Skip to content

Conversation

HandH1998
Copy link

Hello, you guys did an excellent work! I supported w4a8 marlin based your code and achieved a considerable speedup. The main idea can refer to Section 3.3 of our paper QQQ. You can use bench_w4a8.py to benchmark the w4a8 performance and use test_w4a8.py to verify the correctness. Here we provide the benchmark results on A100.
gemm_performance

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant