gpt2-prot 0.1

Single NT/AA resoultion biological GPT2 language modelling

gpt2-prot

Train biological language models at single NT or AA resolution.

Todo

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

Installation

Installation from pypi is on the way

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

gpt2-prot fit --config recipes/cas9_analogues.yml  # Run the demo config for cas9 protein language modelling

Development

Running pre-commit hooks

# Install the hooks:
pre-commit install

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

Running tests

Pytest will find all files with the name "test_.py" or "_test.py", run simply by calling pytest from the repo root.