Search Unity

  1. Unity 6 Preview is now available. To find out what's new, have a look at our Unity 6 Preview blog post.
    Dismiss Notice
  2. Unity is excited to announce that we will be collaborating with TheXPlace for a summer game jam from June 13 - June 19. Learn more.
    Dismiss Notice
  3. Dismiss Notice

Question ML-agents SAC algorithm gets stuck even though PPO works

Discussion in 'ML-Agents' started by jednomije, Nov 24, 2020.

  1. jednomije

    jednomije

    Joined:
    Nov 24, 2020
    Posts:
    1
    I trained my agent to park a car in a simple environment with PPO algorithm and it worked well, but then when I tried to train it with SAC it trains for a bit and after a while it seems to stop doing any actions.
    Agent gets reward for coming closer to the parking spot a gets a big reward after fully parking. Agent also gets reset after moving away from parking spot or not moving for some time.
    Version of Unity: 2019.4.4f1
    Version of ML-agents: 1.0.6
    aa.PNG a.PNG Capture.PNG
     
  2. TreyK-47

    TreyK-47

    Unity Technologies

    Joined:
    Oct 22, 2019
    Posts:
    1,839
    I'll bounce this off the team for some insight and guidance.
     
  3. andrewcoh_unity

    andrewcoh_unity

    Unity Technologies

    Joined:
    Sep 5, 2019
    Posts:
    162
    Hi @jednomije

    Can you also share your console output? It looks like some NaNs are occurring (judging by the entropy plot).