saymrwulf/onnxruntime
1
0
Fork 0
mirror of https://github.com/saymrwulf/onnxruntime.git synced 2026-06-01 23:30:35 +00:00
Code Issues Projects Releases Packages Wiki Activity Actions
3425 commits 2856 branches 57 tags 1.3 GiB
Commit graph

3425 commits

This branch
This branch
All branches
Author SHA1 Message Date
gwang
a6f092300a Fix for CUDA build break 2020-10-12 13:27:37 -07:00
gwang
90e132c5d5 Cherry-pick 5378 2020-10-12 13:27:37 -07:00
Guoyu Wang
c6c4bab7ad Move flatbuffers to 1.12 release (#5392) 2020-10-12 13:27:37 -07:00
stevenlix
bcbcb2552a update onnx-tensorrt submodule (#5442) 2020-10-12 13:27:37 -07:00
Hariharan Seshadri
8206ff4a90 Support trilinear sampling in Resize CPU and CUDA kernels (#5300) 2020-10-12 13:27:37 -07:00
Changming Sun
e4f71abd90 Exclude GPT2_LM_HEAD from OpenVino's model test list (#5356) 
GPT2_LM_HEAD is a new ONNX model zoo model that OpenVino doesn't support.

Error message:1: [ONNXRuntimeError] : 6 : RUNTIME_EXCEPTION : Non-zero status code returned while running OpenVINO-EP-subgraph_1162 node. Name:'OpenVINOExecutionProvider_OpenVINO-EP-subgraph_1162_1' Status Message: _Map_base::at
 2020-10-12 13:27:37 -07:00
Tiago Koji Castro Shibata
d77241e7f1 Fix WinML warnings (#5228) 2020-10-12 13:27:37 -07:00
Tianlei Wu
4227dd7df5 Revert "Move flatbuffers to 1.12 release (#5392)" 
This reverts commit 0294d9aa5624b463ef34aa3ad8458e983715e2a4.
 2020-10-12 13:27:37 -07:00
Tianlei Wu
fc0fc80db2 Revert "Add flatbuffers verifier for ORT format buffer (#5378)" 
This reverts commit e8bf3ba2bb383055b5974e8904324c33e0f4cbb1.
 2020-10-12 13:27:37 -07:00
Guoyu Wang
6782866529 Mitigate pybind11 build break using Xcode 12 on macOS (#5381) 2020-10-12 13:27:37 -07:00
Guoyu Wang
fda4363992 Add flatbuffers verifier for ORT format buffer (#5378) 2020-10-12 13:27:37 -07:00
Pranav Sharma
5f331af157 Include config keys header file in the release packages for Linux and Mac. (#5388) 2020-10-12 13:27:37 -07:00
Guoyu Wang
8adfa7ac70 Move flatbuffers to 1.12 release (#5392) 2020-10-12 13:27:37 -07:00
Tiago Koji Castro Shibata
8283526541 Fix com ptr refcount (#5404) 2020-10-12 13:27:37 -07:00
Tianlei Wu
58bf508ce0 bump version to 1.5.2 (#5420) 2020-10-12 13:27:37 -07:00
Tianlei Wu
4e983634ff clear cudaDelayLoadedLibs since delayload is disabled (#5386) 2020-10-12 13:27:37 -07:00
Yufeng Li
5de47affb1
fix quantization of EmbeddingLayerNorm (#5321) 2020-09-29 01:00:47 -07:00
Tianlei Wu
c00e13a291
Cherry pick (batch 2) to rel-1.5.1 (#5290) 
* remove implicit linking of tensorrt and dnnl ep shared libs (#5262)
* Update DirectML Nuget to 1.3.0 (#5274)
* Update PyTorch TransformerModel sample (#5275)
* Insert telemetry template into GPU build, add telemry build switches. (#5278)
* Synchronize training dependency versions between Docker image and Python wheel (#5261)
* Downgrade GCC (#5269)
* Remove --enable_symbolic_shape_infer_tests to fix linux ci pipeline build error.

Co-authored-by: Edward Chen
Co-authored-by: George Wu <jywu@microsoft.com>
Co-authored-by: Dwayne Robinson <dwayner@microsoft.com>
Co-authored-by: Thiago Crepaldi <thiago.crepaldi@microsoft.com>
Co-authored-by: Dmitri Smirnov <yuslepukhin@users.noreply.github.com>
Co-authored-by: edgchen1 <18449977+edgchen1@users.noreply.github.com>
Co-authored-by: Changming Sun <chasun@microsoft.com>
 2020-09-25 09:26:40 -07:00
Jeff Bloomfield
389cca7a45 Handle missing initializers in allocation planner to fix crashes with DML provider (#5244) 
* Fix memory planning bug with DML EP

* Address PR comments

* Fix typo
 2020-09-23 16:50:58 -07:00
Dwayne Robinson
b648fe5f74 ORT DirectML EP for Iron release, ONNX 1.5 (part 2) (#5263) 
* Merged PR 5195856: Fix broken cases of zero size tensors in Cast/Reduce

 MaskRCNN failed when `Cast` tried to execute `Xor` with emptiness (zero in dimensions). This is perfectly legal and should be treated as a nop.

Ultimately DML itself should treat this case as a nop, just like how C's `memcpy` treats 0 count as a nop, but I'm just addressing it in ORT now, as enabling it in DML would impact more operators to be consistent (probably should incrementally add a flag to tensor validation so operators can be opted in gradually).

Corresponding WindowsAI PR: https://microsoft.visualstudio.com/WindowsAI/_git/WindowsAI/pullrequest/5195850

Related work items: #27469839, #28761382

* Merged PR 5201369: Remove copy of initializers added in DMLXP refactor

When used in ORT, a common method shouldn't copy and return initializer data

Related work items: #29514403

Co-authored-by: Justin Stoecker <justoeck@microsoft.com>
Co-authored-by: Jeff Bloomfield <jeffbloo@microsoft.com>
 2020-09-23 16:50:58 -07:00
Yufeng Li
eb75b492cc Fix bug in the back to back quantization of matmul and conv (#5264) 
* fix bug in the back to back quantization of matmul and conv

* fix bug in back to back gather
 2020-09-23 16:50:58 -07:00
Tianlei Wu
47447da4fd bump version to 1.5.1 (#5258) 2020-09-23 16:50:58 -07:00
Ye Wang
87b15f32ef Fix reshape fusion crash (#5252) 
* fix reshape fusion crash

* handling start_node statelessly

* fix
 2020-09-23 16:50:58 -07:00
Guoyu Wang
fc259de3bc Fix possible ios build break after update to Xcode 12 (#5246) 
* Fix possible ios build break after update to Xcode 12

* Address comments
 2020-09-23 16:50:58 -07:00
Sherlock
9fd76c8693 Place Shape's output in CPU memory (#5245) 
Co-authored-by: Sherlock Huang <bahuang@OrtTrainingDev3.af05slrtruoetgaxwwjv5nsq5e.px.internal.cloudapp.net>
 2020-09-23 16:50:58 -07:00
edgchen1
9158679c43 Update BUILD.md training dependency info. (#5240) 
Update training dependency versions based on Dockerfile.training.
 2020-09-23 16:50:58 -07:00
Changming Sun
b9b7c279fa Update BUILD.md for CUDA versions (#5239) 2020-09-23 16:50:58 -07:00
George Wu
0cbe240ea3 update TensorRT docs (#5238) 
* doc updates TensorRT

* update

* update

* fix warning

* newline

* format
 2020-09-23 16:50:58 -07:00
Scott McKay
c93f292d1f Revert to using release SafeInt repo now that it supports a build with exceptions disabled. (#5233) 2020-09-23 16:50:58 -07:00
edgchen1
6371ad61c5 Fix TransposeScaleMatMul and MatMulScaleFusion issues (#5230) 
- Rename TransposeScaleMatMul back to TransposeMatMul for backwards compatibility
- Fix MatMulScaleFusion issues:
  - Add check for supported execution providers
  - Add check for supported MatMul input types
 2020-09-23 16:50:58 -07:00
stevenlix
c27f461c1d Create profile for all dynamic shape input tensors (#5229) 2020-09-23 16:50:58 -07:00
Adam Pocock
4427b1e2a3 [java] Fixing the buffer semantics. (#5223) 
* [java] Fixing the buffer semantics.
* Renaming bufferCapacity to bufferRemaining.
* Adding a cast to char* so the pointer arithmetic works on Windows.
 2020-09-23 16:50:58 -07:00
George Wu
c909c67701 fix _WIN32 (#5218) 2020-09-23 16:50:58 -07:00
Scott McKay
95b2e31659 Update conversion script and process to simplify creating ORT format models and a minimal build (#5217) 
* Update conversion script and process to simplify creating ORT format models and a minimal build.
 2020-09-23 16:50:58 -07:00
liqunfu
21a7afb2c6 --shm-size=1024m to fix nccl shared memory issue (#5214) 
* --shm-size=256m to fix nccl shared memory issue

Co-authored-by: liqun <liqun@OrtTrainingDev4.af05slrtruoetgaxwwjv5nsq5e.px.internal.cloudapp.net>
 2020-09-23 16:50:58 -07:00
RRRachelllll555
b791402f84 Remove shape inference and fix save large model(>2g) issue (#5210) 
* remove shape inference and fix save large model problem

* remove unnecessary import

* refine code and add external format for quantize_qat

* remove initializers in tensors_to_calibrate

* small refine

Co-authored-by: t-yguo <t-yguo@microsoft.com>
 2020-09-23 16:50:58 -07:00
Pranav Prakash
0a31b9ed3c Fix order of returned values in quantize_weight_per_channel (#5205) 
Must match returned order of `quantize_inputs`
 2020-09-23 16:50:58 -07:00
Tracy Sharpe
f726af34e0 NCHWc optimizer fixes for quantized models (#5203) 
This updates the NCHWc transformer to not interfere with quantized convolution models, based on observations from internal models. The tensor type for MaxPool must be float. The input to GlobalAveragePool/GlobalMaxPool must be in NCHWc format.
 2020-09-23 16:50:58 -07:00
S. Manohar Karlapalem
84ffdbc467 Corrects doc typos and formatting (#5201) 2020-09-23 16:50:58 -07:00
Pranav Sharma
24d111c342 Add API to allow configuration of the global thread pools. (#5199) 2020-09-23 16:50:58 -07:00
Zhang Lei
498483b464 MaxPool versioning in quantization tools. (#5194) 
MaxPool versioning in quantization tools.
 2020-09-23 16:50:58 -07:00
Suffian Khan
39a7f96a44 Fix softmax_warp_backward math when is_log_softmax = True and register LogSoftmax CUDA kernel (#5160) 
* register logsoftmax cuda kernel; fix logsoftmaxgrad cuda kernal; fix tests to invoke dispatch_softmax_*

* forgot to remove axis check

* add tests all axis

Co-authored-by: suffian khan <sukha@OrtTrainingDev1.af05slrtruoetgaxwwjv5nsq5e.px.internal.cloudapp.net>
 2020-09-23 16:50:58 -07:00
Shucai Xiao
8e650c5384 Amdmigraphx improvements (#5158) 
* code backup

* remove unnecessary log info

* code backup

* code backup

* merge changes from master branch

* code backup

* code backup

* merge changes from master branch

* code backup

* code backup for constant folding enhancement

* code backup

* include more scenarios for constant folding

* code backup

* remove unnecessary code

* remove unnecessary log information

* fix an error in comments

* update algorithm to do graph partition

* code backup

* remove unnecessary log information

* remove an unused function

* remove unnecessary changes
 2020-09-23 16:50:58 -07:00
Ye Wang
b693cb1370 Fix a bug in EmbedLayerNorm fusion (#5150) 
* fix embedlayernorm bug

* review comments

* interim checkin

* review comments

* Fix core dump in MacOS

* remove unnecessary lines

* update document

* Update graph_utils.cc

* Update onnx_exporter.py

* resolve comments
 2020-09-23 16:50:58 -07:00
Changming Sun
5b5bcba9e3 Update MCR CUDA docker image to 10.2 (#5181) 2020-09-17 08:39:47 -07:00
Dmitri Smirnov
ece9a7c1fc Refactor TensorAt, prepare for release (#5180) 
* Refactor TensorAt
  locations* must be const and int64_t since our dims are int64_t
  Remove unnecessary copy of locations.
  Remove unnecesary casting and C-casting. Simplify implementation.
  Add a check for string type.
  Make CXX api return T& to fully expose C API in C++, const std::vector& by value as it
  covers more ground and eliminate redundant copy.
  Eliminate inner loop, compute strides first.
 2020-09-17 08:39:47 -07:00
Tracy Sharpe
b2994492af MLAS: add sgemm weight prepacking (#5183) 
Add support to MLAS to prepack weights for the float GEMM. Support for prepacking has been added to MatMul and Attention for this release.
 2020-09-17 08:39:47 -07:00
Tiago Koji Castro Shibata
ecf04d23c4 Fix nuget build (#5163) 
* Fix nuget content

* Revert "Fix nuget content"

This reverts commit e2cdcec4e39964c50eac2fb306c7a4bb84352443.

* Nuget packaging

* skip tests

* msbuild path

* Force msbuild version

* Workaround https://github.com/NuGet/Home/issues/7621

* cleanup
 2020-09-17 08:39:47 -07:00
Tiago Koji Castro Shibata
b523fa08bc Use onecore umbrella lib in onecore builds (#5182) 
* delayload hack

* Skip tests

* Onecore uses onecore umbrella

* Uncomment tests

* cleanup

* Disable dev mode for WinML
 2020-09-17 08:39:47 -07:00
Chun-Wei Chen
393ff2f434 Add GetStartTime() for profiler to get private profiling_start_time_ (#4994) 
* add GetStartTime() for profiler

* add function in inference_session

* remove qualified name

* add the api in cxx_api.h

* rename starttime to StartTimeNs, expost profiling object

* rename GetProfilingStartTime

* move Ortapis to the right place

* move to the end

* add const for session

* const the right place

* use const auto instead of const auto* for session

* remove const for auto getstarttime

* remove const for auto getstarttime

add unit tests

* nit: update test name and add comments
 2020-09-17 08:39:47 -07:00