TF_Distributed_MNIST.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Extracting ./train-images-idx3-ubyte.gz\n",
      "Extracting ./train-labels-idx1-ubyte.gz\n",
      "Extracting ./t10k-images-idx3-ubyte.gz\n",
      "Extracting ./t10k-labels-idx1-ubyte.gz\n"
     ]
    }
   ],
   "source": [
    "from __future__ import print_function\n",
    "\n",
    "import numpy as np\n",
    "import tensorflow as tf\n",
    "import time\n",
    "\n",
    "from tensorflow.examples.tutorials.mnist import input_data\n",
    "mnist = input_data.read_data_sets(\"./\", one_hot=True)\n",
    "\n",
    "learning_rate = 0.001\n",
    "num_input = 784 # MNIST data input (img shape: 28*28)\n",
    "num_classes = 10 # MNIST total classes (0-9 digits)\n",
    "dropout = 0.75 # Dropout, probability to keep units"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "def conv_net(x, n_classes, dropout, reuse, is_training):\n",
    "    with tf.variable_scope('ConvNet', reuse=reuse):\n",
    "        x = tf.reshape(x, shape=[-1, 28, 28, 1])\n",
    "        x = tf.layers.conv2d(x, 64, 5, activation=tf.nn.relu)\n",
    "        x = tf.layers.max_pooling2d(x, 2, 2)\n",
    "        x = tf.layers.conv2d(x, 256, 3, activation=tf.nn.relu)\n",
    "        x = tf.layers.conv2d(x, 512, 3, activation=tf.nn.relu)\n",
    "        x = tf.layers.max_pooling2d(x, 2, 2)\n",
    "        x = tf.contrib.layers.flatten(x)\n",
    "        x = tf.layers.dense(x, 2048)\n",
    "        x = tf.layers.dropout(x, rate=dropout, training=is_training)\n",
    "        x = tf.layers.dense(x, 1024)\n",
    "        x = tf.layers.dropout(x, rate=dropout, training=is_training)\n",
    "        out = tf.layers.dense(x, n_classes)\n",
    "        out = tf.nn.softmax(out) if not is_training else out\n",
    "    return out"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "def average_gradients(tower_grads):\n",
    "    average_grads = []\n",
    "    for grad_and_vars in zip(*tower_grads):\n",
    "        grads = []\n",
    "        for g, _ in grad_and_vars:\n",
    "            expanded_g = tf.expand_dims(g, 0)\n",
    "            grads.append(expanded_g)\n",
    "        grad = tf.concat(grads, 0)\n",
    "        grad = tf.reduce_mean(grad, 0)\n",
    "        v = grad_and_vars[0][1]\n",
    "        grad_and_var = (grad, v)\n",
    "        average_grads.append(grad_and_var)\n",
    "    return average_grads"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "PS_OPS = ['Variable', 'VariableV2', 'AutoReloadVariable']\n",
    "def assign_to_device(device, ps_device='/cpu:0'):\n",
    "    def _assign(op):\n",
    "        node_def = op if isinstance(op, tf.NodeDef) else op.node_def\n",
    "        if node_def.op in PS_OPS:\n",
    "            return \"/\" + ps_device\n",
    "        else:\n",
    "            return device\n",
    "    return _assign"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 66,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Training with 1 GPUs\n",
      "Step 1, Training Accuracy= 0.165\n",
      "Step 80, Training Accuracy= 0.993\n",
      "Step 160, Training Accuracy= 0.998\n",
      "Expected Training time for 10 epoches is 64.75 seconds with 1 GPUs\n",
      "Training with 2 GPUs\n",
      "Step 1, Training Accuracy= 0.458\n",
      "Step 80, Training Accuracy= 0.995\n",
      "Step 160, Training Accuracy= 0.999\n",
      "Expected Training time for 10 epoches is 57.49 seconds with 2 GPUs\n",
      "Training with 3 GPUs\n",
      "Step 1, Training Accuracy= 0.290\n",
      "Step 80, Training Accuracy= 0.996\n",
      "Step 160, Training Accuracy= 1.000\n",
      "Expected Training time for 10 epoches is 52.17 seconds with 3 GPUs\n",
      "Training with 4 GPUs\n",
      "Step 1, Training Accuracy= 0.259\n",
      "Step 80, Training Accuracy= 0.996\n",
      "Step 160, Training Accuracy= 1.000\n",
      "Expected Training time for 10 epoches is 49.59 seconds with 4 GPUs\n"
     ]
    }
   ],
   "source": [
    "num_steps = 160\n",
    "batch_size = 4*1024\n",
    "display_step = 80\n",
    "gpus=[1,2,3,4]\n",
    "t=[]\n",
    "for num_gpus in gpus:\n",
    "    tf.reset_default_graph()\n",
    "    print(\"Training with %i GPUs\" % num_gpus)\n",
    "    with tf.device('/cpu:0'):\n",
    "        tower_grads = []\n",
    "        reuse_vars = False\n",
    "        X = tf.placeholder(tf.float32, [None, num_input])\n",
    "        Y = tf.placeholder(tf.float32, [None, num_classes])\n",
    "        for i in range(num_gpus):\n",
    "            with tf.device(assign_to_device('/gpu:{}'.format(i), ps_device='/cpu:0')):\n",
    "                _x = X[i * batch_size: (i+1) * batch_size]\n",
    "                _y = Y[i * batch_size: (i+1) * batch_size]\n",
    "                logits_train = conv_net(_x, num_classes, dropout,reuse=reuse_vars, is_training=True)\n",
    "                logits_test = conv_net(_x, num_classes, dropout,reuse=True, is_training=False)\n",
    "                loss_op = tf.reduce_mean(tf.nn.softmax_cross_entropy_with_logits(\n",
    "                    logits=logits_train, labels=_y))\n",
    "                optimizer = tf.train.AdamOptimizer(learning_rate=learning_rate)\n",
    "                grads = optimizer.compute_gradients(loss_op)\n",
    "                if i == 0:\n",
    "                    correct_pred = tf.equal(tf.argmax(logits_test, 1), tf.argmax(_y, 1))\n",
    "                    accuracy = tf.reduce_mean(tf.cast(correct_pred, tf.float32))\n",
    "                reuse_vars = True\n",
    "                tower_grads.append(grads)\n",
    "        tower_grads = average_gradients(tower_grads)\n",
    "        train_op = optimizer.apply_gradients(tower_grads)\n",
    "        init = tf.global_variables_initializer()\n",
    "        with tf.Session() as sess:\n",
    "            sess.run(init)\n",
    "            step = 1\n",
    "            ts = time.time()\n",
    "            for step in range(1, num_steps + 1):\n",
    "                batch_x, batch_y = mnist.train.next_batch(batch_size * num_gpus)\n",
    "                sess.run(train_op, feed_dict={X: batch_x, Y: batch_y})\n",
    "                if step % display_step == 0 or step == 1:\n",
    "                    loss, acc = sess.run([loss_op, accuracy], feed_dict={X: batch_x,Y: batch_y})\n",
    "                    print(\"Step \" + str(step) + \", Training Accuracy= \" + \"{:.3f}\".format(acc)) \n",
    "                step += 1\n",
    "        t.append((time.time() - ts)*10*mnist.train.num_examples/(batch_x.shape[0]*num_steps))\n",
    "        print(\"Expected Training time for 10 epoches is %.2f seconds with %d GPUs\"%(t[-1],num_gpus))\n",
    "                        "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 68,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAYIAAAEKCAYAAAAfGVI8AAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\njPFx/w9awqbgBxg4CwAAAABJRU5ErkJggg==\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x3bffffd4efd0>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "import matplotlib.pyplot as plt\n",
    "plt.plot(gpus, t, label='Epochs:10, Batch:%d' % batch_size)\n",
    "plt.xlabel(\"Number of GPUs\")\n",
    "plt.ylabel(\"Training Time in Seconds\")\n",
    "plt.xticks(np.arange(1,5, 1))\n",
    "plt.legend()\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": []
  }
 ],
 "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": "2.7.14"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}