构建一个卷积网络一般需要一下几步：
1、导入相应的工具包
2、得到相应的数据，对数据进行分割，分为训练集和测试集，构建batch数据集，对数据进行批量处理
3、构建卷积网络
4、写出评估函数
5、训练网络模型
import torch
import torch.nn as nn
import torch.optim as optim
import torch.nn.functional as F
import torchvision.datasets as datasets
import torchvision.transforms as transforms
import matplotlib
matplotlib.use('TkAgg')
import matplotlib.pyplot as ply
import numpy as np
# 1、首先读取数据，定义相应的超参数
# 定义超参数
input_size = 28 # 定义图像的大小为28*28
num_classes = 10 # 标签的种类数
num_epochs = 3 # 训练的总循环周期
batch_size = 64 # 一簇撮（批次）的数量，64张图片
# 训练集
train_dataset = datasets.MNIST(root='./data',train= True, transform=transforms.ToTensor(),download=True)
# 测试集
test_dataset = datasets.MNIST(root='./data',train=False,transform=transforms.ToTensor())
# 构建batch数据
train_loader = torch.utils.data.DataLoader(dataset = train_dataset,batch_size = batch_size,shuffle = True)
test_loader = torch.utils.data.DataLoader(dataset = test_dataset,batch_size = batch_size,shuffle = True)
# 卷积网络模块构造
# 一般卷积层，relu层，池化层可以写成一个类
# 注意卷积的结果还是一个特征图，需要将特征图转换成向量才能做分类或者回归任务
class CNN(nn.Module):
def __init__(self):
super(CNN, self).__init__()
# conv1代表的是第一个卷积模块
self.conv1 = nn.Sequential( # 输入大小（1，28，28）
nn.Conv2d(
in_channels=1, # 灰度图
out_channels=16, # 要得到几多少个特征图
kernel_size=5, # 卷积核大小
stride=1, # 步长
padding=2, # 如何希望卷积后大小与原来一样，需要设置padding = (kernel_size-1)/2 if stride = 1
),
nn.ReLU(), # 激活曾relu层
nn.MaxPool2d(kernel_size=2), # 进行池化操作（2*2区域），输出结果（16，14，14）
)
self.conv2 = nn.Sequential( # 下一个卷积模块的输入（16，14，14）
nn.Conv2d(16,32,5,1,2), # 输出（32，14，14）
nn.ReLU(),
nn.Conv2d(32, 32, 5, 1, 2), # 输出（32，14，14）
nn.ReLU(),
nn.MaxPool2d(2) # 输出（32，7，7）
)
self.conv3 = nn.Sequential( # 下一个套餐的输入 (16, 14, 14)
nn.Conv2d(32, 64, 5, 1, 2), # 输出 (32, 14, 14)
nn.ReLU(), # 输出 (32, 7, 7)
)
self.out = nn.Linear(64*7*7,10) # 全连接层得到的结果
def forward(self, x):
x = self.conv1(x)
x = self.conv2(x)
x = self.conv3(x)
x = x.view(x.size(0), -1) # flatten操作，结果为：(batch_size, 32 * 7 * 7)
output = self.out(x)
return output
# 评估函数
def accuracy(predictions, labels):
pred = torch.max(predictions,1)[1]
rights = pred.eq(labels.data.view_as(pred)).sum()
return rights,len(labels)
# 训练网络模型
# 实例化
net = CNN()
# 损失函数
criterion = nn.CrossEntropyLoss()
# 优化器
optimizer = optim.Adam(net.parameters(), lr=0.001) # 定义优化器，普通的随机梯度下降算法
# 开始循环训练
for epoch in range(num_epochs):
train_rights = []
for batch_idx, (data, target) in enumerate(train_loader): # 针对容器中的每一个批进行循环
net.train()
output = net(data)
loss = criterion(output,target)
optimizer.zero_grad()
loss.backward()
optimizer.step()
right = accuracy(output,target)
train_rights.append(right)
if batch_idx % 100 ==0 :
net.eval()
val_rights = []
for (data,target) in test_loader:
output = net(data)
right = accuracy(output,target)
val_rights.append(right)
# 计算准确率
train_r = (sum([tup[0] for tup in train_rights]),sum([tup[1] for tup in train_rights]))
val_r = (sum([tup[0] for tup in val_rights]), sum([tup[1] for tup in val_rights]))
print('当前epoch: {} [{}/{} ({:.0f}%)]\t损失: {:.6f}\t训练集准确率: {:.2f}%\t测试集准确率: {:.2f}%'.format(
epoch,batch_idx*batch_size,len(train_loader.dataset),
100.*batch_idx/len(train_loader),
loss.data,
100.*train_r[0].numpy()/train_r[1],
100.*val_r[0].numpy() / val_r[1]))
训练结果：
PyDev console: starting.
Python 3.11.5 | packaged by Anaconda, Inc. | (main, Sep 11 2023, 13:26:23) [MSC v.1916 64 bit (AMD64)] on win32
runfile('E:/pytarch_pro/test/torch/构建卷积神经网络—简单型.py', wdir='E:/pytarch_pro/test/torch')
当前epoch: 0 [0/60000 (0%)] 损失: 2.302288 训练集准确率: 10.94% 测试集准确率: 9.60%
当前epoch: 0 [6400/60000 (11%)] 损失: 0.130987 训练集准确率: 79.24% 测试集准确率: 92.40%
当前epoch: 0 [12800/60000 (21%)] 损失: 0.281631 训练集准确率: 86.96% 测试集准确率: 96.72%
当前epoch: 0 [19200/60000 (32%)] 损失: 0.070222 训练集准确率: 90.07% 测试集准确率: 97.18%
当前epoch: 0 [25600/60000 (43%)] 损失: 0.034417 训练集准确率: 91.72% 测试集准确率: 97.52%
当前epoch: 0 [32000/60000 (53%)] 损失: 0.084779 训练集准确率: 92.86% 测试集准确率: 97.86%
当前epoch: 0 [38400/60000 (64%)] 损失: 0.097298 训练集准确率: 93.69% 测试集准确率: 98.26%
当前epoch: 0 [44800/60000 (75%)] 损失: 0.035515 训练集准确率: 94.29% 测试集准确率: 98.06%
当前epoch: 0 [51200/60000 (85%)] 损失: 0.060263 训练集准确率: 94.75% 测试集准确率: 98.61%
当前epoch: 0 [57600/60000 (96%)] 损失: 0.061254 训练集准确率: 95.18% 测试集准确率: 98.58%
当前epoch: 1 [0/60000 (0%)] 损失: 0.032130 训练集准确率: 96.88% 测试集准确率: 98.73%
当前epoch: 1 [6400/60000 (11%)] 损失: 0.017042 训练集准确率: 98.70% 测试集准确率: 98.38%
当前epoch: 1 [12800/60000 (21%)] 损失: 0.006646 训练集准确率: 98.66% 测试集准确率: 98.59%
当前epoch: 1 [19200/60000 (32%)] 损失: 0.028878 训练集准确率: 98.68% 测试集准确率: 98.61%
当前epoch: 1 [25600/60000 (43%)] 损失: 0.020657 训练集准确率: 98.61% 测试集准确率: 98.76%
当前epoch: 1 [32000/60000 (53%)] 损失: 0.038819 训练集准确率: 98.63% 测试集准确率: 98.84%
当前epoch: 1 [38400/60000 (64%)] 损失: 0.021729 训练集准确率: 98.59% 测试集准确率: 98.88%
当前epoch: 1 [44800/60000 (75%)] 损失: 0.030117 训练集准确率: 98.58% 测试集准确率: 98.72%
当前epoch: 1 [51200/60000 (85%)] 损失: 0.014634 训练集准确率: 98.62% 测试集准确率: 98.87%
当前epoch: 1 [57600/60000 (96%)] 损失: 0.011667 训练集准确率: 98.66% 测试集准确率: 98.83%
当前epoch: 2 [0/60000 (0%)] 损失: 0.013915 训练集准确率: 100.00% 测试集准确率: 98.90%
当前epoch: 2 [6400/60000 (11%)] 损失: 0.018420 训练集准确率: 99.09% 测试集准确率: 98.99%
当前epoch: 2 [12800/60000 (21%)] 损失: 0.004838 训练集准确率: 99.13% 测试集准确率: 99.03%
当前epoch: 2 [19200/60000 (32%)] 损失: 0.004974 训练集准确率: 99.15% 测试集准确率: 98.95%
当前epoch: 2 [25600/60000 (43%)] 损失: 0.005000 训练集准确率: 99.18% 测试集准确率: 98.85%
当前epoch: 2 [32000/60000 (53%)] 损失: 0.001444 训练集准确率: 99.17% 测试集准确率: 98.82%
当前epoch: 2 [38400/60000 (64%)] 损失: 0.009423 训练集准确率: 99.14% 测试集准确率: 98.46%
当前epoch: 2 [44800/60000 (75%)] 损失: 0.053562 训练集准确率: 99.14% 测试集准确率: 98.96%
当前epoch: 2 [51200/60000 (85%)] 损失: 0.006260 训练集准确率: 99.14% 测试集准确率: 99.09%
当前epoch: 2 [57600/60000 (96%)] 损失: 0.000336 训练集准确率: 99.13% 测试集准确率: 99.05%
被折叠的 条评论
为什么被折叠?
到【灌水乐园】发言
