vllm.models.muse_spark.nvidia.specialized_decode.cutedsl.moe

Modules:

• expert_norm –

  The BF16 edge layers' pre_expert_norm of the pooler's gathered

• pooler_gather –

  The pooler of the Muse Spark specialized decode.

• routing –

  Muse Spark routing for decode batches of at most 8 tokens: sigmoid of the

• tail –

  The MoE tail of the Muse Spark specialized decode.