Skip to content
StarCraft II / PySC2 Deep Reinforcement Learning Agents (A2C)
Branch: master
Clone or download
Latest commit 368a222 Apr 7, 2018
Type Name Latest commit message Commit time
Failed to load latest commit information.
rl 'keep_dims' is deprecated. Using 'keepdims' instead. Feb 8, 2018
.gitignore updated .gitignore for mac Jan 18, 2018
LICENSE add README Jan 5, 2018 fixed default --save_dir path to also work on windows Feb 12, 2018

PySC2 Deep RL Agents

This repository implements a Advantage Actor-Critic agent baseline for the pysc2 environment as described in the DeepMind paper StarCraft II: A New Challenge for Reinforcement Learning. We use a synchronous variant of A3C (A2C) to effectively train on GPUs and otherwise stay as close as possible to the agent described in the paper.

This repository is part of a research project at the Autonomous Systems Labs , TU Darmstadt by Daniel Palenicek, Marcel Hussing, and Simon Meister.


  • A2C agent
  • FullyConv architecture
  • support all spatial screen and minimap observations as well as non-spatial player observations
  • support the full action space as described in the DeepMind paper (predicting all arguments independently)
  • support training on all mini games
  • report results for all mini games
  • LSTM architecture
  • Multi-GPU training


This project is licensed under the MIT License (refer to the LICENSE file for details).


On the mini games, we get the following results:

Map best mean score (ours) best mean score (DeepMind) episodes (ours)
MoveToBeacon 26 26 8K
CollectMineralShards 97 103 300K
FindAndDefeatZerglings 45 45 450K
DefeatRoaches 65 100 -
DefeatZerglingsAndBanelings 68 62 -
CollectMineralsAndGas - 3978 -
BuildMarines - 3 -

In the following we show plots for the score over episodes.




Note that the DeepMind mean scores are their best individual scores after 100 runs for each game, where the initial learning rate was randomly sampled for each run. We use a constant initial learning rate for a much smaller number of runs due to limited hardware. All agents use the same FullyConv agent.

With default settings (32 environments), learning MoveToBeacon well takes between 3K and 8K total episodes. This varies each run depending on random initialization and action sampling.


Hardware Requirements

  • for fast training, a GPU is recommended. We ran each experiment on a single Titan X Pascal (12GB).

Software Requirements

  • Python 3
  • pysc2 (tested with v1.2)
  • TensorFlow (tested with 1.4.0)
  • StarCraft II and mini games (see below or pysc2)

Quick Install Guide

  • pip install numpy tensorflow-gpu pysc2==1.2
  • Install StarCraft II. On Linux, use 3.16.1.
  • Download the mini games and extract them to your StarcraftII/Maps/ directory.

Train & run

  • run and train: python my_experiment --map MoveToBeacon.
  • run and evalutate without training: python my_experiment --map MoveToBeacon --eval.

You can visualize the agents during training or evaluation with the --vis flag. See for all arguments.

Summaries are written to out/summary/<experiment_name> and model checkpoints are written to out/models/<experiment_name>.


The code in rl/ is based on OpenAI baselines, with adaptions from sc2aibot. Some of the code in rl/agents/a2c/ is loosely based on sc2aibot.

You can’t perform that action at this time.