mace-torch
Decision gist · record as of 2026-08-14
Yes. MACE is actively maintained, has low install friction, carries no known vulnerabilities, and is permissively licensed. Install it if you need to train or deploy machine learning interatomic potentials for molecular dynamics or materials simulations. Be aware that PyTorch 2.4.1 is not supported and float64 training requires PyTorch 2.2+; verify your PyTorch version before use.AI-flagged interpretation of the facts on this page — verify before relying
Before you install
- PyTorch >= 1.12 must be installed separately; PyTorch 2.4.1 is not supported; float64 training requires PyTorch 2.2 or later; Python >= 3.9 required
- Low install friction with a pure Python wheel.
- Active maintenance with recent commits and steady updates; requires PyTorch 1.12+ and Python 3.9+, with a note that PyTorch 2.4.1 is not supported and float64 training requires PyTorch 2.2 or later.
License · maintenance · safety
permissive license (permissive) — Licensed permissively, allowing commercial and private use without restriction.
last release 2026-05-10 (96 days) · last repo commit 2026-08-07 · 1,317 stars
0 known vulnerabilities (OSV.dev, 2026-08-14) · 128,560 downloads/mo, #11,702 on PyPI
Alternatives
Verify before relying
pip install mace-torch
import torch
from mace.calculators import MACECalculator
# Load a pretrained model or train your own
calc = MACECalculator(model_paths="path/to/model.model", device="cpu")- Whether pretrained foundation models (MACE-MP, MACE-OFF, MACE-Polar) are included in the package or must be downloaded separately
- GPU memory requirements for training models of different sizes
- Performance comparison with other interatomic potential frameworks
What it is and what it does
MACE is a PyTorch-based framework for training machine learning interatomic potentials using higher-order equivariant message passing. It combines deep learning with physical symmetry constraints to predict atomic forces and energies for molecular dynamics simulations and materials discovery. The package provides command-line tools for training on atomic configurations, evaluating model accuracy, and finetuning pretrained foundation models. It integrates with ASE (Atomic Simulation Environment) for seamless use in molecular dynamics workflows and supports multi-GPU distributed training for large datasets.
The framework handles heterogeneous training data (bulk structures with stress, isolated molecules, etc.) and offers flexible model architecture control through equivariant irreps. It includes utilities for experiment tracking, caching, and visualization of training progress. The package depends on PyTorch, e3nn for equivariant operations, and several scientific libraries (numpy, h5py, pandas, matplotlib) for data handling and analysis.
Use it for
- Train custom interatomic potentials on your own atomic structure datasets for materials-specific simulations
- Finetune pretrained MACE foundation models (MACE-MP, MACE-OFF) on specialized datasets to improve accuracy for your domain
- Run molecular dynamics simulations in ASE using MACE models as the force field calculator
- Evaluate model accuracy on test sets and compare different equivariant architectures for your materials system
- Perform distributed training on large datasets across multiple GPUs using PyTorch's DistributedDataParallel
Worth the install?
AI-flagged interpretation of the facts on this page. Verify before relying on it.
Yes.
MACE is actively maintained, has low install friction, carries no known vulnerabilities, and is permissively licensed. Install it if you need to train or deploy machine learning interatomic potentials for molecular dynamics or materials simulations. Be aware that PyTorch 2.4.1 is not supported and float64 training requires PyTorch 2.2+; verify your PyTorch version before use.
Install
mace-torch on PyPI
Before you install
Low install friction with a pure Python wheel. Active maintenance with recent commits and steady updates; requires PyTorch 1.12+ and Python 3.9+, with a note that PyTorch 2.4.1 is not supported and float64 training requires PyTorch 2.2 or later.
PyTorch >= 1.12 must be installed separately; PyTorch 2.4.1 is not supported; float64 training requires PyTorch 2.2 or later; Python >= 3.9 required
License in practice
Licensed permissively, allowing commercial and private use without restriction.
Quickstart
pip install mace-torch
import torch
from mace.calculators import MACECalculator
# Load a pretrained model or train your own
calc = MACECalculator(model_paths="path/to/model.model", device="cpu")Verify before relying
- Whether pretrained foundation models (MACE-MP, MACE-OFF, MACE-Polar) are included in the package or must be downloaded separately
- GPU memory requirements for training models of different sizes
- Performance comparison with other interatomic potential frameworks
Package facts
| License | permissive license permissive |
| Python support | Supports the current Python release >=3.9 |
| Install friction | Low. Pure-Python wheel |
| Runtime dependencies | 19 packagestorche3nnnumpyopt_einsumasetorch-emaprettytablematscipyh5pytorchmetricspython-hostlistconfigargparseGitPythonpyYAMLtqdmlmdborjsonmatplotlibpandas |
| Maintenance | Actively maintained 96 days since the last release |
| Last repo commit | |
| First released | |
| Downloads | 128,560 / month, #11,702 on PyPI 30-day window, as of 2026-08-14 |
| Known vulnerabilities | None known OSV.dev, checked 2026-08-14 |
| Classifiers | License :: OSI Approved :: MIT LicenseOperating System :: OS IndependentProgramming Language :: Python :: 3Programming Language :: Python :: 3.10Programming Language :: Python :: 3.11Programming Language :: Python :: 3.12Programming Language :: Python :: 3.13Programming Language :: Python :: 3.9 |
Evidence: mace_torch-0.3.16-py3-none-any.whl
Tags
Let your AI agent find packages like this
Example. Real query, live index.
You found this page by searching. An agent finds it by wishing: SkillFed indexes 14,416 PyPI packages by what they can do, searchable in plain language.
wish › “machine learning interatomic potentials”
- mace-torchMACE trains and evaluates machine learning interatomic potentials…
- dpdatadpdata converts and manipulates atomistic simulation data between…
- dscribeDScribe transforms atomic structures into fixed-size numerical…
Give your agent the search over MCP, or paste the wish link into any chat.
More Artificial Intelligence packages
LiteLLM provides a unified Python interface to call 100+ LLM providers (OpenAI, Anthropic, Gemini, Bedrock, Azure, and others) using OpenAI-compatible API format, available as both a Python SDK and a self-hosted AI Gateway proxy server.
Install it if you need to work with multiple LLM providers or want to centralize LLM routing in your organization.
Client library and CLI tool for downloading, uploading, and managing models, datasets, and repositories on the Hugging Face Hub platform.
Install it if you work with Hugging Face Hub models or datasets.
LangChain provides a framework for building agents and LLM-powered applications by composing language models, tools, and memory through a unified API that abstracts over multiple model providers.
hf-xet provides chunk-based deduplication and efficient file transfer for the Hugging Face Hub, enabling faster uploads and downloads of large files with local disk caching.
Tokenizers converts raw text into token sequences for NLP models, with support for training custom vocabularies and using pre-built tokenizers (BPE, WordPiece) optimized for speed via Rust.
Transformers provides a unified framework for loading, fine-tuning, and running state-of-the-art pretrained models across text, vision, audio, video, and multimodal tasks using PyTorch, JAX, or TensorFlow.
Install it if you need to run or train any transformer-based model for NLP, vision, audio, or multimodal tasks.
See also fairchem-core · cuequivariance-torch · chemprop · cuequivariance-ops-torch-cu12 · e3nn · matscipy · e3nn-jax · vesin · pyriemann · cuequivariance