In current implementation, axis in softmax has to be the last, which is
an obvious limitation. This PR removes this limitation and will fix
issues #20710 and #22176.
### Description
<!-- Describe your changes. -->
#21618
This PR optimizes grouped conv by 1) more sequential memory access in
gpu 2) reusing input's data to reduce global memory access times.
See `Conv|GroupedConv` op in
[Wav2Vec2](https://huggingface.co/facebook/wav2vec2-base-960h) becomes
92 ms from 1058 ms on iGPUs with 32 EU.
For the whole model on my iGPUs with 32 EU,
wav2vec2 model becomes 982ms from 1942 ms.
squeezebert-uncased model becomes 71.86ms from 431.77ms.
### Motivation and Context
<!-- - Why is this change required? What problem does it solve?
- If it fixes an open issue, please link to the issue here. -->
### Description
<!-- Describe your changes. -->
Fix bugs in previous implementation and add more situations to go the
optimized path.
Below situations will go to the optimized path.
1. 2d inputs or squeezed 2d inputs
2. channels last or channels first transpose. For example, channel last
transpose: [1, 256, 512, 512] -> [1, 512, 512, 256]
For this case, the transpose becomes [256, 512x512] -> [512x512, 256]
### Motivation and Context
<!-- - Why is this change required? What problem does it solve?
- If it fixes an open issue, please link to the issue here. -->
For SD Turbo demo, the total transpose time becomes 39.98ms from
122.09ms. And the correspnding percents becomes 3.89% from 11.05% in
this demo.
This PR will also help #21618, the total transpose time in that demo
becomes 17.32 ms from 70.25 ms on my iGPUs.
### Description
<!-- Describe your changes. -->
### Motivation and Context
<!-- - Why is this change required? What problem does it solve?
- If it fixes an open issue, please link to the issue here. -->
### Description
<!-- Describe your changes. -->
### Motivation and Context
<!-- - Why is this change required? What problem does it solve?
- If it fixes an open issue, please link to the issue here. -->
---------
Co-authored-by: Yulong Wang <7679871+fs-eire@users.noreply.github.com>
Avoid producing presentKey/presentValue outputs if pastKey/pastValue
don't exists.
### Description
<!-- Describe your changes. -->
### Motivation and Context
<!-- - Why is this change required? What problem does it solve?
- If it fixes an open issue, please link to the issue here. -->
### Description
Previously, MultiHeadAttention supports relative position bias of shape
[1, N, S, T] or [B, N, S, T], and DecoderMaskedMultiHeadAttention
supports [1, N, S, T]. This will extend the support to allow [1, N, S,
T], [B, N, S, T], [B, 1, S, T] and [1, 1, S, T] for CUDA and CPU EPs.
- [x] Rename the input of "relative position bias" to "attention bias"
because it can also be used for other types of bias, like ALiBi
(Attention with Linear Biases) or attention mask.
- [x] Update unfused kernel to support broadcasting 2nd dimension of
attention bias.
- [x] Update efficient attention to support broadcasting 2nd dimension
of attention bias.
- [x] Update operators (MultiHeadAttention,
DecoderMaskedMultiHeadAttention, Attention, PackedAttention,
PackedMultiHeadAttention) to support broadcast attention bias on CUDA
and CPU EPs.
- [x] Update ROCm, DML and WebGPU naming to be consistent. (Note that
those EPs do not support broadcasting attention_bias for now).
- [x] Add attention bias tests for MultiHeadAttention.
- [x] Update operator documents
- [x] Update benchmark script
Other changes:
* Fix some checks in multihead-attention.ts
* Add helper functions to dump tensors given dimensions.
Bug: https://github.com/microsoft/onnxruntime/issues/21386
### Description
<!-- Describe your changes. -->
### Motivation and Context
<!-- - Why is this change required? What problem does it solve?
- If it fixes an open issue, please link to the issue here. -->
### Description
Added DequantizeLinear operator for JSEP.
### Motivation and Context
<!-- - Why is this change required? What problem does it solve?
- If it fixes an open issue, please link to the issue here. -->
### Description
allow op test to use f16 type for inputs/outputs.
This PR introduces "@petamoriken/float16" as Float16Array polyfill but
restricts it to be only used for test runner.
Bug:https://github.com/microsoft/onnxruntime/issues/21467
### Description
<!-- Describe your changes. -->
### Motivation and Context
<!-- - Why is this change required? What problem does it solve?
- If it fixes an open issue, please link to the issue here. -->
### Description
<!-- Describe your changes. -->
### Motivation and Context
<!-- - Why is this change required? What problem does it solve?
- If it fixes an open issue, please link to the issue here. -->
### Description
<!-- Describe your changes. -->
### Motivation and Context
<!-- - Why is this change required? What problem does it solve?
- If it fixes an open issue, please link to the issue here. -->
### Description
Remove explicitly concatinating pastKey with Key and pastValue with
Value.
### Motivation and Context
<!-- - Why is this change required? What problem does it solve?
- If it fixes an open issue, please link to the issue here. -->
### Description
The Key and Value inputs could be 4-dims
### Motivation and Context
<!-- - Why is this change required? What problem does it solve?
- If it fixes an open issue, please link to the issue here. -->
### Description
Fixed pastkey, key and pastvalue, value concatenation condition and
fixed index error. Added new test cases.
### Motivation and Context
<!-- - Why is this change required? What problem does it solve?
- If it fixes an open issue, please link to the issue here. -->
### Description
Enabled more usecases
### Motivation and Context
<!-- - Why is this change required? What problem does it solve?
- If it fixes an open issue, please link to the issue here. -->
### Description
fix test runner with optional input/output.
This change fixes the OP test runner (.jsonc format test) with optional
input(s) and/or output(s).
this fix reveals a problem of dealing with optional outputs:
> Take SkipSimplifiedLayerNorm as example:
>
> if in the ONNX model, the node's outputs are: [ 'output_0', '' ]
instead of [ 'output_0' ], the current implementation will fail. The
difference is, in the first case, context.outputCount == 2, and then the
typescript implementation will try to create a tensor for output[1]. It
will eventually call to C++ function (OpKernelContext::Output), and the
output.DataRaw() will be nullptr. WebGPU backend will fail because it
cannot deal with a TensorView with data == 0.
>
This problem may need to be fixed or workaround in separated PR. This PR
does not fix this problem. Failed test cases are modified to work -
please note this PR does not break those test cases as they never work.
### Description
<!-- Describe your changes. -->
Improve performance using shared memory
### Motivation and Context
<!-- - Why is this change required? What problem does it solve?
- If it fixes an open issue, please link to the issue here. -->
### Description
Avoid using vec4 Matmul implementation for ConvTranspose with channel-last
### Motivation and Context
<!-- - Why is this change required? What problem does it solve?
- If it fixes an open issue, please link to the issue here. -->
### Description
For Concat operation, the zero-size input tensor shape need to be
preserved and, unlike non-zero tensors, the dims are not constrained to
match other input tensors' dims.
### Motivation and Context
<!-- - Why is this change required? What problem does it solve?
- If it fixes an open issue, please link to the issue here. -->
### Description
Fixes build break brought by #19614
Currently WebGL backend does not support zero sized tensor. This change
split test data into 2 parts, and only enable zero sized tensor tests
for WebGPU.
### Description
This PR allows zero-sized output.
To make the implementation simple, it does not support partial
zero-sized tensor. Which means, either all outputs are zero-sized, or an
error will be reported.
added 2 tests:
- op test of `Add` with input T[2,0] T[2,1], and
- test_split_zero_size_splits
### Description
<!-- Describe your changes. -->
1. Fix Where operator to handle Boolean input less than 4 bytes.
2. Fix JSEP test harness to use tensor names consistently.
### Motivation and Context
<!-- - Why is this change required? What problem does it solve?
- If it fixes an open issue, please link to the issue here. -->
### Description
Add MatMulNBits to support MatMul using 4-bit quantized weights
### Motivation and Context
<!-- - Why is this change required? What problem does it solve?
- If it fixes an open issue, please link to the issue here. -->
### Description
This PR 1) adds LeakyRelu activation for fusedConv; 2) makes `vec4<f16>`
value work with `float32` uniforms attributes.
For example:
`clamp(value, vec4<f16>(uniforms.clip_min),
vec4<f16>(uniforms.clip_max)` will throw compilation errors since
`uniforms.clip_min` and `uniforms.clip_min` are `f32` not `f16`. So we
need to change it to `clamp(value, vec4<f16>(f16(uniforms.clip_min)),
vec4<f16>(f16(uniforms.clip_max))`
And above problem was introduced when we make activation attributes as
uniforms instead of constant.
BTW, after adding LeakyRelu, `realesrgan-t256` model can pass.
### Description
```math
\tanh(x)=\frac{e^x-e^{-x}}{e^x+e^{-x}}=
\left\{
\begin{array}{cc}
-\frac{1-e^{-2\cdot(-x)}}{1+e^{-2\cdot(-x)}}, & x<0 \\
0, & x=0 \\
\frac{1-e^{-2x}}{1+e^{-2x}}, & x>0
\end{array}
\right.
```
### Motivation and Context
On some platforms,
$$\tanh(1000)=\frac{e^{1000}-e^{-1000}}{e^{1000}+e^{-1000}}$$ would
produce NaN instead of 0.999... or 1 (imagine $e^{1000}=\infty$ and
$\frac{\infty}{\infty}$ explodes).
### Description
This PR provides a vectorized algorithm for NHWC GroupedConv to improve
performance.
The aggregate time of GroupedConv in mobilenetv2-12 becomes ~1ms from
~4ms on Intel Alder Lake machine. About 20% improvement for the whole
model.
### Description
Also update the op test suite.
### Motivation and Context
Previously the *total* size in case `Expand - last dim is not divisible
by 4` was a multiple of 4, even though the *last dimension* was not, so
the bug has never been caught.
### Description
<!-- Describe your changes. -->
### Motivation and Context
<!-- - Why is this change required? What problem does it solve?
- If it fixes an open issue, please link to the issue here. -->
### Description
Add trilinear interpolation to Resize and changed activation_params attribute as optional for FuseConv.
### Motivation and Context
<!-- - Why is this change required? What problem does it solve?
- If it fixes an open issue, please link to the issue here. -->
