Interactive_grounded_learning

Reinforcement Learning project by Team Markovian Marvels

Problem Statement

Build a 3D target structure based on the conversation between an architect (gives instructions) and a builder (asks clarifying questions).

Installation

Create the python environment

conda create -y -n rl_project python=3.9
conda activate rl_project

All the dependencies can be installed using below command

pip install -r requirements.txt

Training

To train the NLP T5 Model:

python nlp_task/train.py

To train the RL Model:

python training_run.py --config_path iglu_baseline.yaml

Evaluation

To evaluate the NLP Model:

cd nlp_task
python nlp_evaluation.py

To evaluate the RL Model:

python evaluation.py