UEF Summerschool on deep reinforcement learning in video games, 2019

This file contains instructions to setup environment for the practicals. We recommend using pre-made packages below if you are using Windows 10

Pre-made packages for Windows 10

For easiest installation, download this archive file and extract it to location of your choosing (approx. 3GB in size after extraction). Then launch extracted start_uefdrl19.bat. This should open a command prompt with environment activated (text (userdrl19) in the beginning of prompt).

To test if the environment works correctly, launch ipython in the prompt you opened and then command import numpy, tensorflow, keras, gym, vizdoom . If there are no errors, then environment works (don’t mind any FutureWarnings it might print).

Manual installation

Installing miniconda

Following the instruction and install Miniconda from this link: https://conda.io/miniconda.html

On Windows: Launch “Anaconda Prompt”.

Create the environment

Download environment file here, and then command

conda env create -f=environment.yml

You also need to install ViZDoom (Note: You have to activate environment first):

Linux: Install Linux dependencies and then install ViZDoom with pip install vizdoom
MacOS: Install MacOS dependencies and then install ViZdoom with pip install vizdoom
Windows: Download precompiled binaries, and extract the contents into [path_to_environment]/Lib/site-packages

Using installed environment

Activating and deactivating our environment:

For Windows

activate uefdrl19

deactivate uefdrl19

For Linux/Mac

source activate uefdrl19

source deactivate

Listing installed packages:

conda list

Delete environment

conda remove --name uefdrl19 --all