莫烦 pytorch CNN

记录我的视频学习生活

原地址：https://morvanzhou.github.io/tutorials/machine-learning/torch/4-01-CNN/

遇到的问题：
1.
EOFError: Compressed file ended before the end-of-stream marker was reached
找到下载目录下的文件，删除重新下
2.
UserWarning: test_data has been renamed datawarnings.warn(“test_data has been renamed data”)
UserWarning: test_labels has been renamed targets warnings.warn(“test_labels has been renamed targets”)

test_data.test_data → test_data.data
test_data.tast_labels → test_data.targets

test_x = torch.unsqueeze(test_data.data, dim=1).type(torch.FloatTensor)[:10000]/255.
test_y = test_data.targets[:10000]

import torch
import torch.nn as nn
import torch.utils.data as Data
import torchvision
import tensorflow as tf
import matplotlib.pyplot as plt

# hyper parameters
LR = 0.001  # learning rate
BATCH_SIZE = 50
EPOCH = 1
DOWNLOAD_MNIST = False  # 如果下载过 就False

train_data = torchvision.datasets.MNIST(
    root='./mnist',
    train=True,  # training data 就是true
    transform=torchvision.transforms.ToTensor(),
    download=DOWNLOAD_MNIST
)

# print(train_data.train_data.size())
# print(train_data.train_labels.size())
# plt.imshow(train_data.train_data[0].numpy(), cmap='gray')
# plt.title('%i' % train_data.train_labels[0])
# plt.show()

train_loader = Data.DataLoader(
    dataset=train_data,
    batch_size=BATCH_SIZE,
    shuffle=True,
    num_workers=0
)

test_data = torchvision.datasets.MNIST(
    root='./mnist',
    train=False
)
test_x = torch.unsqueeze(test_data.data, dim=1).type(torch.FloatTensor)[:2000] / 255.
test_y = test_data.targets[:2000]


class CNN(nn.Module):
    def __init__(self):
        super(CNN, self).__init__()
        self.conv1 = nn.Sequential(
            nn.Conv2d(  # (1, 28, 28)
                in_channels=1,
                out_channels=16,
                kernel_size=5,
                stride=1,  # 每次跳过几格
                padding=2,  # if stride = 1,  padding = (kernel_size-1)/2 = (5-1)/2
            ),  # → (16, 28, 28)
            nn.ReLU(),  # → (16, 28, 28)
            nn.MaxPool2d(kernel_size=2),  # → (16, 14, 14)
        )
        self.conv2 = nn.Sequential(  # → (16, 14, 14)
            nn.Conv2d(16, 32, 5, 1, 2),  # → (32, 14, 14)
            nn.ReLU(),  # → (32, 14, 14)
            nn.MaxPool2d(2)  # → (32, 7, 7)
        )
        self.out = nn.Linear(32 * 7 * 7, 10)  # 10是分类个数

    def forward(self, x):
        x = self.conv1(x)
        x = self.conv2(x)  # (batch, 32, 7, 7)
        x = x.view(x.size(0), -1)  # (batch, 32 * 7 * 7)  # 展平
        output = self.out(x)
        return output


cnn = CNN()
# print(cnn)  # net architecture
optimizer = torch.optim.Adam(cnn.parameters(), lr=LR)  # optimize all cnn parameters
loss_func = nn.CrossEntropyLoss()  # the target label is not one-hotted  # 分类问题常用的损失函数为交叉熵( Cross Entropy Loss)

for epoch in range(EPOCH):
    for step, (x, y) in enumerate(train_loader):
        output = cnn(x)  # get output from every net
        loss = loss_func(output, y)  # compute loss for every net
        optimizer.zero_grad()  # clear gradient for next train
        loss.backward()  # back propagation, compute gradient
        optimizer.step()  # apply gradients

        if step % 50 == 0:
            test_output = cnn(test_x)
            pred_y = torch.max(test_output, 1)[1].data.numpy()
            accuracy = sum(pred_y == test_y.data.numpy()) / test_y.size(0)
            print('Epoch: ', epoch, '| train loss: %.4f' % loss.data.numpy(), '| test accuracy: %.2f' % accuracy)

# print 10 predictions from test data
test_output = cnn(test_x[:10])
pred_y = torch.max(test_output, 1)[1].data.numpy()
print(pred_y, 'prediction number')
print(test_y[:10].numpy(), 'real number')

具体的一些函数：
1.torchvision.transforms.ToTensor()：
并不是仅仅将图像转化成张量的形式。它的源码：

class ToTensor(object):
    """Convert a ``PIL Image`` or ``numpy.ndarray`` to tensor.

    Converts a PIL Image or numpy.ndarray (H x W x C) in the range
    [0, 255] to a torch.FloatTensor of shape (C x H x W) in the range [0.0, 1.0]
    if the PIL Image belongs to one of the modes (L, LA, P, I, F, RGB, YCbCr, RGBA, CMYK, 1)
    or if the numpy.ndarray has dtype = np.uint8

    In the other cases, tensors are returned without scaling.
    """

    def __call__(self, pic):
        """
        Args:
            pic (PIL Image or numpy.ndarray): Image to be converted to tensor.

        Returns:
            Tensor: Converted image.
        """
        return F.to_tensor(pic)

    def __repr__(self):
        return self.__class__.__name__ + '()'

可以看出：
如果PIL图像属于（L，LA，P，I，F，RGB，YCbCr，RGBA，CMYK，1）其中一种模式或numpy.ndarray具有dtype = np.uint8的情况下，它将范围为[0，255]的PIL图像或（H xW x C）形式的numpy.ndarray转换为范围为[0.0，1.0]的形状为（C xH x W）的Torch.FloatTensor

在其他情况下，将返回张量而不进行缩放。

2.torch.unsqueeze

test_x = torch.unsqueeze(test_data.data, dim=1).type(torch.FloatTensor)[:2000] / 255.

将数据形式由(2000, 28, 28) 转化为 (2000, 1, 28, 28)，且范围为[0,1]

3.x.view(x.size(0), -1)
x.size(0)是batch的大小，所以x = x.view(x.size(0), -1)相当于x = x.view(BATCH_SIZE, -1)。
-1是指列数未知的情况下，根据原来Tensor的数据和BATCH_SIZE自动分配列数