摘要:
## 问题来源 【d2l】9.7 序列到序列学习 ![image](https://img2023.cnblogs.com/blog/3240132/202307/3240132-20230728143550788-834842676.png) ![image](https://img2023.cn 阅读全文
摘要:
**Xavier是一种很好的参数初始化方法** ## 参考1:https://pytorch.org/docs/stable/nn.init.html ![](https://img2023.cnblogs.com/blog/3240132/202307/3240132-20230728113341 阅读全文
摘要:
** nn.CrossEntropyloss自带softmax功能,d2l.cross_entrop不带softmax功能** ## Note reduction = 'none' nn.CrossEntropyloss的常见用法: ``` loss = nn.CrossEntropyloss(re 阅读全文