I have two keras output with different shapes in result of concatenation. for the demonstartion,
firstinput = tf.keras.Input(shape=(128,2304), batch_size=None, name='firstinput', dtype=tf.float32)
secondinput = tf.keras.Input(shape=(2304),batch_size=None,name='secondinput',dtype=tf.float32)
the shapes are:
<KerasTensor: shape=(None, 128, 2304) dtype=float32 (created by layer 'firstinput')>
<KerasTensor: shape=(None, 2304) dtype=float32 (created by layer 'secondinput')>
I have to add them. but for adding them, they should have same shapes. In the paper, they wrote that "by repeated concatenation of secondinput for 128 iterations to form a new matrix of the same shape as firstinput" how can i do this?