AIDD Codebase

A high-level codebase for deep learning development in drug discovery applications using PyTorch-Lightning.

Dependencies

The codebase requires the following additional dependencies

• CUDA >= 11.4
• PyTorch >= 1.9
• Pytorch-Lightning >= 1.5
• RDKit
• Optionally supports: tensorboard and/or wandb

Installation

The codebase can be installed from PyPI using pip, or your package manager of choice, with

$ pip install aidd-codebase

Usage

The codebase is designed to be used in a modular fashion. The main components are the DataModule, Model, and Trainer classes. The DataModule is responsible for loading and preprocessing data, the Model is responsible for defining the model architecture, and the Trainer is responsible for training the model. The Trainer is a subclass of pytorch_lightning.Trainer and can be used as such. The DataModule and Model classes are designed to be used with the Trainer class, but can be used independently if desired.

Starting a new project

$ python -m aidd_codebase.start_project name dir_path

This will create a new project folder with the following structure:

name
├── conf
│   └── config.yaml
├── src
└── main.py

The conf folder contains the configuration file for the project. The src folder contains the source code for the project. The main.py file is the entry point for the project.

Contributors

All fellows of the AIDD consortium have contributed to the packaged.

Code of Conduct

Everyone interacting in the codebase, issue trackers, chat rooms, and mailing lists is expected to follow the PyPA Code of Conduct.