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

torch.float16 and torch.bfloat16 dtypes #45

Open
zhl606 opened this issue Feb 25, 2025 · 0 comments
Open

torch.float16 and torch.bfloat16 dtypes #45

zhl606 opened this issue Feb 25, 2025 · 0 comments

Comments

@zhl606
Copy link

zhl606 commented Feb 25, 2025

Hello,

  1. I encountered the issue described in the image, and I noticed that the script file “run_deepscaler_1.5b_8k” has already set “export VLLM_ATTENTION_BACKEND=XFORMERS”.
  2. I checked the torch_dtype of the loaded models and found that the Actor uses float32, while the reference model uses bfloat16, which likely caused the aforementioned problem. May I ask why they were not configured to use the same precision?

Image
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
1 participant
@zhl606