Skip to content

Conversation

@cfbh-google
Copy link
Collaborator

Notes

This change contains recipes for Gemma3-12B on v6e (Trillium).

Publish the following new recipes:

  • Gemma3-12B on v6e-256
  • Gemma3-12B on 2 slices of v6e-256
  • Gemma3-12B on 4 slices of v6e-256

Added a new MaxText tag tpu-recipes-v0.1.5 so users can check this out directly.

Tests

Optimized the recipes and got following performance:

  • Gemma3-12B on v6e-256 -> MFU=38.06%
  • Gemma3-12B on 2 slices of v6e-256 -> MFU=35.74%
  • Gemma3-12B on 4 slices of v6e-256-> MFU=33.20%

b/451824088

Gemma3-12B on v6e-256
Gemma3-12B on 2 slices of v6e-256
Gemma3-12B on 4 slices of v6e-256
Copy link
Collaborator

@bvandermoon bvandermoon left a comment

Choose a reason for hiding this comment

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

LGTM, just one minor comment

@cfbh-google cfbh-google merged commit 0095923 into main Oct 21, 2025
1 check passed
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