# ML-Agents
## Installation
https://github.com/Unity-Technologies/ml-agents/blob/develop/docs/Installation.md

conda create -n mlagents python=3.8
conda activate mlagents
pip install mlagents==0.30.0
pip install protobuf==3.20.0

test if installation success:

mlagents-learn --help

## Run Train
https://github.com/Unity-Technologies/ml-agents/blob/develop/docs/Getting-Started.md
Install Unity version 2021.3.11f1 first.

git clone --branch release_20 https://github.com/Unity-Technologies/ml-agents.git
cd ml-agents
(open scene at ml-agents/Project/Assets/ML-Agents/Examples/3DBall/Scenes in Unity)
mlagents-learn config/ppo/3DBall.yaml --run-id=first3DBallRun
mlagents-learn config/ppo/FoodCollector.yaml --run-id=firstFoodCollector

## Install Environment Executable
https://github.com/Unity-Technologies/ml-agents/blob/develop/docs/Learning-Environment-Executable.md


Click build. And save to folder named "ball_exe".


Then you can run train:

mlagents-learn config/ppo/3DBall.yaml --env=ball_exe --run-id=firstRunExe

## Python Gym Interface
https://github.com/Unity-Technologies/ml-agents/blob/develop/docs/Python-Gym-API.md

Unity ML-Agents Gym Wrapper only support an environment with a **single agent**. Use 3DBall as example, there are 12 agents by default. You have to delete other 11 agents, and build it to executable following above procedure. I build it and save to folder named “single_ball_exe”.
* For Windows:
* Have to provide folder name in UnityEnvironment.

import gym

from mlagents_envs.environment import UnityEnvironment
from mlagents_envs.envs.unity_gym_env import UnityToGymWrapper


def main():
	unity_env = UnityEnvironment("single_ball_exe")
	env = UnityToGymWrapper(unity_env)
	obs = env.reset()
	print(obs)
	print(env.action_space)


if __name__ == '__main__':
	main()


* For Ubuntu:
* Have to provide file name without .x86_64 in UnityEnvironment.
* Dedicated Server Build:
* 

import gym

from mlagents_envs.environment import UnityEnvironment
from mlagents_envs.envs.unity_gym_env import UnityToGymWrapper


def main():
    unity_env = UnityEnvironment("3dball_single_server/3dball")
    env = UnityToGymWrapper(unity_env)
    obs = env.reset()
    print(obs)
    print(env.action_space)


if __name__ == '__main__':
    main()


## Create Container

podman run -d -v $PWD:/root -w /root --name mlagents -it ct3_release tmux
podman exec -it mlagents tmux a
(Install Anaconda:)
wget https://repo.anaconda.com/archive/Anaconda3-2022.05-Linux-x86_64.sh
chmod 777 Anaconda3-2022.05-Linux-x86_64.sh
./Anaconda3-2022.05-Linux-x86_64.sh
(Install mlagents:)
conda create -n mlagents python=3.8
conda activate mlagents
pip install mlagents==0.30.0
pip install protobuf==3.20.0
(Run test:)
python test_env.py

## Ubuntu Troubleshoot
Problem:
* no usable version of libssl was found.

Solve:
https://forum.unity.com/threads/workaround-for-libssl-issue-on-ubuntu-22-04.1271405/

sudo dpkg -i libssl1.0.0_1.0.2n-1ubuntu5_amd64.deb