[NCCL] Add Environment Variable to guard Async Error Handling feature #44163
Conversation
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
In this PR, we introduce a new environment variable (NCCL_ASYNC_ERROR_HANDLING), which guards the asynchronous error handling feature. We intend to eventually turn this feature on by default for all users, but this is a temporary solution so the change in behavior from hanging to crashing is not the default for users all of a sudden. Differential Revision: [D23517895](https://our.internmc.facebook.com/intern/diff/D23517895/) [ghstack-poisoned]
osalpekar
requested review from
mrshenli,
pietern and
zhaojuanmao
as code owners
This was referenced Sep 4, 2020
osalpekar
added a commit
that referenced
this pull request
Sep 4, 2020
In this PR, we introduce a new environment variable (NCCL_ASYNC_ERROR_HANDLING), which guards the asynchronous error handling feature. We intend to eventually turn this feature on by default for all users, but this is a temporary solution so the change in behavior from hanging to crashing is not the default for users all of a sudden. Differential Revision: [D23517895](https://our.internmc.facebook.com/intern/diff/D23517895/) ghstack-source-id: 111402543 Pull Request resolved: #44163
💊 CI failures summary and remediationsAs of commit 7109ebc (more details on the Dr. CI page):
ci.pytorch.org: 1 failedThis comment was automatically generated by Dr. CI (expand for details).Follow this link to opt-out of these comments for your Pull Requests.Please report bugs/suggestions on the GitHub issue tracker or post in the (internal) Dr. CI Users group. This comment has been revised 8 times. |
jiayisuse
reviewed
Sep 4, 2020
| } | ||
| } | ||
|
|
||
| void ProcessGroupNCCL::parseNcclAsyncErrorHandling() { |
Contributor
There was a problem hiding this comment.
This function looks good. Wondering if we can reuse parseNcclBlockingWait() somehow
Member
Author
There was a problem hiding this comment.
I think it should be doable, I've made an issue to follow up on this: #44205 (mainly because we might be able to make this helper function more broadly available in the PyTorch codebase)
Codecov Report
@@ Coverage Diff @@
## gh/osalpekar/81/base #44163 +/- ##
========================================================
- Coverage 69.24% 69.24% -0.01%
========================================================
Files 381 381
Lines 47573 47573
========================================================
- Hits 32943 32942 -1
- Misses 14630 14631 +1
Continue to review full report at Codecov.
|
jiayisuse
reviewed
Sep 8, 2020
…ing feature" In this PR, we introduce a new environment variable (NCCL_ASYNC_ERROR_HANDLING), which guards the asynchronous error handling feature. We intend to eventually turn this feature on by default for all users, but this is a temporary solution so the change in behavior from hanging to crashing is not the default for users all of a sudden. Differential Revision: [D23517895](https://our.internmc.facebook.com/intern/diff/D23517895/) [ghstack-poisoned]
osalpekar
added a commit
that referenced
this pull request
Sep 8, 2020
Pull Request resolved: #44163 In this PR, we introduce a new environment variable (NCCL_ASYNC_ERROR_HANDLING), which guards the asynchronous error handling feature. We intend to eventually turn this feature on by default for all users, but this is a temporary solution so the change in behavior from hanging to crashing is not the default for users all of a sudden. ghstack-source-id: 111614317 Differential Revision: [D23517895](https://our.internmc.facebook.com/intern/diff/D23517895/)
…ing feature" In this PR, we introduce a new environment variable (NCCL_ASYNC_ERROR_HANDLING), which guards the asynchronous error handling feature. We intend to eventually turn this feature on by default for all users, but this is a temporary solution so the change in behavior from hanging to crashing is not the default for users all of a sudden. Differential Revision: [D23517895](https://our.internmc.facebook.com/intern/diff/D23517895/) [ghstack-poisoned]
osalpekar
added a commit
that referenced
this pull request
Sep 9, 2020
Pull Request resolved: #44163 In this PR, we introduce a new environment variable (NCCL_ASYNC_ERROR_HANDLING), which guards the asynchronous error handling feature. We intend to eventually turn this feature on by default for all users, but this is a temporary solution so the change in behavior from hanging to crashing is not the default for users all of a sudden. ghstack-source-id: 111637788 Differential Revision: [D23517895](https://our.internmc.facebook.com/intern/diff/D23517895/)
Contributor
|
This pull request has been merged in 48c47db. |
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
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.
Stack from ghstack:
In this PR, we introduce a new environment variable
(NCCL_ASYNC_ERROR_HANDLING), which guards the asynchronous error handling
feature. We intend to eventually turn this feature on by default for all users,
but this is a temporary solution so the change in behavior from hanging to
crashing is not the default for users all of a sudden.
Differential Revision: D23517895