Skip to content
This repository has been archived by the owner on Oct 19, 2024. It is now read-only.

[FEATURE] Reduce the required peak RAM on a single node while converting weights #792

Open
zhanyuanucb opened this issue Dec 1, 2022 · 3 comments
Assignees
Labels
good first issue Good for newcomers

Comments

@zhanyuanucb
Copy link
Contributor

System information

  • Alpa version: v0.2.2
  • Are you willing to contribute to it (Yes/No): Yes, but not immediately

Describe the new feature and the current behavior/state
Referring to here, for now, weight conversion for OPT-175B requires a peak RAM usage as large as twice of the model size. It will be great to do this in a distributed way to reduce the required peak RAM on a single node.

Will this change the current API? How?
Changes will mostly happen in the step_2_consolidate_992_shards_to_singleton.py

Describe alternatives you've considered

Additional context

@zhisbug zhisbug added the good first issue Good for newcomers label Dec 19, 2022
@merrymercy merrymercy changed the title Reduce the required peak RAM on a single node while converting weights [Feature] Reduce the required peak RAM on a single node while converting weights Dec 20, 2022
@merrymercy merrymercy changed the title [Feature] Reduce the required peak RAM on a single node while converting weights [FEATURE] Reduce the required peak RAM on a single node while converting weights Dec 20, 2022
@zhisbug
Copy link
Member

zhisbug commented Dec 31, 2022

This is non-trivial to do. I discussed with @merrymercy and he will update on this issue.

@merrymercy
Copy link
Member

@Ying1123

@sammeralomair
Copy link
Contributor

Working on this.
I have a list of questions if anyone is available to disuss over slack

Sign up for free to subscribe to this conversation on GitHub. Already have an account? Sign in.
Labels
good first issue Good for newcomers
Projects
None yet
Development

No branches or pull requests

4 participants