13. Generative Adverserial Networks.ipynb

{
  "nbformat": 4,
  "nbformat_minor": 0,
  "metadata": {
    "colab": {
      "name": "Generative Adverserial Networks.ipynb",
      "version": "0.3.2",
      "views": {},
      "default_view": {},
      "provenance": []
    },
    "kernelspec": {
      "name": "python3",
      "display_name": "Python 3"
    },
    "accelerator": "GPU"
  },
  "cells": [
    {
      "metadata": {
        "id": "o7tBG-FmHnOG",
        "colab_type": "code",
        "colab": {
          "autoexec": {
            "startup": false,
            "wait_interval": 0
          }
        }
      },
      "cell_type": "code",
      "source": [
        "import tensorflow as tf\n",
        "import matplotlib.pyplot as plt\n",
        "import numpy as np\n",
        "%matplotlib inline\n",
        "\n",
        "from tensorflow.examples.tutorials.mnist import input_data"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "metadata": {
        "id": "f7faCe1YHtXg",
        "colab_type": "code",
        "colab": {
          "autoexec": {
            "startup": false,
            "wait_interval": 0
          },
          "output_extras": [
            {
              "item_id": 2
            }
          ],
          "base_uri": "https://localhost:8080/",
          "height": 156
        },
        "outputId": "3b1ea4ad-f3af-4bb0-f747-abef71c945e1",
        "executionInfo": {
          "status": "ok",
          "timestamp": 1522566447968,
          "user_tz": -330,
          "elapsed": 2760,
          "user": {
            "displayName": "Mayur Kulkarni",
            "photoUrl": "https://lh3.googleusercontent.com/a/default-user=s128",
            "userId": "116301151646646358552"
          }
        }
      },
      "cell_type": "code",
      "source": [
        "mnist = input_data.read_data_sets(\".\", one_hot=True)"
      ],
      "execution_count": 2,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Successfully downloaded train-images-idx3-ubyte.gz 9912422 bytes.\n",
            "Extracting ./train-images-idx3-ubyte.gz\n",
            "Successfully downloaded train-labels-idx1-ubyte.gz 28881 bytes.\n",
            "Extracting ./train-labels-idx1-ubyte.gz\n",
            "Successfully downloaded t10k-images-idx3-ubyte.gz 1648877 bytes.\n",
            "Extracting ./t10k-images-idx3-ubyte.gz\n",
            "Successfully downloaded t10k-labels-idx1-ubyte.gz 4542 bytes.\n",
            "Extracting ./t10k-labels-idx1-ubyte.gz\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "metadata": {
        "id": "42pK6mhjIXgH",
        "colab_type": "code",
        "colab": {
          "autoexec": {
            "startup": false,
            "wait_interval": 0
          },
          "output_extras": [
            {
              "item_id": 1
            },
            {
              "item_id": 2
            }
          ],
          "base_uri": "https://localhost:8080/",
          "height": 282
        },
        "outputId": "1532ef0a-f779-4d5e-d69a-9e2adbc35814",
        "executionInfo": {
          "status": "ok",
          "timestamp": 1522566449538,
          "user_tz": -330,
          "elapsed": 1026,
          "user": {
            "displayName": "Mayur Kulkarni",
            "photoUrl": "https://lh3.googleusercontent.com/a/default-user=s128",
            "userId": "116301151646646358552"
          }
        }
      },
      "cell_type": "code",
      "source": [
        "plt.figure().gca().grid(False)\n",
        "plt.imshow(mnist.train.images[12].reshape(28, 28), cmap=\"Greys\")"
      ],
      "execution_count": 3,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "<matplotlib.image.AxesImage at 0x7f8e3c476668>"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 3
        },
        {
          "output_type": "display_data",
          "data": {
            "image/png": "iVBORw0KGgoAAAANSUhEUgAAAPoAAAD4CAYAAADFJPs2AAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nSwUYulTA/wDE65rwG3Bf/AAAAABJRU5ErkJggg==\n",
            "text/plain": [
              "<matplotlib.figure.Figure at 0x7f8e3c476390>"
            ]
          },
          "metadata": {
            "tags": []
          }
        }
      ]
    },
    {
      "metadata": {
        "id": "ZZ8XEIiKIckH",
        "colab_type": "code",
        "colab": {
          "autoexec": {
            "startup": false,
            "wait_interval": 0
          }
        }
      },
      "cell_type": "code",
      "source": [
        "def discriminator(images, reuse=False):\n",
        "    if reuse:\n",
        "        tf.get_variable_scope().reuse_variables()\n",
        "\n",
        "    ####################### Convolutional Layer 1 ##############################\n",
        "    d_w1 = tf.get_variable(\"d_w1\", [5, 5, 1, 32],\n",
        "                       initializer=tf.truncated_normal_initializer(stddev=0.02))\n",
        "    d_b1 = tf.get_variable(\"d_b1\", [32],\n",
        "                           initializer=tf.constant_initializer(0))\n",
        "    d1 = tf.nn.conv2d(input=images, filter=d_w1, strides=[1, 1, 1, 1],\n",
        "                      padding=\"SAME\")\n",
        "    d1 = d1 + d_b1\n",
        "    d1 = tf.nn.relu(d1)\n",
        "    d1 = tf.nn.avg_pool(d1, ksize=[1, 2, 2, 1], strides=[1, 2, 2, 1],\n",
        "                        padding=\"SAME\")\n",
        "\n",
        "    ####################### Convolutional Layer 2 ##############################\n",
        "    d_w2 = tf.get_variable(\"d_w2\", [5, 5, 32, 64],\n",
        "                       initializer=tf.truncated_normal_initializer(stddev=0.02))\n",
        "    d_b2 = tf.get_variable(\"d_b2\", [64],\n",
        "                           initializer=tf.constant_initializer(0))\n",
        "    d2 = tf.nn.conv2d(input=d1, filter=d_w2, strides=[1, 1, 1, 1],\n",
        "                      padding=\"SAME\")\n",
        "    d2 = d2 + d_b2\n",
        "    d2 = tf.nn.relu(d2)\n",
        "    d2 = tf.nn.avg_pool(d2, ksize=[1, 2, 2, 1], strides=[1, 2, 2, 1],\n",
        "                        padding=\"SAME\")\n",
        "    \n",
        "    ########################### Dense Layer 1 ##################################\n",
        "    d_w3 = tf.get_variable(\"d_w3\", [7 * 7 * 64, 1024],\n",
        "                       initializer=tf.truncated_normal_initializer(stddev=0.02))\n",
        "    d_b3 = tf.get_variable(\"d_b3\", [1024],\n",
        "                            initializer=tf.constant_initializer(0))\n",
        "    # flatten\n",
        "    d3 = tf.reshape(d2, [-1, 7 * 7 * 64])\n",
        "    d3 = tf.matmul(d3, d_w3) + d_b3\n",
        "    d3 = tf.nn.relu(d3)\n",
        "    \n",
        "    ########################### Dense Layer 2 ##################################\n",
        "    d_w4 = tf.get_variable(\"d_w4\", [1024, 1],\n",
        "                       initializer=tf.truncated_normal_initializer(stddev=0.02))\n",
        "    d_b4 = tf.get_variable(\"d_b4\", [1],\n",
        "                            initializer=tf.constant_initializer(0))\n",
        "    d4 = tf.matmul(d3, d_w4) + d_b4\n",
        "    # this is unscaled, so it's not probability or something\n",
        "    return d4"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "metadata": {
        "id": "Vr9A8w-lZXDH",
        "colab_type": "code",
        "colab": {
          "autoexec": {
            "startup": false,
            "wait_interval": 0
          }
        }
      },
      "cell_type": "code",
      "source": [
        "def generator(z, batch_size, z_dim):\n",
        "    \"\"\" Takes in random values, and generates images \"\"\"\n",
        "    ################### Dense Layer 1 + Batch Normalization ####################\n",
        "    g_w1 = tf.get_variable(\"g_w1\", [z_dim, 3136], dtype=tf.float32,\n",
        "                    initializer=tf.truncated_normal_initializer(stddev=0.02))\n",
        "    g_b1 = tf.get_variable(\"g_b1\", [3136],\n",
        "                    initializer=tf.truncated_normal_initializer(stddev=0.02))\n",
        "    g1 = tf.matmul(z, g_w1) + g_b1\n",
        "    g1 = tf.reshape(g1, [-1, 56, 56, 1])\n",
        "    g1 = tf.contrib.layers.batch_norm(g1, epsilon=1e-5, scope=\"bn1\")\n",
        "    g1 = tf.nn.relu(g1)\n",
        "\n",
        "    ############## Convolutional Layer 1  + Batch Normalization ################\n",
        "\n",
        "    g_w2 = tf.get_variable(\"g_w2\", [3, 3, 1, z_dim/2], dtype=tf.float32,\n",
        "                    initializer=tf.truncated_normal_initializer(stddev=0.02))\n",
        "    g_b2 = tf.get_variable(\"g_b2\", [z_dim/2],\n",
        "                    initializer=tf.truncated_normal_initializer(stddev=0.02))\n",
        "    g2 = tf.nn.conv2d(g1, g_w2, strides=[1, 2, 2, 1], padding=\"SAME\")\n",
        "    g2 = g2 + g_b2\n",
        "    g2 = tf.contrib.layers.batch_norm(g2, epsilon=1e-5, scope=\"bn2\")\n",
        "    g2 = tf.nn.relu(g2)\n",
        "    g2 = tf.image.resize_images(g2, [56, 56])\n",
        "\n",
        "    ############## Convolutional Layer 2  + Batch Normalization ################\n",
        "    g_w3 = tf.get_variable(\"g_w3\", [3, 3, z_dim/2, z_dim/4], dtype=tf.float32,\n",
        "                    initializer=tf.truncated_normal_initializer(stddev=0.02))\n",
        "    g_b3 = tf.get_variable(\"g_b3\", [z_dim/4],\n",
        "                    initializer=tf.truncated_normal_initializer(stddev=0.02))\n",
        "    g3 = tf.nn.conv2d(g2, g_w3, strides=[1, 2, 2, 1], padding=\"SAME\")\n",
        "    g3 = g3 + g_b3\n",
        "    g3 = tf.contrib.layers.batch_norm(g3, epsilon=1e-5, scope=\"bn3\")\n",
        "    g3 = tf.nn.relu(g3)\n",
        "    g3 = tf.image.resize_images(g3, [56, 56])\n",
        "\n",
        "    ############## Convolutional Layer 3  + Batch Normalization ################\n",
        "    g_w4 = tf.get_variable(\"g_w4\", [1, 1, z_dim/4, 1], dtype=tf.float32,\n",
        "                    initializer=tf.truncated_normal_initializer(stddev=0.02))\n",
        "    g_b4 = tf.get_variable(\"g_b4\", [1],\n",
        "                    initializer=tf.truncated_normal_initializer(stddev=0.02))\n",
        "    g4 = tf.nn.conv2d(g3, g_w4, strides=[1, 2, 2, 1], padding=\"SAME\")\n",
        "    g4 = g4 + g_b4\n",
        "    # output dimensions: batch_size x 28 x 28 x 1\n",
        "    # the reason we're taking sigmoid is to have crispier image\n",
        "    g4 = tf.nn.sigmoid(g4)\n",
        "    return g4"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "metadata": {
        "id": "tgQBG8WrZac9",
        "colab_type": "code",
        "colab": {
          "autoexec": {
            "startup": false,
            "wait_interval": 0
          },
          "output_extras": [
            {
              "item_id": 1
            }
          ],
          "base_uri": "https://localhost:8080/",
          "height": 265
        },
        "outputId": "b966df9a-e26a-468c-e4d5-fdd83208dae7",
        "executionInfo": {
          "status": "ok",
          "timestamp": 1522566458008,
          "user_tz": -330,
          "elapsed": 2846,
          "user": {
            "displayName": "Mayur Kulkarni",
            "photoUrl": "https://lh3.googleusercontent.com/a/default-user=s128",
            "userId": "116301151646646358552"
          }
        }
      },
      "cell_type": "code",
      "source": [
        "def print_random_image():\n",
        "    z_dimensions = 100\n",
        "    z_placeholder = tf.placeholder(tf.float32, shape=[None, z_dimensions])\n",
        "    noisy_image = generator(z_placeholder, 1, z_dimensions)\n",
        "    z_batch = np.random.normal(0, 1, [1, z_dimensions])\n",
        "    with tf.Session() as sess:\n",
        "        sess.run(tf.global_variables_initializer())\n",
        "        generated_image = sess.run(noisy_image, feed_dict={z_placeholder: z_batch})\n",
        "        generated_image = generated_image.reshape([28, 28])\n",
        "        plt.imshow(generated_image, cmap=\"Greys\")\n",
        "print_random_image()"
      ],
      "execution_count": 6,
      "outputs": [
        {
          "output_type": "display_data",
          "data": {
            "image/png": "iVBORw0KGgoAAAANSUhEUgAAAPoAAAD4CAYAAADFJPs2AAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nprtELfDABAAAAABJRU5ErkJggg==\n",
            "text/plain": [
              "<matplotlib.figure.Figure at 0x7f8e3bf93dd8>"
            ]
          },
          "metadata": {
            "tags": []
          }
        }
      ]
    },
    {
      "metadata": {
        "id": "kZGs0Id_akJ7",
        "colab_type": "code",
        "colab": {
          "autoexec": {
            "startup": false,
            "wait_interval": 0
          }
        }
      },
      "cell_type": "code",
      "source": [
        "tf.reset_default_graph()"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "metadata": {
        "id": "5RKgG_ACesN6",
        "colab_type": "code",
        "colab": {
          "autoexec": {
            "startup": false,
            "wait_interval": 0
          }
        }
      },
      "cell_type": "code",
      "source": [
        "# Setup Training\n",
        "z_dimensions = 100\n",
        "batch_size = 50\n",
        "# z_placeholder is for feeding in the noise to the GAN\n",
        "z_placeholder = tf.placeholder(tf.float32, shape=[None, z_dimensions],\n",
        "                               name=\"z_placeholder\")\n",
        "# x_placeholder will feed any image to the disciminator. It can be\n",
        "# from MNIST database, or from our generator function.\n",
        "x_placeholder = tf.placeholder(tf.float32, shape=[None, 28, 28, 1],\n",
        "                               name=\"x_placeholder\")\n",
        "# Gz will hold the newly generated images from generator\n",
        "Gz = generator(z_placeholder, batch_size, z_dimensions)\n",
        "# Dx will hold the probabilities of the actual MNIST images\n",
        "\n",
        "# Dg will hold the probabilities of the new generated images\n",
        "Dg = discriminator(Gz)\n",
        "Dx = discriminator(x_placeholder, reuse=True)"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "metadata": {
        "id": "aFza9yaNlf1K",
        "colab_type": "code",
        "colab": {
          "autoexec": {
            "startup": false,
            "wait_interval": 0
          }
        }
      },
      "cell_type": "code",
      "source": [
        "# Discriminator Loss\n",
        "d_loss_real = tf.reduce_mean(tf.nn\n",
        "                    .sigmoid_cross_entropy_with_logits(logits=Dx, \n",
        "                                                       labels=tf.ones_like(Dx)))\n",
        "d_loss_fake = tf.reduce_mean(tf.nn\n",
        "                    .sigmoid_cross_entropy_with_logits(logits=Dg,\n",
        "                                                       labels=tf.zeros_like(Dg)))"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "metadata": {
        "id": "E_iqMzQ-pYWS",
        "colab_type": "code",
        "colab": {
          "autoexec": {
            "startup": false,
            "wait_interval": 0
          }
        }
      },
      "cell_type": "code",
      "source": [
        "# Generator Loss\n",
        "g_loss = tf.reduce_mean(tf.nn\n",
        "                     .sigmoid_cross_entropy_with_logits(logits=Dg,\n",
        "                                                       labels=tf.ones_like(Dg)))"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "metadata": {
        "id": "pw99rDvBrOfI",
        "colab_type": "code",
        "colab": {
          "autoexec": {
            "startup": false,
            "wait_interval": 0
          },
          "output_extras": [
            {
              "item_id": 1
            }
          ],
          "base_uri": "https://localhost:8080/",
          "height": 52
        },
        "outputId": "61f27f9b-84eb-4891-f683-1dbb393bae21",
        "executionInfo": {
          "status": "ok",
          "timestamp": 1522566466494,
          "user_tz": -330,
          "elapsed": 1108,
          "user": {
            "displayName": "Mayur Kulkarni",
            "photoUrl": "https://lh3.googleusercontent.com/a/default-user=s128",
            "userId": "116301151646646358552"
          }
        }
      },
      "cell_type": "code",
      "source": [
        "# While training Generator, we want to keep the Discriminator variables\n",
        "# constant. And same goes for training Discriminator\n",
        "\n",
        "tvars = tf.trainable_variables()\n",
        "d_vars = [var for var in tvars if \"d_\" in var.name]\n",
        "g_vars = [var for var in tvars if \"g_\" in var.name]\n",
        "\n",
        "print(\"Discriminator vars: {}\".format([var.name for var in d_vars]))\n",
        "print(\"Generator vars: {}\".format([var.name for var in g_vars]))"
      ],
      "execution_count": 11,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Discriminator vars: ['d_w1:0', 'd_b1:0', 'd_w2:0', 'd_b2:0', 'd_w3:0', 'd_b3:0', 'd_w4:0', 'd_b4:0']\n",
            "Generator vars: ['g_w1:0', 'g_b1:0', 'g_w2:0', 'g_b2:0', 'g_w3:0', 'g_b3:0', 'g_w4:0', 'g_b4:0']\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "metadata": {
        "id": "vdxCXQ82sdKb",
        "colab_type": "code",
        "colab": {
          "autoexec": {
            "startup": false,
            "wait_interval": 0
          }
        }
      },
      "cell_type": "code",
      "source": [
        "# Train the Discriminator\n",
        "with tf.variable_scope(tf.get_variable_scope(),reuse=tf.AUTO_REUSE): \n",
        "    d_trainer_real = tf.train.AdamOptimizer(learning_rate=0.0003) \\\n",
        "                             .minimize(d_loss_real, var_list=d_vars)\n",
        "\n",
        "    d_trainer_fake = tf.train.AdamOptimizer(learning_rate=0.0003) \\\n",
        "                             .minimize(d_loss_fake, var_list=d_vars)\n",
        "\n",
        "    # Train the Generator\n",
        "    g_trainer = tf.train.AdamOptimizer(learning_rate=0.0001) \\\n",
        "                        .minimize(g_loss, var_list=g_vars)"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "metadata": {
        "id": "QJE6XwfMtSyX",
        "colab_type": "code",
        "colab": {
          "autoexec": {
            "startup": false,
            "wait_interval": 0
          }
        }
      },
      "cell_type": "code",
      "source": [
        "saver = tf.train.Saver()"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "metadata": {
        "id": "Rt_HRi-Rl4zF",
        "colab_type": "code",
        "colab": {
          "autoexec": {
            "startup": false,
            "wait_interval": 0
          },
          "output_extras": [
            {
              "item_id": 20
            }
          ],
          "base_uri": "https://localhost:8080/",
          "height": 364
        },
        "outputId": "2037fca6-e8d0-401d-cdfc-be0015a6abe7",
        "executionInfo": {
          "status": "ok",
          "timestamp": 1522566536428,
          "user_tz": -330,
          "elapsed": 24828,
          "user": {
            "displayName": "Mayur Kulkarni",
            "photoUrl": "https://lh3.googleusercontent.com/a/default-user=s128",
            "userId": "116301151646646358552"
          }
        }
      },
      "cell_type": "code",
      "source": [
        "init = tf.global_variables_initializer()\n",
        "with tf.Session() as sess:\n",
        "    # Train Discriminator first\n",
        "    sess.run(init)\n",
        "#     saver.restore(sess, \"/content/gan/model\")\n",
        "    for i in range(1000):\n",
        "        z_batch = np.random.normal(0, 1, size=[batch_size, z_dimensions])\n",
        "        real_image_batch = mnist.train.next_batch(batch_size)[0] \\\n",
        "                                      .reshape([batch_size, 28, 28, 1])\n",
        "        _, _, d_loss_real_val, d_loss_fake_val = \\\n",
        "                sess.run([d_trainer_real, d_trainer_fake, d_loss_real, \n",
        "                          d_loss_fake],\n",
        "                         feed_dict={x_placeholder: real_image_batch,\n",
        "                                    z_placeholder: z_batch})\n",
        "        if i % 50 == 0:\n",
        "            print(\"Iteration: {0} \\tLoss Real: {1:.5f}\\tLoss Fake: {2:.5f}\"\n",
        "                       .format(i, d_loss_real_val, d_loss_fake_val))\n",
        "    saver.save(sess, \"/content/gan/model\")"
      ],
      "execution_count": 17,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Iteration: 0 \tLoss Real: 0.69308\tLoss Fake: 0.70976\n",
            "Iteration: 50 \tLoss Real: 0.30527\tLoss Fake: 0.77538\n",
            "Iteration: 100 \tLoss Real: 0.03286\tLoss Fake: 0.08415\n",
            "Iteration: 150 \tLoss Real: 0.00802\tLoss Fake: 0.02482\n",
            "Iteration: 200 \tLoss Real: 0.00065\tLoss Fake: 0.00038\n",
            "Iteration: 250 \tLoss Real: 0.00034\tLoss Fake: 0.00027\n",
            "Iteration: 300 \tLoss Real: 0.00187\tLoss Fake: 0.00017\n",
            "Iteration: 350 \tLoss Real: 0.00008\tLoss Fake: 0.00012\n",
            "Iteration: 400 \tLoss Real: 0.00009\tLoss Fake: 0.00011\n",
            "Iteration: 450 \tLoss Real: 0.00043\tLoss Fake: 0.00011\n",
            "Iteration: 500 \tLoss Real: 0.00009\tLoss Fake: 0.00007\n",
            "Iteration: 550 \tLoss Real: 0.00003\tLoss Fake: 0.00007\n",
            "Iteration: 600 \tLoss Real: 0.00009\tLoss Fake: 0.00006\n",
            "Iteration: 650 \tLoss Real: 0.00002\tLoss Fake: 0.00004\n",
            "Iteration: 700 \tLoss Real: 0.00005\tLoss Fake: 0.00003\n",
            "Iteration: 750 \tLoss Real: 0.00005\tLoss Fake: 0.00003\n",
            "Iteration: 800 \tLoss Real: 0.00008\tLoss Fake: 0.00004\n",
            "Iteration: 850 \tLoss Real: 0.00001\tLoss Fake: 0.00003\n",
            "Iteration: 900 \tLoss Real: 0.00001\tLoss Fake: 0.00002\n",
            "Iteration: 950 \tLoss Real: 0.00002\tLoss Fake: 0.00004\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "metadata": {
        "id": "SBct9YYEpLEK",
        "colab_type": "code",
        "colab": {
          "autoexec": {
            "startup": false,
            "wait_interval": 0
          },
          "output_extras": [
            {
              "item_id": 53
            },
            {
              "item_id": 101
            }
          ],
          "base_uri": "https://localhost:8080/",
          "height": 1771
        },
        "outputId": "b57eda6b-6695-4ebc-83f4-39dc98f326f8",
        "executionInfo": {
          "status": "ok",
          "timestamp": 1522573216932,
          "user_tz": -330,
          "elapsed": 5307638,
          "user": {
            "displayName": "Mayur Kulkarni",
            "photoUrl": "https://lh3.googleusercontent.com/a/default-user=s128",
            "userId": "116301151646646358552"
          }
        }
      },
      "cell_type": "code",
      "source": [
        "# Train generator and discriminator at the same\n",
        "with tf.Session() as sess:\n",
        "    steps = 100000\n",
        "    saver.restore(sess, \"/content/gan/model\")\n",
        "    for i in range(steps):\n",
        "        real_image_batch = mnist.train.next_batch(batch_size)[0] \\\n",
        "                                      .reshape([-1, 28, 28, 1])\n",
        "        # Train Discriminator\n",
        "        z_batch = np.random.normal(0, 1, size=[batch_size, z_dimensions])\n",
        "        _, _, d_loss_real_val, d_loss_fake_val = \\\n",
        "            sess.run(\n",
        "                [d_trainer_real, d_trainer_fake, d_loss_real, d_loss_fake],\n",
        "                feed_dict={x_placeholder: real_image_batch, \n",
        "                           z_placeholder: z_batch}\n",
        "            )\n",
        "        # Train Generator\n",
        "        z_batch = np.random.normal(0, 1, size=[batch_size, z_dimensions])\n",
        "        _, g_loss_val = sess.run([g_trainer, g_loss], \n",
        "                                 feed_dict={z_placeholder: z_batch})\n",
        "\n",
        "        # Debug info\n",
        "        if i % 1000 == 0:\n",
        "            print(\"Iteration: {0}\\tDLoss Real: {1:.4f}\\tDLoss Fake: {2:.4f}\"\n",
        "                  \"\\tGLoss: {3:.4f}\".format(i, d_loss_real_val, d_loss_fake_val,\n",
        "                                           g_loss_val))\n",
        "    saver.save(sess, \"/content/gan/model\")\n",
        "    "
      ],
      "execution_count": 27,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "INFO:tensorflow:Restoring parameters from /content/gan/model\n",
            "Iteration: 0\tDLoss Real: 0.0274\tDLoss Fake: 3.5510\tGLoss: 6.4086\n",
            "Iteration: 1000\tDLoss Real: 0.3001\tDLoss Fake: 4.8100\tGLoss: 1.3463\n",
            "Iteration: 2000\tDLoss Real: 0.0001\tDLoss Fake: 11.8316\tGLoss: 0.2900\n",
            "Iteration: 3000\tDLoss Real: 7.5022\tDLoss Fake: 0.0012\tGLoss: 2.1869\n",
            "Iteration: 4000\tDLoss Real: 1.7415\tDLoss Fake: 1.0484\tGLoss: 2.3957\n",
            "Iteration: 5000\tDLoss Real: 4.0374\tDLoss Fake: 0.6165\tGLoss: 1.9592\n",
            "Iteration: 6000\tDLoss Real: 0.2387\tDLoss Fake: 2.6595\tGLoss: 6.1472\n",
            "Iteration: 7000\tDLoss Real: 1.0244\tDLoss Fake: 7.2106\tGLoss: 2.8867\n",
            "Iteration: 8000\tDLoss Real: 0.6583\tDLoss Fake: 8.4785\tGLoss: 2.7842\n",
            "Iteration: 9000\tDLoss Real: 0.6970\tDLoss Fake: 6.2023\tGLoss: 6.7081\n",
            "Iteration: 10000\tDLoss Real: 3.8503\tDLoss Fake: 3.5546\tGLoss: 0.1147\n",
            "Iteration: 11000\tDLoss Real: 3.2922\tDLoss Fake: 4.9207\tGLoss: 0.4871\n",
            "Iteration: 12000\tDLoss Real: 4.9237\tDLoss Fake: 4.3103\tGLoss: 2.2280\n",
            "Iteration: 13000\tDLoss Real: 23.5717\tDLoss Fake: 0.0315\tGLoss: 3.8023\n",
            "Iteration: 14000\tDLoss Real: 2.0862\tDLoss Fake: 12.6701\tGLoss: 5.0122\n",
            "Iteration: 15000\tDLoss Real: 15.0659\tDLoss Fake: 3.7373\tGLoss: 1.9332\n",
            "Iteration: 16000\tDLoss Real: 9.8914\tDLoss Fake: 0.7287\tGLoss: 0.8998\n",
            "Iteration: 17000\tDLoss Real: 4.3618\tDLoss Fake: 9.1421\tGLoss: 4.6114\n",
            "Iteration: 18000\tDLoss Real: 4.2844\tDLoss Fake: 4.9853\tGLoss: 1.0283\n",
            "Iteration: 19000\tDLoss Real: 5.5525\tDLoss Fake: 2.4970\tGLoss: 5.1407\n",
            "Iteration: 20000\tDLoss Real: 16.2098\tDLoss Fake: 0.3511\tGLoss: 0.8197\n",
            "Iteration: 21000\tDLoss Real: 5.4837\tDLoss Fake: 1.0383\tGLoss: 11.8822\n",
            "Iteration: 22000\tDLoss Real: 9.5747\tDLoss Fake: 2.8820\tGLoss: 0.7144\n",
            "Iteration: 23000\tDLoss Real: 7.0495\tDLoss Fake: 2.7621\tGLoss: 1.0811\n",
            "Iteration: 24000\tDLoss Real: 6.7989\tDLoss Fake: 3.1962\tGLoss: 0.7361\n",
            "Iteration: 25000\tDLoss Real: 2.9073\tDLoss Fake: 10.8696\tGLoss: 4.8520\n",
            "Iteration: 26000\tDLoss Real: 9.8931\tDLoss Fake: 6.6837\tGLoss: 4.4423\n",
            "Iteration: 27000\tDLoss Real: 6.0371\tDLoss Fake: 7.8184\tGLoss: 4.2022\n",
            "Iteration: 28000\tDLoss Real: 15.9837\tDLoss Fake: 4.4098\tGLoss: 1.9826\n",
            "Iteration: 29000\tDLoss Real: 1.4118\tDLoss Fake: 22.2941\tGLoss: 0.9832\n",
            "Iteration: 30000\tDLoss Real: 11.9329\tDLoss Fake: 5.1703\tGLoss: 0.9131\n",
            "Iteration: 31000\tDLoss Real: 7.3483\tDLoss Fake: 11.7756\tGLoss: 11.5158\n",
            "Iteration: 32000\tDLoss Real: 16.6020\tDLoss Fake: 4.6569\tGLoss: 7.8097\n",
            "Iteration: 33000\tDLoss Real: 2.7698\tDLoss Fake: 21.4481\tGLoss: 3.0893\n",
            "Iteration: 34000\tDLoss Real: 2.0851\tDLoss Fake: 22.9645\tGLoss: 2.3089\n",
            "Iteration: 35000\tDLoss Real: 7.3659\tDLoss Fake: 13.8369\tGLoss: 0.6552\n",
            "Iteration: 36000\tDLoss Real: 10.3683\tDLoss Fake: 7.6859\tGLoss: 4.6692\n",
            "Iteration: 37000\tDLoss Real: 14.8475\tDLoss Fake: 6.0724\tGLoss: 3.9881\n",
            "Iteration: 38000\tDLoss Real: 4.8811\tDLoss Fake: 11.7319\tGLoss: 4.6849\n",
            "Iteration: 39000\tDLoss Real: 13.3628\tDLoss Fake: 3.5837\tGLoss: 6.8129\n",
            "Iteration: 40000\tDLoss Real: 16.2911\tDLoss Fake: 8.2786\tGLoss: 0.5563\n",
            "Iteration: 41000\tDLoss Real: 5.6291\tDLoss Fake: 16.3329\tGLoss: 6.1328\n",
            "Iteration: 42000\tDLoss Real: 14.7407\tDLoss Fake: 4.6149\tGLoss: 9.5094\n",
            "Iteration: 43000\tDLoss Real: 4.4418\tDLoss Fake: 14.1768\tGLoss: 4.0872\n",
            "Iteration: 44000\tDLoss Real: 9.1299\tDLoss Fake: 7.7418\tGLoss: 5.1835\n",
            "Iteration: 45000\tDLoss Real: 21.2754\tDLoss Fake: 4.7607\tGLoss: 3.6745\n",
            "Iteration: 46000\tDLoss Real: 10.9355\tDLoss Fake: 12.4470\tGLoss: 9.8953\n",
            "Iteration: 47000\tDLoss Real: 6.9167\tDLoss Fake: 19.8065\tGLoss: 5.0337\n",
            "Iteration: 48000\tDLoss Real: 9.4195\tDLoss Fake: 17.5033\tGLoss: 1.0260\n",
            "Iteration: 49000\tDLoss Real: 16.1238\tDLoss Fake: 9.9893\tGLoss: 6.4204\n",
            "Iteration: 50000\tDLoss Real: 17.1595\tDLoss Fake: 7.1201\tGLoss: 1.5500\n",
            "Iteration: 51000\tDLoss Real: 8.7693\tDLoss Fake: 29.7105\tGLoss: 0.0000\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "stream",
          "text": [
            "Iteration: 52000\tDLoss Real: 21.1157\tDLoss Fake: 5.5879\tGLoss: 1.3915\n",
            "Iteration: 53000\tDLoss Real: 7.5539\tDLoss Fake: 20.7839\tGLoss: 2.6277\n",
            "Iteration: 54000\tDLoss Real: 4.9874\tDLoss Fake: 34.1384\tGLoss: 1.1870\n",
            "Iteration: 55000\tDLoss Real: 3.4666\tDLoss Fake: 34.5087\tGLoss: 8.3575\n",
            "Iteration: 56000\tDLoss Real: 25.1974\tDLoss Fake: 12.5859\tGLoss: 5.3906\n",
            "Iteration: 57000\tDLoss Real: 27.1813\tDLoss Fake: 14.4489\tGLoss: 4.2580\n",
            "Iteration: 58000\tDLoss Real: 17.6728\tDLoss Fake: 19.9264\tGLoss: 9.2541\n",
            "Iteration: 59000\tDLoss Real: 19.6593\tDLoss Fake: 17.6911\tGLoss: 2.5688\n",
            "Iteration: 60000\tDLoss Real: 26.1740\tDLoss Fake: 9.4944\tGLoss: 9.2890\n",
            "Iteration: 61000\tDLoss Real: 20.0087\tDLoss Fake: 14.4073\tGLoss: 3.5832\n",
            "Iteration: 62000\tDLoss Real: 18.1446\tDLoss Fake: 17.8708\tGLoss: 1.4620\n",
            "Iteration: 63000\tDLoss Real: 18.1176\tDLoss Fake: 20.5427\tGLoss: 3.9803\n",
            "Iteration: 64000\tDLoss Real: 13.6793\tDLoss Fake: 33.1709\tGLoss: 1.0754\n",
            "Iteration: 65000\tDLoss Real: 15.2692\tDLoss Fake: 23.2309\tGLoss: 1.2370\n",
            "Iteration: 66000\tDLoss Real: 15.0713\tDLoss Fake: 26.7671\tGLoss: 2.1792\n",
            "Iteration: 67000\tDLoss Real: 27.5811\tDLoss Fake: 11.8021\tGLoss: 20.5527\n",
            "Iteration: 68000\tDLoss Real: 35.0068\tDLoss Fake: 7.6611\tGLoss: 7.9373\n",
            "Iteration: 69000\tDLoss Real: 5.4202\tDLoss Fake: 49.1444\tGLoss: 8.3642\n",
            "Iteration: 70000\tDLoss Real: 23.3354\tDLoss Fake: 15.3683\tGLoss: 6.0408\n",
            "Iteration: 71000\tDLoss Real: 12.4735\tDLoss Fake: 39.4065\tGLoss: 0.7858\n",
            "Iteration: 72000\tDLoss Real: 22.7163\tDLoss Fake: 13.0334\tGLoss: 4.8138\n",
            "Iteration: 73000\tDLoss Real: 22.2810\tDLoss Fake: 20.2573\tGLoss: 4.0959\n",
            "Iteration: 74000\tDLoss Real: 18.8588\tDLoss Fake: 32.3886\tGLoss: 8.8109\n",
            "Iteration: 75000\tDLoss Real: 20.8745\tDLoss Fake: 18.4695\tGLoss: 9.7096\n",
            "Iteration: 76000\tDLoss Real: 7.0409\tDLoss Fake: 38.8243\tGLoss: 4.3664\n",
            "Iteration: 77000\tDLoss Real: 25.7365\tDLoss Fake: 17.1081\tGLoss: 15.6748\n",
            "Iteration: 78000\tDLoss Real: 5.6504\tDLoss Fake: 50.9220\tGLoss: 7.2119\n",
            "Iteration: 79000\tDLoss Real: 14.0125\tDLoss Fake: 23.8258\tGLoss: 7.1845\n",
            "Iteration: 80000\tDLoss Real: 12.7637\tDLoss Fake: 32.4349\tGLoss: 2.2014\n",
            "Iteration: 81000\tDLoss Real: 17.4999\tDLoss Fake: 35.2668\tGLoss: 22.3407\n",
            "Iteration: 82000\tDLoss Real: 15.3259\tDLoss Fake: 34.1550\tGLoss: 3.6180\n",
            "Iteration: 83000\tDLoss Real: 17.1610\tDLoss Fake: 29.6998\tGLoss: 1.3639\n",
            "Iteration: 84000\tDLoss Real: 44.6243\tDLoss Fake: 20.0958\tGLoss: 8.5362\n",
            "Iteration: 85000\tDLoss Real: 22.1712\tDLoss Fake: 18.6301\tGLoss: 7.8616\n",
            "Iteration: 86000\tDLoss Real: 35.6370\tDLoss Fake: 29.1167\tGLoss: 7.7407\n",
            "Iteration: 87000\tDLoss Real: 35.8135\tDLoss Fake: 27.0352\tGLoss: 1.5463\n",
            "Iteration: 88000\tDLoss Real: 12.2598\tDLoss Fake: 57.0171\tGLoss: 4.7215\n",
            "Iteration: 89000\tDLoss Real: 18.8143\tDLoss Fake: 37.2298\tGLoss: 9.8080\n",
            "Iteration: 90000\tDLoss Real: 20.2668\tDLoss Fake: 31.5371\tGLoss: 13.2856\n",
            "Iteration: 91000\tDLoss Real: 47.6391\tDLoss Fake: 3.3685\tGLoss: 17.8624\n",
            "Iteration: 92000\tDLoss Real: 27.6472\tDLoss Fake: 28.2235\tGLoss: 10.8299\n",
            "Iteration: 93000\tDLoss Real: 45.6209\tDLoss Fake: 5.9468\tGLoss: 10.0624\n",
            "Iteration: 94000\tDLoss Real: 39.6811\tDLoss Fake: 17.6792\tGLoss: 16.7481\n",
            "Iteration: 95000\tDLoss Real: 26.6152\tDLoss Fake: 23.4331\tGLoss: 19.6539\n",
            "Iteration: 96000\tDLoss Real: 24.7476\tDLoss Fake: 40.3880\tGLoss: 6.9983\n",
            "Iteration: 97000\tDLoss Real: 21.2450\tDLoss Fake: 61.2738\tGLoss: 9.4444\n",
            "Iteration: 98000\tDLoss Real: 32.9163\tDLoss Fake: 31.1589\tGLoss: 2.9836\n",
            "Iteration: 99000\tDLoss Real: 17.3552\tDLoss Fake: 32.0364\tGLoss: 11.3795\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "metadata": {
        "id": "qFbAiOhttGgJ",
        "colab_type": "code",
        "colab": {
          "autoexec": {
            "startup": false,
            "wait_interval": 0
          },
          "output_extras": [
            {
              "item_id": 1
            },
            {
              "item_id": 2
            },
            {
              "item_id": 3
            },
            {
              "item_id": 4
            },
            {
              "item_id": 5
            },
            {
              "item_id": 6
            },
            {
              "item_id": 7
            },
            {
              "item_id": 8
            },
            {
              "item_id": 9
            },
            {
              "item_id": 10
            },
            {
              "item_id": 11
            }
          ],
          "base_uri": "https://localhost:8080/",
          "height": 1454
        },
        "outputId": "3fca7777-c55d-4e40-dbc2-93bcdd61e1ea",
        "executionInfo": {
          "status": "ok",
          "timestamp": 1522573296954,
          "user_tz": -330,
          "elapsed": 2150,
          "user": {
            "displayName": "Mayur Kulkarni",
            "photoUrl": "https://lh3.googleusercontent.com/a/default-user=s128",
            "userId": "116301151646646358552"
          }
        }
      },
      "cell_type": "code",
      "source": [
        "with tf.Session() as sess:\n",
        "    saver.restore(sess, \"/content/gan/model\")\n",
        "    z_batch = np.random.normal(0, 1, size=[10, z_dimensions])\n",
        "    generated_image = generator(z_placeholder, 10, z_dimensions)\n",
        "    image = sess.run(generated_image, feed_dict={z_placeholder: z_batch})\n",
        "    images = image.reshape([-1, 28, 28])\n",
        "    for image_val in images:\n",
        "        plt.figure(figsize=(2, 2))\n",
        "        plt.imshow(image_val, cmap=\"Greys\")\n",
        "        plt.show()"
      ],
      "execution_count": 28,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "INFO:tensorflow:Restoring parameters from /content/gan/model\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "display_data",
          "data": {
            "image/png": "iVBORw0KGgoAAAANSUhEUgAAAI0AAACOCAYAAAAMyosLAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\n",
            "text/plain": [
              "<matplotlib.figure.Figure at 0x7f8dff0657b8>"
            ]
          },
          "metadata": {
            "tags": []
          }
        },
        {
          "output_type": "display_data",
          "data": {
            "image/png": "iVBORw0KGgoAAAANSUhEUgAAAI0AAACOCAYAAAAMyosLAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz\nAAALEgAACxIB0t1+/AAAADl0RVh0U29mdHdhcmUAbWF0cGxvdGxpYiB2ZXJzaW9uIDIuMS4yLCBo\ndHRwOi8vbWF0cGxvdGxpYi5vcmcvNQv5yAAACP9JREFUeJzt3VtoVdkdx/FvTOItogS12hTxQocV\nUV/0wbGgM3VsnUqNgpeCYxmslwoio2SkUl9KETpYBss4pTgoaFsQfPDSkTJoR9Q3kQEHBbNar4ip\nGjXGiZaQxPQh54S9lsk5Z53sc83vAzL7d3aSswj/WXudffmnoqenB5EQwwo9ACk9KhoJpqKRYCoa\nCaaikWAqGglWle03GmP2A28DPcBH1torsY1KilpWM40x5h3gLWvtAmAj8Fmso5Kilu3h6T3gFIC1\n9gZQa4wZm+Lre/Sv5P4NKNuimQy0RHJL4jUZAuJaCFfE9HOkBGRbNM24M0sd8N/BD0dKQbZFcxZY\nDWCMmQs0W2u/i21UUtQqsr3KbYz5BFgEvAa2WWu/TfHlupReegZccmRdNIFUNKVnwKLRGWEJpqKR\nYCoaCaaikWAqGgmW9VXuoWrHjh192w0NDc6+xYsX53s4BaGZRoKpaCSYikaCaU2Txrp165x87Nix\nvu1nz545+xYuXOjk6urq3A2sgDTTSDAVjQQr+wuWra2tTr569aqTly9fDkB7eztjxozh5cuXzv5h\nw9z/r2pqavq2t27d6uyrq6tz8ty5c508f/58J48YMSLd8AtJFywlPioaCaaikWBlt6Zpa2tz8vr1\n65185syZfr+vp6eHioqKNz4mNzU1OTm6v7m52dnnX0Z49eqVk9euXevko0ePOnnkyJH9jq1AtKaR\n+KhoJJiKRoKV3WWE0aNHO3nWrFlOvnv3rpOXLl3at93Y2MjevXud/anWGVOmTHHy5cuXnTxnzhwn\nHz9+3Mn79u1z8tSpUwd8r2KimUaCqWgkmIpGgpXdeZpC8n+Xw4cPd3J3d7eTX79+nfMxDYLO00h8\nVDQSTEUjwcruPI3v2rVrTp482W3YNXHixKx/tr+GOXnypJPnzZvnZP920FKlmUaCZTTTGGNmA6eB\n/dbaz40xU4C/AZX0dsD6pbW2I3fDlGKSdqYxxtQAB4CvIy//HviztXYhcBP4VW6GJ8Uok5mmA1gG\n/Cby2rtA8gbZL4GPgb/EOrKYLFmyxMmPHz928mDOUz158sTJmzdvdvLz58+dfOfOHSePHz/eybt3\n7856LPmUtmistV1AlzEm+nJN5HD0GPh+DsYmRSqOT09F3Q720aNHOfvZ/ievp0+f5uy9ikm2RdNu\njBllrf0f8AN6W8QWpUmTJjk5zsNTS0uLk+vr653sH54mTJjg5J07dzq5bA5PA/gXsAr4e+K/X8U2\nopi9ePEi5f6Kit6JMnmPcGdnp7O/qsr9FUWvF/nnZWpra5187tw5J/v33wzmHFEhpS0aY8w84FNg\nGtBpjFkNfAAcMcb8GrgHHB34J0i5yWQh/A29n5Z8P4l9NFISdEZYgpX9tSd/HXHz5k0np7sHxt9f\nWVnZt71nzx5n37Zt25wcfe67v+z/7OT6qthpppFgKhoJpqKRYHm5R7gn8SYVFRV950Py5dChQ072\nW57dv38fgAMHDrB9+3YuXLjg7H/48KGTV65c2bftnwO6deuWk2/cuOHkGTNmOHnLli1O9tdEfm+c\nPNM9whIfFY0EK/tHWKZNm+bke/fuOTl5COju7qaysvKNQ6f/2El0f7rfnf/xvaury8n+raeLFi1y\ncrSTaAHo8CTxUdFIMBWNBCv7NU0o/2Oxf4tm9FYJf/1z+vRpJ0+fPt3Ja9ascfL169dTjiX5kb2+\nvp6mpqY37tfJMa1pJD4qGgmmopFgWtOkMXPmTCdH7+v1H1nx1zj+73bZsmVOPnv2rJP91iPJ2zC6\nurqoqqp64zxPjmlNI/FR0UgwFY0EK/vbPQdrw4YNTo62kE13u6Z/3cp/7imdaKsSv21JIWmmkWAq\nGgmmopFgWtN4/HMlu3btcnLIrap+23v/vIy/XvK/Ptqpw+vaUVCaaSSYikaCqWgkmNY0Hv/cy6pV\nq5zstxdJxX+kxe9nc/jwYSf7bU3GjRvXt3379u2M3zfXNNNIsExbwu4DFia+/g/AFdQSdsjKpCXs\nj4HZ1toFwPvAn1BL2CEt7f00xphKYKS19mVi+zHwAqi31nYYYxYAH1trV6X4MSV7P43/6G30zxn6\n7c8OHjzoZP+5peQjwEkdHe7kPHbsWCe3trYGjTVmA56QyqQTVjfwMhE3Av8Elqol7NCV8Z17xpgV\nwG+BnwL/sdZ+L/H6D4G/Wmt/lOLbS3amGcKyn2kAjDFLgT3A+9baNmNMybSEHawhfHgaUCbdPccB\nfwSWWGuTfTpKpiXsYFVXVzv5xIkTfdt+25IjR444ub293cmjRo1ysv8nlhsbG7MdZl5lMtP8ApgA\nHI9cNPsQOKSWsENTJgvhL4Av+tmllrBDlM4ISzA995TGpUuXnNzQ0NC33dbWFvSz6urqnPzgwYPs\nB5Z7eu5J4qOikWAqGgmm+2nSOH/+vJNTrWNWrFjh5FOnTuVkTIWmmUaCqWgkmA5PafiXAlIp18OR\nTzONBFPRSDAVjQTTZYQ0Ll686GRrbd/2pk2bnH0F/qspcdNlBImPikaCqWgkmNY0MhCtaSQ+KhoJ\npqKRYCoaCaaikWAqGgmmopFg+bqfJvM+qlL0NNNIMBWNBFPRSDAVjQRT0UgwFY0Ey9sjLMaY/cDb\n9N4m8ZG19kq+3nuA8cwGTgP7rbWfG2OmUCS9kYu9b3NeZhpjzDvAW4lexBuBz/LxvinGUwMcAL6O\nvFwUvZFLoW9zvg5P7wGnAKy1N4BaY8zY1N+SUx3AMtwGk+8C/0hsfwksyfOYki4BaxLbz4Eaimds\nQP4OT5OBbyK5JfHai/6/PLestV1Al/eHt2qKoTdyKfRtLtRjucV+WaHg40v0bd5Iom9zZFfBx5av\nw1MzvTNLUh29C7pi0m6MSfZsLWhv5Ejf5p9Za9uKaWyQv6I5C6wGMMbMBZqttd/l6b0zleyNDAXs\njRzp2/zzfvo2F3RsSfl6GgFjzCfAIuA1sM1a+21e3rj/scwDPgWmAZ3AA+AD4Agwkt7eyBustZ0F\nGNsW4HfAvyMvfwgcKvTYkvJWNFI+dEZYgqloJJiKRoKpaCSYikaCqWgkmIpGgqloJNj/AQrIMZy6\nNQbtAAAAAElFTkSuQmCC\n",
            "text/plain": [
              "<matplotlib.figure.Figure at 0x7f8de847acc0>"
            ]
          },
          "metadata": {
            "tags": []
          }
        },
        {
          "output_type": "display_data",
          "data": {
            "image/png": "iVBORw0KGgoAAAANSUhEUgAAAI0AAACOCAYAAAAMyosLAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nAABJRU5ErkJggg==\n",
            "text/plain": [
              "<matplotlib.figure.Figure at 0x7f8e0b6bd390>"
            ]
          },
          "metadata": {
            "tags": []
          }
        },
        {
          "output_type": "display_data",
          "data": {
            "image/png": "iVBORw0KGgoAAAANSUhEUgAAAI0AAACOCAYAAAAMyosLAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\n",
            "text/plain": [
              "<matplotlib.figure.Figure at 0x7f8de846a668>"
            ]
          },
          "metadata": {
            "tags": []
          }
        },
        {
          "output_type": "display_data",
          "data": {
            "image/png": "iVBORw0KGgoAAAANSUhEUgAAAI0AAACOCAYAAAAMyosLAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nTEkjwZQ0EkxJI8GUNBJMSSPB/g9E8Caq+4SdygAAAABJRU5ErkJggg==\n",
            "text/plain": [
              "<matplotlib.figure.Figure at 0x7f8de82f75f8>"
            ]
          },
          "metadata": {
            "tags": []
          }
        },
        {
          "output_type": "display_data",
          "data": {
            "image/png": "iVBORw0KGgoAAAANSUhEUgAAAI0AAACOCAYAAAAMyosLAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nETOFRswUGjFTaMRMoREzhUbM/g/ZSfUX56Rn2QAAAABJRU5ErkJggg==\n",
            "text/plain": [
              "<matplotlib.figure.Figure at 0x7f8de82d0668>"
            ]
          },
          "metadata": {
            "tags": []
          }
        },
        {
          "output_type": "display_data",
          "data": {
            "image/png": "iVBORw0KGgoAAAANSUhEUgAAAI0AAACOCAYAAAAMyosLAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\n",
            "text/plain": [
              "<matplotlib.figure.Figure at 0x7f8e0ba7c4e0>"
            ]
          },
          "metadata": {
            "tags": []
          }
        },
        {
          "output_type": "display_data",
          "data": {
            "image/png": "iVBORw0KGgoAAAANSUhEUgAAAI0AAACOCAYAAAAMyosLAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\n",
            "text/plain": [
              "<matplotlib.figure.Figure at 0x7f8e0b684a90>"
            ]
          },
          "metadata": {
            "tags": []
          }
        },
        {
          "output_type": "display_data",
          "data": {
            "image/png": "iVBORw0KGgoAAAANSUhEUgAAAI0AAACOCAYAAAAMyosLAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nGJlGBCPTiGBkGhHM/wGp3SM9hJxMPwAAAABJRU5ErkJggg==\n",
            "text/plain": [
              "<matplotlib.figure.Figure at 0x7f8e0b8f55c0>"
            ]
          },
          "metadata": {
            "tags": []
          }
        },
        {
          "output_type": "display_data",
          "data": {
            "image/png": "iVBORw0KGgoAAAANSUhEUgAAAI0AAACOCAYAAAAMyosLAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nMBKNCOb/aw5eekFYq+AAAAAASUVORK5CYII=\n",
            "text/plain": [
              "<matplotlib.figure.Figure at 0x7f8e0babe710>"
            ]
          },
          "metadata": {
            "tags": []
          }
        }
      ]
    },
    {
      "metadata": {
        "id": "DNaOFnUZUM51",
        "colab_type": "code",
        "colab": {
          "autoexec": {
            "startup": false,
            "wait_interval": 0
          }
        }
      },
      "cell_type": "code",
      "source": [
        "!tar -cf gan.tar gan\n"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "metadata": {
        "id": "1Brf8VvqUVgk",
        "colab_type": "code",
        "colab": {
          "autoexec": {
            "startup": false,
            "wait_interval": 0
          },
          "output_extras": [
            {
              "item_id": 1
            }
          ],
          "base_uri": "https://localhost:8080/",
          "height": 34
        },
        "outputId": "0ef21f54-f44d-467b-feab-637d7dc0c52c",
        "executionInfo": {
          "status": "ok",
          "timestamp": 1522573316192,
          "user_tz": -330,
          "elapsed": 7008,
          "user": {
            "displayName": "Mayur Kulkarni",
            "photoUrl": "https://lh3.googleusercontent.com/a/default-user=s128",
            "userId": "116301151646646358552"
          }
        }
      },
      "cell_type": "code",
      "source": [
        "!curl --upload-file /content/gan.tar https://transfer.sh/gan.tar"
      ],
      "execution_count": 30,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "https://transfer.sh/O9qXJ/gan.tar"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "metadata": {
        "id": "n2wzQ8cAUjFY",
        "colab_type": "code",
        "colab": {
          "autoexec": {
            "startup": false,
            "wait_interval": 0
          }
        }
      },
      "cell_type": "code",
      "source": [
        "# https://transfer.sh/O9qXJ/gan.tar"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "metadata": {
        "id": "GzyTkb54U8a2",
        "colab_type": "code",
        "colab": {
          "autoexec": {
            "startup": false,
            "wait_interval": 0
          },
          "output_extras": [
            {
              "item_id": 36
            }
          ],
          "base_uri": "https://localhost:8080/",
          "height": 69
        },
        "outputId": "e452546d-8a23-4ee0-e682-b2470e225ca7",
        "executionInfo": {
          "status": "ok",
          "timestamp": 1521954720830,
          "user_tz": -330,
          "elapsed": 36131,
          "user": {
            "displayName": "Mayur Kulkarni",
            "photoUrl": "https://lh3.googleusercontent.com/a/default-user=s128",
            "userId": "116301151646646358552"
          }
        }
      },
      "cell_type": "code",
      "source": [
        "! curl --output gan.tar https://transfer.sh/Tsi6K/gan.tar\n"
      ],
      "execution_count": 34,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "  % Total    % Received % Xferd  Average Speed   Time    Time     Time  Current\r\n",
            "                                 Dload  Upload   Total   Spent    Left  Speed\n",
            "100 66.2M  100 66.2M    0     0  1996k      0  0:00:34  0:00:34 --:--:-- 4001k\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "metadata": {
        "id": "NKkQvaUPaWHI",
        "colab_type": "code",
        "colab": {
          "autoexec": {
            "startup": false,
            "wait_interval": 0
          },
          "output_extras": [
            {
              "item_id": 1
            }
          ],
          "base_uri": "https://localhost:8080/",
          "height": 34
        },
        "outputId": "dfa03268-3218-4f4f-b432-24f3be2f14e7",
        "executionInfo": {
          "status": "ok",
          "timestamp": 1521954844241,
          "user_tz": -330,
          "elapsed": 889,
          "user": {
            "displayName": "Mayur Kulkarni",
            "photoUrl": "https://lh3.googleusercontent.com/a/default-user=s128",
            "userId": "116301151646646358552"
          }
        }
      },
      "cell_type": "code",
      "source": [
        "! ls /content/gan/"
      ],
      "execution_count": 42,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "checkpoint  model.data-00000-of-00001  model.index  model.meta\r\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "metadata": {
        "id": "PFiXqxsOU9ny",
        "colab_type": "code",
        "colab": {
          "autoexec": {
            "startup": false,
            "wait_interval": 0
          }
        }
      },
      "cell_type": "code",
      "source": [
        ""
      ],
      "execution_count": 0,
      "outputs": []
    }
  ]
}