huggingface-hub-cpp

This library is a C++ port of the Hugging Face Hub that provides an implementation of hf_hub_download for C++. It enables seamless downloading of models and other assets from the Hugging Face Hub while maintaining full compatibility with its Python counterpart.

With this library, C++ developers can integrate Hugging Face model downloads directly into their applications, ensuring efficient and reliable access to AI resources without requiring Python dependencies.

Table of Contents

• huggingface-hub-cpp
  • Table of Contents
  • Installation
    • Prerequisites
    • Build
  • Usage
    • Integrating the library
    • Running the demo app
  • License

Installation

Prerequisites

Before installing, ensure you have the following dependencies installed:

• CMake (Minimum version recommended: 3.x)
• libcurl (Required for handling downloads)

On Debian-based systems, you can install libcurl using:

sudo apt-get install libcurl4-openssl-dev

Build

To build the project, follow these steps:

git clone https://github.com/agonzc34/huggingface-hub-cpp
mkdir build
cd build
cmake ..
make

Usage

Integrating the library

To use this library in your C++ project, include the necessary headers and call hf_hub_download.

#include "huggingface_hub.h"

int main() {
  std::string repo_id = "<huggingface-repo>";
  std::string filename = "<file-to-download>";

  if (huggingface_hub::hf_hub_download(repo_id, filename).success) {
    std::cout << "Model downloaded!" << std::endl;
  } else {
    std::cout << "Error" << std::endl;
  }

Running the demo app

A demo application is included to showcase the library in action. To build and run the demo:

./build/hfhub_demo

This will execute a sample download for the qwen2.5-0.5b-instruct-q2_k.gguf model

License

This project is licensed under the MIT License. See the LICENSE file for more details.