Multi-Agent Reinforcement learning on StarCraft 2

Reinforcement Learning Agents implmented in Tensorflow, featuring:

• morden Eager Execution (human readable) structure
• Mutiple CPU Support (HPC)

List of implemented Agents are:

• Random Agent
• Modified A3C Agent
• Modified A2C Agent

Important

Checkout the following branch before running

• A3C
• A2C

Benchmarks

See our this repo

Running

Benchmark

Run MoveToBeacon map with random agent

python benchmark.py --map MoveToBeacon --agent_name RandomAgent

Training

Please checkout different branch

Prepare

Python environment

You sould create or activate Virtual Environment now

$ source .env/bin/activate
(.env) ...$ 

Install Dependency

pip install -r requirements.txt

Development

Structure