[distributed] Fix early stopping and DP
#254
Merged
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
|
The documentation is not available anymore as the PR was closed or merged. |
edbeeching
reviewed
Mar 28, 2023
|
If it looks good for @edbeeching I am happy to merge it. |
|
Should have addressed the last comment by now! |
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Outdated suggestions cannot be applied.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.
What does this PR do?
This PR fixes an issue where the training was hanging in multi-GPU setting and using earyl stopping
In fact, in a multi-GPU setting, we need to compute the average of
policyklacross all GPUs, and call early stoppingI can confirm this solution worked, the way I checked was to add a print after the
self.optimizer_zero_gradand forcing the coefficient to be0:if policykl > 0 * self.config.target_kl:cc @edbeeching @lvwerra