Hackable Diffusion is a modular toolbox written in Jax to experiment and educate around Diffusion modeling.

• Gemma Fine-Tuning: Support for text diffusion models (e.g., DiffusionGemma) via a hybrid AR-diffusion implementation. The adapter package and fine-tuning configurations are located in the Gemma repository (GitHub: https://github.com/google-deepmind/gemma/tree/main/gemma/diffusion/hackable_diffusion_adapter).

The core philosophy of this library is hackability. It is designed from the ground up to be modular, composable, and easy to modify, enabling rapid experimentation with new research ideas. Key principles include:

• Composition over Configuration: Build models and training loops by composing small, well-defined Python objects.
• Clear Separation of Concerns: The codebase is organized into logical sub-libraries for architecture, corruption, inference, loss, and sampling.
• Native Multimodality: The library has first-class support for handling multimodal data (e.g., images and text) through a consistent "Nested" component pattern that applies different diffusion parameters to different parts of the data.

The notebooks/ directory contains several tutorials to get you started:

• 2d_training.ipynb: A minimal example on a 2D toy dataset.
• mnist.ipynb: Standard image diffusion on MNIST.
• mnist_discrete.ipynb: An example of discrete diffusion.
• mnist_multimodal.ipynb: A showcase of the multimodal capabilities, generating images and labels jointly.

The kdiff/configs/ directory contains example configurations for training:

• mnist_unet.py: Standard diffusion training configuration on MNIST.

To run a config locally, create a small launcher script (e.g. train.py):

import os
os.environ["XLA_PYTHON_CLIENT_PREALLOCATE"] = "false"

import multiprocessing
from kauldron import konfig

def main():
    import importlib.util
    spec = importlib.util.spec_from_file_location(
        "config", "kdiff/configs/mnist_unet.py"
    )
    config_module = importlib.util.module_from_spec(spec)
    spec.loader.exec_module(config_module)

    cfg = config_module.get_config()
    cfg.workdir = "/tmp/mnist_workdir"
    trainer = konfig.resolve(cfg)
    trainer.train()

if __name__ == "__main__":
    multiprocessing.set_start_method("spawn", force=True)
    main()

Note: XLA_PYTHON_CLIENT_PREALLOCATE=false must be set before importing JAX to prevent GPU memory preallocation conflicts with data loading workers. The if __name__ == "__main__" guard is required for multiprocessing compatibility.

To install the necessary dependencies, you can use pip with the provided pyproject.toml file:

pip install -e .

To install development dependencies (for running tests), use:

pip install -e .[dev]

This will install libraries such as JAX, Flax, and other utilities required to run the code.

Copyright 2025 Google LLC
All software is licensed under the Apache License, Version 2.0 (Apache 2.0); you may not use this file except in compliance with the Apache 2.0 license. You may obtain a copy of the Apache 2.0 license at: https://www.apache.org/licenses/LICENSE-2.0 All other materials are licensed under the Creative Commons Attribution 4.0 International License (CC-BY). You may obtain a copy of the CC-BY license at: https://creativecommons.org/licenses/by/4.0/legalcode Unless required by applicable law or agreed to in writing, all software and materials distributed here under the Apache 2.0 or CC-BY licenses are distributed on an "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. See the licenses for the specific language governing permissions and limitations under those licenses. This is not an official Google product.

If Hackable Diffusion was helpful for a publication, please cite this repository: (authors are included in the alphabetical order by the last name)

@software{hackable_diffusion2026github,
  author = {Crepy, Clement and De Bortoli, Valentin and Galashov, Alexandre and Greff, Klaus and Korshunova, Ira},
  title = {{Hackable Diffusion}: A modular toolbox written in Jax to experiment and educate around Diffusion modeling.},
  url = {https://github.com/google/hackable_diffusion},
  version = {1.0.1},
  year = {2026},
  note = {Authors listed in alphabetical order by the last name},
}

This is not an officially supported Google product.