Module gguf_weights 

Direct GGUF weight loading for Burn inference — no candle.

Reads F32/F16/BF16/Q8_0 tensors with the in-workspace zsfm_gguf reader, decodes everything to F32 host bytes, and builds Burn tensors.

Structs

GgufWeights

All F32-decoded tensors of one GGUF file, keyed by tensor name.

Functions

open_gguf_weights

Open path and decode every tensor to F32.