pengwa 2347a0aca8 Autograd Function Fallback bug fix - moe support (#8105) ... * Support forward inputs orders like "Non_tensor/Tensor/Non_tensor". Correspondingly, support "None/Tensor_Grad/None" fpr backward outputs. * Report RuntimeError when PythonOp detected but _enable_custom_autograd_function is enabled. * Fix "PoliCheck ] - Defect : Term "hang", Component : orttraining\orttraining\python\training\ortmodule\__init__.py (1 issue)" * rename call_convention->input_convention, input_tensor_requires_grads->input_requires_grads * fix minor comment * revert polycheck fix in case of conflict * Update orttraining/orttraining/core/graph/training_op_defs.cc Co-authored-by: Tim Harris <tiharr@microsoft.com> * Apply suggestions from code review Refine the schema description Co-authored-by: Tim Harris <tiharr@microsoft.com> * Resolve review comments Co-authored-by: Tim Harris <tiharr@microsoft.com>		2021-07-07 08:58:01 +08:00
..
contrib_ops	Move kernel implementations outside of lookup table utility functions. (#8306)	2021-07-06 18:31:05 -05:00
core	Autograd Function Fallback bug fix - moe support (#8105)	2021-07-07 08:58:01 +08:00
featurizers_ops/cpu
gsl	Update C++ Standard from 14 to 17 (#8041)	2021-06-25 14:08:01 -07:00
python	fix the shared provider test for training build; expose more symbols to non cuda build (#8249)	2021-07-01 11:03:02 -07:00
test	Fix optimizer crash (#8274)	2021-07-02 17:19:15 -07:00
tool/etw
wasm	Integrate onnxruntime-extensions into onnxruntime. (#8143)	2021-07-01 09:34:03 -07:00
.style.yapf
__init__.py	bumping up version number to 1.8 (#7733)	2021-05-18 09:03:37 -07:00
ReformatSource.ps1
ReformatSourcePython.bat
VSCodeCoverage.runsettings