Commit graph

4896 commits

Author SHA1 Message Date
Ryan Hill
af3824ce25 Merge branch 'master' of https://github.com/microsoft/onnxruntime into ryanunderhill/cuda_shared 2021-05-06 17:00:59 -07:00
Pranav Sharma
bdb2ed7864
Revert "Add log to allow serving platforms to quantify ORT usage. (#7476)" (#7598)
This reverts commits da5c926, 4186233 and be2a304.
2021-05-06 16:21:32 -07:00
Ryan Hill
34511ff862 Fix typos 2021-05-06 15:15:05 -07:00
Ryan Hill
a826b0ca75 Test java fix 2021-05-06 14:42:54 -07:00
Tianlei Wu
d88da44066
Allow flexible order of Add inputs in Attention fusion (#7565) 2021-05-06 09:43:28 -07:00
ashbhandare
4896744638
Improve CPU node placement logic to avoid Memcpys (#7427)
* Modify CPU fallback logic

* Review comments, failing test

* Add test for topological order

* review comment

* Fix test for amd ci

* fix build

* Fix amd test
2021-05-06 09:39:34 -07:00
Ryan Hill
4fc2569224 ROCM 4 2021-05-06 03:34:39 -07:00
Ryan Hill
7d74a92575 ROCM fix 3 2021-05-06 03:16:11 -07:00
Ryan Hill
6878463b1a Merge branch 'master' of https://github.com/microsoft/onnxruntime into ryanunderhill/cuda_shared 2021-05-06 02:58:58 -07:00
Ryan Hill
be540afbea Fix submodule out of sync 2021-05-06 02:58:51 -07:00
Ryan Hill
b4b6c4fac2 Stupid fixes 2021-05-06 02:40:29 -07:00
Scott McKay
097bab8d1e
Cleanup a change to ExecutionFrame a little (#7576)
* Reduce the binary size growth from this change. Minimal build grew by 7KB from this checkin.

Firstly simplify the checking logic a little. Same checks are still done - just without using an extra layer of helpers.

The issue being addressed by the original change only applies if you have a graph output where the shape wasn't able to be inferred. e.g. Reshape node with dynamic input causes downstream shapes to be unknown. If that is not the case, MergeShapeInfo in graph.cc would have resolved any differences between a specified output shape and the inferred output shape during Graph::Resolve.

The issue does not apply to the execution frame used by the optimizer as  the only time it would create a graph output is if it could constant fold all the way through, so MergeShapeInfo would have handled any difference in that case as well.

Due to these considerations, wiring a logger in at the IExecutionFrame level isn't necessary if VerifyOutputSizes optionally overridden by an implementation that cares.

* Address PR comments
2021-05-06 19:29:34 +10:00
Ryan Hill
85e5fddd2a Put back Linux hack, it is necessary 2021-05-06 00:32:09 -07:00
Ryan Hill
3dfec5a339 ROCM fix v2 2021-05-06 00:20:01 -07:00
Pranav Sharma
be2a3046fe
Disable telemetry log for training builds (#7585) 2021-05-06 00:03:19 -07:00
Maajid khan
5413eaa5e4
Additional cmake changes for OpenVINO build (#7579)
->unsetting the CMAKE_MAP_IMPORTED_CONFIG that was
set for OpenVINO EP for Relwithdebinfo build on
windows.

Signed-off-by: MaajidKhan <n.maajidkhan@gmail.com>

Co-authored-by: suryasidd <surya.siddharth.pemmaraju@intel.com>
2021-05-05 23:54:53 -07:00
Ryan Hill
0853a58cfa ROCM fix 2021-05-05 23:29:17 -07:00
Ryan Hill
d74b1e8717 Fix another merge error 2021-05-05 23:27:03 -07:00
Ryan Hill
192d23060e Revert yml changes 2021-05-05 22:34:25 -07:00
Ryan Hill
d2f96e4c59 Cleanup + new training ops 2021-05-05 21:01:17 -07:00
Ryan Hill
e2856cfa21 Undo edit to file 2021-05-05 18:53:46 -07:00
Ryan Hill
605508a071 Comments 2021-05-05 18:42:05 -07:00
Ryan Hill
ad0dde555c Merge branch 'master' of https://github.com/microsoft/onnxruntime into ryanunderhill/cuda_shared 2021-05-05 18:41:45 -07:00
Ryan Hill
ac7679f6c1 Fix merge with master issues. 2021-05-05 17:25:03 -07:00
Edward Chen
830f0b45d0
Update Objective-C API (#7567)
Update Objective-C API to be more usable from Swift. E.g., to allow conversion from Objective-C methods with trailing NSError** parameter to throwing Swift methods.
Update CMake Objective-C framework setup.
2021-05-05 15:56:55 -07:00
Ryan Hill
8ba8c26233 Merge with master 2021-05-05 15:05:00 -07:00
Ryan Hill
0f3252e481 Merge conflicts 2021-05-05 15:01:26 -07:00
Ryan Hill
401751b4c7 Merge with master 2021-05-05 15:01:02 -07:00
baijumeswani
f3a70f1aec
Ignore invalid input argument to install_os_deps.sh (#7566) 2021-05-05 14:33:31 -07:00
Zhang Lei
9465948715
Quantization tools using one more extra_options on interface. (#7293)
handle nnapi special sigmoid options.
2021-05-05 13:51:50 -07:00
Changming Sun
a284eede64
Fix Linux CPU pipeline (#7584) 2021-05-05 13:26:10 -07:00
Ye Wang
8a9ddfe963
Longformer Attention non-determinism issue fix (#7574)
* Fix run-to-run not deterministic bug.

* Remove non-deterministic logic in softmax

* Fix value diff when removing non-deterministic issue.

Co-authored-by: Lei Zhang <zhang.huanning@hotmail.com>
2021-05-05 09:54:25 -07:00
Derek Murray
94c97ac8c2
Fix compiler warnings treated as errors in GistEncodeDecode. (#7568)
* Fix compiler warning in GistEncodeDecode.

* Fix other use of member variable.

* Make `compression_type_` const.

* Change floor to floorf in CUDA code.

* Statically cast size_t to int in GIST CUDA kernels

* Add explicit cast to `long` in gist.cc

Co-authored-by: Derek Murray <demurra@microsoft.com>
2021-05-05 09:05:11 -07:00
Sheil Kumar
91985ab03d
add use_dml (#7569)
Co-authored-by: Sheil Kumar <sheilk@microsoft.com>
2021-05-05 08:55:13 -07:00
Xavier Dupré
2f0479780e
Improves NonMaxSuppression on CPU (#7557)
* improves non max suppression
* use pointer instead of boxes
2021-05-05 09:14:40 +02:00
Xavier Dupré
ade6ed51eb
Speed up Reduce operators for consecutive reduced axes (#7206)
* Improves Reduction for three specific configurations
* Support ReduceMean
* add ReduceMax, ReduceMin
* refactoring
2021-05-05 09:14:00 +02:00
Ryan Hill
401ef4a634 Revert DllMain idea, it didn't work 2021-05-04 22:29:52 -07:00
Pranav Prakash
053bada30f
Add support for setting shape inference function on fused nodes (#7007)
* Add support for setting shape inference function on fused nodes
* Add test for fused node shape inference
2021-05-05 13:32:07 +10:00
Ryan Hill
b2c202d930 Remove extra debug messages
Try a more clean python shutdown through DllMain
2021-05-04 20:10:11 -07:00
Rachel Guo
d8cf960412
Add android test app to validate Java API for ORT-Mobile Android (#7477)
* test

* [gwang] make cmake compile work

* [gwang] enble build apks

* some build update

* add simple sigmoid test android project and cmake

* add build.py

* refine and remove unused import lib

* address CR comments

* remove unnecessary files

* add README.md

* minor update

* remove

* minor change

* fix ci failure and minor update

* fix typo in project folder

* remove

* remove and minor update

* refine

* minor fix

* fix

* fix typo

* add gradle spotlessApply task to fix CI failure

* fix

* enable spotlessApply in build gradle

* revert some changes

* minor fix

* run spotless apply for format

* address CR comments and fix CI version and format

* refine

* Refine

* address comments

* refine

* refine

* modify

* reformat

* resolve version conflicts

* minor update

* minor update

* address comments

* minor update

Co-authored-by: Guoyu Wang <wanggy@outlook.com>
2021-05-04 15:39:14 -07:00
Zhang Lei
f6cefc92e2
Add quantized value map after quantize input node added. (#7558) 2021-05-04 15:27:56 -07:00
Sergii Dymchenko
a647da3e1a
Fix 2 input Gemm grad (#7561)
* Add test for 2 input Gemm grad.

* Fix 2 input Gemm grad.
2021-05-04 12:00:14 -07:00
harshithapv
d812354ebd
Tile grad fix (#7556)
* tile grad fix

* code clean up
2021-05-04 11:16:26 -07:00
Guoyu Wang
e05528a365
Update Android AAR packaging pipeline script (#7559)
* update android package pipeline

* update shell script

* update script

* add kMSExperimentalDomain to reduction
2021-05-04 11:13:33 -07:00
Guoyu Wang
71ff6ff2ec
Disable NNAPI support for dynamic input shape, add warning logs (#7439)
* disable nnapi for graph with dynamic input shape

* Add warning for multiple paritions

* minor update

* update the message logging

* Fix coreml ci failure
2021-05-04 11:09:23 -07:00
Fanny Nina Paravecino
c3c4db2c1b
Upgrade GIST memory compression nodes, kernels, optimizer rule, and cli (#6262)
* Add gist nodes, kernels, optimizer rule, and cli

* Add Gist CUDA kernels

* Added/updated gist compression cli to bert, gpt2, mnist

* Fix decode priority generator for large models

* Fix hardcoded decode priority generator, update gist training test

* Fix incomplete if/else sequence for CI build

* Added MSFP15 for gist compression type

* fix Msfp15 bug

* Resolved azure pipeline errors - unsupported ORT_RETURN macro format, cudastream argument

* Resolved hardcoded cudastream argument, Pack8 zero error

* Resolved PR comments - except gist tests

* Added TypeInference to Gist Nodes, To attribute to Gist Decoder, Updated Gist Test Cases

* Reverted error in merge commit

* Updated logger usage in Gist rule, Updated GistPackMSFP15 compressed tensor's explaination

* Converted onnxruntime::make_unique to std::make_unique based on PR 7502

Co-authored-by: Fanny Nina Paravecino <faninapa@microsoft.com>
Co-authored-by: Aayush Ankit <aayushankit@microsoft.com>
Co-authored-by: Aayush Ankit <Aayush-Ankit@users.noreply.github.com>
Co-authored-by: Fanny Nina Paravecino <fanny.nina@microsoft.com>
2021-05-04 10:33:35 -07:00
Sherlock
c1ed647170
ORTModule enable run_symbolic_shape_infer by default (#7423)
* ORTModule enable run_symbolic_shape_infer by default

* Fix UTs by replacing Relu with Softmax

Co-authored-by: Sherlock Huang <bahuang@OrtTrainingDev3.af05slrtruoetgaxwwjv5nsq5e.px.internal.cloudapp.net>
2021-05-04 10:08:14 -07:00
Chi Lo
a94a893d5e
Update SessionOptions.cs (#7540)
Fix compile warning
2021-05-04 01:51:35 -07:00
Scott McKay
594dde2647
Validate that the conversion script from the python package can be used to convert models. (#7517) 2021-05-04 16:25:04 +10:00
sfatimar
898fff702c
compatibility was broken for myriad config parameter (#7349)
Co-authored-by: MaajidKhan <n.maajidkhan@gmail.com>
Co-authored-by: sfatimar <sahar.fatima@intel/com>
Co-authored-by: suryasidd <surya.siddharth.pemmaraju@intel.com>
2021-05-03 21:13:12 -07:00