meta_modulus.py

import datasets
import torch
from torch_geometric.data import Data


_CITATION = """\
@article{metamatbench,
  title={MetamatBench: Integrating Heterogeneous Data, Computational Tools, and Visual Interface for Metamaterial Discovery},
  author={Chen, Jianpeng and Zhan, Wangzhi and Wang, Haohui and Jia, Zian and Gan, Jingru and Zhang, Junkai and Qi, Jingyuan and Chen, Tingwei and Huang, Lifu and Chen, Muhao and others},
  journal={arXiv preprint arXiv:2505.20299},
  year={2025}
}
"""

_DESCRIPTION = """
This dataset contains lattice structure data for predicting modulus properties, preprocessed into PyTorch Geometric (PyG) compatible format.
"""


class LatticeModulusDataset(datasets.GeneratorBasedBuilder):
    def _info(self):
        return datasets.DatasetInfo(
            description=_DESCRIPTION,
            citation=_CITATION,
            features=datasets.Features({
                'frac_coords': datasets.Array2D(shape=(None, 3), dtype='float32'),
                'cart_coords': datasets.Array2D(shape=(None, 3), dtype='float32'),
                'node_feat': datasets.Array2D(shape=(None, 4), dtype='float32'),
                'node_type': datasets.Sequence(datasets.Value('int64')),
                'edge_feat': datasets.Array2D(shape=(None, 1), dtype='float32'),
                'edge_index': datasets.Array2D(shape=(2, None), dtype='int64'),
                'lengths': datasets.Array2D(shape=(1, 3), dtype='float32'),
                'num_nodes': datasets.Value('int64'),
                'num_atoms': datasets.Value('int64'),
                'angles': datasets.Array2D(shape=(1, 3), dtype='float32'),
                'vector': datasets.Array2D(shape=(1, 9), dtype='float32'),
                'y': datasets.Array2D(shape=(1, 12), dtype='float32'),
                'young': datasets.Array2D(shape=(1, 3), dtype='float32'),
                'shear': datasets.Array2D(shape=(1, 3), dtype='float32'),
                'poisson': datasets.Array2D(shape=(1, 6), dtype='float32'),
            }),
        )

    def _split_generators(self, dl_manager):
        file_path = dl_manager.download_and_extract('data.pkl')
        return [datasets.SplitGenerator(name=datasets.Split.TRAIN, gen_kwargs={'filepath': file_path})]

    def _generate_examples(self, filepath):
        dataset = torch.load(filepath)

        for idx, data in enumerate(dataset):
            yield idx, {
                'frac_coords': data.frac_coords.numpy(),
                'cart_coords': data.cart_coords.numpy(),
                'node_feat': data.node_feat.numpy(),
                'node_type': data.node_type.numpy().tolist(),
                'edge_feat': data.edge_feat.numpy(),
                'edge_index': data.edge_index.numpy(),
                'lengths': data.lengths.numpy(),
                'num_nodes': data.num_nodes,
                'num_atoms': data.num_atoms,
                'angles': data.angles.numpy(),
                'vector': data.vector.numpy(),
                'y': data.y.numpy(),
                'young': data.young.numpy(),
                'shear': data.shear.numpy(),
                'poisson': data.poisson.numpy(),
            }