p30利用GPU训练1

只需要在原来代码中修改几处就可以

GPU训练实现有两种方式

第一种，写法并不太好

# 准备数据集
import torch.optim
import torchvision
from torch import nn
from torch.utils.data import DataLoader
from torch.utils.tensorboard import SummaryWriter

# from model import *

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
train_data_size = len(train_data)
test_data_siz = len(test_data)
print(f"训练数据集的长度为：{train_data_size}")
print(f"测试数据集的长度为：{test_data_siz}")

# 利用DataLoader加载
train_data_loader = DataLoader(train_data, batch_size=256)
test_data_loader = DataLoader(test_data, batch_size=256)

# 创建网络模型
class Lixinyu(nn.Module):
    def __init__(self):
        super(Lixinyu, 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(1024, 64),
            nn.Linear(64, 10)
        )
    def forward(self, x):
        x = self.model(x)
        return x

lixinyu = Lixinyu()
lixinyu = lixinyu.cuda()  ########

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

# 优化器
# learning_rate = 0.01
learning_rate = 1e-2
optimizer = torch.optim.SGD(lixinyu.parameters(), learning_rate)

# 设置训练网络参数
# 记录训练的次数
total_train_step = 0
# 记录测试的次数
total_test_step = 0
# 训练的轮数
epoch = 5

# 添加tensorboard
writer = SummaryWriter("p28")

for i in range(epoch):
    print(f"--------------------第{i+1}轮训练开始了-----------------")

    # 训练步骤开始
    lixinyu.train()   ####
    for data in train_data_loader:
        imgs, targets = data
        imgs = imgs.cuda()   ##############
        targets = targets.cuda()   ###########
        output = lixinyu(imgs)
        loss = loss_fn(output, targets)

        optimizer.zero_grad()
        loss.backward()
        optimizer.step()
        total_train_step += 1
        if total_train_step % 100 == 0:
            print(f"训练次数{total_train_step} 损失值{loss.item()}")  # loss.item()标准写法，这里直接写loss也行
            writer.add_scalar("train_loss", loss.item(), total_train_step)

    # 测试步骤开始
    lixinyu.eval()  #####
    total_test_loss = 0
    total_accuracy = 0
    with torch.no_grad():
        for data in test_data_loader:
            imgs, targets = data
            imgs = imgs.cuda()  ##############
            targets = targets.cuda()  ###########
            outputs = lixinyu(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(f'整体测试loss为：{total_test_loss}')
    print(f'整体测试数据集上的正确率：{total_accuracy/test_data_siz}')
    writer.add_scalar("total_test_loss", total_test_loss, total_test_step)
    writer.add_scalar("total_accuracy", total_accuracy/test_data_siz, total_test_step)
    total_test_step = total_test_step + 1
    torch.save(lixinyu, f"lixinyu_epoch{i}.pth")
    # torch.save(lixinyu.state_dict(), f"lixinyu_{i}.pth")
    print("模型已保存")
writer.close()

第二种

首先进行判断torch.cuda.is_available()

 if torch.cuda.is_available():
            imgs = imgs.cuda()   ##############
            targets = targets.cuda()   ###########

加入time模块比较GPU与CPU的时间

# 准备数据集
import torch.optim
import torchvision
from torch import nn
from torch.utils.data import DataLoader
from torch.utils.tensorboard import SummaryWriter
import time

# from model import *

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
train_data_size = len(train_data)
test_data_siz = len(test_data)
print(f"训练数据集的长度为：{train_data_size}")
print(f"测试数据集的长度为：{test_data_siz}")

# 利用DataLoader加载
train_data_loader = DataLoader(train_data, batch_size=64)
test_data_loader = DataLoader(test_data, batch_size=64)

# 创建网络模型
class Lixinyu(nn.Module):
    def __init__(self):
        super(Lixinyu, 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(1024, 64),
            nn.Linear(64, 10)
        )
    def forward(self, x):
        x = self.model(x)
        return x

lixinyu = Lixinyu()
if torch.cuda.is_available():
    lixinyu = lixinyu.cuda()  ########

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

# 优化器
# learning_rate = 0.01
learning_rate = 1e-2
optimizer = torch.optim.SGD(lixinyu.parameters(), learning_rate)

# 设置训练网络参数
# 记录训练的次数
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("p28")
start_time = time.time()

for i in range(epoch):
    print(f"--------------------第{i+1}轮训练开始了-----------------")

    # 训练步骤开始
    lixinyu.train()   ####
    for data in train_data_loader:
        imgs, targets = data
        if torch.cuda.is_available():
            imgs = imgs.cuda()   ##############
            targets = targets.cuda()   ###########
        output = lixinyu(imgs)
        loss = loss_fn(output, targets)

        optimizer.zero_grad()
        loss.backward()
        optimizer.step()
        total_train_step += 1
        if total_train_step % 100 == 0:
            end_time = time.time()
            print(f"训练次数{total_train_step} 损失值{loss.item()}耗费时间{end_time-start_time}")  # loss.item()标准写法，这里直接写loss也行
            writer.add_scalar("train_loss", loss.item(), total_train_step)

    # 测试步骤开始
    lixinyu.eval()  #####
    total_test_loss = 0
    total_accuracy = 0
    with torch.no_grad():
        for data in test_data_loader:
            imgs, targets = data
            if torch.cuda.is_available():
                imgs = imgs.cuda()  ##############
                targets = targets.cuda()  ###########
            outputs = lixinyu(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(f'整体测试loss为：{total_test_loss}')
    print(f'整体测试数据集上的正确率：{total_accuracy/test_data_siz}')
    writer.add_scalar("total_test_loss", total_test_loss, total_test_step)
    writer.add_scalar("total_accuracy", total_accuracy/test_data_siz, total_test_step)
    total_test_step = total_test_step + 1
    torch.save(lixinyu, f"lixinyu_epoch{i}.pth")
    # torch.save(lixinyu.state_dict(), f"lixinyu_{i}.pth")
    print("模型已保存")

writer.close()

没有GPU的可以使用googleColab