Expand description
2026-09-25: The model engine: the Model trait (traits), the transformer model (model),
the single-request generate loop (engine), the model factory (factory), the Kimi K3 weight
loader (kimi_k3_loader), the prefill chunk planner (prefill_plan) and the startup rank-agreement check (rank_agree).
Owner: model-engine. Invariants: none beyond the types.
Modules§
- engine
- 2026-09-25: A greedy generate loop for one request over the
Modeltrait: prefill, then decode, taking each token withargmax_on_device. - factory
- 2026-09-25: Model factory:
loader_for_configpicks theModelWeightLoaderfor a config’smodel_type, andbuild_modelbuilds the model from the config and the weights. - kimi_
k3_ loader - 2026-09-25: The Kimi K3
ModelWeightLoader: binds BF16 or FP32 checkpoints, and packed MXFP4 experts only whenK3_ALLOW_MXFP4=1. - model
- 2026-09-25: The generic transformer model,
TransformerModel, and the served NLLB model. - prefill_
plan - 2026-09-27: Where prefill chunks end, and which chunk ends carry an SSM snapshot.
- rank_
agree - 2026-09-25: Startup check that every rank holds the same values for the env-read scalars that shape the collective schedule.
- traits
- 2026-09-25: The
Modeltrait, the per-sequence state it works on (SequenceState), and the input and result types of its forward passes.