样例代码

构建神经网络

我们先构建一个简单的神经网络。

import time
import torch
import torch.nn as nn
import torch_npu
from torch.utils.data import Dataset, DataLoader
import torchvision

device = torch.device('npu:0')    # 用户请自行定义训练设备

# 定义一个简单的神经网络
class CNN(nn.Module):
    def __init__(self):
        super(CNN, self).__init__()
        self.net = nn.Sequential(
            nn.Conv2d(in_channels=1, out_channels=16,
                      kernel_size=(3, 3),
                      stride=(1, 1),
                      padding=1),
            nn.MaxPool2d(kernel_size=2),
            nn.Conv2d(16, 32, 3, 1, 1),
            nn.MaxPool2d(2),
            nn.Flatten(),
            nn.Linear(32*7*7, 16),
            nn.ReLU(),
            nn.Linear(16, 10)
        )
    def forward(self, x):
        return self.net(x)

定义训练参数和数据

这一部分我们从torchvision中获取训练数据集，设置训练相关的参数batch_size、epochs。

train_data = torchvision.datasets.MNIST(
    root='mnist',
    download=True,
    train=True,
    transform=torchvision.transforms.ToTensor()
)

batch_size = 64
model = CNN().to(device)
train_dataloader = DataLoader(train_data, batch_size=batch_size)    # 定义DataLoader
loss_func = nn.CrossEntropyLoss().to(device)    # 定义损失函数
optimizer = torch.optim.SGD(model.parameters(), lr=0.1)    # 定义优化器
epochs = 10

进行训练

for epo in range(epochs):
    for imgs, labels in train_dataloader:
        start_time = time.time()
        imgs = imgs.to(device)
        labels = labels.to(device)
        outputs = model(imgs)    # 前向计算
        loss = loss_func(outputs, labels)    # 损失函数计算
        optimizer.zero_grad()
        loss.backward()
        optimizer.step()