Very basic SC2-AI bot playing Terran and using Q-learning
Switch branches/tags
Nothing to show
Clone or download
Fetching latest commit…
Cannot retrieve the latest commit at this time.
Permalink
Failed to load latest commit information.
.gitignore
LICENSE
README.md
buildings.py
identifiers.py
sparse_agent.py
units.py

README.md

TerranEasyBot

A simple Terran Bot with limited build/training options, using Q-learning.
Originially based on: https://github.com/skjb/pysc2-tutorial

Setup

  1. Install game (tested on version 3.16.1) from https://github.com/Blizzard/s2client-proto
  2. Install maps from same repo. We've used Simple64 from the Melee pack for all testing
  3. Install python3 librarias panda and pysc2 with pip
  4. Run agent with python -m pysc2.bin.agent --map Simple64 --agent sparse_agent.SparseAgent --agent_race T --max_agent_steps 0