{"metadata":{"kernelspec":{"display_name":"Python 3","language":"python","name":"python3"},"language_info":{"codemirror_mode":{"name":"ipython","version":3},"file_extension":".py","mimetype":"text/x-python","name":"python","nbconvert_exporter":"python","pygments_lexer":"ipython3","version":"3.10.14"},"kaggle":{"accelerator":"none","dataSources":[{"sourceId":21755,"databundleVersionId":1475600,"sourceType":"competition"}],"isInternetEnabled":true,"language":"python","sourceType":"notebook","isGpuEnabled":false},"papermill":{"default_parameters":{},"duration":1879.373355,"end_time":"2024-11-26T14:08:19.753321","environment_variables":{},"exception":null,"input_path":"__notebook__.ipynb","output_path":"__notebook__.ipynb","parameters":{},"start_time":"2024-11-26T13:37:00.379966","version":"2.6.0"}},"nbformat_minor":4,"nbformat":4,"cells":[{"id":"621b5990","cell_type":"code","source":"!pip3 install tensorflow-addons","metadata":{"execution":{"iopub.execute_input":"2024-11-26T13:37:07.236063Z","iopub.status.busy":"2024-11-26T13:37:07.235635Z","iopub.status.idle":"2024-11-26T13:37:17.801951Z","shell.execute_reply":"2024-11-26T13:37:17.800687Z"},"papermill":{"duration":10.587404,"end_time":"2024-11-26T13:37:17.804322","exception":false,"start_time":"2024-11-26T13:37:07.216918","status":"completed"},"tags":[]},"outputs":[],"execution_count":null},{"id":"57f923a0","cell_type":"code","source":"BATCH_SIZE = 32\n\nimport tensorflow as tf\nfrom tensorflow import keras\nfrom tensorflow.keras import layers\n# import tensorflow_addons as tfa\n\nimport matplotlib.pyplot as plt\nimport numpy as np\nimport re\ntry:\n    from kaggle_datasets import KaggleDatasets\nexcept:\n    pass","metadata":{"execution":{"iopub.execute_input":"2024-11-26T13:37:17.911208Z","iopub.status.busy":"2024-11-26T13:37:17.910889Z","iopub.status.idle":"2024-11-26T13:37:29.182401Z","shell.execute_reply":"2024-11-26T13:37:29.181372Z"},"papermill":{"duration":11.290529,"end_time":"2024-11-26T13:37:29.184720","exception":false,"start_time":"2024-11-26T13:37:17.894191","status":"completed"},"tags":[]},"outputs":[],"execution_count":null},{"id":"e0a1fb87","cell_type":"code","source":"try:\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.execute_input":"2024-11-26T13:37:29.216434Z","iopub.status.busy":"2024-11-26T13:37:29.215756Z","iopub.status.idle":"2024-11-26T13:37:29.222721Z","shell.execute_reply":"2024-11-26T13:37:29.221842Z"},"papermill":{"duration":0.025096,"end_time":"2024-11-26T13:37:29.225023","exception":false,"start_time":"2024-11-26T13:37:29.199927","status":"completed"},"tags":[]},"outputs":[],"execution_count":null},{"id":"7562484f","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'))\nPHOTO_FILENAMES = tf.io.gfile.glob(str(GCS_PATH + '/photo_tfrec/*.tfrec'))","metadata":{"execution":{"iopub.execute_input":"2024-11-26T13:37:29.255478Z","iopub.status.busy":"2024-11-26T13:37:29.255156Z","iopub.status.idle":"2024-11-26T13:37:30.706386Z","shell.execute_reply":"2024-11-26T13:37:30.705237Z"},"papermill":{"duration":1.469969,"end_time":"2024-11-26T13:37:30.709350","exception":false,"start_time":"2024-11-26T13:37:29.239381","status":"completed"},"tags":[]},"outputs":[],"execution_count":null},{"id":"a38bcf63","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\": 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.execute_input":"2024-11-26T13:37:30.739841Z","iopub.status.busy":"2024-11-26T13:37:30.739472Z","iopub.status.idle":"2024-11-26T13:37:30.745746Z","shell.execute_reply":"2024-11-26T13:37:30.744551Z"},"papermill":{"duration":0.023946,"end_time":"2024-11-26T13:37:30.747884","exception":false,"start_time":"2024-11-26T13:37:30.723938","status":"completed"},"tags":[]},"outputs":[],"execution_count":null},{"id":"8f6c0a27","cell_type":"code","source":"def load_dataset(filenames):\n    dataset = tf.data.TFRecordDataset(filenames)\n    dataset = dataset.map(read_tfrecord, num_parallel_calls=AUTOTUNE)\n    return dataset\n\nmonet_ds = load_dataset(MONET_FILENAMES).batch(1)\nphoto_ds = load_dataset(PHOTO_FILENAMES).batch(1)\n\n\nfast_photo_ds = load_dataset(PHOTO_FILENAMES).batch(32*strategy.num_replicas_in_sync).prefetch(32)\nfid_photo_ds = load_dataset(PHOTO_FILENAMES).take(1024).batch(32*strategy.num_replicas_in_sync).prefetch(32)\nfid_monet_ds = load_dataset(MONET_FILENAMES).batch(32*strategy.num_replicas_in_sync).prefetch(32)","metadata":{"execution":{"iopub.execute_input":"2024-11-26T13:37:30.779067Z","iopub.status.busy":"2024-11-26T13:37:30.778152Z","iopub.status.idle":"2024-11-26T13:37:31.015254Z","shell.execute_reply":"2024-11-26T13:37:31.014138Z"},"papermill":{"duration":0.255023,"end_time":"2024-11-26T13:37:31.017658","exception":false,"start_time":"2024-11-26T13:37:30.762635","status":"completed"},"tags":[]},"outputs":[],"execution_count":null},{"id":"4dbddecf","cell_type":"code","source":"\n# def get_gan_dataset(monet_files, photo_files, augment=None, repeat=True, shuffle=True, batch_size=1):\n\n#     monet_ds = load_dataset(monet_files)\n#     photo_ds = load_dataset(photo_files)\n    \n#     if augment:\n#         monet_ds = monet_ds.map(augment, num_parallel_calls=AUTOTUNE)\n#         photo_ds = photo_ds.map(augment, num_parallel_calls=AUTOTUNE)\n\n#     if repeat:\n#         monet_ds = monet_ds.repeat()\n#         photo_ds = photo_ds.repeat()\n        \n#     # if shuffle:\n#     #     monet_ds = monet_ds.shuffle(2048)\n#     #     photo_ds = photo_ds.shuffle(2048)\n        \n#     monet_ds = monet_ds.batch(batch_size, drop_remainder=True)\n#     photo_ds = photo_ds.batch(batch_size, drop_remainder=True)\n# #     monet_ds = monet_ds.cache()\n# #     photo_ds = photo_ds.cache()\n#     monet_ds = monet_ds.prefetch(AUTOTUNE)\n#     photo_ds = photo_ds.prefetch(AUTOTUNE)\n    \n#     gan_ds = tf.data.Dataset.zip((monet_ds, photo_ds))\n    \n#     return gan_ds","metadata":{"execution":{"iopub.execute_input":"2024-11-26T13:37:31.048497Z","iopub.status.busy":"2024-11-26T13:37:31.048117Z","iopub.status.idle":"2024-11-26T13:37:31.052797Z","shell.execute_reply":"2024-11-26T13:37:31.051893Z"},"papermill":{"duration":0.022403,"end_time":"2024-11-26T13:37:31.054734","exception":false,"start_time":"2024-11-26T13:37:31.032331","status":"completed"},"tags":[]},"outputs":[],"execution_count":null},{"id":"f0e17e0a","cell_type":"code","source":"import tensorflow as tf\n\n# Ensure AUTOTUNE is defined\nAUTOTUNE = tf.data.AUTOTUNE\n\ndef get_gan_dataset(monet_files, photo_files, augment=None, repeat=True, shuffle=True, batch_size=1):\n    # Load datasets\n    monet_ds = load_dataset(monet_files)\n    photo_ds = load_dataset(photo_files)\n    \n    # Apply augmentation if provided\n    if augment:\n        monet_ds = monet_ds.map(augment, num_parallel_calls=AUTOTUNE)\n        photo_ds = photo_ds.map(augment, num_parallel_calls=AUTOTUNE)\n\n    # Repeat the datasets if needed\n    if repeat:\n        monet_ds = monet_ds.repeat()\n        photo_ds = photo_ds.repeat()\n\n    # Shuffle datasets if needed\n    if shuffle:\n        monet_ds = monet_ds.shuffle(2048)\n        photo_ds = photo_ds.shuffle(2048)\n        \n    # Batch the datasets\n    monet_ds = monet_ds.batch(batch_size, drop_remainder=True)\n    photo_ds = photo_ds.batch(batch_size, drop_remainder=True)\n    \n    # Prefetch for performance optimization\n    monet_ds = monet_ds.prefetch(AUTOTUNE)\n    photo_ds = photo_ds.prefetch(AUTOTUNE)\n    \n    # Zip the datasets together (pairing Monet and photo images)\n    gan_ds = tf.data.Dataset.zip((monet_ds, photo_ds))\n    \n    return gan_ds\n\ndef load_dataset(file_paths):\n    # Load image files into a TensorFlow dataset\n    def _parse_image(image_path):\n        image = tf.io.read_file(image_path)\n        image = tf.image.decode_jpeg(image, channels=3)\n        image = tf.image.resize(image, [256, 256])  # Resize images to 256x256\n        image = (image / 127.5) - 1  # Normalize to [-1, 1] range\n        return image\n    \n    # Create a TensorFlow Dataset from file paths\n    ds = tf.data.Dataset.from_tensor_slices(file_paths)\n    ds = ds.map(_parse_image, num_parallel_calls=AUTOTUNE)\n    return ds","metadata":{"execution":{"iopub.execute_input":"2024-11-26T13:37:31.086704Z","iopub.status.busy":"2024-11-26T13:37:31.085720Z","iopub.status.idle":"2024-11-26T13:37:31.094292Z","shell.execute_reply":"2024-11-26T13:37:31.093309Z"},"papermill":{"duration":0.026975,"end_time":"2024-11-26T13:37:31.096410","exception":false,"start_time":"2024-11-26T13:37:31.069435","status":"completed"},"tags":[]},"outputs":[],"execution_count":null},{"id":"8d48677a","cell_type":"code","source":"final_dataset = get_gan_dataset(MONET_FILENAMES, PHOTO_FILENAMES, augment=None, repeat=True, shuffle=True, batch_size=BATCH_SIZE)","metadata":{"execution":{"iopub.execute_input":"2024-11-26T13:37:31.127618Z","iopub.status.busy":"2024-11-26T13:37:31.127300Z","iopub.status.idle":"2024-11-26T13:37:31.206699Z","shell.execute_reply":"2024-11-26T13:37:31.205699Z"},"papermill":{"duration":0.097806,"end_time":"2024-11-26T13:37:31.208908","exception":false,"start_time":"2024-11-26T13:37:31.111102","status":"completed"},"tags":[]},"outputs":[],"execution_count":null},{"id":"84aeb40c","cell_type":"code","source":"with strategy.scope():\n\n    inception_model = tf.keras.applications.InceptionV3(input_shape=(256,256,3),pooling=\"avg\",include_top=False)\n\n    mix3  = inception_model.get_layer(\"mixed9\").output\n    f0 = tf.keras.layers.GlobalAveragePooling2D()(mix3)\n\n    inception_model = tf.keras.Model(inputs=inception_model.input, outputs=f0)\n    inception_model.trainable = False\n\n    \n    \n    def calculate_activation_statistics_mod(images,fid_model):\n\n            act=tf.cast(fid_model.predict(images), tf.float32)\n\n            mu = tf.reduce_mean(act, axis=0)\n            mean_x = tf.reduce_mean(act, axis=0, keepdims=True)\n            mx = tf.matmul(tf.transpose(mean_x), mean_x)\n            vx = tf.matmul(tf.transpose(act), act)/tf.cast(tf.shape(act)[0], tf.float32)\n            sigma = vx - mx\n            return mu, sigma\n    myFID_mu2, myFID_sigma2 = calculate_activation_statistics_mod(fid_monet_ds,inception_model)        \n    fids=[]","metadata":{"execution":{"iopub.execute_input":"2024-11-26T13:37:31.239659Z","iopub.status.busy":"2024-11-26T13:37:31.239310Z","iopub.status.idle":"2024-11-26T13:38:08.810862Z","shell.execute_reply":"2024-11-26T13:38:08.809750Z"},"papermill":{"duration":37.589656,"end_time":"2024-11-26T13:38:08.813247","exception":false,"start_time":"2024-11-26T13:37:31.223591","status":"completed"},"tags":[]},"outputs":[],"execution_count":null},{"id":"ababa3c5","cell_type":"code","source":"with strategy.scope():\n    def calculate_frechet_distance(mu1,sigma1,mu2,sigma2):\n        fid_epsilon = 1e-14\n       \n        covmean = tf.linalg.sqrtm(tf.cast(tf.matmul(sigma1,sigma2),tf.complex64))\n#         isgood=tf.cast(tf.math.is_finite(covmean), tf.int32)\n#         if tf.size(isgood)!=tf.math.reduce_sum(isgood):\n#             return 0\n\n        covmean = tf.cast(tf.math.real(covmean),tf.float32)\n  \n        tr_covmean = tf.linalg.trace(covmean)\n\n\n        return tf.matmul(tf.expand_dims(mu1 - mu2, axis=0),tf.expand_dims(mu1 - mu2, axis=1)) + tf.linalg.trace(sigma1) + tf.linalg.trace(sigma2) - 2 * tr_covmean\n\n\n    \n    \n    def FID(images,gen_model,inception_model=inception_model,myFID_mu2=myFID_mu2, myFID_sigma2=myFID_sigma2):\n                inp = layers.Input(shape=[256, 256, 3], name='input_image')\n                x  = gen_model(inp)\n                x=inception_model(x)\n                fid_model = tf.keras.Model(inputs=inp, outputs=x)\n                \n                mu1, sigma1= calculate_activation_statistics_mod(images,fid_model)\n\n                fid_value = calculate_frechet_distance(mu1, sigma1,myFID_mu2, myFID_sigma2)\n\n\n                return fid_value","metadata":{"execution":{"iopub.execute_input":"2024-11-26T13:38:08.852883Z","iopub.status.busy":"2024-11-26T13:38:08.852527Z","iopub.status.idle":"2024-11-26T13:38:08.860406Z","shell.execute_reply":"2024-11-26T13:38:08.859331Z"},"papermill":{"duration":0.030584,"end_time":"2024-11-26T13:38:08.862506","exception":false,"start_time":"2024-11-26T13:38:08.831922","status":"completed"},"tags":[]},"outputs":[],"execution_count":null},{"id":"e2bbad6e","cell_type":"code","source":"OUTPUT_CHANNELS = 3\n\ndef down_sample(filters, size, apply_instancenorm=True):\n    initializer = tf.random_normal_initializer(0., 0.02)\n    gamma_init = keras.initializers.RandomNormal(mean=0.0, stddev=0.02)\n\n    layer = keras.Sequential()\n    layer.add(layers.Conv2D(filters, size, strides=2, padding='same', kernel_initializer=initializer, use_bias=False))\n\n    if apply_instancenorm:\n        # layer.add(tfa.layers.InstanceNormalization(gamma_initializer=gamma_init))\n        layer.add(layers.LayerNormalization(gamma_initializer=gamma_init))\n\n    layer.add(layers.LeakyReLU())\n\n    return layer","metadata":{"execution":{"iopub.execute_input":"2024-11-26T13:38:08.901161Z","iopub.status.busy":"2024-11-26T13:38:08.900228Z","iopub.status.idle":"2024-11-26T13:38:08.906562Z","shell.execute_reply":"2024-11-26T13:38:08.905510Z"},"papermill":{"duration":0.027835,"end_time":"2024-11-26T13:38:08.908629","exception":false,"start_time":"2024-11-26T13:38:08.880794","status":"completed"},"tags":[]},"outputs":[],"execution_count":null},{"id":"aeac2ef1","cell_type":"code","source":"def up_sample(filters, size, apply_dropout=False):\n    initializer = tf.random_normal_initializer(0., 0.02)\n    gamma_init = keras.initializers.RandomNormal(mean=0.0, stddev=0.02)\n\n    layer = keras.Sequential()\n    layer.add(layers.Conv2DTranspose(filters, size, strides=2, padding='same', kernel_initializer=initializer,use_bias=False))\n    # layer.add(tfa.layers.InstanceNormalization(gamma_initializer=gamma_init))\n    layer.add(layers.LayerNormalization(gamma_initializer=gamma_init))\n\n    if apply_dropout:\n        layer.add(layers.Dropout(0.5))\n\n    layer.add(layers.ReLU())\n\n    return layer","metadata":{"execution":{"iopub.execute_input":"2024-11-26T13:38:08.948063Z","iopub.status.busy":"2024-11-26T13:38:08.947628Z","iopub.status.idle":"2024-11-26T13:38:08.954140Z","shell.execute_reply":"2024-11-26T13:38:08.952963Z"},"papermill":{"duration":0.029333,"end_time":"2024-11-26T13:38:08.956034","exception":false,"start_time":"2024-11-26T13:38:08.926701","status":"completed"},"tags":[]},"outputs":[],"execution_count":null},{"id":"72114b92","cell_type":"code","source":"def Generator():\n    inputs = layers.Input(shape=[256,256,3])\n    down_stack = [\n        down_sample(64, 4, apply_instancenorm=False),# (size, 128, 128, 64)\n        down_sample(128, 4),                         # (size, 64, 64, 128)\n        down_sample(256, 4),                         # (size, 32, 32, 256)\n        down_sample(512, 4),                         # (size, 16, 16, 512)\n        down_sample(512, 4),                         # (size, 8, 8, 512)\n        down_sample(512, 4),                         # (size, 4, 4, 512)\n        down_sample(512, 4),                         # (size, 2, 2, 512)\n        down_sample(512, 4),                         # (size, 1, 1, 512)\n    ]\n\n    up_stack = [\n        up_sample(512, 4, apply_dropout=True),       # (size, 2, 2, 1024)\n        up_sample(512, 4, apply_dropout=True),       # (size, 4, 4, 1024)\n        up_sample(512, 4, apply_dropout=True),       # (size, 8, 8, 1024)\n        up_sample(512, 4),                           # (size, 16, 16, 1024)\n        up_sample(256, 4),                           # (size, 32, 32, 512)\n        up_sample(128, 4),                           # (size, 64, 64, 256)\n        up_sample(64, 4),                            # (size, 128, 128, 128)\n    ]\n\n    initializer = tf.random_normal_initializer(0., 0.02)\n    last = layers.Conv2DTranspose(3, 4, strides=2, padding='same', kernel_initializer=initializer, activation='tanh') \n    # (size, 256, 256, 3)\n\n    x = inputs\n\n    # Downsampling through the model\n    skips = []\n    for down in down_stack:\n        x = down(x)\n        skips.append(x)\n\n    skips = reversed(skips[:-1])\n\n    # Upsampling and establishing the skip connections\n    for up, skip in zip(up_stack, skips):\n        x = up(x)\n        x = layers.Concatenate()([x, skip])\n\n    x = last(x)\n\n    return keras.Model(inputs=inputs, outputs=x)","metadata":{"execution":{"iopub.execute_input":"2024-11-26T13:38:08.994728Z","iopub.status.busy":"2024-11-26T13:38:08.994371Z","iopub.status.idle":"2024-11-26T13:38:09.003993Z","shell.execute_reply":"2024-11-26T13:38:09.002857Z"},"papermill":{"duration":0.030761,"end_time":"2024-11-26T13:38:09.005870","exception":false,"start_time":"2024-11-26T13:38:08.975109","status":"completed"},"tags":[]},"outputs":[],"execution_count":null},{"id":"55ec86b2","cell_type":"code","source":"def Discriminator():\n    initializer = tf.random_normal_initializer(0., 0.02)\n    gamma_init = keras.initializers.RandomNormal(mean=0.0, stddev=0.02)\n    \n    inp = layers.Input(shape=[256, 256, 3], name='input_image')\n    x = inp\n    \n    down1 = down_sample(64, 4, False)(x)       # (size, 128, 128, 64)\n    down2 = down_sample(128, 4)(down1)         # (size, 64, 64, 128)\n    down3 = down_sample(256, 4)(down2)         # (size, 32, 32, 256)\n\n    zero_pad1 = layers.ZeroPadding2D()(down3) # (size, 34, 34, 256)\n    conv = layers.Conv2D(512, 4, strides=1, kernel_initializer=initializer, use_bias=False)(zero_pad1) # (size, 31, 31, 512)\n\n    # norm1 = tfa.layers.InstanceNormalization(gamma_initializer=gamma_init)(conv)\n    norm1 = layers.LayerNormalization(gamma_initializer=gamma_init)(conv)\n    leaky_relu = layers.LeakyReLU()(norm1)\n    zero_pad2 = layers.ZeroPadding2D()(leaky_relu) # (size, 33, 33, 512)\n\n    return tf.keras.Model(inputs=inp, outputs=zero_pad2)","metadata":{"execution":{"iopub.execute_input":"2024-11-26T13:38:09.044500Z","iopub.status.busy":"2024-11-26T13:38:09.044155Z","iopub.status.idle":"2024-11-26T13:38:09.050893Z","shell.execute_reply":"2024-11-26T13:38:09.049890Z"},"papermill":{"duration":0.028993,"end_time":"2024-11-26T13:38:09.053016","exception":false,"start_time":"2024-11-26T13:38:09.024023","status":"completed"},"tags":[]},"outputs":[],"execution_count":null},{"id":"06ad207e","cell_type":"code","source":"def DHead():\n    initializer = tf.random_normal_initializer(0., 0.02)\n    \n    inp = layers.Input(shape=[33, 33, 512], name='input_image')\n    x = inp\n    \n    last = layers.Conv2D(1, 4, strides=1, kernel_initializer=initializer)(x) # (size, 30, 30, 1)\n\n    return tf.keras.Model(inputs=inp, outputs=last)","metadata":{"execution":{"iopub.execute_input":"2024-11-26T13:38:09.092709Z","iopub.status.busy":"2024-11-26T13:38:09.092281Z","iopub.status.idle":"2024-11-26T13:38:09.098352Z","shell.execute_reply":"2024-11-26T13:38:09.096999Z"},"papermill":{"duration":0.028481,"end_time":"2024-11-26T13:38:09.100556","exception":false,"start_time":"2024-11-26T13:38:09.072075","status":"completed"},"tags":[]},"outputs":[],"execution_count":null},{"id":"03f90fa6","cell_type":"code","source":"def DiscriminatorP():\n    initializer = tf.random_normal_initializer(0., 0.02)\n    gamma_init = keras.initializers.RandomNormal(mean=0.0, stddev=0.02)\n    \n    inp = layers.Input(shape=[256, 256, 3], name='input_image')\n    x = inp\n    \n    down1 = down_sample(64, 4, False)(x)       # (size, 128, 128, 64)\n    down2 = down_sample(128, 4)(down1)         # (size, 64, 64, 128)\n    down3 = down_sample(256, 4)(down2)         # (size, 32, 32, 256)\n\n    zero_pad1 = layers.ZeroPadding2D()(down3) # (size, 34, 34, 256)\n    conv = layers.Conv2D(512, 4, strides=1, kernel_initializer=initializer, use_bias=False)(zero_pad1) # (size, 31, 31, 512)\n\n    # norm1 = tfa.layers.InstanceNormalization(gamma_initializer=gamma_init)(conv)\n    norm1 = layers.LayerNormalization(gamma_initializer=gamma_init)(conv)\n    leaky_relu = layers.LeakyReLU()(norm1)\n    zero_pad2 = layers.ZeroPadding2D()(leaky_relu) # (size, 33, 33, 512)\n    last = layers.Conv2D(1, 4, strides=1, kernel_initializer=initializer)(zero_pad2) # (size, 30, 30, 1)\n\n    return tf.keras.Model(inputs=inp, outputs=last)","metadata":{"execution":{"iopub.execute_input":"2024-11-26T13:38:09.139785Z","iopub.status.busy":"2024-11-26T13:38:09.138798Z","iopub.status.idle":"2024-11-26T13:38:09.145971Z","shell.execute_reply":"2024-11-26T13:38:09.145023Z"},"papermill":{"duration":0.028755,"end_time":"2024-11-26T13:38:09.147957","exception":false,"start_time":"2024-11-26T13:38:09.119202","status":"completed"},"tags":[]},"outputs":[],"execution_count":null},{"id":"2fed09a3","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 = DiscriminatorP() # differentiates real photos and generated photos\n    dHead1 = DHead() # Head for BCE\n    dHead2 = DHead() # Head for hinge loss","metadata":{"execution":{"iopub.execute_input":"2024-11-26T13:38:09.185583Z","iopub.status.busy":"2024-11-26T13:38:09.184759Z","iopub.status.idle":"2024-11-26T13:38:11.201798Z","shell.execute_reply":"2024-11-26T13:38:11.200750Z"},"papermill":{"duration":2.038001,"end_time":"2024-11-26T13:38:11.203913","exception":false,"start_time":"2024-11-26T13:38:09.165912","status":"completed"},"tags":[]},"outputs":[],"execution_count":null},{"id":"6790c3e1","cell_type":"code","source":"","metadata":{"papermill":{"duration":0.019546,"end_time":"2024-11-26T13:38:11.248373","exception":false,"start_time":"2024-11-26T13:38:11.228827","status":"completed"},"tags":[]},"outputs":[],"execution_count":null},{"id":"5a88e787","cell_type":"code","source":"# BATCH_SIZE = 32\n\n# import tensorflow as tf\n# from tensorflow import keras\n# from tensorflow.keras import layers\n# # Remove tfa import since it's not needed anymore\n# # import tensorflow_addons as tfa\n\n# import matplotlib.pyplot as plt\n# import numpy as np\n# import re\n# try:\n#     from kaggle_datasets import KaggleDatasets\n# except:\n#     pass\n\n\n# try:\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)\n# except:\n#     strategy = tf.distribute.get_strategy()\n# print('Number of replicas:', strategy.num_replicas_in_sync)\n\n# AUTOTUNE = tf.data.experimental.AUTOTUNE\n\n# print(tf.__version__)\n\n# GCS_PATH = KaggleDatasets().get_gcs_path(\"gan-getting-started\")\n\n# MONET_FILENAMES = tf.io.gfile.glob(str(GCS_PATH + '/monet_tfrec/*.tfrec'))\n# PHOTO_FILENAMES = tf.io.gfile.glob(str(GCS_PATH + '/photo_tfrec/*.tfrec'))\n\n# IMAGE_SIZE = [256, 256]\n\n# def 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\n# def read_tfrecord(example):\n#     tfrecord_format = {\n#         \"image\": 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\n\n\n# def load_dataset(filenames):\n#     dataset = tf.data.TFRecordDataset(filenames)\n#     dataset = dataset.map(read_tfrecord, num_parallel_calls=AUTOTUNE)\n#     return dataset\n\n# monet_ds = load_dataset(MONET_FILENAMES).batch(1)\n# photo_ds = load_dataset(PHOTO_FILENAMES).batch(1)\n\n# fast_photo_ds = load_dataset(PHOTO_FILENAMES).batch(32*strategy.num_replicas_in_sync).prefetch(32)\n# fid_photo_ds = load_dataset(PHOTO_FILENAMES).take(1024).batch(32*strategy.num_replicas_in_sync).prefetch(32)\n# fid_monet_ds = load_dataset(MONET_FILENAMES).batch(32*strategy.num_replicas_in_sync).prefetch(32)\n\n\n# def get_gan_dataset(monet_files, photo_files, augment=None, repeat=True, shuffle=True, batch_size=1):\n\n#     monet_ds = load_dataset(monet_files)\n#     photo_ds = load_dataset(photo_files)\n    \n#     if augment:\n#         monet_ds = monet_ds.map(augment, num_parallel_calls=AUTOTUNE)\n#         photo_ds = photo_ds.map(augment, num_parallel_calls=AUTOTUNE)\n\n#     if repeat:\n#         monet_ds = monet_ds.repeat()\n#         photo_ds = photo_ds.repeat()\n\n#     monet_ds = monet_ds.batch(batch_size, drop_remainder=True)\n#     photo_ds = photo_ds.batch(batch_size, drop_remainder=True)\n#     monet_ds = monet_ds.prefetch(AUTOTUNE)\n#     photo_ds = photo_ds.prefetch(AUTOTUNE)\n    \n#     gan_ds = tf.data.Dataset.zip((monet_ds, photo_ds))\n    \n#     return gan_ds\n\n\n# final_dataset = get_gan_dataset(MONET_FILENAMES, PHOTO_FILENAMES, augment=None, repeat=True, shuffle=True, batch_size=BATCH_SIZE)\n\n# with strategy.scope():\n\n#     inception_model = tf.keras.applications.InceptionV3(input_shape=(256,256,3),pooling=\"avg\",include_top=False)\n\n#     mix3  = inception_model.get_layer(\"mixed9\").output\n#     f0 = tf.keras.layers.GlobalAveragePooling2D()(mix3)\n\n#     inception_model = tf.keras.Model(inputs=inception_model.input, outputs=f0)\n#     inception_model.trainable = False\n\n    \n    \n#     def calculate_activation_statistics_mod(images,fid_model):\n\n#             act=tf.cast(fid_model.predict(images), tf.float32)\n\n#             mu = tf.reduce_mean(act, axis=0)\n#             mean_x = tf.reduce_mean(act, axis=0, keepdims=True)\n#             mx = tf.matmul(tf.transpose(mean_x), mean_x)\n#             vx = tf.matmul(tf.transpose(act), act)/tf.cast(tf.shape(act)[0], tf.float32)\n#             sigma = vx - mx\n#             return mu, sigma\n#     myFID_mu2, myFID_sigma2 = calculate_activation_statistics_mod(fid_monet_ds,inception_model)        \n#     fids=[]\n\n# with strategy.scope():\n#     def calculate_frechet_distance(mu1,sigma1,mu2,sigma2):\n#         fid_epsilon = 1e-14\n       \n#         covmean = tf.linalg.sqrtm(tf.cast(tf.matmul(sigma1,sigma2),tf.complex64))\n#         covmean = tf.cast(tf.math.real(covmean),tf.float32)\n#         tr_covmean = tf.linalg.trace(covmean)\n\n#         return tf.matmul(tf.expand_dims(mu1 - mu2, axis=0),tf.expand_dims(mu1 - mu2, axis=1)) + tf.linalg.trace(sigma1) + tf.linalg.trace(sigma2) - 2 * tr_covmean\n\n\n    \n#     def FID(images,gen_model,inception_model=inception_model,myFID_mu2=myFID_mu2, myFID_sigma2=myFID_sigma2):\n#                 inp = layers.Input(shape=[256, 256, 3], name='input_image')\n#                 x  = gen_model(inp)\n#                 x=inception_model(x)\n#                 fid_model = tf.keras.Model(inputs=inp, outputs=x)\n                \n#                 mu1, sigma1= calculate_activation_statistics_mod(images,fid_model)\n\n#                 fid_value = calculate_frechet_distance(mu1, sigma1,myFID_mu2, myFID_sigma2)\n\n#                 return fid_value\n\n# OUTPUT_CHANNELS = 3\n\n# def down_sample(filters, size, apply_instancenorm=True):\n#     initializer = tf.random_normal_initializer(0., 0.02)\n#     gamma_init = keras.initializers.RandomNormal(mean=0.0, stddev=0.02)\n\n#     layer = keras.Sequential()\n#     layer.add(layers.Conv2D(filters, size, strides=2, padding='same', kernel_initializer=initializer, use_bias=False))\n\n#     if apply_instancenorm:\n#         # Replace tfa layers with tf.keras.LayerNormalization\n#         layer.add(layers.LayerNormalization(gamma_initializer=gamma_init))  # Equivalent to InstanceNormalization\n#     layer.add(layers.LeakyReLU())\n\n#     return layer\n\n\n# def up_sample(filters, size, apply_dropout=False):\n#     initializer = tf.random_normal_initializer(0., 0.02)\n#     gamma_init = keras.initializers.RandomNormal(mean=0.0, stddev=0.02)\n\n#     layer = keras.Sequential()\n#     layer.add(layers.Conv2DTranspose(filters, size, strides=2, padding='same', kernel_initializer=initializer,use_bias=False))\n#     layer.add(layers.LayerNormalization(gamma_initializer=gamma_init))  # Equivalent to InstanceNormalization\n\n#     if apply_dropout:\n#         layer.add(layers.Dropout(0.5))\n\n#     layer.add(layers.ReLU())\n\n#     return layer\n\n\n# def Generator():\n#     inputs = layers.Input(shape=[256,256,3])\n#     down_stack = [\n#         down_sample(64, 4, apply_instancenorm=False),  # (size, 128, 128, 64)\n#         down_sample(128, 4),  # (size, 64, 64, 128)\n#         down_sample(256, 4),  # (size, 32, 32, 256)\n#         down_sample(512, 4),  # (size, 16, 16, 512)\n#         down_sample(512, 4),  # (size, 8, 8, 512)\n#         down_sample(512, 4),  # (size, 4, 4, 512)\n#         down_sample(512, 4),  # (size, 2, 2, 512)\n#         down_sample(512, 4),  # (size, 1, 1, 512)\n#     ]\n\n#     up_stack = [\n#         up_sample(512, 4, apply_dropout=True),  # (size, 2, 2, 1024)\n#         up_sample(512, 4, apply_dropout=True),  # (size, 4, 4, 1024)\n#         up_sample(512, 4, apply_dropout=True),  # (size, 8, 8, 1024)\n#         up_sample(512, 4),  # (size, 16, 16, 1024)\n#         up_sample(256, 4),  # (size, 32, 32, 512)\n#         up_sample(128, 4),  # (size, 64, 64, 256)\n#         up_sample(64, 4),  # (size, 128, 128, 128)\n#     ]\n\n#     initializer = tf.random_normal_initializer(0., 0.02)\n#     last = layers.Conv2DTranspose(3, 4, strides=2, padding='same', kernel_initializer=initializer, activation='tanh')\n\n#     x = inputs\n#     for down in down_stack:\n#         x = down(x)\n\n#     for up in up_stack:\n#         x = up(x)\n\n#     x = last(x)\n\n#     return tf.keras.Model(inputs=inputs, outputs=x)\n\n# def Discriminator():\n#     inp = layers.Input(shape=[256, 256, 3], name='input_image')\n\n#     down_stack = [\n#         down_sample(64, 4, apply_instancenorm=False),  # (size, 128, 128, 64)\n#         down_sample(128, 4),  # (size, 64, 64, 128)\n#         down_sample(256, 4),  # (size, 32, 32, 256)\n#         down_sample(512, 4),  # (size, 16, 16, 512)\n#         down_sample(512, 4),  # (size, 8, 8, 512)\n#         down_sample(512, 4),  # (size, 4, 4, 512)\n#         down_sample(512, 4),  # (size, 2, 2, 512)\n#     ]\n    \n#     last = layers.Conv2D(1, 4, strides=1, padding='same')\n\n#     x = inp\n#     for down in down_stack:\n#         x = down(x)\n\n#     x = last(x)\n\n#     return tf.keras.Model(inputs=inp, outputs=x)","metadata":{"execution":{"iopub.execute_input":"2024-11-26T13:38:11.286734Z","iopub.status.busy":"2024-11-26T13:38:11.286408Z","iopub.status.idle":"2024-11-26T13:38:11.295460Z","shell.execute_reply":"2024-11-26T13:38:11.294343Z"},"papermill":{"duration":0.031243,"end_time":"2024-11-26T13:38:11.297580","exception":false,"start_time":"2024-11-26T13:38:11.266337","status":"completed"},"tags":[]},"outputs":[],"execution_count":null},{"id":"16bf383e","cell_type":"code","source":"","metadata":{"papermill":{"duration":0.017981,"end_time":"2024-11-26T13:38:11.334823","exception":false,"start_time":"2024-11-26T13:38:11.316842","status":"completed"},"tags":[]},"outputs":[],"execution_count":null},{"id":"2614648f","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        dhead1,        \n        dhead2,        \n        lambda_cycle=3,\n        lambda_id=3,\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        self.lambda_id = lambda_id\n        self.dhead1 = dhead1\n        self.dhead2 = dhead2\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_fn1,\n        gen_loss_fn2,\n        disc_loss_fn1,\n        disc_loss_fn2,\n        cycle_loss_fn,\n        identity_loss_fn,\n        aug_fn,\n\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_fn1 = gen_loss_fn1\n        self.gen_loss_fn2 = gen_loss_fn2\n        self.disc_loss_fn1 = disc_loss_fn1\n        self.disc_loss_fn2 = disc_loss_fn2\n        self.cycle_loss_fn = cycle_loss_fn\n        self.identity_loss_fn = identity_loss_fn\n        self.aug_fn = aug_fn\n\n        self.step_num = 0\n        \n    def train_step(self, batch_data):\n        real_monet, real_photo = batch_data\n        batch_size = tf.shape(real_monet)[0]\n        with tf.GradientTape(persistent=True) as tape:\n        \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            \n            \n            # Diffaugment\n            both_monet = tf.concat([real_monet, fake_monet], axis=0)            \n            \n            aug_monet = self.aug_fn(both_monet)\n            \n            aug_real_monet = aug_monet[:batch_size]\n            aug_fake_monet = aug_monet[batch_size:]\n            \n            \n            # two-objective discriminator\n            disc_fake_monet1 = self.dhead1(self.m_disc(aug_fake_monet, training=True), training=True)\n            disc_real_monet1 = self.dhead1(self.m_disc(aug_real_monet, training=True), training=True)\n            disc_fake_monet2 = self.dhead2(self.m_disc(aug_fake_monet, training=True), training=True)\n            disc_real_monet2 = self.dhead2(self.m_disc(aug_real_monet, training=True), training=True)\n\n            monet_gen_loss1 = self.gen_loss_fn1(disc_fake_monet1) \n            monet_head_loss1 = self.disc_loss_fn1(disc_real_monet1, disc_fake_monet1)\n            monet_gen_loss2 = self.gen_loss_fn2(disc_fake_monet2)\n            monet_head_loss2 = self.disc_loss_fn2(disc_real_monet2, disc_fake_monet2)\n\n\n\n            monet_gen_loss = (monet_gen_loss1 + monet_gen_loss2) * 0.4\n            monet_disc_loss = monet_head_loss1 + monet_head_loss2\n\n\n           \n            \n            # discriminator used to check, inputing real images\n\n            disc_real_photo = self.p_disc(real_photo, training=True)\n            # discriminator used to check, inputing fake images\n\n            disc_fake_photo = self.p_disc(fake_photo, training=True)\n\n            # evaluates generator loss\n\n            photo_gen_loss = self.gen_loss_fn1(disc_fake_photo)\n            \n            # evaluates discriminator loss\n\n            photo_disc_loss = self.disc_loss_fn1(disc_real_photo, disc_fake_photo)\n\n\n            # evaluates total generator loss\n            total_cycle_loss = self.cycle_loss_fn(real_monet, cycled_monet, self.lambda_cycle/ tf.cast(batch_size,tf.float32)) + self.cycle_loss_fn(real_photo, cycled_photo, self.lambda_cycle/ tf.cast(batch_size,tf.float32))\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_id / tf.cast(batch_size,tf.float32))\n            total_photo_gen_loss = photo_gen_loss + total_cycle_loss + self.identity_loss_fn(real_photo, same_photo, self.lambda_id/ tf.cast(batch_size,tf.float32))\n\n\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\n        # Heads gradients\n        monet_head_gradients = tape.gradient(monet_head_loss1,\n                                                      self.dhead1.trainable_variables)\n\n        self.m_disc_optimizer.apply_gradients(zip(monet_head_gradients,\n                                                  self.dhead1.trainable_variables))       \n\n        monet_head_gradients = tape.gradient(monet_head_loss2,\n                                                      self.dhead2.trainable_variables)\n        self.m_disc_optimizer.apply_gradients(zip(monet_head_gradients,\n                                                  self.dhead2.trainable_variables))     \n        \n        \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_head_loss1\": monet_head_loss1, \n            \"monet_head_loss2\": monet_head_loss2, \n            \"disc_real_monet\": disc_real_monet1, \n            \"disc_fake_monet\": disc_fake_monet1, \n            \"disc_real_monet2\": disc_real_monet2, \n            \"disc_fake_monet2\": disc_fake_monet2, \n            \"monet_gen_loss\": monet_gen_loss, \n            \"photo_disc_loss\": photo_disc_loss, \n            }","metadata":{"execution":{"iopub.execute_input":"2024-11-26T13:38:11.373010Z","iopub.status.busy":"2024-11-26T13:38:11.372623Z","iopub.status.idle":"2024-11-26T13:38:11.390771Z","shell.execute_reply":"2024-11-26T13:38:11.389709Z"},"papermill":{"duration":0.040132,"end_time":"2024-11-26T13:38:11.393021","exception":false,"start_time":"2024-11-26T13:38:11.352889","status":"completed"},"tags":[]},"outputs":[],"execution_count":null},{"id":"e2553858","cell_type":"code","source":"with strategy.scope(): # for TPU\n\n    def discriminator_loss1(real, generated):\n        real_loss = tf.math.minimum(tf.zeros_like(real), real-tf.ones_like(real))\n\n        generated_loss = tf.math.minimum(tf.zeros_like(generated), -generated-tf.ones_like(generated))\n\n        total_disc_loss = real_loss + generated_loss\n\n        return tf.reduce_mean(-total_disc_loss * 0.5)\n\n    def discriminator_loss2(real, generated):\n        generated_loss = tf.keras.losses.BinaryCrossentropy(from_logits=True, reduction=tf.keras.losses.Reduction.NONE)(tf.ones_like(generated), generated)\n        real_loss = tf.keras.losses.BinaryCrossentropy(from_logits=True, reduction=tf.keras.losses.Reduction.NONE)(tf.zeros_like(real), real)\n        total_disc_loss = real_loss + generated_loss\n\n        return tf.reduce_mean(total_disc_loss * 0.5)","metadata":{"execution":{"iopub.execute_input":"2024-11-26T13:38:11.430899Z","iopub.status.busy":"2024-11-26T13:38:11.430547Z","iopub.status.idle":"2024-11-26T13:38:11.437283Z","shell.execute_reply":"2024-11-26T13:38:11.436145Z"},"papermill":{"duration":0.02795,"end_time":"2024-11-26T13:38:11.439161","exception":false,"start_time":"2024-11-26T13:38:11.411211","status":"completed"},"tags":[]},"outputs":[],"execution_count":null},{"id":"ffea447f","cell_type":"code","source":"with strategy.scope():\n    def generator_loss1(generated):\n        return  tf.reduce_mean(-generated)\n\n    def generator_loss2(generated):\n        return tf.reduce_mean(tf.keras.losses.BinaryCrossentropy(from_logits=True, reduction=tf.keras.losses.Reduction.NONE)(tf.zeros_like(generated), generated))","metadata":{"execution":{"iopub.execute_input":"2024-11-26T13:38:11.477256Z","iopub.status.busy":"2024-11-26T13:38:11.476904Z","iopub.status.idle":"2024-11-26T13:38:11.482237Z","shell.execute_reply":"2024-11-26T13:38:11.481183Z"},"papermill":{"duration":0.02715,"end_time":"2024-11-26T13:38:11.484382","exception":false,"start_time":"2024-11-26T13:38:11.457232","status":"completed"},"tags":[]},"outputs":[],"execution_count":null},{"id":"61b5ed49","cell_type":"code","source":"with strategy.scope():\n    def calc_cycle_loss(real_image, cycled_image, LAMBDA):\n        loss1 = tf.reduce_sum(tf.abs(real_image - cycled_image))\n\n        return LAMBDA * loss1 * 0.0000152587890625","metadata":{"execution":{"iopub.execute_input":"2024-11-26T13:38:11.522485Z","iopub.status.busy":"2024-11-26T13:38:11.522135Z","iopub.status.idle":"2024-11-26T13:38:11.527074Z","shell.execute_reply":"2024-11-26T13:38:11.526185Z"},"papermill":{"duration":0.026196,"end_time":"2024-11-26T13:38:11.529095","exception":false,"start_time":"2024-11-26T13:38:11.502899","status":"completed"},"tags":[]},"outputs":[],"execution_count":null},{"id":"ce3f96b6","cell_type":"code","source":"with strategy.scope():\n    def identity_loss(real_image, same_image, LAMBDA):\n        loss = tf.reduce_sum(tf.abs(real_image - same_image))\n        return LAMBDA * 0.5 * loss * 0.0000152587890625","metadata":{"execution":{"iopub.execute_input":"2024-11-26T13:38:11.566905Z","iopub.status.busy":"2024-11-26T13:38:11.566519Z","iopub.status.idle":"2024-11-26T13:38:11.571954Z","shell.execute_reply":"2024-11-26T13:38:11.570783Z"},"papermill":{"duration":0.026875,"end_time":"2024-11-26T13:38:11.574090","exception":false,"start_time":"2024-11-26T13:38:11.547215","status":"completed"},"tags":[]},"outputs":[],"execution_count":null},{"id":"2c907b87","cell_type":"code","source":"with strategy.scope():\n# Differentiable Augmentation for Data-Efficient GAN Training\n# Shengyu Zhao, Zhijian Liu, Ji Lin, Jun-Yan Zhu, and Song Han\n# https://arxiv.org/pdf/2006.10738\n# from https://github.com/mit-han-lab/data-efficient-gans/blob/master/DiffAugment_tf.py\n\n\n\n    def DiffAugment(x, policy='', channels_first=False):\n        if policy:\n            if channels_first:\n                x = tf.transpose(x, [0, 2, 3, 1])\n            for p in policy.split(','):\n                for f in AUGMENT_FNS[p]:\n                    x = f(x)\n            if channels_first:\n                x = tf.transpose(x, [0, 3, 1, 2])\n        return x\n\n\n    def rand_brightness(x):\n        magnitude = tf.random.uniform([tf.shape(x)[0], 1, 1, 1]) - 0.5\n        x = x + magnitude\n        return x\n\n\n    def rand_saturation(x):\n        magnitude = tf.random.uniform([tf.shape(x)[0], 1, 1, 1]) * 2\n        x_mean = tf.reduce_sum(x, axis=3, keepdims=True) * 0.3333333333333333333\n        x = (x - x_mean) * magnitude + x_mean\n        return x\n\n\n    def rand_contrast(x):\n        magnitude = tf.random.uniform([tf.shape(x)[0], 1, 1, 1]) + 0.5\n        x_mean = tf.reduce_sum(x, axis=[1, 2, 3], keepdims=True) * 5.086e-6\n        x = (x - x_mean) * magnitude + x_mean\n        return x\n\n    def rand_translation(x, ratio=0.125):\n        batch_size = tf.shape(x)[0]\n        image_size = tf.shape(x)[1:3]\n        shift = tf.cast(tf.cast(image_size, tf.float32) * ratio + 0.5, tf.int32)\n        translation_x = tf.random.uniform([batch_size, 1], -shift[0], shift[0] + 1, dtype=tf.int32)\n        translation_y = tf.random.uniform([batch_size, 1], -shift[1], shift[1] + 1, dtype=tf.int32)\n        grid_x = tf.clip_by_value(tf.expand_dims(tf.range(image_size[0], dtype=tf.int32), 0) + translation_x + 1, 0, image_size[0] + 1)\n        grid_y = tf.clip_by_value(tf.expand_dims(tf.range(image_size[1], dtype=tf.int32), 0) + translation_y + 1, 0, image_size[1] + 1)\n        x = tf.gather_nd(tf.pad(x, [[0, 0], [1, 1], [0, 0], [0, 0]]), tf.expand_dims(grid_x, -1), batch_dims=1)\n        x = tf.transpose(tf.gather_nd(tf.pad(tf.transpose(x, [0, 2, 1, 3]), [[0, 0], [1, 1], [0, 0], [0, 0]]), tf.expand_dims(grid_y, -1), batch_dims=1), [0, 2, 1, 3])\n        return x\n\n\n    def rand_cutout(x, ratio=0.5):\n        batch_size = tf.shape(x)[0]\n        image_size = tf.shape(x)[1:3]\n        cutout_size = tf.cast(tf.cast(image_size, tf.float32) * ratio + 0.5, tf.int32)\n        offset_x = tf.random.uniform([tf.shape(x)[0], 1, 1], maxval=image_size[0] + (1 - cutout_size[0] % 2), dtype=tf.int32)\n        offset_y = tf.random.uniform([tf.shape(x)[0], 1, 1], maxval=image_size[1] + (1 - cutout_size[1] % 2), dtype=tf.int32)\n        grid_batch, grid_x, grid_y = tf.meshgrid(tf.range(batch_size, dtype=tf.int32), tf.range(cutout_size[0], dtype=tf.int32), tf.range(cutout_size[1], dtype=tf.int32), indexing='ij')\n        cutout_grid = tf.stack([grid_batch, grid_x + offset_x - cutout_size[0] // 2, grid_y + offset_y - cutout_size[1] // 2], axis=-1)\n        mask_shape = tf.stack([batch_size, image_size[0], image_size[1]])\n        cutout_grid = tf.maximum(cutout_grid, 0)\n        cutout_grid = tf.minimum(cutout_grid, tf.reshape(mask_shape - 1, [1, 1, 1, 3]))\n        mask = tf.maximum(1 - tf.scatter_nd(cutout_grid, tf.ones([batch_size, cutout_size[0], cutout_size[1]], dtype=tf.float32), mask_shape), 0)\n        x = x * tf.expand_dims(mask, axis=3)\n        return x\n\n\n    AUGMENT_FNS = {\n        'color': [rand_brightness, rand_saturation, rand_contrast],\n        'translation': [rand_translation],\n        'cutout': [rand_cutout],\n}\n    def aug_fn(image):\n        return DiffAugment(image,\"color,translation,cutout\")","metadata":{"execution":{"iopub.execute_input":"2024-11-26T13:38:11.611950Z","iopub.status.busy":"2024-11-26T13:38:11.611571Z","iopub.status.idle":"2024-11-26T13:38:11.627766Z","shell.execute_reply":"2024-11-26T13:38:11.626876Z"},"papermill":{"duration":0.037451,"end_time":"2024-11-26T13:38:11.629536","exception":false,"start_time":"2024-11-26T13:38:11.592085","status":"completed"},"tags":[]},"outputs":[],"execution_count":null},{"id":"93125643","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)","metadata":{"execution":{"iopub.execute_input":"2024-11-26T13:38:11.668537Z","iopub.status.busy":"2024-11-26T13:38:11.667623Z","iopub.status.idle":"2024-11-26T13:38:11.681342Z","shell.execute_reply":"2024-11-26T13:38:11.680453Z"},"papermill":{"duration":0.035566,"end_time":"2024-11-26T13:38:11.683526","exception":false,"start_time":"2024-11-26T13:38:11.647960","status":"completed"},"tags":[]},"outputs":[],"execution_count":null},{"id":"34c0794c","cell_type":"code","source":"with strategy.scope():\n    cycle_gan_model = CycleGan(\n        monet_generator, photo_generator, monet_discriminator, photo_discriminator, dHead1,  dHead2\n    )","metadata":{"execution":{"iopub.execute_input":"2024-11-26T13:38:11.730359Z","iopub.status.busy":"2024-11-26T13:38:11.729632Z","iopub.status.idle":"2024-11-26T13:38:11.735681Z","shell.execute_reply":"2024-11-26T13:38:11.734642Z"},"papermill":{"duration":0.036703,"end_time":"2024-11-26T13:38:11.738107","exception":false,"start_time":"2024-11-26T13:38:11.701404","status":"completed"},"tags":[]},"outputs":[],"execution_count":null},{"id":"f2091c9a","cell_type":"code","source":"with strategy.scope():\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_fn1 = generator_loss1,\n        gen_loss_fn2 = generator_loss2,\n        disc_loss_fn1 = discriminator_loss1,\n        disc_loss_fn2 = discriminator_loss2,\n        cycle_loss_fn = calc_cycle_loss,\n        identity_loss_fn = identity_loss,\n        aug_fn = aug_fn ,\n\n\n    )","metadata":{"execution":{"iopub.execute_input":"2024-11-26T13:38:11.776256Z","iopub.status.busy":"2024-11-26T13:38:11.775891Z","iopub.status.idle":"2024-11-26T13:38:11.788037Z","shell.execute_reply":"2024-11-26T13:38:11.786999Z"},"papermill":{"duration":0.033495,"end_time":"2024-11-26T13:38:11.790287","exception":false,"start_time":"2024-11-26T13:38:11.756792","status":"completed"},"tags":[]},"outputs":[],"execution_count":null},{"id":"8b63e207","cell_type":"code","source":"# cycle_gan_model.fit(final_dataset,steps_per_epoch=1407, epochs=15)\n# FID(fid_photo_ds,monet_generator) ","metadata":{"execution":{"iopub.execute_input":"2024-11-26T13:38:11.829394Z","iopub.status.busy":"2024-11-26T13:38:11.828456Z","iopub.status.idle":"2024-11-26T13:38:11.833201Z","shell.execute_reply":"2024-11-26T13:38:11.832210Z"},"papermill":{"duration":0.025955,"end_time":"2024-11-26T13:38:11.835166","exception":false,"start_time":"2024-11-26T13:38:11.809211","status":"completed"},"tags":[]},"outputs":[],"execution_count":null},{"id":"8d45e4ad","cell_type":"code","source":"with strategy.scope():\n    monet_generator_optimizer = tf.keras.optimizers.Adam(1e-4, beta_1=0.5)\n    photo_generator_optimizer = tf.keras.optimizers.Adam(1e-4, beta_1=0.5)\n\n    monet_discriminator_optimizer = tf.keras.optimizers.Adam(1e-4, beta_1=0.5)\n    photo_discriminator_optimizer = tf.keras.optimizers.Adam(1e-4, beta_1=0.5)","metadata":{"execution":{"iopub.execute_input":"2024-11-26T13:38:11.873522Z","iopub.status.busy":"2024-11-26T13:38:11.873195Z","iopub.status.idle":"2024-11-26T13:38:11.884909Z","shell.execute_reply":"2024-11-26T13:38:11.883991Z"},"papermill":{"duration":0.033653,"end_time":"2024-11-26T13:38:11.886868","exception":false,"start_time":"2024-11-26T13:38:11.853215","status":"completed"},"tags":[]},"outputs":[],"execution_count":null},{"id":"a0bf2dbe","cell_type":"code","source":"with strategy.scope():\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_fn1 = generator_loss1,\n        gen_loss_fn2 = generator_loss2,\n        disc_loss_fn1 = discriminator_loss1,\n        disc_loss_fn2 = discriminator_loss2,\n        cycle_loss_fn = calc_cycle_loss,\n        identity_loss_fn = identity_loss,\n        aug_fn = aug_fn ,\n    )","metadata":{"execution":{"iopub.execute_input":"2024-11-26T13:38:11.942852Z","iopub.status.busy":"2024-11-26T13:38:11.942176Z","iopub.status.idle":"2024-11-26T13:38:11.955785Z","shell.execute_reply":"2024-11-26T13:38:11.954896Z"},"papermill":{"duration":0.047297,"end_time":"2024-11-26T13:38:11.958642","exception":false,"start_time":"2024-11-26T13:38:11.911345","status":"completed"},"tags":[]},"outputs":[],"execution_count":null},{"id":"e5c5da50","cell_type":"code","source":"# cycle_gan_model.fit(final_dataset,steps_per_epoch=1407, epochs=18)\n# FID(fid_photo_ds,monet_generator)","metadata":{"execution":{"iopub.execute_input":"2024-11-26T13:38:12.008622Z","iopub.status.busy":"2024-11-26T13:38:12.008289Z","iopub.status.idle":"2024-11-26T13:38:12.012400Z","shell.execute_reply":"2024-11-26T13:38:12.011399Z"},"papermill":{"duration":0.026877,"end_time":"2024-11-26T13:38:12.014411","exception":false,"start_time":"2024-11-26T13:38:11.987534","status":"completed"},"tags":[]},"outputs":[],"execution_count":null},{"id":"d3168c69","cell_type":"code","source":"with strategy.scope():\n    monet_generator_optimizer = tf.keras.optimizers.Adam(1e-5, beta_1=0.5)\n    photo_generator_optimizer = tf.keras.optimizers.Adam(1e-5, beta_1=0.5)\n\n    monet_discriminator_optimizer = tf.keras.optimizers.Adam(1e-5, beta_1=0.5)\n    photo_discriminator_optimizer = tf.keras.optimizers.Adam(1e-5, beta_1=0.5)","metadata":{"execution":{"iopub.execute_input":"2024-11-26T13:38:12.054024Z","iopub.status.busy":"2024-11-26T13:38:12.053026Z","iopub.status.idle":"2024-11-26T13:38:12.064802Z","shell.execute_reply":"2024-11-26T13:38:12.063882Z"},"papermill":{"duration":0.033829,"end_time":"2024-11-26T13:38:12.067059","exception":false,"start_time":"2024-11-26T13:38:12.033230","status":"completed"},"tags":[]},"outputs":[],"execution_count":null},{"id":"bae38f74","cell_type":"code","source":"with strategy.scope():\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_fn1 = generator_loss1,\n        gen_loss_fn2 = generator_loss2,\n        disc_loss_fn1 = discriminator_loss1,\n        disc_loss_fn2 = discriminator_loss2,\n        cycle_loss_fn = calc_cycle_loss,\n        identity_loss_fn = identity_loss,\n        aug_fn = aug_fn ,\n    )","metadata":{"execution":{"iopub.execute_input":"2024-11-26T13:38:12.105377Z","iopub.status.busy":"2024-11-26T13:38:12.104561Z","iopub.status.idle":"2024-11-26T13:38:12.113474Z","shell.execute_reply":"2024-11-26T13:38:12.112489Z"},"papermill":{"duration":0.030413,"end_time":"2024-11-26T13:38:12.115591","exception":false,"start_time":"2024-11-26T13:38:12.085178","status":"completed"},"tags":[]},"outputs":[],"execution_count":null},{"id":"90f47e39","cell_type":"code","source":"# cycle_gan_model.fit(final_dataset,steps_per_epoch=1407, epochs=12)\n# FID(fid_photo_ds,monet_generator) ","metadata":{"execution":{"iopub.execute_input":"2024-11-26T13:38:12.153165Z","iopub.status.busy":"2024-11-26T13:38:12.152781Z","iopub.status.idle":"2024-11-26T13:38:12.157331Z","shell.execute_reply":"2024-11-26T13:38:12.156189Z"},"papermill":{"duration":0.025926,"end_time":"2024-11-26T13:38:12.159565","exception":false,"start_time":"2024-11-26T13:38:12.133639","status":"completed"},"tags":[]},"outputs":[],"execution_count":null},{"id":"d22e2c80","cell_type":"code","source":"ds_iter = iter(photo_ds)\nfor n_sample in range(8):\n        example_sample = next(ds_iter)\n        generated_sample = monet_generator(example_sample)\n        \n        f = plt.figure(figsize=(32, 32))\n        \n        plt.subplot(121)\n        plt.title('Input image')\n        plt.imshow(example_sample[0] * 0.5 + 0.5)\n        plt.axis('off')\n        \n        plt.subplot(122)\n        plt.title('Generated image')\n        plt.imshow(generated_sample[0] * 0.5 + 0.5)\n        plt.axis('off')\n        plt.show()","metadata":{"execution":{"iopub.execute_input":"2024-11-26T13:38:12.196833Z","iopub.status.busy":"2024-11-26T13:38:12.196441Z","iopub.status.idle":"2024-11-26T13:38:21.889209Z","shell.execute_reply":"2024-11-26T13:38:21.887548Z"},"papermill":{"duration":9.716964,"end_time":"2024-11-26T13:38:21.894467","exception":false,"start_time":"2024-11-26T13:38:12.177503","status":"completed"},"tags":[]},"outputs":[],"execution_count":null},{"id":"c9e5a889","cell_type":"code","source":"ds_iter = iter(monet_ds)\nfor n_sample in range(8):\n\n        example_sample = next(ds_iter)\n        generated_sample = photo_generator(example_sample)\n        \n        f = plt.figure(figsize=(24, 24))\n        \n        plt.subplot(121)\n        plt.title('Input image')\n        plt.imshow(example_sample[0] * 0.5 + 0.5)\n        plt.axis('off')\n        \n        plt.subplot(122)\n        plt.title('Generated image')\n        plt.imshow(generated_sample[0] * 0.5 + 0.5)\n        plt.axis('off')\n        plt.show()","metadata":{"execution":{"iopub.execute_input":"2024-11-26T13:38:22.063704Z","iopub.status.busy":"2024-11-26T13:38:22.063063Z","iopub.status.idle":"2024-11-26T13:38:30.205893Z","shell.execute_reply":"2024-11-26T13:38:30.204896Z"},"papermill":{"duration":8.221624,"end_time":"2024-11-26T13:38:30.210968","exception":false,"start_time":"2024-11-26T13:38:21.989344","status":"completed"},"tags":[]},"outputs":[],"execution_count":null},{"id":"31982151","cell_type":"code","source":"import PIL\n! mkdir ../images","metadata":{"execution":{"iopub.execute_input":"2024-11-26T13:38:30.455646Z","iopub.status.busy":"2024-11-26T13:38:30.455296Z","iopub.status.idle":"2024-11-26T13:38:31.657637Z","shell.execute_reply":"2024-11-26T13:38:31.656234Z"},"papermill":{"duration":1.326774,"end_time":"2024-11-26T13:38:31.660165","exception":false,"start_time":"2024-11-26T13:38:30.333391","status":"completed"},"tags":[]},"outputs":[],"execution_count":null},{"id":"196fb7cb","cell_type":"code","source":"%%time\ni = 1\nfor img in fast_photo_ds:\n    prediction = monet_generator(img, training=False).numpy()\n    prediction = (prediction * 127.5 + 127.5).astype(np.uint8)\n    for pred in prediction:\n        im = PIL.Image.fromarray(pred)\n        im.save(\"../images/\" + str(i) + \".jpg\")\n        i += 1\n    ","metadata":{"execution":{"iopub.execute_input":"2024-11-26T13:38:31.897862Z","iopub.status.busy":"2024-11-26T13:38:31.896460Z","iopub.status.idle":"2024-11-26T14:08:15.694415Z","shell.execute_reply":"2024-11-26T14:08:15.693049Z"},"papermill":{"duration":1784.033829,"end_time":"2024-11-26T14:08:15.813228","exception":false,"start_time":"2024-11-26T13:38:31.779399","status":"completed"},"tags":[]},"outputs":[],"execution_count":null},{"id":"e330dfc0","cell_type":"code","source":"import shutil\nshutil.make_archive(\"/kaggle/working/images\", 'zip', \"/kaggle/images\")","metadata":{"execution":{"iopub.execute_input":"2024-11-26T14:08:16.039784Z","iopub.status.busy":"2024-11-26T14:08:16.039230Z","iopub.status.idle":"2024-11-26T14:08:17.311230Z","shell.execute_reply":"2024-11-26T14:08:17.310126Z"},"papermill":{"duration":1.387899,"end_time":"2024-11-26T14:08:17.313529","exception":false,"start_time":"2024-11-26T14:08:15.925630","status":"completed"},"tags":[]},"outputs":[],"execution_count":null},{"id":"798e8f67","cell_type":"code","source":"","metadata":{"papermill":{"duration":0.117594,"end_time":"2024-11-26T14:08:17.545272","exception":false,"start_time":"2024-11-26T14:08:17.427678","status":"completed"},"tags":[]},"outputs":[],"execution_count":null},{"id":"e5a0189f","cell_type":"code","source":"","metadata":{"papermill":{"duration":0.121031,"end_time":"2024-11-26T14:08:17.784002","exception":false,"start_time":"2024-11-26T14:08:17.662971","status":"completed"},"tags":[]},"outputs":[],"execution_count":null},{"id":"5d55bbfe","cell_type":"code","source":"","metadata":{"papermill":{"duration":0.114475,"end_time":"2024-11-26T14:08:18.015778","exception":false,"start_time":"2024-11-26T14:08:17.901303","status":"completed"},"tags":[]},"outputs":[],"execution_count":null}]}