transformers
Mamba `slow_forward` gradient fix
#29563
Merged

Mamba `slow_forward` gradient fix #29563

vasqu
vasqu FIX: Cached slow forward in mamba
7e465d50
vasqu formatting
796e0708
asartran
asartran commented on 2024-03-11
javiermcebrian
vasqu fix: use real `slow_forward` call instead of torch module's
1cffa8b1
vasqu
asartran
asartran approved these changes on 2024-03-14
vasqu add shape assertion for mixer block test
aa84a677
vasqu adjust shape assertion
4b8bf0d7
asartran asartran requested a review from ArthurZucker ArthurZucker 2 years ago
asartran
ArthurZucker
ArthurZucker
ArthurZucker approved these changes on 2024-03-27
ArthurZucker ArthurZucker merged cefb819f into main 2 years ago
vasqu vasqu deleted the mamba-slowforward-gradient-fix branch 2 years ago

Login to write a write a comment.

Login via GitHub

Assignees
No one assigned
Labels
Milestone