Forecast with Amazon SageMaker DeepAR 3 (Synthetic data).ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Forecast with Amazon SageMaker DeepAR 3 (Synthetic data)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Often times, you encounter more complex types of data when building a time-series model. This notebook demonstrates how to prepare a synthetic dataset of time series for training DeepAR and how to use the trained model for inference."
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "DeepAR is an algorithm introduced in 2017. It’s quite complex and I won’t go into details here. Let’s just say that unlike other techniques that train a different model for each time-series, DeepAR builds a single model for all time-series and tries to identify similarities across them. Intuitively, this sounds like a good idea for temperature time-series as we could expect them to exhibit similar patterns year after year."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [],
   "source": [
    "import numpy as np                                # For matrix operations and numerical processing\n",
    "import pandas as pd                               # For munging tabular data\n",
    "import matplotlib.pyplot as plt                   # For charts and visualizations\n",
    "from IPython.display import Image                 # For displaying images in the notebook\n",
    "from IPython.display import display               # For displaying outputs in the notebook\n",
    "from time import gmtime, strftime                 # For labeling SageMaker models, endpoints, etc.\n",
    "import sys                                        # For writing outputs to notebook\n",
    "import math                                       # For ceiling function\n",
    "import json                                       # For parsing hosting outputs\n",
    "import os                                         # For manipulating filepath names\n",
    "import boto3\n",
    "import botocore\n",
    "from botocore.exceptions import ClientError\n",
    "\n",
    "import csv\n",
    "import io\n",
    "import re\n",
    "import s3fs\n",
    "\n",
    "\n",
    "import sagemaker                                 \n",
    "from sagemaker.predictor import csv_serializer \n",
    "from sagemaker.predictor import json_deserializer\n",
    "import sagemaker.amazon.common as smac\n",
    "from sagemaker import get_execution_role"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### 1. Preparation (Specifying Sagemaker roles)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Sagemaker session : <sagemaker.session.Session object at 0x0000025F057E1908>\n",
      "S3 bucket : sagemaker-us-west-2-570447867175\n",
      "Prefix : deepar-synthetic\n",
      "Region selected : us-west-2\n",
      "IAM role : arn:aws:iam::570447867175:role/SageMakerNotebookRole\n"
     ]
    }
   ],
   "source": [
    "sagemaker_session = sagemaker.Session()\n",
    "bucket = sagemaker_session.default_bucket()                     \n",
    "prefix = 'deepar-synthetic'\n",
    "region = boto3.Session().region_name\n",
    "role = 'arn:aws:iam::570447867175:role/SageMakerNotebookRole' # pass your IAM role name\n",
    "\n",
    "print('Sagemaker session :', sagemaker_session)\n",
    "print('S3 bucket :', bucket)\n",
    "print('Prefix :', prefix)\n",
    "print('Region selected :', region)\n",
    "print('IAM role :', role)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "'get_image_uri' method will be deprecated in favor of 'ImageURIProvider' class in SageMaker Python SDK v2.\n"
     ]
    }
   ],
   "source": [
    "from sagemaker.amazon.amazon_estimator import get_image_uri\n",
    "image_name = get_image_uri(region, 'forecasting-deepar')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### 2. Load Data"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "##### In this toy example we want to train a model that can predict the next 48 points of syntheticly generated time series. The time series that we use have hourly granularity"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [],
   "source": [
    "freq = 'H'\n",
    "prediction_length = 48"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "We also need to configure the so-called context_length, which determines how much context of the time series the model should take into account when making the prediction, i.e. how many previous points to look at. A typical value to start with is around the same size as the prediction_length. In our example we will use a longer context_length of 72. Note that in addition to the context_length the model also takes into account the values of the time series at typical seasonal windows e.g. for hourly data the model will look at the value of the series 24h ago, one week ago one month ago etc. So it is not necessary to make the context_length span an entire month if you expect monthly seasonalities in your hourly data."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [],
   "source": [
    "context_length = 72"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "For this notebook, we will generate 200 noisy time series, each consisting of 400 data points and with seasonality of 24 hours. In our dummy example, all time series start at the same time point t0. When preparing your data, it is important to use the correct start point for each time series, because the model uses the time-point as a frame of reference, which enables it to learn e.g. that weekdays behave differently from weekends."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {},
   "outputs": [],
   "source": [
    "t0 = '2016-01-01 00:00:00'\n",
    "data_length = 400\n",
    "num_ts = 200\n",
    "period = 24"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Each time series will be a noisy sine wave with a random level"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {},
   "outputs": [],
   "source": [
    "time_series = []\n",
    "for k in range(num_ts):\n",
    "    level = 10 * np.random.rand()\n",
    "    seas_amplitude = (0.1 + 0.3*np.random.rand()) * level\n",
    "    sig = 0.05 * level # noise parameter (constant in time)\n",
    "    time_ticks = np.array(range(data_length))\n",
    "    source = level + seas_amplitude*np.sin(time_ticks*(2*np.pi)/period)\n",
    "    noise = sig*np.random.randn(data_length)\n",
    "    data = source + noise\n",
    "    index = pd.date_range(start=t0, freq=freq, periods=data_length)\n",
    "    time_series.append(pd.Series(data=data, index=index))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAXQAAAEQCAYAAACgBo8fAAAABHNCSVQICAgIfAhkiAAAAAlwSFlzAAALEgAACxIB0t1+/AAAADh0RVh0U29mdHdhcmUAbWF0cGxvdGxpYiB2ZXJzaW9uMy4yLjIsIGh0dHA6Ly9tYXRwbG90bGliLm9yZy+WH4yJAAAgAElEQVR4nOy9Z7QtV3Um+q0KO518s+69kq4SQQgUEDQiP5lHMDYY3LYxz22eh21a2MbYo90eyO7GBhonHm2M6QaDsLuxDR7YJlkBLDcZIQnlnK90cz55pwrr/aiaVatqV1hrVYlzju6eY2jo3nP2XbvCWnN965vfnJNxzjG2sY1tbGPb+Gas9QWMbWxjG9vY6rGxQx/b2MY2tmeIjR362MY2trE9Q2zs0Mc2trGN7RliY4c+trGNbWzPEBs79LGNbWxje4aYtVZfvGXLFr5nz561+vqxjW1sY9uQdvvtt5/gnG/N+t2aOfQ9e/bgtttuW6uvH9vYxja2DWmMsafyfjemXMY2trGN7Rli0g6dMWYyxu5kjF2b8btXM8YWGWN3hf+9r97LHNvYxja2sZWZCuXyHgAPApjO+f13Oec/Uf2Sxja2sY1tbDomhdAZY7sBvBHANU/v5YxtbGMb29h0TZZy+SiA3wXgF3zmCsbY3YyxGxhjz6t+aWMb29jGNjYVK3XojLGfAHCMc357wcfuAHA25/xiAH8J4Ms5Y72TMXYbY+y248ePa13w2MY2trGNLdtkEPrLALyJMfYkgH8AcCVj7O/ED3DOlzjnK+GfrwdgM8a2pAfinH+Kc3455/zyrVszZZRjG9vYxjY2TSt16JzzqznnuznnewC8DcA3OOe/IH6GMbaDMcbCP784HPfk03C9a2r7T3XxoesegO+Pa8iPbWxjW3+mnVjEGLsKADjnnwTw7wG8izHmAugBeBt/BnbOePfn78Rd+xfw5kt24aJdM2t9OWMb29jGljAlh845/xaAb4V//qTw848D+HidF7Yere94AIDgLDK2sY0tbZ7PwQAYxniRrIWNM0UVzA8PHeZ4so5tbJl23u9dj7d9+ua1vozT1sYOXcG8kDt3vWccmzS2sdVmt+49tdaXcNra2KErGEUFBm6RHH9sYxvb2NbGxg5dwbzQow/HDn1sYxvbOrSxQ1cw4tCH3unh0DnneP4ffB1/8/29a30pYxvb2CRs7NAVzA/9uHOaIPS+42N54OL9//LAWl/K2MY2NgkbO3QFo6Do6YLQF3sOAKBhjqfJ2MptnHC39jZeqZLm+fy049DJoTft8TQZW7k5/umxLtazjVeqhJ1cGeC837sex5cHAE4/h96yzTW+krE9XXbh+76Gv/r247WM5YzlvGtuY4cuYXfuW0j8fXCaUS5N6/SYJjc9fgKv+LNvoDt01/pSfiTm+xzdoYc/vuGhWsY7XYDOerbTY6VWtPsOLSb+XsfEvf/QIk6tDiuP83Ta6YbQP/z1h7H/VA/3H1pa60v5kZhbM+ftnCZAZz3b2KFL2F37kwi9Dof+jr/+IT793Scqj/N0WuzQT49pMtu2AQDz63yjrcvcmjnvMUJfezs9VmpFeyCF2OqYuCsDB6uD9X20P91ULnMTDQDAQtdZ4yv50ZiI0N0a0PXpov5az3Z6rNSKlna8Q8+rPKbr8dqPvHXbUujQT5dg11wncOjz3dMDoXvCez282K883phyWXsbO3QJS9duqYrQOQ+cubfOHSUh9IFbfQPbCDbRCGIF6z22UZeJgGL/qW7l8Rx3fc/n08HGDr3EPH8USVdFrFHVxnWO0GOHXh/y+rW/vx2f+d76LCXghO+D5KnPdBM59DoQeh0n1x+V/cI1t+CammJYNz12Yt3M6bFDL7EsNF7VwbmRQ1/fR1Ry6NTYo6rtP9XF9fcewQevXV+lBDjn4JxHJR2OnS4OXQAmS/3qcYPhBkLo9x5cxENHlmsZ6+3X3LJu5vTYoZdYFt1QlXJxBYR+3T2Hcc7V11XSPh9e7OGhI/VL7Y6EqK0uhP4v9xwCAHQa60sGec7V1+M9/3BX9F5OF4TuCSfEpV71AP165tAXukOIXTF7jveMLIM9duglRi+9LWixq0bzSVHgeRz//caHwTlwcL6nPd5H/vUR/Mbn7qx0TWk7utTHwYUeTINh4NQz8X/weNA33FqHHZ++evehyCEdXzk9HLp4QqTTWBVbr7LF+dUhLvnAjfjzf3sUQJBQNXR9DGo6eZKth1jT2KGXGE3SmVCjHPys2osTKRcjbFBahU5f7Dm1LEjR7nhqHgDw0vM2Y+B6qKPn94Fw01rqu+gN137yp40oiIXusHKhKc45Pn/rPiyuYwmkGMOpg3JZrwi9GzruT38n4Mz74fqtA6GLcs/VwdrP6XXn0FcGbm2cbR1Gu+6Lz9kU/awy5eLFlAv1J/UqOJC+46Ffs4O8Y988GpaBS8+ag8+rB3B9n+PgfA9bJgNp4NGl6kG4OkzcqMgh+RxYqZj+/9TJLq7+4r34yt0HK43zdFqCQ68Doa9Th06bcy/0K/3wxFkHoj4q0HPrIa9k3Tn0i/7g63jdR7+z1pcRGb38N77gDHzzd16NF549V51yCY+6ns/BIoSu7zB7Qy+arHXZUye72LO5g6mmBaB6YPT4ygBDz8cLz54DsH4curiROsKfqyLrI+H91XWfDx9Zxr6T1aWFoomb9DOZckmvLVordVzv4YWYKl2pyaE7no9jmvNmXTj0O/bN4yV/9H+iRfRUzRO3itGxrGkZOGfLBBqmUR9C9zgoCbOSQ3c8uD6H4/lY6A5x9RfvqUxpDFwfbduM0v6rHk8PzAfv9PKzg5POkXXi0EUJqnh8rurgyJEfW6qHj3/dR7+DV374m7WMReaFwKJhGVjq1xEUXZ8qFy+1cRE4qYNyOSTIPetC6P90+wFc+ZFva50g1oVDf+TIMo6EQbj1ZsPIoQdB0YZVg0PP4NCrLAaaoD3Hw5/f+Ag+f+t+/PMdBypdY9/x0LTN6L6rO/Tg3b5wz/pC6OJpS3wHVR06OfL1LIEkYLF5olEL5bJeOXQRLN325KkI7NTh0EUkvVyTQz+y2MfKwEVXg5NfFw69Gz7g1XVYtpR2yUZYQrZhGRhWRCJEubg+jxx6lU2CaKH+0IueZVUlSd/10bSMqLlFVUUAOfTn7piGZbB1Uy9FdEKu70elgqteX4TQ17FDJ+Q616nHoYtzeD11LxL3mT/72sMCQq9OU4o0S10Ina6vq7Hm1oVDJ06rLg6qThMpF4AQekWVS7gheEJQtAq66QkIncZpVKxhPnA8tGwzuu9+ReniqdUhOg0T7YaJiaa1LgJIQPK5O56PLZNNADVQLqEjP768Pk4iWUYxg82TDSwPXK3AvO/zKLAsnnb2z3e1lVHzq0P8h8/cgvd95T68/E+/oTWGaHRfLzl3Ex4+uownQ0q3DjluV6A263boOrTpunDodAN1oIS6jVAHcclN06ghKBpMMMfjICBdBaHTi+85XnRtVsUKiX1y6DZRLtU2se7QixKKJptWbcfTqibWH3E8HqlwFnrV6rkQQj+5OqylkuHTYcShbwqrTK4o8ui9oYdzf+96/OU3HgOQ3Bxf9eFv4QdPnNS6rn++4wC+++gJfPYHT0UnuypGlMv26RYA4NhyfQlzqwMXthks4pWaZIu9je7QaZc7ubL+iiINng4OnRKLEhy63pic88QEoGureo19x0fLMiKEXnXy94Yu2oJDXy8IXdycXc/HVMtGwzRq4NADp8E5cGIdzmsgPilSlUnVeybt+md/8BSA0Tmne98TobKK7I+ufxBX/e3t2HtiVWs8kVoC4vhGHZTL6sDFtqlW9Oc6jE7DOsq1deHQ6cLXY5U7eulJyqW+1P+IQ9d06KKjFdOZq8oMBy5RLvUERbtDDx07WKiTLas2eu3wYq+SnC/JoXPYJsN02658Wjy+PMBZmzoAUGtZhjqzEWkebg4RumpyEc1dolbSoMTRnDPp0hCf+s4T+Nr9R/ClO/U0/W7KodPpaeD6lRPmVoceploWmpZRm0Mnf6hTDmR9OHRC6KvrL4BEPFsUFK1Rtihy6LoOU3TcfSdG6FUdet/x0bKNiGqqOl7P8SKEPtG0ajueXvHH36gk5xOd0ND1YZkGZjt2paCo73OsDj383xdux+65Nn7/S/dVol1Ep7Ncg7yQjBzdpkk9hM4R/HufxxSiaLqnzrz1dVhTBUfXNzcRZHtToJrz6lLL7tDFZNOqlUaktaaz5taVQ1+PR1NCzoRULdNIJKDoWKRy8ThCkKM9+cVjWW/oR9dbBVFzztF/OhB66NCnmhZWakg1r8MSOvQQoc+07UpNLkittWO6hV99xbk4uNDDfIUNQrxGVZ67yCIOPUSuqqcSKgVDyyE9R3TndD/PoWuW+E1TLqJktmo8bGXgodO0ag30RyqXjcqhk1M6WUNRJM45/vCr9+POffOVxwKyEDqD41U7qsWp/36E0HVRvxg46TleLGGsgKiHng/OkVC5VJUtig59ommui7oXQIpy8XxYhoHnnjGFu/YvaNNCdG8T4UIH9AJcZKLTqYuq+pvv78Wf3vAwgDgoqkq5EDDx/WzKRVfem1fG4tCiJkJPc+iClLTyvB64mGyatcaFRNWaqq0Phz6sj0Nf6rn4Xzc9ibf8z5tqGW/gerBNFjle2zTAecCd6SIQMfW/alA0gdAdD8v96jXMaVNI6NBrCYqGHHrTruSY+k49xcKAtGyRwzYN/NQlu9B3fPzr/Ue0xqR7m2ia0SZWpTSDyEXXRbm8/18eiLJ1N2tSLjFCz+HQa5jTZA3TwOGFvtZ798J/07QNtG0zAZ505/XKwMULP3gjHj22gk7DwlTLqq1AXhQU3egIvQ7KRUQzukGUxHiun2iSbIeI9aI/+Dp+/lM3a42ZlC1WQ+gJDn3oRcfmKrrxKBBsm2iFssWqHHoQFCWVi4nVoauVfLIycPGc//o1fPwbj9VSsTGdWGSbDC88ew6759ra84eQ2kTDisouV3HodSP0tFOcbtswmHpN9AiY1BwUzXpW526dQM/xtJwmUS4GY5huJxU0ug79ieMrOBkCxomGia1Tzdrq6Ecy5I3q0CmaW8dkFSfVDfcerjzewPUjLTaQzMC87Sk9WkcMihKHrns87Q2F8p1DNwrMVHHARDPVK1tMBkU518uEI1ruE99+PMFz6yL2YUqHbpkMjDG85dJd+P5jJ7SKJEUOvWlFG2KVBibiZr8yqI4C0/XebcMIlD2KlAshc9qX085Wm0PPmBfnb5sEABxaUH8fdJ2mwTDdshO/01UNERADgve8bapVW1YwXdOGpVyy0KRuOVlylns2d3DbU/OVaZeB60VODaiegQkka7kQSq0DoZ9YGYD82iNHl3HT4ycqjdmyzeh0UjXI2nWExKJWgJJ0OEdSn3SHXuLd6l5fOlPUMoL7ffMlO+Fz4MYHjyqPScBksmlF91w1pkF2aKGP/3btA5XGe/xYUs9tmoGjU0W/NI8551jsOrh17ynsmm0L163JoYf3NteJne+5WwOHrlMDiB6fyQJJqmi62aJiwHKiaWHbdBPdoacFSj/xrcfxW/9wJ/a89zr85f95NELmy3317F1p78QYMxljdzLGrs34HWOMfYwx9hhj7B7G2GUqF5G1E+k6OJr8z94xBSBwclVs4PoJh25XzMAExMQiLtAv1Y+nYmW/uw8s4u2fvkVrTNpgW7YJxhiallFJ/zz0fHg+T2SKAnp8sIjKxT/rOrhkUJRHG/ZZmyYA6NV0IZXLRNOMTiWf/s5efP7WfVrXKK6FD3/9YVzzvb34wm37tcYCgMePryT+bhsGZjS09+RsfA786wNH4HgcP3Xpzuj32nN66OGsTR189TdeHv2MMnh15mFEuRjAdKseykU8cXUaJrZNBSUjdE50f/q1h/Dlu4L2jB+58ZFI5fO3Nz+FN37su0pjqXin9wB4MOd3bwBwQfjfOwF8QuUiso6junIimkQUuV+uKI8bun4k3QPqaZ/mCBw68ZDaKhfBkVWR2olGHV1Ig96yzUp1LwhxxEHR6ggdSJZZ1uWo0xy6FQW/GQymt1GQxn6yGXPoP3jiJK7+4r1a15g1N6qcPJ9MZVyaRsAtq5bQJUfp+RzfefQEtk83cemZc9HvdbX3fSco3dwWEoyoY5iObjxBuaQRuiZQEefFRNMSygpUp11EVK7ayFrKoTPGdgN4I4Brcj7yZgCf5YHdDGCWMXaGzNi+zzMpF10HR5RLLMWqxssPXD9Bs9RBuXjiMd+th3KZalqVtM5ZYxL/WxWh0/G0I3DogF7MRNy0xFohugFSkRYIOPTg/TLG0LJNrXFFDr1dQ0PsLKRbRe2S1nNbhh7lIjqeH+49hRefsxmWGQMe7biQ46EVKlLIpiOHrr5O6DpNxrBjJnC8lB1bB+UiIvS1LgttlX8EAPBRAL8LYCrn97sAiGfAA+HPSqOShAYbqaJX2llm4b8jzWlVmVeaQ6+FchEWAjnKKsdTIMj2WwidncGq9SiNg6KhQ7eNSgg97dCrUS6B02nZBq67J55eOkkYwKgSwxZOYG3bjOaniq0OXDA2msKua1mbfZVM1nTfAcNglSgXIGhY8uI9c1EMAqhGI7ZsM+HQZyo4dELohsHw2695Fl574Q64no+f+9TNFSiXeF4MHD+q51KX0kXXSr0TY+wnABzjnN9e9LGMn424FMbYOxljtzHGbjt+/DiA2CFdsH0y8VltjXfNlEvf8SMtNlDdod+1fwFfvCOWw0X9DTXv99TqEA3LwGynES3ymdSxUtXSlEvTMisFRSPKxU4i9J6j7tAXu0NMtyy861XnA4hPTLocOlFeZLawebdsU0v+uTJwMdGwwBhLOCVdo7lhC+j3UIVmMFn/drpdDaEDwGVnz0Hw55VULu2GCcNgEZiqQrmICL1lm3jh2XMRRTL09OaNeHIzDGC6HdRzWev69zLe6WUA3sQYexLAPwC4kjH2d6nPHABwpvD33QAOpQfinH+Kc3455/zyrVu3Aoi5zwvPmE58VpeCcFKUS9VU6SDDMT7IiEdKHfup//F9PHA4LtYUIXRdPeyJVezZ3MFEw4yQf7panaqJQdHg/0YlVQXFSOg5tiMpn/qY810HcxMN/MaV5+OG97wCf/fL/w5AFQ496SDEGEnTNrTGXR24mGgG92jUEXMJ5wahQEA/a3Lo+ji+MsDuuXbi51NNCwPXV1p3aYe+dbJZC0LvO150OiTKqgpCjxx66t0C1SmXq9/wHPzkC3aChXROlY02y0Slj4yVOnTO+dWc892c8z0A3gbgG5zzX0h97KsAfjFUu7wEwCLnXEoETjvdhTtTDr1iUHS6bcM0WGXKpTt0MSEcnRs1UC6iEfLVvd+9J1ZxzpaJBBIUJdk6+mxy3oSOqiJ00pvT4qT/6/DT890hZjsNmAbDc8+YjjMxdTn0NOUivN+WZWqlhq8OvMqbqmg0N7ZNN6OfHV7oayVmHV3qg3PgnC0TiZ/TySR9YikyLzW3Wg0z4TRFjb+KiYXcOmH5CZqLbhWELjp0q5qctOu4aJgG/uOrzoviLrtm27U79KlWzQ49zxhjVzHGrgr/ej2AJwA8BuDTAH5NdhxKhNmzOTnB9BF6WHvFNIIKaBUpl+7Qi9QZQD0cumh0n7rI46mTqzhnyyRaDdGhJwtOqVrk0GsKivZSHHoVJ7zQdRKopVUxEzP93MUTWLuhT7lM1ujQ6Rop8AYETl41EQiI6Za0Q7eizlny8yU9t1qWmTjh6MeF/Ph02AjKJ9C60wE+Hh916ERf6ayP4Bq9kYD3ztl2bX2Rf/7FZwFQVwopzTrO+bcAfCv88yeFn3MAv670zaGdCtP9qZ4EmW5ZS/p3tmlgqmVVR+iDJEK3U5QL5xyM6R+ryXnqbGAH53twPI5zt0wk9PYicnI9DlUal9C4KFusksVLtBc5Ods0YBlMK1N0vjvEeVtjZ1Q1cSftdBIIXZNqWg059LqM5gbxvmQrAxeznUbWP8k1ao1HtdrJ6L5VHIh4QjANlqh5BOg79IHjRSfOtm1i0LCE69NrkwckszuJGtL1M2KxObJds20cWx4E5UIqquF+5vLdYAzK9YTWPFOUHNHmySZe89xt0c91J4MjBJCmWnYl2WKU4djMR+hVU+JjykV9Yu09GeiJ92yZSEwuEXSoHKHT10T0UtOqxqFTqjn16wQC9KuD0LtDL8o0BWI+XpdySS9occNuWaYW8u+7o+itig3dJEKfDU8oqwMP/9/XH8b3H5PPCKax0inw5IhVMhNFdNuyDDDGEiecaiqXYO51QoRuGkFeQF0cOl2nrlY+C6HvmmuDc+CIQpnfvOc92bS0ei+suUOnAjebJxq45h0vwpd//WUAqlMuMULXp1z6TlBGtpNA6EbqM9UKRNGi0LlfkpnNdexcDl0H0bheUKSKTh4B5aK/cR1fHmAqpcnuaDr0oFhaPE7Ex2sGt9JHeDGo12qYWu83XdCtqtFmvy1E6DtngoDmYs/Bx7/5GP6fa+QzgqkGeprjp41Mpda/iNCJIklSLupzz/F8uD6P5vNcpxEJHGzTgKMBUOgyzARCr0a5dIfuCELfHZY9OLAg30Erj8qc7dhoWIbyM1xzh35iJVjsNCFoYukHRQXKpVmt1VmUwl1AuVSpaiiaDvKI4gWWET0/APj0L74w+rPOhHV9nnBsTatapuix5T62CgE9IFC86FAu6eMsBcu0OfSRoGgSoeu8X0coIVCHpRH6ztBxpDM+ZYzmQ6eZdEb0vlUQq0jt0fwzK6pceqkA+gd/6iJ85GcvBhA6dI1Aa0S5CK+EsYAe0jnBAsmWimT0Xg4qNLbOAzWz7QZskyn7wXXg0IcJ/pwWqHamqC9SLtU49O4gmbIejFsvQifTuV/6Nw3LiBZAwzRw6Vlz+OO3Ph+AHuXieH7i6NyyqwVFjy0NsHUy6dCDLEy1d8M5x9Dz0RCujbTeTxeHrnPfQ9fPDJ6bmhLGYUq2uGs2+P9jYU2WLZPyPDod8dMcP71vFUToJRB6cL9Vg6IDoRY/EMQNds8FfL8dNpdRtaygKF2rLkLvO6OUi04yI9VteduLzsRbLt0VXVfDMmCbBjyfK9Fga+7QT64MsFlY7LQQqvYjtEwDUy27EuXSdbIQesqhKy74vPiplkMX6CU6ojZTC0uPcuGJ+2xqJtiQHV8ZRHQBWaehzk8PhROJaLp8PJChQ084dE1ayMsOiunWAXI8HwYDnrV9Er/0sj34yYuDAliPHyOH3iz65wmj+ZCmC6Kgo4psUXA05NyMipRL3JR9NAZhm4YWQPEygqJA6NBrDIrS2lNB1QRErjhvM16weyYYR2hID6j5wnXg0IcJhKFzE6LRJGqYBrZNNbHQc7SdOrUSSwZFq1EuedyqLhIEQoQeJQFR71N9jlAsUgXEskUdTTvnHMeWBgnJHRA4FNXEIvF+RWvbZkSPqdoIQh9J/dc7OYmnCJrfuiWhaYOwTAN/8JPPi2qDE0JPK8SKLELoKQ7d1AAAyaDoKIeuA1IoViNmZ5PZpqGlbfczgqJAsHnrvpNuRlA0KjWt4BPinA9TAGXxaRtQ2yDW3qGvZiN0/eJcMeVy0e4ZcA7cd3Cp5F9lW5zhKCB0qxrl0szhVhd7jvLkihCraUQ6dBqfOFFPA9G4Hh9x6D7X2xxWBi56jjfi0HXQr7hZi7Z5soGTmt2u0g5d3LxbdrDgVcGF4yUplxve80q89sLtcH2utSmmg6zkjKna5FRTPvkkyiYeQejqAEAMihKAqCpbpHWftU5sU4/zjiiXDIReRYmTLutghNJNFXDWF2IGtEG0wnuP2AoFX7imDt3zOU6tDrFlYhSh61ZqoxdkGgwX754FANx9YEFrrHRRKSCoHS2aqkNvZBwlOw0TPg82NxUbCvLCEYSukShC5vp8hHoA9CSaVKxoawZCV6ZcIoSefIbbppraNTTSz0esl62btJQO3G6dauL5u4LjtA4iTFM4tmkkHJ6KE6YNvpNC6DpBUfF76d9X5tAjhz66TizT0CvOFQVF0whdj3L53C37MN8djsxpQD2rui90BxtB6NYGQ+g9x4PPkahR3KiI0IceR8MMNLGbJho4c1Mb92g79GQNEqA65ZKFPKikp2qltqEbUCOGwQSHHi4skxC6nnRMDIrSNesEHqk64txEkhaok3LZNt3SaiwAjC4WcS7SwlK5bz9sWpKOtZgVKLAsGaSYiari5Oj70+hSJyjqC6cNQuZmVQ49fNZZMQjbVJfxAQFCzwpIW4ah9T7+/pan8OztU/jll58z8jvVrGqxVHU7dcqOEfoGCYqma4YA8U386dcewsOKxd2BAGGIzujZ26fwxHF1eRcQc+ji8dQ0WCKwWUX9QXaGpkN3BOTWbsR1VwARoetRLnZKtgjoIfToSJlyIG3bUqZcqDJeelPdNtXEydWhNmcrLvapjKQlFU40L3BbRfc8dP0Rqk9MrlKhIVyPw2CjfLJOUFREt3R/otxVR3o8KKFc9BKLRukWIETomkHWszd3MuusNC21ZCCxEF7TSiJ0HQn3mjr0rOOVaTC8NZTv3HtwUXnMNH850bS0NcpRDRIBDTHGEuOrotasCbRjOtCv6iB0chytEYSunvknXmMCoUeV6TSyJvMcesNAd+gqccp5i51S4nXaDQ4cL7Fhi3ORnqXKOxZrCYmmQ2mQdYejfK342FTrr1jG6LLXCYqKCJ3ojKrlc4cFlIutS7lwjoxb1pYt5j1DIHDGssDH8fyo72/LjhU8cVE8dbZibR16VAQqeRm//8bnAtDrlO6kjrudhhUhbVUj5UR6MYlKCFXKJWvB7JgJuLh0N/YyGwqbV0S5WJTgoc+hOyNBUaIeqieKkHUaFnyuhj5yKZcK3WIGrp9b0Y6epQogiAK3aYRegXLpDkeLfR0R7lUpGcj3M+kHOpGpXJ/4WTPKKjbxu69/Nt508U5wrg4oilUuTC8m5PFshG4YWhus52dTOECwkcue6L5w2378/S37AATrd7Yd0JKXnRW08dORcK8ThJ68jImo56S6I3ZcP3Ekn2iYWhsDECCjlm2MHk8tfYSe9XKmWjYmm5YGQucREoz4NzvJv+kgdC8VFG1Y+trGz+gAACAASURBVPRNurkFmU4NlljlkhyrSj/HvuNFtcvTRs9UZSOjTSfNoccIXf19rA7ckSAmfc+ezR1lJ5ylh9epbSLOLVNYc7/26vPx3LC/geqcGQgdzNJWBaFncugm0z/B5jj0pkIymvjVTdvEhTun8cVfeyl+57XPArChHXpKoG8ZMJgmQk9RLp2QctGpHR3UaxitmmebsZNfVeSBafGJiMsyGLZONdUduudHm2EuQtfNFE2UGtXPDUj3JyUjZ6kSGC1D6DqB0YHr59YuJ8pFDaGXcejqz3B16GEyZ9M5Z8uEcnZnVpMWnVouojNMOzjdEh7DQoRuaNdDzw6K6iF+z8tH6Cp1j0Sqj+baZWfNRWCqsdEol34O5cIYw4QmVeKkJmynYYJz4Pe/fB/uPaDGyQ9dPzs4E7bG6jRMdCVrxRxb7uP3v3QvukMPV73qPFz3my+PfmcaDFsmG8oc8ND1Rjj0KKBCOnRN2aK4KcYOXX2sfMpFw6F72QqIzZNNNCwDX7rzIF75Z99USiTrO15u7fL4pCgPLAYRQh+VyAEVEHoKWHzs5y/FO644Gy3bVJYaUr2Vf7zqCtz4268Mrk+nlkuqfK5oUYKgYqC6ULaoqRvPVbloZp66OZsioCZbpOd33taJ3BMJsMGDomSdpqm0kMicdBJG6Dg+f+s+vONvblUbyxuVnwEB5WKbRsDPSzqkq/729ogvS9eNtgyGiaalJeOj62taBhiL6SuzAiJ0vSTPGqE3Lcol1tmKFmvb1RF62lmaBsO5WyZwx74F7DvVxeOSqibOeaFDJ25dZYOgZ5QGAvH70HPo6Wt808U78f43X6RMQ3hCfORFezbhgu1B33edDcfL4NDJdEFARLlkASlLX4eeTvsHgndSJ+IH1FQuFFT+7C//u8yeCg2NZ7g+gqIZL2+iYWmlc49QLgKyycvSzLOh5484DyCYrA3LwERTjp/3fY479sVaeMswEtdoGoZWOVmxqh9jDC88ay5q5VelI0uwkdVDufQcDw3TSHDyQPwuVKSQRZI2SocH5GumOB6Hz2Mkns6eJAmjSrGlPA5dN6bBOcdqRt0QMksxUOjm0g/qQVEvQ4dOpjtnCjNFdSmSnHsOMk/rVrmoc+hZAVtAj3Kpr62KhqU744img1iB0eOQuBBUmw44OVXzbNNAwzRCWqh8sR9IldO0TBY1vQUCB9Sy9RJtxNPIP73rpdGfdWRoZF5qwlahXPpCswLRIm27RsAxHRQFgAu2TQEI2tjKHlFp4RH6TQceJxsWGINSk5Q8Dj1WHalTEJ7Pc3l+21CjDbyUJJUsVuHUQ7kQIFjoOjhzk/SQGLhBIbKsTVk3KOrxbIQeJBap+5hSlYsi5ZKHP3ROxuuXcmnoUS7p0qXiIlVtC5ZG+2S2ySKELsPzjzZRYIkgoWkwdDSaKQy80YST+BrVERcQIEInteirUC5ZZUYBQduuQLnkSQIB4ILtMUKX5W1JvUKLPY3QDYNhsqHWJCUfoevlBdAaSF8bmWr6eh5CtzVason38qI9Sa/94nM2YaJh4k++9qByrkHTMjMpCFuj4QMQnJDzgqI6ssVClYtC7wCiXNIlCcg2HIeelSlKVgWhZ3HowGjJ0DJLUw9ktmnANlnQpEGGcklN6DT9YJlB6r7q/abjBaLFCF1+Mtx3cBHnXH09nji+mhMU1aNc0pJFQI9yGRbwq5eeNRt/TvI6af6duSlI7PrNH7tg5DOqNfXzMkXNiNJQe4Y0J3IRuiqHXrNscbpl4fvvvRI/FSYDkp0x08Z/eu2z8f3HTuKu/fKlN4r6cdraQdHsWvS6ssVCDl2Bcola4+VQLhsvsehpQOjpOiQih563KPJsKIPQJZxwGkHRJkEbjGkwtBuBvFIFzYiyxbTpJLJ88NoHoj8ng6IVHPrQG5EsAnrlBPKcJRA4kK8oti+k79400cCTf/JGvPWy3SOfUa2pn1cR0takwKjjVt7cVc12dLxY5ZIYR2O+UL7CrrBTT9p+5vLd6DRMfP7WfdJjDlwvd04HskXdoOjozy1Dj8LJ0/IDarLFiHIpQegbiHLJli0C+kHREcqlAkJ3cxoV7JhuYcd0K0DoEptOGgUQP02Nei2huJZqEksemlGVobmej1v2nor+nh0U1ZMtZjv0EKEr0Ex5Khcy1QWQp5EXTRmh52jldVUudALMdeiK2mwvhy7QSXxyc9QjZFMtG6+4YAtue2peesyB42f6A4CqLdaoQ9cIivo+B+fI3BSBAKioqlzy0D7RqRvHoafaTYnWaZpRCzgVCzTUAkIXEjKy0HaR5ckW//itL8DHfv5STEr2LPVGKJfg+igwajCGtkYSy9DNVuGI3yE7YRd7SRSaDIpW5NCzHHrEoesERbPfY0ORxiFAUadDF5uUi2ZpxjRWMgrEiWabDI7vS5/s8jh0KjqnQgn5BUiVbKJpKVEGxKFnWUPxXsnygqI6skV6f/k69KCCowyQoqmQ9wh1Ks+uOeXSsIzMAMhkM0Doqi9vpDhXQ6/MaDzW6LW1GyY6DSsqAVt2jekmEzTmdDu4Nivk4wG17Fgn5wQBqFf3W0g79Bpli5lBUQ3KZeD5UWnkLFPlHPsFgIJMlXLJQ+iWRkwDQHQCLOLQVWqm5HHoQBAYrUMCKVrDVKs+OCiIC1mK90qWFxRVVQgB5ahapYZ5Xmu86PqibNsNokPvO/l8GRVvUi3Zmu6HKaJDZU1sDodONtG04Pq89OWlf03odyYsxjN0edRxSEXpEsgWc9QPUcciSYfeTTqt2jJFMyoFAmJQVEHl4vLcDQxQv86nA6HHfV5Hk58AHYROKpc8ykVt3KIsx0Axo4DQczIwRVMN2g5cL5dy0VVu5WrvNYKiEUIv4NABOTluXms8sqCyq1ogeM0Ret7xigomyVAaoq0O3YTuWQw4qB6v8mSLZFH6egk1lEYBdopyWeo56ETFqtQQq23lLE5F3fNSCqHXlSnad/wSDl0lKOoVOvQ4EUNuk4hrURcj9KW+I31SzKOFImekOAdjlUsO5WKonZ7SOQaiqQZYpRC6otRwkFNuA9CvD+Pn6tDVE5WolEa+ykX+5JnXGk+0lqWWcLjmQdEihA6UO0vR+o6Hha6D7VOtzN+rFB4CAkSYx1EDQq2PEpokfaqjgApRLkt9RyhWJbeBcR70umzmbDiGwWAwBYTeS/bkFEsEM8a062gEOvTRa2QsUAqpcuh5x3FAvWVXFBTNARVAgNAdjyvVuBavhUy3FEOpykUxZb/ICavWNsmjMkSzFSmXINCfFy8INy/FU3u+VFO9STQ9n1KELgEq8lrjiaZatG/NM0Xz0BGhXxWly7Gl4Ma3z+Q4dMWJUIbQ6RhcllyUXiS0CN995QU4vjzAz1x+Jh4PO7jLBkXdMNpehFgtBU40TbmktfK67b/ydOiAeruuoHNP/uSPTxKylEvIoRcgdOoxutR3CqkZ8RqDa8lOLFJF6A8fWca2qWaBPDV0cpKOOE/lAlCijQpC9wvRJRDMz6EXBDLzYh+iFSN0PcrF83mm09QBKZF2vEDlAsjFcfKKhom2Y6aFw4u9ws+ItuYqlzzKRUdUf3Q5KJ9K9bHJbr76x3DRrmlldFTGoZOCphShp47rdEzeNNHAx99+GWbatnJ98Lzgm2gBRyh3zwtdB4zFNFCaZ7VMptzijXNe4tDVGuoOvRKEbqqpXGQQOp0UZd+L4/lgGanrtHBVECHnHLfuPYUXnbMp1xmq6tvdgtKvqpu25xejSyBQpgBqcY2y3ArVeejznAYXGrJFaQ5d4hp9nq9wIdsx08KRRfmy0GtPuZQEQFR2ULrxHSmHvmOmhU0TTWWE6XrFQbgJSVoovdiyglJ0IpFF6HlIUDRTgSNc7DmYaloRkkxP2IZGqdGB64NzRAHftDUt+e4uQPFxHAhpHIUjvtjPMc9UCyQNw6B82gHbikgaCGoAHVnq48V78ouhWIrcfF499GAshp7jKvDx+WifTHUdDwviatR7V7U1ZX499IByUVHSeSWBzIYi5VKk4weAnTNtHF0eSAOBdYDQizXFSgh9KduhA3ppw3myRbIIVZc44RGEniWFtNXqg+dxtcnvkecIF7pDzHYa0URNB85s01DqPg7k9xMlU0mTBgJnWXS/gFoj4aLSE2Sq2vahmx3X0EHo5LioJVmWqQYKPT87UxQINvHr7z2C1330O3JjcQmErpgcQ1LmLHvpeVuwc6aFz90in3ladJ06jbvLdOgqjVuKSgiQ7ZhpwfO5dK+EtUfoeUkEtJAUnPDRpT6alhEFG0VT7Xbi+3yk0UPa4kVafI3pCZOlMlCVLZKDKaIgTEO+q/lCz8Fsx46OpulNx7bUN8ReSSamKuVSJHMlayjUoybNc5FTUq054+QUTNMpfnVyNQhUb59u5n4m5pUlM4ILdOg0L5+QrCevgtClT005MlcgmM9vvnQXvv/4CSUg4PscWf6XTjcqmyyt9TxHTNnfSz25hMOyGASdSg5L0i7rQLaYg9A1ItpHlwbYMdPK5ButMMtM1uizRQ5dNgg3kvpfA0IvqmsSXZ9CkGuh62CmbUe9IUeCooahrBLK6ydKplL3AgAWuw5m29kNnclkHfrQ9XHPgYVSxK9MueQocei5ysY0AGCxGzj0mU7+PVvKHHp2k2ggH3XmmeeXOyQV5RHnHF0nv/Y7AJy/dRKcAwfn5QOFRS3oAOBzt+xTyrQF8qWGs+G7SqvGsszPCdaKtoMc+oLc/a65Q89Db6oSNADYP9+NdrS02aaB/tDDLU+clBqLnHQR5SK7w6d/n1dj3TaZNIdOqpTpAgdnKgR9lnqBQyeUNtoj0lBWCZUjdEOplst8d4i5TqPwM7KJLF+84wBuevxk6YmoqTgPuzmZsSrHe845PnTdA/j2I8fRaZi5p1hAnaMuQuiqEj4ZykAlfX3oBbXfi/oWnLW5AwDYd6orfZ25OvRwbX/g2gdw/6ElqbHcEh06iQrSpTSyTEblsmfzBCabFv75joNS17emDn2hO8RkKz+lGVBLmHjo8HLUbXx0PIZDi3383Kduxp37yosFORJBR9nknVHKJYd/s+WTCIhT2zpZcBw3DGmH3h16mGhYMYdeA+VCQce8Bdq05SkXzjkWug5mJ8oRugxNR3TGJ37hhcXjmfIyNCBoVJ31TlSQ9NDz8env7sUPn5wv3cBUM0WLnLCYESuDWGUcukr2btmJDgDO3hQ49P0KDr0MoQfXp1YhMe8007JNNCwDi91yhy6jcploWnjXq8/Dvz14FI8cXS4ds9ShM8ZajLFbGWN3M8buZ4y9P+Mzr2aMLTLG7gr/e1/ZuD7nmO86OHOuk/l71aPukydX0XM8XJjj0EUK4dRq+XEor8hScky5xeRLcOhA4PhUHfqWAoduKhTwp5ordJRMX6NlGMoZeqVBUQXKZXXoYej5pQ5OVuVC3/ua524rHk+xRMHx5QG2ZnDeKqn/fSFbeLaAbgHi96SC0PPm9JJQs+acq6/HP99+oHCsvDrjoqkERYluLKJctoaafBWE7vEcHbrwHGT9TES55KxhAJht21IIXUblAgTNQgBIJRjJIPQBgCs55xcDuATA6xljL8n43Hc555eE/32gbFB6gLvnsmspq1YaoyPT83bOZP5ezHyUeYjEF5cFHYHyRTqC0HN2907DQleSgqCXu3ky38EFmX9yyC1y6EZ2ULShGFQGZDl0ufudDzfhTWUOXbKRMJUeLkt2Uc2HOLY8wLapUYdO2bYyG6xIu5U5dNWEJVmEDgA33HekZKx8Pj59fTIbNzn0IsqFMYazNnXUKJccrl9E6LL1eiKEXnDfM217JFEvb6yy5wdAKUel1KHzwFbCv9rhf+opgykjtHfmphyEHrVfkvuqBw8vwTZZolmwaMmmzBIOnSiXosxEyZrjox2L8o9rKgh9tmOXUkIyDsTzOYauj7Zt5ssWK6hcslL/AbV2XbRAyhycPEIvV8wAaifF1YGLlYGLbTmlJ2SLQSUdehnloqpyKVemkG0qobc8v16ETie6TkmryN1zbRyUDBIC+Vy1uA5la0a5JSoXIHDoshy6DLhsScqjAUkOnTFmMsbuAnAMwI2c81syPnZFSMvcwBh7XtmYpQhdERnNrwYBs9xysoLjk9kiVCiXskU60rEo57im0lf0xPKwkD+n65M64gvUSB6Hrtrd5fjyAE+dDORvuUFRW55ymQ8VH3MT9SF0JYcuMeax8NSUhdAB+VIM4qZepuqJdOiSOQKyqBAof9aeTOq/wklbhnIBAgSvkp/ieflNosmWJR26DEKf7chTLlIIXSHpUKqWC+fcA3AJY2wWwJcYYxdxzu8TPnIHgLM55yuMsR8H8GUAIw0aGWPvBPBOAJjbeS622yY250waVYfeHRbLnRqCg5KNuANlQVG5gI9YQ7nI+Qd9ReUm1omVQSF/DlB9a/kjfqthRhM1K7FIpS7ziz70b9GfyyiXd3/+Tpy9qYPfed2zc8eLHHopBWFItS4sqvQpWsShS5wkjoWJbdtydOPB+1ejXMqsTh162khTnWeezyM5Ztn1yXHowXsrolyAgL9WapXHObKWsfgcViQpl5hDz7/v6baNBw+XBzBlYhAAhG5mNSF0Ms75AoBvAXh96udLRMtwzq8HYDPGtmT8+09xzi/nnF9uNprYPdfO5TDNsFqgLCrsDj20C45qqgGQvN6QyTHVEov+x9svw4v3bMqVGrZsEz1JCuLEygBbcpAgmawmW+S6aYKlDxENS69DOlCgcgkTi/7l7kP4+DcfKxyDOPTSoKhkoLWofV9iPIXO6zFCz6ZcZto2vvXIcex573W4ryB9XVy4Zfeiop4pa5+WtvLGLfI6dClgIYnQrRJglLaAGhq9Z9EnrAzkmphQ+dw8YQMAzLYbcgidc0gwLvVy6IyxrSEyB2OsDeA1AB5KfWYHCz0zY+zF4biFgu+B6+G8rdl8NxlVapOx7tAtnAgihTD0yh8MTcCiZItYtiinQ3/1s7fiC1ddkbsrdxomepII/fjyAFsKAqKAvIqkl0G5pPco1UYFouUVv2oqZHXOhxz6jERikcx1ynLoKpmiZZTLuVsn8NTJIJh3e0GfTdGhlzlMFQRclrb+W6+5AGdvjmNapfOal6N9uj6loGhJVUvV1nH+jxihz7RtrAzKa+LkBWvTRnOwLg79DADfZIzdA+CHCDj0axljVzHGrgo/8+8B3McYuxvAxwC8jZds7wPXx7O2Fzt0lVrK5ZRLfKsyx2cZHTpjTCq9XoZ3A0IdusRL83yO1aEn5dyUEHrDyK3brVKKN3ENBan1m0o4WtEWew6mWtZIBmvW98mAAFmErlLwi1rV5Z3ARACT5/SBeOG+4oIt+M+vz6ehADUdellhqd96zbPw1d94efT3MofkeeWZjk1LjpYEECm8yigXWYR+5755fOnOA7knCVFhpcyhFwC9GepzUILSZeMZhsHQso16OHTO+T0ALs34+SeFP38cwMdLvy1lF2yfKvx9UwGh94ZeYc0L0ZlKLXgJDh2geilyCL3s5bXDHqVlJlOYi34/9Hxce88hvOKCrbkbgJjRSc8pvWAaFlPWoQPF1QV35GT1ZtnA9aTqkavo0IvotMSYChtj0zJy37Po0IueJW2wf/SW55fGSejoL+PQy5ozAMCU0EhDRo5bZy0XOp2WqVxk1hwAvOV/3gQg8CNZa2VFqJIqK1uUUblEjW8GHjYXYNa8DNYsa9km+nVQLk+nPavEoauUQu06buFEUOXQXQkOHZCrl+L5HAZDqea5LalykSnMRb9/6mQXv/G5O/E7/3h37ueSlEu2gwiKm6k79KJzWlZVzDxzPJ7IJcizQF4pkY3p+oWNLURrSOrly06J522dSHx/nvUlkSqg1ny6LG0dCNDgT1+2u/QageIkJTKV9oWylIslGVwmG7g+NmdsjOI7laVcSOBQtJFF3dac4jFVFEeyp/c161jEAJyzZaLwM7YkHwqEzYhlKRepoGi5Dh2Qa2MVcI3lzqNjm3A8XtopSQWhkxV1PekLCR3vfcNzsNRz8PLzkzFtXcqlyFQQuuv5pXQLEKTqyyL0qZyyE2mT5foDh54/pghgiuZgryTDVjRTUjoLlDdnIPvIz16Mbz58rJRKdNzieQooBkWd4hMOmarKBUCmmu6nL9uNR4+u4L6Di/I6dIlNsSNZQlem/DBZW1IwsWYIfdt0q9whqXLoBQtA5LykVBCSlItMG6ugBVbpV0rrTfMaEadNfL5FygbRgZy/bRJfuOqKkR6WtiblUmRbJpvRwigLUDoFjRlEU6m2KCNbjMaUTIwpajg9N9HAdb/58uj784wahctQTLYC5VLWPi0xrll+8hwWNCmPx5EPivZKTjhksglaomVlVLdsE3/4pufhjNm2tEOP42H5zzBqVlPi0DnnpbVcxGutReXydFlRUIhMdnFSq7NilUv9ssVg3PKJ73pyCL0tORFkuhUBSYdeRFfIIMKGJuVSZKbBouSoqRLNs+v5uQlZopHzTdfPSdvA9aSCooA8sAiUVsWo/9wtAalayKE7wbXJHMdVGmdEHLrExihTu8cpaQkICGWwJSmXsucHyHPoom2eyPc3Uy1LgUOXQegh5VKG0CVVLoA8HbumHHqZ2ZKKhb4TtDrr5HRGB5KJRTJ8qIxsEQgmfmlxLsmdWFZvKku5iJ1zio529H15reKA4D59Ll9iVRZ5UEPvMhqgqBemaLLlbmUzRQF5YNEtof1oLKBYadUv6MOaNhUduqzaCihH6J7P4fNipAoE884y5PrR9obFJxwyVR06UFzzaKppJQqTpW114OJn/+oHeOzYstQzjLsWlXPoapTLBnfo8guJouPFDomsaEzOOa757hNRf9JSysUsly26vhz/K9vSLgqKljl0wSkUTcCyqogAMBE2xJZJmADifqtlRqUfyhCc4/PC2vRksrrxoScnWwTkk5XKTolAgOwsgxXmQvQKuvakzTAYGJPLFJVBl2R2SQ9Z2RgTjSWbKSqL0GV6gYrvogihT7dtDF0/FwEfWujh1r2ncO/BxfgZFsxFWQ49r3l1lslSLmsWFJWxhmlIpcLLRMdNSQ797gOL+G/XPRhfQ8mit6Rki3IVHmX7EUYIXfK4C4x2IBKt53iwDFa4eZHE9OEjy7jivM2F3wvIB3uufsNzcHx5gAcPFzcY8CQ3ReKdg1NYPo0zcOQdukpQdPecXDmBIoROlS9lTWYOAnL8bzSmaRTWh5GdgwAle0no0CVOOAAS0tqiE3TLNiPlUdG4lDew1Hcy4xZ07a7HI3VNEUCaUKBcmhIbIvAMoVxkg1FxUZ/8/UnkVIsDUsmHVhRoBULKpTQoKlfhjjakn/7ETXjyRH5fx6EkQhd/X/T9vaFfigifuyNw6A8dke3sElzjrtns4mtku+c6uOTM2XKE7snVIGlKUBpA0KtWPihqSjXNCCgDOYdexqHLjENG3evLLFZolI9pl5w8445eMgFWuRPO6tBN6ODzLE9amzbqhlVEtwBx9nFeH1B6tm7YZzi4hnLKpSzr21dSucglFq1vh64QjAKKKRcRIRSNKU7ihmmUokKZoKhMmVEguSF9+5HjuZ+T6Sea/n3R9/ccr5A/B4LGApsmGnhIougQEFAkP3nxTlz/nleUftY2y7XjbomUk6wZIfRiWk02UxSQn4cylAtQjvgDDl1+aVoSuRB//b29+Nr9QX3zpox6poQmkalGStaUlB8v993cDmai5SW/iUZCCaCYbgGA6fA78+hE8gmuz6VquTQsA5bBsCpBucjGmta9Dl3GbMmjWk/QUeeZ6KiLgqLiManTlDv+lVMu5Y0AgGTd8J0FyFZ2MYnH4SKuTiYIxxjDc3ZM4UEFhH72pk5peQIguA/P54WJFq6kbDHm0PPfMW2IskFR2UYcshxwGSdfllORNtMsT7T5wLUPRH+W4ecDOW7+vI6VVnIBVpkNcaXvYlIKoZeXOxh6PujXP/ac4q5UMULPc+hEufjRn8uWs0z3MRWVS0uym9kzBKGXV2kTEUzRcVeUL5XRLQB1BSqhXLicskCsFlmEPrR06AULTzYId86WCRyQ6LbukwJCsou8TIEpR1L6SU66X0C5kDNVcehl89D3OfpOOXUVjFec/NRzvNyCZllmGSzqsCVjMqeIhlVMJaogdNmaTMsDt1S+CshJNamN3/t+4kK8+8dGKnknTOTQs8xNcOgB9VeW9d1plJfCVlW5DNxyOe76duiS6gIq6lOEjq587ja8aM8cds22CyfXsvBSZQM05ZSLL/XiJoTvK3TopJEvCajI6tAHrieVBr9tqoVTq0OplHBAbhMD5LTKAeUig9DFoGi2DRUduozaqqeQrt8wixG/FoeukMVbB0JX4dCDRiZlyi0PQ8ns3bwCcqKpvA9C6HmUS5pDlzltTzSsWlUuUU30kue4vh26Kdf2rCfBoU+3bPzjVS/FBdsnCzcJEaGnsyWzTKYrEO3qZTbbaeD9bwqaPRVN1hihF09W0WEVqWxkmx5sDZPBTq4WN6uNE1jkpldc7yP/OQaUi5zzAIo5dFnZJ5lMcL7nlJ8SxWssuj5HQVIJqCfaSAEVaQ5dLthfdGIC4loqMpSLDIeuUj6BGnnkUy4hh+75CvRpeaE9pVoukkmH69uhS8rF9p7owjaZVDnWMhongdClkIyMykW+qtqrn701+jd5pqNyKVrwjmSdFHLoZd3HHU8NodsS9T4cT04pRFRFkcpF9vmRyVB/ZQ2x0+MVOXTXk9Pck1EuRJkum0z2GqUcusQzbNlmKbKk1HslDr0AAETJcpKqo7Zt5gdFPXWELkO5cAWVi2xf0XXt0GUTEn7w+AlceuacvGRMEqFL1ZWQLJ8ryyfLTFZZdNRMOPTqDoQc+rGlEoSuwK+Knyt6L7KnnBih5098+p1KLZcy6k9GOhtfYzGHLrvBkpkGw9fvP4JLPnCjcPqrpAAAIABJREFUVOBM5hrLTp6ypTGA4DmXIXRad1IqF4mCZCqUCwBMt61c2WIUFPV96XnYbljlQVFFlQtQ3oZuXTv0hhWk1Re9uMWeg3sPLkoluwBx27M8Szh0WcqlTHKngNDjHpE1IHSBkincIHwuVbCJ6u8cXymjXMqbAIgmw6E7niTlIpEpKhtUjsc0IxVOnslIZ8nKEL/jcelrAwJQ0Xd8LPYcqUxemdhBUC5ZBlTIIHQj0oTnGSH0unToKicmIODR5Tn08nuekKVcFDl0KtyWZ+vaoVPH81NhP8kse/DwEnwOXHb2nNSYZWhLjHTXpnKR3NUBMYJf4JBkZYuSlIvn+1K1xilBo4xyUQ2KxiqXIg5dNSgqoXJRqIcOyCWkyZwSy2SQsvQSmWxZCyBwDDLHfNssrq5Jv5PZtFt2eZYjASkZlUvdHDoQ8Oi5KpeIQw906HIIvdyh+yoqF8lKrOvaoe+aC/obHlrIl8rR4pyU0IwDJEHLfyhPF+Ui++KiYks1yxaLFSSyGm8Tsx273KHTYpdsRizTBEG2YmVEuRRMfFWELuXQVYKiJUFW1+NS3DRZWb9ckVuXpSDKYkOOwjNs2Qb6JRsNNWmWoVzqVrkAxRUXY9miXyuH7inWcgE2uEPfORtU4jtY4NBV+dqyxsliUFSKcjGKj6aALkIvTpqwTVa6SYgOvWg8Fc5262QTx5b7hZ+JgqKyOnSJComOtGxRRuUScuiS6C2qkFgABIgjlo3j5AVtOefB+1VA6KKDybpv8dXLItZyykVettiyyhF63SqXvkSyYeIabTP3/YqUy3x3KHWNAeJ3CwPVnq/W4ALY4CqX3bPlCD1WVNQjQVNOLDIkmkRzeXkS3UehBliiUwyQRE9Fi9P15dq7AcC26aYE5aK2yUYcemFJBskGF2Y5Qo+dryQIkGjSQMhYhp8uQujkPGSfHZDql5txjeL8lHVwZZRL/I7lKZci50ZNmtV06PVRLkVAL84U5bh7/wKev3umdLzZjg3P54WNMzjnUnV1gPi9beig6HTbwkTDLMxOVNHDAoHD5By5GVfiC5DT6zIcXRrglidO5n5GRW8qy6HLSO5EjrhM5aKC0EuDoqqyxRIOnXPK0Cu/RsZY6SlMNWBGz7HIwRHilnkvRQg9Pt2oqVzIshy6iGRlKKHg+4upRNkmK0Cwcfq8GKSs9F1YBpPaEGkeFCH0+e4QjMVln8usaZm574TWzt4Tqzi5OsSlZ82WjjfbDuJNC938ILWn0CRatrT2unbojDHsmmsXInTVJBZCeVmTNd35SCqxKFxMP/epm3M/o0K5yHDoZT1HyRIIvYxykby+rVMBQi9CWyoKiOBzxRx6fLyX5xuLHDppomWzMek5FnHoKuUEmpaZuzk4CsiXTHzOWfctvnvZe6b6OnnAJ5ItSurQgeIsx8Weg8mWVZpSD8hJex8/vopds21paWpRNit9z61PngIAXHpmuQBjtlOcfQqo5ac8IygXIChSVcShqy72Io564AadjyhBSa6/YfkjVHlxhsFgsGL0MXDLW38BSYdenoUp79D7jh8dkbMs7l2phtDznBxt2jJyMaBcRRIpUhR06ECZQ5fn5RtW4Cyzgo6OAvIlK0XowrtX4dCBeIMZuc4o8C2TG1BOFzx0ZBnnb52UujYZHfpjx1Zw/ja58YBiEJD+HplxZzsxQr/h3sOZSj1f4eTeCgv3bWiEDgTOtWiXcxWCM4CIgEdfXj8qtxm8DKlMUcER5k0wlcSi4BqL29rJtk8zDBZtdEUUToDQ5Z7ftqkgUF3Eo6sGRcs6w7u+2qbdtIsbSNDCbTXk7rkhEWhVqQ9T1CbPrcqhF4wJyFMu9KzzULBSpmhJjXrH83HfwUVcfGY5lQGUq1w8n+OJ4yvSGwQgx6EDQZVFmVMJIfSDC1286+/vwJfuPDh6nQqxtYZpwGAbnEMHyrPMVJACUIzQ6XsIoatQLkDxBJNF6HSNZaoU1YBjWX9IWWcpk/6vGhQtq7aoysmXJY/1HQ+MqSUWAeWUC2Ny11iE+OlnKgBAhUOX71Va/E6imvwKXaTynNHDR5YxcH1ph16mcjk438PA9ZUQerPg1CT+TLZcBOXQHA2zqrNOjL4v18kMCOjntkQbunXv0FslldriZs6qCH10MtBx5nk7ZzDZtHDWpo7EeOXSQI/Lc+h0jUWabJXmDK+5cDsYyz86A4pBUUr/L3LoykHRkEPPaXnmKr7jpmUUIhkqFyzD1wKCAy7oAzoIT00yYxZpimleKmWKCp/NdOgaOvSygmkqJ+PYoWfPwfsOLgIAXrCrXD0ClKtc9p4Mun2dq4TQ8xPSxO+R5eSpJO+JUECQBah8BZULENZYL0Ho67rBBVCOtlSP40bB7k6730W7ZnDf+18nNZ7otNKT/4PXPoB7DyzC9eQTiwBqWFCsAZZ16H/xtkthmwZueuxE/niSmaJAoHIByigXPdlifqBQkXIpUbn0XbXytFJBUceTXuxz4XF8ftXBGTPJRiYxQFEDANF1ZDzDShx6bqDah8Hk4iQkD80LilKG5pap4s5CZGUqF6Jo6TnLGCmZ+o43cjIXv0d23bVsE23bjB16xrWqqFxozGeAQw/qXuSlyaomFskgdKUGvWY+Qv/M9/YCAM6YaSmncpdx6GoqiPwGCEH3dHn0O9uxYZus0KHrBkXzKRe1zNMABBQhdD/idWWsiEP/rX+4E2du6mDoycU1AGBT2BItK1CmuhkC5ZSLSAXKyxaLqbqhAu1XRrnoChvy1shqGLCXoUzJihLSxNOtyslprmPjxHLwjtNUDufBulN16BueQ6fJkIfehorHe1JKZDUE6CsmIwDJBZSXKi2bLkxmGSy3YUFv6OHEygANpY42+U2EyYHIXh9jDJ2GVdgAVxVRl5XPVQ2yltUb77vlPVQT4xUs9i/fdQh/+Y3HMHDkaTCqiZNVV17VuQHliUXiuz9784TUmBHlkqdyceULiJVJ7iJNu+SGHXPo2dem59DzKRdxLcpu2gAw02lECD29/lRBD4BnBode1sXdDTXUsnxoNBkydNRxgSX5x0I1KIB8xKCSWAQUNyz4nX+8G0+cWE2UKJAZry4FCVCebatfyyWHr1UMsk40LJwo0MoPFFu8NaVki/IInVRUJ1ZGEbrqiRNIyjmzEXrwHD7+9kvx1st2SY0pQ7nIbrAx5ZIfYJUpZUFWpkOn5MAJrU07P64ByFMuADDTtqIkvPTcJv+j7NCfKQg9j39T0VADxZmY9B0qCF3MLM2bYCplMoG4YUGW3blvHkCyPkeZ2QUlflWdLxBy1AXKI1cRURMyy3OYqkHWVz17Kw4t9nHHvoXM3/cctSbMsjp06YBZy4ZlMJzKQOhDjfchbsZZgVtCg01LPhBslThNFaUVPZdcykWylEV0bSU69NWBi5ZtKGXbxkXdik84Kgi9ZccVF9P+hv6qRLk0TPRK6sqve4dehtBVJhZQwqErlEAlE2u/FMkWZZNigGKEft62SRgM+NjbLlEYr4hyUUfoZUFHRxFRGwaDZeTXDlENFP74889Ap2HiK3eNan+BQG2hcgqLVS7FCF0WvRkGw9xEAyczETplYNYnW3Sj4730kKUF0xyPK3PoefV1VNdwGYe+MvAw2ZQPiAJlKhd12SIQZ9sCGIlh+SFCVwDoaNtGVHQsz9a/Qy/pQKM9GYo4dAX0Jo5TRLmonCLsgqa/rsdx+dmbpLlQgIKixVlwKmimUaY8UkTUQBz8LrxGyU1xsmnh7M0TOLyYXRWyr0i5xAW/6qFcgIB2OVkQFFVB6OUculqmrThmUexFXvFBCpJ8ykXFUZapXFYHrnQ5bbJCysXTo1zEGEN6PZ++lItVrGF1JQvOkxUd11QrtAHAH77pedg504quJfMafV85sShvc3A8H7YCegOKC5KpJmYBydT6J0+s4nuPJiWRrqIDBorLl6oGReka8474PUctKGqZBkyDlevQFebN5skGTmYUOVMpS0uW4NCzZIvhj3SaZhRTLvK1dYB8ymWoEGAF5FQuKgFRQEDoJZSLynWKzycNqGgtqvgFmaYZ696hiwh9qe/gD796f2JiqBz9gOL2VdTeSYVy2THTwgfefFE4Zv7xtKHY9Dcvgq+Spi+OB2QrFnRSzUU0fc33nsBvf+Gu5JgaWurAARfXclG5xlaB0mXg+EoIHShvGydbjoFs00SzRLaoqUMvoAyUlFZREbvqVKdtGmiYRm79H5XNAShXuazoOPTIz2TJFvUQepGkWU/lYm182aLIbf1w7yn8r5uexP2HFqPfy7YmIys6SvYcD40QjamYWVDBkUwFvRUjdLUNDCgOcJHzVblnsY3f0PVHeD3VnqJAcXEkHQqnZZu5nG0QFFV7hmXNxQeup7TYt001cWSpP3Iq0dGhJzoWFcgWVbOVgfw5PVTILgaAPVs6eOL4SvZYCvEHQAKhD12pJhSiFVEu4sYhG/gGku8wvfaIclFJOJxomlgt6YJU+hQZYy3G2K2MsbsZY/czxt6f8RnGGPsYY+wxxtg9jLHLpK+yxFpCBhdNdlECpNMhHcgvzqUSLCOzS46ngGIqt5GvSlFFM0DxAtALisbO1/NHsxMjp6RwkmgUUCQ6Tq4ocKvKodP1ldVDV0HoLz9/C/qOj5seS9bRV1UIATIIXR0NllEuQ9eLGn/I2AXbp/DosWyHritsyIszrQ5Gsz3LrDAoqs2h59d5or+qqN/aDRMFVasByCH0AYArOecXA7gEwOsZYy9JfeYNAC4I/3sngE9IX2WJiQ+anI+ojXVUOfTC4lxqcjYymR6Hsg2Jaby8gI/rqyN0+nxm4SGqJ68oW6S+rJ7vY+j6Cc2363EYTA19NIsQuibiz9ogOOfhxq1OuZQHReXHfOn5mzHZtHDDfYcTP1cpekVWpkP3ohOOjjQw+557Q7W1csG2Sew71c1MjFHJOgVkVC4VgqKOh8OLPfzuP90dPUud1H8gCUDS61lH5TLRKN+kSq+OB0Zbqx3+l36Sbwbw2fCzNwOYZYydIX+p+SYehcj5iA7drXEy9DQWOhCj26L6K2rBFAOu7+ND1z2A//SFuxO/GypqdoHiph66AccIofPkOPQ9qjx/SwKhK2vls6oZej58rqZkAsLs08KCaZ4SQm9aJi7fM4d7Dy4lfq5aiAyQaUGng9CLk716jqckHrhg2xQ4Bx7PoF1UKRfGWCHoWR24Us5PNJFDv/mJk/jCbQfwVFjky9ENiiaatGdz6CqgR6Zsg9TVMcZMxthdAI4BuJFzfkvqI7sA7Bf+fiD8WWUTK7XFCD3lPBRrjQPZxzWqwqdqMh1UdBD6/YeWcNf++cTvdCiXIj5Up4dl0475ZIrWD1ObrMo7CcYs59BV7jsPoVPgVcX5AuVBUVXZIgBsnmhisZsMjOpQYOKpPVvlouHQzWJpYHfoSdeFAYBnbQ8qHz56bHnkd44n17BFtLw4k+9zdIfqlEskTXVH/UySQ1dw6MLzzkPoKpRLpw6EDgCcc49zfgmA3QBezBi7KPWRrKsaedqMsXcyxm5jjN12/Phxma9OHIViDj1+wKqIlUBevQg9XzlD1jBVaq8Ek3Xo+onEJfoO9aBoAeWiIVtsmDH6pVOT6OxUj9BAgNDzgpg6Ko081Qw5edX3nKWTF2kmHYc+27GxkGreopqUFVxH/Oei4lx6CD0/DqFyyjl78wQsg+HRo6MIPQj0q4OULDqIgoaqQVHLNGAZLGACQkdOz00EakoOPdExLHmtOptsR4JGUpqBnPMFAN8C8PrUrw4AOFP4+24AhzL+/ac455dzzi/funWr1HeKRyE3g0N3FZozAMVJCQPHr4jQCzh0xSi+63EMXD8qLUqmmiYN5FMuX7rzAN5+zS3hZ1QQuinwi8HPREcyUMzEpDHLOHQ12WKga0/Xc9EpwAaQsie54aTnkIqSCQiq8XWHXmJcqgmvcq++cI9ph35iZYDbnwpOeXXlawABQld5hg3LwDlbJjIDoypJSmQMwKe/uxef/cGTiZ+vDoJnOdlSLyRLJS28FLWrW8tFpFzE5/jE8ZWo8YUS5SLxvGVULlsZY7Phn9sAXgPgodTHvgrgF0O1y0sALHLOD6MGo0DTw0eXozrHYiMEV1GXHdVyySrOpatykZAtqmbCuWGwse/4yQWvKNOk8YBRSugLPzwgfEY9sYhzHjkT0ZGo1hsHShC6Zuapz0e5S0LtykHRDISenkOqCH0m7Du52BULvAWdj1SQm3gdacrlM9/bi8/+4CkAamPSZ7PKLqebqcvaBdsn8ejRUcpFJy60GgZXP/pvjyZ+3otOYOrruGmb6LveCOWSaEGn8AwTCF0Y41c+exs+dP0DANQoF6kOahLjnAHgfzPGTAQbwBc459cyxq4CAM75JwFcD+DHATwGoAvgl6SvssRMg4Ex4It3xHU5himVS11JCarBmfgaiXKpB6FbYYMLzw8m53LfRXMyWDxaOnQz+/gs8twqnHfDDJyl6/NosotZlDqywKZt5Fbjo+w4GQ6RLKofktKHxzXvFU8QlomFbvK0lEavqnOH2pQt9Bxsmw6yjXXoqiLKRezHq1p+AgC8jFMnNVNXybYFgsDo1+47MqIyGmpw6GQ7wudG5kQqIfWTNtF0afGF6CtUMjtF2SKNwTnHgfle5MhV5oxU0/qyD3DO7wFwacbPPyn8mQP4dekrU7Q0mHYTDr2+Wi6qY5GVVaYD1F4cBXwo4Ljcd7Flsgnf58p1YcTrSzsgEfGqBkUBRI1HgKR+N0iDV+XQCxKBhi4YU0NdtIH2HR9TwpqPOHTFDWe6ZeGRo8UOXRWhz4UIfV7IGHU9Lt09iizN5YsmJn0pIfSCUydJD2UoANHO2zYJnwNPnezi2Tumop/rUC5kO2aSDp02NJ3x2mEgne454tKFda2CqLMSi5b6bmLTPSN1/UUmA2jWfceiLBtVuejoazM4dE2EXiQLJFPRKFuhyoVe/BJRTRoBs+D6sk8Q4nNTDYoCwfOi55igXGpG6KtDDx2FHqDBeDFCF40QuirfPdtpJKgRIMuhq44ZI3SyoFaP2vsVLyP9jsViTkoleQtKZHQ1itgBwFTIa3dT2Y46lEt6zGgsjdIJZO1G0EAiHavTpVzE9UVjHF9OFozbOZtsQVhktQdF14sN0whdgxvM1mSrKxUAgaOuiUM3DQOOJzj0MDBKG5nq8ZSezzDVhFmc9GoIPewilevQ1RF60zJzu653hy46mqncaaXLQDMoOtdpYHngjgTkRUs7lzKLHLogXQwS5VQdenwdaTmu6NB1OPSs90EIva2q9Y6aZiSvUYdGJEsne1VB6BMNC6tDN3qvsWxRROjy44nri56j2Fy9ZRtKfU9rCYquR0smFmnq0DOcrw5/GYxZr8qFJFmDCKEHiMZx1YteAcDWsPnu0aUkOhAzDFUTi4AA/VJATtxkVbMmgeKuNqqa52C87Ap/cVBUkR6ZIOcbo+l09Urq9C5rsyHlIo4ZaLLV3q94GekCbD1NykWmb4DqppjXalC1fG7iWlLvN+bQ1cfrNINqhm6EzJP/B9QQeiMToccOfedsW+nUaZlG6XPakA5d5LTqLI4/dPWCM2YGjZNe7DqyRXKS1G5Ol3I5c1MHjAF7T6wmfu4IzlO1OBeQT7kMNNRCcfnSeIE+dTJotbc68JQCosF48TWKptMIHBCdr8B3px16S82hTzRM2CbDvKhyUaxNBCQ59DRQSVIu8u/YMBgMlh0XojFVN1mat+Jc4ZyH605tE7v9v7wGF+2aHnHoVRB6JyxPm0bo4jNQccBZHLro0Hcp0C3iNRbZhnDoX/iPVyT+7lQIihapXHT4SyDmG9PcvmhqVdpYxFMC1SmXlm1i50w7SmUmE79DpZAW3Us+5SLfji2+xlGE/qoPfws/88kfoOe4Sv0hg/Gyu+ToBkXpaPyRf30kqhqYdp7TbbVNhzGGmXYDi70k5aLK/9J8aFrGyLzTRehAgAgzOfSQA1eVftJ9DT0fqwM34Ks1cgwAYPNkE1smmyMnMJ1CbmSdhoXuwC3k0JWCotYoQhcpl50z6g69rKTBhnDoLz5nEy4+czb6+zDFY6ogjzyE7vs8rFuuI1sc3SSqSNrSLeOIctGpM052zpYJ7D3ZTfysJwSnlGSLIuWSkfo/cDUSi1IInVDnQ0eWsTpQL5oWyxZTqg9NHTopUr52/xH82t/fASDDoSsidCBAXKLT1aH9fuUV5+KqV52HX3nFOeA8eV0JDl3BGQFU9TM/21YVoTciDt3H8/7g63jlh78ZUySaqpR0sa9BJQ7dxOowrhnlej7+5IaHEqhaqY2f4JdozMOLfWybaqJtmzh3q3zXMbJnBEIHkCjVOcqhayD01FFyWGFiZfGNabWBWmJRcuHFCF0ffezZ0sGTKcpF7H6iWpoWCBYPBeQGaZWLamJRqk2ZeNrpDjWKLUVB0RyVi0aaPhkxHGlQoOrggNGyvN2hqzxOu2HivW94TkRLiXNPvH8V/heIS1CkravLoQsOHQjoBzrZ6czpdpgIJJruKRYIgry9YZxYNPQ4PvntxxOfUVIKiZmiHsf3Hj2B6+45hJedvwU3vOcVeMdL9yhfY9nc2DCyRdEhUqYo5zxENNURepVgShbfKKIky2BaCgMyqucy1EgLJ9uzeQKLPQcL3WHEB4voRjULExjl0P/2B0/idRft0KprEpdJ9kaurTv0pCRbouUh9IETVEVUdW6E0AFge6gd9lMJEir8KpmdKvq1MvAwrZG2DmTnG2SVq5UeL6z6mTZtDp0aT6fq/oi/U7FWw4y6jEXjVUToQ8+PNkHxdPKmi3di00QDP3nxTunxkpmiPq679xCmWjY+9JaLlGNCZGX/bsMg9Kz+fDqVAhkLnO+3HzmOd372Ntz8RNBgoMpEAEb5RhFhqjq3EYQe6pTjVmzqjmP3XMDXHVzoRT/ravKrEeXixA5974lV/Nev3I9f/eztcH2uXvwq1QJMpAp0VC5FCF2n5r34/XSULkokk7UAocfjBA2ONR16qkgcpejrGuVDpI02CdVMUZq3p1ZFVQ8havU53bJGK2pW4tDD5x6vt/jeT64O8Idvep5iPfRktcX9p3rYs2VC25kDwN/80osKf79hHHoCoadqLKhnThq4a/8C/vWBo/jWw0HVxyqUSzBmkm/ULYoPxKoZsjoolx1hAObIYixdFNtZKSXtUFDU82PZYuiIKfCqy6HTAk06dHXKJUu2+MjRZTxydFk5IAoknw+l06cRuo41zbhZCKDX4JhMzFheGbh49+fvRIY/Vhovqx66rmyxGabjH16MQUWlzM6GgZ6TLMBWFaED8XoTTxJZVSLLLC1b3D/fxZlz6oFQ0cqA0gZy6PGNkNxOp9UZkESjdMTXqXInmmUwnFgZYDVshCseVVUVH+L92CaLdegazSjIdoY0wSHBofeGQQ9VVfmUWNI4rXIhekj1ntMyQzGbsO/46g0pMmSLr/3z7+DmJ05pFW4CgP/yxufCMli04IsSyWQtXfRrpQJCFztnfeWug7j2nmr18SzTyEToXceDbTL1zllWcH2HFuI5WAWktO0gGU3cdKpmigKCCEEohPfnP3eJ8njiPXEO7D/VxVmbOsrjqNjGcegZQVFdB5fVg5GKS1WhXL581yG8/i++A6AiQheub8tkcwSh6/D8myebsAyGIyE6Gro+XJ/j3Veej++/90qlsSiBZrHnRPdJaJ/+ruo004g6fZTWR+ijHLBOzXsgUJO89bJdEULPa8+mYmJQlHNejXIROHTdMdLjZdVD7w2r9Q04tDCK0HUcOl2DeJqLELrGeDTH6P0Ow2YX7/mxC/Cy87coj5f2Sz4PckKeTts4Dt2KHw4tAOK6VBe7SGmQ46CAo27VN1pM+08Fk1VEb8ocupl06IR6q6AZ02DYPt3C4RAdxenb6gtzumWhYRk4vjKIHHg3FXzT1qET5ZIKdqkGRc0wEC1WgYy/S8+hA8BM244QXEFisLTZJoucUN8J2uNpUy5R71iuvCayx8vm0AeuXmcvWiMJyqWKbLExmmtAIgmdADXNMQJQVKJXtYwFWZYvOXNu7NABpPioEJnfd2gRAHDhzmmlscQSmDFCp4mlPhGA0UBmXQh961QTKwMXrudXolyAoLLb4ZBy6TqBU9IJ0DDGsHWyiePLg4hHJqqJTBWht1OIOh3M05IECgoSkWfVOY6TTbds9BwvPOHUgdDjZiErA+q2o7fhiPLZvE5DKkY1hdI20KjVAwTzpmEaUWZsyzYi+lQHSFEsRJwrjma2NxCvBQInNKd1x8sCXrsqcuhltnEcujVKudx7cBEN08Cztk/l/bNME1PeBxFCp4mlt5jSjQXEBaWK0EX0s2UykMutCIWhtGtHz7QidBTXGNe7361TgUN38xC6IoIjtEUIPV2RT2fjsc04qCdeH3W10bGZUI++1HdA/vx5O6fx+V99idZ4DTN2muRAdBG6mOCW1/1JxWwzu82bTq0ecUwyz+cC562P0HtphK5Jm6azkel9qM5lsqzr2CTIX58O2zAOPas/370HFvHcM6aUEbDYvX2QDrBqore0k6iC0LcLRfu3TAaFtZZ6biXKBQiyRffP97DUdypRLkDs0NMcOpnqJpZGW1U5dCBAv4NUsBYAlgdO3j8pNcoGXeo5EUL/wJsvwhXnbdYaL2htl0TolVUuQunlKpbXiLmv2dkLSDo5x+NRK7bNk+qOjoDPkyfiDGidhtNk6Yqe9D6a2gh91JeoVuRUtQ3j0EWnOPSCI+U9Bxbx/N0zymOJ6LmfRug1VX1LcuhqTjPTofedypTLS8/bAs/nuPnxkxFPOKXpPLZONXFiZRAVIeumNjRVntowGJqWET3HdEJMupGBjDXMOKgn9mZNN95WsRkhIEx0k0pSVtqC5tPBvcaUS3UOfVAD5WIbRqbWvhpCT64vyl7WKVRFc+yqv7sd+8IHGhvaAAAZxklEQVSyFoMKtdXT5WkJpOly6FnqO9WENlXbkA7d9Xz8cO8prAxcvPICuWbToonyYUJHVWpAZJmI0NNtssps+3Qz+vOWKdGhV6NcLjt7Fm3bxPceOxE1a5jVPAJunWzi5Oowem5phK6j9W43zKjDTje1QZ6nUfdClAQu1+TQqQDXUj8u4qRa9GrkGsP3WpVyeXoQenYtF52+AcDo3N17YhWbJxpagWrxdLkQFjhzPK59benAe0S5aI5nKGaI12EbJ/U/Rbn824PH0LAMLTkRWcs2IoRe1VmmjRb7Nb94ufI1TglFnohDX+rFDkQXgTQtEy/YPYP7Di7iOTuCQDLV+Va1LVNNcA64PI9D1yy2RJSLMN5U09JOq48ReuzE8zrZy9hMOy6jS3OlyqIVVS5Vg6JJDl0/TkBmmQx9Nxuhq9Z+J0sDpidOrCp17RFNjDXRex66nvb6aJgGDBbXl6f3UQXk2YJSSDdepWIbEqE7Hsdd++dx6Zmz2mgGCOpzRCqXpwmhz3RsbZ4aCJAwkEToupQLEBznB66P+bCu91wFhC5a2klqIS7bRE9QuTRMA2++ZCf+8V1XlPzLbEsi9GBxXnjGND7zjsu1xgOATRPB8zq1OoyyZKtQLg3ThB9WSKQjflWEvjLwIsrqJsUcg/R42ZSLPkJP88p7T6xg56w6nQYAz9o+iV9++TnhNcW5KbprmDGW2CRihK6/fm3TiDbaKin/srYhEfrQ87Hcd3H+tslKY852Gji1GgRlqpTxFI2AJB1Vqx654qCoU0kRQNa0gyDcQneIlm1oa7LLCkiptNaKry0uh9odephqWfiLt430J5c2kc4gyuWv/98XafHxZDNtGwYLGjuTc6/Ci9J8G7p+5ECqcujv+Otbg78bDDtn27j7fa/Vkljm1UPvO772vKG5S5Uc+46PXbN62mzGGN74gjPwme/tjTbuoD+p/vtohyV0AUGHXsEnNEwDvsWxOvQwoXnyUrENitD9SjUvyOY69ogOvYqzBIIXeGp1iHsOBBr5KugNCBzIRMPE/lPdypQLEKCNgethoetgtq0voSqShk02LS000hYosJ5GCd60iZUMCaFXVRmYBsNsp4GTq8PoVFIJoQsOfVEzUU68tqyxZzo2NqdOVDKWVw+9GkIP/p1YjlgXoQMx0BsK67jK+siac1VA3mTLiuJUPwqEvmEcerqdU5WaF2RznUbUZLYuymXg+rjsgzfiv9/4CAC1+smi/fr/dR46DROGwfCcM6Zx65Pz+M4jx9GwjEqov2kZGDg+5rtOYlGpWtGi0ZGgAWHXdUG2WJVzbKYQummwWnjMTRMNzHeHEXqtGhQFgMNLPXzu1n249KxZbcSf3liqzmXLNPDosRX875ueTPy87+g1UwdiBywG46nnrY6la/YMXf3+pEC2Q69CufzVf3ghfus1FwDQj42o2IZx6AnZouvXcoSZ7djou15UVx2oLyhKpst3/+fXPQcPfOD1/397Zx4kV3kc8F/v7Oyhc7W6ETpAFuJykECIcCNEKAMucSQYY8ohkFQKMAm4grnLwcHB2Dg4KZOYgsKB2MQ2xMKIsglgytg4BmFBcQlxiUsCIYQk0K29On+893bfjmZmZ77vm9md2f5Vbe3um7e9PW9e9+uvv/76A2D/KaNZtW4LK97dzDeWHOSlT3NjX8rFNX8Oxa9Tkoool9ZsJrWwyK3NbZr0pOjWXVEA4DK5mkv7yCY2buuL0L0cenx//PSZNWza3sG3zvqss6xcPXxSBdD3gPjHZSv7rbTd3eU+ekp0bEtNqo51nGCF/iMc8KtDh/wdJH2u4/5TxvSmTS1CT9G/Dr0nbkDkfiNAZJiq0USKT1Mf2HOVWUKIsqUDpkYVKdPaWjl34QwvWc3ZKOWyeUeHc4ULFG+RMN7RobekqlxcG0ClSS/937Kzs+w9PwvRPiKK0IM49Pi+fvKNDewzYWRv9ZELuaMm7wg99b42bo8m0VXVaQOThKR2Px2hu5bOQsqhd4eJ0PM5dN/rmJT0Wg49RbJaK31xfYcwSWSwqyvqzdHYIM7D3T9cs5ivLJq9x3HfHDpEETrAOYdP95bV0huhd3oZUrGUi1eEHk9E7epyn3hLyDb2Lav/4NNdZa8HKET7qKaoyiVx6B5Rf9JqYvWG7RyxT7uXXnvk0D1Hm+nR5ZpN0cKdju4eVN2XwyfXLJ3ua/OJ0DN7RuheOfQ8gZnvSCdZdBeiYdpA1IxDTybh0pGw76RoclPu7uzxvhHGtmbz1tOGiNAXzGrngUuO4tJFn/GW1ZzNoBpFXD6GVOxaLT5gspPMdA7dp8lSQjpCf3/zTqfViPloH9HEx9s6uHf5u4D7PAn0D1AWejr0PXPofg/E9PtauznqAZTkql2dXGfi0FP3ns9cTm7KxT9C3/NvfSP0ZDX7qX8y1UtOKdRc2eKY1mxvtzZfh94Sf1BPvbWRO59823uSNZ8D8jH2NPNnjAsiJ22IIQwJkgVakUE9eeUi557P6YVFnWXuFZtfR6Gju4eu7h4+3LIrWKe75Lq9Hu9i4/MRp6/jzPF+rVVD59DT8tZsjiL0ZI7DNUJPqmbGpUZx6YV05ZLePSv6rl6BWf4cut+D8YCpY1h906lVWTVaMxF6cuNPTdUQ+zrg5Kb8119HFSnbdrsvCU/LS1Ptpb8D0dwvZRUmQk8bwWSPtEZLNhP1BI/bv4YoIe3o6mF93ETMtd45l9xJPK8IPfUeJziUFqYJnUNPZ5KSPv9JVZjrwyIpvU1fQ9+VtpCucnEvqYQ9q1xE/NotJ1TLD9SMQ09u1r1TDeJdHfq86W1A302Zu+rRlXw3UoibISTpaMNnkqapgEP3XdQBkXF2ekZaEDm0T3d2suiWJ4BwvahPnzeNsw6d1vu736Ro39+61IqnCR2hb0u1S3hl3Ragb8tG1/mNZIN3nwqrNCLSb0VwdN/4T1L3/p5pCFIZVS1qxqGPG5ElmxEOSm1m4ZpyWXrxUay+6dTem3JTPIN/9wA7ag9EvohoyEXoqRyhzyRNv5RL7IgzDW47xSQkD4adnd10dPc4bzaSkDwQkuF4qBx6U2MDf5aaJ/ArW0w9YD3LNHNz6L57WCcj1nnT23hhzSc8/dbG3tSab4TuspK4EM3xSKy7R9nV1U2rx32dO9ryfShWm5rRdvyoZn77tUX9yvZcI/SkC1ryYW3c3sHsiSM5Ye4kLx3zffihcuihSOtY7rZuadJOLHHEvhU9aYfeFSLlkvN5hHLo0D+Y8HnbaR19I8FMTmTqu6NS4tAvOn5fRjZl+OWL67wnRdM9jkIR9ZTvZuuuTlT96tpzo3vXuYLBYmh5mwHYq62131DPt64zHaGH2FQ33+TJkIvQ0ymXQGVUyepLXwecRPo7O7qDpFzSf/8/Fx3pvVApzahUCwEfRxyqGRzs2X/bp6sk9LVLmDi6mUljWti8o6N3hy/nlEs8WhrjMRGaS5JySVon+Dj03IWAIdoQV5Oacui5+DqkfhOEAXYSScs7KR6Sh6hDD0lax1ALHRLj9ukCCX0bDHzwyU7vnhzQ/70eNjNMlVCC68YguYR06LnBQ77GWuWQNDQb3ZJlbGuWT3d2BovQfdcYpEmasIVw6Mk9d8LcaJ+FRGatUDNli/nw3f0j3dcjTITed5Pf9qX5bNi6u+I7lJRLOoceailyX8rFzzkdsW87e41t4YaHVkb1xJ4PiPQDIfTEVogAAMK2mii2UbkLJx80hR88sZrJY1poG5Fl0/aOvrJFx1K+BbPG8cjK9YxoynD9aQdw+Cy/2nvoq2YK6dAPmDqGY+dMZEu9OXQRmQ78FzAF6AHuUNV/yznnBOBB4O340FJV/aewqvYxqrnRu8QQ+lcV+LYRgNza7IxzPXYl6V/lEsih96Zc/Jzm6JYs5x81i289/GosL0wOPXR/HggTAEDfNQsRqecGD/l6mZfDFSfP5aLjZjO2NUtba5a3NmzvjdBd9xT93jnzeHfjDkY2N/I3x+7rpV9C0JRLsutTd09vr/VaopS7sgv4B1V9TkRGA8+KyGOq+krOeU+q6ufDq7gnv7nihN7KFB9GNmV6F8WE2LzVdwFCNeg3KRoop5zI8U25QP+Nel13b09IHHklKhVCzT8kD8PLFs8JIi+Nb4SeaZDeycu2EU18sqOjt2zRdbJwRFNjb2+iUFQi5dLp+TAcLAa8K1V1HbAu/nmriKwCpgG5Dr1qTBzd7NVyM0FEmDCqmbWbdwbJJ9dCiVP6oeMbASck+dB8m+KWS/oa+s4/JA8E101+ixEqldbcmOGdm08LIiuXbt+6xRRjWrNs2dXF+i3RhjA+TjM0TZmoP1EYhx59rp15+sDXAmXd6SIyC5gPLM/z8pEi8oKIPCwifj1eq0gydA6dchmqVMK5tQaaFIX+Dt33evZF6EN/5FQJ8m1O4UrSe2X52xuZ0T4iWMopBM3ZTG/KpSnT4JwOgr5dn3zTVYNFye9cREYBPwcuV9UtOS8/B8xU1UOA7wO/KCDjb0VkhYis2LBhg6vOQUmGvCGrXEJXVISkEqOI3pRLkAg93AgiWZhUCyOnShByDifpX/OH1Rs5MHDKxJckQo9aJGe9JsAXzZ1Ec2MDXz5yZkANq0dJXkxEskTO/F5VXZr7etrBq+qvROQ/RGSCqn6cc94dwB0ACxYsGBKPwJbYgYQoQ2vMNPDAJUcx23Ov00pSiWi1N+USIkJPRVf+vVwivWph5BSa7587n+PmTAwmL3HoqnDgXkPLoTc3NrB28w66unsY69nzfsrYFl775imBNKs+pVS5CHAXsEpVby1wzhRgvaqqiCwkivw3BtW0QiQReqi62FBdEStFJZxba++kaNgcuu8DIvn74Rihn/bZqUFLZsem9p8dchF6YwNbd3Wxdde2fs37hiOlPM6OBr4MvCQiz8fHrgVmAKjq7cBfABeLSBewE/iiasAZmQqS5Ns6anQSpFwqsXK1L+USIofe92D1LTdMJgVrbfl2CEKvf0i3Wj5mzoSgsn1J3ycNNdRIqxKUUuXye6DoVVLV24DbQilVTY7YZzy/eulD9hrmT3YXRKIheGtvyiV0hO4nz3dV40D87+XHsnl7bS08cSWpKjt34YygqzxDkDy8Fu7TzjfPOHiQtRlchs5U9SDxl0fO5MjZ49lv8ujBVqXmyIjQpRps6T/0X7DiW4ee9AuZW6HP1mf/z1pjTEuWZ65dHKRcODQfbdkFwBcWTB/2djzsHbqIDLub4Kz50zjcc7szSIa32ttCIHyVi98D4rCZ4/jPCw7n6NlDK0VQq0wKtCdraJLdlGZPHDnImgw+w96hD0duPWdeEDkNDUB3OuUSuA49QApnkWdLZGPo0xZP2O47cehWl1ULc+iGM8lu961NkeMNU+XSF6GHkGfUP7edN58X13w6pFavDhbm0A1nksmobKaBTIOQDVHlkg1XtmgMDyaNbuGkA4dmOqjaWAhkOJOUQDaI0JRpCDIpmk6zVKJLomHUMxahG84kKZfGjNCSbQhStpiunw7VPGw4MX9GW802ljL8MYduONOQitBvWHJQ8Goh37LF4cgDlxw92CoYg4g5dMOZJELv7lFOnzctuHzLoRtGeVgIZDiT5NB9N1IohOXQDaM8LEI3nLn7gsO5d/l7TGtrrYh8y6EbRnmYQzecmTN5NDcsqdxeJubQDaM8zGKMIYvl0A2jPMyhG0MWn51nDGM4Yg7dMAyjTjCHbhiGUSeYQzcMw6gTzKEbhmHUCVa2aAw5nrxyER9t3TXYahhGzWEO3RhyTG8fwfT2EYOthmHUHJZyMQzDqBPMoRuGYdQJ5tANwzDqBHPohmEYdYI5dMMwjDrBHLphGEadYA7dMAyjTjCHbhiGUSeIamW2DxvwH4tsBV4LKHIs8OkQllcJmSavvuVVQqbJq315c1U1/47sqjooX8CKwPLuGMryakFHkze05NWCjiav+vKK+c56Srk8NMTlVUKmyatveZWQafLqWN5gplxWqOqCQfnnhmEYNUox3zmYEfodg/i/DcMwapWCvnPQHLqqejl0EfmciLwmIm+KyNXxsVtE5FUReVFEHhCRNk95N8aynheRR0VkLx95qdeuEBEVkQme+t0gIu/H+j0vIqf66icifxcfXyki3/HU72cp3d4Rkec95c0TkadjeStEZGGp8orIPEREnhKRl0TkIREZU6KsH4rIRyLycupYu4g8JiJvxN/HlaFbPnlnx59Dj4iUNZotIM/ZPorI9LGRPeSlXnOxkXz6+dhIXv08bCSffmXbSFHfGTLBX60vIAOsBvYFmoAXgAOBk4HG+JxvA9/2lDcmdc7fA7f7yItfmw48ArwLTPDU7wbgioDXbxHwa6A5Pm+S7/tNnfMvwNc99XsUOCU+51TgiQDv+Y/A8fE5FwI3lijvOOBQ4OXUse8AV8c/X13q/VdE3gHAXOAJYEGZn3E+eU72MYBMJxspJM/VRoro52QjReQ52Uix95t6vWQbKfRVlQi9QGTkHH0AC4E3VfUtVe0AfgqcrqqPqmpXfM7TwN6e8rakzhkJlDrhkFde/Nr3gCvLkDWQPBcKybsYuFlVdwOo6kch9BMRAb4A/MRTngJJBD0W+KBEecVkzgV+F5/zGPDnpQhT1d8Bm3IOnw7cE/98D3BGqcrlk6eqq1TVqbS3gDxX+ygm09VGCl1DcLORYvKcKCDP1UaK6udgI3mpuEMXkQzw78ApRBHRuSJyIPAycBZ9xlQO04A1qd/XxsfSXAg87CtPRP5ZRNYA5wFf95EnIkuA91X1hRLlDKgfcGk85P1hGUP8QvL2A44VkeUi8lsROTyAfgDHAutV9Q1PeZcDt8Sfx3eBa0qUV0zmy8CS+NjZRNGhK5NVdR1A/H2Sh6xKU459FMXRRgrJcrWRYrjYSCFcbWQgyrWRvFQjQi8U/TpHH4DkOdb7NBeR64Au4F5feap6napOj2Vd6iGvGbgOtxu+kH4/AGYD84B1REM2H3mNwDjgT4GvAffFkYOrvIRzKS/yKCTvYuCr8efxVeCuADIvBL4iIs8Co4GOMmTWJA72URRHG8mn1wjcbaQQrjZSCFcbGYhybSQv1XDopUTT5bKW/pHU3sTDbxE5H/g8cJ7GiSkfeSn+mxKH4wXkvQfsA7wgIu/Ex54TkSmu+qnqelXtVtUe4E6ih6erfh/Ex5dqxDNAD1DKpFSxz6ORaCT2sxJ1KybvfGBpfOx+Sn+/BWWq6quqerKqHkZkUKvLkJnLehGZChB/L3k4Xi0c7aNUyrGRfMzG3Uby4mEjhXC1kYI42khequHQB4reXPgjMEdE9hGRJuCLwDIR+RxwFbBEVXcEkDcndc4S4FUPeUtVdZKqzlLVWUQ3xqGq+qGHflNT55xJlD5w1W8Z8AvgRAAR2Y9o8vBjD3kAJwGvquraEnUrJu8D4Pj4nBOBcoanha7hJAARaQCuB24vQ2Yuy4geOsTfH/SQFRwP+ygm09VG9kBVX/KwkUL6udpIIVxtpBguNpIfnxnVUr6AI4FHUr9fA1yT+v0JypzBj//uVOB1oojquvjYm0Sjgefjr3Jm3PPJ+znRDfAi0YquaT7ycl5/h/Jm8PPp9yPgpVi/ZcBUT3lNwI/j9/wccKLv+wXuBi4K9PkeAzxLVKGyHDgsgMzL4mOvAzcTL7YrQdZPiIbwnUSO56+B8cDjRA+ax4H2MnTLJ+/M+OfdwPq0HTnKc7aPIjJ9bGQPeZ42kk8/HxvJJ8/HRvK+X1cbyfdV8ZWi8XDidWAx8D5RpPQlVV0Zv/4EUVnRiooqYhiGUedUPOWiUZnUpUR1pauA+1R1pYicKSJriSL4X4rII5XWxTAMo54ZtF4uhmEYRljqqduiYRjGsMYcumEYRp1QjZWi2yr9PwzDMAyL0A3DMOqGajXnGiUij4vIcxK1KT09Pj5LRFaJyJ1xo65HRaS1GjoZhmHUG9WoQ98GtAEjVHWLRP2NnwbmADOJFjssUNXnReQ+YJmq/riiShmGYdQhjVX6PwLcJCLHEfU+mAZMjl97W1WTpu7PArOqpJNhGEZdUS2Hfh4wkWipdmfceKclfm136rxuwFIuhmEYDlRrUnQs8FHszBcRpVoMwzCMgFQ0Qo/7uOwm6pP8kIisIGoK5NyRzTAMw8hPRSdFReQQ4E5V9e1BbBiGYQxAxVIuInIRUbvI6yv1PwzDMIw+rDmXYRhGnWArRQ3DMOqEYA5dRKaLyG/ilZ8rReSy+Hi7iDwmIm/E38fFx8fH528TkdtyZDWJyB0i8rqIvCoiPvsUGoZhDAuCpVzivfumqupzIjKaaJHQGcBfAZtU9WYRuRoYp6pXichIYD5wMHCwql6akvUNIKOq18d7Pbarqu++fYZhGHVNsLJFVV1HtF8eqrpVRFYRrQg9HTghPu0eoj1Er1LV7cDvReQzecRdCOwfy+rBfxNWwzCMuqciOXQRmUUUfS8HJsfOPnH6kwb427b4xxvjZl73i8jkYn9jGIZhVMChi8goop3AL1fVLQ4iGoG9gf9T1UOBp4DvBlTRMAyjLgnq0EUkS+TM71XVpfHh9XF+PcmzfzSAmI3ADuCB+Pf7gUND6mkYhlGPhKxyEeAuYJWq3pp6aRlwfvzz+cCDxeRoNEv7EH1598XAK6H0NAzDqFdCVrkcAzwJvETUIhfgWqI8+n3ADOA94GxV3RT/zTvAGKAJ+AQ4WVVfEZGZwI+I+qhvAC5Q1feCKGoYhlGn2EpRwzCMOsFWihqGYdQJ5tANwzDqBHPohmEYdYI5dMMwjDrBHLphGEadYA7dMAyjTjCHbhiGUSeYQzcMw6gT/h8zIUDry7wfMQAAAABJRU5ErkJggg==\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "time_series[0].plot()\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "##### Often one is interested in tuning or evaluating the model by looking at error metrics on a validation set. \n",
    "For other machine learning tasks such as classification, one typically does this by randomly separating examples into train/test sets. For forecasting it is important to do this train/test split in time rather than by series.\n",
    "\n",
    "In this example, we will leave out the last section of each of the time series we just generated and use only the first part as training data. Here we will predict 48 data points, therefore we take out the trailing 48 points from each time series to define the training set. The test set contains the full range of each time series."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {},
   "outputs": [],
   "source": [
    "time_series_training = []\n",
    "for ts in time_series:\n",
    "    time_series_training.append(ts[:-prediction_length])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "time_series[0].plot(label='test')\n",
    "time_series_training[0].plot(label='train', ls=':')\n",
    "plt.legend()\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Data path : sagemaker-us-west-2-570447867175/deepar-synthetic/data\n",
      "Output path : sagemaker-us-west-2-570447867175/deepar-synthetic/output\n"
     ]
    }
   ],
   "source": [
    "# Set path\n",
    "s3_data_path = \"{}/{}/data\".format(bucket, prefix)\n",
    "s3_output_path = \"{}/{}/output\".format(bucket, prefix)\n",
    "\n",
    "print('Data path :', s3_data_path)\n",
    "print('Output path :', s3_output_path)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### 3. Data Pre-processing"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "The following utility functions convert pandas.Series objects into the appropriate JSON strings that DeepAR can consume. We will use these to write the data to S3."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "metadata": {},
   "outputs": [],
   "source": [
    "def series_to_obj(ts, cat=None):\n",
    "    obj = {\"start\": str(ts.index[0]), \"target\": list(ts)}\n",
    "    if cat is not None:\n",
    "        obj[\"cat\"] = cat\n",
    "    return obj\n",
    "\n",
    "def series_to_jsonline(ts, cat=None):\n",
    "    return json.dumps(series_to_obj(ts, cat))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "metadata": {},
   "outputs": [],
   "source": [
    "encoding = \"utf-8\"\n",
    "s3filesystem = s3fs.S3FileSystem()\n",
    "\n",
    "with s3filesystem.open(s3_data_path + \"/train/train.json\", 'wb') as fp:\n",
    "    for ts in time_series_training:\n",
    "        fp.write(series_to_jsonline(ts).encode(encoding))\n",
    "        fp.write('\\n'.encode(encoding))\n",
    "\n",
    "with s3filesystem.open(s3_data_path + \"/test/test.json\", 'wb') as fp:\n",
    "    for ts in time_series:\n",
    "        fp.write(series_to_jsonline(ts).encode(encoding))\n",
    "        fp.write('\\n'.encode(encoding))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### 4. Start Training"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "WARNING:root:Parameter image_name will be renamed to image_uri in SageMaker Python SDK v2.\n"
     ]
    }
   ],
   "source": [
    "# 1. Create 'Estimator' to start training job\n",
    "estimator = sagemaker.estimator.Estimator(\n",
    "    sagemaker_session=sagemaker_session,\n",
    "    image_name=image_name,\n",
    "    role=role,\n",
    "    train_instance_count=1,\n",
    "    train_instance_type='ml.p2.xlarge',\n",
    "    base_job_name='deepar-synthetic',\n",
    "    output_path=\"s3://\" + s3_output_path\n",
    ")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 20,
   "metadata": {},
   "outputs": [],
   "source": [
    "# 2. Set Hyperparameter \n",
    "### We are NOT going to do hyperparam optimization for time-saving purpose ###\n",
    "hyperparameters = {\n",
    "    \"time_freq\": freq,\n",
    "    \"context_length\": str(context_length),\n",
    "    \"prediction_length\": str(prediction_length),\n",
    "    \"num_cells\": \"40\",\n",
    "    \"num_layers\": \"3\",\n",
    "    \"likelihood\": \"gaussian\",\n",
    "    \"epochs\": \"20\",\n",
    "    \"mini_batch_size\": \"32\",\n",
    "    \"learning_rate\": \"0.001\",\n",
    "    \"dropout_rate\": \"0.05\",\n",
    "    \"early_stopping_patience\": \"10\"\n",
    "}"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 21,
   "metadata": {},
   "outputs": [],
   "source": [
    "estimator.set_hyperparameters(**hyperparameters)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "We are ready to launch the training job. SageMaker will start an EC2 instance, download the data from S3, start training the model and save the trained model.\n",
    "\n",
    "If you provide the test data channel, as we do in this example, DeepAR will also calculate accuracy metrics for the trained model on this test data set. This is done by predicting the last prediction_length points of each time series in the test set and comparing this to the actual value of the time series. The computed error metrics will be included as part of the log output."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 22,
   "metadata": {},
   "outputs": [],
   "source": [
    "data_channels = {\n",
    "    \"train\": \"s3://{}/train/\".format(s3_data_path),\n",
    "    \"test\": \"s3://{}/test/\".format(s3_data_path)\n",
    "}"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 23,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "WARNING:sagemaker:'s3_input' class will be renamed to 'TrainingInput' in SageMaker Python SDK v2.\n",
      "WARNING:sagemaker:'s3_input' class will be renamed to 'TrainingInput' in SageMaker Python SDK v2.\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "2020-07-22 03:27:11 Starting - Starting the training job...\n",
      "2020-07-22 03:27:14 Starting - Launching requested ML instances......\n",
      "2020-07-22 03:28:25 Starting - Preparing the instances for training......\n",
      "2020-07-22 03:29:40 Downloading - Downloading input data...\n",
      "2020-07-22 03:30:05 Training - Downloading the training image...\n",
      "2020-07-22 03:30:36 Training - Training image download completed. Training in progress.\u001b[34mArguments: train\u001b[0m\n",
      "\u001b[34m[07/22/2020 03:30:38 INFO 140662884079424] Reading default configuration from /opt/amazon/lib/python2.7/site-packages/algorithm/resources/default-input.json: {u'num_dynamic_feat': u'auto', u'dropout_rate': u'0.10', u'mini_batch_size': u'128', u'test_quantiles': u'[0.1, 0.2, 0.3, 0.4, 0.5, 0.6, 0.7, 0.8, 0.9]', u'_tuning_objective_metric': u'', u'_num_gpus': u'auto', u'num_eval_samples': u'100', u'learning_rate': u'0.001', u'num_cells': u'40', u'num_layers': u'2', u'embedding_dimension': u'10', u'_kvstore': u'auto', u'_num_kv_servers': u'auto', u'cardinality': u'auto', u'likelihood': u'student-t', u'early_stopping_patience': u''}\u001b[0m\n",
      "\u001b[34m[07/22/2020 03:30:38 INFO 140662884079424] Reading provided configuration from /opt/ml/input/config/hyperparameters.json: {u'dropout_rate': u'0.05', u'learning_rate': u'0.001', u'num_cells': u'40', u'prediction_length': u'48', u'epochs': u'20', u'time_freq': u'H', u'context_length': u'72', u'num_layers': u'3', u'mini_batch_size': u'32', u'likelihood': u'gaussian', u'early_stopping_patience': u'10'}\u001b[0m\n",
      "\u001b[34m[07/22/2020 03:30:38 INFO 140662884079424] Final configuration: {u'dropout_rate': u'0.05', u'test_quantiles': u'[0.1, 0.2, 0.3, 0.4, 0.5, 0.6, 0.7, 0.8, 0.9]', u'_tuning_objective_metric': u'', u'num_eval_samples': u'100', u'learning_rate': u'0.001', u'num_layers': u'3', u'epochs': u'20', u'embedding_dimension': u'10', u'num_cells': u'40', u'_num_kv_servers': u'auto', u'mini_batch_size': u'32', u'likelihood': u'gaussian', u'num_dynamic_feat': u'auto', u'cardinality': u'auto', u'_num_gpus': u'auto', u'prediction_length': u'48', u'time_freq': u'H', u'context_length': u'72', u'_kvstore': u'auto', u'early_stopping_patience': u'10'}\u001b[0m\n",
      "\u001b[34mProcess 1 is a worker.\u001b[0m\n",
      "\u001b[34m[07/22/2020 03:30:38 INFO 140662884079424] Detected entry point for worker worker\u001b[0m\n",
      "\u001b[34m[07/22/2020 03:30:38 INFO 140662884079424] Using early stopping with patience 10\u001b[0m\n",
      "\u001b[34m[07/22/2020 03:30:38 INFO 140662884079424] [cardinality=auto] `cat` field was NOT found in the file `/opt/ml/input/data/train/train.json` and will NOT be used for training.\u001b[0m\n",
      "\u001b[34m[07/22/2020 03:30:38 INFO 140662884079424] [num_dynamic_feat=auto] `dynamic_feat` field was NOT found in the file `/opt/ml/input/data/train/train.json` and will NOT be used for training.\u001b[0m\n",
      "\u001b[34m[07/22/2020 03:30:38 INFO 140662884079424] Training set statistics:\u001b[0m\n",
      "\u001b[34m[07/22/2020 03:30:38 INFO 140662884079424] Real time series\u001b[0m\n",
      "\u001b[34m[07/22/2020 03:30:38 INFO 140662884079424] number of time series: 200\u001b[0m\n",
      "\u001b[34m[07/22/2020 03:30:38 INFO 140662884079424] number of observations: 70400\u001b[0m\n",
      "\u001b[34m[07/22/2020 03:30:38 INFO 140662884079424] mean target length: 352\u001b[0m\n",
      "\u001b[34m[07/22/2020 03:30:38 INFO 140662884079424] min/mean/max target: 0.0611509121954/4.91773288304/14.6525497437\u001b[0m\n",
      "\u001b[34m[07/22/2020 03:30:38 INFO 140662884079424] mean abs(target): 4.91773288304\u001b[0m\n",
      "\u001b[34m[07/22/2020 03:30:38 INFO 140662884079424] contains missing values: no\u001b[0m\n",
      "\u001b[34m[07/22/2020 03:30:38 INFO 140662884079424] Small number of time series. Doing 2 passes over dataset with prob 0.8 per epoch.\u001b[0m\n",
      "\u001b[34m[07/22/2020 03:30:38 INFO 140662884079424] Test set statistics:\u001b[0m\n",
      "\u001b[34m[07/22/2020 03:30:38 INFO 140662884079424] Real time series\u001b[0m\n",
      "\u001b[34m[07/22/2020 03:30:38 INFO 140662884079424] number of time series: 200\u001b[0m\n",
      "\u001b[34m[07/22/2020 03:30:38 INFO 140662884079424] number of observations: 80000\u001b[0m\n",
      "\u001b[34m[07/22/2020 03:30:38 INFO 140662884079424] mean target length: 400\u001b[0m\n",
      "\u001b[34m[07/22/2020 03:30:38 INFO 140662884079424] min/mean/max target: 0.0611509121954/4.91507558107/14.6525497437\u001b[0m\n",
      "\u001b[34m[07/22/2020 03:30:38 INFO 140662884079424] mean abs(target): 4.91507558107\u001b[0m\n",
      "\u001b[34m[07/22/2020 03:30:38 INFO 140662884079424] contains missing values: no\u001b[0m\n",
      "\u001b[34m[07/22/2020 03:30:38 INFO 140662884079424] nvidia-smi took: 0.0755560398102 secs to identify 1 gpus\u001b[0m\n",
      "\u001b[34m[07/22/2020 03:30:38 INFO 140662884079424] Number of GPUs being used: 1\u001b[0m\n",
      "\u001b[34m[07/22/2020 03:30:38 INFO 140662884079424] Create Store: local\u001b[0m\n",
      "\u001b[34m#metrics {\"Metrics\": {\"get_graph.time\": {\"count\": 1, \"max\": 380.0380229949951, \"sum\": 380.0380229949951, \"min\": 380.0380229949951}}, \"EndTime\": 1595388638.927147, \"Dimensions\": {\"Host\": \"algo-1\", \"Operation\": \"training\", \"Algorithm\": \"AWS/DeepAR\"}, \"StartTime\": 1595388638.546081}\n",
      "\u001b[0m\n",
      "\u001b[34m[07/22/2020 03:30:38 INFO 140662884079424] Number of GPUs being used: 1\u001b[0m\n",
      "\u001b[34m#metrics {\"Metrics\": {\"initialize.time\": {\"count\": 1, \"max\": 4615.851879119873, \"sum\": 4615.851879119873, \"min\": 4615.851879119873}}, \"EndTime\": 1595388643.162101, \"Dimensions\": {\"Host\": \"algo-1\", \"Operation\": \"training\", \"Algorithm\": \"AWS/DeepAR\"}, \"StartTime\": 1595388638.927237}\n",
      "\u001b[0m\n",
      "\u001b[34m[07/22/2020 03:30:43 INFO 140662884079424] Epoch[0] Batch[0] avg_epoch_loss=1.892573\u001b[0m\n",
      "\u001b[34m[07/22/2020 03:30:43 INFO 140662884079424] #quality_metric: host=algo-1, epoch=0, batch=0 train loss <loss>=1.89257347584\u001b[0m\n",
      "\u001b[34m[07/22/2020 03:30:46 INFO 140662884079424] Epoch[0] Batch[5] avg_epoch_loss=1.584633\u001b[0m\n",
      "\u001b[34m[07/22/2020 03:30:46 INFO 140662884079424] #quality_metric: host=algo-1, epoch=0, batch=5 train loss <loss>=1.58463297288\u001b[0m\n",
      "\u001b[34m[07/22/2020 03:30:46 INFO 140662884079424] Epoch[0] Batch [5]#011Speed: 73.62 samples/sec#011loss=1.584633\u001b[0m\n",
      "\u001b[34m[07/22/2020 03:30:47 INFO 140662884079424] processed a total of 319 examples\u001b[0m\n",
      "\u001b[34m#metrics {\"Metrics\": {\"epochs\": {\"count\": 1, \"max\": 20, \"sum\": 20.0, \"min\": 20}, \"update.time\": {\"count\": 1, \"max\": 4700.517892837524, \"sum\": 4700.517892837524, \"min\": 4700.517892837524}}, \"EndTime\": 1595388647.863046, \"Dimensions\": {\"Host\": \"algo-1\", \"Operation\": \"training\", \"Algorithm\": \"AWS/DeepAR\"}, \"StartTime\": 1595388643.162204}\n",
      "\u001b[0m\n",
      "\u001b[34m[07/22/2020 03:30:47 INFO 140662884079424] #throughput_metric: host=algo-1, train throughput=67.8626215837 records/second\u001b[0m\n",
      "\u001b[34m[07/22/2020 03:30:47 INFO 140662884079424] #progress_metric: host=algo-1, completed 5 % of epochs\u001b[0m\n",
      "\u001b[34m[07/22/2020 03:30:47 INFO 140662884079424] #quality_metric: host=algo-1, epoch=0, train loss <loss>=1.41188064814\u001b[0m\n",
      "\u001b[34m[07/22/2020 03:30:47 INFO 140662884079424] best epoch loss so far\u001b[0m\n",
      "\u001b[34m[07/22/2020 03:30:47 INFO 140662884079424] Saved checkpoint to \"/opt/ml/model/state_20d6d1d0-fee4-4d65-bd03-84abb8358aca-0000.params\"\u001b[0m\n",
      "\u001b[34m#metrics {\"Metrics\": {\"state.serialize.time\": {\"count\": 1, \"max\": 51.71394348144531, \"sum\": 51.71394348144531, \"min\": 51.71394348144531}}, \"EndTime\": 1595388647.915557, \"Dimensions\": {\"Host\": \"algo-1\", \"Operation\": \"training\", \"Algorithm\": \"AWS/DeepAR\"}, \"StartTime\": 1595388647.863154}\n",
      "\u001b[0m\n",
      "\u001b[34m[07/22/2020 03:30:48 INFO 140662884079424] Epoch[1] Batch[0] avg_epoch_loss=1.104003\u001b[0m\n",
      "\u001b[34m[07/22/2020 03:30:48 INFO 140662884079424] #quality_metric: host=algo-1, epoch=1, batch=0 train loss <loss>=1.10400319099\u001b[0m\n",
      "\u001b[34m[07/22/2020 03:30:50 INFO 140662884079424] Epoch[1] Batch[5] avg_epoch_loss=0.876637\u001b[0m\n",
      "\u001b[34m[07/22/2020 03:30:50 INFO 140662884079424] #quality_metric: host=algo-1, epoch=1, batch=5 train loss <loss>=0.876636932294\u001b[0m\n",
      "\u001b[34m[07/22/2020 03:30:50 INFO 140662884079424] Epoch[1] Batch [5]#011Speed: 75.74 samples/sec#011loss=0.876637\u001b[0m\n",
      "\u001b[34m[07/22/2020 03:30:52 INFO 140662884079424] processed a total of 309 examples\u001b[0m\n",
      "\u001b[34m#metrics {\"Metrics\": {\"update.time\": {\"count\": 1, \"max\": 4359.783172607422, \"sum\": 4359.783172607422, \"min\": 4359.783172607422}}, \"EndTime\": 1595388652.275505, \"Dimensions\": {\"Host\": \"algo-1\", \"Operation\": \"training\", \"Algorithm\": \"AWS/DeepAR\"}, \"StartTime\": 1595388647.915647}\n",
      "\u001b[0m\n",
      "\u001b[34m[07/22/2020 03:30:52 INFO 140662884079424] #throughput_metric: host=algo-1, train throughput=70.8729100313 records/second\u001b[0m\n",
      "\u001b[34m[07/22/2020 03:30:52 INFO 140662884079424] #progress_metric: host=algo-1, completed 10 % of epochs\u001b[0m\n",
      "\u001b[34m[07/22/2020 03:30:52 INFO 140662884079424] #quality_metric: host=algo-1, epoch=1, train loss <loss>=0.802117437124\u001b[0m\n",
      "\u001b[34m[07/22/2020 03:30:52 INFO 140662884079424] best epoch loss so far\u001b[0m\n",
      "\u001b[34m[07/22/2020 03:30:52 INFO 140662884079424] Saved checkpoint to \"/opt/ml/model/state_a9bdb095-f8f7-4340-a742-e78af2fc3580-0000.params\"\u001b[0m\n",
      "\u001b[34m#metrics {\"Metrics\": {\"state.serialize.time\": {\"count\": 1, \"max\": 61.944007873535156, \"sum\": 61.944007873535156, \"min\": 61.944007873535156}}, \"EndTime\": 1595388652.338184, \"Dimensions\": {\"Host\": \"algo-1\", \"Operation\": \"training\", \"Algorithm\": \"AWS/DeepAR\"}, \"StartTime\": 1595388652.275592}\n",
      "\u001b[0m\n",
      "\u001b[34m[07/22/2020 03:30:52 INFO 140662884079424] Epoch[2] Batch[0] avg_epoch_loss=0.686375\u001b[0m\n",
      "\u001b[34m[07/22/2020 03:30:52 INFO 140662884079424] #quality_metric: host=algo-1, epoch=2, batch=0 train loss <loss>=0.686374545097\u001b[0m\n",
      "\u001b[34m[07/22/2020 03:30:54 INFO 140662884079424] Epoch[2] Batch[5] avg_epoch_loss=0.620688\u001b[0m\n",
      "\u001b[34m[07/22/2020 03:30:54 INFO 140662884079424] #quality_metric: host=algo-1, epoch=2, batch=5 train loss <loss>=0.620687961578\u001b[0m\n",
      "\u001b[34m[07/22/2020 03:30:54 INFO 140662884079424] Epoch[2] Batch [5]#011Speed: 75.80 samples/sec#011loss=0.620688\u001b[0m\n",
      "\u001b[34m[07/22/2020 03:30:56 INFO 140662884079424] processed a total of 277 examples\u001b[0m\n",
      "\u001b[34m#metrics {\"Metrics\": {\"update.time\": {\"count\": 1, \"max\": 3922.335147857666, \"sum\": 3922.335147857666, \"min\": 3922.335147857666}}, \"EndTime\": 1595388656.260677, \"Dimensions\": {\"Host\": \"algo-1\", \"Operation\": \"training\", \"Algorithm\": \"AWS/DeepAR\"}, \"StartTime\": 1595388652.338264}\n",
      "\u001b[0m\n",
      "\u001b[34m[07/22/2020 03:30:56 INFO 140662884079424] #throughput_metric: host=algo-1, train throughput=70.6188953782 records/second\u001b[0m\n",
      "\u001b[34m[07/22/2020 03:30:56 INFO 140662884079424] #progress_metric: host=algo-1, completed 15 % of epochs\u001b[0m\n",
      "\u001b[34m[07/22/2020 03:30:56 INFO 140662884079424] #quality_metric: host=algo-1, epoch=2, train loss <loss>=0.544809354676\u001b[0m\n",
      "\u001b[34m[07/22/2020 03:30:56 INFO 140662884079424] best epoch loss so far\u001b[0m\n",
      "\u001b[34m[07/22/2020 03:30:56 INFO 140662884079424] Saved checkpoint to \"/opt/ml/model/state_9915b6f4-f9f8-4c9f-99d3-bcf7223568a7-0000.params\"\u001b[0m\n",
      "\u001b[34m#metrics {\"Metrics\": {\"state.serialize.time\": {\"count\": 1, \"max\": 47.94192314147949, \"sum\": 47.94192314147949, \"min\": 47.94192314147949}}, \"EndTime\": 1595388656.309272, \"Dimensions\": {\"Host\": \"algo-1\", \"Operation\": \"training\", \"Algorithm\": \"AWS/DeepAR\"}, \"StartTime\": 1595388656.260758}\n",
      "\u001b[0m\n",
      "\u001b[34m[07/22/2020 03:30:56 INFO 140662884079424] Epoch[3] Batch[0] avg_epoch_loss=0.481751\u001b[0m\n",
      "\u001b[34m[07/22/2020 03:30:56 INFO 140662884079424] #quality_metric: host=algo-1, epoch=3, batch=0 train loss <loss>=0.481751412153\u001b[0m\n",
      "\u001b[34m[07/22/2020 03:30:58 INFO 140662884079424] Epoch[3] Batch[5] avg_epoch_loss=0.498622\u001b[0m\n",
      "\u001b[34m[07/22/2020 03:30:58 INFO 140662884079424] #quality_metric: host=algo-1, epoch=3, batch=5 train loss <loss>=0.498622447252\u001b[0m\n",
      "\u001b[34m[07/22/2020 03:30:58 INFO 140662884079424] Epoch[3] Batch [5]#011Speed: 75.80 samples/sec#011loss=0.498622\u001b[0m\n",
      "\u001b[34m[07/22/2020 03:31:00 INFO 140662884079424] processed a total of 303 examples\u001b[0m\n",
      "\u001b[34m#metrics {\"Metrics\": {\"update.time\": {\"count\": 1, \"max\": 4354.046821594238, \"sum\": 4354.046821594238, \"min\": 4354.046821594238}}, \"EndTime\": 1595388660.663483, \"Dimensions\": {\"Host\": \"algo-1\", \"Operation\": \"training\", \"Algorithm\": \"AWS/DeepAR\"}, \"StartTime\": 1595388656.309357}\n",
      "\u001b[0m\n",
      "\u001b[34m[07/22/2020 03:31:00 INFO 140662884079424] #throughput_metric: host=algo-1, train throughput=69.5885916847 records/second\u001b[0m\n",
      "\u001b[34m[07/22/2020 03:31:00 INFO 140662884079424] #progress_metric: host=algo-1, completed 20 % of epochs\u001b[0m\n",
      "\u001b[34m[07/22/2020 03:31:00 INFO 140662884079424] #quality_metric: host=algo-1, epoch=3, train loss <loss>=0.302178566158\u001b[0m\n",
      "\u001b[34m[07/22/2020 03:31:00 INFO 140662884079424] best epoch loss so far\u001b[0m\n",
      "\u001b[34m[07/22/2020 03:31:00 INFO 140662884079424] Saved checkpoint to \"/opt/ml/model/state_1fc6ee8f-1519-4edb-afdc-8718d202fa0f-0000.params\"\u001b[0m\n",
      "\u001b[34m#metrics {\"Metrics\": {\"state.serialize.time\": {\"count\": 1, \"max\": 50.2161979675293, \"sum\": 50.2161979675293, \"min\": 50.2161979675293}}, \"EndTime\": 1595388660.714467, \"Dimensions\": {\"Host\": \"algo-1\", \"Operation\": \"training\", \"Algorithm\": \"AWS/DeepAR\"}, \"StartTime\": 1595388660.663553}\n",
      "\u001b[0m\n",
      "\u001b[34m[07/22/2020 03:31:01 INFO 140662884079424] Epoch[4] Batch[0] avg_epoch_loss=0.424608\u001b[0m\n",
      "\u001b[34m[07/22/2020 03:31:01 INFO 140662884079424] #quality_metric: host=algo-1, epoch=4, batch=0 train loss <loss>=0.424607515335\u001b[0m\n",
      "\u001b[34m[07/22/2020 03:31:03 INFO 140662884079424] Epoch[4] Batch[5] avg_epoch_loss=0.237887\u001b[0m\n",
      "\u001b[34m[07/22/2020 03:31:03 INFO 140662884079424] #quality_metric: host=algo-1, epoch=4, batch=5 train loss <loss>=0.237887465085\u001b[0m\n",
      "\u001b[34m[07/22/2020 03:31:03 INFO 140662884079424] Epoch[4] Batch [5]#011Speed: 74.87 samples/sec#011loss=0.237887\u001b[0m\n",
      "\u001b[34m[07/22/2020 03:31:05 INFO 140662884079424] processed a total of 317 examples\u001b[0m\n",
      "\u001b[34m#metrics {\"Metrics\": {\"update.time\": {\"count\": 1, \"max\": 4368.282079696655, \"sum\": 4368.282079696655, \"min\": 4368.282079696655}}, \"EndTime\": 1595388665.082895, \"Dimensions\": {\"Host\": \"algo-1\", \"Operation\": \"training\", \"Algorithm\": \"AWS/DeepAR\"}, \"StartTime\": 1595388660.714542}\n",
      "\u001b[0m\n",
      "\u001b[34m[07/22/2020 03:31:05 INFO 140662884079424] #throughput_metric: host=algo-1, train throughput=72.5666646564 records/second\u001b[0m\n",
      "\u001b[34m[07/22/2020 03:31:05 INFO 140662884079424] #progress_metric: host=algo-1, completed 25 % of epochs\u001b[0m\n",
      "\u001b[34m[07/22/2020 03:31:05 INFO 140662884079424] #quality_metric: host=algo-1, epoch=4, train loss <loss>=0.186028917506\u001b[0m\n",
      "\u001b[34m[07/22/2020 03:31:05 INFO 140662884079424] best epoch loss so far\u001b[0m\n",
      "\u001b[34m[07/22/2020 03:31:05 INFO 140662884079424] Saved checkpoint to \"/opt/ml/model/state_0e081a29-aa4e-470a-b89b-5b144a471e3e-0000.params\"\u001b[0m\n",
      "\u001b[34m#metrics {\"Metrics\": {\"state.serialize.time\": {\"count\": 1, \"max\": 49.88217353820801, \"sum\": 49.88217353820801, \"min\": 49.88217353820801}}, \"EndTime\": 1595388665.133471, \"Dimensions\": {\"Host\": \"algo-1\", \"Operation\": \"training\", \"Algorithm\": \"AWS/DeepAR\"}, \"StartTime\": 1595388665.082967}\n",
      "\u001b[0m\n",
      "\u001b[34m[07/22/2020 03:31:05 INFO 140662884079424] Epoch[5] Batch[0] avg_epoch_loss=0.018948\u001b[0m\n",
      "\u001b[34m[07/22/2020 03:31:05 INFO 140662884079424] #quality_metric: host=algo-1, epoch=5, batch=0 train loss <loss>=0.0189476050436\u001b[0m\n",
      "\u001b[34m[07/22/2020 03:31:07 INFO 140662884079424] Epoch[5] Batch[5] avg_epoch_loss=0.190246\u001b[0m\n",
      "\u001b[34m[07/22/2020 03:31:07 INFO 140662884079424] #quality_metric: host=algo-1, epoch=5, batch=5 train loss <loss>=0.190246071046\u001b[0m\n",
      "\u001b[34m[07/22/2020 03:31:07 INFO 140662884079424] Epoch[5] Batch [5]#011Speed: 74.95 samples/sec#011loss=0.190246\u001b[0m\n",
      "\u001b[34m[07/22/2020 03:31:09 INFO 140662884079424] Epoch[5] Batch[10] avg_epoch_loss=0.120905\u001b[0m\n",
      "\u001b[34m[07/22/2020 03:31:09 INFO 140662884079424] #quality_metric: host=algo-1, epoch=5, batch=10 train loss <loss>=0.0376954376698\u001b[0m\n",
      "\u001b[34m[07/22/2020 03:31:09 INFO 140662884079424] Epoch[5] Batch [10]#011Speed: 74.17 samples/sec#011loss=0.037695\u001b[0m\n",
      "\u001b[34m[07/22/2020 03:31:09 INFO 140662884079424] processed a total of 347 examples\u001b[0m\n",
      "\u001b[34m#metrics {\"Metrics\": {\"update.time\": {\"count\": 1, \"max\": 4811.021089553833, \"sum\": 4811.021089553833, \"min\": 4811.021089553833}}, \"EndTime\": 1595388669.944644, \"Dimensions\": {\"Host\": \"algo-1\", \"Operation\": \"training\", \"Algorithm\": \"AWS/DeepAR\"}, \"StartTime\": 1595388665.133553}\n",
      "\u001b[0m\n",
      "\u001b[34m[07/22/2020 03:31:09 INFO 140662884079424] #throughput_metric: host=algo-1, train throughput=72.1242308678 records/second\u001b[0m\n",
      "\u001b[34m[07/22/2020 03:31:09 INFO 140662884079424] #progress_metric: host=algo-1, completed 30 % of epochs\u001b[0m\n",
      "\u001b[34m[07/22/2020 03:31:09 INFO 140662884079424] #quality_metric: host=algo-1, epoch=5, train loss <loss>=0.120904874057\u001b[0m\n",
      "\u001b[34m[07/22/2020 03:31:09 INFO 140662884079424] best epoch loss so far\u001b[0m\n",
      "\u001b[34m[07/22/2020 03:31:09 INFO 140662884079424] Saved checkpoint to \"/opt/ml/model/state_9c3caa89-8d6f-4631-a8ad-b9675cb8cb0e-0000.params\"\u001b[0m\n",
      "\u001b[34m#metrics {\"Metrics\": {\"state.serialize.time\": {\"count\": 1, \"max\": 48.57897758483887, \"sum\": 48.57897758483887, \"min\": 48.57897758483887}}, \"EndTime\": 1595388669.993903, \"Dimensions\": {\"Host\": \"algo-1\", \"Operation\": \"training\", \"Algorithm\": \"AWS/DeepAR\"}, \"StartTime\": 1595388669.944727}\n",
      "\u001b[0m\n",
      "\u001b[34m[07/22/2020 03:31:10 INFO 140662884079424] Epoch[6] Batch[0] avg_epoch_loss=0.104829\u001b[0m\n",
      "\u001b[34m[07/22/2020 03:31:10 INFO 140662884079424] #quality_metric: host=algo-1, epoch=6, batch=0 train loss <loss>=0.104828581214\u001b[0m\n",
      "\u001b[34m[07/22/2020 03:31:12 INFO 140662884079424] Epoch[6] Batch[5] avg_epoch_loss=0.059250\u001b[0m\n",
      "\u001b[34m[07/22/2020 03:31:12 INFO 140662884079424] #quality_metric: host=algo-1, epoch=6, batch=5 train loss <loss>=0.0592498245339\u001b[0m\n",
      "\u001b[34m[07/22/2020 03:31:12 INFO 140662884079424] Epoch[6] Batch [5]#011Speed: 74.36 samples/sec#011loss=0.059250\u001b[0m\n",
      "\u001b[34m[07/22/2020 03:31:14 INFO 140662884079424] Epoch[6] Batch[10] avg_epoch_loss=0.014011\u001b[0m\n",
      "\u001b[34m[07/22/2020 03:31:14 INFO 140662884079424] #quality_metric: host=algo-1, epoch=6, batch=10 train loss <loss>=-0.0402757398784\u001b[0m\n",
      "\u001b[34m[07/22/2020 03:31:14 INFO 140662884079424] Epoch[6] Batch [10]#011Speed: 74.38 samples/sec#011loss=-0.040276\u001b[0m\n",
      "\u001b[34m[07/22/2020 03:31:14 INFO 140662884079424] processed a total of 325 examples\u001b[0m\n",
      "\u001b[34m#metrics {\"Metrics\": {\"update.time\": {\"count\": 1, \"max\": 4816.555976867676, \"sum\": 4816.555976867676, \"min\": 4816.555976867676}}, \"EndTime\": 1595388674.81062, \"Dimensions\": {\"Host\": \"algo-1\", \"Operation\": \"training\", \"Algorithm\": \"AWS/DeepAR\"}, \"StartTime\": 1595388669.993991}\n",
      "\u001b[0m\n",
      "\u001b[34m[07/22/2020 03:31:14 INFO 140662884079424] #throughput_metric: host=algo-1, train throughput=67.4737053281 records/second\u001b[0m\n",
      "\u001b[34m[07/22/2020 03:31:14 INFO 140662884079424] #progress_metric: host=algo-1, completed 35 % of epochs\u001b[0m\n",
      "\u001b[34m[07/22/2020 03:31:14 INFO 140662884079424] #quality_metric: host=algo-1, epoch=6, train loss <loss>=0.0140109316192\u001b[0m\n",
      "\u001b[34m[07/22/2020 03:31:14 INFO 140662884079424] best epoch loss so far\u001b[0m\n",
      "\u001b[34m[07/22/2020 03:31:14 INFO 140662884079424] Saved checkpoint to \"/opt/ml/model/state_d7d7ceae-5998-470c-b4cc-63496f24b165-0000.params\"\u001b[0m\n",
      "\u001b[34m#metrics {\"Metrics\": {\"state.serialize.time\": {\"count\": 1, \"max\": 50.46200752258301, \"sum\": 50.46200752258301, \"min\": 50.46200752258301}}, \"EndTime\": 1595388674.861777, \"Dimensions\": {\"Host\": \"algo-1\", \"Operation\": \"training\", \"Algorithm\": \"AWS/DeepAR\"}, \"StartTime\": 1595388674.810711}\n",
      "\u001b[0m\n",
      "\u001b[34m[07/22/2020 03:31:15 INFO 140662884079424] Epoch[7] Batch[0] avg_epoch_loss=0.314481\u001b[0m\n",
      "\u001b[34m[07/22/2020 03:31:15 INFO 140662884079424] #quality_metric: host=algo-1, epoch=7, batch=0 train loss <loss>=0.314480960369\u001b[0m\n",
      "\u001b[34m[07/22/2020 03:31:17 INFO 140662884079424] Epoch[7] Batch[5] avg_epoch_loss=0.166182\u001b[0m\n",
      "\u001b[34m[07/22/2020 03:31:17 INFO 140662884079424] #quality_metric: host=algo-1, epoch=7, batch=5 train loss <loss>=0.166181820134\u001b[0m\n",
      "\u001b[34m[07/22/2020 03:31:17 INFO 140662884079424] Epoch[7] Batch [5]#011Speed: 75.16 samples/sec#011loss=0.166182\u001b[0m\n",
      "\u001b[34m[07/22/2020 03:31:18 INFO 140662884079424] processed a total of 288 examples\u001b[0m\n",
      "\u001b[34m#metrics {\"Metrics\": {\"update.time\": {\"count\": 1, \"max\": 3942.960023880005, \"sum\": 3942.960023880005, \"min\": 3942.960023880005}}, \"EndTime\": 1595388678.804902, \"Dimensions\": {\"Host\": \"algo-1\", \"Operation\": \"training\", \"Algorithm\": \"AWS/DeepAR\"}, \"StartTime\": 1595388674.861869}\n",
      "\u001b[0m\n",
      "\u001b[34m[07/22/2020 03:31:18 INFO 140662884079424] #throughput_metric: host=algo-1, train throughput=73.0395010606 records/second\u001b[0m\n",
      "\u001b[34m[07/22/2020 03:31:18 INFO 140662884079424] #progress_metric: host=algo-1, completed 40 % of epochs\u001b[0m\n",
      "\u001b[34m[07/22/2020 03:31:18 INFO 140662884079424] #quality_metric: host=algo-1, epoch=7, train loss <loss>=0.123729121354\u001b[0m\n",
      "\u001b[34m[07/22/2020 03:31:18 INFO 140662884079424] loss did not improve\u001b[0m\n",
      "\u001b[34m[07/22/2020 03:31:19 INFO 140662884079424] Epoch[8] Batch[0] avg_epoch_loss=-0.036453\u001b[0m\n",
      "\u001b[34m[07/22/2020 03:31:19 INFO 140662884079424] #quality_metric: host=algo-1, epoch=8, batch=0 train loss <loss>=-0.0364528782666\u001b[0m\n",
      "\u001b[34m[07/22/2020 03:31:21 INFO 140662884079424] Epoch[8] Batch[5] avg_epoch_loss=0.034744\u001b[0m\n",
      "\u001b[34m[07/22/2020 03:31:21 INFO 140662884079424] #quality_metric: host=algo-1, epoch=8, batch=5 train loss <loss>=0.0347437225282\u001b[0m\n",
      "\u001b[34m[07/22/2020 03:31:21 INFO 140662884079424] Epoch[8] Batch [5]#011Speed: 75.01 samples/sec#011loss=0.034744\u001b[0m\n",
      "\u001b[34m[07/22/2020 03:31:23 INFO 140662884079424] Epoch[8] Batch[10] avg_epoch_loss=0.074241\u001b[0m\n",
      "\u001b[34m[07/22/2020 03:31:23 INFO 140662884079424] #quality_metric: host=algo-1, epoch=8, batch=10 train loss <loss>=0.121637016535\u001b[0m\n",
      "\u001b[34m[07/22/2020 03:31:23 INFO 140662884079424] Epoch[8] Batch [10]#011Speed: 74.32 samples/sec#011loss=0.121637\u001b[0m\n",
      "\u001b[34m[07/22/2020 03:31:23 INFO 140662884079424] processed a total of 333 examples\u001b[0m\n",
      "\u001b[34m#metrics {\"Metrics\": {\"update.time\": {\"count\": 1, \"max\": 4803.326845169067, \"sum\": 4803.326845169067, \"min\": 4803.326845169067}}, \"EndTime\": 1595388683.608849, \"Dimensions\": {\"Host\": \"algo-1\", \"Operation\": \"training\", \"Algorithm\": \"AWS/DeepAR\"}, \"StartTime\": 1595388678.804972}\n",
      "\u001b[0m\n",
      "\u001b[34m[07/22/2020 03:31:23 INFO 140662884079424] #throughput_metric: host=algo-1, train throughput=69.3249404371 records/second\u001b[0m\n",
      "\u001b[34m[07/22/2020 03:31:23 INFO 140662884079424] #progress_metric: host=algo-1, completed 45 % of epochs\u001b[0m\n",
      "\u001b[34m[07/22/2020 03:31:23 INFO 140662884079424] #quality_metric: host=algo-1, epoch=8, train loss <loss>=0.0742406743494\u001b[0m\n",
      "\u001b[34m[07/22/2020 03:31:23 INFO 140662884079424] loss did not improve\u001b[0m\n",
      "\u001b[34m[07/22/2020 03:31:24 INFO 140662884079424] Epoch[9] Batch[0] avg_epoch_loss=0.195052\u001b[0m\n",
      "\u001b[34m[07/22/2020 03:31:24 INFO 140662884079424] #quality_metric: host=algo-1, epoch=9, batch=0 train loss <loss>=0.195052474737\u001b[0m\n",
      "\u001b[34m[07/22/2020 03:31:26 INFO 140662884079424] Epoch[9] Batch[5] avg_epoch_loss=0.040734\u001b[0m\n",
      "\u001b[34m[07/22/2020 03:31:26 INFO 140662884079424] #quality_metric: host=algo-1, epoch=9, batch=5 train loss <loss>=0.040734226505\u001b[0m\n",
      "\u001b[34m[07/22/2020 03:31:26 INFO 140662884079424] Epoch[9] Batch [5]#011Speed: 75.45 samples/sec#011loss=0.040734\u001b[0m\n",
      "\u001b[34m[07/22/2020 03:31:28 INFO 140662884079424] Epoch[9] Batch[10] avg_epoch_loss=0.103009\u001b[0m\n",
      "\u001b[34m[07/22/2020 03:31:28 INFO 140662884079424] #quality_metric: host=algo-1, epoch=9, batch=10 train loss <loss>=0.177738191187\u001b[0m\n",
      "\u001b[34m[07/22/2020 03:31:28 INFO 140662884079424] Epoch[9] Batch [10]#011Speed: 74.26 samples/sec#011loss=0.177738\u001b[0m\n",
      "\u001b[34m[07/22/2020 03:31:28 INFO 140662884079424] processed a total of 331 examples\u001b[0m\n",
      "\u001b[34m#metrics {\"Metrics\": {\"update.time\": {\"count\": 1, \"max\": 4793.823957443237, \"sum\": 4793.823957443237, \"min\": 4793.823957443237}}, \"EndTime\": 1595388688.403284, \"Dimensions\": {\"Host\": \"algo-1\", \"Operation\": \"training\", \"Algorithm\": \"AWS/DeepAR\"}, \"StartTime\": 1595388683.608945}\n",
      "\u001b[0m\n",
      "\u001b[34m[07/22/2020 03:31:28 INFO 140662884079424] #throughput_metric: host=algo-1, train throughput=69.0452173026 records/second\u001b[0m\n",
      "\u001b[34m[07/22/2020 03:31:28 INFO 140662884079424] #progress_metric: host=algo-1, completed 50 % of epochs\u001b[0m\n",
      "\u001b[34m[07/22/2020 03:31:28 INFO 140662884079424] #quality_metric: host=algo-1, epoch=9, train loss <loss>=0.103008755906\u001b[0m\n",
      "\u001b[34m[07/22/2020 03:31:28 INFO 140662884079424] loss did not improve\u001b[0m\n",
      "\u001b[34m[07/22/2020 03:31:28 INFO 140662884079424] Epoch[10] Batch[0] avg_epoch_loss=0.003925\u001b[0m\n",
      "\u001b[34m[07/22/2020 03:31:28 INFO 140662884079424] #quality_metric: host=algo-1, epoch=10, batch=0 train loss <loss>=0.00392451882362\u001b[0m\n",
      "\u001b[34m[07/22/2020 03:31:31 INFO 140662884079424] Epoch[10] Batch[5] avg_epoch_loss=0.024676\u001b[0m\n",
      "\u001b[34m[07/22/2020 03:31:31 INFO 140662884079424] #quality_metric: host=algo-1, epoch=10, batch=5 train loss <loss>=0.024675593401\u001b[0m\n",
      "\u001b[34m[07/22/2020 03:31:31 INFO 140662884079424] Epoch[10] Batch [5]#011Speed: 75.47 samples/sec#011loss=0.024676\u001b[0m\n",
      "\u001b[34m[07/22/2020 03:31:33 INFO 140662884079424] Epoch[10] Batch[10] avg_epoch_loss=0.060751\u001b[0m\n",
      "\u001b[34m[07/22/2020 03:31:33 INFO 140662884079424] #quality_metric: host=algo-1, epoch=10, batch=10 train loss <loss>=0.104041591287\u001b[0m\n",
      "\u001b[34m[07/22/2020 03:31:33 INFO 140662884079424] Epoch[10] Batch [10]#011Speed: 74.20 samples/sec#011loss=0.104042\u001b[0m\n",
      "\u001b[34m[07/22/2020 03:31:33 INFO 140662884079424] processed a total of 355 examples\u001b[0m\n",
      "\u001b[34m#metrics {\"Metrics\": {\"update.time\": {\"count\": 1, \"max\": 5212.39709854126, \"sum\": 5212.39709854126, \"min\": 5212.39709854126}}, \"EndTime\": 1595388693.616274, \"Dimensions\": {\"Host\": \"algo-1\", \"Operation\": \"training\", \"Algorithm\": \"AWS/DeepAR\"}, \"StartTime\": 1595388688.403379}\n",
      "\u001b[0m\n",
      "\u001b[34m[07/22/2020 03:31:33 INFO 140662884079424] #throughput_metric: host=algo-1, train throughput=68.1048757456 records/second\u001b[0m\n",
      "\u001b[34m[07/22/2020 03:31:33 INFO 140662884079424] #progress_metric: host=algo-1, completed 55 % of epochs\u001b[0m\n",
      "\u001b[34m[07/22/2020 03:31:33 INFO 140662884079424] #quality_metric: host=algo-1, epoch=10, train loss <loss>=0.124374741378\u001b[0m\n",
      "\u001b[34m[07/22/2020 03:31:33 INFO 140662884079424] loss did not improve\u001b[0m\n",
      "\u001b[34m[07/22/2020 03:31:34 INFO 140662884079424] Epoch[11] Batch[0] avg_epoch_loss=0.033057\u001b[0m\n",
      "\u001b[34m[07/22/2020 03:31:34 INFO 140662884079424] #quality_metric: host=algo-1, epoch=11, batch=0 train loss <loss>=0.0330572426319\u001b[0m\n",
      "\u001b[34m[07/22/2020 03:31:36 INFO 140662884079424] Epoch[11] Batch[5] avg_epoch_loss=0.191765\u001b[0m\n",
      "\u001b[34m[07/22/2020 03:31:36 INFO 140662884079424] #quality_metric: host=algo-1, epoch=11, batch=5 train loss <loss>=0.191764661421\u001b[0m\n",
      "\u001b[34m[07/22/2020 03:31:36 INFO 140662884079424] Epoch[11] Batch [5]#011Speed: 75.26 samples/sec#011loss=0.191765\u001b[0m\n",
      "\u001b[34m[07/22/2020 03:31:38 INFO 140662884079424] Epoch[11] Batch[10] avg_epoch_loss=0.110276\u001b[0m\n",
      "\u001b[34m[07/22/2020 03:31:38 INFO 140662884079424] #quality_metric: host=algo-1, epoch=11, batch=10 train loss <loss>=0.0124888829887\u001b[0m\n",
      "\u001b[34m[07/22/2020 03:31:38 INFO 140662884079424] Epoch[11] Batch [10]#011Speed: 73.20 samples/sec#011loss=0.012489\u001b[0m\n",
      "\u001b[34m[07/22/2020 03:31:38 INFO 140662884079424] processed a total of 329 examples\u001b[0m\n",
      "\u001b[34m#metrics {\"Metrics\": {\"update.time\": {\"count\": 1, \"max\": 4826.721906661987, \"sum\": 4826.721906661987, \"min\": 4826.721906661987}}, \"EndTime\": 1595388698.443616, \"Dimensions\": {\"Host\": \"algo-1\", \"Operation\": \"training\", \"Algorithm\": \"AWS/DeepAR\"}, \"StartTime\": 1595388693.616352}\n",
      "\u001b[0m\n",
      "\u001b[34m[07/22/2020 03:31:38 INFO 140662884079424] #throughput_metric: host=algo-1, train throughput=68.160466045 records/second\u001b[0m\n",
      "\u001b[34m[07/22/2020 03:31:38 INFO 140662884079424] #progress_metric: host=algo-1, completed 60 % of epochs\u001b[0m\n",
      "\u001b[34m[07/22/2020 03:31:38 INFO 140662884079424] #quality_metric: host=algo-1, epoch=11, train loss <loss>=0.110275671225\u001b[0m\n",
      "\u001b[34m[07/22/2020 03:31:38 INFO 140662884079424] loss did not improve\u001b[0m\n",
      "\u001b[34m[07/22/2020 03:31:38 INFO 140662884079424] Epoch[12] Batch[0] avg_epoch_loss=0.161736\u001b[0m\n",
      "\u001b[34m[07/22/2020 03:31:38 INFO 140662884079424] #quality_metric: host=algo-1, epoch=12, batch=0 train loss <loss>=0.161735534668\u001b[0m\n",
      "\u001b[34m[07/22/2020 03:31:41 INFO 140662884079424] Epoch[12] Batch[5] avg_epoch_loss=-0.008015\u001b[0m\n",
      "\u001b[34m[07/22/2020 03:31:41 INFO 140662884079424] #quality_metric: host=algo-1, epoch=12, batch=5 train loss <loss>=-0.00801530014724\u001b[0m\n",
      "\u001b[34m[07/22/2020 03:31:41 INFO 140662884079424] Epoch[12] Batch [5]#011Speed: 75.10 samples/sec#011loss=-0.008015\u001b[0m\n",
      "\u001b[34m[07/22/2020 03:31:43 INFO 140662884079424] Epoch[12] Batch[10] avg_epoch_loss=-0.045905\u001b[0m\n",
      "\u001b[34m[07/22/2020 03:31:43 INFO 140662884079424] #quality_metric: host=algo-1, epoch=12, batch=10 train loss <loss>=-0.091372872889\u001b[0m\n",
      "\u001b[34m[07/22/2020 03:31:43 INFO 140662884079424] Epoch[12] Batch [10]#011Speed: 74.00 samples/sec#011loss=-0.091373\u001b[0m\n",
      "\u001b[34m[07/22/2020 03:31:43 INFO 140662884079424] processed a total of 345 examples\u001b[0m\n",
      "\u001b[34m#metrics {\"Metrics\": {\"update.time\": {\"count\": 1, \"max\": 4804.527997970581, \"sum\": 4804.527997970581, \"min\": 4804.527997970581}}, \"EndTime\": 1595388703.248789, \"Dimensions\": {\"Host\": \"algo-1\", \"Operation\": \"training\", \"Algorithm\": \"AWS/DeepAR\"}, \"StartTime\": 1595388698.4437}\n",
      "\u001b[0m\n",
      "\u001b[34m[07/22/2020 03:31:43 INFO 140662884079424] #throughput_metric: host=algo-1, train throughput=71.8051844773 records/second\u001b[0m\n",
      "\u001b[34m[07/22/2020 03:31:43 INFO 140662884079424] #progress_metric: host=algo-1, completed 65 % of epochs\u001b[0m\n",
      "\u001b[34m[07/22/2020 03:31:43 INFO 140662884079424] #quality_metric: host=algo-1, epoch=12, train loss <loss>=-0.045905105939\u001b[0m\n",
      "\u001b[34m[07/22/2020 03:31:43 INFO 140662884079424] best epoch loss so far\u001b[0m\n",
      "\u001b[34m[07/22/2020 03:31:43 INFO 140662884079424] Saved checkpoint to \"/opt/ml/model/state_b2285b55-b929-4af0-b4d6-9ce9fc1bb4b8-0000.params\"\u001b[0m\n",
      "\u001b[34m#metrics {\"Metrics\": {\"state.serialize.time\": {\"count\": 1, \"max\": 49.616098403930664, \"sum\": 49.616098403930664, \"min\": 49.616098403930664}}, \"EndTime\": 1595388703.299095, \"Dimensions\": {\"Host\": \"algo-1\", \"Operation\": \"training\", \"Algorithm\": \"AWS/DeepAR\"}, \"StartTime\": 1595388703.248882}\n",
      "\u001b[0m\n",
      "\u001b[34m[07/22/2020 03:31:43 INFO 140662884079424] Epoch[13] Batch[0] avg_epoch_loss=0.108931\u001b[0m\n",
      "\u001b[34m[07/22/2020 03:31:43 INFO 140662884079424] #quality_metric: host=algo-1, epoch=13, batch=0 train loss <loss>=0.108931273222\u001b[0m\n",
      "\u001b[34m[07/22/2020 03:31:45 INFO 140662884079424] Epoch[13] Batch[5] avg_epoch_loss=0.074343\u001b[0m\n",
      "\u001b[34m[07/22/2020 03:31:45 INFO 140662884079424] #quality_metric: host=algo-1, epoch=13, batch=5 train loss <loss>=0.0743429139256\u001b[0m\n",
      "\u001b[34m[07/22/2020 03:31:45 INFO 140662884079424] Epoch[13] Batch [5]#011Speed: 75.46 samples/sec#011loss=0.074343\u001b[0m\n",
      "\u001b[34m[07/22/2020 03:31:48 INFO 140662884079424] Epoch[13] Batch[10] avg_epoch_loss=0.081153\u001b[0m\n",
      "\u001b[34m[07/22/2020 03:31:48 INFO 140662884079424] #quality_metric: host=algo-1, epoch=13, batch=10 train loss <loss>=0.0893260464072\u001b[0m\n",
      "\u001b[34m[07/22/2020 03:31:48 INFO 140662884079424] Epoch[13] Batch [10]#011Speed: 74.10 samples/sec#011loss=0.089326\u001b[0m\n",
      "\u001b[34m[07/22/2020 03:31:48 INFO 140662884079424] processed a total of 325 examples\u001b[0m\n",
      "\u001b[34m#metrics {\"Metrics\": {\"update.time\": {\"count\": 1, \"max\": 4795.884847640991, \"sum\": 4795.884847640991, \"min\": 4795.884847640991}}, \"EndTime\": 1595388708.095145, \"Dimensions\": {\"Host\": \"algo-1\", \"Operation\": \"training\", \"Algorithm\": \"AWS/DeepAR\"}, \"StartTime\": 1595388703.299185}\n",
      "\u001b[0m\n",
      "\u001b[34m[07/22/2020 03:31:48 INFO 140662884079424] #throughput_metric: host=algo-1, train throughput=67.7645479276 records/second\u001b[0m\n",
      "\u001b[34m[07/22/2020 03:31:48 INFO 140662884079424] #progress_metric: host=algo-1, completed 70 % of epochs\u001b[0m\n",
      "\u001b[34m[07/22/2020 03:31:48 INFO 140662884079424] #quality_metric: host=algo-1, epoch=13, train loss <loss>=0.08115342869\u001b[0m\n",
      "\u001b[34m[07/22/2020 03:31:48 INFO 140662884079424] loss did not improve\u001b[0m\n",
      "\u001b[34m[07/22/2020 03:31:48 INFO 140662884079424] Epoch[14] Batch[0] avg_epoch_loss=0.101533\u001b[0m\n",
      "\u001b[34m[07/22/2020 03:31:48 INFO 140662884079424] #quality_metric: host=algo-1, epoch=14, batch=0 train loss <loss>=0.101533025503\u001b[0m\n",
      "\u001b[34m[07/22/2020 03:31:50 INFO 140662884079424] Epoch[14] Batch[5] avg_epoch_loss=0.090569\u001b[0m\n",
      "\u001b[34m[07/22/2020 03:31:50 INFO 140662884079424] #quality_metric: host=algo-1, epoch=14, batch=5 train loss <loss>=0.0905691155543\u001b[0m\n",
      "\u001b[34m[07/22/2020 03:31:50 INFO 140662884079424] Epoch[14] Batch [5]#011Speed: 75.49 samples/sec#011loss=0.090569\u001b[0m\n",
      "\u001b[34m[07/22/2020 03:31:52 INFO 140662884079424] Epoch[14] Batch[10] avg_epoch_loss=0.110064\u001b[0m\n",
      "\u001b[34m[07/22/2020 03:31:52 INFO 140662884079424] #quality_metric: host=algo-1, epoch=14, batch=10 train loss <loss>=0.133458151296\u001b[0m\n",
      "\u001b[34m[07/22/2020 03:31:52 INFO 140662884079424] Epoch[14] Batch [10]#011Speed: 74.44 samples/sec#011loss=0.133458\u001b[0m\n",
      "\u001b[34m[07/22/2020 03:31:52 INFO 140662884079424] processed a total of 325 examples\u001b[0m\n",
      "\u001b[34m#metrics {\"Metrics\": {\"update.time\": {\"count\": 1, \"max\": 4789.314031600952, \"sum\": 4789.314031600952, \"min\": 4789.314031600952}}, \"EndTime\": 1595388712.88505, \"Dimensions\": {\"Host\": \"algo-1\", \"Operation\": \"training\", \"Algorithm\": \"AWS/DeepAR\"}, \"StartTime\": 1595388708.095236}\n",
      "\u001b[0m\n",
      "\u001b[34m[07/22/2020 03:31:52 INFO 140662884079424] #throughput_metric: host=algo-1, train throughput=67.8577631535 records/second\u001b[0m\n",
      "\u001b[34m[07/22/2020 03:31:52 INFO 140662884079424] #progress_metric: host=algo-1, completed 75 % of epochs\u001b[0m\n",
      "\u001b[34m[07/22/2020 03:31:52 INFO 140662884079424] #quality_metric: host=algo-1, epoch=14, train loss <loss>=0.1100641318\u001b[0m\n",
      "\u001b[34m[07/22/2020 03:31:52 INFO 140662884079424] loss did not improve\u001b[0m\n",
      "\u001b[34m[07/22/2020 03:31:53 INFO 140662884079424] Epoch[15] Batch[0] avg_epoch_loss=0.051321\u001b[0m\n",
      "\u001b[34m[07/22/2020 03:31:53 INFO 140662884079424] #quality_metric: host=algo-1, epoch=15, batch=0 train loss <loss>=0.0513211488724\u001b[0m\n",
      "\u001b[34m[07/22/2020 03:31:55 INFO 140662884079424] Epoch[15] Batch[5] avg_epoch_loss=0.033864\u001b[0m\n",
      "\u001b[34m[07/22/2020 03:31:55 INFO 140662884079424] #quality_metric: host=algo-1, epoch=15, batch=5 train loss <loss>=0.0338635047277\u001b[0m\n",
      "\u001b[34m[07/22/2020 03:31:55 INFO 140662884079424] Epoch[15] Batch [5]#011Speed: 75.41 samples/sec#011loss=0.033864\u001b[0m\n",
      "\u001b[34m[07/22/2020 03:31:57 INFO 140662884079424] Epoch[15] Batch[10] avg_epoch_loss=-0.036273\u001b[0m\n",
      "\u001b[34m[07/22/2020 03:31:57 INFO 140662884079424] #quality_metric: host=algo-1, epoch=15, batch=10 train loss <loss>=-0.120436117053\u001b[0m\n",
      "\u001b[34m[07/22/2020 03:31:57 INFO 140662884079424] Epoch[15] Batch [10]#011Speed: 74.31 samples/sec#011loss=-0.120436\u001b[0m\n",
      "\u001b[34m[07/22/2020 03:31:58 INFO 140662884079424] processed a total of 354 examples\u001b[0m\n",
      "\u001b[34m#metrics {\"Metrics\": {\"update.time\": {\"count\": 1, \"max\": 5214.753150939941, \"sum\": 5214.753150939941, \"min\": 5214.753150939941}}, \"EndTime\": 1595388718.100369, \"Dimensions\": {\"Host\": \"algo-1\", \"Operation\": \"training\", \"Algorithm\": \"AWS/DeepAR\"}, \"StartTime\": 1595388712.885132}\n",
      "\u001b[0m\n",
      "\u001b[34m[07/22/2020 03:31:58 INFO 140662884079424] #throughput_metric: host=algo-1, train throughput=67.882596741 records/second\u001b[0m\n",
      "\u001b[34m[07/22/2020 03:31:58 INFO 140662884079424] #progress_metric: host=algo-1, completed 80 % of epochs\u001b[0m\n",
      "\u001b[34m[07/22/2020 03:31:58 INFO 140662884079424] #quality_metric: host=algo-1, epoch=15, train loss <loss>=-0.00787629559636\u001b[0m\n",
      "\u001b[34m[07/22/2020 03:31:58 INFO 140662884079424] loss did not improve\u001b[0m\n",
      "\u001b[34m[07/22/2020 03:31:58 INFO 140662884079424] Epoch[16] Batch[0] avg_epoch_loss=0.102275\u001b[0m\n",
      "\u001b[34m[07/22/2020 03:31:58 INFO 140662884079424] #quality_metric: host=algo-1, epoch=16, batch=0 train loss <loss>=0.102275341749\u001b[0m\n",
      "\u001b[34m[07/22/2020 03:32:00 INFO 140662884079424] Epoch[16] Batch[5] avg_epoch_loss=0.164302\u001b[0m\n",
      "\u001b[34m[07/22/2020 03:32:00 INFO 140662884079424] #quality_metric: host=algo-1, epoch=16, batch=5 train loss <loss>=0.16430216531\u001b[0m\n",
      "\u001b[34m[07/22/2020 03:32:00 INFO 140662884079424] Epoch[16] Batch [5]#011Speed: 75.61 samples/sec#011loss=0.164302\u001b[0m\n",
      "\u001b[34m[07/22/2020 03:32:02 INFO 140662884079424] processed a total of 317 examples\u001b[0m\n",
      "\u001b[34m#metrics {\"Metrics\": {\"update.time\": {\"count\": 1, \"max\": 4357.429981231689, \"sum\": 4357.429981231689, \"min\": 4357.429981231689}}, \"EndTime\": 1595388722.458442, \"Dimensions\": {\"Host\": \"algo-1\", \"Operation\": \"training\", \"Algorithm\": \"AWS/DeepAR\"}, \"StartTime\": 1595388718.100461}\n",
      "\u001b[0m\n",
      "\u001b[34m[07/22/2020 03:32:02 INFO 140662884079424] #throughput_metric: host=algo-1, train throughput=72.7474654273 records/second\u001b[0m\n",
      "\u001b[34m[07/22/2020 03:32:02 INFO 140662884079424] #progress_metric: host=algo-1, completed 85 % of epochs\u001b[0m\n",
      "\u001b[34m[07/22/2020 03:32:02 INFO 140662884079424] #quality_metric: host=algo-1, epoch=16, train loss <loss>=0.0427612952888\u001b[0m\n",
      "\u001b[34m[07/22/2020 03:32:02 INFO 140662884079424] loss did not improve\u001b[0m\n",
      "\u001b[34m[07/22/2020 03:32:02 INFO 140662884079424] Epoch[17] Batch[0] avg_epoch_loss=-0.093357\u001b[0m\n",
      "\u001b[34m[07/22/2020 03:32:02 INFO 140662884079424] #quality_metric: host=algo-1, epoch=17, batch=0 train loss <loss>=-0.0933572500944\u001b[0m\n",
      "\u001b[34m[07/22/2020 03:32:05 INFO 140662884079424] Epoch[17] Batch[5] avg_epoch_loss=-0.092450\u001b[0m\n",
      "\u001b[34m[07/22/2020 03:32:05 INFO 140662884079424] #quality_metric: host=algo-1, epoch=17, batch=5 train loss <loss>=-0.0924500947197\u001b[0m\n",
      "\u001b[34m[07/22/2020 03:32:05 INFO 140662884079424] Epoch[17] Batch [5]#011Speed: 75.51 samples/sec#011loss=-0.092450\u001b[0m\n",
      "\u001b[34m[07/22/2020 03:32:06 INFO 140662884079424] processed a total of 315 examples\u001b[0m\n",
      "\u001b[34m#metrics {\"Metrics\": {\"update.time\": {\"count\": 1, \"max\": 4356.417179107666, \"sum\": 4356.417179107666, \"min\": 4356.417179107666}}, \"EndTime\": 1595388726.815414, \"Dimensions\": {\"Host\": \"algo-1\", \"Operation\": \"training\", \"Algorithm\": \"AWS/DeepAR\"}, \"StartTime\": 1595388722.458511}\n",
      "\u001b[0m\n",
      "\u001b[34m[07/22/2020 03:32:06 INFO 140662884079424] #throughput_metric: host=algo-1, train throughput=72.3054666077 records/second\u001b[0m\n",
      "\u001b[34m[07/22/2020 03:32:06 INFO 140662884079424] #progress_metric: host=algo-1, completed 90 % of epochs\u001b[0m\n",
      "\u001b[34m[07/22/2020 03:32:06 INFO 140662884079424] #quality_metric: host=algo-1, epoch=17, train loss <loss>=0.00101393312216\u001b[0m\n",
      "\u001b[34m[07/22/2020 03:32:06 INFO 140662884079424] loss did not improve\u001b[0m\n",
      "\u001b[34m[07/22/2020 03:32:07 INFO 140662884079424] Epoch[18] Batch[0] avg_epoch_loss=-0.043854\u001b[0m\n",
      "\u001b[34m[07/22/2020 03:32:07 INFO 140662884079424] #quality_metric: host=algo-1, epoch=18, batch=0 train loss <loss>=-0.0438541769981\u001b[0m\n",
      "\u001b[34m[07/22/2020 03:32:09 INFO 140662884079424] Epoch[18] Batch[5] avg_epoch_loss=-0.037309\u001b[0m\n",
      "\u001b[34m[07/22/2020 03:32:09 INFO 140662884079424] #quality_metric: host=algo-1, epoch=18, batch=5 train loss <loss>=-0.0373094500974\u001b[0m\n",
      "\u001b[34m[07/22/2020 03:32:09 INFO 140662884079424] Epoch[18] Batch [5]#011Speed: 74.84 samples/sec#011loss=-0.037309\u001b[0m\n",
      "\n",
      "2020-07-22 03:32:23 Uploading - Uploading generated training model\u001b[34m[07/22/2020 03:32:11 INFO 140662884079424] Epoch[18] Batch[10] avg_epoch_loss=-0.034459\u001b[0m\n",
      "\u001b[34m[07/22/2020 03:32:11 INFO 140662884079424] #quality_metric: host=algo-1, epoch=18, batch=10 train loss <loss>=-0.0310393989086\u001b[0m\n",
      "\u001b[34m[07/22/2020 03:32:11 INFO 140662884079424] Epoch[18] Batch [10]#011Speed: 74.77 samples/sec#011loss=-0.031039\u001b[0m\n",
      "\u001b[34m[07/22/2020 03:32:11 INFO 140662884079424] processed a total of 331 examples\u001b[0m\n",
      "\u001b[34m#metrics {\"Metrics\": {\"update.time\": {\"count\": 1, \"max\": 4797.420978546143, \"sum\": 4797.420978546143, \"min\": 4797.420978546143}}, \"EndTime\": 1595388731.613451, \"Dimensions\": {\"Host\": \"algo-1\", \"Operation\": \"training\", \"Algorithm\": \"AWS/DeepAR\"}, \"StartTime\": 1595388726.815477}\n",
      "\u001b[0m\n",
      "\u001b[34m[07/22/2020 03:32:11 INFO 140662884079424] #throughput_metric: host=algo-1, train throughput=68.9935184644 records/second\u001b[0m\n",
      "\u001b[34m[07/22/2020 03:32:11 INFO 140662884079424] #progress_metric: host=algo-1, completed 95 % of epochs\u001b[0m\n",
      "\u001b[34m[07/22/2020 03:32:11 INFO 140662884079424] #quality_metric: host=algo-1, epoch=18, train loss <loss>=-0.0344594268298\u001b[0m\n",
      "\u001b[34m[07/22/2020 03:32:11 INFO 140662884079424] loss did not improve\u001b[0m\n",
      "\u001b[34m[07/22/2020 03:32:12 INFO 140662884079424] Epoch[19] Batch[0] avg_epoch_loss=-0.042118\u001b[0m\n",
      "\u001b[34m[07/22/2020 03:32:12 INFO 140662884079424] #quality_metric: host=algo-1, epoch=19, batch=0 train loss <loss>=-0.0421176813543\u001b[0m\n",
      "\u001b[34m[07/22/2020 03:32:14 INFO 140662884079424] Epoch[19] Batch[5] avg_epoch_loss=0.045681\u001b[0m\n",
      "\u001b[34m[07/22/2020 03:32:14 INFO 140662884079424] #quality_metric: host=algo-1, epoch=19, batch=5 train loss <loss>=0.0456808765108\u001b[0m\n",
      "\u001b[34m[07/22/2020 03:32:14 INFO 140662884079424] Epoch[19] Batch [5]#011Speed: 74.93 samples/sec#011loss=0.045681\u001b[0m\n",
      "\u001b[34m[07/22/2020 03:32:16 INFO 140662884079424] Epoch[19] Batch[10] avg_epoch_loss=0.046247\u001b[0m\n",
      "\u001b[34m[07/22/2020 03:32:16 INFO 140662884079424] #quality_metric: host=algo-1, epoch=19, batch=10 train loss <loss>=0.046927203238\u001b[0m\n",
      "\u001b[34m[07/22/2020 03:32:16 INFO 140662884079424] Epoch[19] Batch [10]#011Speed: 74.05 samples/sec#011loss=0.046927\u001b[0m\n",
      "\u001b[34m[07/22/2020 03:32:16 INFO 140662884079424] processed a total of 332 examples\u001b[0m\n",
      "\u001b[34m#metrics {\"Metrics\": {\"update.time\": {\"count\": 1, \"max\": 4812.103986740112, \"sum\": 4812.103986740112, \"min\": 4812.103986740112}}, \"EndTime\": 1595388736.426132, \"Dimensions\": {\"Host\": \"algo-1\", \"Operation\": \"training\", \"Algorithm\": \"AWS/DeepAR\"}, \"StartTime\": 1595388731.613536}\n",
      "\u001b[0m\n",
      "\u001b[34m[07/22/2020 03:32:16 INFO 140662884079424] #throughput_metric: host=algo-1, train throughput=68.9908683925 records/second\u001b[0m\n",
      "\u001b[34m[07/22/2020 03:32:16 INFO 140662884079424] #progress_metric: host=algo-1, completed 100 % of epochs\u001b[0m\n",
      "\u001b[34m[07/22/2020 03:32:16 INFO 140662884079424] #quality_metric: host=algo-1, epoch=19, train loss <loss>=0.0462473886596\u001b[0m\n",
      "\u001b[34m[07/22/2020 03:32:16 INFO 140662884079424] loss did not improve\u001b[0m\n",
      "\u001b[34m[07/22/2020 03:32:16 INFO 140662884079424] Final loss: -0.045905105939 (occurred at epoch 12)\u001b[0m\n",
      "\u001b[34m[07/22/2020 03:32:16 INFO 140662884079424] #quality_metric: host=algo-1, train final_loss <loss>=-0.045905105939\u001b[0m\n",
      "\u001b[34m[07/22/2020 03:32:16 INFO 140662884079424] Worker algo-1 finished training.\u001b[0m\n",
      "\u001b[34m[07/22/2020 03:32:16 WARNING 140662884079424] wait_for_all_workers will not sync workers since the kv store is not running distributed\u001b[0m\n",
      "\u001b[34m[07/22/2020 03:32:16 INFO 140662884079424] All workers finished. Serializing model for prediction.\u001b[0m\n",
      "\u001b[34m#metrics {\"Metrics\": {\"get_graph.time\": {\"count\": 1, \"max\": 479.13503646850586, \"sum\": 479.13503646850586, \"min\": 479.13503646850586}}, \"EndTime\": 1595388736.906334, \"Dimensions\": {\"Host\": \"algo-1\", \"Operation\": \"training\", \"Algorithm\": \"AWS/DeepAR\"}, \"StartTime\": 1595388736.426217}\n",
      "\u001b[0m\n",
      "\u001b[34m[07/22/2020 03:32:17 INFO 140662884079424] Number of GPUs being used: 1\u001b[0m\n",
      "\u001b[34m#metrics {\"Metrics\": {\"finalize.time\": {\"count\": 1, \"max\": 723.3760356903076, \"sum\": 723.3760356903076, \"min\": 723.3760356903076}}, \"EndTime\": 1595388737.150533, \"Dimensions\": {\"Host\": \"algo-1\", \"Operation\": \"training\", \"Algorithm\": \"AWS/DeepAR\"}, \"StartTime\": 1595388736.906415}\n",
      "\u001b[0m\n",
      "\u001b[34m[07/22/2020 03:32:17 INFO 140662884079424] Serializing to /opt/ml/model/model_algo-1\u001b[0m\n",
      "\u001b[34m[07/22/2020 03:32:17 INFO 140662884079424] Saved checkpoint to \"/opt/ml/model/model_algo-1-0000.params\"\u001b[0m\n",
      "\u001b[34m#metrics {\"Metrics\": {\"model.serialize.time\": {\"count\": 1, \"max\": 38.20300102233887, \"sum\": 38.20300102233887, \"min\": 38.20300102233887}}, \"EndTime\": 1595388737.188881, \"Dimensions\": {\"Host\": \"algo-1\", \"Operation\": \"training\", \"Algorithm\": \"AWS/DeepAR\"}, \"StartTime\": 1595388737.150623}\n",
      "\u001b[0m\n",
      "\u001b[34m[07/22/2020 03:32:17 INFO 140662884079424] Successfully serialized the model for prediction.\u001b[0m\n",
      "\u001b[34m[07/22/2020 03:32:17 INFO 140662884079424] Evaluating model accuracy on testset using 100 samples\u001b[0m\n",
      "\u001b[34m#metrics {\"Metrics\": {\"model.bind.time\": {\"count\": 1, \"max\": 0.04696846008300781, \"sum\": 0.04696846008300781, \"min\": 0.04696846008300781}}, \"EndTime\": 1595388737.189764, \"Dimensions\": {\"Host\": \"algo-1\", \"Operation\": \"training\", \"Algorithm\": \"AWS/DeepAR\"}, \"StartTime\": 1595388737.188943}\n",
      "\u001b[0m\n",
      "\u001b[34m#metrics {\"Metrics\": {\"model.score.time\": {\"count\": 1, \"max\": 3015.516996383667, \"sum\": 3015.516996383667, \"min\": 3015.516996383667}}, \"EndTime\": 1595388740.205226, \"Dimensions\": {\"Host\": \"algo-1\", \"Operation\": \"training\", \"Algorithm\": \"AWS/DeepAR\"}, \"StartTime\": 1595388737.189826}\n",
      "\u001b[0m\n",
      "\u001b[34m[07/22/2020 03:32:20 INFO 140662884079424] #test_score (algo-1, RMSE): 0.341370279653\u001b[0m\n",
      "\u001b[34m[07/22/2020 03:32:20 INFO 140662884079424] #test_score (algo-1, mean_absolute_QuantileLoss): 1810.5495679287446\u001b[0m\n",
      "\u001b[34m[07/22/2020 03:32:20 INFO 140662884079424] #test_score (algo-1, mean_wQuantileLoss): 0.0385242523222367\u001b[0m\n",
      "\u001b[34m[07/22/2020 03:32:20 INFO 140662884079424] #test_score (algo-1, wQuantileLoss[0.1]): 0.021220547491223547\u001b[0m\n",
      "\u001b[34m[07/22/2020 03:32:20 INFO 140662884079424] #test_score (algo-1, wQuantileLoss[0.2]): 0.034003577308788585\u001b[0m\n",
      "\u001b[34m[07/22/2020 03:32:20 INFO 140662884079424] #test_score (algo-1, wQuantileLoss[0.3]): 0.042499013449546194\u001b[0m\n",
      "\u001b[34m[07/22/2020 03:32:20 INFO 140662884079424] #test_score (algo-1, wQuantileLoss[0.4]): 0.047516044990895547\u001b[0m\n",
      "\u001b[34m[07/22/2020 03:32:20 INFO 140662884079424] #test_score (algo-1, wQuantileLoss[0.5]): 0.04945731068984374\u001b[0m\n",
      "\u001b[34m[07/22/2020 03:32:20 INFO 140662884079424] #test_score (algo-1, wQuantileLoss[0.6]): 0.04837629097505226\u001b[0m\n",
      "\u001b[34m[07/22/2020 03:32:20 INFO 140662884079424] #test_score (algo-1, wQuantileLoss[0.7]): 0.044141901022069295\u001b[0m\n",
      "\u001b[34m[07/22/2020 03:32:20 INFO 140662884079424] #test_score (algo-1, wQuantileLoss[0.8]): 0.036211873306235605\u001b[0m\n",
      "\u001b[34m[07/22/2020 03:32:20 INFO 140662884079424] #test_score (algo-1, wQuantileLoss[0.9]): 0.023291711666475545\u001b[0m\n",
      "\u001b[34m[07/22/2020 03:32:20 INFO 140662884079424] #quality_metric: host=algo-1, test mean_wQuantileLoss <loss>=0.0385242523222\u001b[0m\n",
      "\u001b[34m[07/22/2020 03:32:20 INFO 140662884079424] #quality_metric: host=algo-1, test RMSE <loss>=0.341370279653\u001b[0m\n",
      "\u001b[34m#metrics {\"Metrics\": {\"totaltime\": {\"count\": 1, \"max\": 102015.58780670166, \"sum\": 102015.58780670166, \"min\": 102015.58780670166}, \"setuptime\": {\"count\": 1, \"max\": 9.891986846923828, \"sum\": 9.891986846923828, \"min\": 9.891986846923828}}, \"EndTime\": 1595388740.258599, \"Dimensions\": {\"Host\": \"algo-1\", \"Operation\": \"training\", \"Algorithm\": \"AWS/DeepAR\"}, \"StartTime\": 1595388740.205302}\n",
      "\u001b[0m\n",
      "\n",
      "2020-07-22 03:32:30 Completed - Training job completed\n",
      "Training seconds: 170\n",
      "Billable seconds: 170\n"
     ]
    }
   ],
   "source": [
    "estimator.fit(inputs=data_channels)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### 5. Deploy the model"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 24,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "---------------!"
     ]
    }
   ],
   "source": [
    "job_name = estimator.latest_training_job.name\n",
    "\n",
    "endpoint_name = sagemaker_session.endpoint_from_job(\n",
    "    job_name=job_name,\n",
    "    initial_instance_count=1,\n",
    "    instance_type='ml.p2.xlarge',\n",
    "    deployment_image=image_name,\n",
    "    role=role\n",
    ")"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "To query the endpoint and perform predictions, we can define the following utility class: this allows making requests using pandas.Series objects rather than raw JSON strings. \n",
    "This function is from AWS official blog."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 49,
   "metadata": {},
   "outputs": [],
   "source": [
    "from datetime import timedelta\n",
    "\n",
    "class DeepARPredictor(sagemaker.predictor.RealTimePredictor):\n",
    "    \n",
    "    def set_prediction_parameters(self, freq, prediction_length):\n",
    "        \"\"\"Set the time frequency and prediction length parameters. This method **must** be \n",
    "        called before being able to use `predict`.\n",
    "        \n",
    "        Parameters:\n",
    "        freq -- string indicating the time frequency\n",
    "        prediction_length -- integer, number of predicted time points\n",
    "       \n",
    "        Return value: none.\n",
    "        \"\"\"\n",
    "        self.freq = freq\n",
    "        self.prediction_length = prediction_length\n",
    "        \n",
    "    def predict(self, ts, cat=None, encoding='utf-8', num_samples=100, quantiles=['0.1', '0.5', '0.9']):\n",
    "        \"\"\"Requests the prediction of for the time series listed in `ts`, each with the \n",
    "        (optional) corresponding category listed in `cat`.\n",
    "        \n",
    "        Parameters:\n",
    "        ts -- list of `pandas.Series` objects, the time series to predict\n",
    "        cat -- list of integers (default: None)\n",
    "        encoding -- string, encoding to use for the request (default: 'utf-8')\n",
    "        num_samples -- integer, number of samples to compute at prediction time (default: 100)\n",
    "        quantiles -- list of strings specifying the quantiles to compute (default: ['0.1', '0.5', '0.9'])\n",
    "        \n",
    "        Return value: list of `pandas.DataFrame` objects, each containing the predictions\n",
    "        \"\"\"\n",
    "        prediction_times = [x.index[-1]+timedelta(days=1) for x in ts] \n",
    "        req = self.__encode_request(ts, cat, encoding, num_samples, quantiles)\n",
    "        res = super(DeepARPredictor, self).predict(req)\n",
    "        return self.__decode_response(res, prediction_times, encoding)\n",
    "    \n",
    "    def __encode_request(self, ts, cat, encoding, num_samples, quantiles):\n",
    "        instances = [series_to_obj(ts[k], cat[k] if cat else None) for k in range(len(ts))]\n",
    "        configuration = {'num_samples': num_samples, 'output_types': ['quantiles'], 'quantiles': quantiles}\n",
    "        http_request_data = {'instances': instances, 'configuration': configuration}\n",
    "        return json.dumps(http_request_data).encode(encoding)\n",
    "    \n",
    "    def __decode_response(self, response, prediction_times, encoding):\n",
    "        response_data = json.loads(response.decode(encoding))\n",
    "        list_of_df = []\n",
    "        for k in range(len(prediction_times)):\n",
    "            prediction_index = pd.date_range(prediction_times[k], freq=self.freq, periods=self.prediction_length)\n",
    "            list_of_df.append(pd.DataFrame(data=response_data['predictions'][k]['quantiles'], index=prediction_index))\n",
    "        return list_of_df"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 50,
   "metadata": {},
   "outputs": [],
   "source": [
    "predictor = DeepARPredictor(\n",
    "    endpoint=endpoint_name,\n",
    "    sagemaker_session=sagemaker_session,\n",
    "    content_type=\"application/json\"\n",
    ")\n",
    "predictor.set_prediction_parameters(freq, prediction_length)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 56,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "48"
      ]
     },
     "execution_count": 56,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "predictor.prediction_length"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### 6. Predicting and evaluating with the test set"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "For simplicity, we will predict only the first few time series used for training, and compare the results with the actual data we kept in the test set."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "list_of_df = predictor.predict(time_series_training[:5])\n",
    "actual_data = time_series[:5]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "for k in range(len(list_of_df)):\n",
    "    plt.figure(figsize=(12,6))\n",
    "    actual_data[k][-prediction_length-context_length:].plot(label='target')\n",
    "    p10 = list_of_df[k]['0.1']\n",
    "    p90 = list_of_df[k]['0.9']\n",
    "    plt.fill_between(p10.index, p10, p90, color='y', alpha=0.5, label='80% confidence interval')\n",
    "    list_of_df[k]['0.5'].plot(label='prediction median')\n",
    "    plt.legend()\n",
    "    plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### 7. Close the SageMaker Instance"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 58,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "deepar-synthetic-2020-07-22-03-27-10-217\n"
     ]
    }
   ],
   "source": [
    "print(endpoint_name)\n",
    "sagemaker_session.delete_endpoint(endpoint_name)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Reference:\n",
    ">https://github.com/awslabs/amazon-sagemaker-examples"
   ]
  },
  {
   "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": "3.7.6"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 4
}