Keras KeyError：‘度量’行--执行model.fit()时> 13 callbacks=callbacks

Question

当我偶然发现这个问题时，我正在古瑟拉尝试这门课程。每当我试图运行model.fit()时，它都会显示此错误。

显示的错误：

KeyError                                  Traceback (most recent call last)

<ipython-input-83-0ef54ef3afb9> in <module>()
     11     validation_steps = len(x_val) // batch_size,
     12     epochs=12,
---> 13     callbacks=callbacks
     14 )

3 frames

/usr/local/lib/python3.6/dist-packages/livelossplot/generic_keras.py in on_train_begin(self, logs)
     29 
     30     def on_train_begin(self, logs={}):
---> 31         self.liveplot.set_metrics([metric for metric in self.params['metrics'] if not metric.startswith('val_')])
     32 
     33         # slightly convolved due to model.complie(loss=...) stuff

KeyError: 'metrics'

这里是我的实际代码：

from tensorflow.keras.layers import Dense, Input, Dropout,Flatten, Conv2D
from tensorflow.keras.layers import BatchNormalization, Activation, MaxPooling2D

from tensorflow.keras.models import Model, Sequential
from tensorflow.keras.optimizers import Adam, SGD
from tensorflow.keras.callbacks import ModelCheckpoint

初始化CNN

model = Sequential()

第一卷积

model.add(Conv2D(32,(5,5), padding='same', input_shape=(64, 128, 1)))
model.add(BatchNormalization())
model.add(Activation('relu'))
model.add(MaxPooling2D(pool_size=(2,2)))
model.add(Dropout(0.25))

第二卷积层

model.add(Conv2D(64, (5,5), padding='same'))
model.add(BatchNormalization())
model.add(Activation('relu'))
model.add(MaxPooling2D(pool_size=(2,2)))
model.add(Dropout(0.25))

压平

model.add(Flatten())

全连通层

model.add(Dense(1024))
model.add(BatchNormalization())
model.add(Activation('relu'))
model.add(Dropout(0.4))

model.add(Dense(4, activation='softmax'))

学习速率调度与模型编译

initial_learning_rate=0.005
lr_schedule = tf.keras.optimizers.schedules.ExponentialDecay(
    initial_learning_rate = initial_learning_rate,
    decay_steps=5,
    decay_rate=0.96,
    staircase=True
)
optimizer = Adam(learning_rate=lr_schedule)

model.compile(loss='categorical_crossentropy', optimizer=optimizer , metrics=["accuracy"])
model.summary()

训练模型

checkpoint = ModelCheckpoint('model_weight.h5', monitor='val_loss', 
                             save_weights_only=True, mode='min', verbose=0)
callbacks=[PlotLossesCallback(), checkpoint]

batch_size=32

history = model.fit(
    datagen_train.flow(x_train, y_train, batch_size=batch_size, shuffle=True),
    steps_per_epoch = len(x_train) // batch_size,
    validation_data = datagen_val.flow(x_val, y_val, batch_size=batch_size, shuffle=True),
    validation_steps = len(x_val) // batch_size,
    epochs=12,
    callbacks=callbacks
)

我怎么解决这个问题？

Answer 1

尝试更改导入语句

from livelossplot.tf_keras import PlotLossesCallback

至

from livelossplot.inputs.tf_keras import PlotLossesCallback

Answer 2

livelossplot.tf_keras将不能在Tensorflow版本2.1+中工作，将TensorFlow版本从2.2降级为使用pip install tensorflow==2.1的Tensorflow 2.1，它将工作并绘制模型培训图。

Answer 3

您必须更新livelossplot以使用tensorflow 2.x版本。在工作的最新API中有一些重大变化。不要使用tf_keras，而是使用PlotLossesKeras。

from livelossplot import PlotLossesKeras

model.fit(X_train, Y_train,
          epochs=10,
          validation_data=(X_test, Y_test),
          callbacks=[PlotLossesKeras()],
          verbose=0)