aidd-codebase 0.2.0

High-level codebase for deep learning development in drug discovery.

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

1. Configuration: The coding framework has a number of argument dataclasses in the file arguments.py. This file contains all standard arguments for each of the models. Because they are dataclasses, you can easily adapt them to your own needs.

Does your Seq2Seq adaptation need an extra argument? Import the Seq2SeqArguments from arguments.py, create your own dataclass which inherits it and add your extra argument.

*It is important to note that the order of supplying arguments to a script goes as follows:*
- --flags override config.yaml
- config.yaml overrides default values in arguments.py
- default values from arguments.py are used when no other values are supplied
At the end, it stores all arguments in config.yaml

2. Use: The coding framework has four main parts:
   

• utils
• data_utils
• models
• interpretation

These parts should be used  

3. File Setup: The setup of the files in the system is best used as followed:
   coding_framework
   |-- ..
   ESR X
   |-- project 1
   |-- data
   |-- ..
   |-- Arguments.py
   |-- config.yaml
   |-- main.py
   |-- datamodule.py
   |-- pl_framework.py
   

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.