add fp8 quantization kernels #12

Merged

rogerxfeng8 merged 7 commits intovllm-project:mainfrom

baodii:baodi/fp8_kernels

Aug 18, 2025

Commits on Aug 14, 2025

add fp8 quant kernels

baodii
authored and
zufangzhu
committed
add fp8_e5m2 support and fixing UT

baodii
authored and
zufangzhu
committed
add per-token quanzation

baodii
authored and
zufangzhu
committed
change per-token UT assert_close by percentage of mismatch since fp8 diffence is too huge

baodii
authored and
zufangzhu
committed
merge redundant files
zufangzhu
committed
remove redefine struct
zufangzhu
committed

Commits on Aug 18, 2025

update
zufangzhu
committed