pytorch初步

[chrispher]

http://www.datakit.cn/blog/2017/03/03/pytorch_01_basic.html

[Jason-py]

请问，为什么运行第3部分的代码

简单的case

x = Variable(torch.ones(1), requires_grad = True)
y = x ** 2 + 3
y.backward(retain_variables=True)
print(x.grad) # 梯度值是 2*x = 2

复杂一点,

target = torch.FloatTensor([10])
y.backward(target, retain_variables=True)
print(x.grad) # 梯度值是 2*x = 20, 因为retain的设置为true, grad会加上原来的梯度值2, 结果是22

会报错
TypeError: backward() got an unexpected keyword argument 'retain_variables'