CNN-LSTM 错误初始化时间分布

Question

我正在尝试构建 CNN-LSTM 来预测价格，但在尝试构建模型时收到错误消息。错误消息是

Please initialize 'TimeDistributed layer' with a 'tf.keras.layers.Layer' instance. You passed: <keras.layers.core.Flatten object at 0x7fc70c70f250>

这是我的 CNN-LSTM 模型

model = Sequential()
model.add(TimeDistributed(Conv1D(filters=64, kernel_size=2, activation="relu"), input_shape=(None, n_steps, n_features)))
model.add(TimeDistributed(Conv1D(filters=128, kernel_size=2)))
model.add(TimeDistributed(MaxPooling1D(pool_size=2)))
model.add(TimeDistributed(Flatten()))
model.add(LSTM(100))
model.add(Dense(32))
model.add(Dense(1, activation="relu"))
model.compile(optimizer='Adam', loss='mean_squared_error')

earlyStop = EarlyStopping(monitor='val_loss', mode='min', verbose=1, patience=30)

我已经在寻找解决方案here，但我不太明白，如果能帮到我就太好了。谢谢！

Answer 1

我能够如下所示重现您的问题

from tensorflow import keras
from tensorflow.keras.models import Sequential
from tensorflow.keras.layers import Dense, Conv1D , MaxPooling1D , LSTM, TimeDistributed
from keras.layers import Flatten
from tensorflow.keras.optimizers import Adam

model = Sequential()
model.add(TimeDistributed(Conv1D(filters=64, kernel_size=2, activation="relu"), input_shape=(None, n_steps, n_features)))
model.add(TimeDistributed(Conv1D(filters=128, kernel_size=2)))
model.add(TimeDistributed(MaxPooling1D(pool_size=2)))
model.add(TimeDistributed(Flatten()))
model.add(LSTM(100))
model.add(Dense(32))
model.add(Dense(1, activation="relu"))
model.compile(optimizer='Adam', loss='mean_squared_error')

输出：

---------------------------------------------------------------------------
ValueError                                Traceback (most recent call last)
<ipython-input-9-11436877029b> in <module>()
     11 model.add(TimeDistributed(Conv1D(filters=128, kernel_size=2)))
     12 model.add(TimeDistributed(MaxPooling1D(pool_size=2)))
---> 13 model.add(TimeDistributed(Flatten()))
     14 model.add(LSTM(100))
     15 model.add(Dense(32))

/usr/local/lib/python3.7/dist-packages/tensorflow/python/keras/layers/wrappers.py in __init__(self, layer, **kwargs)
    125           'Please initialize `TimeDistributed` layer with a '
    126           '`tf.keras.layers.Layer` instance. You passed: {input}'.format(
--> 127               input=layer))
    128     super(TimeDistributed, self).__init__(layer, **kwargs)
    129     self.supports_masking = True

ValueError: Please initialize `TimeDistributed` layer with a `tf.keras.layers.Layer` instance. You passed: <keras.layers.core.Flatten object at 0x7fe9c42d5a50>

固定码：

而不是 from keras.layers import Flatten，如果您可以使用 from tensorflow.keras.layers import Flatten，将解决您的问题。

from tensorflow import keras
from tensorflow.keras.models import Sequential
from tensorflow.keras.layers import Dense, Conv1D , MaxPooling1D , LSTM, TimeDistributed, Flatten
#from keras.layers import Flatten
from tensorflow.keras.optimizers import Adam

model = Sequential()
model.add(TimeDistributed(Conv1D(filters=64, kernel_size=2, activation="relu"), input_shape=(None, n_steps, n_features)))
model.add(TimeDistributed(Conv1D(filters=128, kernel_size=2)))
model.add(TimeDistributed(MaxPooling1D(pool_size=2)))
model.add(TimeDistributed(Flatten()))
model.add(LSTM(100))
model.add(Dense(32))
model.add(Dense(1, activation="relu"))
model.compile(optimizer='Adam', loss='mean_squared_error')

CNN-LSTM 错误初始化时间分布

CNN-LSTM error intialized Time Distributed

python

deep-learning

keras

tensorflow

google-colaboratory