[FIXED] How to set the input of a keras subclass model in tensorflow?

Issue

I've created a keras subclass model using tensorflow. Snippets are shown below.

class SubModel(Model):
    def call(self, inputs):
        print(inputs)

model = SubModel()
model.fit(data, labels, ...)

When fit the model, it will get the inputs and input_shape itself. What I want to do is pass inputs to the model myself.Just like the functional API does.

inputs = tf.keras.input(shape=(100,))
model = tf.keras.Model(inputs=inputs, outputs=outputs)

Solution

Something like that?

model_ = SubModel()
inputs = tf.keras.input(shape=(100,))
outputs = model_(inputs)
model = tf.keras.Model(inputs=inputs, outputs=outputs)

Answered By - gdaras