SqueezeNetNN.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "9u3d4Z7uQsmp"
   },
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Using TensorFlow backend.\n"
     ]
    }
   ],
   "source": [
    "import os, math, json, random, time\n",
    "import numpy as np\n",
    "from matplotlib import pyplot as plt\n",
    "import tensorflow as tf\n",
    "AUTO = tf.data.experimental.AUTOTUNE\n",
    "from keras_applications.imagenet_utils import _obtain_input_shape\n",
    "from keras import backend as K\n",
    "from keras.layers import Input, Convolution2D, MaxPooling2D, Activation, concatenate, Dropout\n",
    "from keras.layers import GlobalAveragePooling2D, GlobalMaxPooling2D\n",
    "from keras.models import Model\n",
    "from keras.engine.topology import get_source_inputs\n",
    "from keras.utils import get_file\n",
    "from keras.utils import layer_utils\n",
    "\n",
    "\n",
    "import numpy as np\n",
    "from keras.applications.imagenet_utils import preprocess_input, decode_predictions\n",
    "from keras.preprocessing import image"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "'[[[ 95  69 108]\\n  [135 109 136]\\n  [124 103 112]\\n  ...\\n  [135  77  40]\\n  [127  79  43]\\n  [100  61  44]]\\n\\n [[171 145 154]\\n  [227 205 208]\\n  [235 217 213]\\n  ...\\n  [215 154  91]\\n  [211 161 100]\\n  [169 130  87]]\\n\\n [[170 152 132]\\n  [253 238 219]\\n  [255 254 238]\\n  ...\\n  [216 151  67]\\n  [220 169  88]\\n  [163 125  62]]\\n\\n ...\\n\\n [[ 76  56  58]\\n  [ 51  26  29]\\n  [ 48  18  20]\\n  ...\\n  [ 22  14  12]\\n  [ 27  11  11]\\n  [ 37  16  15]]\\n\\n [[ 44  29  32]\\n  [ 31  16  21]\\n  [ 35  19  22]\\n  ...\\n  [ 23  17  19]\\n  [ 23  11  15]\\n  [ 26  10  13]]\\n\\n [[ 27   5   7]\\n  [ 24   5   7]\\n  [ 25  13  15]\\n  ...\\n  [ 28  17  23]\\n  [ 22   7  14]\\n  [ 19   2  10]]]'"
      ]
     },
     "execution_count": 2,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# Load the data\n",
    "import pandas as pd \n",
    "\n",
    "data = pd.read_csv('combined_json_and_images.csv') \n",
    "\n",
    "# Preview the first 5 lines of the loaded data \n",
    "data.image[0]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Here's our 5 categories that we have to classify.\n",
    "# class_names = ['inside', 'outside', 'food', 'drink', 'menu']\n",
    "from sklearn.model_selection import train_test_split\n",
    "class_names = ['drink', 'food', 'inside', 'menu', 'outside']\n",
    "class_names_label = { 0:'drink',\n",
    "                      1: 'food',\n",
    "                      2: 'inside',\n",
    "                      3: 'menu',\n",
    "                      4: 'outside',\n",
    "                    }\n",
    "nb_classes = 5\n",
    "x= data.drop(columns=['class_names_label'],axis=1)\n",
    "y= data['class_names_label']\n",
    "x_train_o, x_test_o, y_train , y_test  = train_test_split(x, y, test_size=0.20, random_state=42)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 0,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "FpvUOuC3j27n"
   },
   "outputs": [],
   "source": [
    "tpu = tf.contrib.cluster_resolver.TPUClusterResolver()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 0,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "M3G-2aUBQJ-H"
   },
   "outputs": [],
   "source": [
    "GCS_PATTERN = 'gs://flowers-public/tfrecords-jpeg-192x192-2/*.tfrec'\n",
    "IMAGE_SIZE = [192, 192]\n",
    "\n",
    "BATCH_SIZE = 128\n",
    "\n",
    "VALIDATION_SPLIT = 0.19\n",
    "CLASSES = ['drink', 'food', 'inside', 'menu', 'outside']\n",
    "\n",
    "filenames = tf.gfile.Glob(GCS_PATTERN)\n",
    "split = int(len(filenames) * VALIDATION_SPLIT)\n",
    "training_filenames = filenames[split:]\n",
    "validation_filenames = filenames[:split]\n",
    "validation_steps = int(3670 // len(filenames) * len(validation_filenames)) // BATCH_SIZE\n",
    "steps_per_epoch = int(3670 // len(filenames) * len(training_filenames)) // BATCH_SIZE"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 0,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "LtAVr-4CP1rp"
   },
   "outputs": [],
   "source": [
    "def read_tfrecord(example):\n",
    "    features = {\n",
    "        \"image\": tf.io.FixedLenFeature([], tf.string), \n",
    "        \"class\": tf.io.FixedLenFeature([], tf.int64),\n",
    "        \"one_hot_class\": tf.io.VarLenFeature(tf.float32),\n",
    "    }\n",
    "    example = tf.parse_single_example(example, features)\n",
    "    image = tf.image.decode_jpeg(example['image'], channels=3)\n",
    "    image = tf.cast(image, tf.float32) / 255.0  # convert image to floats in [0, 1] range\n",
    "    image = tf.reshape(image, [*IMAGE_SIZE, 3])\n",
    "    class_label = tf.cast(example['class'], tf.int32)\n",
    "    one_hot_class = tf.sparse.to_dense(example['one_hot_class'])\n",
    "    one_hot_class = tf.reshape(one_hot_class, [5])\n",
    "    return image, one_hot_class\n",
    "\n",
    "def load_dataset(filenames):\n",
    "\n",
    "  option_no_order = tf.data.Options()\n",
    "  option_no_order.experimental_deterministic = False\n",
    "\n",
    "  dataset = tf.data.Dataset.from_tensor_slices(filenames)\n",
    "  dataset = dataset.with_options(option_no_order)\n",
    "  dataset = dataset.interleave(tf.data.TFRecordDataset, cycle_length=16, num_parallel_calls=AUTO) # faster\n",
    "  dataset = dataset.map(read_tfrecord, num_parallel_calls=AUTO)\n",
    "  return dataset"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "colab_type": "text",
    "id": "22rVDTx8wCqE"
   },
   "source": [
    "## training and validation datasets"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 0,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "7wxKyCklR4Gh"
   },
   "outputs": [],
   "source": [
    "def data_augment(image, one_hot_class):\n",
    "    image = tf.image.random_flip_left_right(image)\n",
    "    image = tf.image.random_saturation(image, 0, 2)\n",
    "    return image, one_hot_class# Plot a pie chart to check proporation of each observed category\n",
    "\n",
    "sizes = np.bincount(y)\n",
    "explode = (0, 0, 0, 0, 0, 0)  \n",
    "plt.pie(sizes, explode=None, labels=class_names,\n",
    "autopct='%1.1f%%', shadow=True, startangle=150)\n",
    "plt.axis('equal')\n",
    "plt.title('Proportion of each observed category')\n",
    "\n",
    "plt.show()\n",
    "\n",
    "def get_batched_dataset(filenames, augment_data):\n",
    "  dataset = load_dataset(filenames)\n",
    "  dataset = dataset.cache() # This dataset fits in RAM\n",
    "  dataset = dataset.repeat()\n",
    "  if augment_data:\n",
    "      dataset = dataset.map(data_augment, num_parallel_calls=AUTO)\n",
    "  dataset = dataset.shuffle(2048)\n",
    "  dataset = dataset.batch(BATCH_SIZE, drop_remainder=True)\n",
    "  dataset = dataset.prefetch(AUTO)\n",
    "  return dataset\n",
    "\n",
    "def get_training_dataset():\n",
    "  return get_batched_dataset(training_filenames, augment_data=True)\n",
    "\n",
    "def get_validation_dataset():\n",
    "  return get_batched_dataset(validation_filenames, augment_data=False)\n",
    "\n",
    "some_flowers, some_labels = dataset_to_numpy_util(load_dataset(validation_filenames), 8*20)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "# Plot a pie chart to check proporation of each observed category\n",
    "\n",
    "sizes = np.bincount(y)\n",
    "explode = (0, 0, 0, 0, 0, 0)  \n",
    "plt.pie(sizes, explode=None, labels=class_names,\n",
    "autopct='%1.1f%%', shadow=True, startangle=150)\n",
    "plt.axis('equal')\n",
    "plt.title('Proportion of each observed category')\n",
    "\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "colab_type": "text",
    "id": "ALtRUlxhw8Vt"
   },
   "source": [
    "## Model"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 35,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 1000
    },
    "colab_type": "code",
    "id": "XLJNVGwHUDy1",
    "outputId": "7280e4c6-5b0d-47b8-d1a4-db3745f3263f"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Model: \"model_2\"\n",
      "__________________________________________________________________________________________________\n",
      "Layer (type)                    Output Shape         Param #     Connected to                     \n",
      "==================================================================================================\n",
      "input_3 (InputLayer)            [(None, 192, 192, 3) 0                                            \n",
      "__________________________________________________________________________________________________\n",
      "conv2d_32 (Conv2D)              (None, 192, 192, 32) 896         input_3[0][0]                    \n",
      "__________________________________________________________________________________________________\n",
      "batch_normalization_32 (BatchNo (None, 192, 192, 32) 128         conv2d_32[0][0]                  \n",
      "__________________________________________________________________________________________________\n",
      "conv2d_33 (Conv2D)              (None, 192, 192, 24) 792         batch_normalization_32[0][0]     \n",
      "__________________________________________________________________________________________________\n",
      "batch_normalization_33 (BatchNo (None, 192, 192, 24) 96          conv2d_33[0][0]                  \n",
      "__________________________________________________________________________________________________\n",
      "conv2d_34 (Conv2D)              (None, 192, 192, 24) 600         batch_normalization_33[0][0]     \n",
      "__________________________________________________________________________________________________\n",
      "conv2d_35 (Conv2D)              (None, 192, 192, 24) 5208        batch_normalization_33[0][0]     \n",
      "__________________________________________________________________________________________________\n",
      "batch_normalization_34 (BatchNo (None, 192, 192, 24) 96          conv2d_34[0][0]                  \n",
      "__________________________________________________________________________________________________\n",
      "batch_normalization_35 (BatchNo (None, 192, 192, 24) 96          conv2d_35[0][0]                  \n",
      "__________________________________________________________________________________________________\n",
      "concatenate_10 (Concatenate)    (None, 192, 192, 48) 0           batch_normalization_34[0][0]     \n",
      "                                                                 batch_normalization_35[0][0]     \n",
      "__________________________________________________________________________________________________\n",
      "max_pooling2d_8 (MaxPooling2D)  (None, 96, 96, 48)   0           concatenate_10[0][0]             \n",
      "__________________________________________________________________________________________________\n",
      "conv2d_36 (Conv2D)              (None, 96, 96, 48)   2352        max_pooling2d_8[0][0]            \n",
      "__________________________________________________________________________________________________\n",
      "batch_normalization_36 (BatchNo (None, 96, 96, 48)   192         conv2d_36[0][0]                  \n",
      "__________________________________________________________________________________________________\n",
      "conv2d_37 (Conv2D)              (None, 96, 96, 48)   2352        batch_normalization_36[0][0]     \n",
      "__________________________________________________________________________________________________\n",
      "conv2d_38 (Conv2D)              (None, 96, 96, 48)   20784       batch_normalization_36[0][0]     \n",
      "__________________________________________________________________________________________________\n",
      "batch_normalization_37 (BatchNo (None, 96, 96, 48)   192         conv2d_37[0][0]                  \n",
      "__________________________________________________________________________________________________\n",
      "batch_normalization_38 (BatchNo (None, 96, 96, 48)   192         conv2d_38[0][0]                  \n",
      "__________________________________________________________________________________________________\n",
      "concatenate_11 (Concatenate)    (None, 96, 96, 96)   0           batch_normalization_37[0][0]     \n",
      "                                                                 batch_normalization_38[0][0]     \n",
      "__________________________________________________________________________________________________\n",
      "max_pooling2d_9 (MaxPooling2D)  (None, 48, 48, 96)   0           concatenate_11[0][0]             \n",
      "__________________________________________________________________________________________________\n",
      "conv2d_39 (Conv2D)              (None, 48, 48, 64)   6208        max_pooling2d_9[0][0]            \n",
      "__________________________________________________________________________________________________\n",
      "batch_normalization_39 (BatchNo (None, 48, 48, 64)   256         conv2d_39[0][0]                  \n",
      "__________________________________________________________________________________________________\n",
      "conv2d_40 (Conv2D)              (None, 48, 48, 64)   4160        batch_normalization_39[0][0]     \n",
      "__________________________________________________________________________________________________\n",
      "conv2d_41 (Conv2D)              (None, 48, 48, 64)   36928       batch_normalization_39[0][0]     \n",
      "__________________________________________________________________________________________________\n",
      "batch_normalization_40 (BatchNo (None, 48, 48, 64)   256         conv2d_40[0][0]                  \n",
      "__________________________________________________________________________________________________\n",
      "batch_normalization_41 (BatchNo (None, 48, 48, 64)   256         conv2d_41[0][0]                  \n",
      "__________________________________________________________________________________________________\n",
      "concatenate_12 (Concatenate)    (None, 48, 48, 128)  0           batch_normalization_40[0][0]     \n",
      "                                                                 batch_normalization_41[0][0]     \n",
      "__________________________________________________________________________________________________\n",
      "max_pooling2d_10 (MaxPooling2D) (None, 24, 24, 128)  0           concatenate_12[0][0]             \n",
      "__________________________________________________________________________________________________\n",
      "conv2d_42 (Conv2D)              (None, 24, 24, 48)   6192        max_pooling2d_10[0][0]           \n",
      "__________________________________________________________________________________________________\n",
      "batch_normalization_42 (BatchNo (None, 24, 24, 48)   192         conv2d_42[0][0]                  \n",
      "__________________________________________________________________________________________________\n",
      "conv2d_43 (Conv2D)              (None, 24, 24, 48)   2352        batch_normalization_42[0][0]     \n",
      "__________________________________________________________________________________________________\n",
      "conv2d_44 (Conv2D)              (None, 24, 24, 48)   20784       batch_normalization_42[0][0]     \n",
      "__________________________________________________________________________________________________\n",
      "batch_normalization_43 (BatchNo (None, 24, 24, 48)   192         conv2d_43[0][0]                  \n",
      "__________________________________________________________________________________________________\n",
      "batch_normalization_44 (BatchNo (None, 24, 24, 48)   192         conv2d_44[0][0]                  \n",
      "__________________________________________________________________________________________________\n",
      "concatenate_13 (Concatenate)    (None, 24, 24, 96)   0           batch_normalization_43[0][0]     \n",
      "                                                                 batch_normalization_44[0][0]     \n",
      "__________________________________________________________________________________________________\n",
      "max_pooling2d_11 (MaxPooling2D) (None, 12, 12, 96)   0           concatenate_13[0][0]             \n",
      "__________________________________________________________________________________________________\n",
      "conv2d_45 (Conv2D)              (None, 12, 12, 24)   2328        max_pooling2d_11[0][0]           \n",
      "__________________________________________________________________________________________________\n",
      "batch_normalization_45 (BatchNo (None, 12, 12, 24)   96          conv2d_45[0][0]                  \n",
      "__________________________________________________________________________________________________\n",
      "conv2d_46 (Conv2D)              (None, 12, 12, 24)   600         batch_normalization_45[0][0]     \n",
      "__________________________________________________________________________________________________\n",
      "conv2d_47 (Conv2D)              (None, 12, 12, 24)   5208        batch_normalization_45[0][0]     \n",
      "__________________________________________________________________________________________________\n",
      "batch_normalization_46 (BatchNo (None, 12, 12, 24)   96          conv2d_46[0][0]                  \n",
      "__________________________________________________________________________________________________\n",
      "batch_normalization_47 (BatchNo (None, 12, 12, 24)   96          conv2d_47[0][0]                  \n",
      "__________________________________________________________________________________________________\n",
      "concatenate_14 (Concatenate)    (None, 12, 12, 48)   0           batch_normalization_46[0][0]     \n",
      "                                                                 batch_normalization_47[0][0]     \n",
      "__________________________________________________________________________________________________\n",
      "global_average_pooling2d_2 (Glo (None, 48)           0           concatenate_14[0][0]             \n",
      "__________________________________________________________________________________________________\n",
      "dense_2 (Dense)                 (None, 5)            245         global_average_pooling2d_2[0][0] \n",
      "==================================================================================================\n",
      "Total params: 120,613\n",
      "Trainable params: 119,301\n",
      "Non-trainable params: 1,312\n",
      "__________________________________________________________________________________________________\n"
     ]
    }
   ],
   "source": [
    "bnmomemtum=0.9\n",
    "def fire(x, squeeze, expand):\n",
    "  y  = tf.keras.layers.Conv2D(filters=squeeze, kernel_size=1, activation='relu', padding='same')(x)\n",
    "  y = tf.keras.layers.BatchNormalization(momentum=bnmomemtum)(y)\n",
    "  y1 = tf.keras.layers.Conv2D(filters=expand//2, kernel_size=1, activation='relu', padding='same')(y)\n",
    "  y1 = tf.keras.layers.BatchNormalization(momentum=bnmomemtum)(y1)\n",
    "  y3 = tf.keras.layers.Conv2D(filters=expand//2, kernel_size=3, activation='relu', padding='same')(y)\n",
    "  y3 = tf.keras.layers.BatchNormalization(momentum=bnmomemtum)(y3)\n",
    "  return tf.keras.layers.concatenate([y1, y3])\n",
    "\n",
    "def fire_module(squeeze, expand):\n",
    "  return lambda x: fire(x, squeeze, expand)\n",
    "\n",
    "x = tf.keras.layers.Input(shape=[*IMAGE_SIZE, 3]) # input is 192x192 pixels RGB\n",
    "\n",
    "y = tf.keras.layers.Conv2D(kernel_size=3, filters=32, padding='same', use_bias=True, activation='relu')(x)\n",
    "y = tf.keras.layers.BatchNormalization(momentum=bnmomemtum)(y)\n",
    "y = fire_module(24, 48)(y)\n",
    "y = tf.keras.layers.MaxPooling2D(pool_size=2)(y)\n",
    "y = fire_module(48, 96)(y)\n",
    "y = tf.keras.layers.MaxPooling2D(pool_size=2)(y)\n",
    "y = fire_module(64, 128)(y)\n",
    "y = tf.keras.layers.MaxPooling2D(pool_size=2)(y)\n",
    "y = fire_module(48, 96)(y)\n",
    "y = tf.keras.layers.MaxPooling2D(pool_size=2)(y)\n",
    "y = fire_module(24, 48)(y)\n",
    "y = tf.keras.layers.GlobalAveragePooling2D()(y)\n",
    "y = tf.keras.layers.Dense(5, activation='softmax')(y)\n",
    "\n",
    "model = tf.keras.Model(x, y)\n",
    "\n",
    "model.compile(\n",
    "  optimizer='adam',\n",
    "  loss= 'categorical_crossentropy',\n",
    "  metrics=['accuracy'])\n",
    "\n",
    "model.summary()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "colab_type": "text",
    "id": "dMfenMQcxAAb"
   },
   "source": [
    "## Training"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 0,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 306
    },
    "colab_type": "code",
    "id": "qGYZ2uOxLgkr",
    "outputId": "681964c6-2171-48ac-c685-5d5851730f30"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "INFO:tensorflow:Querying Tensorflow master (grpc://192.168.43.2:8470) for TPU system metadata.\n",
      "INFO:tensorflow:Found TPU system:\n",
      "INFO:tensorflow:*** Num TPU Cores: 8\n",
      "INFO:tensorflow:*** Num TPU Workers: 1\n",
      "INFO:tensorflow:*** Num TPU Cores Per Worker: 8\n",
      "INFO:tensorflow:*** Available Device: _DeviceAttributes(/job:worker/replica:0/task:0/device:CPU:0, CPU, -1, 9395547180570580755)\n",
      "INFO:tensorflow:*** Available Device: _DeviceAttributes(/job:worker/replica:0/task:0/device:XLA_CPU:0, XLA_CPU, 17179869184, 6367633851149154879)\n",
      "INFO:tensorflow:*** Available Device: _DeviceAttributes(/job:worker/replica:0/task:0/device:TPU:0, TPU, 17179869184, 11596888001717943185)\n",
      "INFO:tensorflow:*** Available Device: _DeviceAttributes(/job:worker/replica:0/task:0/device:TPU:1, TPU, 17179869184, 9183457543646557935)\n",
      "INFO:tensorflow:*** Available Device: _DeviceAttributes(/job:worker/replica:0/task:0/device:TPU:2, TPU, 17179869184, 10822347198377972777)\n",
      "INFO:tensorflow:*** Available Device: _DeviceAttributes(/job:worker/replica:0/task:0/device:TPU:3, TPU, 17179869184, 5956408950500829060)\n",
      "INFO:tensorflow:*** Available Device: _DeviceAttributes(/job:worker/replica:0/task:0/device:TPU:4, TPU, 17179869184, 922822146319201418)\n",
      "INFO:tensorflow:*** Available Device: _DeviceAttributes(/job:worker/replica:0/task:0/device:TPU:5, TPU, 17179869184, 6791916320445230285)\n",
      "INFO:tensorflow:*** Available Device: _DeviceAttributes(/job:worker/replica:0/task:0/device:TPU:6, TPU, 17179869184, 5871677175911293786)\n",
      "INFO:tensorflow:*** Available Device: _DeviceAttributes(/job:worker/replica:0/task:0/device:TPU:7, TPU, 17179869184, 11941388521539288345)\n",
      "INFO:tensorflow:*** Available Device: _DeviceAttributes(/job:worker/replica:0/task:0/device:TPU_SYSTEM:0, TPU_SYSTEM, 17179869184, 186432556399494142)\n",
      "WARNING:tensorflow:tpu_model (from tensorflow.contrib.tpu.python.tpu.keras_support) is experimental and may change or be removed at any time, and without warning.\n",
      "INFO:tensorflow:Cloning Adam {'beta_2': 0.9990000128746033, 'beta_1': 0.8999999761581421, 'epsilon': 1e-07, 'lr': 0.0010000000474974513, 'decay': 0.0, 'amsgrad': False}\n",
      "INFO:tensorflow:Cloning Adam {'beta_2': 0.9990000128746033, 'beta_1': 0.8999999761581421, 'epsilon': 1e-07, 'lr': 0.0010000000474974513, 'decay': 0.0, 'amsgrad': False}\n"
     ]
    }
   ],
   "source": [
    "strategy = tf.contrib.tpu.TPUDistributionStrategy(tpu)\n",
    "tpu_model = tf.contrib.tpu.keras_to_tpu_model(model, strategy=strategy)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 0,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 904
    },
    "colab_type": "code",
    "id": "M-ID7vP5mIKs",
    "outputId": "1403675f-3853-4aed-94ea-bcc1b68026e7"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 1/15\n",
      "INFO:tensorflow:New input shapes; (re-)compiling: mode=train (# of cores 8), [TensorSpec(shape=(128,), dtype=tf.int32, name=None), TensorSpec(shape=(128, 192, 192, 3), dtype=tf.float32, name=None), TensorSpec(shape=(128, 5), dtype=tf.float32, name=None)]\n",
      "INFO:tensorflow:Overriding default placeholder.\n",
      "INFO:tensorflow:Cloning Adam {'beta_2': 0.9990000128746033, 'beta_1': 0.8999999761581421, 'epsilon': 1e-07, 'lr': 0.0010000000474974513, 'decay': 0.0, 'amsgrad': False}\n",
      "INFO:tensorflow:Remapping placeholder for input_1\n",
      "WARNING:tensorflow:From /usr/local/lib/python3.5/dist-packages/tensorflow/contrib/tpu/python/tpu/keras_support.py:302: to_float (from tensorflow.python.ops.math_ops) is deprecated and will be removed in a future version.\n",
      "Instructions for updating:\n",
      "Use tf.cast instead.\n",
      "INFO:tensorflow:KerasCrossShard: <tensorflow.python.keras.optimizers.Adam object at 0x7f9d324357f0> []\n",
      "WARNING:tensorflow:From /usr/local/lib/python3.5/dist-packages/tensorflow/python/ops/math_ops.py:3066: to_int32 (from tensorflow.python.ops.math_ops) is deprecated and will be removed in a future version.\n",
      "Instructions for updating:\n",
      "Use tf.cast instead.\n",
      "INFO:tensorflow:Started compiling\n",
      "INFO:tensorflow:Finished compiling. Time elapsed: 18.098301887512207 secs\n",
      "INFO:tensorflow:Setting weights on TPU model.\n",
      "INFO:tensorflow:CPU -> TPU beta_2: 0.9990000128746033 {0.999}\n",
      "INFO:tensorflow:CPU -> TPU beta_1: 0.8999999761581421 {0.9}\n",
      "WARNING:tensorflow:Cannot update non-variable config: epsilon\n",
      "INFO:tensorflow:CPU -> TPU lr: 0.0010000000474974513 {0.001}\n",
      "INFO:tensorflow:CPU -> TPU decay: 0.0 {0.0}\n",
      "WARNING:tensorflow:Cannot update non-variable config: amsgrad\n",
      "22/23 [===========================>..] - ETA: 2s - loss: 1.1371 - acc: 0.5558INFO:tensorflow:New input shapes; (re-)compiling: mode=eval (# of cores 8), [TensorSpec(shape=(128,), dtype=tf.int32, name=None), TensorSpec(shape=(128, 192, 192, 3), dtype=tf.float32, name=None), TensorSpec(shape=(128, 5), dtype=tf.float32, name=None)]\n",
      "INFO:tensorflow:Overriding default placeholder.\n",
      "INFO:tensorflow:Cloning Adam {'beta_2': 0.9990000128746033, 'beta_1': 0.8999999761581421, 'epsilon': 1e-07, 'lr': 0.0010000000474974513, 'decay': 0.0, 'amsgrad': False}\n",
      "INFO:tensorflow:Remapping placeholder for input_1\n",
      "INFO:tensorflow:KerasCrossShard: <tensorflow.python.keras.optimizers.Adam object at 0x7f9cd0801f60> []\n",
      "INFO:tensorflow:Started compiling\n",
      "INFO:tensorflow:Finished compiling. Time elapsed: 9.21638822555542 secs\n",
      "23/23 [==============================] - 83s 4s/step - loss: 1.1274 - acc: 0.5604 - val_loss: 2.8444 - val_acc: 0.2369\n",
      "Epoch 2/15\n",
      "23/23 [==============================] - 7s 296ms/step - loss: 0.8550 - acc: 0.6796 - val_loss: 1.6550 - val_acc: 0.4414\n",
      "Epoch 3/15\n",
      "23/23 [==============================] - 7s 297ms/step - loss: 0.7286 - acc: 0.7300 - val_loss: 0.7840 - val_acc: 0.7088\n",
      "Epoch 4/15\n",
      "23/23 [==============================] - 7s 298ms/step - loss: 0.6329 - acc: 0.7731 - val_loss: 0.6621 - val_acc: 0.7547\n",
      "Epoch 5/15\n",
      "23/23 [==============================] - 7s 296ms/step - loss: 0.5823 - acc: 0.7895 - val_loss: 0.6417 - val_acc: 0.7625\n",
      "Epoch 6/15\n",
      "23/23 [==============================] - 7s 297ms/step - loss: 0.5155 - acc: 0.8161 - val_loss: 0.6441 - val_acc: 0.7697\n",
      "Epoch 7/15\n",
      "23/23 [==============================] - 7s 296ms/step - loss: 0.4720 - acc: 0.8362 - val_loss: 0.6346 - val_acc: 0.7668\n",
      "Epoch 8/15\n",
      "23/23 [==============================] - 7s 297ms/step - loss: 0.4279 - acc: 0.8466 - val_loss: 0.6878 - val_acc: 0.7479\n",
      "Epoch 9/15\n",
      "23/23 [==============================] - 7s 297ms/step - loss: 0.3976 - acc: 0.8575 - val_loss: 0.6505 - val_acc: 0.7586\n",
      "Epoch 10/15\n",
      "23/23 [==============================] - 7s 297ms/step - loss: 0.3569 - acc: 0.8753 - val_loss: 0.5721 - val_acc: 0.7990\n",
      "Epoch 11/15\n",
      "23/23 [==============================] - 7s 296ms/step - loss: 0.3207 - acc: 0.8905 - val_loss: 0.6561 - val_acc: 0.7621\n",
      "Epoch 12/15\n",
      "23/23 [==============================] - 7s 296ms/step - loss: 0.2900 - acc: 0.9028 - val_loss: 0.6494 - val_acc: 0.7857\n",
      "Epoch 13/15\n",
      "23/23 [==============================] - 7s 297ms/step - loss: 0.2580 - acc: 0.9130 - val_loss: 0.6280 - val_acc: 0.7836\n",
      "Epoch 14/15\n",
      "23/23 [==============================] - 7s 296ms/step - loss: 0.2463 - acc: 0.9190 - val_loss: 0.6751 - val_acc: 0.7426\n",
      "Epoch 15/15\n",
      "23/23 [==============================] - 7s 298ms/step - loss: 0.1997 - acc: 0.9392 - val_loss: 0.5852 - val_acc: 0.8092\n",
      "Time elapsed:  180.92546439170837  sec\n"
     ]
    }
   ],
   "source": [
    "EPOCHS = 15\n",
    "\n",
    "start_time = time.time()\n",
    "\n",
    "history = tpu_model.fit(get_training_dataset, steps_per_epoch=steps_per_epoch, epochs=EPOCHS,\n",
    "                          validation_data=get_validation_dataset, validation_steps=validation_steps)\n",
    "end_time = time.time()\n",
    "print(\"Time elapsed: \", end_time-start_time, \" sec\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 0,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 772
    },
    "colab_type": "code",
    "id": "VngeUBIdyJ1T",
    "outputId": "b23831cd-15c3-49be-fcee-23ca33f613ae"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "dict_keys(['val_acc', 'loss', 'val_loss', 'acc'])\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 720x720 with 2 Axes>"
      ]
     },
     "metadata": {
      "tags": []
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "print(history.history.keys())\n",
    "display_training_curves(history.history['acc'], history.history['val_acc'], 'accuracy', 211)\n",
    "display_training_curves(history.history['loss'], history.history['val_loss'], 'loss', 212)"
   ]
  }
 ],
 "metadata": {
  "accelerator": "TPU",
  "colab": {
   "collapsed_sections": [],
   "name": "Keras Flowers squeezenet.ipynb",
   "provenance": [],
   "toc_visible": true
  },
  "kernelspec": {
   "display_name": "Python 3",
   "language": "python",
   "name": "python3"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.6.9"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 1
}