marvameraj/Reinforce-Pixelcopter-Unit4

Reinforce Agent playing Pixelcopter-PLE-v0

This is a trained REINFORCE agent playing Pixelcopter-PLE-v0.

This model was trained as part of Unit 4 of the Hugging Face Deep Reinforcement Learning Course.

Environment

Pixelcopter-PLE-v0

Algorithm

REINFORCE / Monte Carlo Policy Gradient

Hyperparameters

• Hidden size: 64
• Training episodes: 50000
• Maximum steps per episode: 10000
• Gamma: 0.99
• Learning rate: 1e-4

Evaluation

• Mean reward: -1.90
• Standard deviation: 1.62
• Course score (mean - std): -3.52

Course requirement: score >= 5.

Course: https://huggingface.co/learn/deep-rl-course/unit4/hands-on