qq_44554842的博客

笔者在学习使用pytorch的backward()时，发现了一个问题。原有代码是这样的import torchN,D_in,H,D_out = 10,1000,100,10w1=torch.randn(D_in,H,requires_grad=True) # D_in * Hw2=torch.randn(H,D_out,requires_grad=True) # H * D_out it...