ONNX Runtime: cross-platform, high performance ML inferencing and training accelerator
Find a file
raoanag b2768bbf23
Enabling support for QAttention (#18326)
[Cherry Pick Reviewed]

https://github.com/microsoft/onnxruntime/pull/16837
https://github.com/microsoft/onnxruntime/pull/16851
https://github.com/microsoft/onnxruntime/pull/17947

### Description
Enabling support for `Past`, `Present` and `unidirectional` for
[QAttention](https://github.com/microsoft/onnxruntime/blob/main/docs/ContribOperators.md#com.microsoft.QAttention)
Contrib Op



```
Note: Google Test filter = *QAttention*
[==========] Running 14 tests from 2 test suites.
[----------] Global test environment set-up.
[----------] 1 test from CPU_U8S8_Precision_Tests
[ RUN      ] CPU_U8S8_Precision_Tests.QAttention
[       OK ] CPU_U8S8_Precision_Tests.QAttention (104 ms)
[----------] 1 test from CPU_U8S8_Precision_Tests (105 ms total)

[----------] 13 tests from QAttentionTest
[ RUN      ] QAttentionTest.QAttentionBatch1
[       OK ] QAttentionTest.QAttentionBatch1 (255 ms)
[ RUN      ] QAttentionTest.QAttentionBatch1_Float16
[       OK ] QAttentionTest.QAttentionBatch1_Float16 (0 ms)
[ RUN      ] QAttentionTest.QAttentionBatch2
[       OK ] QAttentionTest.QAttentionBatch2 (201 ms)
[ RUN      ] QAttentionTest.QAttentionMaskPartialSequence
[       OK ] QAttentionTest.QAttentionMaskPartialSequence (197 ms)
[ RUN      ] QAttentionTest.QAttentionMaskExceedSequence
[       OK ] QAttentionTest.QAttentionMaskExceedSequence (192 ms)
[ RUN      ] QAttentionTest.QAttentionNoMaskIndex
[       OK ] QAttentionTest.QAttentionNoMaskIndex (186 ms)
[ RUN      ] QAttentionTest.QAttentionUnidirectional_U8U8
[       OK ] QAttentionTest.QAttentionUnidirectional_U8U8 (9 ms)
[ RUN      ] QAttentionTest.QAttentionUnidirectional_U8S8
[       OK ] QAttentionTest.QAttentionUnidirectional_U8S8 (9 ms)
[ RUN      ] QAttentionTest.QAttentionUnidirectional_CUDA
[       OK ] QAttentionTest.QAttentionUnidirectional_CUDA (0 ms)
[ RUN      ] QAttentionTest.QAttentionPastState_u8u8
C:\workspace\ORT\onnxruntime\onnxruntime\test\providers\checkers.cc(268): error: The difference between cur_expected[i] and cur_actual[i] is 0.75743968039751053, which exceeds *(params.absolute_error), where
cur_expected[i] evaluates to 0.67312467098236084,
cur_actual[i] evaluates to -0.084315009415149689, and
*(params.absolute_error) evaluates to 0.00019999999494757503.
i:0
Google Test trace:
C:\workspace\ORT\onnxruntime\onnxruntime\test\providers\checkers.cc(484): provider type: DmlExecutionProvider
C:\workspace\ORT\onnxruntime\onnxruntime\test/common/random_generator.h(49): ORT test random seed: 2178993560
C:\workspace\ORT\onnxruntime\onnxruntime\test\providers\checkers.cc(268): error: The difference between cur_expected[i] and cur_actual[i] is 0.75743968039751053, which exceeds *(params.absolute_error), where
cur_expected[i] evaluates to 0.67312467098236084,
cur_actual[i] evaluates to -0.084315009415149689, and
*(params.absolute_error) evaluates to 0.00019999999494757503.
i:0
Google Test trace:
C:\workspace\ORT\onnxruntime\onnxruntime\test\providers\checkers.cc(484): provider type: DmlExecutionProvider
C:\workspace\ORT\onnxruntime\onnxruntime\test/common/random_generator.h(49): ORT test random seed: 2178993560
C:\workspace\ORT\onnxruntime\onnxruntime\test\providers\checkers.cc(268): error: The difference between cur_expected[i] and cur_actual[i] is 0.03001787792891264, which exceeds *(params.absolute_error), where
cur_expected[i] evaluates to -0.021467097103595734,
cur_actual[i] evaluates to 0.008550780825316906, and
*(params.absolute_error) evaluates to 0.00019999999494757503.
i:0
Google Test trace:
C:\workspace\ORT\onnxruntime\onnxruntime\test\providers\checkers.cc(484): provider type: DmlExecutionProvider
C:\workspace\ORT\onnxruntime\onnxruntime\test/common/random_generator.h(49): ORT test random seed: 2178993560
C:\workspace\ORT\onnxruntime\onnxruntime\test\providers\checkers.cc(268): error: The difference between cur_expected[i] and cur_actual[i] is 0.03001787792891264, which exceeds *(params.absolute_error), where
cur_expected[i] evaluates to -0.021467097103595734,
cur_actual[i] evaluates to 0.008550780825316906, and
*(params.absolute_error) evaluates to 0.00019999999494757503.
i:0
Google Test trace:
C:\workspace\ORT\onnxruntime\onnxruntime\test\providers\checkers.cc(484): provider type: DmlExecutionProvider
C:\workspace\ORT\onnxruntime\onnxruntime\test/common/random_generator.h(49): ORT test random seed: 2178993560
[  FAILED  ] QAttentionTest.QAttentionPastState_u8u8 (2067 ms)
[ RUN      ] QAttentionTest.QAttentionPastState_u8s8
C:\workspace\ORT\onnxruntime\onnxruntime\test\providers\checkers.cc(268): error: The difference between cur_expected[i] and cur_actual[i] is 0.74043640494346619, which exceeds *(params.absolute_error), where
cur_expected[i] evaluates to 0.65650326013565063,
cur_actual[i] evaluates to -0.083933144807815552, and
*(params.absolute_error) evaluates to 0.00019999999494757503.
i:0
Google Test trace:
C:\workspace\ORT\onnxruntime\onnxruntime\test\providers\checkers.cc(484): provider type: DmlExecutionProvider
C:\workspace\ORT\onnxruntime\onnxruntime\test/common/random_generator.h(49): ORT test random seed: 2178993560
C:\workspace\ORT\onnxruntime\onnxruntime\test\providers\checkers.cc(268): error: The difference between cur_expected[i] and cur_actual[i] is 0.081788420677185059, which exceeds *(params.absolute_error), where
cur_expected[i] evaluates to 1.0076344013214111,
cur_actual[i] evaluates to 1.0894228219985962, and
*(params.absolute_error) evaluates to 0.00019999999494757503.
i:965
Google Test trace:
C:\workspace\ORT\onnxruntime\onnxruntime\test\providers\checkers.cc(484): provider type: DmlExecutionProvider
C:\workspace\ORT\onnxruntime\onnxruntime\test/common/random_generator.h(49): ORT test random seed: 2178993560
C:\workspace\ORT\onnxruntime\onnxruntime\test\providers\checkers.cc(268): error: The difference between cur_expected[i] and cur_actual[i] is 0.74043640494346619, which exceeds *(params.absolute_error), where
cur_expected[i] evaluates to 0.65650326013565063,
cur_actual[i] evaluates to -0.083933144807815552, and
*(params.absolute_error) evaluates to 0.00019999999494757503.
i:0
Google Test trace:
C:\workspace\ORT\onnxruntime\onnxruntime\test\providers\checkers.cc(484): provider type: DmlExecutionProvider
C:\workspace\ORT\onnxruntime\onnxruntime\test/common/random_generator.h(49): ORT test random seed: 2178993560
C:\workspace\ORT\onnxruntime\onnxruntime\test\providers\checkers.cc(268): error: The difference between cur_expected[i] and cur_actual[i] is 0.081788420677185059, which exceeds *(params.absolute_error), where
cur_expected[i] evaluates to 1.0076344013214111,
cur_actual[i] evaluates to 1.0894228219985962, and
*(params.absolute_error) evaluates to 0.00019999999494757503.
i:965
Google Test trace:
C:\workspace\ORT\onnxruntime\onnxruntime\test\providers\checkers.cc(484): provider type: DmlExecutionProvider
C:\workspace\ORT\onnxruntime\onnxruntime\test/common/random_generator.h(49): ORT test random seed: 2178993560
C:\workspace\ORT\onnxruntime\onnxruntime\test\providers\checkers.cc(268): error: The difference between cur_expected[i] and cur_actual[i] is 0.024714200757443905, which exceeds *(params.absolute_error), where
cur_expected[i] evaluates to -0.016048312187194824,
cur_actual[i] evaluates to 0.0086658885702490807, and
*(params.absolute_error) evaluates to 0.00019999999494757503.
i:0
Google Test trace:
C:\workspace\ORT\onnxruntime\onnxruntime\test\providers\checkers.cc(484): provider type: DmlExecutionProvider
C:\workspace\ORT\onnxruntime\onnxruntime\test/common/random_generator.h(49): ORT test random seed: 2178993560
C:\workspace\ORT\onnxruntime\onnxruntime\test\providers\checkers.cc(268): error: The difference between cur_expected[i] and cur_actual[i] is 0.0092324763536453247, which exceeds *(params.absolute_error), where
cur_expected[i] evaluates to 0.24175386130809784,
cur_actual[i] evaluates to 0.25098633766174316, and
*(params.absolute_error) evaluates to 0.00019999999494757503.
i:979
Google Test trace:
C:\workspace\ORT\onnxruntime\onnxruntime\test\providers\checkers.cc(484): provider type: DmlExecutionProvider
C:\workspace\ORT\onnxruntime\onnxruntime\test/common/random_generator.h(49): ORT test random seed: 2178993560
C:\workspace\ORT\onnxruntime\onnxruntime\test\providers\checkers.cc(268): error: The difference between cur_expected[i] and cur_actual[i] is 0.024714200757443905, which exceeds *(params.absolute_error), where
cur_expected[i] evaluates to -0.016048312187194824,
cur_actual[i] evaluates to 0.0086658885702490807, and
*(params.absolute_error) evaluates to 0.00019999999494757503.
i:0
Google Test trace:
C:\workspace\ORT\onnxruntime\onnxruntime\test\providers\checkers.cc(484): provider type: DmlExecutionProvider
C:\workspace\ORT\onnxruntime\onnxruntime\test/common/random_generator.h(49): ORT test random seed: 2178993560
C:\workspace\ORT\onnxruntime\onnxruntime\test\providers\checkers.cc(268): error: The difference between cur_expected[i] and cur_actual[i] is 0.0092324763536453247, which exceeds *(params.absolute_error), where
cur_expected[i] evaluates to 0.24175386130809784,
cur_actual[i] evaluates to 0.25098633766174316, and
*(params.absolute_error) evaluates to 0.00019999999494757503.
i:979
Google Test trace:
C:\workspace\ORT\onnxruntime\onnxruntime\test\providers\checkers.cc(484): provider type: DmlExecutionProvider
C:\workspace\ORT\onnxruntime\onnxruntime\test/common/random_generator.h(49): ORT test random seed: 2178993560
[  FAILED  ] QAttentionTest.QAttentionPastState_u8s8 (2079 ms)
[ RUN      ] QAttentionTest.QAttentionPrunedModel
[       OK ] QAttentionTest.QAttentionPrunedModel (206 ms)
[ RUN      ] QAttentionTest.SharedPrepackedWeights
[       OK ] QAttentionTest.SharedPrepackedWeights (79 ms)
[----------] 13 tests from QAttentionTest (5492 ms total)

[----------] Global test environment tear-down
[==========] 14 tests from 2 test suites ran. (5600 ms total)
[  PASSED  ] 12 tests.
[  FAILED  ] 2 tests, listed below:
[  FAILED  ] QAttentionTest.QAttentionPastState_u8u8
[  FAILED  ] QAttentionTest.QAttentionPastState_u8s8

 2 FAILED TESTS
memleakdbg:
----- No memory leaks detected -----
```


### Motivation and Context
<!-- - Why is this change required? What problem does it solve?
- If it fixes an open issue, please link to the issue here. -->

---------

Co-authored-by: Xiang Zhang <xianz@microsoft.com>
2023-11-09 09:16:28 -08:00
.config Update tsaoptions.json: update the email alias (#13448) 2022-10-26 15:56:16 -07:00
.devcontainer Remove two lines in the Dockerfile for Github Codespace (#12278) 2022-07-21 20:52:17 -07:00
.gdn Update win-ci-pipeline.yml: enable xnnpack tests (#16244) 2023-06-14 19:12:42 -07:00
.github Merge main to WindowsAI (#18122) 2023-10-27 17:08:01 -07:00
.pipelines Update DML Preview Package to 13.0-dev4c864f8324cef2ff5c39a5822d6c4de05929306d (#18208) 2023-11-01 10:16:56 -07:00
.vscode cpplint & Eager mode: refactor and add comments to empty_* functions, general lint cleanup in ort_aten (#12238) 2022-07-20 11:47:57 -04:00
cgmanifests Merge main to WindowsAI (#18122) 2023-10-27 17:08:01 -07:00
cmake Update DML Preview Package to 13.0-dev4c864f8324cef2ff5c39a5822d6c4de05929306d (#18208) 2023-11-01 10:16:56 -07:00
csharp Merge main to WindowsAI (#18122) 2023-10-27 17:08:01 -07:00
dockerfiles Merge main to WindowsAI (#18122) 2023-10-27 17:08:01 -07:00
docs Merge main to WindowsAI (#18122) 2023-10-27 17:08:01 -07:00
include/onnxruntime/core Merge main to WindowsAI (#18122) 2023-10-27 17:08:01 -07:00
java Merge main to WindowsAI (#18122) 2023-10-27 17:08:01 -07:00
js Merge main to WindowsAI (#18122) 2023-10-27 17:08:01 -07:00
objectivec Objective-C Add Support to Create and Query String ORTValues (#16764) 2023-07-20 17:39:29 -07:00
onnxruntime Enabling support for QAttention (#18326) 2023-11-09 09:16:28 -08:00
orttraining Merge main to WindowsAI (#18122) 2023-10-27 17:08:01 -07:00
rust Merge main to WindowsAI (#18122) 2023-10-27 17:08:01 -07:00
samples Merge main to WindowsAI (#18122) 2023-10-27 17:08:01 -07:00
tools Update DML Preview Package to 13.0-dev4c864f8324cef2ff5c39a5822d6c4de05929306d (#18208) 2023-11-01 10:16:56 -07:00
winml Merge main to WindowsAI (#18122) 2023-10-27 17:08:01 -07:00
.clang-format Merge main to WindowsAI (#18122) 2023-10-27 17:08:01 -07:00
.clang-tidy Create clang-tidy CI (#12653) 2022-09-30 08:05:38 -07:00
.dockerignore
.gitattributes
.gitignore remove 'lib/' from .gitignore (#15613) 2023-04-24 18:43:32 -07:00
.gitmodules Merge main to WindowsAI (#18122) 2023-10-27 17:08:01 -07:00
.lintrunner.toml Merge main to WindowsAI (#18122) 2023-10-27 17:08:01 -07:00
build.bat Merge main to WindowsAI (#18122) 2023-10-27 17:08:01 -07:00
build.sh Upgrade old Python version in packaging pipeline (#16667) 2023-07-17 08:24:47 -07:00
CITATION.cff Fix CITATION.cff and add automatic validation of your citation metadata (#10478) 2022-04-13 10:03:52 -07:00
CODEOWNERS Add owners for public facing API files (#15288) 2023-03-30 17:16:15 -07:00
CONTRIBUTING.md Fix link to High Level Design (#11786) 2023-02-28 11:05:54 -08:00
lgtm.yml Fix lgtm C++ error (#13613) 2022-11-10 10:06:22 -08:00
LICENSE
NuGet.config
ort.wprp
ORT_icon_for_light_bg.png Update nuget icon (#10672) 2022-03-01 09:11:03 -08:00
packages.config Update DML Preview Package to 13.0-dev4c864f8324cef2ff5c39a5822d6c4de05929306d (#18208) 2023-11-01 10:16:56 -07:00
pyproject.toml Merge main to WindowsAI (#18122) 2023-10-27 17:08:01 -07:00
README.md add third-party pipeline status to README.md (#16155) 2023-05-31 22:14:39 -07:00
requirements-dev.txt Merge main to WindowsAI (#18122) 2023-10-27 17:08:01 -07:00
requirements-doc.txt
requirements-lintrunner.txt Merge main to WindowsAI (#18122) 2023-10-27 17:08:01 -07:00
requirements-training.txt Merge main to WindowsAI (#18122) 2023-10-27 17:08:01 -07:00
requirements.txt.in Add additional python requirements (#11522) 2022-05-20 16:16:18 -07:00
SECURITY.md Microsoft mandatory file (#11619) 2022-05-25 13:56:10 -07:00
setup.py Merge main to WindowsAI (#18122) 2023-10-27 17:08:01 -07:00
ThirdPartyNotices.txt Merge main to WindowsAI (#18122) 2023-10-27 17:08:01 -07:00
VERSION_NUMBER Merge main to WindowsAI (#18122) 2023-10-27 17:08:01 -07:00

ONNX Runtime is a cross-platform inference and training machine-learning accelerator.

ONNX Runtime inference can enable faster customer experiences and lower costs, supporting models from deep learning frameworks such as PyTorch and TensorFlow/Keras as well as classical machine learning libraries such as scikit-learn, LightGBM, XGBoost, etc. ONNX Runtime is compatible with different hardware, drivers, and operating systems, and provides optimal performance by leveraging hardware accelerators where applicable alongside graph optimizations and transforms. Learn more →

ONNX Runtime training can accelerate the model training time on multi-node NVIDIA GPUs for transformer models with a one-line addition for existing PyTorch training scripts. Learn more →

Get Started & Resources

Builtin Pipeline Status

System Inference Training
Windows Build Status
Build Status
Build Status
Linux Build Status
Build Status
Build Status
Build Status
Build Status
Build Status
Build Status
Build Status
Mac Build Status
Android Build Status
iOS Build Status
Web Build Status
Other Build Status
Build Status

Third-party Pipeline Status

System Inference Training
Linux Build Status

Data/Telemetry

Windows distributions of this project may collect usage data and send it to Microsoft to help improve our products and services. See the privacy statement for more details.

Contributions and Feedback

We welcome contributions! Please see the contribution guidelines.

For feature requests or bug reports, please file a GitHub Issue.

For general discussion or questions, please use GitHub Discussions.

Code of Conduct

This project has adopted the Microsoft Open Source Code of Conduct. For more information see the Code of Conduct FAQ or contact opencode@microsoft.com with any additional questions or comments.

License

This project is licensed under the MIT License.