Torch learning notes (5) -- autograd

import torch
import torch.nn as nn
import numpy as np
import matplotlib.pyplot as plt

# autograd
# fn1:torch.autograd.backward() Automatically calculate the gradient 
#  Parameters ：tensors: For derivation tensor;retain_graph: Save calculation diagram ;create_graph: Create derivative calculation diagram , For higher-order derivation ;grad_tensors： Multi gradient weight 
# fn2:torch.autograd.grad() Automatically calculate the gradient 
#  Parameters ：outputs： For derivation tensor;input： Requirements grad Of tensor;create_graph: Create derivative calculation diagram , For higher-order derivation ;retain_graph: Save calculation diagram ;grad_outputs： Extract more weights 

#  Case a ：backward() application 
w = torch.tensor([1.], requires_grad=True)
x = torch.tensor([2.], requires_grad=True)
a = torch.add(w, x)
b = torch.add(w, 1)
y0 = torch.mul(a, b)
y1 = torch.add(a, b)
loss = torch.cat([y0, y1], dim=0)
grad_tensors = torch.tensor([1., 1.])  #  Set weight 
loss.backward(gradient=grad_tensors)
# y0.backward(retain_graph=True) #  Keep the calculation chart 
print(w.grad)

#  Case 2 ：grad() application 
x = torch.tensor([3.], requires_grad=True)
y = torch.pow(x, 2)
grad_1 = torch.autograd.grad(y, x, create_graph=True)  # y Yes x Find gradient ,6
print(grad_1)
grad_2 = torch.autograd.grad(grad_1[0], x)  # y Yes x Find the gradient of ,2
print(grad_2)

# autograd Precautions for use 
# 1、grad It doesn't automatically release 
# 2、requires_grad Default True
# 3、leaf Can't execute in-place

#  Case three ：grad Consequences of non release 
w = torch.tensor([1.], requires_grad=True)
x = torch.tensor([2.], requires_grad=True)
for i in range(4):
    a = torch.add(w, x)
    b = torch.add(w, 1)
    y = torch.mul(a, b)
    y.backward()
    # print(w.grad)# requires_grad Default True
    print(b.requires_grad)
    # w.grad.zero_() # grad Zero clearing ,name_ Express in-place（ In situ ） operation 
# in-place operation 
#  Routine operations will open up new addresses 
a = torch.ones((1,))
print(id(a), a)
a = a + torch.ones((1,))
print(id(a), a)
# in-place Operation will not open up new addresses 
a += torch.ones((1,))
print(id(a), a)
# leaf Nodes cannot execute in-place operation