[Train] Disable gathering the full state dict in RayFSDPStrategy for lightning>2.1
#44569
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
Signed-off-by: woshiyyya <xiaoyunxuan1998@gmail.com>
justinvyu
reviewed
Apr 8, 2024
Comment on lines
110
to
112
| # Lightning < 2.1 lacks FSDP state_dict_type support. | ||
| # (PR: https://github.com/Lightning-AI/pytorch-lightning/pull/17623). | ||
| # We need this patch logic to enable FSDP checkpointing between 2.0 and 2.1. |
There was a problem hiding this comment.
Should we just force users to upgrade to versions outside of this range? It's a bit confusing for the behavior to be hardcoded to full state dict ckpt based on the library version.
There was a problem hiding this comment.
Without this fix, the state dict of checkpoint in lightning 2.0.x will be empty.
After offline discussion, we will not raise an error since the hardcoded gathering logic doesn't contradict with the lightning behavior. Instead, we add a notice in RayFSDPStrategy docstring to recommend users upgrade to beyond 2.1 if they want to use FSDP.
Signed-off-by: woshiyyya <xiaoyunxuan1998@gmail.com>
justinvyu
approved these changes
Apr 8, 2024
Co-authored-by: Justin Yu <justinvyu@anyscale.com> Signed-off-by: Yunxuan Xiao <xiaoyunxuan1998@gmail.com>
Signed-off-by: woshiyyya <xiaoyunxuan1998@gmail.com>
…odify_ray_fsdp_strategy
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Outdated suggestions cannot be applied.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.
Why are these changes needed?
lighting 2.0.x does not natively support FSDP
state_dict_type. Therefore, we added default state dict gathering logic (#34967) to enable FSDP checkpointing. After 2.1, Lightning inherently supports FSDPstate_dict_type, so we no longer need this patch logic.This PR restricts the patch's applicability to Lightning versions 2.0 through 2.1, enabling users to leverage Lightning's native FSDP integration in versions beyond 2.1.
Related issue number
Closes #44501
Checks
git commit -s) in this PR.scripts/format.shto lint the changes in this PR.method in Tune, I've added it in
doc/source/tune/api/under thecorresponding
.rstfile.