广州网站推广联盟,广告设计logo标志,百度指数的网站,路由器屏蔽网站怎么做知识点回顾#xff1a; 随机种子内参的初始化神经网络调参指南 参数的分类调参的顺序各部分参数的调整心得 作业#xff1a;对于day41的简单cnn#xff0c;看看是否可以借助调参指南进一步提高精度。 import torch
import torch.nn as nn
import torch.optim as optim
impo… 知识点回顾 随机种子内参的初始化神经网络调参指南 参数的分类调参的顺序各部分参数的调整心得 作业对于day41的简单cnn看看是否可以借助调参指南进一步提高精度。 import torch
import torch.nn as nn
import torch.optim as optim
import torchvision
import torchvision.transforms as transforms
import numpy as np
import matplotlib.pyplot as plt
import random
from tqdm import tqdm# 设置随机种子确保结果可复现
def set_seed(seed42):random.seed(seed)np.random.seed(seed)torch.manual_seed(seed)if torch.cuda.is_available():torch.cuda.manual_seed(seed)torch.cuda.manual_seed_all(seed)torch.backends.cudnn.deterministic Truetorch.backends.cudnn.benchmark False# 定义CNN模型
class SimpleCNN(nn.Module):def __init__(self):super(SimpleCNN, self).__init__()# 第一个卷积层输入通道1输出通道32卷积核3x3self.conv1 nn.Conv2d(1, 32, kernel_size3, padding1)self.bn1 nn.BatchNorm2d(32)self.relu1 nn.ReLU()self.pool1 nn.MaxPool2d(2)# 第二个卷积层输入通道32输出通道64卷积核3x3self.conv2 nn.Conv2d(32, 64, kernel_size3, padding1)self.bn2 nn.BatchNorm2d(64)self.relu2 nn.ReLU()self.pool2 nn.MaxPool2d(2)# 全连接层self.fc1 nn.Linear(64 * 7 * 7, 128)self.dropout nn.Dropout(0.5)self.fc2 nn.Linear(128, 10)def forward(self, x):# 卷积层1x self.conv1(x)x self.bn1(x)x self.relu1(x)x self.pool1(x)# 卷积层2x self.conv2(x)x self.bn2(x)x self.relu2(x)x self.pool2(x)# 展平x x.view(-1, 64 * 7 * 7)# 全连接层x self.fc1(x)x self.relu1(x)x self.dropout(x)x self.fc2(x)return x# 参数初始化
def init_weights(m):if isinstance(m, nn.Conv2d):nn.init.kaiming_normal_(m.weight, modefan_out, nonlinearityrelu)if m.bias is not None:nn.init.constant_(m.bias, 0)elif isinstance(m, nn.BatchNorm2d):nn.init.constant_(m.weight, 1)nn.init.constant_(m.bias, 0)elif isinstance(m, nn.Linear):nn.init.kaiming_normal_(m.weight, modefan_out, nonlinearityrelu)nn.init.constant_(m.bias, 0)# 训练函数
def train(model, train_loader, criterion, optimizer, device):model.train()train_loss 0correct 0total 0for batch_idx, (inputs, targets) in enumerate(tqdm(train_loader)):inputs, targets inputs.to(device), targets.to(device)optimizer.zero_grad()outputs model(inputs)loss criterion(outputs, targets)loss.backward()optimizer.step()train_loss loss.item()_, predicted outputs.max(1)total targets.size(0)correct predicted.eq(targets).sum().item()acc 100. * correct / totalreturn train_loss / len(train_loader), acc# 测试函数
def test(model, test_loader, criterion, device):model.eval()test_loss 0correct 0total 0with torch.no_grad():for batch_idx, (inputs, targets) in enumerate(tqdm(test_loader)):inputs, targets inputs.to(device), targets.to(device)outputs model(inputs)loss criterion(outputs, targets)test_loss loss.item()_, predicted outputs.max(1)total targets.size(0)correct predicted.eq(targets).sum().item()acc 100. * correct / totalreturn test_loss / len(test_loader), acc# 主函数
def main():# 设置随机种子set_seed(42)# 设置设备device cuda if torch.cuda.is_available() else cpuprint(f使用设备: {device})# 数据预处理transform_train transforms.Compose([transforms.RandomRotation(10),transforms.ToTensor(),transforms.Normalize((0.1307,), (0.3081,))])transform_test transforms.Compose([transforms.ToTensor(),transforms.Normalize((0.1307,), (0.3081,))])# 加载数据集train_dataset torchvision.datasets.MNIST(root./data, trainTrue, downloadTrue, transformtransform_train)test_dataset torchvision.datasets.MNIST(root./data, trainFalse, downloadTrue, transformtransform_test)# 创建数据加载器batch_size 128 # 调参参数1train_loader torch.utils.data.DataLoader(train_dataset, batch_sizebatch_size, shuffleTrue, num_workers2)test_loader torch.utils.data.DataLoader(test_dataset, batch_size100, shuffleFalse, num_workers2)# 初始化模型model SimpleCNN().to(device)model.apply(init_weights)# 定义损失函数和优化器criterion nn.CrossEntropyLoss()learning_rate 0.01 # 调参参数2optimizer optim.Adam(model.parameters(), lrlearning_rate, weight_decay1e-5) # 调参参数3scheduler optim.lr_scheduler.ReduceLROnPlateau(optimizer, min, patience3, factor0.5)# 训练模型epochs 15 # 调参参数4train_losses []train_accs []test_losses []test_accs []best_acc 0for epoch in range(epochs):print(f\nEpoch: {epoch1}/{epochs})train_loss, train_acc train(model, train_loader, criterion, optimizer, device)test_loss, test_acc test(model, test_loader, criterion, device)train_losses.append(train_loss)train_accs.append(train_acc)test_losses.append(test_loss)test_accs.append(test_acc)print(fTrain Loss: {train_loss:.4f} | Train Acc: {train_acc:.2f}%)print(fTest Loss: {test_loss:.4f} | Test Acc: {test_acc:.2f}%)# 学习率调整scheduler.step(test_loss)# 保存最佳模型if test_acc best_acc:best_acc test_acctorch.save(model.state_dict(), best_model.pth)print(fModel saved with accuracy: {best_acc:.2f}%)# 加载最佳模型model.load_state_dict(torch.load(best_model.pth))_, final_acc test(model, test_loader, criterion, device)print(f\nFinal Test Accuracy: {final_acc:.2f}%)# 绘制训练过程plt.figure(figsize(12, 5))plt.subplot(1, 2, 1)plt.plot(train_losses, labelTrain Loss)plt.plot(test_losses, labelTest Loss)plt.xlabel(Epoch)plt.ylabel(Loss)plt.legend()plt.title(Loss Curves)plt.subplot(1, 2, 2)plt.plot(train_accs, labelTrain Accuracy)plt.plot(test_accs, labelTest Accuracy)plt.xlabel(Epoch)plt.ylabel(Accuracy (%))plt.legend()plt.title(Accuracy Curves)plt.tight_layout()plt.savefig(training_curves.png)plt.show()if __name__ __main__:main()
