Skip to content

Latest commit

 

History

History
68 lines (46 loc) · 6.38 KB

installationGuide.md

File metadata and controls

68 lines (46 loc) · 6.38 KB
Raw

Detailed Installation Guide

This is a more detailed installation guide for AnimalAI v3, written for users who don't have lots of experience with python dependencies, Github repositories, and/or Unity -- or in case you run into trouble with the installation (it's custom software installation... when isn't there a bug or two?)

This guide is currently designed predominantly for Windows users; a more detailed Mac installation guide is a work-in-progress, but most of the guide will still be useful nonetheless. Linux users tend to have stronger command-line-interface/installation experience (this coming from a Windows user!) so the quick install guide on README.md should be sufficient -- but please get in touch ([email protected] / thanksphil GitHub) should you run into trouble.

Installing Python

We recommend a newer version of Python be installed, but any version 3.x is fine. Python is generally easy to install:

  • Download -- default .exe download is fine, but unpacking a .zip download or Mac .pkg will also work
  • Run the download and follow the installation guide. Default installation settings are fine, but it is recommended to tick the "Add to PATH" box, so that Python can be run via the Windows Command Prompt by typing 'python' in any directory:

  • If you're doing a custom intallation, it is recommended to keep the "install pip" box ticked and use pip to install dependencies later on -- but you can use an alternative method e.g. conda

AnimalAI v3 Repository

It's a good idea to create a 'root' folder ready for cloning the main repository, so that you can also keep your own training scripts, a Python virtual environment, and any other external AAI-related work in the same location, and any unrelated folders are kept separate.

Cloning this repository can then be done either by:

  • Downloading the .zip directly and extracting it into your 'root' folder
  • If you have GitHub Desktop (available on Windows and Mac), you can clone it directly by clicking "Open with GitHub Desktop" on the repo webpage
  • Cloning using GitHub's command line interface (GitHub CLI)

It will be saved by default as "animal-ai-main" but can be renamed to something more obvious like "animal-ai-repo" or "animal-ai-python" if you have other AAI folders (e.g. if you have also cloned the AAI Unity Project repo)

(Optional) Creating/Activating a Virtual Environment

Virtual environments in Python are useful for managing packages for separate projects; sometimes you might only want specific packages for a specific project, so you can create a virtual environment for a project and install packages local to that environment / project, that will not be accessible to Python when you use it outside of the virtual environment.

To create one for AnimalAI:

  • Using Python directly:

    • Open a Command Prompt terminal and navigate to a convenient folder (:C\, :C\Users\Name and :C\Users\Name\NewlyCreatedAAIRootFolder are all good choices!)
    • Type python -m venv your_env_name, where your_env_name can be anything you like e.g. "AAI_venv", so long as it doesn't contain spaces
    • To use the virtual environment, navigate to its folder, then type cd Scripts, then activate. Your Command Prompt should now show your virtual environment name next to the directory you are in

  • For conda users:

    • In any directory run conda create --name your_env_name and by default it will be created in the envs folder wherever you have installed conda
    • To activate it, type conda activate your_env_name

Note: folder navigation in Windows is performed using the cd command, e.g. if the current directory is shown as :C\Users\Name and you want to go to your new AAI root folder called "AAI", you would type cd AAI and it will now show you are at :C\Users\Name\AAI. To go to the parent directory (e.g. in this case :C\Users), you would type cd.. and if your directory name contains spaces, use speech marks e.g. cd "AAI Folder"

Python Dependencies

Everything you need to run scripts in AnimalAI v3 (including the correct version of Unity's ml-agents package) is found in the Python package animalai. You can install this via pip or conda, with or without a virtual environment created -- if you're more familiar with a particular method here it's probably best to stick to it.

  • Open a Command Prompt terminal and, if you are using a virtual environment, make sure to activate the environment first! (See above)
  • Navigate to your AAI root folder, then go to your cloned repository "animal-ai-main" (or whatever you named it)
  • If using pip:
    • Run pip install -e animalai (if this doesn't work, ensure pip is installed correctly; pip --version shows the version you have installed to confirm this)
  • If using conda:
    • AnimalAI does not have a package build that is directly compatible with conda itself, but you can still install it indirectly for use in conda
    • Run conda install pip then (making sure you're in the main repository folder as instructed above) run pip install -e animalai as described for pip users

Downloading the AAI Environment

Find the Windows/Mac/Linux environment download links as seen in README.md, and download a .zip copy of the environment according to your operating system. The entire content of the .zip download needs to then be extracted/unzipped into the env folder found in the main repository (along with configs, docs, examples and so on...)

Sometimes .zip folders unzip themselves into their own sub-folder, so it can get a bit confusing. Your env should look like this:

In other words, the AnimalAI Unity .exe should be directly contained in the env folder along with everything else from the .zip contents.

You can then start using AnimalAI v3! Any problems, please get in touch ([email protected] / thanksphil GitHub) and we can try to help.