saymrwulf/onnxruntime
1
0
Fork 0
mirror of https://github.com/saymrwulf/onnxruntime.git synced 2026-06-26 03:00:54 +00:00
Code Issues Projects Releases Packages Wiki Activity Actions
2551 commits 2856 branches 57 tags 1.3 GiB
Commit graph

2551 commits

This branch
This branch
All branches
Author SHA1 Message Date
Scott McKay
fd8ea4e466
Improve handling of symbolic dimensions in the onnxruntime_test.py script. (#3959) 
If a symbolic dimension is found allow the user to provide a value, or default to 1.

`python .\onnxruntime_test.py --symbolic_dims batch=1,seqlen=4 onnxruntime\test\testdata\transform\fusion\fast_gelu_use_graph_input.onnx`
 2020-05-18 16:51:09 +10:00
Tianlei Wu
523d70f667
Improve Transformer Benchmark for FP16 (#3970) 
Disable ORT in offline optimization script (ORT could generate some fused ops (like FusedGemm) which cannot be converted to fp16).
Remove some models from benchmark until we have optimizations for them.
 2020-05-17 21:50:45 -07:00
Wei-Sheng Chin
0d11649bb3
Address comments from #3823 and polish code (#3964) 
* Address comments from #3823 and polish code

* One line
 2020-05-17 14:08:33 -07:00
Prabhat
4ff73d00b0
Fix python pkg permission issue (#3957) 
* Fix python pkg permission issue

* Run chown with sudo

* Add workspace clean to arm pipeline

* Run docker as current user
 2020-05-17 14:06:55 +05:30
Tianlei Wu
07e9a4c164
Update benchmark to reflect those used in our latest results (#3967) 
Update optimizer for GPT2 models exported from PyTorch 1.5.
Update benchmark to use GPT2 models without Past State inputs/outputs
Update bert_perf_test to allow setting omp_num_threads etc to test only one setting
 2020-05-16 20:17:40 -07:00
Tianlei Wu
56700bec83
Add example of python code to readme of transformers tools (#3966) 
* Use shorter name for tools
* Use optimizer_cli
* Add comments about -i parameter
 2020-05-16 20:13:24 -07:00
Tianlei Wu
769c11f217
Update doc for transformers tools (#3963) 
* update readme for onnxruntime-tools package
* update license section in benchmark
 2020-05-16 11:13:34 -07:00
M. Zeeshan Siddiqui
a296b16719
Prevent divide by zero in CUDA implementation of SoftmaxCrossEntropyLossGrad. (#3962) 2020-05-16 00:33:25 -07:00
KeDengMS
132ce3a561
Fixes for quantizing a BERT from HuggingFace (#3939) 
* Fixes for quantizing a BERT from HuggingFace

* Address CR and some other minor fixes
 2020-05-15 23:41:29 -07:00
Wei-Sheng Chin
33208c9f6b
Modify Pipeline Facilities to Fix PipeDream Deadlock (#3823) 
* Prepare utils for adding Wait's and Record's

* Have a running PipeDream

* Add comments

* Polish comments

* Clean code

* Fix test

* Polish names

* Polish names

* Remove debug headers

* Fix a shape inference bug (not related to pipeline code)

* Fix a warning

* Address some comments

* Address comments

* Only touch consumers of outputs when re-wire edges
 2020-05-15 18:27:19 -07:00
edgchen1
999554cc53 CGManifest - add training entries and generate entries for submodules. (#3933) 
Add cgmanifest.json entries for training dependencies.
Add script to generate git submodule cgmanifest.json entries.
 2020-05-15 13:34:18 -07:00
ytaous
bc441b7e5c
Add cpu/mem usage for perf metrics (#3947) 
* add cpu/mem usage

* on comments

* on comments

* renaming

Co-authored-by: Ethan Tao <ettao@microsoft.com>
 2020-05-15 12:29:40 -07:00
Tracy Sharpe
be003dbab7
Fix ACL build break (#3952) 2020-05-14 23:14:15 -07:00
Pranav Sharma
47ae9691fd
Fix ordering of APIs. (#3951) 2020-05-14 21:27:46 -07:00
Hariharan Seshadri
9ef376880b
Add test for If node with conditional branches only containing Constant nodes (#3949) 2020-05-14 19:21:40 -07:00
Ryan Lai
38467f8c9a
DirectML Nuget package has different time stamp than Native and Managed Nuget (#3950) 
* Fix DirectML nuget creation in Nuget pipeline

* DirectML Nuget package has different timestamp

* remove accidentally changed file
 2020-05-14 18:52:08 -07:00
Jeff Bloomfield
e6da5946d1
Update DML Nuget version and DML EP Doc (#3945) 
Update DML Nuget version and DML EP Doc
 2020-05-14 17:33:46 -07:00
Tianlei Wu
782c6c24b2
Rename bert to transformers (#3946) 
* rename folder bert to transformers
* rename bert_model_optimization.py to optimizer.py
* update URL links in notebooks
 2020-05-14 15:32:59 -07:00
Zhang Lei
3c4f3d01cd
Implement QLinearLeakyRelu (#3648) 
* Implement QLinearRelu and its unit test.
* Add logic to compute table during constructor when all parameters is constant.
* Fix test case rounding result related with rounding mode.
 2020-05-14 14:52:55 -07:00
Scott McKay
5e0928a777
Enable running PEP8 on python scripts using flake8 (#3928) 
* Enable running PEP8 checks via flake8 as part of the build if flake8 is installed.
Update scripts in \tools and \onnxruntime\python. Excluding \onnxruntime\python\tools which needs a lot more work to be PEP8 compliant. Also excluding orttraining\tools for the same reason.
Install flake8 as part of the static_analysis build task in the Win-CPU CI so the checks are run in one CI build.
Update coding standards doc.
 2020-05-15 07:15:06 +10:00
Yulong Wang
39814319b2
[Node.js binding] fix linux build (#3927) 2020-05-14 12:37:32 -07:00
Yufeng Li
50f798dad6
support non-zero zero point for matmulinteger u8s8 (#3883) 
* support non-zero zero point for matmulinteger u8s8
 2020-05-14 12:02:28 -07:00
Changming Sun
9c989c8dd6
Update build doc for cross-compiling (#3672) 2020-05-14 11:43:06 -07:00
manashgoswami
cab21223b3
Updated TPN for OpenMPI and cleanup (#3932) 
* Update README.md

* Update ReleaseManagement.md

* Updated Third Party Notice for training feature

Added Open MPI license
 2020-05-14 11:42:44 -07:00
gwang-msft
cba8bdc790
Make some compile change for Android NNAPI provider using DNNLibrary (#3935) 
* Change compile settings for NNAPI with DNNLib

* update build.py

* update build readme
 2020-05-14 10:53:37 -07:00
Prasanth Pulavarthi
84c108a85e
link to folder instead of READMEs inside folder (#3938) 
otherwise hard to find the source code
 2020-05-14 07:35:23 -07:00
Ryan Lai
48f69cfbb8
Fix DirectML nuget creation in Nuget pipeline (#3929) 2020-05-13 19:34:38 -07:00
Chi Lo
f380460a9e
Update the build steps to support ORT on Jetson (#3869) 
* Update BUILD doc for ARM64 build for TensorRT support on Jetson device

* minor revision

* JetPack 4.4 is in developer preview stage, so we suggest to use JetPack
4.3
 2020-05-13 19:30:27 -07:00
ytaous
93eb9bcfde
Add yaml/perf scripts for new perf test pipeline (#3909) 
* yaml/perf scripts for new pipeline

* yaml/perf scripts for new pipeline

* remove unused imports

* testing some comments change

* testing some comments change

* testing jdbc

* testing jdbc

* testing jdbc

* exclude pwd from jdbc properties

* exclude pwd from jdbc properties

* namedtuple

* on comments

Co-authored-by: Ethan Tao <ettao@microsoft.com>
 2020-05-13 14:15:17 -07:00
Changming Sun
e86214e5c0
Fix the tensorflow performance test (#3847) 2020-05-13 11:52:59 -07:00
Yufeng Li
7c774e967a
support quantization of optimized model with ir<4 (#3853) 2020-05-13 11:16:37 -07:00
Prabhat
25257a661d
Added onnxruntime aarch64 wheel to pypi publishing pipeline (#3903) 
* Added onnxruntime aarch64 wheel to pypi publishing pipeline

* Support nightly build flag

* Add support for nightly build
 2020-05-13 23:20:29 +05:30
Jeff Bloomfield
1c1685ad2b
Fix error handling in LearningModelSession.cpp (#3920) 2020-05-13 10:13:16 -07:00
Tiago Koji Castro Shibata
385073e1cd
Fix DmlCopyTensor test (#3923) 
* Fix heap corruption

* Cleanup
 2020-05-13 09:14:55 -07:00
Zhang Lei
eab61e87ce
Fix quantization tool bugs when model nodes have no name. (#3854) 
Fix bugs when model nodes have no name.
 2020-05-12 20:38:26 -07:00
liqunfu
9b5daa2039
patch torch onnx opset 10 (#3910) 
patch pytorch to export onnx nll_loss opset version 10. add mnist test to covert onnx opset version 10.
 2020-05-12 18:11:25 -07:00
Ori Levari
7b858d60b0
Various changes for automated downlevel test pipeline (#3901) 
Co-authored-by: Ori Levari <orlevari@microsoft.com>
 2020-05-12 17:22:47 -07:00
Hariharan Seshadri
3065219cc1
Changes related to the release binaries requiring Visual C++ 2019 runtime (#3871) 2020-05-12 17:07:06 -07:00
Xiang Zhang
bccbdd03f1
User/xianz/enable batch tests (#3914) 
* enable batch tests in winml_image_test

* copy batchGroundTruth folder

* skip GPU tests when GPU is unavailable
 2020-05-12 15:46:46 -07:00
Dmitri Smirnov
18dc0ec39f
Rework jar by creating os-arch folders (#3849) 
Detect os and arch and move the artifacts to a new folder.
 Remove unnecesary jars so we cam focus on those we publish.
 Add signing
 Make signature simlper.
 Fix indent.
 Halt on 32-bit arch.
 Credits: @Craigacp
 2020-05-12 15:31:20 -07:00
Hariharan Seshadri
c00945ae81
Build ORT by default for Mac OS X versions 10.12+ (#3626) 2020-05-12 14:43:32 -07:00
Scott McKay
99415f09fe
Fix bug where linear_output_ is not cleared when linear_before_reset is true and no bias input is provided. Requires a batch size of 3 or more to trigger if initial_h is not provided. (#3893) 2020-05-13 07:18:32 +10:00
Andrews548
475ea382f9
Fix ACL EP convolution-activation fusion optimization (#3896) 
Co-authored-by: Andrei-Alexandru <andrei-alexandru.avram@nxp.com>
 2020-05-12 14:06:48 -07:00
Jeff Bloomfield
f170f31e1d
Extend workaround with input name matching in DML fused graph kernel (#3918) 2020-05-12 14:04:09 -07:00
Bowen Bao
0f82b42fed
Ensure pt model is set to cpu in ort_trainer (#3867) 
* Ensure pt model is set to cpu in ort_trainer

* add note comment
 2020-05-12 13:32:27 -07:00
Tianlei Wu
29496170d7
Add Benchmark Script for Bert Models (#3829) 
Add benchmark script for Transformer models
* Set intra_op_num_threads=1 for cpu (version <= 1.2.0)
* Add percentiles for latency
* torch.set_num_threads (for intra op) to get fair comparison
* Allow export ONNX model with specified number of inputs
* Add fusion statistics
* Install transformers from source
 2020-05-12 12:26:22 -07:00
Scott McKay
2fed37c8eb
Fix bug in handling of an initializer that provides a graph output. (#3912) 
* Outputs from model execution should always be returned in a newly allocated buffer or an pre-allocated buffer provided in fetches. When an initializer is providing a graph output (e.g. constant folding may result in this) we were returning an OrtValue that pointed to the initializer and not a separately allocated buffer with a copy.

This was wrong as:
  - value wasn't returned in a pre-allocated fetch so whilst the value returned was correct, it was returned in the wrong place
  - user could alter the data in the initializer via the returned value

* Add unit test with and without pre-allocated fetch.

* Add some extra info around why we're handling this special case.
 2020-05-12 20:42:58 +10:00
George Wu
6f729b100f
use LOAD_WITH_ALTERED_SEARCH_PATH for LoadLibraryExA (#3908) 2020-05-11 19:53:34 -07:00
Thiago Crepaldi
70abb120b3
Remove ORTModel from frontend API (#3825) 
* Resolve conflict

* Address review
 2020-05-11 18:20:33 -07:00
Tracy Sharpe
b12d35b679
MLAS: tune softmax kernels for partial vectors (#3906) 2020-05-11 18:02:50 -07:00