Get Started with ML-Agents in Unity - Part 1: Setup & Installation

In this tutorial, we walk through the setup process for Unity ML-Agents, covering both the Unity and Python components. You’ll learn how to install necessary tools, configure the python environment, and run a basic ML-Agents training example within Unity.

▬ Support My Work ▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬

The 'Unity Asset Store' links provided below are affiliate links. If you click on them and make a purchase, I may receive a small commission at no extra cost to you. This support helps me continue creating content for this channel.

Thank you for your support!

▬ Timestamps ▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬

0:00 - Intro
1:22 - Setup Overview
2:07 - Install Visual C++ Redistributable
2:40 - Install CUDA Toolkit
5:14 - Install cuDNN
6:11 - Install Anaconda
7:08 - Anaconda PowerShell Prompt
8:05 - Unity & Python: Key Roles
9:03 - Set Up Virtual Environment
10:21 - Install Dependencies
11:57 - Download ML-Agents Source Code
12:57 - Install ML-Agents Python Package
13:57 - Open Unity Examples Project
15:18 - Import ML-Agents Unity Package
16:01 - Open Basic Example Scene
16:57 - Start Python Backend
18:18 - Play Basic Example in Unity
19:07 - Training Session Output
19:30 - Resume or Restart Training Session
20:09 - Outro

▬ Useful Links ▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬

▬ Credits ▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬▬

#Unity #AI #mlagents #reinforcementlearning

[Geomephysicus]

Thanks a lot! This is the best, and up-to-date ML-Agents tutorial. I've tried to follow a couple of others but they are either outdated or fail to explain certain details.

[okaneskin]

thanks a lot. I've been having a lot of problems with ml agent for a few days. When I followed what you did it worked flawlessly.

[carson]

Thank you very much for the tutorial, you're a lifesaver. This was very clear and understandable, I often found myself pausing to ask myself a question only to unpause it and have you answer the question right after. Great job 👍

[thibaut]

Simple, great voice recording, thank you and keep going

[brionsarachan]

Great video! Very helpful! I had an older version of ML Agents setup about 3 years ago. I ran into some complications having both old and new versions of everything, but I think I have it all figured out.

[Araes]

Very useful, and everything is very well explained!

[Plum-nt]

thanks you so much! i get started with this!

[Akisk]

I had tried to do this 3 times with other tutorials and I didn't work but with this one I did it, it's a good tutorial 😁

[fabis]

Great Video! Extremly helpful to get started with ML-Agents! Hope more content regarding this topic is coming sooner then later :)

[channelblank]

Thanks a lot for the tutorial, it is much better than any other which i have watched. There are a couple problems though:

the miniconda console is giving me errors.

The behavior name (name of ai object) has not been specified in the trainer configuration. Please add an entry in the configuration file for (name of the ai object), or set default_settings.

The output in unity is telling me that there is code which is deprecated.

are these problems linked?
The scenes go straight into inference mode (except for the basic scene, that one works fine)

[mouse]

Thank you for this tutorial Looking forward for the upcoming videos.

[ryanyao]

Hey just want to say thank you for this video and I am really looking forward to the upcoming videos. I do hope that the topic of self-defined environments can be included.

[Ravi-ikvt]

I was trying to install ml agents since 2 days finally found new video

[ardhenismuhammadaflah]

Man, i love this so hard to find this tutorial rn

[ryanyao]

Btw, do you think it is a feasible way to first generate an executable environment from unity rather than training directly in the Unity hub editor (with duplicate agents for parallel training) and then train the MLAgent for faster training?

[ulysseLivet]

Hi, love your work ! 1) When on the training session the agent teleport itself towards the target... Do you have a clue why ?
2) I'm at step 16000 and my mean reward is always 0.070 but you at step 6000 it's at 0.459. Is it computer based ?

[jea]

Hey, are you going to do a full fledged series of this? I’m looking into training robots (robocup football small size) using unity ml agents but most learning resources being old end deprecated scared me. You would be a life saver

[simongentry]

also... i get this window coming up after following instructions... Please move the project folder somewhere readable and writable. can the project NOT be on the C: drive? ty UPDATE: I found if you open HUB at least once as ADMIN and then open and login again, problem with read only goes away.

[sagrgywejhxcvx]

If I already downloaded anaconda IDE than should I download miniconda? And also I wanna say thanks for such high quality tutorial
Edit: you can use Anaconda IDE, but its size is 20gb

[bethedust]

Hey! Thanks a lot for a tutorial, but what should I do if I loaded my working ML-Agents project into GitHub repository and trying to clone it from my laptop? What are the installation steps for doing this? Thank you