Cherry pick PRs to release branch rel-1.3.0#3911
Cherry pick PRs to release branch rel-1.3.0#3911
Conversation
* Merged PR 4616739: Update QLinear Ops fix 1D support layout
Update QLinear Ops fix 1D support layout
Related work items: #26011523
* Merged PR 4617257: Gather operator DML EP fails with scalar indices and 1D inputs
Fix gather with scalar value.
The ONNX conformance test case is in another PR:
// 0D, axis 1, rank 0 indices tensor
{
"op_type": "Gather",
"axis": 0,
"data": [1,2,3],
"indices": 0,
"output": 1,
"T": "float32"
}
* Merged PR 4632178: Re-enable ORT onnx_test_runner test case (DirectML ConvTranspose validation needs to be loosened to comply with ONNX definition of output_padding)
Re-enable 1D convolution tests.
Related work items: #23499747
* Merged PR 4656672: Make DML EP use Direct queue
While a Compute queue has benefits, Direct is consistent with Winml.
Related work items: #26324112
* Update DML nuget version
* Merged PR 4662079: Update DmlDev branch again from github master
Include Sheil's changes to fix namespace and header file include paths. Without this, the ONNX conformance tests all fail with E_NOTIMPL.
* Increment DML nuget version
Co-authored-by: Nick Feeney <nickfe@microsoft.com>
Co-authored-by: Dwayne Robinson <dwayner@microsoft.com>
…3884) * Load provider DLL relative to core runtime executable * Use LoadLibraryEx to fix dependent DLL loading * Fix custom op DLL loading path issue.
… runtime (#3904) * Update dynamic shape range when input shape changed during runtime * Update tensorrt_execution_provider.cc * Update tensorrt_execution_provider.cc
* Added aarch64 build pipeline * Fix build error * Remove auditwheel repair which doesn't work with cross compiling * Statically link C++ * Added auditwheel repair back and fix stdlib.h * Remove extra space
* Use MlasSoftmax in ML ops * Refine when mlas is used based on perf testing.
* avoid unneeded buffer zeroing; includes cleanup * fix includes
* Fix C# log APIs. Fixes github issue #3409. * Fix build error due to accidental duplication of GraphOptimizationLevel * Fix runoptions * Fix broken test. Add --blame switch to dotnet test cmd line to print the failed test in case of crash.
* support more types for MurmurhashOp * resolve comments * fix comments
* Update to ONNX submodule to ONNX 1.7 release branch. * Update to ONNX submodule to ONNX 1.7 release branch. * fix version.
* Re-enable disabled Dropout(12) inferencing tests. * Re-enable disabled Dropout(12) inferencing tests. * Disable Dropout(12) tests that return boolean value on NGraph EP.
* Do not register Dropout(12) as training ONLY kernel. * Move Dropout forward implementation in inference project. * fix inference build test failures. * remove fp16 test since its support is absent on CPU. * build break.
|
Can you add this as well please? #3906 |
|
Were any manual edits required or are they all clean merges of the cherry picked commits? |
@skottmckay I’m hoping only git cherry-pick commit-id was used. I plan on reviewing this PR this morning to ensure my changes are cherry picked correctly, however if @stevenlix has made any manual edits then it would be good to know them. |
|
@stevenlix Why was #3888 manually broken into two part? I see you have created an “updated onnx” commit out of it and then the rest is in the original commit. |
|
@codemzs, @liuziyue, @pranavsharma, @wangyems, @skottmckay, @prabhat00155, @stevenlix, @jywu-msft, @RyanUnderhill, @tracysh, could you please help double check if your PR is picked up correctly? |
codemzs
left a comment
codemzs
left a comment
There was a problem hiding this comment.
@stevenlix Can you please revert "update ONNX" e01ee67 commit that you have manually created? (see below snapshot) because it should already be the part of the my commit that you are cherry-picking "Update ONNX submodule to ONNX 1.7 release branch" #3888
|
We recreated the release branch from master because there are still many PRs coming in for cherry pick. So this PR is closed. |
8 participants
Cherry picked PRs
Register Dropout(12) as inference op, #3859
Re-enable Dropout(12) ONNX tests, #3833
Bump ONNX version, #3888
Huggingface optimization, #3850
MLAS update, #3846
More types for murmurhash, #3827
Fix C# log APIs, #3840
Teams team model, #3868
Use MlasComputeSoftmax in traditional ML ops, #3892
Added aarch64 build pipeline, #3841
Update TensorRT dynamic shape profile when input shape changed during runtime, #3904
fix dnnl ep build break on Linux, #3874
fix dnnl ep shared lib python packaging, #3875
Load provider shared libraries relative to core runtime executable, #3884
use LOAD_WITH_ALTERED_SEARCH_PATH for LoadLibraryExA, #3908
Merge DML Execution Provider updates, #3885
MLAS: tune softmax kernels for partial vectors, #3906