Open Molecules 2025 (OMol25) 4M Training Subset, ColabFit Conversion
- Provider
- Meta FAIR, Lawrence Berkeley National Laboratory and collaborators (Levine et al.); ColabFit conversion
- License
- CC-BY-4.0 (source)
- Size
- 11 files
- Format
- Parquet (in sampled files)
- Stored at
- Hugging Face (https://huggingface.co/datasets/colabfit/OMol25_train_4M)
- Access
- Free and public. No account with the data provider is needed.
- Category
- Chemistry & Materials
- Source
- https://doi.org/10.48550/arXiv.2505.08762
About Open Molecules 2025 (OMol25) 4M Training Subset, ColabFit Conversion
The roughly 4 million structure training subset of Open Molecules 2025 (OMol25), the molecular DFT dataset released by Meta FAIR and Lawrence Berkeley National Laboratory, all computed at the wB97M-V/def2-TZVPD level in ORCA. It samples biomolecules (protein fragments, nucleic acids, ligand pockets), metal complexes, electrolytes and ionic liquids, and recomputed community datasets, with charges and spin multiplicities beyond neutral singlets and systems up to about 350 atoms. ColabFit counts 3,986,754 configurations, about 219 million atoms and 83 elements, labelled with total energies and atomic forces. The 4M split is the one the OMol25 paper uses to compare eSEN, GemNet-OC and MACE baselines at manageable cost, short of the full 100M-structure training set. This is the ColabFit Exchange conversion: configurations sit in Parquet files under co/ (positions, cell, species, computed properties and metadata per row) with dataset-level information in ds.parquet; read them with pyarrow, pandas, polars or Hugging Face datasets following the ColabFit schema at materials.colabfit.org/docs. Meta licenses OMol25 under CC-BY-4.0; cite Levine et al., arXiv:2505.08762 (2025).
How to load Open Molecules 2025 (OMol25) 4M Training Subset, ColabFit Conversion in Python
from datasets import load_dataset
ds = load_dataset("colabfit/OMol25_train_4M")
print(ds)Data structure
File types in sampled files: Parquet
Files
Example file: co/co_0.parquet
Analyze this dataset without downloading it
On Scigantic, Open Molecules 2025 (OMol25) 4M Training Subset, ColabFit Conversion opens in a hosted JupyterLab notebook with the data already mounted, so you can read files where they are stored instead of copying them first. Each notebook includes an AI assistant that can see the mounted files and write and run analysis code. Open the dataset page to launch a notebook, or browse the full catalog.
This page is also available as Markdown and as JSON.
Related Chemistry & Materials datasets
- MassBank 2026.03, open-licence records: 96,949 reference mass spectra of 10,304 compounds as MSP and Parquet
- CoRE MOF 2024: Computation-Ready Experimental Metal-Organic Frameworks (Public SI Subset)
- Open Catalyst 2020 (OC20) S2EF 200K Training Set, ColabFit Conversion
- Stable and Metastable CuInS2 (CIS) Chalcopyrite Polytypes for Solar Cells: DFT Formation Energies (Larsen et al. 2022, Materials Cloud), ColabFit Conversion
- JARVIS-ML CFID Descriptors for 84K Materials Project Crystals: Formation Energies (NIST JARVIS, figshare), ColabFit Conversion
- Open Catalyst 2022 (OC22) S2EF Training Set, ColabFit Conversion
- Alexandria PBE 1D Crystal Geometry Optimization Paths, ColabFit Conversion
- GNPS community spectral libraries: 581,722 reference MS/MS spectra from 65 directly contributed libraries, as MGF and Parquet