Preprint: https://arxiv.org/abs/2305.11014
Tom Silver, Soham Dan, Kavitha Srinivas, Joshua B. Tenenbaum, Leslie Pack Kaelbling, Michael Katz (2023)
- Python 3.11+
- Tested on MacOS Catalina
- Recommended: create and source a virtualenv.
pip install -e ".[develop]"
Run ./run_ci_checks.sh
. It should complete with all green successes in 5-10 seconds.
Run ./scripts/run_all.sh
. This reproduces results using cached chat logs. It will take 12-16 hours to complete.
See llm_genplan/llm_cache
.