Optimizer(1)

Optimizer(1)

pytorchOfficial website link:

https://pytorch.org/docs/stable/optim.html

Use Cases:

import torch
from torch import nn
import torchvision
from torch.utils.data import DataLoader

train_data = torchvision.datasets.CIFAR10(root="./dataset",train=True,transform=torchvision.transforms.ToTensor(),download=True)

# 利用DataLoader加载数据集
train_dataloader = DataLoader(train_data,batch_size=64)

# 定义训练设备
device = torch.device("cuda")  # GPU训练

# 搭建神经网络
class Module(nn.Module):
    def __init__(self):
        super().__init__()
        self.model = nn.Sequential(
            nn.Conv2d(3,32,5,1,2),  # 卷积
            nn.MaxPool2d(2) ,       # 最大池化
            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,input):
        input = self.model(input)
        return input

# 新建网络实例
mymodule = Module()
mymodule = mymodule.to(device)

# 损失函数（交叉熵）
loss = nn.CrossEntropyLoss()
loss = loss.to(device)

# 记录训练的次数
# total_train_step = 0

# 优化器
optim = torch.optim.SGD(mymodule.parameters(),lr = 0.01)    #  mymodule.parameters() 为模型需要优化的参数  lr = 0.01 为学习率

if __name__ == '__main__':
    for epoch in range(20):

        mymodule.train()
        print("------第 {} 轮训练-----".format(epoch+1))
        # 定义每一轮的损失
        running_loss = 0.0

        for data in train_dataloader:
            imgs,targets = data
            imgs = imgs.to(device)
            targets = targets.to(device)

            # 将图片传入网络中进行训练，得到网络输出的结果
            outputs = mymodule(imgs)
            # 使用损失函数
            result_loss = loss(outputs,targets)  # 得到损失函数的值
            
            # 参数优化
            optim.zero_grad()  #  对上一次的梯度进行清零
            result_loss.backward()  #  调用损失函数的反向传播求出每个节点的梯度
            optim.step()  #  对每一个参数进行调优

            # 累加在该轮训练中该数据训练的损失
            running_loss = running_loss + result_loss

            # total_train_step += 1
            # if total_train_step % 100 == 0:
            #     print("训练次数: {},loss: {}".format(total_train_step,result_loss.item()))

        print("第 {} 轮 训练的损失: {}".format(epoch+1,running_loss))

Part of the results:

Files already downloaded and verified
------第 1 轮训练-----
第 1 轮 训练的损失: 1713.95751953125
------第 2 轮训练-----
第 2 轮 训练的损失: 1468.3868408203125
------第 3 轮训练-----
第 3 轮 训练的损失: 1311.38818359375
------第 4 轮训练-----
第 4 轮 训练的损失: 1215.1497802734375

### 经过多轮的训练 损失可能会变小