README.md

metadata

tags:
  - Deathmatch-v1
  - ppo
  - deep-reinforcement-learning
  - reinforcement-learning
  - custom-implementation
  - deep-rl-course
model-index:
  - name: PPO
    results:
      - task:
          type: reinforcement-learning
          name: reinforcement-learning
        dataset:
          name: Deathmatch-v1
          type: Deathmatch-v1
        metrics:
          - type: mean_reward
            value: 38.03 +/- 26.60
            name: mean_reward
            verified: false

PPO Agent Playing Deathmatch-v1

This is a trained model of a PPO agent playing Deathmatch-v1.

Hyperparameters

{'env_id': 'Deathmatch-v1'
'num_envs': 32
'num_steps': 128
'learning_rate': 0.00025
'learning_rate_min': 0.0001
'gamma': 0.99
'gae_lambda': 0.95
'clip_coef': None
'total_timesteps': 10000000
'recurrence': 32
'ent_coef': 0.001
'vf_coef': 0.5
'max_grad_norm': 0.1
'num_minibatches': 4
'update_epochs': 1
'frame_skip': 4}