Enhance new_group doc to mention using NCCL concurrently.
#48872
Conversation
This file contains 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
Using NCCL communicators concurrently is not safe and this is documented in NCCL docs. However, this is not documented in PyTorch and we should add documentation for ProcessGroupNCCL so that users are aware of this limitation. Differential Revision: [D25351778](https://our.internmc.facebook.com/intern/diff/D25351778/) [ghstack-poisoned]
pritamdamania87
requested review from
mingzhe09088,
mrshenli,
rohan-varma and
zhaojuanmao
as code owners
facebook-github-bot
added
cla signed
oncall: distributed
pritamdamania87
pushed a commit
that referenced
this pull request
Dec 5, 2020
Using NCCL communicators concurrently is not safe and this is documented in NCCL docs. However, this is not documented in PyTorch and we should add documentation for ProcessGroupNCCL so that users are aware of this limitation. Differential Revision: [D25351778](https://our.internmc.facebook.com/intern/diff/D25351778/) ghstack-source-id: 117932333 Pull Request resolved: #48872
💊 CI failures summary and remediationsAs of commit 411ba6d (more details on the Dr. CI page): 💚 💚 Looks good so far! There are no failures yet. 💚 💚 This 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 6 times. |
rohan-varma
approved these changes
Dec 5, 2020
There was a problem hiding this comment.
Thanks for clarifying this! I'm assuming "one process group at a time" means that collectives issued by one PG must be completely finished (i.e. not just enqueued, but actually executed by the GPU) before we can kick off a collective via another PG. Is it worth it to make this more explicit?
Codecov Report
@@ Coverage Diff @@
## gh/pritamdamania87/189/base #48872 +/- ##
===============================================================
- Coverage 80.74% 80.74% -0.01%
===============================================================
Files 1868 1868
Lines 201644 201644
===============================================================
- Hits 162823 162820 -3
- Misses 38821 38824 +3 |
Using NCCL communicators concurrently is not safe and this is documented in NCCL docs. However, this is not documented in PyTorch and we should add documentation for ProcessGroupNCCL so that users are aware of this limitation. Differential Revision: [D25351778](https://our.internmc.facebook.com/intern/diff/D25351778/) [ghstack-poisoned]
pritamdamania87
pushed a commit
that referenced
this pull request
Dec 8, 2020
Pull Request resolved: #48872 Using NCCL communicators concurrently is not safe and this is documented in NCCL docs. However, this is not documented in PyTorch and we should add documentation for ProcessGroupNCCL so that users are aware of this limitation. ghstack-source-id: 118060680 Differential Revision: [D25351778](https://our.internmc.facebook.com/intern/diff/D25351778/)
Using NCCL communicators concurrently is not safe and this is documented in NCCL docs. However, this is not documented in PyTorch and we should add documentation for ProcessGroupNCCL so that users are aware of this limitation. Differential Revision: [D25351778](https://our.internmc.facebook.com/intern/diff/D25351778/) [ghstack-poisoned]
pritamdamania87
pushed a commit
that referenced
this pull request
Dec 9, 2020
Pull Request resolved: #48872 Using NCCL communicators concurrently is not safe and this is documented in NCCL docs. However, this is not documented in PyTorch and we should add documentation for ProcessGroupNCCL so that users are aware of this limitation. ghstack-source-id: 118148014 Differential Revision: [D25351778](https://our.internmc.facebook.com/intern/diff/D25351778/)
|
This pull request has been merged in 7584161. |
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:
new_groupdoc to mention using NCCL concurrently. #48872 Enhancenew_groupdoc to mention using NCCL concurrently.Using NCCL communicators concurrently is not safe and this is
documented in NCCL docs.
However, this is not documented in PyTorch and we should add documentation for
ProcessGroupNCCL so that users are aware of this limitation.
Differential Revision: D25351778