#Training multiple brains simultaneously with editor

1 messages · Page 1 of 1 (latest)

fringe willow
#

Has anybody done this and can provide guidance?

devout dust
#

I'm pretty sure you can just duplicate the project, and run it at the same time on the desktop.

fringe willow
#

But duplicating is not what I need. I want to train 2 brains in same environment in parallel.

#

I.e. these agents with different brains will interact with each other

#

Goal signals will not cover my case

#

There is --base-port BASE_PORT propertly for cli, but it does not work lol

#

Note that when training using the Editor rather " "than an executable, the base port will be ignored.

#

VUaaa

neon token
#

Are Brains not are Behaviors in ML-Agents 3? If so, you should be able to simply setup multiple Behaviors in your yaml config, and set your agents in your scene to the appropriate Behavior. Also, —base-port is used in CLI after building a project to specify the starting port. It can resolve port conflicts if running many envs, such as “mlagents-learn aliens.yaml --run-id=run1 --env aliens.exe --num-envs=128 --no-graphics --base-port 50000”

devout dust