Repository navigation

Fix ONNX custom-op metadata and LayerNorm execution - #3659

Merged

pggPL merged 4 commits into

NVIDIA:mainfrom

sanjana658:fix/3636-fp8-fake-scale-shape

Oct 15, 2026

Merged

Fix ONNX custom-op metadata and LayerNorm execution#3659
pggPL merged 4 commits into
NVIDIA:mainfrom
sanjana658:fix/3636-fp8-fake-scale-shape

Commits on Oct 9, 2026
- fix: correct FP8 current scaling fake output shape
  
  Sanjana Thanabalan
  authored and
  gpalme01
  committed
  b3a3b82
- test: include FP8 fake tensor opcheck in L0 CI
  
  Sanjana Thanabalan
  authored and
  gpalme01
  committed
  019b153
Commits on Oct 14, 2026
- Move FP8 current scaling opcheck into ONNX tests
  pggPL
  committed
  2d1d878
- Fix ONNX LayerNorm eager and fake implementations
  pggPL
  committed
  e1f3b74