Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Remove prefetch for megatron_gpt_sft_model.py #7274

Closed
wants to merge 1 commit into from
Closed

Conversation

athitten
Copy link
Collaborator

What does this PR do ?

Replaces super().validation_step in inference_step of megatron_gpt_sft_model.py with the validation_step from megatron_gpt_model but without prefetch as just one batch is passed to the validation_step in megatron_gpt_sft_model and prefetching causes an error.

Collection: [Note which collection this PR will affect]

Changelog

  • Add specific line by line info of high level changes in this PR.

Usage

  • You can potentially add a usage example below
# Add a code snippet demonstrating how to use this 

Before your PR is "Ready for review"

Pre checks:

  • Make sure you read and followed Contributor guidelines
  • Did you write any new necessary tests?
  • Did you add or update any necessary documentation?
  • Does the PR affect components that are optional to install? (Ex: Numba, Pynini, Apex etc)
    • Reviewer: Does the PR have correct import guards for all optional libraries?

PR Type:

  • New Feature
  • Bugfix
  • Documentation

If you haven't finished some of the above items you can still open "Draft" PR.

Who can review?

Anyone in the community is free to review the PR once the checks have passed.
Contributor guidelines contains specific people who can review PRs to various areas.

Additional Information

  • Related to # (issue)

1) Replace super.validation_step in inference_step with the validation_step from megatron_gpt_model but without prefetch as just one batch is passed to the validation_step in megatron_gpt_sft_model and prefetching causes an error.

Signed-off-by: Abhishree <[email protected]>
@github-actions github-actions bot added the NLP label Aug 18, 2023
@athitten athitten marked this pull request as ready for review August 18, 2023 23:08
@@ -400,7 +400,20 @@ def inference_step(self, dataloader_iter, batch_idx, mode, dataloader_idx=0):
self._reconfigure_and_process_inference_batch(batch, data_cfg)
# Meta data from dataset
metadata = batch.get('metadata', [{}] * len(batch['tokens']))
loss = super().validation_step(itertools.chain([batch]), batch_idx)
Copy link
Collaborator

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

if the issue happens because of prefetch-ing when len(dataloader) ==1, what happens during GPT pre-training when val set is small enough to be just 1 batch?

@athitten
Copy link
Collaborator Author

Closing as this is addressed in: #7318

@athitten athitten closed this Aug 29, 2023
@athitten athitten deleted the athitten/sft_fix branch August 29, 2023 20:54
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
Projects
None yet
Development

Successfully merging this pull request may close these issues.

None yet

2 participants