saymrwulf/onnxruntime
1
0
Fork 0
mirror of https://github.com/saymrwulf/onnxruntime.git synced 2026-05-18 21:21:17 +00:00
Code Issues Projects Releases Packages Wiki Activity Actions
2465 commits 2856 branches 57 tags 1.3 GiB
Commit graph

2465 commits

This branch
This branch
All branches
Author SHA1 Message Date
Xueyun Zhu
0e59668c1b
add support for symbolic broadcast for Add/Sub/Mul (#3743) 
* add support for symbolic broadcast

* fix comment

* address feedback
 2020-05-06 10:40:57 -07:00
Scott McKay
687edd702c
Add RelWithDebInfo target to the C# projects so that it correctly finds the native build. (#3839) 
Make the cmake file slightly more consistent for the build c# flag.
 2020-05-06 20:01:04 +10:00
Adam Pocock
d38b79c6e5
[Java] Adding missing methods on Session, SessionOptions and RunOptions (v2) (#3832) 
* java - adding support for custom op libraries.

* Adding support for RunOptions and additional methods for SessionOptions and OrtSession.

As a result OrtEnvironment.LoggingLevel moved to be a top level enum
called OrtLoggingLevel.

* java - adding unit tests for RunOptions and SessionOptions.

* java - removing unused releaseNamesHandle method

* java - add test for custom op library.

* java - adding log verbosity methods, and tests for the same.

* java - fixes for custom op loading test on Windows.

* Cleanup after rebase on master.
 2020-05-06 01:19:46 -07:00
Ryan Hill
d5ec353e58
Ryanunderhill/mkldnn dll (#3314) 
First version of allowing providers to work as DLLs, only implemented for DNNL so far.

More improvements to come next!
 2020-05-06 00:57:09 -07:00
M. Zeeshan Siddiqui
9b02b3df6f
Update ONNX submodule to ONNX 1.7 release candidate 3. (#3838) 2020-05-06 00:55:19 -07:00
Hariharan Seshadri
b45ce92542
Fix 3828 (#3837) 2020-05-06 00:05:15 -07:00
Changming Sun
f0c9fbc051
MaxpoolWithMask (#3831) 2020-05-05 22:19:35 -07:00
airockchip
edaf8a542c
Initial PR for RKNPU execution provider (#3609) 
* Initial RKNPU execution provider

    * Init

    * Support Ops:
        Conv, Relu, Clip, LeakyRelu,
        MaxPool, AveragePool, GlobalAveragePool,
        Concat, Softmax, BatchNormalization, Gemm,
        Add, Mul, Sub,
        Reshape, Squeeze, Unsqueeze,
        Flatten, Transpose,
        QLinearConv, DequantizeLinear

    * Add rknpu unittest

    * Update BUILD.md and Add RKNPU-ExecutionProvider.md

* misc code update

* fix CLIP accuracy issue.

* fix "Error: Duplicate definition of name".

* move rknpu_ddk out of onnxruntime submodule.

* remove temporary code.

* add rknpu namespace.

* update misc of node_attr_helper

* add const & comment for onnx_converter

* add const & comment for shaper

* unify variable name

Co-authored-by: dkm <dkm@rock-chips.com>
Co-authored-by: George Wu <jywu@microsoft.com>
 2020-05-05 20:36:47 -07:00
Tianlei Wu
584facf830
Fix build error when DEBUG_NODE_INPUTS_OUTPUTS is on (#3826) 2020-05-05 16:52:33 -07:00
liuziyue
6e341c002f
Add Epsilon Attribute to Skip and Embed Layer Normalization (#3768) 
* add epsilon to SkipLayerNormalization and EmbedLayerNormalization
 2020-05-05 16:36:26 -07:00
Pranav Sharma
e30d2e38b9
Add guidelines for writing a good PR. (#3830) 2020-05-05 16:28:21 -07:00
M. Zeeshan Siddiqui
ef4d73e887
Update ONNX submodule to ONNX 1.7 release candidate 2. (#3818) 
* Update ONNX submodule to ONNX 1.7 release candidate 2.

* fix build error.

* Update ONNX submodule to latest and disable preview op tests.
 2020-05-05 15:08:40 -07:00
Dmitri Smirnov
5db30a470e
[Java] Tidying up the sample MNIST code (#3824) 
* Updating the Java sample to load MNIST in libsvm format.
* java - code formatting fix.
Co-authored-by: Adam Pocock <adam.pocock@oracle.com>
 2020-05-05 14:34:13 -07:00
Bowen Bao
f7ff5a7aa1
Fix state_dict and save_as_onnx for training (#3774) 2020-05-05 11:47:46 -07:00
Yulong Wang
5dfc91db51
Node.js binding for ONNX Runtime (#3613) 
* initial commit for Node.js binding

* add c++ code

* add inference session impl

* e2e working

* add settings.json

* add test data

* adjust binding declaration

* refine tensor constructor declaration

* update tests

* enable onnx tests

* simply refine readme

* refine cpp impl

* refine tests

* formatting

* add linting

* move bin folder

* fix linux build

* manually update test filter list

* update C++ API headers: fix crash in release build

* make (manually) prebuild work

* add test into prepack script

* specify prebuild runtime type (N-API)

* build.ts: update rebuild and include regex

* fix lazy load on electron.js

* update dev version, git link and binary host

* support session options and run options

* bump dev version

* update README

* add 1 example

* move folder

* adjust path

* update document for examples

* rename example 01

* add example 02

* add session option: log severity level

* add example 04

* resolve comments

* fix typo

* remove double guard in header files

* add copyright banner

* move BUILD outside from README

* consume test filter list from onnxruntime
 2020-05-05 11:45:12 -07:00
Hariharan Seshadri
cffa1b7bf2
Fix (#3812) 2020-05-05 02:08:13 -07:00
Changming Sun
bd78364411
Parallel all the activations ops (#3722) 
1. Parallel all the activations ops.
2. Parallel the performance critical path of the LRN op, which makes the ONNX model zoo googlenet model runs 60% faster(latency reduced from 21ms to 13ms).
3. Make the Gemm-Activation fusion support with all the activations ops. Before this change, it only supports LeakyRelu/Relu/Sigmoid/Tanh.
4. Delete onnxruntime/test/framework/op_kernel_test.cc because the file is almost empty.
5. Remove the loggings in KernelRegistry::TryFindKernel, return Status with error message instead.
 2020-05-05 01:18:17 -07:00
Changming Sun
c11fbf68e4
Publish gpu package to nuget feed (#3816) 2020-05-04 21:49:19 -07:00
Scott McKay
b386b41703
Fix bug in GRU when linear_before_reset is true and no bias input is provided (#3797) 
* Allocate linear_output_ when linear_before_reset is true and there is no bias input.
Add test for this combination.
 2020-05-05 13:15:57 +10:00
M. Zeeshan Siddiqui
a24c71af40
Update Dropout(12) forward kernel with training_mode input. (#3805) 
* Update Dropout(12) forward and backward kernel with training_mode input.

* Revert deleted assert.

* clean up.

* PR feedback.
 2020-05-04 20:05:42 -07:00
Dmitri Smirnov
111469728f
Make Java build and run tests on Windows the box (#3811) 
Incorporate .DLL symbolic link names fix.
  Make unit tests run. Make gradle run on Windows.
 2020-05-04 18:19:35 -07:00
M. Zeeshan Siddiqui
6f95cdfa68
Use new cost based threadpool abstractions in CPU gradient operators. (#3807) 
* Use ThreadPool abstractions instead of OpenMP.

* PR feedback.
 2020-05-04 15:23:10 -07:00
Yufeng Li
156368b67f
Quantize attention with Cuda (#3693) 
* Add definition of QAttention
* implemention of QAttention on GPU
 2020-05-04 14:20:38 -07:00
Tianlei Wu
49f0610447
Add options --disable_layer_norm, --disable_gelu and --enable_gelu_approximation (#3750) 2020-05-04 14:06:57 -07:00
Sherlock
2f8a2364c3
Fix loss function builder (#3801) 2020-05-04 10:41:15 -07:00
Hariharan Seshadri
785b45124d
Add CPU kernel for Einsum op (#3575) 2020-05-03 23:48:38 -07:00
Yulong Wang
c8269e4b89
move backend test filters into data file (#3798) 
* move backend test filters into data file

* update data

* update data

* update document

* fix list for current_failing_tests_OPENVINO_CPU_FP32
 2020-05-02 19:05:58 -07:00
Changming Sun
2684d47fc5
Disable data downloading in linux-nocontribops-ci-pipeline (#3803) 
* Disable data downloading in linux-nocontribops-ci-pipeline

* update

* update
 2020-05-02 12:59:24 -07:00
Sheil Kumar
37b60251ca
test packaging (#3756) 
Co-authored-by: Sheil Kumar <sheilk@microsoft.com>
Co-authored-by: Changming Sun <chasun@microsoft.com>
 2020-05-02 12:23:33 -07:00
Changming Sun
ee8900e21a
Update centos-ci-pipeline.yml (#3800) 
* Update centos-ci-pipeline.yml
 2020-05-02 11:04:23 -07:00
Jeff Bloomfield
d5b2cd7493
Add performance best practices to DML EP doc (#2859) 
* Add performance best practices to DML EP doc


Co-authored-by: Jeff <38966965+jeffbloo@users.noreply.github.com>
 2020-05-02 09:53:33 -07:00
Scott McKay
42cf971ca2
Add a couple of utility scripts to tools/python (#3621) 
* Add a helper script to more easily create a test directory for use with onnx_test_runner or onnxruntime_perf_test.
Add example script that can be used as a base for performance testing a model with a variety of input sizes.
Add __init__.py so files in this directory can be imported in other scripts.

* Fix some flake8 warnings.
Add example of specifying attribute for op.

* Add ability for test dir creation to fill in all missing input data with random values.
Add example of using test dir creation this way
 2020-05-02 17:35:43 +10:00
edgchen1
440f361363
Remove orttraining-linux-gpu-inference-only-ci-pipeline.yml. (#3788) 2020-05-02 00:35:08 -07:00
Sheil Kumar
43a828f0a2
Add tests for WinRT Projection Raw ABI consumption (#3718) 
Add tests for WinRT Projection Raw ABI consumption
Co-authored-by: Sheil Kumar <sheilk@microsoft.com>
 2020-05-02 00:33:17 -07:00
Tianlei Wu
3fab8ebfe9
(MaximKalininMS) Fix Reshape Fusion and Crash in Reshape (#3777) 
* Fix a crash in Reshape
Reshape doesn't handle 0 input dimension properly, which leads to a
division by zero

* Fix reshape fusion
https://github.com/microsoft/onnxruntime/pull/3554 introduced a bug:
initializers can now come before Shape->Gather->Unsqueeze chains; if
those initializers have more than 1 element, expected dimensions in the
chains are now incorrect.

Authored-by: Max Kalinin <makalini@microsoft.com>
 2020-05-02 00:20:00 -07:00
Scott McKay
15eca74d15
Make ThreadPool::PartitionWork a bit more user friendly. Update a few places to use PartitionWork. (#3795) 2020-05-02 17:09:55 +10:00
Pranav Sharma
2b8d9ef0fd
Refactor scatter/gather ops to use the new cost based threadpool abstractions. (#3776) 
* Update Scatter and Gather ops by replacing pragma omp invocations with the new threadpool abstractions.

* Use forward declarations

* PR comments
 2020-05-02 17:09:31 +10:00
M. Zeeshan Siddiqui
4f9f6aedea
CUDA/CPU test for NegativeLogLikelihoodLoss(12) function based loss operator. (#3793) 2020-05-01 21:36:29 -07:00
Sheil Kumar
b1c4d6ff4e
bump dml version (#3792) 
Co-authored-by: Sheil Kumar <sheilk@microsoft.com>
 2020-05-01 18:16:00 -07:00
David Brownell
4b8fad214a
Initial checkin (#3791) 2020-05-01 14:58:49 -07:00
Scott McKay
11b819054b
Fix tree ensemble threading bug (#3778) 
* Fix first instance where the calculation for TryBatchParallelFor was incorrect. Other usages need to be validated.

* Fix some other usages of the threadpool.
 2020-05-02 07:50:35 +10:00
Scott McKay
2fc3984e70
Add test that C is unidirectionally broadcast-able before fusing the MatMul with Add. (#3780) 
Addresses #3764
 2020-05-02 07:36:21 +10:00
Xueyun Zhu
e8e95110d3
add pipeline to distributed context config (#3789) 
* add pipeline to distributed context

* white space
 2020-05-01 13:49:51 -07:00
M. Zeeshan Siddiqui
517bff9675
Function expansion support and Update ONNX to 1.7 release candidate 1. (#3782) 
* Function expansion support, Update ONNX to 1.7 release candidate 1.

* Renable disabled tests.
 2020-05-01 10:35:16 -07:00
edgchen1
047975e404
Address flaky test ReduceApiTest.Sum. (#3716) 
Increase test comparison tolerance. Add output of random seed value for easier debugging later. Unify RandomValueGenerator::Uniform() to consistently use [min, max) interval.
 2020-05-01 09:18:26 -07:00
Changming Sun
edd5855fb7 Remove eigen device from thread pool 2020-05-01 02:21:57 -07:00
George Wu
dcb1a21552
fix python package linux gpu failure (#3786) 
* pin base image for manylinux2010_gpu

* pin base image for Dockerfile.manylinux2010
 2020-05-01 17:04:59 +08:00
stevenlix
99ec93ea42
Apply onnx-tensorrt bug fixes (#3785) 
* merge latest onnx-tensorrt parser

* differentiate kernel names between graph and subgraph

* merge more TRT parser bug fixes

* merge more onnx-tensorrt bug fixes

* fix merge issue

Co-authored-by: stevenlix <stevenlix>
 2020-05-01 16:51:48 +08:00
Pranav Sharma
e42e0d4787
Update documentation + Update mlas threading lib to use the new TrySimpleParallelFor. (#3779) 2020-05-01 00:23:06 -07:00
pengwa
29234458af
disable cublasHgemm for training (#3769) 
* disable cublasHgemm for training
 2020-05-01 13:57:37 +08:00