Module bound 

BF16/FP32 twin layer bind. Decode copies hidden D2H, runs mixer+MLP+AttnRes, copies H2D. LinearAttention / KDA conv+recurrent uses CUDA kda_decode with per-sequence resident recurrence unless K3_CUDA_KDA=0. FullAttention / MLA uses CUDA mla_decode unless K3_CUDA_MLA=0. Packed LatentMoE experts launch DSV4 moe_w4a16_grouped_gemm_ptrtable_e8m0. Router / down / up / shared / SiTU stay on the host.

Structs

K3BoundLayer

One decoder layer whose BF16 (or FP32) tensors were bound from the store.

K3HostShared

Shared across every decoder layer of one loaded K3 model.

WeightMeta

Name + device dtype/numel for a lazy host bind (copy from GPU if needed).