OMC25 / README.md
Pq234's picture
Update README.md
9227455 verified
|
Raw
History Blame Contribute Delete
5.52 kB

Pq234 license: cc-by-4.0 tags:

  • chemistry pretty_name: Open Molecular Crystals 2025 (OMC25) dataset size_categories:
  • 10M<n<100M Pq234

Open Molecular Crystals 2025 (OMC25)
Dataset

Dataset

####LICENSE: The OMC25 dataset is provided under a CC-BY-4.0 license

OMC25 represents the largest high quality molecular crystal DFT dataset. OMC25 was generated at the PBE-D3 level of theory as implemented in Vienna Ab initio Simulation Package (VASP). OMC25 includes structures sampled from relaxation trajectories of molecular crystals generated by Genarris 3.0 starting from molecules in the OE62 dataset. For more details on the dataset, see arXiv.

Here we provide the OMC25 training and validation splits. Dataset files are written as ase-db LMDBDatabase objects.

Splits Size Starting Molecular crystals Starting Molecules Storage Download
Train 24,870,226 207,271 44,403 139GB train.tar.gz
Val 1,386,816 11,570 2,467 7.6GB val.tar.gz

We also provide further detailed information on all unique initial molecular crystal structures that underwent structural relaxations in omc25-starting-crystals.csv (description of its columns is shown below).

| Column name | Description | ||| | csd_refcode | CSD reference code of molecule from OE62 dataset | | z_value | Number of molecular formula units in the crystal unit cell | | genarris_step | Sampled from generation (gener) or Rigid Press (press) step of Genarris 3.0 | | xtal.id | Unique crystal identifier among putative structures from Genarris step | | split | Structure was included in the training (train) or validation (val) split | | nframes | Number of frames sampled from relaxation trajectory | | mol.composition, xtal.composition | Composition of molecule and crystal, respectively | | mol.natoms, xtal.natoms | Number of atoms in molecule and crystal unit cell, respectively | | mol.mass, xtal.mass | Molar mass in g/mol of molecule and crystal unit cell, respectively | | xtal.spacegroup | Crystal space group with SYMPREC=10-5 symmetry tolerance |

How to read the data

The OMC25 datasets can be accessed with the fairchem library. The package can be installed with:

pip install fairchem-core

Once installed, a dataset can be read as follows

from fairchem.core.datasets import AseDBDataset

dataset_path = "/path/to/omc25/dir/train" dataset = AseDBDataset({"src": dataset_path})

index the dataset to get a structure

atoms = dataset.get_atoms(0) atomic_positions = atoms.positions atomic_numbers = atoms.get_atomic_numbers()


Structures are stored as ASE [Atoms objects](https://wiki.fysik.dtu.dk/ase/ase/atoms.html). Each structure contains DFT total energy (eV), force (eV/Å), and stress (eV/Å<sup>3</sup>) labels. Additionally, `atoms.info` contains several other properties and metadata that may be important
for model development:

`atoms.info:`

sid: XAZMUI-2-gener-b8a1abc1a6916c7-84 ## Unique structure identifier consisting of information below and also the index of the structure frame in the filtered relaxation trajectory csd_refcode: XAZMUI ## CSD reference code of molecule from OE62 dataset z_value: 2 ## Number of molecules in the crystal unit cell genarris_step: gener ## Sampled from generation (gener) or Rigid Press (press) step of Genarris 3.0 xtal.id: b8a1abc1a6916c7 ## Unique crystal identifier among putative structures from Genarris step source: omc25_data/dft_relax/batch1/XAZMUI/2/gener/b8a1abc1a6916c7 ## Unique internal identifier


## Model checkpoints

Baseline model checkpoints trained on the full OMC25 dataset can be found at [https://huggingface.co/facebook/OMC25](https://huggingface.co/facebook/OMC25) and [https://huggingface.co/facebook/UMA](https://huggingface.co/facebook/UMA).

## Support
If you run into any issues feel free to post your questions or comments on [Github Issues](https://github.com/facebookresearch/fairchem/issues).

## Citation
If you use this work, please cite:

@misc{gharakhanyan2025openmolecularcrystals2025omc25dataset, title={Open Molecular Crystals 2025 (OMC25) Dataset and Models}, author={Vahe Gharakhanyan and Luis Barroso-Luque and Yi Yang and Muhammed Shuaibi and Kyle Michel and Daniel S. Levine and Misko Dzamba and Xiang Fu and Meng Gao and Xingyu Liu and Haoran Ni and Keian Noori and Brandon M. Wood and Matt Uyttendaele and Arman Boromand and C. Lawrence Zitnick and Noa Marom and Zachary W. Ulissi and Anuroop Sriram}, year={2025}, eprint={2508.02651}, archivePrefix={arXiv}, primaryClass={physics.chem-ph}, url={https://arxiv.org/abs/2508.02651}, } Pq234