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

Which version of transformer engine should I use, when I try to open ub_tp_comm_overlap? #11683

Open

sallyjunjun opened this issue Dec 20, 2024 · 0 comments

Assignees

sallyjunjun commented Dec 20, 2024

I am using NeMo with version v2.0.0rc0. When I set ub_tp_comm_overlap to true with tp and sp 2, I met the following error:

The version of transformer engine 1.6.0+c81733f. Should I update to newer te version?

When I update transformer engine to 1.13.0+e5edd6c. There occurs another error in NeMo:

CUDA 11.8 should be used in NeMo. But transformer engine in 1.13.0+e5edd6c version requires CUDA newer than 12.0.

I'm stuck in these version issues.
Could you please tell me which version of NeMo and TE and CUDA should I use to enable tp_comm_overlap feature?

sallyjunjun assigned okuchaiev

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment