一、网络代码:
# 搭建神经网络 可以单独成一个文件
import torch
from torch import nn
class Tudui(nn.Module):
def __init__(self):
super(Tudui, self).__init__()
self.modle = nn.Sequential(
nn.Conv2d(3, 32, 5, stride=1, padding=2),
nn.MaxPool2d(2),
nn.Conv2d(32, 32, 5, 1, 2),
nn.MaxPool2d(2),
nn.Conv2d(32, 64, 5, 1, 2),
nn.MaxPool2d(2),
nn.Flatten(),
nn.Linear(64 * 4 * 4, 64),
nn.Linear(64, 10)
)
def forward(self, x):
x = self.modle(x)
return x
if __name__ == '__main__':
tudui = Tudui()
input = torch.ones(64, 3, 32, 32)
output = tudui(input)
print(output.shape)
二、训练和测试代码:
import torch
from torch.utils.tensorboard import SummaryWriter
from modle import *
import torchvision
from torch import nn
from torch.utils.data import DataLoader
train_data = torchvision.datasets.CIFAR10("../data_train", train=True, transform=torchvision.transforms.ToTensor(),
download=True)
text_data = torchvision.datasets.CIFAR10("../data_train", train=False, transform=torchvision.transforms.ToTensor(),
download=True)
# length 长度
train_data_size = len(train_data)
text_data_size = len(text_data)
print("训练集长度为:{}".format(train_data_size))
print("测试集长度为:{}".format(text_data_size))
# 利用 Dataloader 加载数据集
train_dataloader = DataLoader(train_data, batch_size=64)
text_dataloader = DataLoader(text_data, batch_size=64)
# 创建网络模型
tudui = Tudui()
# 创建损失函数
loss_fn = nn.CrossEntropyLoss()
# 定义优化器
learning_rate = 0.01 # 太小了会导致训练效果不明显
optimizer = torch.optim.SGD(tudui.parameters(), lr=learning_rate)
# 设置网络的一些参数
# 记录( 训练 )的次数
total_train_step = 0
# 记录( 测试 )的次数
total_text_step = 0
# 训练轮数
epoch = 2
# 添加tensorboard
writer = SummaryWriter("../logs_modle")
# 开始训练以及测试
for i in range(epoch):
print("---第{}轮训练开始---".format(i+1))
# 训练步骤
tudui.train() # 细节
for data in train_dataloader:
imgs, targets = data
outputs = tudui(imgs)
loss = loss_fn(outputs, targets)
# 优化器优化模型
optimizer.zero_grad() # 优化器梯度清零
loss.backward() # 反向传播获取梯度
optimizer.step() # 进行优化
total_train_step += 1
if total_train_step % 100 == 0:
print("训练次数:{}, loss = {}".format(total_train_step, loss.item())) # loss.item()用法 :将tensor数据类型转换成数字
writer.add_scalar("train_loss", loss.item(), total_train_step)
# 测试步骤
tudui.eval() # 细节
total_text_loss = 0
total_accuracy = 0.0
with torch.no_grad(): # 测试部分没有梯度,保证不会进行调优
for data in text_dataloader:
imgs, targets = data
outputs = tudui(imgs)
loss = loss_fn(outputs, targets)
total_text_loss = total_text_loss + loss.item()
accuracy = (outputs.argmax(1) == targets).sum()
total_accuracy += accuracy
print("整体测试集上的loss:{}".format(total_text_loss))
print("整体测试集上的正确率为:{}".format(total_accuracy.float()/text_data_size)) # 正确个数 / 测试集大小
writer.add_scalar("text_loss", total_text_loss, total_text_step)
writer.add_scalar("text_accuraccy", total_accuracy.float()/text_data_size, total_text_step)
total_text_step += 1
# 保存每一轮训练的模型
torch.save(tudui, "tudui_{}_train.pth".format(i)) # 保存在此程序路径文件夹下
print("第{}轮训练模型已保存".format(i+1))
print("老子干完了")
writer.close()
# tensorboard打开:tensorboard --logdir=logs_modle
在按照土堆编写代码时,有些地方出现了小错误,例如:
-
total_accuracy.float()
同时,为了跟上视频进度,我将训练轮数仅设置为 2 次
1.训练结果展示
2.学习率的影响参照
# lr = 0.001
# lr = 0.01
3.tensorboard结果展示:
三、分类时常用的:
1.代码:
import torch
outputs = torch.tensor([[0.1, 0.4, 0.3],
[0.2, 0.3, 0.1],
[0.4, 0.2, 0.6]])
print(outputs.argmax(0)) # argmax: 提取行中概率最大那个类别; 1:行; 0:列
print(outputs.argmax(1))
for i in range(2):
preds = outputs.argmax(i)
targets = torch.tensor([0, 1, 2])
print(preds == targets)
print((preds == targets).sum()) # 正确的个数
通过此代码可以将训练出的概率,选择 行或列 中最大的那个,并给出结果与 targets 进行对比以进行分类。
argmax(1):按行筛选
argmax(0):按列筛选
2.结果:
刚刚开始起步,有不对的欢迎指正一起学习