README.qm9.md

metadata

language: []
pretty_name: QM9
config_name: qm9
tags:
  - graphs
  - molecules
  - chemistry
task_categories:
  - graph-regression
homepage: >-
  https://pytorch-geometric.readthedocs.io/en/latest/_modules/torch_geometric/datasets/qm9.html#QM9

qm9

QM9 molecular graphs with 3D coordinates and 19 regression targets (MoleculeNet).

Splits

Split	File
train	train.parquet
validation	val.parquet
test	test.parquet

Features

• mol_id: int64 unique identifier per molecule
• x: list[float32[11]], shape (num_nodes, 11), atom features from PyG QM9
• z: list[int64], shape (num_nodes,), atomic numbers
• pos: list[float32[3]], shape (num_nodes, 3), 3D coordinates in Angstrom
• edge_index: int64[2, num_edges], COO adjacency
• edge_attr: list[float32[4]], shape (num_edges, 4), bond features from PyG QM9
• smiles: string canonical SMILES
• y: list[float32], length 19, regression targets (converted units per PyG)

References

• PyTorch Geometric QM9 implementation: https://pytorch-geometric.readthedocs.io/en/latest/_modules/torch_geometric/datasets/qm9.html#QM9
• MoleculeNet paper: https://arxiv.org/abs/1703.00564