PixelCNN.ipynb

{
  "nbformat": 4,
  "nbformat_minor": 0,
  "metadata": {
    "colab": {
      "name": "Pixel CNN",
      "provenance": [],
      "collapsed_sections": [],
      "authorship_tag": "ABX9TyPmXUZvp1YEXqdVET9KNOxU",
      "include_colab_link": true
    },
    "kernelspec": {
      "name": "python3",
      "display_name": "Python 3"
    },
    "language_info": {
      "name": "python"
    },
    "accelerator": "GPU",
    "widgets": {
      "application/vnd.jupyter.widget-state+json": {
        "252cabe756f44006b4dfa506d40678aa": {
          "model_module": "@jupyter-widgets/controls",
          "model_name": "HBoxModel",
          "state": {
            "_view_name": "HBoxView",
            "_dom_classes": [],
            "_model_name": "HBoxModel",
            "_view_module": "@jupyter-widgets/controls",
            "_model_module_version": "1.5.0",
            "_view_count": null,
            "_view_module_version": "1.5.0",
            "box_style": "",
            "layout": "IPY_MODEL_815a40ff0e6b434382b751866f620fe0",
            "_model_module": "@jupyter-widgets/controls",
            "children": [
              "IPY_MODEL_0fa4f5ba046e40d7af9dfc35897f4fcf",
              "IPY_MODEL_f31e10d7ffc34af8b44fde5a3f4c12bb"
            ]
          }
        },
        "815a40ff0e6b434382b751866f620fe0": {
          "model_module": "@jupyter-widgets/base",
          "model_name": "LayoutModel",
          "state": {
            "_view_name": "LayoutView",
            "grid_template_rows": null,
            "right": null,
            "justify_content": null,
            "_view_module": "@jupyter-widgets/base",
            "overflow": null,
            "_model_module_version": "1.2.0",
            "_view_count": null,
            "flex_flow": null,
            "width": null,
            "min_width": null,
            "border": null,
            "align_items": null,
            "bottom": null,
            "_model_module": "@jupyter-widgets/base",
            "top": null,
            "grid_column": null,
            "overflow_y": null,
            "overflow_x": null,
            "grid_auto_flow": null,
            "grid_area": null,
            "grid_template_columns": null,
            "flex": null,
            "_model_name": "LayoutModel",
            "justify_items": null,
            "grid_row": null,
            "max_height": null,
            "align_content": null,
            "visibility": null,
            "align_self": null,
            "height": null,
            "min_height": null,
            "padding": null,
            "grid_auto_rows": null,
            "grid_gap": null,
            "max_width": null,
            "order": null,
            "_view_module_version": "1.2.0",
            "grid_template_areas": null,
            "object_position": null,
            "object_fit": null,
            "grid_auto_columns": null,
            "margin": null,
            "display": null,
            "left": null
          }
        },
        "0fa4f5ba046e40d7af9dfc35897f4fcf": {
          "model_module": "@jupyter-widgets/controls",
          "model_name": "FloatProgressModel",
          "state": {
            "_view_name": "ProgressView",
            "style": "IPY_MODEL_51a113008cea414b9d79d139133739b0",
            "_dom_classes": [],
            "description": "Dl Completed...: 100%",
            "_model_name": "FloatProgressModel",
            "bar_style": "success",
            "max": 4,
            "_view_module": "@jupyter-widgets/controls",
            "_model_module_version": "1.5.0",
            "value": 4,
            "_view_count": null,
            "_view_module_version": "1.5.0",
            "orientation": "horizontal",
            "min": 0,
            "description_tooltip": null,
            "_model_module": "@jupyter-widgets/controls",
            "layout": "IPY_MODEL_6b35ae83702d4545a249846d1c49d0ea"
          }
        },
        "f31e10d7ffc34af8b44fde5a3f4c12bb": {
          "model_module": "@jupyter-widgets/controls",
          "model_name": "HTMLModel",
          "state": {
            "_view_name": "HTMLView",
            "style": "IPY_MODEL_4b62c62057f74b78a9e3bd0c0ce6ee80",
            "_dom_classes": [],
            "description": "",
            "_model_name": "HTMLModel",
            "placeholder": "​",
            "_view_module": "@jupyter-widgets/controls",
            "_model_module_version": "1.5.0",
            "value": " 4/4 [00:06&lt;00:00,  1.56s/ file]",
            "_view_count": null,
            "_view_module_version": "1.5.0",
            "description_tooltip": null,
            "_model_module": "@jupyter-widgets/controls",
            "layout": "IPY_MODEL_2fd9815bf24744309deebc7d2be06d2c"
          }
        },
        "51a113008cea414b9d79d139133739b0": {
          "model_module": "@jupyter-widgets/controls",
          "model_name": "ProgressStyleModel",
          "state": {
            "_view_name": "StyleView",
            "_model_name": "ProgressStyleModel",
            "description_width": "initial",
            "_view_module": "@jupyter-widgets/base",
            "_model_module_version": "1.5.0",
            "_view_count": null,
            "_view_module_version": "1.2.0",
            "bar_color": null,
            "_model_module": "@jupyter-widgets/controls"
          }
        },
        "6b35ae83702d4545a249846d1c49d0ea": {
          "model_module": "@jupyter-widgets/base",
          "model_name": "LayoutModel",
          "state": {
            "_view_name": "LayoutView",
            "grid_template_rows": null,
            "right": null,
            "justify_content": null,
            "_view_module": "@jupyter-widgets/base",
            "overflow": null,
            "_model_module_version": "1.2.0",
            "_view_count": null,
            "flex_flow": null,
            "width": null,
            "min_width": null,
            "border": null,
            "align_items": null,
            "bottom": null,
            "_model_module": "@jupyter-widgets/base",
            "top": null,
            "grid_column": null,
            "overflow_y": null,
            "overflow_x": null,
            "grid_auto_flow": null,
            "grid_area": null,
            "grid_template_columns": null,
            "flex": null,
            "_model_name": "LayoutModel",
            "justify_items": null,
            "grid_row": null,
            "max_height": null,
            "align_content": null,
            "visibility": null,
            "align_self": null,
            "height": null,
            "min_height": null,
            "padding": null,
            "grid_auto_rows": null,
            "grid_gap": null,
            "max_width": null,
            "order": null,
            "_view_module_version": "1.2.0",
            "grid_template_areas": null,
            "object_position": null,
            "object_fit": null,
            "grid_auto_columns": null,
            "margin": null,
            "display": null,
            "left": null
          }
        },
        "4b62c62057f74b78a9e3bd0c0ce6ee80": {
          "model_module": "@jupyter-widgets/controls",
          "model_name": "DescriptionStyleModel",
          "state": {
            "_view_name": "StyleView",
            "_model_name": "DescriptionStyleModel",
            "description_width": "",
            "_view_module": "@jupyter-widgets/base",
            "_model_module_version": "1.5.0",
            "_view_count": null,
            "_view_module_version": "1.2.0",
            "_model_module": "@jupyter-widgets/controls"
          }
        },
        "2fd9815bf24744309deebc7d2be06d2c": {
          "model_module": "@jupyter-widgets/base",
          "model_name": "LayoutModel",
          "state": {
            "_view_name": "LayoutView",
            "grid_template_rows": null,
            "right": null,
            "justify_content": null,
            "_view_module": "@jupyter-widgets/base",
            "overflow": null,
            "_model_module_version": "1.2.0",
            "_view_count": null,
            "flex_flow": null,
            "width": null,
            "min_width": null,
            "border": null,
            "align_items": null,
            "bottom": null,
            "_model_module": "@jupyter-widgets/base",
            "top": null,
            "grid_column": null,
            "overflow_y": null,
            "overflow_x": null,
            "grid_auto_flow": null,
            "grid_area": null,
            "grid_template_columns": null,
            "flex": null,
            "_model_name": "LayoutModel",
            "justify_items": null,
            "grid_row": null,
            "max_height": null,
            "align_content": null,
            "visibility": null,
            "align_self": null,
            "height": null,
            "min_height": null,
            "padding": null,
            "grid_auto_rows": null,
            "grid_gap": null,
            "max_width": null,
            "order": null,
            "_view_module_version": "1.2.0",
            "grid_template_areas": null,
            "object_position": null,
            "object_fit": null,
            "grid_auto_columns": null,
            "margin": null,
            "display": null,
            "left": null
          }
        }
      }
    }
  },
  "cells": [
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "view-in-github",
        "colab_type": "text"
      },
      "source": [
        "<a href=\"https://colab.research.google.com/github/dude123studios/AdvancedGenerativeLearning/blob/main/Pixel_CNN.ipynb\" target=\"_parent\"><img src=\"https://colab.research.google.com/assets/colab-badge.svg\" alt=\"Open In Colab\"/></a>"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "N_ntJ6t46AGH"
      },
      "source": [
        "import tensorflow as tf\n",
        "from tensorflow.keras.layers import *\n",
        "import numpy as np\n",
        "import tensorflow_datasets as tfds\n",
        "import matplotlib.pyplot as plt"
      ],
      "execution_count": 1,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 202,
          "referenced_widgets": [
            "252cabe756f44006b4dfa506d40678aa",
            "815a40ff0e6b434382b751866f620fe0",
            "0fa4f5ba046e40d7af9dfc35897f4fcf",
            "f31e10d7ffc34af8b44fde5a3f4c12bb",
            "51a113008cea414b9d79d139133739b0",
            "6b35ae83702d4545a249846d1c49d0ea",
            "4b62c62057f74b78a9e3bd0c0ce6ee80",
            "2fd9815bf24744309deebc7d2be06d2c"
          ]
        },
        "id": "uB8FPESuMnNo",
        "outputId": "3ef22f72-0818-4c33-dafd-1bbcbb9a693b"
      },
      "source": [
        "(ds_train, ds_test), ds_info = tfds.load('mnist', \n",
        "                              split=['test', 'test'], \n",
        "                              shuffle_files=True,\n",
        "                              as_supervised=True,\n",
        "                              with_info=True)\n",
        "\n",
        "def binarize(image, label):\n",
        "    image = tf.cast(image, tf.float32)\n",
        "    image = tf.math.round(image / 255.)\n",
        "    return image, tf.cast(image, tf.int32)\n",
        "\n",
        "ds_train = ds_train.map(binarize)\n",
        "ds_train = ds_train.cache()\n",
        "ds_train = ds_train.shuffle(ds_info.splits['train'].num_examples)\n",
        "ds_train = ds_train.batch(64)\n",
        "ds_test = ds_test.map(binarize).batch(64).cache().prefetch(64)"
      ],
      "execution_count": 2,
      "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": "display_data",
          "data": {
            "application/vnd.jupyter.widget-view+json": {
              "model_id": "252cabe756f44006b4dfa506d40678aa",
              "version_minor": 0,
              "version_major": 2
            },
            "text/plain": [
              "HBox(children=(FloatProgress(value=0.0, description='Dl Completed...', max=4.0, style=ProgressStyle(descriptio…"
            ]
          },
          "metadata": {
            "tags": []
          }
        },
        {
          "output_type": "stream",
          "text": [
            "\n",
            "\n",
            "\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": "zGU-0DisQI0Q"
      },
      "source": [
        "class MaskedConv2D(tf.keras.layers.Layer):\n",
        "\n",
        "    def __init__(self, mask_type, kernel = 5, filters = 1):\n",
        "        super(MaskedConv2D, self).__init__()\n",
        "        self.kernel = kernel\n",
        "        self.filters = filters\n",
        "        self.mask_type = mask_type\n",
        "    \n",
        "    def build(self, input_shape):\n",
        "        self.w = self.add_weight(shape=(self.kernel,\n",
        "                                        self.kernel,\n",
        "                                        input_shape[-1],\n",
        "                                        self.filters),\n",
        "                                  initializer='glorot_normal',\n",
        "                                  trainable=True)\n",
        "        self.b = self.add_weight(shape=(self.filters,), trainable=True, initializer='zeros')\n",
        "\n",
        "        mask = np.ones(self.kernel**2, dtype=np.float32)\n",
        "        center = len(mask) // 2\n",
        "        mask[center+1:] = 0\n",
        "        if self.mask_type == 'A':\n",
        "            mask[center] = 0\n",
        "        mask = mask.reshape((self.kernel, self.kernel, 1, 1))\n",
        "        self.mask  = tf.constant(mask, dtype=tf.float32)\n",
        "    \n",
        "    def call(self, inputs):\n",
        "        masked_w = tf.math.multiply(self.w, self.mask)\n",
        "        output = tf.nn.conv2d(inputs, masked_w, 1, padding='SAME') + self.b\n",
        "        return tf.nn.relu(output)"
      ],
      "execution_count": 3,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "5udfj31wizSK"
      },
      "source": [
        "class ResidualBlock(tf.keras.layers.Layer):\n",
        "    def __init__(self, h=32):\n",
        "        super(ResidualBlock, self).__init__()\n",
        "\n",
        "        self.forward = tf.keras.models.Sequential([\n",
        "            MaskedConv2D('B', kernel=1, filters=h),\n",
        "            MaskedConv2D('B', kernel=3, filters=h),\n",
        "            MaskedConv2D('B', kernel=1, filters=2*h)\n",
        "        ])\n",
        "    \n",
        "    def call(self, inputs):\n",
        "        return self.forward(inputs) + inputs"
      ],
      "execution_count": 4,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "R-heFsPvkq15"
      },
      "source": [
        "def SimplePixelCNN(hidden_features=64,\n",
        "                   output_features=64,\n",
        "                   num_resblocks=7):\n",
        "    inputs = Input(shape=(28, 28, 1), dtype=tf.float32)\n",
        "    x = inputs\n",
        "\n",
        "    x = MaskedConv2D('A', kernel=7, filters=2*hidden_features)(x)\n",
        "\n",
        "    for _ in range(num_resblocks):\n",
        "        x = ResidualBlock(hidden_features)(x)\n",
        "    \n",
        "    x = Conv2D(output_features, (1, 1), padding='same', activation='relu')(x)\n",
        "    x = Conv2D(1, (1, 1), padding='same', activation='sigmoid')(x)\n",
        "\n",
        "    return tf.keras.Model(inputs=inputs, outputs=x, name='PixelCNN')"
      ],
      "execution_count": 5,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "gQQzntV0os7w",
        "outputId": "dc13df7e-2cd6-46d7-fd7f-d45361b7bc66"
      },
      "source": [
        "pixelCNN = SimplePixelCNN()\n",
        "pixelCNN.summary()"
      ],
      "execution_count": 6,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Model: \"PixelCNN\"\n",
            "_________________________________________________________________\n",
            "Layer (type)                 Output Shape              Param #   \n",
            "=================================================================\n",
            "input_1 (InputLayer)         [(None, 28, 28, 1)]       0         \n",
            "_________________________________________________________________\n",
            "masked_conv2d (MaskedConv2D) (None, 28, 28, 128)       6400      \n",
            "_________________________________________________________________\n",
            "residual_block (ResidualBloc (None, 28, 28, 128)       53504     \n",
            "_________________________________________________________________\n",
            "residual_block_1 (ResidualBl (None, 28, 28, 128)       53504     \n",
            "_________________________________________________________________\n",
            "residual_block_2 (ResidualBl (None, 28, 28, 128)       53504     \n",
            "_________________________________________________________________\n",
            "residual_block_3 (ResidualBl (None, 28, 28, 128)       53504     \n",
            "_________________________________________________________________\n",
            "residual_block_4 (ResidualBl (None, 28, 28, 128)       53504     \n",
            "_________________________________________________________________\n",
            "residual_block_5 (ResidualBl (None, 28, 28, 128)       53504     \n",
            "_________________________________________________________________\n",
            "residual_block_6 (ResidualBl (None, 28, 28, 128)       53504     \n",
            "_________________________________________________________________\n",
            "conv2d (Conv2D)              (None, 28, 28, 64)        8256      \n",
            "_________________________________________________________________\n",
            "conv2d_1 (Conv2D)            (None, 28, 28, 1)         65        \n",
            "=================================================================\n",
            "Total params: 389,249\n",
            "Trainable params: 389,249\n",
            "Non-trainable params: 0\n",
            "_________________________________________________________________\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "d7zNzlFQo6nA",
        "outputId": "0220b39f-ec30-485a-b0d4-b5255600f27e"
      },
      "source": [
        "pixelCNN.compile(\n",
        "    optimizer='rmsprop',\n",
        "    loss = tf.keras.losses.BinaryCrossentropy(),\n",
        "    metrics = [tf.keras.losses.BinaryCrossentropy()]\n",
        ")\n",
        "\n",
        "pixelCNN.fit(ds_train, epochs=50, validation_data=ds_test)"
      ],
      "execution_count": 7,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Epoch 1/50\n",
            "157/157 [==============================] - 49s 93ms/step - loss: 0.1405 - binary_crossentropy: 0.1405 - val_loss: 0.1074 - val_binary_crossentropy: 0.1074\n",
            "Epoch 2/50\n",
            "157/157 [==============================] - 14s 89ms/step - loss: 0.1024 - binary_crossentropy: 0.1025 - val_loss: 0.1165 - val_binary_crossentropy: 0.1165\n",
            "Epoch 3/50\n",
            "157/157 [==============================] - 14s 90ms/step - loss: 0.0978 - binary_crossentropy: 0.0978 - val_loss: 0.0954 - val_binary_crossentropy: 0.0954\n",
            "Epoch 4/50\n",
            "157/157 [==============================] - 14s 90ms/step - loss: 0.0946 - binary_crossentropy: 0.0946 - val_loss: 0.0951 - val_binary_crossentropy: 0.0951\n",
            "Epoch 5/50\n",
            "157/157 [==============================] - 14s 90ms/step - loss: 0.0924 - binary_crossentropy: 0.0924 - val_loss: 0.0922 - val_binary_crossentropy: 0.0922\n",
            "Epoch 6/50\n",
            "157/157 [==============================] - 14s 90ms/step - loss: 0.0910 - binary_crossentropy: 0.0910 - val_loss: 0.0903 - val_binary_crossentropy: 0.0903\n",
            "Epoch 7/50\n",
            "157/157 [==============================] - 14s 90ms/step - loss: 0.0900 - binary_crossentropy: 0.0900 - val_loss: 0.0911 - val_binary_crossentropy: 0.0911\n",
            "Epoch 8/50\n",
            "157/157 [==============================] - 14s 91ms/step - loss: 0.0889 - binary_crossentropy: 0.0889 - val_loss: 0.0886 - val_binary_crossentropy: 0.0886\n",
            "Epoch 9/50\n",
            "157/157 [==============================] - 14s 90ms/step - loss: 0.0882 - binary_crossentropy: 0.0882 - val_loss: 0.0870 - val_binary_crossentropy: 0.0870\n",
            "Epoch 10/50\n",
            "157/157 [==============================] - 14s 90ms/step - loss: 0.0876 - binary_crossentropy: 0.0876 - val_loss: 0.0921 - val_binary_crossentropy: 0.0921\n",
            "Epoch 11/50\n",
            "157/157 [==============================] - 14s 91ms/step - loss: 0.0870 - binary_crossentropy: 0.0870 - val_loss: 0.0885 - val_binary_crossentropy: 0.0885\n",
            "Epoch 12/50\n",
            "157/157 [==============================] - 14s 91ms/step - loss: 0.0865 - binary_crossentropy: 0.0865 - val_loss: 0.0867 - val_binary_crossentropy: 0.0867\n",
            "Epoch 13/50\n",
            "157/157 [==============================] - 14s 91ms/step - loss: 0.0860 - binary_crossentropy: 0.0861 - val_loss: 0.0878 - val_binary_crossentropy: 0.0878\n",
            "Epoch 14/50\n",
            "157/157 [==============================] - 14s 91ms/step - loss: 0.0856 - binary_crossentropy: 0.0856 - val_loss: 0.0872 - val_binary_crossentropy: 0.0872\n",
            "Epoch 15/50\n",
            "157/157 [==============================] - 14s 91ms/step - loss: 0.0853 - binary_crossentropy: 0.0853 - val_loss: 0.0869 - val_binary_crossentropy: 0.0869\n",
            "Epoch 16/50\n",
            "157/157 [==============================] - 14s 91ms/step - loss: 0.0849 - binary_crossentropy: 0.0848 - val_loss: 0.0854 - val_binary_crossentropy: 0.0854\n",
            "Epoch 17/50\n",
            "157/157 [==============================] - 14s 91ms/step - loss: 0.0845 - binary_crossentropy: 0.0845 - val_loss: 0.0852 - val_binary_crossentropy: 0.0852\n",
            "Epoch 18/50\n",
            "157/157 [==============================] - 14s 91ms/step - loss: 0.0842 - binary_crossentropy: 0.0842 - val_loss: 0.0838 - val_binary_crossentropy: 0.0838\n",
            "Epoch 19/50\n",
            "157/157 [==============================] - 14s 91ms/step - loss: 0.0839 - binary_crossentropy: 0.0839 - val_loss: 0.0845 - val_binary_crossentropy: 0.0845\n",
            "Epoch 20/50\n",
            "157/157 [==============================] - 14s 91ms/step - loss: 0.0836 - binary_crossentropy: 0.0836 - val_loss: 0.0840 - val_binary_crossentropy: 0.0840\n",
            "Epoch 21/50\n",
            "157/157 [==============================] - 14s 91ms/step - loss: 0.0833 - binary_crossentropy: 0.0833 - val_loss: 0.0841 - val_binary_crossentropy: 0.0841\n",
            "Epoch 22/50\n",
            "157/157 [==============================] - 14s 91ms/step - loss: 0.0830 - binary_crossentropy: 0.0830 - val_loss: 0.0832 - val_binary_crossentropy: 0.0832\n",
            "Epoch 23/50\n",
            "157/157 [==============================] - 14s 91ms/step - loss: 0.0827 - binary_crossentropy: 0.0827 - val_loss: 0.0853 - val_binary_crossentropy: 0.0853\n",
            "Epoch 24/50\n",
            "157/157 [==============================] - 14s 91ms/step - loss: 0.0824 - binary_crossentropy: 0.0825 - val_loss: 0.0834 - val_binary_crossentropy: 0.0834\n",
            "Epoch 25/50\n",
            "157/157 [==============================] - 14s 91ms/step - loss: 0.0822 - binary_crossentropy: 0.0822 - val_loss: 0.0830 - val_binary_crossentropy: 0.0830\n",
            "Epoch 26/50\n",
            "157/157 [==============================] - 14s 91ms/step - loss: 0.0819 - binary_crossentropy: 0.0819 - val_loss: 0.0830 - val_binary_crossentropy: 0.0830\n",
            "Epoch 27/50\n",
            "157/157 [==============================] - 14s 91ms/step - loss: 0.0816 - binary_crossentropy: 0.0816 - val_loss: 0.0822 - val_binary_crossentropy: 0.0822\n",
            "Epoch 28/50\n",
            "157/157 [==============================] - 14s 91ms/step - loss: 0.0814 - binary_crossentropy: 0.0814 - val_loss: 0.0812 - val_binary_crossentropy: 0.0812\n",
            "Epoch 29/50\n",
            "157/157 [==============================] - 14s 91ms/step - loss: 0.0811 - binary_crossentropy: 0.0811 - val_loss: 0.0813 - val_binary_crossentropy: 0.0813\n",
            "Epoch 30/50\n",
            "157/157 [==============================] - 14s 91ms/step - loss: 0.0808 - binary_crossentropy: 0.0808 - val_loss: 0.0806 - val_binary_crossentropy: 0.0806\n",
            "Epoch 31/50\n",
            "157/157 [==============================] - 14s 91ms/step - loss: 0.0806 - binary_crossentropy: 0.0806 - val_loss: 0.0825 - val_binary_crossentropy: 0.0825\n",
            "Epoch 32/50\n",
            "157/157 [==============================] - 14s 92ms/step - loss: 0.0803 - binary_crossentropy: 0.0803 - val_loss: 0.0804 - val_binary_crossentropy: 0.0804\n",
            "Epoch 33/50\n",
            "157/157 [==============================] - 14s 91ms/step - loss: 0.0800 - binary_crossentropy: 0.0800 - val_loss: 0.0807 - val_binary_crossentropy: 0.0807\n",
            "Epoch 34/50\n",
            "157/157 [==============================] - 14s 91ms/step - loss: 0.0797 - binary_crossentropy: 0.0798 - val_loss: 0.0801 - val_binary_crossentropy: 0.0801\n",
            "Epoch 35/50\n",
            "157/157 [==============================] - 14s 91ms/step - loss: 0.0795 - binary_crossentropy: 0.0795 - val_loss: 0.0801 - val_binary_crossentropy: 0.0801\n",
            "Epoch 36/50\n",
            "157/157 [==============================] - 14s 91ms/step - loss: 0.0792 - binary_crossentropy: 0.0793 - val_loss: 0.0801 - val_binary_crossentropy: 0.0801\n",
            "Epoch 37/50\n",
            "157/157 [==============================] - 14s 91ms/step - loss: 0.0791 - binary_crossentropy: 0.0791 - val_loss: 0.0789 - val_binary_crossentropy: 0.0789\n",
            "Epoch 38/50\n",
            "157/157 [==============================] - 14s 91ms/step - loss: 0.0788 - binary_crossentropy: 0.0787 - val_loss: 0.0795 - val_binary_crossentropy: 0.0794\n",
            "Epoch 39/50\n",
            "157/157 [==============================] - 14s 91ms/step - loss: 0.0785 - binary_crossentropy: 0.0785 - val_loss: 0.0785 - val_binary_crossentropy: 0.0785\n",
            "Epoch 40/50\n",
            "157/157 [==============================] - 14s 92ms/step - loss: 0.0783 - binary_crossentropy: 0.0783 - val_loss: 0.0784 - val_binary_crossentropy: 0.0784\n",
            "Epoch 41/50\n",
            "157/157 [==============================] - 14s 91ms/step - loss: 0.0780 - binary_crossentropy: 0.0780 - val_loss: 0.0787 - val_binary_crossentropy: 0.0787\n",
            "Epoch 42/50\n",
            "157/157 [==============================] - 14s 91ms/step - loss: 0.0778 - binary_crossentropy: 0.0778 - val_loss: 0.0777 - val_binary_crossentropy: 0.0777\n",
            "Epoch 43/50\n",
            "157/157 [==============================] - 14s 91ms/step - loss: 0.0776 - binary_crossentropy: 0.0776 - val_loss: 0.0784 - val_binary_crossentropy: 0.0784\n",
            "Epoch 44/50\n",
            "157/157 [==============================] - 14s 91ms/step - loss: 0.0773 - binary_crossentropy: 0.0773 - val_loss: 0.0779 - val_binary_crossentropy: 0.0779\n",
            "Epoch 45/50\n",
            "157/157 [==============================] - 14s 91ms/step - loss: 0.0771 - binary_crossentropy: 0.0771 - val_loss: 0.0776 - val_binary_crossentropy: 0.0776\n",
            "Epoch 46/50\n",
            "157/157 [==============================] - 14s 91ms/step - loss: 0.0769 - binary_crossentropy: 0.0769 - val_loss: 0.0774 - val_binary_crossentropy: 0.0774\n",
            "Epoch 47/50\n",
            "157/157 [==============================] - 14s 91ms/step - loss: 0.0767 - binary_crossentropy: 0.0767 - val_loss: 0.0779 - val_binary_crossentropy: 0.0779\n",
            "Epoch 48/50\n",
            "157/157 [==============================] - 14s 91ms/step - loss: 0.0764 - binary_crossentropy: 0.0764 - val_loss: 0.0771 - val_binary_crossentropy: 0.0771\n",
            "Epoch 49/50\n",
            "157/157 [==============================] - 14s 91ms/step - loss: 0.0761 - binary_crossentropy: 0.0761 - val_loss: 0.0768 - val_binary_crossentropy: 0.0768\n",
            "Epoch 50/50\n",
            "157/157 [==============================] - 14s 91ms/step - loss: 0.0760 - binary_crossentropy: 0.0760 - val_loss: 0.0759 - val_binary_crossentropy: 0.0759\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "<tensorflow.python.keras.callbacks.History at 0x7f6db4319190>"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 7
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "zylENj4A7DDa"
      },
      "source": [
        "grid_row = 5\n",
        "grid_col = 5\n",
        "\n",
        "batch = grid_row * grid_col\n",
        "h = w = 28\n",
        "images = np.ones((batch, w, h, 1), dtype=np.float32)\n",
        "\n",
        "for row in range(h):\n",
        "\n",
        "    for col in range(w):\n",
        "\n",
        "        prob = pixelCNN.predict(images)[:,row,col,0]\n",
        "\n",
        "        pixel_samples = tf.random.categorical(tf.math.log(np.stack([1-prob, prob],1)), 1)\n",
        "\n",
        "        images[:,row,col,0] = tf.reshape(pixel_samples,[batch])"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 423
        },
        "id": "bc-OMP1g-YDq",
        "outputId": "92bfaa4a-83dc-4837-f42e-9ab3ee7345d6"
      },
      "source": [
        "f, axarr = plt.subplots(grid_row, grid_col, figsize=(grid_col*1.1,grid_row))\n",
        "\n",
        "i = 0\n",
        "for row in range(grid_row):\n",
        "    for col in range(grid_col):\n",
        "        axarr[row,col].imshow(images[i,:,:,0], cmap='gray')\n",
        "        axarr[row,col].axis('off')\n",
        "        i += 1\n",
        "f.tight_layout(0.1, h_pad=0.2, w_pad=0.1)        \n",
        "plt.show()"
      ],
      "execution_count": 16,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "/usr/local/lib/python3.7/dist-packages/ipykernel_launcher.py:9: MatplotlibDeprecationWarning: The 'renderer' parameter of tight_layout() is deprecated since Matplotlib 3.2 and will be removed two minor releases later.  If any parameter follows 'renderer', they should be pass as keyword, not positionally.\n",
            "  if __name__ == '__main__':\n"
          ],
          "name": "stderr"
        },
        {
          "output_type": "display_data",
          "data": {
            "image/png": "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\n",
            "text/plain": [
              "<Figure size 396x360 with 25 Axes>"
            ]
          },
          "metadata": {
            "tags": [],
            "needs_background": "light"
          }
        }
      ]
    }
  ]
}