-
Notifications
You must be signed in to change notification settings - Fork 44
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
Add GPU performance tips for high performance LLMs with JAX/XLA #674
Conversation
Add GPU performance tips for high performance LLMs with JAX and XLA.
@abhinavgoel95, can you please review and provide feedback ? |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Happy to seem the flag being better document. Some small comments and suggestion.
@abhinavgoel95 or @nouiz, all conversations are resolved. Can one of you please approve, squash commits, and merge ? |
Remove default NCCL BUFF SIZE vars, add NCCL_NVLS_ENABLE
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Small issues. Most at nits, but the cuda_graph one must be verified.
@abhinavgoel95 anything else or I can merge now? |
Documenting all JAX and XLA flags that we use for high performance LLMs with JAX/XLA.
Will update this periodically, and also link to
Per Yu-Hang's suggestion, will update https://github.com/NVIDIA/JAX-Toolbox/blob/main/README.md#environment-variables to this page as well, once the PR is merged.