Pytorch基础 softmax回归的实现

和之前的逻辑回归类似，针对上篇的Fashion Mnist数据集进行处理；

详细的数学推导这里不给出，直接给出代码；

如果采用自定义方式搭建网络方式：

import torch
import torchvision
import torchvision.transforms as transforms
import matplotlib.pyplot as plt
import time
import sys
import numpy as np
import  requests

#设置训练集和测试集，下载在本地
mnist_train = torchvision.datasets.FashionMNIST(root='~/Datasets/FashionMNIST', train=True, download=True, transform=transforms.ToTensor())
mnist_test = torchvision.datasets.FashionMNIST(root='~/Datasets/FashionMNIST', train=False, download=True, transform=transforms.ToTensor())
batch_size = 256

if sys.platform.startswith('win'):
    num_workers = 0  # 0表示不用额外的进程来加速读取数据
else:
    num_workers = 4
train_iter = torch.utils.data.DataLoader(mnist_train, batch_size=batch_size, shuffle=True, num_workers=num_workers)
test_iter = torch.utils.data.DataLoader(mnist_test, batch_size=batch_size, shuffle=False, num_workers=num_workers)

num_inputs=784
#图片为1*28*28，单通道，所以输入为28*28=784
num_outputs=10
#最终的类别为10个类别，所以输出是10；

w=torch.tensor(np.random.normal(0,0.01,(num_inputs,num_outputs)),dtype=torch.float)
b=torch.zeros(num_outputs,dtype=torch.float)

#使得w,b可以反向传播
w.requires_grad_(requires_grad=True)
b.requires_grad_(requires_grad=True)

def cross_entropy(y_hat, y):
    return - torch.log(y_hat.gather(1, y.view(-1, 1)))

def accuracy(y_hat,y):
    return (y_hat.argmax(dim=1)==y).float().mean().item()
'''
这里注意下正确率计算的写法；
对于y_hat,返回的是batch_size*10的一个矩阵，argemax(dim=1)为选择一个维度上的最大数的索引；
所以y_hat.argmax(dim=1)返回一个batch_size*1的向量；
针对于返回的向量和y进行诸位比较，平均化取值，即可得到该批次的正确率
'''

def sgd(params, lr, batch_size):  # 本函数已保存在d2lzh_pytorch包中方便以后使用
    for param in params:
        param.data -= lr * param.grad / batch_size # 注意这里更改param时用的param.data

def evalute_accuracy(data_iter,net):
    acc_sum,n=0.0,0
    for X,y in data_iter:
        acc_sum+=(net(X).argmax(dim=1)==y).float().sum().item()
        n+=y.shape[0]
    return acc_sum/n

num_epochs,lr=30,0.1

def softmax(X):
    X_exp = X.exp()
    partition = X_exp.sum(dim=1, keepdim=True)
    return X_exp / partition  # 这里应用了广播机制

def net(X):
    return softmax(torch.mm(X.view((-1, num_inputs)), w) + b)

def train(net,train_iter,test_iter,loss,num_epochs,batch_size,params=None,lr=None,optimizer=None):
    for epoch in range(num_epochs):
        train_l_sum,train_acc_sum,n=0.0,0.0,0
        for X,y in train_iter:
            y_hat=net(X)
            l=loss(y_hat,y).sum()

            if optimizer is not None:
                optimizer.zero_grad()
            elif params is not None and params[0].grad is not None:
                for param in params:
                    param.grad.data.zero_()
            l.backward()
            if optimizer is None:
                sgd(params, lr, batch_size)
            else:
                optimizer.step()  # “softmax回归的简洁实现”一节将用到
            train_l_sum+=l.item()
            train_acc_sum+=(y_hat.argmax(dim=1)==y).sum().item()
            n+=y.shape[0]
        test_acc=evalute_accuracy(test_iter,net);
        print('epoch %d, loss %.4f, train acc %.3f, test acc %.3f'
              % (epoch + 1, train_l_sum / n, train_acc_sum / n, test_acc))

if __name__ == '__main__':
    print(mnist_train)
    X, y = [], []
    '''
    for i in range(10):
        X.append(mnist_train[i][0])
        y.append(mnist_train[i][1])
    show_fashion_mnist(X, get_fashion_mnist_labels(y))    
    '''
    train(net, train_iter, test_iter, cross_entropy, num_epochs, batch_size, [w, b], lr)

如果采用典型的层数搭建函数，能有更加简洁的实现版本：

import torch
import torchvision
import torchvision.transforms as transforms
from torch import nn
from collections import OrderedDict
import matplotlib.pyplot as plt
import time
import sys
import numpy as np
import  requests

#设置训练集和测试集，下载在本地
from torch.nn import init

mnist_train = torchvision.datasets.FashionMNIST(root='~/Datasets/FashionMNIST', train=True, download=True, transform=transforms.ToTensor())
mnist_test = torchvision.datasets.FashionMNIST(root='~/Datasets/FashionMNIST', train=False, download=True, transform=transforms.ToTensor())
batch_size = 256

train_iter = torch.utils.data.DataLoader(mnist_train, batch_size=batch_size, shuffle=True)
test_iter = torch.utils.data.DataLoader(mnist_test, batch_size=batch_size, shuffle=False)

num_inputs=784
#图片为1*28*28，单通道，所以输入为28*28=784
num_outputs=10
#最终的类别为10个类别，所以输出是10；

class LinearNet(nn.Module):
    def __init__(self,num_inputs,num_outputs):
        super(LinearNet,self).__init__()
        self.linear=nn.Linear(num_inputs,num_outputs)
    def forward(self,x):
        y=self.linear(x.view(x.shape[0],-1))
        return y

def evalute_accuracy(data_iter,net):
    acc_sum,n=0.0,0
    for X,y in data_iter:
        acc_sum+=(net(X).argmax(dim=1)==y).float().sum().item()
        n+=y.shape[0]
    return acc_sum/n

class FlattenLayer(nn.Module):
    def __init__(self):
        super(FlattenLayer, self).__init__()
    def forward(self,x):
        return x.view(x.shape[0],-1)
    #把1*28*28转化为1*784

net=LinearNet(num_inputs,num_outputs)

#使用orderdict来进行网络结构搭建
#第一层flatten层把1*28*28转化为1*784
#第二层为实际工作层
net=nn.Sequential(
    OrderedDict(
        [
            ('flatten',FlattenLayer()),
            ('linear',nn.Linear(num_inputs,num_outputs))
        ]
    )
)

init.normal_(net.linear.weight, mean=0, std=0.01)
init.constant_(net.linear.bias, val=0)

loss=nn.CrossEntropyLoss()

optimizer=torch.optim.SGD(net.parameters(),lr=0.1)

num_epochs=10

def train(net,train_iter,test_iter,loss,num_epochs,batch_size,params=None,lr=None,optimizer=None):
    for epoch in range(num_epochs):
        train_l_sum,train_acc_sum,n=0.0,0.0,0
        for X,y in train_iter:
            y_hat=net(X)
            l=loss(y_hat,y).sum()

            if optimizer is not None:
                optimizer.zero_grad()
            elif params is not None and params[0].grad is not None:
                for param in params:
                    param.grad.data.zero_()
            l.backward()
            optimizer.step()  # “softmax回归的简洁实现”一节将用到
            train_l_sum+=l.item()
            train_acc_sum+=(y_hat.argmax(dim=1)==y).sum().item()
            n+=y.shape[0]
        test_acc=evalute_accuracy(test_iter,net);
        print('epoch %d, loss %.4f, train acc %.3f, test acc %.3f'
              % (epoch + 1, train_l_sum / n, train_acc_sum / n, test_acc))


if __name__ == '__main__':
    print(mnist_train)
    X, y = [], []
    '''
    for i in range(10):
        X.append(mnist_train[i][0])
        y.append(mnist_train[i][1])
    show_fashion_mnist(X, get_fashion_mnist_labels(y))    
    '''
    train(net, train_iter, test_iter, loss, num_epochs, batch_size,None,None,optimizer)

这里注意一下关于层数构建的问题；

第一层需要进行flatten来进行维度压缩，才能进行784的输出；

对于Flatten和自定义的层，内容都采用类继承的方式，这样更具有灵活和代表性；