Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Cherry-pick 7 changes into 1.14.1 #14762

Merged
merged 8 commits into from
Feb 23, 2023
Merged

Cherry-pick 7 changes into 1.14.1 #14762

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
8 commits
Select commit Hold shift + click to select a range
c945774
Enable Opset11 Sequence Ops on DirectML, and make the CPU implementat…
smk2007 Feb 22, 2023
c49b36b
Some kernel changes for TULR (#14517)
wangyems Feb 7, 2023
a0ab9c2
Add extra include to fix build w/ CUDA 12 (#14659)
mc-nv Feb 12, 2023
999747d
fix error due to () not used on operator priority. (#14699)
zhanghuanrong Feb 16, 2023
71e7e35
Create new stream for data copy for IOBidning input scenario (#14719)
jslhcl Feb 20, 2023
996933b
Disable fused causal attention (#14732)
tianleiwu Feb 21, 2023
bed52f5
[CUDA] Fix FP16 Precision for Sigmoid Op (#14727)
centwang Feb 22, 2023
419e514
Cjian/trying to fix cherry pick on mac os ci (#14792)
jchen351 Feb 23, 2023
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
4 changes: 4 additions & 0 deletions cmake/onnxruntime_rocm_hipify.cmake
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -19,6 +19,10 @@ set(contrib_ops_excluded_files
"bert/fast_gelu_impl.h"
"bert/fast_gelu.cc"
"bert/fast_gelu.h"
"bert/relative_attn_bias.cc"
"bert/relative_attn_bias.h"
"bert/relative_attn_bias_impl.cu"
"bert/relative_attn_bias_impl.h"
"bert/skip_layer_norm.cc"
"bert/skip_layer_norm.h"
"bert/skip_layer_norm_impl.cu"
Expand Down
63 changes: 59 additions & 4 deletions docs/ContribOperators.md
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -30,6 +30,7 @@ Do not modify directly.*
* <a href="#com.microsoft.FusedConv">com.microsoft.FusedConv</a>
* <a href="#com.microsoft.FusedGemm">com.microsoft.FusedGemm</a>
* <a href="#com.microsoft.FusedMatMul">com.microsoft.FusedMatMul</a>
* <a href="#com.microsoft.GatedRelativePositionBias">com.microsoft.GatedRelativePositionBias</a>
* <a href="#com.microsoft.GatherND">com.microsoft.GatherND</a>
* <a href="#com.microsoft.Gelu">com.microsoft.Gelu</a>
* <a href="#com.microsoft.GemmFastGelu">com.microsoft.GemmFastGelu</a>
Expand Down Expand Up @@ -152,7 +153,7 @@ This version of the operator has been available since version 1 of the 'com.micr
<dd>Attention mask with shape (batch_size, 1, max_sequence_length, max_sequence_length), (batch_size, total_sequence_length) or (batch_size, sequence_length, total_sequence_length), or index with shape (batch_size) or (2 * batch_size)</dd>
<dt><tt>past</tt> (optional) : T</dt>
<dd>past state for key and value with shape (2, batch_size, num_heads, past_sequence_length, head_size)When past_present_share_buffer is set, its shape is (2, batch_size, num_heads, max_sequence_length, head_size)</dd>
<dt><tt>extra_add</tt> (optional) : T</dt>
<dt><tt>relative_position_bias</tt> (optional) : T</dt>
<dd>additional add to QxK' with shape (batch_size, num_heads, sequence_length, total_sequence_length)</dd>
<dt><tt>past_sequence_length</tt> (optional) : M</dt>
<dd>When past_present_share_buffer is used, it is required to specify past_sequence_length (could be 0).</dd>
Expand Down Expand Up @@ -1608,6 +1609,58 @@ This version of the operator has been available since version 1 of the 'com.micr
</dl>


### <a name="com.microsoft.GatedRelativePositionBias"></a><a name="com.microsoft.gatedrelativepositionbias">**com.microsoft.GatedRelativePositionBias**</a>

query_layer = (query_layer + query_bias).reshape(batch_size, seq_len, num_heads, head_size).transpose(1, 2)
gate_u, gate_r = torch.sigmoid(
self.gate_ur_linear(query_layer).view(batch_size, num_head, seq_len, 2, D/2).sum(-1, keepdim=False)
).chunk(2, dim=-1)
gate_u_1 = gate_u * (gate_r * self.eco_a - 1.0) + 2.0
rel_pos_bias = gate_u_1 * rel_pos

#### Version

This version of the operator has been available since version 1 of the 'com.microsoft' operator set.

#### Attributes

<dl>
<dt><tt>num_heads</tt> : int (required)</dt>
<dd>Number of attention heads</dd>
</dl>

#### Inputs

<dl>
<dt><tt>query_layer</tt> : T</dt>
<dd>tensor with shape (batch_size, seq_len, num_heads x head_size)</dd>
<dt><tt>query_bias</tt> : T</dt>
<dd>1-d tensor with shape (num_heads x head_size)</dd>
<dt><tt>rel_pos</tt> : T</dt>
<dd>tensor with shape (1, num_head, seq_len, seq_len)</dd>
<dt><tt>weight</tt> : T</dt>
<dd>gemm weight for the gated_ur_linear, shape (head_size, D), D is divisible by 2</dd>
<dt><tt>bias</tt> : T</dt>
<dd>bias for the gated_ur_linear, shape (D)</dd>
<dt><tt>eco_a</tt> : T</dt>
<dd>tensor of shape (1, num_heads, 1, 1)</dd>
</dl>

#### Outputs

<dl>
<dt><tt>output</tt> : T</dt>
<dd>output tensor with shape (batch_size, num_heads, seq_len, seq_len)</dd>
</dl>

#### Type Constraints

<dl>
<dt><tt>T</tt> : tensor(float), tensor(float16)</dt>
<dd>Constrain input and output types to float tensors.</dd>
</dl>


### <a name="com.microsoft.GatherND"></a><a name="com.microsoft.gathernd">**com.microsoft.GatherND**</a>

Given `data` tensor of rank r >= 1, and `indices` tensor of rank q >= 1, gather
Expand Down Expand Up @@ -2222,7 +2275,7 @@ This version of the operator has been available since version 1 of the 'com.micr
<dd>Number of attention heads</dd>
</dl>

#### Inputs (2 - 5)
#### Inputs (2 - 6)

<dl>
<dt><tt>query</tt> : T</dt>
Expand All @@ -2235,6 +2288,8 @@ This version of the operator has been available since version 1 of the 'com.micr
<dd>Bias tensor with shape (hidden_size + hidden_size + v_hidden_size) from input projection</dd>
<dt><tt>key_padding_mask</tt> (optional) : M</dt>
<dd>Key padding mask with shape (batch_size) or (batch_size, kv_sequence_length)</dd>
<dt><tt>relative_position_bias</tt> (optional) : T</dt>
<dd>relative position bias: addition to QxK' with shape (batch_size, num_heads, sequence_length, total_sequence_length) or (1, num_heads, sequence_length, total_sequence_length)</dd>
</dl>

#### Outputs
Expand Down Expand Up @@ -3221,7 +3276,7 @@ This version of the operator has been available since version 1 of the 'com.micr
left-side padding, mask_index has shape (2 * batch_size), where the values are the exclusive end positions followed by
the inclusive start positions. When unidirectional is 1, and each token only attend to previous tokens. For GPT-2, both past
and present state are optional. Present state could appear in output even when past state is not in input.
Current version does not support past/present, extra_add and qkv_hidden_sizes.
Current version does not support past/present, relative_position_bias and qkv_hidden_sizes.
TODO: Support them if needed in the future.

#### Version
Expand Down Expand Up @@ -3286,7 +3341,7 @@ This version of the operator has been available since version 1 of the 'com.micr
<dd>Attention mask with shape (batch_size, 1, max_sequence_length, max_sequence_length), (batch_size, past_sequence_length + sequence_length)or (batch_size, sequence_length, past_sequence_length + sequence_length), or index with shape (batch_size) or (2 * batch_size).</dd>
<dt><tt>past</tt> (optional) : Q</dt>
<dd>past state for key and value with shape (2, batch_size, num_heads, past_sequence_length, head_size).</dd>
<dt><tt>extra_add</tt> (optional) : S</dt>
<dt><tt>relative_position_bias</tt> (optional) : S</dt>
<dd>additional add to QxK' with shape (batch_size, num_heads, sequence_length, sequence_length).</dd>
</dl>

Expand Down
Loading