Repository navigation

cuda: RMS norm rows and token rows for every family, Nemotron's embed the first user - #736

Merged

ashhart merged 2 commits into

ashhart:mainfrom

jschmied:cuda-shared-norms

Oct 13, 2026

Merged

cuda: RMS norm rows and token rows for every family, Nemotron's embed the first user#736
ashhart merged 2 commits into
ashhart:mainfrom
jschmied:cuda-shared-norms

Commits on Oct 12, 2026
- cuda: RMS norm rows and token rows for every family (norms.cu, embed_rows.cu)
  jschmied
  committed
  83d83d9
- Nemotron CUDA: the 4-bit embed on embed_rows.cu (same bits), tf_nemo_embed removed
  jschmied
  committed
  c1138e2