火炬.当最终张量中有多个值时可以使用 autograd 吗? [英] Pytorch. Can autograd be used when the final tensor has more than a single value in it?

问题描述

当最终张量中有多个值时可以使用 autograd 吗?

Can autograd be used when the final tensor has more than a single value in it?

我尝试了以下方法.

x = torch.tensor([4.0, 5.0], requires_grad=True)
y = x ** 2

print(y)

y.backward()

抛出错误

RuntimeError: grad can be implicitly created only for scalar outputs

但是以下有效.

x = torch.tensor([4.0, 5.0], requires_grad=True)
y = x ** 2
y = torch.sum(y)
print(y)

y.backward()
print(x.grad)

输出为

tensor(41., grad_fn=<SumBackward0>)
tensor([ 8., 10.])

我是否在这里遗漏了什么，或者我可以继续假设 autograd 仅在最终张量中有单个值时才有效?

Am I missing something here or can I proceed with the assumption that autograd only works when the final tensor has a single value in it?

推荐答案

参见 https://pytorch.org/tutorials/beginner/blitz/autograd_tutorial.html#gradients

y.backward() 等同于 y.backward(torch.tensor(1.0))

通常，输出是标量，因此标量作为默认选择向后传递.但是，由于您的输出是二维的，因此您应该调用y.backward(torch.tensor([1.0,1.0]))

Usually, the output is scalar and hence the scalar is passed to backward as a default choice. However, since your output is two dimensional you should call y.backward(torch.tensor([1.0,1.0]))

这将给出 x.grad 为 tensor([ 8., 10.])

这篇关于火炬.当最终张量中有多个值时可以使用 autograd 吗?的文章就介绍到这了，希望我们推荐的答案对大家有所帮助，也希望大家多多支持IT屋！