llm4pddl

Code for PDDL Planning with Pretrained Large Language Models. Silver et al. (2022). NeurIPS Foundation Models for Decision Making Workshop.

Requirements

• Python 3.8+
• Tested on MacOS Catalina and Ubuntu 18.04

Installation

First Time

• (Highly recommended) Make a virtual environment:
  • virtualenv venv
  • source venv/bin/activate
• Clone this repository with submodules: git clone --recursive https://github.com/Learning-and-Intelligent-Systems/llm4pddl.git
• Run pip install -e .[develop] to install the main dependencies for development.
  • If you encounter issues with the transformers dependency on MacOS, we recommend doing the following:
    1. Run curl --proto '=https' --tlsv1.2 -sSf https://sh.rustup.rs | sh in terminal.
    2. Restart the terminal.
    3. Run pip install transformers.
  • If problems persist, see huggingface/transformers#2831.
• Run pip install -e llm4pddl/third_party/pyperplan to install our fork of pyperplan.

Example Command

python llm4pddl/main.py --approach llm-standard --env pyperplan-gripper --num_train_tasks 2 --num_eval_tasks 10 --seed 0 --debug

See llm4pddl/flags.py for additional flags.

Instructions for Contributing

• After pulling the latest changes, also run git submodule update --init --recursive to make sure that you have any changes to the pyperplan submodule.
• You can't push directly to master. Make a new branch in this repository (don't use a fork, since that will not properly trigger the checks when you make a PR). When your code is ready for review, make a PR and request reviews from the appropriate people.
• To merge a PR, you need at least one approval, and you have to pass the 4 checks defined in .github/workflows/llm4pddl.yml, which you can run locally as follows:
  • pytest -s tests/ --cov-config=.coveragerc --cov=llm4pddl/ --cov=tests/ --cov-fail-under=100 --cov-report=term-missing:skip-covered --durations=0
  • mypy .
  • pytest . --pylint -m pylint --pylint-rcfile=.llm4pddl_pylintrc
  • ./run_autoformat.sh
• The first one is the unit testing check, which verifies that unit tests pass and that code is adequately covered. The "100" means that all lines in every file must be covered.
• The second one is the static typing check, which uses Mypy to verify type annotations.
• The third one is the linter check, which runs Pylint with the custom config file .llm4pddl_pylintrc in the root of this repository. Feel free to edit this file as necessary.
• The fourth one is the autoformatting check, which uses the custom config files .style.yapf and .isort.cfg in the root of this repository.

Citation

@inproceedings{silver2022pddl,
  title={PDDL Planning with Pretrained Large Language Models},
  author={Silver, Tom and Hariprasad, Varun and Shuttleworth, Reece S and Kumar, Nishanth and Lozano-P{\'e}rez, Tom{\'a}s and Kaelbling, Leslie Pack},
  booktitle={NeurIPS 2022 Foundation Models for Decision Making Workshop},
  year={2022}
}