-
Notifications
You must be signed in to change notification settings - Fork 880
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
[Bug] DataLoader always return batches in the same order (Expected Random) #2157
Comments
I feel like it would be great to fix this quickly as it seems affect a lot of previous experiments that use the newer version of accelerate |
I have the same issue with accelerate version 0.24.1. |
I checked the master branch, the output seems fine. Does it mean the master branch fix it? (but may not be an elegant way |
Thanks for the update! We’ll be pushing a new release next week that should include the fix. #2126 |
Can confirm the master branch fixes this problem, while the latest release 0.24.1 still doesn't. |
Yes. For now, I think I switch to 0.23.0, it seems they haven't fixed this issue to update the version. |
Sorry for the delay on this, a release will be happening Friday with the fix included in it! |
This issue has been automatically marked as stale because it has not had recent activity. If you think this still needs to be addressed please comment on this thread. Please note that issues that do not follow the contributing guidelines are likely to be ignored. |
System Info
Information
Tasks
no_trainer
script in theexamples
folder of thetransformers
repo (such asrun_no_trainer_glue.py
)Reproduction
Using the following example script
Run the above in a distributed environment by
Expected behavior
It will output something like this, suppose we want different epochs have different order of batches, and different batches
But now is "same batch order" and "same batches" for two epochs
Quick fixes that we tried:
accelerate/src/accelerate/data_loader.py
Lines 836 to 846 in 0f2686c
The text was updated successfully, but these errors were encountered: