05-Callbacks_and_Tensorbord.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Callbacks and Tensorboard\n",
    "\n",
    "We'll go over various callbacks such as:\n",
    "\n",
    "* Early Stopping\n",
    "* History (Especially useful for functional models)\n",
    "* Model History\n",
    "* Tensorboard\n",
    "* Model Accuracy and Loss Graphs"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "'2.2.4-tf'"
      ]
     },
     "execution_count": 13,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "import tensorflow\n",
    "import tensorflow.keras as keras\n",
    "from tensorflow.keras.preprocessing.image import ImageDataGenerator\n",
    "from tensorflow.keras.applications import ResNet50\n",
    "from tensorflow.keras.applications.resnet50 import preprocess_input\n",
    "from tensorflow.keras import Model, layers\n",
    "from tensorflow.keras.models import load_model, model_from_json\n",
    "tensorflow.keras.__version__"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {},
   "outputs": [],
   "source": [
    "train_datagen = ImageDataGenerator(\n",
    "    shear_range=10,\n",
    "    zoom_range=0.2,\n",
    "    horizontal_flip=True,\n",
    "    vertical_flip=True,\n",
    "    preprocessing_function=preprocess_input)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Found 12894 images belonging to 10 classes.\n"
     ]
    }
   ],
   "source": [
    "train_generator = train_datagen.flow_from_directory(\n",
    "    'data/train',\n",
    "    batch_size=32,\n",
    "    class_mode='categorical',\n",
    "    target_size=(224, 224))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Found 500 images belonging to 10 classes.\n"
     ]
    }
   ],
   "source": [
    "validation_datagen = ImageDataGenerator(preprocessing_function=preprocess_input)\n",
    "validation_generator = validation_datagen.flow_from_directory(\n",
    "    'data/val',\n",
    "    shuffle=False,\n",
    "    class_mode='categorical',\n",
    "    target_size=(224, 224))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "metadata": {},
   "outputs": [],
   "source": [
    "conv_base = ResNet50(include_top=False, weights='imagenet')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Freeze all of the base layers\n",
    "for layer in conv_base.layers:\n",
    "    layer.trainable = False# Freeze all of the base layers"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "metadata": {},
   "outputs": [],
   "source": [
    "x  = conv_base.output\n",
    "x = layers.GlobalAveragePooling2D()(x)\n",
    "\n",
    "predictions = layers.Dense(10, activation='softmax', name='fc3')(x)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 20,
   "metadata": {},
   "outputs": [],
   "source": [
    "model = Model(conv_base.input, predictions)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 21,
   "metadata": {},
   "outputs": [],
   "source": [
    "#model.summary()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 22,
   "metadata": {},
   "outputs": [],
   "source": [
    "opt = tensorflow.keras.optimizers.Adam()\n",
    "model.compile(loss='categorical_crossentropy',\n",
    "              optimizer=opt,\n",
    "              metrics=['accuracy'])"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Callbacks\n",
    "\n",
    "Callbacks are a good way to:\n",
    "\n",
    "* get feedback about training while training\n",
    "* save history for analysis\n",
    "* checkpoint save your model whenever performance increases\n",
    "\n",
    "You can use Tensorboard for advanced diagnostics and graph view\n",
    "\n",
    "You will need this command for Tensorboard: `logdir=graph/ --host 0.0.0.0 --port 8999`"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 23,
   "metadata": {},
   "outputs": [],
   "source": [
    "filepath = 'models/weights.{epoch:02d}-{val_loss:.2f}.hdf5'\n",
    "\n",
    "stopping_criterion = keras.callbacks.EarlyStopping(monitor='val_accuracy', patience=2)\n",
    "\n",
    "history = keras.callbacks.History()\n",
    "\n",
    "checkpoint = keras.callbacks.ModelCheckpoint(filepath, monitor='val_accuracy', verbose=1, save_best_only=True)\n",
    "\n",
    "tensorboard = keras.callbacks.TensorBoard(log_dir='graph', histogram_freq=0, write_graph=True)\n",
    "\n",
    "callbacks = [stopping_criterion, history, checkpoint, tensorboard]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 24,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 1/4\n",
      "402/403 [============================>.] - ETA: 0s - loss: 0.3583 - accuracy: 0.8905\n",
      "Epoch 00001: val_accuracy improved from -inf to 0.95800, saving model to models/weights.01-0.14.hdf5\n",
      "403/403 [==============================] - 155s 386ms/step - loss: 0.3587 - accuracy: 0.8903 - val_loss: 0.1362 - val_accuracy: 0.9580\n",
      "Epoch 2/4\n",
      "402/403 [============================>.] - ETA: 0s - loss: 0.1832 - accuracy: 0.9418\n",
      "Epoch 00002: val_accuracy improved from 0.95800 to 0.96000, saving model to models/weights.02-0.13.hdf5\n",
      "403/403 [==============================] - 151s 376ms/step - loss: 0.1831 - accuracy: 0.9419 - val_loss: 0.1272 - val_accuracy: 0.9600\n",
      "Epoch 3/4\n",
      "402/403 [============================>.] - ETA: 0s - loss: 0.1529 - accuracy: 0.9510\n",
      "Epoch 00003: val_accuracy did not improve from 0.96000\n",
      "403/403 [==============================] - 153s 380ms/step - loss: 0.1527 - accuracy: 0.9511 - val_loss: 0.1375 - val_accuracy: 0.9580\n",
      "Epoch 4/4\n",
      "402/403 [============================>.] - ETA: 0s - loss: 0.1451 - accuracy: 0.9530\n",
      "Epoch 00004: val_accuracy improved from 0.96000 to 0.96400, saving model to models/weights.04-0.12.hdf5\n",
      "403/403 [==============================] - 155s 384ms/step - loss: 0.1449 - accuracy: 0.9531 - val_loss: 0.1166 - val_accuracy: 0.9640\n"
     ]
    }
   ],
   "source": [
    "history = model.fit_generator(\n",
    "    generator=train_generator,\n",
    "    epochs = 4,\n",
    "    validation_data=validation_generator,\n",
    "    callbacks=callbacks)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Model Accuracy and Loss Graphs\n",
    "\n",
    "These graphs will show us how our model performed\n",
    "\n",
    "We can get and idea of whether it is underfitting or overfitting\n",
    "\n",
    "You can also view these graphs in tensorboard"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 25,
   "metadata": {},
   "outputs": [],
   "source": [
    "import matplotlib.pyplot as plt"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 28,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "# Summarize history for accuracy\n",
    "plt.plot(history.history['accuracy'])\n",
    "plt.plot(history.history['val_accuracy'])\n",
    "plt.title('model accuracy')\n",
    "plt.xlabel('epoch')\n",
    "plt.ylabel('accuracy')\n",
    "plt.legend(['train', 'test'], loc='upper left')\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 29,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "# Summarize history for accuracy\n",
    "plt.plot(history.history['loss'])\n",
    "plt.plot(history.history['val_loss'])\n",
    "plt.title('model loss')\n",
    "plt.xlabel('epoch')\n",
    "plt.ylabel('loss')\n",
    "plt.legend(['train', 'test'], loc='upper left')\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## History Callback Gotcha\n",
    "\n",
    "If you use `load_model()` on your saved model and try to access history it will throw an error\n",
    "\n",
    "So if you want to see these graphs in a notebook you need to do it in the notebook you trained the model in"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  }
 ],
 "metadata": {
  "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": "3.7.3"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 4
}