[Native WebGPU] Add Conv, ConTranspose and FusedConv #24186
Added Conv
de80fadf
lint
4490299b
Added Conv1D
598a7604
Added grouped conv
27e8bf87
Fix shader code.
856f4d37
Removed fused_utils.cc and fused_utils.h
c6e2d1a0
Tmp
27a03df0
tmp2
3c713de7
Handle conv1d output
fd994bc4
Modified CreateMatMulProgram signature.
0c9273bb
Fix shader code.
c34324db
Fix compilation errors.
0ffc9925
Added FusedConv and Activation.
24b40362
Removed DoTranspose, Added shader i32 casts, and input_output_shapes …
171b1013
Fix input_output_shapes usage.
f9b5ae9a
Add FusedConv
a946098b
Enabled WebGpuEP FusedConv tests.
df8f1387
Layout transformation should skip FusedConv for WebGpuEp
234f719b
Revert "Layout transformation should skip FusedConv for WebGpuEp"
02c16ee1
Added sequentially_access_by_threads and transpose_a flags to matmul …
94c6f444
typo
7b9897c1
Added reduced shapes when creating Conv2DMMProgram.
298467fd
Added FusedConv related changes.
8ab59310
fixed merge error
4fa8223c
Added ConvTranspose
2297bc16
More changes to fix issues.
d7a5fed5
Fixed data type in the shader code.
61ab3d38
fix compilation error
3db88f5a
Fix activation
55bf4c3b
Make ConvTranspose test run.
b559ec95
Removed code dividing offset by the components in the shader code.
a82ee44d
Fix unused variable error.
087b407c
lint
b839a7bb
Revert "Fix unused variable error."
ccd43c38
Added activation_snippet
b919e2f7
Added channels last support in Conv and MatMul
bd5322f0
Added ReduceShapeByComponents
0962d299
Added permutation parameter to TransposeKernel function.
da78cca3
Fixed the ordered of AddInput calls.
78f9eb29
typo
5ad45559
Calculate outputIndices from global_idx without multiplying with comp…
9ddbc7d4
convert activation to string
62fade7b
Removed unused variables/members, fixed conv-transpose1d
6fa622d1
Fix ConvTranspose asymmetric pads
3520e256
Fix conv-transpoes padding
ca810e87
Removed ComputeOutputShape defined in Conv, and using functionality a…
65b2e62d
Disabled Conv/ConvTranspose 3D test case and empty input test cases.
27a18e10
Merge branch 'main' of https://github.com/microsoft/onnxruntime into …
eadf892a
Enabled ConvTranspoes_onnx_group testcase.
c1b1af67
Added activation hint to Conv/MatMul.
5627190d
satyajandhyala
marked this pull request as ready for review 297 days ago
Fixed compilation issues.
bf2800e6
lint
de4f84c7
Fixed activation code to handle vector data types
143f2ed3
lint
bc5aed22
Use x_element_t instead of x_value_t
55874c2d
Reverted unintended changes.
4da5aa2c
typo
83296206
typo
c43d1832
Removed default arguments causing f32 vs. f16 related shader code err…
8ee6129f
Removed unused local variable.
08856bf6
Use explicit scope when accessing the template base class member.
36aeffaf
Fixed sigmoid activation, fixed matmul to apply activation after addi…
657ec018
Refactored and fixed issue with calling malmul from conv in one scena…
0d78a8df
lint
d5d45442
clean-up
900d0007
Add missing bias input
705e8bc6
clean-up2
5870ed46
More indexing/components related changes/fixes
4970c7ef
Added stringify
6be25e44
Fixed typos and improved ReduceShapeByComponents functionality to div…
78af3177
Merge branch 'main' of https://github.com/microsoft/onnxruntime into …
1434be4c
Fixed ConvTranspose typed test.
5eb067f7
Enabled some activation fusion test cases.
f51a2046
satyajandhyala
changed the title [WIP][Native WebGPU] Add Conv, ConTranspose and FusedConv [Native WebGPU] Add Conv, ConTranspose and FusedConv 292 days ago
fs-eire
approved these changes
on 2025-04-04
satyajandhyala
deleted the sajandhy/webgpu-ep-add-conv branch 291 days ago
Assignees
No one assigned
Login to write a write a comment.
Login via GitHub