-
Notifications
You must be signed in to change notification settings - Fork 3.8k
Closed
Description
Note
This patch release includes important bug fixes to the function definition of Attention-23 under the Group Query Attention mode and to the reference implementation of RotaryEmbedding-23.
All changes
- Avoid unnecessary re-generating of proto files (Avoid unnecessary re-generating of proto files #7253) in Avoid unnecessary re-generating of proto files (#7253) #7306
- Require ml_dtypes>=0.5.0 (Require ml_dtypes>=0.5.0 #7254) in Require ml_dtypes>=0.5.0 (#7254) #7307
- Cherry pick four attention PRs in Cherry pick four attention PRs #7315
- Update rotary_embedding reference implementation and tests (Update rotary_embedding reference implementation and tests #7304, Rename variables in rotary embedding reference for clarity #7316) in Update rotary_embedding reference implementation and tests (#7304, #7316) #7313
- Override
__repr__for some proto classes (Override__repr__for some proto classes #7259) in Override __repr__ for some proto classes (#7259) #7314 - Update version number to 1.19.1rc1 in Update version number to 1.19.1rc1 #7322
- add check for rc-candidates (Update create_release.yml) (add check for rc-candidates (Update create_release.yml) #7261) in add check for rc-candidates (Update create_release.yml) (#7261) #7323
- Implement repr methods for Model/Graph/Function (Implement repr methods for Model/Graph/Function #7320) in Implement repr methods for Model/Graph/Function (#7320) #7325
Full Changelog: v1.19.0...v1.19.1
Metadata
Metadata
Assignees
Labels
No labels