notebooks/sequence_notebook.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "import tensorflow as tf\n",
    "import numpy as np\n",
    "import sys\n",
    "import os\n",
    "import matplotlib.pyplot as plt\n",
    "\n",
    "sys.path.append(os.path.abspath(\"/notebooks\"))\n",
    "import sq_net"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Our lstm cells have an internal size - hidden size\n",
    "And a size that matches the input    - input size\n",
    "\n",
    "We can change the cell() defintion to try out other cell types"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "hidden_size = 32\n",
    "input_size = 35\n",
    "output_size = 26\n",
    "layer_count = 3\n",
    "train_steps = 120\n",
    "batch_size = 16\n",
    "input_count = 4\n",
    "\n",
    "#beta is the regularization\n",
    "beta_val = 0.0002\n",
    "beta = tf.Variable(beta_val, trainable=False)\n",
    "train_length = train_steps - input_size - 1\n",
    "\n",
    "cell_name = \"gru_cell\"\n",
    "cell_kernel = \"{0}/gates\".format(cell_name)\n",
    "\n",
    "global_dropout = tf.Variable(0.9, trainable=False)\n",
    "\n",
    "# gru_cell/gates, basic_lstm_cell\n",
    "def cell():\n",
    "    lcell = tf.nn.rnn_cell.GRUCell(hidden_size)\n",
    "    return tf.nn.rnn_cell.DropoutWrapper(lcell, input_keep_prob=global_dropout)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Set up the network and a trainer"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "loss, cost, logits, save_dict = sq_net.sequence_layer(train_steps, batch_size, input_size, output_size, \\\n",
    "                        input_count, layer_count, hidden_size, cell, cell_kernel, beta, global_dropout, None)\n",
    "\n",
    "inputs = save_dict['inputs']"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "train_op, grads = sq_net.adam_train(cost)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Import the AirQuality data from the csv file"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "filename_queue = tf.train.string_input_producer([\"data/AirQualityUCI.csv\"])\n",
    "\n",
    "reader = tf.TextLineReader(1)\n",
    "key, value = reader.read_up_to(filename_queue, 8000)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "# Date1;Time2;3CO(GT);4PT08.S1(CO);5NMHC(GT);6C6H6(GT);7PT08.S2(NMHC);8NOx(GT);9PT08.S3(NOx);10NO2(GT);11PT08.S4(NO2);12PT08.S5(O3);12T;14RH;15AH;;\n",
    "record_defaults = [[\"0\"],[\"0\"],[\"0\"],[\"0\"],[\"0\"],[\"0\"],[\"0\"],[\"0\"],[\"0\"],[\"0\"],[\"0\"],[\"0\"],[\"0\"],[\"0\"],[\"0\"],[\"0\"],[\"0\"]]\n",
    "\n",
    "cols = tf.decode_csv(value, record_defaults=record_defaults, field_delim=\";\")\n",
    "\n",
    "ptco = tf.string_to_number(cols[3])\n",
    "ptco_max = tf.reduce_max(ptco)\n",
    "ptco_norm = ptco / ptco_max\n",
    "\n",
    "nmhc = tf.string_to_number(cols[4])\n",
    "nmhc_max = tf.reduce_max(nmhc)\n",
    "nmhc_norm = nmhc / nmhc_max\n",
    "\n",
    "# c6h6 = cols[5]\n",
    "\n",
    "pts2 = tf.string_to_number(cols[6])\n",
    "pts2_max = tf.reduce_max(pts2)\n",
    "pts2_norm = pts2 / pts2_max\n",
    "\n",
    "nox = tf.string_to_number(cols[7])\n",
    "nox_max = tf.reduce_max(nox)\n",
    "nox_norm = nox / nox_max\n",
    "\n",
    "no2 = tf.string_to_number(cols[9])\n",
    "no2_max = tf.reduce_max(no2)\n",
    "no2_norm = no2 / no2_max\n",
    "\n",
    "o3 = tf.string_to_number(cols[11])\n",
    "o3_max = tf.reduce_max(o3)\n",
    "o3_norm = o3 / o3_max"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Read the data using a queue runner"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "with tf.Session() as sess:\n",
    "    coord = tf.train.Coordinator()\n",
    "    threads = tf.train.start_queue_runners(coord=coord)\n",
    "\n",
    "    full_data, ptco_norm_data, pts2_norm_data, nmhc_norm_data, \\\n",
    "        nox_data, nox_norm_data, no2_data, no2_norm_data, o3_data, o3_norm_data = \\\n",
    "        sess.run([cols, ptco_norm, pts2_norm, nmhc_norm, nox, nox_norm, no2, no2_norm, o3, o3_norm])\n",
    "        \n",
    "    coord.request_stop()\n",
    "    coord.join(threads)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "We can choose different gases to train with and target"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "training_set_a = pts2_norm_data[:8000]\n",
    "training_set_b = ptco_norm_data[:8000]\n",
    "training_set_c = nox_norm_data[:8000]\n",
    "training_set_d = no2_norm_data[:8000]\n",
    "\n",
    "target_set = o3_norm_data[:8000]"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Our data feed will adjust based on the data we set with data_in"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "data_in = [training_set_a, training_set_b, training_set_c, training_set_d]"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Lets look at some of the data"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAX8AAAD8CAYAAACfF6SlAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x7f0adbb00110>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "plt.plot(target_set, 'b')\n",
    "plt.plot(training_set_d, 'r')\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Grab a bunch of data for training/testing"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "def feed_data(iterations, steps):\n",
    "    \n",
    "    data_out = []\n",
    "    for dataset in data_in:\n",
    "        data_out.append([])\n",
    "    \n",
    "    targets = []\n",
    "    \n",
    "    for iteration in iterations:\n",
    "        data_t = []\n",
    "        \n",
    "        d_out = []\n",
    "        for dataset in data_out:\n",
    "            d_out.append([])\n",
    "        \n",
    "        offset = iteration * train_length\n",
    "\n",
    "        for i in range(steps):\n",
    "            for idx, dataset in enumerate(data_in):\n",
    "                data = d_out[idx]\n",
    "                input_seq = dataset[offset + i : offset + i + input_size]\n",
    "                data.append(input_seq)\n",
    "\n",
    "            target_seq = target_set[offset + i + input_size : offset + i + input_size + output_size]\n",
    "            data_t.append(target_seq)\n",
    "            \n",
    "        for idx, dataset in enumerate(data_out):\n",
    "            dataset.append(d_out[idx])\n",
    "            \n",
    "        targets.append(data_t)\n",
    "\n",
    "    return (offset, data_out, targets)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Set a few things up for the session"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "from random import sample\n",
    "\n",
    "train_range = range(50)\n",
    "total_cost = 0\n",
    "last_loss = 0\n",
    "\n",
    "currentSession = tf.Session()\n",
    "\n",
    "merged = tf.summary.merge_all()\n",
    "train_writer = tf.summary.FileWriter('summaries/train/b', currentSession.graph)\n",
    "\n",
    "tf.global_variables_initializer().run(session = currentSession)\n",
    "\n",
    "starting_epoch = 0\n",
    "\n",
    "progress = 0"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Play with different learning rates"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "def learning_schedule(step):\n",
    "    flr = 0.005\n",
    "    if (step > 6000):\n",
    "        flr = 0.001\n",
    "    if (step > 8000):\n",
    "        flr = 0.0008\n",
    "    if (step > 14000):\n",
    "        flr = 0.0006\n",
    "    if (step > 16000):\n",
    "        flr = 0.0002\n",
    "    if (step > 21000):\n",
    "        flr = 0.00002\n",
    "    return flr"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "def validate(validation_point):\n",
    "    no_drop = tf.assign(global_dropout, 1.0)\n",
    "    drop = currentSession.run(no_drop)\n",
    "\n",
    "    test_offset = [] \n",
    "    #The network still wants batches at the moment\n",
    "    for i in range(batch_size):\n",
    "        test_offset.append(validation_point)\n",
    "        \n",
    "    offset, data_out, targets = feed_data(test_offset, train_steps)\n",
    "\n",
    "    feed_dict={'targets:0': targets}\n",
    "    for idx in range(input_count):\n",
    "        feed_dict.update({inputs[idx].name: data_out[idx]})\n",
    "\n",
    "    test_logits, current_loss, fstate = currentSession.run([logits, loss, save_dict['final_state']], feed_dict)\n",
    "\n",
    "    distance = 25\n",
    "    #predictions at different distances from the input\n",
    "    far_away = test_logits[0][:,distance]\n",
    "    mid = test_logits[0][:,7]\n",
    "    close = test_logits[0][:,1]\n",
    "\n",
    "    test_size = input_size + distance\n",
    "    offset, data_out, targets = feed_data(test_offset, train_steps)\n",
    "    ts = np.stack(targets[0])\n",
    "    ts = np.array(targets[0])[:,0]\n",
    "    \n",
    "    return current_loss, far_away, ts"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Start training - choose a dropout rate"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "metadata": {},
   "outputs": [],
   "source": [
    "drop_val = 0.8\n",
    "\n",
    "dropit = tf.assign(global_dropout, drop_val)\n",
    "\n",
    "drop = currentSession.run(dropit)\n",
    "\n",
    "for epoch in range(starting_epoch, 395):\n",
    "    starting_epoch = epoch\n",
    "    \n",
    "    iterations = []\n",
    "    \n",
    "    train_ranges = [] \n",
    "    \n",
    "    epoch_loss = 0\n",
    "    \n",
    "    for i in range(batch_size):\n",
    "        train_ranges.append(sample(train_range, 50))\n",
    "        \n",
    "    batch_offsets = np.reshape(train_ranges, (-1, batch_size))\n",
    "    \n",
    "    for offsets in batch_offsets:\n",
    "        offset, data_out, targets = feed_data(offsets, train_steps)\n",
    "        \n",
    "        flr = learning_schedule(progress)\n",
    "        \n",
    "        feed = {'targets:0': targets, 'learning_rate:0': flr, dropit: drop_val}\n",
    "        \n",
    "        for idx in range(input_count):\n",
    "            feed.update({inputs[idx].name: data_out[idx]})\n",
    "        \n",
    "        summary, partial_loss, train, fstate = currentSession.run([merged, loss, train_op, save_dict['final_state']], \\\n",
    "                                                                  feed_dict=feed)\n",
    "        \n",
    "        train_writer.add_summary(summary, progress)\n",
    "        progress = progress + 1\n",
    "        #allow a little time for initial variation to die down\n",
    "        if epoch > 0:\n",
    "            epoch_loss = epoch_loss + partial_loss\n",
    "        \n",
    "    total_cost = epoch_loss + total_cost\n",
    "    \n",
    "    if epoch % 30 == 0:\n",
    "        running_loss = total_cost - last_loss\n",
    "        last_loss = total_cost\n",
    "        av_loss = epoch_loss / 50\n",
    "        print(hidden_size, batch_size, beta_val, epoch_loss, epoch, running_loss, flr, av_loss, partial_loss)\n",
    "        \n",
    "        indices = 12\n",
    "        val_range = sample(xrange(20), indices)\n",
    "        validate_average = 0\n",
    "        \n",
    "        for value in val_range:\n",
    "            validate_offset = value + 50\n",
    "            validations, _, _ = validate(validate_offset)\n",
    "            validate_average = validate_average + validations\n",
    "\n",
    "        no_drop = tf.assign(global_dropout, 0.8)\n",
    "        drop = currentSession.run(no_drop)\n",
    "            \n",
    "        print(validate_average / indices, val_range) "
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "(22, 16, 0.0002, 0.3133568037301302, 300, 9.8163413042202592, 0.0006, 0.006267136074602604, 0.0056910324)\n",
    "(0.020415419712662697, [0, 8, 19, 18, 9, 15, 14, 5, 4, 6, 17, 12])\n",
    "(22, 16, 0.0002, 0.30690161604434252, 330, 9.335203982423991, 0.0002, 0.0061380323208868499, 0.0060125538)\n",
    "(0.024062525869036715, [10, 6, 12, 17, 14, 13, 16, 18, 2, 5, 11, 8])\n",
    "(22, 16, 0.0002, 0.30609454261139035, 360, 9.1924056150019169, 0.0002, 0.006121890852227807, 0.0058105011)\n",
    "(0.022401178876558941, [12, 8, 2, 1, 11, 6, 14, 15, 13, 4, 10, 9])\n",
    "(22, 16, 0.0002, 0.30363314179703593, 390, 9.1560970405116677, 0.0002, 0.0060726628359407189, 0.0057852077)\n",
    "(0.024297634682928521, [19, 11, 8, 1, 0, 13, 14, 15, 12, 7, 3, 4])\n",
    "\n",
    "Results from previous run\n",
    "(22, 4, 0.012, 0.75928797852247953, 30, 28.130740600638092, 0.001, 0.01518575957044959, 0.025324259)\n",
    "(0.01991733970741431, [3, 5, 11, 6, 13, 16, 7, 2, 17, 10, 9, 1])\n",
    "(22, 4, 0.012, 0.56991863530129194, 60, 19.984842267818749, 0.001, 0.011398372706025839, 0.0092041623)\n",
    "(0.019945473720630009, [16, 10, 5, 1, 8, 11, 9, 6, 2, 3, 19, 15])\n",
    "(22, 4, 0.012, 0.53841507947072387, 90, 16.683391402009875, 0.001, 0.010768301589414477, 0.012398911)\n",
    "(0.020400462982555229, [9, 12, 16, 17, 7, 14, 4, 0, 8, 5, 15, 2])\n",
    "(22, 4, 0.012, 0.35388133488595486, 120, 12.493187983520329, 0.001, 0.007077626697719097, 0.011706958)\n",
    "(0.022565561967591446, [0, 17, 3, 6, 4, 7, 15, 19, 5, 8, 16, 18])\n",
    "(22, 4, 0.012, 0.31174778053537011, 150, 11.304671347606927, 0.001, 0.0062349556107074025, 0.0073217414)\n",
    "(0.021213628894959886, [6, 19, 4, 2, 15, 5, 0, 17, 16, 12, 9, 7])\n",
    "(22, 4, 0.012, 0.54646076308563352, 210, 18.03114574495703, 0.0008, 0.01092921526171267, 0.0089054992)\n",
    "(0.018522994980836909, [10, 12, 18, 9, 5, 2, 11, 13, 15, 4, 3, 7])\n",
    "\n",
    "\n",
    "*******\n",
    "(22, 8, 0.016, 0, 0, 0, 0.001, 0, 0.026430946)\n",
    "(0.032205939525738358, [15, 11, 1, 12, 8, 6, 9, 3, 2, 18, 4, 13])\n",
    "(22, 8, 0.016, 0.77701491117477417, 30, 27.029327136464417, 0.001, 0.015540298223495484, 0.012644939)\n",
    "(0.022979985301693279, [13, 5, 0, 14, 12, 1, 4, 7, 11, 2, 17, 6])\n",
    "(22, 8, 0.016, 0.73235377483069897, 60, 23.532660925760865, 0.001, 0.01464707549661398, 0.015665311)\n",
    "(0.019436083035543561, [3, 2, 6, 7, 12, 5, 15, 17, 9, 0, 10, 11])\n",
    "(22, 8, 0.016, 0.6555849714204669, 90, 20.78133162483573, 0.001, 0.013111699428409338, 0.012796191)\n",
    "(0.020664084004238248, [14, 7, 9, 8, 6, 4, 3, 12, 16, 5, 13, 18])\n",
    "(22, 8, 0.016, 0.72552435472607613, 120, 20.182675201445818, 0.001, 0.014510487094521523, 0.010681516)\n",
    "(0.019372141221538186, [11, 0, 6, 8, 17, 16, 14, 10, 5, 7, 2, 3])\n",
    "(22, 8, 0.016, 0.57952790893614292, 150, 18.611567541956902, 0.001, 0.011590558178722858, 0.012818389)\n",
    "(0.021197795247038204, [13, 19, 4, 7, 18, 10, 0, 1, 11, 9, 12, 15])\n",
    "(22, 8, 0.016, 0.59323347639292479, 180, 17.638431374449283, 0.001, 0.011864669527858496, 0.013732762)\n",
    "(0.022313984188561637, [15, 7, 0, 10, 16, 13, 8, 4, 2, 14, 12, 5])\n",
    "(22, 8, 0.016, 0.55298889242112637, 210, 17.036189836449921, 0.0008, 0.011059777848422526, 0.0091538271)\n",
    "(0.023156357385839026, [14, 19, 13, 0, 8, 3, 18, 7, 1, 11, 16, 5])\n",
    "(22, 8, 0.016, 0.50631409976631403, 240, 15.846848386339843, 0.0008, 0.010126281995326281, 0.0092495885)\n",
    "(0.020125681301578879, [0, 12, 16, 14, 19, 4, 1, 5, 7, 3, 9, 15])\n",
    "(22, 8, 0.016, 0.70553555712103844, 270, 17.684835837222636, 0.0008, 0.014110711142420769, 0.012239807)\n",
    "(0.018002962383131187, [7, 4, 12, 19, 6, 2, 11, 9, 8, 1, 10, 5])\n",
    "(22, 8, 0.016, 0.65702436957508326, 300, 21.372702517546713, 0.0008, 0.013140487391501665, 0.01766105)\n",
    "(0.027398040673385065, [8, 18, 12, 7, 1, 19, 2, 9, 14, 6, 0, 4])\n",
    "(22, 8, 0.016, 0.60001945402473211, 330, 19.56549764610827, 0.0008, 0.012000389080494642, 0.012756304)\n",
    "(0.023783902171999216, [12, 14, 1, 4, 18, 17, 3, 9, 10, 0, 8, 2])\n",
    "(22, 8, 0.016, 0.76574821677058935, 360, 17.596865852829069, 0.0008, 0.015314964335411787, 0.015783155)\n",
    "(0.023949965989838045, [14, 17, 13, 3, 6, 12, 4, 16, 8, 15, 0, 7])\n",
    "(22, 8, 0.016, 0.6361575610935688, 390, 20.046616850420833, 0.0006, 0.012723151221871376, 0.012433306)\n",
    "(0.023572704754769802, [5, 0, 9, 6, 14, 7, 18, 8, 1, 16, 12, 10])"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "It's simple to train and restart"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "'checkpoints/gru_cell_394'"
      ]
     },
     "execution_count": 19,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "saver = tf.train.Saver()\n",
    "saver.save(currentSession, 'checkpoints/{0}_{1}'.format(cell_name, epoch))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Without having to create a new network we can take a look at how things are going"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "We can plot the predictions at different prediction points"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 32,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "0.0160164\n"
     ]
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAX0AAAD8CAYAAACb4nSYAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nBUEQahEi+oIgCLUIEX1BEIRaxP8D9WMfHZ6IGt4AAAAASUVORK5CYII=\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x7f0ad9746810>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "validations, far_away, ts = validate(57)\n",
    "\n",
    "print(validations)\n",
    "\n",
    "import matplotlib.pyplot as plt\n",
    "plt.plot(far_away, 'r')\n",
    "# plt.plot(mid[18:], 'y')\n",
    "# plt.plot(close[25:], 'g')\n",
    "plt.plot(ts, 'b')\n",
    "plt.axis([0.0,train_steps,0.0,1.02])\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Some example results at offset 57\n",
    "0.0123539 (22, 4, 0.012, 0.54646076308563352, 210, 18.03114574495703, 0.0008, 0.01092921526171267, 0.0089054992)\n",
    "(16, 0.007, 'offset', [57, 57, 57, 57], 4788, 26, 0.014154767, 126)\n",
    "(16, 0.006, 'offset', [57, 57, 57, 57], 4788, 26, 0.015169268, 122)\n",
    "(16, 0.006, 'offset', [57, 57, 57, 57], 4788, 26, 0.015310187, 98)\n",
    "(16, 0.004, 'offset', [57, 57, 57, 57], 4788, 26, 0.01529764, 66)\n",
    "(16, 0.002, 'offset', [57, 57, 57, 57], 4788, 26, 0.017894622, 195)\n",
    "(16, 0.002, 'offset', [57, 57, 57, 57], 4788, 26, 0.01718206, 160)\n",
    "(16, 0.002, 'offset', [57, 57, 57, 57], 4788, 26, 0.019261345, 122)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 21,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "'checkpoints/predict16_394'"
      ]
     },
     "execution_count": 21,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "saver = tf.train.Saver()\n",
    "saver.save(currentSession, 'checkpoints/predict{0}_{1}'.format(batch_size, epoch))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "We're now going to save the batching network and create a different network with batch size 1 and 1 training step which we can use for making predictions from a stream"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 22,
   "metadata": {
    "collapsed": true
   },
   "outputs": [],
   "source": [
    "p_batch = 1\n",
    "    \n",
    "p_loss, p_cost, p_logits, p_save_dict = sq_net.sequence_layer(1, p_batch, input_size, output_size, \\\n",
    "                       input_count, layer_count, hidden_size, cell, cell_kernel, beta, False, True)\n",
    "\n",
    "p_inputs = p_save_dict['inputs']"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 23,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "INFO:tensorflow:Restoring parameters from checkpoints/predict16_394\n"
     ]
    }
   ],
   "source": [
    "sess = tf.Session()  \n",
    "sess.run(tf.global_variables_initializer())\n",
    "saver = tf.train.Saver()\n",
    "saver.restore(sess, 'checkpoints/predict{0}_{1}'.format(batch_size, epoch))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 24,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "u'input_0_1:0'"
      ]
     },
     "execution_count": 24,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "p_inputs[0].name"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 25,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "[<tf.Tensor 'input_0_1:0' shape=(1, 1, 35) dtype=float32>,\n",
       " <tf.Tensor 'input_1_1:0' shape=(1, 1, 35) dtype=float32>,\n",
       " <tf.Tensor 'input_2_1:0' shape=(1, 1, 35) dtype=float32>,\n",
       " <tf.Tensor 'input_3_1:0' shape=(1, 1, 35) dtype=float32>]"
      ]
     },
     "execution_count": 25,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "p_inputs"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Choose test_offset for in training or out of training predictions"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 28,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "(0,)\n"
     ]
    }
   ],
   "source": [
    "test_offset = [51] \n",
    "\n",
    "offset, input_data, targets = feed_data(test_offset, 420)\n",
    "\n",
    "predictions = []\n",
    "targets_o = []\n",
    "\n",
    "print(np.shape(targets_o))\n",
    "\n",
    "state = None\n",
    "\n",
    "for i in range(400):\n",
    "    feed_dict={}\n",
    "    \n",
    "    for idx in range(input_count):\n",
    "        new_data = np.reshape(input_data[idx][0][i], (1,1,35))\n",
    "        feed_dict.update({p_inputs[idx].name: new_data})\n",
    "\n",
    "    if state is not None:\n",
    "        for idx in range(input_count):\n",
    "            feed_dict.update({p_save_dict['inital_state'][idx]: state[idx]})\n",
    "        \n",
    "    run_logits, state = sess.run([p_logits, p_save_dict['final_state']], feed_dict)\n",
    "    predictions.append(run_logits[0][0][output_size-1])\n",
    "    targets_o.append(targets[0][i][output_size-1])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 27,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAXcAAAD8CAYAAACMwORRAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nlbvFYrHkIf8Pbn9Tl7fsWPsAAAAASUVORK5CYII=\n",
      "text/plain": [
       "<matplotlib.figure.Figure at 0x7f0ab8143d90>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "plt.plot(predictions, 'r')\n",
    "plt.plot(targets_o, 'b')\n",
    "plt.axis([0.0,420,0.0,1.02])\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": "2.7.12"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}