DeepChem officially supports Python 3.6 through 3.7 and requires these packages on any condition.
- joblib
- NumPy
- pandas
- scikit-learn
- SciPy
- TensorFlow
- deepchem>=2.4.0 depends on TensorFlow v2 (2.3.x)
- deepchem<2.4.0 depends on TensorFlow v1 (>=1.14)
DeepChem has a number of "soft" requirements.
| Package name | Version | Location where this package is used (dc: deepchem) |
|---|---|---|
| BioPython | latest | :code:dc.utlis.genomics_utils |
| Deep Graph Library | 0.5.x | dc.feat.graph_data, dc.models.torch_models |
| DGL-LifeSci | 0.2.x | dc.models.torch_models |
| HuggingFace Transformers | Not Testing | dc.feat.smiles_tokenizer |
| LightGBM | latest | :code:dc.models.gbdt_models |
| matminer | latest | :code:dc.feat.materials_featurizers |
| MDTraj | latest | :code:dc.utils.pdbqt_utils |
| Mol2vec | latest | :code:dc.utils.molecule_featurizers |
| Mordred | latest | :code:dc.utils.molecule_featurizers |
| NetworkX | latest | :code:dc.utils.rdkit_utils |
| OpenAI Gym | Not Testing | dc.rl |
| OpenMM | latest | :code:dc.utils.rdkit_utils |
| PDBFixer | latest | :code:dc.utils.rdkit_utils |
| Pillow | latest | :code:dc.data.data_loader, dc.trans.transformers |
| PubChemPy | latest | :code:dc.feat.molecule_featurizers |
| pyGPGO | latest | :code:dc.hyper.gaussian_process |
| Pymatgen | latest | :code:dc.feat.materials_featurizers |
| PyTorch | 1.6.0 | dc.data.datasets |
| PyTorch Geometric | 1.6.x (with PyTorch 1.6.0) | dc.feat.graph_data dc.models.torch_models |
| RDKit | latest | Many modules (we recommend you to install) |
| simdna | latest | :code:dc.metrics.genomic_metrics, dc.molnet.dnasim |
| Tensorflow Probability | 0.11.x | :code:dc.rl |
| Weights & Biases | Not Testing | dc.models.keras_model, dc.models.callbacks |
| XGBoost | latest | :code:dc.models.gbdt_models |
| Tensorflow Addons | latest | :code:dc.models.optimizers |