REINFORCE Agent — CartPole-v1
A REINFORCE policy-gradient agent trained using PyTorch.
Training
- Algorithm: REINFORCE
- Training episodes: 10,000
- Policy architecture: 4 → 64 → 128 → 2
- Optimizer: Adam
- Learning rate: 0.001
- Discount factor: 0.99
Evaluation
- Evaluation episodes: 10
- Mean reward: 500.0
- Standard deviation: 0.0
- Leaderboard score: 500.0
Artifacts
- model.pt: Trained policy weights
- results.json: Evaluation results
- replay.mp4: Recorded agent gameplay
Evaluation results
- mean_reward on CartPole-v1self-reported500.0 +/- 0.0