PyTorch深度学习（12）利用GPU训练及模型验证

一、利用GPU训练

方式一  .cuda()

网络模型

数据（输入，标注）

损失函数

调用  .cuda()  

方式二  .to(device)

.to(device) 

device = torch.device( "cpu" )

torch.device( "cuda" )

电脑中有多张显卡

torch.device( "cuda:0" )

torch.device( "cuda:1" )

只有数据、图片、标注需要.to() 后进行赋值

device = torch.device("cuda" if torch.cuda.is_available() else "cpu")

如果cuda.is_available() 可用，则使用cuda，否则使用cpu

具体代码

# 网络数据   数据（输入、标注）  损失函数   .cuda()
import torch
from torch import nn
import torchvision
from torch.utils.data import DataLoader
from torch.utils.tensorboard import SummaryWriter
import time

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

# 使用指定设备运行
# device = torch.device("cpu")
# device = torch.device("cuda:0")
# device = torch.device("cuda" if torch.cuda.is_available() else "cpu")

# length长度
train_data_size = len(train_data)
test_data_size = len(test_data)
print("训练集数据长度：{}".format(train_data_size))
print("测试集数据长度：{}".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)

# 创建神经网络模型
class TestNN(nn.Module):
    def __init__(self):
        super(TestNN, self).__init__()
        self.model1 = nn.Sequential(
            nn.Conv2d(3, 32, 5, 1, padding=2),
            nn.MaxPool2d(2),
            nn.Conv2d(32, 32, 5, 1, padding=2),
            nn.MaxPool2d(2),
            nn.Conv2d(32, 64, 5, 1, padding=2),
            nn.MaxPool2d(2),
            nn.Flatten(),
            nn.Linear(64 * 4 * 4, 64),
            nn.Linear(64, 10)
        )

    def forward(self, x):
        x = self.model1(x)
        return x


tnn = TestNN()
# tnn.to(device)
if torch.cuda.is_available():
    tnn = tnn.cuda()

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

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

# 设置训练网络的一些参数
# 记录训练的次数
total_train_step = 0
# 记录测试的次数
total_test_step = 0

# 添加tensorboard
writer = SummaryWriter("logs")

epoch = 10
start_time = time.time()
for i in range(epoch):
    print("---第{}轮训练开始---".format(i + 1))

    # 训练步骤开始
    tnn.train()
    for data in train_dataLoader:
        images, targets = data
        # images = images.to(device)
        # targets = targets.to(device)
        if torch.cuda.is_available():
            images = images.cuda()
            targets = targets.cuda()
        outputs = tnn(images)
        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
    tnn.eval()
    total_test_loss = 0
    total_accuracy = 0
    with torch.no_grad():
        for data in test_dataLoader:
            images, targets = data
            if torch.cuda.is_available():
                images = images.cuda()
                targets = targets.cuda()
            output = tnn(images)
            loss = loss_fn(output, targets)
            total_test_loss = total_test_loss + loss
            accuracy = (output.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_test_step)
    total_test_step = total_test_step + 1

    torch.save(tnn, "tnn_{}.pth".format(i))
    print("模型已保存")

writer.close()

二、模型验证

模型验证方法

完整的模型验证（测试、demo），利用已经训练好的模型，给其提供输入

torch.load("加载训练好的数据集", map_location="cpu")  当输入类型为cpu报错时使用

使用中注意图片的大小是否符合训练集和测试集的要求

model.eval()  

with torch.no_grad():

       output = model(image)

对指定图片测试，判断图片的类别

注意：如果图片是png格式，因为png是4通道，要转换image=convert('RGB')

具体代码

import torchvision
from PIL import Image
from test_model import *

image = Image.open("image/plane.jpg")
print(image)

transform = torchvision.transforms.Compose([
    torchvision.transforms.ToTensor(),
    torchvision.transforms.Resize((32, 32))
])

# 如果是png格式，因为Png是4通道，要转换image = image.convert('RGB')

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

model = torch.load("tnn_29.pth", map_location="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))