![Now that pytorch can autograd, why we still need to define backward when defining customer function - autograd - PyTorch Forums Now that pytorch can autograd, why we still need to define backward when defining customer function - autograd - PyTorch Forums](https://discuss.pytorch.org/uploads/default/original/2X/1/1d2712da0016ce57bf913c7643afef7595f6a430.png)
Now that pytorch can autograd, why we still need to define backward when defining customer function - autograd - PyTorch Forums
![pytorch的基本组成结构:torchvision.datasets和models 、 torch.nn 、torch.autograd. Variable 、 torch.optim_torchvision是torch.nn吗_AIchiNiurou的博客-CSDN博客 pytorch的基本组成结构:torchvision.datasets和models 、 torch.nn 、torch.autograd. Variable 、 torch.optim_torchvision是torch.nn吗_AIchiNiurou的博客-CSDN博客](https://img-blog.csdnimg.cn/20200722213813236.png?x-oss-process=image/watermark,type_ZmFuZ3poZW5naGVpdGk,shadow_10,text_aHR0cHM6Ly9ibG9nLmNzZG4ubmV0L3dlaXhpbl80NDUyMzA2Mg==,size_16,color_FFFFFF,t_70)
pytorch的基本组成结构:torchvision.datasets和models 、 torch.nn 、torch.autograd. Variable 、 torch.optim_torchvision是torch.nn吗_AIchiNiurou的博客-CSDN博客
![Autograd not registering tensors : does not require grad and does not have a grad_fn - autograd - PyTorch Forums Autograd not registering tensors : does not require grad and does not have a grad_fn - autograd - PyTorch Forums](https://discuss.pytorch.org/uploads/default/original/3X/5/a/5a7ff69402518b7dcc9b05c93681ee1f5b6e6c3d.png)