新手小白深度学习第P5周
⭐本文为365天深度学习训练营的学习记录博客
⭐原作者为K同学啊
基础配置
🏡 我的环境:
● 语言环境:Python3.8
● 编译器:jupyter notebook
● 深度学习环境:Pytorch
一、前期准备
1.设置GPU
import torch
import torch.nn as nn
import torchvision.transforms as transforms
import torchvision
from torchvision import transforms, datasets
import os,PIL,pathlib
device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
device
输出:
device(type='cpu')
2.导入数据
● 第一步:使用pathlib.Path()函数将字符串类型的文件夹路径转换为pathlib.Path对象。
● 第二步:使用glob()方法获取data_dir路径下的所有文件路径,并以列表形式存储在data_paths中。
● 第三步:通过split()函数对data_paths中的每个文件路径执行分割操作,获得各个文件所属的类别名称,并存储在classeNames中
● 第四步:打印classeNames列表,显示每个文件所属的类别名称。
import os,PIL,random,pathlib
data_dir='./46-data/'
data_dir=pathlib.Path(data_dir)
data_paths=list(data_dir.glob('*'))
classeNames=[str(path).split("\\")[1] for path in data_paths]
classeNames
输出:
['test', 'train']
train_transforms=transforms.Compose(
[
transforms.Resize([224,224]),
transforms.ToTensor(),
transforms.Normalize(mean=[0.485,0.456,0.406],
std=[0.229,0.224,0.225])
]
)
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("./46-data/train/",transform=train_transforms)
test_dataset = datasets.ImageFolder("./46-data/test/",transform=train_transforms)
train_dataset.class_to_idx
输出:
{'adidas': 0, 'nike': 1}
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)
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
二、构建简单的CNN网络
import torch.nn.functional as F
class Model(nn.Module):
def __init__(self):
super(Model, self).__init__()
self.conv1=nn.Sequential(
nn.Conv2d(3, 12, kernel_size=5, padding=0), # 12*220*220
nn.BatchNorm2d(12),
nn.ReLU())
self.conv2=nn.Sequential(
nn.Conv2d(12, 12, kernel_size=5, padding=0), # 12*216*216
nn.BatchNorm2d(12),
nn.ReLU())
self.pool3=nn.Sequential(
nn.MaxPool2d(2)) # 12*108*108
self.conv4=nn.Sequential(
nn.Conv2d(12, 24, kernel_size=5, padding=0), # 24*104*104
nn.BatchNorm2d(24),
nn.ReLU())
self.conv5=nn.Sequential(
nn.Conv2d(24, 24, kernel_size=5, padding=0), # 24*100*100
nn.BatchNorm2d(24),
nn.ReLU())
self.pool6=nn.Sequential(
nn.MaxPool2d(2)) # 24*50*50
self.dropout = nn.Sequential(
nn.Dropout(0.2))
self.fc=nn.Sequential(
nn.Linear(24*50*50, len(classeNames)))
# Sequential函数起封装作用, 是一个有序的容器,它包含了多个网络层,数据会按照在 nn.Sequential 中添加的顺序依次通过这些层。
def forward(self, x):
batch_size = x.size(0)
x = self.conv1(x) # 卷积-BN-激活
x = self.conv2(x) # 卷积-BN-激活
x = self.pool3(x) # 池化
x = self.conv4(x) # 卷积-BN-激活
x = self.conv5(x) # 卷积-BN-激活
x = self.pool6(x) # 池化
x = self.dropout(x)#Dropout层用于在训练过程中随机地将一部分神经元的输出设置为0,这是一种正则化技术,用于防止模型过拟合。参数0.2` 表示在每次前向传播时,有20%的神经元会被随机丢弃。
x = x.view(batch_size, -1) # flatten 变成全连接网络需要的输入 (batch, 24*50*50) ==> (batch, -1), -1 此处自动算出的是24*50*50
x = self.fc(x)
return x
device = "cuda" if torch.cuda.is_available() else "cpu"
print("Using {} device".format(device))
model = Model().to(device)
model
输出:
Using cpu device
Model(
(conv1): Sequential(
(0): Conv2d(3, 12, kernel_size=(5, 5), stride=(1, 1))
(1): BatchNorm2d(12, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
(2): ReLU()
)
(conv2): Sequential(
(0): Conv2d(12, 12, kernel_size=(5, 5), stride=(1, 1))
(1): BatchNorm2d(12, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
(2): ReLU()
)
(pool3): Sequential(
(0): MaxPool2d(kernel_size=2, stride=2, padding=0, dilation=1, ceil_mode=False)
)
(conv4): Sequential(
(0): Conv2d(12, 24, kernel_size=(5, 5), stride=(1, 1))
(1): BatchNorm2d(24, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
(2): ReLU()
)
(conv5): Sequential(
(0): Conv2d(24, 24, kernel_size=(5, 5), stride=(1, 1))
(1): BatchNorm2d(24, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)
(2): ReLU()
)
(pool6): Sequential(
(0): MaxPool2d(kernel_size=2, stride=2, padding=0, dilation=1, ceil_mode=False)
)
(dropout): Sequential(
(0): Dropout(p=0.2, inplace=False)
)
(fc): Sequential(
(0): Linear(in_features=60000, out_features=2, bias=True)
)
)
三、训练模型
1.编写训练模型
# 训练循环
def train(dataloader, model, loss_fn, optimizer):
size = len(dataloader.dataset) # 训练集的大小
num_batches = len(dataloader) # 批次数目, (size/batch_size,向上取整)
train_loss, train_acc = 0, 0 # 初始化训练损失和正确率
for X, y in dataloader: # 获取图片及其标签
X, y = X.to(device), y.to(device)
# 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() # grad属性归零
loss.backward() # 反向传播
optimizer.step() # 每一步自动更新
# 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.编写测试函数
def test (dataloader, model, loss_fn):
size = len(dataloader.dataset) # 测试集的大小
num_batches = len(dataloader) # 批次数目, (size/batch_size,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():
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()
test_acc /= size
test_loss /= num_batches
return test_acc, test_loss
3.设置动态学习率
def adjust_learning_rate(optimizer, epoch, start_lr):
# 每 2 个epoch衰减到原来的 0.98
lr = start_lr * (0.92 ** (epoch // 2))
for param_group in optimizer.param_groups:
param_group['lr'] = lr
# 以下是官方动态学习率接口
learn_rate = 1e-3# 初始学习率
optimizer = torch.optim.SGD(model.parameters(), lr=learn_rate)
scheduler = torch.optim.lr_scheduler.StepLR(optimizer, step_size=5, gamma=0.1)
#scheduler = torch.optim.lr_scheduler.MultiStepLR(optimizer,
# milestones=[2,6,15,27,35], #调整学习率的epoch数
#gamma=0.1)
正式训练
loss_fn = nn.CrossEntropyLoss() # 创建损失函数
epochs = 40
train_loss = []
train_acc = []
test_loss = []
test_acc = []
for epoch in range(epochs):
# 更新学习率(使用自定义学习率时使用)
# adjust_learning_rate(optimizer, epoch, learn_rate)
model.train()
epoch_train_acc, epoch_train_loss = train(train_dl, model, loss_fn, optimizer)
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 = optimizer.state_dict()['param_groups'][0]['lr']
template = ('Epoch:{:2d}, Train_acc:{:.1f}%, Train_loss:{:.3f}, Test_acc:{:.1f}%, Test_loss:{:.3f}, Lr:{:.2E}')
print(template.format(epoch+1, epoch_train_acc*100, epoch_train_loss,
epoch_test_acc*100, epoch_test_loss, lr))
print('Done')
输出:
Epoch: 1, Train_acc:57.6%, Train_loss:0.713, Test_acc:52.6%, Test_loss:0.686, Lr:1.00E-04
Epoch: 2, Train_acc:61.2%, Train_loss:0.668, Test_acc:51.3%, Test_loss:0.699, Lr:1.00E-04
Epoch: 3, Train_acc:67.9%, Train_loss:0.598, Test_acc:81.6%, Test_loss:0.537, Lr:1.00E-04
Epoch: 4, Train_acc:72.5%, Train_loss:0.556, Test_acc:78.9%, Test_loss:0.516, Lr:1.00E-04
Epoch: 5, Train_acc:74.1%, Train_loss:0.529, Test_acc:75.0%, Test_loss:0.503, Lr:1.00E-05
Epoch: 6, Train_acc:77.9%, Train_loss:0.494, Test_acc:82.9%, Test_loss:0.482, Lr:1.00E-05
Epoch: 7, Train_acc:76.7%, Train_loss:0.496, Test_acc:85.5%, Test_loss:0.473, Lr:1.00E-05
Epoch: 8, Train_acc:77.3%, Train_loss:0.495, Test_acc:82.9%, Test_loss:0.458, Lr:1.00E-05
Epoch: 9, Train_acc:77.7%, Train_loss:0.498, Test_acc:84.2%, Test_loss:0.463, Lr:1.00E-05
Epoch:10, Train_acc:79.3%, Train_loss:0.486, Test_acc:84.2%, Test_loss:0.462, Lr:1.00E-06
Epoch:11, Train_acc:80.1%, Train_loss:0.492, Test_acc:85.5%, Test_loss:0.474, Lr:1.00E-06
Epoch:12, Train_acc:80.7%, Train_loss:0.467, Test_acc:84.2%, Test_loss:0.492, Lr:1.00E-06
Epoch:13, Train_acc:80.5%, Train_loss:0.469, Test_acc:84.2%, Test_loss:0.463, Lr:1.00E-06
Epoch:14, Train_acc:81.3%, Train_loss:0.472, Test_acc:82.9%, Test_loss:0.476, Lr:1.00E-06
Epoch:15, Train_acc:79.9%, Train_loss:0.478, Test_acc:82.9%, Test_loss:0.456, Lr:1.00E-07
Epoch:16, Train_acc:79.5%, Train_loss:0.463, Test_acc:82.9%, Test_loss:0.493, Lr:1.00E-07
Epoch:17, Train_acc:78.9%, Train_loss:0.479, Test_acc:82.9%, Test_loss:0.469, Lr:1.00E-07
Epoch:18, Train_acc:82.1%, Train_loss:0.463, Test_acc:82.9%, Test_loss:0.494, Lr:1.00E-07
Epoch:19, Train_acc:80.5%, Train_loss:0.465, Test_acc:82.9%, Test_loss:0.481, Lr:1.00E-07
Epoch:20, Train_acc:81.7%, Train_loss:0.478, Test_acc:82.9%, Test_loss:0.478, Lr:1.00E-08
Epoch:21, Train_acc:81.1%, Train_loss:0.471, Test_acc:82.9%, Test_loss:0.461, Lr:1.00E-08
Epoch:22, Train_acc:79.9%, Train_loss:0.479, Test_acc:81.6%, Test_loss:0.495, Lr:1.00E-08
Epoch:23, Train_acc:82.1%, Train_loss:0.463, Test_acc:82.9%, Test_loss:0.501, Lr:1.00E-08
Epoch:24, Train_acc:78.5%, Train_loss:0.479, Test_acc:81.6%, Test_loss:0.476, Lr:1.00E-08
Epoch:25, Train_acc:80.9%, Train_loss:0.468, Test_acc:81.6%, Test_loss:0.466, Lr:1.00E-09
Epoch:26, Train_acc:82.7%, Train_loss:0.466, Test_acc:82.9%, Test_loss:0.481, Lr:1.00E-09
Epoch:27, Train_acc:79.9%, Train_loss:0.486, Test_acc:82.9%, Test_loss:0.485, Lr:1.00E-09
Epoch:28, Train_acc:79.9%, Train_loss:0.478, Test_acc:82.9%, Test_loss:0.466, Lr:1.00E-09
Epoch:29, Train_acc:81.5%, Train_loss:0.479, Test_acc:82.9%, Test_loss:0.486, Lr:1.00E-09
Epoch:30, Train_acc:80.5%, Train_loss:0.475, Test_acc:84.2%, Test_loss:0.474, Lr:1.00E-10
Epoch:31, Train_acc:80.9%, Train_loss:0.478, Test_acc:84.2%, Test_loss:0.473, Lr:1.00E-10
Epoch:32, Train_acc:80.7%, Train_loss:0.461, Test_acc:82.9%, Test_loss:0.483, Lr:1.00E-10
Epoch:33, Train_acc:80.1%, Train_loss:0.470, Test_acc:85.5%, Test_loss:0.483, Lr:1.00E-10
Epoch:34, Train_acc:79.7%, Train_loss:0.479, Test_acc:82.9%, Test_loss:0.474, Lr:1.00E-10
Epoch:35, Train_acc:79.5%, Train_loss:0.484, Test_acc:81.6%, Test_loss:0.493, Lr:1.00E-11
Epoch:36, Train_acc:81.1%, Train_loss:0.471, Test_acc:82.9%, Test_loss:0.482, Lr:1.00E-11
Epoch:37, Train_acc:83.5%, Train_loss:0.471, Test_acc:82.9%, Test_loss:0.474, Lr:1.00E-11
Epoch:38, Train_acc:83.1%, Train_loss:0.458, Test_acc:82.9%, Test_loss:0.462, Lr:1.00E-11
Epoch:39, Train_acc:81.3%, Train_loss:0.473, Test_acc:84.2%, Test_loss:0.475, Lr:1.00E-11
Epoch:40, Train_acc:80.7%, Train_loss:0.480, Test_acc:82.9%, Test_loss:0.481, Lr:1.00E-12
Done```
# 四、结果可视化
## 1.Loss与Accuracy图
```python
import matplotlib.pyplot as plt
#隐藏警告
import warnings
warnings.filterwarnings("ignore") #忽略警告信息
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
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.图片预测
from PIL import Image
classes = list(train_dataset.class_to_idx)
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()
plt.close()
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}')
# 预测训练集中所有预测结果
folder_path='./46-data/test/adidas/'
image_files=[f for f in os.listdir(folder_path) if f.endswith('.jpg') or f.endswith('.png')]
for image_file in image_files:
predict_one_image(image_path=os.path.join(folder_path, image_file),
model=model,
transform=train_transforms,
classes=classes)
六·、动态学习率
1.torch.optim.lr_scheduler.StepLR
等间隔动态调整方法,每经过step_size个epoch,做一次学习率decay,以gamma值为缩小倍数。
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
示例:
optimizer = torch.optim.SGD(net.parameters(), lr=0.001 )
scheduler = torch.optim.lr_scheduler.StepLR(optimizer, step_size=5, gamma=0.1)
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
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.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
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)
七、个人总结
1.Adam优化器不一定总比SGD优化器好,在此模型中,使用Adam准确率反而不佳。
2.学习率衰减因子,和学习率衰减周期对模型的影响极大。
3.对设置动态学习率的手动和官方方法熟练了解与掌握。