notebook-samples/Watson-ML-pipeline/watson-ml-pipeline.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Train, store and deploy tensorflow model on Watson Machine Learning service"
   ]
  },
  {
   "attachments": {
    "Screen%20Shot%202019-02-15%20at%204.59.42%20PM.png": {
     "image/png": "iVBORw0KGgoAAAANSUhEUgAAASAAAAHUCAYAAACagPWPAAAMRWlDQ1BJQ0MgUHJvZmlsZQAASImVVwdYU8kWnltSIbQAAlJCb6IU6VJCaJEqVbARkkBCCTEhiNiRRQXXLiJgQ1dFFF0LIGvFXhbF3h+KqCi66IoNlTcpoKvfe+975/vm3j9nzvlPydx7ZwDQrOGIxTmoFgC5onxJXFgQY3xKKoPUCRCAAgoAQI3DlYqZsbGREIPB+z/l3Q1oDeWqk5zr5/n/Kto8vpQLABILcTpPys2FeD8AeAlXLMkHgOgF9ZbT8sVyPBFiXQlMEGKxHGcqcYkcpytxpcImIY4F8U4AyOocjiQTAI1mqGcUcDMhj8YtiJ1FPKEIAE0yxP5cAYcHcTjEI3Jz8+QY2gG79O94Mv/BmT7EyeFkDmFlLQohBwul4hzO9P+zHf9bcnNkgzFs4FAXSMLj5DXDvt3KzouQY3WIe0Tp0TEQ60D8QchT2EOMUgWy8ESlPWrMlbJgz4A+xM48TnAExMYQh4pyoiNV+vQMYSgbYrhC0EJhPjtB5buQLw2JV3HWSPLiYgZxhoTFVPk2cCSKuHL7k7LsRKaK/5aAzx7kf1skSEhW5oxRC4RJ0RBrQKwvzY6PUNpgVkUCVvSgjUQWJ8/fCmIfvigsSMmPTc6QhMap7CW50sF6sYUCITtahavyBQnhKp6dXI4ifwOIm/kiZuIgD186PnKwFh4/OERZO3aZL0pU1Yt1iPOD4lS+b8Q5sSp7nMrPCZPrLSA2lhbEq3xx/3y4IJX8eLQ4PzZBmSeensUZG6vMBy8EkYAFggEDyOBIB3kgCwjbepp64C/lTCjgAAnIBHzgpNIMeiQrZkTwGg+KwEuI+EA65BekmOWDAqj/MqRVXp1AhmK2QOGRDZ5AnAsiQA78LVN4iYaiJYHHUCP8KToX5poDh3zuZx0TaiJVGtkgL0Nz0JIYQgwmhhNDifa4Ee6P++KR8BoIhyvuhXsPZvvNnvCE0E54RLhO6CDcniIslvxQDwNEgQ4YIVRVc/r3NeM2kNUdD8L9ID/kxvVxI+CEj4aRmHgAjO0OtSxV5vLqf+T+Rw3fdV1lR3GmoJRhlECK3Y+eGg4a7kMs8p5+3yFlrulDfWUNzfwYn/Vdp3nwHvGjJbYQ24edwY5j57BDWBNgYEexZuwidliOh1bRY8UqGowWp8gnG/IIf4rHUcWUd1LqXO/c7fxZOZfPL5S/HwErTzxdIswU5DOY8M3PZ7BF3JEjGK7OLt4AyL8jytdU7yXF9wEx1P6mm5cHwJiagYGBg990UfcB2P8SAOqdbzrbMvg6KADgbBVXJilQ6nD5hQCoQBM+UYbAFFgCO1iPK/AAviAQhICxIAYkgBQwGXZZANezBEwDM8E8UArKwTKwGlSBDWAz2A52gb2gCRwCx8FpcAFcBtfBXbh6usAL0AvegX4EQUgIDaEjhogZYo04Iq6IF+KPhCCRSBySgqQhmYgIkSEzkflIObICqUI2IXXI78hB5DhyDmlHbiMPkW7kDfIJxVB1VBc1QW3QUagXykQj0AR0EpqJTkWL0BJ0CVqJ1qI70Ub0OHoBvY52oC/QPgxgapg+Zo45YV4YC4vBUrEMTILNxsqwCqwWa8Ba4P98FevAerCPOBGn4wzcCa7gcDwR5+JT8dn4YrwK34434ifxq/hDvBf/SqARjAmOBB8CmzCekEmYRiglVBC2Eg4QTsGnqYvwjkgk6hNtiZ7waUwhZhFnEBcT1xF3E48R24mdxD4SiWRIciT5kWJIHFI+qZS0lrSTdJR0hdRF+kBWI5uRXcmh5FSyiFxMriDvIB8hXyE/JfdTtCjWFB9KDIVHmU5ZStlCaaFconRR+qnaVFuqHzWBmkWdR62kNlBPUe9R/1ZTU7NQ81YbpyZUm6tWqbZH7azaQ7WP6jrqDuos9YnqMvUl6tvUj6nfVv+bRqPZ0AJpqbR82hJaHe0E7QHtgwZdY6QGW4OnMUejWqNR44rGK02KprUmU3OyZpFmheY+zUuaPVoULRstlhZHa7ZWtdZBrZtafdp0bRftGO1c7cXaO7TPaT/TIenY6ITo8HRKdDbrnNDppGN0SzqLzqXPp2+hn6J36RJ1bXXZulm65bq7dNt0e/V09EbrJekV6lXrHdbr0Mf0bfTZ+jn6S/X36t/Q/zTMZBhzGH/YomENw64Me28w3CDQgG9QZrDb4LrBJ0OGYYhhtuFywybD+0a4kYPROKNpRuuNThn1DNcd7jucO7xs+N7hd4xRYwfjOOMZxpuNLxr3mZiahJmITdaanDDpMdU3DTTNMl1lesS024xu5m8mNFtldtTsOUOPwWTkMCoZJxm95sbm4eYy803mbeb9FrYWiRbFFrst7ltSLb0sMyxXWbZa9lqZWUVZzbSqt7pjTbH2shZYr7E+Y/3extYm2WaBTZPNM1sDW7ZtkW297T07ml2A3VS7Wrtr9kR7L/ts+3X2lx1QB3cHgUO1wyVH1NHDUei4zrF9BGGE9wjRiNoRN53UnZhOBU71Tg9H6o+MHFk8smnkq1FWo1JHLR91ZtRXZ3fnHOctzndddFzGuhS7tLi8cXVw5bpWu15zo7mFus1xa3Z7PdpxNH/0+tG33OnuUe4L3Fvdv3h4ekg8Gjy6Pa080zxrPG966XrFei32OutN8A7ynuN9yPujj4dPvs9en798nXyzfXf4PhtjO4Y/ZsuYTj8LP47fJr8Of4Z/mv9G/44A8wBOQG3Ao0DLQF7g1sCnTHtmFnMn81WQc5Ak6EDQe5YPaxbrWDAWHBZcFtwWohOSGFIV8iDUIjQztD60N8w9bEbYsXBCeET48vCbbBM2l13H7h3rOXbW2JMR6hHxEVURjyIdIiWRLVFo1NiolVH3oq2jRdFNMSCGHbMy5n6sbezU2D/GEcfFjqse9yTOJW5m3Jl4evyU+B3x7xKCEpYm3E20S5QltiZpJk1Mqkt6nxycvCK5Y/yo8bPGX0gxShGmNKeSUpNSt6b2TQiZsHpC10T3iaUTb0yynVQ46dxko8k5kw9P0ZzCmbIvjZCWnLYj7TMnhlPL6Utnp9ek93JZ3DXcF7xA3ipeN9+Pv4L/NMMvY0XGs0y/zJWZ3YIAQYWgR8gSVglfZ4Vnbch6nx2TvS17ICc5Z3cuOTct96BIR5QtOplnmleY1y52FJeKO6b6TF09tVcSIdkqRaSTpM35unDDflFmJ/tF9rDAv6C64MO0pGn7CrULRYUXpztMXzT9aVFo0W8z8BncGa0zzWfOm/lwFnPWptnI7PTZrXMs55TM6ZobNnf7POq87Hl/FjsXryh+Oz95fkuJScncks5fwn6pL9UolZTeXOC7YMNCfKFwYdsit0VrF30t45WdL3curyj/vJi7+PyvLr9W/jqwJGNJ21KPpeuXEZeJlt1YHrB8+wrtFUUrOldGrWxcxVhVturt6imrz1WMrtiwhrpGtqajMrKyea3V2mVrP1cJqq5XB1XvrjGuWVTzfh1v3ZX1gesbNphsKN/waaNw461NYZsaa21qKzYTNxdsfrIlacuZ37x+q9tqtLV865dtom0d2+O2n6zzrKvbYbxjaT1aL6vv3jlx5+VdwbuaG5waNu3W312+B+yR7Xn+e9rvN/ZG7G3d57WvYb/1/poD9ANljUjj9MbeJkFTR3NKc/vBsQdbW3xbDvwx8o9th8wPVR/WO7z0CPVIyZGBo0VH+46Jj/Uczzze2Tql9e6J8SeunRx3su1UxKmzp0NPnzjDPHP0rN/ZQ+d8zh0873W+6YLHhcaL7hcP/On+54E2j7bGS56Xmi97X25pH9N+5ErAleNXg6+evsa+duF69PX2G4k3bt2ceLPjFu/Ws9s5t1/fKbjTf3fuPcK9svta9yseGD+o/Zf9v3Z3eHQcfhj88OKj+Ed3O7mdLx5LH3/uKnlCe1Lx1Oxp3TPXZ4e6Q7svP5/wvOuF+EV/T+lL7Zc1r+xe7f8r8K+LveN7u15LXg+8Wfy34d/b3o5+29oX2/fgXe67/vdlHww/bP/o9fHMp+RPT/unfSZ9rvxi/6Xla8TXewO5AwNijoSj2ApgcKAZGQC82QYALQUA+mW4f5igPOcpBFGeTRUI/CesPAsqxAOABniTb9dZxwDYA4dNIOSeC0AMvCcEAtTNbWioRJrh5qrk0qgHgGQ+MPAG7m8ocHwOGxjojx0Y+FIDk70GwJFnyvOlXIjwbLDRWY6umO0DP8q/AcXcf5Gh18WvAAAACXBIWXMAABYlAAAWJQFJUiTwAAACBGlUWHRYTUw6Y29tLmFkb2JlLnhtcAAAAAAAPHg6eG1wbWV0YSB4bWxuczp4PSJhZG9iZTpuczptZXRhLyIgeDp4bXB0az0iWE1QIENvcmUgNS40LjAiPgogICA8cmRmOlJERiB4bWxuczpyZGY9Imh0dHA6Ly93d3cudzMub3JnLzE5OTkvMDIvMjItcmRmLXN5bnRheC1ucyMiPgogICAgICA8cmRmOkRlc2NyaXB0aW9uIHJkZjphYm91dD0iIgogICAgICAgICAgICB4bWxuczpleGlmPSJodHRwOi8vbnMuYWRvYmUuY29tL2V4aWYvMS4wLyIKICAgICAgICAgICAgeG1sbnM6dGlmZj0iaHR0cDovL25zLmFkb2JlLmNvbS90aWZmLzEuMC8iPgogICAgICAgICA8ZXhpZjpQaXhlbFlEaW1lbnNpb24+NzAyPC9leGlmOlBpeGVsWURpbWVuc2lvbj4KICAgICAgICAgPGV4aWY6UGl4ZWxYRGltZW5zaW9uPjQzMjwvZXhpZjpQaXhlbFhEaW1lbnNpb24+CiAgICAgICAgIDx0aWZmOk9yaWVudGF0aW9uPjE8L3RpZmY6T3JpZW50YXRpb24+CiAgICAgIDwvcmRmOkRlc2NyaXB0aW9uPgogICA8L3JkZjpSREY+CjwveDp4bXBtZXRhPgpodN4/AABAAElEQVR4Ae19C5wU1ZX+mQczw/slb2QVISJgJEQUEeShIkIWxeiKq4JEo67ZBdesZqNJ1IgxMcZXDNm/b12MqFndKKxijKhkFRSEoAI/UVbeLwV5zzAz8K/vzJympukepoquruqu7/Lrqapbt+495zv3fvXdW9V0we7duw8IExEgAkQgBAQKQ2iTTRIBIkAEFAESEDsCESACoSFAAgoNejZMBIgACYh9gAgQgdAQIAGFBj0bJgJEgATEPkAEiEBoCJCAQoOeDRMBIkACYh8gAkQgNARIQKFBz4aJABEgAbEPEAEiEBoCJKDQoGfDRIAIkIDYB4gAEQgNARJQaNCzYSJABEhA7ANEgAiEhgAJKDTo2TARIAIkIPYBIkAEQkOgOLSWUzR8QP9rNP7/aCmgYRYRyDoCBQUFgbcZKgEdqGGcwJ1kA0SACHhHINX4zDQpZZ2A3E6l21eolJwcBgYJYx9szC1xYD8IcBxg5NWONR2ENcNPx55zDPJxj9lMkFFWCciMx9Y++/fvr+OUlan1nxsiQARCQsAIprCwUMnHjrHFB2PV8vyamBUCMlJJJp2ioiLBx/KtnF9neB0RIAKZQ8BIprKyUokGYxV5ICQkO38kJJQVAoKxmD2BYKqrq5V0qqqqZO3atbJz507NgxKqKcdFaAWCf4hAyAiAcJo2bSrt27fXMWtjF/kgoUIoIWeLce2XhAInILe6gQMwfPXq1TJnzhxZuHChgHjcxrv3Q8afzROBWCNgM5J+/frJkCFDpHPnzioWiouLBR+HlaQQ4/cISKggyN8FMwdAMkY+y5cvl/vuu0/g1KhRo6R79+7KslY21hGn80QgIghACDjcICtXrpTZs2fLokWL5Oqrr5aePXuqYCgpKVESMjWE8n7EQ+AEBGIx8oHymTp1qowZM0bGjx9fw6IRAZxmEAEikBoBLJc899wMmTlzlkyePFm6du2qMxmQUKNGjXR6ZgTklYQCfxMaBIQPnMC0C8rHyMfWfVK7zVwiQATCRgBjFNOtiy8eL/3795e5c+dKeXm5jmcIC5w/knEcGAEZ8WALmbZx40Zd88G0Cw7BaFtNDxtktk8EiEBqBDBGjYRGjhwpS5culU2bNmkeRMX+/dUqMDDOkWyburZDcwMjIGvK2BFPu7CPNR8kko8hxC0RiDYCNlaPPfZYFQ87duzQZRUQUHV13ff4vHoSOAGZEoJcw/wQj/WYiAARyD0EMHZBRngvyKZfNr5t69WrQAjIZJhtoXxMCVmeV0NZnggQgfARgIjAWDYCwv6RjOlACMgNk9s4ryvk7nq4TwSIQDQQMNLB2HZ//FgXOAHBKDPSj4G8hggQgWghgPFsJHSklgVLQLUr40dqJK8nAkQgegi4hYV734ulwRKQM19kIgJEIM8QSPF1TSyv+BntwRJQCkPzLBR0hwjEDwE/TJMGpWAJCP+5ERMRIAJ5j4BOwXx4GTABZZAqfTjHS4gAEcgOAtGcglEBZSf6bIUIhIwAFVDIAWDzRCDOCFABxTn69J0IhIwAFVDIAWDzRCDOCFABxTn69J0IhIwAFVDIAWDzRCDOCFABxTn69J0IhIwAFVDIAWDzRCDOCFABxTn69J0IhIwAFVDIAWDzRCDOCFABxTn69J0IhIwAFVDIAWDzRCDOCFABxTn69J0IhIwAFVDIAWDzRCDOCFABxTn69J0IhIwAFVDIAWDzRCDOCFABxTn69J0IhIwAFVDIAWDzRCDOCFABxTn69J0IhIwAFVDIAWDzRCDOCFABxTn69J0IhIwAFVDAAQDAmfo52oBNZfVEIOsI+FVAxcFamr3fBVMGdkiisDCYXxpSgPlLr8F2F9aeswiYAvL6Q1zBjNYEjF7NSVzoeQcEERT5wJj58+fLk08+KVu3blXbADgTESACNQj4VUABE1D2Bum6devk008/lcrKyoz3iYqKCpk8ebJMmjRJ5s2bp/WTgDIOMyvMYQRMAXl1IWACCl4BwXGszTz++ONy/PHHy7Zt2xQD5CHZeQWotmzyWo67DM7ZtVqB86e0tFRuvfVWue222+Sb3/ymZoPx3fXYvl5/oKZtu55bIpDvCPhVQDm/BqSOO2TQtm1b6dOnj5IFgo3pGIjFzlsHwLEl93nLty3K2Hnsjx49Wj/ufHdZm/658w44vwxb4PxjIgL5jgDGCuY7Xnt7wATk1RzvYVq/fr188cUX+tm7d6+899570qpVK2nXrp0cd9xxgqnZ559/Lt/61rdk+/bt8te//lV27NghgwcPlt69e6uKwfnPPvtMVq5cKS1btJQePXtI//79paSkJEFCOL9582Y56aSTpGnTpqq0Fi1aJCeeeKI0a9ZMPvzwQ/nkk0+krKxMTjjhBL2+qKgocb13z3gFEcgdBHDj9TPaA56CBbcGBMZFwuLw6aefLi+99JISz7nnniunnXaavPLKK3r+o48+kqFDh8rTTz8t48ePl0suuUSuueYaWbx4sZLDU089Jd/4xjdU3cyYMUMun3C5Xv+zn/1Mvv76a1VQqOjtt9/WdrZs2aL1gtjOPPNMbefnP/+5Etpjjz0mEydOlFNOOUXbw3QMgTFb9UL+IQJ5iIApIK+uBUxAfjixYS7YVGfIkCGqPEAuUDhYJF62bJlcdNFFWhFUDNI///M/y6DTBwkIacmSJTJs2DBZs2aNfO9735Mf//jHsmLFCvnTn/6kiul3v/ud/OpXv5IXX3xRr8WfRo0a6b61a9srr7xSysvL9frZs2fLxx9/LBMmTNB6sSjORATigADGg5/RHvAULDgFpEF1qj/qqKP006FDB9m1a5dOf1q0aJGIua3N/OAHP5Cbf3yzqqTESWdn/Yb10rpVa506Wf51112nCunZZ5+VSy+9VNeVklWMEdBll12mBNa+fXu9HNO/K664QhUQCK5Xr15WLbdEIG8RMAXklYQCJiCv5niMj1O9Ou5Mx/D4HWsuUCMgoOrqaj02ohg5cqSSD/ItD+TUoX0HwVoOpmR4gta8eXMlsa5du+p60e7du+slIEzvQD7V+6sdY0Tb7NKlizpiT+SsPY/esTgRyBkE0Mf9jPaACShgBeSERx2H87VrLTbYbWvKpVOnThpMU0Q4j5cK77rrLrnnnnv08Tqmc1j3ufbaa6V169a64Ix1nPqSKR/AX1BYE4Li4hpYre36ruc5IpAPCMRTASVFDqRixJN0KvGWNICyMi+//LKSzxNPPCHnnXeePt2CQpo6darcfvvtul5kZZPrs2MjNDvGlsTjRoP7cUBAx54PR3NeAbl91pcAaxVLMnG4SQHn8HYzFon79esnY8eOVcWDMli07tatm+zZs8ddNfeJABGoBwGMHcx3vE7DcvYpWDIWeP9m48aNsnbtWj0FJeNOyYSEt5vx8iLWfr788kstijK47g9/+IM8//zzSkruOhq6n9xWQ69jOSKQqwigz3slH/ia8woIzAvn8TUMpF/84hf65AnK5sILL5SqqirNNwWEslBKmDrh/SGk66+/Xt/fwRO1OXPmyJ133qlTslWrVul5/LG1IKvHtpafKOjs1HfOXY77RCBfEECfj7UCGjRoUOLb6iCQnTt3amzt/R3bItPWbQYOHKiEg+kY3iM666yzVBEt+GCBXHzxxdK4ceNEWSgskJpdiyduSMhPTjgHQsT1TEQgDgj4VUAFzmPmjD+qSigAZzpT6SgQGIevKeBN4UcffVQfVWcyKGgPbSDhsTkIBY/TQTpQKJhWgRSMPJLbRnl8zQLn8T4RnmLZddhH3agDeXZs51PVC3ugvFCfEVVymzwmArmGAMYA3pE7++yz9QaL9VJ8LQk3YYwLGws2FhviX8BTMD+zwoaYXbcMHDYSAiD4WAIJpCMeFY0O/WI96Oijj7ZLtK7k6wCum0ySzycudnZgj1txuc9xnwjkIwLo835Ge8CL0BkXV2lj52ZdU2BpC9eeAGR2Ha6x6yzvcNfzPBEgAjUI6PjxAUZeKKBkv/0QiJ9rktvlMRGIKwIYP7FWQHENPP0mAlFAwK8CCngK5ocTowAnbSACRMALAlRAXtBiWSJABDKKABVQRuFkZUSACHhBgArIC1osSwSIQEYRoALKKJysjAgQAS8IUAF5QYtliQARyCgCVEAZhZOVEQEi4AUBKiAvaLEsESACGUWACiijcLIyIkAEvCBABeQFLZYlAkQgowhQAWUUTlZGBIiAFwSogLygxbJEgAhkFAEqoIzCycqIABHwggAVkBe0WJYIEIGMIkAFlFE4WRkRIAJeEKAC8oIWyxIBIpBRBKiAMgonKyMCRMALAlRAXtBiWSJABDKKABVQRuFkZUSACHhBgArIC1osSwSIQEYRoALKKJysjAgQAS8IUAF5QYtliQARyCgCVEAZhZOVEQEi4AUBKiAvaLEsESACGUWACiijcLIyIkAEvCBABeQFLZYlAkQgowhQAWUUTlZGBIiAFwT8KqBiL414L3vA+yW8IqcQ0DvfgYNx1o5YwJ/kzqkgZsBYU0BeIx8sAbEjZiC00a1i//79UlhYKCAdd0JnRErOd5fhfg4jcPB+k3AimgrIdWdMWMqdnEcABIMOB/LZsWOHbNu2TSorK/W4WbNm0r59e/XRyuW8w3SgLgJ17zeJWIOXUpyqe23SUdYVEO+KSRHIsUMjlfLycnnnnXfk1VdflU2bNompniZNmsiQIUNk1KhR0rFjR81nzHMsyA00V1VPrfp17zfwci0WLAE5TcA+GIe7JTop7pSlpaVebGTZiCBg5LNr1y556KGHZPny5TJx4kTp06ePtG3bVvbs2SMrV66UWbNmyQ033CC333679OzZkyQUkfgdqRkYu+gDNu3OxI2l8EiNOvz1BWo0SKe6ulrWr1+vl9gd8/DXs0QUEDDyqaqqkscff1w2btwod999twwfPlynXEVFRdK8eXM56aST5KYf3SSjzz1Xbr31Vi2Hjsp4RyGK/myw2GHs7tu3T8pKy1RUmOqxrZ/aAyEgY0YzDIuV7dq1k27dusn8+fMTdiKfKbcQgOp588035cYbb1TiQQytg8ITHBcXFcv4f7xE+vbtK6+99lpuOUhr6yDgHqMffPCBtG7dWtq0baNloIRMDdW5yMNB4FOwwoJCqTpQJVgbGDZsmEr3Hj16yMknn6wsis7r7sAebGfRLCKAmwk64xtvvCHnn3++dOnSReOGDuhOODYSGjNmjDz44INyrqOGOnTooArYbk7ua7gfTQQQK4vvggUL5IUXXpDvXnihjmXkQ/WayLCtV08CJyDHBzUU0q1Xr15ywQUXyB133CFTpkxREmrRooU64dVwls8+Auhky5Ytk2uvvbbexq3Tdu/eXXbu3CmYtiGhwzLlFgJ4ygnymTZtmk63e3Q/Th1o1KiRFBcXOzE99DUMLx4GRkDorEgHnC320flg8NChQ6Vx48byxBNPyMyZM6V3797StGnThM12XSKDO5FBALFZt26drvXAKCjX+uKFOGPq/dJLL+kTMSxi1lc+Mo7G3BCL6+7du2Xp0qX6lHP06NE6pS4pLZGSkhIBAWFMFxYeVEGAzWt8AyMgiyEMMrmGRWiQ0KmnniqYhuFuig6Nj03FbGvXcxsdBNDh8J7P5s2bNX7pOpt14C+/+lKNX7t2rWzZskWnZtHxhpakQ8DiWlZWpk84x44dqzcdxB8Pk5B/kIBq1oHS1XW4/KwQkJEQjEbCC0tt2rSR0047TdcFINGxboAPOq+Wqd3qAf9EAgHcSF5//XV59913ZdCgQXq3M7JxG2h5y5ct19cusGYE4sINiCn6CGC8GgmZtYg9xi/UjykgiAnkW/nka+za+raBEhAMss4IQy0hv8o5tvcKwKzIQxkSkKEUrS3igg44YMAAeeCBB2TevHkycODABAlZnGE14giVNGPGDDn99NP1HaGKigpVv9HyitakQsCIxMYkxic+pnpAPPggrzAFWaWqM11egTPPS/HNjnTF/eWjc0LQHDhQo3KgdHA33O98qmuVj6kfIyB/LfGqoBEAuUABTZ8+XW655RYloeQ2MaW+9957VfVcdtllKtsZ12SUon1sJIR4JxOREo+Tr+TjbJGsvFevskZAMKyGiGoeuxvh2NbOWTmvjrB8wAjgmYJzE0FHw2fu3Lny1FNPqcIZPHiwPmbHG9KLFy/WhwunDDhFLvqHi3TtAFNsXEMSCjhGGaze4owqCwsR84NEZKSEMki21QOPf7JCQGaTdcAE2dSqIkgw9zkrz200EbDOibeh33rrLX3ZEF/DwCsVA04eIGcMPUMXqbFWYOQTTU9oVXoEQDo1Zy3eRjR2jLOWl76e+s8EugaU3DSMBdGYA7ovNe+GGAElrsGcDd+t1Tuv3nrBUvCY25BxQKzwwXs+eIVizZo1+hkxYoSMGzdO1wp0iu1Mr/XBQ8j2st80dNxg9NWWrR2IykFplM6Rkg+ayCoBocF0RqfLxzVM0UPgwP4DiRcMMY3eu3evGon1AXvahX2m/EIg0+M06wTkDkeyM7hRKgO7C3E/mgg4t0asBaRKyXFNVYZ50UcgG3EMlYCSQ1Cj9FT0JZ/icUQRQCd1d1T3fkRNplkRQiD1LSxCBtKUHEBAlWsO2EkTI4cACShyIclBgyhaczBo0TCZBBSNOOS2FVRAuR2/EK0nAYUIft40TQWUN6HMtiMkoGwjno/tUQHlY1Sz4hMJKCsw53kjVEB5HuDg3CMBBYdtfGqmAopPrDPsKQkow4DGsjoqoFiGPRNOk4AygWLc66ACinsP8O0/Ccg3dLwwgQAVUAIK7nhDgATkDS+WToUAFVAqVJjXAARIQA0AiUUOgwAV0GEA4ul0CJCA0iHD/IYjQAXUcKxYsg4CJKA6cPDAFwJUQL5g40XOf+lCEIjAESNABXTEEMa1AhJQXCOfSb+pgDKJZqzqIgHFKtwBOUsFFBCw+V8tCSj/Yxy8h1RAwWOcpy2QgPI0sNl065BfNMlm42wrpxEgAeV0+MI3Hv8HNP5zepBQuv+kPnwraUFUESABRTUyOWAXyAc/PFheXq6//4Wf5qEayoHARchEElCEgpFrpoCAKioqZPPmzdK4cWPd4jfC+MsYuRbJ8OwlAYWHfV60XFxcrD/JXFlZKS2at0hMx/LCOToROAIkoMAhzt8GoHZKS0ulVatW+suordu0pvrJ33AH4hkJKBBY41Mp1nxs2oUtExHwggAJyAtaLJsSAa75pISFmQ1AgATUAJBY5DAI8E3owwDE0+kQIAGlQ4b5DUeAb0I3HCuWrIMACagOHDzwhQAVkC/YeBH/Ow72gUwgQAWUCRRjWQcVUCzDnmGnqYAyDGh8qiMBxSfWwXlKBRQctnleMwkozwOcFfeogLICcz42QgLKx6hm2ycqoGwjnjftkYDyJpQhOkIFFCL4ud00CSi34xcN66mAohGHHLSCBJSDQYucyVRAkQtJrhhEAsqVSEXZTiqgKEcn0raRgCIdnhwxjgooRwIVPTNJQNGLSe5ZRAWUezGLiMUkoIgEIqfNoALK6fCFaTwJKEz086VtKqB8iWTW/SABZR3yPGyQCigPg5odl0hA2cE5v1uhAsrv+AboHQkoQHBjUzUVUGxCnWlHSUCZRjSO9VEBxTHqGfGZBJQRGGNeCRVQzDuAf/eL/V/KK+OCQKqfW0Ze4iMHGShVWeDEX86IS2/x5icJyBtesSyNXz8tKiqq87vvIJpGjRpJSUmJkguOUQbH7gTiwe+F4TfkmYhAMgKRIiCnDzvp4N002VgeZx8BkMqqVatkw4YNUlZWpmRiVuDcjh07pLy8XH8bfuPGjbJgwQI7rcQE4gEp9e7dmz/bnEAmN3ayoVoLdu/eHdqITyfXcyM8+W8l4oNOWF1dLTNnzpRnnnlGOnTokFAzbtVjSqeiokKVEtTR9u3blXiuueYaadu2rdaTjU6d/5EJz8NMxy/rBOQmnXT7Cq/KIefxCp6wYN8ZCNxmGwfo0ZqpFUjo2WeflbffflvJpLKyMjH1cscRHbSwsFBwvnnz5jJ58mTp1KmTVDgqqcDJT8SW8Yxgf0Z0avtYTaR0+OnYc46TySf5uPYST5usTsGso2JrH6wPWD4sd+978oSFA0MAZIJ1oNFjxsj69etlxYoV0rJlyzrTMXfj+/btkwrnc/XVVytZ7dy1UwoLHPJxSIwpdxAwgsENBft2bPsYq5bn16usKCAjFWzxMdLBGgKS5dvWrzO8LjgEEBtMq7DOM336dFm3bp0eI5buZB1y/Pjx0r9/f52uWZ67HPejjwDihrhD/WIf4xXbVITk15usKSDHj4QzcASLk2vXrpWdO3eqg9aR4TBTNBFAbEpLS6VHjx6ycuVKJSDrpLAYcd21a5f06tVLF56XLFkSTUdoVYMQQDybNm0q7du319iCiJCHD0ioEKrI2aJfoB/4SYErIBhnHzgAw9esWSNz5syRhQsXqhpyG+/e9+MQrwkWAcQQnRKfrVu3KgmhRcR179690qpVK11G+Hr712oI4xlsPIKsHeMWqV+/fjJkyBDp3LmzHmM6bq9mGAnhhJ9YB0pA5gDUjZHP8uXL5b777lOnRo0aJd27d9fObGXVQ/6JNAKIFe6Cc+fOlfvvv1+6du0qmzdvljPOOEMmTZp0yLtAkXaGxqVEAGTiPCFXpTt79mxZtGiRrun17NlTiQavViRIyLn5oHwkCQid1chn9erVMnXqVBnjLGZijQAOMOU2An/84x/lkUcekUGDBukTr9atW+e2Q7T+EASwXPLcczOcVzFmaYxxw4HiBQlhXRA3IyMgryQUOAOAgPCBE5h2Qc4Z+UAZwRGm3EMAMUVnGzt2rHz11VdyzjnnCMjH8nPPI1qcCgGMUQiFiy8e7zwB3aCqd9y4cUo+th6EfoB9Pymw0W/Egy2Mw9MTrPlg2gWHSD5+whWda9DpEFvcBSdMmCDHHHOMGuf1Dhgdj2hJKgQgEIyERo4cKUuXLpVNmzZpHkTF/v3ViTVeXI8+4SUFRkBmBIxHwtMu7GPNB4nKR2HI6T9GQo0bN85pP2h8/QjYWD322GNVPODrN1hWAQFVV9d9j6/+mg49GzgBmRKCweiweHrClD8IGAnlj0f0JB0CGLsgI7yYivEMQWHj27bprk2XHwgBmQyzLQw1JWR56Qxifu4hwGlX7sXMr8WINcayEZCRkN/6AiEgtzFuwmFHdSPDfSKQmwgY6WBsuz9+vAmcgGCUGenHQF5DBIhAtBDAeDYSOlLLgiUgjyviR+oMrycCRCB7CLiFhXvfiwXBEpAzX2QiAkQgzxBI8aQdyyt+RnuwBJTC0DwLBd0hAvFDwA/TpEEpWALif6+aBnZmE4H8QkCnYD5cCpiAMkiVPpzjJUSACGQHgWhOwaiAshN9tkIEQkaACijkALB5IhBnBKiA4hx9+k4EQkaACijkALB5IhBnBKiA4hx9+k4EQkaACijkALB5IhBnBKiA4hx9+k4EQkaACijkALB5IhBnBKiA4hx9+k4EQkaACijkALB5IhBnBKiA4hx9+k4EQkaACijkALB5IhBnBKiA4hx9+k4EQkaACijkALB5IhBnBKiA4hx9+k4EQkaACijkALB5IhBnBKiA4hx9+k4EQkaACijkALB5IhBnBKiA4hx9+k4EQkaACijkALB5IhBnBKiA4hx9+k4EQkaACijkALB5IhBnBKiAGhB9sHSmftO6Ac2xCBGIDQJ+FVBxsAhF63fBlKX5c9HBhpy1xxIBKqB6wg52Rvr6669l6dKlsn379npK8xQRIAJeEfCrgPLml1Hd0ytMs/CxZAS0ZMkS6dOnjyxevFhPVVdXWxHdWh02TbNtnULOgYJdS2rJ11hZy3dv7Ry3RCDfEPCrgAKegtUoj6DBxiB3T6+wbwnnCgtreLZp06aa3bhxY90WFRVZMSUVdx04YfVY/VbY8q2M+9jKuutKdd7q4pYI5AMC6PcY7QdHXsO8CpiAvJrTMKOTS2GAr127Vv72t7/JmjVrpFGjRnL88b3k29/uLyCbHTt2yOeff67nce2HH34oIB+A1r9/fyUo1IEp2sKFC+WLL76QyspK6dy5s57v2rVrgqBwzUcffSRQT1BTH3/8sSqq0tJSGT16tLRu3VrLYqoHpYU6W7VqpWW/+c1valtGUsl+8JgI5CoCesP1YXzABBSsArKBPG/ePBk1apSu7ZxwwgmybNkyheKHP/yh/PSnP9V8EA3S6aefLv/0T/+k+/hTUVEhJSUlsmrVKvn3f/93mTFjRuIcdrp16ybPPvusDBo0SPMxLcMxyO7ss8+WG264IVH+s88+UwJ67rnn5JJLLtH8b33rW7Jo0SLdf+KJJ2TChAkJRZa4kDtEIMcRwFj0o4Bydg3IyGfnzp3y4IMPKkFgoM+dO1fWrVsn06ZNk9/85jeqaDp36SwrVnwmIIb//d//VQLB8fLly5V81q9fr0QC8vmf//kfwfGmTZvk/fffFxAaSAvKCGn/gf3SsmVLefXVV/WaJ5980ql7hZ7v2KmTvPXWW0o+P/nJT2TlypXy+p9f1/N33nmnTJo0SV555ZUc72o0nwgcikDsFJAREJ5oQZFgwPfr1y+BzBVXXCHf+c53pEuXLqo4evQ4zpmK1Tz96t69u+DY0vPPPy8vvvii/OUvf5ERI0ZYtrRv314eeOABufDCC+XXv/61PPzww9KiRQudYqHQ008/LZdffnmi/NatW2XixIlyzTXXqD2YliEd1fYo+bd/+zfZsmWLXHXVVTJw4EDp0KFDYlqXqIA7RCBHEfCrgAKeggW3BgTGRWrbtq1gqjV16lRp1qyZjBw5Uo455hidCh199NFaBtMmLESXl5frMaZdlr766iuZM2dOghiQj/Udq//444/XKdsPfvAD+fnPf64EhPqQhg0bplusF2Hd6dNPP5XVq1cL1oywNoR81IPyTZo0kaOOOkq+/PJLVVckIIWOf/IEgdgpIDgM1sUi8/XXXy8gG2yxjnPiiSfqNGjs2LG6+GsxtqdhRi7IBym9/PLLct999ylJoE57OmZEA0JDsveH9u3bp0oJpIJk9W7btk2PMZXD2lNyOu64GtW1a9cuPeW2I7ksj4lALiEQOwWE4BgJQXFMmTJFzjvvPMFCMNaBbr75Zv289957OuVxB9M98EEe7dq1k927d6vysadj7jJ79+zVy6FykAB2cXFxgng00/mDxWwktD102FCprjqopHBNo5JGUlVZJS1btdRy7jY0g3+IQI4igL7sZ74T8BQs2KdgiBUc37Bhg06/oFTwOeuss+T888/XR+gzZ84UPInCesyB/TX2mLLB9VjTwRrPSy+9JFdeeaV07NhRCQaEAXLCNGrBwgUoqtM93Un6Y0TSyVmERsK0rkvnLkmlRLBgDjtAVKjfrjukIDOIQI4hgP4cu6dgiBHebsb7Oo899pg+Ure4YUqGNZo9e/Yk3oouKq558RAqCQlrPXg5EYvVeMqFJ1qYHoF48AFRgZh++ctfyj333KML2siDSnKTh+336NFDrrvuOpk8ebK8+eabjlQya0SfrF166aV6Du8G4RoEjYkI5AMC6M+xVEAgHzyJ+td//Vdd3MVTLKzRzJo1Sx+J4z0de/MZpNS7d28lgS+clw0xpbrxxhv1BUIQGBQQHs2DkKBUMH2766679HE7nl6BlKqqq5SYyveW1yEQkAmUzS233KIL0WeeeaaSFl4+xCP96dOny+zZs/VJG15MRDLiyocOSB/ijYBfBRTwFMwPJzYskKYg8GQJ7/v07dtXfvvb36paQQ2DBw/WAQ8iQAJAmCI9+uijcu+99+q7Q8OHD0+QCEgM0zEsRl900UV6DaZzOMajdbz7gzoKCwp1utf16K5KSFrQ+WP2gBDxDhLeksajd0uYEr7zzjsyZMgQy+KWCOQNAn4VUIGz+JrxeQAGKtJ+Z4pTWVWlg/OTTz7RaRIIwJ4yZQJ9tGWDH2sveFJVVlam6zXYImF26haIUEiYapU655s6T7KsDpTFOg0elWOqBaWCx/xI7jJVjk9IyVMx5Fk5TO+2fLlFdu102nHUFOqxp2Yox0QEcg0B9GksMeAbAHg9BYofSxgYZ3gog/GAWQLGY0NTziogc9DIB1uoIXwsgQyQ3IAgD8C1adPGiiUIDJPY5s2b6ydx0tkxUrE8gJ0umT0IRscOHUU6HCyZXM/BM9wjArmNAPp9w2nnoK/pR9LBMkewl3FxldIWIxgjHBRSQFIwcZ2yDmIGW518Vyvp6nEVOWQ3VV1+6jmkYmYQgYgigLGH0e6VhAImIK/mHBm6NvAbUku6sunyG1JncplM1pVcN4+JQJQQQF/3M9oD/jJqdhRQlAJBW4hAHBEwBeTV94AJyA8nenWB5YkAEQgbASqgsCPA9olAjBGgAopx8Ok6EQgbASqgsCPA9olAjBGgAopx8Ok6EQgbASqgsCPA9olAjBGgAopx8Ok6EQgbASqgsCPA9olAjBGgAopx8Ok6EQgbASqgsCPA9olAjBGgAopx8Ok6EQgbASqgsCPA9olAjBGgAopx8Ok6EQgbASqgsCPA9olAjBGgAopx8Ok6EQgbASqgsCPA9olAjBGgAopx8Ok6EQgbASqgsCPA9olAjBGgAopx8Ok6EQgbASqgsCPA9olAjBGgAopx8Ok6EQgbASqgsCPA9olAjBGgAopx8Ok6EQgbASqgsCPA9olAjBGgAopx8Ok6EQgbASqgsCPA9olAjBGgAopx8Ok6EQgbAb8KqDhYw/nb8MHiG37teuc7cDDO2hEL+JPc4UcmuxaYAvIa+WAJiB0xu70gy63t379fCgsLBaTjTuiMSMn57jLcz2EEDt5vEk5EUwG57owJS7mT8wiAYNDhQD47duyQbdu2SWVlpR43a9ZM2rdvrz5auZx3mA7URaDu/SYRa/BSilN1r006yroC4l0xKQI5dmikUl5eLu+88468+uqrsmnTJjHV06RJExkyZIiMGjVKOnbsqPmMeY4FuYHmquqpVb/u/QZersWCJSCnCdgH43C3RCfFnbK0tNSLjSwbEQSMfHbt2iUPPfSQLF++XCZOnCh9+vSRtm3byp49e2TlypUya9YsueGGG+T222+Xnj17koQiEr8jNQNjF33Apt2ZuLEUHqlRh7++QI0G6VRXV8v69ev1ErtjHv56logCAkY+VVVV8vjjj8vGjRvl7rvvluHDh+uUq6ioSJo3by4nnXSS3PSjm2T0uefKrbfequXQURnvKETRnw0WO4zdffv2SVlpmYoKUz229VN7IARkzGiGYbGyXbt20q1bN5k/f37CTuQz5RYCUD1vvvmm3HjjjUo8iKF1UHiC4+KiYhn/j5dI37595bXXXsstB2ltHQTcY/SDDz6Q1q1bS5u2bbQMlJCpoToXeTgIfApWWFAoVQeqBGsDw4YNU+neo0cPOfnkk5VF0XndHdiD7SyaRQRwM0FnfOONN+T888+XLl26aNzQAd0Jx0ZCY8aMkQcffFDOddRQhw4dVAHbzcl9DfejiQBiZfFdsGCBvPDCC/LdCy/UsYx8qF4TGbb16kngBOT4oIZCuvXq1UsuuOACueOOO2TKlClKQi1atFAnvBrO8tlHAJ1s2bJlcu2119bbuHXa7t27y86dOwXTNiR0WKbcQgBPOUE+06ZN0+l2j+7HqQONGjWS4uJiJ6aHvobhxcPACAidFemAs8U+Oh8MHjp0qDRu3FieeOIJmTlzpvTu3VuaNm2asNmuS2RwJzIIIDbr1q3TtR4YBeVaX7wQZ0y9X3rpJX0ihkXM+spHxtGYG2Jx3b17tyxdulSfco4ePVqn1CWlJVJSUiIgIIzpwsKDKgiweY1vYARkMYRBJtewCA0SOvXUUwXTMNxN0aHxsamYbe16bqODADoc3vPZvHmzxi9dZ7MO/OVXX6rxa9eulS1btujULDre0JJ0CFhcy8rK9Ann2LFj9aaD+ONhEvIPElDNOlC6ug6XnxUCMhKC0Uh4YalNmzZy2mmn6boAJDrWDfBB59UytVs94J9IIIAbyeuvvy7vvvuuDBo0SO92RjZuAy1v+bLl+toF1oxAXLgBMUUfAYxXIyGzFrHH+IX6MQUEMYF8K598jV1b3zZQAoJB1hlhqCXkVznH9l4BmBV5KEMCMpSitUVc0AEHDBggDzzwgMybN08GDhyYICGLM6xGHKGSZsyYIaeffrq+I1RRUaHqN1pe0ZpUCBiR2JjE+MTHVA+IBx/kFaYgq1R1pssrcOZ5Kb7Zka64v3x0TgiaAwdqVA6UDu6G+51Pda3yMfVjBOSvJV4VNAIgFyig6dOnyy233KIklNwmptT33nuvqp7LLrtMZTvjmoxStI+NhBDvZCJS4nHylXycLZKV9+pV1ggIhtUQUc1jdyMc29o5K+fVEZYPGAE8U3BuIuho+MydO1eeeuopVTiDBw/Wx+x4Q3rx4sX6cOGUAafIRf9wka4dYIqNa0hCAccog9VbnFFlYSFifpCIjJRQBsm2euDxT1YIyGyyDpggm1pVBAnmPmfluY0mAtY58Tb0W2+9pS8b4msYeKViwMkD5IyhZ+giNdYKjHyi6QmtSo8ASKfmrMXbiMaOcdby0tdT/5lA14CSm4axIBpzQPel5t0QI6DENZiz4bu1eufVWy9YCh5zGzIOiBU+eM8Hr1CsWbNGPyNGjJBx48bpWoFOsZ3ptT54CNle9puGjhuMvtqytQNROSiN0jlS8kETWSUgNJjO6HT5uIYpeggc2H8g8YIhptF79+5VI7E+YE+7sM+UXwhkepxmnYDc4Uh2BjdKZWB3Ie5HEwHn1oi1gFQpOa6pyjAv+ghkI46hElByCGqUnoq+5FM8jigC6KTujurej6jJNCtCCKS+hUXIQJqSAwiocs0BO2li5BAgAUUuJDloEEVrDgYtGiaTgKIRh9y2ggoot+MXovUkoBDBz5umqYDyJpTZdoQElG3E87E9KqB8jGpWfCIBZQXmPG+ECijPAxyceySg4LCNT81UQPGJdYY9JQFlGNBYVkcFFMuwZ8JpElAmUIx7HVRAce8Bvv0nAfmGjhcmEKACSkDBHW8IkIC84cXSqRCgAkqFCvMagAAJqAEgschhEKACOgxAPJ0OARJQOmSY33AEqIAajhVL1kGABFQHDh74QoAKyBdsvMj5L10IAhE4YgSogI4YwrhWQAKKa+Qz6TcVUCbRjFVdJKBYhTsgZ6mAAgI2/6slAeV/jIP3kAooeIzztAUSUJ4GNptuHfKLJtlsnG3lNAIkoJwOX/jG4/+Axn9ODxJK95/Uh28lLYgqAiSgqEYmB+wC+eCHB8vLy/X3v/DTPFRDORC4CJlIAopQMHLNFBBQRUWFbN68WRo3bqxb/EYYfxkj1yIZnr0koPCwz4uWi4uL9SeZKysrpUXzFonpWF44RycCR4AEFDjE+dsA1E5paam0atVKfxm1dZvWVD/5G+5APCMBBQJrfCrFmo9Nu7BlIgJeECABeUGLZVMiwDWflLAwswEIkIAaABKLHAYBvgl9GIB4Oh0CJKB0yDC/4QjwTeiGY8WSdRAgAdWBgwe+EKAC8gUbL+J/x8E+kAkEqIAygWIs66ACimXYM+w0FVCGAY1PdSSg+MQ6OE+pgILDNs9rJgHleYCz4h4VUFZgzsdGSED5GNVs+0QFlG3E86Y9ElDehDJER6iAQgQ/t5smAeV2/KJhPRVQNOKQg1aQgHIwaJEzmQoociHJFYNIQLkSqSjbSQUU5ehE2jYSUKTDkyPGUQHlSKCiZyYJKHoxyT2LqIByL2YRsZgEFJFA5LQZVEA5Hb4wjScBhYl+vrRNBZQvkcy6HySgrEOehw1SAeVhULPjEgkoOzjndytUQPkd3wC9IwEFCG5sqqYCik2oM+0oCSjTiMaxPiqgOEY9Iz6TgDICY8wroQKKeQfw736x/0t5ZVwQSPVzy8hLfOQgA6UqC5z4yxlx6S3e/CQBecMrlqXx66dFRUV1fvcdRNOoUSMpKSlRcsExyuDYnUA8+L0w/IY8ExFIRiBSBOT0YScdvJsmG8vj7CMAUlm1apVs2LBBysrKlEzMCpzbsWOHlJeX62/Db9y4URYsWGCnlZhAPCCl3r1782ebE8jkxk42VGvB7t27Qxvx6eR6boQn/61EfNAJq6urZebMmfLMM89Ihw4dEmrGrXpM6VRUVKhSgjravn27Es8111wjbdu21Xqy0anzPzLheZjp+GWdgNykk25f4VU55DxewRMW7DsDgdts4wA9WjO1Agk9++yz8vbbbyuZVFZWJqZe7jiigxYWFgrON2/eXCZPniydOnWSCkclFTj5idgynhHsz4hObR+riZQOPx17znEy+SQf117iaZPVKZh1VGztg/UBy4fl7n1PnrBwYAiATLAONHrMGFm/fr2sWLFCWrZsWWc65m583759UuF8rr76aiWrnbt2SmGBQz4OiTHlDgJGMLihYN+ObR9j1fL8epUVBWSkgi0+RjpYQ0CyfNv6dYbXBYcAYoNpFdZ5pk+fLuvWrdNjxNKdrEOOHz9e+vfvr9M1y3OX4370EUDcEHeoX+xjvGKbipD8epM1BeT4kXAGjmBxcu3atbJz50510DoyHGaKJgKITWlpqfTo0UNWrlypBGSdFBYjrrt27ZJevXrpwvOSJUui6QitahACiGfTpk2lffv2GlsQEfLwAQkVQhU5W/QL9AM/KXAFBOPsAwdg+Jo1a2TOnDmycOFCVUNu4937fhziNcEigBiiU+KzdetWJSG0iLju3btXWrVqpcsIX2//Wg1hPIONR5C1Y9wi9evXT4YMGSKdO3fWY0zH7dUMIyGc8BPrQAnIHIC6MfJZvny53HffferUqFGjpHv37tqZrax6yD+RRgCxwl1w7ty5cv/990vXrl1l8+bNcsYZZ8ikSZMOeRco0s7QuJQIgEycJ+SqdGfPni2LFi3SNb2ePXsq0eDVigQJOTcflI8kAaGzGvmsXr1apk6dKmOcxUysEcABptxG4I9//KM88sgjMmjQIH3i1bp169x2iNYfggCWS557bobzKsYsjTFuOFC8ICGsC+JmZATklYQCZwAQED5wAtMuyDkjHygjOMKUewggpuhsY8eOla+++krOOeccAflYfu55RItTIYAxCqFw8cXjnSegG1T1jhs3TsnH1oPQD7DvJwU2+o14sIVxeHqCNR9Mu+AQycdPuKJzDTodYou74IQJE+SYY45R47zeAaPjES1JhQAEgpHQyJEjZenSpbJp0ybNg6jYv786scaL69EnvKTACMiMgPFIeNqFfaz5IFH5KAw5/cdIqHHjxjntB42vHwEbq8cee6yKB3z9BssqIKDq6rrv8dVf06FnAycgU0IwGB0WT0+Y8gcBI6H88YiepEMAYxdkhBdTMZ4hKGx82zbdtenyAyEgk2G2haGmhCwvnUHMzz0EOO3KvZj5tRixxlg2AjIS8ltfIATkNsZNOOyobmS4TwRyEwEjHYxt98ePN4ETEIwyI/0YyGuIABGIFgIYz0ZCR2pZsATkcUX8SJ3h9USACGQPAbewcO97sSBYAnLmi0xEgAjkGQIpnrRjecXPaA+WgFIYmmehoDtEIH4I+GGaNCgFS0D871XTwM5sIpBfCOgUzIdLARNQBqnSh3O8hAgQgewgEM0pGBVQdqLPVohAyAhQAYUcADZPBOKMABVQnKNP34lAyAhQAYUcADZPBOKMABVQnKNP34lAyAhQAYUcADZPBOKMABVQnKNP34lAyAhQAYUcADZPBOKMABVQnKNP34lAyAhQAYUcADZPBOKMABVQnKNP34lAyAhQAYUcADZPBOKMABVQnKNP34lAyAhQAYUcADZPBOKMABVQnKNP34lAyAhQAYUcADZPBOKMABVQnKNP34lAyAhQAYUcADZPBOKMABVQnKNP34lAyAhQAYUcADZPBOKMABVQnKNP34lAyAhQAdUTAPyMLD7ZSAhEttrKhj9sgwg0BAG/Cqi4IZX7LxONXyYsLAz414dcAGkg+IuwLkS4GwcETAF5/SGugEemV3MyFyoAgrRv3z75y1/+InPnzpXq6urMNZCiJrT57rvvyqxZs6SioiJFCWYRgfxEwK8CCpiAwldAe/fuldtuu02ee+65BAEZOWW6K6De559/Xr7zne8o8aH+oNrKtO2sjwgcCQLo535Ge8BTsOwqIAXBAQJsjHWYoqIixbRr167SqlUrzXeD7C5vRJE8XbN81JlcPtV0q0WLFnLaaac1qK3k62Fzcp7b3sOdd5flPhHIJgLab300GDAB+eFE716AewvwzyEJfJCMfLBfWVl5yMIwyMRd3q5DeTuHfXd+Q8qDJMrLy3FpIll9VpdtUQDnkJCXTH56wvlj/qU7b+W4JQJhIYB+jJ7sVXIETEBezfEHH8gH6fPPP5eFCxfKl19+KVA9/fr1k9atW0tJSUlioFsLGPD/93//JwsWLJBNmzZJs2bN5Pjjj5f+/ftLaWlpgoSWLVsmO3bskFNPPVXWrVsnH338kXzxf18IlM5JJ50kffr0SZCUm1isHWyRv3HjRvnwww9lzerVUuAsiv/d3/2dfPvb35ajjjpKi27fvl0WL14s7dq1k169eiXIyMhr9+7dej38cbfpbof7RCAsBNDH/Yz2gAkoSwrIYd///u//lgsuuOAQ/J955pkEQbjVxjvvvCNDhw7V8ieffLISEQ6mTp0qN9xwgzRu3FhJaObMmfLggw/K73//e7n22muVhFq0aO6Q0k699tFHH5UJEyZIo0aNDiE5FEBglixZIhMnTlSC0Ytq/5x55pnyu9/9TokP6ubxxx+Xp59+WjZs2CAdO3asUx/I64wzzpDp06dL3759EwTpro/7RCAsBPwqoIAXof1wYsMhNELBEy6Qz7nnnivz589XRbNmzRp55OFH5NJLL5U//OEPCVWD2v/2t78p+Xzve9+Tjz/+WF577TVZtWqVPPzww/KTn/xEHnnkEZ2ygTxARFBAf//3fy/XX3+9rFy5Uj777HOt48orr5SrrrpKQFLJyWxbvny52gUVBtLbuGmjqqE33nhD5r8/X2AD6mzevLnainqgyizBhqqqKvmv//ovOfrobjJixAg7xS0RiAwCsVNAGOBwGlMTPOFq16693H333aoOLCpXff8qXT+5+uqrIUV0XQhPxab9fpqccsopctddd0n79u21eNu2beX73/++7NmzR6ZMmSIjR47UqRDaAQHdeeed8i//8i9KZLgAU6Wf/vSnSlxod/jw4brQjfJYB4KiwWP/e+65R9avX68kd+KJJ5pp0qFDB3l99usyaNAgJb5f/vKXOs0Dwfznf/6nEk2TJk20/GeffSYPPPCAKrFOnTppHnxnIgJRQQD93s8aUM4qIDiMtGXLFpk2bZrcdNONiakJCMDe+QGRIB1w8oqLi2Xz5s3y8P97WIYMGaLkAcU0b948ee+993SNBYoHCYMeCXUhjRo1SskH9RrJYB1n/Pjxej3Wk5BQHu8eYREcxPPYY4/pawAgH7vO6hw4cKCSHh7dYx2qZcuWiWOoNEt//etfdffss8/Wrflu57klAmEjEDsFZIBDsSCBDJAwOKE+bJBisfi8884TKB8kKCYkkM5vfvMb3Xf/sXq+/vprzQbhtGnTRjp37qzHqNutPnr27Kn5UElIaBfXoBwWlpGwWG0J+ZZQD5QYpnxoD6po8ODBenr27Nn6OB8L6r/+9a/lxz/+sRx33HF2KbdEIFIIoN/7UUABL0IHP02wx+3Jj74tOlg/wRrLCSecoFlQQUjjxo2TF154QdWKkQJAxHlsy8rKEuW3bt16yKN1Pen8MUKzepEPYrG6cGxEhP1Eqo2WER3WiJDw9O63v/2tTvewvrTaeWr26aefypgxY3ShG+rJ7E3UxR0iEDICfhXQwdtxIA5glAWTTIVA4UC1vP/++6pyMDgxSEEASGvXrpWPPvpISstqHq1jmoNH2XgsjidNuPboo4/WT7du3QRrQSiDrZtEsJhsCfloH+2gXSRbS0I+nojhHPKgkPBVEHw1A+dwrdrmcDNU08svvyxnnXVW4nE86sLTMSQsPGM9aOzYsfpKAfJIPkCBKWoIaL/2YVTABBScAjICAongsflDDz0kL774og58DFJ88N6OTbPwlgJeSMQ05/bbb9fFYSxeQyFZAiHgKRjUBogLbdhaEhaJV6xYoXnWNqZJt912my5aH3vssUosUGRQMzZ1++EPf6gkMmPGDG0f1+KDKeGTTz6p31GD0sFTMCUmx5hvfOMb8qMf/Uifuv3Hf/yH4Glb06ZNE+fNXm6JQFQQ0H7tw5iAp2DBKSD4igELx/EeDqYpl112mSxdulTXTjC1AflAyeCrEZii2QBH+S+++EIuueQSLY8nUZhKQY3gPRy8O2RPm3ANCAFPvbDFek2XLl30nZ6bb75Zv/d100036dQNpIPPzp07E21dfvnlusB8xRVX6PtAePcIZf785z/ru0X333+/TgctdmgPJPbd735XfvWrX8k555wjAwYMsNPcEoFIIoB+W7uq4Mm+gAkoOAUEL0E+cBzf8/rZz36mJIEX+37xi18oCFAnICW84AdVYutFmGLdcsstSjL33nuv3HHHHVoeC8B/+tOf9J0fqxv143E4yuBFQDyqtwR1g0fzWKBGOaguTAlBVNYWroVCw3QMKg3tIeHlRyigi8dfnHhTG21i6oYtCA8J7zeBDFE/8pmIQBQRQN/00zsLnDt/xmUKBgvSfudOX+lMcWDcJ598oo+k8eawDc5MAWmDE1s8zt61a5ejfI5y1npaaROVVZUOOAWqUpBh5bGPx/iYesEmqCVMhZCsDN6CvvXWW/WxPM6jLB7lg1hAKrZobeSAKR2uxTqQux7sb9u2TT8oi3UoEGdyGWsX5DRp0iRVdlhHsny9gH+IQAgIQLlfd911gtdB8LUl3NSxNIAHNhgHGEO4CdtYaIiJOa2AzEE4bAMUa0LupGRQXEMGlu8uD6VhagPnUR7JDSLeDQL4SFA4+Fiydu3Y/TQMee62QDr4uJP7etsHKUJxQaXZo3e3Pe7ruU8EooAA+qcfBRQwAWVcXKXF2gaoEQgKKigOMKlSqvJ2TXJ5zG5Tla+vfncdqa61tuycuzy+WoJXB/CqAO4oRkzuMtwnAlFCAH0Uoz31aEtvacAE5NWc9IY29EyqAV3ftYcrv//Aftm4YWNCGaGuw12Trr36rjOSwTtHeLMbT77wrXgmIpALCKBv+xntARNQ9hRQUEG68LsXytlnna3vBqGN+kgkEzbgvwLB98cwLbRH70G3mQm7WUe8EaACCij+eDMZn6CTkQxIB/+PEZJ76hd0+6yfCBwJAlRAR4JePdeC2ZGMIOopmrFTYbSZMeNZUSwRoAIKKOzZJB5zIYw2rW1uiYAfBPwqoIC/ipH7a0B+gsFriEDcEDAF5NXvgAnIz7q4VxdYnggQgbARoAIKOwJsnwjEGAEqoBgHn64TgbARoAIKOwJsnwjEGAEqoBgHn64TgbARoAIKOwJsnwjEGAEqoBgHn64TgbARoAIKOwJsnwjEGAEqoBgHn64TgbARoAIKOwJsnwjEGAEqoBgHn64TgbARoAIKOwJsnwjEGAEqoBgHn64TgbARoAIKOwJsnwjEGAEqoBgHn64TgbARoAIKOwJsnwjEGAEqoBgHn64TgbARoAIKOwJsnwjEGAEqoBgHn64TgbARoAIKOwJsnwjEGAEqoBgHn64TgbARoAIKOwJsnwjEGAEqoBgHn64TgbARoAIKOwJsnwjEGIFoKqAC/i5YjPskXc9XBFL83mg0FVDt76rnaxzoFxGIJQIpdEXOKCAwJRMRIAK5j4Cqntrx7N734lnAP80sAvtgXGFhoYAlKysrvdjHskSACEQEAYxdjGGMZb+Ek+xK4ATk0I8aftKcJAAABCBJREFUXVpaKtXV1bJ+/Xq1AY4wEQEiEH0EbKxi7O7bt0/KSssSBGREhK2fFAgBmTFm3P79+6Vdu3bSrVs3mT9/fsJO5DMRASIQXQTcY/SDDz6Q1q1bS5u2bdRgKCFTQ349KPZ7YUOvKywolKoDVdKkSRMZNmyYPPTQQ9KjRw85+eSTlUXBrsawDa2T5YgAEQgeAQgIEAzSggUL5IUXXpDvXnihjmXkFxUVHaKEvFpVsHv37kDmQkYsBxyVU+VMvSDd8HnzzTflxRdflClTpigJtWjRwqvNLE8EiECWENixY4eSz7Rp02T48OFyyoBTpLSsVEmocePGUlpaIsXFjZSMjKxsBtQQEwMjIDQOEoKEw9pPVVWVLkBXVFTI+++/L6+88opOy3r37i1NmzZN2OrF+MRF3CECRCBjCGDcYhw64kSWLl0qmzZtkhEjRkjfvn0dwinVTw35lEpJSYmSj1sNeTEkcAJykxAUEFbSQUhfffWVLFu2TNatWyd79+5VskqoJi5Qe4khyxKBjCJgIqCsrEw6duwoxx13nDRv3lyJBgSEfBBPo0aN9OMmH7u2oQYFTkAwBCrIlBAIaB9IyFFElgd1ZGVAQki21QP+IQJEIGsIgESSiQTTKxAOiMfIp7i4ODH1SnVNQwwOdBEaRoFIsLX5IYzCcZXjkL1XYAyKMkY8tm2IEyxDBIhA5hDA+ESycYvxiQ8ICFsQT4J8nLJW3o8FgRKQOVFjGEiodq+WkIodZ6pr1REUEEiHxOMnjLyGCGQWASMViIJkIgIJIb8Q5FM7qK28VysCJyAzqIZUa5QQjMXngONIkYt4jHxsa9dySwSIQHYRsDGKVgsLMV4PEpGRkpGObf1YmBUCchuIfRCM5jnLPQcKHeXjWG6kY1s/zvAaIkAEMoUASKemLoxV+yAnef9IWswKAZmBMNzIJ7EvRXr6EOLRxWh8kcw5jX2gwS1xYD8IcBxgKNaONR2VNcPPmAhj1p2Sj93nGrqfVQKCUemMTpffUEdYjggQgeARyPQ4zToBuSFKdkZFj07I3KW4TwSIQBgIJI/PIGwIlYCSHapReHVlXnIZHhMBIpA/CATybfj8gYeeEAEiECQCJKAg0WXdRIAI1IsACaheeHiSCBCBIBEgAQWJLusmAkSgXgRIQPXCw5NEgAgEiQAJKEh0WTcRIAL1IkACqhceniQCRCBIBEhAQaLLuokAEagXARJQvfDwJBEgAkEiQAIKEl3WTQSIQL0IkIDqhYcniQARCBIBElCQ6LJuIkAE6kWABFQvPDxJBIhAkAiQgIJEl3UTASJQLwIkoHrh4UkiQASCRIAEFCS6rJsIEIF6ESAB1QsPTxIBIhAkAiSgINFl3USACNSLAAmoXnh4kggQgSARIAEFiS7rJgJEoF4E/j8UdE6VCUBPTQAAAABJRU5ErkJggg=="
    }
   },
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "##### Pipeline for Training and deploying with Watson Machine Learning.\n",
    "![Screen%20Shot%202019-02-15%20at%204.59.42%20PM.png](attachment:Screen%20Shot%202019-02-15%20at%204.59.42%20PM.png)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Model for this pipeline\n",
    "We will be training a TensorFlow model that can classify the handwritten digits. This TensorFlow model is a simple convolutional neural network (CNN) with 2 convolutional layers and 1 fully connected layers using the [MNIST](http://yann.lecun.com/exdb/mnist/) dataset. We will be training for 200,000 iterations for the purpose of this demo.\n",
    "\n",
    "<br><br>"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Prior to run this pipeline, need to update the cluster role binding from pipeline-runner to cluster-admin. This is one time modification to the Kubeflow Pipelines service. It allows the secrets to be created inside the container."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "scrolled": false
   },
   "outputs": [],
   "source": [
    "# install the SDK\n",
    "KFP_PACKAGE = 'https://storage.googleapis.com/ml-pipeline/release/0.1.21/kfp.tar.gz'\n",
    "!pip install $KFP_PACKAGE --upgrade"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "# install ai_pipeline_params package\n",
    "!pip install ai_pipeline_params --upgrade --no-cache-dir"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Prereq: users provide the credentials to access the Watson Machine Learning service as well as the S3 Cloud Storage in a <key = value> pair text file. Then run the **create-secret.ipynb** notebook to create the credentials as secret."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "# define secret name\n",
    "secret_name = 'ai-pipeline-creds'"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "* Create an experiment"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "# set up\n",
    "EXPERIMENT_NAME = 'Watson Machine Learning Experiment'"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "# create experiment\n",
    "import kfp\n",
    "\n",
    "client = kfp.Client()\n",
    "exp = client.create_experiment(name=EXPERIMENT_NAME)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "* Construct the pipeline"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "# create pipelines\n",
    "import kfp\n",
    "from kfp import components\n",
    "from kfp import dsl\n",
    "import ai_pipeline_params as params\n",
    "\n",
    "train_op = components.load_component_from_url('https://raw.githubusercontent.com/kubeflow/pipelines/master/components/ibm-components/watson/train/component.yaml')\n",
    "store_op = components.load_component_from_url('https://raw.githubusercontent.com/kubeflow/pipelines/master/components/ibm-components/watson/store/component.yaml')\n",
    "deploy_op = components.load_component_from_url('https://raw.githubusercontent.com/kubeflow/pipelines/master/components/ibm-components/watson/deploy/component.yaml')\n",
    "\n",
    "# create pipelines\n",
    "\n",
    "@dsl.pipeline(\n",
    "    name='KFP on WML training',\n",
    "    description='Kubeflow pipelines running on WML performing tensorflow image recognition.'\n",
    ")\n",
    "def kfp_wml_pipeline(\n",
    "    train_code='tf-model.zip',\n",
    "    execution_command='\\'python3 convolutional_network.py --trainImagesFile ${DATA_DIR}/train-images-idx3-ubyte.gz --trainLabelsFile ${DATA_DIR}/train-labels-idx1-ubyte.gz --testImagesFile ${DATA_DIR}/t10k-images-idx3-ubyte.gz --testLabelsFile ${DATA_DIR}/t10k-labels-idx1-ubyte.gz --learningRate 0.001 --trainingIters 20000\\'',\n",
    "    framework= 'tensorflow',\n",
    "    framework_version = '1.13',\n",
    "    runtime = 'python',\n",
    "    runtime_version = '3.6',\n",
    "    run_definition = 'wml-tensorflow-definition',\n",
    "    run_name = 'wml-tensorflow-run',\n",
    "    model_name='wml-tensorflow-mnist',\n",
    "    scoring_payload='tf-mnist-test-payload.json'\n",
    "):\n",
    "    # op1 - this operation trains the model with the model codes and data saved in the cloud object store\n",
    "    wml_train = train_op(\n",
    "                   train_code=train_code,\n",
    "                   execution_command=execution_command,\n",
    "                   framework=framework,\n",
    "                   framework_version=framework_version,\n",
    "                   runtime=runtime,\n",
    "                   runtime_version=runtime_version,\n",
    "                   run_definition=run_definition,\n",
    "                   run_name=run_name\n",
    "                   ).apply(params.use_ai_pipeline_params(secret_name))\n",
    "\n",
    "    # op2 - this operation stores the model trained above\n",
    "    wml_store = store_op(\n",
    "                   wml_train.outputs['run_uid'],\n",
    "                   model_name\n",
    "                  ).apply(params.use_ai_pipeline_params(secret_name))\n",
    "\n",
    "    # op3 - this operation deploys the model to a web service and run scoring with the payload in the cloud object store\n",
    "    wml_deploy = deploy_op(\n",
    "                  wml_store.output,\n",
    "                  model_name,\n",
    "                  scoring_payload\n",
    "                 ).apply(params.use_ai_pipeline_params(secret_name))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "* Compile the pipeline"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "# compile the pipeline\n",
    "import kfp.compiler as compiler\n",
    "pipeline_filename = kfp_wml_pipeline.__name__ + '.tar.gz'\n",
    "compiler.Compiler().compile(kfp_wml_pipeline, pipeline_filename)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "* Run the pipeline"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "# run the pipeline\n",
    "run_name = kfp_wml_pipeline.__name__ + ' run'\n",
    "run_result = client.run_pipeline(exp.id, run_name, pipeline_filename)\n",
    "import IPython\n",
    "html = ('<p id=\"link\"> </p> <script> document.getElementById(\"link\").innerHTML = \"Actual Run link <a href=//\" + location.host + \"%s/#/runs/details/%s target=_blank >here</a>\"; </script>'\n",
    "          % (client._get_url_prefix(), run_result.id))\n",
    "IPython.display.HTML(html)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Show page with iframe\n",
    "import IPython\n",
    "html = ('<p id=\"web\"> </p> <script> document.getElementById(\"web\").innerHTML = \"<iframe width=1100 height=700 src=http://\" + location.host + \"%s/#/runs/details/%s target=_blank >here</iframe>\"; </script>'\n",
    "          % (client._get_url_prefix(), run_result.id))\n",
    "IPython.display.HTML(html)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python on Kubernetes",
   "language": "python",
   "name": "python_kubernetes"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.6.8"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}