Supplemental code for the thesis Leveraging Large Language Models for Autonomous Task Planning.
conda env create -f env.yml
conda activate roblm
python preprocess.py json_2.1.0/valid_unseen valid_unseen.json --cond knowledge_graph.dot
`python roblm.py --train traindata.json [--chkpt_path <path_to_saved_model>] --model_path <path_to_new_model> --log
- Complete evaluation
python roblm.py --eval validdata.json --chkpt <path_to_saved_model>
- Task based evaluation: validation data split across files
for infile in valid/*.json; do python roblm.py --eval $infile --chkpt <path_to_saved_model>; done