python
复制代码
import torch
from torch import nn
from torch.nn import init
import numpy as np
import sys
import d2lzh_pytorch as d2l
# 1.数据预处理
mnist_train = torchvision.datasets.FashionMNIST(
root='/Users/w/PycharmProjects/DeepLearning_with_LiMu/datasets/FashionMnist', train=True, download=True,
transform=transforms.ToTensor())
mnist_test = torchvision.datasets.FashionMNIST(
root='/Users/w/PycharmProjects/DeepLearning_with_LiMu/datasets/FashionMnist', train=False, download=True,
transform=transforms.ToTensor())
# 1.2 数据集的预处理
batch_size = 256
if sys.platform.startswith('win'):
num_worker = 0
else:
num_worker = 4
train_iter = torch.utils.data.DataLoader(mnist_train, batch_size=batch_size, shuffle=True, num_workers=num_worker)
test_iter = torch.utils.data.DataLoader(mnist_test, batch_size=batch_size, shuffle=False, num_workers=num_worker)
# 封装自定义的结构转换函数
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)
#定义网络结构
num_inputs, num_outputs, num_hiddens = 784, 10, 256
net = nn.Sequential(
FlattenLayer(),
nn.Linear(num_inputs,num_hiddens),
nn.ReLU(),
nn.Linear(num_hiddens,num_outputs)
)
for param in net.parameters():
print(param.shape)
# 在 PyTorch 中，init.normal_ 是一个初始化方法，用于直接将张量中的元素初始化为来自正态分布（高斯分布）随机生成的值。它属于 torch.nn.init 模块，通常在神经网络的权重初始化中使用。
for params in net.parameters():
init.normal_(params, mean=0, std=0.01)
# print 结果 torch.Size([256, 784])
#torch.Size([256])
#torch.Size([10, 256])
#torch.Size([10])
batch_size = 256
loss = torch.nn.CrossEntropyLoss()
optimizer = torch.optim.SGD(net.parameters(), lr=0.5)
num_epochs = 5
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()
# 梯度清零
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 = evaluate_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))
train(net, train_iter, test_iter, loss, num_epochs, batch_size, None, None, optimizer)
