Commit 44f9d9a

committed

enable fp8 rowwise quantization in xpu8

Signed-off-by: Wang, Yi <yi.a.wang@intel.com>

1 parent 3a8d291 commit 44f9d9aCopy full SHA for 44f9d9a

2 files changed

+376

-53

lines changed

src/transformers
- integrations
  - fbgemm_fp8.py
- quantizers
  - quantizer_fbgemm_fp8.py

+376

-53

lines changed

Comments

(0)