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

Training detail #65

Open
zhlhlhlhl opened this issue Dec 10, 2024 · 2 comments
Open

Training detail #65

zhlhlhlhl opened this issue Dec 10, 2024 · 2 comments

Comments

@zhlhlhlhl
Copy link

zhlhlhlhl commented Dec 10, 2024 •
edited
Loading

Hi, really impressive work! I'm curious why using zero3 in pretraining but zero2 in instruction training?
@jpthu17
Copy link
Member

jpthu17 commented Dec 11, 2024

We find using zero3 is very easy to hang because the lengths of the video and image data vary greatly.

@zhlhlhlhl
Copy link
Author

Yes, it's true. Besides, in the instruction training phase, I met CUDA out of memory error in 5258/8056. Are your experiments conducted on A100 GPUs with 40GB or 80GB

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
@jpthu17 @zhlhlhlhl