Pytorch轻松学-构建浅层神经网络
前面我们刚刚组队完毕，更新了第一篇，我说我会坚持写下去，这个是我的第二篇，使用pytorch实现简单神经网络完成手写数字识别。这个是所有深度学习框架入门标配的例子，但是从这个例子上我们可以学到pytorch的很多基础知识点，我罗列一下，大致有如下：
1.开始用torch.nn包里面的函数搭建网络 2.模型保存为pt文件与加载调用 3.Torchvision.transofrms来做数据预处理 4.DataLoader简单调用处理数据集
只有理解和看清以上四点才算入门了这个例子。
数据集：
Mnist数据集，数字为0~9、大小为28x28的灰度图像。
加载数据集代码实现：
train_ts = tv.datasets.MNIST(root='./data', train=True, download=True, transform=transform)
test_ts = tv.datasets.MNIST(root='./data', train=False, download=True, transform=transform)
train_dl = DataLoader(train_ts, batch_size=32, shuffle=True, drop_last=False)
test_dl = DataLoader(test_ts, batch_size=64, shuffle=True, drop_last=False)预处理数据方式
transform = tv.transforms.Compose(
[tv.transforms.ToTensor(),
tv.transforms.Normalize((0.5,), (0.5,)),
])其中
Totensor表示把灰度图像素值从0~255转化为0~1之间
Normalize表示对输入的减去0.5, 除以0.5
网络结构如下：
输入层：784个神经元
隐藏层：100个神经元
输出层：10个神经元
model = t.nn.Sequential(
t.nn.Linear(784, 100),
t.nn.ReLU(),
t.nn.Linear(100, 10),
t.nn.LogSoftmax(dim=1)
)定义损失函数与优化函数
loss_fn = t.nn.NLLLoss(reduction="mean")
optimizer = t.optim.Adam(model.parameters(), lr=1e-3)开启训练
for s in range(5):
print("run in step : %d"%s)
for i, (x_train, y_train) in enumerate(train_dl):
x_train = x_train.view(x_train.shape[0], -1)
y_pred = model(x_train)
train_loss = loss_fn(y_pred, y_train)
if (i + 1) % 100 == 0:
print(i + 1, train_loss.item())
model.zero_grad()
train_loss.backward()
optimizer.step()测试模型准确率
total = 0;
correct_count = 0
for test_images, test_labels in test_dl:
for i in range(len(test_labels)):
image = test_images[i].view(1, 784)
with t.no_grad():
pred_labels = model(image)
plabels = t.exp(pred_labels)
probs = list(plabels.numpy()[0])
pred_label = probs.index(max(probs))
true_label = test_labels.numpy()[i]
if pred_label == true_label:
correct_count += 1
total += 1打印准确率与保存模型
print("total acc : %.2f\n"%(correct_count / total)) t.save(model, './nn_mnist_model.pt')完整演示代码
import torch as t
from torch.utils.data import DataLoader
import torchvision as tv
transform = tv.transforms.Compose([tv.transforms.ToTensor(),
tv.transforms.Normalize((0.5,), (0.5,)),
])
train_ts = tv.datasets.MNIST(root='./data', train=True, download=True, transform=transform)
test_ts = tv.datasets.MNIST(root='./data', train=False, download=True, transform=transform)
train_dl = DataLoader(train_ts, batch_size=32, shuffle=True, drop_last=False)
test_dl = DataLoader(test_ts, batch_size=64, shuffle=True, drop_last=False)
model = t.nn.Sequential(
t.nn.Linear(784, 100),
t.nn.ReLU(),
t.nn.Linear(100, 10),
t.nn.LogSoftmax(dim=1)
)
loss_fn = t.nn.NLLLoss(reduction="mean")
optimizer = t.optim.Adam(model.parameters(), lr=1e-3)
for s in range(5):
print("run in step : %d"%s)
for i, (x_train, y_train) in enumerate(train_dl):
x_train = x_train.view(x_train.shape[0], -1)
y_pred = model(x_train)
train_loss = loss_fn(y_pred, y_train)
if (i + 1) % 100 == 0:
print(i + 1, train_loss.item())
model.zero_grad()
train_loss.backward()
optimizer.step()
total = 0;
correct_count = 0
for test_images, test_labels in test_dl:
for i in range(len(test_labels)):
image = test_images[i].view(1, 784)
with t.no_grad():
pred_labels = model(image)
plabels = t.exp(pred_labels)
probs = list(plabels.numpy()[0])
pred_label = probs.index(max(probs))
true_label = test_labels.numpy()[i]
if pred_label == true_label:
correct_count += 1
total += 1
print("total acc : %.2f\n"%(correct_count / total))
t.save(model, './nn_mnist_model.pt')运行结果:
举报
腾讯云开发者
