《动手学深度学习》7.5 批量归一化-从零开始实现

《动手学深度学习》7.5 批量归一化-从零开始实现

• 导入功能包

import torch
from torch import nn
import MyFunction as MF

• 定义归一化操作

# 归一化操作
def batch_norm(X, gamma, beta, moving_mean, moving_var, eps, momentum):
    # 通过 `is_grad_enabled` 来判断当前模式是训练模式还是预测模式
    if not torch.is_grad_enabled():
        # 如果是在预测模式下，直接使用传入的移动平均所得的均值和方差
        X_hat = (X - moving_mean) / torch.sqrt(moving_var + eps)
    else:
        assert len(X.shape) in (2, 4)
        if len(X.shape) == 2:
            # 使用全连接层的情况，计算特征维上的均值和方差
            mean = X.mean(dim=0)
            var = ((X - mean)**2).mean(dim=0)
        else:
            # 使用二维卷积层的情况，计算通道维上（axis=1）的均值和方差。
            # 这里我们需要保持X的形状以便后面可以做广播运算
            mean = X.mean(dim=(0, 2, 3), keepdim=True)
            var = ((X - mean)**2).mean(dim=(0, 2, 3), keepdim=True)
        # 训练模式下，用当前的均值和方差做标准化
        X_hat = (X - mean) / torch.sqrt(var + eps)
        # 更新移动平均的均值和方差
        moving_mean = momentum * moving_mean + (1.0 - momentum) * mean
        moving_var = momentum * moving_var + (1.0 - momentum) * var
    Y = gamma * X_hat + beta  # 缩放和移位
    return Y, moving_mean.data, moving_var.data

• 定义归一化层

# 归一化层
class BatchNorm(nn.Module):
    # `num_features`：完全连接层的输出数量或卷积层的输出通道数。
    # `num_dims`：2表示完全连接层，4表示卷积层
    def __init__(self, num_features, num_dims):
        super().__init__()
        if num_dims == 2:
            shape = (1, num_features)
        else:
            shape = (1, num_features, 1, 1)
        # 参与求梯度和迭代的拉伸和偏移参数，分别初始化成1和0
        self.gamma = nn.Parameter(torch.ones(shape))
        self.beta = nn.Parameter(torch.zeros(shape))
        # 非模型参数的变量初始化为0和1
        self.moving_mean = torch.zeros(shape)
        self.moving_var = torch.ones(shape)

    def forward(self, X):
        # 如果 `X` 不在内存上，将 `moving_mean` 和 `moving_var`
        # 复制到 `X` 所在显存上
        if self.moving_mean.device != X.device:
            self.moving_mean = self.moving_mean.to(X.device)
            self.moving_var = self.moving_var.to(X.device)
        # 保存更新过的 `moving_mean` 和 `moving_var`
        Y, self.moving_mean, self.moving_var = batch_norm(
            X, self.gamma, self.beta, self.moving_mean, self.moving_var,
            eps=1e-5, momentum=0.9)
        return Y

• 定义模型

# 定义模型
net = nn.Sequential(nn.Conv2d(1, 6, kernel_size=5), BatchNorm(6, num_dims=4),
                    nn.Sigmoid(), nn.MaxPool2d(kernel_size=2, stride=2),
                    nn.Conv2d(6, 16, kernel_size=5), BatchNorm(16, num_dims=4),
                    nn.Sigmoid(), nn.MaxPool2d(kernel_size=2, stride=2),
                    nn.Flatten(), nn.Linear(16 * 4 * 4, 120),
                    BatchNorm(120, num_dims=2), nn.Sigmoid(),
                    nn.Linear(120, 84), BatchNorm(84, num_dims=2),
                    nn.Sigmoid(), nn.Linear(84, 10))

• 训练

# 参数设置
lr, num_epochs, batch_size = 1.0, 10, 256
# 指定设备
device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
# 获取数据
train_iter, test_iter = MF.load_data_fashion_mnist(batch_size)
# 训练
MF.train_ch7_BN(net, train_iter, test_iter, num_epochs, lr, device=device)

• 训练结果
  

预测部分

• 导入相关包

import torch
from torch import nn
import MyFunction as MF
import os
import matplotlib.pyplot as plt

• 模型

# 归一化操作
def batch_norm(X, gamma, beta, moving_mean, moving_var, eps, momentum):
    # 通过 `is_grad_enabled` 来判断当前模式是训练模式还是预测模式
    if not torch.is_grad_enabled():

        # 如果是在预测模式下，直接使用传入的移动平均所得的均值和方差
        X_hat = (X - moving_mean) / torch.sqrt(moving_var + eps)
    else:
        assert len(X.shape) in (2, 4)
        if len(X.shape) == 2:
            # 使用全连接层的情况，计算特征维上的均值和方差
            mean = X.mean(dim=0)
            var = ((X - mean)**2).mean(dim=0)
        else:
            # 使用二维卷积层的情况，计算通道维上（axis=1）的均值和方差。
            # 这里我们需要保持X的形状以便后面可以做广播运算
            mean = X.mean(dim=(0, 2, 3), keepdim=True)
            var = ((X - mean)**2).mean(dim=(0, 2, 3), keepdim=True)
        # 训练模式下，用当前的均值和方差做标准化
        X_hat = (X - mean) / torch.sqrt(var + eps)
        # 更新移动平均的均值和方差
        moving_mean = momentum * moving_mean + (1.0 - momentum) * mean
        moving_var = momentum * moving_var + (1.0 - momentum) * var
    Y = gamma * X_hat + beta  # 缩放和移位
    return Y, moving_mean.data, moving_var.data

# 归一化层
class BatchNorm(nn.Module):
    # `num_features`：完全连接层的输出数量或卷积层的输出通道数。
    # `num_dims`：2表示完全连接层，4表示卷积层
    def __init__(self, num_features, num_dims):
        super().__init__()
        if num_dims == 2:
            shape = (1, num_features)
        else:
            shape = (1, num_features, 1, 1)
        # 参与求梯度和迭代的拉伸和偏移参数，分别初始化成1和0
        self.gamma = nn.Parameter(torch.ones(shape))
        self.beta = nn.Parameter(torch.zeros(shape))
        # 非模型参数的变量初始化为0和1
        self.moving_mean = torch.zeros(shape)
        self.moving_var = torch.ones(shape)

    def forward(self, X):
        # 如果 `X` 不在内存上，将 `moving_mean` 和 `moving_var`
        # 复制到 `X` 所在显存上
        if self.moving_mean.device != X.device:
            self.moving_mean = self.moving_mean.to(X.device)
            self.moving_var = self.moving_var.to(X.device)
        # 保存更新过的 `moving_mean` 和 `moving_var`
        Y, self.moving_mean, self.moving_var = batch_norm(
            X, self.gamma, self.beta, self.moving_mean, self.moving_var,
            eps=1e-5, momentum=0.9)
        return Y

# 定义模型
net = nn.Sequential(nn.Conv2d(1, 6, kernel_size=5), BatchNorm(6, num_dims=4),
                    nn.Sigmoid(), nn.MaxPool2d(kernel_size=2, stride=2),
                    nn.Conv2d(6, 16,kernel_size=5), BatchNorm(16, num_dims=4),
                    nn.Sigmoid(), nn.MaxPool2d(kernel_size=2, stride=2),
                    nn.Flatten(), nn.Linear(16 * 4 * 4, 120),
                    BatchNorm(120, num_dims=2), nn.Sigmoid(),
                    nn.Linear(120, 84), BatchNorm(84, num_dims=2),
                    nn.Sigmoid(), nn.Linear(84, 10))

• 加载模型参数

# 模型参数位置
filename = r'./data/BN-9.pth'
# 判断模型参数是否存在
if os.path.exists(filename):
    # 加载模型参数
    net.load_state_dict(torch.load(filename))
else:
    print(f"No such file or directory: '{filename}'")

• 预测

# 预测
lr, num_epochs, batch_size = 1.0, 10, 256
device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
train_iter, test_iter = MF.load_data_fashion_mnist(batch_size)
MF.predict_ch7_BN(net, test_iter)
plt.show()

• 预测结果
  

MyFunction包里定义的相关函数

# 训练函数
import datetime
def train_ch7_BN(net, train_iter, test_iter, num_epochs, lr, device):
    print("Start Training...")
    nowtime = datetime.datetime.now().strftime('%Y-%m-%d %H:%M:%S')
    print("==========" * 8 + "%s" % nowtime)

    def init_weights(m):
        if type(m) == nn.Linear or type(m) == nn.Conv2d:
            nn.init.xavier_uniform_(m.weight)

    net.apply(init_weights)
    print(f'training on {device}:{torch.cuda.get_device_name()}')
    net.to(device)
    optimizer = torch.optim.SGD(net.parameters(), lr=lr)
    loss = nn.CrossEntropyLoss()

    timer = MF.Timer()
    for epoch in range(num_epochs):
        train_l_sum, train_acc_sum, n = 0.0, 0.0, 0
        net.train()
        with tqdm(train_iter) as t:
            for X, y in t:
                timer.start()
                optimizer.zero_grad()
                X, y = X.to(device), y.to(device)
                y_hat = net(X)
                l = loss(y_hat, y)
                l.backward()
                optimizer.step()
                train_l_sum += l.item() * X.shape[0]
                train_acc_sum += MF.accuracy(y_hat,y)
                n += y.shape[0]
                train_l = train_l_sum / n
                train_acc = train_acc_sum / n
                timer.stop()
                # 设置进度条左边显示的信息
                t.set_description(f"epoch:{epoch}")
                # 设置进度条右边显示的信息
                t.set_postfix(loss="%.3f" % train_l, train_acc="%.3f" % train_acc, time="%.3f sec" % timer.stop())

            torch.save(net.state_dict(),"./data/BN-%d.pth" %(epoch))
            test_acc = MF.evaluate_accuracy_BN(net, test_iter)
    print(f'epoch:{epoch+1},loss {train_l:.3f}, train_acc {train_acc:.3f}, test_acc {test_acc:.3f}, {timer.stop()} sec')
    print(f'{n* num_epochs / timer.sum():.1f} examples/sec '
          f'on {str(device)}')

    nowtime = datetime.datetime.now().strftime('%Y-%m-%d %H:%M:%S')

    print("==========" * 8 + "%s" % nowtime)
    print('Finished Training...')

# 模型评估函数
def evaluate_accuracy_BN(net, data_iter, device=None):
    acc_sum, n = 0.0, 0
    if isinstance(net, torch.nn.Module):
        net.eval()  # 设置为评估模式
        if not device:  # 如果没指定device就使用net的device
            device = next(iter(net.parameters())).device
    # 正确预测的数量，总预测的数量
    for X, y in data_iter:
        if isinstance(X, list):
            # BERT微调所需的
            X = [x.to(device) for x in X]
        else:
            X = X.to(device)
        y = y.to(device)
        acc_sum += MF.accuracy(net(X), y)
        n += y.shape[0]
        result = acc_sum / n
    return result
# 预测函数
def predict_ch7_BN(net, test_iter, n=6):
    if isinstance(net, torch.nn.Module):
        net.eval()  # 设置为评估模式
    """预测标签"""
    device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
    for X, y in test_iter:
        h, w = X.shape[-2:]
        X.to(device)
        y.to(device)
        trues = MF.get_fashion_mnist_labels(y)
        preds = MF.get_fashion_mnist_labels(net(X).argmax(axis=1))
        titles = [true + '\n' + pred for true, pred in zip(trues, preds)]
        MF.show_images(X[0:n].reshape((n,h,w)), 1, n, titles=titles[0:n])
        break