Use boolean flags instead of MLTensorUsage

Adapt to Chromium update:
https://chromium-review.googlesource.com/c/chromium/src/+/5933323

We will keep MLTensorUsage until it is remove from Chromium.

code/samples/matmul.js

@@ -13,7 +13,9 @@ const graph = await builder.build({c});
 const bufferA = new Float32Array(3*4).fill(1.0);
 const bufferB = new Float32Array(4*3).fill(0.8);
 descA.usage = MLTensorUsage.WRITE;
+descA.writable = true;
 descB.usage = MLTensorUsage.WRITE;
+descB.writable = true;
 const tensorA = await context.createTensor(descA);
 const tensorB = await context.createTensor(descB);
 context.writeTensor(tensorA, bufferA);
@@ -23,6 +25,7 @@ const tensorC = await context.createTensor({
   dimensions: [3, 3],
   shape: [3, 3],
   usage: MLTensorUsage.READ,
+  readable: true,
 });
 context.dispatch(graph, {a: tensorA, b: tensorB}, {c: tensorC});
 const results = await context.readTensor(tensorC);

code/samples/mul_add.js

@@ -13,13 +13,15 @@ const graph = await builder.build({'C': C});
 const bufferA = new Float32Array(4).fill(1.0);
 const bufferB = new Float32Array(4).fill(0.8);
 desc.usage = MLTensorUsage.WRITE;
+desc.writable = true;
 const tensorA = await context.createTensor(desc);
 const tensorB = await context.createTensor(desc);
 context.writeTensor(tensorA, bufferA);
 context.writeTensor(tensorB, bufferB);
 const tensorC = await context.createTensor({
   ...desc,
   usage: MLTensorUsage.READ,
+  readable: true,
 });
 const inputs = {'A': tensorA, 'B': tensorB};
 const outputs = {'C': tensorC};

code/samples/simple_graph.js

@@ -52,6 +52,7 @@ const inputBuffer1 = new Float32Array(TENSOR_SIZE).fill(1);
 const inputBuffer2 = new Float32Array(TENSOR_SIZE).fill(1);
 
 desc.usage = MLTensorUsage.WRITE;
+desc.writable = true;
 const inputTensor1 = await context.createTensor(desc);
 const inputTensor2 = await context.createTensor(desc);
 context.writeTensor(inputTensor1, inputBuffer1);
@@ -60,6 +61,7 @@ context.writeTensor(inputTensor2, inputBuffer2);
 const outputTensor = await context.createTensor({
   ...desc,
   usage: MLTensorUsage.READ,
+  readable: true,
 });
 
 // Execute the compiled graph with the specified inputs.

face_recognition/facenet_nchw.js

@@ -148,12 +148,14 @@ export class FaceNetNchw {
     };
     const input = this.builder_.input('input', inputDesc);
     inputDesc.usage = MLTensorUsage.WRITE;
+    inputDesc.writable = true;
     this.inputTensor_ = await this.context_.createTensor(inputDesc);
     this.outputTensor_ = await this.context_.createTensor({
       dataType: 'float32',
       dimensions: this.outputShape_,
       shape: this.outputShape_,
       usage: MLTensorUsage.READ,
+      readable: true,
     });
 
     const poolOptions = {windowDimensions: [3, 3], strides};

face_recognition/facenet_nhwc.js

@@ -149,12 +149,14 @@ export class FaceNetNhwc {
     };
     const input = this.builder_.input('input', inputDesc);
     inputDesc.usage = MLTensorUsage.WRITE;
+    inputDesc.writable = true;
     this.inputTensor_ = await this.context_.createTensor(inputDesc);
     this.outputTensor_ = await this.context_.createTensor({
       dataType: 'float32',
       dimensions: this.outputShape_,
       shape: this.outputShape_,
       usage: MLTensorUsage.READ,
+      readable: true,
     });
 
     const poolOptions = {windowDimensions: [3, 3], strides, layout: 'nhwc'};

facial_landmark_detection/face_landmark_nchw.js

@@ -79,12 +79,14 @@ export class FaceLandmarkNchw {
     };
     const input = this.builder_.input('input', inputDesc);
     inputDesc.usage = MLTensorUsage.WRITE;
+    inputDesc.writable = true;
     this.inputTensor_ = await this.context_.createTensor(inputDesc);
     this.outputTensor_ = await this.context_.createTensor({
       dataType: 'float32',
       dimensions: this.outputShape_,
       shape: this.outputShape_,
       usage: MLTensorUsage.READ,
+      readable: true,
     });
 
     const poolOptions =

facial_landmark_detection/face_landmark_nhwc.js

@@ -80,12 +80,14 @@ export class FaceLandmarkNhwc {
     };
     const input = this.builder_.input('input', inputDesc);
     inputDesc.usage = MLTensorUsage.WRITE;
+    inputDesc.writable = true;
     this.inputTensor_ = await this.context_.createTensor(inputDesc);
     this.outputTensor_ = await this.context_.createTensor({
       dataType: 'float32',
       dimensions: this.outputShape_,
       shape: this.outputShape_,
       usage: MLTensorUsage.READ,
+      readable: true,
     });
 
     const poolOptions =

facial_landmark_detection/ssd_mobilenetv2_face_nchw.js

@@ -122,13 +122,15 @@ ${nameArray[1]}`;
     };
     const input = this.builder_.input('input', inputDesc);
     inputDesc.usage = MLTensorUsage.WRITE;
+    inputDesc.writable = true;
     this.inputTensor_ = await this.context_.createTensor(inputDesc);
     for (const [key, value] of Object.entries(this.outputsInfo)) {
       this.outputTensors_[key] = await this.context_.createTensor({
         dataType: 'float32',
         dimensions: value,
         shape: value,
         usage: MLTensorUsage.READ,
+        readable: true,
       });
     }
 

facial_landmark_detection/ssd_mobilenetv2_face_nhwc.js

@@ -134,13 +134,15 @@ ${nameArray[1]}`;
     };
     const input = this.builder_.input('input', inputDesc);
     inputDesc.usage = MLTensorUsage.WRITE;
+    inputDesc.writable = true;
     this.inputTensor_ = await this.context_.createTensor(inputDesc);
     for (const [key, value] of Object.entries(this.outputsInfo)) {
       this.outputTensors_[key] = await this.context_.createTensor({
         dataType: 'float32',
         dimensions: value,
         shape: value,
         usage: MLTensorUsage.READ,
+        readable: true,
       });
     }
 

image_classification/efficientnet_fp16_nchw.js

@@ -84,12 +84,14 @@ export class EfficientNetFP16Nchw {
     };
     let data = this.builder_.input('input', inputDesc);
     inputDesc.usage = MLTensorUsage.WRITE;
+    inputDesc.writable = true;
     this.inputTensor_ = await this.context_.createTensor(inputDesc);
     this.outputTensor_ = await this.context_.createTensor({
       dataType: 'float32',
       dimensions: this.outputShape_,
       shape: this.outputShape_,
       usage: MLTensorUsage.READ,
+      readable: true,
     });
     data = this.builder_.cast(data, 'float16');
     // Block 0

image_classification/mobilenet_nchw.js

@@ -98,12 +98,14 @@ export class MobileNetV2Nchw {
     };
     let data = this.builder_.input('input', inputDesc);
     inputDesc.usage = MLTensorUsage.WRITE;
+    inputDesc.writable = true;
     this.inputTensor_ = await this.context_.createTensor(inputDesc);
     this.outputTensor_ = await this.context_.createTensor({
       dataType: 'float32',
       dimensions: this.outputShape_,
       shape: this.outputShape_,
       usage: MLTensorUsage.READ,
+      readable: true,
     });
     if (this.dataType_ === 'float16') {
       data = this.builder_.cast(data, 'float16');

image_classification/mobilenet_nhwc.js

@@ -96,12 +96,14 @@ export class MobileNetV2Nhwc {
     };
     const input = this.builder_.input('input', inputDesc);
     inputDesc.usage = MLTensorUsage.WRITE;
+    inputDesc.writable = true;
     this.inputTensor_ = await this.context_.createTensor(inputDesc);
     this.outputTensor_ = await this.context_.createTensor({
       dataType: 'float32',
       dimensions: this.outputShape_,
       shape: this.outputShape_,
       usage: MLTensorUsage.READ,
+      readable: true,
     });
     const conv0 = this.buildConv_(
         input, '90', 'Conv_Conv2D', true, {strides, autoPad, filterLayout});

image_classification/resnet50v1_fp16_nchw.js

@@ -85,12 +85,14 @@ export class ResNet50V1FP16Nchw {
     };
     let data = this.builder_.input('input', inputDesc);
     inputDesc.usage = MLTensorUsage.WRITE;
+    inputDesc.writable = true;
     this.inputTensor_ = await this.context_.createTensor(inputDesc);
     this.outputTensor_ = await this.context_.createTensor({
       dataType: 'float32',
       dimensions: this.outputShape_,
       shape: this.outputShape_,
       usage: MLTensorUsage.READ,
+      readable: true,
     });
     data = this.builder_.cast(data, 'float16');
     const conv1 = await this.buildConv_(

image_classification/resnet50v2_nchw.js

@@ -107,12 +107,14 @@ export class ResNet50V2Nchw {
     };
     const data = this.builder_.input('input', inputDesc);
     inputDesc.usage = MLTensorUsage.WRITE;
+    inputDesc.writable = true;
     this.inputTensor_ = await this.context_.createTensor(inputDesc);
     this.outputTensor_ = await this.context_.createTensor({
       dataType: 'float32',
       dimensions: this.outputShape_,
       shape: this.outputShape_,
       usage: MLTensorUsage.READ,
+      readable: true,
     });
     const bn1 = this.buildBatchNorm_(data, '0', '', false);
     const conv0 = this.buildConv_(

image_classification/resnet50v2_nhwc.js

@@ -129,12 +129,14 @@ export class ResNet50V2Nhwc {
     };
     const input = this.builder_.input('input', inputDesc);
     inputDesc.usage = MLTensorUsage.WRITE;
+    inputDesc.writable = true;
     this.inputTensor_ = await this.context_.createTensor(inputDesc);
     this.outputTensor_ = await this.context_.createTensor({
       dataType: 'float32',
       dimensions: this.outputShape_,
       shape: this.outputShape_,
       usage: MLTensorUsage.READ,
+      readable: true,
     });
     const conv1 = await this.buildConv_(
         input, ['', '', '1'], {strides, padding: [3, 3, 3, 3]}, false);

image_classification/squeezenet_nchw.js

@@ -52,12 +52,14 @@ export class SqueezeNetNchw {
     };
     const data = this.builder_.input('input', inputDesc);
     inputDesc.usage = MLTensorUsage.WRITE;
+    inputDesc.writable = true;
     this.inputTensor_ = await this.context_.createTensor(inputDesc);
     this.outputTensor_ = await this.context_.createTensor({
       dataType: 'float32',
       dimensions: this.outputShape_,
       shape: this.outputShape_,
       usage: MLTensorUsage.READ,
+      readable: true,
     });
     const conv0 = this.buildConv_(data, 'conv0', {strides: [2, 2]});
     const pool0 = this.builder_.maxPool2d(

image_classification/squeezenet_nhwc.js

@@ -63,12 +63,14 @@ export class SqueezeNetNhwc {
     };
     const placeholder = this.builder_.input('input', inputDesc);
     inputDesc.usage = MLTensorUsage.WRITE;
+    inputDesc.writable = true;
     this.inputTensor_ = await this.context_.createTensor(inputDesc);
     this.outputTensor_ = await this.context_.createTensor({
       dataType: 'float32',
       dimensions: this.outputShape_,
       shape: this.outputShape_,
       usage: MLTensorUsage.READ,
+      readable: true,
     });
     const conv1 = this.buildConv_(
         placeholder, 'conv1', {strides, autoPad: 'same-upper'});

lenet/lenet.js

@@ -35,12 +35,14 @@ export class LeNet {
     };
     let input = this.builder_.input('input', inputDesc);
     inputDesc.usage = MLTensorUsage.WRITE;
+    inputDesc.writable = true;
     this.inputTensor_ = await this.context_.createTensor(inputDesc);
     this.outputTensor_ = await this.context_.createTensor({
       dataType: 'float32',
       dimensions: this.outputShape_,
       shape: this.outputShape_,
       usage: MLTensorUsage.READ,
+      readable: true,
     });
     if (this.layout_ === 'nhwc') {
       input = this.builder_.transpose(

nnotepad/js/nnotepad.js

@@ -48,6 +48,7 @@ class WebNNUtil {
       dimensions: operand.shape(),
       shape: operand.shape(),
       usage: MLTensorUsage.READ,
+      readable: true,
     };
     const tensor = await context.createTensor(desc);
     return tensor;

nsnet2/nsnet2.js

@@ -47,6 +47,7 @@ export class NSNet2 {
     const input = this.builder_.input('input', inputDesc);
 
     inputDesc.usage = MLTensorUsage.WRITE;
+    inputDesc.writable = true;
     this.inputTensor_ = await this.context_.createTensor(inputDesc);
 
     const relu20 = this.builder_.relu(this.builder_.add(this.builder_.matmul(input, weight172), biasFcIn0));
@@ -63,6 +64,7 @@ export class NSNet2 {
     const initialState155 = this.builder_.input('initialState155', initialStateDesc);
 
     initialStateDesc.usage = MLTensorUsage.WRITE;
+    initialStateDesc.writable = true;
     this.initialState92Tensor_ = await this.context_.createTensor(initialStateDesc);
     this.initialState155Tensor_ = await this.context_.createTensor(initialStateDesc);
 
@@ -71,13 +73,15 @@ export class NSNet2 {
       dimensions: inputShape,
       shape: inputShape, // Same as inputShape.
       usage: MLTensorUsage.READ,
+      readable: true,
     });
     const gruOutputShape = [1, batchSize, this.hiddenSize];
     const gruOutputDesc = {
       dataType: 'float32',
       dimensions: gruOutputShape,
       shape: gruOutputShape,
       usage: MLTensorUsage.READ,
+      readable: true,
     };
     this.gru94Tensor_ = await this.context_.createTensor(gruOutputDesc);
     this.gru157Tensor_ = await this.context_.createTensor(gruOutputDesc);

object_detection/ssd_mobilenetv1_nchw.js

@@ -91,18 +91,21 @@ ${nameArray[1]}_BatchNorm_batchnorm`;
     };
     let input = this.builder_.input('input', inputDesc);
     inputDesc.usage = MLTensorUsage.WRITE;
+    inputDesc.writable = true;
     this.inputTensor_ = await this.context_.createTensor(inputDesc);
     this.boxesTensor_ = await this.context_.createTensor({
       dataType: 'float32',
       dimensions: this.boxesShape_,
       shape: this.boxesShape_,
       usage: MLTensorUsage.READ,
+      readable: true,
     });
     this.scoresTensor_ = await this.context_.createTensor({
       dataType: 'float32',
       dimensions: this.scoresShape_,
       shape: this.scoresShape_,
       usage: MLTensorUsage.READ,
+      readable: true,
     });
 
     if (this.targetDataType_ === 'float16') {

object_detection/ssd_mobilenetv1_nhwc.js

@@ -97,18 +97,21 @@ ${nameArray[1]}_BatchNorm_batchnorm`;
     };
     const input = this.builder_.input('input', inputDesc);
     inputDesc.usage = MLTensorUsage.WRITE;
+    inputDesc.writable = true;
     this.inputTensor_ = await this.context_.createTensor(inputDesc);
     this.boxesTensor_ = await this.context_.createTensor({
       dataType: 'float32',
       dimensions: this.boxesShape_,
       shape: this.boxesShape_,
       usage: MLTensorUsage.READ,
+      readable: true,
     });
     this.scoresTensor_ = await this.context_.createTensor({
       dataType: 'float32',
       dimensions: this.scoresShape_,
       shape: this.scoresShape_,
       usage: MLTensorUsage.READ,
+      readable: true,
     });
 
     const strides = [2, 2];

object_detection/tiny_yolov2_nchw.js

@@ -70,12 +70,14 @@ export class TinyYoloV2Nchw {
     };
     let image = this.builder_.input('input', inputDesc);
     inputDesc.usage = MLTensorUsage.WRITE;
+    inputDesc.writable = true;
     this.inputTensor_ = await this.context_.createTensor(inputDesc);
     this.outputTensor_ = await this.context_.createTensor({
       dataType: 'float32',
       dimensions: this.outputShape_,
       shape: this.outputShape_,
       usage: MLTensorUsage.READ,
+      readable: true,
     });
 
     let mulScale = this.builder_.constant(

object_detection/tiny_yolov2_nhwc.js

@@ -64,12 +64,14 @@ export class TinyYoloV2Nhwc {
     };
     const input = this.builder_.input('input', inputDesc);
     inputDesc.usage = MLTensorUsage.WRITE;
+    inputDesc.writable = true;
     this.inputTensor_ = await this.context_.createTensor(inputDesc);
     this.outputTensor_ = await this.context_.createTensor({
       dataType: 'float32',
       dimensions: this.outputShape_,
       shape: this.outputShape_,
       usage: MLTensorUsage.READ,
+      readable: true,
     });
 
     const poolOptions = {