Blame for metadata.json

{
  "READY": true,
  "name": "QM9_gap",
  "task": "regression",
  "cat_idx": [],
  "num_samples": 131970,
  "num_features": 841,
  "description": "Geometric, energetic, electronic and thermodynamic properties of DFT-modelled small molecules. Task: predict 'gap' using Coulomb matrix features. Target: z-scored with the training-split mean and SD (stored as target_mean / target_std), following DeepChem's default target normalization for this loader. Split: random 70/15/15 (seed 42).",
  "source": "MoleculeNet (https://moleculenet.org) \u2014 loaded via DeepChem.",
  "label": "Drug Discovery",
  "sub_labels": [
    "Quantum Mechanics",
    "Coulomb Matrix"
  ]
}