# COMP6v2 Benchmark Suite (wB97X/6-31G*), ColabFit Conversion

## Facts

- Provider: Kate Huddleston, Roman Zubatyuk, Justin Smith, Adrian Roitberg, Olexandr Isayev, Ignacio Pickering, Christian Devereux, Kipton Barros; ColabFit conversion
- License: CC-BY-4.0 (https://zenodo.org/api/records/10126157)
- Size: 4 files
- Format: Parquet (in sampled files)
- Stored at: Hugging Face, https://huggingface.co/datasets/colabfit/COMP6v2-wB97X-631Gd
- Access: public, readable without an account
- DOI: https://doi.org/10.60732/cbced4c5
- Category: Chemistry & Materials
- Source: https://doi.org/10.1021/acs.jctc.0c00121
- Page: https://scigantic.com/archive/e35b07d6-3abc-59dd-b166-c204b8b1b66d
- Keywords: Chemistry & Materials, benchmark, Parquet, Hugging Face

## About

COMP6v2 is version 2 of the COmprehensive Machine-learning Potential benchmark suite from the ANI team (Isayev, Roitberg, Smith, Barros and colleagues), extended to molecules containing H, C, N, O, S, F and Cl. It contains the six COMP6 subsets (ANI-MD, DrugBank, GDB07to09, GDB10to13, Tripeptides and S66x8) and is the standard held-out test set for ANI-2x style potentials. This repository is the wB97X/6-31G* portion, the level of theory used to train ANI-2x; other levels (wB97MD3BJ/def2TZVPP, wB97MV/def2TZVPP, B973c/def2mTZVP) are separate ColabFit repositories. ColabFit's card counts 157,718 configurations with energy and atomic forces; the repository is about 1.6 GB (measured 2026-11-14). This is the ColabFit Exchange conversion: configurations are Parquet files under co/ (structure, positions, cell, 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 configuration schema at materials.colabfit.org/docs. The upstream Zenodo release is CC-BY-4.0.

## Load the data

```python
from datasets import load_dataset
ds = load_dataset("colabfit/COMP6v2-wB97X-631Gd")
print(ds)
```

## Data structure

File types in sampled files: Parquet

Example file: `co/co_0.parquet`

## Open it

- In a hosted notebook with the data mounted: https://scigantic.com/archive/e35b07d6-3abc-59dd-b166-c204b8b1b66d
- Full record as JSON (schema card included): https://api.scigantic.com/api/archives/e35b07d6-3abc-59dd-b166-c204b8b1b66d
- Over MCP: `get_schema_card` with id `e35b07d6-3abc-59dd-b166-c204b8b1b66d` at https://api.scigantic.com/mcp

## Related topics

- [Parquet datasets](https://scigantic.com/topics/parquet)
- [Benchmark datasets](https://scigantic.com/topics/benchmark)
- [Chemistry & Materials datasets](https://scigantic.com/topics/chemistry-and-materials)
