Model Overview¶

The torch-molecule library provides a unified interface for molecular property prediction, generation and encoding. All models inherit from the torch_molecule.base.base.BaseModel class and share common methods for model training, evaluation and persistence.

Model Persistence¶

load_from_local: Load a saved model from a local file
save_to_local: Save the current model to a local file
load_from_hf: Load a model from a Hugging Face repository
save_to_hf: Push the current model to a Hugging Face repository
load: Load a model from either local storage or Hugging Face
save: Save the model to either local storage or Hugging Face

(For detailed API documentation of the base class, please refer to Base Modules.)

Molecular Predictor Models

Molecular Property Prediction Models
- Modeling Molecules as Graphs with Graph Neural Networks
- Modeling Molecules as Sequences with RNNs
  - LSTMMolecularPredictor
  - SMILESTransformerMolecularPredictor

Molecular Generator Models

Molecular Generation Models
- Modeling Molecules as Graphs
- Modeling Molecules as Sequences
  - MolGPTMolecularGenerator
  - LSTMMolecularGenerator

Molecular Encoder Models

Molecular Encoder Models