http://blog.csdn.net/xiaojiajia007/article/details/72865764
https://stackoverflow.com/questions/42112260/how-do-i-use-the-tensorboard-callback-of-keras
https://www.tensorflow.org/get_started/summaries_and_tensorboard
直接上代码
tb_cb=keras.callbacks.TensorBoard(log_dir='./logs', histogram_freq=1, write_graph=True, write_images=False, embeddings_freq=0, embeddings_layer_names=None, embeddings_metadata=None)
es_cb=EarlyStopping(monitor='val_loss', min_delta=0.09,patience=5, verbose=0, mode='auto')
cbks=[];
cbks.append(tb_cb);
cbks.append(es_cb);
model.fit(x_train, y_train,batch_size=batch_size,epochs=epochs,verbose=1,callbacks=cbks,validation_data=(x_test, y_test))
需要查看的时候，在命令行窗口 tensorboard --logdir=C:UsersAlexanderlogs（这里是log_dir的位置）
然后在浏览器输入 http://localhost:6006，可以查看
Basically, histogram_freq=2 is the most important parameter to tune when calling this callback: it sets an interval of epochs to call the callback, with the goal of generating fewer files on disks.
Second, I removed write_images=True since at each histogram_freq epoch it would consume more than 5 GB on disks to save those images, and the time to save those images is just outstandingly long on this convolutional neural network that has (just) 22 convolutional layers.