Commit graph

3853 commits

Author SHA1 Message Date
Changming Sun
2d9dcc4576
Add python 3.9 support (#5874)
1. Add python 3.9 support(except Linux ARM)
2. Add Windows GPU python 3.8 to our packaging pipeline.
2020-11-30 12:02:48 -08:00
Wenbing Li
1852ade75d
Enable the xcode build for Apple Silicon (arm64 MacOS) (#5924)
* fix the build script for macos/xcode

* add the version check

* correct the osx-arch configuration

* typo
2020-11-30 11:22:08 -08:00
Jesse Benson
45966d878a Code review feedback 2020-11-30 09:24:22 -08:00
Jesse Benson
86e30a2db6 Update CUDA IsAllFinite kernel 2020-11-30 09:24:22 -08:00
Jesse Benson
bd96f60888 Use CUDA's IsAllFinite kernel for ROCm 2020-11-30 09:24:22 -08:00
Moshe David
06ad516a5d
w (#5947)
Co-authored-by: modav <modav@microsoft.com>
2020-11-30 10:35:44 +10:00
Moshe David
5f5d4a10bd
[OpenVINO]Fix memory leak upon exception throwing (#5954)
1. Make sure to free the output_shape vector even if Output names mismatch between OpenVINO and ONNX exception is thrown
2. Piggy back this PR to remove un-needed call to fstream close method

Authored-by: modav <modav@microsoft.com>
2020-11-26 19:44:51 -08:00
Ivan Stojiljkovic
015fbb3dbb
Add support for Python 3.8+ on Windows when CUDA is enabled (#5956) 2020-11-26 15:52:30 -08:00
Moshe David
e207589631
[OpenVINO]Fix memory leak in IsDebugEnabled() under Windows (#5948)
* w

* w

Co-authored-by: modav <modav@microsoft.com>
2020-11-26 15:10:37 -08:00
Guoyu Wang
4afdced775
[NNAPI EP] Update squeeze ops (#5946)
* [NNAPI EP] Update squeeze ops
2020-11-26 21:00:54 +10:00
Guoyu Wang
d52b9aca68
Enable scalar input/output for NNAPI EP (#5922)
* Enable scalar input for NNAPI EP

* Map scalar output ({1} tensor) of NNAPI back to scalar ({} tensor) if necessary

* Enforce NNAPI scalar output has {1} shape

* address CR comments

* minor update
2020-11-25 21:05:50 -08:00
Changming Sun
5fdd9f0fd2
Fix Python Linux GPU package name (#5943)
Fix Python Linux GPU package name. I accidentally added "noopenmp" to it.
2020-11-25 17:46:11 -08:00
Ryan Hill
27513d1fd7
Update BUILD.md with shared provider information (#5944)
* Update build instructions to include information about shared providers
2020-11-25 17:35:34 -08:00
Dmitri Smirnov
c4b55d29fe
Fix publishing pipelines. (#5942)
Fix publishing pipelines.
2020-11-25 16:23:08 -08:00
Changming Sun
1dbabb2362
Update dockerfiles (#5929)
1. Remove conda from the images. Because conda contains a file named /opt/miniconda/lib/libcrypto.so.1.0.0 which can't pass our security scan. Also, it will be easier for us to manage the third party usage registrations.
2. Remove openssh from the images. Because the official openssh package provided by Ubuntu can't pass our security scan.
3. Reduce the image size to 1/3 by using stages. Also, because it contains less packages, it will be less often needed to update.
4. Put the LICENSE-IMAGE.txt file in right place. It is missed in current images. You can see it was added to a temp folder "/code" but it got deleted afterwards.
5. Update the CPU docker image's base image to Ubuntu 18.04. The GPU one is already 18.04. It's better to keep them the same.
6. Remove the build arg ONNXRUNTIME_REPO/ONNXRUNTIME_BRANCH. Instead, the new one always uses the local source. I feel it can reduce confusion.
2020-11-25 15:38:22 -08:00
Changming Sun
c5b4d9091c
Fix a tiny issue in onnxruntime_unittests.cmake (#5901) 2020-11-25 14:21:13 -08:00
Edward Chen
7546d251e0
Expose parameters in clean build Docker image cache build. (#5941)
Expose some parameters in the clean build Docker image cache build. In particular, whether to do a dry-run and the lifetime of unused cache images.
2020-11-25 14:15:54 -08:00
Tianlei Wu
31a6be3d67
Add Longformer Attention Cuda Op(#5932)
Limitation: Global tokens must be at the beginning of sequence.
2020-11-25 13:52:10 -08:00
Maajid khan
e39e82b43a
Bug fix for MaskRCNN and FasterRCNN (#5935)
Signed-off-by: MaajidKhan <n.maajidkhan@gmail.com>
2020-11-25 12:10:09 -08:00
baijumeswani
69b9368c93
Add unit tests to identify configuration migration scenarios for checkpointing (#5678) 2020-11-25 09:40:26 -08:00
sfatimar
8168c91978
Sahar/fix documentation shared lib (#5926)
* Update OpenVINO-ExecutionProvider.Md

update openvino-executionprovider.md for shared library

* Update Build.md

updated --build_shared_lib flag for building openvino shared provider lib

* Update Dockerfile.openvino 

building for shared library with the new changes for openvino shared lib

* Revert "Update Build.md"

This reverts commit c9cf5fee76be7fdc10cadf07259f1d4ed5b45b93.

* Revert "Update Dockerfile.openvino "

This reverts commit e1624e4f93a4cfb425b6f21d7fb71b299a146740.

* Update OpenVINO-ExecutionProvider.md

fix documentation to the shared library

Co-authored-by: sfatimar <sahar.fatima@intel/com>
2020-11-25 08:50:01 -08:00
Guoyu Wang
87368655e2
Make NNAPI EP reject nodes with no-shape inputs (#5927) 2020-11-25 00:21:00 -08:00
Adam Pocock
fddbd8935c
Adding Java support for getAvailableProviders and other small methods (#5366)
* Adding Java support for getAvailableProviders, addFreeDimensionOverrideByName, disablePerSessionThreads and getProfilingStartTimeNs.

* Fixing copyright years, running spotless and adding javadoc and an accessor to OrtProvider.

* Renaming OrtSession.getProfilingStartTimeInNs.

* Removing ngraph as it's been deprecated.
2020-11-24 21:42:57 -08:00
S. Manohar Karlapalem
40926867c3
Add OpenVINO EP shared lib to Py Wheel (#5920)
* Add OpenVINO EP shared lib to Py Wheel

Include the libonnxruntime_providers_openvino.so/.dll to the wheel

* Follow libs.extend pattern as other EPs
2020-11-24 21:27:13 -08:00
Suffian Khan
4d603e83d7
Remove attention_past.cu and attention_transpose.cu from hipify to fix AMD build (#5921)
* remove attention_transpose.cu and attention_past.cu from hipify

* remove print line

* remove trailing ws for flake test

* fix ws onre mor etime
2020-11-24 20:49:06 -05:00
Adam Pocock
8b83c51a35
[Java] Initial Apple Silicon support (#5891)
* Rearranging checks in onnxruntime_mlas.cmake to pickup Apple Silicon.

On an M1 Macbook Pro clang reports:

$ clang -dumpmachine
arm64-apple-darwin20.1.0

So the regex check needs to look for "arm64" first, as otherwise it
matches 32-bit ARM and you get NEON compilation failures.

* Adding Java side library loading support for Apple Silicon (and other aarch64 architectures).

* Adding Qgemm fix from @tracysh

* Fixes the java packaging on Windows.

* Missed a check in the java platform detector.
2020-11-24 15:51:40 -08:00
KeDengMS
ee908eb0aa
Symbolic shape inference: fix rank for ConstantOfShape (#5912) 2020-11-24 14:50:41 -08:00
Dmitri Smirnov
c2d610066a
C#: Add CreateFromMemory to FixedBufferOnnxValue to allow bind user buffers and pass custom binary compatible types (#5886)
Add CreateFromMemory to FixedBufferOnnxValue so users can bind their own custom binary compatible buffers to feed/fetch data.
2020-11-24 14:10:14 -08:00
Ashwini Khade
705d093167
Update onnx (#5720)
* update onnx

* update docker image for testing
2020-11-24 11:20:15 -08:00
Suffian Khan
9b8189dd0a
Rework AMD CI pipeline to use pool AMD-GPU and disable more tests in order to enable it. (#5885)
Move AMD test pipeline to use self-hosted pool AMD-GPU. For time being, remove failing/flaky unit tests for AMD pipeline.
2020-11-24 09:38:14 -08:00
Guoyu Wang
846c5fb917
Report arm64 minimal baseline binary size only for continuous integration (#5913)
* Report binary size only for continuous integration
2020-11-24 20:24:08 +10:00
Guoyu Wang
c49d5f1d98
Reenable skip flatten/reshape if it's Gemm's input (#5904) 2020-11-24 00:01:23 -08:00
Changming Sun
782303324e
Revert "Improves performance of operator Transpose (#5550)" (#5911)
This reverts commit e5c8040c52.
2020-11-24 00:01:03 -08:00
Zhang Lei
9992f0f812
Implement QLinear GlobalAveragePool with sse2/neon. (#5838)
Add QLinear Global Average Pool for quantization for ARM and SSE2.

Co-authored-by: Tracy Sharpe <tracysh@microsoft.com>
2020-11-23 19:23:58 -08:00
Guoyu Wang
d0a007e4f1
Add info level logging for NNAPI EP partitions and number of supported nodes (#5903) 2020-11-23 18:22:00 -08:00
Guoyu Wang
4137c18d9b
Add ORT minimal with NNAPI EP to Android CI (#5890)
Description: Add ORT minimal with NNAPI EP to Android CI

Motivation and Context

The added build/test to Android CI will only run UT, additional onnx_test_runner with customer .ort models will be added later
2020-11-23 18:21:34 -08:00
sfatimar
916410151c
Fix for hetero multi python binding with new shared library (#5895)
Co-authored-by: sfatimar <sahar.fatima@intel/com>
2020-11-23 15:41:10 -08:00
Xueyun Zhu
58ea7b3572
temporarily disable test (#5868) 2020-11-23 15:18:37 -08:00
Edward Chen
5e8fcda24a
Build docker image cache fixes. (#5902)
Fix Python 3.5 compatibility issue in tools/ci_build/get_docker_image.py.
Fix line endings in tools/ci_build/github/azure-pipelines/clean-build-docker-image-cache-pipeline.yml.
2020-11-23 14:43:12 -08:00
Ye Wang
3d5b48a894
remove use_cdn when loading pretrained model (#5900) 2020-11-23 14:26:55 -08:00
baijumeswani
208f4c1d3c
Azure ci pipeline for distributed environment tests (#5881) 2020-11-23 14:01:00 -08:00
satyajandhyala
353e071b7e
Fuzz testing misc (#5862)
* Run only required steps relevant to fuzz testing.

* Exit status non-zero for any uncaught exception other than ort_exception in the driver code
Co-authored-by: Satya Jandhyala <sajandhy@microsoft.com>
2020-11-23 13:43:44 -08:00
Tracy Sharpe
f473dd295d
Add QLinearMatMul(u8s8) (#5899) 2020-11-23 12:04:32 -08:00
Tracy Sharpe
57c92066c2
Implement missing pieces for ARM QLinearConv support (#5894) 2020-11-22 23:19:27 -08:00
Hariharan Seshadri
d46dbeafd3
Expose knobs to create and share (CPU) allocators across sessions in C# and Python (#5634) 2020-11-21 14:12:33 -08:00
Guoyu Wang
26e6ced172
Temporary fix for Android CI failure (#5889)
* Unblock the Android CI

* Add python to android ci's command
2020-11-21 17:58:32 +10:00
Tianlei Wu
910bbfe1ef
Support 3D attention mask (#5887)
Support 3D attention mask with shape (batch_size, sequence_length, all_sequence_length)
2020-11-20 22:48:01 -08:00
Guoyu Wang
cc6e8fb7cc
Filter initializers for GraphViewer with IndexedSubGraph (#5884)
* fix filtered subgraph initializer issue

* minor fix

* Inlcude implicit input of nodes to see if they are initializers

* Add test case

* minor update

* Address PR comments

* Fix some code error
2020-11-20 18:36:58 -08:00
Ryan Hill
ba739a8000
Convert OpenVINO into a shared provider (#5778)
Same as Dnnl and TensorRT before it, now with more methods and more cleanup.
2020-11-20 17:39:57 -08:00
Edward Chen
8c04ada0d0
Bump training docker image version. (#5428) 2020-11-20 16:38:09 -08:00