Use one param coordinator for both train/inference scenarios #6662
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
…soft/DeepSpeed into tohtana/simplify_param_coordinator
…soft/DeepSpeed into tohtana/simplify_param_coordinator
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Outdated suggestions cannot be applied.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.
The parameter coordinator in ZeRO3 throws a "backward pass is invalid for module in evaluation mode" error when the training mode is unexpected, as it expects all modules to be in training mode during the backward pass. This is an unnecessarily strict restriction.
This PR relaxes the restriction by using a single parameter coordinator (instead of separate ones for training and evaluation modes) and resetting the prefetch state before starting a forward pass.
Use of
is_compilingneeds to be fixed after #6663 is merged.