NEURAL NETWORKS
每个nn.Module需包含网络结构和forward(input)方法。
forward(input)方法返回输出
训练神经网络一般步骤:
- 定义神经网络、可学习参数(权值)
- 遍历输入数据集
- 前向传播
- 计算损失(输出与真实值距离)
- 反向传播梯度
- 更新网络权值, w e i g h t = w e i g h t − l e a r n i n g _ r a t e ∗ g r a d i e n t weight = weight - learning\_rate * gradient weight=weight−learning_rate∗gradient
import torch
import torch.nn as nn
import torch.nn.functional as F
定义网络
class Net(nn.Module):
def __init__(self):
super(Net, self).__init__()
# 1 input image channel, 6 output channels, 3x3 square convolution kernel
self.conv1 = nn.Conv2d(in_channels=1, out_channels=6, kernel_size=3)
self.conv2 = nn.Conv2d(in_channels=6, out_channels=16, kernel_size=3)
# an affine operation: y = Wx + b
self.fc1 = nn.Linear(in_features=16 * 6 * 6, out_features=120) # 6*6 from image dimension
self.fc2 = nn.Linear(in_features=120, out_features=84)
self.fc3 = nn.Linear(in_features=84, out_features=10)
def forward(self, x):
# Max pooling over a (2, 2) window
x = F.max_pool2d(input=F.relu(input=self.conv1(x)), kernel_size=(2, 2))
# If the size is a square you can only specify a single number
x = F.max_pool2d(input=F.relu(input=self.conv2(x)), kernel_size=2)
x = x.view(-1, self.num_flat_features(x))
x = F.relu(input=self.fc1(x))
x = F.relu(input=self.fc2(x))
x = self.fc3(x)
return x
def num_flat_features(self, x):
size = x.size()[1 :] # all dimensions except the batch dimension
num_features = 1
for s in size:
num_features *= s
return num_features
net = Net()
print(net)
Net(
(conv1): Conv2d(1, 6, kernel_size=(3, 3), stride=(1, 1))
(conv2): Conv2d(6, 16, kernel_size=(3, 3), stride=(1, 1))
(fc1): Linear(in_features=576, out_features=120, bias=True)
(fc2): Linear(in_features=120, out_features=84, bias=True)
(fc3): Linear(in_features=84, out_features=10, bias=True)
)
模型的可学习参数由net.parameters()返回
params = list(net.parameters())
print(len(params))
print(params[0].size()) # conv1's .weight
10
torch.Size([6, 1, 3, 3])
随机输入
input = torch.randn(1, 1, 32, 32)
out = net(input)
print(out)
tensor([[-0.0848, -0.0075, 0.0200, 0.0464, 0.0374, 0.0861, 0.0148, 0.0441,
0.0547, -0.0514]], grad_fn=<AddmmBackward>)
所有参数梯度缓存清零,随机梯度反向传播
net.zero_grad()
out.backward(torch.randn(1, 10))
torch.nn仅支持mini-batches,对于单个样本,用input.unsqueeze(0)增加batch维度
- torch.Tensor:一个多维数组、支持autograd运算,如backward(),并保存对张量的梯度。
- nn.Module:神经网络模块,封装参数、协助GPU运算、导出、加载等。
- nn.Parameter:参数张量,作为属性分配给模块时,自动注册为参数。
- autograd.Function:自动求导运算的正、反向定义。
损失函数
损失函数输入:(output, target),计算output与target间的距离。
output = net(input)
target = torch.randn(10) # a dummy target, for example
target - target.view(1, -1) # make it the same shape as output
criterion = nn.MSELoss()
loss = criterion(output, target)
print(loss)
tensor(0.6538, grad_fn=<MseLossBackward>)
print(loss.grad_fn) # MSELoss
print(loss.grad_fn.next_functions[0][0]) # Linear
print(loss.grad_fn.next_functions[0][0].next_functions[0][0]) # ReLU
<MseLossBackward object at 0x000001B7DE3E1630>
<AddmmBackward object at 0x000001B7DE36CE80>
<AccumulateGrad object at 0x000001B7DE36F470>
反向传播(Backprop)
loss.backward()
需要先将梯度清零,否则梯度将会累加。
net.zero_grad() # zeroes the gradient buffers of all parameters
print('conv1.bias.grad before backward')
print(net.conv1.bias.grad)
loss.backward()
print('conv1.bias.grad after backward')
print(net.conv1.bias.grad)
conv1.bias.grad before backward
tensor([0., 0., 0., 0., 0., 0.])
conv1.bias.grad after backward
tensor([-0.0038, -0.0011, -0.0077, -0.0036, -0.0100, -0.0093])
更新权值
随机梯度下降(Stochastic Gradient Descent,SGD)
weight = weight - learning_rate * gradient
learning_rate = 0.01
for f in net.parameters():
f.data.sub_(f.grad.data * learning_rate)
torch.optim
import torch.optim as optim
# create your optimizer
optimizer = optim.SGD(net.parameters(), lr=0.01)
# in your training loop
optimizer.zero_grad() # zero the gradient buffers
output = net(input)
loss = criterion(output, target)
loss.backward()
optimizer.step() # Does the update
D:\ProgramData\Anaconda3\envs\pytorch\lib\site-packages\torch\nn\modules\loss.py:443: UserWarning: Using a target size (torch.Size([10])) that is different to the input size (torch.Size([1, 10])). This will likely lead to incorrect results due to broadcasting. Please ensure they have the same size.
return F.mse_loss(input, target, reduction=self.reduction)