* Change CUDA implementation of Transpose to not use a typed kernel so we can support more types with minimum binary size.
Add support for 8, 16, 32 and 64 bit types.
Add unit tests.
Add method so the implementation can be called directly (will be used by CUDA Scan very soon).
* Disable TensorRT for MLFloat16 and int8 unit tests.
* Address PR comment and add support for calling cublas implementation if type is mlfloat16.
Fuse transpose into MatMul
Implement Pow and constant scalar simplification
Vectorize ReduceMean
Improve symbolic shape inference
Minor updates for better debugging in fused function name
* Direct use python numpy array's memory if already contiguous. This
could greatly improve performance for session with large input,
like big image 1920x1080 fastrcnn, 30~40% speed up could be achieved.
* Add test case enforce contiguous/non-contiguos numpy array as inputs.
* onnxrt server with OVEP
* onnxrt server with OVEP
* Update Dockerfile.server.openvino
* onnxrt server OVEP fix reviews
* onnxrt server OVEP fix reviews
* Update the CUDA Where implementation broadcasting logic to handle a dim with value of 0.
Add unit test
Also add unit test for unary op with dim value of 0
* Exclude ngraph from Where test with 0 dim.
1. refactor the pipeline, remove some duplicated code
2. Move Windows_py_GPU_Wheels job to Win-GPU-CUDA10. We'll deprecated the "Win-GPU" pool
3. Delete cpu-nocontribops-esrp-pipeline.yml and cpu-nocontribops-pipeline.yml
4. In Linux nuget jobs, run "make install" before creating the package. So that extra RPAH info will be removed
* Remove unused initializer from GraphProto as well as name_to_initial_tensor_ in CleanupUnusedInitializers.
This means initializers that have been replaced during graph optimizations are not left in the GraphProto when we save an optimized model.
* Handle edge case where a model has an unused initializer with matching graph input by also removing the graph input.
* Use non-const iterators in std::find_if calls to make centos build happy.
Description: Describe your changes.
Make elementwise op run 4 items per thread
unroll for loop to leverage ILP
remove unnessary N==0 check inside elementwise GPU kernel
Motivation and Context
Why is this change required? What problem does it solve?
It can improve the performance of GPU elementwise ops. ~2% performance gain on popular NLP bert model.
If it fixes an open issue, please link to the issue here.
thread_local/global/static destruction order depends on implementation details of compilers and OS. The bug happens when thread_local is already out of scope while static EP being destructed, thus causing access violation in EP's destructor when accessing thread_local.
The fix is to maintain ownership inside EP with a mapping from tid to ThreadLocalContext, to avoid accessing thread_local in EP's destructor. This way, no matter what the destruction order is, no access violation would be triggered.
* Add logic to try and flatten inner dimensions being copied by Slice and do a block copy if they can be.
Do a block copy for just the inner most dimension where possible (applies even if we don't flatten inner dimensions).
