gpt2-prot 0.2

Single NT/AA resoultion biological GPT2 language modelling

gpt2-prot

Train biological language models at single NT or AA resolution.

Roadmap

• Readme instructions
• AWS spot instances demo
• Update recipe configs with new inference flags
• Add inference mode
• Add config recipes for eg. foundation model training, specific protein modelling etc.
• Github actions for publishing the package to pypi
• Docstrings etc.

Installation

pip install gpt2_prot

From source

micromamba create -f environment.yml  # or conda etc.
micromamba activate gpt2-prot

pip install .  # Basic install
pip install -e ".[dev]"  # Install in editable mode with dev dependencies
pip install ".[test]"  # Install the package and all test dependencies

Usage

From the CLI

gpt2-prot -h

# Run the demo config for cas9 protein language modelling
# Since this uses Lightning you can overwrite parameters from the config using the command line
gpt2-prot fit --config recipes/cas9_analogues.yml --max_epochs 10  

Development

Running pre-commit hooks

# Install the hooks:
pre-commit install

# Run all the hooks:
pre-commit run --all-files

# Run unit tests:
pytest