sc2-rl's Introduction

SC2-RL

Creation of a Reinforcement Learning agent that plays Starcraft II. This project emerges from the interest on reinforcement learning and the complexity of a stratigic game like starcraft.

Diagrams

This section will help you understand how the ecosystem works.

Components diagram

Displays the components and the interactions between each other.

Execution diagram

Since the execution of the exosystem has two different times, since half part is asyncronous and the other is synchronus, the diagram tries to reflect what is the workflow.

Setup steps

If you are using WSL, follow this steps: WSL Installation

Starcraft II installed in your system
Python >= 3.7
Install dependencies poetry install
Add dependencies inside environment (tensorflow && tf_agents)
1. .venv/Script/Activate.ps1 (Windows)
2. source .venv/script/activate (Linux/MacOS)
3. pip install tensorflow tf_agents
Install dependecy python-sc2
Create a folder named logs on parent folder
Create a folder named Maps under your Starcraft II installation folder.
Add maps (.SC2MAP files) under previous folder, they are available for download in sites such as sc2mapster.
1. For this use case, scorpion map is used.

In case you are having trouble with environment, once done step 4.3, repeat step 3.

Run

Using docker run redis service
1. docker run -d --name redis-stack-server -p 6379:6379 redis/redis-stack-server:latest
Activate your environment and execute python main.py
1. Linux source .venv/scripts/activate
2. Windows .venv/Scripts/Activate.ps1

Authors

Setup created following setup tutorial by sendtex

Contribute

Commit && PR Messages

[MODULE][FIX|ADD|DELETE] Summary of modifications

* [MODULE2][ADD] List of modifications from a general perspective

Example

[SC2-RL][FIX] Diagrams && subprocess
* [README][ADD] Execution && Component diagram
* [BOT][ADD] Change from `aioredis` to `redis.asyncio`
* [RUN_GAME][DELETED]
* [SC2ENV][FIX] Subprocess platform

sc2-rl's People