本文结合小土堆的教学视频PyTorch深度学习快速入门教程进行学习
一、完整的模型训练套路
具体代码实现如下
# model.py文件
import torch
from torch import nn
from torch.nn import Conv2d, MaxPool2d, Flatten, Linear, Sequential
# 搭建神经网络
class Model(nn.Module):
def __init__(self) -> None:
super(Model,self).__init__()
self.model = Sequential(
Conv2d(3,32,5,1,2),
MaxPool2d(2),
Conv2d(32,32,5,1,2),
MaxPool2d(2),
Conv2d(32,64,5,1,2),
MaxPool2d(2),
Flatten(),
Linear(1024,64),
Linear(64,10)
)
# 前向传播
def forward(self,x):
x = self.model(x)
return x
if __name__ == '__main__':
model = Model()
input = torch.ones((64,3,32,32))
output = model(input)
print(output.shape)
# train.py文件
import torchvision
from torch.utils.data import DataLoader
from model import * # model.py和train.py必须在同一个文件夹下
from torch.utils.tensorboard import SummaryWriter
# 准备数据集
train_data = torchvision.datasets.CIFAR10("./dataset",train=True,transform=torchvision.transforms.ToTensor(),download=True)
test_data = torchvision.datasets.CIFAR10("./dataset",train=False,transform=torchvision.transforms.ToTensor(),download=True)
# length长度
train_data_len = len(train_data)
test_data_len = len(test_data)
print("训练数据集的长度为:{}".format(train_data_len)) # 格式化字符串
print("测试数据集的长度为:{}".format(test_data_len))
# 加载数据集
train_dataloader = DataLoader(train_data,batch_size=64)
test_dataloader = DataLoader(test_data,batch_size=64)
# 创建网络模型
model = Model()
# 损失函数
loss_fn = nn.CrossEntropyLoss()
# 优化器
# learning_rate = 0.01
learning_rate = 1e-2 # 1e-1 = 1×(10)^(-2) = 0.01
optimizer = torch.optim.SGD(model.parameters(),lr=learning_rate)
# 设置训练网络的一些参数
total_train_step = 0 # 记录训练的次数
total_test_step = 0 # 记录测试的次数
epoch = 10 # 训练的轮数
# 添加tensorboard
writer = SummaryWriter("logs")
for i in range(epoch):
print("--------第{}轮训练开始--------".format(i+1))
# 训练步骤开始
model.train() # 让网络进入训练状态
for data in train_dataloader:
imgs,targets = data
outputs = model(imgs)
loss = loss_fn(outputs,targets) # 计算误差值
# 优化器优化模型
optimizer.zero_grad() # 将参数梯度清0
loss.backward() # 反向传播
optimizer.step() # 对模型参数进行调优
total_train_step += 1
# 展示输出
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_test_step)
# 测试步骤开始
model.eval() # 让网络进入测试状态
total_test_loss = 0
total_accuracy = 0
with torch.no_grad(): # 去除梯度,使模型不会对参数进行调优
for data in test_dataloader:
imgs,targets = data
outputs = model(imgs)
loss = loss_fn(outputs,targets) # 计算误差值
total_test_loss += loss
accuracy = (outputs.argmax(1) == targets).sum() # 计算预算正确的数量
total_accuracy += accuracy
# 展示效果
print("整体测试集上的Loss:{}".format(total_test_loss))
print("整体测试集上的正确率:{}".format(total_accuracy/test_data_len))
writer.add_scalar("test_loss",loss.item(),total_test_step)
writer.add_scalar("test_accuracy",total_accuracy/test_data_len,total_test_step)
total_test_step += 1
# 保存模型
torch.save(model,"model_{}.pth".format(i+1))
print("模型已保存")
writer.close()
程序运行结果显示如下
Files already downloaded and verified
Files already downloaded and verified
训练数据集的长度为:50000
测试数据集的长度为:10000
--------第1轮训练开始--------
训练次数:100, Loss:2.2883501052856445
训练次数:200, Loss:2.283978223800659
训练次数:300, Loss:2.276535749435425
训练次数:400, Loss:2.211817741394043
训练次数:500, Loss:2.121809482574463
训练次数:600, Loss:2.0784034729003906
训练次数:700, Loss:1.9761544466018677
整体测试集上的Loss:311.06243896484375
整体测试集上的正确率:0.288100004196167
模型已保存
--------第2轮训练开始--------
训练次数:800, Loss:1.878362774848938
训练次数:900, Loss:1.832785964012146
训练次数:1000, Loss:1.9041134119033813
训练次数:1100, Loss:1.9213755130767822
训练次数:1200, Loss:1.6937150955200195
训练次数:1300, Loss:1.6661460399627686
训练次数:1400, Loss:1.7335188388824463
训练次数:1500, Loss:1.7996962070465088
整体测试集上的Loss:285.850341796875
整体测试集上的正确率:0.3483000099658966
模型已保存
--------第3轮训练开始--------
训练次数:1600, Loss:1.7524412870407104
训练次数:1700, Loss:1.6654188632965088
训练次数:1800, Loss:1.9707260131835938
训练次数:1900, Loss:1.7274596691131592
训练次数:2000, Loss:1.9246697425842285
训练次数:2100, Loss:1.5304251909255981
训练次数:2200, Loss:1.4880874156951904
训练次数:2300, Loss:1.7865030765533447
整体测试集上的Loss:255.3195037841797
整体测试集上的正确率:0.4147000014781952
模型已保存
--------第4轮训练开始--------
训练次数:2400, Loss:1.732339859008789
训练次数:2500, Loss:1.3916590213775635
训练次数:2600, Loss:1.6185827255249023
训练次数:2700, Loss:1.6978445053100586
训练次数:2800, Loss:1.4863864183425903
训练次数:2900, Loss:1.6016960144042969
训练次数:3000, Loss:1.3664864301681519
训练次数:3100, Loss:1.5545544624328613
整体测试集上的Loss:244.28079223632812
整体测试集上的正确率:0.43779999017715454
模型已保存
--------第5轮训练开始--------
训练次数:3200, Loss:1.3771955966949463
训练次数:3300, Loss:1.4948328733444214
训练次数:3400, Loss:1.4798005819320679
训练次数:3500, Loss:1.563676357269287
训练次数:3600, Loss:1.5787324905395508
训练次数:3700, Loss:1.3048750162124634
训练次数:3800, Loss:1.267425298690796
训练次数:3900, Loss:1.4718506336212158
整体测试集上的Loss:242.99974060058594
整体测试集上的正确率:0.43939998745918274
模型已保存
--------第6轮训练开始--------
训练次数:4000, Loss:1.4343503713607788
训练次数:4100, Loss:1.4455058574676514
训练次数:4200, Loss:1.5285396575927734
训练次数:4300, Loss:1.2184882164001465
训练次数:4400, Loss:1.1589688062667847
训练次数:4500, Loss:1.326550006866455
训练次数:4600, Loss:1.3971928358078003
整体测试集上的Loss:238.05587768554688
整体测试集上的正确率:0.44999998807907104
模型已保存
--------第7轮训练开始--------
训练次数:4700, Loss:1.275310754776001
训练次数:4800, Loss:1.5368099212646484
训练次数:4900, Loss:1.370495319366455
训练次数:5000, Loss:1.370849609375
训练次数:5100, Loss:1.0084483623504639
训练次数:5200, Loss:1.3147143125534058
训练次数:5300, Loss:1.2318178415298462
训练次数:5400, Loss:1.43149733543396
整体测试集上的Loss:232.78636169433594
整体测试集上的正确率:0.4666999876499176
模型已保存
--------第8轮训练开始--------
训练次数:5500, Loss:1.240295648574829
训练次数:5600, Loss:1.2105470895767212
训练次数:5700, Loss:1.2011173963546753
训练次数:5800, Loss:1.2001447677612305
训练次数:5900, Loss:1.3301762342453003
训练次数:6000, Loss:1.5595725774765015
训练次数:6100, Loss:1.070670247077942
训练次数:6200, Loss:1.126779317855835
整体测试集上的Loss:224.64822387695312
整体测试集上的正确率:0.4875999987125397
模型已保存
--------第9轮训练开始--------
训练次数:6300, Loss:1.4885517358779907
训练次数:6400, Loss:1.0875073671340942
训练次数:6500, Loss:1.57513427734375
训练次数:6600, Loss:1.1213852167129517
训练次数:6700, Loss:1.0803751945495605
训练次数:6800, Loss:1.1594667434692383
训练次数:6900, Loss:1.0916932821273804
训练次数:7000, Loss:0.9281116724014282
整体测试集上的Loss:211.3177490234375
整体测试集上的正确率:0.5192999839782715
模型已保存
--------第10轮训练开始--------
训练次数:7100, Loss:1.2686500549316406
训练次数:7200, Loss:1.0005848407745361
训练次数:7300, Loss:1.1087175607681274
训练次数:7400, Loss:0.8756632804870605
训练次数:7500, Loss:1.2323938608169556
训练次数:7600, Loss:1.2329813241958618
训练次数:7700, Loss:0.849784255027771
训练次数:7800, Loss:1.2703667879104614
整体测试集上的Loss:200.145751953125
整体测试集上的正确率:0.5473999977111816
模型已保存
将程序运行后,在终端输入 tensorboard --logdir=logs --port=6007 ,结果显示如下,可见网络模型在测试集上的精确度逐步提升,损失函数值逐渐降低
二、利用GPU训练
方式1:.cuda()
具体代码实现如下
# train.py文件
import torch
from torch import nn
from torch.nn import Conv2d, MaxPool2d, Flatten, Linear, Sequential
import torchvision
from torch.utils.data import DataLoader
from torch.utils.tensorboard import SummaryWriter
# 准备数据集
train_data = torchvision.datasets.CIFAR10("./dataset",train=True,transform=torchvision.transforms.ToTensor(),download=True)
test_data = torchvision.datasets.CIFAR10("./dataset",train=False,transform=torchvision.transforms.ToTensor(),download=True)
# length长度
train_data_len = len(train_data)
test_data_len = len(test_data)
print("训练数据集的长度为:{}".format(train_data_len)) # 格式化字符串
print("测试数据集的长度为:{}".format(test_data_len))
# 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 Model(nn.Module):
def __init__(self) -> None:
super(Model,self).__init__()
self.model = Sequential(
Conv2d(3,32,5,1,2),
MaxPool2d(2),
Conv2d(32,32,5,1,2),
MaxPool2d(2),
Conv2d(32,64,5,1,2),
MaxPool2d(2),
Flatten(),
Linear(1024,64),
Linear(64,10)
)
# 前向传播
def forward(self,x):
x = self.model(x)
return x
# 创建网络模型
model = Model()
if torch.cuda.is_available():
model.cuda() # 将网络模型转移到cuda
# 损失函数
loss_fn = nn.CrossEntropyLoss()
if torch.cuda.is_available():
loss_fn.cuda()
# 优化器
# learning_rate = 0.01
learning_rate = 1e-2 # 1e-1 = 1×(10)^(-2) = 0.01
optimizer = torch.optim.SGD(model.parameters(),lr=learning_rate)
# 设置训练网络的一些参数
total_train_step = 0 # 记录训练的次数
total_test_step = 0 # 记录测试的次数
epoch = 10 # 训练的轮数
# 添加tensorboard
writer = SummaryWriter("logs")
for i in range(epoch):
print("--------第{}轮训练开始--------".format(i+1))
# 训练步骤开始
model.train() # 让网络进入训练状态
for data in train_dataloader:
imgs,targets = data
if torch.cuda.is_available():
imgs = imgs.cuda()
targets = targets.cuda()
outputs = model(imgs)
loss = loss_fn(outputs,targets) # 计算误差值
# 优化器优化模型
optimizer.zero_grad() # 将参数梯度清0
loss.backward() # 反向传播
optimizer.step() # 对模型参数进行调优
total_train_step += 1
# 展示输出
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_test_step)
# 测试步骤开始
model.eval() # 让网络进入测试状态
total_test_loss = 0
total_accuracy = 0
with torch.no_grad(): # 去除梯度,使模型不会对参数进行调优
for data in test_dataloader:
imgs,targets = data
if torch.cuda.is_available():
imgs = imgs.cuda()
targets = targets.cuda()
outputs = model(imgs)
loss = loss_fn(outputs,targets) # 计算误差值
total_test_loss += loss
accuracy = (outputs.argmax(1) == targets).sum() # 计算预算正确的数量
total_accuracy += accuracy
# 展示效果
print("整体测试集上的Loss:{}".format(total_test_loss))
print("整体测试集上的正确率:{}".format(total_accuracy/test_data_len))
writer.add_scalar("test_loss",loss.item(),total_test_step)
writer.add_scalar("test_accuracy",total_accuracy/test_data_len,total_test_step)
total_test_step += 1
# 保存模型
torch.save(model,"model_{}.pth".format(i+1))
print("模型已保存")
writer.close()
方式2:torch.device() & .to()(常用)
# train.py文件
import torch
from torch import nn
from torch.nn import Conv2d, MaxPool2d, Flatten, Linear, Sequential
import torchvision
from torch.utils.data import DataLoader
from torch.utils.tensorboard import SummaryWriter
# 定义一个设备, 通过控制参数来控制程序在哪里运行
device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
# 准备数据集
train_data = torchvision.datasets.CIFAR10("./dataset",train=True,transform=torchvision.transforms.ToTensor(),download=True)
test_data = torchvision.datasets.CIFAR10("./dataset",train=False,transform=torchvision.transforms.ToTensor(),download=True)
# length长度
train_data_len = len(train_data)
test_data_len = len(test_data)
print("训练数据集的长度为:{}".format(train_data_len)) # 格式化字符串
print("测试数据集的长度为:{}".format(test_data_len))
# 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 Model(nn.Module):
def __init__(self) -> None:
super(Model,self).__init__()
self.model = Sequential(
Conv2d(3,32,5,1,2),
MaxPool2d(2),
Conv2d(32,32,5,1,2),
MaxPool2d(2),
Conv2d(32,64,5,1,2),
MaxPool2d(2),
Flatten(),
Linear(1024,64),
Linear(64,10)
)
# 前向传播
def forward(self,x):
x = self.model(x)
return x
# 创建网络模型
model = Model()
model.to(device) # 把网络转移到device设备上
# 损失函数
loss_fn = nn.CrossEntropyLoss()
loss_fn.to(device)
# 优化器
# learning_rate = 0.01
learning_rate = 1e-2 # 1e-1 = 1×(10)^(-2) = 0.01
optimizer = torch.optim.SGD(model.parameters(),lr=learning_rate)
# 设置训练网络的一些参数
total_train_step = 0 # 记录训练的次数
total_test_step = 0 # 记录测试的次数
epoch = 10 # 训练的轮数
# 添加tensorboard
writer = SummaryWriter("logs")
for i in range(epoch):
print("--------第{}轮训练开始--------".format(i+1))
# 训练步骤开始
model.train() # 让网络进入训练状态
for data in train_dataloader:
imgs,targets = data
imgs = imgs.to(device)
targets = targets.to(device)
outputs = model(imgs)
loss = loss_fn(outputs,targets) # 计算误差值
# 优化器优化模型
optimizer.zero_grad() # 将参数梯度清0
loss.backward() # 反向传播
optimizer.step() # 对模型参数进行调优
total_train_step += 1
# 展示输出
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_test_step)
# 测试步骤开始
model.eval() # 让网络进入测试状态
total_test_loss = 0
total_accuracy = 0
with torch.no_grad(): # 去除梯度,使模型不会对参数进行调优
for data in test_dataloader:
imgs,targets = data
imgs = imgs.to(device)
targets = targets.to(device)
outputs = model(imgs)
loss = loss_fn(outputs,targets) # 计算误差值
total_test_loss += loss
accuracy = (outputs.argmax(1) == targets).sum() # 计算预算正确的数量
total_accuracy += accuracy
# 展示效果
print("整体测试集上的Loss:{}".format(total_test_loss))
print("整体测试集上的正确率:{}".format(total_accuracy/test_data_len))
writer.add_scalar("test_loss",loss.item(),total_test_step)
writer.add_scalar("test_accuracy",total_accuracy/test_data_len,total_test_step)
total_test_step += 1
# 保存模型
torch.save(model,"model_{}.pth".format(i+1))
print("模型已保存")
writer.close()
三、完整的模型验证套路
验证:把训练好的模型应用到实际环境中
具体代码实现如下
from PIL import Image
import torchvision
import torch
from torch import nn
from torch.nn import Conv2d, MaxPool2d, Flatten, Linear, Sequential
class Model(nn.Module):
def __init__(self) -> None:
super(Model,self).__init__()
self.model = Sequential(
Conv2d(3,32,5,1,2),
MaxPool2d(2),
Conv2d(32,32,5,1,2),
MaxPool2d(2),
Conv2d(32,64,5,1,2),
MaxPool2d(2),
Flatten(),
Linear(1024,64),
Linear(64,10)
)
def forward(self,x):
x = self.model(x)
return x
image_path = "./image/dog.png"
image = Image.open(image_path)
print(image)
image = image.convert('RGB') # 将图片转化为RGB三通道
transform = torchvision.transforms.Compose([torchvision.transforms.Resize([32,32]),
torchvision.transforms.ToTensor()])
image = transform(image)
print(image.shape)
model = torch.load("model_1.pth")
print(model)
image = torch.reshape(image,(1,3,32,32)) # 添加batch_size
model.eval()
with torch.no_grad():
output = model(image)
print(output)
print(output.argmax(1))
程序运行结果如下,输出索引是6,其对应的种类是frog而不是dog,说明采用第一次训练的模型准确率还有待提高
<PIL.JpegImagePlugin.JpegImageFile image mode=RGB size=1659x1073 at 0x1C8688E2610>
<PIL.Image.Image image mode=RGB size=1659x1073 at 0x1C8688A3040>
torch.Size([3, 32, 32])
Model(
(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([[-1.7079, 0.4696, 0.6549, 0.4952, 0.7226, 0.8485, 1.3298, 0.2228,
-1.4841, 0.0125]])
tensor([6])