有没有pytorch版本的呢？

initial-h / AlphaZero_Gomoku_MPI

An asynchronous/parallel method of AlphaGo Zero algorithm with Gomoku

185 stars 43 forks source link

Closed apocalypse-revelation closed 4 years ago

apocalypse-revelation commented 4 years ago

如果要转为pytorch版的，修改policy_value_net的模块就可以了吧？

initial-h commented 4 years ago

对的，不过模型怎么从tf转到pytorch要搜一下怎么做。

apocalypse-revelation commented 4 years ago

@initial-h 谢谢，另外我想问一下 train.py里的 batch_size、buffer_size、data_buffer以及play_batch_size分别代表什么意思呢？

initial-h commented 4 years ago

data_buffer是存数据的队列，buffer_size是这个队列的大小。batch_size是训网络的batch大小，play_batch_size是每下多少局训练一次网络。

apocalypse-revelation commented 4 years ago

@initial-h 非常感谢！