This repository was archived by the owner on Aug 7, 2024. It is now read-only.
Used functional all-reduce for amax reduction #219
Closed
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
[ghstack-poisoned]
facebook-github-bot
added
the
CLA Signed
This was referenced Feb 16, 2024
added 3 commits
February 16, 2024 07:36
[ghstack-poisoned]
Dynamo cannot remap the in-place max all-reduce to its functional equivalent on PyTorch `main` branch. pytorch/pytorch#120082 [ghstack-poisoned]
|
I will hold off on landing this since @yifuwang will land the Dynamo rewrite fix soon. |
yifuwang
reviewed
Feb 27, 2024
| # https://github.com/pytorch/pytorch/issues/120082 | ||
| # Use functional all-reduce to avoid graph breaking. | ||
| amax = dist._functional_collectives.all_reduce( | ||
| amax, "MAX", list(range(dist.get_world_size())) |
There was a problem hiding this comment.
Ranks + tag as the process group identifier has been deprecated. Can we pass dist.group.WORLD or dist.group.WORLD.group_name here?
There was a problem hiding this comment.
If we are not landing this PR, then is it okay to leave the call as dist.all_reduce(amax, op=dist.ReduceOp.MAX) and wait for your Dynamo rewrite changes?
There was a problem hiding this comment.
Oh the changes you are referring to are for rewriting functional collective. The options I mentioned above should already work :)
Let me know if it doesn't though.
Sign up for free
to subscribe to this conversation on GitHub.
Already have an account?
Sign in.
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Outdated suggestions cannot be applied.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 (oldest at bottom):
modulearg fromfsdp_pre_all_gather#217use_activation_hooks: boolto swap #214amax_and_scale_syncedunconditionally #220Dynamo cannot remap the in-place max all-reduce to its functional equivalent on PyTorch
mainbranch.pytorch/pytorch#120082