TF_workshop2_hint.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Extracting ./MNIST/train-images-idx3-ubyte.gz\n",
      "Extracting ./MNIST/train-labels-idx1-ubyte.gz\n",
      "Extracting ./MNIST/t10k-images-idx3-ubyte.gz\n",
      "Extracting ./MNIST/t10k-labels-idx1-ubyte.gz\n"
     ]
    }
   ],
   "source": [
    "\"\"\"\n",
    "This is the second part of TF tutorial\n",
    "There are two tasks in this section:\n",
    "Visualize the results in tf, and save/restore models of previous training steps\n",
    "\"\"\"\n",
    "\n",
    "import tensorflow as tf\n",
    "import numpy as np\n",
    "\n",
    "from tensorflow.examples.tutorials.mnist import input_data\n",
    "file = \"./MNIST\"\n",
    "mnist = input_data.read_data_sets(file, one_hot=True)\n",
    "\n",
    "# This is the reset command to clear the current \"states\" of tensorflow\n",
    "tf.reset_default_graph()\n",
    "## Task 3\n",
    "# Visualization: See training progress with TensorBoard\n",
    "# Let's reuse the code from Part 1\n",
    "\n",
    "with tf.name_scope(\"input\"):\n",
    "    x = tf.placeholder(tf.float32, shape=[None, 784])\n",
    "    y_ = tf.placeholder(tf.float32, shape=[None, 10])\n",
    "with tf.name_scope(\"param\"):\n",
    "    W = tf.Variable(tf.zeros([784, 10]))\n",
    "    b = tf.Variable(tf.zeros([10]))\n",
    "\n",
    "with tf.name_scope(\"softmax\"):\n",
    "    y = tf.nn.softmax(tf.matmul(x, W) + b)\n",
    "    \n",
    "with tf.name_scope(\"loss\"):\n",
    "    cross_entropy = -tf.reduce_sum(y_*tf.log(y))\n",
    "    train_step = tf.train.GradientDescentOptimizer(0.01).minimize(cross_entropy)\n",
    "\n",
    "with tf.name_scope(\"summary\"):\n",
    "    tf.summary.scalar('loss', cross_entropy)\n",
    "    tf.summary.histogram('histogram', W)\n",
    "    merged = tf.summary.merge_all()\n",
    "\n",
    "with tf.Session() as sess:\n",
    "    sess.run(tf.global_variables_initializer())\n",
    "    writer = tf.summary.FileWriter('./train/plot1', sess.graph)\n",
    "\n",
    "    for i in range(300):\n",
    "        batch = mnist.train.next_batch(100)\n",
    "        summary, _ = sess.run([merged, train_step], feed_dict={x: batch[0], y_: batch[1]})\n",
    "        writer.add_summary(summary, i)\n",
    "        writer.flush()\n",
    "sess.close()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [],
   "source": [
    "### TODO: Use the iteration above, \n",
    "#   Use TensorBoard to describe the change of loss of CNN model for MNIST\n",
    "#   With LR=0.01, 0.001 respectively\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Extracting ./MNIST/train-images-idx3-ubyte.gz\n",
      "Extracting ./MNIST/train-labels-idx1-ubyte.gz\n",
      "Extracting ./MNIST/t10k-images-idx3-ubyte.gz\n",
      "Extracting ./MNIST/t10k-labels-idx1-ubyte.gz\n",
      "test accuracy 0.9091\n"
     ]
    }
   ],
   "source": [
    "### Task 4: Saving and Restoring Sessions\n",
    "\n",
    "import tensorflow as tf\n",
    "import numpy as np\n",
    "\n",
    "from tensorflow.examples.tutorials.mnist import input_data\n",
    "file = \"./MNIST\"\n",
    "mnist = input_data.read_data_sets(file, one_hot=True)\n",
    "\n",
    "# This is the reset command to clear the current \"states\" of tensorflow\n",
    "tf.reset_default_graph()\n",
    "\n",
    "with tf.name_scope(\"input\"):\n",
    "    x = tf.placeholder(tf.float32, shape=[None, 784])\n",
    "    y_ = tf.placeholder(tf.float32, shape=[None, 10])\n",
    "with tf.name_scope(\"param\"):\n",
    "    W = tf.Variable(tf.zeros([784, 10]))\n",
    "    b = tf.Variable(tf.zeros([10]))\n",
    "\n",
    "with tf.name_scope(\"softmax\"):\n",
    "    y = tf.nn.softmax(tf.matmul(x, W) + b)\n",
    "    \n",
    "with tf.name_scope(\"loss\"):\n",
    "    cross_entropy = -tf.reduce_sum(y_*tf.log(y))\n",
    "    train_step = tf.train.GradientDescentOptimizer(0.01).minimize(cross_entropy)\n",
    "saver = tf.train.Saver()\n",
    "correct_prediction = tf.equal(tf.argmax(y,1), tf.argmax(y_,1))\n",
    "accuracy = tf.reduce_mean(tf.cast(correct_prediction, tf.float32))\n",
    "\n",
    "with tf.Session() as sess:\n",
    "    sess.run(tf.global_variables_initializer())\n",
    "    writer = tf.summary.FileWriter('./train/plot1', sess.graph)\n",
    "\n",
    "    for i in range(300):\n",
    "        batch = mnist.train.next_batch(100)\n",
    "        _ = sess.run([train_step], feed_dict={x: batch[0], y_: batch[1]})\n",
    "    print(\"test accuracy %g\"%accuracy.eval(feed_dict={\n",
    "        x: mnist.test.images, y_: mnist.test.labels}))\n",
    "    saver.save(sess, \"./model.ckpt\")\n",
    "\n",
    "sess.close()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Extracting ./MNIST/train-images-idx3-ubyte.gz\n",
      "Extracting ./MNIST/train-labels-idx1-ubyte.gz\n",
      "Extracting ./MNIST/t10k-images-idx3-ubyte.gz\n",
      "Extracting ./MNIST/t10k-labels-idx1-ubyte.gz\n",
      "INFO:tensorflow:Restoring parameters from ./model.ckpt\n",
      "test accuracy 0.9091\n"
     ]
    }
   ],
   "source": [
    "import tensorflow as tf\n",
    "import numpy as np\n",
    "\n",
    "from tensorflow.examples.tutorials.mnist import input_data\n",
    "file = \"./MNIST\"\n",
    "mnist = input_data.read_data_sets(file, one_hot=True)\n",
    "\n",
    "# This is the reset command to clear the current \"states\" of tensorflow\n",
    "tf.reset_default_graph()\n",
    "\n",
    "with tf.name_scope(\"input\"):\n",
    "    x = tf.placeholder(tf.float32, shape=[None, 784])\n",
    "    y_ = tf.placeholder(tf.float32, shape=[None, 10])\n",
    "with tf.name_scope(\"param\"):\n",
    "    W = tf.Variable(tf.zeros([784, 10]))\n",
    "    b = tf.Variable(tf.zeros([10]))\n",
    "\n",
    "with tf.name_scope(\"softmax\"):\n",
    "    y = tf.nn.softmax(tf.matmul(x, W) + b)\n",
    "    \n",
    "with tf.name_scope(\"loss\"):\n",
    "    cross_entropy = -tf.reduce_sum(y_*tf.log(y))\n",
    "    train_step = tf.train.GradientDescentOptimizer(0.01).minimize(cross_entropy)\n",
    "saver = tf.train.Saver()\n",
    "correct_prediction = tf.equal(tf.argmax(y,1), tf.argmax(y_,1))\n",
    "accuracy = tf.reduce_mean(tf.cast(correct_prediction, tf.float32))\n",
    "with tf.Session() as sess:\n",
    "## TODO: Use the previously saved model to find the accuracy on test set\n",
    "#  On previous model, save the result on 50, 100, 150 steps respectively\n",
    "#  Ref:\n",
    "##         saver.restore(sess, \"./model.ckpt\")\n",
    "    saver.restore(sess, \"./model.ckpt\")\n",
    "    print(\"test accuracy %g\"%accuracy.eval(feed_dict={\n",
    "        x: mnist.test.images, y_: mnist.test.labels}))\n",
    "sess.close()"
   ]
  },
  {
   "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": "3.6.4"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}