-
-
Notifications
You must be signed in to change notification settings - Fork 11.7k
[Core][Test] move local_rank to the last arg with default value to keep api compatible #3711
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
Conversation
cadedaniel
left a comment
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.
Thanks for the quick turnaround
| env['LOCAL_RANK'] = str(i) | ||
| env['WORLD_SIZE'] = str(number_of_processes) | ||
| env['LOCAL_WORLD_SIZE'] = str(number_of_processes) |
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.
Should we also test the case where LOCAL_* is not present in env?
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.
The tests using ray already tested this, i.e. the default value of local_rank=-1.
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.
got it, thanks!
|
I can try bringing docker container multi node test next week! |
…project#3711) [Core][Test] move local_rank to the last arg with default value to keep api compatible (vllm-project#3711)
#3686 makes usage of
local_rank, but the API change is breaking. This PR modifies the API to be backward compatible, and add test for the local_rank argument.This is kind of smoke test, as we only test the multi-node code in single node. But that should be fine, because single-node program is also a special case of multi-node program.
#TODO when we find better solution to test multi-node, we can actually test it in multi-node environment.
cc @esmeetu @cadedaniel @simon-mo