深度学习 Day16——P5运动鞋识别

• 本文为365天深度学习训练营 中的学习记录博客
• 原作者：K同学啊 | 接辅导、项目定制

---

前言

本文将采用pytorch框架创建CNN网络，实现运动鞋识别。讲述实现代码与执行结果，并浅谈涉及知识点。
关键字：torch动态学习率

1 我的环境

• 电脑系统：Windows 11
• 语言环境：python 3.8.6
• 编译器：pycharm2020.2.3
• 深度学习环境：
  torch == 1.9.1+cu111
  torchvision == 0.10.1+cu111
• 显卡：NVIDIA GeForce RTX 4070

2 代码实现与执行结果

2.1 前期准备

2.1.1 引入库

import torch
import torch.nn as nn
from torchvision import transforms, datasets
import time
from pathlib import Path
from PIL import Image
from torchinfo import summary
import torch.nn.functional as F
import matplotlib.pyplot as plt

plt.rcParams['font.sans-serif'] = ['SimHei']  # 用来正常显示中文标签
plt.rcParams['axes.unicode_minus'] = False  # 用来正常显示负号
plt.rcParams['figure.dpi'] = 100  # 分辨率
import warnings

warnings.filterwarnings('ignore')  # 忽略一些warning内容，无需打印

2.1.2 设置GPU（如果设备上支持GPU就使用GPU,否则使用CPU）

"""前期准备-设置GPU"""
# 如果设备上支持GPU就使用GPU,否则使用CPU
 device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
 print("Using {} device".format(device))

输出

Using cuda device

2.1.3 导入数据

'''前期工作-导入数据'''
data_dir = r"D:\DeepLearning\data\monkeypox_recognition"
data_dir = Path(data_dir)

data_paths = list(data_dir.glob('./train/*'))
    classeNames = [str(path).split("\\")[-1] for path in data_paths]
    print(classeNames)

输出

['adidas', 'nike']

2.1.4 可视化数据

'''前期工作-可视化数据'''
subfolder = Path(data_dir)/"train"/"nike"
image_files = list(p.resolve() for p in subfolder.glob('*') if p.suffix in [".jpg", ".png", ".jpeg"])
plt.figure(figsize=(10, 6))
for i in range(len(image_files[:12])):
    image_file = image_files[i]
    ax = plt.subplot(3, 4, i + 1)
    img = Image.open(str(image_file))
    plt.imshow(img)
    plt.axis("off")
# 显示图片
plt.tight_layout()
plt.show()

2.1.4 图像数据变换

'''前期工作-图像数据变换'''
    # 关于transforms.Compose的更多介绍可以参考：https://blog.csdn.net/qq_38251616/article/details/124878863
    train_transforms = transforms.Compose([
        transforms.Resize([224, 224]),  # 将输入图片resize成统一尺寸
        # transforms.RandomHorizontalFlip(), # 随机水平翻转
        transforms.ToTensor(),  # 将PIL Image或numpy.ndarray转换为tensor，并归一化到[0,1]之间
        transforms.Normalize(  # 标准化处理-->转换为标准正太分布（高斯分布），使模型更容易收敛
            mean=[0.485, 0.456, 0.406],
            std=[0.229, 0.224, 0.225])  # 其中 mean=[0.485,0.456,0.406]与std=[0.229,0.224,0.225] 从数据集中随机抽样计算得到的。
    ])

    test_transform = transforms.Compose([
        transforms.Resize([224, 224]),  # 将输入图片resize成统一尺寸
        transforms.ToTensor(),  # 将PIL Image或numpy.ndarray转换为tensor，并归一化到[0,1]之间
        transforms.Normalize(  # 标准化处理-->转换为标准正太分布（高斯分布），使模型更容易收敛
            mean=[0.485, 0.456, 0.406],
            std=[0.229, 0.224, 0.225])  # 其中 mean=[0.485,0.456,0.406]与std=[0.229,0.224,0.225] 从数据集中随机抽样计算得到的。
    ])

    train_dataset = datasets.ImageFolder(Path(data_dir)/"train", transform=train_transforms)
    test_dataset = datasets.ImageFolder(Path(data_dir)/"test", transform=train_transforms)
    print(train_dataset.class_to_idx)

输出

{'adidas': 0, 'nike': 1}

2.1.4 加载数据

'''前期工作-加载数据'''
batch_size = 32

train_dl = torch.utils.data.DataLoader(train_dataset,
                                       batch_size=batch_size,
                                       shuffle=True,
                                       num_workers=1)
test_dl = torch.utils.data.DataLoader(test_dataset,
                                      batch_size=batch_size,
                                      shuffle=True,
                                      num_workers=1)

2.1.4 查看数据

'''前期工作-查看数据'''
for X, y in test_dl:
    print("Shape of X [N, C, H, W]: ", X.shape)
    print("Shape of y: ", y.shape, y.dtype)
    break

输出

Shape of X [N, C, H, W]:  torch.Size([32, 3, 224, 224])
Shape of y:  torch.Size([32]) torch.int64

2.2 构建CNN网络模型

"""构建CNN网络"""
class Network_bn(nn.Module):
    def __init__(self):
        super(Network_bn, self).__init__()
        """
        nn.Conv2d()函数：
        第一个参数（in_channels）是输入的channel数量
        第二个参数（out_channels）是输出的channel数量
        第三个参数（kernel_size）是卷积核大小
        第四个参数（stride）是步长，默认为1
        第五个参数（padding）是填充大小，默认为0
        """
        self.conv1 = nn.Conv2d(in_channels=3, out_channels=12, kernel_size=5, stride=1, padding=0)
        self.bn1 = nn.BatchNorm2d(12)
        self.conv2 = nn.Conv2d(in_channels=12, out_channels=12, kernel_size=5, stride=1, padding=0)
        self.bn2 = nn.BatchNorm2d(12)
        self.pool = nn.MaxPool2d(2, 2)
        self.conv4 = nn.Conv2d(in_channels=12, out_channels=24, kernel_size=5, stride=1, padding=0)
        self.bn4 = nn.BatchNorm2d(24)
        self.conv5 = nn.Conv2d(in_channels=24, out_channels=24, kernel_size=5, stride=1, padding=0)
        self.bn5 = nn.BatchNorm2d(24)
        self.fc1 = nn.Linear(24 * 50 * 50, len(classeNames))

    def forward(self, x):
        x = F.relu(self.bn1(self.conv1(x)))
        x = F.relu(self.bn2(self.conv2(x)))
        x = self.pool(x)
        x = F.relu(self.bn4(self.conv4(x)))
        x = F.relu(self.bn5(self.conv5(x)))
        x = self.pool(x)
        x = x.view(-1, 24 * 50 * 50)
        x = self.fc1(x)

        return x

        
model = Network_bn().to(device)
print(model)   

输出

Network_bn(
  (conv1): Conv2d(3, 12, kernel_size=(5, 5), stride=(1, 1))
  (bn1): BatchNorm2d(12, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
  (conv2): Conv2d(12, 12, kernel_size=(5, 5), stride=(1, 1))
  (bn2): BatchNorm2d(12, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
  (pool): MaxPool2d(kernel_size=2, stride=2, padding=0, dilation=1, ceil_mode=False)
  (conv4): Conv2d(12, 24, kernel_size=(5, 5), stride=(1, 1))
  (bn4): BatchNorm2d(24, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
  (conv5): Conv2d(24, 24, kernel_size=(5, 5), stride=(1, 1))
  (bn5): BatchNorm2d(24, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
  (fc1): Linear(in_features=60000, out_features=2, bias=True)
)

2.3 训练模型

2.3.1 设置超参数

"""训练模型--设置超参数"""
loss_fn = nn.CrossEntropyLoss()  # 创建损失函数，计算实际输出和真实相差多少，交叉熵损失函数，事实上，它就是做图片分类任务时常用的损失函数
learn_rate = 1e-4  # 学习率
optimizer1 = torch.optim.SGD(model.parameters(), lr=learn_rate)# 作用是定义优化器，用来训练时候优化模型参数；其中，SGD表示随机梯度下降，用于控制实际输出y与真实y之间的相差有多大
optimizer2 = torch.optim.Adam(model.parameters(), lr=learn_rate)
#测试集acc 84.2%
lr_opt = optimizer1
model_opt = optimizer1

2.3.2 编写训练函数

"""训练模型--编写训练函数"""
# 训练循环
def train(dataloader, model, loss_fn, optimizer):
    size = len(dataloader.dataset)  # 训练集的大小，一共60000张图片
    num_batches = len(dataloader)  # 批次数目，1875（60000/32）

    train_loss, train_acc = 0, 0  # 初始化训练损失和正确率

    for X, y in dataloader:  # 加载数据加载器，得到里面的 X（图片数据）和 y（真实标签）
        X, y = X.to(device), y.to(device) # 用于将数据存到显卡

        # 计算预测误差
        pred = model(X)  # 网络输出
        loss = loss_fn(pred, y)  # 计算网络输出和真实值之间的差距，targets为真实值，计算二者差值即为损失

        # 反向传播
        optimizer.zero_grad()  # 清空过往梯度
        loss.backward()  # 反向传播，计算当前梯度
        optimizer.step()  # 根据梯度更新网络参数

        # 记录acc与loss
        train_acc += (pred.argmax(1) == y).type(torch.float).sum().item()
        train_loss += loss.item()

    train_acc /= size
    train_loss /= num_batches

    return train_acc, train_loss

2.3.3 编写测试函数

"""训练模型--编写测试函数"""
# 测试函数和训练函数大致相同，但是由于不进行梯度下降对网络权重进行更新，所以不需要传入优化器
def test(dataloader, model, loss_fn):
    size = len(dataloader.dataset)  # 测试集的大小，一共10000张图片
    num_batches = len(dataloader)  # 批次数目，313（10000/32=312.5，向上取整）
    test_loss, test_acc = 0, 0

    # 当不进行训练时，停止梯度更新，节省计算内存消耗
    with torch.no_grad(): # 测试时模型参数不用更新，所以 no_grad，整个模型参数正向推就ok，不反向更新参数
        for imgs, target in dataloader:
            imgs, target = imgs.to(device), target.to(device)

            # 计算loss
            target_pred = model(imgs)
            loss = loss_fn(target_pred, target)

            test_loss += loss.item()
            test_acc += (target_pred.argmax(1) == target).type(torch.float).sum().item()#统计预测正确的个数

    test_acc /= size
    test_loss /= num_batches

    return test_acc, test_loss

2.3.4 正式训练

 """训练模型--正式训练"""
    epochs = 40
    train_loss = []
    train_acc = []
    test_loss = []
    test_acc = []
    best_test_acc=0
    PATH = './model.pth'  # 保存的参数文件名
    for epoch in range(epochs):
        milliseconds_t1 = int(time.time() * 1000)

        # 更新学习率（使用自定义学习率时使用）
        adjust_learning_rate(lr_opt, epoch, learn_rate)
        model.train()
        epoch_train_acc, epoch_train_loss = train(train_dl, model, loss_fn, model_opt)
        # scheduler.step() # 更新学习率（调用官方动态学习率接口时使用）

        model.eval()
        epoch_test_acc, epoch_test_loss = test(test_dl, model, loss_fn)

        train_acc.append(epoch_train_acc)
        train_loss.append(epoch_train_loss)
        test_acc.append(epoch_test_acc)
        test_loss.append(epoch_test_loss)

        # 获取当前的学习率
        lr = lr_opt.state_dict()['param_groups'][0]['lr']

        milliseconds_t2 = int(time.time() * 1000)
        template = ('Epoch:{:2d}, duration:{}ms, Train_acc:{:.1f}%, Train_loss:{:.3f}, Test_acc:{:.1f}%，Test_loss:{:.3f}, Lr:{:.2E}')
        if best_test_acc < epoch_test_acc:
            best_test_acc = epoch_test_acc
            # 模型保存
            torch.save(model.state_dict(), PATH)
            template = (
                'Epoch:{:2d}, duration:{}ms, Train_acc:{:.1f}%, Train_loss:{:.3f}, Test_acc:{:.1f}%，Test_loss:{:.3f}, Lr:{:.2E},save model.pth')
        print(
            template.format(epoch + 1, milliseconds_t2-milliseconds_t1, epoch_train_acc * 100, epoch_train_loss, epoch_test_acc * 100, epoch_test_loss, lr))
    print('Done')

输出

Epoch: 1, duration:5323ms, Train_acc:52.2%, Train_loss:0.745, Test_acc:50.0%，Test_loss:0.700, Lr:1.00E-04,save model.pth
Epoch: 2, duration:3125ms, Train_acc:62.9%, Train_loss:0.651, Test_acc:57.9%，Test_loss:0.667, Lr:1.00E-04,save model.pth
Epoch: 3, duration:3130ms, Train_acc:66.9%, Train_loss:0.614, Test_acc:67.1%，Test_loss:0.584, Lr:9.20E-05,save model.pth
Epoch: 4, duration:3136ms, Train_acc:70.5%, Train_loss:0.567, Test_acc:72.4%，Test_loss:0.568, Lr:9.20E-05,save model.pth
Epoch: 5, duration:3346ms, Train_acc:76.1%, Train_loss:0.531, Test_acc:72.4%，Test_loss:0.537, Lr:8.46E-05
Epoch: 6, duration:3087ms, Train_acc:77.5%, Train_loss:0.510, Test_acc:72.4%，Test_loss:0.531, Lr:8.46E-05
Epoch: 7, duration:3215ms, Train_acc:77.7%, Train_loss:0.492, Test_acc:78.9%，Test_loss:0.511, Lr:7.79E-05,save model.pth
Epoch: 8, duration:3520ms, Train_acc:82.3%, Train_loss:0.467, Test_acc:77.6%，Test_loss:0.504, Lr:7.79E-05
Epoch: 9, duration:3662ms, Train_acc:83.1%, Train_loss:0.442, Test_acc:81.6%，Test_loss:0.494, Lr:7.16E-05,save model.pth
Epoch:10, duration:3410ms, Train_acc:85.7%, Train_loss:0.427, Test_acc:80.3%，Test_loss:0.464, Lr:7.16E-05
Epoch:11, duration:3486ms, Train_acc:86.3%, Train_loss:0.413, Test_acc:81.6%，Test_loss:0.469, Lr:6.59E-05
Epoch:12, duration:3356ms, Train_acc:87.6%, Train_loss:0.394, Test_acc:78.9%，Test_loss:0.452, Lr:6.59E-05
Epoch:13, duration:3453ms, Train_acc:87.6%, Train_loss:0.391, Test_acc:81.6%，Test_loss:0.494, Lr:6.06E-05
Epoch:14, duration:3226ms, Train_acc:87.8%, Train_loss:0.385, Test_acc:80.3%，Test_loss:0.450, Lr:6.06E-05
Epoch:15, duration:3290ms, Train_acc:89.0%, Train_loss:0.368, Test_acc:82.9%，Test_loss:0.486, Lr:5.58E-05,save model.pth
Epoch:16, duration:3247ms, Train_acc:90.4%, Train_loss:0.359, Test_acc:81.6%，Test_loss:0.443, Lr:5.58E-05
Epoch:17, duration:3195ms, Train_acc:90.6%, Train_loss:0.358, Test_acc:81.6%，Test_loss:0.452, Lr:5.13E-05
Epoch:18, duration:3294ms, Train_acc:90.6%, Train_loss:0.342, Test_acc:82.9%，Test_loss:0.436, Lr:5.13E-05
Epoch:19, duration:3305ms, Train_acc:91.2%, Train_loss:0.338, Test_acc:81.6%，Test_loss:0.452, Lr:4.72E-05
Epoch:20, duration:3241ms, Train_acc:91.8%, Train_loss:0.332, Test_acc:81.6%，Test_loss:0.418, Lr:4.72E-05
Epoch:21, duration:3371ms, Train_acc:93.0%, Train_loss:0.320, Test_acc:81.6%，Test_loss:0.459, Lr:4.34E-05
Epoch:22, duration:3279ms, Train_acc:92.8%, Train_loss:0.317, Test_acc:81.6%，Test_loss:0.475, Lr:4.34E-05
Epoch:23, duration:3279ms, Train_acc:93.4%, Train_loss:0.310, Test_acc:82.9%，Test_loss:0.438, Lr:4.00E-05
Epoch:24, duration:3225ms, Train_acc:93.0%, Train_loss:0.313, Test_acc:81.6%，Test_loss:0.437, Lr:4.00E-05
Epoch:25, duration:3293ms, Train_acc:94.0%, Train_loss:0.304, Test_acc:81.6%，Test_loss:0.439, Lr:3.68E-05
Epoch:26, duration:3273ms, Train_acc:94.0%, Train_loss:0.297, Test_acc:81.6%，Test_loss:0.414, Lr:3.68E-05
Epoch:27, duration:3249ms, Train_acc:94.2%, Train_loss:0.296, Test_acc:80.3%，Test_loss:0.413, Lr:3.38E-05
Epoch:28, duration:3266ms, Train_acc:94.8%, Train_loss:0.288, Test_acc:84.2%，Test_loss:0.425, Lr:3.38E-05,save model.pth
Epoch:29, duration:3248ms, Train_acc:94.4%, Train_loss:0.288, Test_acc:81.6%，Test_loss:0.400, Lr:3.11E-05
Epoch:30, duration:3243ms, Train_acc:94.6%, Train_loss:0.291, Test_acc:81.6%，Test_loss:0.445, Lr:3.11E-05
Epoch:31, duration:3250ms, Train_acc:96.4%, Train_loss:0.278, Test_acc:81.6%，Test_loss:0.465, Lr:2.86E-05
Epoch:32, duration:3193ms, Train_acc:95.2%, Train_loss:0.275, Test_acc:81.6%，Test_loss:0.438, Lr:2.86E-05
Epoch:33, duration:3283ms, Train_acc:95.2%, Train_loss:0.270, Test_acc:81.6%，Test_loss:0.402, Lr:2.63E-05
Epoch:34, duration:3542ms, Train_acc:94.8%, Train_loss:0.280, Test_acc:81.6%，Test_loss:0.407, Lr:2.63E-05
Epoch:35, duration:3592ms, Train_acc:95.8%, Train_loss:0.269, Test_acc:81.6%，Test_loss:0.442, Lr:2.42E-05
Epoch:36, duration:3592ms, Train_acc:95.4%, Train_loss:0.267, Test_acc:80.3%，Test_loss:0.413, Lr:2.42E-05
Epoch:37, duration:3588ms, Train_acc:95.0%, Train_loss:0.265, Test_acc:81.6%，Test_loss:0.432, Lr:2.23E-05
Epoch:38, duration:3736ms, Train_acc:95.0%, Train_loss:0.267, Test_acc:81.6%，Test_loss:0.438, Lr:2.23E-05
Epoch:39, duration:3431ms, Train_acc:95.2%, Train_loss:0.265, Test_acc:82.9%，Test_loss:0.400, Lr:2.05E-05
Epoch:40, duration:3417ms, Train_acc:95.8%, Train_loss:0.270, Test_acc:81.6%，Test_loss:0.379, Lr:2.05E-05
Done

2.4 结果可视化

"""训练模型--结果可视化"""
epochs_range = range(epochs)

plt.figure(figsize=(12, 3))
plt.subplot(1, 2, 1)

plt.plot(epochs_range, train_acc, label='Training Accuracy')
plt.plot(epochs_range, test_acc, label='Test Accuracy')
plt.legend(loc='lower right')
plt.title('Training and Validation Accuracy')

plt.subplot(1, 2, 2)
plt.plot(epochs_range, train_loss, label='Training Loss')
plt.plot(epochs_range, test_loss, label='Test Loss')
plt.legend(loc='upper right')
plt.title('Training and Validation Loss')
plt.show()

2.4 指定图片进行预测

def predict_one_image(image_path, model, transform, classes):
    test_img = Image.open(image_path).convert('RGB')
    plt.imshow(test_img)  # 展示预测的图片
    plt.show()

    test_img = transform(test_img)
    img = test_img.to(device).unsqueeze(0)

    model.eval()
    output = model(img)

    _, pred = torch.max(output, 1)
    pred_class = classes[pred]
    print(f'预测结果是：{pred_class}')
    
"""指定图片进行预测"""
 classes = list(total_data.class_to_idx)
# 预测训练集中的某张照片
predict_one_image(image_path=str(Path(data_dir)/"test/adidas/1.jpg"),
                  model=model,
                  transform=train_transforms,
                  classes=classes)

如果使用效果最好的模型，就先加载保存好的模型，再调用预测代码

   # 将参数加载到model当中
    model.load_state_dict(torch.load(PATH, map_location=device))

输出

预测结果是：adidas

2.6 保存并加载模型

"""保存并加载模型"""
# 模型保存
PATH = './model.pth'  # 保存的参数文件名
torch.save(model.state_dict(), PATH)

# 将参数加载到model当中
model.load_state_dict(torch.load(PATH, map_location=device))

3 知识点详解

3.1 torch动态学习率

3.1.1 torch.optim.lr_scheduler.StepLR

函数原型：

torch.optim.lr_scheduler.StepLR(optimizer, step_size, gamma=0.1, last_epoch=-1)

关键参数详解：

• optimizer(Optimizer)：是之前定义好的需要优化的优化器的实例名
• step_size(int)：是学习率衰减的周期，每经过每个epoch，做一次学习率decay
• gamma(float)：学习率衰减的乘法因子。Default：0.1

用法示例：

optimizer = torch.optim.SGD(net.parameters(), lr=0.001 )
scheduler = torch.optim.lr_scheduler.StepLR(optimizer, step_size=5, gamma=0.1)

3.1.2 lr_scheduler.LambdaLR

根据自己定义的函数更新学习率。
函数原型：

torch.optim.lr_scheduler.LambdaLR(optimizer, lr_lambda, last_epoch=-1, verbose=False)

关键参数详解：

• optimizer(Optimizer)：是之前定义好的需要优化的优化器的实例名
• lr_lambda(function)：更新学习率的函数
  用法示例：

lambda1 = lambda epoch: (0.92 ** (epoch // 2) # 第二组参数的调整方法
optimizer = torch.optim.SGD(model.parameters(), lr=learn_rate)
scheduler = torch.optim.lr_scheduler.LambdaLR(optimizer, lr_lambda=lambda1) #选定调整方法

3.1.3 lr_scheduler.MultiStepLR

在特定的 epoch 中调整学习率
函数原型：

torch.optim.lr_scheduler.MultiStepLR(optimizer, milestones, gamma=0.1, last_epoch=-1, verbose=False)

关键参数详解：

• optimizer(Optimizer)：是之前定义好的需要优化的优化器的实例名
• milestones(list)：是一个关于epoch数值的list，表示在达到哪个epoch范围内开始变化，必须是升序排列
• gamma(float)：学习率衰减的乘法因子。Default：0.1
  用法示例：

optimizer = torch.optim.SGD(net.parameters(), lr=0.001 )
scheduler = torch.optim.lr_scheduler.MultiStepLR(optimizer, 
                                                 milestones=[2,6,15], #调整学习率的epoch数
                                                 gamma=0.1)

更多的官方动态学习率设置方式可参考：https://pytorch.org/docs/stable/optim.html
调用官方接口示例：

model = [Parameter(torch.randn(2, 2, requires_grad=True))]
optimizer = SGD(model, 0.1)
scheduler = ExponentialLR(optimizer, gamma=0.9)

for epoch in range(20):
    for input, target in dataset:
        optimizer.zero_grad()
        output = model(input)
        loss = loss_fn(output, target)
        loss.backward()
        optimizer.step()
    scheduler.step()

3.2 拔高尝试

尝试变更dropout失活比例为0.5，测试集acc提升至86.8%,实际上也不能确定是因为变更比例导致的效果提升，因为每次运行效果都有不同，有好有坏。

Epoch: 1, duration:4820ms, Train_acc:53.6%, Train_loss:1.133, Test_acc:52.6%，Test_loss:0.712, Lr:1.00E-04,save model.pth
Epoch: 2, duration:3271ms, Train_acc:73.5%, Train_loss:0.547, Test_acc:59.2%，Test_loss:0.697, Lr:1.00E-04,save model.pth
Epoch: 3, duration:3566ms, Train_acc:81.5%, Train_loss:0.396, Test_acc:77.6%，Test_loss:0.485, Lr:9.20E-05,save model.pth
Epoch: 4, duration:3309ms, Train_acc:89.4%, Train_loss:0.287, Test_acc:77.6%，Test_loss:0.470, Lr:9.20E-05
Epoch: 5, duration:3411ms, Train_acc:94.2%, Train_loss:0.219, Test_acc:82.9%，Test_loss:0.418, Lr:8.46E-05,save model.pth
Epoch: 6, duration:3239ms, Train_acc:98.2%, Train_loss:0.172, Test_acc:81.6%，Test_loss:0.385, Lr:8.46E-05
Epoch: 7, duration:3282ms, Train_acc:98.8%, Train_loss:0.127, Test_acc:85.5%，Test_loss:0.374, Lr:7.79E-05,save model.pth
Epoch: 8, duration:3277ms, Train_acc:99.2%, Train_loss:0.115, Test_acc:82.9%，Test_loss:0.332, Lr:7.79E-05
Epoch: 9, duration:3440ms, Train_acc:99.6%, Train_loss:0.091, Test_acc:86.8%，Test_loss:0.356, Lr:7.16E-05,save model.pth
Epoch:10, duration:3570ms, Train_acc:100.0%, Train_loss:0.078, Test_acc:81.6%，Test_loss:0.411, Lr:7.16E-05
Epoch:11, duration:3418ms, Train_acc:99.6%, Train_loss:0.072, Test_acc:84.2%，Test_loss:0.370, Lr:6.59E-05
Epoch:12, duration:3291ms, Train_acc:100.0%, Train_loss:0.064, Test_acc:85.5%，Test_loss:0.339, Lr:6.59E-05
Epoch:13, duration:3273ms, Train_acc:100.0%, Train_loss:0.054, Test_acc:85.5%，Test_loss:0.321, Lr:6.06E-05
Epoch:14, duration:3365ms, Train_acc:100.0%, Train_loss:0.049, Test_acc:85.5%，Test_loss:0.336, Lr:6.06E-05
Epoch:15, duration:3321ms, Train_acc:100.0%, Train_loss:0.046, Test_acc:84.2%，Test_loss:0.311, Lr:5.58E-05
Epoch:16, duration:3273ms, Train_acc:100.0%, Train_loss:0.041, Test_acc:84.2%，Test_loss:0.336, Lr:5.58E-05
Epoch:17, duration:3315ms, Train_acc:100.0%, Train_loss:0.038, Test_acc:85.5%，Test_loss:0.350, Lr:5.13E-05
Epoch:18, duration:3380ms, Train_acc:100.0%, Train_loss:0.034, Test_acc:82.9%，Test_loss:0.314, Lr:5.13E-05
Epoch:19, duration:3275ms, Train_acc:100.0%, Train_loss:0.034, Test_acc:84.2%，Test_loss:0.378, Lr:4.72E-05
Epoch:20, duration:3264ms, Train_acc:100.0%, Train_loss:0.031, Test_acc:82.9%，Test_loss:0.342, Lr:4.72E-05
Epoch:21, duration:3267ms, Train_acc:100.0%, Train_loss:0.029, Test_acc:84.2%，Test_loss:0.299, Lr:4.34E-05
Epoch:22, duration:3243ms, Train_acc:100.0%, Train_loss:0.028, Test_acc:84.2%，Test_loss:0.320, Lr:4.34E-05
Epoch:23, duration:3319ms, Train_acc:100.0%, Train_loss:0.025, Test_acc:82.9%，Test_loss:0.335, Lr:4.00E-05
Epoch:24, duration:3230ms, Train_acc:100.0%, Train_loss:0.025, Test_acc:84.2%，Test_loss:0.351, Lr:4.00E-05
Epoch:25, duration:3251ms, Train_acc:100.0%, Train_loss:0.025, Test_acc:84.2%，Test_loss:0.329, Lr:3.68E-05
Epoch:26, duration:3275ms, Train_acc:100.0%, Train_loss:0.023, Test_acc:84.2%，Test_loss:0.304, Lr:3.68E-05
Epoch:27, duration:3244ms, Train_acc:100.0%, Train_loss:0.022, Test_acc:82.9%，Test_loss:0.324, Lr:3.38E-05
Epoch:28, duration:3319ms, Train_acc:100.0%, Train_loss:0.022, Test_acc:85.5%，Test_loss:0.308, Lr:3.38E-05
Epoch:29, duration:3287ms, Train_acc:100.0%, Train_loss:0.020, Test_acc:85.5%，Test_loss:0.353, Lr:3.11E-05
Epoch:30, duration:3230ms, Train_acc:100.0%, Train_loss:0.019, Test_acc:85.5%，Test_loss:0.346, Lr:3.11E-05
Epoch:31, duration:3285ms, Train_acc:100.0%, Train_loss:0.019, Test_acc:85.5%，Test_loss:0.335, Lr:2.86E-05
Epoch:32, duration:3255ms, Train_acc:100.0%, Train_loss:0.019, Test_acc:85.5%，Test_loss:0.337, Lr:2.86E-05
Epoch:33, duration:3307ms, Train_acc:100.0%, Train_loss:0.018, Test_acc:85.5%，Test_loss:0.334, Lr:2.63E-05
Epoch:34, duration:3281ms, Train_acc:100.0%, Train_loss:0.017, Test_acc:85.5%，Test_loss:0.323, Lr:2.63E-05
Epoch:35, duration:3249ms, Train_acc:100.0%, Train_loss:0.016, Test_acc:85.5%，Test_loss:0.314, Lr:2.42E-05
Epoch:36, duration:3287ms, Train_acc:100.0%, Train_loss:0.017, Test_acc:84.2%，Test_loss:0.368, Lr:2.42E-05
Epoch:37, duration:3337ms, Train_acc:100.0%, Train_loss:0.016, Test_acc:85.5%，Test_loss:0.328, Lr:2.23E-05
Epoch:38, duration:3367ms, Train_acc:100.0%, Train_loss:0.016, Test_acc:84.2%，Test_loss:0.375, Lr:2.23E-05
Epoch:39, duration:3244ms, Train_acc:100.0%, Train_loss:0.015, Test_acc:84.2%，Test_loss:0.329, Lr:2.05E-05
Epoch:40, duration:3277ms, Train_acc:100.0%, Train_loss:0.015, Test_acc:85.5%，Test_loss:0.295, Lr:2.05E-05

尝试变更学习率优化器及模型优化器为（SGD和Adam的四种组合），测试集acc几乎无变化
尝试变更初始学习率，尝试变更学习率不动态更新，测试集acc无提升

总结

通过本文学习到几种动态学习率的设置与调用，要想得到一个比较好的模型效果，对模型相关参数进行不同的尝试，获取一个相对适配该案例的参数。