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.

Sign up for GitHub

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

Jump to bottom

RuntimeError: CUDA error: out of memory when training the model for too long. #13

Open
safranchik opened this issue Aug 30, 2022 · 2 comments
Open

RuntimeError: CUDA error: out of memory when training the model for too long. #13

safranchik opened this issue Aug 30, 2022 · 2 comments

Comments

@safranchik
Copy link
Contributor

There appears to be a memory leak during the validation routine since the GPU I'm using (A40, 48 GB VRAM) runs out of memory from training on the predict flu task for 125 epochs.
@TheMikeMerrill
Copy link
Member

We discussed this on slack, but just wanted to confirm that this is a known bug. I believe there's a memory leak somewhere in on_train_epoch_end but I haven't been able to find it.

@TheMikeMerrill
Copy link
Member

This isn't a complete solution, but I think the memory leak is happening in the metric bootstrapping during the validation loops. I can't figure out where specifically it's going down, but setting --model.val_bootstraps=0 after this commit (370d41f) should stop the leak.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Assignees
No one assigned
Labels
None yet
Projects
None yet
Milestone
No milestone
Development

No branches or pull requests
2 participants
@TheMikeMerrill @safranchik