aboutsummaryrefslogtreecommitdiffhomepage
path: root/tensorflow/contrib/autograph/examples/notebooks/graph_vs_ag_vs_eager_sum_speed_test.ipynb
diff options
context:
space:
mode:
Diffstat (limited to 'tensorflow/contrib/autograph/examples/notebooks/graph_vs_ag_vs_eager_sum_speed_test.ipynb')
-rw-r--r--tensorflow/contrib/autograph/examples/notebooks/graph_vs_ag_vs_eager_sum_speed_test.ipynb519
1 files changed, 519 insertions, 0 deletions
diff --git a/tensorflow/contrib/autograph/examples/notebooks/graph_vs_ag_vs_eager_sum_speed_test.ipynb b/tensorflow/contrib/autograph/examples/notebooks/graph_vs_ag_vs_eager_sum_speed_test.ipynb
new file mode 100644
index 0000000000..32742bec7e
--- /dev/null
+++ b/tensorflow/contrib/autograph/examples/notebooks/graph_vs_ag_vs_eager_sum_speed_test.ipynb
@@ -0,0 +1,519 @@
+{
+ "cells": [
+ {
+ "cell_type": "markdown",
+ "metadata": {
+ "colab_type": "text",
+ "id": "moMkWaT_TTHi"
+ },
+ "source": [
+ "This Colab illustrates the differing overhead* between a custom, vectorized graph operation and a loop over a tensor\n",
+ "that computes the same function. The loop is implemented in TensorFlow Eager mode using Python syntax and control-flow, and using AutoGraph which takes a python function and converts it into graph mode. In AutoGraph the Python loop is converted into a tf.while_loop.\n",
+ "\n",
+ "The actual computation, summing a small number of scalar values, takes very little time to compute, so the graphs below are showing the overhead of the differing approaches. As such, this is more of a \"micro-benchmark\" than a representation of real-world performance of the three approaches.\n",
+ "\n",
+ "*Note the differing scales of the included plots"
+ ]
+ },
+ {
+ "cell_type": "markdown",
+ "metadata": {
+ "colab_type": "text",
+ "id": "a0X_rfvuav98"
+ },
+ "source": [
+ "### Imports"
+ ]
+ },
+ {
+ "cell_type": "code",
+ "execution_count": 0,
+ "metadata": {
+ "colab": {
+ "autoexec": {
+ "startup": false,
+ "wait_interval": 0
+ }
+ },
+ "colab_type": "code",
+ "id": "EdxWv4Vn0ync"
+ },
+ "outputs": [],
+ "source": [
+ "!pip install -U -q tf-nightly"
+ ]
+ },
+ {
+ "cell_type": "code",
+ "execution_count": 0,
+ "metadata": {
+ "colab": {
+ "autoexec": {
+ "startup": false,
+ "wait_interval": 0
+ }
+ },
+ "colab_type": "code",
+ "id": "erq3_S7QsjkU"
+ },
+ "outputs": [],
+ "source": [
+ "from __future__ import absolute_import\n",
+ "from __future__ import division\n",
+ "from __future__ import print_function\n",
+ "\n",
+ "import numpy as np\n",
+ "import tensorflow as tf\n",
+ "import matplotlib.pyplot as plt\n",
+ "import math\n",
+ "import time\n",
+ "import random\n",
+ "from colabtools import adhoc_import\n",
+ "from tensorflow.contrib import autograph as ag\n",
+ "from tensorflow.python.framework import function"
+ ]
+ },
+ {
+ "cell_type": "markdown",
+ "metadata": {
+ "colab_type": "text",
+ "id": "1JgnsXooa2RP"
+ },
+ "source": [
+ "### Testing boilerplate"
+ ]
+ },
+ {
+ "cell_type": "code",
+ "execution_count": 0,
+ "metadata": {
+ "colab": {
+ "autoexec": {
+ "startup": false,
+ "wait_interval": 0
+ }
+ },
+ "colab_type": "code",
+ "id": "UyD5LLjVZzny"
+ },
+ "outputs": [],
+ "source": [
+ "# Test-only parameters. Test checks successful completion not correctness. \n",
+ "burn_ins = 1\n",
+ "trials = 1\n",
+ "batches = 2\n",
+ "max_elements = 2"
+ ]
+ },
+ {
+ "cell_type": "markdown",
+ "metadata": {
+ "colab_type": "text",
+ "id": "4_NBL0RQa8gY"
+ },
+ "source": [
+ "### Speed comparison parameters"
+ ]
+ },
+ {
+ "cell_type": "code",
+ "execution_count": 0,
+ "metadata": {
+ "colab": {
+ "autoexec": {
+ "startup": false,
+ "wait_interval": 0
+ }
+ },
+ "colab_type": "code",
+ "id": "Yq6daecyiJV5"
+ },
+ "outputs": [],
+ "source": [
+ "#@test {\"skip\": true} \n",
+ "burn_ins = 3 # Batches not counted in the average\n",
+ "trials = 10 # Batches run per vector-size (and averaged)\n",
+ "batches = 1000 # Number of random vectors summed over per trial\n",
+ "max_elements = 100 # Vectors of size 0 to this-1 will be executed and plotted"
+ ]
+ },
+ {
+ "cell_type": "markdown",
+ "metadata": {
+ "colab_type": "text",
+ "id": "fiR8m13CbKH2"
+ },
+ "source": [
+ "### Random input"
+ ]
+ },
+ {
+ "cell_type": "code",
+ "execution_count": 0,
+ "metadata": {
+ "colab": {
+ "autoexec": {
+ "startup": false,
+ "wait_interval": 0
+ }
+ },
+ "colab_type": "code",
+ "id": "d8vrTlyNXuxc"
+ },
+ "outputs": [],
+ "source": [
+ "# Construct a random num x 1 tensor\n",
+ "def get_elements(num):\n",
+ " return tf.random_uniform(shape=(num, 1), maxval=1)"
+ ]
+ },
+ {
+ "cell_type": "markdown",
+ "metadata": {
+ "colab_type": "text",
+ "id": "ILJ6SbF3bXFQ"
+ },
+ "source": [
+ "## Graph mode"
+ ]
+ },
+ {
+ "cell_type": "code",
+ "execution_count": 0,
+ "metadata": {
+ "colab": {
+ "autoexec": {
+ "startup": false,
+ "wait_interval": 0
+ }
+ },
+ "colab_type": "code",
+ "id": "vovRf597X55n"
+ },
+ "outputs": [],
+ "source": [
+ "def tf_sum(elements):\n",
+ " # Using custom vectorized op\n",
+ " return tf.reduce_sum(elements)\n",
+ "\n",
+ "def run_trial(num):\n",
+ " elements = get_elements(num)\n",
+ " return tf_sum(elements)\n",
+ "\n",
+ "\n",
+ "\n",
+ "graph_means = []\n",
+ "for num in range(max_elements):\n",
+ " with tf.Graph().as_default():\n",
+ " durations = []\n",
+ " foo = run_trial(num)\n",
+ " \n",
+ " with tf.Session() as sess:\n",
+ " \n",
+ " for _ in range(burn_ins):\n",
+ " for _ in range(batches):\n",
+ " sess.run(foo)\n",
+ " \n",
+ " for _ in range(trials):\n",
+ " \n",
+ " start = time.time()\n",
+ " for _ in range(batches):\n",
+ " sess.run(foo)\n",
+ " \n",
+ " duration = time.time() - start\n",
+ " durations.append(duration) \n",
+ " \n",
+ " graph_means.append(np.mean(durations)) "
+ ]
+ },
+ {
+ "cell_type": "code",
+ "execution_count": 0,
+ "metadata": {
+ "colab": {
+ "autoexec": {
+ "startup": false,
+ "wait_interval": 0
+ },
+ "height": 301
+ },
+ "colab_type": "code",
+ "executionInfo": {
+ "elapsed": 278,
+ "status": "ok",
+ "timestamp": 1532447361278,
+ "user": {
+ "displayName": "",
+ "photoUrl": "",
+ "userId": ""
+ },
+ "user_tz": 240
+ },
+ "id": "Jm9Blkyx90Eq",
+ "outputId": "d83cd51f-7e56-4d73-f7df-bb157dee46df"
+ },
+ "outputs": [
+ {
+ "data": {
+ "image/png": "iVBORw0KGgoAAAANSUhEUgAAAa8AAAEcCAYAAABwNTvaAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nH4gzFtcS9o9MAAAAAElFTkSuQmCC\n",
+ "text/plain": [
+ "\u003cmatplotlib.figure.Figure at 0x7f47b20dd690\u003e"
+ ]
+ },
+ "metadata": {
+ "tags": []
+ },
+ "output_type": "display_data"
+ }
+ ],
+ "source": [
+ "plt.plot(graph_means)\n",
+ "plt.ylabel('Time (seconds)')\n",
+ "plt.xlabel('Length of vector')\n",
+ "_ = plt.title('Time to sum the elements of 1000 vectors (vectorized TF operation)')\n",
+ "_ = plt.ylim(ymin=0)"
+ ]
+ },
+ {
+ "cell_type": "markdown",
+ "metadata": {
+ "colab_type": "text",
+ "id": "4KZg2WXjbhg5"
+ },
+ "source": [
+ "## AutoGraph"
+ ]
+ },
+ {
+ "cell_type": "code",
+ "execution_count": 0,
+ "metadata": {
+ "colab": {
+ "autoexec": {
+ "startup": false,
+ "wait_interval": 0
+ }
+ },
+ "colab_type": "code",
+ "id": "UQJBQWbCbinm"
+ },
+ "outputs": [],
+ "source": [
+ "# Sum written using for loop and converted with AutoGraph\n",
+ "def sum_all(elements):\n",
+ " sum_ = 0.0\n",
+ " length = len(elements)\n",
+ " for i in tf.range(length): \n",
+ " sum_ += elements[i][0]\n",
+ " return sum_\n",
+ "\n",
+ "def run_trial(num):\n",
+ " elements = get_elements(num)\n",
+ " return sum_all(elements)\n",
+ " \n",
+ "ag_means = []\n",
+ "ag_run_trial = ag.to_graph(run_trial)\n",
+ "\n",
+ "for num in range(max_elements):\n",
+ " with tf.Graph().as_default():\n",
+ " durations = []\n",
+ " foo = ag_run_trial(num)\n",
+ " with tf.Session() as sess:\n",
+ " for _ in range(burn_ins):\n",
+ " for _ in range(batches):\n",
+ " sess.run(foo)\n",
+ " \n",
+ " for _ in range(trials):\n",
+ " start = time.time()\n",
+ " for _ in range(batches):\n",
+ " sess.run(foo)\n",
+ " \n",
+ " duration = time.time() - start\n",
+ " durations.append(duration)\n",
+ " ag_means.append(np.mean(durations))"
+ ]
+ },
+ {
+ "cell_type": "code",
+ "execution_count": 0,
+ "metadata": {
+ "colab": {
+ "autoexec": {
+ "startup": false,
+ "wait_interval": 0
+ },
+ "height": 301
+ },
+ "colab_type": "code",
+ "executionInfo": {
+ "elapsed": 310,
+ "status": "ok",
+ "timestamp": 1532448438694,
+ "user": {
+ "displayName": "",
+ "photoUrl": "",
+ "userId": ""
+ },
+ "user_tz": 240
+ },
+ "id": "DLDOmrRW99v5",
+ "outputId": "ae0e0573-39db-4004-a064-efc618dbf867"
+ },
+ "outputs": [
+ {
+ "data": {
+ "image/png": "iVBORw0KGgoAAAANSUhEUgAAAYwAAAEcCAYAAADUX4MJAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\n",
+ "text/plain": [
+ "\u003cmatplotlib.figure.Figure at 0x7f47b218dbd0\u003e"
+ ]
+ },
+ "metadata": {
+ "tags": []
+ },
+ "output_type": "display_data"
+ }
+ ],
+ "source": [
+ "plt.plot(ag_means)\n",
+ "plt.ylabel('Time(s)')\n",
+ "plt.xlabel('Length of vector')\n",
+ "_ = plt.title('Time to sum the elements of 1000 vectors (AutoGraph)')\n",
+ "_ = plt.ylim(ymin=0)"
+ ]
+ },
+ {
+ "cell_type": "markdown",
+ "metadata": {
+ "colab_type": "text",
+ "id": "d7IAJ6Bwbk9t"
+ },
+ "source": [
+ "## Eager"
+ ]
+ },
+ {
+ "cell_type": "code",
+ "execution_count": 0,
+ "metadata": {
+ "colab": {
+ "autoexec": {
+ "startup": false,
+ "wait_interval": 0
+ }
+ },
+ "colab_type": "code",
+ "id": "XMu5-12yoOzY"
+ },
+ "outputs": [],
+ "source": [
+ "from tensorflow.python.eager import context"
+ ]
+ },
+ {
+ "cell_type": "code",
+ "execution_count": 0,
+ "metadata": {
+ "colab": {
+ "autoexec": {
+ "startup": false,
+ "wait_interval": 0
+ }
+ },
+ "colab_type": "code",
+ "id": "_vt9MzpyjQ4T"
+ },
+ "outputs": [],
+ "source": [
+ "# Sum written using for loop and run with tf.eager\n",
+ "def sum_all(elements):\n",
+ " sum_ = 0.0\n",
+ " length = elements.shape[0]\n",
+ " for i in tf.range(length): \n",
+ " sum_ += elements[i][0]\n",
+ " return sum_\n",
+ "\n",
+ "eager_means = []\n",
+ "for num in range(max_elements):\n",
+ " with context.eager_mode():\n",
+ " durations = []\n",
+ " for i in range(trials + burn_ins):\n",
+ " \n",
+ " start = time.time()\n",
+ " for _ in range(batches):\n",
+ " run_trial(num)\n",
+ " \n",
+ " if i \u003c burn_ins:\n",
+ " continue\n",
+ " \n",
+ " duration = time.time() - start\n",
+ " durations.append(duration)\n",
+ " eager_means.append(np.mean(durations))"
+ ]
+ },
+ {
+ "cell_type": "code",
+ "execution_count": 0,
+ "metadata": {
+ "colab": {
+ "autoexec": {
+ "startup": false,
+ "wait_interval": 0
+ },
+ "height": 301
+ },
+ "colab_type": "code",
+ "executionInfo": {
+ "elapsed": 422,
+ "status": "ok",
+ "timestamp": 1532460024499,
+ "user": {
+ "displayName": "",
+ "photoUrl": "",
+ "userId": ""
+ },
+ "user_tz": 240
+ },
+ "id": "5gHVdMlD-A-T",
+ "outputId": "3b581cb7-7ef9-489c-92f1-3e52c0c2dc8a"
+ },
+ "outputs": [
+ {
+ "data": {
+ "image/png": "iVBORw0KGgoAAAANSUhEUgAAAYkAAAEcCAYAAAAydkhNAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz\nAAALEgAACxIB0t1+/AAAIABJREFUeJzs3XlclNX+wPHPsC+CILviCiruhAiiqaAimqnglpZLWpkt\ntmh50+rXvZZ2q9veq5ulZYtlaZp7mQuaorjklkoqiwgKsojsy8yc3x9cBxFUVGBYvu+/mGfmPM93\nzjzMd85zznOORimlEEIIISphYuwAhBBC1F2SJIQQQtyQJAkhhBA3JElCCCHEDUmSEEIIcUOSJIQQ\nQtyQJAlg8eLFvPrqq8YOo14ZOHAge/furfHjzJs3jw8//LDGj1NXxMfHExERQc+ePfnuu++MHU6D\ndvbsWcaMGWPsMCo1btw4YmNjjR0G0EiSxD333IOfnx9+fn506tSJHj16GLZt2LCBxx9/nNdff73G\n40hOTsbHxwe9Xl/jx6pOje2L+lZq8nNcsmQJgYGBHDp0iEmTJlV4fvPmzUyYMAFfX1+mTJlS4flT\np04xevRofH19GTNmDDExMeWef+eddwgMDKR379688847t1W2pk2ePJlVq1bV2vE++ugjHn30UcPj\ngQMH0qNHD/z8/AzfD2+88UatxXOtRx55pM78zzWKJHH48GH+/PNP/vzzT5o3b87ixYsN2+6///5a\ni0MphUajQe5frN9q8nO8cOEC3t7eN3zewcGBhx9+mBkzZlR4rqSkhKeeeorw8HAOHDhAeHg4Tz75\nJFqtFoAVK1awfft21q9fz7p164iMjOTHH3+sUtn64HY+j7S0NKKjoxk0aFC57YsXL+bPP/80fD+8\n8sor1R3mTel0OqA0YUVHR5Oenl6rx69Mo0gS11JKVTiZPvnkE1588UWg7Ffi6tWrCQ4OJjAwkBUr\nVnD8+HFGjhxJQEBAhVbHqlWruO+++wgMDOTRRx/lwoULlR578uTJAPj7++Pn58fRo0dRSvHpp58y\ncOBA+vbty0svvURubm6l5S9fvszMmTPp1asXgYGB5X5p+vj4cP78ecPja3/979+/nwEDBrBkyRL6\n9OlDv3792Lp1Kzt37iQsLIzAwEAWL15c6TF/+ukn1q9fz5IlS/Dz8+OJJ54wPHfq1ClGjhxJr169\nmD17NsXFxYbnduzYQXh4OL169WLixIn8/fffle4fIDY2lunTpxMYGMiwYcPYvHnzDV97s/0OHDiQ\npUuXMnLkSO655x5eeeUVMjIyeOyxx/Dz82P69Onk5OQYXn/kyBEmTJhAr169CA8PZ//+/YbnJk+e\nzIcffsjEiRPx8/PjkUceISsry/AclP8cExMTmTx5Mv7+/gQFBTF79uwbvodt27Zx//33ExAQwJQp\nU4iLiwNg6tSpREdHs2DBAvz8/Dh37lyFskFBQQwdOhQXF5cKz+3fvx+dTseUKVMwNzdn8uTJKKXY\nt28fAL/88gvTp0/H1dUVV1dXpk2bxpo1awCIjo6+adlrbdq0qcJlmmXLlvHkk08CUFxczFtvvUVI\nSAj33nsv//znP8udG1u3biU8PJyePXsyZMgQdu/ezfvvv8+hQ4d4/fXXy/2C//PPPxk7diy9evVi\n3LhxHD58uNxn9P777zNx4kR8fX1JSkpi9erVDB48GD8/PwYPHsyGDRsq/Qz27NlDly5dsLCwKLf9\nRonm/PnzTJ06lcDAQIKCgnjhhRfK/Z+eOHHCcJnw2Wef5fnnny/XErjVefvFF18Yzlu9Xo+FhQVd\nunRh9+7dlcZTq1QjExISoqKiospt+/jjj9WLL76olFIqKSlJdezYUb322muqqKhI7dmzR3Xr1k09\n9dRTKjMzU6WkpKigoCB14MABpZRSv//+uxoyZIiKi4tTOp1O/fe//1UPPPBApcdOSkpSPj4+Sq/X\nG7atXLlSDRkyRCUlJan8/Hz19NNPG2K53rvvvqtee+01pdPplFarVQcPHjQ85+PjoxITEw2PX3rp\nJfXBBx8opZSKjo5WnTt3Vp9++qnSarXqp59+Ur1791Zz5sxR+fn56syZM6pbt27q/PnzlR732n1d\nW4/jxo1TaWlp6sqVK2rYsGFqxYoVSiml/vrrLxUUFKSOHTum9Hq9WrNmjQoJCVHFxcUV9p2fn68G\nDBig1qxZo/R6vTp58qQKDAxUZ8+erXDsW+03JCREPfDAAyojI0OlpqaqoKAgFRERoU6dOqWKi4vV\nlClT1CeffKKUUiolJUUFBASoXbt2KaWUioqKUgEBASozM1MppdSkSZNUaGioOnfunCoqKlKTJk1S\n77777g0/x9mzZ6vPPvtMKaVUUVGROnToUKV1GRcXp3x9fVVUVJTSarXqiy++UKGhoaqkpMRw3JUr\nV1Za9lo//fSTmjx5crltX331lXrsscfKbXv88cfVV199pZRSqmfPnuro0aOG544fP678/PyqVPZa\nBQUFys/PT507d86wbcyYMWrTpk1KKaXeeOMN9cQTT6js7GyVl5enZs6cqd577z2llFJHjx5VPXv2\nNPwPpqamqri4uErfe1ZWlurVq5dat26d0ul0asOGDapXr14qKyvL8PqQkBB19uxZpdPpVE5OjvLz\n81MJCQlKKaXS0tIM59H13nrrLbVgwYJy2yr7brjq3LlzKioqSpWUlKjMzEw1adIktWjRIqWUUsXF\nxSokJER9++23SqvVqi1btqguXbrc1nkbHh6uUlJSVFFRkeGYr7/+uvr3v/9daTy1qdG1JKpCo9Hw\n1FNPYWFhQZ8+fbC2tmb48OE4Ojri5uaGv78/J0+eBODHH39kxowZtG3bFhMTE2bMmEFMTAwXL168\n4f7VNb9WNmzYwMMPP0yLFi2wtrZm9uzZbNq0qdLr3WZmZqSlpZGUlISpqSk9e/asdJ+VMTc3Z+bM\nmZiamnLfffdx+fJlpk6dirW1Nd7e3nh7e9/0135lpkyZgrOzM/b29oSEhHDq1CkAVq5cyYQJE+jW\nrRsajYbw8HAsLCw4evRohX3s2LEDT09PwsPD0Wg0dOrUiSFDhvDrr79WeG1V9jtp0iSaNWuGq6sr\n/v7+9OjRAx8fH8zNzQkNDTXEuG7dOoKDg+nXrx9Q+gu9a9eu7Ny507Cv0aNH06pVKywsLBg2bJih\n7FXX1rmZmRnJycmkpqZiYWGBn59fpXW2efNmgoODCQoKwtTUlEceeYTCwsJyv5DvVH5+PnZ2duW2\nNWnSxPCL9/rn7ezsyM/Pr1LZa1lZWTFo0CDDr/SEhATi4+MNl25WrVrFvHnzsLOzw8bGhhkzZhhe\nu2rVKsaOHUtQUBAArq6utG3bttL3ExkZSZs2bRgxYgQmJiYMHz6cdu3asWPHDsNrIiIi8PLywsTE\nBFNTU0xNTTl9+jRFRUU4Ozvj5eVV6b5zcnKwtbWtsP2pp54iICCAXr16ERAQwMqVKwFo1aoVQUFB\nmJmZ4ejoyNSpUzlw4ABQ2iLV6XRMmjQJU1NTQkND6d69u2GfVTlvp0yZgpubW7mWja2tLdnZ2ZXG\nX5vMjB1AXeXk5GT428rKCmdnZ8NjS0tLwz/XhQsXWLhwIW+99RZQdr06NTUVDw+PWx7n0qVLNG/e\n3PC4RYsWaLVa0tPTcXV1LffaRx99lI8//pjp06ej0WgYN25cpdemK+Pg4IBGozG8n8re49X3VFXX\nlre2tiYtLQ0orZO1a9caRucopdBqtVy6dKnCPi5cuMCRI0cICAgwvFan0xEeHl7pa2+132tjsrS0\nrPD42s9t8+bNhi+cq/u6+uUFlPvMra2tb1o/c+fO5YMPPmDs2LGGfoPKRs5c/3lrNBo8PDxITU29\n4b6rysbGpsKXem5uLk2aNKn0+dzcXGxsbKpU9nrDhw/n7bff5sknn2TDhg0MHjwYCwsLMjMzKSgo\nKPfe9Xq9IaGmpKQwYMCAKr2f6+sKoHnz5uXqyt3d3fC3tbU177//PkuXLmX+/Pn07NmTuXPn0q5d\nuwr7tre3Jy8vr8L2Tz/9lN69e1fYnpmZyRtvvMHBgwfJz89Hp9Ph4OAAlPZvuLm5lXv9tf/7VTlv\nr30fV+Xl5WFvb19he22TJHGX3N3deeKJJ6rUAX71S/parq6u5fowkpOTMTMzK/cFdZWNjQ3/+Mc/\n+Mc//kFsbCyTJ0+me/fu9O7dG2trawoKCgyvTUtLq/TEqw3u7u7MnDmTxx9//Jav9fDwIDAwkKVL\nl1brfqty3PDwcBYsWHDbZSv7HJ2cnAx9VYcOHWLatGkEBATQsmXLcq9zdXXlzJkz5bZdvHixWj6r\n9u3bs2zZsnLbTp8+behD8fb2JiYmhm7dugGlfUrt27e/adnKRlgB3HvvvcybN4+YmBg2btzI/Pnz\nAXB0dMTa2poNGzZU+JEDpZ/htX1n17q+Xl1dXdmyZUu5bRcuXKB///43LNO3b1/69u1LcXEx77//\nPq+++irLly+vcKyOHTuydu3aCttv1CJ/99130Wg0bNiwAXt7e7Zu3WroN3FxcamQ5C9evEirVq0M\n7/lOztu4uDhGjhx5W2VqglxuqsStLt1ca+LEiSxevJizZ88Cpc3Yyi6VADRr1gwTExMSExMN24YP\nH86yZctISkoiLy+P999/n+HDh2NiUvGjiYyMNJS1sbExNK+htON6w4YN6PV6du3aZWgKVwdnZ+cb\n/mNXZvz48axYsYJjx44BpZcydu7cWekv8eDgYOLj41m7di1arZaSkhKOHz9u6My90/3eysiRI9m+\nfTu7d+9Gr9dTVFTE/v37q/SLvrLP8ddffzWUtbe3x8TEpNLPcNiwYURGRrJv3z60Wi1Lly7F0tIS\nX1/fKsWt1+spLi5Gq9WW+xsgICAAExMTvv32W4qLiw2/XAMDAwEIDw9n2bJlpKamkpqayrJlyxg9\nevRNy1b2qxrA1NSUsLAw3n77bbKzs+nbty+AoYW7aNEiMjMzAUhNTTV0wI4dO5bVq1ezb98+lFKk\npqYaPuvrz7MBAwZw7tw5Nm7ciE6nY9OmTcTFxRESElJpTBkZGWzfvp2CggLMzMwM/yOV6du3LydO\nnCjXoX4zeXl52Nra0qRJE1JTU8v9qPH19cXU1JTly5ej0+nYunWr4RyFOztvi4uLOXHihKFejanR\nJYnKfgXe6jU3ezx48GAee+wxnn/+efz9/Rk5ciR//PFHpfu1srJi5syZTJw4kYCAAI4dO8bYsWMZ\nNWoUkyZNIjQ0FGtr6xsOu0tISODhhx/mnnvuYeLEiTz00EP06tULgJdffpnt27fTq1cvNm7cyODB\ng+/qPV5r7NixnD17loCAAJ5++ulbvr5r1668/vrrLFiwgICAAMLCwgyjaK5na2vLl19+yaZNm+jX\nrx/9+vXj3XffrfSf91b7vZ335O7uzqeffsrixYsJCgoiJCSEL7/80vAD4WZlK/scjx8/zrhx4/Dz\n8+Opp57i5ZdfpkWLFhXKtm3blnfeeYfXX3+doKAgIiMj+eyzzzAzM7vlcQHWrl1L9+7dWbBgAYcO\nHaJHjx6GG0HNzc359NNPWbNmDQEBAaxevZpPP/3UsO8JEyYQEhLCyJEjGTlyJCEhIYwfP75KZSsz\nfPhw9u7dy7Bhw8olxBdeeIHWrVszfvx4/P39mT59OgkJCQB0796dRYsWsWjRInr27MmUKVMM/XdT\npkzh119/JTAwkIULF+Lg4MBnn33G0qVL6d27N0uXLmXx4sU0bdq00rrS6/V89dVX9O/fn969e3Pg\nwAFee+21SmN3cnKid+/ebN26tdz2J554wnBPlZ+fH7NmzQLg6aef5q+//sLf35+ZM2cSFhZmKGNu\nbs7HH3/MypUr6dWrFxs2bGDgwIGG/oXbPW+hdARcYGBgpaPYaptG3c7P5ts0f/58IiMjcXJyYv36\n9QDExMTw2muvUVRUhJmZGa+99pqh+SuEELUlNjaWl156ydA5XZ3Gjx/PxIkTiYiIuKPyDzzwAAsX\nLrzpPTO1pUaTxMGDB7G1tWXu3LmGJPHII48wbdo07r33Xnbu3MmSJUv49ttvayoEIYSocQcOHKBt\n27Y4Ojqybt06/vWvf7F169ZK+xbrmxrtuPb39yc5ObncNo1GY7ihKScnp8KoACGEqG/i4+N57rnn\nyM/Pp1WrVnz00UcNIkFADbckoHS0zsyZMw0tidjYWB599FHDnc8rVqyo0lBRIYQQta/WO65/+OEH\nXn75ZSIjI5k3b55h6JwQQoi6p9aTxC+//GIYeTN06NByQ8VupoYbPEIIUe9l5xVz/Gx6tX5f1vjN\ndNcH6+bmxv79+wkICGDv3r20adOmSvvRaDSkpeXc+oWNgIuLndTF/0hdlJG6KNPY6iL9SgFb9p9n\n17ELFJfo+ee0XrRyK51mxcXF7halb65Gk8ScOXOIjo4mKyuL4OBgZs2axeuvv84bb7yBXq/H0tKy\nVtZxEEKIhijjSiG/7I5j71+p6JWimb0lw4Jb09K18ulU7kSNd1xXp8b0y+BmGtuvpJuRuigjdVGm\noddFXmEJG/eeY+vBJLQ6Pc2dbRkW2IrAzm6YmZbvRajTLQkhhBDV61RCJv9de4LcghKa2VsS0a8d\nQV3cMTG59WwSd0KShBBC1BORR5JZvuU0AGODvRjc0xML88rnp6oukiSEEKKO0+n1/LQ9lt8PnqeJ\ntTlPj+5Gh5YOtXJsSRJCCFFHXczIY/fxi+z9K4Ws3GI8nGx4dlwPXB2say0GSRJCCFHHFBRp+WL9\nSY6cTQfA2tKMgX4tGN3fCxur2v3aliQhhBB1SGZ2IR+sPEZSWi7enk0Z5OfJPe2da7zv4UYkSQgh\nRB1x/lIuH6w8yuWcIkL8WvDQ4A41NmqpqiRJCCGEESmlOJt8hT3HU4g+mUpRiY7xId6EBbSs0iJp\nNU2ShBBCGMmhvy+xMjKWS5dL16d3tLNk+vBO9PKpuD64sUiSEEKIWqbXK9b8EcfGvecwNzMhqIsb\nfbp50KmVo9EvL11PkoQQQtSi/MISFq87yfG4DFwdrZk1uhstXKpvrqXqJklCCCFqScaVQt798Qgp\nmfl0bdeMx0d2wdbK3Nhh3ZQkCSGEqAUX0vN498cjXM4pYkivlowP8a5zl5YqI0lCCCFqWNyFbD5Y\neZTcghLGhXgxLLC1sUOqMkkSQghRA5RSJKTksPvYRfb8dZESrZ5pw3zo16O5sUO7LZIkhBCimh2L\nzWBV5FmS0vIAcGhiweQRHbmng4uRI7t9NZok5s+fT2RkJE5OTqxfv96w/dtvv2X58uWYm5szYMAA\nXnjhhZoMQwghas3Wg+f5YdsZTDQaenZ0oV93D7q0bYapicmtC9dBNZokRo8ezeTJk5k7d65hW3R0\nNDt27GDDhg2YmZmRmZlZkyEIIUSt0OsVK7afYevBJOxtLXh2bHfaetgbO6y7VqNJwt/fn+Tk5HLb\nfvjhBx577DHMzEoP3axZs5oMQQghaoRerzgWm0Fyei5pWYWcS83hXEoOzZ1teW5sd5xrcTrvmlTr\nfRIJCQkcPHiQ999/H0tLS+bOnUu3bt1qOwwhhLhjWp2epRtPEX0ytdz27l5OzBjRGZs6fu/D7aj1\nJKHT6cjOzuann37i2LFjPPfcc2zbtq1KZe92Qe+GROqijNRFGamLMjVVF0UlOv799QEOnkqlU5tm\njB3UHrdmNrg52mBl2fDGAtX6O3J3d2fIkCEAdO/eHRMTEy5fvoyjo+Mty6al5dR0ePWCi4ud1MX/\nSF2UkbooU1N1UVCk5eOfjxGTmEWXts14OqIblhal6zzkZBdQF2v/bpNljXe3K6XKPR48eDB79+4F\nID4+Hq1WW6UEIYQQxhR/MZvXvz5ITGIWPTu68MyY7oYE0ZDVaEtizpw5REdHk5WVRXBwMLNmzWLM\nmDHMmzePESNGYG5uzltvvVWTIQghxF3R6fVs3HuO9XsS0OkVQ3q1ZFyIV70d0nq7NOr6n/p1mDSl\nS8llhTJSF2WkLspUV12kXylg8boTxCZn42hnySPDO9G5Tf0akXm3l5saXi+LEEJUgyNn01m64SR5\nhVoCOrkyOaxjnZ+xtSZIkhBCiGsUlehY+0c8v+5PxMzUhKlDO9K/R/M6sZSoMUiSEEIIILeghO2H\nkth6KIncghLcmtnwxKgutHJr3MOKJUkIIRo1vV6xYW8Cm/ado7hEj62VGff3acOwwFZYN8D7Hm6X\n1IAQotG6klfM5+tOcOrcZRyaWDCmf2v69fDAykK+Gq+SmhBCNEp/J17ms7UnuJJXjK+3M4/c36lR\ndkzfiiQJIUSjUqLV88vuOH6NTkSDhnEhXgwNaNVoO6ZvRZKEEKLRSEzNYcmGkySl5eHc1IrHRnSm\nvaeDscOq0yRJCCEavBKtno17E9i49xw6vWKAb3PGh3hLx3QVSA0JIRq00+ez+PrXGC5m5ONoZ8nU\noT5093Iydlj1hiQJIUSDlH6lgLV/xLPnrxQ0wCA/T0YPaCeth9sktSWEaFCu5Baxenc8m6NKJ+Tz\ndLFlylAfvFs0NXZo9ZIkCSFEvafT6zkRn8nu4ykcOZOGVqdwcbAi/N52BHZ2w8RERi7dKUkSQoh6\nq6hYx/Y/k9hy8DxXcosBaO5sS/gAL3zbNcPMtHFM512TJEkIIeqdohIdO/5MZnP0OXLyS7C2NCPE\nrwX3dvOgjbsdrq72Mm16NZEkIYSoV1Iz8/lw1TFSMvOxtjRlZN82DOnVEhu5W7pG1GhbbP78+fTp\n04cRI0ZUeG7p0qX4+PiQlZVVkyEIIRqQUwmZvPHNQVIy8xnU05O3ZvYhvF87SRA1qEaTxOjRo1m6\ndGmF7SkpKURFRdG8efOaPLwQooEo0erYevA87/10lMJiHdPu8+Gh0A40sZbkUNNq9HKTv78/ycnJ\nFbYvWrSIuXPn8sQTT9Tk4YUQ9ZhSiuNxGUSfvMSRs2kUFOloYm3O06O70aGlTKVRW2q9T2L79u14\neHjQsWPH2j60EKIeWRUZy+boRACc7K0Y4NuCQX6eODW1MnJkjUutJonCwkI+++wzvvzyS8M2pVSV\ny9/tgt4NidRFGamLMg2lLk7EZfDr/kQ8nG2Z/aAfHVs53vYsrQ2lLoytVpNEYmIiycnJjBo1CqUU\nqampjBkzhpUrV+LkdOu5VGRIWykXFzupi/+RuijTUOqiqFjHe98fAgXThvrgZGNOenrube2jodRF\ndbjbZFnjSeLalkKHDh3Ys2eP4fHAgQNZs2YNTZvK7fJCiFKrdsZy6XIBQwNa4e0p3w3GVqOjm+bM\nmcOECROIj48nODiYn3/+udzzGo3mti43CSEatphzl9l2KAkPJxvC+7U1djiCGm5JvPvuuzd9ftu2\nbTV5eCFEPXL4TBpfbjyFRgOPDO+MhbmpsUMSyB3XQggj0+r0rIqMZcuB85ibmTD9vk60a25v7LDE\n/0iSEEIYTWJqDl//+jfxF7Nxb2bDE+FdaenaxNhhiWtIkhBC1LrcghLW7Ioj8kgySkFQFzcmh3XE\nykK+kuoa+USEELXm6uytG/cmkFeoxcPJhomD29O1rSwnWldJkhBC1LgSrY7IIxfYtPccV/KKsbY0\nY8JAbwb29JQ1H+o4SRJCiBqhV4qzSVfYdzKVA6dSySvUYmlhyv19WhMW0Apbmbm1XpAkIYSodhcz\n8vho1TFSLxcA0NTWgvt6tyYsoCV2NhZGjk7cDkkSQohqdTmniPd+PEJGdhG9u7jRt6sHnVo7yjrT\n9ZQkCSFEtckvLOG9n0oTRET/dozo08bYIYm7JD1GQohqUVSi46NVx0hOy2OQnyf3B7U2dkiiGkhL\nQghxx/ILtRw9m87hs+n8FZdBYbGOXj6uTBzc/ran9hZ1kyQJIcQdOZt8hU9+PkZ2fgkALg5WDOrp\nyci+baX/oQGRJCGEuG37TqTw5aYY9HrF/X1aE9jJjebOttJ6aIAkSQghqkwpxdrd8azbk4C1pSlP\nhHeTu6UbOEkSQogqUUqx/PfTbP8zGeemVjw7rgctnG2NHZaoYZIkhBC3pFeK5VtOs+NwMp4uTXhh\ngi/2tnJTXGNQo0li/vz5REZG4uTkxPr16wF4++232bFjBxYWFrRq1Yo333yTJk1kamAh6iq9Uny3\n5TSRh5Np6VqaIOSu6cZDo2pw/dCDBw9ia2vL3LlzDUkiKiqK3r17Y2Jiwn/+8x80Gg1z5syp0v5k\nYfNSssh7GamLMtVVF4XFWtbtTiD2whWy84rJzi+moEhHK9cmvDDxHppY1/05l+S8KOPiYndX5Wu0\nJeHv709ycnK5bX369DH87evry2+//VaTIQghbkNs8hW+WH+SS1kFaDRgZ22Ok70VLVya8FBoh3qR\nIET1MmqfxKpVqxg+fLgxQxBCAHq9Yn1UAuv3JKCUYlhgK8L7tcPcTCZlaOyMliT++9//Ym5uzogR\nI6pc5m6bTQ2J1EUZqYsyd1IX+YUlvPPdIQ6eSsXZwZrZE/3o5u1cA9HVLjkvqodRksSaNWvYuXMn\n33zzzW2Vk2uMpeR6axmpizJ3UheZ2YV8uOoY5y/l0qVtM2aO6oKtlXm9r1M5L8rU6T4JKB1bfa1d\nu3axZMkSvvvuOywsZISEEMZyLiWHD1cdJSu3mGDf5jwY2kFWiRMV1GiSmDNnDtHR0WRlZREcHMys\nWbNYvHgxJSUlTJ8+HYAePXrwz3/+sybDEEJc5+/Ey3y46hhFxTrGh3gTFtBSptQQlarRIbDVTZqP\npaQpXUbqokxV6+LI2XT++8tf6PWKx0Z0JqCTWy1EV7vkvChT5y83CSGMK7eghKycIrLzi0lMzWVV\nZCxmphqeGdudbu1k3iVxc5IkhGjA/jh6gW9++xudvuyCgY2lGc+O6057TwcjRibqC0kSQjRQx2Iz\n+PrXv7GxMqNXJ1fsbSywszGnWzsnXBysjR2eqCckSQjRACWkZPPfX/7C1FTDs2O749WiqbFDEvWU\njHcTooFJSsvlg5XHKC7RMWNEF0kQ4q5IS0KIBuDvxMus3h3PgRMppF4uAOCh0A707Ohi5MhEfSdJ\nQoh6rKhEx4/bzxJ5uHQiTUsLU+5p70xAJzcCOze8oa2i9kmSEKKeSrqUy2frTnAhPQ9PF1seH90d\nN3tLuWtXb7olAAAgAElEQVRaVCtJEkLUMyVaPb/tT2TdngS0Oj2D/DwZF+JFi+YOcgOZqHaSJISo\nR04mZPLdltOkZOZjb2PO1GFduKe99DuImiNJQoh6oLBYy/LfT7PneAoaYKBfC0b3b4eNlSwCJGqW\nJAkh6rjE1Bz++8tfpF4uoLWbHVOHdaSNu72xwxKNhCQJIeoopRTb/0zmx+1n0OoUQwNaMXpAO+mY\nFrVKkoQQdZBSip93xrFp3zmaWJvz6P2d6O5V/1eLE/WPJAkh6hi9UqzYeoath5Jwc7TmhQn34NTU\nythhiUbqlkni/PnzrFq1iujoaFJSUrC0tMTHx4ewsDCGDBmCmdmNdzF//nwiIyNxcnJi/fr1AFy5\ncoXnn3+e5ORkPD09+eCDD7Czk7VoReNUotVzObeIgkItCoVSsPNIMruOXqSFiy0vPOBL0yaWxg5T\nNGI3XXTo//7v/zhx4gRDhw7lnnvuwdnZmaKiImJjY9m9ezcnT57kn//8J76+vpWWP3jwILa2tsyd\nO9eQJN555x0cHBx47LHH+Pzzz8nOzuaFF16oUrAyBryULKhSpj7WRVpWAd/8GsP5S7lk55dU+prW\nbnbMmeBLE+uqj16qj3VRU6QuytTookODBg1iwYIFFbZ37NiR++67j6ysLM6fP3/D8v7+/iQnJ5fb\ntm3bNr777jsAIiIimDx5cpWThBD1XWzyFT76+Rg5+SW4OlrTwqUJjnaW2FiZYaLRoNGAjZU5g/xa\nyPBWUSfcNEkMGDDgpoUdHBxwcLi9hUsyMzNxdi7tgHNxceHy5cu3VV6I+upAzCWWbDiJTqeYPKQD\nIX6exg5JiFuq0li6f//73+Tk5KDVannwwQfx9fVl7dq1NR2bEPVadl4xO48k882vMfxr2YHS9R1M\nNDw7rrskCFFvVGl0U1RUFC+99BKRkZG4ubnx/vvvM2PGDEaNGnXbB3RyciI9PR1nZ2fS0tJo1qxZ\nlcve7bW1hkTqokxdrIviEh3zv9hHSkY+AGamJnRq04wnx/agjUfN3QhXF+vCWKQuqsdtDYE9cOAA\noaGhuLm5odFoqlTm+n7xgQMHsnr1ambMmMGaNWsYNGhQlY8vHVGlpFOuTF2ti1+jE0nJyCeoiztD\nerWkhYut4Sa4moq3rtaFMUhdlLnbZFmly01OTk688sorbNq0ib59+6LVatHpdLcsN2fOHCZMmEB8\nfDzBwcH8/PPPzJgxg6ioKMLCwti7dy8zZsy4qzcgRF2TW1DChqgEbK3MeDC0Pa3d7eQuaVFvVakl\n8e6777Ju3TrGjh1L06ZNSUpKYtq0aVUqV5lly5bdVpBC1CcbohLIL9LywEBvbGWEkqjnqpQkmjVr\nxsMPP2x47OnpiaendLwJcb20rAK2/5mEc1MrBkrntGgAbtoGfvLJJzl27Filz+Xm5vL111/z448/\n1khgQtRHa3bFodUpRvdvh7mZXGIS9d9NWxLPPPMM7777LgkJCXTv3h0nJyeKioqIi4sjOTmZCRMm\nMHHixNqKVYg6Sa9XHI1NZ+vBJE6du0xrdzsCZH1p0UDcNEn4+PjwxRdfcPHiRfbv309qaiqWlpYM\nHTqUnj17YmFhUVtxClEnnU26wpINJ7mUVQBAp9aOPBTaAZMqjv4Toq6rUp+Eh4fHHd0TIURDlltQ\nwqe/HCc7r4T+PTwY3LMlnq5NjB2WENWqShdNMzIyeOGFF3jooYcAiImJ4YcffqjRwISo677b8jdZ\nucWE92vLw8M6SYIQDVKVksQrr7xCz549yc7OBqBdu3Z8//33NRqYEHXZvpMp7D91Ce8WTRnWu5Wx\nwxGixlQpSaSmpjJx4kRMTU0BsLCwwMRERm6Ixikzu5DvfjuNpbkpj97fCVP5XxANWJXO7usXFsrO\nzq4w3YYQjUFqZj6frD5OfpGWCYO8cXW0MXZIQtSoKnVcDxkyhP/7v/8jLy+P1atX8/333zNmzJia\njk2IOkOr0/Pb/kTW7k5Aq9PTp6s7/Xs0N3ZYQtS4KiWJRx99lHXr1pGdnc3OnTuZPHmyjHYSjYJS\nimOxGfy8M46ktFzsbS14KLQD/h1dqjzJpRD1WZVngR05ciQjR46syViEqDZanf6uJtXT6vT8FZfJ\nuj3xJKSUziZ6b3cPmY9JNDpVShIZGRl89913JCYmotVqDds//PDDGgtMiDuVmJrDv5Yd4KmIbvh1\ncKlSGaUUv/wRz7HYDC7nFBrWntYAvXxcGdG3DZ4uMsRVND5VShJPPvkknTt3JigoyDDCSYi66kzS\nFZSCY7EZVU4S6/YksD4qAXMzE5rZWdLc2Ra3ZjYM6ukpyUE0alVKEgUFBbz22ms1HYsQ1eLS5dIp\nMs6lVm3RmX0nU1i7Ox7npla8MsUfe1uZbkaIq6qUJHr06MHff/9Nx44dazoeIe5a2v/mUUpOy63Q\nN/H7gfOsj0qgh7cTfbt6YGZmwpcbY7C2NOXZsd0lQQhxnSoliQkTJjBp0iTc3d2xtLQ0bF+1atUd\nH3jZsmWsWrUKjUZDhw4dePPNN2XCQFEtrk62p9UpLqTn0cqtbPnGvSdSyC0oYc/xFPYcTwFAo4Fn\nRvWghVxWEqKCKiWJF198kZkzZ9K5c+dq6ZNITU3l22+/ZfPmzVhYWPDcc8+xadMmwsPD73rfonHT\nK2VoSUDpJaerSaKwWEtiai5eze0ZG+zF7uMXOR6bQXj/dnRt52SskIWo06qUJCwtLXnkkUeq9cB6\nvZ6CggJMTEwoLCzE1dW1WvcvGqcrucWUaPU42VuRkV1IYmqu4bnYC9nolaJDSwc6tnKkYytHI0Yq\nRP1QpYHk/fr1Y9euXdV2UDc3N6ZNm0ZwcDD9+/fHzs6OPn36VNv+ReN16XI+APd0cMZEoynXeX3m\nfBYA7T0djBKbEPVRlVoSP/30E59//jm2trZYWFiglEKj0bB37947Omh2djbbtm1jx44d2NnZ8cwz\nz7B+/XpGjBhxR/sT4qqr/RGeLk3wcLLhfGoueqUw0Wg4k3QFAG/PpsYMUYh6pUpJ4ueff67Wg0ZF\nRdGyZUscHEp/0YWGhnL48OFbJgkXF7ubPt+YSF2UubYu8or1ALRv04xzl3JJPpRECRrcm9kSdzGb\nVu52tG3VzFih1jg5L8pIXVSPKiWJFi1aVOtBmzdvztGjRykqKsLCwoJ9+/bRrVu3W5ZLS6vauPeG\nzsXFTurif66vi4Tk0ktKFoCbgzUAh0+l4OZoQ1GxjnYe9g227uS8KCN1UeZuk+VNk8SLL77IO++8\nw5gxYyqdzOxOh8B2796dsLAwwsPDMTMzo3PnzowfP/6O9iXEtdKyCjAz1eBoZ0lrt9IhrYmpuVzJ\nLQagvVxqEuK23DRJXLp0CYB//OMf1X7gp59+mqeffrra9ysat7SsQpyaWmNioqGla+kvqHMpOVhb\nlp7qHaTTWojbctMkcXW50oCAgFoJRoi7kV+oJbeghLYe9gDYWJnh6mBNYmoOJiYamtlb4tTUyshR\nClG/yLqLosG4ehOd6//6IgBauduRV6glJ79EWhFC3IGbtiROnz5NUFBQhe13OwRWiJpwdfiri2NZ\nkmjt1oSDMaWXTaU/Qojbd9Mk0aZNGz7//PPaikWI25KTX4xdcdn6JldvpLu2JdH6mnmb2reUloQQ\nt+umScLCwqLah78KUR2KS3S8siQab08HZo0uHT6dVklL4uq8TbZWZjR3tq39QIWo526aJMzNZZlG\nUTfFJGaRk1/C4dNpnEvJobW7nWEdCZdrOqftbS3o29UdFwdrTGRNaiFu2007rn/66afaikOI23Is\nNt3w9+8HzwOlLQlHO0sszMvPVPzI/Z0ZeW/bWo1PiIZCRjeJekcpxbHYDKwtzWjh0oTok6mkXykg\nM7sIl2v6I4QQd0+ShKh3Lmbkk36lkC5tmzGqfzt0esWqyFgU4OIg90EIUZ0kSYh651hsBgA9vJwI\n6dkSWysz9p8qHebqKi0JIaqVJAlR71ztj+jazgkrSzMG+JaNwLt2ZJMQ4u5JkhD1Sn6hljNJV2jr\nYUdT29I10Qf6tcDUpHTkkquDjTHDE6LBkSQh6pWTCZno9Ipu16xJ3czeij5d3bG2NMPDSZKEENWp\nSutJCFFXGPojvJ3LbZ8ytCMPDGxvmO1VCFE95D9K1Bt6pTgel4G9jTmt3csvpGJqYoKNlTSMhahu\n8l8l6o0T8ZlcySumWzsnuXtaiFpitJZETk4OL7/8MmfOnMHExIRFixbRo0cPY4Uj6jCtTs+GqAQ2\nRJ1DAwR1dTd2SEI0GkZLEgsXLmTAgAF89NFHaLVaCgsLjRWKqMMupOexdONJ4i/m4GRvySPDO+PT\n2tHYYQnRaBglSeTm5nLw4EH+/e9/lwZhZkaTJk2MEYqoo/ILtazbE8+2Q0no9IqgLu48FNoBGyvp\nRhOiNhnlPy4pKQlHR0fmzZtHTEwMXbt25eWXX8bKSqZUaOwKirTsO5nK2j/iyM4vwbmpFRMHteee\nDi7GDk2IRkmjlFK1fdC//vqLBx54gBUrVtCtWzcWLlyInZ0dzzzzTG2HIuoApRSH/05j+8Hz7P3r\nIsUlOqwsTBk3qAPhA7wqzOoqhKg9RmlJuLu74+7uTrdupYvFhIWFsWTJkluWS0vLqenQ6gUXF7sG\nUxdZuUV8ufEUf8VnAuDmaE1QV3f6dW+Oo50lV7Lyb1q+IdXF3ZK6KCN1UcbFxe7WL7oJoyQJZ2dn\nPDw8iI+Pp23btuzbtw8vLy9jhCKM6PDpNL7aHENuQQld2zVjVN+2tGtuj0aGtwpRZxitF/CVV17h\nhRdeQKvV0rJlS958801jhSJqWVGxjhXbz7DzyAXMzUx4KLQDA/1aSHIQog4yWpLw8fHh559/Ntbh\nhZEkpGSzeN1JUjPz8XRpwuMjO9PCRUa2CVFXyXhCUSu0Oj2/7U/klz/i0ekVYQEtGd3fC3Mzuelf\niLpMkoSoUVeXGv1x+1lSMvNp2sSCR4d3pkvbZsYOTQhRBZIkRI1JTM1hVWQsf8VnotFAiF8LIvq1\no4m1ubFDE0JUkSQJUe0SU3NYtyeBP0+nAdCljSMPDGqPp/Q9CFHvSJIQ1SY9q4CVkbEciCldb9qr\nuT2j7m1Ll7bNZOSSEPWUJAlx1wqKtGzad47f9p9Hq9PT1sOO8H7t6CrJQYh6T5KEuGNKKfaeSGFl\nZCxXcotxtLNkbLAXgZ3dZL0HIRoISRLijpxLyWH576c5m3wFczMTRvZtw7DA1lhayDxLQjQkkiTE\nbSko0rJmVxzbDiWhgJ4dXHhgoDfODtbGDk0IUQMkSYgqOxabwbe/xZCRXYRbMxsmDelAlzZyv4MQ\nDZkkCXFLl3OK+GnHWaJPpmJqouH+Pm0Y0ac15mZyaUmIhk6ShLghrU7P1oNJrN0TT1GxjrYe9kwb\n5oOnq9zvIERjIUlCVKCU4sjZdH7eGceF9DyaWJszYag3/Xo0l1FLQjQykiREOafOXWb1zlhiL2Sj\n0cAA3+aMGeAlU2kI0UhJkhBA6Qpxy7ec5tD/ptLo2cGF8P7taOFsa+TIhBDGJEmikVNK8cexi/y4\n/SwFRVraezZlwqD2tPWwN3ZoQog6wKhJQq/XM2bMGNzc3Pjss8+MGUqjdC4lhxXbzvD3+SysLEyZ\nHNaRAb7S7yCEKGPUJPHNN9/g5eVFbm6uMcNodDKzC1mzK46ov1JQgK+3M5OGdKCZvZWxQxNC1DFG\nSxIpKSns3LmTmTNn8tVXXxkrjEalqETHb9GJbNp3jmKtHk+XJjwwyFtuiBNC3JDRksSiRYuYO3cu\nOTk5xgqh0VBKcSDmEj/tOEtmdhFNbS14KLQdfbt5YGIil5aEEDdmlCQRGRmJs7MznTp1Ijo6usrl\nXFzsajCq+qUqdaGUYv+JFFZsPc3Z81mYmZowdmB7xg1qj41VwxnSKudFGamLMlIX1UOjlFK1fdD3\n3nuPdevWYWpqSlFREXl5eYSGhvL222/ftFxamrQ6oPTkv1VdHDmTzpo/4jh/KRcN4O/jypgB7XB1\ntKmdIGtJVeqisZC6KCN1UeZuk6VRksS19u/fz5dfflml0U3yoZe62T/A5Zwilv9+mj9Pp6HRQGAn\nN4b3adNg73eQL4MyUhdlpC7K3G2SkPskGgi9Uuw8coFVkWcpKNLRoaUDU8I60ryBJgchRO0wepII\nCAggICDA2GHUa38nXuaHbWdITM3F2tKMqUM7yjxLQohqYfQkIe7cpawCVu44y6G/S6fSCOrixthg\nbxztLI0cmRCioZAkUQ/lF5awKjKWLQcS0eoUXi3smTioA+2ay1QaQojqJUmiHlFKEfVXCqt3xXE5\npwhHO0vGhXgR2MkNjVxaEkLUAEkS9URyeh7f/vY3p89nYWFuyqh72zI0sBWW5rI6nBCi5kiSqOOK\ninWsj0rgt/2J6PSKe9o78/QD96DR6owdmhCiEZAkUUcppYg+mcrKyFgu5xThZG/FQ6Ed8G3vjIuj\njYwBF0LUCkkSddC5lByW/36as8lXMDM14f4+bRjeuzWWFnJpSQhRuyRJ1CG5BSWs3hXHzsPJKKBn\nRxceCPHG2cHa2KEJIRopSRJ1gFanZ/exi6zeFUduQQkeTjY8FNqBzjKFtxDCyCRJGJFWpyfqrxQ2\nRCWQfqUQSwtTxod4M9jfEzNTE2OHJ4QQkiSM5fCZNH7Yeob0K4WYmZowuKcn9wW1xqGJ3C0thKg7\nJEnUsit5xXz/+2kOxFzC1ETDoJ6e3Ne7tUylIYSokyRJ1BK9XvHHsQusiowlr1CLVwt7Hh7WqcFO\n4S2EaBgkSdSC0+ez+H7raRJTc7G0MOWh0A6E+LWQWVqFEHWeJIkalJZVwM87Y9l/6hIAQV3cGRvs\nJZeWhBD1hiSJGpBXWMLGqHNsPXQerU7R1sOeBwe3x6tFU2OHJoQQt8UoSSIlJYW5c+eSnp6Oqakp\n48aNY8qUKcYIpVoppfjj2EVW7jhLXqEWJ3tLxgzwIqCzm1xaEkLUS0ZJEqampsybN49OnTqRl5fH\n6NGj6du3L15eXsYIp1qkXs7n680xxCRmYWVhyrhgLwb7e2JuJlNpCCHqL6MkCRcXF1xcXACwtbXF\ny8uLS5cu1cskUVCk5fcD59m47xwlWj2+3s5MGtKBZvZWxg5NCCHumtH7JJKSkoiJiaF79+7GDuW2\nlGh17Dh8gY17E8jJL8He1oJH7++Af0cXWQBICNFgaJRSylgHz8vLY/LkyTz55JMMHjzYWGHclqIS\nHb9Hn+PnHWdJzyrA2tKMiGBvRvVvh42VubHDE0KIamW0JKHVann88cfp378/U6dOrVIZY66hUFyi\nY/ufyfy6P5HsvGIszEwI8WvBfb1bY2djUauxuLjYyXoS/yN1UUbqoozURRkXF7u7Km+0y03z58/H\n29u7ygnCmOIvZrNkw0kuZuRjZWHK8KDWhPq3xN62dpODEELUNqMkiUOHDrF+/Xo6dOhAeHg4Go2G\n559/nv79+xsjnBvS6vSs25PApr3n0CvF4J6ejOrXFlu5rCSEaCSMkiR69uzJqVOnjHHoKjubdIVv\nfoshKS0PJ3srpg/vRKfWjsYOSwghapXRRzfVNbkFJayKPMuuoxcB6N+jOQ8M9MbaUqpKCNH4yDff\nNf48ncayzTHkFpTg6WLL5LCOtPd0MHZYQghhNJIkKL0h7oetZ9h9/CLmZiaMC/Ei1L+lrA4nhGj0\nGn2SOJmQybLNMaRfKaSVWxMeG9FF1ngQQoj/abRJIuNKIT9uP8PBv9PQAMODWjPq3rbSehBCiGs0\nuiRRotXz6/5ENkYlUKzV49XCnkmhHWntfnc3nAghREPUqJLEsdh0vt96hkuXC7C3tWBymBdBXd1l\nGm8hhLiBRpEk0rIKWLHtDIfPpGOi0RDq35JR97bFxqpRvH0hhLhjDfpbsqhEx+Z959i0LxGtTk+H\nlg5MCu2Ap2sTY4cmhBD1QoNMEkopDv2dxo/bz5KRXYhDEwvGD/QmsJObTOMthBC3ocEliYSUbFZs\nPcPppCuYmmgYFtiK+/u0kTumhRDiDjSYb84recX8HBnLnuMXUcA97Z0ZH+KNWzMbY4cmhBD1Vr1P\nElqdnu2Hkli7J56CIh2eLrZMGNSezm2aGTs0IYSo9+p1kvg78TLfbjnNhfQ8bK3MmDSkAwN8m2Nq\nIjfECSFEdaiXSSK3oISfdpxl97GLaIBg3+ZE9G9X6yvECSFEQ2e0JLFr1y4WLVqEUooxY8YwY8aM\nW5ZRShF9MpXvt54ht6CElq5NmDrUh3bN7WshYiGEaHyMkiT0ej2vv/46y5Ytw9XVlbFjxzJo0CC8\nvLxuWCbjSgEfrTrG0dgMLMxNGB/iTWgvT7m0JIQQNcgoSeLYsWO0bt2aFi1aADB8+HC2bdt20yTx\n1NvbySvU0qm1I1OH+eDqYF1b4QohRKNllCSRmpqKh4eH4bGbmxvHjx+/aRm9gqlDO9K/R3O5IU4I\nIWqJUZKEUuq2yyx9JZTCvKIaiEYIIcSNGCVJuLu7c+HCBcPj1NRUXF1db1rGzsZCRi9dw8VFpja/\nSuqijNRFGamL6mGUXt9u3bqRmJhIcnIyxcXFbNy4kUGDBhkjFCGEEDdhlJaEqakpr776KtOnT0cp\nxdixY2/aaS2EEMI4NOpOOgiEEEI0CnKTgRBCiBuSJCGEEOKGJEkIIYS4oTqfJHbt2sXQoUMJCwvj\n888/N3Y4tSolJYUpU6Zw3333MWLECL755hsArly5wvTp0wkLC+ORRx4hJyfHyJHWHr1eT0REBDNn\nzgQgKSmJ8ePHExYWxuzZs9FqtUaOsHbk5OTwzDPPMGzYMIYPH87Ro0cb7XmxbNky7r//fkaMGMGc\nOXMoLi5uNOfF/Pnz6dOnDyNGjDBsu9l58MYbbzBkyBBGjRrFqVOnqnSMOp0krs7xtHTpUjZs2MDG\njRuJjY01dli1xtTUlHnz5rFp0yZWrFjB8uXLiY2N5fPPPycoKIjffvuNwMBAFi9ebOxQa80333xT\nbiTcf/7zH6ZNm8Zvv/2GnZ0dq1atMmJ0tWfhwoUMGDCAzZs3s3btWtq1a9coz4vU1FS+/fZbVq9e\nzfr169HpdGzcuLHRnBejR49m6dKl5bbd6DzYuXMniYmJbNmyhQULFvDaa69V6Rh1OklcO8eTubm5\nYY6nxsLFxYVOnToBYGtri5eXF6mpqWzbto2IiAgAIiIi2Lp1qzHDrDUpKSns3LmTcePGGbbt27eP\nsLAwoLQufv/9d2OFV2tyc3M5ePAgY8aMAcDMzAw7O7tGe17o9XoKCgrQarUUFhbi6upKdHR0ozgv\n/P39sbcvPwv29efB1e/Mbdu2ER4eDkCPHj3IyckhPT39lseo00misjmeLl26ZMSIjCcpKYmYmBh6\n9OhBRkYGzs7OQGkiuXz5spGjqx2LFi1i7ty5hrm7Ll++TNOmTTH530zA7u7ujeL8SEpKwtHRkXnz\n5hEREcGrr75KQUFBozwv3NzcmDZtGsHBwfTv3x87Ozs6d+6Mvb19ozsvrsrMzCx3HmRmZgJw6dIl\n3N3dDa9zc3MjNTX1lvur00lCbuEolZeXxzPPPMP8+fOxtbVtlBMcRkZG4uzsTKdOnQznhVKqwjnS\nGOpGq9Vy8uRJHnzwQdasWYO1tTWff/55o3jv18vOzmbbtm3s2LGDP/74g4KCAnbt2lXhdY2xbq5X\n2fdpVeqlTq9MdydzPDU0Wq2WZ555hlGjRjF48GAAnJycSE9Px9nZmbS0NJo1a/jref/5559s376d\nnTt3UlRURF5eHosWLSInJwe9Xo+JiQkpKSmN4vxwd3fH3d2dbt26ATBkyBC++OKLRnleREVF0bJl\nSxwcHAAYPHgwhw8fJjs7u9GdF1fd6Dxwc3MjJSXF8Lqq1kudbknIHE+loxe8vb2ZOnWqYdvAgQNZ\nvXo1AGvWrGkUdTJ79mwiIyPZtm0b7733HoGBgfznP/8hMDCQX3/9FWg8deHs7IyHhwfx8fFAab+M\nt7d3ozwvmjdvztGjRykqKkIpxb59+2jfvn2jOi+ubyHc6DwYNGgQv/zyCwBHjhzB3t7ecFnqZur8\ntBy7du1i4cKFhjmeqrLMaUNx6NAhJk2aRIcOHdBoNGg0Gp5//nm6d+/Oc889x8WLF2nevDkffvhh\nhc6rhmz//v18+eWXfPbZZ5w/f57Zs2eTnZ1Np06deOeddzA3Nzd2iDUuJiaGl19+Ga1WS8uWLXnz\nzTfR6XSN8rz45JNP2LhxI2ZmZnTu3Jk33niDlJSURnFezJkzh+joaLKysnB2dmbWrFkMHjyYZ599\nttLzYMGCBfzxxx9YW1vz5ptv0qVLl1seo84nCSGEEMZTpy83CSGEMC5JEkIIIW5IkoQQQogbkiQh\nhBDihiRJCCGEuCFJEkIIIW5IkoSocwYOHMjZs2dr5ViffPJJuWmk582bx/Lly+96v/PmzWPEiBHM\nnj37rvd1MzExMWzevLlGjyEaN0kSolH75JNPKCkpqdZ9pqens2XLFtavX897771Xrfu+3smTJ+84\nSej1+mqORjREkiREvREfH89jjz3GuHHjCA8PN0w9AODj48PixYsZO3YsoaGhbNmyxfDcb7/9xrBh\nwxg9ejSLFy/Gx8eHgoICFixYgEajYcKECURERJCbmwvA6dOnmTp1KmFhYbz00ks3jOeXX35hxIgR\njBo1ilmzZpGZmUleXh5Tp06lqKiIiIgIvv7663Jl1q5dy9NPP214rNPp6Nevn2GOsiVLljB+/HhG\njx7NE088QUZGBgAlJSW89dZbjBgxgvDwcGbNmkVWVhYff/wx+/btIyIigoULFwKlsxREREQwatQo\npk2bxvnz54HSO9XDw8N54403mDBhAn/88cfdfByisVBC1DEhISHqzJkz5bZptVoVERGh4uLilFJK\n5ebmqrCwMMPjjh07quXLlyullDp06JDq16+fUkqp9PR0FRAQoBITE5VSSn311VfKx8dH5efnG8oV\nFGMkQHAAAAOxSURBVBQYjvPSSy+pBx98UBUXF6vi4mI1fPhwFRUVVSHG06dPq3vvvVelp6crpZT6\n4IMP1HPPPaeUUiopKUn17t270vdWUFCgevfurS5fvqyUUmr79u1q6tSpSiml1q5dq1599VXDa7//\n/ns1Z84cpZRSH3/8sZo1a5bSarVKKWUov3r1avXMM88YymRkZKjevXur2NhYpZRSK1euVOPGjVNK\nKRUdHa06d+6sjh49WmlsQlRGWhKiXkhISCAuLo7Zs2cTHh7OQw89RElJSbmVCu+77z4AfH19SUtL\no7i4mKNHj9K1a1datmwJwNixYyvsW103M83gwYMxNzfH3Nyczp07k5iYWKFMdHQ0wcHBODk5ATBh\nwgSioqJu+T6srKwYNGgQGzZsAEonYLu6eND27dvZu3cv4eHhhIeH8/3333Px4kWgdKr0KVOmYGpq\nCmCY9fR6R48epVOnTrRr1w6AMWPGcOrUKfLz8wFo3bo13bt3v2WcQlxVp6cKF+IqpRTNmjVjzZo1\nlT6v0WiwtLQEMCw2o9PpKiSA6x9XxsLCwvC3qalppesjK6UqzMV/9bi3Eh4ezptvvsn999/P/v37\neeeddwz7fOKJJxg9enSlx6uKyuK69rGNjU2V9iPEVdKSEPVC27ZtsbKyYu3atYZtcXFx5OXlARW/\nRK8+9vX15cSJE4br8tf2YwA0adKk3ELxVRUUFMTOnTsNfQY//vgjffr0qXD8yvj7+5Obm8t7771H\naGioIbkNHDiQ77//nuzsbACKi4uJiYkBICQkhG+++cbQyX511bkmTZoY+lKuvt9Tp04ZphFfvXo1\nnTt3luQg7pi0JESdo9FoePjhhzEzMzP8Ml6/fj2fffYZCxcu5Msvv0Sn0+Hs7MwHH3xgKHP9PqB0\nAZZ//etfzJgxA0dHR4KDgzEzM8Pa2hqAadOmMWXKFKytrfn222+rHKO3tzezZ8/m4YcfxsTEhJYt\nW7JgwYIKx7+R8PBwPvroo/9v5w5xGASiIAwPBoMhHADNBRCcgtUEzQWQSByChAOgSHB4joVBLqlo\ngnumadK0/T/51LrZyeat1nW9Z2VZ6jgO1XWtIAh0XZeqqlKWZWqaRuM4yjmnMAyVpqmmaVJRFJrn\nWc455Xmurus0DIPatpX3XkmS3E0FeAVfhePnneepKIokPW/W27a9ZRcC+Ac0Cfy8ZVm077u894rj\nWH3ff/pIwNegSQAATDxcAwBMhAQAwERIAABMhAQAwERIAABMhAQAwPQAVSnSA55bZkwAAAAASUVO\nRK5CYII=\n",
+ "text/plain": [
+ "\u003cmatplotlib.figure.Figure at 0x7f47b8e3bd90\u003e"
+ ]
+ },
+ "metadata": {
+ "tags": []
+ },
+ "output_type": "display_data"
+ }
+ ],
+ "source": [
+ "plt.plot(eager_means)\n",
+ "plt.ylabel('Time(s)')\n",
+ "plt.xlabel('Length of vector')\n",
+ "_ = plt.title('Time to sum the elements of 1000 vectors (Eager)')\n",
+ "_ = plt.ylim(ymin=0)"
+ ]
+ }
+ ],
+ "metadata": {
+ "colab": {
+ "collapsed_sections": [],
+ "default_view": {},
+ "name": "Autograph vs. Eager vs Graph sum",
+ "provenance": [
+ {
+ "file_id": "1olZkm32B7n7pQwlIAXR0_w8fZhRHCtkX",
+ "timestamp": 1531755808890
+ }
+ ],
+ "version": "0.3.2",
+ "views": {}
+ }
+ },
+ "nbformat": 4,
+ "nbformat_minor": 0
+}