(Help wanted) problem during copycating a ml-agent project

Unity Engine
,
1

Hi,

I’m copycatting a brilliant ml-agent project called “micro-invader (GitHub - robot-uprising-hq/ai-guide: Micro Invaders: an introduction to the event)”

When doing it, I’ve found out that I need ‘.nn’ file instead of ‘.onnx’ file to run remote-brain.

but, after ‘ml-agent release 10’, it only export ‘.onnx’ file not ‘.nn’ file, when the project requires '.nn. file.

While searching on Google about it, I found out that I can get ‘.nn’ file by doing these following.

  1. add " --tensorflow " to mlagents-learn on CLI
  2. add " framwork: tenserflow " in configuration.yaml

I’ve tried the first method by

mlagents-learn --tensorflow config\trainer_config.yaml --run-id=Test01

but the promt gives me an warning message “the command line argument --tensorflow was removed.”

and for the second method, I have absolutely no idea where to put it.

It would mean a lot to me if you teach me how to, either

  1. how to run ‘ml-agent release 6’ not the latest.
  2. how to export ‘.nn’ file instead of ‘.onnx’ file ( preffered )

forgive my poor English skills… I’m not native English speaker.

thank you guys and have a wonderfulday

2

I’m pretty sure the project must mention what release of ml agents that works.
All older versions are available online on github.
I think it’s simpler to just use old version of ml agents that fiddle with models. Something might break when converting between formats and then there is nothing gained anyway.