Struct GgufShardSet 

pub struct GgufShardSet {
    pub paths: Vec<PathBuf>,
    pub total_tensors: usize,
}

One resolved split-GGUF file set, in split.no order.

A split GGUF is N files that share one logical tensor table. llama.cpp writes the full metadata KV block into shard 0 only; every shard carries split.no, split.count and split.tensors.count, and each holds its own slice of the tensors.

Fields

paths: Vec<PathBuf>

Shard paths ordered by split.no. Length 1 for an unsplit file.

total_tensors: usize

split.tensors.count from shard 0: the total across the whole set.