conda create -n robosense python=3.8.19
conda activate robosense
pip install -r requirements.txt
data: Contains the INK-USC/G-PlanET dataset, and scripts to convert the dataset for huggingface,tapex and iterative training .
script: Includes all the script needed in the main experiment.
evaluation: The proceses needed to run test on the models framework and its code
main: Includes all the code used in the main experiment.
data_init.sh: - Runs initiaiizing dataset procesing steps.bart_train.sh: Trains the mmodel.bart_test.sh: Evaluates model training, validating the model against test data.bart_iter_test.sh: Runs a test on the model with iterative self-training.bart_table_iter_test.sh: Ealuates model trainig, validating the model against iterative training table.tapex_train.sh: Trains and evaluates tapex modeltapex_test.sh: Tests tapex model, validating against test data.
Models used: facebook/bart-base and facebook/bart-large. All models can be found on the HuggingFace Hub.