CAMT5

Efficient Tokenization for Molecular Language Models

Getting Started

• virtual enviroment (whatever you want. PyEnv, Conda or Pyenv)
• dependency
  • use uv: pip install uv
  • uv pip install -r requirements.txt -r requirements-dev.txt

Project Sturcture

CAMT5/
├── .gitignore
├── .pre-commit-config.yaml
├── asset/
├── config/
│   ├── config.yaml
│   └── task/
├── core/
├── main.py
├── metrics/
├── model/
├── README.md
├── requirements-dev.txt
├── requirements.txt
├── scripts/
├── train/
└── utils.py

By default, you can run tasks under the config directory, supporting pretrain, finetune, and evaluation.

config/

This directory contains configuration files for the project.

• config.yaml: The main configuration file that includes settings for the entire project.
• task/: A directory containing task-specific configuration files. Each file in this directory defines the parameters and settings for a specific task.

scripts/

This directory contains utility scripts that assist in various tasks in various settings. e.g.

./scripts/finetune.sh -e ft_frag

Style Guide

• yapf
• isort
• pre-commit

pre-commit run # --all-files