我正在使用Tensorflow Keras在Google colab上使用fit_generator（）在ImageDataGenerator批次中训练二进制Sigmoid分类cnn模型。由于某种原因，模型似乎要训练到完成之前的1步，然后再训练。这发生在每个时代。

https://i.stack.imgur.com/nGBBs.png

可能是什么原因？是在浪费时间吗？应该解决吗？

谢谢

编辑：模型是顺序的。

训练数据生成器：

from keras.preprocessing.image import ImageDataGenerator

train_datagen = ImageDataGenerator(
    rescale = 1./255,rotation_range = 40,width_shift_range = 0.2,height_shift_range = 0.2,shear_range = 0.1,zoom_range = 0.2,fill_mode = 'nearest',horizontal_flip = True,vertical_flip = True,validation_split=0.2
)

data_dir = '/tmp/sorted_images'

train_gen = train_datagen.flow_from_directory(
    data_dir,batch_size=128,target_size = (96,96),color_mode="grayscale",class_mode = 'binary',shuffle=True,subset='training'
)

验证数据生成器：

from keras_preprocessing.image import ImageDataGenerator

valid_datagen = ImageDataGenerator(
    rescale = 1./255,validation_split=0.2
)

valid_gen = valid_datagen.flow_from_directory(
    data_dir,subset='validation'
)

model.compile(loss='binary_crossentropy',optimizer=adam,metrics=['accuracy',tf.keras.metrics.TruePositives(name='truepos'),tf.keras.metrics.FalseNegatives(name='falseneg'),tf.keras.metrics.FalsePositives(name='falsepos'),tf.keras.metrics.TrueNegatives(name='trueneg')])
history = model.fit_generator(train_gen,epochs=15,verbose = 1,callbacks=cb_list,validation_data = valid_gen,validation_freq = 5)