transformers
6f075c56
- [sdpa] Allow prefill to use FA kernel with StaticCache (#47094)
Go
Login via GitHub
Home
Pricing
FAQ
Install
Login
via GitHub
Commit
View On
GitHub
Commit
13 days ago
[sdpa] Allow prefill to use FA kernel with StaticCache (#47094) * allow more conditions * remove xpu specific code: it's the same * remove old cond * fix * faster checks * fix sdpa integration to dispatch to flash
References
#47094 - [sdpa] Allow prefill to use FA kernel with StaticCache
Author
Cyrilvallez
Parents
246ed266
Loading