【deep learning with pytorch】运用pytorch训练模型

【deep learning with pytorch】运用pytorch训练模型

之前的博客已经介绍了如何定义与加载数据集、定义并使用网络。这篇博客将以一个完整的训练分类器为例，介绍如何训练一个神经网络。
总结写一个架构应该有的流程是：
1、构建神经网络架构
2、搭建数据集架构
3、定义loss,优化器
4、将数据输入，训练
5、用测试数据集做测试

1、构建神经网络

import torch.nn as nn
import torch.nn.functional as F

class Net(nn.Module):
	def __init__(self):
		super(Net, self).__init__()
		self.conv1 = nn.Conv2d(3,6,5)
		self.pool = nn.MaxPool2d(2,2)
		self.conv2 = nn.Conv2d(6,16,5)
		self.fc1 = nn.Linear(16*5*5,128)
		self.fc2 = nn.Linear(120,84)
		self.fc3 = nn.Linear(84,10)
	def forward(self,x):
		x = self.pool(F.relu(self.conv1(x)))
		x = self.pool(F.relu(self.conv2(x)))
		x = x.view(-1,16*5*5)
		x = F.relu(self.fc1(x))
		x = F.relu(self.fc2(x))
		x = self.fc3(x)
		return x
net = Net()

2、构建数据集

这里数据集是直接采用torchvision中自带的。

import torch
import torchvision
import trochvision.transforms as transforms
transform = transforms.Compose([transforms.ToTensor(),transforms.Normalize((0.5,0.5,0.5),(0.5,0.5,0.5))])

trainset = torchvision.datasets.CIFAR10(root = './data',train = True,download = True, transform = transform)
trainloader = torch.utils.Dataloder(trainset,batch_size = 4,shuffle = True, num_workers = 2)

testset = torchvision.datasets.CIFAR10(root = './data',train = False,download = True, transform = transform)
testloder = torch.utils.Dataloder(testset,batch_size =4,shuffle = False, num_workers = 2)

class =  ('plane', 'car', 'bird', 'cat',
           'deer', 'dog', 'frog', 'horse', 'ship', 'truck')
 

定义Loss,优化器

import torch.optim as optim

Loss = nn.CrossEntropyLoss()
optimizer = optim.SGD(net.parameters(),lr = 0.001,momentum = 0.9)

训练网络

for epoch in range(2):#迭代整个数据集2次
	running_loss = 0.0
	for i ,data in enumerate（trainloader,0):#这里data是形式是[inputs,labels]
		inputs,labels = data
		optimizer.zero_grad()
		outputs = net(inputs)
		loss = Loss(output,labels)
		loss.backward()
		optimizer.step()
		running_loss +=loss.item()
		if i % 2000 == 1999:#打印数据
			print('[%d, %5d] loss:%.3f'%(epoch+1,i+1,running_loss/2000))
			running_loss =0.0
	

print('Finish')
#存储训练结果：
path = './cifar_net/pth'
torch.save(net.state_dict(),path)

关于存储模型推荐两种存储方式：
1、存储：torch.save(the_model.state_dict(),path)
   加载：the_model = model_class(*args,**kwargs)
   the_model.load_state_dict(torch.load(path))
2、存储：torch.save(the_model.state_dict(),path)
     加载：the_model= torch.load(path)

测试训练结果：

net = Net()
#加载模型
net.load_state_dict(torch.load(path))
with torch.no_grad():
	for data in testloader:
		images, labels = data
		outputs = net(images)
		_,predicted = torch.max(outputs.data,1)
		total +=labels.size(0)
		correct +=(predicted ==labels).sum().item()
print('Accuracy of the network on the 10000 test images: %d %%' % (
    100 * correct / total))

以上就完成了一个网络的训练与测试过程

在GPU上训练

一般情况下，网络的训练，cpu是没有办法处理的，故而需要在gpu上训练。类似于将数据转移到gpu上，可以将整个网络转移到gpu上。

device = torch.device("cuda:0" if torch.cuda.is_available() else "cpu")#检测是否在cuda上
net.to(device)
#同样需要将数据转移到cuda 上面
inputs,labels = data[0].to(device), data[1].to(device)