cherry picked commits for rel-1.8.1#8076
Merged
harshithapv merged 14 commits intorel-1.8.1from Jun 18, 2021
Merged
Conversation
* config parser, default argument values * ut * win build * maxpool2d * fix win build * fix build * unfold atenop
…nput (#7899) * add SoftmaxCrossEntropyLossInternal * bugfix and ut * fix ut * fix ut * support torch1.8.1 * function body for nll_loss_internal
* consolidate copy binary script for gpu/trt tarball package * add provider_options.h * add provider_options.h
* Save module output for backward if needed
…e precise. (#8018) * Address #7981 Reworked the logic around forcing a node to run on fp32 even if it was supported on fp16. The github issue had multiple factors. In ORT 1.8 we remove Identity nodes that produce graph outputs as they're not needed. That resulted in a Loop node no longer having output nodes (it produces graph outputs instead), which meant the check in IsSingleInputNodeFloat16Node returned true as there was no longer a downstream Identity node processing fp16 data. We shouldn't only force a node to fp32 in very specific circumstances, and the changes hopefully check for those more precisely.
* Merged PR 6093117: Fix test_DynamicQuantizedLinear_max_adjusted_expanded by allowing Identity operator to run on non-float inputs Motivation: As part of the OnnxConformance Backend tests, DynamicQuantizedLinear_max_adjusted_expanded is failing. Root Cause: - The test model has `Identity` operator as one of the node. The input of this node is of non-float data type. - In DML, `Identity` operator is registered as operator which requires floating input. - As per `DirectMLSchema.h`, support for non-float input has been added for `Identity` operator in DML but the same has not been reflected in the `OperatorRegistration.cpp`. Changes: - Removed all traces of the requiresFloatFormatsForGraph flag from it's definition and usage. This flag was only used for Identity and it's related operator. - Added null check for the graphOutput nodeArg in GraphDescBuilder.cpp to stop the crash of the test. Related work items: #33076298 * Merged PR 6103324: Remove usage of non-generic error code (FWP_E_NULL_POINTER) Motivation: Addressing Dwayne comment on the previous PR. [Ref: [6093117](https://dev.azure.com/microsoft/WindowsAI/_git/onnxruntime/pullrequest/6093117?discussionId=44292162&path=%2Fonnxruntime%2Fcore%2Fproviders%2Fdml%2FDmlExecutionProvider%2Fsrc%2FGraphPartitioner.cpp)] Changes: Inside the DML EP, we should not use some other platform specific error codes. Instead we should a appropriate generic error code. Related work items: #33076298 Co-authored-by: Sumit Agarwal <sumitagarwal@microsoft.com>
* Change full ort to mobile ort * Update Android example to load mobile ort * Change the format of test models to ort * update ios to use mobile ort * revise README * use onnxruntime-mobile-c CocoaPods in a npm package
should set PATH, not add to the tail the copy of PATH
* reduce optimization * bug fix * add a check * add ut * refactor * add ut cases for keepdims=true
harshithapv
requested review from
a team,
BowenBao,
baijumeswani,
liqunfu,
mrry,
spandantiwari and
thiagocrepaldi
as code owners
faxu
approved these changes
Jun 18, 2021
pranavsharma
approved these changes
Jun 18, 2021
Contributor
pranavsharma
left a comment
pranavsharma
left a comment
There was a problem hiding this comment.
I see a few training PRs that wouldn't qualify to be in a patch release. But I'll let the training team make that call.
Contributor
|
We also need this: #8066 |
Contributor
|
This time, please help us manually check if every GPU package has
|
Lafi7e
approved these changes
Jun 18, 2021
Contributor
@RyanUnderhill can you please help ensure this is the case? thanks! |
Contributor
Author
We will pick this in the next round. |
Contributor
Sure thing, should I run the packaging pipelines on this change or on the 1.8.1 branch? |
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
14 participants
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
The following 15 commits as of 06/16/2021 are cherry-picked in this PR for rel-1.8.1.
Reduce Kernel Optimization component:training-core release:1.8.1
#8067 by iK1D was merged 8 hours ago
• Approved
fix PATH addition in windows release:1.8.1
#8043 by iperov was merged 23 hours ago
• Approved
[js/react_native] Use a mobile ORT instead of a full ORT release:1.8.1
#8042 by hanbitmyths was merged 23 hours ago
• Approved
Fix Memory Leak from DlpackToOrtValue component:training-core release:1.8.1
#8029 by iK1D was merged 6 days ago
• Approved
Make logic in InsertCastTransformer around forcing a node to fp32 more precise. release:1.8.1
#8018 by skottmckay was merged 6 days ago
• Approved
Save module output for backward if needed component:ortmodule release:1.8.1
#8010 by SherlockNoMad was merged 6 days ago
• Approved
Zip Windows GPU Package missing CUDA provider files release:1.8.1
#8002 by RyanUnderhill was merged 7 days ago
• Approved
add missing provider_options.h in packages release:1.8.1
#7995 by jywu-msft was merged 8 days ago
• Approved
Update DirectML EP changes from DmlDev as of 2021-06-07 release:1.8.1
#7987 by sumitsays was merged 5 days ago
• Approved
Update CMakeLists.txt for openvino EP release:1.8.1
#7980 by snnn was merged 9 days ago
• Approved
Override ORTModule named_modules to support extra arg component:ortmodule component:training-frontend release:1.8.1
#7954 by baijumeswani was merged 8 days ago
• Approved
Cache initializers to avoid iterating over them on every forward component:ortmodule component:training-frontend release:1.8 release:1.8.1
#7905 by baijumeswani was merged 14 days ago
• Approved
Add SoftmaxCrossEntropyLossInternal to Support Dynamic ignore_index Input ( conflicts with #7937 and auto grad PR #7513 )
component:ortmodule component:training-core release:1.8.1
#7899 by iK1D was merged 8 days ago
• Approved
Change CMAKE_CUDA_STANDARD to C++17 for Windows GPU build release:1.8.1
#7883 by snnn was merged 15 days ago
• Approved
ATenOp Enhancement component:training-core release:1.8.1 (had conflicts with autograd PR - #7513)
#7725 by iK1D was merged 9 days ago
• Approved