Fix Train Step calculations for Checkpointing #279
Merged
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Outdated suggestions cannot be applied.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.
Currently, the checkpointing during training is implemented at the beginning of the train loop. This causes a training step overlap. For instance, consider a case when
checkpoint_every
is set to 100 and a checkpoint is saved at step 300. Then, a training resumed from the saved checkpoint will start at step 300, causing step 300 to be executed twice.This PR moves the checkpoint at the end of the training loop. Thus, when the training resumes from a checkpoint saved at step 300, it'd start from step 301.