---
dataset_info:
- config_name: Z_1-90
  features:
  - name: material_id
    dtype: string
  - name: formation_energy_per_atom
    dtype: float64
  - name: dft_band_gap
    dtype: float64
  - name: pretty_formula
    dtype: string
  - name: e_above_hull
    dtype: float64
  - name: elements
    sequence: string
  - name: spacegroup_number
    dtype: float64
  - name: azure_bulk_modulus
    dtype: float64
  - name: larsen_score_2d
    dtype: float64
  - name: Si_100_mismatch
    dtype: float64
  - name: azure_band_gap
    dtype: float64
  - name: dft_bulk_modulus
    dtype: float64
  - name: dft_poisson_ratio
    dtype: float64
  - name: dft_mag_density
    dtype: float64
  - name: structure
    dtype: string
  - name: unique_elements
    sequence: string
  - name: num_sites
    dtype: int64
  splits:
  - name: train
    num_bytes: 69299317
    num_examples: 25923
  - name: valid
    num_bytes: 22967236
    num_examples: 8650
  - name: test
    num_bytes: 22926999
    num_examples: 8647
  download_size: 42557908
  dataset_size: 115193552
- config_name: default
  features:
  - name: material_id
    dtype: string
  - name: formation_energy_per_atom
    dtype: float64
  - name: dft_band_gap
    dtype: float64
  - name: pretty_formula
    dtype: string
  - name: e_above_hull
    dtype: float64
  - name: elements
    sequence: string
  - name: spacegroup_number
    dtype: float64
  - name: azure_bulk_modulus
    dtype: float64
  - name: larsen_score_2d
    dtype: float64
  - name: Si_100_mismatch
    dtype: float64
  - name: azure_band_gap
    dtype: float64
  - name: dft_bulk_modulus
    dtype: float64
  - name: dft_poisson_ratio
    dtype: float64
  - name: dft_mag_density
    dtype: float64
  - name: structure
    dtype: string
  - name: unique_elements
    sequence: string
  - name: num_sites
    dtype: int64
  splits:
  - name: train
    num_bytes: 72244062
    num_examples: 27136
  - name: valid
    num_bytes: 23916305
    num_examples: 9047
  - name: test
    num_bytes: 23864869
    num_examples: 9046
  download_size: 44322023
  dataset_size: 120025236
configs:
- config_name: Z_1-90
  data_files:
  - split: train
    path: Z_1-90/train-*
  - split: valid
    path: Z_1-90/valid-*
  - split: test
    path: Z_1-90/test-*
- config_name: default
  data_files:
  - split: train
    path: data/train-*
  - split: valid
    path: data/valid-*
  - split: test
    path: data/test-*
---

The dataset is created with this pipeline:
```bash
git clone https://github.com/microsoft/mattergen.git
cd mattergen
git lfs pull -I data-release/mp-20/ --exclude=""
unzip data-release/mp-20/mp_20.zip -d datasets
```

```python
import pandas as pd
from tqdm.auto import tqdm
from pymatgen.io.cif import CifParser
from datasets import DatasetDict, Dataset

train_path = "./datasets/mp_20/train.csv"
valid_path = "./datasets/mp_20/val.csv"
test_path = "./datasets/mp_20/test.csv"

train_df = pd.read_csv(train_path)
train_structures = [
    CifParser.from_str(s).parse_structures(primitive=True, on_error="ignore")[0]
    for s in tqdm(train_df["cif"], desc="Parsing CIFs", miniters=20)
]
train_df = train_df.drop(columns=['Unnamed: 0', 'cif'])
train_df["structure"] = [s.to_json() for s in tqdm(train_structures, miniters=20)]

valid_df = pd.read_csv(valid_path)
valid_structures = [
    CifParser.from_str(s).parse_structures(primitive=True, on_error="ignore")[0]
    for s in tqdm(valid_df["cif"], desc="Parsing CIFs", miniters=20)
]
valid_df = valid_df.drop(columns=['Unnamed: 0', 'cif'])
valid_df["structure"] = [s.to_json() for s in tqdm(valid_structures, miniters=20)]

test_df = pd.read_csv(test_path)
test_structures = [
    CifParser.from_str(s).parse_structures(primitive=True, on_error="ignore")[0]
    for s in tqdm(test_df["cif"], desc="Parsing CIFs", miniters=20)
]
test_df = test_df.drop(columns=['Unnamed: 0', 'cif'])
test_df["structure"] = [s.to_json() for s in tqdm(test_structures, miniters=20)]



train_dataset = Dataset.from_pandas(train_df)
valid_dataset = Dataset.from_pandas(valid_df)
test_dataset = Dataset.from_pandas(test_df)

dataset_dict = DatasetDict({
    'train': train_dataset,
    'valid': valid_dataset,
    'test': test_dataset
})
dataset_dict.push_to_hub('xpanceo-team/mp-20', private=True)
```