Question Train 2 or more Unity projects simultaneously

Hi everyone,

I am trying to train 2 (or more) projects simultaneously in Unity ML-Agents, but have received the warning: "Couldn't start socket communication because worker number 0 is still in use". I read that by using the command "--worker-id=" and setting that to some value other than "0", it would work out just fine. However I get another warning: "unrecognized arguments: --worker-id=1".

This bug has been found before (https://github.com/Unity-Technologies/ml-agents/issues/5653), but so far I haven't found a way to go around it.. Can someone please help?

I am using ml-agents version: 0.26.0

Many thanks!

 

Guess it was obvious, but I found a way to do it with these links: https://github.com/Unity-Technologi....md#training-using-concurrent-unity-instances and https://github.com/Unity-Technologi...Executable.md#using-an-environment-executable

 

Sorry, problem was not solved. I managed to train two projects by running one training session in Unity Editor and another with a built environment executable.

Trying a third simultaneous training session gives the error message:

"mlagents_envs.exception.UnityWorkerInUseException: Couldn't start socket communication because worker number Couldn't start socket communication because worker number 0 is still in use. You may need to manually close a previously opened environment or use a different worker number. is still in use. You may need to manually close a previously opened environment or use a different worker number."

Is there some way to change the work number?

Many thanks..

 

You can change the base port number in the command line. The base port is set to 5004 by default.

type mlagents-learn --h to get a list of all the CLI Args

PS: you can do something similar with tensorboard - change the port number so you can see multiple projects results

 

Many thanks! changing the base port number did the trick.