Update auto wrap policy and remove duplicate load in trainer.fit #175
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
abhibyreddi
approved these changes
Nov 4, 2024
jdnurme
approved these changes
Nov 4, 2024
…h into mirvine/fsdp5
Yash9060
approved these changes
Nov 4, 2024
abhibyreddi
pushed a commit
that referenced
this pull request
Nov 8, 2024
abhibyreddi
pushed a commit
that referenced
this pull request
Nov 8, 2024
Yash9060
pushed a commit
that referenced
this pull request
Nov 8, 2024
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Outdated suggestions cannot be applied.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.
This updates the benchmark to use the FSDP wrap policy defined in https://lightning.ai/docs/pytorch/stable/advanced/model_parallel/fsdp.html#identify-large-layers.
Testing on a machine with 4 GPUs, before this change I get a CUDA OOM at a certain number of parameters for my model, and after the change I do not.
I also removed the checkpoint restore path from the call to
trainer.fitduring load since this was making the trainer restore from that checkpoint an extra time, which is time consuming, and also explicitly deleting trainer/strategy/model prior to resetting them in case it helps prevent memory leaks.