#unity #mlagents #ArtificialIntelligence #MachineLearning #ReinforcementLearning #AI #NeuralNetworks #gamedev
In this video, I will show you how to install ML-Agents Release 20 for Unity 2022.3 or later, step by step, to train your first Machine Learning project using Unity and Reinforcement Learning.
Unity is a 3d game development engine providing high-precision physics simulation. The tutorial in this video shows a step-by-step description of how to install Unity's Artificial Intelligence (AI) framework. ML Agents allows you to develop a learning environment, including games, robots, wind farms (https://www.youtube.com/watch?v=sjjBpFkaTGw), wildfire management systems (https://www.youtube.com/watch?v=aiVRAK3g4qs&t=4s) and many more use cases with a high level of abstraction or close to reality. At the end of the video, we will train one of my personal "Hello, World!" examples, the Push Block example, where an agent learns to push a block into a specified area. The deep neural networks are used to train a whole school of agents using the state-of-the-art learning algorithm Proximal Policy Optimisation (PPO). Fortunately, Unity's ML-Agents framework provides us with a python communicator, utilising PyTorch for training.
Before you begin, please follow this guide to install:
1. Anaconda and Visual Studio Code: https://www.youtube.com/watch?v=ljFwYKL6-1U
2. Unity and Visual Studio Code Setup: https://www.youtube.com/watch?v=WKDKjaEBhs8&t
00:00 - Intro
00:22 - ML Agents
ML Agents: https://github.com/Unity-Technologies/ml-agents
ML Agents - Getting-Started: https://github.com/Unity-Technologies/ml-agents/blob/develop/docs/Getting-Started.md
ML Agents - Installation: https://github.com/Unity-Technologies/ml-agents/blob/develop/docs/Installation.md
03:20 - Install Git
Git Download: https://git-scm.com/downloads
04:30 - Clone ML Agents
git clone --branch release_20 https://github.com/Unity-Technologies/ml-agents.git
06:20 - Create Anaconda Env
conda create -n mlagents20 python=3.9
conda activate mlagents20
07:10 - Install PyTorch
pip3 install torch -f https://download.pytorch.org/whl/torch_stable.html
07:50 - Install Cuda 11.8
nvcc --version
https://developer.nvidia.com/cuda-11-8-0-download-archive
09:45 - Install ML Agents
pip3 install -e ./ml-agents-envs
pip3 install -e ./ml-agents
mlagents-learn --help
11:10 - Install Protobuf 3.20.3
pip install protobuf==3.20.3
12:20 - Open Unity Project
13:50 - Training PushBlock
mlagents-learn config/ppo/PushBlock.yaml --run-id="testing_pushblock"
15:00 - Install onnx
pip install onnx
15:15 - Re-run Training
mlagents-learn config/ppo/PushBlock.yaml --run-id="testing_pushblock" --force
16:00 - Test Neural Network