Combine two model in tensorflow

[Soumya Mishra]

Hi All,

       I have two models say M1 & M2 both are written in Tensorflow.The output of M1 is given as input to M2 and every time gradient is calculated it should be back propagated till M1's input.Please help me with this issue.

[Biswajit Sahoo]

You can use functional API of Tensorflow(keras) to solve this problem. The following pseudo code might help.

inputs = tf.keras.Input(input_shape = ...)        # Specify input shape

output_1 = M1(inputs)

output_2 = M2(output_1)

new_model = tf.keras.Model(inputs = inputs, outputs = output_2)

new_model.summary( )  # Verify your total trainable parameters

model.compile(...)

model.fit(...)

Hope this will solve your issue.