ChemTorch is a modular framework for developing and benchmarking deep learning models on chemical reaction data. The framework supports multiple families of reaction representations, neural network architectures, and downstream tasks.

The code is provided under MIT license, making it freely available for both academic and commercial use.

The detailed ChemTorch documentation is available at: https://heid-lab.github.io/chemtorch

First clone this repo and navigate to it:

git clone https://github.com/heid-lab/chemtorch.git
cd chemtorch

Then follow the instructions below to install ChemTorch's dependencies using you package manager of choice.

conda create -n chemtorch python=3.10 && \
conda activate chemtorch && \
pip install rdkit numpy==1.26.4 scikit-learn pandas && \
pip install torch && \
pip install hydra-core && \
pip install wandb && \
pip install ipykernel && \
pip install torch_geometric torch_scatter torch_sparse torch_cluster torch_spline_conv -f https://data.pyg.org/whl/torch-2.5.0+cpu.html && \
pip install -e . && \

For GPU usage

pip install torch-scatter torch-sparse -f https://data.pyg.org/whl/torch-${TORCH}+${CUDA}.html

First install uv following the official installation instructions. Then run:

uv sync
uv pip install torch_scatter torch_sparse torch_cluster torch_spline_conv torch_geometric  --no-build-isolation

To also install development and documentation dependencies add the --groups option followed by dev or docs. Alternatively, you can also use --all-groups to install both.

If you encounter bugs, unexpected behaviour, or would like to request a feature, please open an issue on the GitHub issue tracker: https://github.com/heid-lab/chemtorch/issues

Contributions and help are very welcome. If you'd like to contribute a larger change, please open an issue first so we can discuss the best approach.

Get the data from https://github.com/heid-lab/reaction_database and add it to the data folder.

If you use this code in your research, please cite the following paper:

@article{landsheere_chemtorch_2025,
	title = {ChemTorch: A Deep Learning Framework for Benchmarking and Developing Chemical Reaction Property Prediction Models},
	doi = {10.26434/chemrxiv-2025-9mggj},
	journal = {ChemRxiv},
	author = {De Landsheere, Jasper and Zamyatin, Anton and Karwounopoulos, Johannes and Heid, Esther},
	year = {2025},
}

This framework was inspired by:

• GraphGPS
• GraphGym
• lightning-hydra-template