DouZero: Mastering DouDizhu with Self-Play Deep Reinforcement Learning
Paper • 2106.06135 • Published
DouZero Pre-trained Baselines (Model Weights Mirror)
This repository hosts the pre-trained baseline model weights for DouZero, a reinforcement learning framework for DouDizhu (斗地主).
- Original Code & Project: GitHub - kwai/DouZero
- Original Authors: Kuaishou (Kwai) AI Platform & Texas A&M University
- Paper: DouZero: Mastering DouDizhu with Self-Play Deep Reinforcement Learning (ICML 2021)
Why This Repository Exists
The official DouZero model weights are originally hosted on Baidu Netdisk. Due to download speed limits and login requirements on Baidu Netdisk, it is highly inconvenient to programmatically download and deploy these models in cloud environments, remote servers, or platforms like Google Colab.
This repository provides a stable, direct-download, and high-speed mirror of the baselines/ directory, allowing users to easily download and integrate the pre-trained weights into their pipelines.