Fix gradient checkpointing bugs in freezing part of models (requires_grad=False)#3404
Merged
patrickvonplaten merged 5 commits intohuggingface:mainfrom May 17, 2023
Merged
Conversation
Loading
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
4 participants
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
Using
torch.utils.checkpoint.checkpointdirectly will cause the parameters in the checkpoint section to not be learned when part of the model parameters are freezed. As these discussions state:https://discuss.pytorch.org/t/use-of-torch-utils-checkpoint-checkpoint-causes-simple-model-to-diverge/116271
https://discuss.pytorch.org/t/checkpoint-with-no-grad-requiring-inputs-problem/19117/19
In pytroch versions larger than 1.11.0, the ```use_reentrant=False`` can be added to fix this bug.