Super Mario Bros Reinforcement Learning

Let's create an AI that's able to play Super Mario Bros! We'll be using Double Deep Q Network Reinforcement Learning algorithm to do this.

Installation

First, clone this repository

git clone https://github.com/Rahulrai888/Mario-Using-RL.git

Next, create a virtual environment

The command below is for a conda environment, but use whatever you're comfortable with. I'm using Python 3.10.12.

conda create --name smbrl python=3.10.12

Make sure you activate the environment.

conda activate smbrl

Then, install PyTorch v2.0.1

The steps here will be a little different for everyone depending on if you're using a GPU or not. This is why the PyTorch lines in the requirements.txt file are commented out.

If you are using a GPU, it also depends on what version of CUDA you're using (assuming you're using an NVIDIA card). I'm using CUDA 11.8, so I have to go to PyTorch's website and then install PyTorch v2.0.1 for CUDA version 11.8.

For more information, please go to PyTorch's website.

My command looked like:

pip install torch==2.0.1+cu118 torchvision==0.15.2+cu118 torchaudio==2.0.2 --index-url https://download.pytorch.org/whl/cu118

Finally, install the rest of the requirements

pip install -r requirements.txt