Crate forgellm_frontend 

ForgeLLM Frontend — Model parsing and IR construction.

This crate handles parsing model formats (GGUF, SafeTensors) and constructing the intermediate representation (IR) used by the optimizer and code generation backends.

Re-exports

pub use lora::load_lora;

pub use lora::load_lora_from_bytes;

pub use lora::merge_lora;

pub use lora::LoraAdapter;

pub use lora::LoraError;

pub use lora::LoraLayer;

pub use onnx_export::export_onnx;

pub use safetensors_loader::load_safetensors;

pub use ir::*;

Modules

config

HuggingFace model config.json parser.

gguf

GGUF file format parser.

graph_builder

Graph builder — constructs IR computation graphs from model configs.

hub

HuggingFace Hub integration — resolve model IDs to local paths.

ir

Intermediate representation for transformer computation graphs.

lora

LoRA adapter loading and weight merging.

onnx_export

ONNX export — serialize an IR graph + weights to ONNX protobuf format.

safetensors

SafeTensors file format parser.

safetensors_loader

SafeTensors model loader.

weight_loader

Weight loader — reads and dequantizes tensor data from GGUF files.