深度学习与Pytorch入门实战（七）Visdom可视化工具【数字识别实例】

笔记摘抄

1. 安装visdom

安装教程

2. 开启监听进程

python -m visdom.server

3. 访问

用chrome浏览器访问url连接：http://localhost:8097

4. 可视化训练

• 在之前定义网络结构（参考上一节）的基础上加上Visdom可视化。

• 在训练-测试的迭代过程之前，定义两条曲线，在训练-测试的过程中 再不断填充点 以实现 曲线随着训练动态增长：

from visdom import Visdom

viz = Visdom()

viz.line([0.], [0.], win='train_loss', opts=dict(title='train loss'))

viz.line([[0.0, 0.0]], [0.], win='test', opts=dict(title='test loss&acc.', legend=['loss', 'acc.']))

• 第二行Visdom(env="xxx")参数env来设置环境窗口的名称，这里什么都没传(在默认的main窗口下)。

• viz.line的 前两个参数 是曲线的Y和X的坐标（前面是纵轴后面才是横轴）

• 设置了不同的 win参数，它们就会在不同的窗口中展示，

• 第四行定义的是 测试集的loss 和 acc两条曲线，所以在X等于0时，Y给了两个初始值。

开始训练：

• 为了知道训练了多少个batch，设置一个全局的计数器：

global_step = 0

• 在每个batch训练完后，为训练曲线添加点，来让曲线实时增长：

global_step += 1

viz.line([loss.item()], [global_step], win='train_loss', update='append')

• 这里用 win参数 来选择是哪条曲线，用update='append'的方式添加曲线的增长点，前面是Y坐标，后面是X坐标。

• 在每次测试结束后，并在另外两个窗口（用win参数设置）中展示图像（.images） 和 预测值（文本用.text）：

viz.line([[test_loss, correct / len(test_loader.dataset)]], [global_step], 
         win='test', update='append')

viz.images(data.view(-1, 1, 28, 28), win='x')

viz.text(str(pred.detach().numpy()), win='pred', opts=dict(title='pred'))

完整代码：

import  torch
import  torch.nn as nn
import  torch.nn.functional as F
import  torch.optim as optim
from   torchvision import datasets, transforms
from visdom import Visdom

# 超参数
batch_size=200
learning_rate=0.01
epochs=10

# 获取训练数据
train_loader = torch.utils.data.DataLoader(
    datasets.MNIST('../data', train=True, download=True,          #train=True则得到的是训练集
                   transform=transforms.Compose([                 #transform进行数据预处理
                       transforms.ToTensor(),                     #转成Tensor类型的数据
#                        transforms.Normalize((0.1307,), (0.3081,)) # 进行数据标准化(减去均值除以方差),如果要显示数据就不要标准化
                   ])),
    batch_size=batch_size, shuffle=True)                          #按batch_size分出一个batch维度在最前面,shuffle=True打乱顺序

# 获取测试数据
test_loader = torch.utils.data.DataLoader(
    datasets.MNIST('../data', train=False, transform=transforms.Compose([
        transforms.ToTensor(),
#         transforms.Normalize((0.1307,), (0.3081,))
    ])),
    batch_size=batch_size, shuffle=True)


class MLP(nn.Module):

    def __init__(self):
        super(MLP, self).__init__()

        self.model = nn.Sequential(         #定义网络的每一层,
            nn.Linear(784, 200),
            nn.LeakyReLU(inplace=True),
            nn.Linear(200, 200),
            nn.LeakyReLU(inplace=True),
            nn.Linear(200, 10),
            nn.LeakyReLU(inplace=True),
        )

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


net = MLP()
# 定义sgd优化器,指明优化参数、学习率，net.parameters()得到这个类所定义的网络的参数[[w1,b1,w2,b2,...]
optimizer = optim.SGD(net.parameters(), lr=learning_rate)
criteon = nn.CrossEntropyLoss()

# 定义两条曲线，在训练-测试的过程中
# 再不断填充点 以实现 曲线随着训练动态增长
viz = Visdom()
viz.line([0.], [0.], win='train_loss', opts=dict(title='train loss'))
# 定义的是 测试集的loss 和 acc两条曲线，所以在X等于0时，Y给了两个初始值。
viz.line([[0.0, 0.0]], [0.], win='test', opts=dict(title='test loss&acc.',
                                                   legend=['loss', 'acc.']))
global_step = 0


for epoch in range(epochs):

    for batch_idx, (data, target) in enumerate(train_loader):
        data = data.view(-1, 28*28)          # 将二维的图片数据摊平[样本数,784]

        logits = net(data)                   # 前向传播
        loss = criteon(logits, target)       # nn.CrossEntropyLoss()自带Softmax

        optimizer.zero_grad()                # 梯度信息清空
        loss.backward()                      # 反向传播获取梯度
        optimizer.step()                     # 优化器更新

        global_step += 1
        # 绘制训练集的loss的图
        viz.line([loss.item()], [global_step], win='train_loss', update='append')


        if batch_idx % 100 == 0:             # 每100个batch输出一次信息
            print('Train Epoch: {} [{}/{} ({:.0f}%)]	Loss: {:.6f}'.format(
                epoch, batch_idx * len(data), len(train_loader.dataset),
                       100. * batch_idx / len(train_loader), loss.item()))


    test_loss = 0
    correct = 0                                         # correct记录正确分类的样本数
    for data, target in test_loader:
        data = data.view(-1, 28 * 28)
        logits = net(data)
        test_loss += criteon(logits, target).item()     # 其实就是criteon(logits, target)的值，标量

        pred = logits.argmax(dim=1)                
        correct += pred.eq(target.data).float().sum().item()

    # 更新训练集的 loss 和 accuracy
#     print('test accuracy: ', correct / len(test_loader.dataset))   
    viz.line([[test_loss, 100. * correct / len(test_loader.dataset)]], [global_step], 
             win='test', update='append')
    viz.images(data.view(-1, 1, 28, 28), win='x', opts = dict(title = 'Real Image'))
    viz.text(str(pred.detach().cpu().numpy()), win='pred',
             opts=dict(title='pred'))


    test_loss /= len(test_loader.dataset)
    print('
Test set: Average loss: {:.4f}, Accuracy: {}/{} ({:.0f}%)
'.format(
        test_loss, correct, len(test_loader.dataset),
        100. * correct / len(test_loader.dataset)))

view result

Train Epoch: 0 [0/60000 (0%)]	Loss: 2.300833
Train Epoch: 0 [20000/60000 (33%)]	Loss: 2.282752
Train Epoch: 0 [40000/60000 (67%)]	Loss: 2.260350
test accuracy:  0.4326

Test set: Average loss: 0.0111, Accuracy: 4326.0/10000 (43%)

Train Epoch: 1 [0/60000 (0%)]	Loss: 2.243046
Train Epoch: 1 [20000/60000 (33%)]	Loss: 2.179665
Train Epoch: 1 [40000/60000 (67%)]	Loss: 2.034719
test accuracy:  0.5407

Test set: Average loss: 0.0093, Accuracy: 5407.0/10000 (54%)

Train Epoch: 2 [0/60000 (0%)]	Loss: 1.852168
Train Epoch: 2 [20000/60000 (33%)]	Loss: 1.656816
Train Epoch: 2 [40000/60000 (67%)]	Loss: 1.399417
test accuracy:  0.7035

Test set: Average loss: 0.0060, Accuracy: 7035.0/10000 (70%)

Train Epoch: 3 [0/60000 (0%)]	Loss: 1.263554
Train Epoch: 3 [20000/60000 (33%)]	Loss: 1.032210
Train Epoch: 3 [40000/60000 (67%)]	Loss: 0.860783
test accuracy:  0.8003

Test set: Average loss: 0.0038, Accuracy: 8003.0/10000 (80%)

Train Epoch: 4 [0/60000 (0%)]	Loss: 0.833112
Train Epoch: 4 [20000/60000 (33%)]	Loss: 0.625544
Train Epoch: 4 [40000/60000 (67%)]	Loss: 0.610665
test accuracy:  0.8596

Test set: Average loss: 0.0027, Accuracy: 8596.0/10000 (86%)

Train Epoch: 5 [0/60000 (0%)]	Loss: 0.657896
Train Epoch: 5 [20000/60000 (33%)]	Loss: 0.530274
Train Epoch: 5 [40000/60000 (67%)]	Loss: 0.471292
test accuracy:  0.8795

Test set: Average loss: 0.0023, Accuracy: 8795.0/10000 (88%)

Train Epoch: 6 [0/60000 (0%)]	Loss: 0.418883
Train Epoch: 6 [20000/60000 (33%)]	Loss: 0.421887
Train Epoch: 6 [40000/60000 (67%)]	Loss: 0.429563
test accuracy:  0.8908

Test set: Average loss: 0.0020, Accuracy: 8908.0/10000 (89%)

Train Epoch: 7 [0/60000 (0%)]	Loss: 0.429593
Train Epoch: 7 [20000/60000 (33%)]	Loss: 0.368680
Train Epoch: 7 [40000/60000 (67%)]	Loss: 0.389523
test accuracy:  0.8963

Test set: Average loss: 0.0019, Accuracy: 8963.0/10000 (90%)

Train Epoch: 8 [0/60000 (0%)]	Loss: 0.382745
Train Epoch: 8 [20000/60000 (33%)]	Loss: 0.352603
Train Epoch: 8 [40000/60000 (67%)]	Loss: 0.347548
test accuracy:  0.8982

Test set: Average loss: 0.0018, Accuracy: 8982.0/10000 (90%)

Train Epoch: 9 [0/60000 (0%)]	Loss: 0.346902
Train Epoch: 9 [20000/60000 (33%)]	Loss: 0.367318
Train Epoch: 9 [40000/60000 (67%)]	Loss: 0.369855
test accuracy:  0.9021

Test set: Average loss: 0.0017, Accuracy: 9021.0/10000 (90%)