TStarBot-X: An Open-Sourced and Comprehensive Study for Efficient League Training in StarCraft II FG 其他 2021-11-30 01:38 0 阅读 文章目录 前言 Technical Details and Methods Observation, Action and Reward Neural Network Architechture Imitation Learning with Importance Sampling Diversified League Training Rule-Guided Policy Search Stabilized Policy Improvement with DAPO Results Overall Performance Human Evaluation League Evaluation 猜你喜欢