I’m trying to submit to a Kaggle competition. And found a very interesting paper about CycleGANs. Now I’m trying to implement this using TensorFlow:

    def build_cyclegan(
        generator_g, generator_f, discriminator_x, discriminator_y, lambda_cycle=10
    ):
        real_x = tf.keras.layers.Input(shape=(256, 256, 3), name="x_real")
        real_y = tf.keras.layers.Input(shape=(256, 256, 3), name="y_real")
    
        fake_y = generator_g(real_x)
        cycled_x = generator_f(fake_y)
    
        fake_x = generator_f(real_y)
        cycled_y = generator_g(fake_x)
    
        disc_real_x = discriminator_x(real_x)
        disc_real_y = discriminator_y(real_y)
    
        disc_fake_x = discriminator_x(fake_x)
        disc_fake_y = discriminator_y(fake_y)
    
        cycle_gan = tf.keras.Model(
            inputs=[real_x, real_y],
            outputs=[
                disc_real_x,
                disc_real_y,
                cycled_x,
                cycled_y,
                disc_fake_x,
                disc_fake_y,
            ],
        )
    
        cycle_loss_10 = partial(cycle_consistency_loss, LAMBDA=lambda_cycle)
    
        cycle_gan.compile(
            loss=[
                addversarial_loss_discriminator,
                addversarial_loss_discriminator,
                addversarial_loss_generator,
                addversarial_loss_generator,
                cycle_loss_10,
                cycle_loss_10,
            ],
            optimizer=tf.keras.optimizers.Adam(2e-4, beta_1=0.5),
        )
    
        return cycle_gan

And I got a weird exception, which says that my model expects 2 inputs, but I pass only 1 input. This is strange, because I’m trying to pass 2 images into model image

Here how I’m getting data from tfrec files:

    import tensorflow as tf
    
    
    def _parse_image_function(example_proto):
        feature_description = {
            "image": tf.io.FixedLenFeature([], tf.string),
        }
        parsed_features = tf.io.parse_single_example(example_proto, feature_description)
        image = tf.io.decode_jpeg(parsed_features["image"], channels=3)
        image = tf.reshape(image, (256, 256, 3))
        return image
    
    
    def create_dataset(filenames, repeat=True):
        dataset = tf.data.TFRecordDataset(filenames)
        dataset = dataset.map(_parse_image_function)
        if repeat:
            dataset = dataset.repeat()
        return dataset
    
    
    photo_filenames = tf.data.Dataset.list_files(photos_path + "/*.tfrec")
    monet_filenames = tf.data.Dataset.list_files(monet_path + "/*.tfrec")
    
    photo_dataset = create_dataset(photo_filenames)
    monet_dataset = create_dataset(monet_filenames)
    
    combined_dataset = tf.data.Dataset.zip((photo_dataset, monet_dataset))
    
    combined_dataset = combined_dataset.shuffle(
        buffer_size=1000
    )
    
    batch_size = 1
    dataset = combined_dataset.batch(batch_size).prefetch(tf.data.AUTOTUNE)

This exception appears when I’m trying to fit the model with created dataset:

cycle_gan.fit(dataset, epochs=1)

And the exception says:

ValueError                                Traceback (most recent call last)
Cell In[95], line 1
----> 1 cycle_gan.fit(dataset, epochs=1)

File ~/coding/python/kaggle/venv-kaggle/lib64/python3.10/site-packages/keras/src/utils/traceback_utils.py:122, in filter_traceback.<locals>.error_handler(*args, **kwargs)
    119     filtered_tb = _process_traceback_frames(e.__traceback__)
    120     # To get the full stack trace, call:
    121     # `keras.config.disable_traceback_filtering()`
--> 122     raise e.with_traceback(filtered_tb) from None
    123 finally:
    124     del filtered_tb

File ~/coding/python/kaggle/venv-kaggle/lib64/python3.10/site-packages/keras/src/layers/input_spec.py:160, in assert_input_compatibility(input_spec, inputs, layer_name)
    158 inputs = tree.flatten(inputs)
    159 if len(inputs) != len(input_spec):
--> 160     raise ValueError(
    161         f'Layer "{layer_name}" expects {len(input_spec)} input(s),'
    162         f" but it received {len(inputs)} input tensors. "
    163         f"Inputs received: {inputs}"
    164     )
    165 for input_index, (x, spec) in enumerate(zip(inputs, input_spec)):
    166     if spec is None:

ValueError: Layer "functional_2" expects 2 input(s), but it received 1 input tensors. Inputs received: [<tf.Tensor 'data:0' shape=(None, 256, 256, 3) dtype=uint8>]

I don’t know what am I doing wrong. I thought that issue is because of using tf.data.Dataset.zip method, but everywhere on the internet I found this right, so I don’t even know what to look.