About Sequential API and Functional API

[lionwu]

@Oceania2018 Hi Haiping, I have learn Sequential API and Functional API via book "[TensorFlow.NET实战]".

In p.106 to 114, I am tring to build same model via Sequential API and Functional API and except there are equal.

public void SampleBuildModel()
{
        //UseFunctionalAPI to create Model
        Model model = null;
        var inputs = keras.Input(new Tensorflow.Shape(784), name: "I0");
        var outputs = layers.Dense(64, activation:keras.activations.Relu).Apply(inputs);
        outputs = layers.Dense(10, activation: keras.activations.Relu).Apply(outputs);
        model = keras.Model(inputs, outputs, name: "MyNN");

        Console.WriteLine("===== Build Model via Functional API =====");
        model.summary();

        //UseSequentialAPI to create Model
        var sModel = keras.Sequential();
        sModel.add(keras.Input(new Tensorflow.Shape(784), name: "I0"));
        sModel.add(keras.layers.Dense(64, activation: keras.activations.Relu));
        sModel.add(keras.layers.Dense(10, activation: keras.activations.Relu));

        Console.WriteLine("===== Build Model via Sequential API =====");
        sModel.summary();
}

But when I run my program, I got result below:

===== Build Model via Functional API =====
Model: MyNN
_________________________________________________________________
Layer (type)                  Output Shape              Param #
=================================================================
I0 (InputLayer)               (None, 784)               0
_________________________________________________________________
dense (Dense)                 (None, 64)                50240
_________________________________________________________________
dense_1 (Dense)               (None, 10)                650
=================================================================
Total params: 50890
Trainable params: 50890
Non-trainable params: 0
_________________________________________________________________
===== Build Model via Sequential API =====
Model: sequential
_________________________________________________________________
Layer (type)                  Output Shape              Param #
=================================================================
I0 (InputLayer)               (None, 784)               0
=================================================================
Total params: 0
Trainable params: 0
Non-trainable params: 0

Dense layer is not add to model. Am I misunderstand or Sequential is error?

by the way, when I use SequentialAPI, load_weights() can't work correctly, FunctionalAPI can work right.

Thanks for your help! ^^

[Oceania2018]

@lionwu Thanks for reaching out regarding the model.summary bug. It's been fixed in latest commit. You can pull and check if it's working in your end.