notebooks/bigdl_features/visualization.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Visualizing learning in BigDL"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "In this example, we are going to use the [MNIST dataset](http://yann.lecun.com/exdb/mnist/) to train a [lenet-5](http://yann.lecun.com/exdb/lenet/) model, and visualize this training using jupyter and tensorboard."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Populating the interactive namespace from numpy and matplotlib\n",
      "/home/xin/IntelAnalytics/BigDL-Tutorials2/notebooks/neural_networks\n"
     ]
    }
   ],
   "source": [
    "import matplotlib\n",
    "matplotlib.use('Agg')\n",
    "%pylab inline\n",
    "\n",
    "import datetime as dt\n",
    "import tempfile\n",
    "\n",
    "import sys\n",
    "import os\n",
    "print(os.path.abspath(os.path.join('../neural_networks')))\n",
    "sys.path.append(os.path.abspath(os.path.join('../neural_networks')))\n",
    "from utils import get_mnist\n",
    "\n",
    "from bigdl.nn.layer import *\n",
    "from bigdl.nn.criterion import *\n",
    "from bigdl.optim.optimizer import *\n",
    "from bigdl.util.common import *\n",
    "from bigdl.dataset.transformer import *\n",
    "from bigdl.dataset import mnist\n",
    "from matplotlib.pyplot import imshow\n",
    "import matplotlib.pyplot as plt\n",
    "from pyspark import SparkContext\n",
    "from IPython.display import Markdown, display\n",
    "\n",
    "sc=SparkContext.getOrCreate(conf=create_spark_conf().setMaster(\"local[4]\").set(\"spark.driver.memory\",\"2g\"))\n",
    "\n",
    "init_engine()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## 1. Train the network"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "* prepare training and validation samples"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Please edit the \"mnist_path\" accordingly. If the \"mnist_path\" directory does not consist of the mnist data, mnist.read_data_sets method will download the dataset directly to the directory."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "('Extracting', 'datasets/mnist/train-images-idx3-ubyte.gz')\n",
      "('Extracting', 'datasets/mnist/train-labels-idx1-ubyte.gz')\n",
      "('Extracting', 'datasets/mnist/t10k-images-idx3-ubyte.gz')\n",
      "('Extracting', 'datasets/mnist/t10k-labels-idx1-ubyte.gz')\n",
      "60000\n",
      "10000\n"
     ]
    }
   ],
   "source": [
    "# Get and store MNIST into RDD of Sample, please edit the \"mnist_path\" accordingly.\n",
    "mnist_path = \"datasets/mnist\"\n",
    "(train_data, test_data) = get_mnist(sc, mnist_path)\n",
    "\n",
    "print train_data.count()\n",
    "print test_data.count()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "* create the LeNet-5 model"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "creating: createReshape\n",
      "creating: createSpatialConvolution\n",
      "creating: createTanh\n",
      "creating: createSpatialMaxPooling\n",
      "creating: createTanh\n",
      "creating: createSpatialConvolution\n",
      "creating: createSpatialMaxPooling\n",
      "creating: createReshape\n",
      "creating: createLinear\n",
      "creating: createTanh\n",
      "creating: createLinear\n",
      "creating: createLogSoftMax\n",
      "creating: createModel\n"
     ]
    }
   ],
   "source": [
    "# Create a LeNet model\n",
    "def build_model(class_num):\n",
    "    input = Reshape([1, 28, 28]).set_name(\"reshape_1_28_28\")()\n",
    "    conv1 = SpatialConvolution(1, 6, 5, 5).set_name(\"conv1_5x5\")(input)\n",
    "    tanh1 = Tanh().set_name(\"tanh1\")(conv1)\n",
    "    pool1 = SpatialMaxPooling(2, 2, 2, 2).set_name(\"pool1_2x2\")(tanh1)\n",
    "    tanh2 = Tanh().set_name(\"tanh2\")(pool1)\n",
    "    conv2 = SpatialConvolution(6, 12, 5, 5).set_name(\"conv2_5x5\")(tanh2)\n",
    "    pool2 = SpatialMaxPooling(2, 2, 2, 2).set_name(\"pool2_2x2\")(conv2)\n",
    "    reshape = Reshape([12 * 4 * 4]).set_name(\"reshape_192\")(pool2)\n",
    "    fc1 = Linear(12 * 4 * 4, 100).set_name(\"fc_192_100\")(reshape)\n",
    "    tanh3 = Tanh().set_name(\"tanh3\")(fc1)\n",
    "    fc2 = Linear(100, class_num).set_name(\"fc_100_\" + str(class_num))(tanh3)\n",
    "    output = LogSoftMax().set_name(\"classifier\")(fc2)\n",
    "\n",
    "    model = Model(input, output)\n",
    "    return model\n",
    "lenet_model = build_model(10)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "* configure optimizer"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "creating: createClassNLLCriterion\n",
      "creating: createDefault\n",
      "creating: createSGD\n",
      "creating: createMaxEpoch\n",
      "creating: createDistriOptimizer\n",
      "creating: createEveryEpoch\n",
      "creating: createTop1Accuracy\n",
      "creating: createClassNLLCriterion\n",
      "creating: createLoss\n"
     ]
    }
   ],
   "source": [
    "# Create an Optimizer\n",
    "\n",
    "optimizer = Optimizer(\n",
    "    model=lenet_model,\n",
    "    training_rdd=train_data,\n",
    "    criterion=ClassNLLCriterion(),\n",
    "    optim_method=SGD(learningrate=0.4, learningrate_decay=0.0002),\n",
    "    end_trigger=MaxEpoch(5),\n",
    "    batch_size=256)\n",
    "\n",
    "# Set the validation logic\n",
    "optimizer.set_validation(\n",
    "    batch_size=256,\n",
    "    val_rdd=test_data,\n",
    "    trigger=EveryEpoch(),\n",
    "    val_method=[Top1Accuracy(), Loss()]\n",
    ")"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "* Will create a TrainSummary and a ValSummary and set them to Optimizer to enable metrics collecting. By default, BigDL only collect throughput, loss, and validation result during the Training. We can use \n",
    "`train_summary.set_summary_trigger` to enable more metrics, like `Parameters` and `LearningRate`."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "creating: createTrainSummary\n",
      "creating: createSeveralIteration\n",
      "creating: createSeveralIteration\n",
      "creating: createValidationSummary\n",
      "Will saving logs to /tmp/tmpvE6hzB/lenet-20180110-102403\n"
     ]
    },
    {
     "data": {
      "text/markdown": [
       "You can run shell command, **tensorboard --logdir=/tmp/tmpvE6hzB**, to visualize this training"
      ],
      "text/plain": [
       "<IPython.core.display.Markdown object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Optimization Done.\n"
     ]
    }
   ],
   "source": [
    "log_dir=tempfile.mkdtemp() # use a tmp folder\n",
    "app_name='lenet-'+dt.datetime.now().strftime(\"%Y%m%d-%H%M%S\")\n",
    "# create TrainSummary\n",
    "train_summary = TrainSummary(log_dir=log_dir, app_name=app_name)\n",
    "train_summary.set_summary_trigger(\"Parameters\", SeveralIteration(50))\n",
    "train_summary.set_summary_trigger(\"LearningRate\", SeveralIteration(1))\n",
    "# create ValSummary\n",
    "val_summary = ValidationSummary(log_dir=log_dir, app_name=app_name)\n",
    "# set TrainSummary and ValSummary to \n",
    "optimizer.set_train_summary(train_summary)\n",
    "optimizer.set_val_summary(val_summary)\n",
    "\n",
    "print \"Will saving logs to\",log_dir + \"/\" + app_name\n",
    "display(Markdown(\"You can run shell command, **tensorboard --logdir=%s**, to visualize this training\" % log_dir))\n",
    "\n",
    "# Boot training process\n",
    "trained_model = optimizer.optimize()\n",
    "print \"Optimization Done.\""
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## 2. Visualization"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "* During the training, you can run tensorboard command showed below to visualize the training metrics, like LearningRate curve, Loss curve, and Paramters's histogram."
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Run tensorboard command(tensorboard --logdir=/tmp/tmpvE6hzB, notice logdir will change in each run) to start tensorboard: \n",
    "\n",
    "![scalars](pictures/tb-start.png)\n",
    "\n",
    "You can add command flag `--port` to use another port, if the defaut 6006 is unavailable."
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Then, open your browser and type in TensorBoard's http address. From tensorboard's scalar tab, you can see the LearningRate curve and Loss curve like this:\n",
    "![scalars](pictures/tb-scalars.png)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Open the DISTRIBUTIONS tab, you can see the learnable parameters' distribution during the training:\n",
    "![distributions](pictures/tb-distributions.png)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "* You can also use `Model.save_graph_topology` method to save graph's topology, and visualize the graph in TensorBoard."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Saving graph topology to /tmp/tmpvE6hzB/lenet-20180110-102403/lenet-model\n"
     ]
    }
   ],
   "source": [
    "lenet_model.save_graph_topology(log_dir + \"/\" + app_name + \"/lenet-model\")\n",
    "print \"Saving graph topology to\", log_dir + \"/\" + app_name + \"/lenet-model\""
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Open the graph tab in TensorBoard, you can see the model's topology:\n",
    "![graph](pictures/tb-graph.png)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "* BigDL also provide `read_scalar()` method to read scalar type metrics, like Loss and Top1Accuracy.\n",
    "We can read training Loss, learning rate and Top1Accuracy, then draw three curves easily using matplotlib. "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "<matplotlib.text.Text at 0x7f88adb69dd0>"
      ]
     },
     "execution_count": 7,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAswAAALJCAYAAACgHHWpAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nfT0AAAAASUVORK5CYII=\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x7f88adab9090>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "loss = np.array(train_summary.read_scalar(\"Loss\"))\n",
    "lr = np.array(train_summary.read_scalar(\"LearningRate\"))\n",
    "top1 = np.array(val_summary.read_scalar(\"Top1Accuracy\"))\n",
    "\n",
    "plt.figure(figsize = (12,12))\n",
    "plt.subplot(3,1,1)\n",
    "plt.plot(loss[:,0],loss[:,1],label='loss')\n",
    "plt.xlim(0,loss.shape[0]+10)\n",
    "plt.grid(True)\n",
    "plt.title(\"training loss\")\n",
    "\n",
    "plt.subplot(3,1,2)\n",
    "plt.plot(lr[:,0],lr[:,1],label='lr')\n",
    "plt.xlim(0,lr.shape[0]+10)\n",
    "plt.title(\"learning rate\")\n",
    "plt.grid(True)\n",
    "\n",
    "plt.subplot(3,1,3)\n",
    "plt.plot(top1[:,0],top1[:,1],label='top1')\n",
    "plt.xlim(0,loss.shape[0]+10)\n",
    "plt.grid(True)\n",
    "plt.title(\"top1 accuracy\")"
   ]
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 2",
   "language": "python",
   "name": "python2"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 2
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython2",
   "version": "2.7.12"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}