pytorch中的前项计算和反向传播

前项计算1

import torch

# (3*(x+2)^2)/4

#grad_fn 保留计算的过程

x = torch.ones([2,2],requires_grad=True)

print(x)

y = x+2

print(y)

z = 3*y.pow(2)

print(z)

out = z.mean()

print(out)

#带有反向传播属性的tensor不能直接转化为numpy格式，需要先进性detach操作

print(x.detach().numpy())

print(x.numpy())

Traceback (most recent call last):
File "C:/Users/liuxinyu/Desktop/pytorch_test/day2/前向计算.py", line 17, in <module>
print(x.numpy())
RuntimeError: Can't call numpy() on Variable that requires grad. Use var.detach().numpy() instead.
tensor([[1., 1.],
[1., 1.]], requires_grad=True)
tensor([[3., 3.],
[3., 3.]], grad_fn=<AddBackward0>)
tensor([[27., 27.],
[27., 27.]], grad_fn=<MulBackward0>)
tensor(27., grad_fn=<MeanBackward0>)
[[1. 1.]
[1. 1.]]

　前向计算2

import torch

a = torch.randn(2,2)

a = ((a*3)/(a-1))

print(a.requires_grad)

a.requires_grad_(True) #就地修改

print(a.requires_grad)

b = (a*a).sum()

print(b.grad_fn)

with torch.no_grad():

    c = (a*a).sum()

    print(c.requires_grad)

False
True
<SumBackward0 object at 0x000000000249D550>
False

　　反向传播

import torch

# (3*(x+2)^2)/4

#grad_fn 保留计算的过程

x = torch.ones([2,2],requires_grad=True)

print(x)

y = x+2

print(y)

z = 3*y.pow(2)

print(z)

out = z.mean()

print(out)

out.backward()

print(x.grad)

tensor([[1., 1.],

        [1., 1.]], requires_grad=True)

tensor([[3., 3.],

        [3., 3.]], grad_fn=<AddBackward0>)

tensor([[27., 27.],

        [27., 27.]], grad_fn=<MulBackward0>)

tensor(27., grad_fn=<MeanBackward0>)

tensor([[4.5000, 4.5000],

        [4.5000, 4.5000]])