Dataloader prefetch batch #28484
Closed
Dataloader prefetch batch #28484
Conversation
This file contains 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
…#27099) * nits * nits * actual fix * style * ze fix * fix fix fix style
* fix loss computation * compute on GPU if possible
…7973) * fix sdpa with non-contiguous inputs for gpt_bigcode * fix other archs * add currently comment * format
* Fix bug * Write test * Keep back old modification for grad accum steps * Whitespace... * Whitespace again * Race condition * Wait for everyone
|
Hi @qmeeus - thanks for opening a PR! At the moment, there's a very large diff including unrelated changes to those in the PR description. Could you make sure to resolve all conflicts and include the latest updates in |
Yes, this is a work in progress, I was not aware that this was available yet for reviewing. I will sort it out, thank you :) |
|
@amyeroberts I closed this one and created a new one available here |
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Outdated suggestions cannot be applied.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.
What does this PR do?
I added an option to the trainer to prefetch batches during data loading.
When training a model with heavy transformations and an iterable dataset, the dataloader might struggle to deliver fast enough for the GPU. I've found that prefetching batches helps to solve this issue.
The option is implemented in
torch.utils.data.DataLoaderbut not in HF Trainer.Before submitting
Pull Request section?
to it if that's the case.
documentation guidelines, and
here are tips on formatting docstrings.
Who can review?
Anyone in the community is free to review the PR once the tests have passed. Feel free to tag
members/contributors who may be interested in your PR.