[PyTorch小试牛刀]实战三·DNN实现逻辑回归对FashionMNIST数据集进行分类

[PyTorch小试牛刀]实战三·DNN实现逻辑回归对FashionMNIST数据集进行分类

内容还包括了网络模型参数的保存于加载。
数据集
下载地址
代码部分

import torch as t
import torchvision as tv
import numpy as np


# 超参数
EPOCH = 10
BATCH_SIZE = 100
DOWNLOAD_MNIST = True   # 下过数据的话, 就可以设置成 False
N_TEST_IMG = 10          # 到时候显示 5张图片看效果, 如上图一



class DNN(t.nn.Module):
    def __init__(self):
        super(DNN, self).__init__()

        train_data = tv.datasets.FashionMNIST(
        root="./mnist/",
        train=True,
        transform=tv.transforms.ToTensor(),
        download=DOWNLOAD_MNIST
        )

        test_data = tv.datasets.FashionMNIST(
        root="./mnist/",
        train=False,
        transform=tv.transforms.ToTensor(),
        download=DOWNLOAD_MNIST
        )

        print(test_data)


        # Data Loader for easy mini-batch return in training, the image batch shape will be (50, 1, 28, 28)
        self.train_loader = t.utils.data.DataLoader(
            dataset=train_data, 
            batch_size=BATCH_SIZE,
            shuffle=True)

        self.test_loader = t.utils.data.DataLoader(
            dataset=test_data, 
            batch_size=1000,
            shuffle=True)

        self.dnn = t.nn.Sequential(
            t.nn.Linear(28*28,256),
            t.nn.Dropout(0.5),
            t.nn.ELU(),
            t.nn.Linear(256,64),
            t.nn.Dropout(0.5),
            t.nn.ELU(),
            t.nn.Linear(64,10)
        )

        self.lr = 0.001
        self.loss = t.nn.CrossEntropyLoss()
        self.opt = t.optim.Adam(self.parameters(), lr = self.lr)

    def forward(self,x):
        out = self.dnn(x)
        return(out)

def train():
    model = DNN()
    print(model)
    loss = model.loss
    opt = model.opt
    dataloader = model.train_loader
    testloader = model.test_loader

    

    for e in range(EPOCH):
        step = 0
        for (x, y) in (dataloader):
            model.train()# train model dropout used
            step += 1
            b_x = x.view(-1, 28*28)   # batch x, shape (batch, 28*28)
            b_y = y
            out = model(b_x)
            losses = loss(out,b_y)
            opt.zero_grad()
            losses.backward()
            opt.step()
            if(step%100 == 0):
                print(e,step,losses.data.numpy())
                model.eval() # train model dropout not use
                for (tx,ty) in testloader:
                    t_x = tx.view(-1, 28*28)   # batch x, shape (batch, 28*28)
                    t_y = ty
                    t_out = model(t_x)
                    acc = (np.argmax(t_out.data.numpy(),axis=1) == t_y.data.numpy())
                    print(np.sum(acc)/1000)
                    break#只测试前1000个
            


    t.save(model, './model.pkl')  # 保存整个网络
    t.save(model.state_dict(), './model_params.pkl')   # 只保存网络中的参数 (速度快, 占内存少)
    #加载参数的方式
    """net = DNN()
    net.load_state_dict(t.load('./model_params.pkl'))
    net.eval()"""
    #加载整个模型的方式
    net = t.load('./model.pkl')
    net.eval()
    for (tx,ty) in testloader:
        t_x = tx.view(-1, 28*28)   # batch x, shape (batch, 28*28)
        t_y = ty
        t_out = net(t_x)
        acc = (np.argmax(t_out.data.numpy(),axis=1) == t_y.data.numpy())
        print(np.sum(acc)/1000)

if __name__ == "__main__":
    train()

输出结果

9 500 0.42454192
0.875
9 600 0.4553349
0.888
0.876
0.868
0.868
0.881
0.864
0.87
0.87
0.854
0.871
0.879