mario_agent.json

{
  "log_level": "info",
  "supervised_weight": 1.0,
  "expert_margin": 0.9,
  "memory_capacity": 1e3,
  "batch_size": 32,
  "demo_sampling_ratio": 0.2,
  "update_rate": 0.25,
  "update_repeat": 1,
  "use_target_network": true,
  "target_network_update_rate": 0.001,
  "min_replay_size": 5e2,
  "learning_rate": 0.00025,
  "optimizer": {
    "type": "adam",
    "epsilon": 0.01
  },
  "exploration": {
    "type": "epsilon_decay",
    "epsilon": 0.8,
    "epsilon_final": 0.1
  },
  "repeat_actions": 1,
  "alpha": 0.00025,
  "gamma": 0.9,
  "tau": 1.0
}