# COLL Molecular Collisions Dataset (Training Split), ColabFit Conversion

## Facts

- Provider: Johannes Gasteiger, Shankari Giri, Johannes T. Margraf, Stephan Gunnemann; ColabFit conversion
- License: CC-BY-4.0 (https://api.figshare.com/v2/articles/13289165)
- Size: 4 files
- Format: Parquet (in sampled files)
- Stored at: Hugging Face, https://huggingface.co/datasets/colabfit/COLL_train
- Access: public, readable without an account
- DOI: https://doi.org/10.60732/f95867ef
- Category: Chemistry & Materials
- Source: https://doi.org/10.48550/arXiv.2011.14115
- Page: https://scigantic.com/archive/7559bda9-9e63-56ad-90db-3e26f3486557
- Keywords: Chemistry & Materials, DFT, Parquet, Hugging Face

## About

The training split of COLL, a dataset of configurations from molecular collisions of small organic molecules introduced by Gasteiger, Giri, Margraf and Gunnemann to test directional message passing networks (DimeNet++) on high-energy, far-from-equilibrium geometries. Energies and forces for 140,000 random snapshots from collision trajectories were recomputed with DFT at the revPBE/def2-TZVP level with D3 dispersion. ColabFit's card counts 119,965 configurations (C, H, O) with energy, atomization energy and atomic forces; the repository is about 0.22 GB (measured 2026-11-14). Validation and test splits are separate ColabFit repositories. 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 figshare record (COLL v1.2) is CC-BY-4.0.

## Load the data

```python
from datasets import load_dataset
ds = load_dataset("colabfit/COLL_train")
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/7559bda9-9e63-56ad-90db-3e26f3486557
- Full record as JSON (schema card included): https://api.scigantic.com/api/archives/7559bda9-9e63-56ad-90db-3e26f3486557
- Over MCP: `get_schema_card` with id `7559bda9-9e63-56ad-90db-3e26f3486557` at https://api.scigantic.com/mcp

## Related topics

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