import torch
import torch.nn as nn
import torch.nn.functional as F
import torch.optim as optim
# Image datasets and image manipulation
import torchvision
import torchvision.transforms as transforms
# Image display
import matplotlib.pyplot as plt
import numpy as np
# PyTorch TensorBoard support
from torch.utils.tensorboard import SummaryWriter
# 定义转换器
transform = transforms.Compose(
[transforms.ToTensor(),
#均值方差0.5
transforms.Normalize((0.5,), (0.5,))])
# 训练集和验证集
training_set = torchvision.datasets.FashionMNIST('./data',
download=True,
train=True,
transform=transform)
validation_set = torchvision.datasets.FashionMNIST('./data',
download=True,
train=False,
transform=transform)
# shuffle=True 刷新数据,num_workers 开两个进程
training_loader = torch.utils.data.DataLoader(training_set,
batch_size=4,
shuffle=True,
num_workers=2)
validation_loader = torch.utils.data.DataLoader(validation_set,
batch_size=4,
shuffle=False,
num_workers=2)
# Class labels
classes = ('T-shirt/top', 'Trouser', 'Pullover', 'Dress', 'Coat',
'Sandal', 'Shirt', 'Sneaker', 'Bag', 'Ankle Boot')
# Helper function for inline image display
def matplotlib_imshow(img, one_channel=False):
if one_channel:
img = img.mean(dim=0)
img = img / 2 + 0.5 # unnormalize
npimg = img.numpy()
if one_channel:
plt.imshow(npimg, cmap="Greys")
else:
plt.imshow(np.transpose(npimg, (1, 2, 0)))
# Extract a batch of 4 images
dataiter = iter(training_loader)
images, labels = next(dataiter)
# Create a grid from the images and show them
img_grid = torchvision.utils.make_grid(images)
matplotlib_imshow(img_grid, one_channel=True)
writer = SummaryWriter('runs/fashion_mnist_experiment_1')
# Write image data to TensorBoard log dir
writer.add_image('Four Fashion-MNIST Images', img_grid)
writer.flush()
# To view, start TensorBoard on the command line with:
# tensorboard --logdir=./runs
# ...and open a browser tab to http://localhost:6006/
切换到runs父目录
class Net(nn.Module):
def __init__(self):
super(Net, self).__init__()
self.conv1 = nn.Conv2d(1, 6, 5)
self.pool = nn.MaxPool2d(2, 2)
self.conv2 = nn.Conv2d(6, 16, 5)
self.fc1 = nn.Linear(16 * 4 * 4, 120)
self.fc2 = nn.Linear(120, 84)
self.fc3 = nn.Linear(84, 10)
def forward(self, x):
x = self.pool(F.relu(self.conv1(x)))
x = self.pool(F.relu(self.conv2(x)))
x = x.view(-1, 16 * 4 * 4)
x = F.relu(self.fc1(x))
x = F.relu(self.fc2(x))
x = self.fc3(x)
return x
net = Net()
criterion = nn.CrossEntropyLoss() #损失函数，多元交叉熵
optimizer = optim.SGD(net.parameters(), lr=0.001, momentum=0.9)#优化器，SGD
print(len(validation_loader))
for epoch in range(1): # loop over the dataset multiple times
running_loss = 0.0
for i, data in enumerate(training_loader, 0):
# basic training loop
inputs, labels = data #拆包输入和标记
optimizer.zero_grad() # 梯度归0
outputs = net(inputs) #获取模型预测
loss = criterion(outputs, labels) #根据模型预测和真实标签计算批次平均损失
loss.backward() #反向传播
optimizer.step() # 更新参数
running_loss += loss.item() #累积损失
if i % 1000 == 999: # 每隔1000个小批次
print('Batch {}'.format(i + 1)) #打印当前批次
# 初始化验证损失
running_vloss = 0.0
#设置测试模式
net.train(False)
for j, vdata in enumerate(validation_loader, 0):
vinputs, vlabels = vdata#拆包验证批次输入数据和标签，是类似列表
voutputs = net(vinputs) #预测
vloss = criterion(voutputs, vlabels) #计算损失
running_vloss += vloss.item() #累计损失
net.train(True) # 切换回训练模式，训练模式更新参数，测试模式不更新参数
#每1000个小批次的训练平均损失
avg_loss = running_loss / 1000
#每1000个训练批次的验证平均损失
avg_vloss = running_vloss / len(validation_loader)
#每隔1000个小批次，把训练平均损失，验证平均损失，标记是第几个批次
writer.add_scalars('Training vs. Validation Loss',
{ 'Training' : avg_loss, 'Validation' : avg_vloss },
epoch * len(training_loader) + i)
# 最后把训练累计损失归0,因为我们不需要累计之前的批次的损失
running_loss = 0.0
print('Finished Training')
writer.flush()
# Again, grab a single mini-batch of images
dataiter = iter(training_loader)
images, labels = next(dataiter)
# add_graph() will trace the sample input through your model,
# and render it as a graph.
writer.add_graph(net, images)
writer.flush()
# Select a random subset of data and corresponding labels
def select_n_random(data, labels, n=100):
assert len(data) == len(labels) #断言：如果错误，会报错，程序不会向下执行
perm = torch.randperm(len(data)) #刷新过的索引
return data[perm][:n], labels[perm][:n] #返回刷新后的数据切片
# 取的是经过刷新后的前100个数据和标签
images, labels = select_n_random(training_set.data, training_set.targets)
print(images.shape,labels.shape)
#得到文本标签列表
class_labels = [classes[label] for label in labels]
features = images.view(-1, 28 * 28) #扁平处理输入
global_step = 100 # 假设这是当前的训练步骤或迭代次数
writer.add_embedding(features,
metadata=class_labels,
label_img=images.unsqueeze(1),global_step=global_step) #加一维层（通道）
writer.flush()
writer.close()