Code

AMD (MI250X) support#1775

Code

Merged

rasbt merged 7 commits intoLightning-AI:mainLightning-AI/litgpt:mainfrom TensorTemplar:mi250x-supportTensorTemplar/litgpt:mi250x-supportCopy head branch name to clipboardOct 10, 2024

Commits

Commits on Oct 9, 2024

• add debug print and alternative mask

  TensorTemplar

  committed
  bbf3546

  Copy full SHA for bbf3546

• Split nvlink check into nvidia vs amd. Add tests

  TensorTemplar

  committed
  3942bf1

  Copy full SHA for 3942bf1

• Give up and rewrite build_rope_cache with vectorization instead. Add clarification comment in test

  Show description for 232e4fa
  TensorTemplar

  committed
  232e4fa

  Copy full SHA for 232e4fa

• Merge branch 'main' into mi250x-support

  TensorTemplar

  committed
  c29666a

  Copy full SHA for c29666a

• Revert style changes to previous code

  TensorTemplar

  committed
  531e54d

  Copy full SHA for 531e54d

• Use correct fixture for dual-gpu test

  TensorTemplar

  committed
  16d7e5a

  Copy full SHA for 16d7e5a

• Add new mocks to previous tests

  TensorTemplar

  committed
  da906a0

  Copy full SHA for da906a0