Skip to content
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.

Sign up for GitHub

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

Jump to bottom

LayerNorm Refactor #1269

Merged
merged 37 commits into from
Sep 9, 2024
Merged

LayerNorm Refactor #1269

merged 37 commits into from
Sep 9, 2024

Conversation

aurelion-source
Copy link
Contributor

@aurelion-source aurelion-source commented Sep 6, 2024 •
edited
Loading

Adds:

  • Apex's rmsnorm
  • TransformerEngine's rmsnorm and layernorm

@aurelion-source aurelion-source self-assigned this Sep 6, 2024
@aurelion-source
Copy link
Contributor Author

aurelion-source commented Sep 7, 2024 •
edited
Loading

Report comparing performance of norm implementations (including TE and Apex) can be found here: https://api.wandb.ai/links/shetano-personal/o0tz0qy3

Default 125M config was used on 4 x NVIDIA A100-SXM4-80GB

@Quentin-Anthony
Copy link
Member

resolves #1235 and #1265

@Quentin-Anthony
Copy link
Member

Rebased onto upstream main. Doing the close + reopen trick to handle diff.

Quentin-Anthony
Quentin-Anthony approved these changes Sep 9, 2024
View reviewed changes
Copy link
Member

@Quentin-Anthony Quentin-Anthony left a comment

Choose a reason for hiding this comment

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

LGTM

@Quentin-Anthony Quentin-Anthony merged commit 836aefa into EleutherAI:main Sep 9, 2024
2 of 3 checks passed
This was referenced Sep 9, 2024
Closed
Closed
Draft
Merged
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Reviewers

@Quentin-Anthony Quentin-Anthony Quentin-Anthony approved these changes

Assignees

@aurelion-source aurelion-source

Labels
None yet
Projects
None yet
Milestone
No milestone
Development

Successfully merging this pull request may close these issues.
4 participants
@aurelion-source @Quentin-Anthony @lintangsutawika @dmahan93