Skip to content
Projets

deepqmc/deepqmc has 424 GitHub stars and 64 forks. GitHot measured 0 stars in 24 hours and +1 in 7 days. Last checked 2026-11-18 21:47 UTC. Last push 2026-11-11.

Comment ces chiffres sont mesurés

Progression mesurée des étoiles

in 24h
0
in 7 days
+1
in 15 days
Historique en cours
in 1 month
Historique en cours

Variations nettes sur des périodes qui se chevauchent, pas des totaux quotidiens. Les mesures absentes restent vides. Les étoiles indiquent l’intérêt, pas la qualité.

À propos

Deep learning quantum Monte Carlo for electrons in real space

Apprentissage automatique

README.md

Original

DeepQMC

checks coverage python pypi commits since last commit license code style Checked with pyright doi

DeepQMC is an open-source software suite for variational optimization of deep-learning molecular wave functions. It implements the simulation of electronic ground and excited states using deep neural network trial wave functions. The package is based on JAX and Haiku and is configured through a cli build of Hydra.

The program solves the molecular Hamiltonian, allowing the use of effective core potentials. Excited states are obtained via a penalty-based excited-state optimization approach. A spin penalty allows states in a fixed spin sector to be targeted. Furthermore, DeepQMC implements geometric transferability, that is a single ansatz can be optimized across multiple configurations of a set of atoms. This can either be done by training on a fixed set of molecular configurations or by dynamically sampling molecular configurations throughout the optimization, and can be combined with the optimization of electronic excited states.

The software suite includes a general neural network wave function ansatz, that can be configured to obtain a wide range of molecular neural network wave functions. Config files for the instantiation of variants of Psiformer, PauliNet, FermiNet, DeepErwin, LapNet and a transferable extension of the self-attention ansatz TransPsiformer can be found under src/deepqmc/conf/ansatz.

Installation

Install and update to the latest release using Pip:

pip install -U deepqmc

To install DeepQMC from a local Git repository run:

git clone https://github.com/deepqmc/deepqmc
cd deepqmc
pip install -e .[dev]

If Pip complains about setup.py not being found, please update to the latest Pip version.

The above installation will result in the CPU version of JAX. However, running DeepQMC on the GPU is highly recommended. To enable GPU support make sure to upgrade JAX to match the CUDA and cuDNN versions of your system. For most users this can be achieved with:

# CUDA 12 installation
pip install "jax[cuda12]" -f https://storage.googleapis.com/jax-releases/jax_cuda_releases.html

For further help setting up JAX, visit the JAX Install Guide.

To validate the correct installation of DeepQMC and its dependencies run:

deepqmc

Documentation and exemplary usage

For further information about the DeepQMC package and tutorials covering the basic usage visit the documentation.

An introduction to the methodology, implementation details and exemplary experiments can be found in the associated software paper.

The penalty-based excited states approach and its implementation in DeepQMC is discussed in our recent paper on excited state optimization.

The geometrically transferable simulation with DeepQMC on fixed datasets is demonstrated in our paper on tranferable simulation of excited state potential energy surfaces.

An optimization of wave functions across continuous potential energy surfaces and subsequent optimization of molecular geometries is employed in our latest paper on ab initio geometry optimization.

Citation

If you use DeepQMC for your work, please cite our implementation paper:

@article{10.1063/5.0157512,
    author = {Schätzle, Z. and Szabó, P. B. and Mezera, M. and Hermann, J. and Noé, F.},
    title = "{DeepQMC: An open-source software suite for variational optimization of deep-learning molecular wave functions}",
    journal = {The Journal of Chemical Physics},
    volume = {159},
    number = {9},
    pages = {094108},
    year = {2023},
    month = {09},
    issn = {0021-9606},
    doi = {10.1063/5.0157512},
    url = {https://doi.org/10.1063/5.0157512},
}

Experiments including excited state optimization may cite our excited state paper:

@article{10.1021/acs.jctc.4c00678,
    author = {Szabó, P. B. and Schätzle, Z. and Entwistle, M. and Noé, F.},
    title = "{An Improved Penalty-Based Excited-State Variational Monte Carlo Approach with Deep-Learning Ansatzes}",
    journal = {Journal of Chemical Theory and Computation},
    year = {2024},
    month = {08},
    issn = {1549-9618},
    doi = {10.1021/acs.jctc.4c00678},
    url = {https://doi.org/10.1021/acs.jctc.4c00678},
}

When employing transferable optimization consider citing:

@online{schatzle2025a,
  author = {Schätzle, Z. and Szabó, P. B. and Cuzzocrea, A. and Noé, F.},
  title = {Ab-Initio Simulation of Excited-State Potential Energy Surfaces with Transferable Deep Quantum {{Monte Carlo}}},
  year = {2025},
  month = {03},
  doi = {10.48550/arXiv.2503.19847},
  pubstate = {prepublished},
}

The repository can be cited as:

@software{deepqmc,
	author = {Hermann, J. and Schätzle, Z. and Szabó, P. B. and Mezera, M and {DeepQMC Contributors}},
	title = "{DeepQMC}",
	year = {2026},
	publisher = {Zenodo},
	copyright = {MIT},
	url = {https://github.com/deepqmc/deepqmc},
	doi = {10.5281/zenodo.3960826},
}