-
Notifications
You must be signed in to change notification settings - Fork 7
Test Ruisi's commit #138
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
base: main
Are you sure you want to change the base?
Test Ruisi's commit #138
Conversation
[ghstack-poisoned]
[ghstack-poisoned]
ruisizhang123
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.
LGTM! thank you for adding tests.
examples/example_llama3.py
Outdated
| if True: | ||
| backend = "nccl" | ||
| fake_store = None | ||
| kwargs = {} |
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.
qq: is there a reason we want to use a real pg instead of fake pg? iiuc, the test job in github only has 4 gpus. 🤔 we have to update world_size thing to 4 if we decide to use real pg.
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 reason was because I thought the benchmarking might need a real PG, although I haven't really tested to see if it works with a FakeStore
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.
ohhhh, I see. I tried using fake PG and the code went through. The current llama3 tests are use fake PG, but there are new issues with memory estimation caused by a recent change in pytorch. I plan to land my autobucketing PR today, and will land this one after the pytorch issue is fixed.
f2dd1ef to
a368fbc
Compare
a368fbc to
f1cc5fc
Compare
No description provided.