MLIP-liquid-corrections: BSSE and Three-Body Dispersion Labels for OMol25 Molecules

Provider
ByteDance Seed
License
CC-BY-4.0 (source)
Size
13 files
Format
CSV, HDF5 (in sampled files)
Stored at
Hugging Face (https://huggingface.co/datasets/ByteDance-Seed/MLIP-liquid-corrections)
Access
Free and public. No account with the data provider is needed.
Category
Chemistry & Materials
Source
https://huggingface.co/datasets/ByteDance-Seed/MLIP-liquid-corrections

About this dataset

Energy and force correction labels released by ByteDance Seed for improving machine learning interatomic potentials in molecular liquids. Configurations are taken from OMol25, including its SPICE-origin structures, and cover H, C, N, O, F, P, S, Cl and Br. The BSSE subset (419,753 configurations) gives basis-set superposition error energies and forces from the counterpoise method at wB97M-V/def2-TZVPD; the ATM subset (1,606,854 configurations) gives Axilrod-Teller-Muto three-body dispersion energies and forces from the D4 model with wB97M parameters. Labels are corrections to add to or subtract from total energies, not total energies themselves. Each subset is one HDF5 file with a group per configuration (atomic numbers, positions, correction energy and forces, SMILES and atom-mapped SMILES) plus a CSV index of h5_key, atom count, SPICE molecule name and SMILES; about 6.4 GB in total (measured 2026-11-15). Read with h5py. Used to correct density, cohesive energy and liquid structure of MLIPs trained on gas-phase cluster data. Released under CC-BY-4.0; preserve the NOTICE files and cite the OMol25 and SPICE papers as in the repository's CITATION.md.

How to load this dataset in Python

import pandas as pd
df = pd.read_csv("hf://datasets/ByteDance-Seed/MLIP-liquid-corrections/ATM/atm_dataset.csv")
df.head()

Data structure

File types in sampled files: CSV, HDF5

Files

Example file: ATM/atm_dataset.csv

Analyze this dataset without downloading it

On Scigantic, MLIP-liquid-corrections: BSSE and Three-Body Dispersion Labels for OMol25 Molecules 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