0a80b9769115d291f15c244429793eda4cb8ecad,tests/test_layer_transformer.py,,test_conv_to_wider_layer,#,60

Before Change

def test_conv_to_wider_layer():
    a = Conv2D(20, kernel_size=(1, 1),
               activation="relu",
               input_shape=(28, 28, 1),
               padding="same")
    b = Conv2D(30, kernel_size=(1, 1),
               activation="relu",
               padding="same")
    model = Sequential([a, b])
    model.compile(loss=categorical_crossentropy,
                  optimizer=Adadelta(),
                  metrics=["accuracy"])
    a2, b2 = conv_to_wider_layer(a, b, 5)

After Change

def test_conv_to_wider_layer():
    model = get_conv_model()
    conv1 = model.layers[1]
    conv2 = model.layers[4]
    bn1 = model.layers[2]
    new_conv1, [new_conv2], [new_bn1] = conv_to_wider_layer(conv1, [conv2], [bn1], 3)

    new_input = Input(shape=get_int_tuple(model.inputs[0].shape[1:]))
    temp_tensor = new_conv1(new_input)
    temp_tensor = new_bn1(temp_tensor)
    temp_tensor = Activation("relu")(temp_tensor)
    temp_tensor = new_conv2(temp_tensor)
    temp_tensor = copy_layer(model.layers[5])(temp_tensor)
    temp_tensor = Activation("relu")(temp_tensor)
    model2 = Model(inputs=new_input, outputs=temp_tensor)

    random_input = get_conv_data()
    output1 = model.predict_on_batch(random_input)
    output2 = model2.predict_on_batch(random_input)

Instances