{"metadata":{"kernelspec":{"language":"python","display_name":"Python 3","name":"python3"},"language_info":{"pygments_lexer":"ipython3","nbconvert_exporter":"python","version":"3.6.4","file_extension":".py","codemirror_mode":{"name":"ipython","version":3},"name":"python","mimetype":"text/x-python"}},"nbformat_minor":4,"nbformat":4,"cells":[{"cell_type":"markdown","source":"# Introduction and Setup","metadata":{}},{"cell_type":"code","source":"import tensorflow as tf\nfrom tensorflow import keras\nfrom tensorflow.keras import layers\nimport tensorflow_addons as tfa\n\nfrom kaggle_datasets import KaggleDatasets\nimport matplotlib.pyplot as plt\nimport numpy as np\n\ntry:\n    tpu = tf.distribute.cluster_resolver.TPUClusterResolver()\n    print('Device:', tpu.master())\n    tf.config.experimental_connect_to_cluster(tpu)\n    tf.tpu.experimental.initialize_tpu_system(tpu)\n    strategy = tf.distribute.experimental.TPUStrategy(tpu)\nexcept:\n    strategy = tf.distribute.get_strategy()\nprint('Number of replicas:', strategy.num_replicas_in_sync)\n\nAUTOTUNE = tf.data.experimental.AUTOTUNE\n    \nprint(tf.__version__)","metadata":{"execution":{"iopub.status.busy":"2021-12-09T02:55:29.679545Z","iopub.execute_input":"2021-12-09T02:55:29.680366Z","iopub.status.idle":"2021-12-09T02:55:29.68881Z","shell.execute_reply.started":"2021-12-09T02:55:29.680328Z","shell.execute_reply":"2021-12-09T02:55:29.688216Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":"# Load in the data\n\nWe want to keep our photo dataset and our Monet dataset separate. First, load in the filenames of the TFRecords.","metadata":{}},{"cell_type":"code","source":"GCS_PATH = KaggleDatasets().get_gcs_path('gan-getting-started')\n\nMONET_FILENAMES = tf.io.gfile.glob(str(GCS_PATH + '/monet_tfrec/*.tfrec'))\nprint('Monet TFRecord Files:', len(MONET_FILENAMES))\n\nPHOTO_FILENAMES = tf.io.gfile.glob(str(GCS_PATH + '/photo_tfrec/*.tfrec'))\nprint('Photo TFRecord Files:', len(PHOTO_FILENAMES))","metadata":{"execution":{"iopub.status.busy":"2021-12-09T02:57:59.559094Z","iopub.execute_input":"2021-12-09T02:57:59.559403Z","iopub.status.idle":"2021-12-09T02:58:19.633302Z","shell.execute_reply.started":"2021-12-09T02:57:59.55937Z","shell.execute_reply":"2021-12-09T02:58:19.631855Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":"All the images for the competition are already sized to 256x256. As these images are RGB images, set the channel to 3. Additionally, we need to scale the images to a [-1, 1] scale. Because we are building a generative model, we don't need the labels or the image id so we'll only return the image from the TFRecord.","metadata":{}},{"cell_type":"code","source":"IMAGE_SIZE = [256, 256]\n\ndef decode_image(image):\n    image = tf.image.decode_jpeg(image, channels=3)\n    image = (tf.cast(image, tf.float32) / 127.5) - 1\n    image = tf.reshape(image, [*IMAGE_SIZE, 3])\n    return image\n\ndef read_tfrecord(example):\n    tfrecord_format = {\n        \"image_name\": tf.io.FixedLenFeature([], tf.string),\n        \"image\": tf.io.FixedLenFeature([], tf.string),\n        \"target\": tf.io.FixedLenFeature([], tf.string)\n    }\n    example = tf.io.parse_single_example(example, tfrecord_format)\n    image = decode_image(example['image'])\n    return image","metadata":{"execution":{"iopub.status.busy":"2021-12-09T02:51:04.387887Z","iopub.status.idle":"2021-12-09T02:51:04.38854Z","shell.execute_reply.started":"2021-12-09T02:51:04.388247Z","shell.execute_reply":"2021-12-09T02:51:04.388281Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":"Define the function to extract the image from the files.","metadata":{}},{"cell_type":"code","source":"def load_dataset(filenames, labeled=True, ordered=False):\n    dataset = tf.data.TFRecordDataset(filenames)\n    dataset = dataset.map(read_tfrecord, num_parallel_calls=AUTOTUNE)\n    return dataset","metadata":{"execution":{"iopub.status.busy":"2021-12-09T02:51:04.389917Z","iopub.status.idle":"2021-12-09T02:51:04.390567Z","shell.execute_reply.started":"2021-12-09T02:51:04.39028Z","shell.execute_reply":"2021-12-09T02:51:04.390325Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":"Let's load in our datasets.","metadata":{}},{"cell_type":"code","source":"monet_ds = load_dataset(MONET_FILENAMES, labeled=True).batch(1)\nphoto_ds = load_dataset(PHOTO_FILENAMES, labeled=True).batch(1)","metadata":{"execution":{"iopub.status.busy":"2021-12-09T02:51:04.391827Z","iopub.status.idle":"2021-12-09T02:51:04.392472Z","shell.execute_reply.started":"2021-12-09T02:51:04.392186Z","shell.execute_reply":"2021-12-09T02:51:04.392217Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"example_monet = next(iter(monet_ds))\nexample_photo = next(iter(photo_ds))","metadata":{"execution":{"iopub.status.busy":"2021-12-09T02:51:04.393689Z","iopub.status.idle":"2021-12-09T02:51:04.394347Z","shell.execute_reply.started":"2021-12-09T02:51:04.394056Z","shell.execute_reply":"2021-12-09T02:51:04.394086Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":"Let's  visualize a photo example and a Monet example.","metadata":{}},{"cell_type":"code","source":"plt.subplot(121)\nplt.title('Photo')\nplt.imshow(example_photo[0] * 0.5 + 0.5)\n\nplt.subplot(122)\nplt.title('Monet')\nplt.imshow(example_monet[0] * 0.5 + 0.5)","metadata":{"execution":{"iopub.status.busy":"2021-12-09T02:51:04.395589Z","iopub.status.idle":"2021-12-09T02:51:04.396236Z","shell.execute_reply.started":"2021-12-09T02:51:04.395958Z","shell.execute_reply":"2021-12-09T02:51:04.395988Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":"# Build the generator","metadata":{}},{"cell_type":"markdown","source":"## Reflection Padding\nAs I saw some artifacts and white edges in generated images, I'll try to give some reflection padding in generator.\n\nKeras only supports zero padding. The reference for reflection padding is [here](https://www.machinecurve.com/index.php/2020/02/10/using-constant-padding-reflection-padding-and-replication-padding-with-keras/#keras-reflectionpadding2d).","metadata":{}},{"cell_type":"code","source":"from tensorflow import pad\n\n'''\n  2D Reflection Padding\n  Attributes:\n    - padding: (padding_width, padding_height) tuple\n'''\nclass ReflectionPadding2D(layers.Layer):\n    def __init__(self, padding=(1, 1), **kwargs):\n        self.padding = tuple(padding)\n        super(ReflectionPadding2D, self).__init__(**kwargs)\n\n    def compute_output_shape(self, input_shape):\n        return (input_shape[0], input_shape[1] + 2 * self.padding[0], input_shape[2] + 2 * self.padding[1], input_shape[3])\n\n    def call(self, input_tensor, mask=None):\n        padding_width, padding_height = self.padding\n        return pad(input_tensor, [[0,0], [padding_height, padding_height], [padding_width, padding_width], [0,0] ], 'REFLECT')","metadata":{"execution":{"iopub.status.busy":"2021-12-09T02:51:04.397483Z","iopub.status.idle":"2021-12-09T02:51:04.398122Z","shell.execute_reply.started":"2021-12-09T02:51:04.397856Z","shell.execute_reply":"2021-12-09T02:51:04.397887Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"# class c7s1_64(layers.Layer):\n#     def __init__(self, **kwargs):\n#         super(c7s1_64, self).__init__(**kwargs)\n        \n#     def call(self, x):\n#         initializer = tf.random_normal_initializer(0., 0.02)\n#         gamma_init = keras.initializers.RandomNormal(mean=0.0, stddev=0.02)\n\n#         x = ReflectionPadding2D(padding=(3, 3))(x)\n#         x = layers.Conv2D(64, 7, kernel_initializer=initializer, use_bias=False)(x)\n#         x = tfa.layers.InstanceNormalization(gamma_initializer=gamma_init)(x)\n#         x = layers.ReLU()(x)\n\n#         return x\n    \n#     def compute_output_shape(self, input_shape):\n#         return input_shape\n    \n# class c7s1_3(layers.Layer):\n#     def __init__(self, **kwargs):\n#         super(c7s1_3, self).__init__(**kwargs)\n        \n#     def call(self, x):\n#         initializer = tf.random_normal_initializer(0., 0.02)\n#         gamma_init = keras.initializers.RandomNormal(mean=0.0, stddev=0.02)\n\n#         x = ReflectionPadding2D(padding=(3, 3))(x)\n#         x = layers.Conv2D(3, 7, kernel_initializer=initializer, use_bias=False, activation='tanh')(x)\n\n#         return x\n    \n#     def compute_output_shape(self, input_shape):\n#         return input_shape\n\n# class d128(layers.Layer):\n#     def __init__(self, **kwargs):\n#         super(d128, self).__init__(**kwargs)\n        \n#     def call(self, x):\n#         initializer = tf.random_normal_initializer(0., 0.02)\n#         gamma_init = keras.initializers.RandomNormal(mean=0.0, stddev=0.02)\n\n#         x = layers.Conv2D(128, 3, strides=2, kernel_initializer=initializer, padding='same', use_bias=False)(x)\n#         x = tfa.layers.InstanceNormalization(gamma_initializer=gamma_init)(x)\n#         x = layers.ReLU()(x)\n\n#         return x\n    \n#     def compute_output_shape(self, input_shape):\n#         return input_shape\n    \n# class d256(layers.Layer):\n#     def __init__(self, **kwargs):\n#         super(d256, self).__init__(**kwargs)\n        \n#     def call(self, x):\n#         initializer = tf.random_normal_initializer(0., 0.02)\n#         gamma_init = keras.initializers.RandomNormal(mean=0.0, stddev=0.02)\n\n#         x = layers.Conv2D(256, 3, strides=2, kernel_initializer=initializer, padding='same', use_bias=False)(x)\n#         x = tfa.layers.InstanceNormalization(gamma_initializer=gamma_init)(x)\n#         x = layers.ReLU()(x)\n\n#         return x\n    \n#     def compute_output_shape(self, input_shape):\n#         return input_shape\n    \n# class R256(layers.Layer):\n#     def __init__(self, **kwargs):\n#         super(R256, self).__init__(**kwargs)\n        \n#     def call(self, x):\n#         initializer = tf.random_normal_initializer(0., 0.02)\n#         gamma_init = keras.initializers.RandomNormal(mean=0.0, stddev=0.02)\n\n#         skip = x\n#         x = ReflectionPadding2D(padding=(1, 1))(x)\n#         x = layers.Conv2D(256, 3, kernel_initializer=initializer, use_bias=False)(x)\n#         x = tfa.layers.InstanceNormalization(gamma_initializer=gamma_init)(x)\n#         x = layers.ReLU()(x)\n        \n#         x = ReflectionPadding2D(padding=(1, 1))(x)\n#         x = layers.Conv2D(256, 3, kernel_initializer=initializer, use_bias=False)(x)\n#         x = tfa.layers.InstanceNormalization(gamma_initializer=gamma_init)(x)\n        \n#         x = layers.Add()([x, skip])\n\n#         return x\n    \n#     def compute_output_shape(self, input_shape):\n#         return input_shape\n    \n# class u128(layers.Layer):\n#     def __init__(self, **kwargs):\n#         super(u128, self).__init__(**kwargs)\n        \n#     def call(self, x):\n#         initializer = tf.random_normal_initializer(0., 0.02)\n#         gamma_init = keras.initializers.RandomNormal(mean=0.0, stddev=0.02)\n\n#         x = layers.Conv2DTranspose(128, 3, strides=2, padding='same', kernel_initializer=initializer, use_bias=False)(x)\n#         x = tfa.layers.InstanceNormalization(gamma_initializer=gamma_init)(x)\n#         x = layers.ReLU()(x)\n#         return x\n    \n#     def compute_output_shape(self, input_shape):\n#         return input_shape\n    \n# class u64(layers.Layer):\n#     def __init__(self, **kwargs):\n#         super(u64, self).__init__(**kwargs)\n        \n#     def call(self, x):\n#         initializer = tf.random_normal_initializer(0., 0.02)\n#         gamma_init = keras.initializers.RandomNormal(mean=0.0, stddev=0.02)\n\n#         x = layers.Conv2DTranspose(64, 3, strides=2, padding='same', kernel_initializer=initializer, use_bias=False)(x)\n#         x = tfa.layers.InstanceNormalization(gamma_initializer=gamma_init)(x)\n#         x = layers.ReLU()(x)\n\n#         return x\n    \n#     def compute_output_shape(self, input_shape):\n#         return input_shape\n    \n# def Generator(n_blocks=9):\n#     inputs = layers.Input(shape=[256,256,3])\n\n#     x = inputs\n\n#     x = c7s1_64()(x)\n    \n#     x = d128()(x)\n#     x = d256()(x)\n    \n#     for i in range(n_blocks):\n#         x = R256()(x)\n    \n#     x = u128()(x)\n#     x = u64()(x)\n    \n#     x = c7s1_3()(x)\n\n#     return keras.Model(inputs=inputs, outputs=x)","metadata":{"_kg_hide-input":true,"execution":{"iopub.status.busy":"2021-12-09T02:51:04.399457Z","iopub.status.idle":"2021-12-09T02:51:04.400099Z","shell.execute_reply.started":"2021-12-09T02:51:04.399835Z","shell.execute_reply":"2021-12-09T02:51:04.399866Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"def Generator(n_blocks=9):\n    \n    initializer = tf.random_normal_initializer(0., 0.02)\n    gamma_init = keras.initializers.RandomNormal(mean=0.0, stddev=0.02)\n    \n    inputs = layers.Input(shape=[256,256,3])\n    x = inputs\n\n    # c7s1-64\n    x = ReflectionPadding2D(padding=(3, 3))(x)\n    x = layers.Conv2D(64, 7, kernel_initializer=initializer, use_bias=False)(x)\n    x = tfa.layers.InstanceNormalization(gamma_initializer=gamma_init)(x)\n    x = layers.ReLU()(x)\n    \n    # d128\n    x = layers.Conv2D(128, 3, strides=2, padding='same', kernel_initializer=initializer, use_bias=False)(x)\n    x = tfa.layers.InstanceNormalization(gamma_initializer=gamma_init)(x)\n    x = layers.ReLU()(x)\n    \n    # d256\n    x = layers.Conv2D(256, 3, strides=2, padding='same', kernel_initializer=initializer, use_bias=False)(x)\n    x = tfa.layers.InstanceNormalization(gamma_initializer=gamma_init)(x)\n    x = layers.ReLU()(x)\n    \n    # R256\n    for i in range(n_blocks):\n        skip = x\n        x = ReflectionPadding2D(padding=(1, 1))(x)\n        x = layers.Conv2D(256, 3, kernel_initializer=initializer, use_bias=False)(x)\n        x = tfa.layers.InstanceNormalization(gamma_initializer=gamma_init)(x)\n        x = layers.ReLU()(x)\n        \n        x = ReflectionPadding2D(padding=(1, 1))(x)\n        x = layers.Conv2D(256, 3, kernel_initializer=initializer, use_bias=False)(x)\n        x = tfa.layers.InstanceNormalization(gamma_initializer=gamma_init)(x)\n        \n        x = layers.Add()([x, skip])\n    \n    # u128\n    x = layers.Conv2DTranspose(128, 3, strides=2, padding='same', kernel_initializer=initializer, use_bias=False)(x)\n    x = tfa.layers.InstanceNormalization(gamma_initializer=gamma_init)(x)\n    x = layers.ReLU()(x)\n    \n    # u64\n    x = layers.Conv2DTranspose(64, 3, strides=2, padding='same', kernel_initializer=initializer, use_bias=False)(x)\n    x = tfa.layers.InstanceNormalization(gamma_initializer=gamma_init)(x)\n    x = layers.ReLU()(x)\n    \n    # c7s1-3\n    x = ReflectionPadding2D(padding=(3, 3))(x)\n    x = layers.Conv2D(3, 7, kernel_initializer=initializer, activation='tanh')(x)\n\n    return keras.Model(inputs=inputs, outputs=x)","metadata":{"execution":{"iopub.status.busy":"2021-12-09T02:51:04.401395Z","iopub.status.idle":"2021-12-09T02:51:04.402035Z","shell.execute_reply.started":"2021-12-09T02:51:04.401767Z","shell.execute_reply":"2021-12-09T02:51:04.401796Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":"# Build the discriminator\n\nThe discriminator takes in the input image and classifies it as real or fake (generated). Instead of outputing a single node, the discriminator outputs a smaller 2D image with higher pixel values indicating a real classification and lower values indicating a fake classification.","metadata":{}},{"cell_type":"code","source":"def Discriminator():\n    \n    initializer = tf.random_normal_initializer(0., 0.02)\n    gamma_init = keras.initializers.RandomNormal(mean=0.0, stddev=0.02)\n    \n    inputs = layers.Input(shape=[256,256,3])\n    x = inputs\n    \n    # C64\n    x = layers.Conv2D(64, 4, strides=2, padding='same', kernel_initializer=initializer, use_bias=False)(x)\n    x = layers.LeakyReLU(0.2)(x)\n    \n    # C128\n    x = layers.Conv2D(128, 4, strides=2, padding='same', kernel_initializer=initializer, use_bias=False)(x)\n    x = tfa.layers.InstanceNormalization(gamma_initializer=gamma_init)(x)\n    x = layers.LeakyReLU(0.2)(x)\n    \n    # C256\n    x = layers.Conv2D(256, 4, strides=2, padding='same', kernel_initializer=initializer, use_bias=False)(x)\n    x = tfa.layers.InstanceNormalization(gamma_initializer=gamma_init)(x)\n    x = layers.LeakyReLU(0.2)(x)\n\n    # C512\n    x = layers.ZeroPadding2D()(x)\n    x = layers.Conv2D(512, 4, strides=1, kernel_initializer=initializer, use_bias=False)(x)\n    x = tfa.layers.InstanceNormalization(gamma_initializer=gamma_init)(x)\n    x = layers.LeakyReLU(0.2)(x)\n    \n    x = layers.ZeroPadding2D()(x)\n    x = layers.Conv2D(1, 4, strides=1, kernel_initializer=initializer)(x)\n\n    return tf.keras.Model(inputs=inputs, outputs=x)","metadata":{"execution":{"iopub.status.busy":"2021-12-09T02:51:04.403274Z","iopub.status.idle":"2021-12-09T02:51:04.403874Z","shell.execute_reply.started":"2021-12-09T02:51:04.403628Z","shell.execute_reply":"2021-12-09T02:51:04.403661Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"with strategy.scope():\n    monet_generator = Generator() # transforms photos to Monet-esque paintings\n    photo_generator = Generator() # transforms Monet paintings to be more like photos\n\n    monet_discriminator = Discriminator() # differentiates real Monet paintings and generated Monet paintings\n    photo_discriminator = Discriminator() # differentiates real photos and generated photos","metadata":{"execution":{"iopub.status.busy":"2021-12-09T02:51:04.404829Z","iopub.status.idle":"2021-12-09T02:51:04.405335Z","shell.execute_reply.started":"2021-12-09T02:51:04.4051Z","shell.execute_reply":"2021-12-09T02:51:04.405127Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":"Since our generators are not trained yet, the generated Monet-esque photo does not show what is expected at this point.","metadata":{}},{"cell_type":"code","source":"to_monet = monet_generator(example_photo)\n\nplt.subplot(1, 2, 1)\nplt.title(\"Original Photo\")\nplt.imshow(example_photo[0] * 0.5 + 0.5)\n\nplt.subplot(1, 2, 2)\nplt.title(\"Monet-esque Photo\")\nplt.imshow(to_monet[0] * 0.5 + 0.5)\nplt.show()","metadata":{"execution":{"iopub.status.busy":"2021-12-09T02:51:04.406258Z","iopub.status.idle":"2021-12-09T02:51:04.40657Z","shell.execute_reply.started":"2021-12-09T02:51:04.406417Z","shell.execute_reply":"2021-12-09T02:51:04.406433Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":"# Build the CycleGAN model\n\nWe will subclass a `tf.keras.Model` so that we can run `fit()` later to train our model. During the training step, the model transforms a photo to a Monet painting and then back to a photo. The difference between the original photo and the twice-transformed photo is the cycle-consistency loss. We want the original photo and the twice-transformed photo to be similar to one another.\n\nThe losses are defined in the next section.","metadata":{}},{"cell_type":"code","source":"class CycleGan(keras.Model):\n    def __init__(\n        self,\n        monet_generator,\n        photo_generator,\n        monet_discriminator,\n        photo_discriminator,\n        lambda_cycle=10,\n    ):\n        super(CycleGan, self).__init__()\n        self.m_gen = monet_generator\n        self.p_gen = photo_generator\n        self.m_disc = monet_discriminator\n        self.p_disc = photo_discriminator\n        self.lambda_cycle = lambda_cycle\n        \n    def compile(\n        self,\n        m_gen_optimizer,\n        p_gen_optimizer,\n        m_disc_optimizer,\n        p_disc_optimizer,\n        gen_loss_fn,\n        disc_loss_fn,\n        cycle_loss_fn,\n        identity_loss_fn\n    ):\n        super(CycleGan, self).compile()\n        self.m_gen_optimizer = m_gen_optimizer\n        self.p_gen_optimizer = p_gen_optimizer\n        self.m_disc_optimizer = m_disc_optimizer\n        self.p_disc_optimizer = p_disc_optimizer\n        self.gen_loss_fn = gen_loss_fn\n        self.disc_loss_fn = disc_loss_fn\n        self.cycle_loss_fn = cycle_loss_fn\n        self.identity_loss_fn = identity_loss_fn\n        \n    def train_step(self, batch_data):\n        real_monet, real_photo = batch_data\n        \n        with tf.GradientTape(persistent=True) as tape:\n            # photo to monet back to photo\n            fake_monet = self.m_gen(real_photo, training=True)\n            cycled_photo = self.p_gen(fake_monet, training=True)\n\n            # monet to photo back to monet\n            fake_photo = self.p_gen(real_monet, training=True)\n            cycled_monet = self.m_gen(fake_photo, training=True)\n\n            # generating itself\n            same_monet = self.m_gen(real_monet, training=True)\n            same_photo = self.p_gen(real_photo, training=True)\n\n            # discriminator used to check, inputing real images\n            disc_real_monet = self.m_disc(real_monet, training=True)\n            disc_real_photo = self.p_disc(real_photo, training=True)\n\n            # discriminator used to check, inputing fake images\n            disc_fake_monet = self.m_disc(fake_monet, training=True)\n            disc_fake_photo = self.p_disc(fake_photo, training=True)\n\n            # evaluates generator loss\n            monet_gen_loss = self.gen_loss_fn(disc_fake_monet)\n            photo_gen_loss = self.gen_loss_fn(disc_fake_photo)\n\n            # evaluates total cycle consistency loss\n            total_cycle_loss = self.cycle_loss_fn(real_monet, cycled_monet, self.lambda_cycle) + self.cycle_loss_fn(real_photo, cycled_photo, self.lambda_cycle)\n\n            # evaluates total generator loss\n            total_monet_gen_loss = monet_gen_loss + total_cycle_loss + self.identity_loss_fn(real_monet, same_monet, self.lambda_cycle)\n            total_photo_gen_loss = photo_gen_loss + total_cycle_loss + self.identity_loss_fn(real_photo, same_photo, self.lambda_cycle)\n\n            # evaluates discriminator loss\n            monet_disc_loss = self.disc_loss_fn(disc_real_monet, disc_fake_monet)\n            photo_disc_loss = self.disc_loss_fn(disc_real_photo, disc_fake_photo)\n\n        # Calculate the gradients for generator and discriminator\n        monet_generator_gradients = tape.gradient(total_monet_gen_loss,\n                                                  self.m_gen.trainable_variables)\n        photo_generator_gradients = tape.gradient(total_photo_gen_loss,\n                                                  self.p_gen.trainable_variables)\n\n        monet_discriminator_gradients = tape.gradient(monet_disc_loss,\n                                                      self.m_disc.trainable_variables)\n        photo_discriminator_gradients = tape.gradient(photo_disc_loss,\n                                                      self.p_disc.trainable_variables)\n\n        # Apply the gradients to the optimizer\n        self.m_gen_optimizer.apply_gradients(zip(monet_generator_gradients,\n                                                 self.m_gen.trainable_variables))\n\n        self.p_gen_optimizer.apply_gradients(zip(photo_generator_gradients,\n                                                 self.p_gen.trainable_variables))\n\n        self.m_disc_optimizer.apply_gradients(zip(monet_discriminator_gradients,\n                                                  self.m_disc.trainable_variables))\n\n        self.p_disc_optimizer.apply_gradients(zip(photo_discriminator_gradients,\n                                                  self.p_disc.trainable_variables))\n        \n        return {\n            \"monet_gen_loss\": total_monet_gen_loss,\n            \"photo_gen_loss\": total_photo_gen_loss,\n            \"monet_disc_loss\": monet_disc_loss,\n            \"photo_disc_loss\": photo_disc_loss\n        }","metadata":{"execution":{"iopub.status.busy":"2021-12-09T02:51:04.407843Z","iopub.status.idle":"2021-12-09T02:51:04.408139Z","shell.execute_reply.started":"2021-12-09T02:51:04.407987Z","shell.execute_reply":"2021-12-09T02:51:04.408003Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":"# Define loss functions","metadata":{}},{"cell_type":"code","source":"with strategy.scope():\n    def discriminator_loss(real, generated):\n        real_loss = tf.keras.losses.BinaryCrossentropy(from_logits=True, reduction=tf.keras.losses.Reduction.NONE)(tf.ones_like(real), real)\n        generated_loss = tf.keras.losses.BinaryCrossentropy(from_logits=True, reduction=tf.keras.losses.Reduction.NONE)(tf.zeros_like(generated), generated)\n        total_disc_loss = real_loss + generated_loss\n        return total_disc_loss * 0.5\n    \n    def generator_loss(generated):\n        return tf.keras.losses.BinaryCrossentropy(from_logits=True, reduction=tf.keras.losses.Reduction.NONE)(tf.ones_like(generated), generated)\n    \n    def calc_cycle_loss(real_image, cycled_image, LAMBDA):\n        loss1 = tf.reduce_mean(tf.abs(real_image - cycled_image))\n        return LAMBDA * loss1\n    \n    def identity_loss(real_image, same_image, LAMBDA):\n        loss = tf.reduce_mean(tf.abs(real_image - same_image))\n        return LAMBDA * 0.5 * loss","metadata":{"execution":{"iopub.status.busy":"2021-12-09T02:51:04.409596Z","iopub.status.idle":"2021-12-09T02:51:04.410159Z","shell.execute_reply.started":"2021-12-09T02:51:04.409965Z","shell.execute_reply":"2021-12-09T02:51:04.409988Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":"# Train the CycleGAN\n\nLet's compile our model. Since we used `tf.keras.Model` to build our CycleGAN, we can just ude the `fit` function to train our model.","metadata":{}},{"cell_type":"code","source":"with strategy.scope():\n    monet_generator_optimizer = tf.keras.optimizers.Adam(2e-4, beta_1=0.5)\n    photo_generator_optimizer = tf.keras.optimizers.Adam(2e-4, beta_1=0.5)\n\n    monet_discriminator_optimizer = tf.keras.optimizers.Adam(2e-4, beta_1=0.5)\n    photo_discriminator_optimizer = tf.keras.optimizers.Adam(2e-4, beta_1=0.5)\n    \n    cycle_gan_model = CycleGan(\n        monet_generator, photo_generator, monet_discriminator, photo_discriminator\n    )\n\n    cycle_gan_model.compile(\n        m_gen_optimizer = monet_generator_optimizer,\n        p_gen_optimizer = photo_generator_optimizer,\n        m_disc_optimizer = monet_discriminator_optimizer,\n        p_disc_optimizer = photo_discriminator_optimizer,\n        gen_loss_fn = generator_loss,\n        disc_loss_fn = discriminator_loss,\n        cycle_loss_fn = calc_cycle_loss,\n        identity_loss_fn = identity_loss\n    )","metadata":{"execution":{"iopub.status.busy":"2021-12-09T02:51:04.411563Z","iopub.status.idle":"2021-12-09T02:51:04.412172Z","shell.execute_reply.started":"2021-12-09T02:51:04.412004Z","shell.execute_reply":"2021-12-09T02:51:04.412021Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"cycle_gan_model.fit(\n    tf.data.Dataset.zip((monet_ds, photo_ds)),\n    epochs=100\n)","metadata":{"execution":{"iopub.status.busy":"2021-12-09T02:51:04.412901Z","iopub.status.idle":"2021-12-09T02:51:04.413627Z","shell.execute_reply.started":"2021-12-09T02:51:04.413436Z","shell.execute_reply":"2021-12-09T02:51:04.413456Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":"# Visualize Monet-esque photos with 100 epoch trained model","metadata":{}},{"cell_type":"code","source":"_, ax = plt.subplots(5, 2, figsize=(12, 12))\nfor i, img in enumerate(photo_ds.take(5)):\n    prediction = monet_generator(img, training=False)[0].numpy()\n    prediction = (prediction * 127.5 + 127.5).astype(np.uint8)\n    img = (img[0] * 127.5 + 127.5).numpy().astype(np.uint8)\n\n    ax[i, 0].imshow(img)\n    ax[i, 1].imshow(prediction)\n    ax[i, 0].set_title(\"Input Photo\")\n    ax[i, 1].set_title(\"Monet-esque\")\n    ax[i, 0].axis(\"off\")\n    ax[i, 1].axis(\"off\")\nplt.savefig('first_monetesque_image.png', dpi=200)\nplt.show()","metadata":{"execution":{"iopub.status.busy":"2021-12-09T02:51:04.414608Z","iopub.status.idle":"2021-12-09T02:51:04.415171Z","shell.execute_reply.started":"2021-12-09T02:51:04.414987Z","shell.execute_reply":"2021-12-09T02:51:04.415006Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":"# Save weights\nhttps://www.kaggle.com/amyjang/monet-cyclegan-tutorial/comments#1195949","metadata":{}},{"cell_type":"code","source":"with strategy.scope():\n    monet_generator.save_weights('./monet_generator_wr_00.h5')\n    photo_generator.save_weights('./photo_generator_wr_00.h5')\n    \n    monet_discriminator.save_weights('./monet_discriminator_wr_00.h5')\n    photo_discriminator.save_weights('./photo_discriminator_wr_00.h5')","metadata":{"execution":{"iopub.status.busy":"2021-12-09T02:51:04.416671Z","iopub.status.idle":"2021-12-09T02:51:04.417164Z","shell.execute_reply.started":"2021-12-09T02:51:04.416986Z","shell.execute_reply":"2021-12-09T02:51:04.417005Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":"# Re-train with loaded weights for decaying learing rate","metadata":{}},{"cell_type":"code","source":"with strategy.scope():\n    monet_generator = Generator() # transforms photos to Monet-esque paintings\n    photo_generator = Generator() # transforms Monet paintings to be more like photos\n\n    monet_discriminator = Discriminator() # differentiates real Monet paintings and generated Monet paintings\n    photo_discriminator = Discriminator() # differentiates real photos and generated photos\n    \n    monet_generator.load_weights('./monet_generator_wr_00.h5')\n    photo_generator.load_weights('./photo_generator_wr_00.h5')\n    \n    monet_discriminator.load_weights('./monet_discriminator_wr_00.h5')\n    photo_discriminator.load_weights('./photo_discriminator_wr_00.h5')","metadata":{"execution":{"iopub.status.busy":"2021-12-09T02:51:04.418331Z","iopub.status.idle":"2021-12-09T02:51:04.418659Z","shell.execute_reply.started":"2021-12-09T02:51:04.418491Z","shell.execute_reply":"2021-12-09T02:51:04.418514Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"with strategy.scope():\n    lr_fn = tf.keras.optimizers.schedules.PolynomialDecay(\n        initial_learning_rate=2e-4,\n        decay_steps=100*300,    # 100 epoch with 300 step\n        end_learning_rate=0\n    )\n\n    monet_generator_optimizer = tf.keras.optimizers.Adam(beta_1=0.5, learning_rate=lr_fn)\n    photo_generator_optimizer = tf.keras.optimizers.Adam(beta_1=0.5, learning_rate=lr_fn)\n\n    monet_discriminator_optimizer = tf.keras.optimizers.Adam(beta_1=0.5, learning_rate=lr_fn)\n    photo_discriminator_optimizer = tf.keras.optimizers.Adam(beta_1=0.5, learning_rate=lr_fn)\n\n    cycle_gan_model = CycleGan(\n        monet_generator, photo_generator, monet_discriminator, photo_discriminator\n    )\n    \n    cycle_gan_model.compile(\n        m_gen_optimizer = monet_generator_optimizer,\n        p_gen_optimizer = photo_generator_optimizer,\n        m_disc_optimizer = monet_discriminator_optimizer,\n        p_disc_optimizer = photo_discriminator_optimizer,\n        gen_loss_fn = generator_loss,\n        disc_loss_fn = discriminator_loss,\n        cycle_loss_fn = calc_cycle_loss,\n        identity_loss_fn = identity_loss\n    )\n\ncycle_gan_model.fit(\n    tf.data.Dataset.zip((monet_ds, photo_ds)),\n    epochs=100\n)","metadata":{"execution":{"iopub.status.busy":"2021-12-09T02:51:04.419664Z","iopub.status.idle":"2021-12-09T02:51:04.420424Z","shell.execute_reply.started":"2021-12-09T02:51:04.420174Z","shell.execute_reply":"2021-12-09T02:51:04.420205Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":"# Visualize final Monet-esque photos","metadata":{}},{"cell_type":"code","source":"_, ax = plt.subplots(5, 2, figsize=(12, 12))\nfor i, img in enumerate(photo_ds.take(5)):\n    prediction = monet_generator(img, training=False)[0].numpy()\n    prediction = (prediction * 127.5 + 127.5).astype(np.uint8)\n    img = (img[0] * 127.5 + 127.5).numpy().astype(np.uint8)\n\n    ax[i, 0].imshow(img)\n    ax[i, 1].imshow(prediction)\n    ax[i, 0].set_title(\"Input Photo\")\n    ax[i, 1].set_title(\"Monet-esque\")\n    ax[i, 0].axis(\"off\")\n    ax[i, 1].axis(\"off\")\nplt.savefig('final_monetesque_image.png', dpi=200)\nplt.show()","metadata":{"execution":{"iopub.status.busy":"2021-12-09T02:51:04.421451Z","iopub.status.idle":"2021-12-09T02:51:04.42208Z","shell.execute_reply.started":"2021-12-09T02:51:04.421871Z","shell.execute_reply":"2021-12-09T02:51:04.421898Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":"# Create submission file","metadata":{}},{"cell_type":"code","source":"import PIL\n! mkdir ../images","metadata":{"execution":{"iopub.status.busy":"2021-12-09T02:51:04.423287Z","iopub.status.idle":"2021-12-09T02:51:04.423636Z","shell.execute_reply.started":"2021-12-09T02:51:04.423467Z","shell.execute_reply":"2021-12-09T02:51:04.42349Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"i = 1\nfor img in photo_ds:\n    prediction = monet_generator(img, training=False)[0].numpy()\n    prediction = (prediction * 127.5 + 127.5).astype(np.uint8)\n    im = PIL.Image.fromarray(prediction)\n    im.save(\"../images/\" + str(i) + \".jpg\")\n    i += 1","metadata":{"execution":{"iopub.status.busy":"2021-12-09T02:51:04.424913Z","iopub.status.idle":"2021-12-09T02:51:04.425226Z","shell.execute_reply.started":"2021-12-09T02:51:04.425058Z","shell.execute_reply":"2021-12-09T02:51:04.425081Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":"import shutil\nshutil.make_archive(\"/kaggle/working/images\", 'zip', \"/kaggle/images\")","metadata":{"execution":{"iopub.status.busy":"2021-12-09T01:19:33.644687Z","iopub.status.idle":"2021-12-09T01:19:33.645586Z","shell.execute_reply.started":"2021-12-09T01:19:33.645359Z","shell.execute_reply":"2021-12-09T01:19:33.645387Z"}}}]}