{"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":"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":{"_uuid":"8f2839f25d086af736a60e9eeb907d3b93b6e0e5","_cell_guid":"b1076dfc-b9ad-4769-8c92-a6c4dae69d19","execution":{"iopub.status.busy":"2022-03-30T11:58:24.297843Z","iopub.execute_input":"2022-03-30T11:58:24.298105Z","iopub.status.idle":"2022-03-30T11:58:24.307187Z","shell.execute_reply.started":"2022-03-30T11:58:24.298075Z","shell.execute_reply":"2022-03-30T11:58:24.306225Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"GCS_PATH = KaggleDatasets().get_gcs_path()","metadata":{"execution":{"iopub.status.busy":"2022-03-30T11:58:24.333261Z","iopub.execute_input":"2022-03-30T11:58:24.333581Z","iopub.status.idle":"2022-03-30T11:58:25.050474Z","shell.execute_reply.started":"2022-03-30T11:58:24.33355Z","shell.execute_reply":"2022-03-30T11:58:25.049708Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"MONET_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":"2022-03-30T11:58:25.053617Z","iopub.execute_input":"2022-03-30T11:58:25.053857Z","iopub.status.idle":"2022-03-30T11:58:25.804562Z","shell.execute_reply.started":"2022-03-30T11:58:25.053829Z","shell.execute_reply":"2022-03-30T11:58:25.803722Z"},"trusted":true},"execution_count":null,"outputs":[]},{"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\n\ndef 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":"2022-03-30T11:58:25.805772Z","iopub.execute_input":"2022-03-30T11:58:25.806214Z","iopub.status.idle":"2022-03-30T11:58:25.816444Z","shell.execute_reply.started":"2022-03-30T11:58:25.806174Z","shell.execute_reply":"2022-03-30T11:58:25.815546Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"monet_ds = load_dataset(MONET_FILENAMES, labeled=True).batch(1)\nphoto_ds = load_dataset(PHOTO_FILENAMES, labeled=True).batch(1)\nfast_photo_ds = load_dataset(PHOTO_FILENAMES).batch(32*strategy.num_replicas_in_sync).prefetch(32)\nexample_monet = next(iter(monet_ds))\nexample_photo = next(iter(photo_ds))","metadata":{"execution":{"iopub.status.busy":"2022-03-30T13:27:18.525537Z","iopub.execute_input":"2022-03-30T13:27:18.525894Z","iopub.status.idle":"2022-03-30T13:27:20.464762Z","shell.execute_reply.started":"2022-03-30T13:27:18.525857Z","shell.execute_reply":"2022-03-30T13:27:20.463983Z"},"trusted":true},"execution_count":null,"outputs":[]},{"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":"2022-03-30T11:58:27.136109Z","iopub.execute_input":"2022-03-30T11:58:27.13661Z","iopub.status.idle":"2022-03-30T11:58:27.45628Z","shell.execute_reply.started":"2022-03-30T11:58:27.136582Z","shell.execute_reply":"2022-03-30T11:58:27.455602Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":"### CycleGAN 에서 UNET 아키텍처 가 뭐지?\n\n\n\n\n\n\n\n\n참조:\n* https://chacha95.github.io/2019-08-30-Image-Translation/\n","metadata":{}},{"cell_type":"code","source":"OUTPUT_CHANNELS = 3\n\ndef downsample(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    result = keras.Sequential()\n    result.add(layers.Conv2D(filters, size, strides=2, padding='same',\n                            kernel_initializer=initializer, use_bias=False\n                            ))\n    if apply_instancenorm:\n        result.add(tfa.layers.InstanceNormalization(gamma_initializer=gamma_init))\n\n    result.add(layers.LeakyReLU())\n\n    return result\n        ","metadata":{"execution":{"iopub.status.busy":"2022-03-30T11:58:27.457527Z","iopub.execute_input":"2022-03-30T11:58:27.458377Z","iopub.status.idle":"2022-03-30T11:58:27.465819Z","shell.execute_reply.started":"2022-03-30T11:58:27.458342Z","shell.execute_reply":"2022-03-30T11:58:27.465201Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":"# tf.random_normal_initializer 설명\n* 이니셜라이저를 사용하면 초기화되는 변수의 모양과 dtype을 몰라도 Initializer 객체에 인코딩된 초기화 전략을 미리 지정\n","metadata":{}},{"cell_type":"markdown","source":"# 다운샘플 업샘플이 무슨 정확히 무슨 의미이고 하는 이유가 무엇일까?\n다운샘플 이미지의 폭과 높이를 stride 단위로 줄여준다고 한다.\n사진을 일정하게 맞춰주기 위함으로 보인다.\n업샘플은 반대의 의미","metadata":{}},{"cell_type":"markdown","source":"# Build Generator\n> * Generator는 먼저 입력 이미지를 다운샘플링한 다음 롱 스킵 연결을 설정하는 동안 업샘플링을 수행합니다. \n> * 연결 건너뛰기는 한 계층의 출력을 한 계층이 아닌 여러 계층으로 연결하여 사라지는 그라데이션 문제를 우회하는 방법입니다. \n> * 여기서는 다운샘플 레이어의 출력을 업샘플 레이어에 대칭적으로 연결","metadata":{}},{"cell_type":"code","source":"def upsample(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    result = keras.Sequential()\n    result.add(layers.Conv2DTranspose(filters, size, strides=2,\n                                      padding='same',\n                                      kernel_initializer=initializer,\n                                      use_bias=False))\n\n    result.add(tfa.layers.InstanceNormalization(gamma_initializer=gamma_init))\n\n    if apply_dropout:\n        result.add(layers.Dropout(0.5))\n\n    result.add(layers.ReLU())\n\n    return result","metadata":{"execution":{"iopub.status.busy":"2022-03-30T11:58:27.467065Z","iopub.execute_input":"2022-03-30T11:58:27.467424Z","iopub.status.idle":"2022-03-30T11:58:27.476763Z","shell.execute_reply.started":"2022-03-30T11:58:27.467375Z","shell.execute_reply":"2022-03-30T11:58:27.4759Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"upsample(512, 4, apply_dropout=True)","metadata":{"execution":{"iopub.status.busy":"2022-03-30T11:58:27.478608Z","iopub.execute_input":"2022-03-30T11:58:27.478904Z","iopub.status.idle":"2022-03-30T11:58:27.494849Z","shell.execute_reply.started":"2022-03-30T11:58:27.478871Z","shell.execute_reply":"2022-03-30T11:58:27.494191Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"def Generator():\n    inputs = layers.Input(shape=[256,256,3])\n\n    # bs = batch size\n    down_stack = [\n        downsample(64, 4, apply_instancenorm=False), # (bs, 128, 128, 64)\n        downsample(128, 4), # (bs, 64, 64, 128)\n        downsample(256, 4), # (bs, 32, 32, 256)\n        downsample(512, 4), # (bs, 16, 16, 512)\n        downsample(512, 4), # (bs, 8, 8, 512)\n        downsample(512, 4), # (bs, 4, 4, 512)\n        downsample(512, 4), # (bs, 2, 2, 512)\n        downsample(512, 4), # (bs, 1, 1, 512)\n    ]\n\n    up_stack = [\n        upsample(512, 4, apply_dropout=True), # (bs, 2, 2, 1024)\n        upsample(512, 4, apply_dropout=True), # (bs, 4, 4, 1024)\n        upsample(512, 4, apply_dropout=True), # (bs, 8, 8, 1024)\n        upsample(512, 4), # (bs, 16, 16, 1024)\n        upsample(256, 4), # (bs, 32, 32, 512)\n        upsample(128, 4), # (bs, 64, 64, 256)\n        upsample(64, 4), # (bs, 128, 128, 128)\n    ]\n\n    initializer = tf.random_normal_initializer(0., 0.02)\n    last = layers.Conv2DTranspose(OUTPUT_CHANNELS, 4,\n                                  strides=2,\n                                  padding='same',\n                                  kernel_initializer=initializer,\n                                  activation='tanh') # (bs, 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.status.busy":"2022-03-30T11:58:27.495943Z","iopub.execute_input":"2022-03-30T11:58:27.496937Z","iopub.status.idle":"2022-03-30T11:58:27.508276Z","shell.execute_reply.started":"2022-03-30T11:58:27.496901Z","shell.execute_reply":"2022-03-30T11:58:27.50761Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":"# Build discriminator\n> * discriminator는 입력 이미지를 가져와 실제 또는 가짜(생성)로 분류합니다. \n> * discriminator는 단일 노드를 출력하는 대신 실제 분류를 나타내는 더 높은 픽셀 값과 가짜 분류를 나타내는 더 낮은 값으로 더 작은 2D 이미지를 출력","metadata":{}},{"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    \n    x = inp\n    \n    down1 = downsample(64, 4, False)(x) # (bs, 128, 128, 64)\n    down2 = downsample(128, 4)(down1) # (bs, 64, 64, 128)\n    down3 = downsample(256, 4)(down2) # (bs, 32, 32, 256)\n    \n    zero_pad1 = layers.ZeroPadding2D()(down3) # (bs, 34, 34, 256)\n    conv = layers.Conv2D(512, 4, strides=1,\n                        kernel_initializer=initializer,\n                        use_bias=False)(zero_pad1) # (bs, 31, 31, 512)\n    norm1 = tfa.layers.InstanceNormalization(gamma_initializer=gamma_init)(conv)\n    \n    leaky_relu = layers.LeakyReLU()(norm1)\n    \n    zero_pad2 = layers.ZeroPadding2D()(leaky_relu) # (bs, 33, 33, 512)\n\n    \n    last = layers.Conv2D(1, 4, strides=1,\n                        kernel_initializer=initializer)(zero_pad2) # (bs, 30, 30, 1)\n    \n    return tf.keras.Model(inputs=inp, outputs=last)\n    ","metadata":{"execution":{"iopub.status.busy":"2022-03-30T11:58:27.511546Z","iopub.execute_input":"2022-03-30T11:58:27.511763Z","iopub.status.idle":"2022-03-30T11:58:27.522403Z","shell.execute_reply.started":"2022-03-30T11:58:27.511716Z","shell.execute_reply":"2022-03-30T11:58:27.521627Z"},"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":"2022-03-30T11:58:27.523454Z","iopub.execute_input":"2022-03-30T11:58:27.523811Z","iopub.status.idle":"2022-03-30T11:58:29.482715Z","shell.execute_reply.started":"2022-03-30T11:58:27.523753Z","shell.execute_reply":"2022-03-30T11:58:29.482022Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":"# strategy.scope() ??\n* https://www.tensorflow.org/api_docs/python/tf/distribute/Strategy\n* https://www.tensorflow.org/guide/distributed_training?hl=ko\n\n> * 분산전략? 분산과관련된 처리를 한다?\n> * 모든 변수와 모델, 그래프는 strategy(지정한 변수)에 복제된다.\n> * strategy는 주어지는 입력에 대해서 손실(loss)와 gradient(가중치)를 계산한다.\n> * gradient(가중치)를 전부 더함으로써 모든 장치(GPU,TPU)간에 gradient(가중치)가 동기화 된다.\n> * 동기화된 후에, 동일한 업데이트가 각 장치에 있는 변수의 복사본(copies)에 동일하게 적용","metadata":{}},{"cell_type":"markdown","source":"# 결론\n> * 훈련을 여러 TPU,GPU로 나눠 처리하고 각자 장치간의 훈련손실, 가중치를 기록하고 계산하기 위함\n> * 나눠 처리된 훈련을 테스트 손실과 훈련 정확도,체크포인트,등 이해하고 처리 할 수 있다.\n> * MirroedStrategy를 사용하면 미러링된 변수들을 하나의 가상변수에 대응되어서, 모두 같은 값을 유지 된다고 한다.","metadata":{}},{"cell_type":"code","source":"# print('장치의 수 ':{}.format(strategy.num_replicas_in_sync))","metadata":{"execution":{"iopub.status.busy":"2022-03-30T11:58:29.48383Z","iopub.execute_input":"2022-03-30T11:58:29.485904Z","iopub.status.idle":"2022-03-30T11:58:29.48946Z","shell.execute_reply.started":"2022-03-30T11:58:29.485874Z","shell.execute_reply":"2022-03-30T11:58:29.488824Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":"> Generators가 교육을 받지 않았기 때문에 생성된 모네사진은 이 시점에서 예상되는 것을 보여주지 않는다.","metadata":{}},{"cell_type":"code","source":"# \nto_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":"2022-03-30T11:58:29.490526Z","iopub.execute_input":"2022-03-30T11:58:29.491287Z","iopub.status.idle":"2022-03-30T11:58:29.811282Z","shell.execute_reply.started":"2022-03-30T11:58:29.491243Z","shell.execute_reply":"2022-03-30T11:58:29.810591Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":"# Build the CycleGAN model\n> * tf.keras를 분류하겠습니다.나중에 fit()을 실행하여 모델을 교육할 수 있도록 모델을 만드십시오. \n> * 교육 단계에서 모델은 사진을 모네 그림으로 변환한 다음 다시 사진으로 변환합니다. \n> * 원본 사진과 두 번 변환된 사진의 차이는 주기 일관성 손실입니다. 원본 사진과 두 번 변형된 사진이 서로 비슷했으면 합니다.\n\n손실은 다음 섹션에서 정의합니다.","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":"2022-03-30T11:58:29.812813Z","iopub.execute_input":"2022-03-30T11:58:29.8135Z","iopub.status.idle":"2022-03-30T11:58:29.832053Z","shell.execute_reply.started":"2022-03-30T11:58:29.813462Z","shell.execute_reply":"2022-03-30T11:58:29.83131Z"},"trusted":true},"execution_count":null,"outputs":[]},{"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\n        generated_loss = tf.keras.losses.BinaryCrossentropy(from_logits=True, reduction=tf.keras.losses.Reduction.NONE)(tf.zeros_like(generated), generated)\n\n        total_disc_loss = real_loss + generated_loss\n\n        return total_disc_loss * 0.5","metadata":{"execution":{"iopub.status.busy":"2022-03-30T11:58:29.833564Z","iopub.execute_input":"2022-03-30T11:58:29.833838Z","iopub.status.idle":"2022-03-30T11:58:29.845957Z","shell.execute_reply.started":"2022-03-30T11:58:29.833803Z","shell.execute_reply":"2022-03-30T11:58:29.845168Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"with strategy.scope():\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)","metadata":{"execution":{"iopub.status.busy":"2022-03-30T11:58:29.8472Z","iopub.execute_input":"2022-03-30T11:58:29.847832Z","iopub.status.idle":"2022-03-30T11:58:29.854706Z","shell.execute_reply.started":"2022-03-30T11:58:29.847793Z","shell.execute_reply":"2022-03-30T11:58:29.853997Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"with strategy.scope():\n    def calc_cycle_loss(real_image, cycled_image, LAMBDA):\n        loss1 = tf.reduce_mean(tf.abs(real_image - cycled_image))\n\n        return LAMBDA * loss1","metadata":{"execution":{"iopub.status.busy":"2022-03-30T11:58:29.856039Z","iopub.execute_input":"2022-03-30T11:58:29.856575Z","iopub.status.idle":"2022-03-30T11:58:29.86389Z","shell.execute_reply.started":"2022-03-30T11:58:29.856538Z","shell.execute_reply":"2022-03-30T11:58:29.863094Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"with strategy.scope():\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":"2022-03-30T11:58:29.864766Z","iopub.execute_input":"2022-03-30T11:58:29.865254Z","iopub.status.idle":"2022-03-30T11:58:29.872414Z","shell.execute_reply.started":"2022-03-30T11:58:29.865218Z","shell.execute_reply":"2022-03-30T11:58:29.871766Z"},"trusted":true},"execution_count":null,"outputs":[]},{"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.status.busy":"2022-03-30T11:58:29.873538Z","iopub.execute_input":"2022-03-30T11:58:29.874243Z","iopub.status.idle":"2022-03-30T11:58:29.88176Z","shell.execute_reply.started":"2022-03-30T11:58:29.874157Z","shell.execute_reply":"2022-03-30T11:58:29.881114Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"with strategy.scope():\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":"2022-03-30T11:58:29.883923Z","iopub.execute_input":"2022-03-30T11:58:29.885049Z","iopub.status.idle":"2022-03-30T11:58:29.900958Z","shell.execute_reply.started":"2022-03-30T11:58:29.885019Z","shell.execute_reply":"2022-03-30T11:58:29.900201Z"},"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=25\n)","metadata":{"execution":{"iopub.status.busy":"2022-03-30T11:58:29.903172Z","iopub.execute_input":"2022-03-30T11:58:29.904387Z","iopub.status.idle":"2022-03-30T12:31:44.547261Z","shell.execute_reply.started":"2022-03-30T11:58:29.904358Z","shell.execute_reply":"2022-03-30T12:31:44.546557Z"},"trusted":true},"execution_count":null,"outputs":[]},{"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.show()","metadata":{"execution":{"iopub.status.busy":"2022-03-30T12:31:44.551185Z","iopub.execute_input":"2022-03-30T12:31:44.55316Z","iopub.status.idle":"2022-03-30T12:31:45.865253Z","shell.execute_reply.started":"2022-03-30T12:31:44.553122Z","shell.execute_reply":"2022-03-30T12:31:45.864637Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"import PIL\n! mkdir ../images","metadata":{"execution":{"iopub.status.busy":"2022-03-30T13:26:22.441689Z","iopub.execute_input":"2022-03-30T13:26:22.442017Z","iopub.status.idle":"2022-03-30T13:26:23.242487Z","shell.execute_reply.started":"2022-03-30T13:26:22.441973Z","shell.execute_reply":"2022-03-30T13:26:23.241529Z"},"trusted":true},"execution_count":null,"outputs":[]},{"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","metadata":{"execution":{"iopub.status.busy":"2022-03-30T13:27:25.290828Z","iopub.execute_input":"2022-03-30T13:27:25.291519Z","iopub.status.idle":"2022-03-30T13:28:26.780518Z","shell.execute_reply.started":"2022-03-30T13:27:25.291484Z","shell.execute_reply":"2022-03-30T13:28:26.779745Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"import shutil\nshutil.make_archive(\"/kaggle/working/images\", 'zip', \"/kaggle/images\")","metadata":{"execution":{"iopub.status.busy":"2022-03-30T13:28:35.101959Z","iopub.execute_input":"2022-03-30T13:28:35.102225Z","iopub.status.idle":"2022-03-30T13:28:38.297214Z","shell.execute_reply.started":"2022-03-30T13:28:35.102197Z","shell.execute_reply":"2022-03-30T13:28:38.296372Z"},"trusted":true},"execution_count":null,"outputs":[]},{"cell_type":"code","source":"","metadata":{},"execution_count":null,"outputs":[]},{"cell_type":"markdown","source":"## ","metadata":{}},{"cell_type":"code","source":"","metadata":{},"execution_count":null,"outputs":[]}]}