超级会员免费看
本文详细介绍了在PyTorch中如何对`torch.tensor`进行求导，包括开启`requires_grad`属性、矩阵求导、清零导数、保留计算图以及使用`torch.autograd.backward()`、`torch.autograd.grad()`进行反向传播。还讨论了如何进行二次求导，验证了链式法则。
可以结合这篇文章，更好地了解理论知识：https://blog.csdn.net/weixin_42468475/article/details/108369795
目录
方法1 tensor.backward()
Central to all neural networks in PyTorch is the autograd package.
Tensor
torch.Tensor is the central class of the package. If you set its attribute .requires_grad as True, it starts to track all operations on it. When you finish your computation you can call .backward() and have all the gradients computed automatically. The gradient for this tensor will be accumulated into .grad attribute.
To stop a tensor from tracking history, you can call .deta
订阅专栏 解锁全文
扫一扫
356
被折叠的 条评论
为什么被折叠?
到【灌水乐园】发言
