diff --git a/server/Makefile-awq b/server/Makefile-awq index 80e78c08..11517386 100644 --- a/server/Makefile-awq +++ b/server/Makefile-awq @@ -1,8 +1,10 @@ -awq_commit := f084f40bd996f3cf3a0633c1ad7d9d476c318aaa +# Fork that adds only the correct stream to this kernel in order +# to make cuda graphs work. +awq_commit := d2d7cfbfa0de87fbbc17a2984c69cd6681149885 awq: rm -rf llm-awq - git clone https://github.com/mit-han-lab/llm-awq + git clone https://github.com/huggingface/llm-awq build-awq: awq cd llm-awq/ && git fetch && git checkout $(awq_commit)