Search Unity

Search Results

Search Again
Page 1 of 2
  1. ChrissCrass
    Post

    Question What happens when end episode is called when one agent is training and another is using inference?

    Episodes are local to the agent, so calling EndEpisode on one agent should not affect others.
    Post by: ChrissCrass, Aug 16, 2020 in forum: ML-Agents
  2. ChrissCrass
    Post

    Can I train ML-Agent Brains within Unity? (During gameplay)

    Here's the thing: most ML algos are very complicated to implement. On-top of that they're very difficult to make computationally efficient....
    Post by: ChrissCrass, May 11, 2020 in forum: ML-Agents
  3. ChrissCrass
    Post

    Using multiple envs not speeding up training and not using more than 30% of CPU

    What's strange is that this bottleneck only actually occurs when training. According to reports from rz_0lento, the 8 parallel environment setup...
    Post by: ChrissCrass, May 8, 2020 in forum: ML-Agents
  4. ChrissCrass
    Post

    Why should batch size be >= recurrence's sequence length? Why not horizon?

    AFAIK: You can fit multiple sequences within a single batch and learn from that, but since batch is the chunk size of state-action pairs used for...
    Post by: ChrissCrass, May 3, 2020 in forum: ML-Agents
  5. ChrissCrass
    Post

    Problem with Multiple (unique) Agent Env (python only loads one behavior)

    Another ml-agents user in the discord followed the repro steps (mlagents version 14) and confirmed the problem. EDIT: I went ahead and opened up...
    Post by: ChrissCrass, Apr 25, 2020 in forum: ML-Agents
  6. ChrissCrass
    Post

    Disk space keeps reducing during the training

    It sounds like the problem is with your scene. There is no way you are building up more than a gig of observations...
    Post by: ChrissCrass, Apr 25, 2020 in forum: ML-Agents
  7. ChrissCrass
    Post

    Problem with Multiple (unique) Agent Env (python only loads one behavior)

    Did a bit of playing around, the issue is definitely a strange one. A sloppy fix to cut through the strangeness is simply adding a...
    Post by: ChrissCrass, Apr 25, 2020 in forum: ML-Agents
  8. ChrissCrass
    Post

    Problem with Multiple (unique) Agent Env (python only loads one behavior)

    I think I found the final ingredient that should let you reproduce this issue in the 3D ball env: 1: Add a single 3DBallHardNew agent to the...
    Post by: ChrissCrass, Apr 25, 2020 in forum: ML-Agents
  9. ChrissCrass
    Post

    Problem with Multiple (unique) Agent Env (python only loads one behavior)

    Yep, am using different behavior names. I will try re-installing at some point today I think. The agents have their own prefab. The prefabs are...
    Post by: ChrissCrass, Apr 24, 2020 in forum: ML-Agents
  10. ChrissCrass
    Post

    What controls Time.timeScale during training?

    That's for the low-level python API I think. If you're using the terminal to launch training, try out this command: mlagents-learn --help Time...
    Post by: ChrissCrass, Apr 24, 2020 in forum: ML-Agents
  11. ChrissCrass
    Post

    Problem with Multiple (unique) Agent Env (python only loads one behavior)

    This is exactly the problem. I am indeed requesting decisions for both agents, and both agent classes work fine if used individually. The sole...
    Post by: ChrissCrass, Apr 24, 2020 in forum: ML-Agents
  12. ChrissCrass
    Post

    Model behavior vs training

    If you're doing anything in Update rather than FixedUpdate (and are using physics at all) then you will probably see deviations...
    Post by: ChrissCrass, Apr 24, 2020 in forum: ML-Agents
  13. ChrissCrass
    Post

    New Reinforcement Learning Algorithm

    You can definitely implement your own models using TF2, but it isn't simple to set up. Look at the models.py file and go from there (I can't...
    Post by: ChrissCrass, Apr 24, 2020 in forum: ML-Agents
  14. ChrissCrass
    Thread

    Problem with Multiple (unique) Agent Env (python only loads one behavior)

    I am trying to train two different sets of agents in the same environment, but at training time, only one behavior parameter will be loaded in the...
    Thread by: ChrissCrass, Apr 24, 2020, 9 replies, in forum: ML-Agents
  15. ChrissCrass
    Post

    Disk space keeps reducing during the training

    If you are using visual observations then this can be a cause of large buffer sizes, but there should be an upper limit to how much data is being...
    Post by: ChrissCrass, Apr 18, 2020 in forum: ML-Agents
  16. ChrissCrass
    Post

    Does training increase time scale?

    Training does not necessarily use fixedDeltaTime to govern its updates, so training may not appear to be the same speed as when you inference on...
    Post by: ChrissCrass, Apr 3, 2020 in forum: ML-Agents
  17. ChrissCrass
    Post

    Minimum visual observation size

    Only the final encoded result. That sounds plausible, but I'm not at all sure if thin features are sometimes biased out or not. to little...
    Post by: ChrissCrass, Apr 3, 2020 in forum: ML-Agents
  18. ChrissCrass
    Post

    Any possibilities accessing values from the agent's brain?

    Yup, sorry about that.
    Post by: ChrissCrass, Apr 3, 2020 in forum: ML-Agents
  19. ChrissCrass
    Post

    Very important question

    Nope. You could never define a "take over the world reward function" correctly. Only an evolutionary approach could write something so...
    Post by: ChrissCrass, Mar 31, 2020 in forum: ML-Agents
  20. ChrissCrass
    Post

    Minimum visual observation size

    The problem is that if you pool and convolve too many times on a small image, it will end up with negative or puny dimensions on the other end....
    Post by: ChrissCrass, Mar 31, 2020 in forum: ML-Agents
Showing results 1 to 20 of 31 Next >
Search Again
Page 1 of 2