Repository navigation

Fused QKV ROPE - CS FP8 quantization - #3675

Open

vasunvidia wants to merge 3 commits into

NVIDIA:mainfrom

vasunvidia:vrengasamy/fused_rope_optimization

Open

Fused QKV ROPE - CS FP8 quantization#3675
vasunvidia wants to merge 3 commits into
NVIDIA:mainfrom
vasunvidia:vrengasamy/fused_rope_optimization

Commits on Nov 7, 2026
- Add two-phase Triton FP8 current-scaling quantize
  
  vasunvidia
  and
  claude
  committed
  3769101
- Fuse fused-QKV RoPE with FP8 current-scaling Q/K/V quantize
  
  vasunvidia
  and
  claude
  committed
  ab4d0ec
- [pre-commit.ci] auto fixes from pre-commit.com hooks
  
  pre-commit-ci[bot]
  authored and
  vasunvidia
  committed
  a7b5c99