Repository navigation

[PyTorch] [torch.compile] torch.compile support for LayerNormLinear and LayerNormMLP - #3534

Open

pggPL wants to merge 7 commits into

NVIDIA:mainfrom

pggPL:layernorm_compile

Open

[PyTorch] [torch.compile] torch.compile support for LayerNormLinear and LayerNormMLP#3534
pggPL wants to merge 7 commits into
NVIDIA:mainfrom
pggPL:layernorm_compile

Commits on Oct 29, 2026
- [PyTorch] Add torch.compile support for LayerNormLinear and LayerNormMLP
  pggPL
  committed
  83641e8
Commits on Oct 30, 2026
- Fix captured FP8 state and share normalization output layout policy
  pggPL
  committed
  3fda6d9
- Gate custom compile recipe on FP8 support
  pggPL
  committed
  0a7f8bf
- [pre-commit.ci] auto fixes from pre-commit.com hooks
  pre-commit-ci[bot]
  committed
  3add45c
- Clarify normalization scale layout override naming
  pggPL
  committed
  9108ab5
- Use normalization_scale_layout and explain returned metadata
  pggPL
  committed
  267b0d4
- Merge main into layernorm_compile
  pggPL
  committed
  a2f424e