This is the "v_1.01" version, and the updates are as follows:
- Change to using the win-loss reward in winloss_baseline (the thing AlphaStar should do);
- Refine pseudo reward, scale Leven reward, right log prob ( which should be negative CrossEntropy);
- Fix some problems due to wrong original codes of AlphaStar;
- Add analysis of move camera count in the AlphaStar replay;
We release the mini-AlphaStar project, which is a mini source version of the original AlphaStar program. AlphaStar is an intelligent AI proposed by DeepMind to play StarCraft II. StarCraft II is an RTS game developed by Blizzard.
"mini" means that we make the original AlphaStar hyperparameter adjustable so that it can run on a small scale.
The readme for the Chinese version is at here.
The below table shows the corresponding packages in the project.
Packages | Content |
---|---|
alphastarmini.core.arch | the alphaStar architecture |
alphastarmini.core.sl | surpervised learning |
alphastarmini.core.rl | reinforcement learning |
alphastarmini.core.ma | multi-agent league traning |
alphastarmini.lib | lib functions |
alphastarmini.third | third party functions |
res | other useful resources |
Pytorch >= 1.5, others please see requirements.txt.
The codes are in these places:
Location | URL |
---|---|
Github | https://github.com/liuruoze/mini-AlphaStar |
Gitee | https://gitee.com/liuruoze/mini-AlphaStar |
There are still some todos (very few) that need to be filled up and improved.
If you find this repository useful, please cite our project:
@misc{mini-AlphaStar,
author = {Ruo{-}Ze Liu and Wenhai Wang and Yang Yu and Tong Lu},
title = {mini-AlphaStar},
year = {2021},
publisher = {GitHub},
journal = {GitHub repository},
howpublished = {\url{https://github.com/liuruoze/mini-AlphaStar}},
}
The technical report is now on arxiv named as An Introduction of mini-AlphaStar.
We will give two to three updates for the report, to make it more complete and clear.
If you find this report useful, please cite the report:
@misc{report_mini-AlphaStar,
title={An Introduction of mini-AlphaStar},
author={Ruo-Ze Liu and Wenhai Wang and Yanjie Shen and Zhiqi Li and Yang Yu and Tong Lu},
year={2021},
journal={CoRR},
eprint={2104.06890},
archivePrefix={arXiv},
}
We will give a paper which may be available in the future presenting the experiments and evaluations on using it.