Module format 

Model Format Detection and Metadata

Detects model file formats and extracts metadata from model files.

Supported Formats

• GGUF: GGML Universal Format (llama.cpp, ollama)
• SafeTensors: HuggingFace safe tensor format
• APR: Aprender native format
• ONNX: Open Neural Network Exchange
• PyTorch: .pt/.pth files (detection only)

Example

use pacha::format::{detect_format, ModelFormat};

let format = detect_format(&data)?;
match format {
    ModelFormat::Gguf(info) => println!("GGUF: {} params", info.parameters),
    ModelFormat::SafeTensors(info) => println!("SafeTensors: {} tensors", info.tensor_count),
    _ => println!("Other format"),
}

Structs

AprInfo

APR (Aprender) file information

GgufInfo

GGUF file information

OnnxInfo

ONNX file information

SafeTensorsInfo

SafeTensors file information

TensorInfo

Tensor information

Enums

ModelFormat

Detected model format

QuantType

Common quantization types (GGUF spec naming convention)

Functions

detect_format

Detect model format from bytes

detect_format_from_path

Detect format from file path extension