# LeMat-Bulk MLIP Hull Reference Datasets

This dataset contains materials close to the convex hull computed using various ML interatomic potentials (MLIPs).

## Dataset Splits

- **`all`**: Contains ALL materials with hull energies for all MLIPs (no threshold filtering)
- **`dft`, `orb`, `uma`, `mace_mp`, `mace_omat`**: Materials within 0.001 eV/atom of respective hulls

## Energy Types

- `dft`: DFT reference energies
- `orb`: ORB model energies
- `uma`: UMA model energies
- `mace_mp`: MACE-MP model energies
- `mace_omat`: MACE-OMAT model energies

## File Structure

- `all_hull_energies.parquet`: All materials with hull energies for all energy types
- `{energy_type}_above_hull_dataset.parquet`: Filtered materials within threshold
- `{energy_type}_above_hull_composition_matrix.npz`: Sparse composition matrix
- `{energy_type}_above_hull_metadata.npz`: Metadata including element mappings

## Usage

```python
from datasets import load_dataset
import pandas as pd
from scipy import sparse
import numpy as np

# Load dataset
ds = load_dataset("LeMaterial/LeMat-Bulk-MLIP-Hull")

# Or load specific files
df = pd.read_parquet("hf://datasets/LeMaterial/LeMat-Bulk-DFT-Hull/threshold_0_001/dft_above_hull_dataset.parquet")
comp_matrix = sparse.load_npz("path/to/dft_above_hull_composition_matrix.npz")
metadata = np.load("path/to/dft_above_hull_metadata.npz", allow_pickle=True)
```