dennisrotondi / deep_rl_world_models_grid_world

Project for the course Deep Learning and Applied AI a.y. 2021/2022, Dept. of Computer Science, Prof. Emanuele Rodolà

Deep Learning and Applied Ai Project:

Dennis Rotondi, World Models in MiniGrid

How to reproduce the results:

• clone this repository on your machine

git clone https://github.com/DennisRotondi/dlai_project.git

• install the requirements inside the cloned folder

pip3 install -r requirements.txt

• install the package

python3 setup.py install

• create the dataset

python3 src/env/grid_save.py 

• train the models {vae, mdnrnn, controller} (pick one each time, in this order)

python3 src/models/{vae, mdnrnn, controller}_train.py
eg.
python3 src/models/vae_train.py

• • or request access to my dvc remote folder (dvc on google drive requires it, also if the folder were public it still need to be shared individually) and pull my uploaded files for dataset and ckpt

dvc pull

• now enjoy watching the agent solving the game... or at least trying

python3 src/env/play.py test_env.visualize=True