Best of Atomistic Machine Learning
Best of Atomistic Machine Learning βοΈπ§¬π
π A ranked list of awesome atomistic machine learning (AML) projects. Updated regularly.
This curated list contains 430 awesome open-source projects with a total of 200K stars grouped into 22 categories. All projects are ranked by a project-quality score, which is calculated based on various metrics automatically collected from GitHub and different package managers. If you like to add or update projects, feel free to open an issue, submit a pull request, or directly edit the projects.yaml.
The current focus of this list is more on simulation data rather than experimental data, and more on materials rather than drug design. Nevertheless, contributions from other fields are warmly welcome!
How to cite. See the button "Cite this repository" on the right side-bar.
π§ββοΈ Discover other best-of lists or create your own.
Contents
- Active learning 6 projects
- Community resources 31 projects
- Datasets 46 projects
- Data Structures 4 projects
- Density functional theory (ML-DFT) 33 projects
- Educational Resources 28 projects
- Explainable Artificial intelligence (XAI) 3 projects
- Electronic structure methods (ML-ESM) 5 projects
- General Tools 22 projects
- Generative Models 14 projects
- Interatomic Potentials (ML-IAP) 71 projects
- Language Models 22 projects
- Materials Discovery 12 projects
- Mathematical tools 11 projects
- Molecular Dynamics 10 projects
- Reinforcement Learning 2 projects
- Representation Engineering 25 projects
- Representation Learning 58 projects
- Universal Potentials 12 projects
- Unsupervised Learning 7 projects
- Visualization 6 projects
- Wavefunction methods (ML-WFT) 5 projects
- Others 1 projects
Explanation
- π₯π₯π₯ Combined project-quality score
- βοΈ Star count from GitHub
- π£ New project (less than 6 months old)
- π€ Inactive project (6 months no activity)
- π Dead project (12 months no activity)
- ππ Project is trending up or down
- β Project was recently added
- π¨βπ» Contributors count from GitHub
- π Fork count from GitHub
- π Issue count from GitHub
- β±οΈ Last update timestamp on package manager
- π₯ Download count from package manager
- π¦ Number of dependent projects
Active learning
Projects that focus on enabling active learning, iterative learning schemes for atomistic ML.
FLARE (π₯21 Β· β 300) - An open-source Python package for creating fast and accurate interatomic potentials. MIT C++ ML-IAP
- [GitHub](https://github.com/mir-group/flare) (π¨βπ» 43 Β· π 71 Β· π₯ 8 Β· π¦ 12 Β· π 220 - 16% open Β· β±οΈ 01.11.2024):
IPSuite (π₯17 Β· β 19) - A Python toolkit for FAIR development and deployment of machine-learned interatomic potentials. EPL-2.0 ML-IAP MD workflows HTC FAIR
- [GitHub](https://github.com/zincware/IPSuite) (π¨βπ» 8 Β· π 11 Β· π¦ 7 Β· π 140 - 50% open Β· β±οΈ 17.12.2024):
- [PyPi](https://pypi.org/project/ipsuite) (π₯ 500 / month Β· π¦ 2 Β· β±οΈ 04.12.2024):
Finetuna (π₯10 Β· β 46 Β· π€) - Active Learning for Machine Learning Potentials. MIT
- [GitHub](https://github.com/ulissigroup/finetuna) (π¨βπ» 11 Β· π 11 Β· π¦ 1 Β· π 20 - 25% open Β· β±οΈ 15.05.2024):
Show 3 hidden projects...
- flare++ (π₯13 Β· β 35 Β· π) - A many-body extension of the FLARE code.MIT C++ ML-IAP
- ACEHAL (π₯5 Β· β 11 Β· π) - Hyperactive Learning (HAL) Python interface for building Atomic Cluster Expansion potentials. Unlicensed Julia
- ALEBREW (π₯2 Β· β 14) - Official repository for the paper Uncertainty-biased molecular dynamics for learning uniformly accurate interatomic.. Custom ML-IAP MD
Community resources
Projects that collect atomistic ML resources or foster communication within community.
π AI for Science Map - Interactive mindmap of the AI4Science research field, including atomistic machine learning, including papers,..
π Atomic Cluster Expansion - Atomic Cluster Expansion (ACE) community homepage.
π CrystaLLM - Generate a crystal structure from a composition. language-models generative pretrained transformer
π GAP-ML.org community homepage ML-IAP
π matsci.org - A community forum for the discussion of anything materials science, with a focus on computational materials science..
π Matter Modeling Stack Exchange - Machine Learning - Forum StackExchange, site Matter Modeling, ML-tagged questions.
π ACE / GRACE support - Support forum for the Atomic Cluster Expansion (ACE) and extensions.
Best-of Machine Learning with Python (π₯23 Β· β 18K) - A ranked list of awesome machine learning Python libraries. Updated weekly. CC-BY-4.0 general-ml Python
- [GitHub](https://github.com/ml-tooling/best-of-ml-python) (π¨βπ» 50 Β· π 2.5K Β· π 61 - 44% open Β· β±οΈ 27.12.2024):
OpenML (π₯19 Β· β 680) - Open Machine Learning. BSD-3 datasets
- [GitHub](https://github.com/openml/OpenML) (π¨βπ» 35 Β· π 90 Β· π 930 - 39% open Β· β±οΈ 07.12.2024):
MatBench Discovery (π₯19 Β· β 120) - An evaluation framework for machine learning models simulating high-throughput materials discovery. MIT datasets benchmarking model-repository
- [GitHub](https://github.com/janosh/matbench-discovery) (π¨βπ» 11 Β· π 22 Β· π¦ 4 Β· π 49 - 8% open Β· β±οΈ 27.12.2024):
- [PyPi](https://pypi.org/project/matbench-discovery) (π₯ 900 / month Β· β±οΈ 11.09.2024):
Graph-based Deep Learning Literature (π₯18 Β· β 4.9K) - links to conference publications in graph-based deep learning. MIT general-ml rep-learn
- [GitHub](https://github.com/naganandy/graph-based-deep-learning-literature) (π¨βπ» 12 Β· π 770 Β· β±οΈ 12.12.2024):
MatBench (π₯17 Β· β 140 Β· π€) - Matbench: Benchmarks for materials science property prediction. MIT datasets benchmarking model-repository
- [GitHub](https://github.com/materialsproject/matbench) (π¨βπ» 25 Β· π 45 Β· π¦ 20 Β· π 65 - 60% open Β· β±οΈ 20.01.2024):
- [PyPi](https://pypi.org/project/matbench) (π₯ 330 / month Β· π¦ 2 Β· β±οΈ 27.07.2022):
GT4SD - Generative Toolkit for Scientific Discovery (π₯15 Β· β 340) - Gradio apps of generative models in GT4SD. MIT generative pretrained drug-discovery model-repository
- [GitHub](https://github.com/GT4SD/gt4sd-core) (π¨βπ» 20 Β· π 72 Β· π 120 - 12% open Β· β±οΈ 12.09.2024):
AI for Science Resources (π₯13 Β· β 550) - List of resources for AI4Science research, including learning resources. GPL-3.0 license
- [GitHub](https://github.com/divelab/AIRS) (π¨βπ» 30 Β· π 63 Β· π 20 - 15% open Β· β±οΈ 15.11.2024):
GNoME Explorer (π₯10 Β· β 920) - Graph Networks for Materials Exploration Database. Apache-2 datasets materials-discovery
- [GitHub](https://github.com/google-deepmind/materials_discovery) (π¨βπ» 2 Β· π 150 Β· π 25 - 84% open Β· β±οΈ 09.12.2024):
Neural-Network-Models-for-Chemistry (π₯10 Β· β 100) - A collection of Nerual Network Models for chemistry. Unlicensed rep-learn
- [GitHub](https://github.com/Eipgen/Neural-Network-Models-for-Chemistry) (π¨βπ» 3 Β· π 16 Β· π 2 - 50% open Β· β±οΈ 31.12.2024):
Awesome Materials Informatics (π₯9 Β· β 400) - Curated list of known efforts in materials informatics, i.e. in modern materials science. Custom
- [GitHub](https://github.com/tilde-lab/awesome-materials-informatics) (π¨βπ» 19 Β· π 85 Β· β±οΈ 18.09.2024):
Awesome Neural Geometry (π₯8 Β· β 940) - A curated collection of resources and research related to the geometry of representations in the brain, deep networks,.. Unlicensed educational rep-learn
- [GitHub](https://github.com/neurreps/awesome-neural-geometry) (π¨βπ» 12 Β· π 59 Β· β±οΈ 25.09.2024):
Awesome-Crystal-GNNs (π₯8 Β· β 76) - This repository contains a collection of resources and papers on GNN Models on Crystal Solid State Materials. MIT
- [GitHub](https://github.com/kdmsit/Awesome-Crystal-GNNs) (π¨βπ» 2 Β· π 9 Β· β±οΈ 22.12.2024):
optimade.science (π₯8 Β· β 8 Β· π€) - A sky-scanner Optimade browser-only GUI. MIT datasets
- [GitHub](https://github.com/tilde-lab/optimade.science) (π¨βπ» 8 Β· π 2 Β· π 26 - 26% open Β· β±οΈ 10.06.2024):
Awesome Neural SBI (π₯7 Β· β 100) - Community-sourced list of papers and resources on neural simulation-based inference. MIT active-learning
- [GitHub](https://github.com/smsharma/awesome-neural-sbi) (π¨βπ» 3 Β· π 7 Β· π 2 - 50% open Β· β±οΈ 23.11.2024):
AI for Science paper collection (π₯7 Β· β 84) - List the AI for Science papers accepted by top conferences. Apache-2
- [GitHub](https://github.com/sherrylixuecheng/AI_for_Science_paper_collection) (π¨βπ» 5 Β· π 9 Β· β±οΈ 14.09.2024):
Awesome-Graph-Generation (π₯6 Β· β 310) - A curated list of up-to-date graph generation papers and resources. Unlicensed rep-learn
- [GitHub](https://github.com/yuanqidu/awesome-graph-generation) (π¨βπ» 4 Β· π 19 Β· β±οΈ 14.10.2024):
The Collection of Database and Dataset Resources in Materials Science (π₯6 Β· β 280) - A list of databases, datasets and books/handbooks where you can find materials properties for machine learning.. Unlicensed datasets
- [GitHub](https://github.com/sedaoturak/data-resources-for-materials-science) (π¨βπ» 2 Β· π 48 Β· π 2 - 50% open Β· β±οΈ 18.12.2024):
Show 7 hidden projects...
- MoLFormers UI (π₯9 Β· β 280 Β· π) - A family of foundation models trained on chemicals.Apache-2 transformer language-models pretrained drug-discovery
- A Highly Opinionated List of Open-Source Materials Informatics Resources (π₯7 Β· β 120 Β· π) - A Highly Opinionated List of Open Source Materials Informatics Resources. MIT
- MADICES Awesome Interoperability (π₯7 Β· β 1) - Linked data interoperability resources of the Machine-actionable data interoperability for the chemical sciences.. MIT datasets
- Geometric-GNNs (π₯4 Β· β 96 Β· π€) - List of Geometric GNNs for 3D atomic systems. Unlicensed datasets educational rep-learn
- Does this material exist? (π₯4 Β· β 15 Β· π€) - Vote on whether you think predicted crystal structures could be synthesised. MIT for-fun materials-discovery
- GitHub topic materials-informatics (π₯1) - GitHub topic materials-informatics. Unlicensed
- MateriApps (π₯1) - A Portal Site of Materials Science Simulation. Unlicensed
Datasets
Datasets, databases and trained models for atomistic ML.
π Alexandria Materials Database - A database of millions of theoretical crystal structures (3D, 2D and 1D) discovered by machine learning accelerated..
π Catalysis Hub - A web-platform for sharing data and software for computational catalysis research!.
π Citrination Datasets - AI-Powered Materials Data Platform. Open Citrination has been decommissioned.
π crystals.ai - Curated datasets for reproducible AI in materials science.
π DeepChem Models - DeepChem models on HuggingFace. model-repository pretrained language-models
π Graphs of Materials Project 20190401 - The dataset used to train the MEGNet interatomic potential. ML-IAP
π HME21 Dataset - High-temperature multi-element 2021 dataset for the PreFerred Potential (PFP).. UIP
π JARVIS-Leaderboard ( β 62) - A large scale benchmark of materials design methods: https://www.nature.com/articles/s41524-024-01259-w. model-repository benchmarking community-resource educational
π Materials Project - Charge Densities - Materials Project has started offering charge density information available for download via their public API.
π Materials Project Trajectory (MPtrj) Dataset - The dataset used to train the CHGNet universal potential. UIP
π matterverse.ai - Database of yet-to-be-sythesized materials predicted using state-of-the-art machine learning algorithms.
π MPF.2021.2.8 - The dataset used to train the M3GNet universal potential. UIP
π NRELMatDB - Computational materials database with the specific focus on materials for renewable energy applications including, but..
π Quantum-Machine.org Datasets - Collection of datasets, including QM7, QM9, etc. MD, DFT. Small organic molecules, mostly.
π sGDML Datasets - MD17, MD22, DFT datasets.
π MoleculeNet - A Benchmark for Molecular Machine Learning. benchmarking
π ZINC15 - A free database of commercially-available compounds for virtual screening. ZINC contains over 230 million purchasable.. graph biomolecules
π ZINC20 - A free database of commercially-available compounds for virtual screening. ZINC contains over 230 million purchasable.. graph biomolecules
FAIR Chemistry datasets (π₯25 Β· β 940 Β· π) - Datasets OC20, OC22, etc. Formerly known as Open Catalyst Project. MIT catalysis
- [GitHub](https://github.com/FAIR-Chem/fairchem) (π¨βπ» 43 Β· π 260 Β· π 250 - 11% open Β· β±οΈ 20.12.2024):
- [PyPi](https://pypi.org/project/fairchem-core) (π₯ 4.8K / month Β· π¦ 3 Β· β±οΈ 19.12.2024):
OPTIMADE Python tools (π₯25 Β· β 72) - Tools for implementing and consuming OPTIMADE APIs in Python. MIT
- [GitHub](https://github.com/Materials-Consortia/optimade-python-tools) (π¨βπ» 28 Β· π 44 Β· π¦ 61 Β· π 470 - 24% open Β· β±οΈ 27.12.2024):
- [PyPi](https://pypi.org/project/optimade) (π¦ 4 Β· β±οΈ 27.12.2024):
- [Conda](https://anaconda.org/conda-forge/optimade) (π₯ 100K Β· β±οΈ 28.12.2024):
MPContribs (π₯22 Β· β 37 Β· π) - Platform for materials scientists to contribute and disseminate their materials data through Materials Project. MIT
- [GitHub](https://github.com/materialsproject/MPContribs) (π¨βπ» 25 Β· π 23 Β· π¦ 41 Β· π 100 - 22% open Β· β±οΈ 30.12.2024):
- [PyPi](https://pypi.org/project/mpcontribs-client) (π¦ 3 Β· β±οΈ 17.10.2024):
load-atoms (π₯18 Β· β 39) - download and manipulate atomistic datasets. MIT data-structures
- [GitHub](https://github.com/jla-gardner/load-atoms) (π¨βπ» 4 Β· π 3 Β· π¦ 5 Β· π 32 - 6% open Β· β±οΈ 16.12.2024):
- [PyPi](https://pypi.org/project/load-atoms) (π₯ 2.3K / month Β· π¦ 2 Β· β±οΈ 13.12.2024):
Open Databases Integration for Materials Design (OPTIMADE) (π₯17 Β· β 83 Β· π€) - Specification of a common REST API for access to materials databases. CC-BY-4.0
- [GitHub](https://github.com/Materials-Consortia/OPTIMADE) (π¨βπ» 21 Β· π 35 Β· π 240 - 28% open Β· β±οΈ 12.06.2024):
Meta Open Materials 2024 (OMat24) Dataset (π₯15 Β· β 930) - Contains over 100 million Density Functional Theory calculations focused on structural and compositional diversity. CC-BY-4.0
- [GitHub]() (π 260):
- [PyPi](https://pypi.org/project/fairchem-core) (π₯ 4.8K / month Β· π¦ 3 Β· β±οΈ 19.12.2024):
QH9 (π₯13 Β· β 550) - A Quantum Hamiltonian Prediction Benchmark. CC-BY-NC-SA-4.0 ML-DFT
- [GitHub](https://github.com/divelab/AIRS) (π¨βπ» 30 Β· π 63 Β· π 20 - 15% open Β· β±οΈ 15.11.2024):
SPICE (π₯11 Β· β 160) - A collection of QM data for training potential functions. MIT ML-IAP MD
- [GitHub](https://github.com/openmm/spice-dataset) (π¨βπ» 1 Β· π 9 Β· π₯ 280 Β· π 69 - 24% open Β· β±οΈ 19.08.2024):
AIS Square (π₯9 Β· β 13) - A collaborative and open-source platform for sharing AI for Science datasets, models, and workflows. Home of the.. LGPL-3.0 community-resource model-repository
- [GitHub](https://github.com/deepmodeling/AIS-Square) (π¨βπ» 8 Β· π 8 Β· π 6 - 83% open Β· β±οΈ 28.12.2024):
Materials Data Facility (MDF) (π₯9 Β· β 10 Β· π€) - A simple way to publish, discover, and access materials datasets. Publication of very large datasets supported (e.g.,.. Apache-2
- [GitHub](https://github.com/materials-data-facility/connect_client) (π¨βπ» 7 Β· π 1 Β· π 7 - 14% open Β· β±οΈ 05.02.2024):
3DSC Database (π₯6 Β· β 16) - Repo for the paper publishing the superconductor database with 3D crystal structures. Custom superconductors materials-discovery
- [GitHub](https://github.com/aimat-lab/3DSC) (π 5 Β· π 2 - 50% open Β· β±οΈ 21.11.2024):
The Perovskite Database Project (π₯5 Β· β 60 Β· π€) - Perovskite Database Project aims at making all perovskite device data, both past and future, available in a form.. Unlicensed community-resource
- [GitHub](https://github.com/Jesperkemist/perovskitedatabase) (π¨βπ» 2 Β· π 20 Β· β±οΈ 07.03.2024):
Show 16 hidden projects...
- ATOM3D (π₯17 Β· β 300 Β· π) - ATOM3D: tasks on molecules in three dimensions.MIT biomolecules benchmarking
- OpenKIM (π₯10 Β· β 32 Β· π) - The Open Knowledgebase of Interatomic Models (OpenKIM) aims to be an online resource for standardized testing, long-.. LGPL-2.1 model-repository knowledge-base pretrained
- 2DMD dataset (π₯9 Β· β 6 Β· π) - Code for Kazeev, N., Al-Maeeni, A.R., Romanov, I. et al. Sparse representation for machine learning the properties of.. Apache-2 material-defect
- ANI-1 Dataset (π₯8 Β· β 96 Β· π) - A data set of 20 million calculated off-equilibrium conformations for organic molecules. MIT
- MoleculeNet Leaderboard (π₯8 Β· β 92 Β· π) - MIT benchmarking
- GEOM (π₯7 Β· β 200 Β· π) - GEOM: Energy-annotated molecular conformations. Unlicensed drug-discovery
- ANI-1x Datasets (π₯6 Β· β 62 Β· π) - The ANI-1ccx and ANI-1x data sets, coupled-cluster and density functional theory properties for organic molecules. MIT
- COMP6 Benchmark dataset (π₯6 Β· β 39 Β· π) - COMP6 Benchmark dataset for ML potentials. MIT
- SciGlass (π₯5 Β· β 12 Β· π) - The database contains a vast set of data on the properties of glass materials. MIT
- GDB-9-Ex9 and ORNL_AISD-Ex (π₯5 Β· β 6 Β· π) - Distributed computing workflow for generation and analysis of large scale molecular datasets obtained running multi-.. Unlicensed
- linear-regression-benchmarks (π₯5 Β· β 1 Β· π) - Data sets used for linear regression benchmarks. MIT benchmarking single-paper
- paper-data-redundancy (π₯4 Β· β 9) - Repo for the paper Exploiting redundancy in large materials datasets for efficient machine learning with less data. BSD-3 small-data single-paper
- Visual Graph Datasets (π₯4 Β· β 2) - Datasets for the training of graph neural networks (GNNs) and subsequent visualization of attributional explanations.. MIT XAI rep-learn
- OPTIMADE providers dashboard (π₯4 Β· β 1) - A dashboard of known providers. Unlicensed
- nep-data (π₯2 Β· β 14 Β· π) - Data related to the NEP machine-learned potential of GPUMD. Unlicensed ML-IAP MD transport-phenomena
- tmQM_wB97MV Dataset (π₯2 Β· β 6 Β· π€) - Code for Applying Large Graph Neural Networks to Predict Transition Metal Complex Energies Using the tmQM_wB97MV.. Unlicensed catalysis rep-learn
Data Structures
Projects that focus on providing data structures used in atomistic machine learning.
dpdata (π₯23 Β· β 200) - A Python package for manipulating atomistic data of software in computational science. LGPL-3.0
- [GitHub](https://github.com/deepmodeling/dpdata) (π¨βπ» 61 Β· π 130 Β· π¦ 130 Β· π 120 - 27% open Β· β±οΈ 20.09.2024):
- [PyPi](https://pypi.org/project/dpdata) (π₯ 21K / month Β· π¦ 40 Β· β±οΈ 20.09.2024):
- [Conda](https://anaconda.org/deepmodeling/dpdata) (π₯ 250 Β· β±οΈ 27.09.2023):
Metatensor (π₯22 Β· β 57) - Self-describing sparse tensor data format for atomistic machine learning and beyond. BSD-3 Rust C-lang C++ Python
- [GitHub](https://github.com/metatensor/metatensor) (π¨βπ» 26 Β· π 18 Β· π₯ 37K Β· π¦ 13 Β· π 220 - 29% open Β· β±οΈ 19.12.2024):
mp-pyrho (π₯17 Β· β 37) - Tools for re-griding volumetric quantum chemistry data for machine-learning purposes. Custom ML-DFT
- [GitHub](https://github.com/materialsproject/pyrho) (π¨βπ» 10 Β· π 7 Β· π¦ 26 Β· π 5 - 40% open Β· β±οΈ 22.10.2024):
- [PyPi](https://pypi.org/project/mp-pyrho) (π₯ 6.6K / month Β· π¦ 5 Β· β±οΈ 22.10.2024):
dlpack (π₯15 Β· β 920) - common in-memory tensor structure. Apache-2 C++
- [GitHub](https://github.com/dmlc/dlpack) (π¨βπ» 24 Β· π 130 Β· π 72 - 41% open Β· β±οΈ 28.09.2024):
Density functional theory (ML-DFT)
Projects and models that focus on quantities of DFT, such as density functional approximations (ML-DFA), the charge density, density of states, the Hamiltonian, etc.
π IKS-PIML - Code and generated data for the paper Inverting the Kohn-Sham equations with physics-informed machine learning.. neural-operator pinn datasets single-paper
JAX-DFT (π₯25 Β· β 35K) - This library provides basic building blocks that can construct DFT calculations as a differentiable program. Apache-2
- [GitHub](https://github.com/google-research/google-research) (π¨βπ» 820 Β· π 7.9K Β· π 1.8K - 81% open Β· β±οΈ 13.12.2024):
MALA (π₯20 Β· β 82) - Materials Learning Algorithms. A framework for machine learning materials properties from first-principles data. BSD-3
- [GitHub](https://github.com/mala-project/mala) (π¨βπ» 44 Β· π 26 Β· π¦ 2 Β· π 290 - 10% open Β· β±οΈ 13.12.2024):
QHNet (π₯13 Β· β 550) - Artificial Intelligence Research for Science (AIRS). GPL-3.0 rep-learn
- [GitHub](https://github.com/divelab/AIRS) (π¨βπ» 30 Β· π 63 Β· π 20 - 15% open Β· β±οΈ 15.11.2024):
SALTED (π₯12 Β· β 32) - Symmetry-Adapted Learning of Three-dimensional Electron Densities. GPL-3.0
- [GitHub](https://github.com/andreagrisafi/SALTED) (π¨βπ» 17 Β· π 4 Β· π 7 - 28% open Β· β±οΈ 27.09.2024):
DeepH-pack (π₯11 Β· β 250) - Deep neural networks for density functional theory Hamiltonian. LGPL-3.0 Julia
- [GitHub](https://github.com/mzjb/DeepH-pack) (π¨βπ» 8 Β· π 44 Β· π 55 - 29% open Β· β±οΈ 07.10.2024):
Grad DFT (π₯10 Β· β 82 Β· π€) - GradDFT is a JAX-based library enabling the differentiable design and experimentation of exchange-correlation.. Apache-2
- [GitHub](https://github.com/XanaduAI/GradDFT) (π¨βπ» 4 Β· π 8 Β· π 54 - 20% open Β· β±οΈ 13.02.2024):
DeePKS-kit (π₯9 Β· β 100 Β· π€) - a package for developing machine learning-based chemically accurate energy and density functional models. LGPL-3.0
- [GitHub](https://github.com/deepmodeling/deepks-kit) (π¨βπ» 7 Β· π 36 Β· π 24 - 41% open Β· β±οΈ 13.04.2024):
Q-stack (π₯9 Β· β 15) - Stack of codes for dedicated pre- and post-processing tasks for Quantum Machine Learning (QML). MIT excited-states general-tool
- [GitHub](https://github.com/lcmd-epfl/Q-stack) (π¨βπ» 7 Β· π 5 Β· π 29 - 27% open Β· β±οΈ 11.12.2024):
HamGNN (π₯8 Β· β 72) - An E(3) equivariant Graph Neural Network for predicting electronic Hamiltonian matrix. GPL-3.0 rep-learn magnetism C-lang
- [GitHub](https://github.com/QuantumLab-ZY/HamGNN) (π¨βπ» 2 Β· π 15 Β· π 35 - 82% open Β· β±οΈ 27.12.2024):
ChargE3Net (π₯5 Β· β 41) - Higher-order equivariant neural networks for charge density prediction in materials. MIT rep-learn
- [GitHub](https://github.com/AIforGreatGood/charge3net) (π¨βπ» 2 Β· π 12 Β· π 7 - 42% open Β· β±οΈ 30.10.2024):
Show 22 hidden projects...
- DM21 (π₯20 Β· β 13K Β· π) - This package provides a PySCF interface to the DM21 (DeepMind 21) family of exchange-correlation functionals described..Apache-2
- NeuralXC (π₯10 Β· β 34 Β· π) - Implementation of a machine learned density functional. BSD-3
- ACEhamiltonians (π₯10 Β· β 15 Β· π) - Provides tools for constructing, fitting, and predicting self-consistent Hamiltonian and overlap matrices in solid-.. MIT Julia
- PROPhet (π₯9 Β· β 64 Β· π) - PROPhet is a code to integrate machine learning techniques with first-principles quantum chemistry approaches. GPL-3.0 ML-IAP MD single-paper C++
- DeepH-E3 (π₯7 Β· β 83 Β· π) - General framework for E(3)-equivariant neural network representation of density functional theory Hamiltonian. MIT magnetism
- Mat2Spec (π₯7 Β· β 28 Β· π) - Density of States Prediction for Materials Discovery via Contrastive Learning from Probabilistic Embeddings. MIT spectroscopy
- Libnxc (π₯7 Β· β 17 Β· π) - A library for using machine-learned exchange-correlation functionals for density-functional theory. MPL-2.0 C++ Fortran
- DeepDFT (π₯6 Β· β 66 Β· π) - Official implementation of DeepDFT model. MIT
- charge-density-models (π₯6 Β· β 10 Β· π) - Tools to build charge density models using [fairchem](https://github.com/FAIR-Chem/fairchem). MIT rep-learn
- KSR-DFT (π₯6 Β· β 4 Β· π) - Kohn-Sham regularizer for machine-learned DFT functionals. Apache-2
- xDeepH (π₯5 Β· β 34 Β· π) - Extended DeepH (xDeepH) method for magnetic materials. LGPL-3.0 magnetism Julia
- ML-DFT (π₯5 Β· β 23 Β· π) - A package for density functional approximation using machine learning. MIT
- InfGCN for Electron Density Estimation (π₯5 Β· β 12 Β· π) - Official implementation of the NeurIPS 23 spotlight paper of InfGCN. MIT rep-learn neural-operator
- rho_learn (π₯5 Β· β 4 Β· π) - A proof-of-concept workflow for torch-based electron density learning. MIT
- DeepCDP (π₯4 Β· β 6 Β· π) - DeepCDP: Deep learning Charge Density Prediction. Unlicensed
- gprep (π₯4 Β· π) - Fitting DFTB repulsive potentials with GPR. MIT single-paper
- APET (π₯3 Β· β 4 Β· π) - Atomic Positional Embedding-based Transformer. GPL-3.0 density-of-states transformer
- CSNN (π₯3 Β· β 2 Β· π) - Primary codebase of CSNN - Concentric Spherical Neural Network for 3D Representation Learning. BSD-3
- MALADA (π₯3 Β· β 1) - MALA Data Acquisition: Helpful tools to build data for MALA. BSD-3
- A3MD (π₯2 Β· β 8 Β· π) - MPNN-like + Analytic Density Model = Accurate electron densities. Unlicensed rep-learn single-paper
- MLDensity (π₯1 Β· β 3 Β· π) - Linear Jacobi-Legendre expansion of the charge density for machine learning-accelerated electronic structure.. Unlicensed
- kdft (π₯1 Β· β 2 Β· π) - The Kernel Density Functional (KDF) code allows generating ML based DFT functionals. Unlicensed
Educational Resources
Tutorials, guides, cookbooks, recipes, etc.
π AI for Science 101 community-resource rep-learn
π AL4MS 2023 workshop tutorials active-learning
π Quantum Chemistry in the Age of Machine Learning - Book, 2022.
AI4Chemistry course (π₯11 Β· β 160 Β· π€) - EPFL AI for chemistry course, Spring 2023. https://schwallergroup.github.io/ai4chem_course. MIT chemistry
- [GitHub](https://github.com/schwallergroup/ai4chem_course) (π¨βπ» 6 Β· π 37 Β· π 4 - 25% open Β· β±οΈ 02.05.2024):
jarvis-tools-notebooks (π₯9 Β· β 70) - A Google-Colab Notebook Collection for Materials Design: https://jarvis.nist.gov/. NIST
- [GitHub](https://github.com/JARVIS-Materials-Design/jarvis-tools-notebooks) (π¨βπ» 5 Β· π 26 Β· β±οΈ 14.08.2024):
DSECOP (π₯9 Β· β 44 Β· π€) - This repository contains data science educational materials developed by DSECOP Fellows. CCO-1.0
- [GitHub](https://github.com/GDS-Education-Community-of-Practice/DSECOP) (π¨βπ» 14 Β· π 26 Β· π 8 - 12% open Β· β±οΈ 26.06.2024):
iam-notebooks (π₯8 Β· β 26) - Jupyter notebooks for the lectures of the Introduction to Atomistic Modeling. Apache-2
- [GitHub](https://github.com/ceriottm/iam-notebooks) (π¨βπ» 6 Β· π 5 Β· β±οΈ 09.10.2024):
COSMO Software Cookbook (π₯8 Β· β 17) - A cookbook with recipes for atomic-scale modeling of materials and molecules. BSD-3
- [GitHub](https://github.com/lab-cosmo/atomistic-cookbook) (π¨βπ» 11 Β· π 1 Β· π 12 - 8% open Β· β±οΈ 20.12.2024):
MACE-tutorials (π₯6 Β· β 43) - Another set of tutorials for the MACE interatomic potential by one of the authors. MIT ML-IAP rep-learn MD
- [GitHub](https://github.com/ilyes319/mace-tutorials) (π¨βπ» 2 Β· π 11 Β· β±οΈ 16.07.2024):
Show 19 hidden projects...
- Geometric GNN Dojo (π₯12 Β· β 480 Β· π) - New to geometric GNNs: try our practical notebook, prepared for MPhil students at the University of Cambridge.MIT rep-learn
- DeepLearningLifeSciences (π₯12 Β· β 360 Β· π) - Example code from the book Deep Learning for the Life Sciences. MIT
- Deep Learning for Molecules and Materials Book (π₯11 Β· β 630 Β· π) - Deep learning for molecules and materials book. Custom
- OPTIMADE Tutorial Exercises (π₯9 Β· β 15 Β· π) - Tutorial exercises for the OPTIMADE API. MIT datasets
- RDKit Tutorials (π₯8 Β· β 270 Β· π) - Tutorials to learn how to work with the RDKit. Custom
- BestPractices (π₯8 Β· β 180 Β· π) - Things that you should (and should not) do in your Materials Informatics research. MIT
- MAChINE (π₯7 Β· β 1 Β· π) - Client-Server Web App to introduce usage of ML in materials science to beginners. MIT
- Applied AI for Materials (π₯6 Β· β 59 Β· π) - Course materials for Applied AI for Materials Science and Engineering. Unlicensed
- ML for catalysis tutorials (π₯6 Β· β 8 Β· π) - A jupyter book repo for tutorial on how to use OCP ML models for catalysis. MIT
- AI4Science101 (π₯5 Β· β 86 Β· π) - AI for Science. Unlicensed
- Machine Learning for Materials Hard and Soft (π₯5 Β· β 35 Β· π) - ESI-DCAFM-TACO-VDSP Summer School on Machine Learning for Materials Hard and Soft. Unlicensed
- Data Handling, DoE and Statistical Analysis for Material Chemists (π₯5 Β· β 2 Β· π) - Notebooks for workshops of DoE course, hosted by the Computational Materials Chemistry group at Uppsala University. GPL-3.0
- ML-in-chemistry-101 (π₯4 Β· β 72 Β· π) - The course materials for Machine Learning in Chemistry 101. Unlicensed
- chemrev-gpr (π₯4 Β· β 10 Β· π) - Notebooks accompanying the paper on GPR in materials and molecules in Chemical Reviews 2020. Unlicensed
- PiNN Lab (π₯4 Β· β 3 Β· π) - Material for running a lab session on atomic neural networks. GPL-3.0
- AI4ChemMat Hands-On Series (π₯4 Β· β 1 Β· π€) - Hands-On Series organized by Chemistry and Materials working group at Argonne Nat Lab. MPL-2.0
- MLDensity_tutorial (π₯2 Β· β 9 Β· π) - Tutorial files to work with ML for the charge density in molecules and solids. Unlicensed
- LAMMPS-style pair potentials with GAP (π₯2 Β· β 4 Β· π) - A tutorial on how to create LAMMPS-style pair potentials and use them in combination with GAP potentials to run MD.. Unlicensed ML-IAP MD rep-eng
- MALA Tutorial (π₯2 Β· β 2 Β· π) - A full MALA hands-on tutorial. Unlicensed
Explainable Artificial intelligence (XAI)
Projects that focus on explainability and model interpretability in atomistic ML.
exmol (π₯21 Β· β 290) - Explainer for black box models that predict molecule properties. MIT
- [GitHub](https://github.com/ur-whitelab/exmol) (π¨βπ» 7 Β· π 42 Β· π¦ 23 Β· π 71 - 16% open Β· β±οΈ 22.11.2024):
- [PyPi](https://pypi.org/project/exmol) (π₯ 1.3K / month Β· π¦ 1 Β· β±οΈ 22.11.2024):
MEGAN: Multi Explanation Graph Attention Student (π₯5 Β· β 8) - Minimal implementation of graph attention student model architecture. MIT rep-learn
- [GitHub](https://github.com/aimat-lab/graph_attention_student) (π¨βπ» 2 Β· π 1 Β· π 3 - 33% open Β· β±οΈ 07.10.2024):
Show 1 hidden projects...
- Linear vs blackbox (π₯3 Β· β 2 Β· π) - Code and data related to the publication: Interpretable models for extrapolation in scientific machine learning.MIT XAI single-paper rep-eng
Electronic structure methods (ML-ESM)
Projects and models that focus on quantities of electronic structure methods, which do not fit into either of the categories ML-WFT or ML-DFT.
Show 5 hidden projects...
- QDF for molecule (π₯8 Β· β 210 Β· π) - Quantum deep field: data-driven wave function, electron density generation, and energy prediction and extrapolation..MIT
- QMLearn (π₯5 Β· β 11 Β· π) - Quantum Machine Learning by learning one-body reduced density matrices in the AO basis... MIT
- q-pac (π₯5 Β· β 4 Β· π) - Kernel charge equilibration method. MIT electrostatics
- halex (π₯5 Β· β 3 Β· π€) - Hamiltonian Learning for Excited States https://doi.org/10.48550/arXiv.2311.00844. Unlicensed excited-states
- e3psi (π₯3 Β· β 3 Β· π€) - Equivariant machine learning library for learning from electronic structures. LGPL-3.0
General Tools
General tools for atomistic machine learning.
RDKit (π₯36 Β· β 2.7K) - BSD-3 C++
- [GitHub](https://github.com/rdkit/rdkit) (π¨βπ» 240 Β· π 880 Β· π₯ 870 Β· π¦ 3 Β· π 3.7K - 18% open Β· β±οΈ 25.12.2024):
- [PyPi](https://pypi.org/project/rdkit) (π₯ 1.4M / month Β· π¦ 840 Β· β±οΈ 29.12.2024):
- [Conda](https://anaconda.org/rdkit/rdkit) (π₯ 2.6M Β· β±οΈ 16.06.2023):
DeepChem (π₯34 Β· β 5.6K) - Democratizing Deep-Learning for Drug Discovery, Quantum Chemistry, Materials Science and Biology. MIT
- [GitHub](https://github.com/deepchem/deepchem) (π¨βπ» 250 Β· π 1.7K Β· π¦ 480 Β· π 1.9K - 34% open Β· β±οΈ 24.12.2024):
- [PyPi](https://pypi.org/project/deepchem) (π₯ 51K / month Β· π¦ 14 Β· β±οΈ 24.12.2024):
- [Conda](https://anaconda.org/conda-forge/deepchem) (π₯ 110K Β· β±οΈ 05.04.2024):
- [Docker Hub](https://hub.docker.com/r/deepchemio/deepchem) (π₯ 8K Β· β 5 Β· β±οΈ 24.12.2024):
Matminer (π₯28 Β· β 490) - Data mining for materials science. Custom
- [GitHub](https://github.com/hackingmaterials/matminer) (π¨βπ» 56 Β· π 190 Β· π¦ 350 Β· π 230 - 13% open Β· β±οΈ 11.10.2024):
- [PyPi](https://pypi.org/project/matminer) (π₯ 15K / month Β· π¦ 60 Β· β±οΈ 06.10.2024):
- [Conda](https://anaconda.org/conda-forge/matminer) (π₯ 78K Β· β±οΈ 21.12.2024):
QUIP (π₯24 Β· β 360) - libAtoms/QUIP molecular dynamics framework: https://libatoms.github.io. GPL-2.0 MD ML-IAP rep-eng Fortran
- [GitHub](https://github.com/libAtoms/QUIP) (π¨βπ» 85 Β· π 120 Β· π₯ 730 Β· π¦ 45 Β· π 470 - 22% open Β· β±οΈ 27.09.2024):
- [PyPi](https://pypi.org/project/quippy-ase) (π₯ 2.6K / month Β· π¦ 4 Β· β±οΈ 15.01.2023):
- [Docker Hub](https://hub.docker.com/r/libatomsquip/quip) (π₯ 10K Β· β 4 Β· β±οΈ 24.04.2023):
JARVIS-Tools (π₯23 Β· β 320) - JARVIS-Tools: an open-source software package for data-driven atomistic materials design. Publications:.. Custom
- [GitHub](https://github.com/usnistgov/jarvis) (π¨βπ» 15 Β· π 120 Β· π¦ 110 Β· π 92 - 51% open Β· β±οΈ 20.11.2024):
- [PyPi](https://pypi.org/project/jarvis-tools) (π₯ 19K / month Β· π¦ 31 Β· β±οΈ 20.11.2024):
- [Conda](https://anaconda.org/conda-forge/jarvis-tools) (π₯ 87K Β· β±οΈ 20.11.2024):
MAML (π₯21 Β· β 380) - Python for Materials Machine Learning, Materials Descriptors, Machine Learning Force Fields, Deep Learning, etc. BSD-3
- [GitHub](https://github.com/materialsvirtuallab/maml) (π¨βπ» 33 Β· π 79 Β· π¦ 12 Β· π 71 - 12% open Β· β±οΈ 06.11.2024):
- [PyPi](https://pypi.org/project/maml) (π₯ 460 / month Β· π¦ 2 Β· β±οΈ 13.06.2024):
MAST-ML (π₯19 Β· β 110) - MAterials Simulation Toolkit for Machine Learning (MAST-ML). MIT
- [GitHub](https://github.com/uw-cmg/MAST-ML) (π¨βπ» 19 Β· π 61 Β· π₯ 140 Β· π¦ 45 Β· π 220 - 14% open Β· β±οΈ 09.10.2024):
QML (π₯18 Β· β 200) - QML: Quantum Machine Learning. MIT
- [GitHub](https://github.com/qmlcode/qml) (π¨βπ» 10 Β· π 84 Β· π 59 - 64% open Β· β±οΈ 08.12.2024):
- [PyPi](https://pypi.org/project/qml) (π₯ 390 / month Β· β±οΈ 13.08.2018):
Scikit-Matter (π₯17 Β· β 77) - A collection of scikit-learn compatible utilities that implement methods born out of the materials science and.. BSD-3 scikit-learn
- [GitHub](https://github.com/scikit-learn-contrib/scikit-matter) (π¨βπ» 15 Β· π 19 Β· π¦ 11 Β· π 70 - 20% open Β· β±οΈ 09.10.2024):
- [PyPi](https://pypi.org/project/skmatter) (π₯ 1.8K / month Β· β±οΈ 24.08.2023):
- [Conda](https://anaconda.org/conda-forge/skmatter) (π₯ 2.6K Β· β±οΈ 24.08.2023):
MLatom (π₯16 Β· β 72) - AI-enhanced computational chemistry. MIT UIP ML-IAP MD ML-DFT ML-ESM transfer-learning active-learning spectroscopy structure-optimization
- [GitHub](https://github.com/dralgroup/mlatom) (π¨βπ» 4 Β· π 11 Β· π 5 - 20% open Β· β±οΈ 18.12.2024):
- [PyPi](https://pypi.org/project/mlatom) (π₯ 3.4K / month Β· β±οΈ 18.12.2024):
XenonPy (π₯15 Β· β 140 Β· π€) - XenonPy is a Python Software for Materials Informatics. BSD-3
- [GitHub](https://github.com/yoshida-lab/XenonPy) (π¨βπ» 10 Β· π 61 Β· π₯ 1.5K Β· π 87 - 24% open Β· β±οΈ 21.04.2024):
- [PyPi](https://pypi.org/project/xenonpy) (π₯ 890 / month Β· π¦ 1 Β· β±οΈ 31.10.2022):
Artificial Intelligence for Science (AIRS) (π₯13 Β· β 550) - Artificial Intelligence Research for Science (AIRS). GPL-3.0 license rep-learn generative ML-IAP MD ML-DFT ML-WFT biomolecules
- [GitHub](https://github.com/divelab/AIRS) (π¨βπ» 30 Β· π 63 Β· π 20 - 15% open Β· β±οΈ 15.11.2024):
Show 10 hidden projects...
- Automatminer (π₯15 Β· β 140 Β· π) - An automatic engine for predicting materials properties.Custom autoML
- AMPtorch (π₯11 Β· β 60 Β· π) - AMPtorch: Atomistic Machine Learning Package (AMP) - PyTorch. GPL-3.0
- OpenChem (π₯10 Β· β 680 Β· π) - OpenChem: Deep Learning toolkit for Computational Chemistry and Drug Design Research. MIT
- JAXChem (π₯7 Β· β 79 Β· π) - JAXChem is a JAX-based deep learning library for complex and versatile chemical modeling. MIT
- uncertainty_benchmarking (π₯7 Β· β 41 Β· π) - Various code/notebooks to benchmark different ways we could estimate uncertainty in ML predictions. Unlicensed benchmarking probabilistic
- torchchem (π₯7 Β· β 35 Β· π) - An experimental repo for experimenting with PyTorch models. MIT
- Equisolve (π₯6 Β· β 5 Β· π) - A ML toolkit package utilizing the metatensor data format to build models for the prediction of equivariant properties.. BSD-3 ML-IAP
- ACEatoms (π₯4 Β· β 2 Β· π) - Generic code for modelling atomic properties using ACE. Custom Julia
- Magpie (π₯3) - Materials Agnostic Platform for Informatics and Exploration (Magpie). MIT Java
- quantum-structure-ml (π₯2 Β· β 2 Β· π) - Multi-class classification model for predicting the magnetic order of magnetic structures and a binary classification.. Unlicensed magnetism benchmarking
Generative Models
Projects that implement generative models for atomistic ML.
GT4SD (π₯18 Β· β 340 Β· π) - GT4SD, an open-source library to accelerate hypothesis generation in the scientific discovery process. MIT pretrained drug-discovery rep-learn
- [GitHub](https://github.com/GT4SD/gt4sd-core) (π¨βπ» 20 Β· π 72 Β· π 120 - 12% open Β· β±οΈ 12.09.2024):
- [PyPi](https://pypi.org/project/gt4sd) (π₯ 2.4K / month Β· β±οΈ 12.09.2024):
MoLeR (π₯15 Β· β 280 Β· π€) - Implementation of MoLeR: a generative model of molecular graphs which supports scaffold-constrained generation. MIT
- [GitHub](https://github.com/microsoft/molecule-generation) (π¨βπ» 5 Β· π 41 Β· π 40 - 22% open Β· β±οΈ 03.01.2024):
- [PyPi](https://pypi.org/project/molecule-generation) (π₯ 240 / month Β· π¦ 1 Β· β±οΈ 05.01.2024):
PMTransformer (π₯14 Β· β 89 Β· π€) - Universal Transfer Learning in Porous Materials, including MOFs. MIT transfer-learning pretrained transformer
- [GitHub](https://github.com/hspark1212/MOFTransformer) (π¨βπ» 2 Β· π 13 Β· π¦ 8 Β· β±οΈ 20.06.2024):
- [PyPi](https://pypi.org/project/moftransformer) (π₯ 570 / month Β· π¦ 1 Β· β±οΈ 20.06.2024):
SiMGen (π₯13 Β· β 17) - Zero Shot Molecular Generation via Similarity Kernels. MIT viz
- [GitHub](https://github.com/RokasEl/simgen) (π¨βπ» 4 Β· π 2 Β· π¦ 2 Β· π 4 - 25% open Β· β±οΈ 13.12.2024):
- [PyPi](https://pypi.org/project/simgen) (π₯ 200 / month Β· β±οΈ 13.12.2024):
SchNetPack G-SchNet (π₯12 Β· β 52) - G-SchNet extension for SchNetPack. MIT
- [GitHub](https://github.com/atomistic-machine-learning/schnetpack-gschnet) (π¨βπ» 3 Β· π 8 Β· π 16 - 6% open Β· β±οΈ 07.11.2024):
COATI (π₯5 Β· β 100 Β· π€) - COATI: multi-modal contrastive pre-training for representing and traversing chemical space. Apache-2 drug-discovery multimodal pretrained rep-learn
- [GitHub](https://github.com/terraytherapeutics/COATI) (π¨βπ» 5 Β· π 6 Β· π 3 - 33% open Β· β±οΈ 23.03.2024):
Show 8 hidden projects...
- synspace (π₯12 Β· β 36 Β· π) - Synthesis generative model.MIT
- EDM (π₯9 Β· β 460 Β· π) - E(3) Equivariant Diffusion Model for Molecule Generation in 3D. MIT
- G-SchNet (π₯8 Β· β 130 Β· π) - G-SchNet - a generative model for 3d molecular structures. MIT
- bVAE-IM (π₯8 Β· β 11 Β· π) - Implementation of Chemical Design with GPU-based Ising Machine. MIT QML single-paper
- cG-SchNet (π₯7 Β· β 54 Β· π) - cG-SchNet - a conditional generative neural network for 3d molecular structures. MIT
- rxngenerator (π₯6 Β· β 12 Β· π) - A generative model for molecular generation via multi-step chemical reactions. MIT
- MolSLEPA (π₯5 Β· β 5 Β· π) - Interpretable Fragment-based Molecule Design with Self-learning Entropic Population Annealing. MIT XAI
- Mapping out phase diagrams with generative classifiers (π₯4 Β· β 7 Β· π) - Repository for our ``Mapping out phase diagrams with generative models paper. MIT phase-transition
Interatomic Potentials (ML-IAP)
Machine learning interatomic potentials (aka ML-IAP, MLIAP, MLIP, MLP) and force fields (ML-FF) for molecular dynamics.
DeePMD-kit (π₯28 Β· β 1.5K) - A deep learning package for many-body potential energy representation and molecular dynamics. LGPL-3.0 C++
- [GitHub](https://github.com/deepmodeling/deepmd-kit) (π¨βπ» 73 Β· π 520 Β· π₯ 46K Β· π¦ 22 Β· π 870 - 10% open Β· β±οΈ 23.12.2024):
- [PyPi](https://pypi.org/project/deepmd-kit) (π₯ 6K / month Β· π¦ 4 Β· β±οΈ 23.12.2024):
- [Conda](https://anaconda.org/deepmodeling/deepmd-kit) (π₯ 1.7K Β· β±οΈ 06.04.2024):
- [Docker Hub](https://hub.docker.com/r/deepmodeling/deepmd-kit) (π₯ 3.3K Β· β 1 Β· β±οΈ 25.11.2024):
fairchem (π₯25 Β· β 940) - FAIR Chemistrys library of machine learning methods for chemistry. Formerly known as Open Catalyst Project. MIT pretrained UIP rep-learn catalysis
- [GitHub](https://github.com/FAIR-Chem/fairchem) (π¨βπ» 43 Β· π 260 Β· π 250 - 11% open Β· β±οΈ 20.12.2024):
- [PyPi](https://pypi.org/project/fairchem-core) (π₯ 4.8K / month Β· π¦ 3 Β· β±οΈ 19.12.2024):
DP-GEN (π₯23 Β· β 320) - The deep potential generator to generate a deep-learning based model of interatomic potential energy and force field. LGPL-3.0 workflows
- [GitHub](https://github.com/deepmodeling/dpgen) (π¨βπ» 69 Β· π 180 Β· π₯ 1.9K Β· π¦ 7 Β· π 310 - 14% open Β· β±οΈ 23.11.2024):
- [PyPi](https://pypi.org/project/dpgen) (π₯ 870 / month Β· π¦ 2 Β· β±οΈ 23.11.2024):
- [Conda](https://anaconda.org/deepmodeling/dpgen) (π₯ 220 Β· β±οΈ 16.06.2023):
NequIP (π₯22 Β· β 660) - NequIP is a code for building E(3)-equivariant interatomic potentials. MIT
- [GitHub](https://github.com/mir-group/nequip) (π¨βπ» 12 Β· π 140 Β· π¦ 33 Β· π 98 - 25% open Β· β±οΈ 14.11.2024):
- [PyPi](https://pypi.org/project/nequip) (π₯ 1.6K / month Β· π¦ 1 Β· β±οΈ 09.07.2024):
- [Conda](https://anaconda.org/conda-forge/nequip) (π₯ 7.1K Β· β±οΈ 31.12.2024):
MACE (π₯22 Β· β 580) - MACE - Fast and accurate machine learning interatomic potentials with higher order equivariant message passing. MIT
- [GitHub](https://github.com/ACEsuit/mace) (π¨βπ» 47 Β· π 210 Β· π 320 - 21% open Β· β±οΈ 20.12.2024):
GPUMD (π₯22 Β· β 500) - GPUMD is a highly efficient general-purpose molecular dynamic (MD) package and enables machine-learned potentials.. GPL-3.0 MD C++ electrostatics
- [GitHub](https://github.com/brucefan1983/GPUMD) (π¨βπ» 42 Β· π 120 Β· π 190 - 11% open Β· β±οΈ 02.01.2025):
TorchMD-NET (π₯21 Β· β 350) - Training neural network potentials. MIT MD rep-learn transformer pretrained
- [GitHub](https://github.com/torchmd/torchmd-net) (π¨βπ» 16 Β· π 75 Β· π 130 - 34% open Β· β±οΈ 03.12.2024):
- [Conda](https://anaconda.org/conda-forge/torchmd-net) (π₯ 270K Β· β±οΈ 03.12.2024):
apax (π₯19 Β· β 19) - A flexible and performant framework for training machine learning potentials. MIT
- [GitHub](https://github.com/apax-hub/apax) (π¨βπ» 8 Β· π 3 Β· π¦ 3 Β· π 140 - 13% open Β· β±οΈ 17.12.2024):
- [PyPi](https://pypi.org/project/apax) (π₯ 600 / month Β· β±οΈ 03.12.2024):
Neural Force Field (π₯16 Β· β 250) - Neural Network Force Field based on PyTorch. MIT pretrained
- [GitHub](https://github.com/learningmatter-mit/NeuralForceField) (π¨βπ» 42 Β· π 51 Β· π 21 - 14% open Β· β±οΈ 06.12.2024):
n2p2 (π₯16 Β· β 230) - n2p2 - A Neural Network Potential Package. GPL-3.0 C++
- [GitHub](https://github.com/CompPhysVienna/n2p2) (π¨βπ» 11 Β· π 78 Β· π 150 - 44% open Β· β±οΈ 24.11.2024):
NNPOps (π₯15 Β· β 88) - High-performance operations for neural network potentials. MIT MD C++
- [GitHub](https://github.com/openmm/NNPOps) (π¨βπ» 9 Β· π 18 Β· π 57 - 38% open Β· β±οΈ 10.07.2024):
- [Conda](https://anaconda.org/conda-forge/nnpops) (π₯ 310K Β· β±οΈ 14.11.2024):
PyXtalFF (π₯15 Β· β 87 Β· π€) - Machine Learning Interatomic Potential Predictions. MIT
- [GitHub](https://github.com/MaterSim/PyXtal_FF) (π¨βπ» 9 Β· π 23 Β· π 63 - 19% open Β· β±οΈ 07.01.2024):
- [PyPi](https://pypi.org/project/pyxtal_ff) (π₯ 210 / month Β· β±οΈ 21.12.2022):
KLIFF (π₯15 Β· β 34) - KIM-based Learning-Integrated Fitting Framework for interatomic potentials. LGPL-2.1 probabilistic workflows
- [GitHub](https://github.com/openkim/kliff) (π¨βπ» 9 Β· π 19 Β· π¦ 4 Β· π 42 - 54% open Β· β±οΈ 08.10.2024):
- [PyPi](https://pypi.org/project/kliff) (π₯ 270 / month Β· β±οΈ 17.12.2023):
- [Conda](https://anaconda.org/conda-forge/kliff) (π₯ 130K Β· β±οΈ 10.09.2024):
Ultra-Fast Force Fields (UF3) (π₯14 Β· β 62) - UF3: a python library for generating ultra-fast interatomic potentials. Apache-2
- [GitHub](https://github.com/uf3/uf3) (π¨βπ» 10 Β· π 22 Β· π¦ 2 Β· π 50 - 38% open Β· β±οΈ 04.10.2024):
- [PyPi](https://pypi.org/project/uf3) (π₯ 57 / month Β· β±οΈ 27.10.2023):
MLIPX - Machine-Learned Interatomic Potential eXploration (π₯14 Β· β 62 Β· π£) - Machine-Learned Interatomic Potential eXploration (mlipx) is designed at BASF for evaluating machine-learned.. MIT benchmarking viz workflows
- [GitHub](https://github.com/basf/mlipx) (π¨βπ» 4 Β· π 4 Β· π 4 - 50% open Β· β±οΈ 12.12.2024):
- [PyPi](https://pypi.org/project/mlipx) (π₯ 860 / month Β· β±οΈ 12.12.2024):
wfl (π₯14 Β· β 36) - Workflow is a Python toolkit for building interatomic potential creation and atomistic simulation workflows. GPL-2.0 workflows HTC
- [GitHub](https://github.com/libAtoms/workflow) (π¨βπ» 19 Β· π 19 Β· π¦ 2 Β· π 160 - 41% open Β· β±οΈ 04.12.2024):
PiNN (π₯13 Β· β 110) - A Python library for building atomic neural networks. BSD-3
- [GitHub](https://github.com/Teoroo-CMC/PiNN) (π¨βπ» 6 Β· π 33 Β· π 7 - 14% open Β· β±οΈ 20.12.2024):
- [Docker Hub](https://hub.docker.com/r/teoroo/pinn) (π₯ 380 Β· β±οΈ 20.12.2024):
So3krates (MLFF) (π₯13 Β· β 100) - Build neural networks for machine learning force fields with JAX. MIT
- [GitHub](https://github.com/thorben-frank/mlff) (π¨βπ» 4 Β· π 22 Β· π 10 - 40% open Β· β±οΈ 23.08.2024):
ANI-1 (π₯12 Β· β 220 Β· π€) - ANI-1 neural net potential with python interface (ASE). MIT
- [GitHub](https://github.com/isayev/ASE_ANI) (π¨βπ» 6 Β· π 54 Β· π 37 - 43% open Β· β±οΈ 11.03.2024):
DMFF (π₯12 Β· β 160 Β· π€) - DMFF (Differentiable Molecular Force Field) is a Jax-based python package that provides a full differentiable.. LGPL-3.0
- [GitHub](https://github.com/deepmodeling/DMFF) (π¨βπ» 14 Β· π 45 Β· π 27 - 40% open Β· β±οΈ 12.01.2024):
Pacemaker (π₯12 Β· β 73) - Python package for fitting atomic cluster expansion (ACE) potentials. Custom
- [GitHub](https://github.com/ICAMS/python-ace) (π¨βπ» 7 Β· π 19 Β· π 58 - 34% open Β· β±οΈ 20.11.2024):
- [PyPi](https://pypi.org/project/python-ace) (π₯ 15 / month Β· β±οΈ 24.10.2022):
CCS_fit (π₯12 Β· β 8 Β· π€) - Curvature Constrained Splines. GPL-3.0
- [GitHub](https://github.com/Teoroo-CMC/CCS) (π¨βπ» 8 Β· π 11 Β· π₯ 750 Β· π 14 - 57% open Β· β±οΈ 16.02.2024):
- [PyPi](https://pypi.org/project/ccs_fit) (π₯ 2.5K / month Β· β±οΈ 16.02.2024):
PyNEP (π₯11 Β· β 50) - A python interface of the machine learning potential NEP used in GPUMD. MIT
- [GitHub](https://github.com/bigd4/PyNEP) (π¨βπ» 9 Β· π 16 Β· π 11 - 36% open Β· β±οΈ 15.12.2024):
calorine (π₯11 Β· β 14) - A Python package for constructing and sampling neuroevolution potential models. https://doi.org/10.21105/joss.06264. Custom
- [PyPi](https://pypi.org/project/calorine) (π₯ 1.4K / month Β· π¦ 4 Β· β±οΈ 25.10.2024):
- [GitLab](https://gitlab.com/materials-modeling/calorine) (π 4 Β· π 91 - 5% open Β· β±οΈ 25.10.2024):
Allegro (π₯10 Β· β 370) - Allegro is an open-source code for building highly scalable and accurate equivariant deep learning interatomic.. MIT
- [GitHub](https://github.com/mir-group/allegro) (π¨βπ» 2 Β· π 46 Β· π 40 - 52% open Β· β±οΈ 14.11.2024):
ACE.jl (π₯10 Β· β 65) - Parameterisation of Equivariant Properties of Particle Systems. Custom Julia
- [GitHub](https://github.com/ACEsuit/ACE.jl) (π¨βπ» 12 Β· π 15 Β· π 82 - 29% open Β· β±οΈ 17.12.2024):
Asparagus (π₯10 Β· β 9 Β· π£) - Program Package for Sampling, Training and Applying ML-based Potential models https://doi.org/10.48550/arXiv.2407.15175. MIT workflows sampling MD
- [GitHub](https://github.com/MMunibas/Asparagus) (π¨βπ» 7 Β· π 3 Β· β±οΈ 13.12.2024):
tinker-hp (π₯9 Β· β 82 Β· π) - Tinker-HP: High-Performance Massively Parallel Evolution of Tinker on CPUs & GPUs. Custom
- [GitHub](https://github.com/TinkerTools/tinker-hp) (π¨βπ» 12 Β· π 22 Β· π 22 - 22% open Β· β±οΈ 26.10.2024):
ACE1.jl (π₯9 Β· β 21) - Atomic Cluster Expansion for Modelling Invariant Atomic Properties. Custom Julia
- [GitHub](https://github.com/ACEsuit/ACE1.jl) (π¨βπ» 9 Β· π 7 Β· π 46 - 47% open Β· β±οΈ 11.09.2024):
Point Edge Transformer (PET) (π₯9 Β· β 19) - Point Edge Transformer. MIT rep-learn transformer
- [GitHub](https://github.com/spozdn/pet) (π¨βπ» 7 Β· π 5 Β· β±οΈ 02.07.2024):
ACEfit (π₯9 Β· β 7) - MIT Julia
- [GitHub](https://github.com/ACEsuit/ACEfit.jl) (π¨βπ» 8 Β· π 7 Β· π 57 - 38% open Β· β±οΈ 14.09.2024):
GAP (π₯8 Β· β 40) - Gaussian Approximation Potential (GAP). Custom
- [GitHub](https://github.com/libAtoms/GAP) (π¨βπ» 13 Β· π 20 Β· β±οΈ 17.08.2024):
ALF (π₯8 Β· β 31) - A framework for performing active learning for training machine-learned interatomic potentials. Custom active-learning
- [GitHub](https://github.com/lanl/ALF) (π¨βπ» 5 Β· π 12 Β· β±οΈ 04.11.2024):
TurboGAP (π₯8 Β· β 16) - The TurboGAP code. Custom Fortran
- [GitHub](https://github.com/mcaroba/turbogap) (π¨βπ» 8 Β· π 10 Β· π 11 - 72% open Β· β±οΈ 17.12.2024):
MLXDM (π₯6 Β· β 7) - A Neural Network Potential with Rigorous Treatment of Long-Range Dispersion https://doi.org/10.1039/D2DD00150K. MIT long-range
- [GitHub](https://github.com/RowleyGroup/MLXDM) (π¨βπ» 7 Β· π 2 Β· β±οΈ 18.12.2024):
TensorPotential (π₯5 Β· β 10) - Tensorpotential is a TensorFlow based tool for development, fitting ML interatomic potentials from electronic.. Custom
- [GitHub](https://github.com/ICAMS/TensorPotential) (π¨βπ» 4 Β· π 4 Β· β±οΈ 12.09.2024):
Show 35 hidden projects...
- TorchANI (π₯24 Β· β 480 Β· π) - Accurate Neural Network Potential on PyTorch.MIT
- MEGNet (π₯23 Β· β 510 Β· π) - Graph Networks as a Universal Machine Learning Framework for Molecules and Crystals. BSD-3 multifidelity
- sGDML (π₯16 Β· β 140 Β· π) - sGDML - Reference implementation of the Symmetric Gradient Domain Machine Learning model. MIT
- TensorMol (π₯12 Β· β 270 Β· π) - Tensorflow + Molecules = TensorMol. GPL-3.0 single-paper
- SIMPLE-NN (π₯11 Β· β 47 Β· π) - SIMPLE-NN(SNU Interatomic Machine-learning PotentiaL packagE version Neural Network). GPL-3.0
- NNsforMD (π₯10 Β· β 10 Β· π) - Neural network class for molecular dynamics to predict potential energy, forces and non-adiabatic couplings. MIT
- DimeNet (π₯9 Β· β 300 Β· π) - DimeNet and DimeNet++ models, as proposed in Directional Message Passing for Molecular Graphs (ICLR 2020) and Fast and.. Custom
- SchNet (π₯9 Β· β 230 Β· π) - SchNet - a deep learning architecture for quantum chemistry. MIT
- GemNet (π₯9 Β· β 190 Β· π) - GemNet model in PyTorch, as proposed in GemNet: Universal Directional Graph Neural Networks for Molecules (NeurIPS.. Custom
- AIMNet (π₯8 Β· β 100 Β· π) - Atoms In Molecules Neural Network Potential. MIT single-paper
- MACE-Jax (π₯8 Β· β 64 Β· π) - Equivariant machine learning interatomic potentials in JAX. MIT
- SIMPLE-NN v2 (π₯8 Β· β 41 Β· π) - SIMPLE-NN is an open package that constructs Behler-Parrinello-type neural-network interatomic potentials from ab.. GPL-3.0
- SNAP (π₯8 Β· β 37 Β· π) - Repository for spectral neighbor analysis potential (SNAP) model development. BSD-3
- Atomistic Adversarial Attacks (π₯8 Β· β 34 Β· π) - Code for performing adversarial attacks on atomistic systems using NN potentials. MIT probabilistic
- MEGNetSparse (π₯8 Β· β 2) - A library imlementing a graph neural network with sparse representation from Code for Kazeev, N., Al-Maeeni, A.R.,.. MIT material-defect
- PhysNet (π₯7 Β· β 94 Β· π) - Code for training PhysNet models. MIT electrostatics
- MLIP-3 (π₯6 Β· β 26 Β· π) - MLIP-3: Active learning on atomic environments with Moment Tensor Potentials (MTP). BSD-2 C++
- testing-framework (π₯6 Β· β 11 Β· π) - The purpose of this repository is to aid the testing of a large number of interatomic potentials for a variety of.. Unlicensed benchmarking
- PANNA (π₯6 Β· β 10 Β· π) - A package to train and validate all-to-all connected network models for BP[1] and modified-BP[2] type local atomic.. MIT benchmarking
- GN-MM (π₯5 Β· β 10 Β· π) - The Gaussian Moment Neural Network (GM-NN) package developed for large-scale atomistic simulations employing atomistic.. MIT active-learning MD rep-eng magnetism
- Alchemical learning (π₯5 Β· β 2 Β· π) - Code for the Modeling high-entropy transition metal alloys with alchemical compression article. BSD-3
- ACE1Pack.jl (π₯5 Β· β 1 Β· π) - Provides convenience functionality for the usage of ACE1.jl, ACEfit.jl, JuLIP.jl for fitting interatomic potentials.. MIT Julia
- NequIP-JAX (π₯4 Β· β 20 Β· π) - JAX implementation of the NequIP interatomic potential. Unlicensed
- Allegro-Legato (π₯4 Β· β 19 Β· π) - An extension of Allegro with enhanced robustness and time-to-failure. MIT MD
- glp (π₯4 Β· β 18 Β· π€) - tools for graph-based machine-learning potentials in jax. MIT
- ACE Workflows (π₯4 Β· π) - Workflow Examples for ACE Models. Unlicensed Julia workflows
- PeriodicPotentials (π₯4 Β· π) - A Periodic table app that displays potentials based on the selected elements. MIT community-resource viz JavaScript
- PyFLAME (π₯3 Β· π) - An automated approach for developing neural network interatomic potentials with FLAME.. Unlicensed active-learning structure-prediction structure-optimization rep-eng Fortran
- SingleNN (π₯2 Β· β 9 Β· π) - An efficient package for training and executing neural-network interatomic potentials. Unlicensed C++
- AisNet (π₯2 Β· β 3 Β· π) - A Universal Interatomic Potential Neural Network with Encoded Local Environment Features.. MIT
- RuNNer (π₯2) - The RuNNer Neural Network Energy Representation is a Fortran-based framework for the construction of Behler-.. GPL-3.0 Fortran
- Allegro-JAX (π₯1 Β· β 21 Β· π€) - JAX implementation of the Allegro interatomic potential. Unlicensed
- nnp-pre-training (π₯1 Β· β 6 Β· π) - Synthetic pre-training for neural-network interatomic potentials. Unlicensed pretrained MD
- mag-ace (π₯1 Β· β 2 Β· π) - Magnetic ACE potential. FORTRAN interface for LAMMPS SPIN package. Unlicensed magnetism MD Fortran
- mlp (π₯1 Β· β 1 Β· π) - Proper orthogonal descriptors for efficient and accurate interatomic potentials... Unlicensed Julia
Language Models
Projects that use (large) language models (LMs, LLMs) or natural language procesing (NLP) techniques for atomistic ML.
paper-qa (π₯30 Β· β 6.7K) - High accuracy RAG for answering questions from scientific documents with citations. Apache-2 ai-agent
- [GitHub](https://github.com/Future-House/paper-qa) (π¨βπ» 31 Β· π 640 Β· π¦ 89 Β· π 290 - 43% open Β· β±οΈ 30.12.2024):
- [PyPi](https://pypi.org/project/paper-qa) (π₯ 17K / month Β· π¦ 10 Β· β±οΈ 11.12.2024):
ChemCrow (π₯18 Β· β 660 Β· π) - Open source package for the accurate solution of reasoning-intensive chemical tasks. MIT ai-agent
- [GitHub](https://github.com/ur-whitelab/chemcrow-public) (π¨βπ» 3 Β· π 98 Β· π¦ 8 Β· π 22 - 36% open Β· β±οΈ 19.12.2024):
- [PyPi](https://pypi.org/project/chemcrow) (π₯ 1.2K / month Β· β±οΈ 27.03.2024):
OpenBioML ChemNLP (π₯18 Β· β 150) - ChemNLP project. MIT datasets
- [GitHub](https://github.com/OpenBioML/chemnlp) (π¨βπ» 27 Β· π 45 Β· π 250 - 44% open Β· β±οΈ 19.08.2024):
- [PyPi](https://pypi.org/project/chemnlp) (π₯ 270 / month Β· π¦ 1 Β· β±οΈ 07.08.2023):
NIST ChemNLP (π₯12 Β· β 73) - ChemNLP: A Natural Language Processing based Library for Materials Chemistry Text Data. MIT literature-data
- [GitHub](https://github.com/usnistgov/chemnlp) (π¨βπ» 2 Β· π 17 Β· π¦ 4 Β· β±οΈ 19.08.2024):
- [PyPi](https://pypi.org/project/chemnlp) (π₯ 270 / month Β· π¦ 1 Β· β±οΈ 07.08.2023):
ChatMOF (π₯11 Β· β 67) - Predict and Inverse design for metal-organic framework with large-language models (llms). MIT generative
- [GitHub](https://github.com/Yeonghun1675/ChatMOF) (π¨βπ» 1 Β· π 12 Β· π¦ 3 Β· β±οΈ 01.07.2024):
- [PyPi](https://pypi.org/project/chatmof) (π₯ 840 / month Β· β±οΈ 01.07.2024):
AtomGPT (π₯11 Β· β 36) - AtomGPT: Atomistic Generative Pretrained Transformer for Forward and Inverse Materials Design.. Custom generative pretrained transformer
- [GitHub](https://github.com/usnistgov/atomgpt) (π¨βπ» 2 Β· π 6 Β· π¦ 2 Β· β±οΈ 12.12.2024):
- [PyPi](https://pypi.org/project/atomgpt) (π₯ 180 / month Β· β±οΈ 22.09.2024):
LLaMP (π₯7 Β· β 71) - A web app and Python API for multi-modal RAG framework to ground LLMs on high-fidelity materials informatics. An.. BSD-3 materials-discovery cheminformatics generative MD multimodal language-models Python general-tool
- [GitHub](https://github.com/chiang-yuan/llamp) (π¨βπ» 6 Β· π 12 Β· π 25 - 32% open Β· β±οΈ 14.10.2024):
LLM-Prop (π₯7 Β· β 30 Β· π€) - A repository for the LLM-Prop implementation. MIT
- [GitHub](https://github.com/vertaix/LLM-Prop) (π¨βπ» 6 Β· π 6 Β· π 2 - 50% open Β· β±οΈ 26.04.2024):
crystal-text-llm (π₯5 Β· β 90 Β· π€) - Large language models to generate stable crystals. CC-BY-NC-4.0 materials-discovery
- [GitHub](https://github.com/facebookresearch/crystal-text-llm) (π¨βπ» 3 Β· π 17 Β· π 11 - 81% open Β· β±οΈ 18.06.2024):
SciBot (π₯5 Β· β 30) - SciBot is a simple demo of building a domain-specific chatbot for science. Unlicensed ai-agent
- [GitHub](https://github.com/CFN-softbio/SciBot) (π¨βπ» 1 Β· π 9 Β· π¦ 2 Β· β±οΈ 03.09.2024):
MAPI_LLM (π₯5 Β· β 9 Β· π€) - A LLM application developed during the LLM March MADNESS Hackathon https://doi.org/10.1039/D3DD00113J. MIT ai-agent dataset
- [GitHub](https://github.com/maykcaldas/MAPI_LLM) (π¨βπ» 2 Β· π 2 Β· β±οΈ 11.04.2024):
Cephalo (π₯5 Β· β 9) - Multimodal Vision-Language Models for Bio-Inspired Materials Analysis and Design. Apache-2 generative multimodal pretrained
- [GitHub](https://github.com/lamm-mit/Cephalo) (π 1 Β· β±οΈ 23.07.2024):
Show 10 hidden projects...
- ChemDataExtractor (π₯16 Β· β 310 Β· π) - Automatically extract chemical information from scientific documents.MIT literature-data
- gptchem (π₯13 Β· β 240 Β· π) - Use GPT-3 to solve chemistry problems. MIT
- mat2vec (π₯12 Β· β 620 Β· π) - Supplementary Materials for Tshitoyan et al. Unsupervised word embeddings capture latent knowledge from materials.. MIT rep-learn
- nlcc (π₯12 Β· β 44 Β· π) - Natural language computational chemistry command line interface. MIT single-paper
- MoLFormer (π₯9 Β· β 280 Β· π) - Repository for MolFormer. Apache-2 transformer pretrained drug-discovery
- MolSkill (π₯9 Β· β 100 Β· π) - Extracting medicinal chemistry intuition via preference machine learning. MIT drug-discovery recommender
- chemlift (π₯7 Β· β 32 Β· π) - Language-interfaced fine-tuning for chemistry. MIT
- BERT-PSIE-TC (π₯5 Β· β 12 Β· π) - A dataset of Curie temperatures automatically extracted from scientific literature with the use of the BERT-PSIE.. MIT magnetism
- CatBERTa (π₯4 Β· β 22 Β· π€) - Large Language Model for Catalyst Property Prediction. Unlicensed transformer catalysis
- ChemDataWriter (π₯4 Β· β 14 Β· π) - ChemDataWriter is a transformer-based library for automatically generating research books in the chemistry area. MIT literature-data
Materials Discovery
Projects that implement materials discovery methods using atomistic ML.
π MatterGen - A generative model for inorganic materials design https://doi.org/10.48550/arXiv.2312.03687. generative proprietary
BOSS (π₯14 Β· β 21) - Bayesian Optimization Structure Search (BOSS). Apache-2 probabilistic
- [PyPi](https://pypi.org/project/aalto-boss) (π₯ 1.7K / month Β· β±οΈ 13.11.2024):
- [GitLab](https://gitlab.com/cest-group/boss) (π 11 Β· π 31 - 6% open Β· β±οΈ 13.11.2024):
aviary (π₯13 Β· β 48) - The Wren sits on its Roost in the Aviary. MIT
- [GitHub](https://github.com/CompRhys/aviary) (π¨βπ» 5 Β· π 12 Β· π 31 - 12% open Β· β±οΈ 15.12.2024):
Materials Discovery: GNoME (π₯10 Β· β 920) - Graph Networks for Materials Science (GNoME) and dataset of 381,000 novel stable materials. Apache-2 UIP datasets rep-learn proprietary
- [GitHub](https://github.com/google-deepmind/materials_discovery) (π¨βπ» 2 Β· π 150 Β· π 25 - 84% open Β· β±οΈ 09.12.2024):
AGOX (π₯9 Β· β 14) - AGOX is a package for global optimization of atomic system using e.g. the energy calculated from density functional.. GPL-3.0 structure-optimization
- [PyPi](https://pypi.org/project/agox) (π₯ 240 / month Β· β±οΈ 23.10.2024):
- [GitLab](https://gitlab.com/agox/agox) (π 5 Β· π 26 - 38% open Β· β±οΈ 23.10.2024):
CSPML (crystal structure prediction with machine learning-based element substitution) (π₯6 Β· β 22) - Original implementation of CSPML. MIT structure-prediction
- [GitHub](https://github.com/Minoru938/CSPML) (π¨βπ» 1 Β· π 8 Β· π 3 - 66% open Β· β±οΈ 22.12.2024):
Show 6 hidden projects...
- Computational Autonomy for Materials Discovery (CAMD) (π₯6 Β· β 1 Β· π) - Agent-based sequential learning software for materials discovery.Apache-2
- MAGUS (π₯4 Β· β 63 Β· π) - Machine learning And Graph theory assisted Universal structure Searcher. Unlicensed structure-prediction active-learning
- ML-atomate (π₯4 Β· β 5 Β· π) - Machine learning-assisted Atomate code for autonomous computational materials screening. GPL-3.0 active-learning workflows
- closed-loop-acceleration-benchmarks (π₯4 Β· π) - Data and scripts in support of the publication By how much can closed-loop frameworks accelerate computational.. MIT materials-discovery active-learning single-paper
- SPINNER (π₯3 Β· β 12 Β· π) - SPINNER (Structure Prediction of Inorganic crystals using Neural Network potentials with Evolutionary and Random.. GPL-3.0 C++ structure-prediction
- sl_discovery (π₯3 Β· β 5 Β· π) - Data processing and models related to Quantifying the performance of machine learning models in materials discovery. Apache-2 materials-discovery single-paper
Mathematical tools
Projects that implement mathematical objects used in atomistic machine learning.
KFAC-JAX (π₯19 Β· β 260) - Second Order Optimization and Curvature Estimation with K-FAC in JAX. Apache-2
- [GitHub](https://github.com/google-deepmind/kfac-jax) (π¨βπ» 17 Β· π 23 Β· π¦ 11 Β· π 20 - 45% open Β· β±οΈ 19.12.2024):
- [PyPi](https://pypi.org/project/kfac-jax) (π₯ 660 / month Β· π¦ 1 Β· β±οΈ 04.04.2024):
gpax (π₯17 Β· β 220 Β· π€) - Gaussian Processes for Experimental Sciences. MIT probabilistic active-learning
- [GitHub](https://github.com/ziatdinovmax/gpax) (π¨βπ» 6 Β· π 26 Β· π¦ 3 Β· π 40 - 20% open Β· β±οΈ 21.05.2024):
- [PyPi](https://pypi.org/project/gpax) (π₯ 520 / month Β· β±οΈ 20.03.2024):
SpheriCart (π₯16 Β· β 75) - Multi-language library for the calculation of spherical harmonics in Cartesian coordinates. MIT
- [GitHub](https://github.com/lab-cosmo/sphericart) (π¨βπ» 11 Β· π 12 Β· π₯ 100 Β· π¦ 5 Β· π 41 - 56% open Β· β±οΈ 07.11.2024):
- [PyPi](https://pypi.org/project/sphericart) (π₯ 700 / month Β· β±οΈ 04.09.2024):
Polynomials4ML.jl (π₯11 Β· β 12 Β· π€) - Polynomials for ML: fast evaluation, batching, differentiation. MIT Julia
- [GitHub](https://github.com/ACEsuit/Polynomials4ML.jl) (π¨βπ» 10 Β· π 5 Β· π 51 - 33% open Β· β±οΈ 22.06.2024):
GElib (π₯9 Β· β 21) - C++/CUDA library for SO(3) equivariant operations. MPL-2.0 C++
- [GitHub](https://github.com/risi-kondor/GElib) (π¨βπ» 4 Β· π 3 Β· π 8 - 50% open Β· β±οΈ 27.07.2024):
COSMO Toolbox (π₯6 Β· β 7 Β· π€) - Assorted libraries and utilities for atomistic simulation analysis. Unlicensed C++
- [GitHub](https://github.com/lab-cosmo/toolbox) (π¨βπ» 9 Β· π 7 Β· β±οΈ 19.03.2024):
Show 5 hidden projects...
- lie-nn (π₯9 Β· β 27 Β· π) - Tools for building equivariant polynomials on reductive Lie groups.MIT rep-learn
- EquivariantOperators.jl (π₯6 Β· β 19 Β· π) - This package is deprecated. Functionalities are migrating to Porcupine.jl. MIT Julia
- cnine (π₯5 Β· β 4) - Cnine tensor library. Unlicensed C++
- torch_spex (π₯3 Β· β 3 Β· π€) - Spherical expansions in PyTorch. Unlicensed
- Wigner Kernels (π₯1 Β· β 2 Β· π) - Collection of programs to benchmark Wigner kernels. Unlicensed benchmarking
Molecular Dynamics
Projects that simplify the integration of molecular dynamics and atomistic machine learning.
JAX-MD (π₯25 Β· β 1.2K) - Differentiable, Hardware Accelerated, Molecular Dynamics. Apache-2
- [GitHub](https://github.com/jax-md/jax-md) (π¨βπ» 36 Β· π 200 Β· π¦ 64 Β· π 160 - 49% open Β· β±οΈ 26.11.2024):
- [PyPi](https://pypi.org/project/jax-md) (π₯ 3.7K / month Β· π¦ 3 Β· β±οΈ 09.08.2023):
mlcolvar (π₯19 Β· β 95) - A unified framework for machine learning collective variables for enhanced sampling simulations. MIT sampling
- [GitHub](https://github.com/luigibonati/mlcolvar) (π¨βπ» 8 Β· π 26 Β· π¦ 3 Β· π 74 - 17% open Β· β±οΈ 25.11.2024):
- [PyPi](https://pypi.org/project/mlcolvar) (π₯ 200 / month Β· β±οΈ 12.06.2024):
FitSNAP (π₯18 Β· β 160) - Software for generating machine-learning interatomic potentials for LAMMPS. GPL-2.0
- [GitHub](https://github.com/FitSNAP/FitSNAP) (π¨βπ» 24 Β· π 54 Β· π₯ 13 Β· π 73 - 21% open Β· β±οΈ 02.12.2024):
- [Conda](https://anaconda.org/conda-forge/fitsnap3) (π₯ 9.9K Β· β±οΈ 16.06.2023):
openmm-torch (π₯17 Β· β 190) - OpenMM plugin to define forces with neural networks. Custom ML-IAP C++
- [GitHub](https://github.com/openmm/openmm-torch) (π¨βπ» 8 Β· π 24 Β· π 96 - 29% open Β· β±οΈ 11.11.2024):
- [Conda](https://anaconda.org/conda-forge/openmm-torch) (π₯ 590K Β· β±οΈ 12.11.2024):
OpenMM-ML (π₯12 Β· β 85) - High level API for using machine learning models in OpenMM simulations. MIT ML-IAP
- [GitHub](https://github.com/openmm/openmm-ml) (π¨βπ» 5 Β· π 20 Β· π 55 - 36% open Β· β±οΈ 06.08.2024):
- [Conda](https://anaconda.org/conda-forge/openmm-ml) (π₯ 6.4K Β· β±οΈ 07.06.2024):
pair_nequip (π₯10 Β· β 41 Β· π€) - LAMMPS pair style for NequIP. MIT ML-IAP rep-learn
- [GitHub](https://github.com/mir-group/pair_nequip) (π¨βπ» 3 Β· π 13 Β· π 31 - 35% open Β· β±οΈ 05.06.2024):
PACE (π₯10 Β· β 28) - The LAMMPS ML-IAP `pair_style pace`, aka Atomic Cluster Expansion (ACE), aka ML-PACE,.. Custom
- [GitHub](https://github.com/ICAMS/lammps-user-pace) (π¨βπ» 8 Β· π 12 Β· π 8 - 25% open Β· β±οΈ 17.12.2024):
pair_allegro (π₯7 Β· β 39 Β· π€) - LAMMPS pair style for Allegro deep learning interatomic potentials with parallelization support. MIT ML-IAP rep-learn
- [GitHub](https://github.com/mir-group/pair_allegro) (π¨βπ» 2 Β· π 8 Β· π 33 - 45% open Β· β±οΈ 05.06.2024):
SOMD (π₯6 Β· β 14) - Molecular dynamics package designed for the SIESTA DFT code. AGPL-3.0 ML-IAP active-learning
- [GitHub](https://github.com/initqp/somd) (π 2 Β· β±οΈ 04.11.2024):
Show 1 hidden projects...
- interface-lammps-mlip-3 (π₯3 Β· β 5 Β· π) - An interface between LAMMPS and MLIP (version 3).GPL-2.0
Reinforcement Learning
Projects that focus on reinforcement learning for atomistic ML.
Show 2 hidden projects...
- ReLeaSE (π₯11 Β· β 350 Β· π) - Deep Reinforcement Learning for de-novo Drug Design.MIT drug-discovery
- CatGym (π₯6 Β· β 11 Β· π) - Surface segregation using Deep Reinforcement Learning. GPL
Representation Engineering
Projects that offer implementations of representations aka descriptors, fingerprints of atomistic systems, and models built with them, aka feature engineering.
cdk (π₯26 Β· β 500) - The Chemistry Development Kit. LGPL-2.1 cheminformatics Java
- [GitHub](https://github.com/cdk/cdk) (π¨βπ» 170 Β· π 160 Β· π₯ 24K Β· π 300 - 10% open Β· β±οΈ 17.12.2024):
- [Maven](https://search.maven.org/artifact/org.openscience.cdk/cdk-bundle) (π¦ 16 Β· β±οΈ 21.08.2023):
DScribe (π₯25 Β· β 410 Β· π€) - DScribe is a python package for creating machine learning descriptors for atomistic systems. Apache-2
- [GitHub](https://github.com/SINGROUP/dscribe) (π¨βπ» 18 Β· π 88 Β· π¦ 220 Β· π 100 - 11% open Β· β±οΈ 28.05.2024):
- [PyPi](https://pypi.org/project/dscribe) (π₯ 63K / month Β· π¦ 35 Β· β±οΈ 28.05.2024):
- [Conda](https://anaconda.org/conda-forge/dscribe) (π₯ 160K Β· β±οΈ 28.05.2024):
MODNet (π₯16 Β· β 82) - MODNet: a framework for machine learning materials properties. MIT pretrained small-data transfer-learning
- [GitHub](https://github.com/ppdebreuck/modnet) (π¨βπ» 11 Β· π 33 Β· π¦ 10 Β· π 56 - 46% open Β· β±οΈ 28.11.2024):
Rascaline (π₯16 Β· β 49 Β· π) - Computing representations for atomistic machine learning. BSD-3 Rust C++
- [GitHub](https://github.com/metatensor/featomic) (π¨βπ» 14 Β· π 14 Β· π₯ 22 Β· π 71 - 46% open Β· β±οΈ 20.12.2024):
GlassPy (π₯14 Β· β 29) - Python module for scientists working with glass materials. GPL-3.0
- [GitHub](https://github.com/drcassar/glasspy) (π¨βπ» 2 Β· π 7 Β· π¦ 7 Β· π 15 - 46% open Β· β±οΈ 13.10.2024):
- [PyPi](https://pypi.org/project/glasspy) (π₯ 720 / month Β· β±οΈ 05.09.2024):
SISSO (π₯12 Β· β 260) - A data-driven method combining symbolic regression and compressed sensing for accurate & interpretable models. Apache-2 Fortran
- [GitHub](https://github.com/rouyang2017/SISSO) (π¨βπ» 3 Β· π 85 Β· π 77 - 23% open Β· β±οΈ 20.09.2024):
fplib (π₯8 Β· β 7 Β· π) - libfp is a library for calculating crystalline fingerprints and measuring similarities of materials. MIT C-lang single-paper
- [GitHub](https://github.com/Rutgers-ZRG/libfp) (π 1 Β· π¦ 1 Β· β±οΈ 15.10.2024):
NICE (π₯7 Β· β 12 Β· π€) - NICE (N-body Iteratively Contracted Equivariants) is a set of tools designed for the calculation of invariant and.. MIT
- [GitHub](https://github.com/lab-cosmo/nice) (π¨βπ» 4 Β· π 3 Β· π 3 - 66% open Β· β±οΈ 15.04.2024):
milad (π₯6 Β· β 31) - Moment Invariants Local Atomic Descriptor. GPL-3.0 generative
- [GitHub](https://github.com/muhrin/milad) (π¨βπ» 1 Β· π 2 Β· π¦ 3 Β· β±οΈ 20.08.2024):
SA-GPR (π₯6 Β· β 19) - Public repository for symmetry-adapted Gaussian Process Regression (SA-GPR). LGPL-3.0 C-lang
- [GitHub](https://github.com/dilkins/TENSOAP) (π¨βπ» 5 Β· π 14 Β· π 7 - 28% open Β· β±οΈ 23.07.2024):
Show 15 hidden projects...
- CatLearn (π₯16 Β· β 100 Β· π) -GPL-3.0 surface-science
- Librascal (π₯13 Β· β 80 Β· π) - A scalable and versatile library to generate representations for atomic-scale learning. LGPL-2.1
- BenchML (π₯12 Β· β 15 Β· π) - ML benchmarking and pipeling framework. Apache-2 benchmarking
- cmlkit (π₯11 Β· β 34 Β· π) - tools for machine learning in condensed matter physics and quantum chemistry. MIT benchmarking
- CBFV (π₯11 Β· β 27 Β· π) - Tool to quickly create a composition-based feature vector. Unlicensed
- SkipAtom (π₯10 Β· β 24 Β· π) - Distributed representations of atoms, inspired by the Skip-gram model. MIT
- SOAPxx (π₯6 Β· β 7 Β· π) - A SOAP implementation. GPL-2.0 C++
- pyLODE (π₯6 Β· β 3 Β· π) - Pythonic implementation of LOng Distance Equivariants. Apache-2 electrostatics
- AMP (π₯6 Β· π) - Amp is an open-source package designed to easily bring machine-learning to atomistic calculations. Unlicensed
- MXenes4HER (π₯5 Β· β 6 Β· π) - Predicting hydrogen evolution (HER) activity over 4500 MXene materials https://doi.org/10.1039/D3TA00344B. GPL-3.0 materials-discovery catalysis scikit-learn single-paper
- soap_turbo (π₯5 Β· β 5 Β· π) - soap_turbo comprises a series of libraries to be used in combination with QUIP/GAP and TurboGAP. Custom Fortran
- SISSO++ (π₯5 Β· β 3 Β· π) - C++ Implementation of SISSO with python bindings. Apache-2 C++
- automl-materials (π₯4 Β· β 5 Β· π) - AutoML for Regression Tasks on Small Tabular Data in Materials Design. MIT autoML benchmarking single-paper
- magnetism-prediction (π₯4 Β· β 1 Β· π) - DFT-aided Machine Learning Search for Magnetism in Fe-based Bimetallic Chalcogenides. Apache-2 magnetism single-paper
- ML-for-CurieTemp-Predictions (π₯3 Β· β 1 Β· π) - Machine Learning Predictions of High-Curie-Temperature Materials. MIT single-paper magnetism
Representation Learning
General models that learn a representations aka embeddings of atomistic systems, such as message-passing neural networks (MPNN).
PyG Models (π₯35 Β· β 22K) - Representation learning models implemented in PyTorch Geometric. MIT general-ml
- [GitHub](https://github.com/pyg-team/pytorch_geometric) (π¨βπ» 530 Β· π 3.7K Β· π¦ 7.4K Β· π 3.8K - 29% open Β· β±οΈ 30.12.2024):
Deep Graph Library (DGL) (π₯35 Β· β 14K) - Python package built to ease deep learning on graph, on top of existing DL frameworks. Apache-2
- [GitHub](https://github.com/dmlc/dgl) (π¨βπ» 300 Β· π 3K Β· π¦ 330 Β· π 2.9K - 18% open Β· β±οΈ 18.10.2024):
- [PyPi](https://pypi.org/project/dgl) (π₯ 95K / month Β· π¦ 150 Β· β±οΈ 13.05.2024):
- [Conda](https://anaconda.org/dglteam/dgl) (π₯ 400K Β· β±οΈ 03.09.2024):
e3nn (π₯28 Β· β 1K) - A modular framework for neural networks with Euclidean symmetry. MIT
- [GitHub](https://github.com/e3nn/e3nn) (π¨βπ» 34 Β· π 140 Β· π¦ 370 Β· π 160 - 14% open Β· β±οΈ 23.12.2024):
- [PyPi](https://pypi.org/project/e3nn) (π₯ 170K / month Β· π¦ 34 Β· β±οΈ 06.11.2024):
- [Conda](https://anaconda.org/conda-forge/e3nn) (π₯ 28K Β· β±οΈ 21.12.2024):
SchNetPack (π₯26 Β· β 800) - SchNetPack - Deep Neural Networks for Atomistic Systems. MIT
- [GitHub](https://github.com/atomistic-machine-learning/schnetpack) (π¨βπ» 36 Β· π 210 Β· π¦ 96 Β· π 260 - 2% open Β· β±οΈ 26.11.2024):
- [PyPi](https://pypi.org/project/schnetpack) (π₯ 830 / month Β· π¦ 4 Β· β±οΈ 05.09.2024):
MatGL (Materials Graph Library) (π₯24 Β· β 300) - Graph deep learning library for materials. BSD-3 multifidelity
- [GitHub](https://github.com/materialsvirtuallab/matgl) (π¨βπ» 17 Β· π 68 Β· π¦ 59 Β· π 110 - 6% open Β· β±οΈ 31.12.2024):
- [PyPi](https://pypi.org/project/m3gnet) (π₯ 880 / month Β· π¦ 5 Β· β±οΈ 17.11.2022):
e3nn-jax (π₯22 Β· β 190) - jax library for E3 Equivariant Neural Networks. Apache-2
- [GitHub](https://github.com/e3nn/e3nn-jax) (π¨βπ» 7 Β· π 18 Β· π¦ 46 Β· π 23 - 4% open Β· β±οΈ 15.12.2024):
- [PyPi](https://pypi.org/project/e3nn-jax) (π₯ 2.9K / month Β· π¦ 13 Β· β±οΈ 14.08.2024):
ALIGNN (π₯21 Β· β 240) - Atomistic Line Graph Neural Network https://scholar.google.com/citations?user=9Q-tNnwAAAAJ&hl=en.. Custom
- [GitHub](https://github.com/usnistgov/alignn) (π¨βπ» 7 Β· π 86 Β· π¦ 17 Β· π 70 - 61% open Β· β±οΈ 02.12.2024):
- [PyPi](https://pypi.org/project/alignn) (π₯ 6.1K / month Β· π¦ 8 Β· β±οΈ 02.12.2024):
NVIDIA Deep Learning Examples for Tensor Cores (π₯20 Β· β 14K Β· π€) - State-of-the-Art Deep Learning scripts organized by models - easy to train and deploy with reproducible accuracy and.. Custom educational drug-discovery
- [GitHub](https://github.com/NVIDIA/DeepLearningExamples) (π¨βπ» 120 Β· π 3.2K Β· π 910 - 37% open Β· β±οΈ 04.04.2024):
DIG: Dive into Graphs (π₯20 Β· β 1.9K Β· π€) - A library for graph deep learning research. GPL-3.0
- [GitHub](https://github.com/divelab/DIG) (π¨βπ» 50 Β· π 280 Β· π 210 - 16% open Β· β±οΈ 04.02.2024):
- [PyPi](https://pypi.org/project/dive-into-graphs) (π₯ 840 / month Β· β±οΈ 27.06.2022):
matsciml (π₯19 Β· β 160) - Open MatSci ML Toolkit is a framework for prototyping and scaling out deep learning models for materials discovery.. MIT workflows benchmarking
- [GitHub](https://github.com/IntelLabs/matsciml) (π¨βπ» 12 Β· π 23 Β· π 66 - 34% open Β· β±οΈ 20.12.2024):
Uni-Mol (π₯18 Β· β 760) - Official Repository for the Uni-Mol Series Methods. MIT pretrained
- [GitHub](https://github.com/deepmodeling/Uni-Mol) (π¨βπ» 19 Β· π 130 Β· π₯ 17K Β· π 180 - 44% open Β· β±οΈ 02.01.2025):
kgcnn (π₯18 Β· β 110 Β· π€) - Graph convolutions in Keras with TensorFlow, PyTorch or Jax. MIT
- [GitHub](https://github.com/aimat-lab/gcnn_keras) (π¨βπ» 7 Β· π 31 Β· π¦ 19 Β· π 87 - 14% open Β· β±οΈ 06.05.2024):
- [PyPi](https://pypi.org/project/kgcnn) (π₯ 630 / month Β· π¦ 3 Β· β±οΈ 27.02.2024):
escnn (π₯16 Β· β 380) - Equivariant Steerable CNNs Library for Pytorch https://quva-lab.github.io/escnn/. Custom
- [GitHub](https://github.com/QUVA-Lab/escnn) (π¨βπ» 10 Β· π 47 Β· π 75 - 50% open Β· β±οΈ 31.10.2024):
- [PyPi](https://pypi.org/project/escnn) (π₯ 1.1K / month Β· π¦ 6 Β· β±οΈ 01.04.2022):
Graphormer (π₯15 Β· β 2.2K Β· π€) - Graphormer is a general-purpose deep learning backbone for molecular modeling. MIT transformer pretrained
- [GitHub](https://github.com/microsoft/Graphormer) (π¨βπ» 14 Β· π 330 Β· π 160 - 57% open Β· β±οΈ 28.05.2024):
HydraGNN (π₯14 Β· β 68) - Distributed PyTorch implementation of multi-headed graph convolutional neural networks. BSD-3
- [GitHub](https://github.com/ORNL/HydraGNN) (π¨βπ» 15 Β· π 28 Β· π¦ 2 Β· π 49 - 34% open Β· β±οΈ 31.12.2024):
Compositionally-Restricted Attention-Based Network (CrabNet) (π₯13 Β· β 15) - Predict materials properties using only the composition information!. MIT
- [GitHub](https://github.com/sparks-baird/CrabNet) (π¨βπ» 6 Β· π 5 Β· π¦ 14 Β· π 19 - 84% open Β· β±οΈ 09.09.2024):
- [PyPi](https://pypi.org/project/crabnet) (π₯ 1.1K / month Β· π¦ 2 Β· β±οΈ 10.01.2023):
hippynn (π₯12 Β· β 72) - python library for atomistic machine learning. Custom workflows
- [GitHub](https://github.com/lanl/hippynn) (π¨βπ» 14 Β· π 23 Β· π¦ 2 Β· π 22 - 45% open Β· β±οΈ 31.10.2024):
Atom2Vec (π₯10 Β· β 36 Β· π€) - Atom2Vec: a simple way to describe atoms for machine learning. MIT
- [GitHub](https://github.com/idocx/Atom2Vec) (π¨βπ» 1 Β· π 9 Β· π¦ 3 Β· π 4 - 75% open Β· β±οΈ 23.02.2024):
- [PyPi](https://pypi.org/project/atom2vec) (π₯ 120 / month Β· β±οΈ 23.02.2024):
GATGNN: Global Attention Graph Neural Network (π₯9 Β· β 72) - Pytorch Repository for our work: Graph convolutional neural networks with global attention for improved materials.. MIT
- [GitHub](https://github.com/superlouis/GATGNN) (π¨βπ» 4 Β· π 16 Β· π 7 - 57% open Β· β±οΈ 17.12.2024):
EquiformerV2 (π₯8 Β· β 230) - [ICLR 2024] EquiformerV2: Improved Equivariant Transformer for Scaling to Higher-Degree Representations. MIT
- [GitHub](https://github.com/atomicarchitects/equiformer_v2) (π¨βπ» 2 Β· π 32 Β· π 19 - 68% open Β· β±οΈ 16.07.2024):
Equiformer (π₯8 Β· β 220) - [ICLR 2023 Spotlight] Equiformer: Equivariant Graph Attention Transformer for 3D Atomistic Graphs. MIT transformer
- [GitHub](https://github.com/atomicarchitects/equiformer) (π¨βπ» 2 Β· π 40 Β· π 18 - 50% open Β· β±οΈ 18.07.2024):
graphite (π₯8 Β· β 66) - A repository for implementing graph network models based on atomic structures. MIT
- [GitHub](https://github.com/LLNL/graphite) (π¨βπ» 2 Β· π 9 Β· π¦ 15 Β· π 4 - 75% open Β· β±οΈ 08.08.2024):
DeeperGATGNN (π₯8 Β· β 49 Β· π€) - Scalable graph neural networks for materials property prediction. MIT
- [GitHub](https://github.com/usccolumbia/deeperGATGNN) (π¨βπ» 3 Β· π 7 Β· π 12 - 33% open Β· β±οΈ 19.01.2024):
T-e3nn (π₯8 Β· β 12) - Time-reversal Euclidean neural networks based on e3nn. MIT magnetism
- [GitHub](https://github.com/Hongyu-yu/T-e3nn) (π¨βπ» 26 Β· π 1 Β· β±οΈ 29.09.2024):
Show 34 hidden projects...
- dgl-lifesci (π₯24 Β· β 740 Β· π) - Python package for graph neural networks in chemistry and biology.Apache-2
- benchmarking-gnns (π₯14 Β· β 2.5K Β· π) - Repository for benchmarking graph neural networks. MIT single-paper benchmarking
- Crystal Graph Convolutional Neural Networks (CGCNN) (π₯13 Β· β 670 Β· π) - Crystal graph convolutional neural networks for predicting material properties. MIT
- Neural fingerprint (nfp) (π₯12 Β· β 57 Β· π) - Keras layers for end-to-end learning with rdkit and pymatgen. Custom
- FAENet (π₯11 Β· β 33 Β· π) - Frame Averaging Equivariant GNN for materials modeling. MIT
- pretrained-gnns (π₯10 Β· β 980 Β· π) - Strategies for Pre-training Graph Neural Networks. MIT pretrained
- GDC (π₯10 Β· β 270 Β· π) - Graph Diffusion Convolution, as proposed in Diffusion Improves Graph Learning (NeurIPS 2019). MIT generative
- SE(3)-Transformers (π₯9 Β· β 500 Β· π) - code for the SE3 Transformers paper: https://arxiv.org/abs/2006.10503. MIT single-paper transformer
- ai4material_design (π₯9 Β· β 6 Β· π) - Code for Kazeev, N., Al-Maeeni, A.R., Romanov, I. et al. Sparse representation for machine learning the properties of.. Apache-2 pretrained material-defect
- molecularGNN_smiles (π₯8 Β· β 300 Β· π) - The code of a graph neural network (GNN) for molecules, which is based on learning representations of r-radius.. Apache-2
- CGAT (π₯8 Β· β 27 Β· π) - Crystal graph attention neural networks for materials prediction. MIT
- UVVisML (π₯8 Β· β 26 Β· π) - Predict optical properties of molecules with machine learning. MIT optical-properties single-paper probabilistic
- tensorfieldnetworks (π₯7 Β· β 150 Β· π) - Rotation- and translation-equivariant neural networks for 3D point clouds. MIT
- DTNN (π₯7 Β· β 78 Β· π) - Deep Tensor Neural Network. MIT
- Cormorant (π₯7 Β· β 60 Β· π) - Codebase for Cormorant Neural Networks. Custom
- AdsorbML (π₯7 Β· β 39 Β· π) - MIT surface-science single-paper
- escnn_jax (π₯7 Β· β 29 Β· π) - Equivariant Steerable CNNs Library for Pytorch https://quva-lab.github.io/escnn/. Custom
- ML4pXRDs (π₯7 Β· π) - Contains code to train neural networks based on simulated powder XRDs from synthetic crystals. MIT XRD single-paper
- MACE-Layer (π₯6 Β· β 33 Β· π) - Higher order equivariant graph neural networks for 3D point clouds. MIT
- charge_transfer_nnp (π₯6 Β· β 33 Β· π) - Graph neural network potential with charge transfer. MIT electrostatics
- GLAMOUR (π₯6 Β· β 21 Β· π) - Graph Learning over Macromolecule Representations. MIT single-paper
- Autobahn (π₯5 Β· β 29 Β· π) - Repository for Autobahn: Automorphism Based Graph Neural Networks. MIT
- FieldSchNet (π₯5 Β· β 19 Β· π) - Deep neural network for molecules in external fields. MIT
- SCFNN (π₯5 Β· β 14 Β· π) - Self-consistent determination of long-range electrostatics in neural network potentials. MIT C++ electrostatics single-paper
- CraTENet (π₯5 Β· β 14 Β· π) - An attention-based deep neural network for thermoelectric transport properties. MIT transport-phenomena
- EGraFFBench (π₯5 Β· β 10 Β· π) - Unlicensed single-paper benchmarking ML-IAP
- Per-Site CGCNN (π₯5 Β· β 1 Β· π) - Crystal graph convolutional neural networks for predicting material properties. MIT pretrained single-paper
- Per-site PAiNN (π₯5 Β· β 1 Β· π) - Fork of PaiNN for PerovskiteOrderingGCNNs. MIT probabilistic pretrained single-paper
- Graph Transport Network (π₯4 Β· β 16 Β· π) - Graph transport network (GTN), as proposed in Scalable Optimal Transport in High Dimensions for Graph Distances,.. Custom transport-phenomena
- gkx: Green-Kubo Method in JAX (π₯4 Β· β 5 Β· π€) - Green-Kubo + JAX + MLPs = Anharmonic Thermal Conductivities Done Fast. MIT transport-phenomena
- atom_by_atom (π₯3 Β· β 9 Β· π) - Atom-by-atom design of metal oxide catalysts for the oxygen evolution reaction with Machine Learning. Unlicensed surface-science single-paper
- Element encoder (π₯3 Β· β 6 Β· π) - Autoencoder neural network to compress properties of atomic species into a vector representation. GPL-3.0 single-paper
- Point Edge Transformer (π₯2) - Smooth, exact rotational symmetrization for deep learning on point clouds. CC-BY-4.0
- SphericalNet (π₯1 Β· β 3 Β· π) - Implementation of Clebsch-Gordan Networks (CGnet: https://arxiv.org/pdf/1806.09231.pdf) by GElib & cnine libraries in.. Unlicensed
Universal Potentials
Machine-learned interatomic potentials (ML-IAP) that have been trained on large, chemically and structural diverse datasets. For materials, this means e.g. datasets that include a majority of the periodic table.
π TeaNet - Universal neural network interatomic potential inspired by iterative electronic relaxations.. ML-IAP
π PreFerred Potential (PFP) - Universal neural network potential for material discovery https://doi.org/10.1038/s41467-022-30687-9. ML-IAP proprietary
π MatterSim - A Deep Learning Atomistic Model Across Elements, Temperatures and Pressures https://doi.org/10.48550/arXiv.2405.04967. ML-IAP active-learning proprietary
DPA-2 (π₯27 Β· β 1.5K) - Towards a universal large atomic model for molecular and material simulation https://doi.org/10.48550/arXiv.2312.15492. LGPL-3.0 ML-IAP pretrained workflows datasets
- [GitHub](https://github.com/deepmodeling/deepmd-kit) (π¨βπ» 73 Β· π 520 Β· π₯ 46K Β· π¦ 22 Β· π 870 - 10% open Β· β±οΈ 23.12.2024):
CHGNet (π₯22 Β· β 260) - Pretrained universal neural network potential for charge-informed atomistic modeling https://chgnet.lbl.gov. Custom ML-IAP MD pretrained electrostatics magnetism structure-relaxation
- [GitHub](https://github.com/CederGroupHub/chgnet) (π¨βπ» 10 Β· π 68 Β· π¦ 43 Β· π 62 - 4% open Β· β±οΈ 16.11.2024):
- [PyPi](https://pypi.org/project/chgnet) (π₯ 24K / month Β· π¦ 21 Β· β±οΈ 16.09.2024):
MACE-MP (π₯18 Β· β 560) - Pretrained foundation models for materials chemistry. MIT ML-IAP pretrained rep-learn MD
- [GitHub](https://github.com/ACEsuit/mace-mp) (π¨βπ» 2 Β· π 210 Β· π₯ 46K Β· π 10 - 10% open Β· β±οΈ 15.11.2024):
- [PyPi](https://pypi.org/project/mace-torch) (π₯ 8.9K / month Β· π¦ 23 Β· β±οΈ 07.12.2024):
M3GNet (π₯18 Β· β 260) - Materials graph network with 3-body interactions featuring a DFT surrogate crystal relaxer and a state-of-the-art.. BSD-3 ML-IAP pretrained
- [GitHub](https://github.com/materialsvirtuallab/m3gnet) (π¨βπ» 16 Β· π 66 Β· π¦ 30 Β· π 35 - 42% open Β· β±οΈ 04.10.2024):
- [PyPi](https://pypi.org/project/m3gnet) (π₯ 880 / month Β· π¦ 5 Β· β±οΈ 17.11.2022):
Orb Models (π₯18 Β· β 220 Β· π£) - ORB forcefield models from Orbital Materials. Custom ML-IAP pretrained
- [GitHub](https://github.com/orbital-materials/orb-models) (π¨βπ» 7 Β· π 23 Β· π¦ 6 Β· π 19 - 10% open Β· β±οΈ 19.12.2024):
- [PyPi](https://pypi.org/project/orb-models) (π₯ 1.9K / month Β· π¦ 4 Β· β±οΈ 20.12.2024):
SevenNet (π₯17 Β· β 140) - SevenNet (Scalable EquiVariance Enabled Neural Network) is a graph neural network interatomic potential package that.. GPL-3.0 ML-IAP MD pretrained
- [GitHub](https://github.com/MDIL-SNU/SevenNet) (π¨βπ» 14 Β· π 21 Β· π¦ 8 Β· π 33 - 30% open Β· β±οΈ 19.12.2024):
MLIP Arena Leaderboard (π₯13 Β· β 53) - Fair and transparent benchmark of machine-learned interatomic potentials (MLIPs), beyond basic error metrics. Apache-2 ML-IAP community-resource
- [GitHub](https://github.com/atomind-ai/mlip-arena) (π¨βπ» 3 Β· π 2 Β· π¦ 2 Β· π 11 - 63% open Β· β±οΈ 25.12.2024):
GRACE (π₯10 Β· β 27 Β· π£) - GRACE models and gracemaker (as implemented in TensorPotential package). Custom ML-IAP pretrained MD rep-learn rep-eng
- [GitHub](https://github.com/ICAMS/grace-tensorpotential) (π¨βπ» 3 Β· π 3 Β· π¦ 1 Β· π 2 - 50% open Β· β±οΈ 13.12.2024):
Joint Multidomain Pre-Training (JMP) (π₯5 Β· β 43) - Code for From Molecules to Materials Pre-training Large Generalizable Models for Atomic Property Prediction. CC-BY-NC-4.0 pretrained ML-IAP general-tool
- [GitHub](https://github.com/facebookresearch/JMP) (π¨βπ» 2 Β· π 6 Β· π 5 - 40% open Β· β±οΈ 22.10.2024):
Unsupervised Learning
Projects that focus on unsupervised learning (USL) for atomistic ML, such as dimensionality reduction, clustering and visualization.
DADApy (π₯19 Β· β 110) - Distance-based Analysis of DAta-manifolds in python. Apache-2
- [GitHub](https://github.com/sissa-data-science/DADApy) (π¨βπ» 20 Β· π 18 Β· π¦ 12 Β· π 37 - 27% open Β· β±οΈ 20.11.2024):
- [PyPi](https://pypi.org/project/dadapy) (π₯ 240 / month Β· β±οΈ 20.11.2024):
ASAP (π₯11 Β· β 140 Β· π€) - ASAP is a package that can quickly analyze and visualize datasets of crystal or molecular structures. MIT
- [GitHub](https://github.com/BingqingCheng/ASAP) (π¨βπ» 6 Β· π 28 Β· π¦ 7 Β· π 25 - 24% open Β· β±οΈ 27.06.2024):
Show 5 hidden projects...
- Sketchmap (π₯8 Β· β 46 Β· π) - Suite of programs to perform non-linear dimensionality reduction -- sketch-map in particular.GPL-3.0 C++
- Coarse-Graining-Auto-encoders (π₯5 Β· β 21 Β· π) - Implementation of coarse-graining Autoencoders. Unlicensed single-paper
- paper-ml-robustness-material-property (π₯5 Β· β 4 Β· π) - A critical examination of robustness and generalizability of machine learning prediction of materials properties. BSD-3 datasets single-paper
- KmdPlus (π₯4 Β· β 4) - This module contains a class for treating kernel mean descriptor (KMD), and a function for generating descriptors with.. MIT
- Descriptor Embedding and Clustering for Atomisitic-environment Framework (DECAF) ( β 2) - Provides a workflow to obtain clustering of local environments in dataset of structures. Unlicensed
Visualization
Projects that focus on visualization (viz.) for atomistic ML.
Crystal Toolkit (π₯24 Β· β 160) - Crystal Toolkit is a framework for building web apps for materials science and is currently powering the new Materials.. MIT
- [GitHub](https://github.com/materialsproject/crystaltoolkit) (π¨βπ» 31 Β· π 57 Β· π¦ 41 Β· π 110 - 47% open Β· β±οΈ 02.01.2025):
- [PyPi](https://pypi.org/project/crystal-toolkit) (π₯ 2.8K / month Β· π¦ 10 Β· β±οΈ 22.10.2024):
pymatviz (π₯22 Β· β 180) - A toolkit for visualizations in materials informatics. MIT general-tool probabilistic
- [GitHub](https://github.com/janosh/pymatviz) (π¨βπ» 9 Β· π 16 Β· π¦ 17 Β· π 54 - 22% open Β· β±οΈ 31.12.2024):
- [PyPi](https://pypi.org/project/pymatviz) (π₯ 6.7K / month Β· π¦ 6 Β· β±οΈ 20.12.2024):
ZnDraw (π₯21 Β· β 38) - A powerful tool for visualizing, modifying, and analysing atomistic systems. EPL-2.0 MD generative JavaScript
- [GitHub](https://github.com/zincware/ZnDraw) (π¨βπ» 7 Β· π 4 Β· π¦ 10 Β· π 360 - 27% open Β· β±οΈ 13.12.2024):
- [PyPi](https://pypi.org/project/zndraw) (π₯ 1.9K / month Β· π¦ 4 Β· β±οΈ 13.12.2024):
Chemiscope (π₯19 Β· β 140) - An interactive structure/property explorer for materials and molecules. BSD-3 JavaScript
- [GitHub](https://github.com/lab-cosmo/chemiscope) (π¨βπ» 24 Β· π 34 Β· π₯ 400 Β· π¦ 6 Β· π 140 - 28% open Β· β±οΈ 14.11.2024):
- [npm](https://www.npmjs.com/package/chemiscope) (π₯ 27 / month Β· π¦ 3 Β· β±οΈ 15.03.2023):
Elementari (π₯12 Β· β 140) - Interactive browser visualizations for materials science: periodic tables, 3d crystal structures, Bohr atoms, nuclei,.. MIT JavaScript
- [GitHub](https://github.com/janosh/elementari) (π¨βπ» 2 Β· π 13 Β· π¦ 3 Β· π 7 - 28% open Β· β±οΈ 07.10.2024):
- [npm](https://www.npmjs.com/package/elementari) (π₯ 170 / month Β· π¦ 1 Β· β±οΈ 15.01.2024):
Show 1 hidden projects...
- Atomvision (π₯12 Β· β 30 Β· π) - Deep learning framework for atomistic image data.Custom computer-vision experimental-data rep-learn
Wavefunction methods (ML-WFT)
Projects and models that focus on quantities of wavefunction theory methods, such as Monte Carlo techniques like deep learning variational Monte Carlo (DL-VMC), quantum chemistry methods, etc.
DeepQMC (π₯20 Β· β 360 Β· π) - Deep learning quantum Monte Carlo for electrons in real space. MIT
- [GitHub](https://github.com/deepqmc/deepqmc) (π¨βπ» 13 Β· π 62 Β· π¦ 3 Β· π 51 - 5% open Β· β±οΈ 23.10.2024):
- [PyPi](https://pypi.org/project/deepqmc) (π₯ 450 / month Β· β±οΈ 24.09.2024):
FermiNet (π₯13 Β· β 750) - An implementation of the Fermionic Neural Network for ab-initio electronic structure calculations. Apache-2 transformer
- [GitHub](https://github.com/google-deepmind/ferminet) (π¨βπ» 18 Β· π 130 Β· π 57 - 1% open Β· β±οΈ 08.12.2024):
DeepErwin (π₯10 Β· β 54) - DeepErwin is a python 3.8+ package that implements and optimizes JAX 2.x wave function models for numerical solutions.. Custom
- [GitHub](https://github.com/mdsunivie/deeperwin) (π¨βπ» 7 Β· π 8 Β· π₯ 13 Β· π¦ 2 Β· β±οΈ 19.12.2024):
- [PyPi](https://pypi.org/project/deeperwin) (π₯ 190 / month Β· β±οΈ 14.12.2021):
Show 2 hidden projects...
- ACEpsi.jl (π₯6 Β· β 2 Β· π) - ACE wave function parameterizations.MIT rep-eng Julia
- SchNOrb (π₯5 Β· β 61 Β· π) - Unifying machine learning and quantum chemistry with a deep neural network for molecular wavefunctions. MIT
Others
Show 1 hidden projects...
Contribution
Contributions are encouraged and always welcome! If you like to add or update projects, choose one of the following ways:
- Open an issue by selecting one of the provided categories from the issue page and fill in the requested information.
- Modify the projects.yaml with your additions or changes, and submit a pull request. This can also be done directly via the Github UI.
If you like to contribute to or share suggestions regarding the project metadata collection or markdown generation, please refer to the best-of-generator repository. If you like to create your own best-of list, we recommend to follow this guide.
For more information on how to add or update projects, please read the contribution guidelines. By participating in this project, you agree to abide by its Code of Conduct.
License
