Software Architecture

Good day. I was hoping someone could explain how the Editor and Standalone player interact with the ml-agent learning process.

I played with it a little but all from the same PC, meaning I just followed the installation instructions.

Reading the command line parameters this particular port (see below) description confused my understanding of how the learning process works. It seems to me that the Editor or Standalone require the ml-agent training process to be local and it also provides the unity instance some direction on how it hosts an API as it seems the ml-agent helps each instance with assigning a port for communication. I would have assumed that there would have been a configuration on the actual Unity editor/standalone side that pointed to a running agent and not the other way around (as it seems from the description below)

What Id ultimately like to understand is if and how I can have either the Editor or Standalone to learn from a remote server instance of ml-agents learn function.

Outside of the Editor add-in does unity require python or tensorflow to be installed locally or can that be on a separate remote machine?

----- command line description ------


--base-port: Specifies the starting port. Each concurrent Unity environment instance will get assigned a port sequentially, starting from the base-port. Each instance will use the port (base_port + worker_id), where the worker_id is sequential IDs given to each instance from 0 to num_envs - 1. Default is 5005. Note: When training using the Editor rather than an executable, the base port will be ignored.