神经网络-搭建小实战

最新推荐文章于 2024-08-20 17:20:20 发布

于小猿Sup

最新推荐文章于 2024-08-20 17:20:20 发布

阅读量286

点赞数

文章标签：神经网络深度学习人工智能

本文链接：https://blog.csdn.net/weixin_51002159/article/details/126810341

版权

本文详细介绍了如何使用PyTorch构建CIFAR10分类网络，包括模型搭建、保存与加载，以及利用GPU进行高效训练的技巧。从构建基本的卷积神经网络到完整的训练流程，以及验证模型的完整步骤都有详尽的演示。

摘要由CSDN通过智能技术生成

文章目录

神经网络搭建CIFAR10

# -*- coding: UTF-8 -*-  
# @Time : 2022/8/1 21:49
# @File : nn_seq.py
# @Software: PyCharm
import torch
from tensorboardX import SummaryWriter
from torch import nn
from torch.nn import Conv2d, MaxPool2d, Flatten, Linear, Sequential


class GHigher(nn.Module):
    def __init__(self):
        super(GHigher, self).__init__()
        self.conv1 = Conv2d(3, 32, 5, padding=2)
        self.maxpool1 = MaxPool2d(2)
        self.conv2 = Conv2d(32, 32, 5, padding=2)
        self.maxpool2 = MaxPool2d(2)
        self.conv3 = Conv2d(32, 64, 5, padding=2)
        self.maxpool3 = MaxPool2d(2)
        self.flatten = Flatten()
        self.linear1 = Linear(1024, 64)
        self.linear2 = Linear(64, 10)
        # Sequential
        self.model = Sequential(
            Conv2d(3, 32, 5, padding=2),
            MaxPool2d(2),
            Conv2d(32, 32, 5, padding=2),
            MaxPool2d(2),
            Conv2d(32, 64, 5, padding=2),
            MaxPool2d(2),
            Flatten(),
            Linear(1024, 64),
            Linear(64, 10)
        )
    def forward(self, x):
        # x = self.conv1(x)
        # x = self.maxpool1(x)
        # x = self.conv2(x)
        # x = self.maxpool2(x)
        # x = self.conv3(x)
        # x = self.maxpool3(x)
        # x = self.flatten(x)
        # x = self.linear1(x)
        # x = self.linear2(x)
        x = self.model(x)
        return x


net_gh = GHigher()
print(net_gh)
# 检验网络是否正确
test = torch.ones((64, 3, 32, 32))
result = net_gh(test)
print(result.shape)
writer = SummaryWriter("logs_seq")
writer.add_graph(net_gh, test)
writer.close()

网络模型的保存

import torch
import torchvision
# train_data = torchvision.datasets.ImageNet("./Image_Net", split='train', transform=torchvision.transforms.ToTensor())
vgg16 = torchvision.models.vgg16(pretrained=False)
# 保存方式1
torch.save(vgg16, "vgg16_method1.pth")
# 保存方式2
torch.save(vgg16.state_dict(), "vgg16_method2.pth")

一般保存方式2比保存方式1所占的内存小，所以我们通常使用方式2保存

网络模型的加载

import torch
import torchvision

vgg16 = torchvision.models.vgg16(pretrained=False)
# 加载方式1
model1 = torch.load("vgg16_method1.pth")
print(model1)
# 加载方式2
model2 = torch.load("vgg16_method2.pth")
print(model2)
vgg16.load_state_dict(model2)
print(vgg16)

model1返回

VGG(
  (features): Sequential(
    (0): Conv2d(3, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
    (1): ReLU(inplace=True)
    (2): Conv2d(64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
    (3): ReLU(inplace=True)
    (4): MaxPool2d(kernel_size=2, stride=2, padding=0, dilation=1, ceil_mode=False)
    (5): Conv2d(64, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
    (6): ReLU(inplace=True)
    (7): Conv2d(128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
    (8): ReLU(inplace=True)
    (9): MaxPool2d(kernel_size=2, stride=2, padding=0, dilation=1, ceil_mode=False)
    (10): Conv2d(128, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
    (11): ReLU(inplace=True)
    (12): Conv2d(256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
    (13): ReLU(inplace=True)
    (14): Conv2d(256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
    (15): ReLU(inplace=True)
    (16): MaxPool2d(kernel_size=2, stride=2, padding=0, dilation=1, ceil_mode=False)
    (17): Conv2d(256, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
    (18): ReLU(inplace=True)
    (19): Conv2d(512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
    (20): ReLU(inplace=True)
    (21): Conv2d(512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
    (22): ReLU(inplace=True)
    (23): MaxPool2d(kernel_size=2, stride=2, padding=0, dilation=1, ceil_mode=False)
    (24): Conv2d(512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
    (25): ReLU(inplace=True)
    (26): Conv2d(512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
    (27): ReLU(inplace=True)
    (28): Conv2d(512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1))
    (29): ReLU(inplace=True)
    (30): MaxPool2d(kernel_size=2, stride=2, padding=0, dilation=1, ceil_mode=False)
  )
  (avgpool): AdaptiveAvgPool2d(output_size=(7, 7))
  (classifier): Sequential(
    (0): Linear(in_features=25088, out_features=4096, bias=True)
    (1): ReLU(inplace=True)
    (2): Dropout(p=0.5, inplace=False)
    (3): Linear(in_features=4096, out_features=4096, bias=True)
    (4): ReLU(inplace=True)
    (5): Dropout(p=0.5, inplace=False)
    (6): Linear(in_features=4096, out_features=1000, bias=True)
  )
)

model2返回为字典类型

OrderedDict([('features.0.weight', tensor([[[[-4.1368e-02,  1.2598e-02, -3.7834e-03],
          [ 2.3529e-02,  4.8841e-02, -3.7602e-02],
          [ 3.2726e-02,  5.5254e-02,  6.1646e-02]],

         [[ 3.3375e-02, -9.8072e-02,  3.3444e-02],
          [-3.6574e-02, -5.2286e-02, -8.1664e-03],
          [-3.8757e-02,  3.0309e-02,  9.3537e-03]],

         [[ 6.7703e-03, -2.8736e-02, -1.2290e-02],
          [ 3.2700e-02, -4.2516e-02, -2.9173e-02],
          [-1.2818e-01, -1.7044e-02,  3.0051e-02]]],


        [[[ 5.4985e-02, -6.2916e-03,  6.8897e-02],
        .....]))])

通常我们保存使用方式2，加载使用vgg16.load_state_dict(model2)这种方式，

print(vgg16)同model1

注意！！！

class conv(nn.Module):
    def __init__(self):
        super(conv, self).__init__()
        self.conv1 = Conv2d(3, 32, kernel_size=3)

    def forward(self, x):
        return self.conv1(x)


test = conv()
torch.save(test, "conv_method.pth")

当我们自己定义神经网络，保存之后，再在另一个文件加载是会报错，找不到自己定义的类

AttributeError: Can't get attribute 'conv' on <module '__main__' from 'C:/Users/25354/Desktop/Python/pytorch/model_load.py'>

解决方法

import torch

from model_save import *
test_model = torch.load("conv_method.pth")
print(test_model)
"""
conv(
  (conv1): Conv2d(3, 32, kernel_size=(3, 3), stride=(1, 1))
)
"""

完整的模型训练套路(以CIFAR10数据集)

import torch
import torchvision
from torch import nn
from tensorboardX import SummaryWriter
from torch.nn import Conv2d, MaxPool2d, Flatten, Linear, Sequential
from torch.utils.data import DataLoader

train_data = torchvision.datasets.CIFAR10("./datasets", train=True, transform=torchvision.transforms.ToTensor())
test_data = torchvision.datasets.CIFAR10("./datasets", train=False, transform=torchvision.transforms.ToTensor())
# length长度
train_data_size = len(train_data)
test_data_size = len(test_data)
print("训练数据集的长度为:{}".format(train_data_size))
print("测试数据集的长度为:{}".format(test_data_size))
# 利用DataLoader来加载数据集
train_dataloader = DataLoader(train_data, batch_size=64)
test_dataloader = DataLoader(test_data, batch_size=64)


# 搭建神经网络
class cifiar(nn.Module):
    def __init__(self):
        super(cifiar, self).__init__()
        # Sequential
        self.model = Sequential(
            Conv2d(3, 32, 5, padding=2),
            MaxPool2d(2),
            Conv2d(32, 32, 5, padding=2),
            MaxPool2d(2),
            Conv2d(32, 64, 5, padding=2),
            MaxPool2d(2),
            Flatten(),
            Linear(1024, 64),
            Linear(64, 10)
        )

    def forward(self, x):
        x = self.model(x)
        return x


# 创建网络模型
nn_cifai = cifiar()

# 损失函数
loss_fn = nn.CrossEntropyLoss()

# 优化器
learning_rate =1e-2
optimizer = torch.optim.SGD(nn_cifai.parameters(), lr=learning_rate)

# 设置训练网络的一些参数
# 记录训练的次数
total_train_step = 0
total_test_step = 0
# 训练的论数
epoch = 10

# 添加tensorboard
writer = SummaryWriter("log_train")

for i in range(epoch):
    print("---------第{}轮训练开始---------".format(i+1))
    # 训练步骤开始
    for data in train_dataloader:
        img, target = data
        output = nn_cifai(img)
        loss = loss_fn(output, target)
        optimizer.zero_grad()
        loss.backward()
        optimizer.step()

        total_train_step = total_train_step + 1
        if total_train_step % 100 == 0:
            print("训练次数:{}, Loss:{}".format(total_train_step, loss.item()))
            writer.add_scalar("train_loss", loss.item(), total_train_step)

    # 测试步骤开始
    total_test_loss = 0
    total_accuracy = 0
    with torch.no_grad():
        for data in test_dataloader:
            img, target = data
            output = nn_cifai(img)
            loss = loss_fn(output, target)
            total_test_loss = total_test_loss + loss.item()
            accuray = (output.argmax(1)==target).sum()
            total_accuracy = total_accuracy + accuray
    print("整体测试集上的Loss:{}".format(total_test_loss))
    print("整体测试集的正确率:{}".format(total_accuracy/test_data_size))
    writer.add_scalar("test_loss", total_test_loss, total_train_step)
    total_test_step += 1

    # torch.save(nn_cifai, "cifar_{}.pth".format(i))
    print("模型已保存")

writer.close()

利用GPU训练

电脑存在GPU

只需修改以下几项即可

网络模型
损失函数
数据(输入，标注)

两种方法

①后缀加.cuda()

# 网络模型
if torch.cuda.is_available():
    nn_cifai = nn_cifai.cuda()
# 损失函数 
if torch.cuda.is_available():
    loss_fn = loss_fn.cuda()
# 数据(输入，标注)  
if torch.cuda.is_available():
    img = img.cuda()
    target = target.cuda()

②后缀加.to(device)

#定义训练的设备,如果gpu存在这样gpu否则用cpu
device = torch.device("cuda" if torch.cuda.is_available() else "cpu")

nn_cifai = nn_cifai.to(device)
loss_fn = loss_fn.to(device)
img = img.to(device)
target = target.to(device)

如果电脑上有多个GPU，"cuda"可以写为"cuda:0"或"cuda:1"…

在任务管理器->性能可以查看

电脑不存在GPU

在线GPU

通过测试发现

GPU运行时间为169.24s
CPU运行时间为788.99s

因此可以得出采用GPU训练可以大大提高运行时间

完整的模型验证套路

利用已经训练好的模型，然后它提供输入

# -*- coding: UTF-8 -*-  
# @Time : 2022/8/2 21:28
# @File : test.py.py
# @Software: PyCharm
import torch
import torchvision
from PIL import Image
from torch import nn
from torch.nn import Conv2d, MaxPool2d, Flatten, Linear, Sequential

train_label = ['airplane', 'automobile', 'bird', 'cat', 'deer', 'dog', 'frog', 'horse', 'ship', 'truck']
img_path = "airplane1.jpg"
img1_path = "truck.png"
image = Image.open(img1_path)
image = image.convert("RGB")
print(image)

transform = torchvision.transforms.Compose([torchvision.transforms.Resize((32, 32)),
                                            torchvision.transforms.ToTensor()])
image = transform(image)
print(image.shape)

# 搭建神经网络
class cifiar(nn.Module):
    def __init__(self):
        super(cifiar, self).__init__()
        # Sequential
        self.model = Sequential(
            Conv2d(3, 32, 5, padding=2),
            MaxPool2d(2),
            Conv2d(32, 32, 5, padding=2),
            MaxPool2d(2),
            Conv2d(32, 64, 5, padding=2),
            MaxPool2d(2),
            Flatten(),
            Linear(1024, 64),
            Linear(64, 10)
        )

    def forward(self, x):
        x = self.model(x)
        return x


model = torch.load("cifar_30.pth", map_location=torch.device("cpu"))
print(model)
image = torch.reshape(image, (1, 3, 32, 32))
model.eval()
with torch.no_grad():
    output = model(image)
print(output)
index = output.argmax(1).item()
print(index, train_label[index])
"""
<PIL.Image.Image image mode=RGB size=416x380 at 0x24E8447A820>
torch.Size([3, 32, 32])
cifiar(
  (model): Sequential(
    (0): Conv2d(3, 32, kernel_size=(5, 5), stride=(1, 1), padding=(2, 2))
    (1): MaxPool2d(kernel_size=2, stride=2, padding=0, dilation=1, ceil_mode=False)
    (2): Conv2d(32, 32, kernel_size=(5, 5), stride=(1, 1), padding=(2, 2))
    (3): MaxPool2d(kernel_size=2, stride=2, padding=0, dilation=1, ceil_mode=False)
    (4): Conv2d(32, 64, kernel_size=(5, 5), stride=(1, 1), padding=(2, 2))
    (5): MaxPool2d(kernel_size=2, stride=2, padding=0, dilation=1, ceil_mode=False)
    (6): Flatten(start_dim=1, end_dim=-1)
    (7): Linear(in_features=1024, out_features=64, bias=True)
    (8): Linear(in_features=64, out_features=10, bias=True)
  )
)
tensor([[-0.7932,  1.5045, -0.6240,  1.7350, -5.7618,  0.1339, -0.3017, -2.6942,
          0.0614,  5.5677]])
9 truck

Process finished with exit code 0

"""