Module weights 

Model-agnostic weight I/O — paths, formats, drain policy only.

Architecture checks and tensor renaming live in model crates (gguf_validate_arch, register_gguf_tensor_resolver), not here.

use rlx_models_core::weights::{self, LoadOpts};

let (path, map) = weights::open_map_with(LoadOpts::map().prefer_q4_k_m(), "weights/")?;
let loaded = weights::open_with(LoadOpts::loader(), "model.gguf")?;

Structs

GgufDirGuide

GgufLoader

LlamaFamilyGgufResolver

HF model.layers.N.* ↔ GGUF blk.N.* (Llama, Qwen3, Qwen35, …).

LoadWeightsOptions

Options for load_weights_resolved — prefer crate::weights::LoadOpts presets at call sites.

PrefixStripGgufResolver

Strip common HF prefixes and match verbatim tensor names (architecture-agnostic fallback).

Qwen35NativeGgufResolver

Qwen3.5 native blk.N.* names; also accept HF aliases via the Llama mapper.

RegisteredFormat

One registered on-disk format (built-in or custom).

ResolveWeightsOptions

Options for resolve_weights_file_with_options.

WeightFormatRegistration

Describes one on-disk weight format.

WeightMap

Map of tensor name → (f32 data, shape).

Enums

LoadedWeights

Result of resolving and opening weights.

WeightDrainPolicy

How WeightMap::drain_loader / WeightMap::from_weight_loader handle leftovers.

Traits

GgufTensorNameResolver

Resolve a builder-requested tensor name to the name stored in a GGUF file.

WeightLoader

Common interface every weight format must satisfy. Mirrors the existing WeightMap API so the safetensors impl is a one-line adapter.

Functions

default_resolve_opts

Default directory resolve: prefer DEFAULT_GGUF_PREFER_SUBSTR.

format_for_extension

Extension → format id (last registration wins).

gguf_dir_guide

Numbered .gguf listing + resolve hints for a directory (CLI / errors).

gguf_split_siblings

Other parts of the same multi-file GGUF split in path’s directory (sorted by split.no).

gguf_validate_arch

Ensure a GGUF file’s general.architecture is in allowed (call from model runners, not the loader).

init

Idempotent: ensure built-in GGUF tensor resolvers are registered (safe to call from main).

list_gguf_files_in_dir

Sorted .gguf paths in a directory (non-recursive).

list_registered_formats

All registered formats (built-ins first, then custom registrations).

load_gguf_file

Open a GGUF file, merging multi-part splits when all siblings are present in the directory.

load_weight_map

Resolve a weights path, validate GGUF general.architecture when applicable, drain to F32.

load_weight_map_resolved

Convenience: resolve + drain to F32 WeightMap.

load_weights_resolved

Resolve a file or directory, enforce GGUF arch policy, open via registry, optionally drain.

open

Resolve + open (live WeightLoader).

open_map

Resolve + drain to F32 WeightMap.

open_map_with

Resolve + drain with options.

open_weight_loader

Open a single file via the format registry.

open_with

Resolve + open with options.

pick

Resolve a file or weights directory to one on-disk path.

pick_default

pick with default_resolve_opts.

register_gguf_tensor_resolver

Register a custom resolver (call before first GGUF load). Later registrations win among resolvers that match the same architecture.

register_weight_format

Register a custom weight format (call before the first load). Later entries override built-ins when the same extension is registered twice.

resolve_weights_file

Resolve --weights to a single file: pass-through for files, or pick one .gguf / model.safetensors inside a directory.

resolve_weights_file_with_options

Resolve with optional GGUF file selection inside a directory.

Type Aliases

LoadOpts

Alias for LoadWeightsOptions.

PassThroughGgufResolver

Alias for PrefixStripGgufResolver (older name).

ResolveOpts

Alias for ResolveWeightsOptions.