Tensorflow2.0 使用Tensorboard回调函数可视化保存和控制学习速率

本文链接：https://blog.csdn.net/qq_42536854/article/details/106491629

直接使用代码展示


log_dir = os.path.join('logs',datetime.datetime.now().strftime("%Y%d%d-%H%M%S"))

tensorboard = tf.keras.callbacks.TensorBoard(log_dir,histogram_freq=1)
# histogram_freq记录直方图的频率

file_writer = tf.summary.create_file_writer(log_dir + '/lr') # 文件编写器

file_writer.set_as_default()# 设置默认编写器
def lr_sche(epoch):  # 控制学习速率的函数
    learning_rate = 0.02
    if epoch>2:
        learning_rate = 0.01
    if epoch > 4:
        learning_rate = 0.005
    if epoch>5:
        learning_rate = 0.001
    tf.summary.scalar("learning_rate",data=learning_rate,step=epoch)
    return  learning_rate

lr_callback = tf.keras.callbacks.LearningRateScheduler(lr_sche) # 把函数当做参数


model.fit(dataset,epochs=6,steps_per_epoch=60000/128,validation_data=dataset_test,validation_steps=10000/128,
          callbacks=[tensorboard,lr_callback])