Code

Support size_per_head=112#660

Code

Merged

byshiue merged 3 commits intoNVIDIA:mainNVIDIA/FasterTransformer:mainfrom dskhudia:support_112dskhudia/FasterTransformer:support_112Copy head branch name to clipboardJun 29, 2023

+113-9

Commits

Commits on May 17, 2023

• fix multi-gpu build

  dskhudia

  committed
  eef4cf6

Commits on Jun 9, 2023

• Merge pull request #1 from dskhudia/multi_gpu_build_fix

  Show description for 29a376a
  dskhudia

  authored
  29a376a

• add support for size_per_head=112 for gpt decoder

  dskhudia

  committed
  7a71b18