Fix resuming training from checkpoint#70
Open
marcandrelarochelle wants to merge 1 commit into
Open
Conversation
This file contains hidden or 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
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
Description
When saving checkpoints using Hugging-Face trl / Unsloth, the optimizer is saved, upon being loaded back, the
self.ortho_matrixisn't necessarily loaded on the right device (if you train on GPU, it will be loaded on CPU when unpickled).Bug
self.ortho_matrixis unpickled on the 'CPU' device, but there is no check to verify on the first pass if it should be on the 'CPU' after being unpickledSolution
Introduce a variable
self.resumeset to True when unpickling the class and check if theself.ortho_matrixwhen resuming training, if not send it to the same device as the weights (similar to the functionget_orthogonal_matrix).