[ML-Agents] 3D Ball 실습

22.07.11

 

 

Unity ML

---

 

• PPO (Proximal Policy Optimization)
• SAC (Soft Actor-Critic)
• Curiosity
• MA-POCA (MultiAgent POsthumous Credit Assignment)

3D Ball 예제

---

C:\ml-agents-main\ml-agents-main\config\ppo

3DBall.yaml

behaviors:
  3DBall:
    trainer_type: ppo
    hyperparameters:
      batch_size: 64
      buffer_size: 12000
      learning_rate: 0.0003 // 
      beta: 0.001
      epsilon: 0.2
      lambd: 0.99
      num_epoch: 3
      learning_rate_schedule: linear
    network_settings:
      normalize: true
      hidden_units: 128
      num_layers: 2
      vis_encode_type: simple
    reward_signals:
      extrinsic:
        gamma: 0.99
        strength: 1.0
    keep_checkpoints: 5
    **max_steps: 500000** // 몇번 돌릴지
    time_horizon: 1000
    **summary_freq: 12000** // 결과를 몇번마다 볼 것 인지

 

• Episode
  • Behavior 기준
• Step
  • Agent 기준
  Creating new learning environments