Draft
Conversation
fixed test issues formatted gated non-wrapped to FSDP1 updated for FSDP2 propagated changes to trainer added minor test fix formatted formatted once more addressed comments formatted minor fix
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
1 participant
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
Draft PR for supporting automicrobatching on FSDP2
This isn't added yet because we ran into some hiccups with how FSDP2 handles state transitions. As FSDP2 is stateful, it expects the program to stop when training runs into an OOM issue. Since we just restart training with a reduced microbatch size,
torch.distributed.fsdp._fully_shard._fsdp_common.TrainingStatecan be in an illegal state which can cause hangs/unexpected errors. Until we have a clear API for something like this, it would be finnicky to mess around with this state.