search

Unity-Technologies / ml-agents

The Unity Machine Learning Agents Toolkit (ML-Agents) is an open-source project that enables games and simulations to serve as environments for training intelligent agents using deep reinforcement learning and imitation learning.
https://unity.com/products/machine-learning-agents
Other
16.93k stars 4.14k forks source link

Failure to train example VisualPyramids #2072

Closed fradino closed 5 years ago

fradino commented 5 years ago

I failed in training VisualPyramids. I didn't change anything, but just run the training. It always shows :Mean Reward: -1.000

shihzy commented 5 years ago

hi @fradino - can you provide some additional details. are you using the editor to train, or did you build the project and use parameter in mlagents-learn?

fradino commented 5 years ago

I used the editor to train, and use the parameter in trainer-config.yaml

shihzy commented 5 years ago

can you send a screenshot or dump of what you are seeing in after running mlagents-learn?

fradino commented 5 years ago

图片 图片 The reward is always -1.

shihzy commented 5 years ago

hi @fradino - if you are running training in the editor, you need to remove the --env=C:\Users... argument when you run mlagents-learn. Only use --env when you want to run training outside of the editor on a separate build of the project. Let me know if this helps answer your question.

fradino commented 5 years ago

Oh, I'm sorry, I ran training outside of the editor. But it's not the main problem. The main problem is I cannot train the example VisualPyramids by the parameters in trainer-config.yaml. They are officially provided.

shihzy commented 5 years ago

in the academy game object, before you built the Unity scene, did you make sure "control" is checked in the broadcast hub?

xiaomaogy commented 5 years ago

Thank you for the discussion. We are closing this issue due to inactivity. Feel free to reopen it if you’d like to continue the discussion though.

github-actions[bot] commented 3 years ago

This thread has been automatically locked since there has not been any recent activity after it was closed. Please open a new issue for related bugs.