search

ROCm / flash-attention

Fast and memory-efficient exact attention
BSD 3-Clause "New" or "Revised" License
139 stars 46 forks source link

Remove patch #8

Closed groenenboomj closed 1 year ago

sabreshao commented 1 year ago

Our main branch is flash_attention_for_rocm. Could you close this PR and create a new PR to flash_attention_for_rocm?

groenenboomj commented 1 year ago

Our main branch is flash_attention_for_rocm. Could you close this PR and create a new PR to flash_attention_for_rocm?

Done

sabreshao commented 1 year ago

So you add some patches inside latest Pytorch docker so no more patch is needed during FA building? Does it work for both Pyt 1.13 and 2.0+? Would you please update the build instruction part of README.md? I prefer adding the simplified building step and keeping the existing building step for legacy docker image. We can remove the legacy building step in the future.