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

GPU memory #7

Open
DhavalTaunk08 opened this issue Jun 4, 2022 · 3 comments
Open

GPU memory #7

DhavalTaunk08 opened this issue Jun 4, 2022 · 3 comments

Comments

@DhavalTaunk08
Copy link

Hi, can you please tell, how much gpu memory is required while finetuning the model? i am trying to finetune it using the Nivida 2080Ti with a memory of 12GB. But I am getting Cuda out of memory error.

@MarkusSagen
Copy link
Owner

Hi @DhavalTaunk08,
I would you mind expanding on some of your settings a bit?

When fine-tuning, what is your dataset? How large is your batch size? What do you set as the sequence length? Do you run it with mixed precision (fp16) and does your gpu have support and installed NVIDIA apex?

@DhavalTaunk08
Copy link
Author

Hi @MarkusSagen, I am using my own custom built dataset having same format as Wikisum dataset. I tried different batch sizes varying from 2 to 16. I have tried sequence length from 2 to 4096. I am using mixed precision (fp16) and the gpu is also well setup.

@MarkusSagen
Copy link
Owner

I think it is unlikely to work on with those specs unfortunately. xlm-r, just to initialize the weights and train for a minimal text sample is likely to take up more then 18gb I would estimate

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
None yet
Projects
None yet
Development

No branches or pull requests

2 participants