OpenViSTAI · Agents · Blockchain · Crypto
Submit a project
Menu
AI Models

Omatg

OMatG is an open-source Python generative framework for predicting crystal structures and generating inorganic crystals de novo. It provides training, generation and evaluation workflows, YAML-based configuration, materials datasets and pretrained checkpoints.

Explore the practical guide ↓

Independent OpenViST project profile, based on linked sources.

Omatg — OpenViST project profile in AI Models.
OpenViST project illustration. See the linked sources for product details.
Understand the project

Practical overview

When it fits

Use OMatG when a research workflow needs generative modeling for inorganic crystal structures, either for a fixed composition or for de novo generation. Its command-line and configuration approach suits users prepared to install and configure a Python machine-learning package.

Who it is for

  • Computational materials researchers studying crystal structure prediction.
  • Materials-design researchers exploring candidate inorganic crystals.
  • Machine-learning practitioners training or evaluating generative models on materials datasets.

Documented uses

  • Generate candidate crystal coordinates and lattice vectors for a specified atomic composition.
  • Generate inorganic crystals in which atomic species, coordinates and lattice vectors can all change.
  • Configure model training, datasets and hyperparameters with YAML files.
  • Use pretrained checkpoints or train and evaluate models with materials datasets.
Before you choose

Requirements & limits

The repository is public and identifies the project as MIT-licensed. Installation is through the Python package in the repository; pretrained checkpoints and configuration files are linked through Hugging Face.

  • The two generation modes differ in whether atomic species are fixed or generated alongside coordinates and lattice vectors.
  • Workflow choices and hyperparameters are configured through the PyTorch Lightning CLI, typically using YAML files.
  • The repository includes several standard materials datasets and supports dataset configuration based on LMDB or CSV files.
  • OMatG-IRL is a separate package for aligning pretrained models to downstream objectives with black-box reward functions and policy-gradient reinforcement learning.

Check current requirements in the official documentation. Source-based guidance does not mean OpenViST has run or independently tested this project.

Source-based practical guide

Install OMatG for local research workflows

Install the repository package and its command-line entry point for training, generation and evaluation.

Before you start

  • Python 3.11–3.13
  • The OMatG repository and its dependencies
  • PyTorch 2.8; the README recommends choosing the build for the relevant compute platform
  • PyG 2.7 and PyTorch Scatter 2.1 are also called out in the installation notes

Steps

  1. Install the dependencies listed in the repository's pyproject.toml.

  2. From the repository base directory, run the command below.

    Documented command
    pip install .

  3. For developer editable mode, run the command below instead.

    Documented command
    pip install -e .

Expected outcome

The `omg` command is installed for training, generation and evaluation.

Guide limitations

  • The README says to rerun pip install . after repository code changes when installed normally; editable installs make code changes available directly.
Documentation 1 ↗
Evidence and transparency

Sources & review

Official repository ↗
Source record details
Source availability
Active
Sources checked
16 October 2026
Original website title
FERMat-ML/OMatG · Official homepage title · retained in its source language
How we review sourcesSuggest a correction