Mac上使用GPU加速训练模型

前言

上一篇文章中我介绍了使用pytorch的一个完整模型训练套路，其中没有使用gpu，如果要使用gpu的话，win上我们可以使用cuda，mac上可以使用mps，而我自己是mac电脑，需要进行如下修改。

使用GPU

import torchvision
from torch import nn
from torch.utils.data import DataLoader
from torch.utils.tensorboard import SummaryWriter
from model2 import *
import time

# 创建数据集
train_data = torchvision.datasets.CIFAR10("./source", train=True,
                                          transform=torchvision.transforms.ToTensor(), download=True)
test_data = torchvision.datasets.CIFAR10("./source", train=False,
                                          transform=torchvision.transforms.ToTensor(), download=True)

# 加载数据集
train_loader = DataLoader(train_data, batch_size=64)
test_loader = DataLoader(test_data, batch_size=64)

# 查看数据集长度
train_data_size = len(train_data)
test_data_size = len(test_data)
print(f"训练数据集的大小为{train_data_size}")
print(f"测试数据集的大小为{test_data_size}")

# 创建网络模型 搭建神经网络
# class Aniu(nn.Module):
#     def __init__(self):
#         super(Aniu, 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

# 定义训练的设备
device = torch.device("mps" if torch.backends.mps.is_available() else "cpu")
# 创建神经网络模型
aniu = Aniu()
aniu = aniu.to(device)

# 定义损失函数
loss_fn = nn.CrossEntropyLoss()
loss_fn = loss_fn.to(device)
# 定义优化器
learning_rate = 1e-2
optimizer = torch.optim.SGD(aniu.parameters(), lr=learning_rate)

# 训练网络:

# 设置训练网络的一些参数:
# 总共训练次数
total_train_step = 0
# 总共测试次数
total_test_step = 0
# 总轮次
epoch = 50

# 添加 tensorboard 以便观察
writer = SummaryWriter("./log_train2")
start_time = time.time()
for i in range(epoch):
    print(f"------------第{i+1}轮训练开始------------")

    # 训练开始
    aniu.train()
    for data in train_loader:
        imgs, targets = data
        imgs = imgs.to(device)
        targets = targets.to(device)
        output = aniu(imgs)
        loss = loss_fn(output, targets)

        # 优化器优化模型
        optimizer.zero_grad() # 优化器梯度清零
        loss.backward()
        optimizer.step()

        total_train_step += 1
        if total_train_step % 100 == 0 :
            end_time = time.time()
            print(end_time - start_time)
            print(f"训练次数{total_train_step}，Loss:{loss.item()}")
            writer.add_scalar("train_loss", loss.item(), total_train_step)

    # 测试步骤开始：
    aniu.eval()
    total_test_loss = 0

    # 整体正确的个数
    total_accuracy = 0
    with torch.no_grad():
        for data in test_loader:
            imgs, targets = data
            imgs = imgs.to(device)
            targets = targets.to(device)
            output = aniu(imgs)
            loss = loss_fn(output, targets)
            total_test_loss += loss.item()
            accuracy = (output.argmax(1) == targets).sum()
            total_accuracy = total_accuracy + accuracy

    print(f"整体测试集上的Loss为{total_test_loss}")
    print(f"整体测试集上的正确率：{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 += 1

    torch.save(aniu, f"aniu_{i}.pth")
    print("模型已保存")

writer.close()

总的来说就是添加了几行代码：

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

aniu = aniu.to(device)

loss_fn = loss_fn.to(device)

output = aniu(imgs)

loss = loss_fn(output, targets)

速度大概快了10几倍。