deep_learning/generative/TF2Keras_VanillaGAN_DCGAN.ipynb

{"nbformat":4,"nbformat_minor":0,"metadata":{"colab":{"name":"TF2Keras_VanillaGAN_DCGAN.ipynb","provenance":[],"collapsed_sections":[],"toc_visible":true,"authorship_tag":"ABX9TyPxuLiNciqfHiluzabTBzrW"},"kernelspec":{"name":"python3","display_name":"Python 3"},"accelerator":"GPU"},"cells":[{"cell_type":"markdown","metadata":{"id":"6QMb0FZoYYQu","colab_type":"text"},"source":["This tutorial shows how to generate an image of handwritten digits using Deep Convolutional Generative Adversarial Network (DCGAN).\n","\n","Generative Adversarial Networks (GANs) are one of the most interesting fields in machine learning. The standard GAN consists of two models, a generative and a discriminator one. Two models are trained simultaneously by an adversarial process. A generative model (`the artist`) learns to generate images that look real, while the discriminator (`the art critic`) one learns to tell real images apart from the fakes.\n","\n","![](https://www.tensorflow.org/tutorials/generative/images/gan1.png)\n","\n","Refer to Tensorflow.org (2020).\n","\n","During training, the generative model becomes progressively creating images that look real, and the discriminator model becomes progressively telling them apart. The whole process reaches equilibrium when the discriminator is no longer able to distinguish real images from fakes.\n","\n","![](https://www.tensorflow.org/tutorials/generative/images/gan2.png)\n","\n","Refer to Tensorflow.org (2020).\n","\n","In this demo, we show how to train a GAN model on MNIST and FASHION MNIST dataset.\n"]},{"cell_type":"code","metadata":{"id":"-HcC6Eg1GtMe","colab_type":"code","colab":{}},"source":["!pip uninstall -y tensorflow\n","!pip install -q tf-nightly tfds-nightly"],"execution_count":null,"outputs":[]},{"cell_type":"code","metadata":{"id":"S04ibrTKHbMO","colab_type":"code","colab":{"base_uri":"https://localhost:8080/","height":51},"executionInfo":{"status":"ok","timestamp":1594087709034,"user_tz":-480,"elapsed":3860,"user":{"displayName":"王DevOps","photoUrl":"https://lh3.googleusercontent.com/a-/AOh14GgXVAgB9A4OE1KXeAfp5b-xUS2OSsbqSRVEe_UETw=s64","userId":"04300517850278510646"}},"outputId":"7b141115-ecee-4f83-934a-cd4795068017"},"source":["import glob\n","import tensorflow as tf\n","import tensorflow_datasets as tfds\n","import matplotlib.pyplot as plt\n","from tensorflow.keras.layers import Conv2D, Conv2DTranspose, Dense, Flatten, BatchNormalization, ELU, LeakyReLU, Reshape, Dropout\n","import numpy as np\n","import IPython.display as display\n","from IPython.display import clear_output\n","import os\n","import time\n","import imageio\n","\n","tfds.disable_progress_bar()\n","\n","print(\"Tensorflow Version: {}\".format(tf.__version__))\n","print(\"GPU {} available.\".format(\"is\" if tf.config.experimental.list_physical_devices(\"GPU\") else \"not\"))"],"execution_count":2,"outputs":[{"output_type":"stream","text":["Tensorflow Version: 2.4.0-dev20200706\n","GPU is available.\n"],"name":"stdout"}]},{"cell_type":"markdown","metadata":{"id":"fZOIZTpDJzkM","colab_type":"text"},"source":["# Data Preprocessing"]},{"cell_type":"code","metadata":{"id":"CjNWA9d9KvYS","colab_type":"code","colab":{},"executionInfo":{"status":"ok","timestamp":1594087776201,"user_tz":-480,"elapsed":1722,"user":{"displayName":"王DevOps","photoUrl":"https://lh3.googleusercontent.com/a-/AOh14GgXVAgB9A4OE1KXeAfp5b-xUS2OSsbqSRVEe_UETw=s64","userId":"04300517850278510646"}}},"source":["def normalize(image):\n","  img = image['image']\n","  img = (tf.cast(img, tf.float32) - 127.5) / 127.5\n","  return img"],"execution_count":3,"outputs":[]},{"cell_type":"markdown","metadata":{"id":"Q7r4irSJeoam","colab_type":"text"},"source":["## MNIST Dataset"]},{"cell_type":"code","metadata":{"id":"0nxoek5Ceu1W","colab_type":"code","colab":{"base_uri":"https://localhost:8080/","height":156},"executionInfo":{"status":"ok","timestamp":1594087783374,"user_tz":-480,"elapsed":5602,"user":{"displayName":"王DevOps","photoUrl":"https://lh3.googleusercontent.com/a-/AOh14GgXVAgB9A4OE1KXeAfp5b-xUS2OSsbqSRVEe_UETw=s64","userId":"04300517850278510646"}},"outputId":"1c2fe527-698d-426c-ea3c-9dfe30ee7734"},"source":["raw_datasets, metadata = tfds.load(name=\"mnist\", with_info=True)\n","raw_train_datasets, raw_test_datasets = raw_datasets['train'], raw_datasets['test']"],"execution_count":4,"outputs":[{"output_type":"stream","text":["\u001b[1mDownloading and preparing dataset mnist/3.0.1 (download: 11.06 MiB, generated: 21.00 MiB, total: 32.06 MiB) to /root/tensorflow_datasets/mnist/3.0.1...\u001b[0m\n"],"name":"stdout"},{"output_type":"stream","text":["WARNING:absl:Dataset mnist is hosted on GCS. It will automatically be downloaded to your\n","local data directory. If you'd instead prefer to read directly from our public\n","GCS bucket (recommended if you're running on GCP), you can instead pass\n","`try_gcs=True` to `tfds.load` or set `data_dir=gs://tfds-data/datasets`.\n","\n"],"name":"stderr"},{"output_type":"stream","text":["\u001b[1mDataset mnist downloaded and prepared to /root/tensorflow_datasets/mnist/3.0.1. Subsequent calls will reuse this data.\u001b[0m\n"],"name":"stdout"}]},{"cell_type":"code","metadata":{"id":"HK22JzVsdeOA","colab_type":"code","colab":{"base_uri":"https://localhost:8080/","height":442},"executionInfo":{"status":"ok","timestamp":1594087785229,"user_tz":-480,"elapsed":1103,"user":{"displayName":"王DevOps","photoUrl":"https://lh3.googleusercontent.com/a-/AOh14GgXVAgB9A4OE1KXeAfp5b-xUS2OSsbqSRVEe_UETw=s64","userId":"04300517850278510646"}},"outputId":"a1bdffa4-0f3a-4189-b436-6e1e0645ba7b"},"source":["raw_test_datasets, metadata"],"execution_count":5,"outputs":[{"output_type":"execute_result","data":{"text/plain":["(<PrefetchDataset shapes: {image: (28, 28, 1), label: ()}, types: {image: tf.uint8, label: tf.int64}>,\n"," tfds.core.DatasetInfo(\n","     name='mnist',\n","     version=3.0.1,\n","     description='The MNIST database of handwritten digits.',\n","     homepage='http://yann.lecun.com/exdb/mnist/',\n","     features=FeaturesDict({\n","         'image': Image(shape=(28, 28, 1), dtype=tf.uint8),\n","         'label': ClassLabel(shape=(), dtype=tf.int64, num_classes=10),\n","     }),\n","     total_num_examples=70000,\n","     splits={\n","         'test': 10000,\n","         'train': 60000,\n","     },\n","     supervised_keys=('image', 'label'),\n","     citation=\"\"\"@article{lecun2010mnist,\n","       title={MNIST handwritten digit database},\n","       author={LeCun, Yann and Cortes, Corinna and Burges, CJ},\n","       journal={ATT Labs [Online]. Available: http://yann.lecun.com/exdb/mnist},\n","       volume={2},\n","       year={2010}\n","     }\"\"\",\n","     redistribution_info=,\n"," ))"]},"metadata":{"tags":[]},"execution_count":5}]},{"cell_type":"code","metadata":{"id":"ryS6VxK6daI0","colab_type":"code","colab":{},"executionInfo":{"status":"ok","timestamp":1594087828334,"user_tz":-480,"elapsed":797,"user":{"displayName":"王DevOps","photoUrl":"https://lh3.googleusercontent.com/a-/AOh14GgXVAgB9A4OE1KXeAfp5b-xUS2OSsbqSRVEe_UETw=s64","userId":"04300517850278510646"}}},"source":["BUFFER_SIZE = 10000\n","BATCH_SIZE = 256\n","\n","train_datasets = raw_train_datasets.map(normalize).cache().shuffle(BUFFER_SIZE).batch(BATCH_SIZE)\n","test_datasets = raw_test_datasets.map(normalize).batch(BATCH_SIZE)"],"execution_count":6,"outputs":[]},{"cell_type":"code","metadata":{"id":"krRI_KrLd-NZ","colab_type":"code","colab":{"base_uri":"https://localhost:8080/","height":248},"executionInfo":{"status":"ok","timestamp":1594087853043,"user_tz":-480,"elapsed":2019,"user":{"displayName":"王DevOps","photoUrl":"https://lh3.googleusercontent.com/a-/AOh14GgXVAgB9A4OE1KXeAfp5b-xUS2OSsbqSRVEe_UETw=s64","userId":"04300517850278510646"}},"outputId":"595e72dc-6411-4c08-c562-ff7371c1cd37"},"source":["for imgs in train_datasets.take(1):\n","  img = imgs[0]\n","  plt.imshow(tf.keras.preprocessing.image.array_to_img(img))\n","  plt.axis(\"off\")\n","  plt.show()"],"execution_count":7,"outputs":[{"output_type":"display_data","data":{"image/png":"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\n","text/plain":["<Figure size 432x288 with 1 Axes>"]},"metadata":{"tags":[],"needs_background":"light"}}]},{"cell_type":"markdown","metadata":{"id":"GD1bFuR7eglh","colab_type":"text"},"source":["## Fashion_MNIST Dataset"]},{"cell_type":"code","metadata":{"id":"chYWY7UwH2H3","colab_type":"code","colab":{"base_uri":"https://localhost:8080/","height":105},"executionInfo":{"status":"ok","timestamp":1594088257207,"user_tz":-480,"elapsed":23989,"user":{"displayName":"王DevOps","photoUrl":"https://lh3.googleusercontent.com/a-/AOh14GgXVAgB9A4OE1KXeAfp5b-xUS2OSsbqSRVEe_UETw=s64","userId":"04300517850278510646"}},"outputId":"a0164092-0bdc-48c2-cade-e65fa3b4dd95"},"source":["raw_datasets, metadata = tfds.load(name=\"fashion_mnist\", with_info=True)"],"execution_count":8,"outputs":[{"output_type":"stream","text":["\u001b[1mDownloading and preparing dataset fashion_mnist/3.0.1 (download: 29.45 MiB, generated: 36.42 MiB, total: 65.87 MiB) to /root/tensorflow_datasets/fashion_mnist/3.0.1...\u001b[0m\n","Shuffling and writing examples to /root/tensorflow_datasets/fashion_mnist/3.0.1.incompleteQB6NVK/fashion_mnist-train.tfrecord\n","Shuffling and writing examples to /root/tensorflow_datasets/fashion_mnist/3.0.1.incompleteQB6NVK/fashion_mnist-test.tfrecord\n","\u001b[1mDataset fashion_mnist downloaded and prepared to /root/tensorflow_datasets/fashion_mnist/3.0.1. Subsequent calls will reuse this data.\u001b[0m\n"],"name":"stdout"}]},{"cell_type":"code","metadata":{"id":"_bVstXyAKFIY","colab_type":"code","colab":{"base_uri":"https://localhost:8080/","height":34},"executionInfo":{"status":"ok","timestamp":1594088259727,"user_tz":-480,"elapsed":998,"user":{"displayName":"王DevOps","photoUrl":"https://lh3.googleusercontent.com/a-/AOh14GgXVAgB9A4OE1KXeAfp5b-xUS2OSsbqSRVEe_UETw=s64","userId":"04300517850278510646"}},"outputId":"a6d6703e-d486-408f-ebcc-e9d93bcbc4a6"},"source":["raw_train_datasets, raw_test_datasets = raw_datasets['train'], raw_datasets['test']\n","raw_train_datasets"],"execution_count":9,"outputs":[{"output_type":"execute_result","data":{"text/plain":["<PrefetchDataset shapes: {image: (28, 28, 1), label: ()}, types: {image: tf.uint8, label: tf.int64}>"]},"metadata":{"tags":[]},"execution_count":9}]},{"cell_type":"code","metadata":{"id":"7dZvIaDcKPUg","colab_type":"code","colab":{"base_uri":"https://localhost:8080/","height":264},"executionInfo":{"status":"ok","timestamp":1594088279117,"user_tz":-480,"elapsed":842,"user":{"displayName":"王DevOps","photoUrl":"https://lh3.googleusercontent.com/a-/AOh14GgXVAgB9A4OE1KXeAfp5b-xUS2OSsbqSRVEe_UETw=s64","userId":"04300517850278510646"}},"outputId":"1d94421e-3f5c-4b30-b319-ac291fc41ac8"},"source":["for image in raw_train_datasets.take(1):\n","  plt.imshow(tf.keras.preprocessing.image.array_to_img(image['image']))\n","  plt.axis(\"off\")\n","  plt.title(\"Label: {}\".format(image['label']))\n","  plt.show()"],"execution_count":10,"outputs":[{"output_type":"display_data","data":{"image/png":"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\n","text/plain":["<Figure size 432x288 with 1 Axes>"]},"metadata":{"tags":[],"needs_background":"light"}}]},{"cell_type":"code","metadata":{"id":"qW7itH0DVxOz","colab_type":"code","colab":{},"executionInfo":{"status":"ok","timestamp":1594088291237,"user_tz":-480,"elapsed":1137,"user":{"displayName":"王DevOps","photoUrl":"https://lh3.googleusercontent.com/a-/AOh14GgXVAgB9A4OE1KXeAfp5b-xUS2OSsbqSRVEe_UETw=s64","userId":"04300517850278510646"}}},"source":["BUFFER_SIZE = 10000\n","BATCH_SIZE = 256\n","\n","train_datasets = raw_train_datasets.map(normalize).cache().prefetch(BUFFER_SIZE).batch(BATCH_SIZE)\n","test_datasets = raw_test_datasets.map(normalize).batch(BATCH_SIZE)"],"execution_count":11,"outputs":[]},{"cell_type":"code","metadata":{"id":"goPh96zBn8WY","colab_type":"code","colab":{"base_uri":"https://localhost:8080/","height":248},"executionInfo":{"status":"ok","timestamp":1594088299135,"user_tz":-480,"elapsed":1064,"user":{"displayName":"王DevOps","photoUrl":"https://lh3.googleusercontent.com/a-/AOh14GgXVAgB9A4OE1KXeAfp5b-xUS2OSsbqSRVEe_UETw=s64","userId":"04300517850278510646"}},"outputId":"227f95f4-0e49-4f11-dcfe-8d2ef9bfa1e4"},"source":["for imgs in train_datasets.take(1):\n","  img = imgs[0]\n","  plt.imshow(tf.keras.preprocessing.image.array_to_img(img))\n","  plt.axis(\"off\")\n","  plt.show()"],"execution_count":12,"outputs":[{"output_type":"display_data","data":{"image/png":"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\n","text/plain":["<Figure size 432x288 with 1 Axes>"]},"metadata":{"tags":[],"needs_background":"light"}}]},{"cell_type":"markdown","metadata":{"id":"2GK6Lk9LWfWY","colab_type":"text"},"source":["# Build the GAN Model"]},{"cell_type":"markdown","metadata":{"id":"TEfcpaFNfWMw","colab_type":"text"},"source":["## The Generator\n","\n","The generator uses the `tf.keras.layers.Conv2DTranspose` (upsampling) layer to produce an image from a seed input (a random noise). Start from this seed input, upsample it several times to reach the desired output (28x28x1)."]},{"cell_type":"code","metadata":{"id":"3Dmy6GYkNvsi","colab_type":"code","colab":{},"executionInfo":{"status":"ok","timestamp":1594088483797,"user_tz":-480,"elapsed":1222,"user":{"displayName":"王DevOps","photoUrl":"https://lh3.googleusercontent.com/a-/AOh14GgXVAgB9A4OE1KXeAfp5b-xUS2OSsbqSRVEe_UETw=s64","userId":"04300517850278510646"}}},"source":["def build_generator_model():\n","  model = tf.keras.Sequential()\n","  \n","  model.add(Dense(units=7 * 7 * 256, use_bias=False, input_shape=(100,)))\n","  model.add(BatchNormalization())\n","  model.add(LeakyReLU())\n","\n","  model.add(Reshape(target_shape=[7,7,256]))\n","  assert model.output_shape == (None, 7, 7, 256)\n","\n","  model.add(Conv2DTranspose(filters=128, kernel_size=(5,5), strides=(1,1), padding=\"same\", use_bias=False))\n","  model.add(BatchNormalization())\n","  model.add(LeakyReLU())\n","  assert model.output_shape == (None, 7, 7, 128)\n","\n","  model.add(Conv2DTranspose(filters=64, kernel_size=(5,5), strides=(2,2), padding='same', use_bias=False))\n","  model.add(BatchNormalization())\n","  model.add(LeakyReLU())\n","  assert model.output_shape == (None, 14, 14, 64)\n","\n","  model.add(Conv2DTranspose(filters=1, kernel_size=(5,5), strides=(2,2), padding='same', use_bias=False,\n","                            activation=\"tanh\"))\n","  assert model.output_shape == (None, 28, 28, 1)\n","\n","  return model"],"execution_count":13,"outputs":[]},{"cell_type":"code","metadata":{"id":"wCNMu-sFhLiA","colab_type":"code","colab":{"base_uri":"https://localhost:8080/","height":265},"executionInfo":{"status":"ok","timestamp":1594088493623,"user_tz":-480,"elapsed":7377,"user":{"displayName":"王DevOps","photoUrl":"https://lh3.googleusercontent.com/a-/AOh14GgXVAgB9A4OE1KXeAfp5b-xUS2OSsbqSRVEe_UETw=s64","userId":"04300517850278510646"}},"outputId":"ad885183-1351-4148-b922-edfe0de2271c"},"source":["generator = build_generator_model()\n","\n","generator_input = tf.random.normal(shape=[1, 100])\n","generator_outputs = generator(generator_input, training=False)\n","\n","plt.imshow(generator_outputs[0, :, :, 0], cmap='gray')\n","plt.show()"],"execution_count":14,"outputs":[{"output_type":"display_data","data":{"image/png":"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\n","text/plain":["<Figure size 432x288 with 1 Axes>"]},"metadata":{"tags":[],"needs_background":"light"}}]},{"cell_type":"markdown","metadata":{"id":"w8ngI-RpnZQI","colab_type":"text"},"source":["## The Discriminator\n","\n","The discriminator is basically a CNN network."]},{"cell_type":"code","metadata":{"id":"Y-s-9asfi5nl","colab_type":"code","colab":{},"executionInfo":{"status":"ok","timestamp":1594088777529,"user_tz":-480,"elapsed":1077,"user":{"displayName":"王DevOps","photoUrl":"https://lh3.googleusercontent.com/a-/AOh14GgXVAgB9A4OE1KXeAfp5b-xUS2OSsbqSRVEe_UETw=s64","userId":"04300517850278510646"}}},"source":["def build_discriminator_model():\n","  model = tf.keras.Sequential()\n","\n","  # [None, 28, 28, 64]\n","  model.add(Conv2D(filters=64, kernel_size=(5,5), strides=(1,1), padding=\"same\", \n","                   input_shape=[28,28,1]))\n","  model.add(LeakyReLU())\n","  model.add(Dropout(rate=0.3))\n","\n","  # [None, 14, 14, 128]\n","  model.add(Conv2D(filters=128, kernel_size=(3,3), strides=(2,2), padding='same'))\n","  model.add(LeakyReLU())\n","  model.add(Dropout(rate=0.3))\n","\n","  model.add(Flatten())\n","  model.add(Dense(units=1))\n","\n","  return model"],"execution_count":15,"outputs":[]},{"cell_type":"markdown","metadata":{"id":"UGjAbJEjpIAx","colab_type":"text"},"source":["The output of the discriminator was trained that the negative values are for the fake images and the positive values are for real ones."]},{"cell_type":"code","metadata":{"id":"cu58-QX6ojB6","colab_type":"code","colab":{"base_uri":"https://localhost:8080/","height":34},"executionInfo":{"status":"ok","timestamp":1594088784848,"user_tz":-480,"elapsed":1076,"user":{"displayName":"王DevOps","photoUrl":"https://lh3.googleusercontent.com/a-/AOh14GgXVAgB9A4OE1KXeAfp5b-xUS2OSsbqSRVEe_UETw=s64","userId":"04300517850278510646"}},"outputId":"06fbe674-03b8-4dcd-8ed4-dd7d29b706bb"},"source":["discriminator = build_discriminator_model()\n","discriminator_outputs = discriminator(generator_outputs)\n","discriminator_outputs"],"execution_count":16,"outputs":[{"output_type":"execute_result","data":{"text/plain":["<tf.Tensor: shape=(1, 1), dtype=float32, numpy=array([[0.00205505]], dtype=float32)>"]},"metadata":{"tags":[]},"execution_count":16}]},{"cell_type":"markdown","metadata":{"id":"j3LHMK0mqOAj","colab_type":"text"},"source":["# Define the losses and optimizers\n","\n","Define the loss functions and the optimizers for both models."]},{"cell_type":"code","metadata":{"id":"QYytzMq_pSh2","colab_type":"code","colab":{},"executionInfo":{"status":"ok","timestamp":1594089109332,"user_tz":-480,"elapsed":1080,"user":{"displayName":"王DevOps","photoUrl":"https://lh3.googleusercontent.com/a-/AOh14GgXVAgB9A4OE1KXeAfp5b-xUS2OSsbqSRVEe_UETw=s64","userId":"04300517850278510646"}}},"source":["# define the cross entropy as the helper function\n","cross_entropy = tf.keras.losses.BinaryCrossentropy(from_logits=True)"],"execution_count":17,"outputs":[]},{"cell_type":"markdown","metadata":{"id":"YRrKyGSqqt0y","colab_type":"text"},"source":["## Discriminator Loss\n","\n","The discriminator's loss quantifies how well the discriminator can tell the real images from fakes. It compares the discriminator's predictions on real images to an array of 1s, and the discriminator's predictions on fake images to an array of 0s."]},{"cell_type":"code","metadata":{"id":"lLw6eUcbqsYK","colab_type":"code","colab":{},"executionInfo":{"status":"ok","timestamp":1594089132664,"user_tz":-480,"elapsed":1025,"user":{"displayName":"王DevOps","photoUrl":"https://lh3.googleusercontent.com/a-/AOh14GgXVAgB9A4OE1KXeAfp5b-xUS2OSsbqSRVEe_UETw=s64","userId":"04300517850278510646"}}},"source":["def discriminator_loss(real_output, fake_output):\n","  real_loss = cross_entropy(tf.ones_like(real_output), real_output)\n","  fake_loss = cross_entropy(tf.zeros_like(fake_output), fake_output)\n","  total_loss = real_loss + fake_loss\n","  return total_loss"],"execution_count":18,"outputs":[]},{"cell_type":"markdown","metadata":{"id":"3RqltBaHsMJY","colab_type":"text"},"source":["## Generator Loss"]},{"cell_type":"markdown","metadata":{"id":"cHLHC4Ewttxl","colab_type":"text"},"source":["The generator's loss quantifies how well the generator model can trick the discriminator model. If the generator performs well, the discriminator will classify the fake images as real (or 1). Here, we will compare the discriminator decisions on the generated images to an array of 1s."]},{"cell_type":"code","metadata":{"id":"MehxvEWMsIiK","colab_type":"code","colab":{},"executionInfo":{"status":"ok","timestamp":1594089192091,"user_tz":-480,"elapsed":815,"user":{"displayName":"王DevOps","photoUrl":"https://lh3.googleusercontent.com/a-/AOh14GgXVAgB9A4OE1KXeAfp5b-xUS2OSsbqSRVEe_UETw=s64","userId":"04300517850278510646"}}},"source":["def generator_loss(fake_output):\n","  # the generator learns to make the discriminator predictions became real \n","  # (or an array of 1s) on the fake images\n","  return cross_entropy(tf.ones_like(fake_output), fake_output)"],"execution_count":19,"outputs":[]},{"cell_type":"markdown","metadata":{"id":"9zS2NWjkuNyb","colab_type":"text"},"source":["## Define optimizers."]},{"cell_type":"code","metadata":{"id":"8y7EO5dNuLUg","colab_type":"code","colab":{},"executionInfo":{"status":"ok","timestamp":1594089198453,"user_tz":-480,"elapsed":1062,"user":{"displayName":"王DevOps","photoUrl":"https://lh3.googleusercontent.com/a-/AOh14GgXVAgB9A4OE1KXeAfp5b-xUS2OSsbqSRVEe_UETw=s64","userId":"04300517850278510646"}}},"source":["generator_optimizer = tf.keras.optimizers.Adam(learning_rate=1e-4)\n","discriminator_optimizer = tf.keras.optimizers.Adam(learning_rate=1e-4)"],"execution_count":20,"outputs":[]},{"cell_type":"markdown","metadata":{"id":"7fmrruZ3ugYW","colab_type":"text"},"source":["## Save Checkpoints"]},{"cell_type":"code","metadata":{"id":"0nLMo4fGucp5","colab_type":"code","colab":{"base_uri":"https://localhost:8080/","height":34},"executionInfo":{"status":"ok","timestamp":1594089519715,"user_tz":-480,"elapsed":1055,"user":{"displayName":"王DevOps","photoUrl":"https://lh3.googleusercontent.com/a-/AOh14GgXVAgB9A4OE1KXeAfp5b-xUS2OSsbqSRVEe_UETw=s64","userId":"04300517850278510646"}},"outputId":"144caa8f-bd36-4cb8-9f08-82db6ae6caf7"},"source":["ckpt_dir = \"./gan_ckpt\"\n","ckpt_prefix = os.path.join(ckpt_dir, \"ckpt\")\n","ckpt = tf.train.Checkpoint(generator_optimizer=generator_optimizer,\n","                           discriminator_optimizer=discriminator_optimizer,\n","                           generator=generator,\n","                           discriminator=discriminator)\n","ckpt"],"execution_count":21,"outputs":[{"output_type":"execute_result","data":{"text/plain":["<tensorflow.python.training.tracking.util.Checkpoint at 0x7f94f60b9f28>"]},"metadata":{"tags":[]},"execution_count":21}]},{"cell_type":"markdown","metadata":{"id":"jS3oLFaFjO5l","colab_type":"text"},"source":["# Define the training loop"]},{"cell_type":"code","metadata":{"id":"0JLDYl7KvGA4","colab_type":"code","colab":{},"executionInfo":{"status":"ok","timestamp":1594090112776,"user_tz":-480,"elapsed":1143,"user":{"displayName":"王DevOps","photoUrl":"https://lh3.googleusercontent.com/a-/AOh14GgXVAgB9A4OE1KXeAfp5b-xUS2OSsbqSRVEe_UETw=s64","userId":"04300517850278510646"}}},"source":["EPOCHS = 50\n","noise_dim = 100\n","num_generated_examples = 16\n","\n","# You will reuse the seed overtime to visualize progress in the animated GIF.\n","seed = tf.random.normal(shape=[num_generated_examples, noise_dim])"],"execution_count":23,"outputs":[]},{"cell_type":"markdown","metadata":{"id":"yqqDzMlYqiGa","colab_type":"text"},"source":["In the training loop, the generator model takes the noise as the input to generate the fake images. The discriminator model takes real images and fake images to give the discriminations (or outputs) for them. Calculate the generator and discriminator losses each using the real outputs and the fake outputs. Calculate the gradients of the model trainable variables based on these losses and then apply gradients back to them."]},{"cell_type":"code","metadata":{"id":"H_zApEtLkEy2","colab_type":"code","colab":{},"executionInfo":{"status":"ok","timestamp":1594090115029,"user_tz":-480,"elapsed":980,"user":{"displayName":"王DevOps","photoUrl":"https://lh3.googleusercontent.com/a-/AOh14GgXVAgB9A4OE1KXeAfp5b-xUS2OSsbqSRVEe_UETw=s64","userId":"04300517850278510646"}}},"source":["@tf.function\n","def train_step(images):\n","  fake_noises = tf.random.normal(shape=[BATCH_SIZE, noise_dim])\n","\n","  with tf.GradientTape() as disc_tape, tf.GradientTape() as gen_tape:\n","    fake_images = generator(fake_noises, training=True)\n","\n","    fake_outputs = discriminator(fake_images, training=True)\n","    real_outputs = discriminator(images, training=True)\n","\n","    disc_loss = discriminator_loss(real_output=real_outputs, \n","                                   fake_output=fake_outputs)\n","    gen_loss = generator_loss(fake_output=fake_outputs)\n","\n","  disc_gradients = disc_tape.gradient(disc_loss, discriminator.trainable_variables)\n","  gen_gradients = gen_tape.gradient(gen_loss, generator.trainable_variables)\n","\n","  discriminator_optimizer.apply_gradients(zip(disc_gradients, discriminator.trainable_variables))\n","  generator_optimizer.apply_gradients(zip(gen_gradients, generator.trainable_variables))"],"execution_count":24,"outputs":[]},{"cell_type":"code","metadata":{"id":"6J_ctXX3pAgv","colab_type":"code","colab":{},"executionInfo":{"status":"ok","timestamp":1594090116222,"user_tz":-480,"elapsed":648,"user":{"displayName":"王DevOps","photoUrl":"https://lh3.googleusercontent.com/a-/AOh14GgXVAgB9A4OE1KXeAfp5b-xUS2OSsbqSRVEe_UETw=s64","userId":"04300517850278510646"}}},"source":["def generate_and_save_images(model, epoch, test_input):\n","  \"\"\"Helps to generate the images from a fixed seed.\"\"\"\n","  predictions = model(test_input, training=False)\n","\n","  fig = plt.figure(figsize=(8,8))\n","  for i in range(predictions.shape[0]):\n","    plt.subplot(4, 4, i+1)\n","    plt.imshow(predictions[i, :, :, 0] * 127.5 + 127.5, cmap='gray')\n","    plt.axis(\"off\")\n","  plt.savefig('image_epoch_{:04d}.png'.format(epoch))\n","  plt.show()"],"execution_count":25,"outputs":[]},{"cell_type":"code","metadata":{"id":"WQOR4HCMtL0M","colab_type":"code","colab":{},"executionInfo":{"status":"ok","timestamp":1594090124038,"user_tz":-480,"elapsed":1034,"user":{"displayName":"王DevOps","photoUrl":"https://lh3.googleusercontent.com/a-/AOh14GgXVAgB9A4OE1KXeAfp5b-xUS2OSsbqSRVEe_UETw=s64","userId":"04300517850278510646"}}},"source":["def train(dataset, epochs):\n","  for epoch in range(epochs):\n","    start = time.time()\n","    for batch_dataset in dataset:\n","      train_step(batch_dataset)\n","    \n","    clear_output(wait=True)\n","    generate_and_save_images(generator, epoch+1, seed)\n","\n","    if (epoch+1) % 15 == 0:\n","      ckpt.save(file_prefix=ckpt_prefix)\n","    print(\"Epoch {} in time {}.\".format(epoch + 1, time.time()-start))\n","\n","  # after the training\n","  clear_output(wait=True)\n","  generate_and_save_images(generator, epoch+1, seed)"],"execution_count":26,"outputs":[]},{"cell_type":"markdown","metadata":{"id":"viNE-YAouox7","colab_type":"text"},"source":["## Train the Model\n","\n","Call the `train()` function to start the model training. Note, training GANs can be tricky. It's important that the generator and discriminator do not overpower each other (e.g. they train at a similar rate)."]},{"cell_type":"code","metadata":{"id":"LwZngQf5unsc","colab_type":"code","colab":{"base_uri":"https://localhost:8080/","height":466},"executionInfo":{"status":"ok","timestamp":1594091604167,"user_tz":-480,"elapsed":1179718,"user":{"displayName":"王DevOps","photoUrl":"https://lh3.googleusercontent.com/a-/AOh14GgXVAgB9A4OE1KXeAfp5b-xUS2OSsbqSRVEe_UETw=s64","userId":"04300517850278510646"}},"outputId":"8faa273f-df55-4b38-ec5e-08ed4d2a3027"},"source":["train(train_datasets, epochs=EPOCHS)"],"execution_count":27,"outputs":[{"output_type":"display_data","data":{"image/png":"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\n","text/plain":["<Figure size 576x576 with 16 Axes>"]},"metadata":{"tags":[],"needs_background":"light"}}]},{"cell_type":"markdown","metadata":{"id":"34DTWzGAylSz","colab_type":"text"},"source":["# Create a GIF"]},{"cell_type":"code","metadata":{"id":"z5e4qDWVvqIl","colab_type":"code","colab":{},"executionInfo":{"status":"ok","timestamp":1594091755285,"user_tz":-480,"elapsed":1148,"user":{"displayName":"王DevOps","photoUrl":"https://lh3.googleusercontent.com/a-/AOh14GgXVAgB9A4OE1KXeAfp5b-xUS2OSsbqSRVEe_UETw=s64","userId":"04300517850278510646"}}},"source":["def display_image(epoch_no):\n","  image_path = 'image_epoch_{:04d}.png'.format(epoch_no)\n","  img = plt.imread(fname=image_path)\n","  plt.imshow(img)\n","  plt.margins(0)\n","  plt.axis(\"off\")\n","  plt.tight_layout()\n","  plt.show()"],"execution_count":33,"outputs":[]},{"cell_type":"code","metadata":{"id":"55zkvDKlzD9w","colab_type":"code","colab":{"base_uri":"https://localhost:8080/","height":297},"executionInfo":{"status":"ok","timestamp":1594091757074,"user_tz":-480,"elapsed":1346,"user":{"displayName":"王DevOps","photoUrl":"https://lh3.googleusercontent.com/a-/AOh14GgXVAgB9A4OE1KXeAfp5b-xUS2OSsbqSRVEe_UETw=s64","userId":"04300517850278510646"}},"outputId":"e53220f9-39a4-4432-8c17-78aa9229dc22"},"source":["display_image(50)"],"execution_count":34,"outputs":[{"output_type":"display_data","data":{"image/png":"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\n","text/plain":["<Figure size 432x288 with 1 Axes>"]},"metadata":{"tags":[],"needs_background":"light"}}]},{"cell_type":"code","metadata":{"id":"d13hfUqSzGZu","colab_type":"code","colab":{},"executionInfo":{"status":"ok","timestamp":1594091717027,"user_tz":-480,"elapsed":2569,"user":{"displayName":"王DevOps","photoUrl":"https://lh3.googleusercontent.com/a-/AOh14GgXVAgB9A4OE1KXeAfp5b-xUS2OSsbqSRVEe_UETw=s64","userId":"04300517850278510646"}}},"source":["anim_file = 'dcgan.gif'\n","\n","with imageio.get_writer(anim_file, mode=\"I\") as writer:\n","  filenames = glob.glob('image*.png')\n","  filenames = sorted(filenames)\n","\n","  for _, filename in enumerate(filenames):\n","    image = imageio.imread(filename)\n","    writer.append_data(image)"],"execution_count":30,"outputs":[]},{"cell_type":"code","metadata":{"id":"mIQg-tmL0Vy8","colab_type":"code","colab":{}},"source":["try:\n","  from google.colab import files\n","except ImportError:\n","  pass\n","else:\n","  files.download(anim_file)"],"execution_count":null,"outputs":[]},{"cell_type":"code","metadata":{"id":"dS7h1FtXN5Xj","colab_type":"code","colab":{}},"source":[""],"execution_count":null,"outputs":[]}]}