ShineiArakawa/diffusion-model-cpp

A C++ implementation of a diffusion model using libtorch.

diffusion-model-cpp
_{A C++ implementation of a diffusion model using libtorch}

Features

• Diffusion model implemented using libtorch
• Support for training and inference
• Customizable parameters for training and model configuration
• Pure C++ implementation

Tested environments

• C++17
• GCC 11.4.0
• CMake 3.30
• libtorch 2.1.0
• OpenCV 4.7.0 (for image processing)
• CUDAToolkit 12.3

Getting Started

Build

1. Clone the repository:

   git clone --recursive https://github.com/your_username/diffusion-model-cpp.git
   cd diffusion-model-cpp

2. Install dependencies:

   sudo apt-get update
   sudo apt-get install libopencv-dev

3. Download and extract libtorch:

   wget https://download.pytorch.org/libtorch/cu121/libtorch-cxx11-abi-shared-with-deps-2.3.1%2Bcu121.zip
   unzip libtorch-cxx11-abi-shared-with-deps-2.3.1+cu121.zip

4. Build the project using CMake:

   mkdir build
   cd build
   cmake -DCMAKE_PREFIX_PATH=/path/to/libtorch ..
   make

Training

1. Prepare your dataset and update the dataset path and log dir path in the configuration file configs/sample.json.

2. Run the training program:

   ./build/src/train configs/sample.json

3. The model and training logs will be saved in the log directory.

Acknowledgements

• This project uses libtorch for implementing the diffusion model.
• OpenCV is used for image processing tasks.
• Inspiration and algorithms are based on recent research in the field of diffusion models.