手写数字识别（利用TensorBoard监视训练过程）

1、利用TensorBoard实现监视的步骤

①导入TensorBoard类

import tensorflow as tf
from tensorflow.keras.callbacks import TensorBoard

②定义日志文件的保存路径

log_dir = 'logs/fit_logs'

③实例化TensorBoard对象，指定日志保存路径

tensorboard_callback = TensorBoard(log_dir=log_dir)

④模型编译后，将TensorBoard对象作为回调函数传入

model.compile(....)
model.fit(..., callbacks=[tensorboard_callback])

⑤启动TensorBoard,指明日志目录

tensorboard --logdir logs/fit_logs

⑥在浏览器中打开TensorBoard的网址，既可以看到训练过程中指标的变化过程

2、代码举例 （完整代码可直接运行）

# 手写数字识别
import os
os.environ['TF_CPP_MIN_LOG_LEVEL'] = '2'
import tensorflow as tf
from keras.models import Sequential
from keras.layers import Dense, Dropout, Convolution2D, MaxPooling2D, Flatten
from keras.optimizers import Adam
from keras.callbacks import TensorBoard # 导入TensorBoard类

# 载入数据
mnist = tf.keras.datasets.mnist
(train_data, train_target), (test_data, test_target) = mnist.load_data()
log_dir = 'logs/mnist_convnet'# 设置日志目录
# 处理数据维度
train_data = train_data.reshape(-1, 28, 28, 1)
test_data = test_data.reshape(-1, 28, 28, 1)
# 归一化
train_data = train_data / 255.0
test_data = test_data / 255.0
# 4、one-hot编码
train_target = tf.keras.utils.to_categorical(train_target, num_classes=10)
test_target = tf.keras.utils.to_categorical(test_target, num_classes=10)  # 10种结果

# 搭建CNN卷积神经网络
model = Sequential()
# 卷积层
model.add(
    Convolution2D(input_shape=(28, 28, 1), filters=32, kernel_size=5, strides=1, padding='same', activation='relu'))  
# 池化层
model.add(MaxPooling2D(pool_size=2, strides=2, padding='same', ))
# 第二个卷积层
model.add(Convolution2D(64, 5, strides=1, padding='same', activation='relu'))
# 池化层
model.add(MaxPooling2D(2, 2, 'same'))
# 扁平化 （相当于把(64,7,7,64)数据->(64,7*7*64)）
model.add(Flatten())
# 全连接层
model.add(Dense(1024, activation='relu'))
model.add(Dropout(0.5))
# 全连接层（输出层）
model.add(Dense(10, activation='softmax'))

# 编译
model.compile(optimizer=Adam(lr=1e-4), loss='categorical_crossentropy', metrics=['accuracy'])
tensorboard_callback = TensorBoard(log_dir=log_dir)

# 训练
model.fit(train_data, train_target, batch_size=64, epochs=100, validation_data=(test_data, ...
test_target), callbacks=tensorboard_callback)# 模型编译后添加TensorBoard回调，自己设置需要观察的内容

# 保存模型
model.save('mnist.h1')

3、查看训练过程

① 打开终端

②进入logs文件所在的文件夹（cd 文件夹名称）
接着输入指令：tensorboard --logdir=logs/mnist_convnet

tensorboard --logdir=logs/mnist_convnet

③点击网址即可查看