Pytorch笔记20 完整的模型训练套路

20.1 代码展示

  train.py

import torch
import torchvision
from torch.utils.tensorboard import SummaryWriter
from model import *    # 引入 model 中的所有内容(需在同一文件夹下)
from torch import nn
from torch.utils.data import DataLoader

# 准备数据集
train_data = torchvision.datasets.CIFAR10(root="../data", train=True, transform=torchvision.transforms.ToTensor(),
                                          download=True)
test_data = torchvision.datasets.CIFAR10(root="../data", train=False, transform=torchvision.transforms.ToTensor(),
                                          download=True)

# length 长度
train_data_size = len(train_data)
test_data_size = len(test_data)
# 若 train_data_size=10,训练数据集的长度为：10
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
avlon = Avlon()

# 损失函数
loss_fn = nn.CrossEntropyLoss()

# 优化器
# 1e-2=1 × (10)^(-2) = 1/100 = 0.01
learning_rate = 1e-2    # 便于修改
optimizer = torch.optim.SGD(avlon.parameters(), learning_rate)

# 设置训练网络的一些参数
# 记录训练次数
total_train_step = 0
# 记录测试的次数
total_test_step = 0
# 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
    # avlon.train()    # 训练模式只对 Dropout, BatchNorm 等部分层有效
    for data in train_dataloader:
        imgs, targets = data    # 针对批量数据进行操作
        outputs = avlon(imgs)    # imgs 数据特征
        loss = loss_fn(outputs, targets)    # targets 样本对应的真实标签

        optimizer.zero_grad()    # 梯度清零
        loss.backward()    # 获取梯度
        optimizer.step()    # 对每一参数进行调优

        total_train_step = total_train_step + 1
        if total_train_step % 100 == 0:    # 每逢 100 打印一次
            print("训练次数： {}, loss: {}".format(total_train_step, loss.item()))
            writer.add_scalar("train_loss", loss.item(), total_train_step)

    # 测试开始
    avlon.eval()    # 测试模式同样只对部分层有效
    total_test_loss = 0
    total_accuracy = 0
    with torch.no_grad():    # 测试中不调优
        for data in test_dataloader:
            imgs, targets = data
            outputs = avlon(imgs)
            loss = loss_fn(outputs, targets)
            total_test_loss = total_test_loss + loss.item()
            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Accuracy: {}".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_test_step)
    total_test_step = total_test_step + 1

    torch.save(avlon, "avlon_{}.pth".format(i))
    # torch.save(avlon.state_dict(), "avlon_{}.pth".format(i))
    print("模型已保存")

writer.close()

  model.py

import torch
from torch import nn


# 搭建神经网络
class Avlon(nn.Module):
    def __init__(self):
        super().__init__()
        self.model = nn.Sequential(
            nn.Conv2d(3, 32, 5, 1, 2),
            nn.MaxPool2d(2),
            nn.Conv2d(32, 32, 5, 1, 2),
            nn.MaxPool2d(2),
            nn.Conv2d(32, 64, 5, 1, 2),
            nn.MaxPool2d(2),
            nn.Flatten(),
            nn.Linear(64*4*4, 64)
        )

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


# 测试正确性
if __name__ == '__main__':    # 输入"main"
    avlon = Avlon()
    input = torch.ones((64, 3, 32, 32))
    output = avlon(input)
    print(output.shape)
    print(output)

20.2 代码细节

  # “Ctrl + F” 打开搜索框

20.2.1 item() 的作用

  在"Scratches and Consoles" 下新建一 python 文件

import torch
a = torch.tensor(5)
print(a)
print(a.item())

  运行结果：
  tensor(5)
  5
  可见 item() 可将 tensor 数据转化成数字输出

20.2.2 正确率的计算

  同样在"Scratches and Consoles" 下新建一 python 文件

import torch

outputs = torch.tensor([[0.1, 0.2],
                       [0.3, 0.4]])

print(outputs.argmax(1))    # argmax() 获取每个样本输出的预测值; 1 表示返回行上最大值所在的列索引
preds = outputs.argmax(1)
targets = torch.tensor([0, 1])    # 定义两个样本的目标值
print((preds == targets).sum())    # 比较预测值与目标值; sum() 统计预测正确的数量

  运行结果：
  tensor([1, 1])
  tensor(1)

  本人的理解