[FP8] Support FP8 MatrixCore Code gen and related test

microsoft / BitBLAS

BitBLAS is a library to support mixed-precision matrix multiplications, especially for quantized LLM deployment.

MIT License

190 stars 21 forks source link

Closed LeiWang1999 closed 2 months ago

LeiWang1999 commented 2 months ago

TODO Items:

[ ] fp8 torch tensor doesn't support cast to numpy/tvm, should be handled in another way, currently we disable ladder tranform for fp8.