transformers
119e04d5
- Simplify _flash_attention_forward
Go
Login via GitHub
Home
Pricing
FAQ
Install
Login
via GitHub
Commit
View On
GitHub
Commit
3 days ago
Simplify _flash_attention_forward
References
#49071 - [Refactor] Make some flash-attention utils more readable
#49088 - 🚨 [CB] 🚨 Fuse update for index and block table path
#49101 - 🚨 Attention 🚨 Make regular attention support CB
#49112 - [Attention] Deprecate "paged|" prefix for SDPA and flash
Author
remi-or
Parents
8d9e7416
Loading