AlphaGogoo

``` kl = torch.sum(prob * (input_log_sofmax - torch.log(target_sofmax)), 1) ``` 看了一个文章，感觉这一行代码应该改成下面的样子，不知道理解的对不对 [loss函数之KLDivLoss](https://blog.csdn.net/ltochange/article/details/118300003) ``` kl = torch.sum(prob * (torch.log(target_sofmax)-input_log_sofmax), 1) ```