Repository navigation

perf(nemotron_h): prefill single-GPU prompts in multi-token chunks - #2207

Merged

chaofengw-nv merged 2 commits into

NVIDIA:mainfrom

chaofengw-nv:perf/nemotron-h-chunked-prefill

Nov 11, 2026

Merged

perf(nemotron_h): prefill single-GPU prompts in multi-token chunks#2207
chaofengw-nv merged 2 commits into
NVIDIA:mainfrom
chaofengw-nv:perf/nemotron-h-chunked-prefill

Commits on Nov 10, 2026
- perf(nemotron_h): prefill single-GPU prompts in multi-token chunks
  chaofengw-nv
  committed
  d27a33d
- fix(nemotron_h): take the chunk decay total from the cumulative sum
  chaofengw-nv
  committed
  2f2cc06