vllm.models.deepseek_v41.nvidia.specialized_decode.moe

Modules:

• routed_expert_front_cute –

  Fused MoE front of the DeepSeek-V4.1 low-batch routed experts in CuTe DSL.

• routed_expert_gemm –

  Low-batch routed-expert GEMMs for DeepSeek-V4.1 decode (1-8 rows).

• routed_expert_gemm_cute –

  Routed-expert GEMMs of the DeepSeek-V4.1 low-batch MoE in CuTe DSL.

• routed_expert_route_cute –

  Routing record of the DeepSeek-V4.1 low-batch routed-expert GEMMs in CuTe DSL.

• step –

  The MoE of a specialized decode step.