saymrwulf/onnxruntime
1
0
Fork 0
mirror of https://github.com/saymrwulf/onnxruntime.git synced 2026-05-23 22:13:38 +00:00
Code Issues Projects Releases Packages Wiki Activity Actions
4886 commits 2856 branches 57 tags 1.3 GiB
Commit graph

4886 commits

This branch
This branch
All branches
Author SHA1 Message Date
baijumeswani
e161213f8e
Handle model with no parameters (#7736) 
* Handle model with no parameters

* Set the minimum module_output_grads as 0 to handle parameterless models
 2021-05-18 09:33:57 -07:00
Yulong Wang
96deec596f
fix npm test for webgl without wasm artifacts (#7742) 2021-05-18 09:19:19 -07:00
Xueyun Zhu
e92b3c1394
bumping up version number to 1.8 (#7733) 
* bump to 1.8

* fix windows AI
 2021-05-18 09:03:37 -07:00
Du Li
e4a985ff17
[JS/Web] WebGL Profiling Tool (#7724) 2021-05-18 06:31:00 -07:00
Hariharan Seshadri
43e2ee37f2
Some cosmetic changes (#7741) 2021-05-18 00:02:07 -07:00
stevenlix
a6972c8782
Fix issues in TensorRT provider options (#7738) 
* add legacy env variable support in pybind

* formating code
 2021-05-17 23:07:27 -07:00
Young Jin Kim
e9057d2e49
ZCode FastFormers changes (#5827) 
* Add FBGEMM submodule

* Add fbgemm based per-channel quantization

* Add missing logic for pre-layernorm transformer model fusion

* add support for structured pruning architecture -fastformers

* Fix windows build

* Add a default behavior when head_size is not present for the backward compatibility

* Remove FBGEMM and default to tensor-wise quantization, column-wise quantization will be enabled later

* Fixed some unit test errors

* Fix windows compile error and unit test errors

* delete the option removed from the upstream

* Addresses review comments and fixes a merge error

* Remove commented out code

* add non-zero zp support

* support A and B scale with any dimensions

* fix build breaks

* fix warning in MSVC

* Fix bug for not checking original float value names when treat it as not existing.

* Clean up head size

* Clean up python tools

* Enable per column quantization

* fix quant weight cleanup bug

* A few code clean up

* Some code clean-up

* Some code clean-up

* Change option name

* update default value

* Rename option and parameter names

* Missing argument name change

* Add tests for quantization options for attention and matmul

Co-authored-by: Yufeng Li <liyufeng1987@gmail.com>
Co-authored-by: Lei Zhang <zhang.huanning@hotmail.com>
 2021-05-17 21:12:21 -07:00
Changming Sun
38d90b0f15
Cleanup install_deps.sh (#7734) 2021-05-17 19:27:47 -07:00
Yufeng Li
9075488368
Add per-column support for QLinearMatMul (#7729) 
* Add per-column support for QLinearMatMul
 2021-05-17 18:57:42 -07:00
Du Li
d3c4b70ede
[Web/JS] Fixing two bugs in reshape_pack and im2col_pack (#7689) 
* fixing two bugs in reshape_pack and im2col_pack

* minor fix

* fix lint complaints
 2021-05-17 18:28:09 -07:00
Xueyun Zhu
79854dda8f
bumping up verson (#7731) 2021-05-17 17:55:42 -07:00
Jesse Benson
f977644324 ROCM support int reductions 2021-05-17 16:42:06 -07:00
Sunghoon
da5f24bd2d
Support additional session options and run options in WebAssembly (#7712) 
* add all session options and run options in C API except AddInitializer and AddFreeDimensionOverride

* remove unnecessary comment

* change extra session and run options to object notation

* resolve comments

* use an optional chaining for options

* resolve comments
 2021-05-17 14:57:19 -07:00
Tixxx
6d9f541442
[JS]moved logging level flag to global env (#7700) 
* moved logging level flag to global env

* added setter and getter for loggingLevel in Env

* moved implementation of env to a separate file
 2021-05-17 14:16:59 -07:00
Ye Wang
5e8086ad8e
Support fusions inside subgraphs in optimizer tool (#7701) 
* skip subgraph when updating model

* intreim checkin

* interim checkin 2

* support transformers optimizations in subgraph

* change more files

* fix comments typo
 2021-05-17 12:43:55 -07:00
Zhang Lei
ad95b19322
Help build pass on ubuntu 16.04 with old g++. (#7706) 2021-05-17 11:37:00 -07:00
Yufeng Li
0e48187b4e
Add type checks for QDQ transformer (#7715) 2021-05-17 10:48:20 -07:00
baijumeswani
c873f5589d
Fix bug where the output names were sorted lexicographically (#7709) 2021-05-17 10:27:20 -07:00
Thiago Crepaldi
6c41ed597b
Add custom autograd function to prevent input passthrough on ORTModule (#7694) 
* Changes for investigation

* Gradient for Identity

* Keep Identity betwen YieldOp and GraphOutput

* Revert debugging changes

* Add custom autograd fn to prevent input passthrough on ORTModule

* Add comment

Co-authored-by: Sherlock Huang <bahuang@OrtTrainingDev3.af05slrtruoetgaxwwjv5nsq5e.px.internal.cloudapp.net>
 2021-05-17 09:56:02 -07:00
Thiago Crepaldi
4fe2ffae16
Fix ORTModule python doc generation (#7704) 
* Fix ORTModule python doc generation

* Address comment
 2021-05-17 09:55:49 -07:00
Hariharan Seshadri
ebee380911
Partially revert PR #7348 (#7702) 2021-05-16 23:56:13 -07:00
Scott McKay
2b73163690
Exclude training specific logic from BatchNormalization to reduce binary size. (#7703) 2021-05-17 16:42:33 +10:00
stevenlix
557b94637d
Add more TensorRT env variables to provider options (#7698) 
* add all trt env variables to provider options

* add python test

* Update onnxruntime_c_api.h

* fix issues

* validate values for options
 2021-05-16 22:09:52 -07:00
Tracy Sharpe
943ab9dcef
undo clang-format corruption (#7720) 2021-05-16 15:32:30 -07:00
liqunfu
d604281a86
Liqun/training pkg to run tests (#7662) 2021-05-16 09:10:57 -07:00
liqunfu
3ead2f2f39
update pt lightning version (#7711) 
Co-authored-by: liqun <liqun@OrtTrainingDev4.af05slrtruoetgaxwwjv5nsq5e.px.internal.cloudapp.net>
 2021-05-15 21:46:16 -07:00
Yulong Wang
017b94ab36
Specify correct dependency for CI pipeline of nodejs binding (#7717) 2021-05-15 08:56:58 -07:00
Yufeng Li
6b0a7905ed
fix quant weight cleanup bug (#7707) 2021-05-14 22:04:35 -07:00
Sergii Dymchenko
9ba8da65d2
Fix BiasDropoutFusion when there are multiple Dropout consumers. (#7708) 
* Add tests for multiple consimers bias dropout fusion.

* Don't fuse resudial Add for multiple Dropout consumers.

* Remove duplicate code.

* Fix comment typo.

Co-authored-by: Derek Murray <Derek.Murray@microsoft.com>

Co-authored-by: Derek Murray <Derek.Murray@microsoft.com>
 2021-05-14 21:03:23 -07:00
Hariharan Seshadri
53d1d55ea8
Add ability for pre-packed weights of shared initializers to be shared across sessions (#7421) 2021-05-14 20:44:42 -07:00
Ashwini Khade
c78a40c0c7
add opset14 registrations for cuda provider (#7699) 
* add opset14 registrations for cuda provider

* update macros

* fix typo

* updates to rocm provider
 2021-05-14 19:08:45 -07:00
Yulong Wang
97d9bcd644
[js/web] fix bundle for multi-thread, add e2e test and support nodejs (#7688) 
* fix bundle for multi-thread, add e2e test and support nodejs

* add copyright banner

* resolve comments

* add comments for isMultiThreadSupported()
 2021-05-14 18:15:38 -07:00
Yufeng Li
a74e41e47d
Add non-zero zp support for quant matmul and attention (#7570) 
* add non-zero zp support
* support A and B scale with any dimensions
 2021-05-14 16:50:31 -07:00
Yulong Wang
c53b5be509
force multi steps to use the same commit in CI (#7697) 2021-05-14 15:13:38 -07:00
Zhang Lei
0f7721a019
Fix bug for not checking original float value names when treat it as not existing. (#7695) 2021-05-14 12:50:30 -07:00
Changming Sun
1d403ba03b
Fix a compile warning in EigenNonBlockingThreadPool.h (#7638) 2021-05-14 11:38:34 -07:00
Zhang Lei
033f0b3b7c
fix typo. (#7690) 2021-05-14 10:25:34 -07:00
Sunghoon
a9b47ca8e7
update debug.keystore for a public usage (#7696) 2021-05-14 09:54:54 -07:00
liqunfu
359fe1d197
Liqun/ort training version (#7620) 2021-05-14 09:54:19 -07:00
ashbhandare
bfbcc89db1
Add MLFloat16 support for SoftmaxCrossEntropyLoss for CUDA EP (#7679) 
* Forward op changes

* Add tests, improve kernel

* add opset 13 registration, remove unnecessary changes

* Add fp16 grad for SCELoss, review comments
 2021-05-14 09:00:27 -07:00
Changming Sun
39fac6d304
Fix a buffer reuse bug in allocation_planner.cc (#7645) 2021-05-14 07:31:35 -07:00
Ashwini Khade
442c7300eb
add opset14 rnn ops (#7687) 
* add opset14 rnn ops

* update kernel hashes
 2021-05-14 05:52:54 -07:00
Yufeng Li
3725d0211f
support maxpool QDQ fusion from opset 12 (#7693) 2021-05-13 23:58:15 -07:00
Zhang Lei
50c5edcf13
Add nhwc support for QLinearAveragePool operator (#7656) 
* Add nhwc support for QLinearAveragePool operator

* Update ContribOperators.md

* Update OperatorKernels.md with cpu,dnnl and cuda enabled.
 2021-05-13 22:05:30 -07:00
baijumeswani
37f69fcee5
Regain performance by caching initializer names in ORTModule (#7685) 2021-05-13 20:54:49 -07:00
Edward Chen
19704aedbb
Update Objective-C API (#7675) 
- Add session/run configuration
- Add additional supported tensor data types
- Clean up
 2021-05-13 18:47:22 -07:00
ashbhandare
56e993a434
Bump to rel-1.9.1 (#7684) 2021-05-13 18:41:28 -07:00
Xueyun Zhu
32d8278c2f
reshape fix (#7678) 2021-05-13 16:34:03 -07:00
raviskolli
4b37901f10
Aten support for rocm (#7680) 
* Aten support for rocm

* Removed aten_ops.cc as it is resued from cuda version
 2021-05-13 15:56:03 -07:00
Aswin John Mathews
4afdc19958
ROCm optimized layernorm for MI100 (#7682) 
* layernorm optimizations

* Changed HIP flag from HIP_VERSION to __HIP_PLATFORM_HCC__
 2021-05-13 15:54:06 -07:00