How can I achieve fusion of two consecutive GEMMs using CUTE? #1720
-
|
I am trying to implement fusion of two consecutive GEMMs (mxn * nxk * kxn) using cute. How can I set the result of the first GEMM with tiled MMA to be compatible with the second GEMM? In other words, how can I achieve something similar to examples/13_two_tensor_op_fusion using cute? |
Beta Was this translation helpful? Give feedback.
Answered by
thakkarV
Aug 19, 2024
Replies: 1 comment
-
|
https://github.com/Dao-AILab/flash-attention/tree/main/hopper this is the best example we have |
Beta Was this translation helpful? Give feedback.
0 replies
Answer selected by
luliyucoordinate
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
https://github.com/Dao-AILab/flash-attention/tree/main/hopper
this is the best example we have