Agent is Not Challenging Error

[fbkhalif]

Hello,
Currently, if I try to combine the simple rl example code with the experimental self play code in the poke-env examples, and try to create an environment with two rl actors playing against one another, I get an error that says "Agent is not challenging". I've tried a number of work arounds such as threading, instantiating separate simple RL players with "none" as opponents and trying to make them manually challenge one another, making changes in the openai api code, etc etc, but still circle back to this error/issue. Was wondering what might be causing this/what could be done to fix/work around this?

[hsahovic]

Hey @fbkhalif - have you looked at https://github.com/hsahovic/poke-env/blob/master/examples/experimental-self-play.py ?

[fbkhalif]

hello, yes - I checked it out and it was quite helpful, and self play with two random agents works well! The only issue is that as soon as I attempt to implement self-play between two simple RL agents, I get an "agent is not challenging" error

[ghost]

Hello @fbkhalif, I am having the same issue and wanted to ask whether you have found a solution to this. I have tried various things you have also mentioned in your first comment and I am thinking the problem lays on the API from the OpenAI.