pytorch中模型训练套路

项目结构如下

首先我们需要定义一个网络，定义网络模型 model.py文件

import torch
from torch import nn


class Train(nn.Module):
    def __init__(self):
        super(Train, self).__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),
            nn.Linear(64, 10)
        )

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


if __name__ == '__main__':
    train = Train()
    input = torch.ones((64, 3, 32, 32))
    output = train(input)
    print(output.shape)

再接着定义train文件

import torch
import torchvision
from torch import nn
from torch.utils.data import DataLoader
from torch.utils.tensorboard import SummaryWriter

from model import *

# 对数据集进行导入操作
train_data = torchvision.datasets.CIFAR10(root="./dataset1", train=True, transform=torchvision.transforms.ToTensor())

test_data = torchvision.datasets.CIFAR10(root="./dataset1", train=False, transform=torchvision.transforms.ToTensor())

train_data_size = len(train_data)
test_data_size = len(test_data)
print(train_data_size)
print(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
train = Train()

#使用gpu的方式
if torch.cuda.is_available():
    train = train.cuda()

# 损失函数
loss_fn = nn.CrossEntropyLoss()
loss_fn = loss_fn.cuda()

# 优化器
optimizer = torch.optim.SGD(train.parameters(), lr=0.01)

# 设置训练网络的一些参数
total_train = 0
total_test = 0
epoch = 10

writer = SummaryWriter("./logs_train")

for i in range(epoch):
    print(f"-----------第{i + 1}轮------------------：")
    for data in train_dataloader:
        imgs, targets = data
        imgs = imgs.cuda()
        targets = targets.cuda()
        output = train(imgs)
        loss = loss_fn(output, targets)

        # 优化器优化模型
        optimizer.zero_grad()
        loss.backward()
        optimizer.step()

        total_train += 1
        if total_train % 100 == 0:
            print("训练次数:{},loss:{}".format(total_train, loss))
            writer.add_scalar("train_loss", loss.item(), total_train)

    # 测试步骤开始
    total_test_loss = 0
    total_acc = 0
    with torch.no_grad():
        for data in test_dataloader:
            imgs, targets = data
            imgs = imgs.cuda()
            targets = targets.cuda()
            output = train(imgs)
            loss = loss_fn(output, targets)
            total_test_loss = total_test_loss + loss
            acc = (output.argmax(1) == targets).sum()
            total_acc = acc + total_acc
    print("整体测试即的loss:{}".format(total_test_loss))
    print("整体测试即的acc:{}".format(total_acc / test_data_size))
    writer.add_scalar("test_loss", total_test_loss, total_test)
    writer.add_scalar("test_acc", total_acc / test_data_size, total_test)
    total_test += 1

	# 跑出来的权重文件保存在这个目录下
    torch.save(train, "weight_{}.pth".format(i))
    print("模型已保存")
writer.close()

测试文件test.py

import torch
import torchvision
from PIL import Image
from torch import nn

image_path = "./imgs/dog.jpg"
image = Image.open(image_path)
print(image)
image = image.convert('RGB')
transform = torchvision.transforms.Compose([torchvision.transforms.Resize((32, 32)),
                                            torchvision.transforms.ToTensor()])

image = transform(image)
print(image.shape)


class Tudui(nn.Module):
    def __init__(self):
        super(Tudui, self).__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),
            nn.Linear(64, 10)
        )

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


model = torch.load("weight_1.pth", map_location=torch.device('cpu'))
print(model)
image = torch.reshape(image, (1, 3, 32, 32))
model.eval()
with torch.no_grad():
    output = model(image)
print(output)

print(output.argmax(1))