PyTorch基础完整模型训练套路（土堆老师版）详细注释及讲解！小白学习必看！
目录
1、准备数据集
2、利用dataloader加载数据集
3、创建网络模型
model.py
4、损失函数
5、优化器
6、设置训练网络的参数
7 、添加tensorboard
8 、训练过程并保存模型结构及参数
① cyx.train()和cyx.eval()
② loss.item()
③ with torch.no_grad():
④ argmax()
9、关闭writer
完整代码
1、准备数据集
# 准备数据集
from torch.utils.data import DataLoader
train_data = torchvision.datasets.CIFAR10(root="datasets", train=True,
transform=torchvision.transforms.ToTensor(),
download=False)
test_data = torchvision.datasets.CIFAR10(root="datasets", train=False,
transform=torchvision.transforms.ToTensor(),
download=False)
train_data_size = len(train_data)
test_data_size = len(test_data)
print("训练数据集的长度：{}".format(train_data_size))
print("测试数据集的长度：{}".format(test_data_size))2、利用dataloader加载数据集
train_dataloader = DataLoader(train_data, batch_size=64)
test_dataloader = DataLoader(test_data, batch_size=64)3、创建网络模型
from model import *
cyx = CYX()model.py
# 搭建神经网络
import torch
from torch import nn
from torch.nn import Sequential, Conv2d, MaxPool2d, Flatten, Linear
class CYX(nn.Module):
def __init__(self):
super(CYX, self).__init__()
self.model1 = Sequential(
Conv2d(in_channels=3, out_channels=32, kernel_size=5,
padding=2, stride=1),
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.model1(x)
return x
# 验证模型是否成功搭建
# if __name__ == '__main__':
# cyx = CYX()
# input = torch.ones((64, 3, 32, 32))
# output = cyx(input)
# print(output.shape)4、损失函数
loss_fn = nn.CrossEntropyLoss()5、优化器
learning_rate = 0.01
optimizer = torch.optim.SGD(cyx.parameters(), lr=learning_rate)6、设置训练网络的参数
# 记录训练的次数
total_train_step = 0
# 记录测试的次数
total_test_step = 0
# 记录训练的轮数
epoch = 107 、添加tensorboard
writer = SummaryWriter("logs_train")8 、训练过程并保存模型结构及参数
for i in range(epoch):
print("---------第{}轮训练开始---------".format(i+1))
# 训练步骤开始
# 只对特定层起作用
cyx.train()
for data in train_dataloader:
imgs, targets = data
outputs = cyx(imgs)
loss = loss_fn(outputs, targets)
# 优化器优化
optimizer.zero_grad()
loss.backward()
optimizer.step()
total_train_step = total_train_step + 1
# 每逢100的时候再打印（好看）
if total_train_step % 100 == 0:
# .item()会把tensor里面的变成数字，比如tensor(5)->5
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
# 只对特定层起作用
cyx.eval()
# 看的是整个数据集的损失
total_test_loss = 0
total_accuracy = 0
# 不累计梯度
with torch.no_grad():
for data in test_dataloader:
imgs, targets = data
outputs = cyx(imgs)
loss = loss_fn(outputs, targets)
total_test_loss = total_test_loss + loss.item()
# 见截图
accuracy = (outputs.argmax(1) == targets).sum()
total_accuracy = total_accuracy + accuracy
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_size))
writer.add_scalar("test_loss", total_test_loss, total_test_step)
writer.add_scalar("test_accuracy", total_accuracy/test_data_size, total_train_step)
total_test_step = total_test_step + 1
torch.save(cyx, "cyx_cpu_{}.pth".format(i))
# torch.save(cyx.state_dict(), "cyx_{}.pth".format(i))
print("模型已保存")注意细节：
① cyx.train()和cyx.eval()
是否启用 Batch Normalization 和 Dropout，分别在训练时和测试时添加。
参考文章：
【Pytorch】model.train() 和 model.eval() 原理与用法
② loss.item()
把tensor里面的变成数字，比如tensor(5)->5
参考文章：
【Pytorch】.item() 方法介绍
③ with torch.no_grad():
with 语句适用于对资源进行访问的场合，确保不管使用过程中是否发生异常都会执行必要的“清理”操作，释放资源，比如文件使用后自动关闭／线程中锁的自动获取和释放等。
因为是测试，不累计梯度
参考文章：
【pytorch】 with torch.no_grad():用法详解
④ argmax()
arg = 1时，沿 axis=1 比较 
arg = 0时，沿 axis=0 比较
9、关闭writer
writer.close()
完整代码
import torch
import torchvision
from torch import nn
from torch.nn import Conv2d, MaxPool2d, Flatten, Linear, Sequential
from torch.utils.tensorboard import SummaryWriter
from model import *
# 准备数据集
from torch.utils.data import DataLoader
train_data = torchvision.datasets.CIFAR10(root="datasets", train=True,
transform=torchvision.transforms.ToTensor(),
download=False)
test_data = torchvision.datasets.CIFAR10(root="datasets", train=False,
transform=torchvision.transforms.ToTensor(),
download=False)
train_data_size = len(train_data)
test_data_size = len(test_data)
print("训练数据集的长度：{}".format(train_data_size))
print("测试数据集的长度：{}".format(test_data_size))
# 利用dataloader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
cyx = CYX()
# 损失函数
loss_fn = nn.CrossEntropyLoss()
# 优化器
learning_rate = 0.01
optimizer = torch.optim.SGD(cyx.parameters(), lr=learning_rate)
# 设置训练网络的参数
# 记录训练的次数
total_train_step = 0
# 记录测试的次数
total_test_step = 0
# 记录训练的轮数
epoch = 10
# 添加tensorboard
writer = SummaryWriter("logs_train")
for i in range(epoch):
print("---------第{}轮训练开始---------".format(i+1))
# 训练步骤开始
# 只对特定层起作用
cyx.train()
for data in train_dataloader:
imgs, targets = data
outputs = cyx(imgs)
loss = loss_fn(outputs, targets)
# 优化器优化
optimizer.zero_grad()
loss.backward()
optimizer.step()
total_train_step = total_train_step + 1
# 每逢100的时候再打印（好看）
if total_train_step % 100 == 0:
# .item()会把tensor里面的变成数字，比如tensor(5)->5
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
# 只对特定层起作用
cyx.eval()
# 看的是整个数据集的损失
total_test_loss = 0
total_accuracy = 0
# 不累计梯度
with torch.no_grad():
for data in test_dataloader:
imgs, targets = data
outputs = cyx(imgs)
loss = loss_fn(outputs, targets)
total_test_loss = total_test_loss + loss.item()
# 见截图
accuracy = (outputs.argmax(1) == targets).sum()
total_accuracy = total_accuracy + accuracy
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_size))
writer.add_scalar("test_loss", total_test_loss, total_test_step)
writer.add_scalar("test_accuracy", total_accuracy/test_data_size, total_train_step)
total_test_step = total_test_step + 1
torch.save(cyx, "cyx_cpu_{}.pth".format(i))
# torch.save(cyx.state_dict(), "cyx_{}.pth".format(i))
print("模型已保存")
writer.close()