模型訓(xùn)練

在同一個包下創(chuàng)建train.py和model.py，按照步驟先從數(shù)據(jù)處理，模型架構(gòu)搭建，訓(xùn)練測試，統(tǒng)計損失，如下面代碼所示

1. train.py

import torch.optim
import torchvision
from torch import nn
from torch.utils.data import DataLoader
from torch.utils.tensorboard import SummaryWriter

from model import NNN

# 1. 準備數(shù)據(jù)集
train_data = torchvision.datasets.CIFAR10("./data", train=True, transform=torchvision.transforms.ToTensor(),
                                          download=True)
test_data = torchvision.datasets.CIFAR10("./data", train=False, transform=torchvision.transforms.ToTensor(),
                                         download=True)

train_data_size = len(train_data)
test_data_size = len(test_data)
print(f"訓(xùn)練數(shù)據(jù)集的長度：{train_data_size}")
print(f"測試數(shù)據(jù)集的長度：{test_data_size}")

# 2. 利用DataLoader 加載數(shù)據(jù)集
train_dataloader = DataLoader(train_data, batch_size=64)
test_dataloader = DataLoader(test_data, batch_size=64)

# 3. 搭建神經(jīng)網(wǎng)絡(luò)
# 引入model.py
nnn = NNN()

# 4. 創(chuàng)建損失函數(shù)loss
loss_fn = nn.CrossEntropyLoss()  # 交叉熵

# 5. 優(yōu)化器
learning_rate = 0.01
optimizer = torch.optim.SGD(nnn.parameters(), lr=learning_rate)  # 隨機梯度下降

# 6. 設(shè)置訓(xùn)練網(wǎng)絡(luò)的一些參數(shù)
total_train_step = 0  # 記錄訓(xùn)練次數(shù)
total_test_step = 0  # 訓(xùn)練測試次數(shù)
epoch = 10  # 訓(xùn)練輪數(shù)

# 補充tensorboard
writer = SummaryWriter("../logs")

# 開始訓(xùn)練
for i in range(epoch):
    print(f"--------第{i+1}輪訓(xùn)練開始--------")
    # 訓(xùn)練
    nnn.train()
    for data in train_dataloader:
        imgs, targets = data
        outputs = nnn(imgs)
        loss = loss_fn(outputs, targets)

        optimizer.zero_grad()
        loss.backward()
        optimizer.step()

        total_train_step += 1
        if total_train_step % 100 == 0:
            print(f"訓(xùn)練次數(shù)：{total_train_step}---loss：{loss.item()}")
            writer.add_scalar("train_loss", loss.item(), total_train_step)

    # 測試
    nnn.eval()
    total_test_loss = 0  # 總體的誤差
    total_accuracy = 0  # 總體的正確率
    with torch.no_grad():
        for data in test_dataloader:
            imgs, targets = data
            outputs = nnn(imgs)
            loss = loss_fn(outputs, targets)
            total_test_loss += loss.item()
            accuracy = (outputs.argmax(1) == targets).sum()
            total_accuracy += accuracy
    print(f"整體測試集上的loss：{total_test_loss}")
    print(f"整體測試集上的準確率：{total_accuracy/test_data_size}")
    writer.add_scalar("test_loss", total_test_loss, total_test_step)
    writer.add_scalar("total_accuracy", total_accuracy/test_data_size, total_test_step)
    total_test_step += 1

    # 保存每一輪訓(xùn)練的模型
    torch.save(nnn, f"nnn_{i+1}.pth")
    print("模式已保存")


writer.close()

2. model.py

import torch
from torch import nn


# 搭建神經(jīng)網(wǎng)絡(luò)
class NNN(nn.Module):
    def __init__(self):
        super(NNN, self).__init__()
        self.model = nn.Sequential(
            nn.Conv2d(3, 32, 5, stride=1, padding=2),
            nn.MaxPool2d(kernel_size=2),
            nn.Conv2d(32, 32, 5, stride=1, padding=2),
            nn.MaxPool2d(kernel_size=2),
            nn.Conv2d(32, 64, 5, stride=1, padding=2),
            nn.MaxPool2d(2),
            nn.Flatten(),
            nn.Linear(1024, 64),
            nn.Linear(64, 10)
        )

    def forward(self, x):
        x = self.model(x)
        return x


if __name__ == '__main__':
    nnn = NNN()
    input = torch.ones((64, 3, 32, 32))
    output = nnn(input)
    print(output.shape)

運行train.py后可以通過啟動tensorboard進行查看我們的loss情況，損失是不斷下降的。

補充argmax函數(shù)的使用
我們模型預(yù)測處理的是概率，我們需要使用argmax函數(shù)還得到預(yù)測的結(jié)果，就是選出概率最大的，上面測試準確率的計算使用到了。
簡單代碼示例：

import torch
# 模型輸出的概率
outputs = torch.tensor([[0.1, 0.3],
                        [0.7, 0.2]])
# 真實的分類
targets = torch.tensor([[1, 1]])
# 對概率進行預(yù)測
preds = outputs.argmax(1)  # 1:橫向比較 0：豎向比較

# 預(yù)測與真實進行比較
print(preds == targets)
print((preds == targets).sum().item())  # 統(tǒng)計正確的個數(shù)

輸出：文章來源地址http://www.zghlxwxcb.cn/news/detail-733294.html

tensor([[ True, False]])
1

到了這里，關(guān)于Pytorch學習筆記（模型訓(xùn)練）的文章就介紹完了。如果您還想了解更多內(nèi)容，請在右上角搜索TOY模板網(wǎng)以前的文章或繼續(xù)瀏覽下面的相關(guān)文章，希望大家以后多多支持TOY模板網(wǎng)！