate\u2468"}, {"id": "ref-for-map-iterate\u2460\u24ea"}], "title": "7.6.3.2. The buildSync(outputs) method"}, {"refs": [{"id": "ref-for-map-iterate\u2460\u2460"}, {"id": "ref-for-map-iterate\u2460\u2461"}], "title": "7.6.7. The concat() method"}, {"refs": [{"id": "ref-for-map-iterate\u2460\u2462"}], "title": "7.6.23. The matmul() method"}], "external": true};
+window.dfnpanelData['860300d4'] = {"dfnID": "860300d4", "url": "https://infra.spec.whatwg.org/#implementation-defined", "dfnText": "implementation-defined", "refSections": [{"refs": [{"id": "ref-for-implementation-defined"}, {"id": "ref-for-implementation-defined\u2460"}], "title": "7.3.2. The MLOperand interface"}, {"refs": [{"id": "ref-for-implementation-defined\u2461"}], "title": "7.3.3. The MLActivation interface"}, {"refs": [{"id": "ref-for-implementation-defined\u2462"}, {"id": "ref-for-implementation-defined\u2463"}], "title": "7.3.3.1. Creating MLActivation"}, {"refs": [{"id": "ref-for-implementation-defined\u2464"}], "title": "7.6.2. The input() method"}, {"refs": [{"id": "ref-for-implementation-defined\u2465"}], "title": "7.6.3.2. The buildSync(outputs) method"}, {"refs": [{"id": "ref-for-implementation-defined\u2466"}], "title": "7.6.4.1. The constant(descriptor, bufferView) method"}, {"refs": [{"id": "ref-for-implementation-defined\u2467"}], "title": "7.6.4.2. The constant(value, type) method"}, {"refs": [{"id": "ref-for-implementation-defined\u2468"}], "title": "7.6.5. The batchNormalization() method"}, {"refs": [{"id": "ref-for-implementation-defined\u2460\u24ea"}, {"id": "ref-for-implementation-defined\u2460\u2460"}], "title": "7.6.6.1. The clamp(operand, options) method"}, {"refs": [{"id": "ref-for-implementation-defined\u2460\u2461"}, {"id": "ref-for-implementation-defined\u2460\u2462"}], "title": "7.6.7. The concat() method"}, {"refs": [{"id": "ref-for-implementation-defined\u2460\u2463"}, {"id": "ref-for-implementation-defined\u2460\u2464"}], "title": "7.6.8. The conv2d() method"}, {"refs": [{"id": "ref-for-implementation-defined\u2460\u2465"}, {"id": "ref-for-implementation-defined\u2460\u2466"}], "title": "7.6.9. The convTranspose2d() method"}, {"refs": [{"id": "ref-for-implementation-defined\u2460\u2467"}, {"id": "ref-for-implementation-defined\u2460\u2468"}, {"id": "ref-for-implementation-defined\u2461\u24ea"}], "title": "7.6.10. Element-wise binary operations"}, {"refs": [{"id": "ref-for-implementation-defined\u2461\u2460"}, {"id": "ref-for-implementation-defined\u2461\u2461"}], "title": "7.6.11. Element-wise unary operations"}, {"refs": [{"id": "ref-for-implementation-defined\u2461\u2462"}, {"id": "ref-for-implementation-defined\u2461\u2463"}], "title": "7.6.12.1. The elu(input, options) method"}, {"refs": [{"id": "ref-for-implementation-defined\u2461\u2464"}, {"id": "ref-for-implementation-defined\u2461\u2465"}], "title": "7.6.13. The gemm() method"}, {"refs": [{"id": "ref-for-implementation-defined\u2461\u2466"}], "title": "7.6.14. The gru() method"}, {"refs": [{"id": "ref-for-implementation-defined\u2461\u2467"}, {"id": "ref-for-implementation-defined\u2461\u2468"}], "title": "7.6.15. The gruCell() method"}, {"refs": [{"id": "ref-for-implementation-defined\u2462\u24ea"}, {"id": "ref-for-implementation-defined\u2462\u2460"}], "title": "7.6.16.1. The hardSigmoid(input, options) method"}, {"refs": [{"id": "ref-for-implementation-defined\u2462\u2461"}, {"id": "ref-for-implementation-defined\u2462\u2462"}], "title": "7.6.17.1. The hardSwish(input) method"}, {"refs": [{"id": "ref-for-implementation-defined\u2462\u2463"}, {"id": "ref-for-implementation-defined\u2462\u2464"}], "title": "7.6.18. The instanceNormalization() method"}, {"refs": [{"id": "ref-for-implementation-defined\u2462\u2465"}, {"id": "ref-for-implementation-defined\u2462\u2466"}], "title": "7.6.19.1. The leakyRelu(input, options) method"}, {"refs": [{"id": "ref-for-implementation-defined\u2462\u2467"}, {"id": "ref-for-implementation-defined\u2462\u2468"}], "title": "7.6.20.1. The linear(input, options) method"}, {"refs": [{"id": "ref-for-implementation-defined\u2463\u24ea"}, {"id": "ref-for-implementation-defined\u2463\u2460"}], "title": "7.6.21. The lstm() method"}, {"refs": [{"id": "ref-for-implementation-defined\u2463\u2461"}, {"id": "ref-for-implementation-defined\u2463\u2462"}], "title": "7.6.22. The lstmCell() method"}, {"refs": [{"id": "ref-for-implementation-defined\u2463\u2463"}, {"id": "ref-for-implementation-defined\u2463\u2464"}], "title": "7.6.23. The matmul() method"}, {"refs": [{"id": "ref-for-implementation-defined\u2463\u2465"}, {"id": "ref-for-implementation-defined\u2463\u2466"}], "title": "7.6.24. The pad() method"}, {"refs": [{"id": "ref-for-implementation-defined\u2463\u2467"}, {"id": "ref-for-implementation-defined\u2463\u2468"}], "title": "7.6.25. Pooling operations"}, {"refs": [{"id": "ref-for-implementation-defined\u2464\u24ea"}, {"id": "ref-for-implementation-defined\u2464\u2460"}], "title": "7.6.26. The prelu() method"}, {"refs": [{"id": "ref-for-implementation-defined\u2464\u2461"}, {"id": "ref-for-implementation-defined\u2464\u2462"}], "title": "7.6.27. Reduction operations"}, {"refs": [{"id": "ref-for-implementation-defined\u2464\u2463"}, {"id": "ref-for-implementation-defined\u2464\u2464"}], "title": "7.6.28.1. The relu(input) method"}, {"refs": [{"id": "ref-for-implementation-defined\u2464\u2465"}, {"id": "ref-for-implementation-defined\u2464\u2466"}], "title": "7.6.29. The resample2d() method"}, {"refs": [{"id": "ref-for-implementation-defined\u2464\u2467"}, {"id": "ref-for-implementation-defined\u2464\u2468"}], "title": "7.6.30. The reshape() method"}, {"refs": [{"id": "ref-for-implementation-defined\u2465\u24ea"}, {"id": "ref-for-implementation-defined\u2465\u2460"}], "title": "7.6.31.1. The sigmoid(input) method"}, {"refs": [{"id": "ref-for-implementation-defined\u2465\u2461"}, {"id": "ref-for-implementation-defined\u2465\u2462"}], "title": "7.6.32. The slice() method"}, {"refs": [{"id": "ref-for-implementation-defined\u2465\u2463"}, {"id": "ref-for-implementation-defined\u2465\u2464"}], "title": "7.6.33.1. The softmax(input) method"}, {"refs": [{"id": "ref-for-implementation-defined\u2465\u2465"}, {"id": "ref-for-implementation-defined\u2465\u2466"}], "title": "7.6.34.1. The softplus(input, options) method"}, {"refs": [{"id": "ref-for-implementation-defined\u2465\u2467"}, {"id": "ref-for-implementation-defined\u2465\u2468"}], "title": "7.6.35.1. The softsign(input) method"}, {"refs": [{"id": "ref-for-implementation-defined\u2466\u24ea"}, {"id": "ref-for-implementation-defined\u2466\u2460"}], "title": "7.6.36. The split() method"}, {"refs": [{"id": "ref-for-implementation-defined\u2466\u2461"}, {"id": "ref-for-implementation-defined\u2466\u2462"}], "title": "7.6.37. The squeeze() method"}, {"refs": [{"id": "ref-for-implementation-defined\u2466\u2463"}, {"id": "ref-for-implementation-defined\u2466\u2464"}], "title": "7.6.38.1. The tanh(input) method"}, {"refs": [{"id": "ref-for-implementation-defined\u2466\u2465"}, {"id": "ref-for-implementation-defined\u2466\u2466"}], "title": "7.6.39. The transpose() method"}], "external": true};
+window.dfnpanelData['0204d188'] = {"dfnID": "0204d188", "url": "https://infra.spec.whatwg.org/#list-size", "dfnText": "size", "refSections": [{"refs": [{"id": "ref-for-list-size"}], "title": "7.3.2. The MLOperand interface"}, {"refs": [{"id": "ref-for-list-size\u2460"}, {"id": "ref-for-list-size\u2461"}], "title": "7.3.2.1. Creating MLOperand"}, {"refs": [{"id": "ref-for-list-size\u2462"}, {"id": "ref-for-list-size\u2463"}, {"id": "ref-for-list-size\u2464"}, {"id": "ref-for-list-size\u2465"}, {"id": "ref-for-list-size\u2466"}, {"id": "ref-for-list-size\u2467"}, {"id": "ref-for-list-size\u2468"}], "title": "7.6.5. The batchNormalization() method"}, {"refs": [{"id": "ref-for-list-size\u2460\u24ea"}, {"id": "ref-for-list-size\u2460\u2460"}, {"id": "ref-for-list-size\u2460\u2461"}, {"id": "ref-for-list-size\u2460\u2462"}, {"id": "ref-for-list-size\u2460\u2463"}, {"id": "ref-for-list-size\u2460\u2464"}], "title": "7.6.8. The conv2d() method"}, {"refs": [{"id": "ref-for-list-size\u2460\u2465"}, {"id": "ref-for-list-size\u2460\u2466"}, {"id": "ref-for-list-size\u2460\u2467"}, {"id": "ref-for-list-size\u2460\u2468"}, {"id": "ref-for-list-size\u2461\u24ea"}, {"id": "ref-for-list-size\u2461\u2460"}, {"id": "ref-for-list-size\u2461\u2461"}, {"id": "ref-for-list-size\u2461\u2462"}], "title": "7.6.9. The convTranspose2d() method"}, {"refs": [{"id": "ref-for-list-size\u2461\u2463"}, {"id": "ref-for-list-size\u2461\u2464"}], "title": "7.6.13. The gemm() method"}, {"refs": [{"id": "ref-for-list-size\u2461\u2465"}], "title": "7.6.14. The gru() method"}, {"refs": [{"id": "ref-for-list-size\u2461\u2466"}], "title": "7.6.15. The gruCell() method"}, {"refs": [{"id": "ref-for-list-size\u2461\u2467"}, {"id": "ref-for-list-size\u2461\u2468"}, {"id": "ref-for-list-size\u2462\u24ea"}, {"id": "ref-for-list-size\u2462\u2460"}, {"id": "ref-for-list-size\u2462\u2461"}, {"id": "ref-for-list-size\u2462\u2462"}], "title": "7.6.18. The instanceNormalization() method"}, {"refs": [{"id": "ref-for-list-size\u2462\u2463"}], "title": "7.6.21. The lstm() method"}, {"refs": [{"id": "ref-for-list-size\u2462\u2464"}], "title": "7.6.22. The lstmCell() method"}, {"refs": [{"id": "ref-for-list-size\u2462\u2465"}, {"id": "ref-for-list-size\u2462\u2466"}], "title": "7.6.23. The matmul() method"}, {"refs": [{"id": "ref-for-list-size\u2462\u2467"}], "title": "7.6.24. The pad() method"}, {"refs": [{"id": "ref-for-list-size\u2462\u2468"}, {"id": "ref-for-list-size\u2463\u24ea"}, {"id": "ref-for-list-size\u2463\u2460"}, {"id": "ref-for-list-size\u2463\u2461"}, {"id": "ref-for-list-size\u2463\u2462"}, {"id": "ref-for-list-size\u2463\u2463"}], "title": "7.6.25. Pooling operations"}, {"refs": [{"id": "ref-for-list-size\u2463\u2464"}], "title": "7.6.27. Reduction operations"}, {"refs": [{"id": "ref-for-list-size\u2463\u2465"}, {"id": "ref-for-list-size\u2463\u2466"}, {"id": "ref-for-list-size\u2463\u2467"}], "title": "7.6.29. The resample2d() method"}, {"refs": [{"id": "ref-for-list-size\u2463\u2468"}], "title": "7.6.30. The reshape() method"}, {"refs": [{"id": "ref-for-list-size\u2464\u24ea"}], "title": "7.6.32. The slice() method"}, {"refs": [{"id": "ref-for-list-size\u2464\u2460"}], "title": "7.6.33.1. The softmax(input) method"}, {"refs": [{"id": "ref-for-list-size\u2464\u2461"}, {"id": "ref-for-list-size\u2464\u2462"}, {"id": "ref-for-list-size\u2464\u2463"}], "title": "7.6.36. The split() method"}, {"refs": [{"id": "ref-for-list-size\u2464\u2464"}], "title": "7.6.37. The squeeze() method"}], "external": true};
+window.dfnpanelData['0698d556'] = {"dfnID": "0698d556", "url": "https://infra.spec.whatwg.org/#string", "dfnText": "string", "refSections": [{"refs": [{"id": "ref-for-string"}], "title": "7.3.2. The MLOperand interface"}, {"refs": [{"id": "ref-for-string\u2460"}], "title": "7.3.3. The MLActivation interface"}, {"refs": [{"id": "ref-for-string\u2461"}], "title": "7.6.2. The input() method"}, {"refs": [{"id": "ref-for-string\u2462"}], "title": "7.6.3.2. The buildSync(outputs) method"}, {"refs": [{"id": "ref-for-string\u2463"}, {"id": "ref-for-string\u2464"}, {"id": "ref-for-string\u2465"}], "title": "7.6.8. The conv2d() method"}, {"refs": [{"id": "ref-for-string\u2466"}, {"id": "ref-for-string\u2467"}, {"id": "ref-for-string\u2468"}], "title": "7.6.9. The convTranspose2d() method"}, {"refs": [{"id": "ref-for-string\u2460\u24ea"}], "title": "7.6.24. The pad() method"}, {"refs": [{"id": "ref-for-string\u2460\u2460"}, {"id": "ref-for-string\u2460\u2461"}, {"id": "ref-for-string\u2460\u2462"}], "title": "7.6.25. Pooling operations"}, {"refs": [{"id": "ref-for-string\u2460\u2463"}], "title": "7.6.29. The resample2d() method"}], "external": true};
+window.dfnpanelData['febebe0e'] = {"dfnID": "febebe0e", "url": "https://infra.spec.whatwg.org/#the-range", "dfnText": "the range", "refSections": [{"refs": [{"id": "ref-for-the-range"}], "title": "7.6.5. The batchNormalization() method"}, {"refs": [{"id": "ref-for-the-range\u2460"}, {"id": "ref-for-the-range\u2461"}], "title": "7.6.7. The concat() method"}, {"refs": [{"id": "ref-for-the-range\u2462"}], "title": "7.6.23. The matmul() method"}, {"refs": [{"id": "ref-for-the-range\u2463"}], "title": "7.6.24. The pad() method"}, {"refs": [{"id": "ref-for-the-range\u2464"}], "title": "7.6.27. Reduction operations"}, {"refs": [{"id": "ref-for-the-range\u2465"}], "title": "7.6.29. The resample2d() method"}, {"refs": [{"id": "ref-for-the-range\u2466"}], "title": "7.6.37. The squeeze() method"}, {"refs": [{"id": "ref-for-the-range\u2467"}], "title": "7.6.39. The transpose() method"}], "external": true};
 window.dfnpanelData['1570624a'] = {"dfnID": "1570624a", "url": "https://w3c.github.io/webappsec-permissions-policy/#policy-controlled-feature-default-allowlist", "dfnText": "default allowlist", "refSections": [{"refs": [{"id": "ref-for-policy-controlled-feature-default-allowlist"}], "title": "7.2.1. Permissions Policy Integration"}], "external": true};
 window.dfnpanelData['cc890cc1'] = {"dfnID": "cc890cc1", "url": "https://w3c.github.io/webappsec-permissions-policy/#policy-controlled-feature", "dfnText": "policy-controlled feature", "refSections": [{"refs": [{"id": "ref-for-policy-controlled-feature"}], "title": "7.2.1. Permissions Policy Integration"}], "external": true};
-window.dfnpanelData['4acc25f4'] = {"dfnID": "4acc25f4", "url": "https://gpuweb.github.io/gpuweb/#buffer-interface", "dfnText": "GPUBuffer", "refSections": [{"refs": [{"id": "ref-for-buffer-interface"}], "title": "6.2. Device Selection"}, {"refs": [{"id": "ref-for-buffer-interface\u2460"}], "title": "7.5. The MLCommandEncoder interface"}, {"refs": [{"id": "ref-for-buffer-interface\u2461"}, {"id": "ref-for-buffer-interface\u2462"}], "title": "7.5.2. Dispatch Execution Commands"}, {"refs": [{"id": "ref-for-buffer-interface\u2463"}], "title": "7.6. The MLGraphBuilder interface"}], "external": true};
-window.dfnpanelData['12d473bb'] = {"dfnID": "12d473bb", "url": "https://gpuweb.github.io/gpuweb/#command-buffers", "dfnText": "GPUCommandBuffer", "refSections": [{"refs": [{"id": "ref-for-command-buffers"}], "title": "4. Privacy Considerations"}, {"refs": [{"id": "ref-for-command-buffers\u2460"}], "title": "6.1. Overview"}, {"refs": [{"id": "ref-for-command-buffers\u2461"}], "title": "7.4.5. WebGPU Interoperability"}, {"refs": [{"id": "ref-for-command-buffers\u2462"}], "title": "7.5. The MLCommandEncoder interface"}, {"refs": [{"id": "ref-for-command-buffers\u2463"}, {"id": "ref-for-command-buffers\u2464"}, {"id": "ref-for-command-buffers\u2465"}], "title": "7.5.3. Generate GPU Command Buffer"}], "external": true};
+window.dfnpanelData['be5908a3'] = {"dfnID": "be5908a3", "url": "https://gpuweb.github.io/gpuweb/#gpubuffer", "dfnText": "GPUBuffer", "refSections": [{"refs": [{"id": "ref-for-gpubuffer"}], "title": "6.2. Device Selection"}, {"refs": [{"id": "ref-for-gpubuffer\u2460"}], "title": "7.5. The MLCommandEncoder interface"}, {"refs": [{"id": "ref-for-gpubuffer\u2461"}, {"id": "ref-for-gpubuffer\u2462"}], "title": "7.5.2. Dispatch Execution Commands"}, {"refs": [{"id": "ref-for-gpubuffer\u2463"}, {"id": "ref-for-gpubuffer\u2464"}, {"id": "ref-for-gpubuffer\u2465"}], "title": "7.6. The MLGraphBuilder interface"}], "external": true};
+window.dfnpanelData['819f210c'] = {"dfnID": "819f210c", "url": "https://gpuweb.github.io/gpuweb/#gpucommandbuffer", "dfnText": "GPUCommandBuffer", "refSections": [{"refs": [{"id": "ref-for-gpucommandbuffer"}], "title": "4. Privacy Considerations"}, {"refs": [{"id": "ref-for-gpucommandbuffer\u2460"}], "title": "6.1. Overview"}, {"refs": [{"id": "ref-for-gpucommandbuffer\u2461"}], "title": "7.4.5. WebGPU Interoperability"}, {"refs": [{"id": "ref-for-gpucommandbuffer\u2462"}], "title": "7.5. The MLCommandEncoder interface"}, {"refs": [{"id": "ref-for-gpucommandbuffer\u2463"}, {"id": "ref-for-gpucommandbuffer\u2464"}, {"id": "ref-for-gpucommandbuffer\u2465"}, {"id": "ref-for-gpucommandbuffer\u2466"}], "title": "7.5.3. Generate GPU Command Buffer"}], "external": true};
 window.dfnpanelData['30267b02'] = {"dfnID": "30267b02", "url": "https://gpuweb.github.io/gpuweb/#dictdef-gpucommandbufferdescriptor", "dfnText": "GPUCommandBufferDescriptor", "refSections": [{"refs": [{"id": "ref-for-dictdef-gpucommandbufferdescriptor"}, {"id": "ref-for-dictdef-gpucommandbufferdescriptor\u2460"}], "title": "7.5.3. Generate GPU Command Buffer"}], "external": true};
-window.dfnpanelData['09adced4'] = {"dfnID": "09adced4", "url": "https://gpuweb.github.io/gpuweb/#gpu-device", "dfnText": "GPUDevice", "refSections": [{"refs": [{"id": "ref-for-gpu-device"}], "title": "3. Security Considerations"}, {"refs": [{"id": "ref-for-gpu-device\u2460"}], "title": "4. Privacy Considerations"}, {"refs": [{"id": "ref-for-gpu-device\u2461"}], "title": "6.2. Device Selection"}, {"refs": [{"id": "ref-for-gpu-device\u2462"}, {"id": "ref-for-gpu-device\u2463"}], "title": "7.2. The ML interface"}, {"refs": [{"id": "ref-for-gpu-device\u2464"}], "title": "7.2.2. The createContext() method"}, {"refs": [{"id": "ref-for-gpu-device\u2465"}], "title": "7.4.5. WebGPU Interoperability"}], "external": true};
-window.dfnpanelData['2e08cf49'] = {"dfnID": "2e08cf49", "url": "https://gpuweb.github.io/gpuweb/#queues", "dfnText": "GPUQueue", "refSections": [{"refs": [{"id": "ref-for-queues"}], "title": "6.1. Overview"}], "external": true};
-window.dfnpanelData['4abf7301'] = {"dfnID": "4abf7301", "url": "https://gpuweb.github.io/gpuweb/#texture-interface", "dfnText": "GPUTexture", "refSections": [{"refs": [{"id": "ref-for-texture-interface"}], "title": "6.2. Device Selection"}, {"refs": [{"id": "ref-for-texture-interface\u2460"}], "title": "7.5. The MLCommandEncoder interface"}], "external": true};
+window.dfnpanelData['189809f1'] = {"dfnID": "189809f1", "url": "https://gpuweb.github.io/gpuweb/#gpudevice", "dfnText": "GPUDevice", "refSections": [{"refs": [{"id": "ref-for-gpudevice"}], "title": "3. Security Considerations"}, {"refs": [{"id": "ref-for-gpudevice\u2460"}], "title": "4. Privacy Considerations"}, {"refs": [{"id": "ref-for-gpudevice\u2461"}], "title": "6.2. Device Selection"}, {"refs": [{"id": "ref-for-gpudevice\u2462"}, {"id": "ref-for-gpudevice\u2463"}], "title": "7.2. The ML interface"}, {"refs": [{"id": "ref-for-gpudevice\u2464"}], "title": "7.2.2. The createContext() method"}, {"refs": [{"id": "ref-for-gpudevice\u2465"}], "title": "7.4.5. WebGPU Interoperability"}], "external": true};
+window.dfnpanelData['899c078f'] = {"dfnID": "899c078f", "url": "https://gpuweb.github.io/gpuweb/#gpuqueue", "dfnText": "GPUQueue", "refSections": [{"refs": [{"id": "ref-for-gpuqueue"}], "title": "6.1. Overview"}], "external": true};
+window.dfnpanelData['4b3142af'] = {"dfnID": "4b3142af", "url": "https://gpuweb.github.io/gpuweb/#gputexture", "dfnText": "GPUTexture", "refSections": [{"refs": [{"id": "ref-for-gputexture"}], "title": "6.2. Device Selection"}, {"refs": [{"id": "ref-for-gputexture\u2460"}], "title": "7.5. The MLCommandEncoder interface"}], "external": true};
 window.dfnpanelData['49f6fafc'] = {"dfnID": "49f6fafc", "url": "https://gpuweb.github.io/gpuweb/#dom-gpubuffer-size", "dfnText": "size", "refSections": [{"refs": [{"id": "ref-for-dom-gpubuffer-size"}, {"id": "ref-for-dom-gpubuffer-size\u2460"}], "title": "7.5.2. Dispatch Execution Commands"}], "external": true};
 window.dfnpanelData['6c9ab516'] = {"dfnID": "6c9ab516", "url": "https://webidl.spec.whatwg.org/#ArrayBufferView", "dfnText": "ArrayBufferView", "refSections": [{"refs": [{"id": "ref-for-ArrayBufferView"}, {"id": "ref-for-ArrayBufferView\u2460"}], "title": "6.2. Device Selection"}, {"refs": [{"id": "ref-for-ArrayBufferView\u2461"}], "title": "7.3.1. The MLOperandDescriptor dictionary"}, {"refs": [{"id": "ref-for-ArrayBufferView\u2462"}, {"id": "ref-for-ArrayBufferView\u2463"}], "title": "7.4. The MLContext interface"}, {"refs": [{"id": "ref-for-ArrayBufferView\u2464"}], "title": "7.4.2. Synchronous Execution"}, {"refs": [{"id": "ref-for-ArrayBufferView\u2465"}], "title": "7.4.3. The MLNamedArrayBufferViews transfer algorithm"}, {"refs": [{"id": "ref-for-ArrayBufferView\u2466"}], "title": "7.6. The MLGraphBuilder interface"}, {"refs": [{"id": "ref-for-ArrayBufferView\u2467"}, {"id": "ref-for-ArrayBufferView\u2468"}], "title": "9.1. MLOperandType and ArrayBufferView compatibility"}], "external": true};
-window.dfnpanelData['dca2de17'] = {"dfnID": "dca2de17", "url": "https://webidl.spec.whatwg.org/#idl-DOMException", "dfnText": "DOMException", "refSections": [{"refs": [{"id": "ref-for-idl-DOMException"}], "title": "4. Privacy Considerations"}, {"refs": [{"id": "ref-for-idl-DOMException\u2460"}, {"id": "ref-for-idl-DOMException\u2461"}], "title": "7.2.2. The createContext() method"}, {"refs": [{"id": "ref-for-idl-DOMException\u2462"}, {"id": "ref-for-idl-DOMException\u2463"}], "title": "7.2.3. The createContextSync() method"}, {"refs": [{"id": "ref-for-idl-DOMException\u2464"}, {"id": "ref-for-idl-DOMException\u2465"}], "title": "7.3.2.1. Creating MLOperand"}, {"refs": [{"id": "ref-for-idl-DOMException\u2466"}, {"id": "ref-for-idl-DOMException\u2467"}, {"id": "ref-for-idl-DOMException\u2468"}, {"id": "ref-for-idl-DOMException\u2460\u24ea"}, {"id": "ref-for-idl-DOMException\u2460\u2460"}, {"id": "ref-for-idl-DOMException\u2460\u2461"}, {"id": "ref-for-idl-DOMException\u2460\u2462"}], "title": "7.4.2. Synchronous Execution"}, {"refs": [{"id": "ref-for-idl-DOMException\u2460\u2463"}, {"id": "ref-for-idl-DOMException\u2460\u2464"}, {"id": "ref-for-idl-DOMException\u2460\u2465"}, {"id": "ref-for-idl-DOMException\u2460\u2466"}], "title": "7.4.4. Asynchronous Execution"}, {"refs": [{"id": "ref-for-idl-DOMException\u2460\u2467"}], "title": "7.4.5. WebGPU Interoperability"}, {"refs": [{"id": "ref-for-idl-DOMException\u2460\u2468"}, {"id": "ref-for-idl-DOMException\u2461\u24ea"}], "title": "7.5.2. Dispatch Execution Commands"}, {"refs": [{"id": "ref-for-idl-DOMException\u2461\u2460"}], "title": "7.6.1. The MLGraphBuilder constructor"}, {"refs": [{"id": "ref-for-idl-DOMException\u2461\u2461"}, {"id": "ref-for-idl-DOMException\u2461\u2462"}, {"id": "ref-for-idl-DOMException\u2461\u2463"}, {"id": "ref-for-idl-DOMException\u2461\u2464"}, {"id": "ref-for-idl-DOMException\u2461\u2465"}], "title": "7.6.2. The input() method"}, {"refs": [{"id": "ref-for-idl-DOMException\u2461\u2466"}, {"id": "ref-for-idl-DOMException\u2461\u2467"}, {"id": "ref-for-idl-DOMException\u2461\u2468"}, {"id": "ref-for-idl-DOMException\u2462\u24ea"}, {"id": "ref-for-idl-DOMException\u2462\u2460"}], "title": "7.6.3.1. The constant(descriptor, bufferView) method"}, {"refs": [{"id": "ref-for-idl-DOMException\u2462\u2461"}, {"id": "ref-for-idl-DOMException\u2462\u2462"}], "title": "7.6.3.2. The constant(value, type) method"}], "external": true};
+window.dfnpanelData['dca2de17'] = {"dfnID": "dca2de17", "url": "https://webidl.spec.whatwg.org/#idl-DOMException", "dfnText": "DOMException", "refSections": [{"refs": [{"id": "ref-for-idl-DOMException"}], "title": "4. Privacy Considerations"}, {"refs": [{"id": "ref-for-idl-DOMException\u2460"}, {"id": "ref-for-idl-DOMException\u2461"}, {"id": "ref-for-idl-DOMException\u2462"}, {"id": "ref-for-idl-DOMException\u2463"}], "title": "7.2.2. The createContext() method"}, {"refs": [{"id": "ref-for-idl-DOMException\u2464"}, {"id": "ref-for-idl-DOMException\u2465"}, {"id": "ref-for-idl-DOMException\u2466"}, {"id": "ref-for-idl-DOMException\u2467"}], "title": "7.2.3. The createContextSync() method"}, {"refs": [{"id": "ref-for-idl-DOMException\u2468"}], "title": "7.3.3.1. Creating MLActivation"}, {"refs": [{"id": "ref-for-idl-DOMException\u2460\u24ea"}, {"id": "ref-for-idl-DOMException\u2460\u2460"}, {"id": "ref-for-idl-DOMException\u2460\u2461"}, {"id": "ref-for-idl-DOMException\u2460\u2462"}, {"id": "ref-for-idl-DOMException\u2460\u2463"}, {"id": "ref-for-idl-DOMException\u2460\u2464"}, {"id": "ref-for-idl-DOMException\u2460\u2465"}], "title": "7.4.2. Synchronous Execution"}, {"refs": [{"id": "ref-for-idl-DOMException\u2460\u2466"}, {"id": "ref-for-idl-DOMException\u2460\u2467"}, {"id": "ref-for-idl-DOMException\u2460\u2468"}, {"id": "ref-for-idl-DOMException\u2461\u24ea"}], "title": "7.4.4. Asynchronous Execution"}, {"refs": [{"id": "ref-for-idl-DOMException\u2461\u2460"}], "title": "7.4.5. WebGPU Interoperability"}, {"refs": [{"id": "ref-for-idl-DOMException\u2461\u2461"}, {"id": "ref-for-idl-DOMException\u2461\u2462"}], "title": "7.5.2. Dispatch Execution Commands"}, {"refs": [{"id": "ref-for-idl-DOMException\u2461\u2463"}], "title": "7.5.3. Generate GPU Command Buffer"}, {"refs": [{"id": "ref-for-idl-DOMException\u2461\u2464"}], "title": "7.6.1. The MLGraphBuilder constructor"}, {"refs": [{"id": "ref-for-idl-DOMException\u2461\u2465"}, {"id": "ref-for-idl-DOMException\u2461\u2466"}, {"id": "ref-for-idl-DOMException\u2461\u2467"}], "title": "7.6.2. The input() method"}, {"refs": [{"id": "ref-for-idl-DOMException\u2461\u2468"}], "title": "7.6.3.2. The buildSync(outputs) method"}, {"refs": [{"id": "ref-for-idl-DOMException\u2462\u24ea"}, {"id": "ref-for-idl-DOMException\u2462\u2460"}, {"id": "ref-for-idl-DOMException\u2462\u2461"}], "title": "7.6.4.1. The constant(descriptor, bufferView) method"}, {"refs": [{"id": "ref-for-idl-DOMException\u2462\u2462"}], "title": "7.6.4.2. The constant(value, type) method"}, {"refs": [{"id": "ref-for-idl-DOMException\u2462\u2463"}], "title": "7.6.5. The batchNormalization() method"}, {"refs": [{"id": "ref-for-idl-DOMException\u2462\u2464"}], "title": "7.6.6.1. The clamp(operand, options) method"}, {"refs": [{"id": "ref-for-idl-DOMException\u2462\u2465"}, {"id": "ref-for-idl-DOMException\u2462\u2466"}, {"id": "ref-for-idl-DOMException\u2462\u2467"}, {"id": "ref-for-idl-DOMException\u2462\u2468"}, {"id": "ref-for-idl-DOMException\u2463\u24ea"}, {"id": "ref-for-idl-DOMException\u2463\u2460"}], "title": "7.6.7. The concat() method"}, {"refs": [{"id": "ref-for-idl-DOMException\u2463\u2461"}, {"id": "ref-for-idl-DOMException\u2463\u2462"}, {"id": "ref-for-idl-DOMException\u2463\u2463"}, {"id": "ref-for-idl-DOMException\u2463\u2464"}, {"id": "ref-for-idl-DOMException\u2463\u2465"}, {"id": "ref-for-idl-DOMException\u2463\u2466"}, {"id": "ref-for-idl-DOMException\u2463\u2467"}, {"id": "ref-for-idl-DOMException\u2463\u2468"}], "title": "7.6.8. The conv2d() method"}, {"refs": [{"id": "ref-for-idl-DOMException\u2464\u24ea"}, {"id": "ref-for-idl-DOMException\u2464\u2460"}, {"id": "ref-for-idl-DOMException\u2464\u2461"}, {"id": "ref-for-idl-DOMException\u2464\u2462"}, {"id": "ref-for-idl-DOMException\u2464\u2463"}, {"id": "ref-for-idl-DOMException\u2464\u2464"}, {"id": "ref-for-idl-DOMException\u2464\u2465"}, {"id": "ref-for-idl-DOMException\u2464\u2466"}], "title": "7.6.9. The convTranspose2d() method"}, {"refs": [{"id": "ref-for-idl-DOMException\u2464\u2467"}, {"id": "ref-for-idl-DOMException\u2464\u2468"}, {"id": "ref-for-idl-DOMException\u2465\u24ea"}], "title": "7.6.10. Element-wise binary operations"}, {"refs": [{"id": "ref-for-idl-DOMException\u2465\u2460"}], "title": "7.6.11. Element-wise unary operations"}, {"refs": [{"id": "ref-for-idl-DOMException\u2465\u2461"}], "title": "7.6.12.1. The elu(input, options) method"}, {"refs": [{"id": "ref-for-idl-DOMException\u2465\u2462"}, {"id": "ref-for-idl-DOMException\u2465\u2463"}, {"id": "ref-for-idl-DOMException\u2465\u2464"}, {"id": "ref-for-idl-DOMException\u2465\u2465"}], "title": "7.6.13. The gemm() method"}, {"refs": [{"id": "ref-for-idl-DOMException\u2465\u2466"}, {"id": "ref-for-idl-DOMException\u2465\u2467"}, {"id": "ref-for-idl-DOMException\u2465\u2468"}, {"id": "ref-for-idl-DOMException\u2466\u24ea"}, {"id": "ref-for-idl-DOMException\u2466\u2460"}], "title": "7.6.14. The gru() method"}, {"refs": [{"id": "ref-for-idl-DOMException\u2466\u2461"}, {"id": "ref-for-idl-DOMException\u2466\u2462"}, {"id": "ref-for-idl-DOMException\u2466\u2463"}, {"id": "ref-for-idl-DOMException\u2466\u2464"}, {"id": "ref-for-idl-DOMException\u2466\u2465"}, {"id": "ref-for-idl-DOMException\u2466\u2466"}], "title": "7.6.15. The gruCell() method"}, {"refs": [{"id": "ref-for-idl-DOMException\u2466\u2467"}], "title": "7.6.16.1. The hardSigmoid(input, options) method"}, {"refs": [{"id": "ref-for-idl-DOMException\u2466\u2468"}], "title": "7.6.17.1. The hardSwish(input) method"}, {"refs": [{"id": "ref-for-idl-DOMException\u2467\u24ea"}, {"id": "ref-for-idl-DOMException\u2467\u2460"}, {"id": "ref-for-idl-DOMException\u2467\u2461"}, {"id": "ref-for-idl-DOMException\u2467\u2462"}], "title": "7.6.18. The instanceNormalization() method"}, {"refs": [{"id": "ref-for-idl-DOMException\u2467\u2463"}], "title": "7.6.19.1. The leakyRelu(input, options) method"}, {"refs": [{"id": "ref-for-idl-DOMException\u2467\u2464"}], "title": "7.6.20.1. The linear(input, options) method"}, {"refs": [{"id": "ref-for-idl-DOMException\u2467\u2465"}, {"id": "ref-for-idl-DOMException\u2467\u2466"}, {"id": "ref-for-idl-DOMException\u2467\u2467"}, {"id": "ref-for-idl-DOMException\u2467\u2468"}, {"id": "ref-for-idl-DOMException\u2468\u24ea"}, {"id": "ref-for-idl-DOMException\u2468\u2460"}, {"id": "ref-for-idl-DOMException\u2468\u2461"}, {"id": "ref-for-idl-DOMException\u2468\u2462"}, {"id": "ref-for-idl-DOMException\u2468\u2463"}, {"id": "ref-for-idl-DOMException\u2468\u2464"}, {"id": "ref-for-idl-DOMException\u2468\u2465"}, {"id": "ref-for-idl-DOMException\u2468\u2466"}, {"id": "ref-for-idl-DOMException\u2468\u2467"}, {"id": "ref-for-idl-DOMException\u2468\u2468"}, {"id": "ref-for-idl-DOMException\u2460\u24ea\u24ea"}, {"id": "ref-for-idl-DOMException\u2460\u24ea\u2460"}, {"id": "ref-for-idl-DOMException\u2460\u24ea\u2461"}, {"id": "ref-for-idl-DOMException\u2460\u24ea\u2462"}, {"id": "ref-for-idl-DOMException\u2460\u24ea\u2463"}], "title": "7.6.21. The lstm() method"}, {"refs": [{"id": "ref-for-idl-DOMException\u2460\u24ea\u2464"}, {"id": "ref-for-idl-DOMException\u2460\u24ea\u2465"}, {"id": "ref-for-idl-DOMException\u2460\u24ea\u2466"}, {"id": "ref-for-idl-DOMException\u2460\u24ea\u2467"}, {"id": "ref-for-idl-DOMException\u2460\u24ea\u2468"}, {"id": "ref-for-idl-DOMException\u2460\u2460\u24ea"}, {"id": "ref-for-idl-DOMException\u2460\u2460\u2460"}, {"id": "ref-for-idl-DOMException\u2460\u2460\u2461"}], "title": "7.6.22. The lstmCell() method"}, {"refs": [{"id": "ref-for-idl-DOMException\u2460\u2460\u2462"}, {"id": "ref-for-idl-DOMException\u2460\u2460\u2463"}, {"id": "ref-for-idl-DOMException\u2460\u2460\u2464"}], "title": "7.6.23. The matmul() method"}, {"refs": [{"id": "ref-for-idl-DOMException\u2460\u2460\u2465"}], "title": "7.6.24. The pad() method"}, {"refs": [{"id": "ref-for-idl-DOMException\u2460\u2460\u2466"}, {"id": "ref-for-idl-DOMException\u2460\u2460\u2467"}, {"id": "ref-for-idl-DOMException\u2460\u2460\u2468"}, {"id": "ref-for-idl-DOMException\u2460\u2461\u24ea"}, {"id": "ref-for-idl-DOMException\u2460\u2461\u2460"}, {"id": "ref-for-idl-DOMException\u2460\u2461\u2461"}, {"id": "ref-for-idl-DOMException\u2460\u2461\u2462"}, {"id": "ref-for-idl-DOMException\u2460\u2461\u2463"}, {"id": "ref-for-idl-DOMException\u2460\u2461\u2464"}], "title": "7.6.25. Pooling operations"}, {"refs": [{"id": "ref-for-idl-DOMException\u2460\u2461\u2465"}], "title": "7.6.26. The prelu() method"}, {"refs": [{"id": "ref-for-idl-DOMException\u2460\u2461\u2466"}, {"id": "ref-for-idl-DOMException\u2460\u2461\u2467"}], "title": "7.6.27. Reduction operations"}, {"refs": [{"id": "ref-for-idl-DOMException\u2460\u2461\u2468"}], "title": "7.6.28.1. The relu(input) method"}, {"refs": [{"id": "ref-for-idl-DOMException\u2460\u2462\u24ea"}, {"id": "ref-for-idl-DOMException\u2460\u2462\u2460"}, {"id": "ref-for-idl-DOMException\u2460\u2462\u2461"}], "title": "7.6.29. The resample2d() method"}, {"refs": [{"id": "ref-for-idl-DOMException\u2460\u2462\u2462"}, {"id": "ref-for-idl-DOMException\u2460\u2462\u2463"}, {"id": "ref-for-idl-DOMException\u2460\u2462\u2464"}, {"id": "ref-for-idl-DOMException\u2460\u2462\u2465"}, {"id": "ref-for-idl-DOMException\u2460\u2462\u2466"}], "title": "7.6.30. The reshape() method"}, {"refs": [{"id": "ref-for-idl-DOMException\u2460\u2462\u2467"}], "title": "7.6.31.1. The sigmoid(input) method"}, {"refs": [{"id": "ref-for-idl-DOMException\u2460\u2462\u2468"}], "title": "7.6.32. The slice() method"}, {"refs": [{"id": "ref-for-idl-DOMException\u2460\u2463\u24ea"}, {"id": "ref-for-idl-DOMException\u2460\u2463\u2460"}], "title": "7.6.33.1. The softmax(input) method"}, {"refs": [{"id": "ref-for-idl-DOMException\u2460\u2463\u2461"}], "title": "7.6.34.1. The softplus(input, options) method"}, {"refs": [{"id": "ref-for-idl-DOMException\u2460\u2463\u2462"}], "title": "7.6.35.1. The softsign(input) method"}, {"refs": [{"id": "ref-for-idl-DOMException\u2460\u2463\u2463"}], "title": "7.6.36. The split() method"}, {"refs": [{"id": "ref-for-idl-DOMException\u2460\u2463\u2464"}], "title": "7.6.37. The squeeze() method"}, {"refs": [{"id": "ref-for-idl-DOMException\u2460\u2463\u2465"}], "title": "7.6.38.1. The tanh(input) method"}, {"refs": [{"id": "ref-for-idl-DOMException\u2460\u2463\u2466"}], "title": "7.6.39. The transpose() method"}], "external": true};
 window.dfnpanelData['8855a9aa'] = {"dfnID": "8855a9aa", "url": "https://webidl.spec.whatwg.org/#idl-DOMString", "dfnText": "DOMString", "refSections": [{"refs": [{"id": "ref-for-idl-DOMString"}, {"id": "ref-for-idl-DOMString\u2460"}], "title": "7.3. The MLGraph interface"}, {"refs": [{"id": "ref-for-idl-DOMString\u2461"}], "title": "7.4. The MLContext interface"}, {"refs": [{"id": "ref-for-idl-DOMString\u2462"}], "title": "7.5. The MLCommandEncoder interface"}, {"refs": [{"id": "ref-for-idl-DOMString\u2463"}, {"id": "ref-for-idl-DOMString\u2464"}], "title": "7.6. The MLGraphBuilder interface"}], "external": true};
-window.dfnpanelData['24fec0b4'] = {"dfnID": "24fec0b4", "url": "https://webidl.spec.whatwg.org/#dataerror", "dfnText": "DataError", "refSections": [{"refs": [{"id": "ref-for-dataerror"}, {"id": "ref-for-dataerror\u2460"}, {"id": "ref-for-dataerror\u2461"}, {"id": "ref-for-dataerror\u2462"}], "title": "7.4.2. Synchronous Execution"}, {"refs": [{"id": "ref-for-dataerror\u2463"}, {"id": "ref-for-dataerror\u2464"}], "title": "7.4.4. Asynchronous Execution"}, {"refs": [{"id": "ref-for-dataerror\u2465"}], "title": "7.5.2. Dispatch Execution Commands"}, {"refs": [{"id": "ref-for-dataerror\u2466"}, {"id": "ref-for-dataerror\u2467"}], "title": "7.6.2. The input() method"}, {"refs": [{"id": "ref-for-dataerror\u2468"}, {"id": "ref-for-dataerror\u2460\u24ea"}], "title": "7.6.3.1. The constant(descriptor, bufferView) method"}], "external": true};
+window.dfnpanelData['24fec0b4'] = {"dfnID": "24fec0b4", "url": "https://webidl.spec.whatwg.org/#dataerror", "dfnText": "DataError", "refSections": [{"refs": [{"id": "ref-for-dataerror"}, {"id": "ref-for-dataerror\u2460"}, {"id": "ref-for-dataerror\u2461"}, {"id": "ref-for-dataerror\u2462"}], "title": "7.4.2. Synchronous Execution"}, {"refs": [{"id": "ref-for-dataerror\u2463"}, {"id": "ref-for-dataerror\u2464"}], "title": "7.4.4. Asynchronous Execution"}, {"refs": [{"id": "ref-for-dataerror\u2465"}], "title": "7.5.2. Dispatch Execution Commands"}, {"refs": [{"id": "ref-for-dataerror\u2466"}, {"id": "ref-for-dataerror\u2467"}], "title": "7.6.2. The input() method"}, {"refs": [{"id": "ref-for-dataerror\u2468"}, {"id": "ref-for-dataerror\u2460\u24ea"}], "title": "7.6.4.1. The constant(descriptor, bufferView) method"}, {"refs": [{"id": "ref-for-dataerror\u2460\u2460"}, {"id": "ref-for-dataerror\u2460\u2461"}, {"id": "ref-for-dataerror\u2460\u2462"}, {"id": "ref-for-dataerror\u2460\u2463"}, {"id": "ref-for-dataerror\u2460\u2464"}], "title": "7.6.7. The concat() method"}, {"refs": [{"id": "ref-for-dataerror\u2460\u2465"}, {"id": "ref-for-dataerror\u2460\u2466"}, {"id": "ref-for-dataerror\u2460\u2467"}, {"id": "ref-for-dataerror\u2460\u2468"}, {"id": "ref-for-dataerror\u2461\u24ea"}, {"id": "ref-for-dataerror\u2461\u2460"}, {"id": "ref-for-dataerror\u2461\u2461"}], "title": "7.6.8. The conv2d() method"}, {"refs": [{"id": "ref-for-dataerror\u2461\u2462"}, {"id": "ref-for-dataerror\u2461\u2463"}, {"id": "ref-for-dataerror\u2461\u2464"}, {"id": "ref-for-dataerror\u2461\u2465"}, {"id": "ref-for-dataerror\u2461\u2466"}, {"id": "ref-for-dataerror\u2461\u2467"}, {"id": "ref-for-dataerror\u2461\u2468"}], "title": "7.6.9. The convTranspose2d() method"}, {"refs": [{"id": "ref-for-dataerror\u2462\u24ea"}, {"id": "ref-for-dataerror\u2462\u2460"}], "title": "7.6.10. Element-wise binary operations"}, {"refs": [{"id": "ref-for-dataerror\u2462\u2461"}, {"id": "ref-for-dataerror\u2462\u2462"}, {"id": "ref-for-dataerror\u2462\u2463"}], "title": "7.6.13. The gemm() method"}, {"refs": [{"id": "ref-for-dataerror\u2462\u2464"}, {"id": "ref-for-dataerror\u2462\u2465"}, {"id": "ref-for-dataerror\u2462\u2466"}, {"id": "ref-for-dataerror\u2462\u2467"}], "title": "7.6.14. The gru() method"}, {"refs": [{"id": "ref-for-dataerror\u2462\u2468"}, {"id": "ref-for-dataerror\u2463\u24ea"}, {"id": "ref-for-dataerror\u2463\u2460"}, {"id": "ref-for-dataerror\u2463\u2461"}, {"id": "ref-for-dataerror\u2463\u2462"}], "title": "7.6.15. The gruCell() method"}, {"refs": [{"id": "ref-for-dataerror\u2463\u2463"}, {"id": "ref-for-dataerror\u2463\u2464"}, {"id": "ref-for-dataerror\u2463\u2465"}], "title": "7.6.18. The instanceNormalization() method"}, {"refs": [{"id": "ref-for-dataerror\u2463\u2466"}, {"id": "ref-for-dataerror\u2463\u2467"}, {"id": "ref-for-dataerror\u2463\u2468"}, {"id": "ref-for-dataerror\u2464\u24ea"}, {"id": "ref-for-dataerror\u2464\u2460"}, {"id": "ref-for-dataerror\u2464\u2461"}, {"id": "ref-for-dataerror\u2464\u2462"}, {"id": "ref-for-dataerror\u2464\u2463"}, {"id": "ref-for-dataerror\u2464\u2464"}, {"id": "ref-for-dataerror\u2464\u2465"}, {"id": "ref-for-dataerror\u2464\u2466"}, {"id": "ref-for-dataerror\u2464\u2467"}, {"id": "ref-for-dataerror\u2464\u2468"}, {"id": "ref-for-dataerror\u2465\u24ea"}, {"id": "ref-for-dataerror\u2465\u2460"}, {"id": "ref-for-dataerror\u2465\u2461"}, {"id": "ref-for-dataerror\u2465\u2462"}, {"id": "ref-for-dataerror\u2465\u2463"}], "title": "7.6.21. The lstm() method"}, {"refs": [{"id": "ref-for-dataerror\u2465\u2464"}, {"id": "ref-for-dataerror\u2465\u2465"}, {"id": "ref-for-dataerror\u2465\u2466"}, {"id": "ref-for-dataerror\u2465\u2467"}, {"id": "ref-for-dataerror\u2465\u2468"}, {"id": "ref-for-dataerror\u2466\u24ea"}, {"id": "ref-for-dataerror\u2466\u2460"}], "title": "7.6.22. The lstmCell() method"}, {"refs": [{"id": "ref-for-dataerror\u2466\u2461"}, {"id": "ref-for-dataerror\u2466\u2462"}, {"id": "ref-for-dataerror\u2466\u2463"}, {"id": "ref-for-dataerror\u2466\u2464"}, {"id": "ref-for-dataerror\u2466\u2465"}, {"id": "ref-for-dataerror\u2466\u2466"}, {"id": "ref-for-dataerror\u2466\u2467"}, {"id": "ref-for-dataerror\u2466\u2468"}], "title": "7.6.25. Pooling operations"}, {"refs": [{"id": "ref-for-dataerror\u2467\u24ea"}], "title": "7.6.27. Reduction operations"}, {"refs": [{"id": "ref-for-dataerror\u2467\u2460"}, {"id": "ref-for-dataerror\u2467\u2461"}], "title": "7.6.29. The resample2d() method"}, {"refs": [{"id": "ref-for-dataerror\u2467\u2462"}, {"id": "ref-for-dataerror\u2467\u2463"}, {"id": "ref-for-dataerror\u2467\u2464"}, {"id": "ref-for-dataerror\u2467\u2465"}], "title": "7.6.30. The reshape() method"}, {"refs": [{"id": "ref-for-dataerror\u2467\u2466"}], "title": "7.6.33.1. The softmax(input) method"}], "external": true};
 window.dfnpanelData['889e932f'] = {"dfnID": "889e932f", "url": "https://webidl.spec.whatwg.org/#Exposed", "dfnText": "Exposed", "refSections": [{"refs": [{"id": "ref-for-Exposed"}, {"id": "ref-for-Exposed\u2460"}, {"id": "ref-for-Exposed\u2461"}], "title": "7.2. The ML interface"}, {"refs": [{"id": "ref-for-Exposed\u2462"}], "title": "7.3. The MLGraph interface"}, {"refs": [{"id": "ref-for-Exposed\u2463"}], "title": "7.3.2. The MLOperand interface"}, {"refs": [{"id": "ref-for-Exposed\u2464"}], "title": "7.3.3. The MLActivation interface"}, {"refs": [{"id": "ref-for-Exposed\u2465"}], "title": "7.4. The MLContext interface"}, {"refs": [{"id": "ref-for-Exposed\u2466"}], "title": "7.4.2. Synchronous Execution"}, {"refs": [{"id": "ref-for-Exposed\u2467"}], "title": "7.5. The MLCommandEncoder interface"}, {"refs": [{"id": "ref-for-Exposed\u2468"}, {"id": "ref-for-Exposed\u2460\u24ea"}], "title": "7.6. The MLGraphBuilder interface"}], "external": true};
 window.dfnpanelData['0a296dfe'] = {"dfnID": "0a296dfe", "url": "https://webidl.spec.whatwg.org/#idl-Float32Array", "dfnText": "Float32Array", "refSections": [{"refs": [{"id": "ref-for-idl-Float32Array"}], "title": "9.1. MLOperandType and ArrayBufferView compatibility"}], "external": true};
 window.dfnpanelData['0e951fad'] = {"dfnID": "0e951fad", "url": "https://webidl.spec.whatwg.org/#idl-Int32Array", "dfnText": "Int32Array", "refSections": [{"refs": [{"id": "ref-for-idl-Int32Array"}], "title": "9.1. MLOperandType and ArrayBufferView compatibility"}], "external": true};
 window.dfnpanelData['1c26b9f3'] = {"dfnID": "1c26b9f3", "url": "https://webidl.spec.whatwg.org/#idl-Int8Array", "dfnText": "Int8Array", "refSections": [{"refs": [{"id": "ref-for-idl-Int8Array"}], "title": "9.1. MLOperandType and ArrayBufferView compatibility"}], "external": true};
-window.dfnpanelData['3182eb71'] = {"dfnID": "3182eb71", "url": "https://webidl.spec.whatwg.org/#notsupportederror", "dfnText": "NotSupportedError", "refSections": [{"refs": [{"id": "ref-for-notsupportederror"}], "title": "7.2.2. The createContext() method"}, {"refs": [{"id": "ref-for-notsupportederror\u2460"}], "title": "7.2.3. The createContextSync() method"}], "external": true};
-window.dfnpanelData['34443090'] = {"dfnID": "34443090", "url": "https://webidl.spec.whatwg.org/#operationerror", "dfnText": "OperationError", "refSections": [{"refs": [{"id": "ref-for-operationerror"}], "title": "4. Privacy Considerations"}, {"refs": [{"id": "ref-for-operationerror\u2460"}, {"id": "ref-for-operationerror\u2461"}, {"id": "ref-for-operationerror\u2462"}], "title": "7.4.2. Synchronous Execution"}, {"refs": [{"id": "ref-for-operationerror\u2463"}, {"id": "ref-for-operationerror\u2464"}], "title": "7.4.4. Asynchronous Execution"}, {"refs": [{"id": "ref-for-operationerror\u2465"}], "title": "7.4.5. WebGPU Interoperability"}, {"refs": [{"id": "ref-for-operationerror\u2466"}], "title": "7.5.2. Dispatch Execution Commands"}, {"refs": [{"id": "ref-for-operationerror\u2467"}], "title": "7.6.2. The input() method"}, {"refs": [{"id": "ref-for-operationerror\u2468"}], "title": "7.6.3.1. The constant(descriptor, bufferView) method"}], "external": true};
+window.dfnpanelData['3182eb71'] = {"dfnID": "3182eb71", "url": "https://webidl.spec.whatwg.org/#notsupportederror", "dfnText": "NotSupportedError", "refSections": [{"refs": [{"id": "ref-for-notsupportederror"}, {"id": "ref-for-notsupportederror\u2460"}], "title": "7.2.2. The createContext() method"}, {"refs": [{"id": "ref-for-notsupportederror\u2461"}, {"id": "ref-for-notsupportederror\u2462"}], "title": "7.2.3. The createContextSync() method"}], "external": true};
+window.dfnpanelData['34443090'] = {"dfnID": "34443090", "url": "https://webidl.spec.whatwg.org/#operationerror", "dfnText": "OperationError", "refSections": [{"refs": [{"id": "ref-for-operationerror"}], "title": "4. Privacy Considerations"}, {"refs": [{"id": "ref-for-operationerror\u2460"}], "title": "7.3.3.1. Creating MLActivation"}, {"refs": [{"id": "ref-for-operationerror\u2461"}, {"id": "ref-for-operationerror\u2462"}, {"id": "ref-for-operationerror\u2463"}], "title": "7.4.2. Synchronous Execution"}, {"refs": [{"id": "ref-for-operationerror\u2464"}, {"id": "ref-for-operationerror\u2465"}], "title": "7.4.4. Asynchronous Execution"}, {"refs": [{"id": "ref-for-operationerror\u2466"}], "title": "7.4.5. WebGPU Interoperability"}, {"refs": [{"id": "ref-for-operationerror\u2467"}], "title": "7.5.2. Dispatch Execution Commands"}, {"refs": [{"id": "ref-for-operationerror\u2468"}], "title": "7.5.3. Generate GPU Command Buffer"}, {"refs": [{"id": "ref-for-operationerror\u2460\u24ea"}], "title": "7.6.2. The input() method"}, {"refs": [{"id": "ref-for-operationerror\u2460\u2460"}], "title": "7.6.3.2. The buildSync(outputs) method"}, {"refs": [{"id": "ref-for-operationerror\u2460\u2461"}], "title": "7.6.4.1. The constant(descriptor, bufferView) method"}, {"refs": [{"id": "ref-for-operationerror\u2460\u2462"}], "title": "7.6.4.2. The constant(value, type) method"}, {"refs": [{"id": "ref-for-operationerror\u2460\u2463"}], "title": "7.6.5. The batchNormalization() method"}, {"refs": [{"id": "ref-for-operationerror\u2460\u2464"}], "title": "7.6.6.1. The clamp(operand, options) method"}, {"refs": [{"id": "ref-for-operationerror\u2460\u2465"}], "title": "7.6.7. The concat() method"}, {"refs": [{"id": "ref-for-operationerror\u2460\u2466"}], "title": "7.6.8. The conv2d() method"}, {"refs": [{"id": "ref-for-operationerror\u2460\u2467"}], "title": "7.6.9. The convTranspose2d() method"}, {"refs": [{"id": "ref-for-operationerror\u2460\u2468"}], "title": "7.6.10. Element-wise binary operations"}, {"refs": [{"id": "ref-for-operationerror\u2461\u24ea"}], "title": "7.6.11. Element-wise unary operations"}, {"refs": [{"id": "ref-for-operationerror\u2461\u2460"}], "title": "7.6.12.1. The elu(input, options) method"}, {"refs": [{"id": "ref-for-operationerror\u2461\u2461"}], "title": "7.6.13. The gemm() method"}, {"refs": [{"id": "ref-for-operationerror\u2461\u2462"}], "title": "7.6.14. The gru() method"}, {"refs": [{"id": "ref-for-operationerror\u2461\u2463"}], "title": "7.6.15. The gruCell() method"}, {"refs": [{"id": "ref-for-operationerror\u2461\u2464"}], "title": "7.6.16.1. The hardSigmoid(input, options) method"}, {"refs": [{"id": "ref-for-operationerror\u2461\u2465"}], "title": "7.6.17.1. The hardSwish(input) method"}, {"refs": [{"id": "ref-for-operationerror\u2461\u2466"}], "title": "7.6.18. The instanceNormalization() method"}, {"refs": [{"id": "ref-for-operationerror\u2461\u2467"}], "title": "7.6.19.1. The leakyRelu(input, options) method"}, {"refs": [{"id": "ref-for-operationerror\u2461\u2468"}], "title": "7.6.20.1. The linear(input, options) method"}, {"refs": [{"id": "ref-for-operationerror\u2462\u24ea"}], "title": "7.6.21. The lstm() method"}, {"refs": [{"id": "ref-for-operationerror\u2462\u2460"}], "title": "7.6.22. The lstmCell() method"}, {"refs": [{"id": "ref-for-operationerror\u2462\u2461"}, {"id": "ref-for-operationerror\u2462\u2462"}, {"id": "ref-for-operationerror\u2462\u2463"}], "title": "7.6.23. The matmul() method"}, {"refs": [{"id": "ref-for-operationerror\u2462\u2464"}], "title": "7.6.24. The pad() method"}, {"refs": [{"id": "ref-for-operationerror\u2462\u2465"}], "title": "7.6.25. Pooling operations"}, {"refs": [{"id": "ref-for-operationerror\u2462\u2466"}], "title": "7.6.26. The prelu() method"}, {"refs": [{"id": "ref-for-operationerror\u2462\u2467"}], "title": "7.6.27. Reduction operations"}, {"refs": [{"id": "ref-for-operationerror\u2462\u2468"}], "title": "7.6.28.1. The relu(input) method"}, {"refs": [{"id": "ref-for-operationerror\u2463\u24ea"}], "title": "7.6.29. The resample2d() method"}, {"refs": [{"id": "ref-for-operationerror\u2463\u2460"}], "title": "7.6.30. The reshape() method"}, {"refs": [{"id": "ref-for-operationerror\u2463\u2461"}], "title": "7.6.31.1. The sigmoid(input) method"}, {"refs": [{"id": "ref-for-operationerror\u2463\u2462"}], "title": "7.6.32. The slice() method"}, {"refs": [{"id": "ref-for-operationerror\u2463\u2463"}], "title": "7.6.33.1. The softmax(input) method"}, {"refs": [{"id": "ref-for-operationerror\u2463\u2464"}], "title": "7.6.34.1. The softplus(input, options) method"}, {"refs": [{"id": "ref-for-operationerror\u2463\u2465"}], "title": "7.6.35.1. The softsign(input) method"}, {"refs": [{"id": "ref-for-operationerror\u2463\u2466"}], "title": "7.6.36. The split() method"}, {"refs": [{"id": "ref-for-operationerror\u2463\u2467"}], "title": "7.6.37. The squeeze() method"}, {"refs": [{"id": "ref-for-operationerror\u2463\u2468"}], "title": "7.6.38.1. The tanh(input) method"}, {"refs": [{"id": "ref-for-operationerror\u2464\u24ea"}], "title": "7.6.39. The transpose() method"}], "external": true};
 window.dfnpanelData['bdbd19d1'] = {"dfnID": "bdbd19d1", "url": "https://webidl.spec.whatwg.org/#idl-promise", "dfnText": "Promise", "refSections": [{"refs": [{"id": "ref-for-idl-promise"}], "title": "6.1. Overview"}, {"refs": [{"id": "ref-for-idl-promise\u2460"}, {"id": "ref-for-idl-promise\u2461"}], "title": "7.2. The ML interface"}, {"refs": [{"id": "ref-for-idl-promise\u2462"}], "title": "7.4.4. Asynchronous Execution"}, {"refs": [{"id": "ref-for-idl-promise\u2463"}], "title": "7.6. The MLGraphBuilder interface"}], "external": true};
 window.dfnpanelData['a5c91173'] = {"dfnID": "a5c91173", "url": "https://webidl.spec.whatwg.org/#SameObject", "dfnText": "SameObject", "refSections": [{"refs": [{"id": "ref-for-SameObject"}], "title": "7.1. The navigator.ml interface"}], "external": true};
 window.dfnpanelData['b75bb3bd'] = {"dfnID": "b75bb3bd", "url": "https://webidl.spec.whatwg.org/#SecureContext", "dfnText": "SecureContext", "refSections": [{"refs": [{"id": "ref-for-SecureContext"}], "title": "7.1. The navigator.ml interface"}, {"refs": [{"id": "ref-for-SecureContext\u2460"}], "title": "7.2. The ML interface"}, {"refs": [{"id": "ref-for-SecureContext\u2461"}], "title": "7.3. The MLGraph interface"}, {"refs": [{"id": "ref-for-SecureContext\u2462"}], "title": "7.3.2. The MLOperand interface"}, {"refs": [{"id": "ref-for-SecureContext\u2463"}], "title": "7.3.3. The MLActivation interface"}, {"refs": [{"id": "ref-for-SecureContext\u2464"}], "title": "7.4. The MLContext interface"}, {"refs": [{"id": "ref-for-SecureContext\u2465"}], "title": "7.5. The MLCommandEncoder interface"}, {"refs": [{"id": "ref-for-SecureContext\u2466"}], "title": "7.6. The MLGraphBuilder interface"}], "external": true};
-window.dfnpanelData['c3e881ef'] = {"dfnID": "c3e881ef", "url": "https://webidl.spec.whatwg.org/#securityerror", "dfnText": "SecurityError", "refSections": [{"refs": [{"id": "ref-for-securityerror"}], "title": "7.2.2. The createContext() method"}, {"refs": [{"id": "ref-for-securityerror\u2460"}], "title": "7.2.3. The createContextSync() method"}, {"refs": [{"id": "ref-for-securityerror\u2461"}], "title": "7.6.1. The MLGraphBuilder constructor"}], "external": true};
-window.dfnpanelData['82ca3efc'] = {"dfnID": "82ca3efc", "url": "https://webidl.spec.whatwg.org/#exceptiondef-typeerror", "dfnText": "TypeError", "refSections": [{"refs": [{"id": "ref-for-exceptiondef-typeerror"}, {"id": "ref-for-exceptiondef-typeerror\u2460"}], "title": "7.3.2.1. Creating MLOperand"}, {"refs": [{"id": "ref-for-exceptiondef-typeerror\u2461"}], "title": "7.6.1. The MLGraphBuilder constructor"}, {"refs": [{"id": "ref-for-exceptiondef-typeerror\u2462"}, {"id": "ref-for-exceptiondef-typeerror\u2463"}], "title": "7.6.2. The input() method"}, {"refs": [{"id": "ref-for-exceptiondef-typeerror\u2464"}, {"id": "ref-for-exceptiondef-typeerror\u2465"}], "title": "7.6.3.1. The constant(descriptor, bufferView) method"}, {"refs": [{"id": "ref-for-exceptiondef-typeerror\u2466"}, {"id": "ref-for-exceptiondef-typeerror\u2467"}], "title": "7.6.3.2. The constant(value, type) method"}], "external": true};
+window.dfnpanelData['c3e881ef'] = {"dfnID": "c3e881ef", "url": "https://webidl.spec.whatwg.org/#securityerror", "dfnText": "SecurityError", "refSections": [{"refs": [{"id": "ref-for-securityerror"}, {"id": "ref-for-securityerror\u2460"}], "title": "7.2.2. The createContext() method"}, {"refs": [{"id": "ref-for-securityerror\u2461"}, {"id": "ref-for-securityerror\u2462"}], "title": "7.2.3. The createContextSync() method"}, {"refs": [{"id": "ref-for-securityerror\u2463"}], "title": "7.6.1. The MLGraphBuilder constructor"}], "external": true};
+window.dfnpanelData['82ca3efc'] = {"dfnID": "82ca3efc", "url": "https://webidl.spec.whatwg.org/#exceptiondef-typeerror", "dfnText": "TypeError", "refSections": [{"refs": [{"id": "ref-for-exceptiondef-typeerror"}], "title": "7.3.3.1. Creating MLActivation"}, {"refs": [{"id": "ref-for-exceptiondef-typeerror\u2460"}], "title": "7.6.1. The MLGraphBuilder constructor"}, {"refs": [{"id": "ref-for-exceptiondef-typeerror\u2461"}], "title": "7.6.2. The input() method"}, {"refs": [{"id": "ref-for-exceptiondef-typeerror\u2462"}, {"id": "ref-for-exceptiondef-typeerror\u2463"}, {"id": "ref-for-exceptiondef-typeerror\u2464"}, {"id": "ref-for-exceptiondef-typeerror\u2465"}], "title": "7.6.3.2. The buildSync(outputs) method"}, {"refs": [{"id": "ref-for-exceptiondef-typeerror\u2466"}], "title": "7.6.4.1. The constant(descriptor, bufferView) method"}, {"refs": [{"id": "ref-for-exceptiondef-typeerror\u2467"}, {"id": "ref-for-exceptiondef-typeerror\u2468"}, {"id": "ref-for-exceptiondef-typeerror\u2460\u24ea"}, {"id": "ref-for-exceptiondef-typeerror\u2460\u2460"}, {"id": "ref-for-exceptiondef-typeerror\u2460\u2461"}, {"id": "ref-for-exceptiondef-typeerror\u2460\u2462"}, {"id": "ref-for-exceptiondef-typeerror\u2460\u2463"}, {"id": "ref-for-exceptiondef-typeerror\u2460\u2464"}, {"id": "ref-for-exceptiondef-typeerror\u2460\u2465"}], "title": "7.6.5. The batchNormalization() method"}, {"refs": [{"id": "ref-for-exceptiondef-typeerror\u2460\u2466"}], "title": "7.6.6.1. The clamp(operand, options) method"}, {"refs": [{"id": "ref-for-exceptiondef-typeerror\u2460\u2467"}], "title": "7.6.6.2. The clamp(options) method"}, {"refs": [{"id": "ref-for-exceptiondef-typeerror\u2460\u2468"}, {"id": "ref-for-exceptiondef-typeerror\u2461\u24ea"}, {"id": "ref-for-exceptiondef-typeerror\u2461\u2460"}, {"id": "ref-for-exceptiondef-typeerror\u2461\u2461"}, {"id": "ref-for-exceptiondef-typeerror\u2461\u2462"}, {"id": "ref-for-exceptiondef-typeerror\u2461\u2463"}], "title": "7.6.8. The conv2d() method"}, {"refs": [{"id": "ref-for-exceptiondef-typeerror\u2461\u2464"}, {"id": "ref-for-exceptiondef-typeerror\u2461\u2465"}, {"id": "ref-for-exceptiondef-typeerror\u2461\u2466"}, {"id": "ref-for-exceptiondef-typeerror\u2461\u2467"}, {"id": "ref-for-exceptiondef-typeerror\u2461\u2468"}, {"id": "ref-for-exceptiondef-typeerror\u2462\u24ea"}, {"id": "ref-for-exceptiondef-typeerror\u2462\u2460"}, {"id": "ref-for-exceptiondef-typeerror\u2462\u2461"}], "title": "7.6.9. The convTranspose2d() method"}, {"refs": [{"id": "ref-for-exceptiondef-typeerror\u2462\u2462"}, {"id": "ref-for-exceptiondef-typeerror\u2462\u2463"}], "title": "7.6.14. The gru() method"}, {"refs": [{"id": "ref-for-exceptiondef-typeerror\u2462\u2464"}], "title": "7.6.15. The gruCell() method"}, {"refs": [{"id": "ref-for-exceptiondef-typeerror\u2462\u2465"}], "title": "7.6.21. The lstm() method"}, {"refs": [{"id": "ref-for-exceptiondef-typeerror\u2462\u2466"}], "title": "7.6.22. The lstmCell() method"}, {"refs": [{"id": "ref-for-exceptiondef-typeerror\u2462\u2467"}], "title": "7.6.24. The pad() method"}, {"refs": [{"id": "ref-for-exceptiondef-typeerror\u2462\u2468"}], "title": "7.6.25. Pooling operations"}, {"refs": [{"id": "ref-for-exceptiondef-typeerror\u2463\u24ea"}, {"id": "ref-for-exceptiondef-typeerror\u2463\u2460"}], "title": "7.6.32. The slice() method"}, {"refs": [{"id": "ref-for-exceptiondef-typeerror\u2463\u2461"}, {"id": "ref-for-exceptiondef-typeerror\u2463\u2462"}], "title": "7.6.36. The split() method"}, {"refs": [{"id": "ref-for-exceptiondef-typeerror\u2463\u2463"}], "title": "7.6.37. The squeeze() method"}, {"refs": [{"id": "ref-for-exceptiondef-typeerror\u2463\u2464"}, {"id": "ref-for-exceptiondef-typeerror\u2463\u2465"}, {"id": "ref-for-exceptiondef-typeerror\u2463\u2466"}], "title": "7.6.39. The transpose() method"}], "external": true};
 window.dfnpanelData['49f6acb5'] = {"dfnID": "49f6acb5", "url": "https://webidl.spec.whatwg.org/#idl-Uint16Array", "dfnText": "Uint16Array", "refSections": [{"refs": [{"id": "ref-for-idl-Uint16Array"}], "title": "9.1. MLOperandType and ArrayBufferView compatibility"}], "external": true};
 window.dfnpanelData['13d0450b'] = {"dfnID": "13d0450b", "url": "https://webidl.spec.whatwg.org/#idl-Uint32Array", "dfnText": "Uint32Array", "refSections": [{"refs": [{"id": "ref-for-idl-Uint32Array"}], "title": "9.1. MLOperandType and ArrayBufferView compatibility"}], "external": true};
 window.dfnpanelData['95d7775a'] = {"dfnID": "95d7775a", "url": "https://webidl.spec.whatwg.org/#idl-Uint8Array", "dfnText": "Uint8Array", "refSections": [{"refs": [{"id": "ref-for-idl-Uint8Array"}], "title": "9.1. MLOperandType and ArrayBufferView compatibility"}], "external": true};
-window.dfnpanelData['dacde8b5'] = {"dfnID": "dacde8b5", "url": "https://webidl.spec.whatwg.org/#a-new-promise", "dfnText": "a new promise", "refSections": [{"refs": [{"id": "ref-for-a-new-promise"}, {"id": "ref-for-a-new-promise\u2460"}], "title": "7.2.2. The createContext() method"}, {"refs": [{"id": "ref-for-a-new-promise\u2461"}], "title": "7.4.4. Asynchronous Execution"}], "external": true};
-window.dfnpanelData['5372cca8'] = {"dfnID": "5372cca8", "url": "https://webidl.spec.whatwg.org/#idl-boolean", "dfnText": "boolean", "refSections": [{"refs": [{"id": "ref-for-idl-boolean"}, {"id": "ref-for-idl-boolean\u2460"}, {"id": "ref-for-idl-boolean\u2461"}, {"id": "ref-for-idl-boolean\u2462"}], "title": "7.6.12. The gemm() method"}, {"refs": [{"id": "ref-for-idl-boolean\u2463"}, {"id": "ref-for-idl-boolean\u2464"}, {"id": "ref-for-idl-boolean\u2465"}, {"id": "ref-for-idl-boolean\u2466"}], "title": "7.6.13. The gru() method"}, {"refs": [{"id": "ref-for-idl-boolean\u2467"}, {"id": "ref-for-idl-boolean\u2468"}], "title": "7.6.14. The gruCell() method"}, {"refs": [{"id": "ref-for-idl-boolean\u2460\u24ea"}, {"id": "ref-for-idl-boolean\u2460\u2460"}], "title": "7.6.20. The lstm() method"}, {"refs": [{"id": "ref-for-idl-boolean\u2460\u2461"}, {"id": "ref-for-idl-boolean\u2460\u2462"}], "title": "7.6.26. Reduction operations"}], "external": true};
-window.dfnpanelData['e9acf661'] = {"dfnID": "e9acf661", "url": "https://webidl.spec.whatwg.org/#buffersource-byte-length", "dfnText": "buffer byte length", "refSections": [{"refs": [{"id": "ref-for-buffersource-byte-length"}], "title": "7.4.3. The MLNamedArrayBufferViews transfer algorithm"}], "external": true};
+window.dfnpanelData['dacde8b5'] = {"dfnID": "dacde8b5", "url": "https://webidl.spec.whatwg.org/#a-new-promise", "dfnText": "a new promise", "refSections": [{"refs": [{"id": "ref-for-a-new-promise"}, {"id": "ref-for-a-new-promise\u2460"}, {"id": "ref-for-a-new-promise\u2461"}, {"id": "ref-for-a-new-promise\u2462"}], "title": "7.2.2. The createContext() method"}, {"refs": [{"id": "ref-for-a-new-promise\u2463"}], "title": "7.4.4. Asynchronous Execution"}, {"refs": [{"id": "ref-for-a-new-promise\u2464"}], "title": "7.6.3.1. The build(outputs) method"}], "external": true};
+window.dfnpanelData['5372cca8'] = {"dfnID": "5372cca8", "url": "https://webidl.spec.whatwg.org/#idl-boolean", "dfnText": "boolean", "refSections": [{"refs": [{"id": "ref-for-idl-boolean"}, {"id": "ref-for-idl-boolean\u2460"}, {"id": "ref-for-idl-boolean\u2461"}, {"id": "ref-for-idl-boolean\u2462"}, {"id": "ref-for-idl-boolean\u2463"}, {"id": "ref-for-idl-boolean\u2464"}], "title": "7.6.13. The gemm() method"}, {"refs": [{"id": "ref-for-idl-boolean\u2465"}, {"id": "ref-for-idl-boolean\u2466"}, {"id": "ref-for-idl-boolean\u2467"}, {"id": "ref-for-idl-boolean\u2468"}, {"id": "ref-for-idl-boolean\u2460\u24ea"}, {"id": "ref-for-idl-boolean\u2460\u2460"}], "title": "7.6.14. The gru() method"}, {"refs": [{"id": "ref-for-idl-boolean\u2460\u2461"}, {"id": "ref-for-idl-boolean\u2460\u2462"}, {"id": "ref-for-idl-boolean\u2460\u2463"}], "title": "7.6.15. The gruCell() method"}, {"refs": [{"id": "ref-for-idl-boolean\u2460\u2464"}, {"id": "ref-for-idl-boolean\u2460\u2465"}, {"id": "ref-for-idl-boolean\u2460\u2466"}], "title": "7.6.21. The lstm() method"}, {"refs": [{"id": "ref-for-idl-boolean\u2460\u2467"}, {"id": "ref-for-idl-boolean\u2460\u2468"}], "title": "7.6.27. Reduction operations"}], "external": true};
+window.dfnpanelData['495737df'] = {"dfnID": "495737df", "url": "https://webidl.spec.whatwg.org/#buffersource-byte-length", "dfnText": "byte length", "refSections": [{"refs": [{"id": "ref-for-buffersource-byte-length"}], "title": "7.4.3. The MLNamedArrayBufferViews transfer algorithm"}], "external": true};
 window.dfnpanelData['8c800cdf'] = {"dfnID": "8c800cdf", "url": "https://webidl.spec.whatwg.org/#idl-double", "dfnText": "double", "refSections": [{"refs": [{"id": "ref-for-idl-double"}], "title": "7.6. The MLGraphBuilder interface"}], "external": true};
-window.dfnpanelData['eec1be2d'] = {"dfnID": "eec1be2d", "url": "https://webidl.spec.whatwg.org/#idl-float", "dfnText": "float", "refSections": [{"refs": [{"id": "ref-for-idl-float"}, {"id": "ref-for-idl-float\u2460"}], "title": "7.6.4. The batchNormalization() method"}, {"refs": [{"id": "ref-for-idl-float\u2461"}, {"id": "ref-for-idl-float\u2462"}, {"id": "ref-for-idl-float\u2463"}, {"id": "ref-for-idl-float\u2464"}], "title": "7.6.5. The clamp() method"}, {"refs": [{"id": "ref-for-idl-float\u2465"}, {"id": "ref-for-idl-float\u2466"}], "title": "7.6.11. The elu() method"}, {"refs": [{"id": "ref-for-idl-float\u2467"}, {"id": "ref-for-idl-float\u2468"}, {"id": "ref-for-idl-float\u2460\u24ea"}, {"id": "ref-for-idl-float\u2460\u2460"}], "title": "7.6.12. The gemm() method"}, {"refs": [{"id": "ref-for-idl-float\u2460\u2461"}, {"id": "ref-for-idl-float\u2460\u2462"}, {"id": "ref-for-idl-float\u2460\u2463"}, {"id": "ref-for-idl-float\u2460\u2464"}], "title": "7.6.15. The hardSigmoid() method"}, {"refs": [{"id": "ref-for-idl-float\u2460\u2465"}, {"id": "ref-for-idl-float\u2460\u2466"}], "title": "7.6.17. The instanceNormalization() method"}, {"refs": [{"id": "ref-for-idl-float\u2460\u2467"}, {"id": "ref-for-idl-float\u2460\u2468"}], "title": "7.6.18. The leakyRelu() method"}, {"refs": [{"id": "ref-for-idl-float\u2461\u24ea"}, {"id": "ref-for-idl-float\u2461\u2460"}, {"id": "ref-for-idl-float\u2461\u2461"}, {"id": "ref-for-idl-float\u2461\u2462"}], "title": "7.6.19. The linear() method"}, {"refs": [{"id": "ref-for-idl-float\u2461\u2463"}, {"id": "ref-for-idl-float\u2461\u2464"}], "title": "7.6.23. The pad() method"}, {"refs": [{"id": "ref-for-idl-float\u2461\u2465"}, {"id": "ref-for-idl-float\u2461\u2466"}], "title": "7.6.28. The resample2d() method"}, {"refs": [{"id": "ref-for-idl-float\u2461\u2467"}], "title": "7.6.33. The softplus() method"}], "external": true};
-window.dfnpanelData['7b05fe24'] = {"dfnID": "7b05fe24", "url": "https://webidl.spec.whatwg.org/#dfn-get-buffer-source-copy", "dfnText": "get a copy of the bytes held by the buffer source", "refSections": [{"refs": [{"id": "ref-for-dfn-get-buffer-source-copy"}], "title": "7.6.3.1. The constant(descriptor, bufferView) method"}], "external": true};
-window.dfnpanelData['a32c65d4'] = {"dfnID": "a32c65d4", "url": "https://webidl.spec.whatwg.org/#implements", "dfnText": "implements", "refSections": [{"refs": [{"id": "ref-for-implements"}], "title": "7.3.2.1. Creating MLOperand"}, {"refs": [{"id": "ref-for-implements\u2460"}], "title": "7.6.2. The input() method"}, {"refs": [{"id": "ref-for-implements\u2461"}], "title": "7.6.3.1. The constant(descriptor, bufferView) method"}], "external": true};
+window.dfnpanelData['eec1be2d'] = {"dfnID": "eec1be2d", "url": "https://webidl.spec.whatwg.org/#idl-float", "dfnText": "float", "refSections": [{"refs": [{"id": "ref-for-idl-float"}, {"id": "ref-for-idl-float\u2460"}, {"id": "ref-for-idl-float\u2461"}], "title": "7.6.5. The batchNormalization() method"}, {"refs": [{"id": "ref-for-idl-float\u2462"}, {"id": "ref-for-idl-float\u2463"}], "title": "7.6.6. The clamp() method"}, {"refs": [{"id": "ref-for-idl-float\u2464"}, {"id": "ref-for-idl-float\u2465"}], "title": "7.6.6.1. The clamp(operand, options) method"}, {"refs": [{"id": "ref-for-idl-float\u2466"}, {"id": "ref-for-idl-float\u2467"}], "title": "7.6.6.2. The clamp(options) method"}, {"refs": [{"id": "ref-for-idl-float\u2468"}], "title": "7.6.12. The elu() method"}, {"refs": [{"id": "ref-for-idl-float\u2460\u24ea"}], "title": "7.6.12.1. The elu(input, options) method"}, {"refs": [{"id": "ref-for-idl-float\u2460\u2460"}], "title": "7.6.12.2. The elu(options) method"}, {"refs": [{"id": "ref-for-idl-float\u2460\u2461"}, {"id": "ref-for-idl-float\u2460\u2462"}, {"id": "ref-for-idl-float\u2460\u2463"}, {"id": "ref-for-idl-float\u2460\u2464"}, {"id": "ref-for-idl-float\u2460\u2465"}, {"id": "ref-for-idl-float\u2460\u2466"}], "title": "7.6.13. The gemm() method"}, {"refs": [{"id": "ref-for-idl-float\u2460\u2467"}, {"id": "ref-for-idl-float\u2460\u2468"}, {"id": "ref-for-idl-float\u2461\u24ea"}, {"id": "ref-for-idl-float\u2461\u2460"}, {"id": "ref-for-idl-float\u2461\u2461"}, {"id": "ref-for-idl-float\u2461\u2462"}], "title": "7.6.16. The hardSigmoid() method"}, {"refs": [{"id": "ref-for-idl-float\u2461\u2463"}, {"id": "ref-for-idl-float\u2461\u2464"}, {"id": "ref-for-idl-float\u2461\u2465"}], "title": "7.6.18. The instanceNormalization() method"}, {"refs": [{"id": "ref-for-idl-float\u2461\u2466"}, {"id": "ref-for-idl-float\u2461\u2467"}, {"id": "ref-for-idl-float\u2461\u2468"}], "title": "7.6.19. The leakyRelu() method"}, {"refs": [{"id": "ref-for-idl-float\u2462\u24ea"}, {"id": "ref-for-idl-float\u2462\u2460"}, {"id": "ref-for-idl-float\u2462\u2461"}, {"id": "ref-for-idl-float\u2462\u2462"}, {"id": "ref-for-idl-float\u2462\u2463"}, {"id": "ref-for-idl-float\u2462\u2464"}], "title": "7.6.20. The linear() method"}, {"refs": [{"id": "ref-for-idl-float\u2462\u2465"}, {"id": "ref-for-idl-float\u2462\u2466"}, {"id": "ref-for-idl-float\u2462\u2467"}], "title": "7.6.24. The pad() method"}, {"refs": [{"id": "ref-for-idl-float\u2462\u2468"}, {"id": "ref-for-idl-float\u2463\u24ea"}, {"id": "ref-for-idl-float\u2463\u2460"}], "title": "7.6.29. The resample2d() method"}, {"refs": [{"id": "ref-for-idl-float\u2463\u2461"}, {"id": "ref-for-idl-float\u2463\u2462"}, {"id": "ref-for-idl-float\u2463\u2463"}], "title": "7.6.34. The softplus() method"}], "external": true};
+window.dfnpanelData['7b05fe24'] = {"dfnID": "7b05fe24", "url": "https://webidl.spec.whatwg.org/#dfn-get-buffer-source-copy", "dfnText": "get a copy of the bytes held by the buffer source", "refSections": [{"refs": [{"id": "ref-for-dfn-get-buffer-source-copy"}], "title": "7.6.4.1. The constant(descriptor, bufferView) method"}], "external": true};
+window.dfnpanelData['f8de33a3'] = {"dfnID": "f8de33a3", "url": "https://webidl.spec.whatwg.org/#idl-long", "dfnText": "long", "refSections": [{"refs": [{"id": "ref-for-idl-long"}], "title": "7.6.5. The batchNormalization() method"}], "external": true};
 window.dfnpanelData['56f81a8e'] = {"dfnID": "56f81a8e", "url": "https://webidl.spec.whatwg.org/#new", "dfnText": "new", "refSections": [{"refs": [{"id": "ref-for-new"}], "title": "7.6.1. The MLGraphBuilder constructor"}], "external": true};
-window.dfnpanelData['aedfadcb'] = {"dfnID": "aedfadcb", "url": "https://webidl.spec.whatwg.org/#idl-nullable-type", "dfnText": "nullable", "refSections": [{"refs": [{"id": "ref-for-idl-nullable-type"}], "title": "7.6.29. The reshape() method"}], "external": true};
+window.dfnpanelData['51978e94'] = {"dfnID": "51978e94", "url": "https://webidl.spec.whatwg.org/#dfn-nullable-type", "dfnText": "nullable type", "refSections": [{"refs": [{"id": "ref-for-dfn-nullable-type"}], "title": "7.6.30. The reshape() method"}], "external": true};
 window.dfnpanelData['7787d359'] = {"dfnID": "7787d359", "url": "https://webidl.spec.whatwg.org/#idl-record", "dfnText": "record", "refSections": [{"refs": [{"id": "ref-for-idl-record"}, {"id": "ref-for-idl-record\u2460"}], "title": "7.3. The MLGraph interface"}, {"refs": [{"id": "ref-for-idl-record\u2461"}], "title": "7.4. The MLContext interface"}, {"refs": [{"id": "ref-for-idl-record\u2462"}], "title": "7.4.2. Synchronous Execution"}, {"refs": [{"id": "ref-for-idl-record\u2463"}], "title": "7.5. The MLCommandEncoder interface"}, {"refs": [{"id": "ref-for-idl-record\u2464"}], "title": "7.6. The MLGraphBuilder interface"}], "external": true};
-window.dfnpanelData['b262501e'] = {"dfnID": "b262501e", "url": "https://webidl.spec.whatwg.org/#reject", "dfnText": "reject", "refSections": [{"refs": [{"id": "ref-for-reject"}, {"id": "ref-for-reject\u2460"}], "title": "7.2.2. The createContext() method"}, {"refs": [{"id": "ref-for-reject\u2461"}, {"id": "ref-for-reject\u2462"}, {"id": "ref-for-reject\u2463"}, {"id": "ref-for-reject\u2464"}], "title": "7.4.4. Asynchronous Execution"}], "external": true};
-window.dfnpanelData['3b90bdcd'] = {"dfnID": "3b90bdcd", "url": "https://webidl.spec.whatwg.org/#resolve", "dfnText": "resolve", "refSections": [{"refs": [{"id": "ref-for-resolve"}], "title": "7.2.2. The createContext() method"}, {"refs": [{"id": "ref-for-resolve\u2460"}], "title": "7.4.4. Asynchronous Execution"}], "external": true};
-window.dfnpanelData['9cce47fd'] = {"dfnID": "9cce47fd", "url": "https://webidl.spec.whatwg.org/#idl-sequence", "dfnText": "sequence", "refSections": [{"refs": [{"id": "ref-for-idl-sequence"}], "title": "7.3.1. The MLOperandDescriptor dictionary"}, {"refs": [{"id": "ref-for-idl-sequence\u2460"}], "title": "7.6.6. The concat() method"}, {"refs": [{"id": "ref-for-idl-sequence\u2461"}, {"id": "ref-for-idl-sequence\u2462"}, {"id": "ref-for-idl-sequence\u2463"}], "title": "7.6.7. The conv2d() method"}, {"refs": [{"id": "ref-for-idl-sequence\u2464"}, {"id": "ref-for-idl-sequence\u2465"}, {"id": "ref-for-idl-sequence\u2466"}, {"id": "ref-for-idl-sequence\u2467"}, {"id": "ref-for-idl-sequence\u2468"}], "title": "7.6.8. The convTranspose2d() method"}, {"refs": [{"id": "ref-for-idl-sequence\u2460\u24ea"}, {"id": "ref-for-idl-sequence\u2460\u2460"}], "title": "7.6.13. The gru() method"}, {"refs": [{"id": "ref-for-idl-sequence\u2460\u2461"}], "title": "7.6.14. The gruCell() method"}, {"refs": [{"id": "ref-for-idl-sequence\u2460\u2462"}, {"id": "ref-for-idl-sequence\u2460\u2463"}], "title": "7.6.20. The lstm() method"}, {"refs": [{"id": "ref-for-idl-sequence\u2460\u2464"}, {"id": "ref-for-idl-sequence\u2460\u2465"}], "title": "7.6.21. The lstmCell() method"}, {"refs": [{"id": "ref-for-idl-sequence\u2460\u2466"}, {"id": "ref-for-idl-sequence\u2460\u2467"}], "title": "7.6.23. The pad() method"}, {"refs": [{"id": "ref-for-idl-sequence\u2460\u2468"}, {"id": "ref-for-idl-sequence\u2461\u24ea"}, {"id": "ref-for-idl-sequence\u2461\u2460"}, {"id": "ref-for-idl-sequence\u2461\u2461"}, {"id": "ref-for-idl-sequence\u2461\u2462"}], "title": "7.6.24. Pooling operations"}, {"refs": [{"id": "ref-for-idl-sequence\u2461\u2463"}], "title": "7.6.26. Reduction operations"}, {"refs": [{"id": "ref-for-idl-sequence\u2461\u2464"}, {"id": "ref-for-idl-sequence\u2461\u2465"}, {"id": "ref-for-idl-sequence\u2461\u2466"}], "title": "7.6.28. The resample2d() method"}, {"refs": [{"id": "ref-for-idl-sequence\u2461\u2467"}], "title": "7.6.29. The reshape() method"}, {"refs": [{"id": "ref-for-idl-sequence\u2461\u2468"}, {"id": "ref-for-idl-sequence\u2462\u24ea"}], "title": "7.6.31. The slice() method"}, {"refs": [{"id": "ref-for-idl-sequence\u2462\u2460"}, {"id": "ref-for-idl-sequence\u2462\u2461"}], "title": "7.6.35. The split() method"}, {"refs": [{"id": "ref-for-idl-sequence\u2462\u2462"}], "title": "7.6.36. The squeeze() method"}, {"refs": [{"id": "ref-for-idl-sequence\u2462\u2463"}], "title": "7.6.38. The transpose() method"}], "external": true};
-window.dfnpanelData['4013a022'] = {"dfnID": "4013a022", "url": "https://webidl.spec.whatwg.org/#this", "dfnText": "this", "refSections": [{"refs": [{"id": "ref-for-this"}], "title": "7.2.2. The createContext() method"}, {"refs": [{"id": "ref-for-this\u2460"}], "title": "7.2.3. The createContextSync() method"}, {"refs": [{"id": "ref-for-this\u2461"}], "title": "7.6.1. The MLGraphBuilder constructor"}, {"refs": [{"id": "ref-for-this\u2462"}], "title": "7.6.2. The input() method"}, {"refs": [{"id": "ref-for-this\u2463"}], "title": "7.6.3.1. The constant(descriptor, bufferView) method"}, {"refs": [{"id": "ref-for-this\u2464"}], "title": "7.6.3.2. The constant(value, type) method"}], "external": true};
+window.dfnpanelData['b262501e'] = {"dfnID": "b262501e", "url": "https://webidl.spec.whatwg.org/#reject", "dfnText": "reject", "refSections": [{"refs": [{"id": "ref-for-reject"}, {"id": "ref-for-reject\u2460"}, {"id": "ref-for-reject\u2461"}, {"id": "ref-for-reject\u2462"}], "title": "7.2.2. The createContext() method"}, {"refs": [{"id": "ref-for-reject\u2463"}, {"id": "ref-for-reject\u2464"}, {"id": "ref-for-reject\u2465"}, {"id": "ref-for-reject\u2466"}], "title": "7.4.4. Asynchronous Execution"}], "external": true};
+window.dfnpanelData['3b90bdcd'] = {"dfnID": "3b90bdcd", "url": "https://webidl.spec.whatwg.org/#resolve", "dfnText": "resolve", "refSections": [{"refs": [{"id": "ref-for-resolve"}, {"id": "ref-for-resolve\u2460"}], "title": "7.2.2. The createContext() method"}, {"refs": [{"id": "ref-for-resolve\u2461"}], "title": "7.4.4. Asynchronous Execution"}], "external": true};
+window.dfnpanelData['9cce47fd'] = {"dfnID": "9cce47fd", "url": "https://webidl.spec.whatwg.org/#idl-sequence", "dfnText": "sequence", "refSections": [{"refs": [{"id": "ref-for-idl-sequence"}], "title": "7.3.1. The MLOperandDescriptor dictionary"}, {"refs": [{"id": "ref-for-idl-sequence\u2460"}], "title": "7.6.7. The concat() method"}, {"refs": [{"id": "ref-for-idl-sequence\u2461"}, {"id": "ref-for-idl-sequence\u2462"}, {"id": "ref-for-idl-sequence\u2463"}], "title": "7.6.8. The conv2d() method"}, {"refs": [{"id": "ref-for-idl-sequence\u2464"}, {"id": "ref-for-idl-sequence\u2465"}, {"id": "ref-for-idl-sequence\u2466"}, {"id": "ref-for-idl-sequence\u2467"}, {"id": "ref-for-idl-sequence\u2468"}], "title": "7.6.9. The convTranspose2d() method"}, {"refs": [{"id": "ref-for-idl-sequence\u2460\u24ea"}, {"id": "ref-for-idl-sequence\u2460\u2460"}], "title": "7.6.14. The gru() method"}, {"refs": [{"id": "ref-for-idl-sequence\u2460\u2461"}], "title": "7.6.15. The gruCell() method"}, {"refs": [{"id": "ref-for-idl-sequence\u2460\u2462"}, {"id": "ref-for-idl-sequence\u2460\u2463"}], "title": "7.6.21. The lstm() method"}, {"refs": [{"id": "ref-for-idl-sequence\u2460\u2464"}, {"id": "ref-for-idl-sequence\u2460\u2465"}], "title": "7.6.22. The lstmCell() method"}, {"refs": [{"id": "ref-for-idl-sequence\u2460\u2466"}, {"id": "ref-for-idl-sequence\u2460\u2467"}], "title": "7.6.24. The pad() method"}, {"refs": [{"id": "ref-for-idl-sequence\u2460\u2468"}, {"id": "ref-for-idl-sequence\u2461\u24ea"}, {"id": "ref-for-idl-sequence\u2461\u2460"}, {"id": "ref-for-idl-sequence\u2461\u2461"}, {"id": "ref-for-idl-sequence\u2461\u2462"}], "title": "7.6.25. Pooling operations"}, {"refs": [{"id": "ref-for-idl-sequence\u2461\u2463"}], "title": "7.6.27. Reduction operations"}, {"refs": [{"id": "ref-for-idl-sequence\u2461\u2464"}, {"id": "ref-for-idl-sequence\u2461\u2465"}, {"id": "ref-for-idl-sequence\u2461\u2466"}], "title": "7.6.29. The resample2d() method"}, {"refs": [{"id": "ref-for-idl-sequence\u2461\u2467"}], "title": "7.6.30. The reshape() method"}, {"refs": [{"id": "ref-for-idl-sequence\u2461\u2468"}, {"id": "ref-for-idl-sequence\u2462\u24ea"}], "title": "7.6.32. The slice() method"}, {"refs": [{"id": "ref-for-idl-sequence\u2462\u2460"}, {"id": "ref-for-idl-sequence\u2462\u2461"}], "title": "7.6.36. The split() method"}, {"refs": [{"id": "ref-for-idl-sequence\u2462\u2462"}], "title": "7.6.37. The squeeze() method"}, {"refs": [{"id": "ref-for-idl-sequence\u2462\u2463"}], "title": "7.6.39. The transpose() method"}], "external": true};
+window.dfnpanelData['4013a022'] = {"dfnID": "4013a022", "url": "https://webidl.spec.whatwg.org/#this", "dfnText": "this", "refSections": [{"refs": [{"id": "ref-for-this"}, {"id": "ref-for-this\u2460"}], "title": "7.2.2. The createContext() method"}, {"refs": [{"id": "ref-for-this\u2461"}, {"id": "ref-for-this\u2462"}], "title": "7.2.3. The createContextSync() method"}, {"refs": [{"id": "ref-for-this\u2463"}], "title": "7.6.1. The MLGraphBuilder constructor"}, {"refs": [{"id": "ref-for-this\u2464"}], "title": "7.6.2. The input() method"}, {"refs": [{"id": "ref-for-this\u2465"}, {"id": "ref-for-this\u2466"}], "title": "7.6.3.2. The buildSync(outputs) method"}, {"refs": [{"id": "ref-for-this\u2467"}], "title": "7.6.4.1. The constant(descriptor, bufferView) method"}, {"refs": [{"id": "ref-for-this\u2468"}], "title": "7.6.4.2. The constant(value, type) method"}, {"refs": [{"id": "ref-for-this\u2460\u24ea"}], "title": "7.6.5. The batchNormalization() method"}, {"refs": [{"id": "ref-for-this\u2460\u2460"}], "title": "7.6.6.2. The clamp(options) method"}, {"refs": [{"id": "ref-for-this\u2460\u2461"}, {"id": "ref-for-this\u2460\u2462"}], "title": "7.6.7. The concat() method"}, {"refs": [{"id": "ref-for-this\u2460\u2463"}], "title": "7.6.8. The conv2d() method"}, {"refs": [{"id": "ref-for-this\u2460\u2464"}], "title": "7.6.9. The convTranspose2d() method"}, {"refs": [{"id": "ref-for-this\u2460\u2465"}], "title": "7.6.10. Element-wise binary operations"}, {"refs": [{"id": "ref-for-this\u2460\u2466"}], "title": "7.6.12.2. The elu(options) method"}, {"refs": [{"id": "ref-for-this\u2460\u2467"}], "title": "7.6.13. The gemm() method"}, {"refs": [{"id": "ref-for-this\u2460\u2468"}], "title": "7.6.15. The gruCell() method"}, {"refs": [{"id": "ref-for-this\u2461\u24ea"}], "title": "7.6.16.2. The hardSigmoid(options) method"}, {"refs": [{"id": "ref-for-this\u2461\u2460"}], "title": "7.6.17.2. The hardSwish() method"}, {"refs": [{"id": "ref-for-this\u2461\u2461"}], "title": "7.6.19.2. The leakyRelu(options) method"}, {"refs": [{"id": "ref-for-this\u2461\u2462"}], "title": "7.6.20.2. The linear(options) method"}, {"refs": [{"id": "ref-for-this\u2461\u2463"}, {"id": "ref-for-this\u2461\u2464"}, {"id": "ref-for-this\u2461\u2465"}], "title": "7.6.21. The lstm() method"}, {"refs": [{"id": "ref-for-this\u2461\u2466"}, {"id": "ref-for-this\u2461\u2467"}], "title": "7.6.22. The lstmCell() method"}, {"refs": [{"id": "ref-for-this\u2461\u2468"}], "title": "7.6.23. The matmul() method"}, {"refs": [{"id": "ref-for-this\u2462\u24ea"}], "title": "7.6.24. The pad() method"}, {"refs": [{"id": "ref-for-this\u2462\u2460"}], "title": "7.6.25. Pooling operations"}, {"refs": [{"id": "ref-for-this\u2462\u2461"}], "title": "7.6.26. The prelu() method"}, {"refs": [{"id": "ref-for-this\u2462\u2462"}], "title": "7.6.28.2. The relu() method"}, {"refs": [{"id": "ref-for-this\u2462\u2463"}], "title": "7.6.29. The resample2d() method"}, {"refs": [{"id": "ref-for-this\u2462\u2464"}], "title": "7.6.30. The reshape() method"}, {"refs": [{"id": "ref-for-this\u2462\u2465"}], "title": "7.6.31.2. The sigmoid() method"}, {"refs": [{"id": "ref-for-this\u2462\u2466"}], "title": "7.6.34.2. The softplus(options) method"}, {"refs": [{"id": "ref-for-this\u2462\u2467"}], "title": "7.6.35.2. The softsign() method"}, {"refs": [{"id": "ref-for-this\u2462\u2468"}], "title": "7.6.38.2. The tanh() method"}], "external": true};
+window.dfnpanelData['b4cfa5ce'] = {"dfnID": "b4cfa5ce", "url": "https://webidl.spec.whatwg.org/#dfn-throw", "dfnText": "throw", "refSections": [{"refs": [{"id": "ref-for-dfn-throw"}, {"id": "ref-for-dfn-throw\u2460"}, {"id": "ref-for-dfn-throw\u2461"}, {"id": "ref-for-dfn-throw\u2462"}], "title": "7.2.3. The createContextSync() method"}, {"refs": [{"id": "ref-for-dfn-throw\u2463"}, {"id": "ref-for-dfn-throw\u2464"}], "title": "7.3.3.1. Creating MLActivation"}, {"refs": [{"id": "ref-for-dfn-throw\u2465"}, {"id": "ref-for-dfn-throw\u2466"}, {"id": "ref-for-dfn-throw\u2467"}, {"id": "ref-for-dfn-throw\u2468"}, {"id": "ref-for-dfn-throw\u2460\u24ea"}, {"id": "ref-for-dfn-throw\u2460\u2460"}, {"id": "ref-for-dfn-throw\u2460\u2461"}, {"id": "ref-for-dfn-throw\u2460\u2462"}, {"id": "ref-for-dfn-throw\u2460\u2463"}], "title": "7.4.2. Synchronous Execution"}, {"refs": [{"id": "ref-for-dfn-throw\u2460\u2464"}, {"id": "ref-for-dfn-throw\u2460\u2465"}], "title": "7.4.4. Asynchronous Execution"}, {"refs": [{"id": "ref-for-dfn-throw\u2460\u2466"}], "title": "7.4.5. WebGPU Interoperability"}, {"refs": [{"id": "ref-for-dfn-throw\u2460\u2467"}], "title": "7.5.2. Dispatch Execution Commands"}, {"refs": [{"id": "ref-for-dfn-throw\u2460\u2468"}], "title": "7.5.3. Generate GPU Command Buffer"}, {"refs": [{"id": "ref-for-dfn-throw\u2461\u24ea"}, {"id": "ref-for-dfn-throw\u2461\u2460"}], "title": "7.6.1. The MLGraphBuilder constructor"}, {"refs": [{"id": "ref-for-dfn-throw\u2461\u2461"}, {"id": "ref-for-dfn-throw\u2461\u2462"}, {"id": "ref-for-dfn-throw\u2461\u2463"}, {"id": "ref-for-dfn-throw\u2461\u2464"}], "title": "7.6.2. The input() method"}, {"refs": [{"id": "ref-for-dfn-throw\u2461\u2465"}, {"id": "ref-for-dfn-throw\u2461\u2466"}], "title": "7.6.3.1. The build(outputs) method"}, {"refs": [{"id": "ref-for-dfn-throw\u2461\u2467"}, {"id": "ref-for-dfn-throw\u2461\u2468"}, {"id": "ref-for-dfn-throw\u2462\u24ea"}, {"id": "ref-for-dfn-throw\u2462\u2460"}, {"id": "ref-for-dfn-throw\u2462\u2461"}], "title": "7.6.3.2. The buildSync(outputs) method"}, {"refs": [{"id": "ref-for-dfn-throw\u2462\u2462"}, {"id": "ref-for-dfn-throw\u2462\u2463"}, {"id": "ref-for-dfn-throw\u2462\u2464"}, {"id": "ref-for-dfn-throw\u2462\u2465"}], "title": "7.6.4.1. The constant(descriptor, bufferView) method"}, {"refs": [{"id": "ref-for-dfn-throw\u2462\u2466"}], "title": "7.6.4.2. The constant(value, type) method"}, {"refs": [{"id": "ref-for-dfn-throw\u2462\u2467"}, {"id": "ref-for-dfn-throw\u2462\u2468"}, {"id": "ref-for-dfn-throw\u2463\u24ea"}, {"id": "ref-for-dfn-throw\u2463\u2460"}, {"id": "ref-for-dfn-throw\u2463\u2461"}, {"id": "ref-for-dfn-throw\u2463\u2462"}, {"id": "ref-for-dfn-throw\u2463\u2463"}, {"id": "ref-for-dfn-throw\u2463\u2464"}, {"id": "ref-for-dfn-throw\u2463\u2465"}, {"id": "ref-for-dfn-throw\u2463\u2466"}], "title": "7.6.5. The batchNormalization() method"}, {"refs": [{"id": "ref-for-dfn-throw\u2463\u2467"}, {"id": "ref-for-dfn-throw\u2463\u2468"}], "title": "7.6.6.1. The clamp(operand, options) method"}, {"refs": [{"id": "ref-for-dfn-throw\u2464\u24ea"}, {"id": "ref-for-dfn-throw\u2464\u2460"}, {"id": "ref-for-dfn-throw\u2464\u2461"}], "title": "7.6.6.2. The clamp(options) method"}, {"refs": [{"id": "ref-for-dfn-throw\u2464\u2462"}, {"id": "ref-for-dfn-throw\u2464\u2463"}, {"id": "ref-for-dfn-throw\u2464\u2464"}, {"id": "ref-for-dfn-throw\u2464\u2465"}, {"id": "ref-for-dfn-throw\u2464\u2466"}, {"id": "ref-for-dfn-throw\u2464\u2467"}], "title": "7.6.7. The concat() method"}, {"refs": [{"id": "ref-for-dfn-throw\u2464\u2468"}, {"id": "ref-for-dfn-throw\u2465\u24ea"}, {"id": "ref-for-dfn-throw\u2465\u2460"}, {"id": "ref-for-dfn-throw\u2465\u2461"}, {"id": "ref-for-dfn-throw\u2465\u2462"}, {"id": "ref-for-dfn-throw\u2465\u2463"}, {"id": "ref-for-dfn-throw\u2465\u2464"}, {"id": "ref-for-dfn-throw\u2465\u2465"}, {"id": "ref-for-dfn-throw\u2465\u2466"}, {"id": "ref-for-dfn-throw\u2465\u2467"}, {"id": "ref-for-dfn-throw\u2465\u2468"}, {"id": "ref-for-dfn-throw\u2466\u24ea"}, {"id": "ref-for-dfn-throw\u2466\u2460"}, {"id": "ref-for-dfn-throw\u2466\u2461"}], "title": "7.6.8. The conv2d() method"}, {"refs": [{"id": "ref-for-dfn-throw\u2466\u2462"}, {"id": "ref-for-dfn-throw\u2466\u2463"}, {"id": "ref-for-dfn-throw\u2466\u2464"}, {"id": "ref-for-dfn-throw\u2466\u2465"}, {"id": "ref-for-dfn-throw\u2466\u2466"}, {"id": "ref-for-dfn-throw\u2466\u2467"}, {"id": "ref-for-dfn-throw\u2466\u2468"}, {"id": "ref-for-dfn-throw\u2467\u24ea"}, {"id": "ref-for-dfn-throw\u2467\u2460"}, {"id": "ref-for-dfn-throw\u2467\u2461"}, {"id": "ref-for-dfn-throw\u2467\u2462"}, {"id": "ref-for-dfn-throw\u2467\u2463"}, {"id": "ref-for-dfn-throw\u2467\u2464"}, {"id": "ref-for-dfn-throw\u2467\u2465"}, {"id": "ref-for-dfn-throw\u2467\u2466"}, {"id": "ref-for-dfn-throw\u2467\u2467"}], "title": "7.6.9. The convTranspose2d() method"}, {"refs": [{"id": "ref-for-dfn-throw\u2467\u2468"}, {"id": "ref-for-dfn-throw\u2468\u24ea"}, {"id": "ref-for-dfn-throw\u2468\u2460"}, {"id": "ref-for-dfn-throw\u2468\u2461"}, {"id": "ref-for-dfn-throw\u2468\u2462"}, {"id": "ref-for-dfn-throw\u2468\u2463"}, {"id": "ref-for-dfn-throw\u2468\u2464"}, {"id": "ref-for-dfn-throw\u2468\u2465"}, {"id": "ref-for-dfn-throw\u2468\u2466"}, {"id": "ref-for-dfn-throw\u2468\u2467"}, {"id": "ref-for-dfn-throw\u2468\u2468"}, {"id": "ref-for-dfn-throw\u2460\u24ea\u24ea"}, {"id": "ref-for-dfn-throw\u2460\u24ea\u2460"}, {"id": "ref-for-dfn-throw\u2460\u24ea\u2461"}, {"id": "ref-for-dfn-throw\u2460\u24ea\u2462"}, {"id": "ref-for-dfn-throw\u2460\u24ea\u2463"}, {"id": "ref-for-dfn-throw\u2460\u24ea\u2464"}, {"id": "ref-for-dfn-throw\u2460\u24ea\u2465"}, {"id": "ref-for-dfn-throw\u2460\u24ea\u2466"}], "title": "7.6.10. Element-wise binary operations"}, {"refs": [{"id": "ref-for-dfn-throw\u2460\u24ea\u2467"}, {"id": "ref-for-dfn-throw\u2460\u24ea\u2468"}, {"id": "ref-for-dfn-throw\u2460\u2460\u24ea"}, {"id": "ref-for-dfn-throw\u2460\u2460\u2460"}, {"id": "ref-for-dfn-throw\u2460\u2460\u2461"}, {"id": "ref-for-dfn-throw\u2460\u2460\u2462"}, {"id": "ref-for-dfn-throw\u2460\u2460\u2463"}, {"id": "ref-for-dfn-throw\u2460\u2460\u2464"}, {"id": "ref-for-dfn-throw\u2460\u2460\u2465"}, {"id": "ref-for-dfn-throw\u2460\u2460\u2466"}, {"id": "ref-for-dfn-throw\u2460\u2460\u2467"}, {"id": "ref-for-dfn-throw\u2460\u2460\u2468"}, {"id": "ref-for-dfn-throw\u2460\u2461\u24ea"}, {"id": "ref-for-dfn-throw\u2460\u2461\u2460"}, {"id": "ref-for-dfn-throw\u2460\u2461\u2461"}, {"id": "ref-for-dfn-throw\u2460\u2461\u2462"}, {"id": "ref-for-dfn-throw\u2460\u2461\u2463"}, {"id": "ref-for-dfn-throw\u2460\u2461\u2464"}, {"id": "ref-for-dfn-throw\u2460\u2461\u2465"}], "title": "7.6.11. Element-wise unary operations"}, {"refs": [{"id": "ref-for-dfn-throw\u2460\u2461\u2466"}], "title": "7.6.12.1. The elu(input, options) method"}, {"refs": [{"id": "ref-for-dfn-throw\u2460\u2461\u2467"}, {"id": "ref-for-dfn-throw\u2460\u2461\u2468"}, {"id": "ref-for-dfn-throw\u2460\u2462\u24ea"}, {"id": "ref-for-dfn-throw\u2460\u2462\u2460"}], "title": "7.6.13. The gemm() method"}, {"refs": [{"id": "ref-for-dfn-throw\u2460\u2462\u2461"}, {"id": "ref-for-dfn-throw\u2460\u2462\u2462"}, {"id": "ref-for-dfn-throw\u2460\u2462\u2463"}, {"id": "ref-for-dfn-throw\u2460\u2462\u2464"}, {"id": "ref-for-dfn-throw\u2460\u2462\u2465"}, {"id": "ref-for-dfn-throw\u2460\u2462\u2466"}, {"id": "ref-for-dfn-throw\u2460\u2462\u2467"}], "title": "7.6.14. The gru() method"}, {"refs": [{"id": "ref-for-dfn-throw\u2460\u2462\u2468"}, {"id": "ref-for-dfn-throw\u2460\u2463\u24ea"}, {"id": "ref-for-dfn-throw\u2460\u2463\u2460"}, {"id": "ref-for-dfn-throw\u2460\u2463\u2461"}, {"id": "ref-for-dfn-throw\u2460\u2463\u2462"}, {"id": "ref-for-dfn-throw\u2460\u2463\u2463"}, {"id": "ref-for-dfn-throw\u2460\u2463\u2464"}], "title": "7.6.15. The gruCell() method"}, {"refs": [{"id": "ref-for-dfn-throw\u2460\u2463\u2465"}], "title": "7.6.16.1. The hardSigmoid(input, options) method"}, {"refs": [{"id": "ref-for-dfn-throw\u2460\u2463\u2466"}, {"id": "ref-for-dfn-throw\u2460\u2463\u2467"}], "title": "7.6.16.2. The hardSigmoid(options) method"}, {"refs": [{"id": "ref-for-dfn-throw\u2460\u2463\u2468"}], "title": "7.6.17.1. The hardSwish(input) method"}, {"refs": [{"id": "ref-for-dfn-throw\u2460\u2464\u24ea"}, {"id": "ref-for-dfn-throw\u2460\u2464\u2460"}], "title": "7.6.17.2. The hardSwish() method"}, {"refs": [{"id": "ref-for-dfn-throw\u2460\u2464\u2461"}, {"id": "ref-for-dfn-throw\u2460\u2464\u2462"}, {"id": "ref-for-dfn-throw\u2460\u2464\u2463"}, {"id": "ref-for-dfn-throw\u2460\u2464\u2464"}], "title": "7.6.18. The instanceNormalization() method"}, {"refs": [{"id": "ref-for-dfn-throw\u2460\u2464\u2465"}], "title": "7.6.19.1. The leakyRelu(input, options) method"}, {"refs": [{"id": "ref-for-dfn-throw\u2460\u2464\u2466"}, {"id": "ref-for-dfn-throw\u2460\u2464\u2467"}], "title": "7.6.19.2. The leakyRelu(options) method"}, {"refs": [{"id": "ref-for-dfn-throw\u2460\u2464\u2468"}], "title": "7.6.20.1. The linear(input, options) method"}, {"refs": [{"id": "ref-for-dfn-throw\u2460\u2465\u24ea"}, {"id": "ref-for-dfn-throw\u2460\u2465\u2460"}], "title": "7.6.20.2. The linear(options) method"}, {"refs": [{"id": "ref-for-dfn-throw\u2460\u2465\u2461"}, {"id": "ref-for-dfn-throw\u2460\u2465\u2462"}, {"id": "ref-for-dfn-throw\u2460\u2465\u2463"}, {"id": "ref-for-dfn-throw\u2460\u2465\u2464"}, {"id": "ref-for-dfn-throw\u2460\u2465\u2465"}, {"id": "ref-for-dfn-throw\u2460\u2465\u2466"}, {"id": "ref-for-dfn-throw\u2460\u2465\u2467"}, {"id": "ref-for-dfn-throw\u2460\u2465\u2468"}, {"id": "ref-for-dfn-throw\u2460\u2466\u24ea"}, {"id": "ref-for-dfn-throw\u2460\u2466\u2460"}, {"id": "ref-for-dfn-throw\u2460\u2466\u2461"}, {"id": "ref-for-dfn-throw\u2460\u2466\u2462"}, {"id": "ref-for-dfn-throw\u2460\u2466\u2463"}, {"id": "ref-for-dfn-throw\u2460\u2466\u2464"}, {"id": "ref-for-dfn-throw\u2460\u2466\u2465"}, {"id": "ref-for-dfn-throw\u2460\u2466\u2466"}, {"id": "ref-for-dfn-throw\u2460\u2466\u2467"}, {"id": "ref-for-dfn-throw\u2460\u2466\u2468"}, {"id": "ref-for-dfn-throw\u2460\u2467\u24ea"}, {"id": "ref-for-dfn-throw\u2460\u2467\u2460"}], "title": "7.6.21. The lstm() method"}, {"refs": [{"id": "ref-for-dfn-throw\u2460\u2467\u2461"}, {"id": "ref-for-dfn-throw\u2460\u2467\u2462"}, {"id": "ref-for-dfn-throw\u2460\u2467\u2463"}, {"id": "ref-for-dfn-throw\u2460\u2467\u2464"}, {"id": "ref-for-dfn-throw\u2460\u2467\u2465"}, {"id": "ref-for-dfn-throw\u2460\u2467\u2466"}, {"id": "ref-for-dfn-throw\u2460\u2467\u2467"}, {"id": "ref-for-dfn-throw\u2460\u2467\u2468"}, {"id": "ref-for-dfn-throw\u2460\u2468\u24ea"}], "title": "7.6.22. The lstmCell() method"}, {"refs": [{"id": "ref-for-dfn-throw\u2460\u2468\u2460"}, {"id": "ref-for-dfn-throw\u2460\u2468\u2461"}, {"id": "ref-for-dfn-throw\u2460\u2468\u2462"}, {"id": "ref-for-dfn-throw\u2460\u2468\u2463"}], "title": "7.6.23. The matmul() method"}, {"refs": [{"id": "ref-for-dfn-throw\u2460\u2468\u2464"}, {"id": "ref-for-dfn-throw\u2460\u2468\u2465"}], "title": "7.6.24. The pad() method"}, {"refs": [{"id": "ref-for-dfn-throw\u2460\u2468\u2466"}, {"id": "ref-for-dfn-throw\u2460\u2468\u2467"}, {"id": "ref-for-dfn-throw\u2460\u2468\u2468"}, {"id": "ref-for-dfn-throw\u2461\u24ea\u24ea"}, {"id": "ref-for-dfn-throw\u2461\u24ea\u2460"}, {"id": "ref-for-dfn-throw\u2461\u24ea\u2461"}, {"id": "ref-for-dfn-throw\u2461\u24ea\u2462"}, {"id": "ref-for-dfn-throw\u2461\u24ea\u2463"}, {"id": "ref-for-dfn-throw\u2461\u24ea\u2464"}, {"id": "ref-for-dfn-throw\u2461\u24ea\u2465"}, {"id": "ref-for-dfn-throw\u2461\u24ea\u2466"}, {"id": "ref-for-dfn-throw\u2461\u24ea\u2467"}, {"id": "ref-for-dfn-throw\u2461\u24ea\u2468"}, {"id": "ref-for-dfn-throw\u2461\u2460\u24ea"}, {"id": "ref-for-dfn-throw\u2461\u2460\u2460"}, {"id": "ref-for-dfn-throw\u2461\u2460\u2461"}], "title": "7.6.25. Pooling operations"}, {"refs": [{"id": "ref-for-dfn-throw\u2461\u2460\u2462"}, {"id": "ref-for-dfn-throw\u2461\u2460\u2463"}, {"id": "ref-for-dfn-throw\u2461\u2460\u2464"}], "title": "7.6.26. The prelu() method"}, {"refs": [{"id": "ref-for-dfn-throw\u2461\u2460\u2465"}, {"id": "ref-for-dfn-throw\u2461\u2460\u2466"}, {"id": "ref-for-dfn-throw\u2461\u2460\u2467"}, {"id": "ref-for-dfn-throw\u2461\u2460\u2468"}, {"id": "ref-for-dfn-throw\u2461\u2461\u24ea"}, {"id": "ref-for-dfn-throw\u2461\u2461\u2460"}, {"id": "ref-for-dfn-throw\u2461\u2461\u2461"}, {"id": "ref-for-dfn-throw\u2461\u2461\u2462"}, {"id": "ref-for-dfn-throw\u2461\u2461\u2463"}, {"id": "ref-for-dfn-throw\u2461\u2461\u2464"}, {"id": "ref-for-dfn-throw\u2461\u2461\u2465"}, {"id": "ref-for-dfn-throw\u2461\u2461\u2466"}, {"id": "ref-for-dfn-throw\u2461\u2461\u2467"}, {"id": "ref-for-dfn-throw\u2461\u2461\u2468"}, {"id": "ref-for-dfn-throw\u2461\u2462\u24ea"}, {"id": "ref-for-dfn-throw\u2461\u2462\u2460"}, {"id": "ref-for-dfn-throw\u2461\u2462\u2461"}, {"id": "ref-for-dfn-throw\u2461\u2462\u2462"}, {"id": "ref-for-dfn-throw\u2461\u2462\u2463"}, {"id": "ref-for-dfn-throw\u2461\u2462\u2464"}, {"id": "ref-for-dfn-throw\u2461\u2462\u2465"}, {"id": "ref-for-dfn-throw\u2461\u2462\u2466"}], "title": "7.6.27. Reduction operations"}, {"refs": [{"id": "ref-for-dfn-throw\u2461\u2462\u2467"}], "title": "7.6.28.1. The relu(input) method"}, {"refs": [{"id": "ref-for-dfn-throw\u2461\u2462\u2468"}, {"id": "ref-for-dfn-throw\u2461\u2463\u24ea"}], "title": "7.6.28.2. The relu() method"}, {"refs": [{"id": "ref-for-dfn-throw\u2461\u2463\u2460"}, {"id": "ref-for-dfn-throw\u2461\u2463\u2461"}, {"id": "ref-for-dfn-throw\u2461\u2463\u2462"}, {"id": "ref-for-dfn-throw\u2461\u2463\u2463"}, {"id": "ref-for-dfn-throw\u2461\u2463\u2464"}], "title": "7.6.29. The resample2d() method"}, {"refs": [{"id": "ref-for-dfn-throw\u2461\u2463\u2465"}, {"id": "ref-for-dfn-throw\u2461\u2463\u2466"}, {"id": "ref-for-dfn-throw\u2461\u2463\u2467"}, {"id": "ref-for-dfn-throw\u2461\u2463\u2468"}, {"id": "ref-for-dfn-throw\u2461\u2464\u24ea"}], "title": "7.6.30. The reshape() method"}, {"refs": [{"id": "ref-for-dfn-throw\u2461\u2464\u2460"}], "title": "7.6.31.1. The sigmoid(input) method"}, {"refs": [{"id": "ref-for-dfn-throw\u2461\u2464\u2461"}, {"id": "ref-for-dfn-throw\u2461\u2464\u2462"}], "title": "7.6.31.2. The sigmoid() method"}, {"refs": [{"id": "ref-for-dfn-throw\u2461\u2464\u2463"}, {"id": "ref-for-dfn-throw\u2461\u2464\u2464"}, {"id": "ref-for-dfn-throw\u2461\u2464\u2465"}], "title": "7.6.32. The slice() method"}, {"refs": [{"id": "ref-for-dfn-throw\u2461\u2464\u2466"}, {"id": "ref-for-dfn-throw\u2461\u2464\u2467"}], "title": "7.6.33.1. The softmax(input) method"}, {"refs": [{"id": "ref-for-dfn-throw\u2461\u2464\u2468"}, {"id": "ref-for-dfn-throw\u2461\u2465\u24ea"}], "title": "7.6.33.2. The softmax() method"}, {"refs": [{"id": "ref-for-dfn-throw\u2461\u2465\u2460"}], "title": "7.6.34.1. The softplus(input, options) method"}, {"refs": [{"id": "ref-for-dfn-throw\u2461\u2465\u2461"}, {"id": "ref-for-dfn-throw\u2461\u2465\u2462"}], "title": "7.6.34.2. The softplus(options) method"}, {"refs": [{"id": "ref-for-dfn-throw\u2461\u2465\u2463"}], "title": "7.6.35.1. The softsign(input) method"}, {"refs": [{"id": "ref-for-dfn-throw\u2461\u2465\u2464"}, {"id": "ref-for-dfn-throw\u2461\u2465\u2465"}], "title": "7.6.35.2. The softsign() method"}, {"refs": [{"id": "ref-for-dfn-throw\u2461\u2465\u2466"}, {"id": "ref-for-dfn-throw\u2461\u2465\u2467"}, {"id": "ref-for-dfn-throw\u2461\u2465\u2468"}], "title": "7.6.36. The split() method"}, {"refs": [{"id": "ref-for-dfn-throw\u2461\u2466\u24ea"}, {"id": "ref-for-dfn-throw\u2461\u2466\u2460"}], "title": "7.6.37. The squeeze() method"}, {"refs": [{"id": "ref-for-dfn-throw\u2461\u2466\u2461"}], "title": "7.6.38.1. The tanh(input) method"}, {"refs": [{"id": "ref-for-dfn-throw\u2461\u2466\u2462"}, {"id": "ref-for-dfn-throw\u2461\u2466\u2463"}], "title": "7.6.38.2. The tanh() method"}, {"refs": [{"id": "ref-for-dfn-throw\u2461\u2466\u2464"}, {"id": "ref-for-dfn-throw\u2461\u2466\u2465"}, {"id": "ref-for-dfn-throw\u2461\u2466\u2466"}, {"id": "ref-for-dfn-throw\u2461\u2466\u2467"}], "title": "7.6.39. The transpose() method"}], "external": true};
 window.dfnpanelData['69c80471'] = {"dfnID": "69c80471", "url": "https://webidl.spec.whatwg.org/#arraybuffer-transfer", "dfnText": "transfer", "refSections": [{"refs": [{"id": "ref-for-arraybuffer-transfer"}], "title": "7.4.3. The MLNamedArrayBufferViews transfer algorithm"}], "external": true};
 window.dfnpanelData['5f90bbfb'] = {"dfnID": "5f90bbfb", "url": "https://webidl.spec.whatwg.org/#idl-undefined", "dfnText": "undefined", "refSections": [{"refs": [{"id": "ref-for-idl-undefined"}, {"id": "ref-for-idl-undefined\u2460"}, {"id": "ref-for-idl-undefined\u2461"}, {"id": "ref-for-idl-undefined\u2462"}], "title": "7.4.2. Synchronous Execution"}, {"refs": [{"id": "ref-for-idl-undefined\u2463"}, {"id": "ref-for-idl-undefined\u2464"}], "title": "7.5.1. Graph Initialization"}, {"refs": [{"id": "ref-for-idl-undefined\u2465"}, {"id": "ref-for-idl-undefined\u2466"}, {"id": "ref-for-idl-undefined\u2467"}], "title": "7.5.2. Dispatch Execution Commands"}], "external": true};
 window.dfnpanelData['f32e046f'] = {"dfnID": "f32e046f", "url": "https://webidl.spec.whatwg.org/#buffersource-underlying-buffer", "dfnText": "underlying buffer", "refSections": [{"refs": [{"id": "ref-for-buffersource-underlying-buffer"}], "title": "7.4.3. The MLNamedArrayBufferViews transfer algorithm"}], "external": true};
-window.dfnpanelData['e97a9688'] = {"dfnID": "e97a9688", "url": "https://webidl.spec.whatwg.org/#idl-unsigned-long", "dfnText": "unsigned long", "refSections": [{"refs": [{"id": "ref-for-idl-unsigned-long"}], "title": "7.3.1. The MLOperandDescriptor dictionary"}, {"refs": [{"id": "ref-for-idl-unsigned-long\u2460"}, {"id": "ref-for-idl-unsigned-long\u2461"}], "title": "7.6.4. The batchNormalization() method"}, {"refs": [{"id": "ref-for-idl-unsigned-long\u2462"}, {"id": "ref-for-idl-unsigned-long\u2463"}], "title": "7.6.6. The concat() method"}, {"refs": [{"id": "ref-for-idl-unsigned-long\u2464"}, {"id": "ref-for-idl-unsigned-long\u2465"}, {"id": "ref-for-idl-unsigned-long\u2466"}, {"id": "ref-for-idl-unsigned-long\u2467"}, {"id": "ref-for-idl-unsigned-long\u2468"}, {"id": "ref-for-idl-unsigned-long\u2460\u24ea"}, {"id": "ref-for-idl-unsigned-long\u2460\u2460"}, {"id": "ref-for-idl-unsigned-long\u2460\u2461"}], "title": "7.6.7. The conv2d() method"}, {"refs": [{"id": "ref-for-idl-unsigned-long\u2460\u2462"}, {"id": "ref-for-idl-unsigned-long\u2460\u2463"}, {"id": "ref-for-idl-unsigned-long\u2460\u2464"}, {"id": "ref-for-idl-unsigned-long\u2460\u2465"}, {"id": "ref-for-idl-unsigned-long\u2460\u2466"}, {"id": "ref-for-idl-unsigned-long\u2460\u2467"}, {"id": "ref-for-idl-unsigned-long\u2460\u2468"}, {"id": "ref-for-idl-unsigned-long\u2461\u24ea"}, {"id": "ref-for-idl-unsigned-long\u2461\u2460"}, {"id": "ref-for-idl-unsigned-long\u2461\u2461"}, {"id": "ref-for-idl-unsigned-long\u2461\u2462"}, {"id": "ref-for-idl-unsigned-long\u2461\u2463"}], "title": "7.6.8. The convTranspose2d() method"}, {"refs": [{"id": "ref-for-idl-unsigned-long\u2461\u2464"}, {"id": "ref-for-idl-unsigned-long\u2461\u2465"}, {"id": "ref-for-idl-unsigned-long\u2461\u2466"}, {"id": "ref-for-idl-unsigned-long\u2461\u2467"}], "title": "7.6.13. The gru() method"}, {"refs": [{"id": "ref-for-idl-unsigned-long\u2461\u2468"}, {"id": "ref-for-idl-unsigned-long\u2462\u24ea"}], "title": "7.6.14. The gruCell() method"}, {"refs": [{"id": "ref-for-idl-unsigned-long\u2462\u2460"}, {"id": "ref-for-idl-unsigned-long\u2462\u2461"}, {"id": "ref-for-idl-unsigned-long\u2462\u2462"}, {"id": "ref-for-idl-unsigned-long\u2462\u2463"}], "title": "7.6.20. The lstm() method"}, {"refs": [{"id": "ref-for-idl-unsigned-long\u2462\u2464"}, {"id": "ref-for-idl-unsigned-long\u2462\u2465"}], "title": "7.6.21. The lstmCell() method"}, {"refs": [{"id": "ref-for-idl-unsigned-long\u2462\u2466"}, {"id": "ref-for-idl-unsigned-long\u2462\u2467"}, {"id": "ref-for-idl-unsigned-long\u2462\u2468"}, {"id": "ref-for-idl-unsigned-long\u2463\u24ea"}], "title": "7.6.23. The pad() method"}, {"refs": [{"id": "ref-for-idl-unsigned-long\u2463\u2460"}, {"id": "ref-for-idl-unsigned-long\u2463\u2461"}, {"id": "ref-for-idl-unsigned-long\u2463\u2462"}, {"id": "ref-for-idl-unsigned-long\u2463\u2463"}, {"id": "ref-for-idl-unsigned-long\u2463\u2464"}, {"id": "ref-for-idl-unsigned-long\u2463\u2465"}, {"id": "ref-for-idl-unsigned-long\u2463\u2466"}, {"id": "ref-for-idl-unsigned-long\u2463\u2467"}, {"id": "ref-for-idl-unsigned-long\u2463\u2468"}, {"id": "ref-for-idl-unsigned-long\u2464\u24ea"}], "title": "7.6.24. Pooling operations"}, {"refs": [{"id": "ref-for-idl-unsigned-long\u2464\u2460"}, {"id": "ref-for-idl-unsigned-long\u2464\u2461"}], "title": "7.6.26. Reduction operations"}, {"refs": [{"id": "ref-for-idl-unsigned-long\u2464\u2462"}, {"id": "ref-for-idl-unsigned-long\u2464\u2463"}, {"id": "ref-for-idl-unsigned-long\u2464\u2464"}, {"id": "ref-for-idl-unsigned-long\u2464\u2465"}], "title": "7.6.28. The resample2d() method"}, {"refs": [{"id": "ref-for-idl-unsigned-long\u2464\u2466"}, {"id": "ref-for-idl-unsigned-long\u2464\u2467"}], "title": "7.6.29. The reshape() method"}, {"refs": [{"id": "ref-for-idl-unsigned-long\u2464\u2468"}, {"id": "ref-for-idl-unsigned-long\u2465\u24ea"}, {"id": "ref-for-idl-unsigned-long\u2465\u2460"}, {"id": "ref-for-idl-unsigned-long\u2465\u2461"}], "title": "7.6.31. The slice() method"}, {"refs": [{"id": "ref-for-idl-unsigned-long\u2465\u2462"}, {"id": "ref-for-idl-unsigned-long\u2465\u2463"}, {"id": "ref-for-idl-unsigned-long\u2465\u2464"}, {"id": "ref-for-idl-unsigned-long\u2465\u2465"}, {"id": "ref-for-idl-unsigned-long\u2465\u2466"}, {"id": "ref-for-idl-unsigned-long\u2465\u2467"}, {"id": "ref-for-idl-unsigned-long\u2465\u2468"}, {"id": "ref-for-idl-unsigned-long\u2466\u24ea"}, {"id": "ref-for-idl-unsigned-long\u2466\u2460"}, {"id": "ref-for-idl-unsigned-long\u2466\u2461"}], "title": "7.6.35. The split() method"}, {"refs": [{"id": "ref-for-idl-unsigned-long\u2466\u2462"}, {"id": "ref-for-idl-unsigned-long\u2466\u2463"}], "title": "7.6.36. The squeeze() method"}, {"refs": [{"id": "ref-for-idl-unsigned-long\u2466\u2464"}, {"id": "ref-for-idl-unsigned-long\u2466\u2465"}], "title": "7.6.38. The transpose() method"}], "external": true};
-window.dfnpanelData['f14b47b8'] = {"dfnID": "f14b47b8", "url": "https://webidl.spec.whatwg.org/#idl-unsigned-long-long", "dfnText": "unsigned long long", "refSections": [{"refs": [{"id": "ref-for-idl-unsigned-long-long"}, {"id": "ref-for-idl-unsigned-long-long\u2460"}], "title": "7.6. The MLGraphBuilder interface"}], "external": true};
+window.dfnpanelData['e97a9688'] = {"dfnID": "e97a9688", "url": "https://webidl.spec.whatwg.org/#idl-unsigned-long", "dfnText": "unsigned long", "refSections": [{"refs": [{"id": "ref-for-idl-unsigned-long"}], "title": "7.3.1. The MLOperandDescriptor dictionary"}, {"refs": [{"id": "ref-for-idl-unsigned-long\u2460"}, {"id": "ref-for-idl-unsigned-long\u2461"}], "title": "7.6.5. The batchNormalization() method"}, {"refs": [{"id": "ref-for-idl-unsigned-long\u2462"}, {"id": "ref-for-idl-unsigned-long\u2463"}], "title": "7.6.7. The concat() method"}, {"refs": [{"id": "ref-for-idl-unsigned-long\u2464"}, {"id": "ref-for-idl-unsigned-long\u2465"}, {"id": "ref-for-idl-unsigned-long\u2466"}, {"id": "ref-for-idl-unsigned-long\u2467"}, {"id": "ref-for-idl-unsigned-long\u2468"}, {"id": "ref-for-idl-unsigned-long\u2460\u24ea"}, {"id": "ref-for-idl-unsigned-long\u2460\u2460"}, {"id": "ref-for-idl-unsigned-long\u2460\u2461"}, {"id": "ref-for-idl-unsigned-long\u2460\u2462"}], "title": "7.6.8. The conv2d() method"}, {"refs": [{"id": "ref-for-idl-unsigned-long\u2460\u2463"}, {"id": "ref-for-idl-unsigned-long\u2460\u2464"}, {"id": "ref-for-idl-unsigned-long\u2460\u2465"}, {"id": "ref-for-idl-unsigned-long\u2460\u2466"}, {"id": "ref-for-idl-unsigned-long\u2460\u2467"}, {"id": "ref-for-idl-unsigned-long\u2460\u2468"}, {"id": "ref-for-idl-unsigned-long\u2461\u24ea"}, {"id": "ref-for-idl-unsigned-long\u2461\u2460"}, {"id": "ref-for-idl-unsigned-long\u2461\u2461"}, {"id": "ref-for-idl-unsigned-long\u2461\u2462"}, {"id": "ref-for-idl-unsigned-long\u2461\u2463"}, {"id": "ref-for-idl-unsigned-long\u2461\u2464"}, {"id": "ref-for-idl-unsigned-long\u2461\u2465"}], "title": "7.6.9. The convTranspose2d() method"}, {"refs": [{"id": "ref-for-idl-unsigned-long\u2461\u2466"}, {"id": "ref-for-idl-unsigned-long\u2461\u2467"}, {"id": "ref-for-idl-unsigned-long\u2461\u2468"}, {"id": "ref-for-idl-unsigned-long\u2462\u24ea"}], "title": "7.6.14. The gru() method"}, {"refs": [{"id": "ref-for-idl-unsigned-long\u2462\u2460"}, {"id": "ref-for-idl-unsigned-long\u2462\u2461"}], "title": "7.6.15. The gruCell() method"}, {"refs": [{"id": "ref-for-idl-unsigned-long\u2462\u2462"}, {"id": "ref-for-idl-unsigned-long\u2462\u2463"}, {"id": "ref-for-idl-unsigned-long\u2462\u2464"}, {"id": "ref-for-idl-unsigned-long\u2462\u2465"}], "title": "7.6.21. The lstm() method"}, {"refs": [{"id": "ref-for-idl-unsigned-long\u2462\u2466"}, {"id": "ref-for-idl-unsigned-long\u2462\u2467"}], "title": "7.6.22. The lstmCell() method"}, {"refs": [{"id": "ref-for-idl-unsigned-long\u2462\u2468"}, {"id": "ref-for-idl-unsigned-long\u2463\u24ea"}, {"id": "ref-for-idl-unsigned-long\u2463\u2460"}, {"id": "ref-for-idl-unsigned-long\u2463\u2461"}], "title": "7.6.24. The pad() method"}, {"refs": [{"id": "ref-for-idl-unsigned-long\u2463\u2462"}, {"id": "ref-for-idl-unsigned-long\u2463\u2463"}, {"id": "ref-for-idl-unsigned-long\u2463\u2464"}, {"id": "ref-for-idl-unsigned-long\u2463\u2465"}, {"id": "ref-for-idl-unsigned-long\u2463\u2466"}, {"id": "ref-for-idl-unsigned-long\u2463\u2467"}, {"id": "ref-for-idl-unsigned-long\u2463\u2468"}, {"id": "ref-for-idl-unsigned-long\u2464\u24ea"}, {"id": "ref-for-idl-unsigned-long\u2464\u2460"}, {"id": "ref-for-idl-unsigned-long\u2464\u2461"}], "title": "7.6.25. Pooling operations"}, {"refs": [{"id": "ref-for-idl-unsigned-long\u2464\u2462"}, {"id": "ref-for-idl-unsigned-long\u2464\u2463"}], "title": "7.6.27. Reduction operations"}, {"refs": [{"id": "ref-for-idl-unsigned-long\u2464\u2464"}, {"id": "ref-for-idl-unsigned-long\u2464\u2465"}, {"id": "ref-for-idl-unsigned-long\u2464\u2466"}, {"id": "ref-for-idl-unsigned-long\u2464\u2467"}], "title": "7.6.29. The resample2d() method"}, {"refs": [{"id": "ref-for-idl-unsigned-long\u2464\u2468"}, {"id": "ref-for-idl-unsigned-long\u2465\u24ea"}, {"id": "ref-for-idl-unsigned-long\u2465\u2460"}, {"id": "ref-for-idl-unsigned-long\u2465\u2461"}, {"id": "ref-for-idl-unsigned-long\u2465\u2462"}], "title": "7.6.30. The reshape() method"}, {"refs": [{"id": "ref-for-idl-unsigned-long\u2465\u2463"}, {"id": "ref-for-idl-unsigned-long\u2465\u2464"}, {"id": "ref-for-idl-unsigned-long\u2465\u2465"}, {"id": "ref-for-idl-unsigned-long\u2465\u2466"}], "title": "7.6.32. The slice() method"}, {"refs": [{"id": "ref-for-idl-unsigned-long\u2465\u2467"}, {"id": "ref-for-idl-unsigned-long\u2465\u2468"}, {"id": "ref-for-idl-unsigned-long\u2466\u24ea"}, {"id": "ref-for-idl-unsigned-long\u2466\u2460"}, {"id": "ref-for-idl-unsigned-long\u2466\u2461"}, {"id": "ref-for-idl-unsigned-long\u2466\u2462"}, {"id": "ref-for-idl-unsigned-long\u2466\u2463"}, {"id": "ref-for-idl-unsigned-long\u2466\u2464"}, {"id": "ref-for-idl-unsigned-long\u2466\u2465"}, {"id": "ref-for-idl-unsigned-long\u2466\u2466"}, {"id": "ref-for-idl-unsigned-long\u2466\u2467"}, {"id": "ref-for-idl-unsigned-long\u2466\u2468"}, {"id": "ref-for-idl-unsigned-long\u2467\u24ea"}], "title": "7.6.36. The split() method"}, {"refs": [{"id": "ref-for-idl-unsigned-long\u2467\u2460"}, {"id": "ref-for-idl-unsigned-long\u2467\u2461"}], "title": "7.6.37. The squeeze() method"}, {"refs": [{"id": "ref-for-idl-unsigned-long\u2467\u2462"}, {"id": "ref-for-idl-unsigned-long\u2467\u2463"}], "title": "7.6.39. The transpose() method"}], "external": true};
+window.dfnpanelData['f14b47b8'] = {"dfnID": "f14b47b8", "url": "https://webidl.spec.whatwg.org/#idl-unsigned-long-long", "dfnText": "unsigned long long", "refSections": [{"refs": [{"id": "ref-for-idl-unsigned-long-long"}, {"id": "ref-for-idl-unsigned-long-long\u2460"}, {"id": "ref-for-idl-unsigned-long-long\u2461"}, {"id": "ref-for-idl-unsigned-long-long\u2462"}, {"id": "ref-for-idl-unsigned-long-long\u2463"}, {"id": "ref-for-idl-unsigned-long-long\u2464"}], "title": "7.6. The MLGraphBuilder interface"}], "external": true};
 window.dfnpanelData['f3fd9d04'] = {"dfnID": "f3fd9d04", "url": "https://webidl.spec.whatwg.org/#arraybufferview-write", "dfnText": "write", "refSections": [{"refs": [{"id": "ref-for-arraybufferview-write"}], "title": "7.4.4. Asynchronous Execution"}], "external": true};
 window.dfnpanelData['navigatorml'] = {"dfnID": "navigatorml", "url": "#navigatorml", "dfnText": "NavigatorML", "refSections": [{"refs": [{"id": "ref-for-navigatorml"}, {"id": "ref-for-navigatorml\u2460"}], "title": "7.1. The navigator.ml interface"}], "external": false};
 window.dfnpanelData['enumdef-mldevicetype'] = {"dfnID": "enumdef-mldevicetype", "url": "#enumdef-mldevicetype", "dfnText": "MLDeviceType", "refSections": [{"refs": [{"id": "ref-for-enumdef-mldevicetype"}], "title": "6.2. Device Selection"}, {"refs": [{"id": "ref-for-enumdef-mldevicetype\u2460"}], "title": "7.2. The ML interface"}], "external": false};
@@ -5903,35 +9317,46 @@ <h2 class="no-num no-ref heading settled" id="issues-index"><span class="content
 window.dfnpanelData['dom-mlcontextoptions-devicetype'] = {"dfnID": "dom-mlcontextoptions-devicetype", "url": "#dom-mlcontextoptions-devicetype", "dfnText": "deviceType", "refSections": [{"refs": [{"id": "ref-for-dom-mlcontextoptions-devicetype"}, {"id": "ref-for-dom-mlcontextoptions-devicetype\u2460"}], "title": "7.2.2. The createContext() method"}, {"refs": [{"id": "ref-for-dom-mlcontextoptions-devicetype\u2461"}], "title": "7.4. The MLContext interface"}], "external": false};
 window.dfnpanelData['dom-mlcontextoptions-powerpreference'] = {"dfnID": "dom-mlcontextoptions-powerpreference", "url": "#dom-mlcontextoptions-powerpreference", "dfnText": "powerPreference", "refSections": [{"refs": [{"id": "ref-for-dom-mlcontextoptions-powerpreference"}, {"id": "ref-for-dom-mlcontextoptions-powerpreference\u2460"}], "title": "7.2.2. The createContext() method"}], "external": false};
 window.dfnpanelData['ml'] = {"dfnID": "ml", "url": "#ml", "dfnText": "ML", "refSections": [{"refs": [{"id": "ref-for-ml"}, {"id": "ref-for-ml\u2460"}], "title": "7.1. The navigator.ml interface"}], "external": false};
-window.dfnpanelData['dom-ml-createcontext'] = {"dfnID": "dom-ml-createcontext", "url": "#dom-ml-createcontext", "dfnText": "createContext", "refSections": [{"refs": [{"id": "ref-for-dom-ml-createcontext"}, {"id": "ref-for-dom-ml-createcontext\u2460"}], "title": "7.2.2. The createContext() method"}], "external": false};
-window.dfnpanelData['dom-ml-createcontextsync'] = {"dfnID": "dom-ml-createcontextsync", "url": "#dom-ml-createcontextsync", "dfnText": "createContextSync", "refSections": [{"refs": [{"id": "ref-for-dom-ml-createcontextsync"}, {"id": "ref-for-dom-ml-createcontextsync\u2460"}], "title": "7.2.3. The createContextSync() method"}], "external": false};
-window.dfnpanelData['webnn-feature'] = {"dfnID": "webnn-feature", "url": "#webnn-feature", "dfnText": "webnn", "refSections": [{"refs": [{"id": "ref-for-webnn-feature"}], "title": "7.2.2. The createContext() method"}, {"refs": [{"id": "ref-for-webnn-feature\u2460"}], "title": "7.2.3. The createContextSync() method"}, {"refs": [{"id": "ref-for-webnn-feature\u2461"}], "title": "7.6.1. The MLGraphBuilder constructor"}], "external": false};
-window.dfnpanelData['create-context'] = {"dfnID": "create-context", "url": "#create-context", "dfnText": "create context", "refSections": [{"refs": [{"id": "ref-for-create-context"}], "title": "7.2.3. The createContextSync() method"}], "external": false};
-window.dfnpanelData['mlgraph'] = {"dfnID": "mlgraph", "url": "#mlgraph", "dfnText": "MLGraph", "refSections": [{"refs": [{"id": "ref-for-mlgraph"}, {"id": "ref-for-mlgraph\u2460"}, {"id": "ref-for-mlgraph\u2461"}, {"id": "ref-for-mlgraph\u2462"}, {"id": "ref-for-mlgraph\u2463"}, {"id": "ref-for-mlgraph\u2464"}], "title": "6.1. Overview"}, {"refs": [{"id": "ref-for-mlgraph\u2465"}, {"id": "ref-for-mlgraph\u2466"}, {"id": "ref-for-mlgraph\u2467"}, {"id": "ref-for-mlgraph\u2468"}, {"id": "ref-for-mlgraph\u2460\u24ea"}], "title": "7.3. The MLGraph interface"}, {"refs": [{"id": "ref-for-mlgraph\u2460\u2460"}, {"id": "ref-for-mlgraph\u2460\u2461"}, {"id": "ref-for-mlgraph\u2460\u2462"}], "title": "7.4.2. Synchronous Execution"}, {"refs": [{"id": "ref-for-mlgraph\u2460\u2463"}, {"id": "ref-for-mlgraph\u2460\u2464"}, {"id": "ref-for-mlgraph\u2460\u2465"}], "title": "7.4.4. Asynchronous Execution"}, {"refs": [{"id": "ref-for-mlgraph\u2460\u2466"}], "title": "7.5. The MLCommandEncoder interface"}, {"refs": [{"id": "ref-for-mlgraph\u2460\u2467"}, {"id": "ref-for-mlgraph\u2460\u2468"}, {"id": "ref-for-mlgraph\u2461\u24ea"}], "title": "7.5.1. Graph Initialization"}, {"refs": [{"id": "ref-for-mlgraph\u2461\u2460"}, {"id": "ref-for-mlgraph\u2461\u2461"}, {"id": "ref-for-mlgraph\u2461\u2462"}], "title": "7.5.2. Dispatch Execution Commands"}, {"refs": [{"id": "ref-for-mlgraph\u2461\u2463"}, {"id": "ref-for-mlgraph\u2461\u2464"}, {"id": "ref-for-mlgraph\u2461\u2465"}], "title": "7.6. The MLGraphBuilder interface"}], "external": false};
-window.dfnpanelData['dom-mlgraph-context-slot'] = {"dfnID": "dom-mlgraph-context-slot", "url": "#dom-mlgraph-context-slot", "dfnText": "[[context]]", "refSections": [{"refs": [{"id": "ref-for-dom-mlgraph-context-slot"}], "title": "7.4.2. Synchronous Execution"}, {"refs": [{"id": "ref-for-dom-mlgraph-context-slot\u2460"}], "title": "7.4.4. Asynchronous Execution"}], "external": false};
-window.dfnpanelData['dom-mlgraph-inputdescriptors-slot'] = {"dfnID": "dom-mlgraph-inputdescriptors-slot", "url": "#dom-mlgraph-inputdescriptors-slot", "dfnText": "[[inputDescriptors]]", "refSections": [{"refs": [{"id": "ref-for-dom-mlgraph-inputdescriptors-slot"}, {"id": "ref-for-dom-mlgraph-inputdescriptors-slot\u2460"}], "title": "7.4.2. Synchronous Execution"}, {"refs": [{"id": "ref-for-dom-mlgraph-inputdescriptors-slot\u2461"}], "title": "7.4.4. Asynchronous Execution"}, {"refs": [{"id": "ref-for-dom-mlgraph-inputdescriptors-slot\u2462"}, {"id": "ref-for-dom-mlgraph-inputdescriptors-slot\u2463"}], "title": "7.5.2. Dispatch Execution Commands"}], "external": false};
+window.dfnpanelData['webnn-feature'] = {"dfnID": "webnn-feature", "url": "#webnn-feature", "dfnText": "webnn", "refSections": [{"refs": [{"id": "ref-for-webnn-feature"}, {"id": "ref-for-webnn-feature\u2460"}], "title": "7.2.2. The createContext() method"}, {"refs": [{"id": "ref-for-webnn-feature\u2461"}, {"id": "ref-for-webnn-feature\u2462"}], "title": "7.2.3. The createContextSync() method"}, {"refs": [{"id": "ref-for-webnn-feature\u2463"}], "title": "7.6.1. The MLGraphBuilder constructor"}], "external": false};
+window.dfnpanelData['create-a-context'] = {"dfnID": "create-a-context", "url": "#create-a-context", "dfnText": "create a context", "refSections": [{"refs": [{"id": "ref-for-create-a-context"}, {"id": "ref-for-create-a-context\u2460"}], "title": "7.2.2. The createContext() method"}, {"refs": [{"id": "ref-for-create-a-context\u2461"}, {"id": "ref-for-create-a-context\u2462"}], "title": "7.2.3. The createContextSync() method"}], "external": false};
+window.dfnpanelData['dom-ml-createcontext'] = {"dfnID": "dom-ml-createcontext", "url": "#dom-ml-createcontext", "dfnText": "createContext(options)", "refSections": [{"refs": [{"id": "ref-for-dom-ml-createcontext"}], "title": "7.2. The ML interface"}, {"refs": [{"id": "ref-for-dom-ml-createcontext\u2460"}], "title": "7.2.2. The createContext() method"}], "external": false};
+window.dfnpanelData['dom-ml-createcontext-gpudevice'] = {"dfnID": "dom-ml-createcontext-gpudevice", "url": "#dom-ml-createcontext-gpudevice", "dfnText": "createContext(gpuDevice)", "refSections": [{"refs": [{"id": "ref-for-dom-ml-createcontext-gpudevice"}], "title": "7.2. The ML interface"}], "external": false};
+window.dfnpanelData['dom-ml-createcontextsync'] = {"dfnID": "dom-ml-createcontextsync", "url": "#dom-ml-createcontextsync", "dfnText": "createContextSync(options)", "refSections": [{"refs": [{"id": "ref-for-dom-ml-createcontextsync"}], "title": "7.2. The ML interface"}, {"refs": [{"id": "ref-for-dom-ml-createcontextsync\u2460"}], "title": "7.2.3. The createContextSync() method"}], "external": false};
+window.dfnpanelData['dom-ml-createcontextsync-gpudevice'] = {"dfnID": "dom-ml-createcontextsync-gpudevice", "url": "#dom-ml-createcontextsync-gpudevice", "dfnText": "createContextSync(gpuDevice)", "refSections": [{"refs": [{"id": "ref-for-dom-ml-createcontextsync-gpudevice"}], "title": "7.2. The ML interface"}], "external": false};
+window.dfnpanelData['mlgraph'] = {"dfnID": "mlgraph", "url": "#mlgraph", "dfnText": "MLGraph", "refSections": [{"refs": [{"id": "ref-for-mlgraph"}, {"id": "ref-for-mlgraph\u2460"}, {"id": "ref-for-mlgraph\u2461"}, {"id": "ref-for-mlgraph\u2462"}, {"id": "ref-for-mlgraph\u2463"}, {"id": "ref-for-mlgraph\u2464"}], "title": "6.1. Overview"}, {"refs": [{"id": "ref-for-mlgraph\u2465"}, {"id": "ref-for-mlgraph\u2466"}, {"id": "ref-for-mlgraph\u2467"}, {"id": "ref-for-mlgraph\u2468"}, {"id": "ref-for-mlgraph\u2460\u24ea"}], "title": "7.3. The MLGraph interface"}, {"refs": [{"id": "ref-for-mlgraph\u2460\u2460"}, {"id": "ref-for-mlgraph\u2460\u2461"}, {"id": "ref-for-mlgraph\u2460\u2462"}], "title": "7.4.2. Synchronous Execution"}, {"refs": [{"id": "ref-for-mlgraph\u2460\u2463"}, {"id": "ref-for-mlgraph\u2460\u2464"}, {"id": "ref-for-mlgraph\u2460\u2465"}], "title": "7.4.4. Asynchronous Execution"}, {"refs": [{"id": "ref-for-mlgraph\u2460\u2466"}], "title": "7.5. The MLCommandEncoder interface"}, {"refs": [{"id": "ref-for-mlgraph\u2460\u2467"}, {"id": "ref-for-mlgraph\u2460\u2468"}, {"id": "ref-for-mlgraph\u2461\u24ea"}], "title": "7.5.1. Graph Initialization"}, {"refs": [{"id": "ref-for-mlgraph\u2461\u2460"}, {"id": "ref-for-mlgraph\u2461\u2461"}, {"id": "ref-for-mlgraph\u2461\u2462"}], "title": "7.5.2. Dispatch Execution Commands"}, {"refs": [{"id": "ref-for-mlgraph\u2461\u2463"}, {"id": "ref-for-mlgraph\u2461\u2464"}, {"id": "ref-for-mlgraph\u2461\u2465"}], "title": "7.6. The MLGraphBuilder interface"}, {"refs": [{"id": "ref-for-mlgraph\u2461\u2466"}], "title": "7.6.3.2. The buildSync(outputs) method"}], "external": false};
+window.dfnpanelData['dom-mlgraph-context-slot'] = {"dfnID": "dom-mlgraph-context-slot", "url": "#dom-mlgraph-context-slot", "dfnText": "[[context]]", "refSections": [{"refs": [{"id": "ref-for-dom-mlgraph-context-slot"}], "title": "7.4.2. Synchronous Execution"}, {"refs": [{"id": "ref-for-dom-mlgraph-context-slot\u2460"}], "title": "7.4.4. Asynchronous Execution"}, {"refs": [{"id": "ref-for-dom-mlgraph-context-slot\u2461"}], "title": "7.6.3.2. The buildSync(outputs) method"}], "external": false};
+window.dfnpanelData['dom-mlgraph-inputdescriptors-slot'] = {"dfnID": "dom-mlgraph-inputdescriptors-slot", "url": "#dom-mlgraph-inputdescriptors-slot", "dfnText": "[[inputDescriptors]]", "refSections": [{"refs": [{"id": "ref-for-dom-mlgraph-inputdescriptors-slot"}, {"id": "ref-for-dom-mlgraph-inputdescriptors-slot\u2460"}], "title": "7.4.2. Synchronous Execution"}, {"refs": [{"id": "ref-for-dom-mlgraph-inputdescriptors-slot\u2461"}], "title": "7.4.4. Asynchronous Execution"}, {"refs": [{"id": "ref-for-dom-mlgraph-inputdescriptors-slot\u2462"}, {"id": "ref-for-dom-mlgraph-inputdescriptors-slot\u2463"}], "title": "7.5.2. Dispatch Execution Commands"}, {"refs": [{"id": "ref-for-dom-mlgraph-inputdescriptors-slot\u2464"}], "title": "7.6.3.2. The buildSync(outputs) method"}], "external": false};
 window.dfnpanelData['dom-mlgraph-outputdescriptors-slot'] = {"dfnID": "dom-mlgraph-outputdescriptors-slot", "url": "#dom-mlgraph-outputdescriptors-slot", "dfnText": "[[outputDescriptors]]", "refSections": [{"refs": [{"id": "ref-for-dom-mlgraph-outputdescriptors-slot"}, {"id": "ref-for-dom-mlgraph-outputdescriptors-slot\u2460"}], "title": "7.4.2. Synchronous Execution"}, {"refs": [{"id": "ref-for-dom-mlgraph-outputdescriptors-slot\u2461"}], "title": "7.4.4. Asynchronous Execution"}, {"refs": [{"id": "ref-for-dom-mlgraph-outputdescriptors-slot\u2462"}, {"id": "ref-for-dom-mlgraph-outputdescriptors-slot\u2463"}], "title": "7.5.2. Dispatch Execution Commands"}], "external": false};
-window.dfnpanelData['dom-mlgraph-implementation-slot'] = {"dfnID": "dom-mlgraph-implementation-slot", "url": "#dom-mlgraph-implementation-slot", "dfnText": "[[implementation]]", "refSections": [{"refs": [{"id": "ref-for-dom-mlgraph-implementation-slot"}, {"id": "ref-for-dom-mlgraph-implementation-slot\u2460"}, {"id": "ref-for-dom-mlgraph-implementation-slot\u2461"}], "title": "7.4.2. Synchronous Execution"}, {"refs": [{"id": "ref-for-dom-mlgraph-implementation-slot\u2462"}, {"id": "ref-for-dom-mlgraph-implementation-slot\u2463"}, {"id": "ref-for-dom-mlgraph-implementation-slot\u2464"}, {"id": "ref-for-dom-mlgraph-implementation-slot\u2465"}], "title": "7.5.2. Dispatch Execution Commands"}], "external": false};
-window.dfnpanelData['enumdef-mlinputoperandlayout'] = {"dfnID": "enumdef-mlinputoperandlayout", "url": "#enumdef-mlinputoperandlayout", "dfnText": "MLInputOperandLayout", "refSections": [{"refs": [{"id": "ref-for-enumdef-mlinputoperandlayout"}, {"id": "ref-for-enumdef-mlinputoperandlayout\u2460"}], "title": "7.6.7. The conv2d() method"}, {"refs": [{"id": "ref-for-enumdef-mlinputoperandlayout\u2461"}, {"id": "ref-for-enumdef-mlinputoperandlayout\u2462"}], "title": "7.6.8. The convTranspose2d() method"}, {"refs": [{"id": "ref-for-enumdef-mlinputoperandlayout\u2463"}, {"id": "ref-for-enumdef-mlinputoperandlayout\u2464"}], "title": "7.6.17. The instanceNormalization() method"}, {"refs": [{"id": "ref-for-enumdef-mlinputoperandlayout\u2465"}, {"id": "ref-for-enumdef-mlinputoperandlayout\u2466"}], "title": "7.6.24. Pooling operations"}], "external": false};
-window.dfnpanelData['enumdef-mloperandtype'] = {"dfnID": "enumdef-mloperandtype", "url": "#enumdef-mloperandtype", "dfnText": "MLOperandType", "refSections": [{"refs": [{"id": "ref-for-enumdef-mloperandtype"}], "title": "4. Privacy Considerations"}, {"refs": [{"id": "ref-for-enumdef-mloperandtype\u2460"}], "title": "7.3.1. The MLOperandDescriptor dictionary"}, {"refs": [{"id": "ref-for-enumdef-mloperandtype\u2461"}], "title": "7.6. The MLGraphBuilder interface"}, {"refs": [{"id": "ref-for-enumdef-mloperandtype\u2462"}, {"id": "ref-for-enumdef-mloperandtype\u2463"}], "title": "7.6.3.2. The constant(value, type) method"}, {"refs": [{"id": "ref-for-enumdef-mloperandtype\u2464"}, {"id": "ref-for-enumdef-mloperandtype\u2465"}], "title": "9.1. MLOperandType and ArrayBufferView compatibility"}], "external": false};
+window.dfnpanelData['dom-mlgraph-implementation-slot'] = {"dfnID": "dom-mlgraph-implementation-slot", "url": "#dom-mlgraph-implementation-slot", "dfnText": "[[implementation]]", "refSections": [{"refs": [{"id": "ref-for-dom-mlgraph-implementation-slot"}, {"id": "ref-for-dom-mlgraph-implementation-slot\u2460"}, {"id": "ref-for-dom-mlgraph-implementation-slot\u2461"}], "title": "7.4.2. Synchronous Execution"}, {"refs": [{"id": "ref-for-dom-mlgraph-implementation-slot\u2462"}, {"id": "ref-for-dom-mlgraph-implementation-slot\u2463"}, {"id": "ref-for-dom-mlgraph-implementation-slot\u2464"}, {"id": "ref-for-dom-mlgraph-implementation-slot\u2465"}], "title": "7.5.2. Dispatch Execution Commands"}, {"refs": [{"id": "ref-for-dom-mlgraph-implementation-slot\u2466"}], "title": "7.6.3.2. The buildSync(outputs) method"}], "external": false};
+window.dfnpanelData['enumdef-mlinputoperandlayout'] = {"dfnID": "enumdef-mlinputoperandlayout", "url": "#enumdef-mlinputoperandlayout", "dfnText": "MLInputOperandLayout", "refSections": [{"refs": [{"id": "ref-for-enumdef-mlinputoperandlayout"}, {"id": "ref-for-enumdef-mlinputoperandlayout\u2460"}, {"id": "ref-for-enumdef-mlinputoperandlayout\u2461"}], "title": "7.6.8. The conv2d() method"}, {"refs": [{"id": "ref-for-enumdef-mlinputoperandlayout\u2462"}, {"id": "ref-for-enumdef-mlinputoperandlayout\u2463"}, {"id": "ref-for-enumdef-mlinputoperandlayout\u2464"}], "title": "7.6.9. The convTranspose2d() method"}, {"refs": [{"id": "ref-for-enumdef-mlinputoperandlayout\u2465"}, {"id": "ref-for-enumdef-mlinputoperandlayout\u2466"}, {"id": "ref-for-enumdef-mlinputoperandlayout\u2467"}], "title": "7.6.18. The instanceNormalization() method"}, {"refs": [{"id": "ref-for-enumdef-mlinputoperandlayout\u2468"}, {"id": "ref-for-enumdef-mlinputoperandlayout\u2460\u24ea"}, {"id": "ref-for-enumdef-mlinputoperandlayout\u2460\u2460"}], "title": "7.6.25. Pooling operations"}], "external": false};
+window.dfnpanelData['enumdef-mloperandtype'] = {"dfnID": "enumdef-mloperandtype", "url": "#enumdef-mloperandtype", "dfnText": "MLOperandType", "refSections": [{"refs": [{"id": "ref-for-enumdef-mloperandtype"}], "title": "4. Privacy Considerations"}, {"refs": [{"id": "ref-for-enumdef-mloperandtype\u2460"}], "title": "7.3.1. The MLOperandDescriptor dictionary"}, {"refs": [{"id": "ref-for-enumdef-mloperandtype\u2461"}], "title": "7.6. The MLGraphBuilder interface"}, {"refs": [{"id": "ref-for-enumdef-mloperandtype\u2462"}], "title": "7.6.4.2. The constant(value, type) method"}, {"refs": [{"id": "ref-for-enumdef-mloperandtype\u2463"}, {"id": "ref-for-enumdef-mloperandtype\u2464"}], "title": "9.1. MLOperandType and ArrayBufferView compatibility"}], "external": false};
 window.dfnpanelData['dom-mloperandtype-float32'] = {"dfnID": "dom-mloperandtype-float32", "url": "#dom-mloperandtype-float32", "dfnText": "\"float32\"", "refSections": [{"refs": [{"id": "ref-for-dom-mloperandtype-float32"}], "title": "9.1. MLOperandType and ArrayBufferView compatibility"}], "external": false};
 window.dfnpanelData['dom-mloperandtype-float16'] = {"dfnID": "dom-mloperandtype-float16", "url": "#dom-mloperandtype-float16", "dfnText": "\"float16\"", "refSections": [{"refs": [{"id": "ref-for-dom-mloperandtype-float16"}], "title": "9.1. MLOperandType and ArrayBufferView compatibility"}], "external": false};
 window.dfnpanelData['dom-mloperandtype-int32'] = {"dfnID": "dom-mloperandtype-int32", "url": "#dom-mloperandtype-int32", "dfnText": "\"int32\"", "refSections": [{"refs": [{"id": "ref-for-dom-mloperandtype-int32"}], "title": "9.1. MLOperandType and ArrayBufferView compatibility"}], "external": false};
 window.dfnpanelData['dom-mloperandtype-uint32'] = {"dfnID": "dom-mloperandtype-uint32", "url": "#dom-mloperandtype-uint32", "dfnText": "\"uint32\"", "refSections": [{"refs": [{"id": "ref-for-dom-mloperandtype-uint32"}], "title": "9.1. MLOperandType and ArrayBufferView compatibility"}], "external": false};
 window.dfnpanelData['dom-mloperandtype-int8'] = {"dfnID": "dom-mloperandtype-int8", "url": "#dom-mloperandtype-int8", "dfnText": "\"int8\"", "refSections": [{"refs": [{"id": "ref-for-dom-mloperandtype-int8"}], "title": "9.1. MLOperandType and ArrayBufferView compatibility"}], "external": false};
 window.dfnpanelData['dom-mloperandtype-uint8'] = {"dfnID": "dom-mloperandtype-uint8", "url": "#dom-mloperandtype-uint8", "dfnText": "\"uint8\"", "refSections": [{"refs": [{"id": "ref-for-dom-mloperandtype-uint8"}], "title": "9.1. MLOperandType and ArrayBufferView compatibility"}], "external": false};
-window.dfnpanelData['dictdef-mloperanddescriptor'] = {"dfnID": "dictdef-mloperanddescriptor", "url": "#dictdef-mloperanddescriptor", "dfnText": "MLOperandDescriptor", "refSections": [{"refs": [{"id": "ref-for-dictdef-mloperanddescriptor"}, {"id": "ref-for-dictdef-mloperanddescriptor\u2460"}, {"id": "ref-for-dictdef-mloperanddescriptor\u2461"}, {"id": "ref-for-dictdef-mloperanddescriptor\u2462"}], "title": "7.3. The MLGraph interface"}, {"refs": [{"id": "ref-for-dictdef-mloperanddescriptor\u2463"}], "title": "7.3.1. The MLOperandDescriptor dictionary"}, {"refs": [{"id": "ref-for-dictdef-mloperanddescriptor\u2464"}], "title": "7.3.2. The MLOperand interface"}, {"refs": [{"id": "ref-for-dictdef-mloperanddescriptor\u2465"}], "title": "7.3.2.1. Creating MLOperand"}, {"refs": [{"id": "ref-for-dictdef-mloperanddescriptor\u2466"}, {"id": "ref-for-dictdef-mloperanddescriptor\u2467"}], "title": "7.6. The MLGraphBuilder interface"}, {"refs": [{"id": "ref-for-dictdef-mloperanddescriptor\u2468"}, {"id": "ref-for-dictdef-mloperanddescriptor\u2460\u24ea"}], "title": "7.6.2. The input() method"}, {"refs": [{"id": "ref-for-dictdef-mloperanddescriptor\u2460\u2460"}, {"id": "ref-for-dictdef-mloperanddescriptor\u2460\u2461"}], "title": "7.6.3.1. The constant(descriptor, bufferView) method"}, {"refs": [{"id": "ref-for-dictdef-mloperanddescriptor\u2460\u2462"}], "title": "7.6.3.2. The constant(value, type) method"}], "external": false};
-window.dfnpanelData['dom-mloperanddescriptor-type'] = {"dfnID": "dom-mloperanddescriptor-type", "url": "#dom-mloperanddescriptor-type", "dfnText": "type", "refSections": [{"refs": [{"id": "ref-for-dom-mloperanddescriptor-type"}], "title": "7.3.1. The MLOperandDescriptor dictionary"}, {"refs": [{"id": "ref-for-dom-mloperanddescriptor-type\u2460"}], "title": "7.4.2. Synchronous Execution"}, {"refs": [{"id": "ref-for-dom-mloperanddescriptor-type\u2461"}], "title": "7.6.2. The input() method"}, {"refs": [{"id": "ref-for-dom-mloperanddescriptor-type\u2462"}], "title": "7.6.3.1. The constant(descriptor, bufferView) method"}, {"refs": [{"id": "ref-for-dom-mloperanddescriptor-type\u2463"}], "title": "7.6.3.2. The constant(value, type) method"}], "external": false};
-window.dfnpanelData['dom-mloperanddescriptor-dimensions'] = {"dfnID": "dom-mloperanddescriptor-dimensions", "url": "#dom-mloperanddescriptor-dimensions", "dfnText": "dimensions", "refSections": [{"refs": [{"id": "ref-for-dom-mloperanddescriptor-dimensions"}], "title": "7.3.1. The MLOperandDescriptor dictionary"}, {"refs": [{"id": "ref-for-dom-mloperanddescriptor-dimensions\u2460"}], "title": "7.4.2. Synchronous Execution"}, {"refs": [{"id": "ref-for-dom-mloperanddescriptor-dimensions\u2461"}, {"id": "ref-for-dom-mloperanddescriptor-dimensions\u2462"}, {"id": "ref-for-dom-mloperanddescriptor-dimensions\u2463"}], "title": "7.6.2. The input() method"}, {"refs": [{"id": "ref-for-dom-mloperanddescriptor-dimensions\u2464"}], "title": "7.6.3.1. The constant(descriptor, bufferView) method"}, {"refs": [{"id": "ref-for-dom-mloperanddescriptor-dimensions\u2465"}, {"id": "ref-for-dom-mloperanddescriptor-dimensions\u2466"}], "title": "7.6.3.2. The constant(value, type) method"}], "external": false};
-window.dfnpanelData['mloperanddescriptor-byte-length'] = {"dfnID": "mloperanddescriptor-byte-length", "url": "#mloperanddescriptor-byte-length", "dfnText": "byte length", "refSections": [{"refs": [{"id": "ref-for-mloperanddescriptor-byte-length"}, {"id": "ref-for-mloperanddescriptor-byte-length\u2460"}], "title": "7.4.2. Synchronous Execution"}, {"refs": [{"id": "ref-for-mloperanddescriptor-byte-length\u2461"}, {"id": "ref-for-mloperanddescriptor-byte-length\u2462"}], "title": "7.5.2. Dispatch Execution Commands"}, {"refs": [{"id": "ref-for-mloperanddescriptor-byte-length\u2463"}], "title": "7.6.2. The input() method"}, {"refs": [{"id": "ref-for-mloperanddescriptor-byte-length\u2464"}], "title": "7.6.3.1. The constant(descriptor, bufferView) method"}], "external": false};
-window.dfnpanelData['mloperand'] = {"dfnID": "mloperand", "url": "#mloperand", "dfnText": "MLOperand", "refSections": [{"refs": [{"id": "ref-for-mloperand"}, {"id": "ref-for-mloperand\u2460"}, {"id": "ref-for-mloperand\u2461"}, {"id": "ref-for-mloperand\u2462"}, {"id": "ref-for-mloperand\u2463"}, {"id": "ref-for-mloperand\u2464"}, {"id": "ref-for-mloperand\u2465"}], "title": "6.1. Overview"}, {"refs": [{"id": "ref-for-mloperand\u2466"}, {"id": "ref-for-mloperand\u2467"}, {"id": "ref-for-mloperand\u2468"}, {"id": "ref-for-mloperand\u2460\u24ea"}], "title": "7.3. The MLGraph interface"}, {"refs": [{"id": "ref-for-mloperand\u2460\u2460"}, {"id": "ref-for-mloperand\u2460\u2461"}, {"id": "ref-for-mloperand\u2460\u2462"}, {"id": "ref-for-mloperand\u2460\u2463"}, {"id": "ref-for-mloperand\u2460\u2464"}, {"id": "ref-for-mloperand\u2460\u2465"}, {"id": "ref-for-mloperand\u2460\u2466"}, {"id": "ref-for-mloperand\u2460\u2467"}, {"id": "ref-for-mloperand\u2460\u2468"}], "title": "7.3.2. The MLOperand interface"}, {"refs": [{"id": "ref-for-mloperand\u2461\u24ea"}, {"id": "ref-for-mloperand\u2461\u2460"}], "title": "7.3.2.1. Creating MLOperand"}, {"refs": [{"id": "ref-for-mloperand\u2461\u2461"}, {"id": "ref-for-mloperand\u2461\u2462"}, {"id": "ref-for-mloperand\u2461\u2463"}, {"id": "ref-for-mloperand\u2461\u2464"}], "title": "7.6. The MLGraphBuilder interface"}, {"refs": [{"id": "ref-for-mloperand\u2461\u2465"}, {"id": "ref-for-mloperand\u2461\u2466"}], "title": "7.6.2. The input() method"}, {"refs": [{"id": "ref-for-mloperand\u2461\u2467"}], "title": "7.6.3. The constant() method"}, {"refs": [{"id": "ref-for-mloperand\u2461\u2468"}], "title": "7.6.3.1. The constant(descriptor, bufferView) method"}, {"refs": [{"id": "ref-for-mloperand\u2462\u24ea"}], "title": "7.6.3.2. The constant(value, type) method"}, {"refs": [{"id": "ref-for-mloperand\u2462\u2460"}, {"id": "ref-for-mloperand\u2462\u2461"}, {"id": "ref-for-mloperand\u2462\u2462"}, {"id": "ref-for-mloperand\u2462\u2463"}, {"id": "ref-for-mloperand\u2462\u2464"}, {"id": "ref-for-mloperand\u2462\u2465"}, {"id": "ref-for-mloperand\u2462\u2466"}, {"id": "ref-for-mloperand\u2462\u2467"}, {"id": "ref-for-mloperand\u2462\u2468"}, {"id": "ref-for-mloperand\u2463\u24ea"}, {"id": "ref-for-mloperand\u2463\u2460"}, {"id": "ref-for-mloperand\u2463\u2461"}], "title": "7.6.4. The batchNormalization() method"}, {"refs": [{"id": "ref-for-mloperand\u2463\u2462"}, {"id": "ref-for-mloperand\u2463\u2463"}, {"id": "ref-for-mloperand\u2463\u2464"}, {"id": "ref-for-mloperand\u2463\u2465"}], "title": "7.6.5. The clamp() method"}, {"refs": [{"id": "ref-for-mloperand\u2463\u2466"}, {"id": "ref-for-mloperand\u2463\u2467"}, {"id": "ref-for-mloperand\u2463\u2468"}, {"id": "ref-for-mloperand\u2464\u24ea"}], "title": "7.6.6. The concat() method"}, {"refs": [{"id": "ref-for-mloperand\u2464\u2460"}, {"id": "ref-for-mloperand\u2464\u2461"}, {"id": "ref-for-mloperand\u2464\u2462"}, {"id": "ref-for-mloperand\u2464\u2463"}, {"id": "ref-for-mloperand\u2464\u2464"}, {"id": "ref-for-mloperand\u2464\u2465"}, {"id": "ref-for-mloperand\u2464\u2466"}, {"id": "ref-for-mloperand\u2464\u2467"}], "title": "7.6.7. The conv2d() method"}, {"refs": [{"id": "ref-for-mloperand\u2464\u2468"}, {"id": "ref-for-mloperand\u2465\u24ea"}, {"id": "ref-for-mloperand\u2465\u2460"}, {"id": "ref-for-mloperand\u2465\u2461"}, {"id": "ref-for-mloperand\u2465\u2462"}, {"id": "ref-for-mloperand\u2465\u2463"}, {"id": "ref-for-mloperand\u2465\u2464"}, {"id": "ref-for-mloperand\u2465\u2465"}], "title": "7.6.8. The convTranspose2d() method"}, {"refs": [{"id": "ref-for-mloperand\u2465\u2466"}, {"id": "ref-for-mloperand\u2465\u2467"}, {"id": "ref-for-mloperand\u2465\u2468"}, {"id": "ref-for-mloperand\u2466\u24ea"}, {"id": "ref-for-mloperand\u2466\u2460"}, {"id": "ref-for-mloperand\u2466\u2461"}, {"id": "ref-for-mloperand\u2466\u2462"}, {"id": "ref-for-mloperand\u2466\u2463"}, {"id": "ref-for-mloperand\u2466\u2464"}, {"id": "ref-for-mloperand\u2466\u2465"}, {"id": "ref-for-mloperand\u2466\u2466"}, {"id": "ref-for-mloperand\u2466\u2467"}, {"id": "ref-for-mloperand\u2466\u2468"}, {"id": "ref-for-mloperand\u2467\u24ea"}, {"id": "ref-for-mloperand\u2467\u2460"}, {"id": "ref-for-mloperand\u2467\u2461"}, {"id": "ref-for-mloperand\u2467\u2462"}, {"id": "ref-for-mloperand\u2467\u2463"}, {"id": "ref-for-mloperand\u2467\u2464"}, {"id": "ref-for-mloperand\u2467\u2465"}, {"id": "ref-for-mloperand\u2467\u2466"}, {"id": "ref-for-mloperand\u2467\u2467"}, {"id": "ref-for-mloperand\u2467\u2468"}, {"id": "ref-for-mloperand\u2468\u24ea"}], "title": "7.6.9. Element-wise binary operations"}, {"refs": [{"id": "ref-for-mloperand\u2468\u2460"}, {"id": "ref-for-mloperand\u2468\u2461"}, {"id": "ref-for-mloperand\u2468\u2462"}, {"id": "ref-for-mloperand\u2468\u2463"}, {"id": "ref-for-mloperand\u2468\u2464"}, {"id": "ref-for-mloperand\u2468\u2465"}, {"id": "ref-for-mloperand\u2468\u2466"}, {"id": "ref-for-mloperand\u2468\u2467"}, {"id": "ref-for-mloperand\u2468\u2468"}, {"id": "ref-for-mloperand\u2460\u24ea\u24ea"}, {"id": "ref-for-mloperand\u2460\u24ea\u2460"}, {"id": "ref-for-mloperand\u2460\u24ea\u2461"}, {"id": "ref-for-mloperand\u2460\u24ea\u2462"}, {"id": "ref-for-mloperand\u2460\u24ea\u2463"}, {"id": "ref-for-mloperand\u2460\u24ea\u2464"}, {"id": "ref-for-mloperand\u2460\u24ea\u2465"}, {"id": "ref-for-mloperand\u2460\u24ea\u2466"}, {"id": "ref-for-mloperand\u2460\u24ea\u2467"}, {"id": "ref-for-mloperand\u2460\u24ea\u2468"}, {"id": "ref-for-mloperand\u2460\u2460\u24ea"}], "title": "7.6.10. Element-wise unary operations"}, {"refs": [{"id": "ref-for-mloperand\u2460\u2460\u2460"}, {"id": "ref-for-mloperand\u2460\u2460\u2461"}, {"id": "ref-for-mloperand\u2460\u2460\u2462"}, {"id": "ref-for-mloperand\u2460\u2460\u2463"}], "title": "7.6.11. The elu() method"}, {"refs": [{"id": "ref-for-mloperand\u2460\u2460\u2464"}, {"id": "ref-for-mloperand\u2460\u2460\u2465"}, {"id": "ref-for-mloperand\u2460\u2460\u2466"}, {"id": "ref-for-mloperand\u2460\u2460\u2467"}, {"id": "ref-for-mloperand\u2460\u2460\u2468"}, {"id": "ref-for-mloperand\u2460\u2461\u24ea"}, {"id": "ref-for-mloperand\u2460\u2461\u2460"}, {"id": "ref-for-mloperand\u2460\u2461\u2461"}], "title": "7.6.12. The gemm() method"}, {"refs": [{"id": "ref-for-mloperand\u2460\u2461\u2462"}, {"id": "ref-for-mloperand\u2460\u2461\u2463"}, {"id": "ref-for-mloperand\u2460\u2461\u2464"}, {"id": "ref-for-mloperand\u2460\u2461\u2465"}, {"id": "ref-for-mloperand\u2460\u2461\u2466"}, {"id": "ref-for-mloperand\u2460\u2461\u2467"}, {"id": "ref-for-mloperand\u2460\u2461\u2468"}, {"id": "ref-for-mloperand\u2460\u2462\u24ea"}, {"id": "ref-for-mloperand\u2460\u2462\u2460"}, {"id": "ref-for-mloperand\u2460\u2462\u2461"}, {"id": "ref-for-mloperand\u2460\u2462\u2462"}, {"id": "ref-for-mloperand\u2460\u2462\u2463"}, {"id": "ref-for-mloperand\u2460\u2462\u2464"}, {"id": "ref-for-mloperand\u2460\u2462\u2465"}], "title": "7.6.13. The gru() method"}, {"refs": [{"id": "ref-for-mloperand\u2460\u2462\u2466"}, {"id": "ref-for-mloperand\u2460\u2462\u2467"}, {"id": "ref-for-mloperand\u2460\u2462\u2468"}, {"id": "ref-for-mloperand\u2460\u2463\u24ea"}, {"id": "ref-for-mloperand\u2460\u2463\u2460"}, {"id": "ref-for-mloperand\u2460\u2463\u2461"}, {"id": "ref-for-mloperand\u2460\u2463\u2462"}, {"id": "ref-for-mloperand\u2460\u2463\u2463"}, {"id": "ref-for-mloperand\u2460\u2463\u2464"}, {"id": "ref-for-mloperand\u2460\u2463\u2465"}, {"id": "ref-for-mloperand\u2460\u2463\u2466"}, {"id": "ref-for-mloperand\u2460\u2463\u2467"}, {"id": "ref-for-mloperand\u2460\u2463\u2468"}, {"id": "ref-for-mloperand\u2460\u2464\u24ea"}], "title": "7.6.14. The gruCell() method"}, {"refs": [{"id": "ref-for-mloperand\u2460\u2464\u2460"}, {"id": "ref-for-mloperand\u2460\u2464\u2461"}, {"id": "ref-for-mloperand\u2460\u2464\u2462"}, {"id": "ref-for-mloperand\u2460\u2464\u2463"}], "title": "7.6.15. The hardSigmoid() method"}, {"refs": [{"id": "ref-for-mloperand\u2460\u2464\u2464"}, {"id": "ref-for-mloperand\u2460\u2464\u2465"}, {"id": "ref-for-mloperand\u2460\u2464\u2466"}, {"id": "ref-for-mloperand\u2460\u2464\u2467"}], "title": "7.6.16. The hardSwish() method"}, {"refs": [{"id": "ref-for-mloperand\u2460\u2464\u2468"}, {"id": "ref-for-mloperand\u2460\u2465\u24ea"}, {"id": "ref-for-mloperand\u2460\u2465\u2460"}, {"id": "ref-for-mloperand\u2460\u2465\u2461"}, {"id": "ref-for-mloperand\u2460\u2465\u2462"}, {"id": "ref-for-mloperand\u2460\u2465\u2463"}, {"id": "ref-for-mloperand\u2460\u2465\u2464"}, {"id": "ref-for-mloperand\u2460\u2465\u2465"}], "title": "7.6.17. The instanceNormalization() method"}, {"refs": [{"id": "ref-for-mloperand\u2460\u2465\u2466"}, {"id": "ref-for-mloperand\u2460\u2465\u2467"}, {"id": "ref-for-mloperand\u2460\u2465\u2468"}, {"id": "ref-for-mloperand\u2460\u2466\u24ea"}], "title": "7.6.18. The leakyRelu() method"}, {"refs": [{"id": "ref-for-mloperand\u2460\u2466\u2460"}, {"id": "ref-for-mloperand\u2460\u2466\u2461"}, {"id": "ref-for-mloperand\u2460\u2466\u2462"}, {"id": "ref-for-mloperand\u2460\u2466\u2463"}], "title": "7.6.19. The linear() method"}, {"refs": [{"id": "ref-for-mloperand\u2460\u2466\u2464"}, {"id": "ref-for-mloperand\u2460\u2466\u2465"}, {"id": "ref-for-mloperand\u2460\u2466\u2466"}, {"id": "ref-for-mloperand\u2460\u2466\u2467"}, {"id": "ref-for-mloperand\u2460\u2466\u2468"}, {"id": "ref-for-mloperand\u2460\u2467\u24ea"}, {"id": "ref-for-mloperand\u2460\u2467\u2460"}, {"id": "ref-for-mloperand\u2460\u2467\u2461"}, {"id": "ref-for-mloperand\u2460\u2467\u2462"}, {"id": "ref-for-mloperand\u2460\u2467\u2463"}, {"id": "ref-for-mloperand\u2460\u2467\u2464"}, {"id": "ref-for-mloperand\u2460\u2467\u2465"}, {"id": "ref-for-mloperand\u2460\u2467\u2466"}, {"id": "ref-for-mloperand\u2460\u2467\u2467"}, {"id": "ref-for-mloperand\u2460\u2467\u2468"}, {"id": "ref-for-mloperand\u2460\u2468\u24ea"}, {"id": "ref-for-mloperand\u2460\u2468\u2460"}, {"id": "ref-for-mloperand\u2460\u2468\u2461"}], "title": "7.6.20. The lstm() method"}, {"refs": [{"id": "ref-for-mloperand\u2460\u2468\u2462"}, {"id": "ref-for-mloperand\u2460\u2468\u2463"}, {"id": "ref-for-mloperand\u2460\u2468\u2464"}, {"id": "ref-for-mloperand\u2460\u2468\u2465"}, {"id": "ref-for-mloperand\u2460\u2468\u2466"}, {"id": "ref-for-mloperand\u2460\u2468\u2467"}, {"id": "ref-for-mloperand\u2460\u2468\u2468"}, {"id": "ref-for-mloperand\u2461\u24ea\u24ea"}, {"id": "ref-for-mloperand\u2461\u24ea\u2460"}, {"id": "ref-for-mloperand\u2461\u24ea\u2461"}, {"id": "ref-for-mloperand\u2461\u24ea\u2462"}, {"id": "ref-for-mloperand\u2461\u24ea\u2463"}, {"id": "ref-for-mloperand\u2461\u24ea\u2464"}, {"id": "ref-for-mloperand\u2461\u24ea\u2465"}, {"id": "ref-for-mloperand\u2461\u24ea\u2466"}, {"id": "ref-for-mloperand\u2461\u24ea\u2467"}, {"id": "ref-for-mloperand\u2461\u24ea\u2468"}, {"id": "ref-for-mloperand\u2461\u2460\u24ea"}], "title": "7.6.21. The lstmCell() method"}, {"refs": [{"id": "ref-for-mloperand\u2461\u2460\u2460"}, {"id": "ref-for-mloperand\u2461\u2460\u2461"}, {"id": "ref-for-mloperand\u2461\u2460\u2462"}, {"id": "ref-for-mloperand\u2461\u2460\u2463"}, {"id": "ref-for-mloperand\u2461\u2460\u2464"}, {"id": "ref-for-mloperand\u2461\u2460\u2465"}], "title": "7.6.22. The matmul() method"}, {"refs": [{"id": "ref-for-mloperand\u2461\u2460\u2466"}, {"id": "ref-for-mloperand\u2461\u2460\u2467"}, {"id": "ref-for-mloperand\u2461\u2460\u2468"}, {"id": "ref-for-mloperand\u2461\u2461\u24ea"}], "title": "7.6.23. The pad() method"}, {"refs": [{"id": "ref-for-mloperand\u2461\u2461\u2460"}, {"id": "ref-for-mloperand\u2461\u2461\u2461"}, {"id": "ref-for-mloperand\u2461\u2461\u2462"}, {"id": "ref-for-mloperand\u2461\u2461\u2463"}, {"id": "ref-for-mloperand\u2461\u2461\u2464"}, {"id": "ref-for-mloperand\u2461\u2461\u2465"}, {"id": "ref-for-mloperand\u2461\u2461\u2466"}, {"id": "ref-for-mloperand\u2461\u2461\u2467"}], "title": "7.6.24. Pooling operations"}, {"refs": [{"id": "ref-for-mloperand\u2461\u2461\u2468"}, {"id": "ref-for-mloperand\u2461\u2462\u24ea"}, {"id": "ref-for-mloperand\u2461\u2462\u2460"}, {"id": "ref-for-mloperand\u2461\u2462\u2461"}, {"id": "ref-for-mloperand\u2461\u2462\u2462"}, {"id": "ref-for-mloperand\u2461\u2462\u2463"}], "title": "7.6.25. The prelu() method"}, {"refs": [{"id": "ref-for-mloperand\u2461\u2462\u2464"}, {"id": "ref-for-mloperand\u2461\u2462\u2465"}, {"id": "ref-for-mloperand\u2461\u2462\u2466"}, {"id": "ref-for-mloperand\u2461\u2462\u2467"}, {"id": "ref-for-mloperand\u2461\u2462\u2468"}, {"id": "ref-for-mloperand\u2461\u2463\u24ea"}, {"id": "ref-for-mloperand\u2461\u2463\u2460"}, {"id": "ref-for-mloperand\u2461\u2463\u2461"}, {"id": "ref-for-mloperand\u2461\u2463\u2462"}, {"id": "ref-for-mloperand\u2461\u2463\u2463"}, {"id": "ref-for-mloperand\u2461\u2463\u2464"}, {"id": "ref-for-mloperand\u2461\u2463\u2465"}, {"id": "ref-for-mloperand\u2461\u2463\u2466"}, {"id": "ref-for-mloperand\u2461\u2463\u2467"}, {"id": "ref-for-mloperand\u2461\u2463\u2468"}, {"id": "ref-for-mloperand\u2461\u2464\u24ea"}, {"id": "ref-for-mloperand\u2461\u2464\u2460"}, {"id": "ref-for-mloperand\u2461\u2464\u2461"}, {"id": "ref-for-mloperand\u2461\u2464\u2462"}, {"id": "ref-for-mloperand\u2461\u2464\u2463"}, {"id": "ref-for-mloperand\u2461\u2464\u2464"}, {"id": "ref-for-mloperand\u2461\u2464\u2465"}], "title": "7.6.26. Reduction operations"}, {"refs": [{"id": "ref-for-mloperand\u2461\u2464\u2466"}, {"id": "ref-for-mloperand\u2461\u2464\u2467"}, {"id": "ref-for-mloperand\u2461\u2464\u2468"}, {"id": "ref-for-mloperand\u2461\u2465\u24ea"}], "title": "7.6.27. The relu() method"}, {"refs": [{"id": "ref-for-mloperand\u2461\u2465\u2460"}, {"id": "ref-for-mloperand\u2461\u2465\u2461"}, {"id": "ref-for-mloperand\u2461\u2465\u2462"}, {"id": "ref-for-mloperand\u2461\u2465\u2463"}], "title": "7.6.28. The resample2d() method"}, {"refs": [{"id": "ref-for-mloperand\u2461\u2465\u2464"}, {"id": "ref-for-mloperand\u2461\u2465\u2465"}, {"id": "ref-for-mloperand\u2461\u2465\u2466"}, {"id": "ref-for-mloperand\u2461\u2465\u2467"}], "title": "7.6.29. The reshape() method"}, {"refs": [{"id": "ref-for-mloperand\u2461\u2465\u2468"}, {"id": "ref-for-mloperand\u2461\u2466\u24ea"}, {"id": "ref-for-mloperand\u2461\u2466\u2460"}, {"id": "ref-for-mloperand\u2461\u2466\u2461"}], "title": "7.6.30. The sigmoid() method"}, {"refs": [{"id": "ref-for-mloperand\u2461\u2466\u2462"}, {"id": "ref-for-mloperand\u2461\u2466\u2463"}, {"id": "ref-for-mloperand\u2461\u2466\u2464"}, {"id": "ref-for-mloperand\u2461\u2466\u2465"}], "title": "7.6.31. The slice() method"}, {"refs": [{"id": "ref-for-mloperand\u2461\u2466\u2466"}, {"id": "ref-for-mloperand\u2461\u2466\u2467"}, {"id": "ref-for-mloperand\u2461\u2466\u2468"}, {"id": "ref-for-mloperand\u2461\u2467\u24ea"}], "title": "7.6.32. The softmax() method"}, {"refs": [{"id": "ref-for-mloperand\u2461\u2467\u2460"}, {"id": "ref-for-mloperand\u2461\u2467\u2461"}, {"id": "ref-for-mloperand\u2461\u2467\u2462"}, {"id": "ref-for-mloperand\u2461\u2467\u2463"}], "title": "7.6.33. The softplus() method"}, {"refs": [{"id": "ref-for-mloperand\u2461\u2467\u2464"}, {"id": "ref-for-mloperand\u2461\u2467\u2465"}, {"id": "ref-for-mloperand\u2461\u2467\u2466"}, {"id": "ref-for-mloperand\u2461\u2467\u2467"}], "title": "7.6.34. The softsign() method"}, {"refs": [{"id": "ref-for-mloperand\u2461\u2467\u2468"}, {"id": "ref-for-mloperand\u2461\u2468\u24ea"}, {"id": "ref-for-mloperand\u2461\u2468\u2460"}, {"id": "ref-for-mloperand\u2461\u2468\u2461"}], "title": "7.6.35. The split() method"}, {"refs": [{"id": "ref-for-mloperand\u2461\u2468\u2462"}, {"id": "ref-for-mloperand\u2461\u2468\u2463"}, {"id": "ref-for-mloperand\u2461\u2468\u2464"}, {"id": "ref-for-mloperand\u2461\u2468\u2465"}], "title": "7.6.36. The squeeze() method"}, {"refs": [{"id": "ref-for-mloperand\u2461\u2468\u2466"}, {"id": "ref-for-mloperand\u2461\u2468\u2467"}, {"id": "ref-for-mloperand\u2461\u2468\u2468"}, {"id": "ref-for-mloperand\u2462\u24ea\u24ea"}], "title": "7.6.37. The tanh() method"}, {"refs": [{"id": "ref-for-mloperand\u2462\u24ea\u2460"}, {"id": "ref-for-mloperand\u2462\u24ea\u2461"}, {"id": "ref-for-mloperand\u2462\u24ea\u2462"}, {"id": "ref-for-mloperand\u2462\u24ea\u2463"}], "title": "7.6.38. The transpose() method"}], "external": false};
-window.dfnpanelData['dom-mloperand-builder-slot'] = {"dfnID": "dom-mloperand-builder-slot", "url": "#dom-mloperand-builder-slot", "dfnText": "[[builder]]", "refSections": [{"refs": [{"id": "ref-for-dom-mloperand-builder-slot"}], "title": "7.3.2. The MLOperand interface"}, {"refs": [{"id": "ref-for-dom-mloperand-builder-slot\u2460"}], "title": "7.3.2.1. Creating MLOperand"}], "external": false};
-window.dfnpanelData['dom-mloperand-descriptor-slot'] = {"dfnID": "dom-mloperand-descriptor-slot", "url": "#dom-mloperand-descriptor-slot", "dfnText": "[[descriptor]]", "refSections": [{"refs": [{"id": "ref-for-dom-mloperand-descriptor-slot"}], "title": "7.3.2.1. Creating MLOperand"}], "external": false};
-window.dfnpanelData['dom-mloperand-name-slot'] = {"dfnID": "dom-mloperand-name-slot", "url": "#dom-mloperand-name-slot", "dfnText": "[[name]]", "refSections": [{"refs": [{"id": "ref-for-dom-mloperand-name-slot"}], "title": "7.6.2. The input() method"}], "external": false};
-window.dfnpanelData['dom-mloperand-operand-slot'] = {"dfnID": "dom-mloperand-operand-slot", "url": "#dom-mloperand-operand-slot", "dfnText": "[[operand]]", "refSections": [{"refs": [{"id": "ref-for-dom-mloperand-operand-slot"}], "title": "7.6.2. The input() method"}, {"refs": [{"id": "ref-for-dom-mloperand-operand-slot\u2460"}], "title": "7.6.3.1. The constant(descriptor, bufferView) method"}, {"refs": [{"id": "ref-for-dom-mloperand-operand-slot\u2461"}], "title": "7.6.3.2. The constant(value, type) method"}], "external": false};
-window.dfnpanelData['create-mloperand'] = {"dfnID": "create-mloperand", "url": "#create-mloperand", "dfnText": "create MLOperand", "refSections": [{"refs": [{"id": "ref-for-create-mloperand"}], "title": "7.6.2. The input() method"}, {"refs": [{"id": "ref-for-create-mloperand\u2460"}], "title": "7.6.3.1. The constant(descriptor, bufferView) method"}, {"refs": [{"id": "ref-for-create-mloperand\u2461"}], "title": "7.6.3.2. The constant(value, type) method"}], "external": false};
-window.dfnpanelData['check-dimensions'] = {"dfnID": "check-dimensions", "url": "#check-dimensions", "dfnText": "check dimensions", "refSections": [{"refs": [{"id": "ref-for-check-dimensions"}], "title": "7.6.2. The input() method"}, {"refs": [{"id": "ref-for-check-dimensions\u2460"}], "title": "7.6.3.1. The constant(descriptor, bufferView) method"}], "external": false};
-window.dfnpanelData['mlactivation'] = {"dfnID": "mlactivation", "url": "#mlactivation", "dfnText": "MLActivation", "refSections": [{"refs": [{"id": "ref-for-mlactivation"}, {"id": "ref-for-mlactivation\u2460"}], "title": "7.3.3. The MLActivation interface"}, {"refs": [{"id": "ref-for-mlactivation\u2461"}, {"id": "ref-for-mlactivation\u2462"}], "title": "7.6.4. The batchNormalization() method"}, {"refs": [{"id": "ref-for-mlactivation\u2463"}, {"id": "ref-for-mlactivation\u2464"}], "title": "7.6.5. The clamp() method"}, {"refs": [{"id": "ref-for-mlactivation\u2465"}, {"id": "ref-for-mlactivation\u2466"}], "title": "7.6.7. The conv2d() method"}, {"refs": [{"id": "ref-for-mlactivation\u2467"}, {"id": "ref-for-mlactivation\u2468"}], "title": "7.6.8. The convTranspose2d() method"}, {"refs": [{"id": "ref-for-mlactivation\u2460\u24ea"}, {"id": "ref-for-mlactivation\u2460\u2460"}], "title": "7.6.11. The elu() method"}, {"refs": [{"id": "ref-for-mlactivation\u2460\u2461"}, {"id": "ref-for-mlactivation\u2460\u2462"}], "title": "7.6.13. The gru() method"}, {"refs": [{"id": "ref-for-mlactivation\u2460\u2463"}, {"id": "ref-for-mlactivation\u2460\u2464"}], "title": "7.6.14. The gruCell() method"}, {"refs": [{"id": "ref-for-mlactivation\u2460\u2465"}, {"id": "ref-for-mlactivation\u2460\u2466"}], "title": "7.6.15. The hardSigmoid() method"}, {"refs": [{"id": "ref-for-mlactivation\u2460\u2467"}, {"id": "ref-for-mlactivation\u2460\u2468"}], "title": "7.6.16. The hardSwish() method"}, {"refs": [{"id": "ref-for-mlactivation\u2461\u24ea"}, {"id": "ref-for-mlactivation\u2461\u2460"}], "title": "7.6.18. The leakyRelu() method"}, {"refs": [{"id": "ref-for-mlactivation\u2461\u2461"}, {"id": "ref-for-mlactivation\u2461\u2462"}], "title": "7.6.19. The linear() method"}, {"refs": [{"id": "ref-for-mlactivation\u2461\u2463"}, {"id": "ref-for-mlactivation\u2461\u2464"}], "title": "7.6.20. The lstm() method"}, {"refs": [{"id": "ref-for-mlactivation\u2461\u2465"}, {"id": "ref-for-mlactivation\u2461\u2466"}], "title": "7.6.21. The lstmCell() method"}, {"refs": [{"id": "ref-for-mlactivation\u2461\u2467"}, {"id": "ref-for-mlactivation\u2461\u2468"}], "title": "7.6.27. The relu() method"}, {"refs": [{"id": "ref-for-mlactivation\u2462\u24ea"}, {"id": "ref-for-mlactivation\u2462\u2460"}], "title": "7.6.30. The sigmoid() method"}, {"refs": [{"id": "ref-for-mlactivation\u2462\u2461"}, {"id": "ref-for-mlactivation\u2462\u2462"}], "title": "7.6.32. The softmax() method"}, {"refs": [{"id": "ref-for-mlactivation\u2462\u2463"}, {"id": "ref-for-mlactivation\u2462\u2464"}], "title": "7.6.33. The softplus() method"}, {"refs": [{"id": "ref-for-mlactivation\u2462\u2465"}, {"id": "ref-for-mlactivation\u2462\u2466"}], "title": "7.6.34. The softsign() method"}, {"refs": [{"id": "ref-for-mlactivation\u2462\u2467"}, {"id": "ref-for-mlactivation\u2462\u2468"}], "title": "7.6.37. The tanh() method"}], "external": false};
+window.dfnpanelData['dictdef-mloperanddescriptor'] = {"dfnID": "dictdef-mloperanddescriptor", "url": "#dictdef-mloperanddescriptor", "dfnText": "MLOperandDescriptor", "refSections": [{"refs": [{"id": "ref-for-dictdef-mloperanddescriptor"}, {"id": "ref-for-dictdef-mloperanddescriptor\u2460"}, {"id": "ref-for-dictdef-mloperanddescriptor\u2461"}, {"id": "ref-for-dictdef-mloperanddescriptor\u2462"}], "title": "7.3. The MLGraph interface"}, {"refs": [{"id": "ref-for-dictdef-mloperanddescriptor\u2463"}], "title": "7.3.1. The MLOperandDescriptor dictionary"}, {"refs": [{"id": "ref-for-dictdef-mloperanddescriptor\u2464"}], "title": "7.3.2. The MLOperand interface"}, {"refs": [{"id": "ref-for-dictdef-mloperanddescriptor\u2465"}], "title": "7.3.2.1. Creating MLOperand"}, {"refs": [{"id": "ref-for-dictdef-mloperanddescriptor\u2466"}, {"id": "ref-for-dictdef-mloperanddescriptor\u2467"}], "title": "7.6. The MLGraphBuilder interface"}, {"refs": [{"id": "ref-for-dictdef-mloperanddescriptor\u2468"}, {"id": "ref-for-dictdef-mloperanddescriptor\u2460\u24ea"}], "title": "7.6.2. The input() method"}, {"refs": [{"id": "ref-for-dictdef-mloperanddescriptor\u2460\u2460"}, {"id": "ref-for-dictdef-mloperanddescriptor\u2460\u2461"}], "title": "7.6.4.1. The constant(descriptor, bufferView) method"}, {"refs": [{"id": "ref-for-dictdef-mloperanddescriptor\u2460\u2462"}], "title": "7.6.4.2. The constant(value, type) method"}, {"refs": [{"id": "ref-for-dictdef-mloperanddescriptor\u2460\u2463"}], "title": "7.6.8. The conv2d() method"}, {"refs": [{"id": "ref-for-dictdef-mloperanddescriptor\u2460\u2464"}], "title": "7.6.9. The convTranspose2d() method"}, {"refs": [{"id": "ref-for-dictdef-mloperanddescriptor\u2460\u2465"}], "title": "7.6.10. Element-wise binary operations"}, {"refs": [{"id": "ref-for-dictdef-mloperanddescriptor\u2460\u2466"}], "title": "7.6.13. The gemm() method"}, {"refs": [{"id": "ref-for-dictdef-mloperanddescriptor\u2460\u2467"}], "title": "7.6.15. The gruCell() method"}, {"refs": [{"id": "ref-for-dictdef-mloperanddescriptor\u2460\u2468"}], "title": "7.6.21. The lstm() method"}, {"refs": [{"id": "ref-for-dictdef-mloperanddescriptor\u2461\u24ea"}], "title": "7.6.22. The lstmCell() method"}, {"refs": [{"id": "ref-for-dictdef-mloperanddescriptor\u2461\u2460"}], "title": "7.6.23. The matmul() method"}, {"refs": [{"id": "ref-for-dictdef-mloperanddescriptor\u2461\u2461"}], "title": "7.6.26. The prelu() method"}, {"refs": [{"id": "ref-for-dictdef-mloperanddescriptor\u2461\u2462"}], "title": "7.6.29. The resample2d() method"}], "external": false};
+window.dfnpanelData['dom-mloperanddescriptor-type'] = {"dfnID": "dom-mloperanddescriptor-type", "url": "#dom-mloperanddescriptor-type", "dfnText": "type", "refSections": [{"refs": [{"id": "ref-for-dom-mloperanddescriptor-type"}], "title": "7.3.1. The MLOperandDescriptor dictionary"}, {"refs": [{"id": "ref-for-dom-mloperanddescriptor-type\u2460"}], "title": "7.3.2.1. Creating MLOperand"}, {"refs": [{"id": "ref-for-dom-mloperanddescriptor-type\u2461"}], "title": "7.4.2. Synchronous Execution"}, {"refs": [{"id": "ref-for-dom-mloperanddescriptor-type\u2462"}], "title": "7.6.2. The input() method"}, {"refs": [{"id": "ref-for-dom-mloperanddescriptor-type\u2463"}], "title": "7.6.4.1. The constant(descriptor, bufferView) method"}, {"refs": [{"id": "ref-for-dom-mloperanddescriptor-type\u2464"}], "title": "7.6.4.2. The constant(value, type) method"}, {"refs": [{"id": "ref-for-dom-mloperanddescriptor-type\u2465"}, {"id": "ref-for-dom-mloperanddescriptor-type\u2466"}, {"id": "ref-for-dom-mloperanddescriptor-type\u2467"}], "title": "7.6.7. The concat() method"}, {"refs": [{"id": "ref-for-dom-mloperanddescriptor-type\u2468"}, {"id": "ref-for-dom-mloperanddescriptor-type\u2460\u24ea"}, {"id": "ref-for-dom-mloperanddescriptor-type\u2460\u2460"}, {"id": "ref-for-dom-mloperanddescriptor-type\u2460\u2461"}, {"id": "ref-for-dom-mloperanddescriptor-type\u2460\u2462"}, {"id": "ref-for-dom-mloperanddescriptor-type\u2460\u2463"}], "title": "7.6.8. The conv2d() method"}, {"refs": [{"id": "ref-for-dom-mloperanddescriptor-type\u2460\u2464"}, {"id": "ref-for-dom-mloperanddescriptor-type\u2460\u2465"}, {"id": "ref-for-dom-mloperanddescriptor-type\u2460\u2466"}, {"id": "ref-for-dom-mloperanddescriptor-type\u2460\u2467"}, {"id": "ref-for-dom-mloperanddescriptor-type\u2460\u2468"}, {"id": "ref-for-dom-mloperanddescriptor-type\u2461\u24ea"}], "title": "7.6.9. The convTranspose2d() method"}, {"refs": [{"id": "ref-for-dom-mloperanddescriptor-type\u2461\u2460"}, {"id": "ref-for-dom-mloperanddescriptor-type\u2461\u2461"}, {"id": "ref-for-dom-mloperanddescriptor-type\u2461\u2462"}, {"id": "ref-for-dom-mloperanddescriptor-type\u2461\u2463"}], "title": "7.6.10. Element-wise binary operations"}, {"refs": [{"id": "ref-for-dom-mloperanddescriptor-type\u2461\u2464"}, {"id": "ref-for-dom-mloperanddescriptor-type\u2461\u2465"}], "title": "7.6.13. The gemm() method"}, {"refs": [{"id": "ref-for-dom-mloperanddescriptor-type\u2461\u2466"}, {"id": "ref-for-dom-mloperanddescriptor-type\u2461\u2467"}], "title": "7.6.15. The gruCell() method"}, {"refs": [{"id": "ref-for-dom-mloperanddescriptor-type\u2461\u2468"}, {"id": "ref-for-dom-mloperanddescriptor-type\u2462\u24ea"}], "title": "7.6.21. The lstm() method"}, {"refs": [{"id": "ref-for-dom-mloperanddescriptor-type\u2462\u2460"}, {"id": "ref-for-dom-mloperanddescriptor-type\u2462\u2461"}], "title": "7.6.22. The lstmCell() method"}, {"refs": [{"id": "ref-for-dom-mloperanddescriptor-type\u2462\u2462"}, {"id": "ref-for-dom-mloperanddescriptor-type\u2462\u2463"}], "title": "7.6.23. The matmul() method"}, {"refs": [{"id": "ref-for-dom-mloperanddescriptor-type\u2462\u2464"}, {"id": "ref-for-dom-mloperanddescriptor-type\u2462\u2465"}], "title": "7.6.26. The prelu() method"}], "external": false};
+window.dfnpanelData['dom-mloperanddescriptor-dimensions'] = {"dfnID": "dom-mloperanddescriptor-dimensions", "url": "#dom-mloperanddescriptor-dimensions", "dfnText": "dimensions", "refSections": [{"refs": [{"id": "ref-for-dom-mloperanddescriptor-dimensions"}], "title": "7.3.1. The MLOperandDescriptor dictionary"}, {"refs": [{"id": "ref-for-dom-mloperanddescriptor-dimensions\u2460"}], "title": "7.3.2. The MLOperand interface"}, {"refs": [{"id": "ref-for-dom-mloperanddescriptor-dimensions\u2461"}, {"id": "ref-for-dom-mloperanddescriptor-dimensions\u2462"}], "title": "7.3.2.1. Creating MLOperand"}, {"refs": [{"id": "ref-for-dom-mloperanddescriptor-dimensions\u2463"}], "title": "7.4.2. Synchronous Execution"}, {"refs": [{"id": "ref-for-dom-mloperanddescriptor-dimensions\u2464"}, {"id": "ref-for-dom-mloperanddescriptor-dimensions\u2465"}, {"id": "ref-for-dom-mloperanddescriptor-dimensions\u2466"}], "title": "7.6.2. The input() method"}, {"refs": [{"id": "ref-for-dom-mloperanddescriptor-dimensions\u2467"}], "title": "7.6.4.1. The constant(descriptor, bufferView) method"}, {"refs": [{"id": "ref-for-dom-mloperanddescriptor-dimensions\u2468"}, {"id": "ref-for-dom-mloperanddescriptor-dimensions\u2460\u24ea"}], "title": "7.6.4.2. The constant(value, type) method"}, {"refs": [{"id": "ref-for-dom-mloperanddescriptor-dimensions\u2460\u2460"}, {"id": "ref-for-dom-mloperanddescriptor-dimensions\u2460\u2461"}, {"id": "ref-for-dom-mloperanddescriptor-dimensions\u2460\u2462"}, {"id": "ref-for-dom-mloperanddescriptor-dimensions\u2460\u2463"}, {"id": "ref-for-dom-mloperanddescriptor-dimensions\u2460\u2464"}, {"id": "ref-for-dom-mloperanddescriptor-dimensions\u2460\u2465"}, {"id": "ref-for-dom-mloperanddescriptor-dimensions\u2460\u2466"}, {"id": "ref-for-dom-mloperanddescriptor-dimensions\u2460\u2467"}, {"id": "ref-for-dom-mloperanddescriptor-dimensions\u2460\u2468"}, {"id": "ref-for-dom-mloperanddescriptor-dimensions\u2461\u24ea"}], "title": "7.6.5. The batchNormalization() method"}, {"refs": [{"id": "ref-for-dom-mloperanddescriptor-dimensions\u2461\u2460"}, {"id": "ref-for-dom-mloperanddescriptor-dimensions\u2461\u2461"}, {"id": "ref-for-dom-mloperanddescriptor-dimensions\u2461\u2462"}, {"id": "ref-for-dom-mloperanddescriptor-dimensions\u2461\u2463"}, {"id": "ref-for-dom-mloperanddescriptor-dimensions\u2461\u2464"}, {"id": "ref-for-dom-mloperanddescriptor-dimensions\u2461\u2465"}, {"id": "ref-for-dom-mloperanddescriptor-dimensions\u2461\u2466"}], "title": "7.6.7. The concat() method"}, {"refs": [{"id": "ref-for-dom-mloperanddescriptor-dimensions\u2461\u2467"}, {"id": "ref-for-dom-mloperanddescriptor-dimensions\u2461\u2468"}, {"id": "ref-for-dom-mloperanddescriptor-dimensions\u2462\u24ea"}, {"id": "ref-for-dom-mloperanddescriptor-dimensions\u2462\u2460"}, {"id": "ref-for-dom-mloperanddescriptor-dimensions\u2462\u2461"}], "title": "7.6.8. The conv2d() method"}, {"refs": [{"id": "ref-for-dom-mloperanddescriptor-dimensions\u2462\u2462"}, {"id": "ref-for-dom-mloperanddescriptor-dimensions\u2462\u2463"}, {"id": "ref-for-dom-mloperanddescriptor-dimensions\u2462\u2464"}, {"id": "ref-for-dom-mloperanddescriptor-dimensions\u2462\u2465"}, {"id": "ref-for-dom-mloperanddescriptor-dimensions\u2462\u2466"}], "title": "7.6.9. The convTranspose2d() method"}, {"refs": [{"id": "ref-for-dom-mloperanddescriptor-dimensions\u2462\u2467"}, {"id": "ref-for-dom-mloperanddescriptor-dimensions\u2462\u2468"}, {"id": "ref-for-dom-mloperanddescriptor-dimensions\u2463\u24ea"}], "title": "7.6.10. Element-wise binary operations"}, {"refs": [{"id": "ref-for-dom-mloperanddescriptor-dimensions\u2463\u2460"}, {"id": "ref-for-dom-mloperanddescriptor-dimensions\u2463\u2461"}, {"id": "ref-for-dom-mloperanddescriptor-dimensions\u2463\u2462"}], "title": "7.6.13. The gemm() method"}, {"refs": [{"id": "ref-for-dom-mloperanddescriptor-dimensions\u2463\u2463"}, {"id": "ref-for-dom-mloperanddescriptor-dimensions\u2463\u2464"}, {"id": "ref-for-dom-mloperanddescriptor-dimensions\u2463\u2465"}], "title": "7.6.14. The gru() method"}, {"refs": [{"id": "ref-for-dom-mloperanddescriptor-dimensions\u2463\u2466"}, {"id": "ref-for-dom-mloperanddescriptor-dimensions\u2463\u2467"}, {"id": "ref-for-dom-mloperanddescriptor-dimensions\u2463\u2468"}, {"id": "ref-for-dom-mloperanddescriptor-dimensions\u2464\u24ea"}], "title": "7.6.15. The gruCell() method"}, {"refs": [{"id": "ref-for-dom-mloperanddescriptor-dimensions\u2464\u2460"}, {"id": "ref-for-dom-mloperanddescriptor-dimensions\u2464\u2461"}], "title": "7.6.18. The instanceNormalization() method"}, {"refs": [{"id": "ref-for-dom-mloperanddescriptor-dimensions\u2464\u2462"}, {"id": "ref-for-dom-mloperanddescriptor-dimensions\u2464\u2463"}, {"id": "ref-for-dom-mloperanddescriptor-dimensions\u2464\u2464"}, {"id": "ref-for-dom-mloperanddescriptor-dimensions\u2464\u2465"}, {"id": "ref-for-dom-mloperanddescriptor-dimensions\u2464\u2466"}, {"id": "ref-for-dom-mloperanddescriptor-dimensions\u2464\u2467"}, {"id": "ref-for-dom-mloperanddescriptor-dimensions\u2464\u2468"}, {"id": "ref-for-dom-mloperanddescriptor-dimensions\u2465\u24ea"}, {"id": "ref-for-dom-mloperanddescriptor-dimensions\u2465\u2460"}, {"id": "ref-for-dom-mloperanddescriptor-dimensions\u2465\u2461"}, {"id": "ref-for-dom-mloperanddescriptor-dimensions\u2465\u2462"}, {"id": "ref-for-dom-mloperanddescriptor-dimensions\u2465\u2463"}, {"id": "ref-for-dom-mloperanddescriptor-dimensions\u2465\u2464"}, {"id": "ref-for-dom-mloperanddescriptor-dimensions\u2465\u2465"}, {"id": "ref-for-dom-mloperanddescriptor-dimensions\u2465\u2466"}, {"id": "ref-for-dom-mloperanddescriptor-dimensions\u2465\u2467"}], "title": "7.6.21. The lstm() method"}, {"refs": [{"id": "ref-for-dom-mloperanddescriptor-dimensions\u2465\u2468"}, {"id": "ref-for-dom-mloperanddescriptor-dimensions\u2466\u24ea"}, {"id": "ref-for-dom-mloperanddescriptor-dimensions\u2466\u2460"}, {"id": "ref-for-dom-mloperanddescriptor-dimensions\u2466\u2461"}, {"id": "ref-for-dom-mloperanddescriptor-dimensions\u2466\u2462"}], "title": "7.6.22. The lstmCell() method"}, {"refs": [{"id": "ref-for-dom-mloperanddescriptor-dimensions\u2466\u2463"}, {"id": "ref-for-dom-mloperanddescriptor-dimensions\u2466\u2464"}, {"id": "ref-for-dom-mloperanddescriptor-dimensions\u2466\u2465"}], "title": "7.6.23. The matmul() method"}, {"refs": [{"id": "ref-for-dom-mloperanddescriptor-dimensions\u2466\u2466"}, {"id": "ref-for-dom-mloperanddescriptor-dimensions\u2466\u2467"}], "title": "7.6.24. The pad() method"}, {"refs": [{"id": "ref-for-dom-mloperanddescriptor-dimensions\u2466\u2468"}, {"id": "ref-for-dom-mloperanddescriptor-dimensions\u2467\u24ea"}], "title": "7.6.25. Pooling operations"}, {"refs": [{"id": "ref-for-dom-mloperanddescriptor-dimensions\u2467\u2460"}, {"id": "ref-for-dom-mloperanddescriptor-dimensions\u2467\u2461"}, {"id": "ref-for-dom-mloperanddescriptor-dimensions\u2467\u2462"}], "title": "7.6.26. The prelu() method"}, {"refs": [{"id": "ref-for-dom-mloperanddescriptor-dimensions\u2467\u2463"}, {"id": "ref-for-dom-mloperanddescriptor-dimensions\u2467\u2464"}, {"id": "ref-for-dom-mloperanddescriptor-dimensions\u2467\u2465"}, {"id": "ref-for-dom-mloperanddescriptor-dimensions\u2467\u2466"}], "title": "7.6.29. The resample2d() method"}, {"refs": [{"id": "ref-for-dom-mloperanddescriptor-dimensions\u2467\u2467"}, {"id": "ref-for-dom-mloperanddescriptor-dimensions\u2467\u2468"}], "title": "7.6.30. The reshape() method"}, {"refs": [{"id": "ref-for-dom-mloperanddescriptor-dimensions\u2468\u24ea"}], "title": "7.6.33.1. The softmax(input) method"}, {"refs": [{"id": "ref-for-dom-mloperanddescriptor-dimensions\u2468\u2460"}, {"id": "ref-for-dom-mloperanddescriptor-dimensions\u2468\u2461"}], "title": "7.6.36. The split() method"}, {"refs": [{"id": "ref-for-dom-mloperanddescriptor-dimensions\u2468\u2462"}], "title": "7.6.37. The squeeze() method"}, {"refs": [{"id": "ref-for-dom-mloperanddescriptor-dimensions\u2468\u2463"}, {"id": "ref-for-dom-mloperanddescriptor-dimensions\u2468\u2464"}, {"id": "ref-for-dom-mloperanddescriptor-dimensions\u2468\u2465"}], "title": "7.6.39. The transpose() method"}], "external": false};
+window.dfnpanelData['mloperanddescriptor-byte-length'] = {"dfnID": "mloperanddescriptor-byte-length", "url": "#mloperanddescriptor-byte-length", "dfnText": "byte length", "refSections": [{"refs": [{"id": "ref-for-mloperanddescriptor-byte-length"}, {"id": "ref-for-mloperanddescriptor-byte-length\u2460"}], "title": "7.4.2. Synchronous Execution"}, {"refs": [{"id": "ref-for-mloperanddescriptor-byte-length\u2461"}, {"id": "ref-for-mloperanddescriptor-byte-length\u2462"}], "title": "7.5.2. Dispatch Execution Commands"}, {"refs": [{"id": "ref-for-mloperanddescriptor-byte-length\u2463"}], "title": "7.6.2. The input() method"}, {"refs": [{"id": "ref-for-mloperanddescriptor-byte-length\u2464"}], "title": "7.6.4.1. The constant(descriptor, bufferView) method"}], "external": false};
+window.dfnpanelData['mloperand'] = {"dfnID": "mloperand", "url": "#mloperand", "dfnText": "MLOperand", "refSections": [{"refs": [{"id": "ref-for-mloperand"}, {"id": "ref-for-mloperand\u2460"}, {"id": "ref-for-mloperand\u2461"}, {"id": "ref-for-mloperand\u2462"}, {"id": "ref-for-mloperand\u2463"}, {"id": "ref-for-mloperand\u2464"}, {"id": "ref-for-mloperand\u2465"}], "title": "6.1. Overview"}, {"refs": [{"id": "ref-for-mloperand\u2466"}, {"id": "ref-for-mloperand\u2467"}, {"id": "ref-for-mloperand\u2468"}, {"id": "ref-for-mloperand\u2460\u24ea"}], "title": "7.3. The MLGraph interface"}, {"refs": [{"id": "ref-for-mloperand\u2460\u2460"}, {"id": "ref-for-mloperand\u2460\u2461"}, {"id": "ref-for-mloperand\u2460\u2462"}, {"id": "ref-for-mloperand\u2460\u2463"}, {"id": "ref-for-mloperand\u2460\u2464"}, {"id": "ref-for-mloperand\u2460\u2465"}, {"id": "ref-for-mloperand\u2460\u2466"}, {"id": "ref-for-mloperand\u2460\u2467"}, {"id": "ref-for-mloperand\u2460\u2468"}, {"id": "ref-for-mloperand\u2461\u24ea"}], "title": "7.3.2. The MLOperand interface"}, {"refs": [{"id": "ref-for-mloperand\u2461\u2460"}, {"id": "ref-for-mloperand\u2461\u2461"}, {"id": "ref-for-mloperand\u2461\u2462"}], "title": "7.3.2.1. Creating MLOperand"}, {"refs": [{"id": "ref-for-mloperand\u2461\u2463"}, {"id": "ref-for-mloperand\u2461\u2464"}, {"id": "ref-for-mloperand\u2461\u2465"}, {"id": "ref-for-mloperand\u2461\u2466"}], "title": "7.6. The MLGraphBuilder interface"}, {"refs": [{"id": "ref-for-mloperand\u2461\u2467"}, {"id": "ref-for-mloperand\u2461\u2468"}], "title": "7.6.2. The input() method"}, {"refs": [{"id": "ref-for-mloperand\u2462\u24ea"}], "title": "7.6.3.2. The buildSync(outputs) method"}, {"refs": [{"id": "ref-for-mloperand\u2462\u2460"}], "title": "7.6.4. The constant() method"}, {"refs": [{"id": "ref-for-mloperand\u2462\u2461"}], "title": "7.6.4.1. The constant(descriptor, bufferView) method"}, {"refs": [{"id": "ref-for-mloperand\u2462\u2462"}], "title": "7.6.4.2. The constant(value, type) method"}, {"refs": [{"id": "ref-for-mloperand\u2462\u2463"}, {"id": "ref-for-mloperand\u2462\u2464"}, {"id": "ref-for-mloperand\u2462\u2465"}, {"id": "ref-for-mloperand\u2462\u2466"}, {"id": "ref-for-mloperand\u2462\u2467"}, {"id": "ref-for-mloperand\u2462\u2468"}, {"id": "ref-for-mloperand\u2463\u24ea"}, {"id": "ref-for-mloperand\u2463\u2460"}, {"id": "ref-for-mloperand\u2463\u2461"}, {"id": "ref-for-mloperand\u2463\u2462"}, {"id": "ref-for-mloperand\u2463\u2463"}, {"id": "ref-for-mloperand\u2463\u2464"}, {"id": "ref-for-mloperand\u2463\u2465"}, {"id": "ref-for-mloperand\u2463\u2466"}, {"id": "ref-for-mloperand\u2463\u2467"}], "title": "7.6.5. The batchNormalization() method"}, {"refs": [{"id": "ref-for-mloperand\u2463\u2468"}, {"id": "ref-for-mloperand\u2464\u24ea"}], "title": "7.6.6. The clamp() method"}, {"refs": [{"id": "ref-for-mloperand\u2464\u2460"}, {"id": "ref-for-mloperand\u2464\u2461"}, {"id": "ref-for-mloperand\u2464\u2462"}], "title": "7.6.6.1. The clamp(operand, options) method"}, {"refs": [{"id": "ref-for-mloperand\u2464\u2463"}, {"id": "ref-for-mloperand\u2464\u2464"}, {"id": "ref-for-mloperand\u2464\u2465"}, {"id": "ref-for-mloperand\u2464\u2466"}, {"id": "ref-for-mloperand\u2464\u2467"}], "title": "7.6.7. The concat() method"}, {"refs": [{"id": "ref-for-mloperand\u2464\u2468"}, {"id": "ref-for-mloperand\u2465\u24ea"}, {"id": "ref-for-mloperand\u2465\u2460"}, {"id": "ref-for-mloperand\u2465\u2461"}, {"id": "ref-for-mloperand\u2465\u2462"}, {"id": "ref-for-mloperand\u2465\u2463"}, {"id": "ref-for-mloperand\u2465\u2464"}, {"id": "ref-for-mloperand\u2465\u2465"}, {"id": "ref-for-mloperand\u2465\u2466"}, {"id": "ref-for-mloperand\u2465\u2467"}, {"id": "ref-for-mloperand\u2465\u2468"}], "title": "7.6.8. The conv2d() method"}, {"refs": [{"id": "ref-for-mloperand\u2466\u24ea"}, {"id": "ref-for-mloperand\u2466\u2460"}, {"id": "ref-for-mloperand\u2466\u2461"}, {"id": "ref-for-mloperand\u2466\u2462"}, {"id": "ref-for-mloperand\u2466\u2463"}, {"id": "ref-for-mloperand\u2466\u2464"}, {"id": "ref-for-mloperand\u2466\u2465"}, {"id": "ref-for-mloperand\u2466\u2466"}, {"id": "ref-for-mloperand\u2466\u2467"}, {"id": "ref-for-mloperand\u2466\u2468"}, {"id": "ref-for-mloperand\u2467\u24ea"}], "title": "7.6.9. The convTranspose2d() method"}, {"refs": [{"id": "ref-for-mloperand\u2467\u2460"}, {"id": "ref-for-mloperand\u2467\u2461"}, {"id": "ref-for-mloperand\u2467\u2462"}, {"id": "ref-for-mloperand\u2467\u2463"}, {"id": "ref-for-mloperand\u2467\u2464"}, {"id": "ref-for-mloperand\u2467\u2465"}, {"id": "ref-for-mloperand\u2467\u2466"}, {"id": "ref-for-mloperand\u2467\u2467"}, {"id": "ref-for-mloperand\u2467\u2468"}, {"id": "ref-for-mloperand\u2468\u24ea"}, {"id": "ref-for-mloperand\u2468\u2460"}, {"id": "ref-for-mloperand\u2468\u2461"}, {"id": "ref-for-mloperand\u2468\u2462"}, {"id": "ref-for-mloperand\u2468\u2463"}, {"id": "ref-for-mloperand\u2468\u2464"}, {"id": "ref-for-mloperand\u2468\u2465"}, {"id": "ref-for-mloperand\u2468\u2466"}, {"id": "ref-for-mloperand\u2468\u2467"}, {"id": "ref-for-mloperand\u2468\u2468"}, {"id": "ref-for-mloperand\u2460\u24ea\u24ea"}, {"id": "ref-for-mloperand\u2460\u24ea\u2460"}, {"id": "ref-for-mloperand\u2460\u24ea\u2461"}, {"id": "ref-for-mloperand\u2460\u24ea\u2462"}, {"id": "ref-for-mloperand\u2460\u24ea\u2463"}, {"id": "ref-for-mloperand\u2460\u24ea\u2464"}], "title": "7.6.10. Element-wise binary operations"}, {"refs": [{"id": "ref-for-mloperand\u2460\u24ea\u2465"}, {"id": "ref-for-mloperand\u2460\u24ea\u2466"}, {"id": "ref-for-mloperand\u2460\u24ea\u2467"}, {"id": "ref-for-mloperand\u2460\u24ea\u2468"}, {"id": "ref-for-mloperand\u2460\u2460\u24ea"}, {"id": "ref-for-mloperand\u2460\u2460\u2460"}, {"id": "ref-for-mloperand\u2460\u2460\u2461"}, {"id": "ref-for-mloperand\u2460\u2460\u2462"}, {"id": "ref-for-mloperand\u2460\u2460\u2463"}, {"id": "ref-for-mloperand\u2460\u2460\u2464"}, {"id": "ref-for-mloperand\u2460\u2460\u2465"}, {"id": "ref-for-mloperand\u2460\u2460\u2466"}, {"id": "ref-for-mloperand\u2460\u2460\u2467"}, {"id": "ref-for-mloperand\u2460\u2460\u2468"}, {"id": "ref-for-mloperand\u2460\u2461\u24ea"}, {"id": "ref-for-mloperand\u2460\u2461\u2460"}, {"id": "ref-for-mloperand\u2460\u2461\u2461"}, {"id": "ref-for-mloperand\u2460\u2461\u2462"}, {"id": "ref-for-mloperand\u2460\u2461\u2463"}, {"id": "ref-for-mloperand\u2460\u2461\u2464"}, {"id": "ref-for-mloperand\u2460\u2461\u2465"}], "title": "7.6.11. Element-wise unary operations"}, {"refs": [{"id": "ref-for-mloperand\u2460\u2461\u2466"}, {"id": "ref-for-mloperand\u2460\u2461\u2467"}], "title": "7.6.12. The elu() method"}, {"refs": [{"id": "ref-for-mloperand\u2460\u2461\u2468"}, {"id": "ref-for-mloperand\u2460\u2462\u24ea"}], "title": "7.6.12.1. The elu(input, options) method"}, {"refs": [{"id": "ref-for-mloperand\u2460\u2462\u2460"}, {"id": "ref-for-mloperand\u2460\u2462\u2461"}, {"id": "ref-for-mloperand\u2460\u2462\u2462"}, {"id": "ref-for-mloperand\u2460\u2462\u2463"}, {"id": "ref-for-mloperand\u2460\u2462\u2464"}, {"id": "ref-for-mloperand\u2460\u2462\u2465"}, {"id": "ref-for-mloperand\u2460\u2462\u2466"}, {"id": "ref-for-mloperand\u2460\u2462\u2467"}, {"id": "ref-for-mloperand\u2460\u2462\u2468"}, {"id": "ref-for-mloperand\u2460\u2463\u24ea"}], "title": "7.6.13. The gemm() method"}, {"refs": [{"id": "ref-for-mloperand\u2460\u2463\u2460"}, {"id": "ref-for-mloperand\u2460\u2463\u2461"}, {"id": "ref-for-mloperand\u2460\u2463\u2462"}, {"id": "ref-for-mloperand\u2460\u2463\u2463"}, {"id": "ref-for-mloperand\u2460\u2463\u2464"}, {"id": "ref-for-mloperand\u2460\u2463\u2465"}, {"id": "ref-for-mloperand\u2460\u2463\u2466"}, {"id": "ref-for-mloperand\u2460\u2463\u2467"}, {"id": "ref-for-mloperand\u2460\u2463\u2468"}, {"id": "ref-for-mloperand\u2460\u2464\u24ea"}, {"id": "ref-for-mloperand\u2460\u2464\u2460"}, {"id": "ref-for-mloperand\u2460\u2464\u2461"}, {"id": "ref-for-mloperand\u2460\u2464\u2462"}, {"id": "ref-for-mloperand\u2460\u2464\u2463"}, {"id": "ref-for-mloperand\u2460\u2464\u2464"}, {"id": "ref-for-mloperand\u2460\u2464\u2465"}, {"id": "ref-for-mloperand\u2460\u2464\u2466"}, {"id": "ref-for-mloperand\u2460\u2464\u2467"}, {"id": "ref-for-mloperand\u2460\u2464\u2468"}, {"id": "ref-for-mloperand\u2460\u2465\u24ea"}, {"id": "ref-for-mloperand\u2460\u2465\u2460"}, {"id": "ref-for-mloperand\u2460\u2465\u2461"}], "title": "7.6.14. The gru() method"}, {"refs": [{"id": "ref-for-mloperand\u2460\u2465\u2462"}, {"id": "ref-for-mloperand\u2460\u2465\u2463"}, {"id": "ref-for-mloperand\u2460\u2465\u2464"}, {"id": "ref-for-mloperand\u2460\u2465\u2465"}, {"id": "ref-for-mloperand\u2460\u2465\u2466"}, {"id": "ref-for-mloperand\u2460\u2465\u2467"}, {"id": "ref-for-mloperand\u2460\u2465\u2468"}, {"id": "ref-for-mloperand\u2460\u2466\u24ea"}, {"id": "ref-for-mloperand\u2460\u2466\u2460"}, {"id": "ref-for-mloperand\u2460\u2466\u2461"}, {"id": "ref-for-mloperand\u2460\u2466\u2462"}, {"id": "ref-for-mloperand\u2460\u2466\u2463"}, {"id": "ref-for-mloperand\u2460\u2466\u2464"}, {"id": "ref-for-mloperand\u2460\u2466\u2465"}, {"id": "ref-for-mloperand\u2460\u2466\u2466"}, {"id": "ref-for-mloperand\u2460\u2466\u2467"}, {"id": "ref-for-mloperand\u2460\u2466\u2468"}, {"id": "ref-for-mloperand\u2460\u2467\u24ea"}, {"id": "ref-for-mloperand\u2460\u2467\u2460"}], "title": "7.6.15. The gruCell() method"}, {"refs": [{"id": "ref-for-mloperand\u2460\u2467\u2461"}, {"id": "ref-for-mloperand\u2460\u2467\u2462"}], "title": "7.6.16. The hardSigmoid() method"}, {"refs": [{"id": "ref-for-mloperand\u2460\u2467\u2463"}, {"id": "ref-for-mloperand\u2460\u2467\u2464"}, {"id": "ref-for-mloperand\u2460\u2467\u2465"}], "title": "7.6.16.1. The hardSigmoid(input, options) method"}, {"refs": [{"id": "ref-for-mloperand\u2460\u2467\u2466"}, {"id": "ref-for-mloperand\u2460\u2467\u2467"}], "title": "7.6.17. The hardSwish() method"}, {"refs": [{"id": "ref-for-mloperand\u2460\u2467\u2468"}, {"id": "ref-for-mloperand\u2460\u2468\u24ea"}], "title": "7.6.17.1. The hardSwish(input) method"}, {"refs": [{"id": "ref-for-mloperand\u2460\u2468\u2460"}, {"id": "ref-for-mloperand\u2460\u2468\u2461"}, {"id": "ref-for-mloperand\u2460\u2468\u2462"}, {"id": "ref-for-mloperand\u2460\u2468\u2463"}, {"id": "ref-for-mloperand\u2460\u2468\u2464"}, {"id": "ref-for-mloperand\u2460\u2468\u2465"}, {"id": "ref-for-mloperand\u2460\u2468\u2466"}, {"id": "ref-for-mloperand\u2460\u2468\u2467"}, {"id": "ref-for-mloperand\u2460\u2468\u2468"}, {"id": "ref-for-mloperand\u2461\u24ea\u24ea"}, {"id": "ref-for-mloperand\u2461\u24ea\u2460"}, {"id": "ref-for-mloperand\u2461\u24ea\u2461"}, {"id": "ref-for-mloperand\u2461\u24ea\u2462"}], "title": "7.6.18. The instanceNormalization() method"}, {"refs": [{"id": "ref-for-mloperand\u2461\u24ea\u2463"}, {"id": "ref-for-mloperand\u2461\u24ea\u2464"}], "title": "7.6.19. The leakyRelu() method"}, {"refs": [{"id": "ref-for-mloperand\u2461\u24ea\u2465"}, {"id": "ref-for-mloperand\u2461\u24ea\u2466"}], "title": "7.6.19.1. The leakyRelu(input, options) method"}, {"refs": [{"id": "ref-for-mloperand\u2461\u24ea\u2467"}, {"id": "ref-for-mloperand\u2461\u24ea\u2468"}], "title": "7.6.20. The linear() method"}, {"refs": [{"id": "ref-for-mloperand\u2461\u2460\u24ea"}, {"id": "ref-for-mloperand\u2461\u2460\u2460"}], "title": "7.6.20.1. The linear(input, options) method"}, {"refs": [{"id": "ref-for-mloperand\u2461\u2460\u2461"}, {"id": "ref-for-mloperand\u2461\u2460\u2462"}, {"id": "ref-for-mloperand\u2461\u2460\u2463"}, {"id": "ref-for-mloperand\u2461\u2460\u2464"}, {"id": "ref-for-mloperand\u2461\u2460\u2465"}, {"id": "ref-for-mloperand\u2461\u2460\u2466"}, {"id": "ref-for-mloperand\u2461\u2460\u2467"}, {"id": "ref-for-mloperand\u2461\u2460\u2468"}, {"id": "ref-for-mloperand\u2461\u2461\u24ea"}, {"id": "ref-for-mloperand\u2461\u2461\u2460"}, {"id": "ref-for-mloperand\u2461\u2461\u2461"}, {"id": "ref-for-mloperand\u2461\u2461\u2462"}, {"id": "ref-for-mloperand\u2461\u2461\u2463"}, {"id": "ref-for-mloperand\u2461\u2461\u2464"}, {"id": "ref-for-mloperand\u2461\u2461\u2465"}, {"id": "ref-for-mloperand\u2461\u2461\u2466"}, {"id": "ref-for-mloperand\u2461\u2461\u2467"}, {"id": "ref-for-mloperand\u2461\u2461\u2468"}, {"id": "ref-for-mloperand\u2461\u2462\u24ea"}, {"id": "ref-for-mloperand\u2461\u2462\u2460"}, {"id": "ref-for-mloperand\u2461\u2462\u2461"}, {"id": "ref-for-mloperand\u2461\u2462\u2462"}, {"id": "ref-for-mloperand\u2461\u2462\u2463"}, {"id": "ref-for-mloperand\u2461\u2462\u2464"}, {"id": "ref-for-mloperand\u2461\u2462\u2465"}, {"id": "ref-for-mloperand\u2461\u2462\u2466"}, {"id": "ref-for-mloperand\u2461\u2462\u2467"}, {"id": "ref-for-mloperand\u2461\u2462\u2468"}, {"id": "ref-for-mloperand\u2461\u2463\u24ea"}], "title": "7.6.21. The lstm() method"}, {"refs": [{"id": "ref-for-mloperand\u2461\u2463\u2460"}, {"id": "ref-for-mloperand\u2461\u2463\u2461"}, {"id": "ref-for-mloperand\u2461\u2463\u2462"}, {"id": "ref-for-mloperand\u2461\u2463\u2463"}, {"id": "ref-for-mloperand\u2461\u2463\u2464"}, {"id": "ref-for-mloperand\u2461\u2463\u2465"}, {"id": "ref-for-mloperand\u2461\u2463\u2466"}, {"id": "ref-for-mloperand\u2461\u2463\u2467"}, {"id": "ref-for-mloperand\u2461\u2463\u2468"}, {"id": "ref-for-mloperand\u2461\u2464\u24ea"}, {"id": "ref-for-mloperand\u2461\u2464\u2460"}, {"id": "ref-for-mloperand\u2461\u2464\u2461"}, {"id": "ref-for-mloperand\u2461\u2464\u2462"}, {"id": "ref-for-mloperand\u2461\u2464\u2463"}, {"id": "ref-for-mloperand\u2461\u2464\u2464"}, {"id": "ref-for-mloperand\u2461\u2464\u2465"}, {"id": "ref-for-mloperand\u2461\u2464\u2466"}, {"id": "ref-for-mloperand\u2461\u2464\u2467"}, {"id": "ref-for-mloperand\u2461\u2464\u2468"}, {"id": "ref-for-mloperand\u2461\u2465\u24ea"}, {"id": "ref-for-mloperand\u2461\u2465\u2460"}, {"id": "ref-for-mloperand\u2461\u2465\u2461"}, {"id": "ref-for-mloperand\u2461\u2465\u2462"}, {"id": "ref-for-mloperand\u2461\u2465\u2463"}, {"id": "ref-for-mloperand\u2461\u2465\u2464"}], "title": "7.6.22. The lstmCell() method"}, {"refs": [{"id": "ref-for-mloperand\u2461\u2465\u2465"}, {"id": "ref-for-mloperand\u2461\u2465\u2466"}, {"id": "ref-for-mloperand\u2461\u2465\u2467"}, {"id": "ref-for-mloperand\u2461\u2465\u2468"}, {"id": "ref-for-mloperand\u2461\u2466\u24ea"}, {"id": "ref-for-mloperand\u2461\u2466\u2460"}, {"id": "ref-for-mloperand\u2461\u2466\u2461"}], "title": "7.6.23. The matmul() method"}, {"refs": [{"id": "ref-for-mloperand\u2461\u2466\u2462"}, {"id": "ref-for-mloperand\u2461\u2466\u2463"}, {"id": "ref-for-mloperand\u2461\u2466\u2464"}, {"id": "ref-for-mloperand\u2461\u2466\u2465"}, {"id": "ref-for-mloperand\u2461\u2466\u2466"}], "title": "7.6.24. The pad() method"}, {"refs": [{"id": "ref-for-mloperand\u2461\u2466\u2467"}, {"id": "ref-for-mloperand\u2461\u2466\u2468"}, {"id": "ref-for-mloperand\u2461\u2467\u24ea"}, {"id": "ref-for-mloperand\u2461\u2467\u2460"}, {"id": "ref-for-mloperand\u2461\u2467\u2461"}, {"id": "ref-for-mloperand\u2461\u2467\u2462"}, {"id": "ref-for-mloperand\u2461\u2467\u2463"}, {"id": "ref-for-mloperand\u2461\u2467\u2464"}, {"id": "ref-for-mloperand\u2461\u2467\u2465"}], "title": "7.6.25. Pooling operations"}, {"refs": [{"id": "ref-for-mloperand\u2461\u2467\u2466"}, {"id": "ref-for-mloperand\u2461\u2467\u2467"}, {"id": "ref-for-mloperand\u2461\u2467\u2468"}, {"id": "ref-for-mloperand\u2461\u2468\u24ea"}, {"id": "ref-for-mloperand\u2461\u2468\u2460"}, {"id": "ref-for-mloperand\u2461\u2468\u2461"}, {"id": "ref-for-mloperand\u2461\u2468\u2462"}], "title": "7.6.26. The prelu() method"}, {"refs": [{"id": "ref-for-mloperand\u2461\u2468\u2463"}, {"id": "ref-for-mloperand\u2461\u2468\u2464"}, {"id": "ref-for-mloperand\u2461\u2468\u2465"}, {"id": "ref-for-mloperand\u2461\u2468\u2466"}, {"id": "ref-for-mloperand\u2461\u2468\u2467"}, {"id": "ref-for-mloperand\u2461\u2468\u2468"}, {"id": "ref-for-mloperand\u2462\u24ea\u24ea"}, {"id": "ref-for-mloperand\u2462\u24ea\u2460"}, {"id": "ref-for-mloperand\u2462\u24ea\u2461"}, {"id": "ref-for-mloperand\u2462\u24ea\u2462"}, {"id": "ref-for-mloperand\u2462\u24ea\u2463"}, {"id": "ref-for-mloperand\u2462\u24ea\u2464"}, {"id": "ref-for-mloperand\u2462\u24ea\u2465"}, {"id": "ref-for-mloperand\u2462\u24ea\u2466"}, {"id": "ref-for-mloperand\u2462\u24ea\u2467"}, {"id": "ref-for-mloperand\u2462\u24ea\u2468"}, {"id": "ref-for-mloperand\u2462\u2460\u24ea"}, {"id": "ref-for-mloperand\u2462\u2460\u2460"}, {"id": "ref-for-mloperand\u2462\u2460\u2461"}, {"id": "ref-for-mloperand\u2462\u2460\u2462"}, {"id": "ref-for-mloperand\u2462\u2460\u2463"}, {"id": "ref-for-mloperand\u2462\u2460\u2464"}, {"id": "ref-for-mloperand\u2462\u2460\u2465"}], "title": "7.6.27. Reduction operations"}, {"refs": [{"id": "ref-for-mloperand\u2462\u2460\u2466"}, {"id": "ref-for-mloperand\u2462\u2460\u2467"}], "title": "7.6.28. The relu() method"}, {"refs": [{"id": "ref-for-mloperand\u2462\u2460\u2468"}, {"id": "ref-for-mloperand\u2462\u2461\u24ea"}, {"id": "ref-for-mloperand\u2462\u2461\u2460"}], "title": "7.6.28.1. The relu(input) method"}, {"refs": [{"id": "ref-for-mloperand\u2462\u2461\u2461"}, {"id": "ref-for-mloperand\u2462\u2461\u2462"}, {"id": "ref-for-mloperand\u2462\u2461\u2463"}, {"id": "ref-for-mloperand\u2462\u2461\u2464"}], "title": "7.6.29. The resample2d() method"}, {"refs": [{"id": "ref-for-mloperand\u2462\u2461\u2465"}, {"id": "ref-for-mloperand\u2462\u2461\u2466"}, {"id": "ref-for-mloperand\u2462\u2461\u2467"}, {"id": "ref-for-mloperand\u2462\u2461\u2468"}, {"id": "ref-for-mloperand\u2462\u2462\u24ea"}], "title": "7.6.30. The reshape() method"}, {"refs": [{"id": "ref-for-mloperand\u2462\u2462\u2460"}, {"id": "ref-for-mloperand\u2462\u2462\u2461"}], "title": "7.6.31. The sigmoid() method"}, {"refs": [{"id": "ref-for-mloperand\u2462\u2462\u2462"}, {"id": "ref-for-mloperand\u2462\u2462\u2463"}, {"id": "ref-for-mloperand\u2462\u2462\u2464"}], "title": "7.6.31.1. The sigmoid(input) method"}, {"refs": [{"id": "ref-for-mloperand\u2462\u2462\u2465"}, {"id": "ref-for-mloperand\u2462\u2462\u2466"}, {"id": "ref-for-mloperand\u2462\u2462\u2467"}, {"id": "ref-for-mloperand\u2462\u2462\u2468"}, {"id": "ref-for-mloperand\u2462\u2463\u24ea"}], "title": "7.6.32. The slice() method"}, {"refs": [{"id": "ref-for-mloperand\u2462\u2463\u2460"}, {"id": "ref-for-mloperand\u2462\u2463\u2461"}], "title": "7.6.33. The softmax() method"}, {"refs": [{"id": "ref-for-mloperand\u2462\u2463\u2462"}, {"id": "ref-for-mloperand\u2462\u2463\u2463"}, {"id": "ref-for-mloperand\u2462\u2463\u2464"}], "title": "7.6.33.1. The softmax(input) method"}, {"refs": [{"id": "ref-for-mloperand\u2462\u2463\u2465"}, {"id": "ref-for-mloperand\u2462\u2463\u2466"}], "title": "7.6.34. The softplus() method"}, {"refs": [{"id": "ref-for-mloperand\u2462\u2463\u2467"}, {"id": "ref-for-mloperand\u2462\u2463\u2468"}], "title": "7.6.34.1. The softplus(input, options) method"}, {"refs": [{"id": "ref-for-mloperand\u2462\u2464\u24ea"}, {"id": "ref-for-mloperand\u2462\u2464\u2460"}], "title": "7.6.35. The softsign() method"}, {"refs": [{"id": "ref-for-mloperand\u2462\u2464\u2461"}, {"id": "ref-for-mloperand\u2462\u2464\u2462"}, {"id": "ref-for-mloperand\u2462\u2464\u2463"}], "title": "7.6.35.1. The softsign(input) method"}, {"refs": [{"id": "ref-for-mloperand\u2462\u2464\u2464"}, {"id": "ref-for-mloperand\u2462\u2464\u2465"}, {"id": "ref-for-mloperand\u2462\u2464\u2466"}, {"id": "ref-for-mloperand\u2462\u2464\u2467"}, {"id": "ref-for-mloperand\u2462\u2464\u2468"}], "title": "7.6.36. The split() method"}, {"refs": [{"id": "ref-for-mloperand\u2462\u2465\u24ea"}, {"id": "ref-for-mloperand\u2462\u2465\u2460"}, {"id": "ref-for-mloperand\u2462\u2465\u2461"}, {"id": "ref-for-mloperand\u2462\u2465\u2462"}, {"id": "ref-for-mloperand\u2462\u2465\u2463"}], "title": "7.6.37. The squeeze() method"}, {"refs": [{"id": "ref-for-mloperand\u2462\u2465\u2464"}, {"id": "ref-for-mloperand\u2462\u2465\u2465"}], "title": "7.6.38. The tanh() method"}, {"refs": [{"id": "ref-for-mloperand\u2462\u2465\u2466"}, {"id": "ref-for-mloperand\u2462\u2465\u2467"}, {"id": "ref-for-mloperand\u2462\u2465\u2468"}], "title": "7.6.38.1. The tanh(input) method"}, {"refs": [{"id": "ref-for-mloperand\u2462\u2466\u24ea"}, {"id": "ref-for-mloperand\u2462\u2466\u2460"}, {"id": "ref-for-mloperand\u2462\u2466\u2461"}, {"id": "ref-for-mloperand\u2462\u2466\u2462"}, {"id": "ref-for-mloperand\u2462\u2466\u2463"}], "title": "7.6.39. The transpose() method"}], "external": false};
+window.dfnpanelData['dom-mloperand-builder-slot'] = {"dfnID": "dom-mloperand-builder-slot", "url": "#dom-mloperand-builder-slot", "dfnText": "[[builder]]", "refSections": [{"refs": [{"id": "ref-for-dom-mloperand-builder-slot"}], "title": "7.3.2. The MLOperand interface"}, {"refs": [{"id": "ref-for-dom-mloperand-builder-slot\u2460"}, {"id": "ref-for-dom-mloperand-builder-slot\u2461"}, {"id": "ref-for-dom-mloperand-builder-slot\u2462"}, {"id": "ref-for-dom-mloperand-builder-slot\u2463"}, {"id": "ref-for-dom-mloperand-builder-slot\u2464"}], "title": "7.3.2.1. Creating MLOperand"}], "external": false};
+window.dfnpanelData['dom-mloperand-descriptor-slot'] = {"dfnID": "dom-mloperand-descriptor-slot", "url": "#dom-mloperand-descriptor-slot", "dfnText": "[[descriptor]]", "refSections": [{"refs": [{"id": "ref-for-dom-mloperand-descriptor-slot"}], "title": "7.3.2. The MLOperand interface"}, {"refs": [{"id": "ref-for-dom-mloperand-descriptor-slot\u2460"}, {"id": "ref-for-dom-mloperand-descriptor-slot\u2461"}, {"id": "ref-for-dom-mloperand-descriptor-slot\u2462"}, {"id": "ref-for-dom-mloperand-descriptor-slot\u2463"}], "title": "7.3.2.1. Creating MLOperand"}, {"refs": [{"id": "ref-for-dom-mloperand-descriptor-slot\u2464"}], "title": "7.6.3.2. The buildSync(outputs) method"}, {"refs": [{"id": "ref-for-dom-mloperand-descriptor-slot\u2465"}, {"id": "ref-for-dom-mloperand-descriptor-slot\u2466"}, {"id": "ref-for-dom-mloperand-descriptor-slot\u2467"}, {"id": "ref-for-dom-mloperand-descriptor-slot\u2468"}, {"id": "ref-for-dom-mloperand-descriptor-slot\u2460\u24ea"}, {"id": "ref-for-dom-mloperand-descriptor-slot\u2460\u2460"}, {"id": "ref-for-dom-mloperand-descriptor-slot\u2460\u2461"}, {"id": "ref-for-dom-mloperand-descriptor-slot\u2460\u2462"}, {"id": "ref-for-dom-mloperand-descriptor-slot\u2460\u2463"}, {"id": "ref-for-dom-mloperand-descriptor-slot\u2460\u2464"}, {"id": "ref-for-dom-mloperand-descriptor-slot\u2460\u2465"}], "title": "7.6.5. The batchNormalization() method"}, {"refs": [{"id": "ref-for-dom-mloperand-descriptor-slot\u2460\u2466"}, {"id": "ref-for-dom-mloperand-descriptor-slot\u2460\u2467"}, {"id": "ref-for-dom-mloperand-descriptor-slot\u2460\u2468"}, {"id": "ref-for-dom-mloperand-descriptor-slot\u2461\u24ea"}], "title": "7.6.7. The concat() method"}, {"refs": [{"id": "ref-for-dom-mloperand-descriptor-slot\u2461\u2460"}, {"id": "ref-for-dom-mloperand-descriptor-slot\u2461\u2461"}, {"id": "ref-for-dom-mloperand-descriptor-slot\u2461\u2462"}, {"id": "ref-for-dom-mloperand-descriptor-slot\u2461\u2463"}, {"id": "ref-for-dom-mloperand-descriptor-slot\u2461\u2464"}, {"id": "ref-for-dom-mloperand-descriptor-slot\u2461\u2465"}, {"id": "ref-for-dom-mloperand-descriptor-slot\u2461\u2466"}, {"id": "ref-for-dom-mloperand-descriptor-slot\u2461\u2467"}, {"id": "ref-for-dom-mloperand-descriptor-slot\u2461\u2468"}], "title": "7.6.8. The conv2d() method"}, {"refs": [{"id": "ref-for-dom-mloperand-descriptor-slot\u2462\u24ea"}, {"id": "ref-for-dom-mloperand-descriptor-slot\u2462\u2460"}, {"id": "ref-for-dom-mloperand-descriptor-slot\u2462\u2461"}, {"id": "ref-for-dom-mloperand-descriptor-slot\u2462\u2462"}, {"id": "ref-for-dom-mloperand-descriptor-slot\u2462\u2463"}, {"id": "ref-for-dom-mloperand-descriptor-slot\u2462\u2464"}, {"id": "ref-for-dom-mloperand-descriptor-slot\u2462\u2465"}, {"id": "ref-for-dom-mloperand-descriptor-slot\u2462\u2466"}, {"id": "ref-for-dom-mloperand-descriptor-slot\u2462\u2467"}], "title": "7.6.9. The convTranspose2d() method"}, {"refs": [{"id": "ref-for-dom-mloperand-descriptor-slot\u2462\u2468"}, {"id": "ref-for-dom-mloperand-descriptor-slot\u2463\u24ea"}, {"id": "ref-for-dom-mloperand-descriptor-slot\u2463\u2460"}, {"id": "ref-for-dom-mloperand-descriptor-slot\u2463\u2461"}, {"id": "ref-for-dom-mloperand-descriptor-slot\u2463\u2462"}], "title": "7.6.10. Element-wise binary operations"}, {"refs": [{"id": "ref-for-dom-mloperand-descriptor-slot\u2463\u2463"}, {"id": "ref-for-dom-mloperand-descriptor-slot\u2463\u2464"}, {"id": "ref-for-dom-mloperand-descriptor-slot\u2463\u2465"}], "title": "7.6.13. The gemm() method"}, {"refs": [{"id": "ref-for-dom-mloperand-descriptor-slot\u2463\u2466"}, {"id": "ref-for-dom-mloperand-descriptor-slot\u2463\u2467"}, {"id": "ref-for-dom-mloperand-descriptor-slot\u2463\u2468"}], "title": "7.6.14. The gru() method"}, {"refs": [{"id": "ref-for-dom-mloperand-descriptor-slot\u2464\u24ea"}, {"id": "ref-for-dom-mloperand-descriptor-slot\u2464\u2460"}, {"id": "ref-for-dom-mloperand-descriptor-slot\u2464\u2461"}], "title": "7.6.15. The gruCell() method"}, {"refs": [{"id": "ref-for-dom-mloperand-descriptor-slot\u2464\u2462"}, {"id": "ref-for-dom-mloperand-descriptor-slot\u2464\u2463"}], "title": "7.6.18. The instanceNormalization() method"}, {"refs": [{"id": "ref-for-dom-mloperand-descriptor-slot\u2464\u2464"}, {"id": "ref-for-dom-mloperand-descriptor-slot\u2464\u2465"}, {"id": "ref-for-dom-mloperand-descriptor-slot\u2464\u2466"}, {"id": "ref-for-dom-mloperand-descriptor-slot\u2464\u2467"}, {"id": "ref-for-dom-mloperand-descriptor-slot\u2464\u2468"}, {"id": "ref-for-dom-mloperand-descriptor-slot\u2465\u24ea"}, {"id": "ref-for-dom-mloperand-descriptor-slot\u2465\u2460"}, {"id": "ref-for-dom-mloperand-descriptor-slot\u2465\u2461"}, {"id": "ref-for-dom-mloperand-descriptor-slot\u2465\u2462"}, {"id": "ref-for-dom-mloperand-descriptor-slot\u2465\u2463"}, {"id": "ref-for-dom-mloperand-descriptor-slot\u2465\u2464"}, {"id": "ref-for-dom-mloperand-descriptor-slot\u2465\u2465"}, {"id": "ref-for-dom-mloperand-descriptor-slot\u2465\u2466"}, {"id": "ref-for-dom-mloperand-descriptor-slot\u2465\u2467"}, {"id": "ref-for-dom-mloperand-descriptor-slot\u2465\u2468"}], "title": "7.6.21. The lstm() method"}, {"refs": [{"id": "ref-for-dom-mloperand-descriptor-slot\u2466\u24ea"}, {"id": "ref-for-dom-mloperand-descriptor-slot\u2466\u2460"}, {"id": "ref-for-dom-mloperand-descriptor-slot\u2466\u2461"}, {"id": "ref-for-dom-mloperand-descriptor-slot\u2466\u2462"}, {"id": "ref-for-dom-mloperand-descriptor-slot\u2466\u2463"}], "title": "7.6.22. The lstmCell() method"}, {"refs": [{"id": "ref-for-dom-mloperand-descriptor-slot\u2466\u2464"}, {"id": "ref-for-dom-mloperand-descriptor-slot\u2466\u2465"}, {"id": "ref-for-dom-mloperand-descriptor-slot\u2466\u2466"}], "title": "7.6.23. The matmul() method"}, {"refs": [{"id": "ref-for-dom-mloperand-descriptor-slot\u2466\u2467"}, {"id": "ref-for-dom-mloperand-descriptor-slot\u2466\u2468"}], "title": "7.6.24. The pad() method"}, {"refs": [{"id": "ref-for-dom-mloperand-descriptor-slot\u2467\u24ea"}, {"id": "ref-for-dom-mloperand-descriptor-slot\u2467\u2460"}], "title": "7.6.25. Pooling operations"}, {"refs": [{"id": "ref-for-dom-mloperand-descriptor-slot\u2467\u2461"}, {"id": "ref-for-dom-mloperand-descriptor-slot\u2467\u2462"}, {"id": "ref-for-dom-mloperand-descriptor-slot\u2467\u2463"}], "title": "7.6.26. The prelu() method"}, {"refs": [{"id": "ref-for-dom-mloperand-descriptor-slot\u2467\u2464"}, {"id": "ref-for-dom-mloperand-descriptor-slot\u2467\u2465"}, {"id": "ref-for-dom-mloperand-descriptor-slot\u2467\u2466"}, {"id": "ref-for-dom-mloperand-descriptor-slot\u2467\u2467"}, {"id": "ref-for-dom-mloperand-descriptor-slot\u2467\u2468"}], "title": "7.6.29. The resample2d() method"}, {"refs": [{"id": "ref-for-dom-mloperand-descriptor-slot\u2468\u24ea"}, {"id": "ref-for-dom-mloperand-descriptor-slot\u2468\u2460"}], "title": "7.6.30. The reshape() method"}, {"refs": [{"id": "ref-for-dom-mloperand-descriptor-slot\u2468\u2461"}], "title": "7.6.33.1. The softmax(input) method"}, {"refs": [{"id": "ref-for-dom-mloperand-descriptor-slot\u2468\u2462"}], "title": "7.6.37. The squeeze() method"}, {"refs": [{"id": "ref-for-dom-mloperand-descriptor-slot\u2468\u2463"}, {"id": "ref-for-dom-mloperand-descriptor-slot\u2468\u2464"}, {"id": "ref-for-dom-mloperand-descriptor-slot\u2468\u2465"}], "title": "7.6.39. The transpose() method"}], "external": false};
+window.dfnpanelData['dom-mloperand-name-slot'] = {"dfnID": "dom-mloperand-name-slot", "url": "#dom-mloperand-name-slot", "dfnText": "[[name]]", "refSections": [{"refs": [{"id": "ref-for-dom-mloperand-name-slot"}, {"id": "ref-for-dom-mloperand-name-slot\u2460"}, {"id": "ref-for-dom-mloperand-name-slot\u2461"}], "title": "7.3.2.1. Creating MLOperand"}, {"refs": [{"id": "ref-for-dom-mloperand-name-slot\u2462"}], "title": "7.6.2. The input() method"}, {"refs": [{"id": "ref-for-dom-mloperand-name-slot\u2463"}, {"id": "ref-for-dom-mloperand-name-slot\u2464"}], "title": "7.6.3.2. The buildSync(outputs) method"}], "external": false};
+window.dfnpanelData['dom-mloperand-operand-slot'] = {"dfnID": "dom-mloperand-operand-slot", "url": "#dom-mloperand-operand-slot", "dfnText": "[[operand]]", "refSections": [{"refs": [{"id": "ref-for-dom-mloperand-operand-slot"}], "title": "7.6.2. The input() method"}, {"refs": [{"id": "ref-for-dom-mloperand-operand-slot\u2460"}], "title": "7.6.3.2. The buildSync(outputs) method"}, {"refs": [{"id": "ref-for-dom-mloperand-operand-slot\u2461"}], "title": "7.6.4.1. The constant(descriptor, bufferView) method"}, {"refs": [{"id": "ref-for-dom-mloperand-operand-slot\u2462"}], "title": "7.6.4.2. The constant(value, type) method"}, {"refs": [{"id": "ref-for-dom-mloperand-operand-slot\u2463"}, {"id": "ref-for-dom-mloperand-operand-slot\u2464"}, {"id": "ref-for-dom-mloperand-operand-slot\u2465"}], "title": "7.6.6.1. The clamp(operand, options) method"}, {"refs": [{"id": "ref-for-dom-mloperand-operand-slot\u2466"}, {"id": "ref-for-dom-mloperand-operand-slot\u2467"}], "title": "7.6.7. The concat() method"}, {"refs": [{"id": "ref-for-dom-mloperand-operand-slot\u2468"}, {"id": "ref-for-dom-mloperand-operand-slot\u2460\u24ea"}, {"id": "ref-for-dom-mloperand-operand-slot\u2460\u2460"}], "title": "7.6.8. The conv2d() method"}, {"refs": [{"id": "ref-for-dom-mloperand-operand-slot\u2460\u2461"}, {"id": "ref-for-dom-mloperand-operand-slot\u2460\u2462"}, {"id": "ref-for-dom-mloperand-operand-slot\u2460\u2463"}], "title": "7.6.9. The convTranspose2d() method"}, {"refs": [{"id": "ref-for-dom-mloperand-operand-slot\u2460\u2464"}, {"id": "ref-for-dom-mloperand-operand-slot\u2460\u2465"}, {"id": "ref-for-dom-mloperand-operand-slot\u2460\u2466"}, {"id": "ref-for-dom-mloperand-operand-slot\u2460\u2467"}], "title": "7.6.10. Element-wise binary operations"}, {"refs": [{"id": "ref-for-dom-mloperand-operand-slot\u2460\u2468"}, {"id": "ref-for-dom-mloperand-operand-slot\u2461\u24ea"}, {"id": "ref-for-dom-mloperand-operand-slot\u2461\u2460"}], "title": "7.6.11. Element-wise unary operations"}, {"refs": [{"id": "ref-for-dom-mloperand-operand-slot\u2461\u2461"}, {"id": "ref-for-dom-mloperand-operand-slot\u2461\u2462"}, {"id": "ref-for-dom-mloperand-operand-slot\u2461\u2463"}], "title": "7.6.12.1. The elu(input, options) method"}, {"refs": [{"id": "ref-for-dom-mloperand-operand-slot\u2461\u2464"}, {"id": "ref-for-dom-mloperand-operand-slot\u2461\u2465"}, {"id": "ref-for-dom-mloperand-operand-slot\u2461\u2466"}, {"id": "ref-for-dom-mloperand-operand-slot\u2461\u2467"}], "title": "7.6.13. The gemm() method"}, {"refs": [{"id": "ref-for-dom-mloperand-operand-slot\u2461\u2468"}], "title": "7.6.14. The gru() method"}, {"refs": [{"id": "ref-for-dom-mloperand-operand-slot\u2462\u24ea"}, {"id": "ref-for-dom-mloperand-operand-slot\u2462\u2460"}, {"id": "ref-for-dom-mloperand-operand-slot\u2462\u2461"}], "title": "7.6.15. The gruCell() method"}, {"refs": [{"id": "ref-for-dom-mloperand-operand-slot\u2462\u2462"}, {"id": "ref-for-dom-mloperand-operand-slot\u2462\u2463"}, {"id": "ref-for-dom-mloperand-operand-slot\u2462\u2464"}], "title": "7.6.16.1. The hardSigmoid(input, options) method"}, {"refs": [{"id": "ref-for-dom-mloperand-operand-slot\u2462\u2465"}, {"id": "ref-for-dom-mloperand-operand-slot\u2462\u2466"}, {"id": "ref-for-dom-mloperand-operand-slot\u2462\u2467"}], "title": "7.6.17.1. The hardSwish(input) method"}, {"refs": [{"id": "ref-for-dom-mloperand-operand-slot\u2462\u2468"}, {"id": "ref-for-dom-mloperand-operand-slot\u2463\u24ea"}, {"id": "ref-for-dom-mloperand-operand-slot\u2463\u2460"}], "title": "7.6.18. The instanceNormalization() method"}, {"refs": [{"id": "ref-for-dom-mloperand-operand-slot\u2463\u2461"}, {"id": "ref-for-dom-mloperand-operand-slot\u2463\u2462"}, {"id": "ref-for-dom-mloperand-operand-slot\u2463\u2463"}], "title": "7.6.19.1. The leakyRelu(input, options) method"}, {"refs": [{"id": "ref-for-dom-mloperand-operand-slot\u2463\u2464"}, {"id": "ref-for-dom-mloperand-operand-slot\u2463\u2465"}, {"id": "ref-for-dom-mloperand-operand-slot\u2463\u2466"}], "title": "7.6.20.1. The linear(input, options) method"}, {"refs": [{"id": "ref-for-dom-mloperand-operand-slot\u2463\u2467"}, {"id": "ref-for-dom-mloperand-operand-slot\u2463\u2468"}, {"id": "ref-for-dom-mloperand-operand-slot\u2464\u24ea"}, {"id": "ref-for-dom-mloperand-operand-slot\u2464\u2460"}], "title": "7.6.21. The lstm() method"}, {"refs": [{"id": "ref-for-dom-mloperand-operand-slot\u2464\u2461"}, {"id": "ref-for-dom-mloperand-operand-slot\u2464\u2462"}, {"id": "ref-for-dom-mloperand-operand-slot\u2464\u2463"}], "title": "7.6.22. The lstmCell() method"}, {"refs": [{"id": "ref-for-dom-mloperand-operand-slot\u2464\u2464"}, {"id": "ref-for-dom-mloperand-operand-slot\u2464\u2465"}, {"id": "ref-for-dom-mloperand-operand-slot\u2464\u2466"}, {"id": "ref-for-dom-mloperand-operand-slot\u2464\u2467"}], "title": "7.6.23. The matmul() method"}, {"refs": [{"id": "ref-for-dom-mloperand-operand-slot\u2464\u2468"}, {"id": "ref-for-dom-mloperand-operand-slot\u2465\u24ea"}, {"id": "ref-for-dom-mloperand-operand-slot\u2465\u2460"}], "title": "7.6.24. The pad() method"}, {"refs": [{"id": "ref-for-dom-mloperand-operand-slot\u2465\u2461"}, {"id": "ref-for-dom-mloperand-operand-slot\u2465\u2462"}, {"id": "ref-for-dom-mloperand-operand-slot\u2465\u2463"}], "title": "7.6.25. Pooling operations"}, {"refs": [{"id": "ref-for-dom-mloperand-operand-slot\u2465\u2464"}, {"id": "ref-for-dom-mloperand-operand-slot\u2465\u2465"}, {"id": "ref-for-dom-mloperand-operand-slot\u2465\u2466"}], "title": "7.6.26. The prelu() method"}, {"refs": [{"id": "ref-for-dom-mloperand-operand-slot\u2465\u2467"}, {"id": "ref-for-dom-mloperand-operand-slot\u2465\u2468"}, {"id": "ref-for-dom-mloperand-operand-slot\u2466\u24ea"}], "title": "7.6.27. Reduction operations"}, {"refs": [{"id": "ref-for-dom-mloperand-operand-slot\u2466\u2460"}, {"id": "ref-for-dom-mloperand-operand-slot\u2466\u2461"}, {"id": "ref-for-dom-mloperand-operand-slot\u2466\u2462"}], "title": "7.6.28.1. The relu(input) method"}, {"refs": [{"id": "ref-for-dom-mloperand-operand-slot\u2466\u2463"}, {"id": "ref-for-dom-mloperand-operand-slot\u2466\u2464"}, {"id": "ref-for-dom-mloperand-operand-slot\u2466\u2465"}], "title": "7.6.29. The resample2d() method"}, {"refs": [{"id": "ref-for-dom-mloperand-operand-slot\u2466\u2466"}, {"id": "ref-for-dom-mloperand-operand-slot\u2466\u2467"}, {"id": "ref-for-dom-mloperand-operand-slot\u2466\u2468"}], "title": "7.6.30. The reshape() method"}, {"refs": [{"id": "ref-for-dom-mloperand-operand-slot\u2467\u24ea"}, {"id": "ref-for-dom-mloperand-operand-slot\u2467\u2460"}, {"id": "ref-for-dom-mloperand-operand-slot\u2467\u2461"}], "title": "7.6.31.1. The sigmoid(input) method"}, {"refs": [{"id": "ref-for-dom-mloperand-operand-slot\u2467\u2462"}, {"id": "ref-for-dom-mloperand-operand-slot\u2467\u2463"}, {"id": "ref-for-dom-mloperand-operand-slot\u2467\u2464"}], "title": "7.6.32. The slice() method"}, {"refs": [{"id": "ref-for-dom-mloperand-operand-slot\u2467\u2465"}, {"id": "ref-for-dom-mloperand-operand-slot\u2467\u2466"}, {"id": "ref-for-dom-mloperand-operand-slot\u2467\u2467"}], "title": "7.6.33.1. The softmax(input) method"}, {"refs": [{"id": "ref-for-dom-mloperand-operand-slot\u2467\u2468"}, {"id": "ref-for-dom-mloperand-operand-slot\u2468\u24ea"}, {"id": "ref-for-dom-mloperand-operand-slot\u2468\u2460"}], "title": "7.6.34.1. The softplus(input, options) method"}, {"refs": [{"id": "ref-for-dom-mloperand-operand-slot\u2468\u2461"}, {"id": "ref-for-dom-mloperand-operand-slot\u2468\u2462"}, {"id": "ref-for-dom-mloperand-operand-slot\u2468\u2463"}], "title": "7.6.35.1. The softsign(input) method"}, {"refs": [{"id": "ref-for-dom-mloperand-operand-slot\u2468\u2464"}, {"id": "ref-for-dom-mloperand-operand-slot\u2468\u2465"}, {"id": "ref-for-dom-mloperand-operand-slot\u2468\u2466"}], "title": "7.6.36. The split() method"}, {"refs": [{"id": "ref-for-dom-mloperand-operand-slot\u2468\u2467"}, {"id": "ref-for-dom-mloperand-operand-slot\u2468\u2468"}, {"id": "ref-for-dom-mloperand-operand-slot\u2460\u24ea\u24ea"}], "title": "7.6.37. The squeeze() method"}, {"refs": [{"id": "ref-for-dom-mloperand-operand-slot\u2460\u24ea\u2460"}, {"id": "ref-for-dom-mloperand-operand-slot\u2460\u24ea\u2461"}, {"id": "ref-for-dom-mloperand-operand-slot\u2460\u24ea\u2462"}], "title": "7.6.38.1. The tanh(input) method"}, {"refs": [{"id": "ref-for-dom-mloperand-operand-slot\u2460\u24ea\u2463"}, {"id": "ref-for-dom-mloperand-operand-slot\u2460\u24ea\u2464"}, {"id": "ref-for-dom-mloperand-operand-slot\u2460\u24ea\u2465"}], "title": "7.6.39. The transpose() method"}], "external": false};
+window.dfnpanelData['dom-mloperand-operator-slot'] = {"dfnID": "dom-mloperand-operator-slot", "url": "#dom-mloperand-operator-slot", "dfnText": "[[operator]]", "refSections": [{"refs": [{"id": "ref-for-dom-mloperand-operator-slot"}], "title": "7.6.3.2. The buildSync(outputs) method"}, {"refs": [{"id": "ref-for-dom-mloperand-operator-slot\u2460"}], "title": "7.6.6.1. The clamp(operand, options) method"}, {"refs": [{"id": "ref-for-dom-mloperand-operator-slot\u2461"}], "title": "7.6.7. The concat() method"}, {"refs": [{"id": "ref-for-dom-mloperand-operator-slot\u2462"}], "title": "7.6.8. The conv2d() method"}, {"refs": [{"id": "ref-for-dom-mloperand-operator-slot\u2463"}], "title": "7.6.9. The convTranspose2d() method"}, {"refs": [{"id": "ref-for-dom-mloperand-operator-slot\u2464"}], "title": "7.6.10. Element-wise binary operations"}, {"refs": [{"id": "ref-for-dom-mloperand-operator-slot\u2465"}], "title": "7.6.11. Element-wise unary operations"}, {"refs": [{"id": "ref-for-dom-mloperand-operator-slot\u2466"}], "title": "7.6.12.1. The elu(input, options) method"}, {"refs": [{"id": "ref-for-dom-mloperand-operator-slot\u2467"}], "title": "7.6.13. The gemm() method"}, {"refs": [{"id": "ref-for-dom-mloperand-operator-slot\u2468"}], "title": "7.6.15. The gruCell() method"}, {"refs": [{"id": "ref-for-dom-mloperand-operator-slot\u2460\u24ea"}], "title": "7.6.16.1. The hardSigmoid(input, options) method"}, {"refs": [{"id": "ref-for-dom-mloperand-operator-slot\u2460\u2460"}], "title": "7.6.17.1. The hardSwish(input) method"}, {"refs": [{"id": "ref-for-dom-mloperand-operator-slot\u2460\u2461"}], "title": "7.6.18. The instanceNormalization() method"}, {"refs": [{"id": "ref-for-dom-mloperand-operator-slot\u2460\u2462"}], "title": "7.6.19.1. The leakyRelu(input, options) method"}, {"refs": [{"id": "ref-for-dom-mloperand-operator-slot\u2460\u2463"}], "title": "7.6.20.1. The linear(input, options) method"}, {"refs": [{"id": "ref-for-dom-mloperand-operator-slot\u2460\u2464"}, {"id": "ref-for-dom-mloperand-operator-slot\u2460\u2465"}, {"id": "ref-for-dom-mloperand-operator-slot\u2460\u2466"}], "title": "7.6.21. The lstm() method"}, {"refs": [{"id": "ref-for-dom-mloperand-operator-slot\u2460\u2467"}, {"id": "ref-for-dom-mloperand-operator-slot\u2460\u2468"}], "title": "7.6.22. The lstmCell() method"}, {"refs": [{"id": "ref-for-dom-mloperand-operator-slot\u2461\u24ea"}], "title": "7.6.23. The matmul() method"}, {"refs": [{"id": "ref-for-dom-mloperand-operator-slot\u2461\u2460"}], "title": "7.6.24. The pad() method"}, {"refs": [{"id": "ref-for-dom-mloperand-operator-slot\u2461\u2461"}], "title": "7.6.25. Pooling operations"}, {"refs": [{"id": "ref-for-dom-mloperand-operator-slot\u2461\u2462"}], "title": "7.6.26. The prelu() method"}, {"refs": [{"id": "ref-for-dom-mloperand-operator-slot\u2461\u2463"}], "title": "7.6.27. Reduction operations"}, {"refs": [{"id": "ref-for-dom-mloperand-operator-slot\u2461\u2464"}], "title": "7.6.28.1. The relu(input) method"}, {"refs": [{"id": "ref-for-dom-mloperand-operator-slot\u2461\u2465"}], "title": "7.6.29. The resample2d() method"}, {"refs": [{"id": "ref-for-dom-mloperand-operator-slot\u2461\u2466"}], "title": "7.6.30. The reshape() method"}, {"refs": [{"id": "ref-for-dom-mloperand-operator-slot\u2461\u2467"}], "title": "7.6.31.1. The sigmoid(input) method"}, {"refs": [{"id": "ref-for-dom-mloperand-operator-slot\u2461\u2468"}], "title": "7.6.32. The slice() method"}, {"refs": [{"id": "ref-for-dom-mloperand-operator-slot\u2462\u24ea"}], "title": "7.6.33.1. The softmax(input) method"}, {"refs": [{"id": "ref-for-dom-mloperand-operator-slot\u2462\u2460"}], "title": "7.6.34.1. The softplus(input, options) method"}, {"refs": [{"id": "ref-for-dom-mloperand-operator-slot\u2462\u2461"}], "title": "7.6.35.1. The softsign(input) method"}, {"refs": [{"id": "ref-for-dom-mloperand-operator-slot\u2462\u2462"}], "title": "7.6.36. The split() method"}, {"refs": [{"id": "ref-for-dom-mloperand-operator-slot\u2462\u2463"}], "title": "7.6.37. The squeeze() method"}, {"refs": [{"id": "ref-for-dom-mloperand-operator-slot\u2462\u2464"}], "title": "7.6.38.1. The tanh(input) method"}, {"refs": [{"id": "ref-for-dom-mloperand-operator-slot\u2462\u2465"}], "title": "7.6.39. The transpose() method"}], "external": false};
+window.dfnpanelData['mloperand-rank'] = {"dfnID": "mloperand-rank", "url": "#mloperand-rank", "dfnText": "rank", "refSections": [{"refs": [{"id": "ref-for-mloperand-rank"}, {"id": "ref-for-mloperand-rank\u2460"}], "title": "7.6.5. The batchNormalization() method"}, {"refs": [{"id": "ref-for-mloperand-rank\u2461"}, {"id": "ref-for-mloperand-rank\u2462"}, {"id": "ref-for-mloperand-rank\u2463"}, {"id": "ref-for-mloperand-rank\u2464"}], "title": "7.6.7. The concat() method"}, {"refs": [{"id": "ref-for-mloperand-rank\u2465"}, {"id": "ref-for-mloperand-rank\u2466"}], "title": "7.6.10. Element-wise binary operations"}, {"refs": [{"id": "ref-for-mloperand-rank\u2467"}], "title": "7.6.14. The gru() method"}, {"refs": [{"id": "ref-for-mloperand-rank\u2468"}], "title": "7.6.15. The gruCell() method"}, {"refs": [{"id": "ref-for-mloperand-rank\u2460\u24ea"}, {"id": "ref-for-mloperand-rank\u2460\u2460"}, {"id": "ref-for-mloperand-rank\u2460\u2461"}], "title": "7.6.18. The instanceNormalization() method"}, {"refs": [{"id": "ref-for-mloperand-rank\u2460\u2462"}], "title": "7.6.22. The lstmCell() method"}, {"refs": [{"id": "ref-for-mloperand-rank\u2460\u2463"}], "title": "7.6.23. The matmul() method"}, {"refs": [{"id": "ref-for-mloperand-rank\u2460\u2464"}, {"id": "ref-for-mloperand-rank\u2460\u2465"}, {"id": "ref-for-mloperand-rank\u2460\u2466"}, {"id": "ref-for-mloperand-rank\u2460\u2467"}], "title": "7.6.24. The pad() method"}, {"refs": [{"id": "ref-for-mloperand-rank\u2460\u2468"}, {"id": "ref-for-mloperand-rank\u2461\u24ea"}], "title": "7.6.27. Reduction operations"}, {"refs": [{"id": "ref-for-mloperand-rank\u2461\u2460"}, {"id": "ref-for-mloperand-rank\u2461\u2461"}, {"id": "ref-for-mloperand-rank\u2461\u2462"}], "title": "7.6.32. The slice() method"}, {"refs": [{"id": "ref-for-mloperand-rank\u2461\u2463"}], "title": "7.6.36. The split() method"}, {"refs": [{"id": "ref-for-mloperand-rank\u2461\u2464"}, {"id": "ref-for-mloperand-rank\u2461\u2465"}, {"id": "ref-for-mloperand-rank\u2461\u2466"}], "title": "7.6.37. The squeeze() method"}, {"refs": [{"id": "ref-for-mloperand-rank\u2461\u2467"}, {"id": "ref-for-mloperand-rank\u2461\u2468"}, {"id": "ref-for-mloperand-rank\u2462\u24ea"}, {"id": "ref-for-mloperand-rank\u2462\u2460"}, {"id": "ref-for-mloperand-rank\u2462\u2461"}], "title": "7.6.39. The transpose() method"}], "external": false};
+window.dfnpanelData['create-an-mloperand'] = {"dfnID": "create-an-mloperand", "url": "#create-an-mloperand", "dfnText": "create an MLOperand", "refSections": [{"refs": [{"id": "ref-for-create-an-mloperand"}], "title": "7.6.2. The input() method"}, {"refs": [{"id": "ref-for-create-an-mloperand\u2460"}], "title": "7.6.4.1. The constant(descriptor, bufferView) method"}, {"refs": [{"id": "ref-for-create-an-mloperand\u2461"}], "title": "7.6.4.2. The constant(value, type) method"}, {"refs": [{"id": "ref-for-create-an-mloperand\u2462"}], "title": "7.6.5. The batchNormalization() method"}, {"refs": [{"id": "ref-for-create-an-mloperand\u2463"}], "title": "7.6.7. The concat() method"}, {"refs": [{"id": "ref-for-create-an-mloperand\u2464"}], "title": "7.6.8. The conv2d() method"}, {"refs": [{"id": "ref-for-create-an-mloperand\u2465"}], "title": "7.6.9. The convTranspose2d() method"}, {"refs": [{"id": "ref-for-create-an-mloperand\u2466"}], "title": "7.6.10. Element-wise binary operations"}, {"refs": [{"id": "ref-for-create-an-mloperand\u2467"}], "title": "7.6.13. The gemm() method"}, {"refs": [{"id": "ref-for-create-an-mloperand\u2468"}], "title": "7.6.15. The gruCell() method"}, {"refs": [{"id": "ref-for-create-an-mloperand\u2460\u24ea"}, {"id": "ref-for-create-an-mloperand\u2460\u2460"}, {"id": "ref-for-create-an-mloperand\u2460\u2461"}], "title": "7.6.21. The lstm() method"}, {"refs": [{"id": "ref-for-create-an-mloperand\u2460\u2462"}, {"id": "ref-for-create-an-mloperand\u2460\u2463"}], "title": "7.6.22. The lstmCell() method"}, {"refs": [{"id": "ref-for-create-an-mloperand\u2460\u2464"}], "title": "7.6.23. The matmul() method"}, {"refs": [{"id": "ref-for-create-an-mloperand\u2460\u2465"}], "title": "7.6.24. The pad() method"}, {"refs": [{"id": "ref-for-create-an-mloperand\u2460\u2466"}], "title": "7.6.25. Pooling operations"}, {"refs": [{"id": "ref-for-create-an-mloperand\u2460\u2467"}], "title": "7.6.26. The prelu() method"}, {"refs": [{"id": "ref-for-create-an-mloperand\u2460\u2468"}], "title": "7.6.29. The resample2d() method"}, {"refs": [{"id": "ref-for-create-an-mloperand\u2461\u24ea"}], "title": "7.6.30. The reshape() method"}], "external": false};
+window.dfnpanelData['copy-an-mloperand'] = {"dfnID": "copy-an-mloperand", "url": "#copy-an-mloperand", "dfnText": "copy an MLOperand", "refSections": [{"refs": [{"id": "ref-for-copy-an-mloperand"}], "title": "7.6.6.1. The clamp(operand, options) method"}, {"refs": [{"id": "ref-for-copy-an-mloperand\u2460"}], "title": "7.6.11. Element-wise unary operations"}, {"refs": [{"id": "ref-for-copy-an-mloperand\u2461"}], "title": "7.6.12.1. The elu(input, options) method"}, {"refs": [{"id": "ref-for-copy-an-mloperand\u2462"}], "title": "7.6.16.1. The hardSigmoid(input, options) method"}, {"refs": [{"id": "ref-for-copy-an-mloperand\u2463"}], "title": "7.6.17.1. The hardSwish(input) method"}, {"refs": [{"id": "ref-for-copy-an-mloperand\u2464"}], "title": "7.6.18. The instanceNormalization() method"}, {"refs": [{"id": "ref-for-copy-an-mloperand\u2465"}], "title": "7.6.19.1. The leakyRelu(input, options) method"}, {"refs": [{"id": "ref-for-copy-an-mloperand\u2466"}], "title": "7.6.20.1. The linear(input, options) method"}, {"refs": [{"id": "ref-for-copy-an-mloperand\u2467"}], "title": "7.6.27. Reduction operations"}, {"refs": [{"id": "ref-for-copy-an-mloperand\u2468"}], "title": "7.6.28.1. The relu(input) method"}, {"refs": [{"id": "ref-for-copy-an-mloperand\u2460\u24ea"}], "title": "7.6.31.1. The sigmoid(input) method"}, {"refs": [{"id": "ref-for-copy-an-mloperand\u2460\u2460"}], "title": "7.6.32. The slice() method"}, {"refs": [{"id": "ref-for-copy-an-mloperand\u2460\u2461"}], "title": "7.6.33.1. The softmax(input) method"}, {"refs": [{"id": "ref-for-copy-an-mloperand\u2460\u2462"}], "title": "7.6.34.1. The softplus(input, options) method"}, {"refs": [{"id": "ref-for-copy-an-mloperand\u2460\u2463"}], "title": "7.6.35.1. The softsign(input) method"}, {"refs": [{"id": "ref-for-copy-an-mloperand\u2460\u2464"}], "title": "7.6.36. The split() method"}, {"refs": [{"id": "ref-for-copy-an-mloperand\u2460\u2465"}], "title": "7.6.37. The squeeze() method"}, {"refs": [{"id": "ref-for-copy-an-mloperand\u2460\u2466"}], "title": "7.6.38.1. The tanh(input) method"}, {"refs": [{"id": "ref-for-copy-an-mloperand\u2460\u2467"}], "title": "7.6.39. The transpose() method"}], "external": false};
+window.dfnpanelData['check-dimensions'] = {"dfnID": "check-dimensions", "url": "#check-dimensions", "dfnText": "check dimensions", "refSections": [{"refs": [{"id": "ref-for-check-dimensions"}], "title": "7.3.2.1. Creating MLOperand"}, {"refs": [{"id": "ref-for-check-dimensions\u2460"}], "title": "7.6.2. The input() method"}, {"refs": [{"id": "ref-for-check-dimensions\u2461"}], "title": "7.6.4.1. The constant(descriptor, bufferView) method"}], "external": false};
+window.dfnpanelData['mloperand-validate-mloperand'] = {"dfnID": "mloperand-validate-mloperand", "url": "#mloperand-validate-mloperand", "dfnText": "validate MLOperand", "refSections": [{"refs": [{"id": "ref-for-mloperand-validate-mloperand"}], "title": "7.6.3.2. The buildSync(outputs) method"}, {"refs": [{"id": "ref-for-mloperand-validate-mloperand\u2460"}], "title": "7.6.7. The concat() method"}], "external": false};
+window.dfnpanelData['mlactivation'] = {"dfnID": "mlactivation", "url": "#mlactivation", "dfnText": "MLActivation", "refSections": [{"refs": [{"id": "ref-for-mlactivation"}, {"id": "ref-for-mlactivation\u2460"}, {"id": "ref-for-mlactivation\u2461"}, {"id": "ref-for-mlactivation\u2462"}, {"id": "ref-for-mlactivation\u2463"}, {"id": "ref-for-mlactivation\u2464"}], "title": "7.3.3. The MLActivation interface"}, {"refs": [{"id": "ref-for-mlactivation\u2465"}, {"id": "ref-for-mlactivation\u2466"}], "title": "7.3.3.1. Creating MLActivation"}, {"refs": [{"id": "ref-for-mlactivation\u2467"}, {"id": "ref-for-mlactivation\u2468"}, {"id": "ref-for-mlactivation\u2460\u24ea"}], "title": "7.6.5. The batchNormalization() method"}, {"refs": [{"id": "ref-for-mlactivation\u2460\u2460"}], "title": "7.6.6. The clamp() method"}, {"refs": [{"id": "ref-for-mlactivation\u2460\u2461"}], "title": "7.6.6.2. The clamp(options) method"}, {"refs": [{"id": "ref-for-mlactivation\u2460\u2462"}, {"id": "ref-for-mlactivation\u2460\u2463"}, {"id": "ref-for-mlactivation\u2460\u2464"}, {"id": "ref-for-mlactivation\u2460\u2465"}], "title": "7.6.8. The conv2d() method"}, {"refs": [{"id": "ref-for-mlactivation\u2460\u2466"}, {"id": "ref-for-mlactivation\u2460\u2467"}, {"id": "ref-for-mlactivation\u2460\u2468"}, {"id": "ref-for-mlactivation\u2461\u24ea"}], "title": "7.6.9. The convTranspose2d() method"}, {"refs": [{"id": "ref-for-mlactivation\u2461\u2460"}], "title": "7.6.12. The elu() method"}, {"refs": [{"id": "ref-for-mlactivation\u2461\u2461"}], "title": "7.6.12.2. The elu(options) method"}, {"refs": [{"id": "ref-for-mlactivation\u2461\u2462"}, {"id": "ref-for-mlactivation\u2461\u2463"}, {"id": "ref-for-mlactivation\u2461\u2464"}], "title": "7.6.14. The gru() method"}, {"refs": [{"id": "ref-for-mlactivation\u2461\u2465"}, {"id": "ref-for-mlactivation\u2461\u2466"}, {"id": "ref-for-mlactivation\u2461\u2467"}], "title": "7.6.15. The gruCell() method"}, {"refs": [{"id": "ref-for-mlactivation\u2461\u2468"}], "title": "7.6.16. The hardSigmoid() method"}, {"refs": [{"id": "ref-for-mlactivation\u2462\u24ea"}], "title": "7.6.16.2. The hardSigmoid(options) method"}, {"refs": [{"id": "ref-for-mlactivation\u2462\u2460"}], "title": "7.6.17. The hardSwish() method"}, {"refs": [{"id": "ref-for-mlactivation\u2462\u2461"}], "title": "7.6.17.2. The hardSwish() method"}, {"refs": [{"id": "ref-for-mlactivation\u2462\u2462"}], "title": "7.6.19. The leakyRelu() method"}, {"refs": [{"id": "ref-for-mlactivation\u2462\u2463"}], "title": "7.6.19.2. The leakyRelu(options) method"}, {"refs": [{"id": "ref-for-mlactivation\u2462\u2464"}], "title": "7.6.20. The linear() method"}, {"refs": [{"id": "ref-for-mlactivation\u2462\u2465"}], "title": "7.6.20.2. The linear(options) method"}, {"refs": [{"id": "ref-for-mlactivation\u2462\u2466"}, {"id": "ref-for-mlactivation\u2462\u2467"}, {"id": "ref-for-mlactivation\u2462\u2468"}, {"id": "ref-for-mlactivation\u2463\u24ea"}], "title": "7.6.21. The lstm() method"}, {"refs": [{"id": "ref-for-mlactivation\u2463\u2460"}, {"id": "ref-for-mlactivation\u2463\u2461"}, {"id": "ref-for-mlactivation\u2463\u2462"}, {"id": "ref-for-mlactivation\u2463\u2463"}], "title": "7.6.22. The lstmCell() method"}, {"refs": [{"id": "ref-for-mlactivation\u2463\u2464"}], "title": "7.6.28. The relu() method"}, {"refs": [{"id": "ref-for-mlactivation\u2463\u2465"}], "title": "7.6.28.2. The relu() method"}, {"refs": [{"id": "ref-for-mlactivation\u2463\u2466"}], "title": "7.6.31. The sigmoid() method"}, {"refs": [{"id": "ref-for-mlactivation\u2463\u2467"}], "title": "7.6.31.2. The sigmoid() method"}, {"refs": [{"id": "ref-for-mlactivation\u2463\u2468"}], "title": "7.6.33. The softmax() method"}, {"refs": [{"id": "ref-for-mlactivation\u2464\u24ea"}], "title": "7.6.33.2. The softmax() method"}, {"refs": [{"id": "ref-for-mlactivation\u2464\u2460"}], "title": "7.6.34. The softplus() method"}, {"refs": [{"id": "ref-for-mlactivation\u2464\u2461"}], "title": "7.6.34.2. The softplus(options) method"}, {"refs": [{"id": "ref-for-mlactivation\u2464\u2462"}], "title": "7.6.35. The softsign() method"}, {"refs": [{"id": "ref-for-mlactivation\u2464\u2463"}], "title": "7.6.35.2. The softsign() method"}, {"refs": [{"id": "ref-for-mlactivation\u2464\u2464"}], "title": "7.6.38. The tanh() method"}, {"refs": [{"id": "ref-for-mlactivation\u2464\u2465"}], "title": "7.6.38.2. The tanh() method"}], "external": false};
+window.dfnpanelData['dom-mlactivation-name-slot'] = {"dfnID": "dom-mlactivation-name-slot", "url": "#dom-mlactivation-name-slot", "dfnText": "[[name]]", "refSections": [{"refs": [{"id": "ref-for-dom-mlactivation-name-slot"}], "title": "7.3.3.1. Creating MLActivation"}], "external": false};
+window.dfnpanelData['dom-mlactivation-builder-slot'] = {"dfnID": "dom-mlactivation-builder-slot", "url": "#dom-mlactivation-builder-slot", "dfnText": "[[builder]]", "refSections": [{"refs": [{"id": "ref-for-dom-mlactivation-builder-slot"}], "title": "7.3.3.1. Creating MLActivation"}], "external": false};
+window.dfnpanelData['dom-mlactivation-options-slot'] = {"dfnID": "dom-mlactivation-options-slot", "url": "#dom-mlactivation-options-slot", "dfnText": "[[options]]", "refSections": [{"refs": [{"id": "ref-for-dom-mlactivation-options-slot"}], "title": "7.3.3.1. Creating MLActivation"}], "external": false};
+window.dfnpanelData['dom-mlactivation-operator-slot'] = {"dfnID": "dom-mlactivation-operator-slot", "url": "#dom-mlactivation-operator-slot", "dfnText": "[[operator]]", "refSections": [{"refs": [{"id": "ref-for-dom-mlactivation-operator-slot"}, {"id": "ref-for-dom-mlactivation-operator-slot\u2460"}], "title": "7.3.3.1. Creating MLActivation"}], "external": false};
+window.dfnpanelData['create-an-mlactivation'] = {"dfnID": "create-an-mlactivation", "url": "#create-an-mlactivation", "dfnText": "create an MLActivation", "refSections": [{"refs": [{"id": "ref-for-create-an-mlactivation"}], "title": "7.6.6.2. The clamp(options) method"}, {"refs": [{"id": "ref-for-create-an-mlactivation\u2460"}], "title": "7.6.12.2. The elu(options) method"}, {"refs": [{"id": "ref-for-create-an-mlactivation\u2461"}], "title": "7.6.16.2. The hardSigmoid(options) method"}, {"refs": [{"id": "ref-for-create-an-mlactivation\u2462"}], "title": "7.6.17.2. The hardSwish() method"}, {"refs": [{"id": "ref-for-create-an-mlactivation\u2463"}], "title": "7.6.19.2. The leakyRelu(options) method"}, {"refs": [{"id": "ref-for-create-an-mlactivation\u2464"}], "title": "7.6.20.2. The linear(options) method"}, {"refs": [{"id": "ref-for-create-an-mlactivation\u2465"}], "title": "7.6.28.2. The relu() method"}, {"refs": [{"id": "ref-for-create-an-mlactivation\u2466"}], "title": "7.6.31.2. The sigmoid() method"}, {"refs": [{"id": "ref-for-create-an-mlactivation\u2467"}], "title": "7.6.33.2. The softmax() method"}, {"refs": [{"id": "ref-for-create-an-mlactivation\u2468"}], "title": "7.6.34.2. The softplus(options) method"}, {"refs": [{"id": "ref-for-create-an-mlactivation\u2460\u24ea"}], "title": "7.6.35.2. The softsign() method"}, {"refs": [{"id": "ref-for-create-an-mlactivation\u2460\u2460"}], "title": "7.6.38.2. The tanh() method"}], "external": false};
 window.dfnpanelData['context-type'] = {"dfnID": "context-type", "url": "#context-type", "dfnText": "context type", "refSections": [{"refs": [{"id": "ref-for-context-type"}, {"id": "ref-for-context-type\u2460"}, {"id": "ref-for-context-type\u2461"}], "title": "7.4. The MLContext interface"}], "external": false};
 window.dfnpanelData['default-context'] = {"dfnID": "default-context", "url": "#default-context", "dfnText": "default", "refSections": [{"refs": [{"id": "ref-for-default-context"}], "title": "7.2.2. The createContext() method"}, {"refs": [{"id": "ref-for-default-context\u2460"}], "title": "7.4. The MLContext interface"}, {"refs": [{"id": "ref-for-default-context\u2461"}], "title": "7.4.1. The MLContext validation algorithm"}, {"refs": [{"id": "ref-for-default-context\u2462"}], "title": "7.4.2. Synchronous Execution"}, {"refs": [{"id": "ref-for-default-context\u2463"}], "title": "7.4.4. Asynchronous Execution"}, {"refs": [{"id": "ref-for-default-context\u2464"}], "title": "7.6. The MLGraphBuilder interface"}], "external": false};
 window.dfnpanelData['webgpu-context'] = {"dfnID": "webgpu-context", "url": "#webgpu-context", "dfnText": "webgpu", "refSections": [{"refs": [{"id": "ref-for-webgpu-context"}], "title": "7.2.2. The createContext() method"}, {"refs": [{"id": "ref-for-webgpu-context\u2460"}], "title": "7.4.1. The MLContext validation algorithm"}], "external": false};
@@ -5943,68 +9368,242 @@ <h2 class="no-num no-ref heading settled" id="issues-index"><span class="content
 window.dfnpanelData['power-preference-high-performance'] = {"dfnID": "power-preference-high-performance", "url": "#power-preference-high-performance", "dfnText": "high-performance", "refSections": [{"refs": [{"id": "ref-for-power-preference-high-performance"}], "title": "7.4.1. The MLContext validation algorithm"}], "external": false};
 window.dfnpanelData['power-preference-low-power'] = {"dfnID": "power-preference-low-power", "url": "#power-preference-low-power", "dfnText": "low-power", "refSections": [{"refs": [{"id": "ref-for-power-preference-low-power"}], "title": "7.4.1. The MLContext validation algorithm"}], "external": false};
 window.dfnpanelData['typedefdef-mlnamedarraybufferviews'] = {"dfnID": "typedefdef-mlnamedarraybufferviews", "url": "#typedefdef-mlnamedarraybufferviews", "dfnText": "MLNamedArrayBufferViews", "refSections": [{"refs": [{"id": "ref-for-typedefdef-mlnamedarraybufferviews"}, {"id": "ref-for-typedefdef-mlnamedarraybufferviews\u2460"}], "title": "6.1. Overview"}, {"refs": [{"id": "ref-for-typedefdef-mlnamedarraybufferviews\u2461"}, {"id": "ref-for-typedefdef-mlnamedarraybufferviews\u2462"}, {"id": "ref-for-typedefdef-mlnamedarraybufferviews\u2463"}, {"id": "ref-for-typedefdef-mlnamedarraybufferviews\u2464"}, {"id": "ref-for-typedefdef-mlnamedarraybufferviews\u2465"}, {"id": "ref-for-typedefdef-mlnamedarraybufferviews\u2466"}, {"id": "ref-for-typedefdef-mlnamedarraybufferviews\u2467"}], "title": "7.4.2. Synchronous Execution"}, {"refs": [{"id": "ref-for-typedefdef-mlnamedarraybufferviews\u2468"}, {"id": "ref-for-typedefdef-mlnamedarraybufferviews\u2460\u24ea"}, {"id": "ref-for-typedefdef-mlnamedarraybufferviews\u2460\u2460"}], "title": "7.4.3. The MLNamedArrayBufferViews transfer algorithm"}, {"refs": [{"id": "ref-for-typedefdef-mlnamedarraybufferviews\u2460\u2461"}, {"id": "ref-for-typedefdef-mlnamedarraybufferviews\u2460\u2462"}, {"id": "ref-for-typedefdef-mlnamedarraybufferviews\u2460\u2463"}, {"id": "ref-for-typedefdef-mlnamedarraybufferviews\u2460\u2464"}, {"id": "ref-for-typedefdef-mlnamedarraybufferviews\u2460\u2465"}, {"id": "ref-for-typedefdef-mlnamedarraybufferviews\u2460\u2466"}, {"id": "ref-for-typedefdef-mlnamedarraybufferviews\u2460\u2467"}, {"id": "ref-for-typedefdef-mlnamedarraybufferviews\u2460\u2468"}, {"id": "ref-for-typedefdef-mlnamedarraybufferviews\u2461\u24ea"}], "title": "7.4.4. Asynchronous Execution"}], "external": false};
-window.dfnpanelData['mlcontext'] = {"dfnID": "mlcontext", "url": "#mlcontext", "dfnText": "MLContext", "refSections": [{"refs": [{"id": "ref-for-mlcontext"}, {"id": "ref-for-mlcontext\u2460"}, {"id": "ref-for-mlcontext\u2461"}], "title": "3. Security Considerations"}, {"refs": [{"id": "ref-for-mlcontext\u2462"}, {"id": "ref-for-mlcontext\u2463"}, {"id": "ref-for-mlcontext\u2464"}, {"id": "ref-for-mlcontext\u2465"}, {"id": "ref-for-mlcontext\u2466"}], "title": "6.1. Overview"}, {"refs": [{"id": "ref-for-mlcontext\u2467"}, {"id": "ref-for-mlcontext\u2468"}, {"id": "ref-for-mlcontext\u2460\u24ea"}, {"id": "ref-for-mlcontext\u2460\u2460"}], "title": "6.2. Device Selection"}, {"refs": [{"id": "ref-for-mlcontext\u2460\u2461"}, {"id": "ref-for-mlcontext\u2460\u2462"}, {"id": "ref-for-mlcontext\u2460\u2463"}, {"id": "ref-for-mlcontext\u2460\u2464"}], "title": "7.2. The ML interface"}, {"refs": [{"id": "ref-for-mlcontext\u2460\u2465"}], "title": "7.2.2. The createContext() method"}, {"refs": [{"id": "ref-for-mlcontext\u2460\u2466"}, {"id": "ref-for-mlcontext\u2460\u2467"}], "title": "7.3. The MLGraph interface"}, {"refs": [{"id": "ref-for-mlcontext\u2460\u2468"}, {"id": "ref-for-mlcontext\u2461\u24ea"}], "title": "7.3.2. The MLOperand interface"}, {"refs": [{"id": "ref-for-mlcontext\u2461\u2460"}, {"id": "ref-for-mlcontext\u2461\u2461"}, {"id": "ref-for-mlcontext\u2461\u2462"}, {"id": "ref-for-mlcontext\u2461\u2463"}, {"id": "ref-for-mlcontext\u2461\u2464"}, {"id": "ref-for-mlcontext\u2461\u2465"}], "title": "7.4. The MLContext interface"}, {"refs": [{"id": "ref-for-mlcontext\u2461\u2466"}, {"id": "ref-for-mlcontext\u2461\u2467"}], "title": "7.4.1. The MLContext validation algorithm"}, {"refs": [{"id": "ref-for-mlcontext\u2461\u2468"}, {"id": "ref-for-mlcontext\u2462\u24ea"}], "title": "7.4.2. Synchronous Execution"}, {"refs": [{"id": "ref-for-mlcontext\u2462\u2460"}, {"id": "ref-for-mlcontext\u2462\u2461"}], "title": "7.4.4. Asynchronous Execution"}, {"refs": [{"id": "ref-for-mlcontext\u2462\u2462"}, {"id": "ref-for-mlcontext\u2462\u2463"}], "title": "7.4.5. WebGPU Interoperability"}, {"refs": [{"id": "ref-for-mlcontext\u2462\u2464"}, {"id": "ref-for-mlcontext\u2462\u2465"}], "title": "7.5. The MLCommandEncoder interface"}, {"refs": [{"id": "ref-for-mlcontext\u2462\u2466"}, {"id": "ref-for-mlcontext\u2462\u2467"}, {"id": "ref-for-mlcontext\u2462\u2468"}, {"id": "ref-for-mlcontext\u2463\u24ea"}, {"id": "ref-for-mlcontext\u2463\u2460"}], "title": "7.6. The MLGraphBuilder interface"}], "external": false};
+window.dfnpanelData['mlcontext'] = {"dfnID": "mlcontext", "url": "#mlcontext", "dfnText": "MLContext", "refSections": [{"refs": [{"id": "ref-for-mlcontext"}, {"id": "ref-for-mlcontext\u2460"}, {"id": "ref-for-mlcontext\u2461"}], "title": "3. Security Considerations"}, {"refs": [{"id": "ref-for-mlcontext\u2462"}, {"id": "ref-for-mlcontext\u2463"}, {"id": "ref-for-mlcontext\u2464"}, {"id": "ref-for-mlcontext\u2465"}, {"id": "ref-for-mlcontext\u2466"}], "title": "6.1. Overview"}, {"refs": [{"id": "ref-for-mlcontext\u2467"}, {"id": "ref-for-mlcontext\u2468"}, {"id": "ref-for-mlcontext\u2460\u24ea"}, {"id": "ref-for-mlcontext\u2460\u2460"}], "title": "6.2. Device Selection"}, {"refs": [{"id": "ref-for-mlcontext\u2460\u2461"}, {"id": "ref-for-mlcontext\u2460\u2462"}, {"id": "ref-for-mlcontext\u2460\u2463"}, {"id": "ref-for-mlcontext\u2460\u2464"}], "title": "7.2. The ML interface"}, {"refs": [{"id": "ref-for-mlcontext\u2460\u2465"}], "title": "7.2.2. The createContext() method"}, {"refs": [{"id": "ref-for-mlcontext\u2460\u2466"}, {"id": "ref-for-mlcontext\u2460\u2467"}], "title": "7.3. The MLGraph interface"}, {"refs": [{"id": "ref-for-mlcontext\u2460\u2468"}, {"id": "ref-for-mlcontext\u2461\u24ea"}], "title": "7.3.2. The MLOperand interface"}, {"refs": [{"id": "ref-for-mlcontext\u2461\u2460"}, {"id": "ref-for-mlcontext\u2461\u2461"}, {"id": "ref-for-mlcontext\u2461\u2462"}, {"id": "ref-for-mlcontext\u2461\u2463"}, {"id": "ref-for-mlcontext\u2461\u2464"}, {"id": "ref-for-mlcontext\u2461\u2465"}], "title": "7.4. The MLContext interface"}, {"refs": [{"id": "ref-for-mlcontext\u2461\u2466"}], "title": "7.4.1. The MLContext validation algorithm"}, {"refs": [{"id": "ref-for-mlcontext\u2461\u2467"}, {"id": "ref-for-mlcontext\u2461\u2468"}], "title": "7.4.2. Synchronous Execution"}, {"refs": [{"id": "ref-for-mlcontext\u2462\u24ea"}, {"id": "ref-for-mlcontext\u2462\u2460"}], "title": "7.4.4. Asynchronous Execution"}, {"refs": [{"id": "ref-for-mlcontext\u2462\u2461"}, {"id": "ref-for-mlcontext\u2462\u2462"}], "title": "7.4.5. WebGPU Interoperability"}, {"refs": [{"id": "ref-for-mlcontext\u2462\u2463"}, {"id": "ref-for-mlcontext\u2462\u2464"}], "title": "7.5. The MLCommandEncoder interface"}, {"refs": [{"id": "ref-for-mlcontext\u2462\u2465"}, {"id": "ref-for-mlcontext\u2462\u2466"}, {"id": "ref-for-mlcontext\u2462\u2467"}, {"id": "ref-for-mlcontext\u2462\u2468"}, {"id": "ref-for-mlcontext\u2463\u24ea"}], "title": "7.6. The MLGraphBuilder interface"}], "external": false};
 window.dfnpanelData['dom-mlcontext-contexttype-slot'] = {"dfnID": "dom-mlcontext-contexttype-slot", "url": "#dom-mlcontext-contexttype-slot", "dfnText": "[[contextType]]", "refSections": [{"refs": [{"id": "ref-for-dom-mlcontext-contexttype-slot"}, {"id": "ref-for-dom-mlcontext-contexttype-slot\u2460"}], "title": "7.2.2. The createContext() method"}, {"refs": [{"id": "ref-for-dom-mlcontext-contexttype-slot\u2461"}], "title": "7.4. The MLContext interface"}, {"refs": [{"id": "ref-for-dom-mlcontext-contexttype-slot\u2462"}, {"id": "ref-for-dom-mlcontext-contexttype-slot\u2463"}], "title": "7.4.1. The MLContext validation algorithm"}, {"refs": [{"id": "ref-for-dom-mlcontext-contexttype-slot\u2464"}], "title": "7.4.2. Synchronous Execution"}, {"refs": [{"id": "ref-for-dom-mlcontext-contexttype-slot\u2465"}], "title": "7.4.4. Asynchronous Execution"}, {"refs": [{"id": "ref-for-dom-mlcontext-contexttype-slot\u2466"}], "title": "7.6. The MLGraphBuilder interface"}], "external": false};
 window.dfnpanelData['dom-mlcontext-devicetype-slot'] = {"dfnID": "dom-mlcontext-devicetype-slot", "url": "#dom-mlcontext-devicetype-slot", "dfnText": "[[deviceType]]", "refSections": [{"refs": [{"id": "ref-for-dom-mlcontext-devicetype-slot"}, {"id": "ref-for-dom-mlcontext-devicetype-slot\u2460"}, {"id": "ref-for-dom-mlcontext-devicetype-slot\u2461"}], "title": "7.2.2. The createContext() method"}, {"refs": [{"id": "ref-for-dom-mlcontext-devicetype-slot\u2462"}, {"id": "ref-for-dom-mlcontext-devicetype-slot\u2463"}], "title": "7.4.1. The MLContext validation algorithm"}], "external": false};
 window.dfnpanelData['dom-mlcontext-powerpreference-slot'] = {"dfnID": "dom-mlcontext-powerpreference-slot", "url": "#dom-mlcontext-powerpreference-slot", "dfnText": "[[powerPreference]]", "refSections": [{"refs": [{"id": "ref-for-dom-mlcontext-powerpreference-slot"}, {"id": "ref-for-dom-mlcontext-powerpreference-slot\u2460"}, {"id": "ref-for-dom-mlcontext-powerpreference-slot\u2461"}], "title": "7.2.2. The createContext() method"}, {"refs": [{"id": "ref-for-dom-mlcontext-powerpreference-slot\u2462"}, {"id": "ref-for-dom-mlcontext-powerpreference-slot\u2463"}], "title": "7.4.1. The MLContext validation algorithm"}], "external": false};
-window.dfnpanelData['validate-mlcontext'] = {"dfnID": "validate-mlcontext", "url": "#validate-mlcontext", "dfnText": "validate MLContext", "refSections": [{"refs": [{"id": "ref-for-validate-mlcontext"}], "title": "7.2.2. The createContext() method"}, {"refs": [{"id": "ref-for-validate-mlcontext\u2460"}], "title": "7.2.3. The createContextSync() method"}, {"refs": [{"id": "ref-for-validate-mlcontext\u2461"}], "title": "7.6.1. The MLGraphBuilder constructor"}], "external": false};
-window.dfnpanelData['dom-mlcontext-computesync'] = {"dfnID": "dom-mlcontext-computesync", "url": "#dom-mlcontext-computesync", "dfnText": "computeSync", "refSections": [{"refs": [{"id": "ref-for-dom-mlcontext-computesync"}, {"id": "ref-for-dom-mlcontext-computesync\u2460"}], "title": "6.1. Overview"}, {"refs": [{"id": "ref-for-dom-mlcontext-computesync\u2461"}], "title": "7.4.2. Synchronous Execution"}], "external": false};
+window.dfnpanelData['validate-mlcontext'] = {"dfnID": "validate-mlcontext", "url": "#validate-mlcontext", "dfnText": "validate MLContext", "refSections": [{"refs": [{"id": "ref-for-validate-mlcontext"}, {"id": "ref-for-validate-mlcontext\u2460"}], "title": "7.2.2. The createContext() method"}, {"refs": [{"id": "ref-for-validate-mlcontext\u2461"}, {"id": "ref-for-validate-mlcontext\u2462"}], "title": "7.2.3. The createContextSync() method"}, {"refs": [{"id": "ref-for-validate-mlcontext\u2463"}], "title": "7.6.1. The MLGraphBuilder constructor"}], "external": false};
+window.dfnpanelData['dom-mlcontext-computesync'] = {"dfnID": "dom-mlcontext-computesync", "url": "#dom-mlcontext-computesync", "dfnText": "computeSync(graph, inputs, outputs)", "refSections": [{"refs": [{"id": "ref-for-dom-mlcontext-computesync"}, {"id": "ref-for-dom-mlcontext-computesync\u2460"}], "title": "6.1. Overview"}, {"refs": [{"id": "ref-for-dom-mlcontext-computesync\u2461"}], "title": "7.4.2. Synchronous Execution"}], "external": false};
 window.dfnpanelData['validate-graph-resources'] = {"dfnID": "validate-graph-resources", "url": "#validate-graph-resources", "dfnText": "validate graph resources", "refSections": [{"refs": [{"id": "ref-for-validate-graph-resources"}, {"id": "ref-for-validate-graph-resources\u2460"}], "title": "7.4.2. Synchronous Execution"}, {"refs": [{"id": "ref-for-validate-graph-resources\u2461"}, {"id": "ref-for-validate-graph-resources\u2462"}], "title": "7.4.4. Asynchronous Execution"}], "external": false};
-window.dfnpanelData['validate-buffer-with-descriptor'] = {"dfnID": "validate-buffer-with-descriptor", "url": "#validate-buffer-with-descriptor", "dfnText": "validate buffer with descriptor", "refSections": [{"refs": [{"id": "ref-for-validate-buffer-with-descriptor"}], "title": "7.4.2. Synchronous Execution"}, {"refs": [{"id": "ref-for-validate-buffer-with-descriptor\u2460"}], "title": "7.6.3.1. The constant(descriptor, bufferView) method"}], "external": false};
+window.dfnpanelData['validate-buffer-with-descriptor'] = {"dfnID": "validate-buffer-with-descriptor", "url": "#validate-buffer-with-descriptor", "dfnText": "validate buffer with descriptor", "refSections": [{"refs": [{"id": "ref-for-validate-buffer-with-descriptor"}], "title": "7.4.2. Synchronous Execution"}, {"refs": [{"id": "ref-for-validate-buffer-with-descriptor\u2460"}], "title": "7.6.4.1. The constant(descriptor, bufferView) method"}], "external": false};
 window.dfnpanelData['execute-graph'] = {"dfnID": "execute-graph", "url": "#execute-graph", "dfnText": "execute graph", "refSections": [{"refs": [{"id": "ref-for-execute-graph"}], "title": "7.4.2. Synchronous Execution"}, {"refs": [{"id": "ref-for-execute-graph\u2460"}, {"id": "ref-for-execute-graph\u2461"}], "title": "7.4.4. Asynchronous Execution"}], "external": false};
 window.dfnpanelData['mlnamedarraybufferviews-transfer'] = {"dfnID": "mlnamedarraybufferviews-transfer", "url": "#mlnamedarraybufferviews-transfer", "dfnText": "transfer", "refSections": [{"refs": [{"id": "ref-for-mlnamedarraybufferviews-transfer"}, {"id": "ref-for-mlnamedarraybufferviews-transfer\u2460"}, {"id": "ref-for-mlnamedarraybufferviews-transfer\u2461"}, {"id": "ref-for-mlnamedarraybufferviews-transfer\u2462"}, {"id": "ref-for-mlnamedarraybufferviews-transfer\u2463"}], "title": "7.4.4. Asynchronous Execution"}], "external": false};
 window.dfnpanelData['dictdef-mlcomputeresult'] = {"dfnID": "dictdef-mlcomputeresult", "url": "#dictdef-mlcomputeresult", "dfnText": "MLComputeResult", "refSections": [{"refs": [{"id": "ref-for-dictdef-mlcomputeresult"}, {"id": "ref-for-dictdef-mlcomputeresult\u2460"}, {"id": "ref-for-dictdef-mlcomputeresult\u2461"}], "title": "7.4.4. Asynchronous Execution"}], "external": false};
 window.dfnpanelData['dom-mlcomputeresult-inputs'] = {"dfnID": "dom-mlcomputeresult-inputs", "url": "#dom-mlcomputeresult-inputs", "dfnText": "inputs", "refSections": [{"refs": [{"id": "ref-for-dom-mlcomputeresult-inputs"}], "title": "7.4.4. Asynchronous Execution"}], "external": false};
 window.dfnpanelData['dom-mlcomputeresult-outputs'] = {"dfnID": "dom-mlcomputeresult-outputs", "url": "#dom-mlcomputeresult-outputs", "dfnText": "outputs", "refSections": [{"refs": [{"id": "ref-for-dom-mlcomputeresult-outputs"}], "title": "7.4.4. Asynchronous Execution"}], "external": false};
-window.dfnpanelData['dom-mlcontext-compute'] = {"dfnID": "dom-mlcontext-compute", "url": "#dom-mlcontext-compute", "dfnText": "compute", "refSections": [{"refs": [{"id": "ref-for-dom-mlcontext-compute"}, {"id": "ref-for-dom-mlcontext-compute\u2460"}], "title": "3. Security Considerations"}, {"refs": [{"id": "ref-for-dom-mlcontext-compute\u2461"}, {"id": "ref-for-dom-mlcontext-compute\u2462"}], "title": "6.1. Overview"}, {"refs": [{"id": "ref-for-dom-mlcontext-compute\u2463"}], "title": "7.4.4. Asynchronous Execution"}], "external": false};
+window.dfnpanelData['dom-mlcontext-compute'] = {"dfnID": "dom-mlcontext-compute", "url": "#dom-mlcontext-compute", "dfnText": "compute(graph, inputs, outputs)", "refSections": [{"refs": [{"id": "ref-for-dom-mlcontext-compute"}, {"id": "ref-for-dom-mlcontext-compute\u2460"}], "title": "3. Security Considerations"}, {"refs": [{"id": "ref-for-dom-mlcontext-compute\u2461"}, {"id": "ref-for-dom-mlcontext-compute\u2462"}], "title": "6.1. Overview"}, {"refs": [{"id": "ref-for-dom-mlcontext-compute\u2463"}], "title": "7.4.4. Asynchronous Execution"}], "external": false};
 window.dfnpanelData['dom-mlcontext-createcommandencoder'] = {"dfnID": "dom-mlcontext-createcommandencoder", "url": "#dom-mlcontext-createcommandencoder", "dfnText": "createCommandEncoder", "refSections": [{"refs": [{"id": "ref-for-dom-mlcontext-createcommandencoder"}], "title": "6.1. Overview"}], "external": false};
 window.dfnpanelData['typedefdef-mlgpuresource'] = {"dfnID": "typedefdef-mlgpuresource", "url": "#typedefdef-mlgpuresource", "dfnText": "MLGPUResource", "refSections": [{"refs": [{"id": "ref-for-typedefdef-mlgpuresource"}], "title": "7.5. The MLCommandEncoder interface"}], "external": false};
 window.dfnpanelData['typedefdef-mlnamedgpuresources'] = {"dfnID": "typedefdef-mlnamedgpuresources", "url": "#typedefdef-mlnamedgpuresources", "dfnText": "MLNamedGPUResources", "refSections": [{"refs": [{"id": "ref-for-typedefdef-mlnamedgpuresources"}, {"id": "ref-for-typedefdef-mlnamedgpuresources\u2460"}, {"id": "ref-for-typedefdef-mlnamedgpuresources\u2461"}, {"id": "ref-for-typedefdef-mlnamedgpuresources\u2462"}, {"id": "ref-for-typedefdef-mlnamedgpuresources\u2463"}, {"id": "ref-for-typedefdef-mlnamedgpuresources\u2464"}], "title": "7.5.2. Dispatch Execution Commands"}], "external": false};
 window.dfnpanelData['mlcommandencoder'] = {"dfnID": "mlcommandencoder", "url": "#mlcommandencoder", "dfnText": "MLCommandEncoder", "refSections": [{"refs": [{"id": "ref-for-mlcommandencoder\u2460"}], "title": "6.1. Overview"}, {"refs": [{"id": "ref-for-mlcommandencoder\u2461"}, {"id": "ref-for-mlcommandencoder\u2462"}, {"id": "ref-for-mlcommandencoder\u2463"}], "title": "7.4.5. WebGPU Interoperability"}, {"refs": [{"id": "ref-for-mlcommandencoder\u2464"}, {"id": "ref-for-mlcommandencoder\u2465"}, {"id": "ref-for-mlcommandencoder\u2466"}], "title": "7.5. The MLCommandEncoder interface"}, {"refs": [{"id": "ref-for-mlcommandencoder\u2467"}], "title": "7.5.1. Graph Initialization"}, {"refs": [{"id": "ref-for-mlcommandencoder\u2468"}], "title": "7.5.2. Dispatch Execution Commands"}, {"refs": [{"id": "ref-for-mlcommandencoder\u2460\u24ea"}], "title": "7.5.3. Generate GPU Command Buffer"}], "external": false};
-window.dfnpanelData['typedefdef-mlnamedoperands'] = {"dfnID": "typedefdef-mlnamedoperands", "url": "#typedefdef-mlnamedoperands", "dfnText": "MLNamedOperands", "refSections": [{"refs": [{"id": "ref-for-typedefdef-mlnamedoperands"}, {"id": "ref-for-typedefdef-mlnamedoperands\u2460"}], "title": "7.6. The MLGraphBuilder interface"}], "external": false};
-window.dfnpanelData['dictdef-mlbufferresourceview'] = {"dfnID": "dictdef-mlbufferresourceview", "url": "#dictdef-mlbufferresourceview", "dfnText": "MLBufferResourceView", "refSections": [{"refs": [{"id": "ref-for-dictdef-mlbufferresourceview"}], "title": "7.6. The MLGraphBuilder interface"}], "external": false};
-window.dfnpanelData['typedefdef-mlbufferview'] = {"dfnID": "typedefdef-mlbufferview", "url": "#typedefdef-mlbufferview", "dfnText": "MLBufferView", "refSections": [{"refs": [{"id": "ref-for-typedefdef-mlbufferview"}], "title": "7.4.2. Synchronous Execution"}, {"refs": [{"id": "ref-for-typedefdef-mlbufferview\u2460"}], "title": "7.6. The MLGraphBuilder interface"}, {"refs": [{"id": "ref-for-typedefdef-mlbufferview\u2461"}], "title": "7.6.3.1. The constant(descriptor, bufferView) method"}], "external": false};
-window.dfnpanelData['mlgraphbuilder'] = {"dfnID": "mlgraphbuilder", "url": "#mlgraphbuilder", "dfnText": "MLGraphBuilder", "refSections": [{"refs": [{"id": "ref-for-mlgraphbuilder"}], "title": "3. Security Considerations"}, {"refs": [{"id": "ref-for-mlgraphbuilder\u2460"}, {"id": "ref-for-mlgraphbuilder\u2461"}, {"id": "ref-for-mlgraphbuilder\u2462"}, {"id": "ref-for-mlgraphbuilder\u2463"}, {"id": "ref-for-mlgraphbuilder\u2464"}, {"id": "ref-for-mlgraphbuilder\u2465"}, {"id": "ref-for-mlgraphbuilder\u2466"}], "title": "6.1. Overview"}, {"refs": [{"id": "ref-for-mlgraphbuilder\u2467"}], "title": "7.3.2. The MLOperand interface"}, {"refs": [{"id": "ref-for-mlgraphbuilder\u2468"}, {"id": "ref-for-mlgraphbuilder\u2460\u24ea"}], "title": "7.3.2.1. Creating MLOperand"}, {"refs": [{"id": "ref-for-mlgraphbuilder\u2460\u2460"}, {"id": "ref-for-mlgraphbuilder\u2460\u2461"}, {"id": "ref-for-mlgraphbuilder\u2460\u2462"}, {"id": "ref-for-mlgraphbuilder\u2460\u2463"}, {"id": "ref-for-mlgraphbuilder\u2460\u2464"}, {"id": "ref-for-mlgraphbuilder\u2460\u2465"}], "title": "7.6. The MLGraphBuilder interface"}, {"refs": [{"id": "ref-for-mlgraphbuilder\u2460\u2466"}, {"id": "ref-for-mlgraphbuilder\u2460\u2467"}], "title": "7.6.1. The MLGraphBuilder constructor"}, {"refs": [{"id": "ref-for-mlgraphbuilder\u2460\u2468"}], "title": "7.6.3. The constant() method"}, {"refs": [{"id": "ref-for-mlgraphbuilder\u2461\u24ea"}], "title": "7.6.4. The batchNormalization() method"}, {"refs": [{"id": "ref-for-mlgraphbuilder\u2461\u2460"}], "title": "7.6.5. The clamp() method"}, {"refs": [{"id": "ref-for-mlgraphbuilder\u2461\u2461"}], "title": "7.6.6. The concat() method"}, {"refs": [{"id": "ref-for-mlgraphbuilder\u2461\u2462"}], "title": "7.6.7. The conv2d() method"}, {"refs": [{"id": "ref-for-mlgraphbuilder\u2461\u2463"}], "title": "7.6.8. The convTranspose2d() method"}, {"refs": [{"id": "ref-for-mlgraphbuilder\u2461\u2464"}], "title": "7.6.9. Element-wise binary operations"}, {"refs": [{"id": "ref-for-mlgraphbuilder\u2461\u2465"}], "title": "7.6.10. Element-wise unary operations"}, {"refs": [{"id": "ref-for-mlgraphbuilder\u2461\u2466"}], "title": "7.6.11. The elu() method"}, {"refs": [{"id": "ref-for-mlgraphbuilder\u2461\u2467"}], "title": "7.6.12. The gemm() method"}, {"refs": [{"id": "ref-for-mlgraphbuilder\u2461\u2468"}], "title": "7.6.13. The gru() method"}, {"refs": [{"id": "ref-for-mlgraphbuilder\u2462\u24ea"}], "title": "7.6.14. The gruCell() method"}, {"refs": [{"id": "ref-for-mlgraphbuilder\u2462\u2460"}], "title": "7.6.15. The hardSigmoid() method"}, {"refs": [{"id": "ref-for-mlgraphbuilder\u2462\u2461"}], "title": "7.6.16. The hardSwish() method"}, {"refs": [{"id": "ref-for-mlgraphbuilder\u2462\u2462"}], "title": "7.6.17. The instanceNormalization() method"}, {"refs": [{"id": "ref-for-mlgraphbuilder\u2462\u2463"}], "title": "7.6.18. The leakyRelu() method"}, {"refs": [{"id": "ref-for-mlgraphbuilder\u2462\u2464"}], "title": "7.6.19. The linear() method"}, {"refs": [{"id": "ref-for-mlgraphbuilder\u2462\u2465"}], "title": "7.6.20. The lstm() method"}, {"refs": [{"id": "ref-for-mlgraphbuilder\u2462\u2466"}], "title": "7.6.21. The lstmCell() method"}, {"refs": [{"id": "ref-for-mlgraphbuilder\u2462\u2467"}], "title": "7.6.22. The matmul() method"}, {"refs": [{"id": "ref-for-mlgraphbuilder\u2462\u2468"}], "title": "7.6.23. The pad() method"}, {"refs": [{"id": "ref-for-mlgraphbuilder\u2463\u24ea"}], "title": "7.6.24. Pooling operations"}, {"refs": [{"id": "ref-for-mlgraphbuilder\u2463\u2460"}], "title": "7.6.25. The prelu() method"}, {"refs": [{"id": "ref-for-mlgraphbuilder\u2463\u2461"}], "title": "7.6.26. Reduction operations"}, {"refs": [{"id": "ref-for-mlgraphbuilder\u2463\u2462"}], "title": "7.6.27. The relu() method"}, {"refs": [{"id": "ref-for-mlgraphbuilder\u2463\u2463"}], "title": "7.6.28. The resample2d() method"}, {"refs": [{"id": "ref-for-mlgraphbuilder\u2463\u2464"}], "title": "7.6.29. The reshape() method"}, {"refs": [{"id": "ref-for-mlgraphbuilder\u2463\u2465"}], "title": "7.6.30. The sigmoid() method"}, {"refs": [{"id": "ref-for-mlgraphbuilder\u2463\u2466"}], "title": "7.6.31. The slice() method"}, {"refs": [{"id": "ref-for-mlgraphbuilder\u2463\u2467"}], "title": "7.6.32. The softmax() method"}, {"refs": [{"id": "ref-for-mlgraphbuilder\u2463\u2468"}], "title": "7.6.33. The softplus() method"}, {"refs": [{"id": "ref-for-mlgraphbuilder\u2464\u24ea"}], "title": "7.6.34. The softsign() method"}, {"refs": [{"id": "ref-for-mlgraphbuilder\u2464\u2460"}], "title": "7.6.35. The split() method"}, {"refs": [{"id": "ref-for-mlgraphbuilder\u2464\u2461"}], "title": "7.6.36. The squeeze() method"}, {"refs": [{"id": "ref-for-mlgraphbuilder\u2464\u2462"}], "title": "7.6.37. The tanh() method"}, {"refs": [{"id": "ref-for-mlgraphbuilder\u2464\u2463"}], "title": "7.6.38. The transpose() method"}], "external": false};
-window.dfnpanelData['dom-mlgraphbuilder-mlgraphbuilder'] = {"dfnID": "dom-mlgraphbuilder-mlgraphbuilder", "url": "#dom-mlgraphbuilder-mlgraphbuilder", "dfnText": "constructor", "refSections": [{"refs": [{"id": "ref-for-dom-mlgraphbuilder-mlgraphbuilder"}], "title": "7.3.2. The MLOperand interface"}], "external": false};
-window.dfnpanelData['dom-mlgraphbuilder-input'] = {"dfnID": "dom-mlgraphbuilder-input", "url": "#dom-mlgraphbuilder-input", "dfnText": "input", "refSections": [{"refs": [{"id": "ref-for-dom-mlgraphbuilder-input"}, {"id": "ref-for-dom-mlgraphbuilder-input\u2460"}], "title": "7.6.2. The input() method"}], "external": false};
-window.dfnpanelData['dom-mlgraphbuilder-constant'] = {"dfnID": "dom-mlgraphbuilder-constant", "url": "#dom-mlgraphbuilder-constant", "dfnText": "constant", "refSections": [{"refs": [{"id": "ref-for-dom-mlgraphbuilder-constant"}], "title": "7.5.1. Graph Initialization"}, {"refs": [{"id": "ref-for-dom-mlgraphbuilder-constant\u2460"}, {"id": "ref-for-dom-mlgraphbuilder-constant\u2461"}], "title": "7.6.3.1. The constant(descriptor, bufferView) method"}], "external": false};
-window.dfnpanelData['dom-mlgraphbuilder-constant-value-type'] = {"dfnID": "dom-mlgraphbuilder-constant-value-type", "url": "#dom-mlgraphbuilder-constant-value-type", "dfnText": "constant", "refSections": [{"refs": [{"id": "ref-for-dom-mlgraphbuilder-constant-value-type"}, {"id": "ref-for-dom-mlgraphbuilder-constant-value-type\u2460"}], "title": "7.6.3.2. The constant(value, type) method"}], "external": false};
-window.dfnpanelData['dom-mlgraphbuilder-build'] = {"dfnID": "dom-mlgraphbuilder-build", "url": "#dom-mlgraphbuilder-build", "dfnText": "build", "refSections": [{"refs": [{"id": "ref-for-dom-mlgraphbuilder-build"}], "title": "6.1. Overview"}, {"refs": [{"id": "ref-for-dom-mlgraphbuilder-build\u2460"}], "title": "7.6. The MLGraphBuilder interface"}], "external": false};
-window.dfnpanelData['dom-mlgraphbuilder-buildsync'] = {"dfnID": "dom-mlgraphbuilder-buildsync", "url": "#dom-mlgraphbuilder-buildsync", "dfnText": "buildSync", "refSections": [{"refs": [{"id": "ref-for-dom-mlgraphbuilder-buildsync"}], "title": "6.1. Overview"}, {"refs": [{"id": "ref-for-dom-mlgraphbuilder-buildsync\u2460"}, {"id": "ref-for-dom-mlgraphbuilder-buildsync\u2461"}], "title": "7.6. The MLGraphBuilder interface"}], "external": false};
-window.dfnpanelData['dom-mlgraphbuilder-context-slot'] = {"dfnID": "dom-mlgraphbuilder-context-slot", "url": "#dom-mlgraphbuilder-context-slot", "dfnText": "[[context]]", "refSections": [{"refs": [{"id": "ref-for-dom-mlgraphbuilder-context-slot"}], "title": "7.6.1. The MLGraphBuilder constructor"}], "external": false};
-window.dfnpanelData['dictdef-mlbatchnormalizationoptions'] = {"dfnID": "dictdef-mlbatchnormalizationoptions", "url": "#dictdef-mlbatchnormalizationoptions", "dfnText": "MLBatchNormalizationOptions", "refSections": [{"refs": [{"id": "ref-for-dictdef-mlbatchnormalizationoptions"}, {"id": "ref-for-dictdef-mlbatchnormalizationoptions\u2460"}], "title": "7.6.4. The batchNormalization() method"}], "external": false};
-window.dfnpanelData['dictdef-mlclampoptions'] = {"dfnID": "dictdef-mlclampoptions", "url": "#dictdef-mlclampoptions", "dfnText": "MLClampOptions", "refSections": [{"refs": [{"id": "ref-for-dictdef-mlclampoptions"}, {"id": "ref-for-dictdef-mlclampoptions\u2460"}, {"id": "ref-for-dictdef-mlclampoptions\u2461"}], "title": "7.6.5. The clamp() method"}], "external": false};
-window.dfnpanelData['enumdef-mlconv2dfilteroperandlayout'] = {"dfnID": "enumdef-mlconv2dfilteroperandlayout", "url": "#enumdef-mlconv2dfilteroperandlayout", "dfnText": "MLConv2dFilterOperandLayout", "refSections": [{"refs": [{"id": "ref-for-enumdef-mlconv2dfilteroperandlayout"}, {"id": "ref-for-enumdef-mlconv2dfilteroperandlayout\u2460"}], "title": "7.6.7. The conv2d() method"}], "external": false};
-window.dfnpanelData['enumdef-mlautopad'] = {"dfnID": "enumdef-mlautopad", "url": "#enumdef-mlautopad", "dfnText": "MLAutoPad", "refSections": [{"refs": [{"id": "ref-for-enumdef-mlautopad"}, {"id": "ref-for-enumdef-mlautopad\u2460"}], "title": "7.6.7. The conv2d() method"}, {"refs": [{"id": "ref-for-enumdef-mlautopad\u2461"}, {"id": "ref-for-enumdef-mlautopad\u2462"}], "title": "7.6.8. The convTranspose2d() method"}, {"refs": [{"id": "ref-for-enumdef-mlautopad\u2463"}, {"id": "ref-for-enumdef-mlautopad\u2464"}], "title": "7.6.24. Pooling operations"}], "external": false};
-window.dfnpanelData['dictdef-mlconv2doptions'] = {"dfnID": "dictdef-mlconv2doptions", "url": "#dictdef-mlconv2doptions", "dfnText": "MLConv2dOptions", "refSections": [{"refs": [{"id": "ref-for-dictdef-mlconv2doptions"}, {"id": "ref-for-dictdef-mlconv2doptions\u2460"}], "title": "7.6.7. The conv2d() method"}], "external": false};
-window.dfnpanelData['enumdef-mlconvtranspose2dfilteroperandlayout'] = {"dfnID": "enumdef-mlconvtranspose2dfilteroperandlayout", "url": "#enumdef-mlconvtranspose2dfilteroperandlayout", "dfnText": "MLConvTranspose2dFilterOperandLayout", "refSections": [{"refs": [{"id": "ref-for-enumdef-mlconvtranspose2dfilteroperandlayout"}, {"id": "ref-for-enumdef-mlconvtranspose2dfilteroperandlayout\u2460"}], "title": "7.6.8. The convTranspose2d() method"}], "external": false};
-window.dfnpanelData['dictdef-mlconvtranspose2doptions'] = {"dfnID": "dictdef-mlconvtranspose2doptions", "url": "#dictdef-mlconvtranspose2doptions", "dfnText": "MLConvTranspose2dOptions", "refSections": [{"refs": [{"id": "ref-for-dictdef-mlconvtranspose2doptions"}, {"id": "ref-for-dictdef-mlconvtranspose2doptions\u2460"}], "title": "7.6.8. The convTranspose2d() method"}], "external": false};
-window.dfnpanelData['dictdef-mleluoptions'] = {"dfnID": "dictdef-mleluoptions", "url": "#dictdef-mleluoptions", "dfnText": "MLEluOptions", "refSections": [{"refs": [{"id": "ref-for-dictdef-mleluoptions"}, {"id": "ref-for-dictdef-mleluoptions\u2460"}, {"id": "ref-for-dictdef-mleluoptions\u2461"}], "title": "7.6.11. The elu() method"}], "external": false};
-window.dfnpanelData['dictdef-mlgemmoptions'] = {"dfnID": "dictdef-mlgemmoptions", "url": "#dictdef-mlgemmoptions", "dfnText": "MLGemmOptions", "refSections": [{"refs": [{"id": "ref-for-dictdef-mlgemmoptions"}, {"id": "ref-for-dictdef-mlgemmoptions\u2460"}], "title": "7.6.12. The gemm() method"}], "external": false};
-window.dfnpanelData['dom-mlgraphbuilder-gemm'] = {"dfnID": "dom-mlgraphbuilder-gemm", "url": "#dom-mlgraphbuilder-gemm", "dfnText": "gemm", "refSections": [{"refs": [{"id": "ref-for-dom-mlgraphbuilder-gemm"}], "title": "6.1. Overview"}], "external": false};
-window.dfnpanelData['enumdef-mlgruweightlayout'] = {"dfnID": "enumdef-mlgruweightlayout", "url": "#enumdef-mlgruweightlayout", "dfnText": "MLGruWeightLayout", "refSections": [{"refs": [{"id": "ref-for-enumdef-mlgruweightlayout"}, {"id": "ref-for-enumdef-mlgruweightlayout\u2460"}], "title": "7.6.13. The gru() method"}, {"refs": [{"id": "ref-for-enumdef-mlgruweightlayout\u2461"}, {"id": "ref-for-enumdef-mlgruweightlayout\u2462"}], "title": "7.6.14. The gruCell() method"}], "external": false};
-window.dfnpanelData['enumdef-mlrecurrentnetworkdirection'] = {"dfnID": "enumdef-mlrecurrentnetworkdirection", "url": "#enumdef-mlrecurrentnetworkdirection", "dfnText": "MLRecurrentNetworkDirection", "refSections": [{"refs": [{"id": "ref-for-enumdef-mlrecurrentnetworkdirection"}, {"id": "ref-for-enumdef-mlrecurrentnetworkdirection\u2460"}], "title": "7.6.13. The gru() method"}, {"refs": [{"id": "ref-for-enumdef-mlrecurrentnetworkdirection\u2461"}, {"id": "ref-for-enumdef-mlrecurrentnetworkdirection\u2462"}], "title": "7.6.20. The lstm() method"}], "external": false};
-window.dfnpanelData['dictdef-mlgruoptions'] = {"dfnID": "dictdef-mlgruoptions", "url": "#dictdef-mlgruoptions", "dfnText": "MLGruOptions", "refSections": [{"refs": [{"id": "ref-for-dictdef-mlgruoptions"}, {"id": "ref-for-dictdef-mlgruoptions\u2460"}], "title": "7.6.13. The gru() method"}, {"refs": [{"id": "ref-for-dictdef-mlgruoptions\u2461"}], "title": "7.6.20. The lstm() method"}], "external": false};
-window.dfnpanelData['dictdef-mlgrucelloptions'] = {"dfnID": "dictdef-mlgrucelloptions", "url": "#dictdef-mlgrucelloptions", "dfnText": "MLGruCellOptions", "refSections": [{"refs": [{"id": "ref-for-dictdef-mlgrucelloptions"}, {"id": "ref-for-dictdef-mlgrucelloptions\u2460"}], "title": "7.6.14. The gruCell() method"}], "external": false};
-window.dfnpanelData['dictdef-mlhardsigmoidoptions'] = {"dfnID": "dictdef-mlhardsigmoidoptions", "url": "#dictdef-mlhardsigmoidoptions", "dfnText": "MLHardSigmoidOptions", "refSections": [{"refs": [{"id": "ref-for-dictdef-mlhardsigmoidoptions"}, {"id": "ref-for-dictdef-mlhardsigmoidoptions\u2460"}, {"id": "ref-for-dictdef-mlhardsigmoidoptions\u2461"}], "title": "7.6.15. The hardSigmoid() method"}], "external": false};
-window.dfnpanelData['dictdef-mlinstancenormalizationoptions'] = {"dfnID": "dictdef-mlinstancenormalizationoptions", "url": "#dictdef-mlinstancenormalizationoptions", "dfnText": "MLInstanceNormalizationOptions", "refSections": [{"refs": [{"id": "ref-for-dictdef-mlinstancenormalizationoptions"}, {"id": "ref-for-dictdef-mlinstancenormalizationoptions\u2460"}], "title": "7.6.17. The instanceNormalization() method"}], "external": false};
-window.dfnpanelData['dictdef-mlleakyreluoptions'] = {"dfnID": "dictdef-mlleakyreluoptions", "url": "#dictdef-mlleakyreluoptions", "dfnText": "MLLeakyReluOptions", "refSections": [{"refs": [{"id": "ref-for-dictdef-mlleakyreluoptions"}, {"id": "ref-for-dictdef-mlleakyreluoptions\u2460"}, {"id": "ref-for-dictdef-mlleakyreluoptions\u2461"}], "title": "7.6.18. The leakyRelu() method"}], "external": false};
-window.dfnpanelData['dictdef-mllinearoptions'] = {"dfnID": "dictdef-mllinearoptions", "url": "#dictdef-mllinearoptions", "dfnText": "MLLinearOptions", "refSections": [{"refs": [{"id": "ref-for-dictdef-mllinearoptions"}, {"id": "ref-for-dictdef-mllinearoptions\u2460"}, {"id": "ref-for-dictdef-mllinearoptions\u2461"}], "title": "7.6.19. The linear() method"}], "external": false};
-window.dfnpanelData['enumdef-mllstmweightlayout'] = {"dfnID": "enumdef-mllstmweightlayout", "url": "#enumdef-mllstmweightlayout", "dfnText": "MLLstmWeightLayout", "refSections": [{"refs": [{"id": "ref-for-enumdef-mllstmweightlayout"}, {"id": "ref-for-enumdef-mllstmweightlayout\u2460"}], "title": "7.6.20. The lstm() method"}, {"refs": [{"id": "ref-for-enumdef-mllstmweightlayout\u2461"}, {"id": "ref-for-enumdef-mllstmweightlayout\u2462"}], "title": "7.6.21. The lstmCell() method"}], "external": false};
-window.dfnpanelData['dictdef-mllstmoptions'] = {"dfnID": "dictdef-mllstmoptions", "url": "#dictdef-mllstmoptions", "dfnText": "MLLstmOptions", "refSections": [{"refs": [{"id": "ref-for-dictdef-mllstmoptions"}], "title": "7.6.20. The lstm() method"}], "external": false};
-window.dfnpanelData['dictdef-mllstmcelloptions'] = {"dfnID": "dictdef-mllstmcelloptions", "url": "#dictdef-mllstmcelloptions", "dfnText": "MLLstmCellOptions", "refSections": [{"refs": [{"id": "ref-for-dictdef-mllstmcelloptions"}, {"id": "ref-for-dictdef-mllstmcelloptions\u2460"}], "title": "7.6.21. The lstmCell() method"}], "external": false};
-window.dfnpanelData['enumdef-mlpaddingmode'] = {"dfnID": "enumdef-mlpaddingmode", "url": "#enumdef-mlpaddingmode", "dfnText": "MLPaddingMode", "refSections": [{"refs": [{"id": "ref-for-enumdef-mlpaddingmode"}, {"id": "ref-for-enumdef-mlpaddingmode\u2460"}], "title": "7.6.23. The pad() method"}], "external": false};
-window.dfnpanelData['dictdef-mlpadoptions'] = {"dfnID": "dictdef-mlpadoptions", "url": "#dictdef-mlpadoptions", "dfnText": "MLPadOptions", "refSections": [{"refs": [{"id": "ref-for-dictdef-mlpadoptions"}, {"id": "ref-for-dictdef-mlpadoptions\u2460"}], "title": "7.6.23. The pad() method"}], "external": false};
-window.dfnpanelData['enumdef-mlroundingtype'] = {"dfnID": "enumdef-mlroundingtype", "url": "#enumdef-mlroundingtype", "dfnText": "MLRoundingType", "refSections": [{"refs": [{"id": "ref-for-enumdef-mlroundingtype"}, {"id": "ref-for-enumdef-mlroundingtype\u2460"}], "title": "7.6.24. Pooling operations"}], "external": false};
-window.dfnpanelData['dictdef-mlpool2doptions'] = {"dfnID": "dictdef-mlpool2doptions", "url": "#dictdef-mlpool2doptions", "dfnText": "MLPool2dOptions", "refSections": [{"refs": [{"id": "ref-for-dictdef-mlpool2doptions"}, {"id": "ref-for-dictdef-mlpool2doptions\u2460"}, {"id": "ref-for-dictdef-mlpool2doptions\u2461"}, {"id": "ref-for-dictdef-mlpool2doptions\u2462"}], "title": "7.6.24. Pooling operations"}], "external": false};
-window.dfnpanelData['dictdef-mlreduceoptions'] = {"dfnID": "dictdef-mlreduceoptions", "url": "#dictdef-mlreduceoptions", "dfnText": "MLReduceOptions", "refSections": [{"refs": [{"id": "ref-for-dictdef-mlreduceoptions"}, {"id": "ref-for-dictdef-mlreduceoptions\u2460"}, {"id": "ref-for-dictdef-mlreduceoptions\u2461"}, {"id": "ref-for-dictdef-mlreduceoptions\u2462"}, {"id": "ref-for-dictdef-mlreduceoptions\u2463"}, {"id": "ref-for-dictdef-mlreduceoptions\u2464"}, {"id": "ref-for-dictdef-mlreduceoptions\u2465"}, {"id": "ref-for-dictdef-mlreduceoptions\u2466"}, {"id": "ref-for-dictdef-mlreduceoptions\u2467"}, {"id": "ref-for-dictdef-mlreduceoptions\u2468"}, {"id": "ref-for-dictdef-mlreduceoptions\u2460\u24ea"}], "title": "7.6.26. Reduction operations"}], "external": false};
-window.dfnpanelData['enumdef-mlinterpolationmode'] = {"dfnID": "enumdef-mlinterpolationmode", "url": "#enumdef-mlinterpolationmode", "dfnText": "MLInterpolationMode", "refSections": [{"refs": [{"id": "ref-for-enumdef-mlinterpolationmode"}, {"id": "ref-for-enumdef-mlinterpolationmode\u2460"}], "title": "7.6.28. The resample2d() method"}], "external": false};
-window.dfnpanelData['dictdef-mlresample2doptions'] = {"dfnID": "dictdef-mlresample2doptions", "url": "#dictdef-mlresample2doptions", "dfnText": "MLResample2dOptions", "refSections": [{"refs": [{"id": "ref-for-dictdef-mlresample2doptions"}, {"id": "ref-for-dictdef-mlresample2doptions\u2460"}], "title": "7.6.28. The resample2d() method"}], "external": false};
-window.dfnpanelData['dom-mlgraphbuilder-softmax①'] = {"dfnID": "dom-mlgraphbuilder-softmax\u2460", "url": "#dom-mlgraphbuilder-softmax\u2460", "dfnText": "softmax", "refSections": [{"refs": [{"id": "ref-for-dom-mlgraphbuilder-softmax\u2460"}], "title": "6.1. Overview"}], "external": false};
-window.dfnpanelData['dictdef-mlsoftplusoptions'] = {"dfnID": "dictdef-mlsoftplusoptions", "url": "#dictdef-mlsoftplusoptions", "dfnText": "MLSoftplusOptions", "refSections": [{"refs": [{"id": "ref-for-dictdef-mlsoftplusoptions"}, {"id": "ref-for-dictdef-mlsoftplusoptions\u2460"}], "title": "7.6.33. The softplus() method"}], "external": false};
-window.dfnpanelData['dictdef-mlsplitoptions'] = {"dfnID": "dictdef-mlsplitoptions", "url": "#dictdef-mlsplitoptions", "dfnText": "MLSplitOptions", "refSections": [{"refs": [{"id": "ref-for-dictdef-mlsplitoptions"}, {"id": "ref-for-dictdef-mlsplitoptions\u2460"}], "title": "7.6.35. The split() method"}], "external": false};
-window.dfnpanelData['dictdef-mlsqueezeoptions'] = {"dfnID": "dictdef-mlsqueezeoptions", "url": "#dictdef-mlsqueezeoptions", "dfnText": "MLSqueezeOptions", "refSections": [{"refs": [{"id": "ref-for-dictdef-mlsqueezeoptions"}, {"id": "ref-for-dictdef-mlsqueezeoptions\u2460"}], "title": "7.6.36. The squeeze() method"}], "external": false};
-window.dfnpanelData['dictdef-mltransposeoptions'] = {"dfnID": "dictdef-mltransposeoptions", "url": "#dictdef-mltransposeoptions", "dfnText": "MLTransposeOptions", "refSections": [{"refs": [{"id": "ref-for-dictdef-mltransposeoptions"}, {"id": "ref-for-dictdef-mltransposeoptions\u2460"}], "title": "7.6.38. The transpose() method"}], "external": false};
+window.dfnpanelData['dom-mlcommandencoder-initializegraph'] = {"dfnID": "dom-mlcommandencoder-initializegraph", "url": "#dom-mlcommandencoder-initializegraph", "dfnText": "initializeGraph(graph)", "refSections": [{"refs": [{"id": "ref-for-dom-mlcommandencoder-initializegraph"}], "title": "7.5.1. Graph Initialization"}], "external": false};
+window.dfnpanelData['dom-mlcommandencoder-dispatch'] = {"dfnID": "dom-mlcommandencoder-dispatch", "url": "#dom-mlcommandencoder-dispatch", "dfnText": "dispatch(graph, inputs, outputs)", "refSections": [{"refs": [{"id": "ref-for-dom-mlcommandencoder-dispatch"}], "title": "7.5.2. Dispatch Execution Commands"}], "external": false};
+window.dfnpanelData['dom-mlcommandencoder-finish'] = {"dfnID": "dom-mlcommandencoder-finish", "url": "#dom-mlcommandencoder-finish", "dfnText": "finish(descriptor)", "refSections": [{"refs": [{"id": "ref-for-dom-mlcommandencoder-finish"}], "title": "7.5.3. Generate GPU Command Buffer"}], "external": false};
+window.dfnpanelData['typedefdef-mlnamedoperands'] = {"dfnID": "typedefdef-mlnamedoperands", "url": "#typedefdef-mlnamedoperands", "dfnText": "MLNamedOperands", "refSections": [{"refs": [{"id": "ref-for-typedefdef-mlnamedoperands"}, {"id": "ref-for-typedefdef-mlnamedoperands\u2460"}], "title": "7.6. The MLGraphBuilder interface"}, {"refs": [{"id": "ref-for-typedefdef-mlnamedoperands\u2461"}], "title": "7.6.3.2. The buildSync(outputs) method"}], "external": false};
+window.dfnpanelData['dictdef-mlbufferresourceview'] = {"dfnID": "dictdef-mlbufferresourceview", "url": "#dictdef-mlbufferresourceview", "dfnText": "MLBufferResourceView", "refSections": [{"refs": [{"id": "ref-for-dictdef-mlbufferresourceview"}, {"id": "ref-for-dictdef-mlbufferresourceview\u2460"}], "title": "7.6. The MLGraphBuilder interface"}], "external": false};
+window.dfnpanelData['typedefdef-mlbufferview'] = {"dfnID": "typedefdef-mlbufferview", "url": "#typedefdef-mlbufferview", "dfnText": "MLBufferView", "refSections": [{"refs": [{"id": "ref-for-typedefdef-mlbufferview"}], "title": "7.4.2. Synchronous Execution"}, {"refs": [{"id": "ref-for-typedefdef-mlbufferview\u2460"}], "title": "7.6. The MLGraphBuilder interface"}, {"refs": [{"id": "ref-for-typedefdef-mlbufferview\u2461"}], "title": "7.6.4.1. The constant(descriptor, bufferView) method"}], "external": false};
+window.dfnpanelData['mlgraphbuilder'] = {"dfnID": "mlgraphbuilder", "url": "#mlgraphbuilder", "dfnText": "MLGraphBuilder", "refSections": [{"refs": [{"id": "ref-for-mlgraphbuilder"}], "title": "3. Security Considerations"}, {"refs": [{"id": "ref-for-mlgraphbuilder\u2460"}, {"id": "ref-for-mlgraphbuilder\u2461"}, {"id": "ref-for-mlgraphbuilder\u2462"}, {"id": "ref-for-mlgraphbuilder\u2463"}, {"id": "ref-for-mlgraphbuilder\u2464"}, {"id": "ref-for-mlgraphbuilder\u2465"}, {"id": "ref-for-mlgraphbuilder\u2466"}], "title": "6.1. Overview"}, {"refs": [{"id": "ref-for-mlgraphbuilder\u2467"}], "title": "7.3.2. The MLOperand interface"}, {"refs": [{"id": "ref-for-mlgraphbuilder\u2468"}, {"id": "ref-for-mlgraphbuilder\u2460\u24ea"}, {"id": "ref-for-mlgraphbuilder\u2460\u2460"}], "title": "7.3.2.1. Creating MLOperand"}, {"refs": [{"id": "ref-for-mlgraphbuilder\u2460\u2461"}], "title": "7.3.3. The MLActivation interface"}, {"refs": [{"id": "ref-for-mlgraphbuilder\u2460\u2462"}, {"id": "ref-for-mlgraphbuilder\u2460\u2463"}], "title": "7.3.3.1. Creating MLActivation"}, {"refs": [{"id": "ref-for-mlgraphbuilder\u2460\u2464"}, {"id": "ref-for-mlgraphbuilder\u2460\u2465"}, {"id": "ref-for-mlgraphbuilder\u2460\u2466"}, {"id": "ref-for-mlgraphbuilder\u2460\u2467"}, {"id": "ref-for-mlgraphbuilder\u2460\u2468"}, {"id": "ref-for-mlgraphbuilder\u2461\u24ea"}], "title": "7.6. The MLGraphBuilder interface"}, {"refs": [{"id": "ref-for-mlgraphbuilder\u2461\u2460"}], "title": "7.6.1. The MLGraphBuilder constructor"}, {"refs": [{"id": "ref-for-mlgraphbuilder\u2461\u2461"}], "title": "7.6.4. The constant() method"}, {"refs": [{"id": "ref-for-mlgraphbuilder\u2461\u2462"}], "title": "7.6.5. The batchNormalization() method"}, {"refs": [{"id": "ref-for-mlgraphbuilder\u2461\u2463"}], "title": "7.6.6. The clamp() method"}, {"refs": [{"id": "ref-for-mlgraphbuilder\u2461\u2464"}], "title": "7.6.7. The concat() method"}, {"refs": [{"id": "ref-for-mlgraphbuilder\u2461\u2465"}], "title": "7.6.8. The conv2d() method"}, {"refs": [{"id": "ref-for-mlgraphbuilder\u2461\u2466"}], "title": "7.6.9. The convTranspose2d() method"}, {"refs": [{"id": "ref-for-mlgraphbuilder\u2461\u2467"}], "title": "7.6.10. Element-wise binary operations"}, {"refs": [{"id": "ref-for-mlgraphbuilder\u2461\u2468"}], "title": "7.6.11. Element-wise unary operations"}, {"refs": [{"id": "ref-for-mlgraphbuilder\u2462\u24ea"}], "title": "7.6.12. The elu() method"}, {"refs": [{"id": "ref-for-mlgraphbuilder\u2462\u2460"}], "title": "7.6.13. The gemm() method"}, {"refs": [{"id": "ref-for-mlgraphbuilder\u2462\u2461"}], "title": "7.6.14. The gru() method"}, {"refs": [{"id": "ref-for-mlgraphbuilder\u2462\u2462"}], "title": "7.6.15. The gruCell() method"}, {"refs": [{"id": "ref-for-mlgraphbuilder\u2462\u2463"}], "title": "7.6.16. The hardSigmoid() method"}, {"refs": [{"id": "ref-for-mlgraphbuilder\u2462\u2464"}], "title": "7.6.17. The hardSwish() method"}, {"refs": [{"id": "ref-for-mlgraphbuilder\u2462\u2465"}], "title": "7.6.18. The instanceNormalization() method"}, {"refs": [{"id": "ref-for-mlgraphbuilder\u2462\u2466"}], "title": "7.6.19. The leakyRelu() method"}, {"refs": [{"id": "ref-for-mlgraphbuilder\u2462\u2467"}], "title": "7.6.20. The linear() method"}, {"refs": [{"id": "ref-for-mlgraphbuilder\u2462\u2468"}], "title": "7.6.21. The lstm() method"}, {"refs": [{"id": "ref-for-mlgraphbuilder\u2463\u24ea"}], "title": "7.6.22. The lstmCell() method"}, {"refs": [{"id": "ref-for-mlgraphbuilder\u2463\u2460"}], "title": "7.6.23. The matmul() method"}, {"refs": [{"id": "ref-for-mlgraphbuilder\u2463\u2461"}], "title": "7.6.24. The pad() method"}, {"refs": [{"id": "ref-for-mlgraphbuilder\u2463\u2462"}], "title": "7.6.25. Pooling operations"}, {"refs": [{"id": "ref-for-mlgraphbuilder\u2463\u2463"}], "title": "7.6.26. The prelu() method"}, {"refs": [{"id": "ref-for-mlgraphbuilder\u2463\u2464"}], "title": "7.6.27. Reduction operations"}, {"refs": [{"id": "ref-for-mlgraphbuilder\u2463\u2465"}], "title": "7.6.28. The relu() method"}, {"refs": [{"id": "ref-for-mlgraphbuilder\u2463\u2466"}], "title": "7.6.29. The resample2d() method"}, {"refs": [{"id": "ref-for-mlgraphbuilder\u2463\u2467"}], "title": "7.6.30. The reshape() method"}, {"refs": [{"id": "ref-for-mlgraphbuilder\u2463\u2468"}], "title": "7.6.31. The sigmoid() method"}, {"refs": [{"id": "ref-for-mlgraphbuilder\u2464\u24ea"}], "title": "7.6.32. The slice() method"}, {"refs": [{"id": "ref-for-mlgraphbuilder\u2464\u2460"}], "title": "7.6.33. The softmax() method"}, {"refs": [{"id": "ref-for-mlgraphbuilder\u2464\u2461"}], "title": "7.6.34. The softplus() method"}, {"refs": [{"id": "ref-for-mlgraphbuilder\u2464\u2462"}], "title": "7.6.35. The softsign() method"}, {"refs": [{"id": "ref-for-mlgraphbuilder\u2464\u2463"}], "title": "7.6.36. The split() method"}, {"refs": [{"id": "ref-for-mlgraphbuilder\u2464\u2464"}], "title": "7.6.37. The squeeze() method"}, {"refs": [{"id": "ref-for-mlgraphbuilder\u2464\u2465"}], "title": "7.6.38. The tanh() method"}, {"refs": [{"id": "ref-for-mlgraphbuilder\u2464\u2466"}], "title": "7.6.39. The transpose() method"}], "external": false};
+window.dfnpanelData['dom-mlbufferresourceview-resource'] = {"dfnID": "dom-mlbufferresourceview-resource", "url": "#dom-mlbufferresourceview-resource", "dfnText": "resource", "refSections": [{"refs": [{"id": "ref-for-dom-mlbufferresourceview-resource"}], "title": "7.6. The MLGraphBuilder interface"}], "external": false};
+window.dfnpanelData['dom-mlbufferresourceview-offset'] = {"dfnID": "dom-mlbufferresourceview-offset", "url": "#dom-mlbufferresourceview-offset", "dfnText": "offset", "refSections": [{"refs": [{"id": "ref-for-dom-mlbufferresourceview-offset"}], "title": "7.6. The MLGraphBuilder interface"}], "external": false};
+window.dfnpanelData['dom-mlbufferresourceview-size'] = {"dfnID": "dom-mlbufferresourceview-size", "url": "#dom-mlbufferresourceview-size", "dfnText": "size", "refSections": [{"refs": [{"id": "ref-for-dom-mlbufferresourceview-size"}], "title": "7.6. The MLGraphBuilder interface"}], "external": false};
+window.dfnpanelData['dom-mlgraphbuilder-context-slot'] = {"dfnID": "dom-mlgraphbuilder-context-slot", "url": "#dom-mlgraphbuilder-context-slot", "dfnText": "[[context]]", "refSections": [{"refs": [{"id": "ref-for-dom-mlgraphbuilder-context-slot"}], "title": "7.6.1. The MLGraphBuilder constructor"}, {"refs": [{"id": "ref-for-dom-mlgraphbuilder-context-slot\u2460"}], "title": "7.6.3.2. The buildSync(outputs) method"}], "external": false};
+window.dfnpanelData['dom-mlgraphbuilder-mlgraphbuilder'] = {"dfnID": "dom-mlgraphbuilder-mlgraphbuilder", "url": "#dom-mlgraphbuilder-mlgraphbuilder", "dfnText": "MLGraphBuilder(context)", "refSections": [{"refs": [{"id": "ref-for-dom-mlgraphbuilder-mlgraphbuilder"}], "title": "7.3.2. The MLOperand interface"}, {"refs": [{"id": "ref-for-dom-mlgraphbuilder-mlgraphbuilder\u2460"}], "title": "7.6. The MLGraphBuilder interface"}], "external": false};
+window.dfnpanelData['dom-mlgraphbuilder-input'] = {"dfnID": "dom-mlgraphbuilder-input", "url": "#dom-mlgraphbuilder-input", "dfnText": "input(name, descriptor)", "refSections": [{"refs": [{"id": "ref-for-dom-mlgraphbuilder-input"}], "title": "7.6. The MLGraphBuilder interface"}, {"refs": [{"id": "ref-for-dom-mlgraphbuilder-input\u2460"}], "title": "7.6.2. The input() method"}], "external": false};
+window.dfnpanelData['dom-mlgraphbuilder-build'] = {"dfnID": "dom-mlgraphbuilder-build", "url": "#dom-mlgraphbuilder-build", "dfnText": "build(outputs)", "refSections": [{"refs": [{"id": "ref-for-dom-mlgraphbuilder-build"}], "title": "6.1. Overview"}, {"refs": [{"id": "ref-for-dom-mlgraphbuilder-build\u2460"}, {"id": "ref-for-dom-mlgraphbuilder-build\u2461"}], "title": "7.6. The MLGraphBuilder interface"}, {"refs": [{"id": "ref-for-dom-mlgraphbuilder-build\u2462"}], "title": "7.6.3.1. The build(outputs) method"}], "external": false};
+window.dfnpanelData['dom-mlgraphbuilder-buildsync'] = {"dfnID": "dom-mlgraphbuilder-buildsync", "url": "#dom-mlgraphbuilder-buildsync", "dfnText": "buildSync(outputs)", "refSections": [{"refs": [{"id": "ref-for-dom-mlgraphbuilder-buildsync"}], "title": "6.1. Overview"}, {"refs": [{"id": "ref-for-dom-mlgraphbuilder-buildsync\u2460"}, {"id": "ref-for-dom-mlgraphbuilder-buildsync\u2461"}, {"id": "ref-for-dom-mlgraphbuilder-buildsync\u2462"}], "title": "7.6. The MLGraphBuilder interface"}, {"refs": [{"id": "ref-for-dom-mlgraphbuilder-buildsync\u2463"}], "title": "7.6.3.1. The build(outputs) method"}, {"refs": [{"id": "ref-for-dom-mlgraphbuilder-buildsync\u2464"}], "title": "7.6.3.2. The buildSync(outputs) method"}], "external": false};
+window.dfnpanelData['dom-mlgraphbuilder-constant'] = {"dfnID": "dom-mlgraphbuilder-constant", "url": "#dom-mlgraphbuilder-constant", "dfnText": "constant(descriptor, bufferView)", "refSections": [{"refs": [{"id": "ref-for-dom-mlgraphbuilder-constant"}], "title": "7.5.1. Graph Initialization"}, {"refs": [{"id": "ref-for-dom-mlgraphbuilder-constant\u2460"}], "title": "7.6. The MLGraphBuilder interface"}, {"refs": [{"id": "ref-for-dom-mlgraphbuilder-constant\u2461"}], "title": "7.6.4.1. The constant(descriptor, bufferView) method"}], "external": false};
+window.dfnpanelData['dom-mlgraphbuilder-constant-value-type'] = {"dfnID": "dom-mlgraphbuilder-constant-value-type", "url": "#dom-mlgraphbuilder-constant-value-type", "dfnText": "constant(value, type)", "refSections": [{"refs": [{"id": "ref-for-dom-mlgraphbuilder-constant-value-type"}], "title": "7.6. The MLGraphBuilder interface"}, {"refs": [{"id": "ref-for-dom-mlgraphbuilder-constant-value-type\u2460"}], "title": "7.6.4.2. The constant(value, type) method"}], "external": false};
+window.dfnpanelData['dictdef-mlbatchnormalizationoptions'] = {"dfnID": "dictdef-mlbatchnormalizationoptions", "url": "#dictdef-mlbatchnormalizationoptions", "dfnText": "MLBatchNormalizationOptions", "refSections": [{"refs": [{"id": "ref-for-dictdef-mlbatchnormalizationoptions"}, {"id": "ref-for-dictdef-mlbatchnormalizationoptions\u2460"}, {"id": "ref-for-dictdef-mlbatchnormalizationoptions\u2461"}], "title": "7.6.5. The batchNormalization() method"}], "external": false};
+window.dfnpanelData['dom-mlbatchnormalizationoptions-scale'] = {"dfnID": "dom-mlbatchnormalizationoptions-scale", "url": "#dom-mlbatchnormalizationoptions-scale", "dfnText": "scale", "refSections": [{"refs": [{"id": "ref-for-dom-mlbatchnormalizationoptions-scale"}, {"id": "ref-for-dom-mlbatchnormalizationoptions-scale\u2460"}, {"id": "ref-for-dom-mlbatchnormalizationoptions-scale\u2461"}], "title": "7.6.5. The batchNormalization() method"}], "external": false};
+window.dfnpanelData['dom-mlbatchnormalizationoptions-bias'] = {"dfnID": "dom-mlbatchnormalizationoptions-bias", "url": "#dom-mlbatchnormalizationoptions-bias", "dfnText": "bias", "refSections": [{"refs": [{"id": "ref-for-dom-mlbatchnormalizationoptions-bias"}, {"id": "ref-for-dom-mlbatchnormalizationoptions-bias\u2460"}, {"id": "ref-for-dom-mlbatchnormalizationoptions-bias\u2461"}], "title": "7.6.5. The batchNormalization() method"}], "external": false};
+window.dfnpanelData['dom-mlbatchnormalizationoptions-axis'] = {"dfnID": "dom-mlbatchnormalizationoptions-axis", "url": "#dom-mlbatchnormalizationoptions-axis", "dfnText": "axis", "refSections": [{"refs": [{"id": "ref-for-dom-mlbatchnormalizationoptions-axis"}, {"id": "ref-for-dom-mlbatchnormalizationoptions-axis\u2460"}, {"id": "ref-for-dom-mlbatchnormalizationoptions-axis\u2461"}, {"id": "ref-for-dom-mlbatchnormalizationoptions-axis\u2462"}, {"id": "ref-for-dom-mlbatchnormalizationoptions-axis\u2463"}, {"id": "ref-for-dom-mlbatchnormalizationoptions-axis\u2464"}, {"id": "ref-for-dom-mlbatchnormalizationoptions-axis\u2465"}, {"id": "ref-for-dom-mlbatchnormalizationoptions-axis\u2466"}, {"id": "ref-for-dom-mlbatchnormalizationoptions-axis\u2467"}], "title": "7.6.5. The batchNormalization() method"}], "external": false};
+window.dfnpanelData['dom-mlbatchnormalizationoptions-epsilon'] = {"dfnID": "dom-mlbatchnormalizationoptions-epsilon", "url": "#dom-mlbatchnormalizationoptions-epsilon", "dfnText": "epsilon", "refSections": [{"refs": [{"id": "ref-for-dom-mlbatchnormalizationoptions-epsilon"}], "title": "7.6.5. The batchNormalization() method"}], "external": false};
+window.dfnpanelData['dom-mlbatchnormalizationoptions-activation'] = {"dfnID": "dom-mlbatchnormalizationoptions-activation", "url": "#dom-mlbatchnormalizationoptions-activation", "dfnText": "activation", "refSections": [{"refs": [{"id": "ref-for-dom-mlbatchnormalizationoptions-activation"}], "title": "7.6.5. The batchNormalization() method"}], "external": false};
+window.dfnpanelData['dom-mlgraphbuilder-batchnormalization'] = {"dfnID": "dom-mlgraphbuilder-batchnormalization", "url": "#dom-mlgraphbuilder-batchnormalization", "dfnText": "batchNormalization(input, mean, variance, options)", "refSections": [{"refs": [{"id": "ref-for-dom-mlgraphbuilder-batchnormalization"}], "title": "7.6.5. The batchNormalization() method"}], "external": false};
+window.dfnpanelData['dictdef-mlclampoptions'] = {"dfnID": "dictdef-mlclampoptions", "url": "#dictdef-mlclampoptions", "dfnText": "MLClampOptions", "refSections": [{"refs": [{"id": "ref-for-dictdef-mlclampoptions"}, {"id": "ref-for-dictdef-mlclampoptions\u2460"}], "title": "7.6.6. The clamp() method"}, {"refs": [{"id": "ref-for-dictdef-mlclampoptions\u2461"}], "title": "7.6.6.1. The clamp(operand, options) method"}, {"refs": [{"id": "ref-for-dictdef-mlclampoptions\u2462"}], "title": "7.6.6.2. The clamp(options) method"}], "external": false};
+window.dfnpanelData['dom-mlclampoptions-minvalue'] = {"dfnID": "dom-mlclampoptions-minvalue", "url": "#dom-mlclampoptions-minvalue", "dfnText": "minValue", "refSections": [{"refs": [{"id": "ref-for-dom-mlclampoptions-minvalue"}], "title": "7.6.6. The clamp() method"}, {"refs": [{"id": "ref-for-dom-mlclampoptions-minvalue\u2460"}], "title": "7.6.6.1. The clamp(operand, options) method"}], "external": false};
+window.dfnpanelData['dom-mlclampoptions-maxvalue'] = {"dfnID": "dom-mlclampoptions-maxvalue", "url": "#dom-mlclampoptions-maxvalue", "dfnText": "maxValue", "refSections": [{"refs": [{"id": "ref-for-dom-mlclampoptions-maxvalue"}], "title": "7.6.6. The clamp() method"}, {"refs": [{"id": "ref-for-dom-mlclampoptions-maxvalue\u2460"}], "title": "7.6.6.1. The clamp(operand, options) method"}], "external": false};
+window.dfnpanelData['check-clamp-options'] = {"dfnID": "check-clamp-options", "url": "#check-clamp-options", "dfnText": "check clamp options", "refSections": [{"refs": [{"id": "ref-for-check-clamp-options"}], "title": "7.6.6.1. The clamp(operand, options) method"}, {"refs": [{"id": "ref-for-check-clamp-options\u2460"}], "title": "7.6.6.2. The clamp(options) method"}], "external": false};
+window.dfnpanelData['dom-mlgraphbuilder-clamp'] = {"dfnID": "dom-mlgraphbuilder-clamp", "url": "#dom-mlgraphbuilder-clamp", "dfnText": "clamp(operand, options)", "refSections": [{"refs": [{"id": "ref-for-dom-mlgraphbuilder-clamp"}], "title": "7.6.6. The clamp() method"}, {"refs": [{"id": "ref-for-dom-mlgraphbuilder-clamp\u2460"}], "title": "7.6.6.1. The clamp(operand, options) method"}], "external": false};
+window.dfnpanelData['dom-mlgraphbuilder-clamp-options'] = {"dfnID": "dom-mlgraphbuilder-clamp-options", "url": "#dom-mlgraphbuilder-clamp-options", "dfnText": "clamp(options)", "refSections": [{"refs": [{"id": "ref-for-dom-mlgraphbuilder-clamp-options"}], "title": "7.6.6. The clamp() method"}, {"refs": [{"id": "ref-for-dom-mlgraphbuilder-clamp-options\u2460"}], "title": "7.6.6.2. The clamp(options) method"}], "external": false};
+window.dfnpanelData['dom-mlgraphbuilder-concat'] = {"dfnID": "dom-mlgraphbuilder-concat", "url": "#dom-mlgraphbuilder-concat", "dfnText": "concat(inputs, axis)", "refSections": [{"refs": [{"id": "ref-for-dom-mlgraphbuilder-concat"}], "title": "7.6.7. The concat() method"}], "external": false};
+window.dfnpanelData['enumdef-mlconv2dfilteroperandlayout'] = {"dfnID": "enumdef-mlconv2dfilteroperandlayout", "url": "#enumdef-mlconv2dfilteroperandlayout", "dfnText": "MLConv2dFilterOperandLayout", "refSections": [{"refs": [{"id": "ref-for-enumdef-mlconv2dfilteroperandlayout"}, {"id": "ref-for-enumdef-mlconv2dfilteroperandlayout\u2460"}, {"id": "ref-for-enumdef-mlconv2dfilteroperandlayout\u2461"}], "title": "7.6.8. The conv2d() method"}], "external": false};
+window.dfnpanelData['enumdef-mlautopad'] = {"dfnID": "enumdef-mlautopad", "url": "#enumdef-mlautopad", "dfnText": "MLAutoPad", "refSections": [{"refs": [{"id": "ref-for-enumdef-mlautopad"}, {"id": "ref-for-enumdef-mlautopad\u2460"}, {"id": "ref-for-enumdef-mlautopad\u2461"}], "title": "7.6.8. The conv2d() method"}, {"refs": [{"id": "ref-for-enumdef-mlautopad\u2462"}, {"id": "ref-for-enumdef-mlautopad\u2463"}, {"id": "ref-for-enumdef-mlautopad\u2464"}], "title": "7.6.9. The convTranspose2d() method"}, {"refs": [{"id": "ref-for-enumdef-mlautopad\u2465"}, {"id": "ref-for-enumdef-mlautopad\u2466"}, {"id": "ref-for-enumdef-mlautopad\u2467"}], "title": "7.6.25. Pooling operations"}], "external": false};
+window.dfnpanelData['dictdef-mlconv2doptions'] = {"dfnID": "dictdef-mlconv2doptions", "url": "#dictdef-mlconv2doptions", "dfnText": "MLConv2dOptions", "refSections": [{"refs": [{"id": "ref-for-dictdef-mlconv2doptions"}, {"id": "ref-for-dictdef-mlconv2doptions\u2460"}, {"id": "ref-for-dictdef-mlconv2doptions\u2461"}], "title": "7.6.8. The conv2d() method"}], "external": false};
+window.dfnpanelData['dom-mlconv2doptions-padding'] = {"dfnID": "dom-mlconv2doptions-padding", "url": "#dom-mlconv2doptions-padding", "dfnText": "padding", "refSections": [{"refs": [{"id": "ref-for-dom-mlconv2doptions-padding"}, {"id": "ref-for-dom-mlconv2doptions-padding\u2460"}, {"id": "ref-for-dom-mlconv2doptions-padding\u2461"}, {"id": "ref-for-dom-mlconv2doptions-padding\u2462"}, {"id": "ref-for-dom-mlconv2doptions-padding\u2463"}], "title": "7.6.8. The conv2d() method"}], "external": false};
+window.dfnpanelData['dom-mlconv2doptions-strides'] = {"dfnID": "dom-mlconv2doptions-strides", "url": "#dom-mlconv2doptions-strides", "dfnText": "strides", "refSections": [{"refs": [{"id": "ref-for-dom-mlconv2doptions-strides"}, {"id": "ref-for-dom-mlconv2doptions-strides\u2460"}, {"id": "ref-for-dom-mlconv2doptions-strides\u2461"}, {"id": "ref-for-dom-mlconv2doptions-strides\u2462"}], "title": "7.6.8. The conv2d() method"}, {"refs": [{"id": "ref-for-dom-mlconv2doptions-strides\u2463"}], "title": "7.6.9. The convTranspose2d() method"}], "external": false};
+window.dfnpanelData['dom-mlconv2doptions-dilations'] = {"dfnID": "dom-mlconv2doptions-dilations", "url": "#dom-mlconv2doptions-dilations", "dfnText": "dilations", "refSections": [{"refs": [{"id": "ref-for-dom-mlconv2doptions-dilations"}, {"id": "ref-for-dom-mlconv2doptions-dilations\u2460"}, {"id": "ref-for-dom-mlconv2doptions-dilations\u2461"}], "title": "7.6.8. The conv2d() method"}], "external": false};
+window.dfnpanelData['dom-mlconv2doptions-autopad'] = {"dfnID": "dom-mlconv2doptions-autopad", "url": "#dom-mlconv2doptions-autopad", "dfnText": "autoPad", "refSections": [{"refs": [{"id": "ref-for-dom-mlconv2doptions-autopad"}, {"id": "ref-for-dom-mlconv2doptions-autopad\u2460"}], "title": "7.6.8. The conv2d() method"}], "external": false};
+window.dfnpanelData['dom-mlconv2doptions-groups'] = {"dfnID": "dom-mlconv2doptions-groups", "url": "#dom-mlconv2doptions-groups", "dfnText": "groups", "refSections": [{"refs": [{"id": "ref-for-dom-mlconv2doptions-groups"}, {"id": "ref-for-dom-mlconv2doptions-groups\u2460"}, {"id": "ref-for-dom-mlconv2doptions-groups\u2461"}, {"id": "ref-for-dom-mlconv2doptions-groups\u2462"}, {"id": "ref-for-dom-mlconv2doptions-groups\u2463"}], "title": "7.6.8. The conv2d() method"}], "external": false};
+window.dfnpanelData['dom-mlconv2doptions-inputlayout'] = {"dfnID": "dom-mlconv2doptions-inputlayout", "url": "#dom-mlconv2doptions-inputlayout", "dfnText": "inputLayout", "refSections": [{"refs": [{"id": "ref-for-dom-mlconv2doptions-inputlayout"}, {"id": "ref-for-dom-mlconv2doptions-inputlayout\u2460"}, {"id": "ref-for-dom-mlconv2doptions-inputlayout\u2461"}], "title": "7.6.8. The conv2d() method"}], "external": false};
+window.dfnpanelData['dom-mlconv2doptions-filterlayout'] = {"dfnID": "dom-mlconv2doptions-filterlayout", "url": "#dom-mlconv2doptions-filterlayout", "dfnText": "filterLayout", "refSections": [{"refs": [{"id": "ref-for-dom-mlconv2doptions-filterlayout"}, {"id": "ref-for-dom-mlconv2doptions-filterlayout\u2460"}], "title": "7.6.8. The conv2d() method"}], "external": false};
+window.dfnpanelData['dom-mlconv2doptions-bias'] = {"dfnID": "dom-mlconv2doptions-bias", "url": "#dom-mlconv2doptions-bias", "dfnText": "bias", "refSections": [{"refs": [{"id": "ref-for-dom-mlconv2doptions-bias"}, {"id": "ref-for-dom-mlconv2doptions-bias\u2460"}, {"id": "ref-for-dom-mlconv2doptions-bias\u2461"}, {"id": "ref-for-dom-mlconv2doptions-bias\u2462"}, {"id": "ref-for-dom-mlconv2doptions-bias\u2463"}, {"id": "ref-for-dom-mlconv2doptions-bias\u2464"}], "title": "7.6.8. The conv2d() method"}], "external": false};
+window.dfnpanelData['dom-mlconv2doptions-activation'] = {"dfnID": "dom-mlconv2doptions-activation", "url": "#dom-mlconv2doptions-activation", "dfnText": "activation", "refSections": [{"refs": [{"id": "ref-for-dom-mlconv2doptions-activation"}, {"id": "ref-for-dom-mlconv2doptions-activation\u2460"}, {"id": "ref-for-dom-mlconv2doptions-activation\u2461"}, {"id": "ref-for-dom-mlconv2doptions-activation\u2462"}], "title": "7.6.8. The conv2d() method"}], "external": false};
+window.dfnpanelData['dom-mlgraphbuilder-conv2d'] = {"dfnID": "dom-mlgraphbuilder-conv2d", "url": "#dom-mlgraphbuilder-conv2d", "dfnText": "conv2d(input, filter, options)", "refSections": [{"refs": [{"id": "ref-for-dom-mlgraphbuilder-conv2d"}], "title": "7.6.8. The conv2d() method"}], "external": false};
+window.dfnpanelData['enumdef-mlconvtranspose2dfilteroperandlayout'] = {"dfnID": "enumdef-mlconvtranspose2dfilteroperandlayout", "url": "#enumdef-mlconvtranspose2dfilteroperandlayout", "dfnText": "MLConvTranspose2dFilterOperandLayout", "refSections": [{"refs": [{"id": "ref-for-enumdef-mlconvtranspose2dfilteroperandlayout"}, {"id": "ref-for-enumdef-mlconvtranspose2dfilteroperandlayout\u2460"}, {"id": "ref-for-enumdef-mlconvtranspose2dfilteroperandlayout\u2461"}], "title": "7.6.9. The convTranspose2d() method"}], "external": false};
+window.dfnpanelData['dictdef-mlconvtranspose2doptions'] = {"dfnID": "dictdef-mlconvtranspose2doptions", "url": "#dictdef-mlconvtranspose2doptions", "dfnText": "MLConvTranspose2dOptions", "refSections": [{"refs": [{"id": "ref-for-dictdef-mlconvtranspose2doptions"}, {"id": "ref-for-dictdef-mlconvtranspose2doptions\u2460"}, {"id": "ref-for-dictdef-mlconvtranspose2doptions\u2461"}], "title": "7.6.9. The convTranspose2d() method"}], "external": false};
+window.dfnpanelData['dom-mlconvtranspose2doptions-padding'] = {"dfnID": "dom-mlconvtranspose2doptions-padding", "url": "#dom-mlconvtranspose2doptions-padding", "dfnText": "padding", "refSections": [{"refs": [{"id": "ref-for-dom-mlconvtranspose2doptions-padding"}, {"id": "ref-for-dom-mlconvtranspose2doptions-padding\u2460"}, {"id": "ref-for-dom-mlconvtranspose2doptions-padding\u2461"}, {"id": "ref-for-dom-mlconvtranspose2doptions-padding\u2462"}, {"id": "ref-for-dom-mlconvtranspose2doptions-padding\u2463"}], "title": "7.6.9. The convTranspose2d() method"}], "external": false};
+window.dfnpanelData['dom-mlconvtranspose2doptions-strides'] = {"dfnID": "dom-mlconvtranspose2doptions-strides", "url": "#dom-mlconvtranspose2doptions-strides", "dfnText": "strides", "refSections": [{"refs": [{"id": "ref-for-dom-mlconvtranspose2doptions-strides"}, {"id": "ref-for-dom-mlconvtranspose2doptions-strides\u2460"}, {"id": "ref-for-dom-mlconvtranspose2doptions-strides\u2461"}, {"id": "ref-for-dom-mlconvtranspose2doptions-strides\u2462"}, {"id": "ref-for-dom-mlconvtranspose2doptions-strides\u2463"}], "title": "7.6.9. The convTranspose2d() method"}], "external": false};
+window.dfnpanelData['dom-mlconvtranspose2doptions-dilations'] = {"dfnID": "dom-mlconvtranspose2doptions-dilations", "url": "#dom-mlconvtranspose2doptions-dilations", "dfnText": "dilations", "refSections": [{"refs": [{"id": "ref-for-dom-mlconvtranspose2doptions-dilations"}, {"id": "ref-for-dom-mlconvtranspose2doptions-dilations\u2460"}, {"id": "ref-for-dom-mlconvtranspose2doptions-dilations\u2461"}], "title": "7.6.9. The convTranspose2d() method"}], "external": false};
+window.dfnpanelData['dom-mlconvtranspose2doptions-outputpadding'] = {"dfnID": "dom-mlconvtranspose2doptions-outputpadding", "url": "#dom-mlconvtranspose2doptions-outputpadding", "dfnText": "outputPadding", "refSections": [{"refs": [{"id": "ref-for-dom-mlconvtranspose2doptions-outputpadding"}, {"id": "ref-for-dom-mlconvtranspose2doptions-outputpadding\u2460"}, {"id": "ref-for-dom-mlconvtranspose2doptions-outputpadding\u2461"}, {"id": "ref-for-dom-mlconvtranspose2doptions-outputpadding\u2462"}, {"id": "ref-for-dom-mlconvtranspose2doptions-outputpadding\u2463"}], "title": "7.6.9. The convTranspose2d() method"}], "external": false};
+window.dfnpanelData['dom-mlconvtranspose2doptions-outputsizes'] = {"dfnID": "dom-mlconvtranspose2doptions-outputsizes", "url": "#dom-mlconvtranspose2doptions-outputsizes", "dfnText": "outputSizes", "refSections": [{"refs": [{"id": "ref-for-dom-mlconvtranspose2doptions-outputsizes"}, {"id": "ref-for-dom-mlconvtranspose2doptions-outputsizes\u2460"}, {"id": "ref-for-dom-mlconvtranspose2doptions-outputsizes\u2461"}, {"id": "ref-for-dom-mlconvtranspose2doptions-outputsizes\u2462"}, {"id": "ref-for-dom-mlconvtranspose2doptions-outputsizes\u2463"}], "title": "7.6.9. The convTranspose2d() method"}], "external": false};
+window.dfnpanelData['dom-mlconvtranspose2doptions-autopad'] = {"dfnID": "dom-mlconvtranspose2doptions-autopad", "url": "#dom-mlconvtranspose2doptions-autopad", "dfnText": "autoPad", "refSections": [{"refs": [{"id": "ref-for-dom-mlconvtranspose2doptions-autopad"}], "title": "7.6.9. The convTranspose2d() method"}], "external": false};
+window.dfnpanelData['dom-mlconvtranspose2doptions-groups'] = {"dfnID": "dom-mlconvtranspose2doptions-groups", "url": "#dom-mlconvtranspose2doptions-groups", "dfnText": "groups", "refSections": [{"refs": [{"id": "ref-for-dom-mlconvtranspose2doptions-groups"}, {"id": "ref-for-dom-mlconvtranspose2doptions-groups\u2460"}, {"id": "ref-for-dom-mlconvtranspose2doptions-groups\u2461"}, {"id": "ref-for-dom-mlconvtranspose2doptions-groups\u2462"}], "title": "7.6.9. The convTranspose2d() method"}], "external": false};
+window.dfnpanelData['dom-mlconvtranspose2doptions-inputlayout'] = {"dfnID": "dom-mlconvtranspose2doptions-inputlayout", "url": "#dom-mlconvtranspose2doptions-inputlayout", "dfnText": "inputLayout", "refSections": [{"refs": [{"id": "ref-for-dom-mlconvtranspose2doptions-inputlayout"}, {"id": "ref-for-dom-mlconvtranspose2doptions-inputlayout\u2460"}, {"id": "ref-for-dom-mlconvtranspose2doptions-inputlayout\u2461"}], "title": "7.6.9. The convTranspose2d() method"}], "external": false};
+window.dfnpanelData['dom-mlconvtranspose2doptions-filterlayout'] = {"dfnID": "dom-mlconvtranspose2doptions-filterlayout", "url": "#dom-mlconvtranspose2doptions-filterlayout", "dfnText": "filterLayout", "refSections": [{"refs": [{"id": "ref-for-dom-mlconvtranspose2doptions-filterlayout"}, {"id": "ref-for-dom-mlconvtranspose2doptions-filterlayout\u2460"}], "title": "7.6.9. The convTranspose2d() method"}], "external": false};
+window.dfnpanelData['dom-mlconvtranspose2doptions-bias'] = {"dfnID": "dom-mlconvtranspose2doptions-bias", "url": "#dom-mlconvtranspose2doptions-bias", "dfnText": "bias", "refSections": [{"refs": [{"id": "ref-for-dom-mlconvtranspose2doptions-bias"}, {"id": "ref-for-dom-mlconvtranspose2doptions-bias\u2460"}, {"id": "ref-for-dom-mlconvtranspose2doptions-bias\u2461"}, {"id": "ref-for-dom-mlconvtranspose2doptions-bias\u2462"}, {"id": "ref-for-dom-mlconvtranspose2doptions-bias\u2463"}, {"id": "ref-for-dom-mlconvtranspose2doptions-bias\u2464"}], "title": "7.6.9. The convTranspose2d() method"}], "external": false};
+window.dfnpanelData['dom-mlconvtranspose2doptions-activation'] = {"dfnID": "dom-mlconvtranspose2doptions-activation", "url": "#dom-mlconvtranspose2doptions-activation", "dfnText": "activation", "refSections": [{"refs": [{"id": "ref-for-dom-mlconvtranspose2doptions-activation"}, {"id": "ref-for-dom-mlconvtranspose2doptions-activation\u2460"}, {"id": "ref-for-dom-mlconvtranspose2doptions-activation\u2461"}, {"id": "ref-for-dom-mlconvtranspose2doptions-activation\u2462"}], "title": "7.6.9. The convTranspose2d() method"}], "external": false};
+window.dfnpanelData['dom-mlgraphbuilder-convtranspose2d'] = {"dfnID": "dom-mlgraphbuilder-convtranspose2d", "url": "#dom-mlgraphbuilder-convtranspose2d", "dfnText": "convTranspose2d(input, filter, options)", "refSections": [{"refs": [{"id": "ref-for-dom-mlgraphbuilder-convtranspose2d"}], "title": "7.6.9. The convTranspose2d() method"}], "external": false};
+window.dfnpanelData['mlgraphbuilder-element-wise-binary-op'] = {"dfnID": "mlgraphbuilder-element-wise-binary-op", "url": "#mlgraphbuilder-element-wise-binary-op", "dfnText": "create element-wise binary operation", "refSections": [{"refs": [{"id": "ref-for-mlgraphbuilder-element-wise-binary-op"}, {"id": "ref-for-mlgraphbuilder-element-wise-binary-op\u2460"}, {"id": "ref-for-mlgraphbuilder-element-wise-binary-op\u2461"}, {"id": "ref-for-mlgraphbuilder-element-wise-binary-op\u2462"}, {"id": "ref-for-mlgraphbuilder-element-wise-binary-op\u2463"}, {"id": "ref-for-mlgraphbuilder-element-wise-binary-op\u2464"}, {"id": "ref-for-mlgraphbuilder-element-wise-binary-op\u2465"}, {"id": "ref-for-mlgraphbuilder-element-wise-binary-op\u2466"}], "title": "7.6.10. Element-wise binary operations"}], "external": false};
+window.dfnpanelData['mlgraphbuilder-broadcast-shapes'] = {"dfnID": "mlgraphbuilder-broadcast-shapes", "url": "#mlgraphbuilder-broadcast-shapes", "dfnText": "broadcast-shapes", "refSections": [{"refs": [{"id": "ref-for-mlgraphbuilder-broadcast-shapes"}], "title": "7.6.10. Element-wise binary operations"}, {"refs": [{"id": "ref-for-mlgraphbuilder-broadcast-shapes\u2460"}], "title": "7.6.26. The prelu() method"}], "external": false};
+window.dfnpanelData['dom-mlgraphbuilder-add'] = {"dfnID": "dom-mlgraphbuilder-add", "url": "#dom-mlgraphbuilder-add", "dfnText": "add(a, b)", "refSections": [{"refs": [{"id": "ref-for-dom-mlgraphbuilder-add"}], "title": "7.6.10. Element-wise binary operations"}], "external": false};
+window.dfnpanelData['dom-mlgraphbuilder-sub'] = {"dfnID": "dom-mlgraphbuilder-sub", "url": "#dom-mlgraphbuilder-sub", "dfnText": "sub(a, b)", "refSections": [{"refs": [{"id": "ref-for-dom-mlgraphbuilder-sub"}], "title": "7.6.10. Element-wise binary operations"}], "external": false};
+window.dfnpanelData['dom-mlgraphbuilder-mul'] = {"dfnID": "dom-mlgraphbuilder-mul", "url": "#dom-mlgraphbuilder-mul", "dfnText": "mul(a, b)", "refSections": [{"refs": [{"id": "ref-for-dom-mlgraphbuilder-mul"}], "title": "7.6.10. Element-wise binary operations"}], "external": false};
+window.dfnpanelData['dom-mlgraphbuilder-div'] = {"dfnID": "dom-mlgraphbuilder-div", "url": "#dom-mlgraphbuilder-div", "dfnText": "div(a, b)", "refSections": [{"refs": [{"id": "ref-for-dom-mlgraphbuilder-div"}], "title": "7.6.10. Element-wise binary operations"}], "external": false};
+window.dfnpanelData['dom-mlgraphbuilder-max'] = {"dfnID": "dom-mlgraphbuilder-max", "url": "#dom-mlgraphbuilder-max", "dfnText": "max(a, b)", "refSections": [{"refs": [{"id": "ref-for-dom-mlgraphbuilder-max"}], "title": "7.6.10. Element-wise binary operations"}], "external": false};
+window.dfnpanelData['dom-mlgraphbuilder-min'] = {"dfnID": "dom-mlgraphbuilder-min", "url": "#dom-mlgraphbuilder-min", "dfnText": "min(a, b)", "refSections": [{"refs": [{"id": "ref-for-dom-mlgraphbuilder-min"}], "title": "7.6.10. Element-wise binary operations"}], "external": false};
+window.dfnpanelData['dom-mlgraphbuilder-pow'] = {"dfnID": "dom-mlgraphbuilder-pow", "url": "#dom-mlgraphbuilder-pow", "dfnText": "pow(a, b)", "refSections": [{"refs": [{"id": "ref-for-dom-mlgraphbuilder-pow"}], "title": "7.6.10. Element-wise binary operations"}], "external": false};
+window.dfnpanelData['mlgraphbuilder-element-wise-unary-op'] = {"dfnID": "mlgraphbuilder-element-wise-unary-op", "url": "#mlgraphbuilder-element-wise-unary-op", "dfnText": "create element-wise unary operation", "refSections": [{"refs": [{"id": "ref-for-mlgraphbuilder-element-wise-unary-op"}, {"id": "ref-for-mlgraphbuilder-element-wise-unary-op\u2460"}, {"id": "ref-for-mlgraphbuilder-element-wise-unary-op\u2461"}, {"id": "ref-for-mlgraphbuilder-element-wise-unary-op\u2462"}, {"id": "ref-for-mlgraphbuilder-element-wise-unary-op\u2463"}, {"id": "ref-for-mlgraphbuilder-element-wise-unary-op\u2464"}, {"id": "ref-for-mlgraphbuilder-element-wise-unary-op\u2465"}, {"id": "ref-for-mlgraphbuilder-element-wise-unary-op\u2466"}, {"id": "ref-for-mlgraphbuilder-element-wise-unary-op\u2467"}, {"id": "ref-for-mlgraphbuilder-element-wise-unary-op\u2468"}], "title": "7.6.11. Element-wise unary operations"}], "external": false};
+window.dfnpanelData['dom-mlgraphbuilder-abs'] = {"dfnID": "dom-mlgraphbuilder-abs", "url": "#dom-mlgraphbuilder-abs", "dfnText": "abs(input)", "refSections": [{"refs": [{"id": "ref-for-dom-mlgraphbuilder-abs"}], "title": "7.6.11. Element-wise unary operations"}], "external": false};
+window.dfnpanelData['dom-mlgraphbuilder-ceil'] = {"dfnID": "dom-mlgraphbuilder-ceil", "url": "#dom-mlgraphbuilder-ceil", "dfnText": "ceil(input)", "refSections": [{"refs": [{"id": "ref-for-dom-mlgraphbuilder-ceil"}], "title": "7.6.11. Element-wise unary operations"}], "external": false};
+window.dfnpanelData['dom-mlgraphbuilder-cos'] = {"dfnID": "dom-mlgraphbuilder-cos", "url": "#dom-mlgraphbuilder-cos", "dfnText": "cos(input)", "refSections": [{"refs": [{"id": "ref-for-dom-mlgraphbuilder-cos"}], "title": "7.6.11. Element-wise unary operations"}], "external": false};
+window.dfnpanelData['dom-mlgraphbuilder-exp'] = {"dfnID": "dom-mlgraphbuilder-exp", "url": "#dom-mlgraphbuilder-exp", "dfnText": "exp(input)", "refSections": [{"refs": [{"id": "ref-for-dom-mlgraphbuilder-exp"}], "title": "7.6.11. Element-wise unary operations"}], "external": false};
+window.dfnpanelData['dom-mlgraphbuilder-floor'] = {"dfnID": "dom-mlgraphbuilder-floor", "url": "#dom-mlgraphbuilder-floor", "dfnText": "floor(input)", "refSections": [{"refs": [{"id": "ref-for-dom-mlgraphbuilder-floor"}], "title": "7.6.11. Element-wise unary operations"}], "external": false};
+window.dfnpanelData['dom-mlgraphbuilder-log'] = {"dfnID": "dom-mlgraphbuilder-log", "url": "#dom-mlgraphbuilder-log", "dfnText": "log(input)", "refSections": [{"refs": [{"id": "ref-for-dom-mlgraphbuilder-log"}], "title": "7.6.11. Element-wise unary operations"}], "external": false};
+window.dfnpanelData['dom-mlgraphbuilder-neg'] = {"dfnID": "dom-mlgraphbuilder-neg", "url": "#dom-mlgraphbuilder-neg", "dfnText": "neg(input)", "refSections": [{"refs": [{"id": "ref-for-dom-mlgraphbuilder-neg"}], "title": "7.6.11. Element-wise unary operations"}], "external": false};
+window.dfnpanelData['dom-mlgraphbuilder-sin'] = {"dfnID": "dom-mlgraphbuilder-sin", "url": "#dom-mlgraphbuilder-sin", "dfnText": "sin(input)", "refSections": [{"refs": [{"id": "ref-for-dom-mlgraphbuilder-sin"}], "title": "7.6.11. Element-wise unary operations"}], "external": false};
+window.dfnpanelData['dom-mlgraphbuilder-tan'] = {"dfnID": "dom-mlgraphbuilder-tan", "url": "#dom-mlgraphbuilder-tan", "dfnText": "tan(input)", "refSections": [{"refs": [{"id": "ref-for-dom-mlgraphbuilder-tan"}], "title": "7.6.11. Element-wise unary operations"}], "external": false};
+window.dfnpanelData['dictdef-mleluoptions'] = {"dfnID": "dictdef-mleluoptions", "url": "#dictdef-mleluoptions", "dfnText": "MLEluOptions", "refSections": [{"refs": [{"id": "ref-for-dictdef-mleluoptions"}, {"id": "ref-for-dictdef-mleluoptions\u2460"}], "title": "7.6.12. The elu() method"}, {"refs": [{"id": "ref-for-dictdef-mleluoptions\u2461"}], "title": "7.6.12.1. The elu(input, options) method"}, {"refs": [{"id": "ref-for-dictdef-mleluoptions\u2462"}], "title": "7.6.12.2. The elu(options) method"}], "external": false};
+window.dfnpanelData['dom-mlgraphbuilder-elu'] = {"dfnID": "dom-mlgraphbuilder-elu", "url": "#dom-mlgraphbuilder-elu", "dfnText": "elu(input, options)", "refSections": [{"refs": [{"id": "ref-for-dom-mlgraphbuilder-elu"}], "title": "7.6.12. The elu() method"}, {"refs": [{"id": "ref-for-dom-mlgraphbuilder-elu\u2460"}], "title": "7.6.12.1. The elu(input, options) method"}], "external": false};
+window.dfnpanelData['dom-mlgraphbuilder-elu-options'] = {"dfnID": "dom-mlgraphbuilder-elu-options", "url": "#dom-mlgraphbuilder-elu-options", "dfnText": "elu(options)", "refSections": [{"refs": [{"id": "ref-for-dom-mlgraphbuilder-elu-options"}], "title": "7.6.12. The elu() method"}, {"refs": [{"id": "ref-for-dom-mlgraphbuilder-elu-options\u2460"}], "title": "7.6.12.2. The elu(options) method"}], "external": false};
+window.dfnpanelData['dictdef-mlgemmoptions'] = {"dfnID": "dictdef-mlgemmoptions", "url": "#dictdef-mlgemmoptions", "dfnText": "MLGemmOptions", "refSections": [{"refs": [{"id": "ref-for-dictdef-mlgemmoptions"}, {"id": "ref-for-dictdef-mlgemmoptions\u2460"}, {"id": "ref-for-dictdef-mlgemmoptions\u2461"}], "title": "7.6.13. The gemm() method"}], "external": false};
+window.dfnpanelData['dom-mlgemmoptions-c'] = {"dfnID": "dom-mlgemmoptions-c", "url": "#dom-mlgemmoptions-c", "dfnText": "c", "refSections": [{"refs": [{"id": "ref-for-dom-mlgemmoptions-c"}, {"id": "ref-for-dom-mlgemmoptions-c\u2460"}, {"id": "ref-for-dom-mlgemmoptions-c\u2461"}, {"id": "ref-for-dom-mlgemmoptions-c\u2462"}], "title": "7.6.13. The gemm() method"}], "external": false};
+window.dfnpanelData['dom-mlgemmoptions-alpha'] = {"dfnID": "dom-mlgemmoptions-alpha", "url": "#dom-mlgemmoptions-alpha", "dfnText": "alpha", "refSections": [{"refs": [{"id": "ref-for-dom-mlgemmoptions-alpha"}], "title": "7.6.13. The gemm() method"}], "external": false};
+window.dfnpanelData['dom-mlgemmoptions-beta'] = {"dfnID": "dom-mlgemmoptions-beta", "url": "#dom-mlgemmoptions-beta", "dfnText": "beta", "refSections": [{"refs": [{"id": "ref-for-dom-mlgemmoptions-beta"}], "title": "7.6.13. The gemm() method"}], "external": false};
+window.dfnpanelData['dom-mlgemmoptions-atranspose'] = {"dfnID": "dom-mlgemmoptions-atranspose", "url": "#dom-mlgemmoptions-atranspose", "dfnText": "aTranspose", "refSections": [{"refs": [{"id": "ref-for-dom-mlgemmoptions-atranspose"}, {"id": "ref-for-dom-mlgemmoptions-atranspose\u2460"}], "title": "7.6.13. The gemm() method"}], "external": false};
+window.dfnpanelData['dom-mlgemmoptions-btranspose'] = {"dfnID": "dom-mlgemmoptions-btranspose", "url": "#dom-mlgemmoptions-btranspose", "dfnText": "bTranspose", "refSections": [{"refs": [{"id": "ref-for-dom-mlgemmoptions-btranspose"}, {"id": "ref-for-dom-mlgemmoptions-btranspose\u2460"}], "title": "7.6.13. The gemm() method"}], "external": false};
+window.dfnpanelData['dom-mlgraphbuilder-gemm'] = {"dfnID": "dom-mlgraphbuilder-gemm", "url": "#dom-mlgraphbuilder-gemm", "dfnText": "gemm(a, b, options)", "refSections": [{"refs": [{"id": "ref-for-dom-mlgraphbuilder-gemm"}], "title": "6.1. Overview"}, {"refs": [{"id": "ref-for-dom-mlgraphbuilder-gemm\u2460"}], "title": "7.6.13. The gemm() method"}], "external": false};
+window.dfnpanelData['enumdef-mlgruweightlayout'] = {"dfnID": "enumdef-mlgruweightlayout", "url": "#enumdef-mlgruweightlayout", "dfnText": "MLGruWeightLayout", "refSections": [{"refs": [{"id": "ref-for-enumdef-mlgruweightlayout"}, {"id": "ref-for-enumdef-mlgruweightlayout\u2460"}, {"id": "ref-for-enumdef-mlgruweightlayout\u2461"}], "title": "7.6.14. The gru() method"}, {"refs": [{"id": "ref-for-enumdef-mlgruweightlayout\u2462"}, {"id": "ref-for-enumdef-mlgruweightlayout\u2463"}, {"id": "ref-for-enumdef-mlgruweightlayout\u2464"}], "title": "7.6.15. The gruCell() method"}], "external": false};
+window.dfnpanelData['enumdef-mlrecurrentnetworkdirection'] = {"dfnID": "enumdef-mlrecurrentnetworkdirection", "url": "#enumdef-mlrecurrentnetworkdirection", "dfnText": "MLRecurrentNetworkDirection", "refSections": [{"refs": [{"id": "ref-for-enumdef-mlrecurrentnetworkdirection"}, {"id": "ref-for-enumdef-mlrecurrentnetworkdirection\u2460"}, {"id": "ref-for-enumdef-mlrecurrentnetworkdirection\u2461"}], "title": "7.6.14. The gru() method"}, {"refs": [{"id": "ref-for-enumdef-mlrecurrentnetworkdirection\u2462"}, {"id": "ref-for-enumdef-mlrecurrentnetworkdirection\u2463"}, {"id": "ref-for-enumdef-mlrecurrentnetworkdirection\u2464"}], "title": "7.6.21. The lstm() method"}], "external": false};
+window.dfnpanelData['dictdef-mlgruoptions'] = {"dfnID": "dictdef-mlgruoptions", "url": "#dictdef-mlgruoptions", "dfnText": "MLGruOptions", "refSections": [{"refs": [{"id": "ref-for-dictdef-mlgruoptions"}, {"id": "ref-for-dictdef-mlgruoptions\u2460"}, {"id": "ref-for-dictdef-mlgruoptions\u2461"}], "title": "7.6.14. The gru() method"}], "external": false};
+window.dfnpanelData['dom-mlgruoptions-bias'] = {"dfnID": "dom-mlgruoptions-bias", "url": "#dom-mlgruoptions-bias", "dfnText": "bias", "refSections": [{"refs": [{"id": "ref-for-dom-mlgruoptions-bias"}, {"id": "ref-for-dom-mlgruoptions-bias\u2460"}, {"id": "ref-for-dom-mlgruoptions-bias\u2461"}], "title": "7.6.14. The gru() method"}, {"refs": [{"id": "ref-for-dom-mlgruoptions-bias\u2462"}], "title": "7.6.15. The gruCell() method"}], "external": false};
+window.dfnpanelData['dom-mlgruoptions-recurrentbias'] = {"dfnID": "dom-mlgruoptions-recurrentbias", "url": "#dom-mlgruoptions-recurrentbias", "dfnText": "recurrentBias", "refSections": [{"refs": [{"id": "ref-for-dom-mlgruoptions-recurrentbias"}, {"id": "ref-for-dom-mlgruoptions-recurrentbias\u2460"}, {"id": "ref-for-dom-mlgruoptions-recurrentbias\u2461"}], "title": "7.6.14. The gru() method"}, {"refs": [{"id": "ref-for-dom-mlgruoptions-recurrentbias\u2462"}], "title": "7.6.15. The gruCell() method"}], "external": false};
+window.dfnpanelData['dom-mlgruoptions-initialhiddenstate'] = {"dfnID": "dom-mlgruoptions-initialhiddenstate", "url": "#dom-mlgruoptions-initialhiddenstate", "dfnText": "initialHiddenState", "refSections": [{"refs": [{"id": "ref-for-dom-mlgruoptions-initialhiddenstate"}, {"id": "ref-for-dom-mlgruoptions-initialhiddenstate\u2460"}], "title": "7.6.14. The gru() method"}], "external": false};
+window.dfnpanelData['dom-mlgruoptions-resetafter'] = {"dfnID": "dom-mlgruoptions-resetafter", "url": "#dom-mlgruoptions-resetafter", "dfnText": "resetAfter", "refSections": [{"refs": [{"id": "ref-for-dom-mlgruoptions-resetafter"}], "title": "7.6.14. The gru() method"}], "external": false};
+window.dfnpanelData['dom-mlgruoptions-returnsequence'] = {"dfnID": "dom-mlgruoptions-returnsequence", "url": "#dom-mlgruoptions-returnsequence", "dfnText": "returnSequence", "refSections": [{"refs": [{"id": "ref-for-dom-mlgruoptions-returnsequence"}, {"id": "ref-for-dom-mlgruoptions-returnsequence\u2460"}], "title": "7.6.14. The gru() method"}], "external": false};
+window.dfnpanelData['dom-mlgruoptions-direction'] = {"dfnID": "dom-mlgruoptions-direction", "url": "#dom-mlgruoptions-direction", "dfnText": "direction", "refSections": [{"refs": [{"id": "ref-for-dom-mlgruoptions-direction"}], "title": "7.6.14. The gru() method"}], "external": false};
+window.dfnpanelData['dom-mlgruoptions-layout'] = {"dfnID": "dom-mlgruoptions-layout", "url": "#dom-mlgruoptions-layout", "dfnText": "layout", "refSections": [{"refs": [{"id": "ref-for-dom-mlgruoptions-layout"}, {"id": "ref-for-dom-mlgruoptions-layout\u2460"}, {"id": "ref-for-dom-mlgruoptions-layout\u2461"}, {"id": "ref-for-dom-mlgruoptions-layout\u2462"}, {"id": "ref-for-dom-mlgruoptions-layout\u2463"}], "title": "7.6.14. The gru() method"}, {"refs": [{"id": "ref-for-dom-mlgruoptions-layout\u2464"}, {"id": "ref-for-dom-mlgruoptions-layout\u2465"}], "title": "7.6.15. The gruCell() method"}], "external": false};
+window.dfnpanelData['dom-mlgruoptions-activations'] = {"dfnID": "dom-mlgruoptions-activations", "url": "#dom-mlgruoptions-activations", "dfnText": "activations", "refSections": [{"refs": [{"id": "ref-for-dom-mlgruoptions-activations"}, {"id": "ref-for-dom-mlgruoptions-activations\u2460"}], "title": "7.6.14. The gru() method"}, {"refs": [{"id": "ref-for-dom-mlgruoptions-activations\u2461"}], "title": "7.6.15. The gruCell() method"}], "external": false};
+window.dfnpanelData['dom-mlgraphbuilder-gru'] = {"dfnID": "dom-mlgraphbuilder-gru", "url": "#dom-mlgraphbuilder-gru", "dfnText": "gru(input, weight, recurrentWeight, steps, hiddenSize, options)", "refSections": [{"refs": [{"id": "ref-for-dom-mlgraphbuilder-gru"}], "title": "7.6.14. The gru() method"}], "external": false};
+window.dfnpanelData['dictdef-mlgrucelloptions'] = {"dfnID": "dictdef-mlgrucelloptions", "url": "#dictdef-mlgrucelloptions", "dfnText": "MLGruCellOptions", "refSections": [{"refs": [{"id": "ref-for-dictdef-mlgrucelloptions"}, {"id": "ref-for-dictdef-mlgrucelloptions\u2460"}, {"id": "ref-for-dictdef-mlgrucelloptions\u2461"}], "title": "7.6.15. The gruCell() method"}], "external": false};
+window.dfnpanelData['dom-mlgrucelloptions-bias'] = {"dfnID": "dom-mlgrucelloptions-bias", "url": "#dom-mlgrucelloptions-bias", "dfnText": "bias", "refSections": [{"refs": [{"id": "ref-for-dom-mlgrucelloptions-bias"}], "title": "7.6.15. The gruCell() method"}], "external": false};
+window.dfnpanelData['dom-mlgrucelloptions-recurrentbias'] = {"dfnID": "dom-mlgrucelloptions-recurrentbias", "url": "#dom-mlgrucelloptions-recurrentbias", "dfnText": "recurrentBias", "refSections": [{"refs": [{"id": "ref-for-dom-mlgrucelloptions-recurrentbias"}], "title": "7.6.15. The gruCell() method"}], "external": false};
+window.dfnpanelData['dom-mlgrucelloptions-resetafter'] = {"dfnID": "dom-mlgrucelloptions-resetafter", "url": "#dom-mlgrucelloptions-resetafter", "dfnText": "resetAfter", "refSections": [{"refs": [{"id": "ref-for-dom-mlgrucelloptions-resetafter"}], "title": "7.6.15. The gruCell() method"}], "external": false};
+window.dfnpanelData['dom-mlgrucelloptions-layout'] = {"dfnID": "dom-mlgrucelloptions-layout", "url": "#dom-mlgrucelloptions-layout", "dfnText": "layout", "refSections": [{"refs": [{"id": "ref-for-dom-mlgrucelloptions-layout"}], "title": "7.6.15. The gruCell() method"}], "external": false};
+window.dfnpanelData['dom-mlgrucelloptions-activations'] = {"dfnID": "dom-mlgrucelloptions-activations", "url": "#dom-mlgrucelloptions-activations", "dfnText": "activations", "refSections": [{"refs": [{"id": "ref-for-dom-mlgrucelloptions-activations"}], "title": "7.6.15. The gruCell() method"}], "external": false};
+window.dfnpanelData['dom-mlgraphbuilder-grucell'] = {"dfnID": "dom-mlgraphbuilder-grucell", "url": "#dom-mlgraphbuilder-grucell", "dfnText": "gruCell(input, weight, recurrentWeight, hiddenState, hiddenSize, options)", "refSections": [{"refs": [{"id": "ref-for-dom-mlgraphbuilder-grucell"}], "title": "7.6.15. The gruCell() method"}], "external": false};
+window.dfnpanelData['dictdef-mlhardsigmoidoptions'] = {"dfnID": "dictdef-mlhardsigmoidoptions", "url": "#dictdef-mlhardsigmoidoptions", "dfnText": "MLHardSigmoidOptions", "refSections": [{"refs": [{"id": "ref-for-dictdef-mlhardsigmoidoptions"}, {"id": "ref-for-dictdef-mlhardsigmoidoptions\u2460"}, {"id": "ref-for-dictdef-mlhardsigmoidoptions\u2461"}], "title": "7.6.16. The hardSigmoid() method"}, {"refs": [{"id": "ref-for-dictdef-mlhardsigmoidoptions\u2462"}], "title": "7.6.16.1. The hardSigmoid(input, options) method"}, {"refs": [{"id": "ref-for-dictdef-mlhardsigmoidoptions\u2463"}], "title": "7.6.16.2. The hardSigmoid(options) method"}], "external": false};
+window.dfnpanelData['dom-mlhardsigmoidoptions-alpha'] = {"dfnID": "dom-mlhardsigmoidoptions-alpha", "url": "#dom-mlhardsigmoidoptions-alpha", "dfnText": "alpha", "refSections": [{"refs": [{"id": "ref-for-dom-mlhardsigmoidoptions-alpha"}], "title": "7.6.16. The hardSigmoid() method"}], "external": false};
+window.dfnpanelData['dom-mlhardsigmoidoptions-beta'] = {"dfnID": "dom-mlhardsigmoidoptions-beta", "url": "#dom-mlhardsigmoidoptions-beta", "dfnText": "beta", "refSections": [{"refs": [{"id": "ref-for-dom-mlhardsigmoidoptions-beta"}], "title": "7.6.16. The hardSigmoid() method"}], "external": false};
+window.dfnpanelData['dom-mlgraphbuilder-hardsigmoid'] = {"dfnID": "dom-mlgraphbuilder-hardsigmoid", "url": "#dom-mlgraphbuilder-hardsigmoid", "dfnText": "hardSigmoid(input, options)", "refSections": [{"refs": [{"id": "ref-for-dom-mlgraphbuilder-hardsigmoid"}], "title": "7.6.16. The hardSigmoid() method"}, {"refs": [{"id": "ref-for-dom-mlgraphbuilder-hardsigmoid\u2460"}], "title": "7.6.16.1. The hardSigmoid(input, options) method"}], "external": false};
+window.dfnpanelData['dom-mlgraphbuilder-hardsigmoid-options'] = {"dfnID": "dom-mlgraphbuilder-hardsigmoid-options", "url": "#dom-mlgraphbuilder-hardsigmoid-options", "dfnText": "hardSigmoid(options)", "refSections": [{"refs": [{"id": "ref-for-dom-mlgraphbuilder-hardsigmoid-options"}], "title": "7.6.16. The hardSigmoid() method"}, {"refs": [{"id": "ref-for-dom-mlgraphbuilder-hardsigmoid-options\u2460"}], "title": "7.6.16.2. The hardSigmoid(options) method"}], "external": false};
+window.dfnpanelData['dom-mlgraphbuilder-hardswish'] = {"dfnID": "dom-mlgraphbuilder-hardswish", "url": "#dom-mlgraphbuilder-hardswish", "dfnText": "hardSwish(input)", "refSections": [{"refs": [{"id": "ref-for-dom-mlgraphbuilder-hardswish"}], "title": "7.6.17. The hardSwish() method"}, {"refs": [{"id": "ref-for-dom-mlgraphbuilder-hardswish\u2460"}], "title": "7.6.17.1. The hardSwish(input) method"}], "external": false};
+window.dfnpanelData['hardswish-noargs'] = {"dfnID": "hardswish-noargs", "url": "#hardswish-noargs", "dfnText": "hardSwish()", "refSections": [{"refs": [{"id": "ref-for-hardswish-noargs"}], "title": "7.6.17. The hardSwish() method"}, {"refs": [{"id": "ref-for-hardswish-noargs\u2460"}], "title": "7.6.17.2. The hardSwish() method"}], "external": false};
+window.dfnpanelData['dictdef-mlinstancenormalizationoptions'] = {"dfnID": "dictdef-mlinstancenormalizationoptions", "url": "#dictdef-mlinstancenormalizationoptions", "dfnText": "MLInstanceNormalizationOptions", "refSections": [{"refs": [{"id": "ref-for-dictdef-mlinstancenormalizationoptions"}, {"id": "ref-for-dictdef-mlinstancenormalizationoptions\u2460"}, {"id": "ref-for-dictdef-mlinstancenormalizationoptions\u2461"}], "title": "7.6.18. The instanceNormalization() method"}], "external": false};
+window.dfnpanelData['dom-mlinstancenormalizationoptions-scale'] = {"dfnID": "dom-mlinstancenormalizationoptions-scale", "url": "#dom-mlinstancenormalizationoptions-scale", "dfnText": "scale", "refSections": [{"refs": [{"id": "ref-for-dom-mlinstancenormalizationoptions-scale"}, {"id": "ref-for-dom-mlinstancenormalizationoptions-scale\u2460"}, {"id": "ref-for-dom-mlinstancenormalizationoptions-scale\u2461"}], "title": "7.6.18. The instanceNormalization() method"}], "external": false};
+window.dfnpanelData['dom-mlinstancenormalizationoptions-bias'] = {"dfnID": "dom-mlinstancenormalizationoptions-bias", "url": "#dom-mlinstancenormalizationoptions-bias", "dfnText": "bias", "refSections": [{"refs": [{"id": "ref-for-dom-mlinstancenormalizationoptions-bias"}, {"id": "ref-for-dom-mlinstancenormalizationoptions-bias\u2460"}, {"id": "ref-for-dom-mlinstancenormalizationoptions-bias\u2461"}], "title": "7.6.18. The instanceNormalization() method"}], "external": false};
+window.dfnpanelData['dom-mlinstancenormalizationoptions-epsilon'] = {"dfnID": "dom-mlinstancenormalizationoptions-epsilon", "url": "#dom-mlinstancenormalizationoptions-epsilon", "dfnText": "epsilon", "refSections": [{"refs": [{"id": "ref-for-dom-mlinstancenormalizationoptions-epsilon"}], "title": "7.6.18. The instanceNormalization() method"}], "external": false};
+window.dfnpanelData['dom-mlinstancenormalizationoptions-layout'] = {"dfnID": "dom-mlinstancenormalizationoptions-layout", "url": "#dom-mlinstancenormalizationoptions-layout", "dfnText": "layout", "refSections": [{"refs": [{"id": "ref-for-dom-mlinstancenormalizationoptions-layout"}], "title": "7.6.18. The instanceNormalization() method"}], "external": false};
+window.dfnpanelData['dom-mlgraphbuilder-instancenormalization'] = {"dfnID": "dom-mlgraphbuilder-instancenormalization", "url": "#dom-mlgraphbuilder-instancenormalization", "dfnText": "instanceNormalization(input, options)", "refSections": [{"refs": [{"id": "ref-for-dom-mlgraphbuilder-instancenormalization"}], "title": "7.6.18. The instanceNormalization() method"}], "external": false};
+window.dfnpanelData['dictdef-mlleakyreluoptions'] = {"dfnID": "dictdef-mlleakyreluoptions", "url": "#dictdef-mlleakyreluoptions", "dfnText": "MLLeakyReluOptions", "refSections": [{"refs": [{"id": "ref-for-dictdef-mlleakyreluoptions"}, {"id": "ref-for-dictdef-mlleakyreluoptions\u2460"}, {"id": "ref-for-dictdef-mlleakyreluoptions\u2461"}], "title": "7.6.19. The leakyRelu() method"}, {"refs": [{"id": "ref-for-dictdef-mlleakyreluoptions\u2462"}], "title": "7.6.19.1. The leakyRelu(input, options) method"}, {"refs": [{"id": "ref-for-dictdef-mlleakyreluoptions\u2463"}], "title": "7.6.19.2. The leakyRelu(options) method"}], "external": false};
+window.dfnpanelData['dom-mlleakyreluoptions-alpha'] = {"dfnID": "dom-mlleakyreluoptions-alpha", "url": "#dom-mlleakyreluoptions-alpha", "dfnText": "alpha", "refSections": [{"refs": [{"id": "ref-for-dom-mlleakyreluoptions-alpha"}], "title": "7.6.19. The leakyRelu() method"}], "external": false};
+window.dfnpanelData['dom-mlgraphbuilder-leakyrelu'] = {"dfnID": "dom-mlgraphbuilder-leakyrelu", "url": "#dom-mlgraphbuilder-leakyrelu", "dfnText": "leakyRelu(input, options)", "refSections": [{"refs": [{"id": "ref-for-dom-mlgraphbuilder-leakyrelu"}], "title": "7.6.19. The leakyRelu() method"}, {"refs": [{"id": "ref-for-dom-mlgraphbuilder-leakyrelu\u2460"}], "title": "7.6.19.1. The leakyRelu(input, options) method"}], "external": false};
+window.dfnpanelData['dom-mlgraphbuilder-leakyrelu-options'] = {"dfnID": "dom-mlgraphbuilder-leakyrelu-options", "url": "#dom-mlgraphbuilder-leakyrelu-options", "dfnText": "leakyRelu(options)", "refSections": [{"refs": [{"id": "ref-for-dom-mlgraphbuilder-leakyrelu-options"}], "title": "7.6.19. The leakyRelu() method"}, {"refs": [{"id": "ref-for-dom-mlgraphbuilder-leakyrelu-options\u2460"}], "title": "7.6.19.2. The leakyRelu(options) method"}], "external": false};
+window.dfnpanelData['dictdef-mllinearoptions'] = {"dfnID": "dictdef-mllinearoptions", "url": "#dictdef-mllinearoptions", "dfnText": "MLLinearOptions", "refSections": [{"refs": [{"id": "ref-for-dictdef-mllinearoptions"}, {"id": "ref-for-dictdef-mllinearoptions\u2460"}, {"id": "ref-for-dictdef-mllinearoptions\u2461"}], "title": "7.6.20. The linear() method"}, {"refs": [{"id": "ref-for-dictdef-mllinearoptions\u2462"}], "title": "7.6.20.1. The linear(input, options) method"}, {"refs": [{"id": "ref-for-dictdef-mllinearoptions\u2463"}], "title": "7.6.20.2. The linear(options) method"}], "external": false};
+window.dfnpanelData['dom-mllinearoptions-alpha'] = {"dfnID": "dom-mllinearoptions-alpha", "url": "#dom-mllinearoptions-alpha", "dfnText": "alpha", "refSections": [{"refs": [{"id": "ref-for-dom-mllinearoptions-alpha"}], "title": "7.6.20. The linear() method"}], "external": false};
+window.dfnpanelData['dom-mllinearoptions-beta'] = {"dfnID": "dom-mllinearoptions-beta", "url": "#dom-mllinearoptions-beta", "dfnText": "beta", "refSections": [{"refs": [{"id": "ref-for-dom-mllinearoptions-beta"}], "title": "7.6.20. The linear() method"}], "external": false};
+window.dfnpanelData['dom-mlgraphbuilder-linear'] = {"dfnID": "dom-mlgraphbuilder-linear", "url": "#dom-mlgraphbuilder-linear", "dfnText": "linear(input, options)", "refSections": [{"refs": [{"id": "ref-for-dom-mlgraphbuilder-linear"}], "title": "7.6.20. The linear() method"}, {"refs": [{"id": "ref-for-dom-mlgraphbuilder-linear\u2460"}], "title": "7.6.20.1. The linear(input, options) method"}], "external": false};
+window.dfnpanelData['dom-mlgraphbuilder-linear-options'] = {"dfnID": "dom-mlgraphbuilder-linear-options", "url": "#dom-mlgraphbuilder-linear-options", "dfnText": "linear(options)", "refSections": [{"refs": [{"id": "ref-for-dom-mlgraphbuilder-linear-options"}], "title": "7.6.20. The linear() method"}, {"refs": [{"id": "ref-for-dom-mlgraphbuilder-linear-options\u2460"}], "title": "7.6.20.2. The linear(options) method"}], "external": false};
+window.dfnpanelData['enumdef-mllstmweightlayout'] = {"dfnID": "enumdef-mllstmweightlayout", "url": "#enumdef-mllstmweightlayout", "dfnText": "MLLstmWeightLayout", "refSections": [{"refs": [{"id": "ref-for-enumdef-mllstmweightlayout"}, {"id": "ref-for-enumdef-mllstmweightlayout\u2460"}, {"id": "ref-for-enumdef-mllstmweightlayout\u2461"}], "title": "7.6.21. The lstm() method"}, {"refs": [{"id": "ref-for-enumdef-mllstmweightlayout\u2462"}, {"id": "ref-for-enumdef-mllstmweightlayout\u2463"}, {"id": "ref-for-enumdef-mllstmweightlayout\u2464"}], "title": "7.6.22. The lstmCell() method"}], "external": false};
+window.dfnpanelData['dictdef-mllstmoptions'] = {"dfnID": "dictdef-mllstmoptions", "url": "#dictdef-mllstmoptions", "dfnText": "MLLstmOptions", "refSections": [{"refs": [{"id": "ref-for-dictdef-mllstmoptions"}, {"id": "ref-for-dictdef-mllstmoptions\u2460"}, {"id": "ref-for-dictdef-mllstmoptions\u2461"}], "title": "7.6.21. The lstm() method"}], "external": false};
+window.dfnpanelData['dom-mllstmoptions-bias'] = {"dfnID": "dom-mllstmoptions-bias", "url": "#dom-mllstmoptions-bias", "dfnText": "bias", "refSections": [{"refs": [{"id": "ref-for-dom-mllstmoptions-bias"}, {"id": "ref-for-dom-mllstmoptions-bias\u2460"}, {"id": "ref-for-dom-mllstmoptions-bias\u2461"}, {"id": "ref-for-dom-mllstmoptions-bias\u2462"}], "title": "7.6.21. The lstm() method"}], "external": false};
+window.dfnpanelData['dom-mllstmoptions-recurrentbias'] = {"dfnID": "dom-mllstmoptions-recurrentbias", "url": "#dom-mllstmoptions-recurrentbias", "dfnText": "recurrentBias", "refSections": [{"refs": [{"id": "ref-for-dom-mllstmoptions-recurrentbias"}, {"id": "ref-for-dom-mllstmoptions-recurrentbias\u2460"}, {"id": "ref-for-dom-mllstmoptions-recurrentbias\u2461"}, {"id": "ref-for-dom-mllstmoptions-recurrentbias\u2462"}], "title": "7.6.21. The lstm() method"}], "external": false};
+window.dfnpanelData['dom-mllstmoptions-peepholeweight'] = {"dfnID": "dom-mllstmoptions-peepholeweight", "url": "#dom-mllstmoptions-peepholeweight", "dfnText": "peepholeWeight", "refSections": [{"refs": [{"id": "ref-for-dom-mllstmoptions-peepholeweight"}, {"id": "ref-for-dom-mllstmoptions-peepholeweight\u2460"}, {"id": "ref-for-dom-mllstmoptions-peepholeweight\u2461"}, {"id": "ref-for-dom-mllstmoptions-peepholeweight\u2462"}], "title": "7.6.21. The lstm() method"}], "external": false};
+window.dfnpanelData['dom-mllstmoptions-initialhiddenstate'] = {"dfnID": "dom-mllstmoptions-initialhiddenstate", "url": "#dom-mllstmoptions-initialhiddenstate", "dfnText": "initialHiddenState", "refSections": [{"refs": [{"id": "ref-for-dom-mllstmoptions-initialhiddenstate"}, {"id": "ref-for-dom-mllstmoptions-initialhiddenstate\u2460"}, {"id": "ref-for-dom-mllstmoptions-initialhiddenstate\u2461"}, {"id": "ref-for-dom-mllstmoptions-initialhiddenstate\u2462"}, {"id": "ref-for-dom-mllstmoptions-initialhiddenstate\u2463"}], "title": "7.6.21. The lstm() method"}], "external": false};
+window.dfnpanelData['dom-mllstmoptions-initialcellstate'] = {"dfnID": "dom-mllstmoptions-initialcellstate", "url": "#dom-mllstmoptions-initialcellstate", "dfnText": "initialCellState", "refSections": [{"refs": [{"id": "ref-for-dom-mllstmoptions-initialcellstate"}, {"id": "ref-for-dom-mllstmoptions-initialcellstate\u2460"}, {"id": "ref-for-dom-mllstmoptions-initialcellstate\u2461"}, {"id": "ref-for-dom-mllstmoptions-initialcellstate\u2462"}, {"id": "ref-for-dom-mllstmoptions-initialcellstate\u2463"}], "title": "7.6.21. The lstm() method"}], "external": false};
+window.dfnpanelData['dom-mllstmoptions-returnsequence'] = {"dfnID": "dom-mllstmoptions-returnsequence", "url": "#dom-mllstmoptions-returnsequence", "dfnText": "returnSequence", "refSections": [{"refs": [{"id": "ref-for-dom-mllstmoptions-returnsequence"}, {"id": "ref-for-dom-mllstmoptions-returnsequence\u2460"}, {"id": "ref-for-dom-mllstmoptions-returnsequence\u2461"}], "title": "7.6.21. The lstm() method"}], "external": false};
+window.dfnpanelData['dom-mllstmoptions-direction'] = {"dfnID": "dom-mllstmoptions-direction", "url": "#dom-mllstmoptions-direction", "dfnText": "direction", "refSections": [{"refs": [{"id": "ref-for-dom-mllstmoptions-direction"}, {"id": "ref-for-dom-mllstmoptions-direction\u2460"}], "title": "7.6.21. The lstm() method"}], "external": false};
+window.dfnpanelData['dom-mllstmoptions-layout'] = {"dfnID": "dom-mllstmoptions-layout", "url": "#dom-mllstmoptions-layout", "dfnText": "layout", "refSections": [{"refs": [{"id": "ref-for-dom-mllstmoptions-layout"}, {"id": "ref-for-dom-mllstmoptions-layout\u2460"}, {"id": "ref-for-dom-mllstmoptions-layout\u2461"}, {"id": "ref-for-dom-mllstmoptions-layout\u2462"}, {"id": "ref-for-dom-mllstmoptions-layout\u2463"}], "title": "7.6.21. The lstm() method"}], "external": false};
+window.dfnpanelData['dom-mllstmoptions-activations'] = {"dfnID": "dom-mllstmoptions-activations", "url": "#dom-mllstmoptions-activations", "dfnText": "activations", "refSections": [{"refs": [{"id": "ref-for-dom-mllstmoptions-activations"}, {"id": "ref-for-dom-mllstmoptions-activations\u2460"}], "title": "7.6.21. The lstm() method"}], "external": false};
+window.dfnpanelData['dom-mlgraphbuilder-lstm'] = {"dfnID": "dom-mlgraphbuilder-lstm", "url": "#dom-mlgraphbuilder-lstm", "dfnText": "lstm(input, weight, recurrentWeight, steps, hiddenSize, options)", "refSections": [{"refs": [{"id": "ref-for-dom-mlgraphbuilder-lstm"}], "title": "7.6.21. The lstm() method"}], "external": false};
+window.dfnpanelData['dictdef-mllstmcelloptions'] = {"dfnID": "dictdef-mllstmcelloptions", "url": "#dictdef-mllstmcelloptions", "dfnText": "MLLstmCellOptions", "refSections": [{"refs": [{"id": "ref-for-dictdef-mllstmcelloptions"}, {"id": "ref-for-dictdef-mllstmcelloptions\u2460"}, {"id": "ref-for-dictdef-mllstmcelloptions\u2461"}], "title": "7.6.22. The lstmCell() method"}], "external": false};
+window.dfnpanelData['dom-mllstmcelloptions-bias'] = {"dfnID": "dom-mllstmcelloptions-bias", "url": "#dom-mllstmcelloptions-bias", "dfnText": "bias", "refSections": [{"refs": [{"id": "ref-for-dom-mllstmcelloptions-bias"}, {"id": "ref-for-dom-mllstmcelloptions-bias\u2460"}, {"id": "ref-for-dom-mllstmcelloptions-bias\u2461"}], "title": "7.6.22. The lstmCell() method"}], "external": false};
+window.dfnpanelData['dom-mllstmcelloptions-recurrentbias'] = {"dfnID": "dom-mllstmcelloptions-recurrentbias", "url": "#dom-mllstmcelloptions-recurrentbias", "dfnText": "recurrentBias", "refSections": [{"refs": [{"id": "ref-for-dom-mllstmcelloptions-recurrentbias"}, {"id": "ref-for-dom-mllstmcelloptions-recurrentbias\u2460"}, {"id": "ref-for-dom-mllstmcelloptions-recurrentbias\u2461"}], "title": "7.6.22. The lstmCell() method"}], "external": false};
+window.dfnpanelData['dom-mllstmcelloptions-peepholeweight'] = {"dfnID": "dom-mllstmcelloptions-peepholeweight", "url": "#dom-mllstmcelloptions-peepholeweight", "dfnText": "peepholeWeight", "refSections": [{"refs": [{"id": "ref-for-dom-mllstmcelloptions-peepholeweight"}, {"id": "ref-for-dom-mllstmcelloptions-peepholeweight\u2460"}, {"id": "ref-for-dom-mllstmcelloptions-peepholeweight\u2461"}], "title": "7.6.22. The lstmCell() method"}], "external": false};
+window.dfnpanelData['dom-mllstmcelloptions-layout'] = {"dfnID": "dom-mllstmcelloptions-layout", "url": "#dom-mllstmcelloptions-layout", "dfnText": "layout", "refSections": [{"refs": [{"id": "ref-for-dom-mllstmcelloptions-layout"}, {"id": "ref-for-dom-mllstmcelloptions-layout\u2460"}, {"id": "ref-for-dom-mllstmcelloptions-layout\u2461"}], "title": "7.6.22. The lstmCell() method"}], "external": false};
+window.dfnpanelData['dom-mllstmcelloptions-activations'] = {"dfnID": "dom-mllstmcelloptions-activations", "url": "#dom-mllstmcelloptions-activations", "dfnText": "activations", "refSections": [{"refs": [{"id": "ref-for-dom-mllstmcelloptions-activations"}, {"id": "ref-for-dom-mllstmcelloptions-activations\u2460"}], "title": "7.6.22. The lstmCell() method"}], "external": false};
+window.dfnpanelData['dom-mlgraphbuilder-lstmcell'] = {"dfnID": "dom-mlgraphbuilder-lstmcell", "url": "#dom-mlgraphbuilder-lstmcell", "dfnText": "lstmCell(input, weight, recurrentWeight, hiddenState, cellState, hiddenSize, options)", "refSections": [{"refs": [{"id": "ref-for-dom-mlgraphbuilder-lstmcell"}], "title": "7.6.22. The lstmCell() method"}], "external": false};
+window.dfnpanelData['mlgraphbuilder-calculate-matmul-output-sizes'] = {"dfnID": "mlgraphbuilder-calculate-matmul-output-sizes", "url": "#mlgraphbuilder-calculate-matmul-output-sizes", "dfnText": "calculate matmul output sizes", "refSections": [{"refs": [{"id": "ref-for-mlgraphbuilder-calculate-matmul-output-sizes"}], "title": "7.6.23. The matmul() method"}], "external": false};
+window.dfnpanelData['dom-mlgraphbuilder-matmul'] = {"dfnID": "dom-mlgraphbuilder-matmul", "url": "#dom-mlgraphbuilder-matmul", "dfnText": "matmul(a, b)", "refSections": [{"refs": [{"id": "ref-for-dom-mlgraphbuilder-matmul"}], "title": "7.6.23. The matmul() method"}], "external": false};
+window.dfnpanelData['enumdef-mlpaddingmode'] = {"dfnID": "enumdef-mlpaddingmode", "url": "#enumdef-mlpaddingmode", "dfnText": "MLPaddingMode", "refSections": [{"refs": [{"id": "ref-for-enumdef-mlpaddingmode"}, {"id": "ref-for-enumdef-mlpaddingmode\u2460"}, {"id": "ref-for-enumdef-mlpaddingmode\u2461"}], "title": "7.6.24. The pad() method"}], "external": false};
+window.dfnpanelData['dictdef-mlpadoptions'] = {"dfnID": "dictdef-mlpadoptions", "url": "#dictdef-mlpadoptions", "dfnText": "MLPadOptions", "refSections": [{"refs": [{"id": "ref-for-dictdef-mlpadoptions"}, {"id": "ref-for-dictdef-mlpadoptions\u2460"}, {"id": "ref-for-dictdef-mlpadoptions\u2461"}], "title": "7.6.24. The pad() method"}], "external": false};
+window.dfnpanelData['dom-mlpadoptions-mode'] = {"dfnID": "dom-mlpadoptions-mode", "url": "#dom-mlpadoptions-mode", "dfnText": "mode", "refSections": [{"refs": [{"id": "ref-for-dom-mlpadoptions-mode"}, {"id": "ref-for-dom-mlpadoptions-mode\u2460"}], "title": "7.6.24. The pad() method"}], "external": false};
+window.dfnpanelData['dom-mlpadoptions-value'] = {"dfnID": "dom-mlpadoptions-value", "url": "#dom-mlpadoptions-value", "dfnText": "value", "refSections": [{"refs": [{"id": "ref-for-dom-mlpadoptions-value"}], "title": "7.6.24. The pad() method"}], "external": false};
+window.dfnpanelData['mlgraphbuilder-calculate-padding-output-sizes'] = {"dfnID": "mlgraphbuilder-calculate-padding-output-sizes", "url": "#mlgraphbuilder-calculate-padding-output-sizes", "dfnText": "calculate padding output sizes", "refSections": [{"refs": [{"id": "ref-for-mlgraphbuilder-calculate-padding-output-sizes"}], "title": "7.6.24. The pad() method"}], "external": false};
+window.dfnpanelData['dom-mlgraphbuilder-pad'] = {"dfnID": "dom-mlgraphbuilder-pad", "url": "#dom-mlgraphbuilder-pad", "dfnText": "pad(input, beginningPadding, endingPadding, options)", "refSections": [{"refs": [{"id": "ref-for-dom-mlgraphbuilder-pad"}], "title": "7.6.24. The pad() method"}], "external": false};
+window.dfnpanelData['enumdef-mlroundingtype'] = {"dfnID": "enumdef-mlroundingtype", "url": "#enumdef-mlroundingtype", "dfnText": "MLRoundingType", "refSections": [{"refs": [{"id": "ref-for-enumdef-mlroundingtype"}, {"id": "ref-for-enumdef-mlroundingtype\u2460"}, {"id": "ref-for-enumdef-mlroundingtype\u2461"}], "title": "7.6.25. Pooling operations"}], "external": false};
+window.dfnpanelData['dictdef-mlpool2doptions'] = {"dfnID": "dictdef-mlpool2doptions", "url": "#dictdef-mlpool2doptions", "dfnText": "MLPool2dOptions", "refSections": [{"refs": [{"id": "ref-for-dictdef-mlpool2doptions"}, {"id": "ref-for-dictdef-mlpool2doptions\u2460"}, {"id": "ref-for-dictdef-mlpool2doptions\u2461"}, {"id": "ref-for-dictdef-mlpool2doptions\u2462"}, {"id": "ref-for-dictdef-mlpool2doptions\u2463"}], "title": "7.6.25. Pooling operations"}], "external": false};
+window.dfnpanelData['dom-mlpool2doptions-windowdimensions'] = {"dfnID": "dom-mlpool2doptions-windowdimensions", "url": "#dom-mlpool2doptions-windowdimensions", "dfnText": "windowDimensions", "refSections": [{"refs": [{"id": "ref-for-dom-mlpool2doptions-windowdimensions"}, {"id": "ref-for-dom-mlpool2doptions-windowdimensions\u2460"}, {"id": "ref-for-dom-mlpool2doptions-windowdimensions\u2461"}], "title": "7.6.25. Pooling operations"}], "external": false};
+window.dfnpanelData['dom-mlpool2doptions-padding'] = {"dfnID": "dom-mlpool2doptions-padding", "url": "#dom-mlpool2doptions-padding", "dfnText": "padding", "refSections": [{"refs": [{"id": "ref-for-dom-mlpool2doptions-padding"}, {"id": "ref-for-dom-mlpool2doptions-padding\u2460"}, {"id": "ref-for-dom-mlpool2doptions-padding\u2461"}, {"id": "ref-for-dom-mlpool2doptions-padding\u2462"}, {"id": "ref-for-dom-mlpool2doptions-padding\u2463"}, {"id": "ref-for-dom-mlpool2doptions-padding\u2464"}, {"id": "ref-for-dom-mlpool2doptions-padding\u2465"}], "title": "7.6.25. Pooling operations"}], "external": false};
+window.dfnpanelData['dom-mlpool2doptions-strides'] = {"dfnID": "dom-mlpool2doptions-strides", "url": "#dom-mlpool2doptions-strides", "dfnText": "strides", "refSections": [{"refs": [{"id": "ref-for-dom-mlpool2doptions-strides"}, {"id": "ref-for-dom-mlpool2doptions-strides\u2460"}, {"id": "ref-for-dom-mlpool2doptions-strides\u2461"}, {"id": "ref-for-dom-mlpool2doptions-strides\u2462"}, {"id": "ref-for-dom-mlpool2doptions-strides\u2463"}, {"id": "ref-for-dom-mlpool2doptions-strides\u2464"}], "title": "7.6.25. Pooling operations"}], "external": false};
+window.dfnpanelData['dom-mlpool2doptions-dilations'] = {"dfnID": "dom-mlpool2doptions-dilations", "url": "#dom-mlpool2doptions-dilations", "dfnText": "dilations", "refSections": [{"refs": [{"id": "ref-for-dom-mlpool2doptions-dilations"}, {"id": "ref-for-dom-mlpool2doptions-dilations\u2460"}, {"id": "ref-for-dom-mlpool2doptions-dilations\u2461"}, {"id": "ref-for-dom-mlpool2doptions-dilations\u2462"}, {"id": "ref-for-dom-mlpool2doptions-dilations\u2463"}], "title": "7.6.25. Pooling operations"}], "external": false};
+window.dfnpanelData['dom-mlpool2doptions-autopad'] = {"dfnID": "dom-mlpool2doptions-autopad", "url": "#dom-mlpool2doptions-autopad", "dfnText": "autoPad", "refSections": [{"refs": [{"id": "ref-for-dom-mlpool2doptions-autopad"}, {"id": "ref-for-dom-mlpool2doptions-autopad\u2460"}], "title": "7.6.25. Pooling operations"}], "external": false};
+window.dfnpanelData['dom-mlpool2doptions-layout'] = {"dfnID": "dom-mlpool2doptions-layout", "url": "#dom-mlpool2doptions-layout", "dfnText": "layout", "refSections": [{"refs": [{"id": "ref-for-dom-mlpool2doptions-layout"}], "title": "7.6.25. Pooling operations"}], "external": false};
+window.dfnpanelData['dom-mlpool2doptions-roundingtype'] = {"dfnID": "dom-mlpool2doptions-roundingtype", "url": "#dom-mlpool2doptions-roundingtype", "dfnText": "roundingType", "refSections": [{"refs": [{"id": "ref-for-dom-mlpool2doptions-roundingtype"}, {"id": "ref-for-dom-mlpool2doptions-roundingtype\u2460"}], "title": "7.6.25. Pooling operations"}], "external": false};
+window.dfnpanelData['dom-mlpool2doptions-outputsizes'] = {"dfnID": "dom-mlpool2doptions-outputsizes", "url": "#dom-mlpool2doptions-outputsizes", "dfnText": "outputSizes", "refSections": [{"refs": [{"id": "ref-for-dom-mlpool2doptions-outputsizes"}, {"id": "ref-for-dom-mlpool2doptions-outputsizes\u2460"}, {"id": "ref-for-dom-mlpool2doptions-outputsizes\u2461"}, {"id": "ref-for-dom-mlpool2doptions-outputsizes\u2462"}, {"id": "ref-for-dom-mlpool2doptions-outputsizes\u2463"}], "title": "7.6.25. Pooling operations"}], "external": false};
+window.dfnpanelData['mlgraphbuilder-pooling-op'] = {"dfnID": "mlgraphbuilder-pooling-op", "url": "#mlgraphbuilder-pooling-op", "dfnText": "create pooling operation", "refSections": [{"refs": [{"id": "ref-for-mlgraphbuilder-pooling-op"}, {"id": "ref-for-mlgraphbuilder-pooling-op\u2460"}, {"id": "ref-for-mlgraphbuilder-pooling-op\u2461"}], "title": "7.6.25. Pooling operations"}], "external": false};
+window.dfnpanelData['dom-mlgraphbuilder-averagepool2d'] = {"dfnID": "dom-mlgraphbuilder-averagepool2d", "url": "#dom-mlgraphbuilder-averagepool2d", "dfnText": "averagePool2d(input, options)", "refSections": [{"refs": [{"id": "ref-for-dom-mlgraphbuilder-averagepool2d"}], "title": "7.6.25. Pooling operations"}], "external": false};
+window.dfnpanelData['dom-mlgraphbuilder-l2pool2d'] = {"dfnID": "dom-mlgraphbuilder-l2pool2d", "url": "#dom-mlgraphbuilder-l2pool2d", "dfnText": "l2Pool2d(input, options)", "refSections": [{"refs": [{"id": "ref-for-dom-mlgraphbuilder-l2pool2d"}], "title": "7.6.25. Pooling operations"}], "external": false};
+window.dfnpanelData['dom-mlgraphbuilder-maxpool2d'] = {"dfnID": "dom-mlgraphbuilder-maxpool2d", "url": "#dom-mlgraphbuilder-maxpool2d", "dfnText": "maxPool2d(input, options)", "refSections": [{"refs": [{"id": "ref-for-dom-mlgraphbuilder-maxpool2d"}], "title": "7.6.25. Pooling operations"}], "external": false};
+window.dfnpanelData['dom-mlgraphbuilder-prelu'] = {"dfnID": "dom-mlgraphbuilder-prelu", "url": "#dom-mlgraphbuilder-prelu", "dfnText": "prelu(input, slope)", "refSections": [{"refs": [{"id": "ref-for-dom-mlgraphbuilder-prelu"}], "title": "7.6.26. The prelu() method"}], "external": false};
+window.dfnpanelData['dictdef-mlreduceoptions'] = {"dfnID": "dictdef-mlreduceoptions", "url": "#dictdef-mlreduceoptions", "dfnText": "MLReduceOptions", "refSections": [{"refs": [{"id": "ref-for-dictdef-mlreduceoptions"}, {"id": "ref-for-dictdef-mlreduceoptions\u2460"}, {"id": "ref-for-dictdef-mlreduceoptions\u2461"}, {"id": "ref-for-dictdef-mlreduceoptions\u2462"}, {"id": "ref-for-dictdef-mlreduceoptions\u2463"}, {"id": "ref-for-dictdef-mlreduceoptions\u2464"}, {"id": "ref-for-dictdef-mlreduceoptions\u2465"}, {"id": "ref-for-dictdef-mlreduceoptions\u2466"}, {"id": "ref-for-dictdef-mlreduceoptions\u2467"}, {"id": "ref-for-dictdef-mlreduceoptions\u2468"}, {"id": "ref-for-dictdef-mlreduceoptions\u2460\u24ea"}], "title": "7.6.27. Reduction operations"}], "external": false};
+window.dfnpanelData['dom-mlreduceoptions-axes'] = {"dfnID": "dom-mlreduceoptions-axes", "url": "#dom-mlreduceoptions-axes", "dfnText": "axes", "refSections": [{"refs": [{"id": "ref-for-dom-mlreduceoptions-axes"}, {"id": "ref-for-dom-mlreduceoptions-axes\u2460"}], "title": "7.6.27. Reduction operations"}], "external": false};
+window.dfnpanelData['dom-mlreduceoptions-keepdimensions'] = {"dfnID": "dom-mlreduceoptions-keepdimensions", "url": "#dom-mlreduceoptions-keepdimensions", "dfnText": "keepDimensions", "refSections": [{"refs": [{"id": "ref-for-dom-mlreduceoptions-keepdimensions"}], "title": "7.6.27. Reduction operations"}], "external": false};
+window.dfnpanelData['mlgraphbuilder-reduce-op'] = {"dfnID": "mlgraphbuilder-reduce-op", "url": "#mlgraphbuilder-reduce-op", "dfnText": "create reduce operation", "refSections": [{"refs": [{"id": "ref-for-mlgraphbuilder-reduce-op"}, {"id": "ref-for-mlgraphbuilder-reduce-op\u2460"}, {"id": "ref-for-mlgraphbuilder-reduce-op\u2461"}, {"id": "ref-for-mlgraphbuilder-reduce-op\u2462"}, {"id": "ref-for-mlgraphbuilder-reduce-op\u2463"}, {"id": "ref-for-mlgraphbuilder-reduce-op\u2464"}, {"id": "ref-for-mlgraphbuilder-reduce-op\u2465"}, {"id": "ref-for-mlgraphbuilder-reduce-op\u2466"}, {"id": "ref-for-mlgraphbuilder-reduce-op\u2467"}, {"id": "ref-for-mlgraphbuilder-reduce-op\u2468"}], "title": "7.6.27. Reduction operations"}], "external": false};
+window.dfnpanelData['dom-mlgraphbuilder-reducel1'] = {"dfnID": "dom-mlgraphbuilder-reducel1", "url": "#dom-mlgraphbuilder-reducel1", "dfnText": "reduceL1(input, options)", "refSections": [{"refs": [{"id": "ref-for-dom-mlgraphbuilder-reducel1"}], "title": "7.6.27. Reduction operations"}], "external": false};
+window.dfnpanelData['dom-mlgraphbuilder-reducel2'] = {"dfnID": "dom-mlgraphbuilder-reducel2", "url": "#dom-mlgraphbuilder-reducel2", "dfnText": "reduceL2(input, options)", "refSections": [{"refs": [{"id": "ref-for-dom-mlgraphbuilder-reducel2"}], "title": "7.6.27. Reduction operations"}], "external": false};
+window.dfnpanelData['dom-mlgraphbuilder-reducelogsum'] = {"dfnID": "dom-mlgraphbuilder-reducelogsum", "url": "#dom-mlgraphbuilder-reducelogsum", "dfnText": "reduceLogSum(input, options)", "refSections": [{"refs": [{"id": "ref-for-dom-mlgraphbuilder-reducelogsum"}], "title": "7.6.27. Reduction operations"}], "external": false};
+window.dfnpanelData['dom-mlgraphbuilder-reducelogsumexp'] = {"dfnID": "dom-mlgraphbuilder-reducelogsumexp", "url": "#dom-mlgraphbuilder-reducelogsumexp", "dfnText": "reduceLogSumExp(input, options)", "refSections": [{"refs": [{"id": "ref-for-dom-mlgraphbuilder-reducelogsumexp"}], "title": "7.6.27. Reduction operations"}], "external": false};
+window.dfnpanelData['dom-mlgraphbuilder-reducemax'] = {"dfnID": "dom-mlgraphbuilder-reducemax", "url": "#dom-mlgraphbuilder-reducemax", "dfnText": "reduceMax(input, options)", "refSections": [{"refs": [{"id": "ref-for-dom-mlgraphbuilder-reducemax"}], "title": "7.6.27. Reduction operations"}], "external": false};
+window.dfnpanelData['dom-mlgraphbuilder-reducemean'] = {"dfnID": "dom-mlgraphbuilder-reducemean", "url": "#dom-mlgraphbuilder-reducemean", "dfnText": "reduceMean(input, options)", "refSections": [{"refs": [{"id": "ref-for-dom-mlgraphbuilder-reducemean"}], "title": "7.6.27. Reduction operations"}], "external": false};
+window.dfnpanelData['dom-mlgraphbuilder-reducemin'] = {"dfnID": "dom-mlgraphbuilder-reducemin", "url": "#dom-mlgraphbuilder-reducemin", "dfnText": "reduceMin(input, options)", "refSections": [{"refs": [{"id": "ref-for-dom-mlgraphbuilder-reducemin"}], "title": "7.6.27. Reduction operations"}], "external": false};
+window.dfnpanelData['dom-mlgraphbuilder-reduceproduct'] = {"dfnID": "dom-mlgraphbuilder-reduceproduct", "url": "#dom-mlgraphbuilder-reduceproduct", "dfnText": "reduceProduct(input, options)", "refSections": [{"refs": [{"id": "ref-for-dom-mlgraphbuilder-reduceproduct"}], "title": "7.6.27. Reduction operations"}], "external": false};
+window.dfnpanelData['dom-mlgraphbuilder-reducesum'] = {"dfnID": "dom-mlgraphbuilder-reducesum", "url": "#dom-mlgraphbuilder-reducesum", "dfnText": "reduceSum(input, options)", "refSections": [{"refs": [{"id": "ref-for-dom-mlgraphbuilder-reducesum"}], "title": "7.6.27. Reduction operations"}], "external": false};
+window.dfnpanelData['dom-mlgraphbuilder-reducesumsquare'] = {"dfnID": "dom-mlgraphbuilder-reducesumsquare", "url": "#dom-mlgraphbuilder-reducesumsquare", "dfnText": "reduceSumSquare(input, options)", "refSections": [{"refs": [{"id": "ref-for-dom-mlgraphbuilder-reducesumsquare"}], "title": "7.6.27. Reduction operations"}], "external": false};
+window.dfnpanelData['dom-mlgraphbuilder-relu'] = {"dfnID": "dom-mlgraphbuilder-relu", "url": "#dom-mlgraphbuilder-relu", "dfnText": "relu(input)", "refSections": [{"refs": [{"id": "ref-for-dom-mlgraphbuilder-relu"}], "title": "7.6.28. The relu() method"}, {"refs": [{"id": "ref-for-dom-mlgraphbuilder-relu\u2460"}], "title": "7.6.28.1. The relu(input) method"}], "external": false};
+window.dfnpanelData['relu-noargs'] = {"dfnID": "relu-noargs", "url": "#relu-noargs", "dfnText": "relu()", "refSections": [{"refs": [{"id": "ref-for-relu-noargs"}], "title": "7.6.28. The relu() method"}, {"refs": [{"id": "ref-for-relu-noargs\u2460"}], "title": "7.6.28.2. The relu() method"}], "external": false};
+window.dfnpanelData['enumdef-mlinterpolationmode'] = {"dfnID": "enumdef-mlinterpolationmode", "url": "#enumdef-mlinterpolationmode", "dfnText": "MLInterpolationMode", "refSections": [{"refs": [{"id": "ref-for-enumdef-mlinterpolationmode"}, {"id": "ref-for-enumdef-mlinterpolationmode\u2460"}, {"id": "ref-for-enumdef-mlinterpolationmode\u2461"}], "title": "7.6.29. The resample2d() method"}], "external": false};
+window.dfnpanelData['dictdef-mlresample2doptions'] = {"dfnID": "dictdef-mlresample2doptions", "url": "#dictdef-mlresample2doptions", "dfnText": "MLResample2dOptions", "refSections": [{"refs": [{"id": "ref-for-dictdef-mlresample2doptions"}, {"id": "ref-for-dictdef-mlresample2doptions\u2460"}, {"id": "ref-for-dictdef-mlresample2doptions\u2461"}], "title": "7.6.29. The resample2d() method"}], "external": false};
+window.dfnpanelData['dom-mlresample2doptions-mode'] = {"dfnID": "dom-mlresample2doptions-mode", "url": "#dom-mlresample2doptions-mode", "dfnText": "mode", "refSections": [{"refs": [{"id": "ref-for-dom-mlresample2doptions-mode"}], "title": "7.6.29. The resample2d() method"}], "external": false};
+window.dfnpanelData['dom-mlresample2doptions-scales'] = {"dfnID": "dom-mlresample2doptions-scales", "url": "#dom-mlresample2doptions-scales", "dfnText": "scales", "refSections": [{"refs": [{"id": "ref-for-dom-mlresample2doptions-scales"}, {"id": "ref-for-dom-mlresample2doptions-scales\u2460"}, {"id": "ref-for-dom-mlresample2doptions-scales\u2461"}, {"id": "ref-for-dom-mlresample2doptions-scales\u2462"}], "title": "7.6.29. The resample2d() method"}], "external": false};
+window.dfnpanelData['dom-mlresample2doptions-sizes'] = {"dfnID": "dom-mlresample2doptions-sizes", "url": "#dom-mlresample2doptions-sizes", "dfnText": "sizes", "refSections": [{"refs": [{"id": "ref-for-dom-mlresample2doptions-sizes"}, {"id": "ref-for-dom-mlresample2doptions-sizes\u2460"}, {"id": "ref-for-dom-mlresample2doptions-sizes\u2461"}, {"id": "ref-for-dom-mlresample2doptions-sizes\u2462"}], "title": "7.6.29. The resample2d() method"}], "external": false};
+window.dfnpanelData['dom-mlresample2doptions-axes'] = {"dfnID": "dom-mlresample2doptions-axes", "url": "#dom-mlresample2doptions-axes", "dfnText": "axes", "refSections": [{"refs": [{"id": "ref-for-dom-mlresample2doptions-axes"}, {"id": "ref-for-dom-mlresample2doptions-axes\u2460"}, {"id": "ref-for-dom-mlresample2doptions-axes\u2461"}, {"id": "ref-for-dom-mlresample2doptions-axes\u2462"}, {"id": "ref-for-dom-mlresample2doptions-axes\u2463"}], "title": "7.6.29. The resample2d() method"}], "external": false};
+window.dfnpanelData['mlgraphbuilder-check-resample-options'] = {"dfnID": "mlgraphbuilder-check-resample-options", "url": "#mlgraphbuilder-check-resample-options", "dfnText": "check resample options", "refSections": [{"refs": [{"id": "ref-for-mlgraphbuilder-check-resample-options"}], "title": "7.6.29. The resample2d() method"}], "external": false};
+window.dfnpanelData['mlgraphbuilder-resample-output-sizes'] = {"dfnID": "mlgraphbuilder-resample-output-sizes", "url": "#mlgraphbuilder-resample-output-sizes", "dfnText": "resample output sizes", "refSections": [{"refs": [{"id": "ref-for-mlgraphbuilder-resample-output-sizes"}], "title": "7.6.29. The resample2d() method"}], "external": false};
+window.dfnpanelData['dom-mlgraphbuilder-resample2d'] = {"dfnID": "dom-mlgraphbuilder-resample2d", "url": "#dom-mlgraphbuilder-resample2d", "dfnText": "resample2d(input, options)", "refSections": [{"refs": [{"id": "ref-for-dom-mlgraphbuilder-resample2d"}], "title": "7.6.29. The resample2d() method"}], "external": false};
+window.dfnpanelData['dom-mlgraphbuilder-reshape'] = {"dfnID": "dom-mlgraphbuilder-reshape", "url": "#dom-mlgraphbuilder-reshape", "dfnText": "reshape(input, newShape)", "refSections": [{"refs": [{"id": "ref-for-dom-mlgraphbuilder-reshape"}], "title": "7.6.30. The reshape() method"}], "external": false};
+window.dfnpanelData['dom-mlgraphbuilder-sigmoid'] = {"dfnID": "dom-mlgraphbuilder-sigmoid", "url": "#dom-mlgraphbuilder-sigmoid", "dfnText": "sigmoid(input)", "refSections": [{"refs": [{"id": "ref-for-dom-mlgraphbuilder-sigmoid"}], "title": "7.6.31. The sigmoid() method"}, {"refs": [{"id": "ref-for-dom-mlgraphbuilder-sigmoid\u2460"}], "title": "7.6.31.1. The sigmoid(input) method"}], "external": false};
+window.dfnpanelData['sigmoid-noargs'] = {"dfnID": "sigmoid-noargs", "url": "#sigmoid-noargs", "dfnText": "sigmoid()", "refSections": [{"refs": [{"id": "ref-for-sigmoid-noargs"}], "title": "7.6.31. The sigmoid() method"}, {"refs": [{"id": "ref-for-sigmoid-noargs\u2460"}], "title": "7.6.31.2. The sigmoid() method"}], "external": false};
+window.dfnpanelData['dom-mlgraphbuilder-slice'] = {"dfnID": "dom-mlgraphbuilder-slice", "url": "#dom-mlgraphbuilder-slice", "dfnText": "slice(input, starts, sizes)", "refSections": [{"refs": [{"id": "ref-for-dom-mlgraphbuilder-slice"}], "title": "7.6.32. The slice() method"}], "external": false};
+window.dfnpanelData['dom-mlgraphbuilder-softmax'] = {"dfnID": "dom-mlgraphbuilder-softmax", "url": "#dom-mlgraphbuilder-softmax", "dfnText": "softmax(input)", "refSections": [{"refs": [{"id": "ref-for-dom-mlgraphbuilder-softmax"}], "title": "7.6.33. The softmax() method"}, {"refs": [{"id": "ref-for-dom-mlgraphbuilder-softmax\u2460"}], "title": "7.6.33.1. The softmax(input) method"}], "external": false};
+window.dfnpanelData['softmax-noargs'] = {"dfnID": "softmax-noargs", "url": "#softmax-noargs", "dfnText": "softmax()", "refSections": [{"refs": [{"id": "ref-for-softmax-noargs"}], "title": "6.1. Overview"}, {"refs": [{"id": "ref-for-softmax-noargs\u2460"}], "title": "7.6.33. The softmax() method"}, {"refs": [{"id": "ref-for-softmax-noargs\u2461"}], "title": "7.6.33.2. The softmax() method"}], "external": false};
+window.dfnpanelData['dictdef-mlsoftplusoptions'] = {"dfnID": "dictdef-mlsoftplusoptions", "url": "#dictdef-mlsoftplusoptions", "dfnText": "MLSoftplusOptions", "refSections": [{"refs": [{"id": "ref-for-dictdef-mlsoftplusoptions"}, {"id": "ref-for-dictdef-mlsoftplusoptions\u2460"}, {"id": "ref-for-dictdef-mlsoftplusoptions\u2461"}], "title": "7.6.34. The softplus() method"}, {"refs": [{"id": "ref-for-dictdef-mlsoftplusoptions\u2462"}], "title": "7.6.34.1. The softplus(input, options) method"}, {"refs": [{"id": "ref-for-dictdef-mlsoftplusoptions\u2463"}], "title": "7.6.34.2. The softplus(options) method"}], "external": false};
+window.dfnpanelData['dom-mlsoftplusoptions-steepness'] = {"dfnID": "dom-mlsoftplusoptions-steepness", "url": "#dom-mlsoftplusoptions-steepness", "dfnText": "steepness", "refSections": [{"refs": [{"id": "ref-for-dom-mlsoftplusoptions-steepness"}], "title": "7.6.34. The softplus() method"}], "external": false};
+window.dfnpanelData['dom-mlgraphbuilder-softplus'] = {"dfnID": "dom-mlgraphbuilder-softplus", "url": "#dom-mlgraphbuilder-softplus", "dfnText": "softplus(input, options)", "refSections": [{"refs": [{"id": "ref-for-dom-mlgraphbuilder-softplus"}], "title": "7.6.34. The softplus() method"}, {"refs": [{"id": "ref-for-dom-mlgraphbuilder-softplus\u2460"}], "title": "7.6.34.1. The softplus(input, options) method"}], "external": false};
+window.dfnpanelData['dom-mlgraphbuilder-softplus-options'] = {"dfnID": "dom-mlgraphbuilder-softplus-options", "url": "#dom-mlgraphbuilder-softplus-options", "dfnText": "softplus(options)", "refSections": [{"refs": [{"id": "ref-for-dom-mlgraphbuilder-softplus-options"}], "title": "7.6.34. The softplus() method"}, {"refs": [{"id": "ref-for-dom-mlgraphbuilder-softplus-options\u2460"}], "title": "7.6.34.2. The softplus(options) method"}], "external": false};
+window.dfnpanelData['dom-mlgraphbuilder-softsign'] = {"dfnID": "dom-mlgraphbuilder-softsign", "url": "#dom-mlgraphbuilder-softsign", "dfnText": "softsign(input)", "refSections": [{"refs": [{"id": "ref-for-dom-mlgraphbuilder-softsign"}], "title": "7.6.35. The softsign() method"}, {"refs": [{"id": "ref-for-dom-mlgraphbuilder-softsign\u2460"}], "title": "7.6.35.1. The softsign(input) method"}], "external": false};
+window.dfnpanelData['softsign-noargs'] = {"dfnID": "softsign-noargs", "url": "#softsign-noargs", "dfnText": "softsign()", "refSections": [{"refs": [{"id": "ref-for-softsign-noargs"}], "title": "7.6.35. The softsign() method"}, {"refs": [{"id": "ref-for-softsign-noargs\u2460"}], "title": "7.6.35.2. The softsign() method"}], "external": false};
+window.dfnpanelData['dictdef-mlsplitoptions'] = {"dfnID": "dictdef-mlsplitoptions", "url": "#dictdef-mlsplitoptions", "dfnText": "MLSplitOptions", "refSections": [{"refs": [{"id": "ref-for-dictdef-mlsplitoptions"}, {"id": "ref-for-dictdef-mlsplitoptions\u2460"}, {"id": "ref-for-dictdef-mlsplitoptions\u2461"}], "title": "7.6.36. The split() method"}], "external": false};
+window.dfnpanelData['dom-mlsplitoptions-axis'] = {"dfnID": "dom-mlsplitoptions-axis", "url": "#dom-mlsplitoptions-axis", "dfnText": "axis", "refSections": [{"refs": [{"id": "ref-for-dom-mlsplitoptions-axis"}, {"id": "ref-for-dom-mlsplitoptions-axis\u2460"}, {"id": "ref-for-dom-mlsplitoptions-axis\u2461"}], "title": "7.6.36. The split() method"}], "external": false};
+window.dfnpanelData['dom-mlgraphbuilder-split'] = {"dfnID": "dom-mlgraphbuilder-split", "url": "#dom-mlgraphbuilder-split", "dfnText": "split(input, splits, options)", "refSections": [{"refs": [{"id": "ref-for-dom-mlgraphbuilder-split"}], "title": "7.6.36. The split() method"}], "external": false};
+window.dfnpanelData['dictdef-mlsqueezeoptions'] = {"dfnID": "dictdef-mlsqueezeoptions", "url": "#dictdef-mlsqueezeoptions", "dfnText": "MLSqueezeOptions", "refSections": [{"refs": [{"id": "ref-for-dictdef-mlsqueezeoptions"}, {"id": "ref-for-dictdef-mlsqueezeoptions\u2460"}, {"id": "ref-for-dictdef-mlsqueezeoptions\u2461"}], "title": "7.6.37. The squeeze() method"}], "external": false};
+window.dfnpanelData['dom-mlsqueezeoptions-axes'] = {"dfnID": "dom-mlsqueezeoptions-axes", "url": "#dom-mlsqueezeoptions-axes", "dfnText": "axes", "refSections": [{"refs": [{"id": "ref-for-dom-mlsqueezeoptions-axes"}, {"id": "ref-for-dom-mlsqueezeoptions-axes\u2460"}, {"id": "ref-for-dom-mlsqueezeoptions-axes\u2461"}, {"id": "ref-for-dom-mlsqueezeoptions-axes\u2462"}], "title": "7.6.37. The squeeze() method"}], "external": false};
+window.dfnpanelData['dom-mlgraphbuilder-squeeze'] = {"dfnID": "dom-mlgraphbuilder-squeeze", "url": "#dom-mlgraphbuilder-squeeze", "dfnText": "squeeze(input, options)", "refSections": [{"refs": [{"id": "ref-for-dom-mlgraphbuilder-squeeze"}], "title": "7.6.37. The squeeze() method"}], "external": false};
+window.dfnpanelData['dom-mlgraphbuilder-tanh'] = {"dfnID": "dom-mlgraphbuilder-tanh", "url": "#dom-mlgraphbuilder-tanh", "dfnText": "tanh(input)", "refSections": [{"refs": [{"id": "ref-for-dom-mlgraphbuilder-tanh"}], "title": "7.6.38. The tanh() method"}, {"refs": [{"id": "ref-for-dom-mlgraphbuilder-tanh\u2460"}], "title": "7.6.38.1. The tanh(input) method"}], "external": false};
+window.dfnpanelData['tanh-noargs'] = {"dfnID": "tanh-noargs", "url": "#tanh-noargs", "dfnText": "tanh()", "refSections": [{"refs": [{"id": "ref-for-tanh-noargs"}], "title": "7.6.38. The tanh() method"}, {"refs": [{"id": "ref-for-tanh-noargs\u2460"}], "title": "7.6.38.2. The tanh() method"}], "external": false};
+window.dfnpanelData['dictdef-mltransposeoptions'] = {"dfnID": "dictdef-mltransposeoptions", "url": "#dictdef-mltransposeoptions", "dfnText": "MLTransposeOptions", "refSections": [{"refs": [{"id": "ref-for-dictdef-mltransposeoptions"}, {"id": "ref-for-dictdef-mltransposeoptions\u2460"}, {"id": "ref-for-dictdef-mltransposeoptions\u2461"}], "title": "7.6.39. The transpose() method"}], "external": false};
+window.dfnpanelData['dom-mltransposeoptions-permutation'] = {"dfnID": "dom-mltransposeoptions-permutation", "url": "#dom-mltransposeoptions-permutation", "dfnText": "permutation", "refSections": [{"refs": [{"id": "ref-for-dom-mltransposeoptions-permutation"}, {"id": "ref-for-dom-mltransposeoptions-permutation\u2460"}, {"id": "ref-for-dom-mltransposeoptions-permutation\u2461"}, {"id": "ref-for-dom-mltransposeoptions-permutation\u2462"}, {"id": "ref-for-dom-mltransposeoptions-permutation\u2463"}, {"id": "ref-for-dom-mltransposeoptions-permutation\u2464"}, {"id": "ref-for-dom-mltransposeoptions-permutation\u2465"}], "title": "7.6.39. The transpose() method"}], "external": false};
+window.dfnpanelData['dom-mlgraphbuilder-transpose'] = {"dfnID": "dom-mlgraphbuilder-transpose", "url": "#dom-mlgraphbuilder-transpose", "dfnText": "transpose(input, options)", "refSections": [{"refs": [{"id": "ref-for-dom-mlgraphbuilder-transpose"}], "title": "7.6.39. The transpose() method"}], "external": false};
 </script>
 <script>/* Boilerplate: script-dom-helper */
 function query(sel) { return document.querySelector(sel); }