This curated list contains 430 awesome open-source projects with a total of 190K 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.
Explanation
Combined project-quality score
Star count from GitHub
New project (less than 6 months old)
Inactive project (6 months no activity)
Dead project (12 months no activity)
Project is trending up or down
Project was recently added
Contributors count from GitHub
Fork count from GitHub
Issue count from GitHub
Last update timestamp on package manager
Download count from package manager
Number of dependent projects
Active learning
Projects that focus on enabling active learning, iterative learning schemes for atomistic ML.
FLARE (21 · 300) - An open-source Python package for creating fast and accurate interatomic potentials. MIT C++ ML-IAP
IPSuite (15 · 19) - A Python toolkit for FAIR development and deployment of machine-learned interatomic potentials. EPL-2.0 ML-IAP MD workflows HTC FAIR
Finetuna (10 · 46 · ) - Active Learning for Machine Learning Potentials. MIT
Show 3 hidden projects...
Community resources
Projects that collect atomistic ML resources or foster communication within community.
AI for Science Map - Interactive mindmap of the AI4Science research field, including atomistic machine learning, including papers,..
Atomic Cluster Expansion - Atomic Cluster Expansion (ACE) community homepage.
CrystaLLM - Generate a crystal structure from a composition. language-models generative pretrained transformer
GAP-ML.org community homepage ML-IAP
matsci.org - A community forum for the discussion of anything materials science, with a focus on computational materials science..
Matter Modeling Stack Exchange - Machine Learning - Forum StackExchange, site Matter Modeling, ML-tagged questions.
ACE / GRACE support - Support forum for the Atomic Cluster Expansion (ACE) and extensions.
Best-of Machine Learning with Python (23 · 18K) - A ranked list of awesome machine learning Python libraries. Updated weekly. CC-BY-4.0 general-ml Python
OpenML (19 · 670) - Open Machine Learning. BSD-3 datasets
Graph-based Deep Learning Literature (18 · 4.8K) - links to conference publications in graph-based deep learning. MIT general-ml rep-learn
MatBench Discovery (16 · 110) - An evaluation framework for machine learning models simulating high-throughput materials discovery. MIT datasets benchmarking model-repository
MatBench (15 · 130 · ) - Matbench: Benchmarks for materials science property prediction. MIT datasets benchmarking model-repository
GT4SD - Generative Toolkit for Scientific Discovery (14 · 340) - Gradio apps of generative models in GT4SD. MIT generative pretrained drug-discovery model-repository
AI for Science Resources (13 · 530) - List of resources for AI4Science research, including learning resources. GPL-3.0 license
Neural-Network-Models-for-Chemistry (11 · 97) - A collection of Nerual Network Models for chemistry. Unlicensed rep-learn
GNoME Explorer (9 · 900) - Graph Networks for Materials Exploration Database. Apache-2 datasets materials-discovery
Awesome Materials Informatics (8 · 390) - Curated list of known efforts in materials informatics, i.e. in modern materials science. Custom
optimade.science (8 · 8) - A sky-scanner Optimade browser-only GUI. MIT datasets
Awesome Neural Geometry (7 · 920) - A curated collection of resources and research related to the geometry of representations in the brain, deep networks,.. Unlicensed educational rep-learn
Awesome-Graph-Generation (7 · 300) - A curated list of up-to-date graph generation papers and resources. Unlicensed rep-learn
Awesome Neural SBI (7 · 100) - Community-sourced list of papers and resources on neural simulation-based inference. MIT active-learning
Awesome-Crystal-GNNs (7 · 72) - This repository contains a collection of resources and papers on GNN Models on Crystal Solid State Materials. MIT
AI for Science paper collection (7 · 69 · ) - List the AI for Science papers accepted by top conferences. Apache-2
The Collection of Database and Dataset Resources in Materials Science (6 · 280) - A list of databases, datasets and books/handbooks where you can find materials properties for machine learning.. Unlicensed datasets
Show 7 hidden projects...
Datasets
Datasets, databases and trained models for atomistic ML.
Alexandria Materials Database - A database of millions of theoretical crystal structures (3D, 2D and 1D) discovered by machine learning accelerated..
Catalysis Hub - A web-platform for sharing data and software for computational catalysis research!.
Citrination Datasets - AI-Powered Materials Data Platform. Open Citrination has been decommissioned.
crystals.ai - Curated datasets for reproducible AI in materials science.
DeepChem Models - DeepChem models on HuggingFace. model-repository pretrained language-models
Graphs of Materials Project 20190401 - The dataset used to train the MEGNet interatomic potential. ML-IAP
HME21 Dataset - High-temperature multi-element 2021 dataset for the PreFerred Potential (PFP).. UIP
JARVIS-Leaderboard ( 61) - A large scale benchmark of materials design methods: https://www.nature.com/articles/s41524-024-01259-w. model-repository benchmarking community-resource educational
Materials Project - Charge Densities - Materials Project has started offering charge density information available for download via their public API.
Materials Project Trajectory (MPtrj) Dataset - The dataset used to train the CHGNet universal potential. UIP
matterverse.ai - Database of yet-to-be-sythesized materials predicted using state-of-the-art machine learning algorithms.
MPF.2021.2.8 - The dataset used to train the M3GNet universal potential. UIP
NRELMatDB - Computational materials database with the specific focus on materials for renewable energy applications including, but..
Quantum-Machine.org Datasets - Collection of datasets, including QM7, QM9, etc. MD, DFT. Small organic molecules, mostly.
sGDML Datasets - MD17, MD22, DFT datasets.
MoleculeNet - A Benchmark for Molecular Machine Learning. benchmarking
ZINC15 - A free database of commercially-available compounds for virtual screening. ZINC contains over 230 million purchasable.. graph biomolecules
ZINC20 - A free database of commercially-available compounds for virtual screening. ZINC contains over 230 million purchasable.. graph biomolecules
FAIR Chemistry datasets (26 · 910 · ) - Datasets OC20, OC22, etc. Formerly known as Open Catalyst Project. MIT catalysis
OPTIMADE Python tools (25 · 71) - Tools for implementing and consuming OPTIMADE APIs in Python. MIT
MPContribs (25 · 37) - Platform for materials scientists to contribute and disseminate their materials data through Materials Project. MIT
Open Databases Integration for Materials Design (OPTIMADE) (17 · 83) - Specification of a common REST API for access to materials databases. CC-BY-4.0
load-atoms (17 · 38) - download and manipulate atomistic datasets. MIT data-structures
Meta Open Materials 2024 (OMat24) Dataset (15 · 910 · ) - Contains over 100 million Density Functional Theory calculations focused on structural and compositional diversity. CC-BY-4.0
QH9 (13 · 530) - A Quantum Hamiltonian Prediction Benchmark. CC-BY-NC-SA-4.0 ML-DFT
SPICE (11 · 160) - A collection of QM data for training potential functions. MIT ML-IAP MD
AIS Square (9 · 12) - 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
Materials Data Facility (MDF) (9 · 10 · ) - A simple way to publish, discover, and access materials datasets. Publication of very large datasets supported (e.g.,.. Apache-2
3DSC Database (6 · 15) - Repo for the paper publishing the superconductor database with 3D crystal structures. Custom superconductors materials-discovery
The Perovskite Database Project (5 · 60 · ) - Perovskite Database Project aims at making all perovskite device data, both past and future, available in a form.. Unlicensed community-resource
Show 16 hidden projects...
Data Structures
Projects that focus on providing data structures used in atomistic machine learning.
dpdata (23 · 200) - A Python package for manipulating atomistic data of software in computational science. LGPL-3.0
Metatensor (22 · 54) - Self-describing sparse tensor data format for atomistic machine learning and beyond. BSD-3 Rust C-lang C++ Python
mp-pyrho (18 · 37) - Tools for re-griding volumetric quantum chemistry data for machine-learning purposes. Custom ML-DFT
dlpack (14 · 910) - common in-memory tensor structure. Apache-2 C++
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.
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 (25 · 34K) - This library provides basic building blocks that can construct DFT calculations as a differentiable program. Apache-2
MALA (20 · 82) - Materials Learning Algorithms. A framework for machine learning materials properties from first-principles data. BSD-3
QHNet (13 · 530) - Artificial Intelligence Research for Science (AIRS). GPL-3.0 rep-learn
DeepH-pack (12 · 250) - Deep neural networks for density functional theory Hamiltonian. LGPL-3.0 Julia
SALTED (12 · 30) - Symmetry-Adapted Learning of Three-dimensional Electron Densities. GPL-3.0
Grad DFT (10 · 79 · ) - GradDFT is a JAX-based library enabling the differentiable design and experimentation of exchange-correlation.. Apache-2
DeePKS-kit (9 · 100 · ) - a package for developing machine learning-based chemically accurate energy and density functional models. LGPL-3.0
HamGNN (8 · 64) - An E(3) equivariant Graph Neural Network for predicting electronic Hamiltonian matrix. GPL-3.0 rep-learn magnetism C-lang
Q-stack (7 · 15) - Stack of codes for dedicated pre- and post-processing tasks for Quantum Machine Learning (QML). MIT excited-states general-tool
ChargE3Net (6 · 38) - Higher-order equivariant neural networks for charge density prediction in materials. MIT rep-learn
InfGCN for Electron Density Estimation (5 · 11 · ) - Official implementation of the NeurIPS 23 spotlight paper of InfGCN. MIT rep-learn neural-operator
Show 21 hidden projects...
Educational Resources
Tutorials, guides, cookbooks, recipes, etc.
AI for Science 101 community-resource rep-learn
AL4MS 2023 workshop tutorials active-learning
Quantum Chemistry in the Age of Machine Learning - Book, 2022.
AI4Chemistry course (11 · 150 · ) - EPFL AI for chemistry course, Spring 2023. https://schwallergroup.github.io/ai4chem_course. MIT chemistry
jarvis-tools-notebooks (9 · 67) - A Google-Colab Notebook Collection for Materials Design: https://jarvis.nist.gov/. NIST
DSECOP (9 · 44) - This repository contains data science educational materials developed by DSECOP Fellows. CCO-1.0
iam-notebooks (9 · 26) - Jupyter notebooks for the lectures of the Introduction to Atomistic Modeling. Apache-2
COSMO Software Cookbook (9 · 17) - A cookbook wtih recipes for atomic-scale modeling of materials and molecules. BSD-3
MACE-tutorials (6 · 41) - Another set of tutorials for the MACE interatomic potential by one of the authors. MIT ML-IAP rep-learn MD
Show 19 hidden projects...
Explainable Artificial intelligence (XAI)
Projects that focus on explainability and model interpretability in atomistic ML.
exmol (18 · 290) - Explainer for black box models that predict molecule properties. MIT
MEGAN: Multi Explanation Graph Attention Student (6 · 8) - Minimal implementation of graph attention student model architecture. MIT rep-learn
Show 1 hidden projects...
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...
General Tools
General tools for atomistic machine learning.
RDKit (36 · 2.7K) - BSD-3 C++
DeepChem (33 · 5.6K) - Democratizing Deep-Learning for Drug Discovery, Quantum Chemistry, Materials Science and Biology. MIT
Matminer (28 · 480) - Data mining for materials science. Custom
JARVIS-Tools (23 · 320) - JARVIS-Tools: an open-source software package for data-driven atomistic materials design. Publications:.. Custom
QUIP (21 · 360) - libAtoms/QUIP molecular dynamics framework: https://libatoms.github.io. GPL-2.0 MD ML-IAP rep-eng Fortran
MAML (20 · 370 · ) - Python for Materials Machine Learning, Materials Descriptors, Machine Learning Force Fields, Deep Learning, etc. BSD-3
MAST-ML (20 · 110) - MAterials Simulation Toolkit for Machine Learning (MAST-ML). MIT
XenonPy (16 · 140 · ) - XenonPy is a Python Software for Materials Informatics. BSD-3
Scikit-Matter (15 · 76) - A collection of scikit-learn compatible utilities that implement methods born out of the materials science and.. BSD-3 scikit-learn
Artificial Intelligence for Science (AIRS) (13 · 530) - Artificial Intelligence Research for Science (AIRS). GPL-3.0 license rep-learn generative ML-IAP MD ML-DFT ML-WFT biomolecules
MLatom (13 · 69) - AI-enhanced computational chemistry. MIT UIP ML-IAP MD ML-DFT ML-ESM transfer-learning active-learning spectroscopy structure-optimization
Show 11 hidden projects...
Generative Models
Projects that implement generative models for atomistic ML.
GT4SD (14 · 340) - GT4SD, an open-source library to accelerate hypothesis generation in the scientific discovery process. MIT pretrained drug-discovery rep-learn
MoLeR (13 · 280 · ) - Implementation of MoLeR: a generative model of molecular graphs which supports scaffold-constrained generation. MIT
SchNetPack G-SchNet (13 · 49) - G-SchNet extension for SchNetPack. MIT
PMTransformer (12 · 86) - Universal Transfer Learning in Porous Materials, including MOFs. MIT transfer-learning pretrained transformer
SiMGen (9 · 17 · ) - Zero Shot Molecular Generation via Similarity Kernels. MIT viz
COATI (5 · 100 · ) - COATI: multi-modal contrastive pre-training for representing and traversing chemical space. Apache-2 drug-discovery multimodal pretrained rep-learn
Show 8 hidden projects...
Interatomic Potentials (ML-IAP)
Machine learning interatomic potentials (aka ML-IAP, MLIAP, MLIP, MLP) and force fields (ML-FF) for molecular dynamics.
DeePMD-kit (28 · 1.5K) - A deep learning package for many-body potential energy representation and molecular dynamics. LGPL-3.0 C++
fairchem (26 · 910 · ) - FAIR Chemistrys library of machine learning methods for chemistry. Formerly known as Open Catalyst Project. MIT pretrained UIP rep-learn catalysis
NequIP (22 · 650) - NequIP is a code for building E(3)-equivariant interatomic potentials. MIT
MACE (22 · 560) - MACE - Fast and accurate machine learning interatomic potentials with higher order equivariant message passing. MIT
GPUMD (22 · 480) - GPUMD is a highly efficient general-purpose molecular dynamic (MD) package and enables machine-learned potentials.. GPL-3.0 MD C++ electrostatics
DP-GEN (22 · 310) - The deep potential generator to generate a deep-learning based model of interatomic potential energy and force field. LGPL-3.0 workflows
TorchMD-NET (21 · 340) - Training neural network potentials. MIT MD rep-learn transformer pretrained
apax (20 · 18) - A flexible and performant framework for training machine learning potentials. MIT
n2p2 (16 · 220 · ) - n2p2 - A Neural Network Potential Package. GPL-3.0 C++
KLIFF (16 · 34) - KIM-based Learning-Integrated Fitting Framework for interatomic potentials. LGPL-2.1 probabilistic workflows
Neural Force Field (15 · 250) - Neural Network Force Field based on PyTorch. MIT pretrained
PyXtalFF (15 · 87 · ) - Machine Learning Interatomic Potential Predictions. MIT
NNPOps (15 · 86) - High-performance operations for neural network potentials. MIT MD C++
Ultra-Fast Force Fields (UF3) (14 · 62) - UF3: a python library for generating ultra-fast interatomic potentials. Apache-2
wfl (14 · 36) - Workflow is a Python toolkit for building interatomic potential creation and atomistic simulation workflows. GPL-2.0 workflows HTC
DMFF (13 · 160 · ) - DMFF (Differentiable Molecular Force Field) is a Jax-based python package that provides a full differentiable.. LGPL-3.0
ANI-1 (12 · 220 · ) - ANI-1 neural net potential with python interface (ASE). MIT
PiNN (12 · 110) - A Python library for building atomic neural networks. BSD-3
So3krates (MLFF) (12 · 93) - Build neural networks for machine learning force fields with JAX. MIT
Pacemaker (12 · 72) - Python package for fitting atomic cluster expansion (ACE) potentials. Custom
tinker-hp (11 · 82) - Tinker-HP: High-Performance Massively Parallel Evolution of Tinker on CPUs & GPUs. Custom
PyNEP (11 · 49) - A python interface of the machine learning potential NEP used in GPUMD. MIT
Allegro (10 · 360) - Allegro is an open-source code for building highly scalable and accurate equivariant deep learning interatomic.. MIT
calorine (10 · 14) - A Python package for constructing and sampling neuroevolution potential models. https://doi.org/10.21105/joss.06264. Custom
CCS_fit (10 · 8 · ) - Curvature Constrained Splines. GPL-3.0
ACEfit (10 · 7) - MIT Julia
ACE.jl (9 · 65) - Parameterisation of Equivariant Properties of Particle Systems. Custom Julia
Point Edge Transformer (PET) (9 · 19) - Point Edge Transformer. MIT rep-learn transformer
Asparagus (9 · 8 · ) - Program Package for Sampling, Training and Applying ML-based Potential models https://doi.org/10.48550/arXiv.2407.15175. MIT workflows sampling MD
SIMPLE-NN v2 (8 · 41 · ) - SIMPLE-NN is an open package that constructs Behler-Parrinello-type neural-network interatomic potentials from ab.. GPL-3.0
GAP (8 · 40) - Gaussian Approximation Potential (GAP). Custom
ALF (8 · 31) - A framework for performing active learning for training machine-learned interatomic potentials. Custom active-learning
ACE1.jl (8 · 20) - Atomic Cluster Expansion for Modelling Invariant Atomic Properties. Custom Julia
TurboGAP (8 · 16) - The TurboGAP code. Custom Fortran
MLXDM (6 · 7) - A Neural Network Potential with Rigorous Treatment of Long-Range Dispersion https://doi.org/10.1039/D2DD00150K. MIT long-range
Show 35 hidden projects...
Language Models
Projects that use (large) language models (LMs, LLMs) or natural language procesing (NLP) techniques for atomistic ML.
paper-qa (27 · 6.5K · ) - High accuracy RAG for answering questions from scientific documents with citations. Apache-2 ai-agent
OpenBioML ChemNLP (18 · 150) - ChemNLP project. MIT datasets
ChemCrow (13 · 640 · ) - Open source package for the accurate solution of reasoning-intensive chemical tasks. MIT ai-agent
NIST ChemNLP (12 · 73) - ChemNLP: A Natural Language Processing based Library for Materials Chemistry Text Data. MIT literature-data
AtomGPT (11 · 34) - AtomGPT: Atomistic Generative Pretrained Transformer for Forward and Inverse Materials Design.. Custom generative pretrained transformer
ChatMOF (9 · 66) - Predict and Inverse design for metal-organic framework with large-language models (llms). MIT generative
LLaMP (8 · 68) - 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
LLM-Prop (7 · 29 · ) - A repository for the LLM-Prop implementation. MIT
crystal-text-llm (5 · 85) - Large language models to generate stable crystals. CC-BY-NC-4.0 materials-discovery
SciBot (5 · 29) - SciBot is a simple demo of building a domain-specific chatbot for science. Unlicensed ai-agent
MAPI_LLM (5 · 9 · ) - A LLM application developed during the LLM March MADNESS Hackathon https://doi.org/10.1039/D3DD00113J. MIT ai-agent dataset
Cephalo (5 · 8) - Multimodal Vision-Language Models for Bio-Inspired Materials Analysis and Design. Apache-2 generative multimodal pretrained
Show 10 hidden projects...
Materials Discovery
Projects that implement materials discovery methods using atomistic ML.
MatterGen - A generative model for inorganic materials design https://doi.org/10.48550/arXiv.2312.03687. generative proprietary
aviary (14 · 48) - The Wren sits on its Roost in the Aviary. MIT
BOSS (11 · 21) - Bayesian Optimization Structure Search (BOSS). Apache-2 probabilistic
AGOX (10 · 13) - AGOX is a package for global optimization of atomic system using e.g. the energy calculated from density functional.. GPL-3.0 structure-optimization
Materials Discovery: GNoME (9 · 900) - Graph Networks for Materials Science (GNoME) and dataset of 381,000 novel stable materials. Apache-2 UIP datasets rep-learn proprietary
Show 7 hidden projects...
Mathematical tools
Projects that implement mathematical objects used in atomistic machine learning.
KFAC-JAX (19 · 250) - Second Order Optimization and Curvature Estimation with K-FAC in JAX. Apache-2
gpax (18 · 210 · ) - Gaussian Processes for Experimental Sciences. MIT probabilistic active-learning
SpheriCart (16 · 73) - Multi-language library for the calculation of spherical harmonics in Cartesian coordinates. MIT
Polynomials4ML.jl (11 · 12) - Polynomials for ML: fast evaluation, batching, differentiation. MIT Julia
GElib (8 · 19) - C++/CUDA library for SO(3) equivariant operations. MPL-2.0 C++
COSMO Toolbox (6 · 7 · ) - Assorted libraries and utilities for atomistic simulation analysis. Unlicensed C++
Show 5 hidden projects...
Molecular Dynamics
Projects that simplify the integration of molecular dynamics and atomistic machine learning.
JAX-MD (25 · 1.2K) - Differentiable, Hardware Accelerated, Molecular Dynamics. Apache-2
FitSNAP (19 · 160) - Software for generating machine-learning interatomic potentials for LAMMPS. GPL-2.0
mlcolvar (18 · 93) - A unified framework for machine learning collective variables for enhanced sampling simulations. MIT sampling
openmm-torch (17 · 190) - OpenMM plugin to define forces with neural networks. Custom ML-IAP C++
OpenMM-ML (12 · 84) - High level API for using machine learning models in OpenMM simulations. MIT ML-IAP
pair_nequip (10 · 41) - LAMMPS pair style for NequIP. MIT ML-IAP rep-learn
PACE (9 · 27) - The LAMMPS ML-IAP `pair_style pace`, aka Atomic Cluster Expansion (ACE), aka ML-PACE,.. Custom
pair_allegro (8 · 37) - LAMMPS pair style for Allegro deep learning interatomic potentials with parallelization support. MIT ML-IAP rep-learn
SOMD (6 · 13) - Molecular dynamics package designed for the SIESTA DFT code. AGPL-3.0 ML-IAP active-learning
Show 1 hidden projects...
Reinforcement Learning
Projects that focus on reinforcement learning for atomistic ML.
Show 2 hidden projects...
Representation Engineering
Projects that offer implementations of representations aka descriptors, fingerprints of atomistic systems, and models built with them, aka feature engineering.
cdk (26 · 500) - The Chemistry Development Kit. LGPL-2.1 cheminformatics Java
DScribe (23 · 400 · ) - DScribe is a python package for creating machine learning descriptors for atomistic systems. Apache-2
MODNet (16 · 82) - MODNet: a framework for machine learning materials properties. MIT pretrained small-data transfer-learning
GlassPy (13 · 29) - Python module for scientists working with glass materials. GPL-3.0
Rascaline (12 · 48) - Computing representations for atomistic machine learning. BSD-3 Rust C++
SISSO (11 · 250) - A data-driven method combining symbolic regression and compressed sensing for accurate & interpretable models. Apache-2 Fortran
fplib (11 · 7) - libfp is a library for calculating crystalline fingerprints and measuring similarities of materials. MIT C-lang single-paper
NICE (7 · 12 · ) - NICE (N-body Iteratively Contracted Equivariants) is a set of tools designed for the calculation of invariant and.. MIT
milad (6 · 30) - Moment Invariants Local Atomic Descriptor. GPL-3.0 generative
SA-GPR (6 · 19) - Public repository for symmetry-adapted Gaussian Process Regression (SA-GPR). LGPL-3.0 C-lang
Show 15 hidden projects...
Representation Learning
General models that learn a representations aka embeddings of atomistic systems, such as message-passing neural networks (MPNN).
PyG Models (35 · 22K) - Representation learning models implemented in PyTorch Geometric. MIT general-ml
Deep Graph Library (DGL) (35 · 14K) - Python package built to ease deep learning on graph, on top of existing DL frameworks. Apache-2
e3nn (25 · 980) - A modular framework for neural networks with Euclidean symmetry. MIT
SchNetPack (24 · 790 · ) - SchNetPack - Deep Neural Networks for Atomistic Systems. MIT
MatGL (Materials Graph Library) (22 · 290) - Graph deep learning library for materials. BSD-3 multifidelity
ALIGNN (21 · 240) - Atomistic Line Graph Neural Network https://scholar.google.com/citations?user=9Q-tNnwAAAAJ&hl=en.. Custom
NVIDIA Deep Learning Examples for Tensor Cores (20 · 14K · ) - State-of-the-Art Deep Learning scripts organized by models - easy to train and deploy with reproducible accuracy and.. Custom educational drug-discovery
e3nn-jax (19 · 190) - jax library for E3 Equivariant Neural Networks. Apache-2
matsciml (19 · 160) - Open MatSci ML Toolkit is a framework for prototyping and scaling out deep learning models for materials discovery.. MIT workflows benchmarking
DIG: Dive into Graphs (18 · 1.9K · ) - A library for graph deep learning research. GPL-3.0
Uni-Mol (18 · 730) - Official Repository for the Uni-Mol Series Methods. MIT pretrained
kgcnn (16 · 110 · ) - Graph convolutions in Keras with TensorFlow, PyTorch or Jax. MIT
Graphormer (15 · 2.2K · ) - Graphormer is a general-purpose deep learning backbone for molecular modeling. MIT transformer pretrained
escnn (15 · 370) - Equivariant Steerable CNNs Library for Pytorch https://quva-lab.github.io/escnn/. Custom
HydraGNN (14 · 68) - Distributed PyTorch implementation of multi-headed graph convolutional neural networks. BSD-3
hippynn (12 · 72) - python library for atomistic machine learning. Custom workflows
Atom2Vec (10 · 35 · ) - Atom2Vec: a simple way to describe atoms for machine learning. MIT
Compositionally-Restricted Attention-Based Network (CrabNet) (10 · 15) - Predict materials properties using only the composition information!. MIT
EquiformerV2 (8 · 220) - [ICLR 2024] EquiformerV2: Improved Equivariant Transformer for Scaling to Higher-Degree Representations. MIT
Equiformer (8 · 210) - [ICLR 2023 Spotlight] Equiformer: Equivariant Graph Attention Transformer for 3D Atomistic Graphs. MIT transformer
graphite (8 · 64) - A repository for implementing graph network models based on atomic structures. MIT
DeeperGATGNN (8 · 48 · ) - Scalable graph neural networks for materials property prediction. MIT
T-e3nn (7 · 11) - Time-reversal Euclidean neural networks based on e3nn. MIT magnetism
Show 35 hidden projects...
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.
TeaNet - Universal neural network interatomic potential inspired by iterative electronic relaxations.. ML-IAP
PreFerred Potential (PFP) - Universal neural network potential for material discovery https://doi.org/10.1038/s41467-022-30687-9. ML-IAP proprietary
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 (27 · 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
CHGNet (19 · 260 · ) - Pretrained universal neural network potential for charge-informed atomistic modeling https://chgnet.lbl.gov. Custom ML-IAP MD pretrained electrostatics magnetism structure-relaxation
MACE-MP (18 · 540) - Pretrained foundation models for materials chemistry. MIT ML-IAP pretrained rep-learn MD
SevenNet (16 · 140) - SevenNet (Scalable EquiVariance Enabled Neural Network) is a graph neural network interatomic potential package that.. GPL-3.0 ML-IAP MD pretrained
M3GNet (15 · 250) - Materials graph network with 3-body interactions featuring a DFT surrogate crystal relaxer and a state-of-the-art.. BSD-3 ML-IAP pretrained
Orb Models (14 · 210 · ) - ORB forcefield models from Orbital Materials. Custom ML-IAP pretrained
MLIP Arena Leaderboard (13 · 49) - Fair and transparent benchmark of machine-learned interatomic potentials (MLIPs), beyond basic error metrics. Apache-2 ML-IAP community-resource
GRACE (9 · 19 · ) - GRACE models and gracemaker (as implemented in TensorPotential package). Custom ML-IAP pretrained MD rep-learn rep-eng
Joint Multidomain Pre-Training (JMP) (5 · 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
Unsupervised Learning
Projects that focus on unsupervised learning (USL) for atomistic ML, such as dimensionality reduction, clustering and visualization.
DADApy (19 · 110) - Distance-based Analysis of DAta-manifolds in python. Apache-2
ASAP (11 · 140) - ASAP is a package that can quickly analyze and visualize datasets of crystal or molecular structures. MIT
Show 5 hidden projects...
Visualization
Projects that focus on visualization (viz.) for atomistic ML.
Crystal Toolkit (24 · 160) - Crystal Toolkit is a framework for building web apps for materials science and is currently powering the new Materials.. MIT
pymatviz (22 · 180) - A toolkit for visualizations in materials informatics. MIT general-tool probabilistic
Chemiscope (19 · 130) - An interactive structure/property explorer for materials and molecules. BSD-3 JavaScript
ZnDraw (18 · 34) - A powerful tool for visualizing, modifying, and analysing atomistic systems. EPL-2.0 MD generative JavaScript
Elementari (13 · 140) - Interactive browser visualizations for materials science: periodic tables, 3d crystal structures, Bohr atoms, nuclei,.. MIT JavaScript
Show 1 hidden projects...
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 (22 · 360) - Deep learning quantum Monte Carlo for electrons in real space. MIT
FermiNet (13 · 740) - An implementation of the Fermionic Neural Network for ab-initio electronic structure calculations. Apache-2 transformer
DeepErwin (6 · 53) - DeepErwin is a python 3.8+ package that implements and optimizes JAX 2.x wave function models for numerical solutions.. Custom
Show 2 hidden projects...
来源：
https://github.com/JuDFTteam/best-of-atomistic-machine-learning