Skip to content

Conversation

samsja
Copy link
Member

@samsja samsja commented Sep 10, 2024

poc of pure fsdp diloco

@Jackmin801
Copy link
Member

lgtm! a lot cleaner. just need to make sure it has loss curve parity with the previous implementation :)

@samsja
Copy link
Member Author

samsja commented Sep 11, 2024

Screenshot from 2024-09-11 18-48-46
Screenshot from 2024-09-11 18-49-17

check against our hivemind impl, seems to be equivalent

MixedPrecision,
)
from torch.distributed.device_mesh import init_device_mesh
from hivemind.optim.optimizer import logger
Copy link
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

I think we might not need this logger from hivemind?

Copy link
Member Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

we do, otherwise we don;t have a nice logger

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
None yet
Projects
None yet
Development

Successfully merging this pull request may close these issues.

2 participants