saymrwulf/onnxruntime
1
0
Fork 0
mirror of https://github.com/saymrwulf/onnxruntime.git synced 2026-05-31 23:27:43 +00:00
Code Issues Projects Releases Packages Wiki Activity Actions
5842 commits 2856 branches 57 tags 1.3 GiB
Commit graph

5842 commits

This branch
This branch
All branches
Author SHA1 Message Date
raviskolli
9f4e8cf6a0
Update training pipelines to pytorch 1.10 (#9709) 
* Update training pipelines to pytorch 1.10

* Fixed a typo in cuda version.

* Downgraded gcc to 8 for cuda 10.2
 2021-11-15 11:21:55 -08:00
Tang, Cheng
99257eb8e3
support build option to include external graph transformers (#9478) 
* temp code

* support external graph transformer  from build script

* remove debug code

* add test case

* support register rewrite rule

* fix source_group issue if external source is not share any common prefix

* fix python code style checker

* resolve merge conflict

Co-authored-by: Cheng Tang <chenta@microsoft.com>
 2021-11-15 08:16:20 -08:00
pengwa
6e09fc5152
Implement block wise softmax for reduction dimention > 1024 cases. (#9696) 
* implement block wise softmax for reduction dimention > 1024 cases.

* fix builds

* fix

* fix amd build

* fix amd build

* fix win-gpu build

* add tests

* remove cudnn path/add python tests
 2021-11-14 11:47:58 +08:00
Sheil Kumar
3d0bd2596f
Enable creating OrtValues from ID3D12Resources from the onnxruntime C-API (#9686) 
* Add onnxruntime-windows api.

* minor fixes

* add to package headers

* Build ort_dml_api for provider extensions.

* Cleanup

* misc comment

* remove winml specific comments

* use dml check in onnxruntime

* Update include/onnxruntime/core/providers/dml/dml_provider_factory.h

Co-authored-by: Dwayne Robinson <dwayner@microsoft.com>

* Update include/onnxruntime/core/session/onnxruntime_c_api.h

Co-authored-by: Dwayne Robinson <dwayner@microsoft.com>

* Update include/onnxruntime/core/providers/dml/dml_provider_factory.h

Co-authored-by: Dwayne Robinson <dwayner@microsoft.com>

* Update include/onnxruntime/core/providers/dml/dml_provider_factory.h

Co-authored-by: Dwayne Robinson <dwayner@microsoft.com>

* Update onnxruntime/core/session/onnxruntime_c_api.cc

Co-authored-by: Dwayne Robinson <dwayner@microsoft.com>

* Update onnxruntime/core/session/ort_apis.h

Co-authored-by: Dwayne Robinson <dwayner@microsoft.com>

* Update winml/test/adapter/AdapterSessionTest.cpp

Co-authored-by: Dwayne Robinson <dwayner@microsoft.com>

* Update onnxruntime/core/session/onnxruntime_c_api.cc

Co-authored-by: Dwayne Robinson <dwayner@microsoft.com>

* Update winml/adapter/winml_adapter_c_api.cpp

Co-authored-by: Dwayne Robinson <dwayner@microsoft.com>

* Update include/onnxruntime/core/session/onnxruntime_c_api.h

Co-authored-by: Pranav Sharma <prs@microsoft.com>

* Update onnxruntime/core/session/onnxruntime_c_api.cc

Co-authored-by: Pranav Sharma <prs@microsoft.com>

* Update winml/adapter/winml_adapter_c_api.cpp

* PR feedback

* Update include/onnxruntime/core/providers/dml/dml_provider_factory.h

Co-authored-by: Dwayne Robinson <dwayner@microsoft.com>

* Update include/onnxruntime/core/providers/dml/dml_provider_factory.h

Co-authored-by: Dwayne Robinson <dwayner@microsoft.com>

* Update include/onnxruntime/core/providers/dml/dml_provider_factory.h

Co-authored-by: Dwayne Robinson <dwayner@microsoft.com>

* PR feedback

* merge resolution and unreference param

* (naming) Remove Dml prefix

* maybe unused version

* move DML code into DML path. CIs failing because DML is not available when --use_dml is not on

* fix warning causing local build failures after merging

* Change getvaluememoryinfo to gettensormemoryinfo

* minor breaks

* fix comment paste

* fix comment

Co-authored-by: Sheil Kumar <sheilk@microsoft.com>
Co-authored-by: Dwayne Robinson <dwayner@microsoft.com>
Co-authored-by: Pranav Sharma <prs@microsoft.com>
 2021-11-13 03:34:54 -08:00
RandySheriffH
21eb747a0f
Custom thread creation and join hooks (#9426) 2021-11-12 19:10:31 -08:00
Edward Chen
9f69d8bbae
Disable partial runtime optimization implementation by default (#9748) 
* Only serialize runtime optimization records container if non-empty.

* Remove runtime optimizations from onnxruntime/core/flatbuffers/schema/README.md as it's not completely implemented yet.

* Disable partial runtime optimization implementation by default.
 2021-11-12 17:37:29 -08:00
Sheil Kumar
a17bdaf725
Enable JoinModels API in WinML+RT Experimental API (#9746) 
* Dynamic onnx model fusion

* empty node names shoudl remain empty

* comments and cleanup

* logic reversed for promoting_unlined_outputs

* PR feedback

* type

* typo

* fix model outputs with promote unlinked output

* remove disembodied model

Co-authored-by: Sheil Kumar <sheilk@microsoft.com>
 2021-11-12 16:56:31 -08:00
Guoyu Wang
5ad6dbb314
Remove experimental from ORT format namespace (#9729) 
* schema change

* cc channges

* remove temp debug code

* Adding fbs namespace to session_state_flatbuffers_utils.h

* Add fbs namepsace to all ort format utils
 2021-11-11 19:46:30 -08:00
Aidan Beggs
f6edf13513
Implement a Gemm/Sum fusion pattern (#9699) 
When the pattern Sum(Gemm(A, B), C) exists, we can convert it to
Gemm(A, B, C), assuming that C the output of the original Gemm is
not used elsewhere, and this change does not break broadcasting.
 2021-11-11 18:33:13 -08:00
Edward Chen
997266a620
Add build.py option to disable ORT format model runtime optimization (#9723) 
ORT format model runtime optimization implementation is in progress.
This change adds a build.py option to disable the partial runtime optimization implementation, adds CI builds to test it, and disables runtime optimizations in mobile package builds.
 2021-11-11 18:05:45 -08:00
Gary Miguel
93e239747f
Construct valid graphs for ONNX checker for IR version < 4. (#9665) 
* Construct valid graphs for ONNX checker for IR version < 4.

Previously the constructed graph was not guaranteed to have its
initializers be a subset of its inputs, which is required for IR
version < 4. This resulted in spurious failures.

Fixes #9663
 2021-11-12 09:13:28 +10:00
zhaoyang-intel
32c896df6d
Add DynamicQuantizeLinear to DNNL EP (#9620) 
implement dynamicquantizelinear in DNNL EP

add debug log in EP for operator coverage

block gpu elementwise op with 5 dims or more

Signed-off-by: Wang <zhaoyang.wang@intel.com>
 2021-11-11 14:01:09 -08:00
Nick Kreeger
559adbb534
Replace deprecated PyEval_CallObject() call with PyObject_CallObject(). (#9737) 
This API was changed in 3.9:
https://docs.python.org/3/whatsnew/3.9.html#changes-in-the-c-api
https://docs.python.org/3/c-api/call.html#c.PyObject_Call
 2021-11-11 15:37:55 -06:00
Scott McKay
724009289b
Fix Issue #9671 (#9691) 
* Fix #9671 by running the level 1 rewrite rules first and allowing the transpose optimizer to run multiple times to ensure it completes in level 1.

Removed unnecessary call to GenerateRuleBasedGraphTransformer as there are no level 2 rewrite rules.
 2021-11-12 07:18:16 +10:00
Yi-Hong Lyu
03f9d77e17
Fix index out of bound bug of generated Gather (#9673) 
Handle the NHWC support for quantized Resize that roi and scales are
rank 0
 2021-11-11 11:32:50 -08:00
Suffian Khan
e6f0fdd653
Strip AMD libraries bundled with Python package due to libonnxruntime_providers_rocm.so change (#9679) 
* remove AMD library depedence from libonnxruntime_providers_rocm.so

* fix flake error

* remove rocm dependency from original library as well
 2021-11-11 09:32:09 -08:00
Sherlock
acb5459268
Random CPU kernels should respect global seed (#9726) 
* Random CPU kernels should respect global seed
 2021-11-10 22:28:48 -08:00
Viswanath Boga
9d84811fb6
fixing pypi pipeline for release (#9716) 
* fixing pypi pipeline for release

* updated the script and correct python version

* updated the version correctly with script changes

* Remove 1.9.1
 2021-11-10 17:33:51 -08:00
Edward Chen
962feff7f1
Update linux-ci-pipeline.yml to run debug unit tests. (#9722) 2021-11-10 16:43:40 -08:00
Yulong Wang
31dc768e07
update ONNX Runtime Web CI to use same script for package versioning (#9698) 
* use ps script for package generating

* allow e2e test has no common pkg
 2021-11-10 12:52:34 -08:00
George Wu
1541784f6c
[python api] align api with other language bindings' treatment of explicit provider registrations. enforce use of providers param in python InferenceSession when execution providers other than default CPU are enabled. (#9712) 
* remove default python ep registration. raise exception if providers are not explicitly set if there are available providers

* temporarily disable exception

* fix python tests

* explicitly set CUDAProvider for python iobinding tests

* explicitly set providers param for InferenceSession())

* onnxrt

* raise ValueError if not explicitly set providers when creating InferenceSession

* add required providers param

* explicitly set providers

* typo
 2021-11-10 12:17:53 -08:00
Guoyu Wang
517fff0a39
Add opset 15 support for NNAPI/CoreML EPs (#9711) 2021-11-09 17:21:42 -08:00
Vincent Wang
adf98feb2c
ATenOp Support for BCEWithLogitsLoss (#9670) 2021-11-10 08:36:57 +08:00
Tiago Koji Castro Shibata
1b70a14c51
Remove usage of wstring_convert (#9251) 
* Remove usage of wstring_convert

* Use local codepage per PR feedback
 2021-11-09 13:52:57 -08:00
Chi Lo
3db2390dab
QDQ tool modification (#9616) 
* Add 1.option for enable qdq for node's output 2.force qdq appear as a pair

* modify description

* modify description

* Revert the logic of variable

* Revert the logic of variable

* Code refactor based on review's suggestions

* Update init

* Code refactor for able to specify nodes to exclude output quantization

* rename variable

* Fix bug

* code refactor

* remove the exposure of APIs

* fix bug

* fix bug

* fix bug

* fix bug

* exposure one API

Co-authored-by: Ubuntu <onnxruntime@ort-trt-ep-linux-t4.bxgbzpva45kedp3rhbsbit4phb.jx.internal.cloudapp.net>
Co-authored-by: Chi Lo <Chi.Lo@gmail.com>
 2021-11-09 12:22:12 -08:00
Sunghoon
e65f284476
[js/web] Support WebGL for ort format models in benchmarks (#9661) 
* add p50 in test

* Support FusedConv in WebGL

* resolve comments

* add a comment for longToNumber change

Co-authored-by: Yulong Wang <yulongw@microsoft.com>
 2021-11-09 11:58:47 -08:00
Changming Sun
de018f58e8
Update manylinux build scripts (#9701) 2021-11-09 11:55:49 -08:00
Wei-Sheng Chin
bdc279a7ed
Use the same allocator following Pytorch (#9697) 
* Use the same allocator following Pytorch

* Polish

* Fix AMD build
 2021-11-09 11:25:16 -08:00
satyajandhyala
229c9a4e1c
Added Trilu CUDA kernel. (#9633) 
* Added Trilu CUDA kernel.

* Added TriluGrad.

* Added a training testcase for Trilu.

* Added Trilu gradient checker test.
 2021-11-09 11:20:17 -08:00
Tang, Cheng
6420530b3a
fix the mkl dependency for eager mode (#9702) 
* explicit link with libtorch instead of use cmake var to avoid introduce mkl dependency

* use find_lib to get libtorch lib name

* temp fix

* add missing libraries

Co-authored-by: Cheng Tang <chenta@microsoft.com>
 2021-11-09 08:52:55 -08:00
Guoyu Wang
a70ae24475
Add QDQ::Selector::Select to use const GraphViewer instead of mutable Graph (#9621) 
* Move qdq selector to use const GraphViewer

* minor update

* Move qdq logic from NodeSelector to QDQ Selectors

* Fix build break

* Move selector result to NodesToOptimizeIndexes

* fix build break

* address CR comments

* move indexes -> indices

* Pass  graph_viewer to avoid recreating many times

* Update after merge master

* update graph viewer remarks

* update comments

* Add ut for new qdq selector logic

* Increase minimal binary size limit

* UT minor update

* Address CR comments
 2021-11-08 21:36:29 -08:00
Hariharan Seshadri
65590b049c
Expose an API to query the CUDA compute stream to launch a custom kernel (#9141) 2021-11-08 21:10:34 -08:00
mindest
c579ebfbc3
change a for iteration (#9678) 
Co-authored-by: Min Lin <linmin@microsoft.com>
 2021-11-09 08:33:50 +08:00
Changming Sun
53afaefe3b
Refactor Windows CI pipeline yaml files (#9672) 2021-11-08 11:11:49 -08:00
Ryan Hill
24e35fba32
Change TensorShape to typically not allocate heap memory (#9542) 2021-11-08 10:29:54 -08:00
Ginés Hidalgo
13e64f8ff7
Remove all warnings C4800: Implicit conversion from 'int32_t/int64_t' to bool. Possible information loss (#9535) 2021-11-08 10:12:27 -08:00
Xavier Dupré
7e207ba3be
Use ORTMODULE_ONNX_OPSET_VERSION to modify the opset version in OrtModule (#9529) 
* Use environment variable to change the ONNX opset in ORTModule
* overwrite ONNX_OPSET_VERSION
* store envvar in module constant
 2021-11-08 17:03:16 +01:00
ashari4
1151c661eb
Add gi overload (#9690) 2021-11-07 16:04:00 -08:00
Yulong Wang
c6fddb263f
Add Node.js binding support to packaging pipeline (#9577) 2021-11-05 15:29:40 -07:00
Changming Sun
1cbbafdbe0
Change the default value of onnxruntime_DISABLE_RTTI (#9674) 2021-11-05 15:27:04 -07:00
Edward Chen
3e5dbfd94f
Increase binary size limit and publish binary size data even if limit is exceeded. (#9675) 2021-11-05 09:41:20 -07:00
Abhishek Jindal
dfe4d0a330
Abjindal/eager windows ci pipeline (#9587) 
* adding eager ci pipelines files

* adding import torch before onnxruntime

* finding os environ path

* finding os environ path corrected

* print OS environ path variables

* adding environ path for torch

* changing python version

* changing python python for torch libs

* removing import torch statements

* removing unncecessary torch path

* removing path variable

* add dll_path

* test for python 3.7

* adding dll directory path for python 3.8+

* print dll directory path for python 3.8+

* adding requirements file

* change requirements directory

* print more

* adding dll dir path

* removing setup eager file

* adding details for dll directory

* adding details for dll directory more

* adding import torch in onnxruntime init file

* removing dll dir path and moving requirements file

* enabling pipeline for py3.7

* remove enter

* removing debug build

* removing openmp

* adding comments for torch dll loading and cases of failure

* cleaning up the pipeline
 2021-11-05 09:09:09 -07:00
Changming Sun
74dc48f61c
Remove python 3.6 from macos pipeline (#9602) 2021-11-04 19:36:44 -07:00
pengwa
ee167bd078
Optimize _TileKernel for non-memcpy case (#9648) 
* optimize _TileKernel for non-memcpy case

* fallback shape_rank >MAX_DIMS
 2021-11-05 09:22:09 +08:00
Edward Chen
a355bcbd73
Clarify cgmanifest.json update process. (#9664) 2021-11-04 16:23:52 -07:00
Ginés Hidalgo
90de3c8e7c
Update MLOperatorAuthorPrivate.h (#9575) 
Updated MLOperatorAuthorPrivate.h to remove `enum DML_TENSOR_DATA_TYPE;` to avoid warning "C4471: 'DML_TENSOR_DATA_TYPE': a forward declaration of an unscoped enumeration must have an underlying type"
 2021-11-04 16:19:22 -07:00
Hariharan Seshadri
bbeceb7541
Support optional type in ORT (#8339) 2021-11-04 15:01:42 -07:00
Ginés Hidalgo
2fb03769bc
Updated OperatorUtility to avoid C2672 and C2783 (#9651) 
Updated OperatorUtility to avoid compiler error errors C2672 and C2783.

- Error C2672: 'TryMapStringToIndex': no matching overloaded function found
- Error C2783: 'std::optional<_Ty> Dml::TryMapStringToIndex(std::string_view,gsl::span<const Dml::NameAndIndex>)': could not deduce template argument for 'T'. note: see declaration of 'Dml::TryMapStringToIndex'. 'TryMapStringToIndex': function declaration must be available as none of the arguments depend on a template parameter
 2021-11-04 14:04:32 -07:00
Ginés Hidalgo
1128bf282d
Updated cpuid_info.cc to fix PVS-Studio error C1012 (#9581) 
Updated cpuid_info.cc to fix PVS-Studio error: "fatal error C1012: unmatched parenthesis: missing ')'"
 2021-11-04 14:03:49 -07:00
Edward Chen
ddb4c05852
Save graph runtime optimizations for minimal build (#9508) 
Add support for saving graph runtime optimizations in an ORT format model. The idea is to allow some optimizations to be "replayed" at runtime in a minimal build. The replaying part will be in a future change.
 2021-11-04 10:49:46 -07:00