--- pretty_name: Molecules license: other task_categories: - other --- # Molecules This repository stores downloaded 3D molecular datasets collected into one gated Hugging Face dataset repo. ## Included datasets - `geom_drugs`: GEOM-Drugs (~7GB including QM9, ~304K). Status: `uploaded`. Uploaded: `yes`. Cleaned locally: `yes`, staged size 39.8 GB. - `geom_qm9`: GEOM-QM9 (included in GEOM, ~130K). Status: `uploaded`. Uploaded: `yes`. Cleaned locally: `yes`, staged size 148.0 B. - `spice`: SPICE v2 (~7GB, ~19K molecules / 1.1M conformers). Status: `uploaded`. Uploaded: `yes`. Cleaned locally: `yes`, staged size 34.9 GB. - `molecule3d`: Molecule3D (~15GB, ~3.9M). Status: `uploaded`. Uploaded: `yes`. Cleaned locally: `yes`, staged size 1.9 GB. - `pcqm4mv2`: PCQM4Mv2 (~17GB SDF, ~3.8M). Status: `uploaded`. Uploaded: `yes`. Cleaned locally: `yes`, staged size 1.5 GB. - `zinc3d`: ZINC20 3D Conformers (SDF) (~50-200GB depending on tranche selection, hundreds of millions). Status: `uploaded`. Uploaded: `yes`. Cleaned locally: `yes`, staged size 463.4 GB. - `pubchem3d`: PubChem 3D Conformer (hundreds of GB, ~100M+). Status: `uploaded`. Uploaded: `yes`. Cleaned locally: `yes`, staged size 165.4 GB. ## Layout Each dataset is stored under its own top-level folder in this repo. ## Notes - `pubchem3d` is very large. The downloader defaults to a limited subset unless `--pubchem3d-full` is used. - `zinc3d` can also become very large depending on tranche selection. - `geom_qm9` reuses the GEOM archive downloaded for `geom_drugs`.