fix(train): save model checkpoint before evaluation to prevent losing results on eval crash#609
Open
aniruddhaadak80 wants to merge 1 commit into
Open
Conversation
… results on eval crash (karpathy#7)
There was a problem hiding this comment.
Pull request overview
Note
Copilot was unable to run its full agentic suite in this review.
Adds a “pre-eval” checkpoint dump right before final evaluation, and deletes it afterward to keep the workspace clean if evaluation completes.
Changes:
- Save
model.state_dict()to a fixedcheckpoint_preeval.ptbefore final eval - Remove the checkpoint file after eval completes
💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.
Comment on lines
+612
to
+613
| checkpoint_path = "checkpoint_preeval.pt" | ||
| torch.save(model.state_dict(), checkpoint_path) |
Comment on lines
614
to
+619
| with autocast_ctx: | ||
| val_bpb = evaluate_bpb(model, tokenizer, DEVICE_BATCH_SIZE) | ||
|
|
||
| # Clean up checkpoint if evaluation succeeded | ||
| if os.path.exists(checkpoint_path): | ||
| os.remove(checkpoint_path) |
Comment on lines
+618
to
+619
| if os.path.exists(checkpoint_path): | ||
| os.remove(checkpoint_path) |
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
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
Saves model state dict to checkpoint_preeval.pt right after the training loop finishes and before evaluation begins. The checkpoint is deleted on successful evaluation, preserving it on evaluation crashes or OOMs for recovery/debugging (Fixes #7).