Skip to main content

Module source

ferrum_quantization::gguf

Module source 

Source

Structs§

GgufWeightComponentSource
Schema-addressed, mmap-backed GGUF source for vNext static weights. Fixed-block payloads borrow the immutable file mapping without dequantization or repacking. Dense floating-point payloads are borrowed when their type matches the schema and materialized on a cold-path source request when the typed execution plan requires another floating-point type.

Functions§

block_quantization_format