^^Youtube^^

Training parameters are in parameters.py. Adjust as needed.

Run train_model.py to train the model.

Found in the models directory.

• model_weights_mps.pth (Best Performer) was trained on apple silicon with MPS.
• model_weights_cuda.pth was trained on an Nvidia GPU with CUDA. Cuda trains much faster than cpu and mps.

Use test_model.py with appropriate weights path in agent.load_model() to test the model.

• Good introduction to RL and DQN's: Professor Nguyen Slides

• Pytorch Cartpole DQN tutorial

• Original Deepmind DQN Atari paper: Playing Atari with Deep Reinforcement Learning

• Run manual_drive.py to drive the car yourself using the arrow keys. Can you beat the DQN? 🤔