softmax回归的简洁实现

import torch
from torch  import nn
from torch.nn import init
import numpy as np
import sys
sys.path.append("..")
import painting as pt

#1.获取和读取数据
batch_size=256
train_iter,test_iter=pt.load_data_fashion_mnist(batch_size)

"""
softmax回归的输出层是一个全连接层，因此用一个线性模块就行，由于前面我们数据
返回的每个batch样本x的形状为(batch_size,1,28,28)，因此要先用view()将x的
形状转换成(batch_size,784)才送入全连接层
"""
#2.定义和初始化模型
num_inputs=784
num_outputs=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): # x shape:(batch,1,28,28)
        y=self.linear(x.view(x.shape[0],-1))
        return y

net=LinearNet(num_inputs,num_outputs)
#我们将对x的形状转换的这个功能自定义一个FlattenLayer并记录在painting中方便以后使用
class FlattenLayer(nn.Module):
    def __init__(self):
        super(FlattenLayer, self).__init__()
    def forward(self, x): # x shape:(batch,*,*,...)
        return x.view(x.shape[0],-1)
#3.这样就可以更方便的定义我们的模型
from collections import OrderedDict
net=nn.Sequential(
    #FlattenLayer(),
    #nn.Linear(num_inputs,num_outputs)
    OrderedDict([
        ('flatten',FlattenLayer()),
        ('linear',nn.Linear(num_inputs,num_outputs))
    ])
)
#然后我们使用均值为0、标准差为0.01的正态分布随机初始化模型的权重参数
init.normal_(net.linear.weight,mean=0,std=0.01)
init.constant_(net.linear.bias,val=0)

#4.softmax和交叉熵损失函数
loss=nn.CrossEntropyLoss()

#5.定义优化算法
optimizer=torch.optim.SGD(net.parameters(),lr=0.1)

#6.训练模型
num_epochs=5
pt.train(net,train_iter,test_iter,loss,num_epochs,batch_size,None,None,optimizer)