如何采用GPU训练

如何采用GPU训练

方法1：对网络模型，数据（数据、标注），损失函数调用.cuda()即可

import torch
import torchvision
from torch import nn
from torch.utils.data import DataLoader
from torch.utils.tensorboard import SummaryWriter
import time
# 将model文件夹中有的东西都引入过来
# from model import *

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

# 看一下训练数据集和测试数据集有多少张 len-length 长度
train_data_size = len(train_data)
test_data_size = len(test_data)

# python中常用的写法：字符串格式化
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))

# 用DataLoader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 Peipei(nn.Module):
    def __init__(self):
        super(Peipei, 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


# 创建网络模型
peipei = Peipei()
if torch.cuda.is_available():
    peipei = peipei.cuda()

# 创建损失函数
loss_fn = nn.CrossEntropyLoss()
if torch.cuda.is_available():
    loss_fn = loss_fn.cuda()

# 定义优化器
# 1e-2 = 0.01
learning_rate = 1e-2
optimizer = torch.optim.SGD(peipei.parameters(), lr=learning_rate)

# 设置训练网络的一些参数
# 记录训练次数
total_train_step = 0
# 记录测试次数
total_test_step = 0
# 训练的轮数
epoch = 10

# 添加tensorboard
writer = SummaryWriter("logs_train")

start_time = time.time()

# i从0-9
for i in range(epoch):
    print("--------------------第{}轮训练开始--------------------".format(i + 1))

    # 训练步骤开始
    # 使模型进入训练状态，但只对特定层（Dropout,BatchNorm层）起作用
    peipei.train()
    for data in train_dataloader:
        imgs, targets = data
        if torch.cuda.is_available():
            imgs = imgs.cuda()
            targets = targets.cuda()
        outputs = peipei(imgs)
        # 计算损失函数
        loss = loss_fn(outputs, targets)

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

        total_train_step = total_train_step + 1
        if total_train_step % 100 == 0:
            end_time = time.time()
            print(end_time-start_time)
            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
    # 使模型进入验证状态，但只对特定层（Dropout,BatchNorm层）起作用
    peipei.eval()
    total_test_loss = 0
    totel_accuracy = 0
    with torch.no_grad():
        for data in test_dataloader:
            imgs, targets = data
            if torch.cuda.is_available():
                imgs = imgs.cuda()
                targets = targets.cuda()
            outputs = peipei(imgs)
            loss = loss_fn(outputs, targets)
            # 计算整体测试集损失函数
            total_test_loss = total_test_loss + loss.item()
            # 计算整体正确率
            accuracy = (outputs.argmax(1) == targets).sum()
            totel_accuracy = totel_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(totel_accuracy / test_data_size))
    writer.add_scalar("test_loss", total_test_loss, total_test_step)
    writer.add_scalar("test_accuracy", totel_accuracy / test_data_size, total_test_step)
    total_test_step = total_test_step + 1

    # 对每轮训练完的模型保存
    torch.save(peipei, "peipei_{}.pth".format(i))
    torch.save(peipei.state_dict(), "peipei_{}.pth".format(i))

writer.close()

CPU

GPU

方法2

import torchvision
from torch import nn
from torch.utils.data import DataLoader
from torch.utils.tensorboard import SummaryWriter
import time
# 将model文件夹中有的东西都引入过来
# from model import *

# 定义训练的设备
device = torch.device("cuda" if torch.cuda.is_available() else "cpu")

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

# 看一下训练数据集和测试数据集有多少张 len-length 长度
train_data_size = len(train_data)
test_data_size = len(test_data)

# python中常用的写法：字符串格式化
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))

# 用DataLoader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 Peipei(nn.Module):
    def __init__(self):
        super(Peipei, 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


# 创建网络模型
peipei = Peipei()
peipei = peipei.to(device)

# 创建损失函数
loss_fn = nn.CrossEntropyLoss()
loss_fn = loss_fn.to(device)

# 定义优化器
# 1e-2 = 0.01
learning_rate = 1e-2
optimizer = torch.optim.SGD(peipei.parameters(), lr=learning_rate)

# 设置训练网络的一些参数
# 记录训练次数
total_train_step = 0
# 记录测试次数
total_test_step = 0
# 训练的轮数
epoch = 10

# 添加tensorboard
writer = SummaryWriter("logs_train")

start_time = time.time()

# i从0-9
for i in range(epoch):
    print("--------------------第{}轮训练开始--------------------".format(i + 1))

    # 训练步骤开始
    # 使模型进入训练状态，但只对特定层（Dropout,BatchNorm层）起作用
    peipei.train()
    for data in train_dataloader:
        imgs, targets = data
        imgs = imgs.to(device)
        targets = targets.to(device)
        outputs = peipei(imgs)
        # 计算损失函数
        loss = loss_fn(outputs, targets)

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

        total_train_step = total_train_step + 1
        if total_train_step % 100 == 0:
            end_time = time.time()
            print(end_time - start_time)
            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
    # 使模型进入验证状态，但只对特定层（Dropout,BatchNorm层）起作用
    peipei.eval()
    total_test_loss = 0
    totel_accuracy = 0
    with torch.no_grad():
        for data in test_dataloader:
            imgs, targets = data
            imgs = imgs.to(device)
            targets = targets.to(device)
            outputs = peipei(imgs)
            loss = loss_fn(outputs, targets)
            # 计算整体测试集损失函数
            total_test_loss = total_test_loss + loss.item()
            # 计算整体正确率
            accuracy = (outputs.argmax(1) == targets).sum()
            totel_accuracy = totel_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(totel_accuracy / test_data_size))
    writer.add_scalar("test_loss", total_test_loss, total_test_step)
    writer.add_scalar("test_accuracy", totel_accuracy / test_data_size, total_test_step)
    total_test_step = total_test_step + 1

    # 对每轮训练完的模型保存
    torch.save(peipei, "peipei_{}.pth".format(i))
    torch.save(peipei.state_dict(), "peipei_{}.pth".format(i))

writer.close()

GPU

CPU

完整的模型验证（测试，demo）套路：利用已经训练好的模型，然后给它提供输入（对外应用）

GitHub常见代码