Repository navigation

Enable inference CUDA graphs with Shortcut MoE and wide residuals - #7772

Open

susavlsh10 wants to merge 6 commits into

NVIDIA:mainfrom

susavlsh10:nemotron4-inference-cuda-graphs

Commits on Oct 6, 2026

Enable Nemotron 4 Shortcut MoE inference CUDA graphs
susavlsh10
committed
2cb2653
Remove redundant inference RMSNorm dtype adaptation
susavlsh10
committed
5a6a958
Align checkpoint routing test with upstream MIMO detection
susavlsh10
committed
cfa345b
Handle SSM token padding independently of batch-invariant guards
susavlsh10
committed
7083a3b
Prepare global-batch QB routing in shared inference configuration
susavlsh10
committed
4884d25
Remove QB inference workaround in favor of native support

Susav Shrestha
authored and
susavlsh10
committed
1d5f7d4