丽水做网站企业,一个网站完整的html代码,火车头采集wordpress发布,中国宣布取消新冠免费治疗1、import所需要的torch库和包 2、加载mnist手写数字数据集#xff0c;划分训练集和测试集#xff0c;转化数据格式#xff0c;batch_size设置为200 3、定义三层线性网络参数w#xff0c;b#xff0c;设置求导信息 4、初始化参数#xff0c;这一步比较关键#xff0c;…1、import所需要的torch库和包 2、加载mnist手写数字数据集划分训练集和测试集转化数据格式batch_size设置为200 3、定义三层线性网络参数wb设置求导信息 4、初始化参数这一步比较关键是否初始化影响到数据质量以及后续网络学习效果 5、自定义三层线性网络 6、选定优化器激活函数和loss函数 7、训练及测试并记录每轮训练的loss变化和在测试集上的效果。第一轮就达到了98的准确度判断是初始化效果较好在前几次测试中根据初始化的情况不同初始准确率为50%-85%不等 完整代码
import torch
import torchvision
import torch.nn.functional as Ftrain_loader torch.utils.data.DataLoader(torchvision.datasets.MNIST(mnist_data, trainTrue, downloadTrue,transformtorchvision.transforms.Compose([torchvision.transforms.ToTensor(),torchvision.transforms.Normalize((0.1307, ), (0.3081, ))])),batch_size200, shuffleTrue)test_loader torch.utils.data.DataLoader(torchvision.datasets.MNIST(mnist_data, trainFalse, downloadTrue,transformtorchvision.transforms.Compose([torchvision.transforms.ToTensor(),torchvision.transforms.Normalize((0.1307, ), (0.3081, ))])),batch_size200, shuffleTrue)w1 torch.randn(200, 784, requires_gradTrue)
b1 torch.randn(200, requires_gradTrue)
w2 torch.randn(200, 200, requires_gradTrue)
b2 torch.randn(200, requires_gradTrue)
w3 torch.randn(10, 200, requires_gradTrue)
b3 torch.randn(10, requires_gradTrue)torch.nn.init.kaiming_normal_(w1)
torch.nn.init.kaiming_normal_(w2)
torch.nn.init.kaiming_normal_(w3)def forward(x):x xw1.t() b1x F.relu(x)x xw2.t() b2x F.relu(x)x xw3.t() b3x F.relu(x)return xoptimizer torch.optim.Adam([w1, b1, w2, b2, w3, b3], lr0.001)
criterion torch.nn.CrossEntropyLoss()for epoch in range(10):for batch_idx, (data, target) in enumerate(train_loader):data data.view(-1, 28*28)logits forward(data)loss criterion(logits, target)optimizer.zero_grad()loss.backward()optimizer.step()if (batch_idx1) % 150 0:print(Train Epoch:{} [{}/{}({:.0f}%)]\tLoss:{:.6f}.format(epoch, (batch_idx1) * len(data), len(train_loader.dataset),100. * (batch_idx1) / len(train_loader), loss.item()))test_loss 0correct 0for data, target in test_loader:data data.view(-1, 28*28)logits forward(data)test_loss criterion(logits, target).item()pred logits.data.max(1)[1]correct pred.eq(target.data).sum()test_loss / len(test_loader)print(\nTest Set:Average Loss:{:.4f}, Accuracy:{}/{}({:.0f}%)\n.format(test_loss, correct, len(test_loader.dataset),100. * correct / len(test_loader.dataset)))
