Python keras.models.Model.from_config() Examples

def test_layer_sharing_at_heterogeneous_depth_with_concat():
    input_shape = (16, 9, 3)
    input_layer = Input(shape=input_shape)

    A = Dense(3, name='dense_A')
    B = Dense(3, name='dense_B')
    C = Dense(3, name='dense_C')

    x1 = B(A(input_layer))
    x2 = A(C(input_layer))
    output = layers.concatenate([x1, x2])

    M = Model(inputs=input_layer, outputs=output)

    x_val = np.random.random((10, 16, 9, 3))
    output_val = M.predict(x_val)

    config = M.get_config()
    weights = M.get_weights()

    M2 = Model.from_config(config)
    M2.set_weights(weights)

    output_val_2 = M2.predict(x_val)
    np.testing.assert_allclose(output_val, output_val_2, atol=1e-6) 

def test_activity_regularization():
    layer = layers.ActivityRegularization(l1=0.01, l2=0.01)

    # test in functional API
    x = layers.Input(shape=(3,))
    z = layers.Dense(2)(x)
    y = layer(z)
    model = Model(x, y)
    model.compile('rmsprop', 'mse')

    model.predict(np.random.random((2, 3)))

    # test serialization
    model_config = model.get_config()
    model = Model.from_config(model_config)
    model.compile('rmsprop', 'mse') 

def test_layer_sharing_at_heterogeneous_depth_with_concat():
    input_shape = (16, 9, 3)
    input_layer = Input(shape=input_shape)

    A = Dense(3, name='dense_A')
    B = Dense(3, name='dense_B')
    C = Dense(3, name='dense_C')

    x1 = B(A(input_layer))
    x2 = A(C(input_layer))
    output = layers.concatenate([x1, x2])

    M = Model(inputs=input_layer, outputs=output)

    x_val = np.random.random((10, 16, 9, 3))
    output_val = M.predict(x_val)

    config = M.get_config()
    weights = M.get_weights()

    M2 = Model.from_config(config)
    M2.set_weights(weights)

    output_val_2 = M2.predict(x_val)
    np.testing.assert_allclose(output_val, output_val_2, atol=1e-6) 

def test_layer_sharing_at_heterogeneous_depth():
    x_val = np.random.random((10, 5))

    x = Input(shape=(5,))
    A = Dense(5, name='A')
    B = Dense(5, name='B')
    output = A(B(A(B(x))))
    M = Model(x, output)

    output_val = M.predict(x_val)

    config = M.get_config()
    weights = M.get_weights()

    M2 = Model.from_config(config)
    M2.set_weights(weights)

    output_val_2 = M2.predict(x_val)
    np.testing.assert_allclose(output_val, output_val_2, atol=1e-6) 

def test_layer_call_arguments():
    # Test the ability to pass and serialize arguments to `call`.
    inp = layers.Input(shape=(2,))
    x = layers.Dense(3)(inp)
    x = layers.Dropout(0.5)(x, training=True)
    model = Model(inp, x)
    assert not model.uses_learning_phase

    # Test that argument is kept when applying the model
    inp2 = layers.Input(shape=(2,))
    out2 = model(inp2)
    assert not out2._uses_learning_phase

    # Test that argument is kept after loading a model
    config = model.get_config()
    model = Model.from_config(config)
    assert not model.uses_learning_phase 

def test_layer_sharing_at_heterogeneous_depth_with_concat():
    input_shape = (16, 9, 3)
    input_layer = Input(shape=input_shape)

    A = Dense(3, name='dense_A')
    B = Dense(3, name='dense_B')
    C = Dense(3, name='dense_C')

    x1 = B(A(input_layer))
    x2 = A(C(input_layer))
    output = layers.concatenate([x1, x2])

    M = Model(inputs=input_layer, outputs=output)

    x_val = np.random.random((10, 16, 9, 3))
    output_val = M.predict(x_val)

    config = M.get_config()
    weights = M.get_weights()

    M2 = Model.from_config(config)
    M2.set_weights(weights)

    output_val_2 = M2.predict(x_val)
    np.testing.assert_allclose(output_val, output_val_2, atol=1e-6) 

def test_layer_sharing_at_heterogeneous_depth():
    x_val = np.random.random((10, 5))

    x = Input(shape=(5,))
    A = Dense(5, name='A')
    B = Dense(5, name='B')
    output = A(B(A(B(x))))
    M = Model(x, output)

    output_val = M.predict(x_val)

    config = M.get_config()
    weights = M.get_weights()

    M2 = Model.from_config(config)
    M2.set_weights(weights)

    output_val_2 = M2.predict(x_val)
    np.testing.assert_allclose(output_val, output_val_2, atol=1e-6) 

def test_layer_call_arguments():
    # Test the ability to pass and serialize arguments to `call`.
    inp = layers.Input(shape=(2,))
    x = layers.Dense(3)(inp)
    x = layers.Dropout(0.5)(x, training=True)
    model = Model(inp, x)
    assert not model.uses_learning_phase

    # Test that argument is kept when applying the model
    inp2 = layers.Input(shape=(2,))
    out2 = model(inp2)
    assert not out2._uses_learning_phase

    # Test that argument is kept after loading a model
    config = model.get_config()
    model = Model.from_config(config)
    assert not model.uses_learning_phase 

def test_activity_regularization():
    layer = layers.ActivityRegularization(l1=0.01, l2=0.01)

    # test in functional API
    x = layers.Input(shape=(3,))
    z = layers.Dense(2)(x)
    y = layer(z)
    model = Model(x, y)
    model.compile('rmsprop', 'mse')

    model.predict(np.random.random((2, 3)))

    # test serialization
    model_config = model.get_config()
    model = Model.from_config(model_config)
    model.compile('rmsprop', 'mse') 

def test_layer_sharing_at_heterogeneous_depth_with_concat():
    input_shape = (16, 9, 3)
    input_layer = Input(shape=input_shape)

    A = Dense(3, name='dense_A')
    B = Dense(3, name='dense_B')
    C = Dense(3, name='dense_C')

    x1 = B(A(input_layer))
    x2 = A(C(input_layer))
    output = layers.concatenate([x1, x2])

    M = Model(inputs=input_layer, outputs=output)

    x_val = np.random.random((10, 16, 9, 3))
    output_val = M.predict(x_val)

    config = M.get_config()
    weights = M.get_weights()

    M2 = Model.from_config(config)
    M2.set_weights(weights)

    output_val_2 = M2.predict(x_val)
    np.testing.assert_allclose(output_val, output_val_2, atol=1e-6) 

def test_layer_call_arguments():
    # Test the ability to pass and serialize arguments to `call`.
    inp = layers.Input(shape=(2,))
    x = layers.Dense(3)(inp)
    x = layers.Dropout(0.5)(x, training=True)
    model = Model(inp, x)
    assert not model.uses_learning_phase

    # Test that argument is kept when applying the model
    inp2 = layers.Input(shape=(2,))
    out2 = model(inp2)
    assert not out2._uses_learning_phase

    # Test that argument is kept after loading a model
    config = model.get_config()
    model = Model.from_config(config)
    assert not model.uses_learning_phase 

def test_activity_regularization():
    layer = layers.ActivityRegularization(l1=0.01, l2=0.01)

    # test in functional API
    x = layers.Input(shape=(3,))
    z = layers.Dense(2)(x)
    y = layer(z)
    model = Model(x, y)
    model.compile('rmsprop', 'mse')

    model.predict(np.random.random((2, 3)))

    # test serialization
    model_config = model.get_config()
    model = Model.from_config(model_config)
    model.compile('rmsprop', 'mse') 

def test_layer_sharing_at_heterogeneous_depth_with_concat():
    input_shape = (16, 9, 3)
    input_layer = Input(shape=input_shape)

    A = Dense(3, name='dense_A')
    B = Dense(3, name='dense_B')
    C = Dense(3, name='dense_C')

    x1 = B(A(input_layer))
    x2 = A(C(input_layer))
    output = layers.concatenate([x1, x2])

    M = Model(inputs=input_layer, outputs=output)

    x_val = np.random.random((10, 16, 9, 3))
    output_val = M.predict(x_val)

    config = M.get_config()
    weights = M.get_weights()

    M2 = Model.from_config(config)
    M2.set_weights(weights)

    output_val_2 = M2.predict(x_val)
    np.testing.assert_allclose(output_val, output_val_2, atol=1e-6) 

def test_layer_sharing_at_heterogeneous_depth():
    x_val = np.random.random((10, 5))

    x = Input(shape=(5,))
    A = Dense(5, name='A')
    B = Dense(5, name='B')
    output = A(B(A(B(x))))
    M = Model(x, output)

    output_val = M.predict(x_val)

    config = M.get_config()
    weights = M.get_weights()

    M2 = Model.from_config(config)
    M2.set_weights(weights)

    output_val_2 = M2.predict(x_val)
    np.testing.assert_allclose(output_val, output_val_2, atol=1e-6) 

def test_layer_call_arguments():
    # Test the ability to pass and serialize arguments to `call`.
    inp = layers.Input(shape=(2,))
    x = layers.Dense(3)(inp)
    x = layers.Dropout(0.5)(x, training=True)
    model = Model(inp, x)
    assert not model.uses_learning_phase

    # Test that argument is kept when applying the model
    inp2 = layers.Input(shape=(2,))
    out2 = model(inp2)
    assert not out2._uses_learning_phase

    # Test that argument is kept after loading a model
    config = model.get_config()
    model = Model.from_config(config)
    assert not model.uses_learning_phase 

def test_layer_sharing_at_heterogeneous_depth():
    x_val = np.random.random((10, 5))

    x = Input(shape=(5,))
    A = Dense(5, name='A')
    B = Dense(5, name='B')
    output = A(B(A(B(x))))
    M = Model(x, output)

    output_val = M.predict(x_val)

    config = M.get_config()
    weights = M.get_weights()

    M2 = Model.from_config(config)
    M2.set_weights(weights)

    output_val_2 = M2.predict(x_val)
    np.testing.assert_allclose(output_val, output_val_2, atol=1e-6) 

def test_layer_call_arguments():
    # Test the ability to pass and serialize arguments to `call`.
    inp = layers.Input(shape=(2,))
    x = layers.Dense(3)(inp)
    x = layers.Dropout(0.5)(x, training=True)
    model = Model(inp, x)
    assert not model.uses_learning_phase

    # Test that argument is kept when applying the model
    inp2 = layers.Input(shape=(2,))
    out2 = model(inp2)
    assert not out2._uses_learning_phase

    # Test that argument is kept after loading a model
    config = model.get_config()
    model = Model.from_config(config)
    assert not model.uses_learning_phase 

def test_activity_regularization():
    layer = layers.ActivityRegularization(l1=0.01, l2=0.01)

    # test in functional API
    x = layers.Input(shape=(3,))
    z = layers.Dense(2)(x)
    y = layer(z)
    model = Model(x, y)
    model.compile('rmsprop', 'mse')

    model.predict(np.random.random((2, 3)))

    # test serialization
    model_config = model.get_config()
    model = Model.from_config(model_config)
    model.compile('rmsprop', 'mse') 

def test_layer_sharing_at_heterogeneous_depth_with_concat():
    input_shape = (16, 9, 3)
    input_layer = Input(shape=input_shape)

    A = Dense(3, name='dense_A')
    B = Dense(3, name='dense_B')
    C = Dense(3, name='dense_C')

    x1 = B(A(input_layer))
    x2 = A(C(input_layer))
    output = layers.concatenate([x1, x2])

    M = Model(inputs=input_layer, outputs=output)

    x_val = np.random.random((10, 16, 9, 3))
    output_val = M.predict(x_val)

    config = M.get_config()
    weights = M.get_weights()

    M2 = Model.from_config(config)
    M2.set_weights(weights)

    output_val_2 = M2.predict(x_val)
    np.testing.assert_allclose(output_val, output_val_2, atol=1e-6) 

def test_layer_call_arguments():
    # Test the ability to pass and serialize arguments to `call`.
    inp = layers.Input(shape=(2,))
    x = layers.Dense(3)(inp)
    x = layers.Dropout(0.5)(x, training=True)
    model = Model(inp, x)
    assert not model.uses_learning_phase

    # Test that argument is kept when applying the model
    inp2 = layers.Input(shape=(2,))
    out2 = model(inp2)
    assert not out2._uses_learning_phase

    # Test that argument is kept after loading a model
    config = model.get_config()
    model = Model.from_config(config)
    assert not model.uses_learning_phase 

def test_activity_regularization():
    layer = layers.ActivityRegularization(l1=0.01, l2=0.01)

    # test in functional API
    x = layers.Input(shape=(3,))
    z = layers.Dense(2)(x)
    y = layer(z)
    model = Model(x, y)
    model.compile('rmsprop', 'mse')

    model.predict(np.random.random((2, 3)))

    # test serialization
    model_config = model.get_config()
    model = Model.from_config(model_config)
    model.compile('rmsprop', 'mse') 

def test_layer_sharing_at_heterogeneous_depth_with_concat():
    input_shape = (16, 9, 3)
    input_layer = Input(shape=input_shape)

    A = Dense(3, name='dense_A')
    B = Dense(3, name='dense_B')
    C = Dense(3, name='dense_C')

    x1 = B(A(input_layer))
    x2 = A(C(input_layer))
    output = layers.concatenate([x1, x2])

    M = Model(inputs=input_layer, outputs=output)

    x_val = np.random.random((10, 16, 9, 3))
    output_val = M.predict(x_val)

    config = M.get_config()
    weights = M.get_weights()

    M2 = Model.from_config(config)
    M2.set_weights(weights)

    output_val_2 = M2.predict(x_val)
    np.testing.assert_allclose(output_val, output_val_2, atol=1e-6) 

def test_layer_sharing_at_heterogeneous_depth():
    x_val = np.random.random((10, 5))

    x = Input(shape=(5,))
    A = Dense(5, name='A')
    B = Dense(5, name='B')
    output = A(B(A(B(x))))
    M = Model(x, output)

    output_val = M.predict(x_val)

    config = M.get_config()
    weights = M.get_weights()

    M2 = Model.from_config(config)
    M2.set_weights(weights)

    output_val_2 = M2.predict(x_val)
    np.testing.assert_allclose(output_val, output_val_2, atol=1e-6) 

def test_layer_call_arguments():
    # Test the ability to pass and serialize arguments to `call`.
    inp = layers.Input(shape=(2,))
    x = layers.Dense(3)(inp)
    x = layers.Dropout(0.5)(x, training=True)
    model = Model(inp, x)
    assert not model.uses_learning_phase

    # Test that argument is kept when applying the model
    inp2 = layers.Input(shape=(2,))
    out2 = model(inp2)
    assert not out2._uses_learning_phase

    # Test that argument is kept after loading a model
    config = model.get_config()
    model = Model.from_config(config)
    assert not model.uses_learning_phase 

def test_layer_call_arguments():
    # Test the ability to pass and serialize arguments to `call`.
    inp = layers.Input(shape=(2,))
    x = layers.Dense(3)(inp)
    x = layers.Dropout(0.5)(x, training=True)
    model = Model(inp, x)
    assert not model.uses_learning_phase

    # Test that argument is kept when applying the model
    inp2 = layers.Input(shape=(2,))
    out2 = model(inp2)
    assert not out2._uses_learning_phase

    # Test that argument is kept after loading a model
    config = model.get_config()
    model = Model.from_config(config)
    assert not model.uses_learning_phase 

def test_layer_sharing_at_heterogeneous_depth_with_concat():
    input_shape = (16, 9, 3)
    input_layer = Input(shape=input_shape)

    A = Dense(3, name='dense_A')
    B = Dense(3, name='dense_B')
    C = Dense(3, name='dense_C')

    x1 = B(A(input_layer))
    x2 = A(C(input_layer))
    output = layers.concatenate([x1, x2])

    M = Model(inputs=input_layer, outputs=output)

    x_val = np.random.random((10, 16, 9, 3))
    output_val = M.predict(x_val)

    config = M.get_config()
    weights = M.get_weights()

    M2 = Model.from_config(config)
    M2.set_weights(weights)

    output_val_2 = M2.predict(x_val)
    np.testing.assert_allclose(output_val, output_val_2, atol=1e-6) 

def test_activity_regularization():
    layer = layers.ActivityRegularization(l1=0.01, l2=0.01)

    # test in functional API
    x = layers.Input(shape=(3,))
    z = layers.Dense(2)(x)
    y = layer(z)
    model = Model(x, y)
    model.compile('rmsprop', 'mse')

    model.predict(np.random.random((2, 3)))

    # test serialization
    model_config = model.get_config()
    model = Model.from_config(model_config)
    model.compile('rmsprop', 'mse') 

def from_config(cls, config, custom_objects={}):
        if type(custom_objects) is list:
            custom_objects = {obj.__name__: obj for obj in custom_objects}
        custom_objects.update(_get_cells())
        config = config.copy()
        model_config = config.pop('model_config')
        if model_config is None:
            model = None
        else:
            model = Model.from_config(model_config, custom_objects)
        if type(model.input) is list:
            input = model.input[0]
            initial_states = model.input[1:]
        else:
            input = model.input
            initial_states = None
        if type(model.output) is list:
            output = model.output[0]
            final_states = model.output[1:]
        else:
            output = model.output
            final_states = None
        return cls(input, output, initial_states, final_states, **config) 

def reset_pingpong(self):
        """ Reset the models for pingpong training """
        logger.debug("Resetting models")

        # Clear models and graph
        self.predictors = dict()
        K.clear_session()

        # Load Models for current training run
        for model in self.networks.values():
            model.network = Model.from_config(model.config)
            model.network.set_weights(model.weights)

        inputs = self.get_inputs()
        self.build_autoencoders(inputs)
        self.compile_predictors(initialize=False)
        logger.debug("Reset models") 

def from_config(cls, config, custom_objects=None):
        """
        Keras' API isn't really extendible at this point
        therefore we need to use a bit hacky solution to
        be able to correctly reconstruct the HAN model
        from a config. This therefore does not reconstruct
        a instance of HAN model, but actually a standard
        Keras model that behaves exactly the same.
        """
        base_config = config.pop('base_config')

        return Model.from_config(
            base_config, custom_objects=custom_objects
        )