Module nemotron

Module nemotron 

Source
Expand description

2026-09-25: Nemotron-H weight structs and loaders for the Mamba-2 mixer, attention and MoE layers.

Owner: model-layers (weight loading). Invariants: none beyond the types.

Structs§

NemotronExpertWeight
2026-09-25: Nemotron-H expert: up_proj and down_proj, no gate_proj.
NemotronMoeWeights
2026-09-25: Nemotron-H MoE layer weights.
NemotronSsmWeights
2026-09-25: Nemotron-H Mamba-2 mixer weights.

Enums§

NemotronSsmQuant
2026-09-25: Format of a Mamba-2 layer’s in_proj, detected at load.

Functions§

load_nemotron_attention
2026-09-25: Load a Nemotron-H attention mixer.
load_nemotron_moe
2026-09-25: Load a Nemotron-H MoE mixer.
load_nemotron_ssm
2026-09-25: Load a Nemotron-H Mamba-2 mixer ({layer_prefix}.mixer).