from torch import nn
from softmax回归 import train_ch3
import torch
import torchvision
from torch.utils import data
from torchvision import transforms# ①准备数据集
def load_data_fashion_mnist(batch_size, resize=None):# PyTorch中的一个转换函数，它的作用是将一个PIL Image或numpy.ndarray图像转换为一个Tensor数据类型。trans = [transforms.ToTensor()]# 是否需要改变大小if resize:trans.insert(0, transforms.Resize(resize))# 函数compose将这些转换操作组合起来trans = transforms.Compose(trans)# 训练数据mnist_train = torchvision.datasets.FashionMNIST(root="../data", train=True, transform=trans, download=True)# 测试数据mnist_test = torchvision.datasets.FashionMNIST(root="../data", train=False, transform=trans, download=True)# 返回值return (torch.utils.data.DataLoader(mnist_train, batch_size, shuffle=True,num_workers=4),torch.utils.data.DataLoader(mnist_test, batch_size, shuffle=False,num_workers=4))
# 批量大小为256
batch_size = 256
# 获取训练数据集和测试数据集
train_iter, test_iter = load_data_fashion_mnist(batch_size)# ②实现一个具有单隐藏层的多层感知机，它包含256个隐藏单元
# 定义输入，输出，隐藏层大小
num_inputs, num_outputs, num_hiddens = 784, 10, 256
# 定义W1、b1、W2、b2
W1 = nn.Parameter(torch.randn(num_inputs, num_hiddens, requires_grad=True) * 0.01)
b1 = nn.Parameter(torch.zeros(num_hiddens, requires_grad=True))
W2 = nn.Parameter(torch.randn(num_hiddens, num_outputs, requires_grad=True) * 0.01)
b2 = nn.Parameter(torch.zeros(num_outputs, requires_grad=True))params = [W1, b1, W2, b2]# ③实现ReLU激活函数
def relu(X):a = torch.zeros_like(X)return torch.max(X, a)# ④实现模型
def net(X):# x=256*784X = X.reshape((-1, num_inputs))# torch.matmul(X,W1)= 256*256H = relu(torch.matmul(X,W1) + b1)# torch.matmul(H,W2) = 256*10return (torch.matmul(H,W2) + b2)# ⑤定义损失函数
loss = nn.CrossEntropyLoss()# ⑥训练
# 定义学习率
lr = 0.1
# 优化函数
updater = torch.optim.SGD(params, lr=lr)# 训练
if __name__ == '__main__':num_epochs = 10train_ch3(net, train_iter, test_iter, loss, num_epochs, updater)训练结果
训练损失:0.0015049066459139188
训练精度:0.86405
测试精度:0.8453
貌似是比softmax十次好一些
