Using standard layernorm cuda kernel for skiplayernorm. #15076
Using standard layernorm cuda kernel for skiplayernorm.
d993649a
better resource calc.
ac105925
Fix no enough resource issue. Ensure power of 2 on the threads_y.
b8182259
Strictly using internal calculation type with gamma and beta.
c6d3e53d
yufenglee
approved these changes
on 2023-03-23
yufenglee
merged
910fc09d
into main 3 years ago
yufenglee
deleted the zhalei/stable_skip_layernorm branch 3 years ago
Assignees
No one assigned
Login to write a write a comment.
Login via GitHub