ResultsExp2.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "The autoreload extension is already loaded. To reload it, use:\n",
      "  %reload_ext autoreload\n"
     ]
    }
   ],
   "source": [
    "import pandas as pd\n",
    "import matplotlib\n",
    "matplotlib.use('Agg')\n",
    "import matplotlib.pyplot as plt\n",
    "from matplotlib.backends.backend_pdf import PdfPages\n",
    "import seaborn as sns\n",
    "%matplotlib inline\n",
    "%load_ext autoreload\n",
    "%autoreload 2\n",
    "from scripts.parsing_utils import *\n",
    "from utils.utils import *\n",
    "from scripts.parse_results import *\n",
    "import pickle"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {},
   "outputs": [],
   "source": [
    "RES_DIR1 = \"/home/pari/learned-cardinalities/all_results/inl_fixed_scan_ops/nested_loop_index7/default/final_results2/\"\n",
    "RES_DIR2 = \"/home/pari/learned-cardinalities/all_results/inl_fixed_scan_ops/nested_loop_index7/default/large_bitmap/\"\n",
    "\n",
    "#RES_DIR2 = \"/home/pari/learned-cardinalities/all_results/inl_fixed_scan_ops/nested_loop_index7/test_diff/subq_imp2\"\n",
    "\n",
    "SAVE_DIR =\"./figures/results2/\"\n",
    "make_dir(SAVE_DIR)\n",
    "FN_TMP = \"{ERROR_TYPE}-{SAMPLES_TYPE}-{PART}.png\"\n",
    "\n",
    "COLORS = {}\n",
    "COLORS[\"Flow Loss\"] = \"blue\"\n",
    "COLORS[\"Q-Error\"] = \"green\""
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {},
   "outputs": [],
   "source": [
    "def arg_filters(df):\n",
    "    #tmp = df2[df2[\"template\"] == \"all_job\"]\n",
    "    #tmp = tmp[tmp[\"partition\"] == 2]\n",
    "    #tmp = tmp[tmp[\"alg\"] == \"flow_loss2\"]\n",
    "    tmp = df[df[\"flow_features\"] == 1]\n",
    "    tmp = tmp[tmp[\"hidden_layer_size\"] == 512]\n",
    "    tmp = tmp[tmp[\"feat_rel_pg_ests_onehot\"] == 1]\n",
    "    return tmp"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {},
   "outputs": [],
   "source": [
    "#df1 = get_all_objects(RES_DIR, \"query_qerr.pkl\")\n",
    "#df3 = get_all_objects(RES_DIR, \"flow_err.pkl\")\n",
    "\n",
    "RES_DIRS = []\n",
    "RES_DIRS.append(RES_DIR1)\n",
    "#RES_DIRS.append(RES_DIR2)\n",
    "\n",
    "df1s = []\n",
    "df2s = []\n",
    "for res_dir in RES_DIRS:\n",
    "    qerr = get_all_objects(res_dir, \"all_qerr.pkl\")\n",
    "    jerr = get_all_objects(res_dir, \"cm1_jerr.pkl\")\n",
    "    df1s.append(qerr)\n",
    "    df2s.append(jerr)\n",
    "    \n",
    "df1 = pd.concat(df1s, ignore_index=True)\n",
    "df2 = pd.concat(df2s, ignore_index=True)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "27212 6807966\n"
     ]
    }
   ],
   "source": [
    "#df1 = arg_filters(df1)\n",
    "#df2 = arg_filters(df2)\n",
    "#len(df2) == len(df1)\n",
    "\n",
    "# FIXME: this should not happen\n",
    "#df1 = df1[~df1.qerr.isna()]\n",
    "#df2 = df2[~df2.loss.isna()]\n",
    "\n",
    "print(len(df2), len(df1))\n",
    "ALG_KEY = {}\n",
    "ALG_KEY[\"flow_loss2\"] = \"Flow Loss\"\n",
    "ALG_KEY[\"mse\"] = \"Q-Error\"\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {},
   "outputs": [],
   "source": [
    "#print(set(df[\"samples_type\"]))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "{'mse', 'flow_loss2'}\n",
      "{'mse0', 'flow_loss20'}\n"
     ]
    }
   ],
   "source": [
    "set(df2[\"samples_type\"])\n",
    "print(set(df2[\"alg\"]))\n",
    "df2[\"alg_name\"] = df2.apply(lambda x: x[\"alg\"]+ str(x[\"sample_bitmap\"]), axis=1)\n",
    "print(set(df2[\"alg_name\"]))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "metadata": {},
   "outputs": [],
   "source": [
    "def compute_PDF(arr, num_bins=600):\n",
    "    \"\"\"\n",
    "    usage: x, y = compute_PDF(arr):\n",
    "           plt.plot(x, y)\n",
    "    \"\"\"\n",
    "    values, base = np.histogram(arr, bins=num_bins)\n",
    "    cumulative = np.cumsum(values)\n",
    "    return base[:-1], cumulative / float(cumulative[-1])  # this is for CDF"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "metadata": {},
   "outputs": [],
   "source": [
    "#TITLE_FMT = \"{SAMPLES}-{PARTITION}\"\n",
    "TITLE_KEYS = {}\n",
    "TITLE_KEYS[\"job\"] = \"Join Order Benchmark\"\n",
    "TITLE_KEYS[\"test\"] = \"Test Set\"\n",
    "TITLE_KEYS[\"train\"] = \"Train Set\"\n",
    "\n",
    "def plot_cdf_jerr(samples_type, df, part):\n",
    "    #plt.gcf()\n",
    "    fn = FN_TMP.format(ERROR_TYPE = \"postgres\",\n",
    "                       SAMPLES_TYPE = st,\n",
    "                       PART = part)\n",
    "    fn = SAVE_DIR + fn\n",
    "    #plt.figure(figsize=(5,4))\n",
    "    plt.figure(figsize=(5,4))\n",
    "    plt.style.use('ggplot')   \n",
    "\n",
    "    df = df[df.samples_type == samples_type]\n",
    "    algs = set(df[\"alg\"])\n",
    "    \n",
    "    plt.title(TITLE_KEYS[samples_type] + \"-\" + str(part))\n",
    "\n",
    "    for alg in algs:\n",
    "        if alg == \"true\" or alg == \"postgres\":\n",
    "            continue\n",
    "        tmp_df = df[df.alg == alg]\n",
    "        \n",
    "        alg_label = ALG_KEY[alg]\n",
    "\n",
    "        losses = tmp_df[\"loss\"].values\n",
    "        #x, y = compute_PDF(losses, num_bins=len(losses))\n",
    "        x, y = compute_PDF(losses, num_bins=len(losses))\n",
    "        plt.plot(x, y, label = alg_label, linestyle=\"--\", linewidth=3, color = COLORS[alg_label])\n",
    "        plt.xscale(\"symlog\")\n",
    "        plt.legend(frameon=False)\n",
    "    plt.savefig(fn)\n",
    "    plt.show()\n",
    "\n",
    "def plot_cdf_qerr(samples_type, df,part):\n",
    "    fn = FN_TMP.format(ERROR_TYPE = \"qerror\",\n",
    "                       SAMPLES_TYPE = st,\n",
    "                       PART = part)\n",
    "    fn = SAVE_DIR + fn\n",
    "\n",
    "    plt.figure(figsize=(5,4))\n",
    "    plt.title(TITLE_KEYS[samples_type] + \"-\" + str(part))\n",
    "    \n",
    "    algs = set(df[\"alg\"])\n",
    "    for alg in algs:\n",
    "        if alg == \"true\" or alg == \"postgres\":\n",
    "            continue\n",
    "        #alg = ALG_KEYS[alg]\n",
    "        tmp_df = df[df.alg == alg]\n",
    "        \n",
    "        alg_label = ALG_KEY[alg]\n",
    "\n",
    "        #losses = tmp_df[\"qerr\"].values\n",
    "        losses = tmp_df[\"loss\"].values\n",
    "        x, y = compute_PDF(losses, num_bins=600)\n",
    "        plt.plot(x, y, label = alg_label, linestyle=\"--\", linewidth=3, color = COLORS[alg_label])\n",
    "        plt.xscale(\"symlog\")\n",
    "        #plt.ylim([None, 0.999])\n",
    "        #plt.xlim([1.0,1000.0])\n",
    "        #plt.yscale(\"log\")\n",
    "        plt.legend(frameon=False)\n",
    "    plt.savefig(fn)\n",
    "    plt.show()\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAUUAAAENCAYAAACYf3dcAAAABHNCSVQICAgIfAhkiAAAAAlwSFlzAAALEgAACxIB0t1+/AAAADh0RVh0U29mdHdhcmUAbWF0cGxvdGxpYiB2ZXJzaW9uMy4yLjEsIGh0dHA6Ly9tYXRwbG90bGliLm9yZy+j8jraAAAgAElEQVR4nO3deXxU1d348c+dyTaBEJYQlrBvsi+JLApaFLRAkVWPKEpVfqVarQo8D5Zqq49apba12j4+okXb2rodbWJRQUDUAiISCfsisgkhQAghELJn5v7+uEMyk3UCM5NM8n2/XvNK7r3nzPfcZPLNvffcc65hmiZCCCEstvpugBBCNCSSFIUQwoMkRSGE8CBJUQghPEhSFEIID5IUhRDCgyRFUW8MwxhrGIZpGEan+m6LEBdJUhS1cieuml5HLvGtNwIdgAw/tPFuwzC2GIZx3jCMXMMw9hqG8Zc6vsdjddkXwzBGGoax0TCMQsMwThiG8axhGPY6N140KGH13QAREjp4fD8C+Lf76zH3OqdnYcMwIkzTLK7tTd1lTl5u4wzDuAtYCiwEVrtX9wOmXu571xCzM7AG+BfwE6A38DpgAL8IVFwRBKZpyktePr+AMYAJdPNYZwIPAm8B54D33Ot/A+wF8rES6FIg1qPeWHfdThWWbwDWuevtAX5YS5s+AN73oe1JWEnzAnAaSAa6urfd5Y7t+Xqihvd6BkgHbB7r7gfygGb1/XuS16W/5PRZ+MvjwFdAIvCoe10BMA/oj5V0xgJ/8uG9fo+VdIYA3wDvGobRsobyJ4ArDcPoU10BwzD6A/9xt/FK4HqsI9w1hmFEAe8Cv8VKdB3cr9/XEHM0sNo0TZfHuk+AaGBYjXsnGjRJisJfPjBN88+maR40TXM/gGmaT5umud40zSOmaa4FFgOzDMOo7XP3P6ZpfmKa5nfAIiAWGFlTeaxk9q1hGEcMw3jXMIx5hmFEe5RZBHxkmubjpmnuM01zJ3AH0AmYYJpmAdYRpNM0zZPu14UaYnag8qn/SY9tIkRJUhT+srniCsMwZhiGsc4wjAzDMC4AbwIRQPta3mvbxW9M0zyJdUTXrrrC7gQ2BuuI9FmsU9jngN2GYcS7iw0HphuGceHiCzgDRGFdD6ySYRizPesYhjG7hnabFb6KECQdLcJf8jwXDMMYCbyHlaT+GzgLjAL+jpUYa1JVJ02t/8BN09yLdQ3zFcMwngL2A/dhHUnagH8AS6qoeqaGt10OfO2xfMr99QSVk/vF5cvuPBL1R5KiCJQxQJZpmo9dXGEYxs1BjH8Eq6Pm4pHiN8Bg4KBpmtUdyRUDXrfUmKaZC+RWUfZL4E7DMGwe1xUnuGNuvbymi/okp88iUL4F2hqGMdcwjB6GYcwBfhaIQIZhvGwYxuOGYVxjGEZXwzCSsI5IW2D1TIPVcdMP+KdhGCMMw+huGMZ1hmG8aBhGD3eZw0B7wzCuMgwjrsI1yYpexrrW+RfDMAYYhjEFeAr4s2maeTXUEw2cJEUREKZpfoR1S84zwE5gFtZpdCCswbrd5m2sU+YVWJ0dk0zTXONuz17gaqA5sArrVp+/AA4gx/0+H2Cd8n+MdcvOouoCmqZ5DLgRK9FuAV51vx6tro4IDUb1ZxJCCNH0yJGiEEJ4kKQohBAeJCkKIYQHSYpCCOFBkqIQQnhoqDdvS5e4ECJQjJo2NtSkSEaG7/OOxsXFkZWVdUlxLqeuvwSzDcGK1djiBDOW7FPgYnXs2LHW+nL6LIQQHiQpCiGEB0mKQgjhQZKiEEJ4qLWjRSn1OjAZyNRaD6xiuwG8CEzCmjbpLq11mnvbBPc2O7BMa13VXHZCCNFg+HKk+DeseeKqMxFr5uLeWM/jeBlAKWUHXnJv7w/cppTqfzmNFUI0bU6Xk0BPYlPrkaLWep1SqlsNRaYCb2itTWCTUqqlUqoD0A04oLU+BKCUesddds9lt1oIETSFhRAV5b1u+/ZwTpywU1QEZ8/aKCgwME0D06TS67rrirj+eu/6//xnNBkZdlwuq8y5ovMUO0vIvpBP5tkics5BbraD5n2/Jnzc02TkHaewtBATkx137KCNo03A9tcf9ykmUP78X7AeIJRQzfpqHz6klJqHdaSJ1pq4uDifGxAWFlan8pdaNz09nYceeoi9e/fidDqZMGECzz33HJGRkV7l3njjDRYvXux1T9Qbb7xBv379/N7+ugpWrMYWJ5ix6nOfXC7IyoKvvzZYsdLFRytMBlz7LZN/vpYSVwklzhISOyTy2rIJJCf71iXxp30Lsae9gtN04nQ5eWniS7z//lxSUz3rx1RZN/tQD2j/EnQ6UF6yZQxxMdX/fC735+ePpFjV3eFmDeurpLW+OEkngFmXGz2DcfO2aZrMnDmTOXPm8Morr+B0Olm0aBHz58/nySef9Cqbm5vL5MmT+c1vfuO13jOO0+nEbreXteHUqVNlyzUpLS0lLOzSf22N7WbdhnJTcCjEySnK4VTeKU7ln+LaTtcC4HDE8emn5/n759tYvek0pQfGYuZ6P4ww89AGPv90YdnyPQPuweUai/U019oVlhRBSflk5DnncygtLaX2R/UArghI+QfcNxjCirEZNk6fOU1EUfV1L/fmbX8kxXSgs8dyJyADa4+rWh+SNmzYQGRkJLfeeisAdrudJ554gpEjR/LII4/QrFmzWt9j48aNPP/887Rr147du3fzzDPP8Pzzz9OlSxfS0tL45JNPWLx4MTt27MBut/P4448zevRo3n33XdauXUtRURH5+fm89957gd5dEaIKSgvIKcrh2+xv2X1mN1+f/JrTBac5kXeC0wWnAejcvDPLx21m4sS2nDxpB+KA8dW/aWEsuGxgsx5FU+IqYfDgEgoKCggLg8hIk2/OfcbRgr2ACYbp/bXjFq+3c5kuZs/OZ9y4QgwDDAOSD7xPTtFZwsPCaduiGQnxkcS3CaNbQjh9ukUwMCGVVpGtsNtqP3C4XP5IisuBB9zXDEcC57TWJ5RSp4HeSqnuwHGs6ehv90M8AP6w5Q88n/a8T2Vn953Nc9c857Vu0fpFvLnvzbLlBYkLWJi0sGLVMvv372fQoEFe62JiYujUqROHDx9m4EDvjvnly5ezefNmr2WAbdu28dlnn9GlSxc2btzItm3b+Otf/0pMTAxLly4FYO3atRw4cIDbbruN9evXA7BlyxY+/fRTWrVq5dM+i6blyPkjTFs+rSzx1eRs0VkiI013QqyCrRhaHYZOm2g14BsSf3CEzq3mEG4LJ9wWTlK7JCaMyeOnPy0/+juYE8u54n7YDTt2w47NsGE37DjCHMRGTic+Lp6csznYDBvhtnBs/fK9Qj5UY19ucPlyS87bwFggTimVDjwOhANorZdiPQ9jEnAA65acu93bSpVSD2A9D8MOvK613h2AfQgK0zQxjMpXBKrrCZsyZUql02eAoUOH0qVLF6/l7t27k5WVRWpqKnfffTcAvXr1olOnThw6dAiAa6+9VhJiE1bqKuW9/e+x+uhq0jLTWDpuKVd1uKpse7vodlUnxJJI2DULeqwlvNUpEponEO+IJ7pZKYZhYpoGXbuWMmRYAT365DF4SCHDhxcRFWEnzPYDIuw3+NS+ni171rg9JjKGorCiOu1zffGl9/m2WrabwP3VbFuBlTRDXp8+fVixwntXcnNzycrK4ptvvmH+/PkA/OMf/6jxfaKjo6tdrulWg4r1RNNwIOcA/7Ppf1iXvo5Ss7RsfU5hjlc5R5iD9tHtOV1wmjZRbWjjaEPnM3fwzev3kX2iFT+44Qz/eD3f6/Tziy9OM3RoSy5c8Lz+FoFP1/oasQY7S05tFiYtLDvdvZQL089d8xzPXfOcz3WvueYann32Wd577z1uueUWnE4nTz75JHfddVfZ63KNHDmSlJQUxowZw8GDBzl+/Dg9e/Zk586dl/3eIrTsOL2Dv+/5Ox8c/IBCZ2Gl7edLzldat2rGKmIjYynMi+CRR2L597/L/5H+Z00bPv8Mxo8vP1rr1auUqCi4cCEw+xCqQjYpBpthGCxbtoxHH32UF154gezsbG666SYeeuihKstXvKb4zDPP1Brjxz/+Mb/4xS8YN24cdrudP/7xj5Vu9xGN26n8U9z59p18duSzSttaRbbizn53MrP3TLq26Fppe0RJWxb/qgX/+lc0xcXll3piY1385CcXSEwsCWjbG4uG+ohTs6HPp5iamsr999/PsmXLGDx48CXFvtw2NORYjS1OMGLtP7uf2Stnk5Hn/dkfHDeYZ8c8y5C4IVVe1wY4etTOLbe0IT3d+zinfXsnKSlZdOnirLJeU/s9uW/JCc1JZhu64cOHex0JCnG5EponEGaz/iQNDKb1nMZdA+4iKT6p2mQIsGVLOHPntub06fLrhVFRLhYvzuXWW/OJiWmQBz4NliRFIRqIZuHN+MO1f+Cna3/Kn8b+ies6X+dTvRdeiPFKiPfcc4GHH75AmzauQDW1UZOkKEQ9KSwtJCrMe1Dx1R2vZv/P9lOU6/vtKz/72QU++yyKli1dvPpqNqNHF/u7qU2KzKcoRD3Yc2YPo98dzbr0dZW2xURWPQ64OqNGFTN37gVWrjwtCdEPJCkKEWQXii8w79N5nMw/yZxVc1h1ZJXPdd9918G+fd4neIYBTz55vtrOFFE3khSFCCLTNFn85WIOnz8MQLgtnM4xnWupZVmzJpLHHotl5sw4tm0LD2QzmzRJinXQuXNnbrjhhrLXsWPH2LhxI3PmzPFbjGPHjnF9xcnnRKOhv9MkH0guW14yZgn929Q897JpwmuvNWPu3Nbk59vIybHx2GOxgW5qkyUdLXUQFRXFmjVrvNYdO3asmtJCePvu7Hc8+uWjZcuqj2Jm75k11ikshMWLW6J1+eiUjh1Lef75nBpqicshR4p+dPbsWe655x7Gjx/P5MmT2bPHmmR83LhxnDt3DtM0GTBgQNnUXz//+c9Zt67yhfaqrF+/nhtvvJFx48axYMECioqs3slnnnmGsWPHMn78+LJ5HT/88EOuv/56xo8fz4wZMwKwp6KuCkoLuO+z+ygoLQCgV8te/ObqyhOGeMrLM7jlljivhDh0aDEff5xFnz6lNdQUlyNkjxT/8IcYnn/es5eu+skjZ8/O47nnznmtW7QoljffvDgHYkcWLMhl4cLcGmMWFhZyww3WrCFdunThtddeq9CmPzBw4EBef/11NmzYwEMPPcSaNWu48sorSU1NpVOnTnTt2pXNmzdzyy23kJaWxpIltT/Lq7CwkPnz5/Puu+/Ss2dPHnzwQd544w1uvvlmVq5cybp16zAMg3PnrH184YUXePPNN+nQoUPZOlG//pj2R/Zm7wUgyh7F0nFLiQ6veZKPF19sTlpa+eQMt9ySz5IlOZUeDSD8S44U6+Di6fOaNWsqJUSAzZs3M3OmdTo0ZswYzp49y/nz5xk5ciRff/01mzZtYs6cOezdu5cTJ07QqlUrnyanPXjwIF26dKFnT2t6pltuuYWvv/6amJgYIiMj+a//+i9WrFiBw+EA4Morr2T+/Pm8+eabOJ3SI1nf9mbv5ZUdr5Qt/3rUr+nXuupHU1z01lvR/N//NS9bXrjwPH/8oyTEYJCk6EdVjSM3DKMsKW7evJmrrrqKNm3a8PHHHzNixIhLfl+wnkXx8ccfM2nSJD755BNmz54NwG9/+1sWLVpERkYGN954I9nZ2Ze+U+KyuEwXi9YvKpv2a1T7UczpV3PH3KefGvz3f7fENK2hfcOGFfPwwxeoYaSf8KOQPX1euLD8dPeSpg577hzPPXfOrwPVR40aRXJyMvPnz2fjxo20bt2amJgYYmJiyM7OpqSkhK5duzJixAiWLl1a5SS0VenVqxfHjh3j8OHDdO/enX/961+MGjWKvLw8CgoKGDduHImJiYwZMwaAI0eOkJiYSGJiImvWrCEjI4PWrVv7ZR9F3bhMF5O6TWLPmT04TSdLxiypcRwzwGuvlQ/ZGzSomDfeOINNDl+CJmSTYkO0YMECFixYwPjx44mKiuKFF14o2zZs2DBcLmss6ogRI3j22WcZPnx4le9z8OBBkpKSypafeOIJnn/+eX7605/idDoZMmQId955Jzk5Odxzzz0UFRVhmiaPP/44AE8//TSHDx/GNE3GjBnDgAEDArjXoiZhtjDuG3IfP+r+I9Iy0+jdqnetdd56q5S33rrAX/7SjJdfPkvr1jKhQzDJ1GFBnNKoIbShsU0VJT+70IjVUPbJl6nD5KBcCCE8SFIUIgB2nN5BxgXfz3bef99BZqb8OTYE8lsQws9M02ThuoVc9c5VPPTFQ5zIO1Fj+e3bw3n44ZZMmtSW7dtlTHN986mjRSk1AXgR61Gly7TWSypsbwW8DvQECoF7tNa73NuOALmAEyjVWl/pt9YL0QCtP76ePdnWaKaPD3/M46Mer7asacJTT7XANA1OnLDz+9/HMG5csFoqqlLrkaJSyg68BEwE+gO3KaUqjmD/JbBNaz0YmIOVQD1dp7UeKglRNAUv73i57PtZfWbROqr626HWro3kq6+sh5PZ7SZPPCEjkOqbL6fPI4ADWutDWuti4B1gaoUy/YG1AFrrfUA3pVQ7v7ZUiBCwM2sn645b49ltho15g+ZVW7a0FH7zmxZly3fckU/PnjICqb75cvqcAHhOBZMOjKxQZjswA9iglBoBdAU6AacAE1itlDKBV7TWr1YVRCk1D5gHoLUmLi7O950IC6tTeX/V9ZdgtiFYsRpbHF9j/XHtH8u+n9F3Bok9Eqst+/77Nvbvt/4Emzc3eeqpcOLi4hrcPoVSHH/E8iUpVnVPT8WbG5cALyqltgE7ga3AxWk8RmutM5RS8cAapdQ+rXWlqWHcyfJiwjTrck+T3KfY8GI1tji+xPr6xNesOmTNom1gcP+A+2ss//LLbbj4Jzh37gXs9lyyshrWPoVanNpiue9TrJEvSTEd8JwauBPgda+B1vo8cDeAUsoADrtfaK0z3F8zlVIpWKfjvs2XJUSIME2TJanl/Y8ze8+kT6s+1ZY/csTOhg3WtUSbzWT27LyAt1H4xpekmAr0Vkp1B44Ds4DbPQsopVoC+e5rjv8PWKe1Pq+UagbYtNa57u9vBJ706x4I0QB8kf4Fm09ZzwEPt4WzMHFhjeXffrt82rCxY4tISJDHkTYUtXa0aK1LgQeAVcBea5XerZS6Vyl1r7tYP2C3UmofVi/1Q+717bCuM24HNgMfa60/8fdOCFGfXKbL6yhxdt/ZdGnRpdryJSV4TRx7xx35AW2fqBuf7lPUWq8AVlRYt9Tj+6+ASiPdtdaHgCGX2UYhGrTswmzCbNafUpQ9igeHPVhj+b17wzl3zjoeadfOybhxhQFvo/CdzJIjxGWKc8Tx0dSPWPX9Kk7knaBddM13ow0eXEJa2kmSk6Ox203C5K+wQZFfhxB+YBgGE7pN8Ll8y5Ym99wjnSsNkYx9FkIID5IUhbhEqSdTKXYW13czhJ9JUhTiEnx//ntuXXErNyTfwLp03267XbcugkWLYtm+PZyGObezAEmKQlySJzY9QZGziAM5B3jum+eqfbiYp7feasabbzZj0qS2vPJK7U9xFPVDkqIQdfT5sc9Z/f3qsuWnrn6q1odRnT9vsHp1+fNJx44tClj7xOWRpChEHRQ7i/n1V78uW57VZxbD4ofVWu/jjx0UFVmJc8CAEvr2La2lhqgvkhSFqINlu5Zx6NwhAFpEtGDxiMU+1fvXvxxl38+cKSNYGjJJikL4KCM3gxe2lj+2dmHSQuIctU9RlZFhK5tI1mYzmTatIGBtFJdPkqIQPvrl578kr8S64fqKVlfw4/4/9qnexo2RZd+PGlVMu3Yy+UNDJklRCB9sPrmZt3e/Xbb81NVPEW7z7SFTmzZFlH0/Zox0sDR0khSFqIXT5eSxjY+VLU/uPpnRHUf7XH/TJu8jRdGwSVIUohZO08nEbhNxhDmIskfx61G/rr2S26lTNg4ftqYYiIw0GTJEkmJDJxNCCFGLCHsE8xPnM2/kPP6z/z8kNE/wue7mzeWnzomJxURF1VBYNAiSFIXw4DJdfHf2O9Iy04gOj2Zqz/IHV3aN7cqk7pPq9H4/+lEha9dmsmlTBG3aSAdLKJCkKJq07MJstmZuJS0zjS2ZW9iWuY3cklwAEuMTvZLipbDZoG/fUrlZO4RIUhRNSnZhNh8e+tBKgqe2cPj84WrL7sraRZGziEh7ZLVlROMjSVE0Wpn5mcQ54rAZ5f2JZwvP8ssvf1ljvbaOtiTFJ5EYn0iJs0SSYhMjSVE0CsXOYnaf2V12Gpx2Ko1jF47xxc1f0LtV+eODesT2oGVkS3KKcgDryXsD4waSGJ9Ylgg7Ne9U6wQPvti5M5zYWBedOzvxw9uJIPEpKSqlJgAvAnZgmdZ6SYXtrYDXgZ5AIXCP1nqXL3WFuBTHLxwnLTONvdv3suHIBnadsU51K0rLTPNKioZhcN/g+4iwR5AYn8jANgOJCgtMl/Cjj8ayZUsE7ds7ee21bIYOLQlIHOFftSZFpZQdeAm4AUgHUpVSy7XWezyK/RLYprWerpTq6y4/zse6QtTJPavvYdX3q2otF2WPIqsgq9L6B4Y+EIhmeSkogB07rBEvJ0/a6dzZGfCYwj98OVIcARxwP64UpdQ7wFTAM7H1B54F0FrvU0p1U0q1A3r4UFeIMqZpcjT3aFlHSK+WvbhrwF1eZbq16FZl3W4tupEYn1j26t+mv89D8fxtx44ISkqsc+aePUvkdpwQ4ktSTACOeSynAyMrlNkOzMB68P0IoCvQyce6ognLK8lj2+ltZUkwLTONM4VnyraPbD+yUlJMapdEs33NGNp2KGO6jaFv874kxif6NGNNsHjetD1ihIxiCSW+JMWqLhFXnHt9CfCiUmobsBPYCpT6WBcApdQ8YB6A1pq4ON8/4GFhYXUq76+6/hLMNgQrVk1xvsv+jhc2v8Dm45vZdXoXLrP6o6gdWTuIbRVLuL38iO/2VrdzR9Id2G12wsLCKC0Nzj2AdfnZbd9e/qd13XWRQfs811VD+Dw0tFi+JMV0oLPHcicgw7OA1vo8cDeAUsoADrtf0bXV9XiPV4FX3YtmVlbla0HViYuLoy7l/VXXX4LZhmDFiouL4+Dxg2w/vZ1rEq7x6s3NzM5k2dZl1daNjYhlWPywstPgrDNZ1Z4GN8SfncsFGze2L1vu1+8MWVm+X1NsiPsUKnFqi9WxY8da6/uSFFOB3kqp7sBxYBZwu2cBpVRLIF9rXQz8P2Cd1vq8UqrWuqJxcLqc7M/ZT1pmGmmn0th2Zhv7zuwDYIPaQPfY7mVle7fqTUx4DLkludgMG1e0usLrlpieLXt63VsYavbvD+PcOav9cXFOuneXTpZQUmtS1FqXKqUeAFZh3VbzutZ6t1LqXvf2pUA/4A2llBOrE2VuTXUDsysimLILs72uA247vY0LJReqLJuWmeaVFG2GjWfGPEO8I56hbYfSPKJ5sJodFKmp5dcThw8vlnsUQ4xP9ylqrVcAKyqsW+rx/VdA74r1qqsrQt+cT+aw9fTWGsvYDTsD2gyo8tR3Rq8ZgWpavfPsZBk+XDpZQo2MaBGVZOZneh0FXpNwDQ8nPuxVJjE+sVJSjHfEk9QuiaT4JK7rcx1dw7viCHPQ1HgeKUrPc+iRpNjEFTmLyofHuZNg+oV0rzI2w1YpKY5oP4K002ll1wGT4pNIaJ5Q1qHSEDqw6kNJCYwfX0hqagRHjoQxcKCMYgk1khSbqNRTqTy56Ul2Ze2i2FXz0cy209twupzYbfaydZN7TGZyj8mBbmbICQ+Hp58+D0BhobUsQoskxUasoLSAnVk72ZW1i7sH3O11W4wjzEFaZlqV9aLsUQxpO8RrdIhnQhS+kVm2Q5MkxUbCNE2+z/3e6zR4z5k9lJrWjc0/7PZDr2n0+7bqiyPMQUFpQdnwuIunwv3a9Ku34XFC1DdJiiHsqxNfkXoy1ZoqKzON7MLsast+c+obr6QYZgvj7Ylv0yO2B20cbYLRXCFCgiTFEOAyXRQ5iyr15D7+1ePsPlPzbZ+9W/YmMT6Rjs0r38k/vP1wv7azqdu4MYIXX4xhxIhirruukMRE6WQJRZIUG6Ccohy2ZW4rOwLcmrmVuQPnsjBpoVe5pPgkr6QYGxHrdR1wWPwwYiNjg938Jmvjxkg2bLBe2dk2EhPP1XeTxCWQpFjPnC4nOzN3svbbtWVJ8EDOgUrlquoUGddlHC7TRWI763pgj9geIT08LtRVHMkiQpMkxXq04vAKHv7Pw+SV5NVa9mjuUUzT9OpBHt9lPOO7jA9kE4WPSkogLa28c2r48MqzgIvQIEkxgEpcJew9s5ctmVs4mHOQp0c/7bW9Y/OOVSbEMCOMAW0GWKfB7qPALjFd/PLcEBEYe/aEk59vHaUnJJSSkCCTyoYqSYp+dDLvpNeDk3Zk7aDQWVi2/YGhD9C+WfmUUv1b9yfSHkkbRxuGth1adkvMoLhBTXJ4XCiTU+fGQ5LiZTBNk9d2v0bqyVTSMtPIyKtyqsgyWzO3MrH7xLLlCHsEm2/bTN/OfZvkkLjGRCaBaDwkKfrANE2OXzhOpD2SttFty9YbhsFb+97i27PfVlu3c/POJLazeoMHxg2stL0hTaEvLo1pyiQQjYkkxSoUlBaw/fR2r9EhmQWZLLpyEQ8Ne8irbGJ8YllSjLJHWafB7ZLKbolpF92uPnZBBNHRo3YyM61hkC1auLjiiuA8HkEERpNPiqZpcvjc4fJrge7hcU6z8mzJVd0WM6PXDAbGDSQpPom+rfvK8LgmaNu28t/50KHF2GWYeEhr8knxpW9eYuGnC2st1zy8OdFh0ZXWX93xaq7ueHUgmiZCxLZt5afOw4bJKJZQ16iTost0cTDnYNkR4Mm8k7wx4Q2vMkkdkqqs26dln7LT4MT4RHq37C0zxYgq/eIX55k6tYBt28JJSpKkGOoaVVI8W3iWbae3ld0Ss/X0Vs4Xn/cqc6bgjNcECEPbDaWtoy2D4gaVzRQzpO0QGR4nfBYZCUOHljB0qCTExiDkk6LT5eQnH/2EL49+ycFzB2stvyVzCzd2vbFs2RHuYOvsrXJjtBACaARJ0VZ9W6wAABZISURBVG6zszF9Y7UJsU1Um7LT4KT4JIa2HVqpjCREIcRFPiVFpdQE4EWsx5Qu01ovqbA9Fvgn0MX9nr/XWv/Vve0IkAs4gVKt9ZV+a73bqIRRHDh7gDAjjIFxA8uuAybFJ9E5prMkPREwu3aFERfnon17GdbXWNSaFJVSduAl4AYgHUhVSi3XWu/xKHY/sEdrfZNSqi3wrVLqTa31xbtYr9NaB2zIxoMjHuTm7jczMG6gDI8TQfXww63Yuzec9u2d/POfZ+jXT+5RDHW+HCmOAA5orQ8BKKXeAaZiPfT+IhOIUUoZQHMgGwjap2NIuyEk2BNqLyiEH+XnG+zfb/0JnTplo1Onyve2itDjS1JMAI55LKcDIyuU+V9gOZABxAC3aq0vnk+YwGqllAm8orV+taogSql5wDwArTVxcb4PfwsLC6tTeX/V9ZdgtiFYsRpbnKpiffmlgdNpXZrp29eke3f/PNZBPg/1G8uXpFjVBTmzwvIPgW3A9UBPYI1Sar3W+jwwWmudoZSKd6/fp7VeV/EN3cnyYsI06zJBwuU8Y7ghPJ84mG0IVqzGFqeqWF980Qywbt0aOLCQrKycgMQJpKbwe/LUsWPlx3JU5Ms0zelAZ4/lTlhHhJ7uBpK11qbW+gBwGOgLoLXOcH/NBFKwTseFCHnbt3sP7xONgy9HiqlAb6VUd+A4MAu4vUKZo8A4YL1Sqh1wBXBIKdUMsGmtc93f3wg86bfWC1GPZHhf41TrkaLWuhR4AFgF7LVW6d1KqXuVUve6iz0FXK2U2gmsBR5x9za3AzYopbYDm4GPtdafBGJHhAim7Gwb339vHVNERJj06ydJsbHw6T5FrfUKYEWFdUs9vs/AOgqsWO8QMOQy2yhEg+N56jxgQAkRETUUFiFFHv0mxCXwnC5syBA5SmxMJCkKcQm2bi0/NJROlsZFkqIQl6B371L69i3BMEzpZGlkQn5CCCHqw69+dZ5f/Qry8gwcjoq37YpQJklRiMvQrJkkxMZGTp+FEMKDJEUhhPAgp89C1IFpwoMPtqRnz1IGDCjh+uuL5Ol9jYwkRSHqID3dTnKy9VTHli1d7Np1sp5bJPxNTp+FqIM9e8pv2u7XrwSZ1L3xkaQoRB3s3l1+ctW/v9yf2BhJUhSiDjyPFAcMkKTYGElSFKIOPJNi//7yPJbGSJKiED46f56y6cLCwkx695YjxcZIkqIQPtq1q7xXpVevUqKi6rExImAkKQrho+3by5OidLI0XpIUhfDRjh2SFJsCSYpC+GjnTs+kKJ0sjZWMaBHCR7/9rZN163LZsyecgQPlSLGxkqQohI9Gjza54oq8+m6GCDA5fRZCCA8+HSkqpSYALwJ2YJnWekmF7bHAP4Eu7vf8vdb6r77UFUKIhqTWI0WllB14CZgI9AduU0r1r1DsfmCP1noIMBb4g1Iqwse6QjR4hYX13QIRLL6cPo8ADmitD2mti4F3gKkVyphAjFLKAJoD2UCpj3WFaPCuuSaeK64IZ+7cVmRny9Q4jZkvp88JwDGP5XRgZIUy/wssBzKAGOBWrbVLKeVLXQCUUvOAeQBaa+Li4nzaAYCwsLA6lfdXXX8JZhuCFasxxTlzBjIyrD+VEyei6NGjDWEB7KKUz0P9xvLlV1vVv8WKT+v5IbANuB7oCaxRSq33sS4AWutXgVcvlsnKyvKhaZa4uDjqUt5fdf0lmG0IVqzGFGfDhgjA+iO74ooScnICG08+D4GL1bFjx1rr+3L6nA509ljuhHVE6OluIFlrbWqtDwCHgb4+1hWiQfOeGUfuT2zsfDlSTAV6K6W6A8eBWcDtFcocBcYB65VS7YArgENAjg91hWjQZLqwpqXWI0WtdSnwALAK2Gut0ruVUvcqpe51F3sKuFoptRNYCzyitc6qrm4gdkSIQNm9WyaWbUoM02yQD/M2MzJ8P8uWa4oNL1ZjiVNcDH36dKCkxLo8vmfPCWJjA/s3I5+HwMVyX1Os8fYBGdEiRA0OHAgrS4hdu5oBT4ii/klSFKIGntcTBw2ShNgUSFIUogae1xMHD5ak2BRIUhSiBllZ5X8iQ4a46rElIlgkKQpRgz//OYe9e0+QnJzFtdfKkWJTIElRiFq0aGEycmQxrVvXd0tEMEhSFEIID5IUhRDCgzyOQIhqrF0bSVSUSf/+JbRqJdcTmwpJikJU46mnWvDdd9YtOStXnub66+u5QSIo5PRZiCoUFMDBg9Yxg2GY9OolE0E0FZIUhajCt9+G43JZw/u6d3cSHS2nz02FJEUhqiBzKDZdkhSFqIJMF9Z0SVIUogp79pT3QcqRYtMiSVGICkwT9u6V0+emSpKiEBUcO2YnN9f602jZ0kWHDjIRRFMiSVGICjyvJ/bvX4Ihj3luUiQpClGB9Dw3bZIUhajgiitK+NGPCujevZSBAyUpNjU+DfNTSk0AXgTswDKt9ZIK2/8bmO3xnv2AtlrrbKXUESAXcAKlWusr/dR2IS6L0wmbNkWQkuLgxz/OZ9AgKwFOnlzI5MmF9dw6UV9qTYpKKTvwEnAD1sPtU5VSy7XWey6W0Vr/Dvidu/xNwHytdbbH21ynta7fR+YJgdWzvHt3GMnJ0fz73w5OnrQDEB1tliVF0bT5cqQ4AjigtT4EoJR6B5gK7Kmm/G3A2/5pnhD+8f33dlJSHKSkODhwILzS9uXLHTz++Hns9nponGhQfEmKCcAxj+V0YGRVBZVS0cAE4AGP1SawWillAq9orV+9xLYKUSc5OYY7EUazZUtElWVat3YyZUoh06fnY5Mr7ALfkmJVNyRUNzr+JuDLCqfOo7XWGUqpeGCNUmqf1npdxYpKqXnAPACtNXFxcT40zRIWFlan8v6q6y/BbEOwYjWEOFlZ8NhjlZNhs2YmU6a4mDXLxbhxJuHh4UDsZcXyJ/k81G8sX5JiOtDZY7kTkFFN2VlUOHXWWme4v2YqpVKwTscrJUX3EeTFo0gzK8v3S5BxcXHUpby/6vpLMNsQrFjBjHPiRBb/+U8kgwaV0K6dy2MbDBoUx86dEYSFmfzgB0XMmFHAjTcWls16c+5c3WI1pp9dMGM1lH3q2LFjrfV9SYqpQG+lVHfgOFbiu71iIaVULPAD4A6Pdc0Am9Y61/39jcCTPsQUokamCd98E8HKlXbee68d2dl2fvnL89x//wWvcj/72QXOnrVx002FtG4tI1NE7WpNilrrUqXUA8AqrFtyXtda71ZK3evevtRddDqwWmud51G9HZCilLoY6y2t9Sf+3AHRtHz7bRgpKQ4++MDBsWPeH9+UFEelpDhlitxaI+rGp/sUtdYrgBUV1i2tsPw34G8V1h0ChlxWC0WTd/y4jeXLHSQnR3uNNvHUoYOTsWOLKCmB8KqLCOETeUaLaNB27gxn4sQ4TLNyf19srIuZM00mTTrLyJHF0nss/EKSomgwCgrAZoPIyPJ1AwZYnScXb7KOijIZP76QGTMKGDu2kISEOLKyiuupxaIxkqQo6pXTCV9+GUFKSjQrV0bx1FPnuOWWgrLtNhvMnJnPzp3hTJtWwKRJhcTEyPNSROBIUhRBZ5qwY0c4yckOli93kJlZPowkJcXhlRQBFi/Olem7RNBIUhRBc/iwvWyEyaFDVX/00tPtFBV5n0JLQhTBJElRBNyhQ3YefLAVW7dWPdQuLs7J1KkFTJ9ewNChMqmrqF+SFEXAtW/v4ttvvT9qzZq5mDjR6jAZPbqIMPkkigZCPorCL4qL4YsvIklJieaOO/KYOrV8W3S0yYQJhXz4oYPrritk+vQCbrihCIdDOkxEwyNJUVwylws2b7Ymaf3oIwc5OdaNgpGRpldSBHjkkVyefPIcrVpJIhQNmyRFUWd795YPtTt+vPJHaOXKKAoLS73WderkDFbzhLgskhSFT7KzDd56qxkpKQ727at6HF1CQinTpxcwbVoBUVEtuXChymJCNGiSFIVPiooMliyJqTTcrmVLFzfdZPUcDx8uQ+1E6JOkKLwUFBisXh3J4MEldO9efsrboYOLq68u5ssvI4mKcnHjjUVMn57P2LFFRFR9p40QIUmSoqC0FDZsiCQ52cEnn0SRl2fjgQdyWbw416vc/fdfQKl8JkwopHlz6TARjZMkxSbKNGHr1nBSUqyhdllZ3k9sSklx8MgjuV6nwz/4QVGQWylE8ElSbGK+/Rb++tcYUlIcHDlS9a+/R49SZszIp6gIHI4gN1CIeiZJsQn57rswxo6NACpfBGzXzsmUKQXMmFHAoEEy1E40XZIUG6nz5w1sNryu/fXqVUrfvib79lkZLybGxaRJ1uM9r766WJ55LASSFBuVoiL4/PMokpMdfPppFI8+ep65c8sfmWMY8OMfO1m3roTp0wu4/vpCOT0WogJJiiHO5YJNm6yhdh9/7ODcufKekZQUh1dSBFiwwMWcOWeD3UwhQoYkxRBkmrB7dxgpKdF88IGjbKr+ikpLIT/fKHvGsRCidj4lRaXUBOBFrEecLtNaL6mw/b+B2R7v2Q9oq7XOrq2uqJujR+3MmdOa776reqhdly6lTJtmdZj07l1aZRkhRPVqTYpKKTvwEnADkA6kKqWWa633XCyjtf4d8Dt3+ZuA+e6EWGtdUTcdOjg5c8Z7LF3r1k6mTLE6TJKSpOdYiMvhy5HiCOCA+xnOKKXeAaYC1SW224C3L7GuwDrlXbXK6jC57bZ8Jk0qf6B7eLj1gPd333UwcWIh06YVcO21RfKsYyH8xJekmAAc81hOB0ZWVVApFQ1MAB64hLrzgHkAWmvi4uJ8aJolLCysTuX9VddfwsLCiI2N49NPDd55x8by5Tby863DvWbNIpgzx/s0+Omn4fnnS2nWLAyIcb98jxWM/W1scYIZS/apfmP5khSrOhmr7sr9TcCXWuvsutbVWr8KvHqxTFZWlg9Ns8TFxVGX8v6qe7lME775JpxPPmmN1naysyt3mKxaZXD48Bmvx3oahvWM5IKCSsVrFaz9bWxxghlL9ilwsTp27FhrfV+SYjrQ2WO5E5BRTdlZlJ8617Vuk5GdbfCXvzTngw8cHD1a9a+gT5+SsrkJ5TnHQgSPL0kxFeitlOoOHMdKfLdXLKSUigV+ANxR17pNTVgYvPJKc4qKvA+k27d3Mm1aAdOn5zNgQKl0mAhRD2qdElRrXYp1jXAVsNdapXcrpe5VSt3rUXQ6sFprnVdbXX/uQEN27pzB229Hs2uX9/+eFi1Mxo2zOk9iY13cc4+T997LYvPmU/zqV+cZOFASohD1xTDNBnlqZmZk+H6W3ZCuKRYWwtq1UaSkOFi7NoriYoM778xjyZJzXuW2bAknM9PO9dcXkpDQMK63SJyGEUv2KXCx3NcUazzkkBEtfuB0wsaNEaSkRLNiRRS5ud4H4B9+6ODJJ895zVCdlFQClAS3oUKIWklSvESmCTt3hpOcbE3SeupU1UPthgwpZtq0ApxOg+o77YUQDYUkxUt0/LidiRPbVrmtW7eLT7XLp1cvebSnEKFEkqIPTp+2YbdD69ausnWdOjlJSipmyxbrnDguzsnUqdZT7YYOlaF2QoQqSYrVuHDBYOVKq8Nk/fpIFizIZf587wcZz56dR7dupcyYUcCYMUWEyU9TiJAnf8Yeiovhiy8iSUmJZvXqSAoLyztMkpOjefjhC15HgLfeWsCtt17CsBIhRIPV5JOiywVffx1BcrKDjz5ykJNT9a2bbds6OXfOoGVL6SwRojFr0knxxAkbV10VztGjVQ8e79u3fKhdp07SYSJEU9CkkqJp4nX62769q9LDmjp2tHqOp08voF8/maRViKam0SfFs2cNPvrIQUqKg5tvLuD22/PLthkGzJrl4uWXDSZPthLhiBHFXg+AF0I0LY0yKRYUGKxeHckHHzj4/PMoSkqsw0ObDa+kCLBggZN587K8RpsIIZquRpMUS0vhyy8jSU52sHJlFHl5lQ/3UlMjyM62ed1v2KKF1esshBDQCJLi2bMGzzxjR+t2nD5d9VC7YcOKmT69gClTCrwSohBCVBTySdHhMPnnP22cO+c9hKRHj1JmzMhn2rQCuneXnmMhhG9CPilGRcH06S7+9jc78fFOpkyxHu85eLAMtRNC1F3IJ0WABx908cMfnmX06OJKt9gIIURdNIqkOGCASbt20lsihLh8ckeeEEJ4kKQohBAeJCkKIYQHSYpCCOFBkqIQQniQpCiEEB4kKQohhAfDNBvkTNINslFCiEahxrFuDfVI0ajLSym1pa51/FHXX69gtiFYsRpbHNmn0IjjY6waNdSkKIQQ9UKSohBCeGgsSfHVeqrrL8FsQ7BiNbY4wYwl+1SPsRpqR4sQQtSLxnKkKIQQfiFJUQghPEhSFEIID41ikll/Uko1A/4PKAa+0Fq/WQ9t6AE8CsRqrW8OYJxpwI+AeOAlrfXqAMbqBzwExAFrtdYvBzBWM2Ad8LjW+qMAxRgLPAXsBt7RWn8RiDjuWDZ3rBbAN1rrvwcozjXAbKy80F9rfXUg4rhjdQH+F8gC9mutlwQoTn/gCeAM1ufu/drqhHRSVEpNAF4E7MCy6n6wSqnXgclAptZ6YC31ZwDva60/VEq9C/glKdalDVrrQ8BcpVStv8DLjPMB8IFSqhXwe6BOSbGOsfYC97r/wP8SqDjuTY8Aui4xLiGOCVwAooD0AMeaCiQA2XWNVcff0XpgvfufZWqA96kP8LHW+hWl1BsBjDMR+LPWer1SajlQ699UyJ4+K6XswEtYO90fuM39X6EqfwMm+Fi/E3DMXcyfjwGsSxuCHecx9/aAxlJKTQE2AGsDFUcpNR7YA5yqY4w6xQHWa60nYiXg/wlwrCuAr7TWC4D7AhjnotuBt+sYp66xtgKzlFKfAZ8HMM4/3HF+B7Tx5c1DNikCI4ADWutDWuti4B2s/6iVaK3XYf2X9aV+OlZiBD/+fOrYhqDEUUoZSqnfAiu11mmBjOUuv9x9SjY7gHGuA0Zh/WH/xH1k6vc4WuuLDxA/C0TWZX8uYZ/S3XGgjv+o6/o7cp/WntNan69LnEuIdTfW5Y3rsS7hBCSO1jpTa30/8AusU/VahfLpcwLlR3RgfXBG+qH+n4D/VUr9CPjwcht5KW1QSrUBfgMMU0ot1lo/G4g4wM+B8UCsUqqX1nrpZcapNpb7GtwMrASyIlBxtNYPACil7gKyPJKXX+MopWYAPwRaYl0b84fqfk8vAn92X/NbF8A4AHOBv/ohRm2xlgJPKKVuB44EKo5SqhvwS6AZ8Dtf3iiUk2JVA7vrcid6lfW11nlY/8WCobo2nAHuDUKcP2H9E/Cn6mJ9AXwR6DgXv9Fa/y2QcbTWyUCyn2LUFisfK1kFNA6A1vpxP8apNpbWehfgz07E6uIcAebV5Y1C+fQ5HejssdwJyAhifX8IVhuCua+NbZ/kZxcasfwWJ5SPFFOB3kqp7sBxYBbWdaRg1feHYLUhmPva2PZJfnahEctvcUJ67LNSahLwAlYX/Ota699UU+5tYCzWPXKnsC7wvuZrfT+1NShtCOa+NrZ9kp/d5Wks+xTSSVEIIfwtlK8pCiGE30lSFEIID5IUhRDCgyRFIYTwIElRCCE8SFIUQggPkhSFEMKDJEUhhPAgSVEIITz8f4SZm51HBzqUAAAAAElFTkSuQmCC\n",
      "text/plain": [
       "<Figure size 360x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 360x288 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "SAMPLES_TYPES = [\"train\", \"test\"]\n",
    "#SAMPLES_TYPES = [\"test\", \"job\"]\n",
    "\n",
    "for st in SAMPLES_TYPES:\n",
    "    for part in set(df2[\"partition\"]):\n",
    "        tmp_df = df2[df2.partition == part]\n",
    "        plot_cdf_jerr(st, tmp_df, part)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 360x288 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 360x288 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "# for part in set(df1[\"partition\"]):\n",
    "#     tmp_df = df1[df1.partition == part]\n",
    "#     plot_cdf_qerr(\"test\", tmp_df)\n",
    "\n",
    "for st in SAMPLES_TYPES:\n",
    "    cur_df = df1[df1.samples_type == st]\n",
    "    for part in set(df1[\"partition\"]):\n",
    "        tmp_df = cur_df[cur_df.partition == part]\n",
    "        plot_cdf_qerr(st, tmp_df, part)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 20,
   "metadata": {},
   "outputs": [
    {
     "ename": "KeyError",
     "evalue": "'feat_rel_pg_ests_onehot'",
     "output_type": "error",
     "traceback": [
      "\u001b[0;31m---------------------------------------------------------------------------\u001b[0m",
      "\u001b[0;31mKeyError\u001b[0m                                  Traceback (most recent call last)",
      "\u001b[0;32m/usr/lib/python3.8/site-packages/pandas/core/indexes/base.py\u001b[0m in \u001b[0;36mget_loc\u001b[0;34m(self, key, method, tolerance)\u001b[0m\n\u001b[1;32m   2645\u001b[0m             \u001b[0;32mtry\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m-> 2646\u001b[0;31m                 \u001b[0;32mreturn\u001b[0m \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0m_engine\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mget_loc\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mkey\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m   2647\u001b[0m             \u001b[0;32mexcept\u001b[0m \u001b[0mKeyError\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n",
      "\u001b[0;32mpandas/_libs/index.pyx\u001b[0m in \u001b[0;36mpandas._libs.index.IndexEngine.get_loc\u001b[0;34m()\u001b[0m\n",
      "\u001b[0;32mpandas/_libs/index.pyx\u001b[0m in \u001b[0;36mpandas._libs.index.IndexEngine.get_loc\u001b[0;34m()\u001b[0m\n",
      "\u001b[0;32mpandas/_libs/hashtable_class_helper.pxi\u001b[0m in \u001b[0;36mpandas._libs.hashtable.PyObjectHashTable.get_item\u001b[0;34m()\u001b[0m\n",
      "\u001b[0;32mpandas/_libs/hashtable_class_helper.pxi\u001b[0m in \u001b[0;36mpandas._libs.hashtable.PyObjectHashTable.get_item\u001b[0;34m()\u001b[0m\n",
      "\u001b[0;31mKeyError\u001b[0m: 'feat_rel_pg_ests_onehot'",
      "\nDuring handling of the above exception, another exception occurred:\n",
      "\u001b[0;31mKeyError\u001b[0m                                  Traceback (most recent call last)",
      "\u001b[0;32m<ipython-input-20-1b61ba249b4e>\u001b[0m in \u001b[0;36m<module>\u001b[0;34m\u001b[0m\n\u001b[1;32m      3\u001b[0m \u001b[0mtmp\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mtmp\u001b[0m\u001b[0;34m[\u001b[0m\u001b[0mtmp\u001b[0m\u001b[0;34m[\u001b[0m\u001b[0;34m\"alg\"\u001b[0m\u001b[0;34m]\u001b[0m \u001b[0;34m==\u001b[0m \u001b[0;34m\"flow_loss2\"\u001b[0m\u001b[0;34m]\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m      4\u001b[0m \u001b[0mtmp\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mtmp\u001b[0m\u001b[0;34m[\u001b[0m\u001b[0mtmp\u001b[0m\u001b[0;34m[\u001b[0m\u001b[0;34m\"flow_features\"\u001b[0m\u001b[0;34m]\u001b[0m \u001b[0;34m==\u001b[0m \u001b[0;36m1\u001b[0m\u001b[0;34m]\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m----> 5\u001b[0;31m \u001b[0mtmp\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mtmp\u001b[0m\u001b[0;34m[\u001b[0m\u001b[0mtmp\u001b[0m\u001b[0;34m[\u001b[0m\u001b[0;34m\"feat_rel_pg_ests_onehot\"\u001b[0m\u001b[0;34m]\u001b[0m \u001b[0;34m==\u001b[0m \u001b[0;36m1\u001b[0m\u001b[0;34m]\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m",
      "\u001b[0;32m/usr/lib/python3.8/site-packages/pandas/core/frame.py\u001b[0m in \u001b[0;36m__getitem__\u001b[0;34m(self, key)\u001b[0m\n\u001b[1;32m   2798\u001b[0m             \u001b[0;32mif\u001b[0m \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mcolumns\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mnlevels\u001b[0m \u001b[0;34m>\u001b[0m \u001b[0;36m1\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m   2799\u001b[0m                 \u001b[0;32mreturn\u001b[0m \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0m_getitem_multilevel\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mkey\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m-> 2800\u001b[0;31m             \u001b[0mindexer\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mcolumns\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mget_loc\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mkey\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m   2801\u001b[0m             \u001b[0;32mif\u001b[0m \u001b[0mis_integer\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mindexer\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m   2802\u001b[0m                 \u001b[0mindexer\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0;34m[\u001b[0m\u001b[0mindexer\u001b[0m\u001b[0;34m]\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n",
      "\u001b[0;32m/usr/lib/python3.8/site-packages/pandas/core/indexes/base.py\u001b[0m in \u001b[0;36mget_loc\u001b[0;34m(self, key, method, tolerance)\u001b[0m\n\u001b[1;32m   2646\u001b[0m                 \u001b[0;32mreturn\u001b[0m \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0m_engine\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mget_loc\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mkey\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m   2647\u001b[0m             \u001b[0;32mexcept\u001b[0m \u001b[0mKeyError\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m-> 2648\u001b[0;31m                 \u001b[0;32mreturn\u001b[0m \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0m_engine\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mget_loc\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0m_maybe_cast_indexer\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mkey\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m   2649\u001b[0m         \u001b[0mindexer\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mget_indexer\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m[\u001b[0m\u001b[0mkey\u001b[0m\u001b[0;34m]\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mmethod\u001b[0m\u001b[0;34m=\u001b[0m\u001b[0mmethod\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mtolerance\u001b[0m\u001b[0;34m=\u001b[0m\u001b[0mtolerance\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m   2650\u001b[0m         \u001b[0;32mif\u001b[0m \u001b[0mindexer\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mndim\u001b[0m \u001b[0;34m>\u001b[0m \u001b[0;36m1\u001b[0m \u001b[0;32mor\u001b[0m \u001b[0mindexer\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0msize\u001b[0m \u001b[0;34m>\u001b[0m \u001b[0;36m1\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n",
      "\u001b[0;32mpandas/_libs/index.pyx\u001b[0m in \u001b[0;36mpandas._libs.index.IndexEngine.get_loc\u001b[0;34m()\u001b[0m\n",
      "\u001b[0;32mpandas/_libs/index.pyx\u001b[0m in \u001b[0;36mpandas._libs.index.IndexEngine.get_loc\u001b[0;34m()\u001b[0m\n",
      "\u001b[0;32mpandas/_libs/hashtable_class_helper.pxi\u001b[0m in \u001b[0;36mpandas._libs.hashtable.PyObjectHashTable.get_item\u001b[0;34m()\u001b[0m\n",
      "\u001b[0;32mpandas/_libs/hashtable_class_helper.pxi\u001b[0m in \u001b[0;36mpandas._libs.hashtable.PyObjectHashTable.get_item\u001b[0;34m()\u001b[0m\n",
      "\u001b[0;31mKeyError\u001b[0m: 'feat_rel_pg_ests_onehot'"
     ]
    }
   ],
   "source": [
    "tmp = df2[df2[\"template\"] == \"all_job\"]\n",
    "tmp = tmp[tmp[\"partition\"] == 2]\n",
    "tmp = tmp[tmp[\"alg\"] == \"flow_loss2\"]\n",
    "tmp = tmp[tmp[\"flow_features\"] == 1]\n",
    "tmp = tmp[tmp[\"feat_rel_pg_ests_onehot\"] == 1]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 21,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "0\n"
     ]
    },
    {
     "ename": "KeyError",
     "evalue": "'num_hidden_layers'",
     "output_type": "error",
     "traceback": [
      "\u001b[0;31m---------------------------------------------------------------------------\u001b[0m",
      "\u001b[0;31mKeyError\u001b[0m                                  Traceback (most recent call last)",
      "\u001b[0;32m/usr/lib/python3.8/site-packages/pandas/core/indexes/base.py\u001b[0m in \u001b[0;36mget_loc\u001b[0;34m(self, key, method, tolerance)\u001b[0m\n\u001b[1;32m   2645\u001b[0m             \u001b[0;32mtry\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m-> 2646\u001b[0;31m                 \u001b[0;32mreturn\u001b[0m \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0m_engine\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mget_loc\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mkey\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m   2647\u001b[0m             \u001b[0;32mexcept\u001b[0m \u001b[0mKeyError\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n",
      "\u001b[0;32mpandas/_libs/index.pyx\u001b[0m in \u001b[0;36mpandas._libs.index.IndexEngine.get_loc\u001b[0;34m()\u001b[0m\n",
      "\u001b[0;32mpandas/_libs/index.pyx\u001b[0m in \u001b[0;36mpandas._libs.index.IndexEngine.get_loc\u001b[0;34m()\u001b[0m\n",
      "\u001b[0;32mpandas/_libs/hashtable_class_helper.pxi\u001b[0m in \u001b[0;36mpandas._libs.hashtable.PyObjectHashTable.get_item\u001b[0;34m()\u001b[0m\n",
      "\u001b[0;32mpandas/_libs/hashtable_class_helper.pxi\u001b[0m in \u001b[0;36mpandas._libs.hashtable.PyObjectHashTable.get_item\u001b[0;34m()\u001b[0m\n",
      "\u001b[0;31mKeyError\u001b[0m: 'num_hidden_layers'",
      "\nDuring handling of the above exception, another exception occurred:\n",
      "\u001b[0;31mKeyError\u001b[0m                                  Traceback (most recent call last)",
      "\u001b[0;32m<ipython-input-21-4426bdb23ae3>\u001b[0m in \u001b[0;36m<module>\u001b[0;34m\u001b[0m\n\u001b[1;32m      1\u001b[0m \u001b[0mprint\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mlen\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mtmp\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m----> 2\u001b[0;31m \u001b[0mprint\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mset\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mtmp\u001b[0m\u001b[0;34m[\u001b[0m\u001b[0;34m\"num_hidden_layers\"\u001b[0m\u001b[0;34m]\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m      3\u001b[0m \u001b[0mprint\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mset\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mtmp\u001b[0m\u001b[0;34m[\u001b[0m\u001b[0;34m\"hidden_layer_size\"\u001b[0m\u001b[0;34m]\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m      4\u001b[0m \u001b[0mprint\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mset\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mtmp\u001b[0m\u001b[0;34m[\u001b[0m\u001b[0;34m\"flow_features\"\u001b[0m\u001b[0;34m]\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m      5\u001b[0m \u001b[0mprint\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mset\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mtmp\u001b[0m\u001b[0;34m[\u001b[0m\u001b[0;34m\"feat_rel_pg_ests_onehot\"\u001b[0m\u001b[0;34m]\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n",
      "\u001b[0;32m/usr/lib/python3.8/site-packages/pandas/core/frame.py\u001b[0m in \u001b[0;36m__getitem__\u001b[0;34m(self, key)\u001b[0m\n\u001b[1;32m   2798\u001b[0m             \u001b[0;32mif\u001b[0m \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mcolumns\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mnlevels\u001b[0m \u001b[0;34m>\u001b[0m \u001b[0;36m1\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m   2799\u001b[0m                 \u001b[0;32mreturn\u001b[0m \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0m_getitem_multilevel\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mkey\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m-> 2800\u001b[0;31m             \u001b[0mindexer\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mcolumns\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mget_loc\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mkey\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m   2801\u001b[0m             \u001b[0;32mif\u001b[0m \u001b[0mis_integer\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mindexer\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m   2802\u001b[0m                 \u001b[0mindexer\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0;34m[\u001b[0m\u001b[0mindexer\u001b[0m\u001b[0;34m]\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n",
      "\u001b[0;32m/usr/lib/python3.8/site-packages/pandas/core/indexes/base.py\u001b[0m in \u001b[0;36mget_loc\u001b[0;34m(self, key, method, tolerance)\u001b[0m\n\u001b[1;32m   2646\u001b[0m                 \u001b[0;32mreturn\u001b[0m \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0m_engine\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mget_loc\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mkey\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m   2647\u001b[0m             \u001b[0;32mexcept\u001b[0m \u001b[0mKeyError\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m-> 2648\u001b[0;31m                 \u001b[0;32mreturn\u001b[0m \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0m_engine\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mget_loc\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0m_maybe_cast_indexer\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mkey\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m   2649\u001b[0m         \u001b[0mindexer\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mget_indexer\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m[\u001b[0m\u001b[0mkey\u001b[0m\u001b[0;34m]\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mmethod\u001b[0m\u001b[0;34m=\u001b[0m\u001b[0mmethod\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mtolerance\u001b[0m\u001b[0;34m=\u001b[0m\u001b[0mtolerance\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m   2650\u001b[0m         \u001b[0;32mif\u001b[0m \u001b[0mindexer\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mndim\u001b[0m \u001b[0;34m>\u001b[0m \u001b[0;36m1\u001b[0m \u001b[0;32mor\u001b[0m \u001b[0mindexer\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0msize\u001b[0m \u001b[0;34m>\u001b[0m \u001b[0;36m1\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n",
      "\u001b[0;32mpandas/_libs/index.pyx\u001b[0m in \u001b[0;36mpandas._libs.index.IndexEngine.get_loc\u001b[0;34m()\u001b[0m\n",
      "\u001b[0;32mpandas/_libs/index.pyx\u001b[0m in \u001b[0;36mpandas._libs.index.IndexEngine.get_loc\u001b[0;34m()\u001b[0m\n",
      "\u001b[0;32mpandas/_libs/hashtable_class_helper.pxi\u001b[0m in \u001b[0;36mpandas._libs.hashtable.PyObjectHashTable.get_item\u001b[0;34m()\u001b[0m\n",
      "\u001b[0;32mpandas/_libs/hashtable_class_helper.pxi\u001b[0m in \u001b[0;36mpandas._libs.hashtable.PyObjectHashTable.get_item\u001b[0;34m()\u001b[0m\n",
      "\u001b[0;31mKeyError\u001b[0m: 'num_hidden_layers'"
     ]
    }
   ],
   "source": [
    "print(len(tmp))\n",
    "print(set(tmp[\"num_hidden_layers\"]))\n",
    "print(set(tmp[\"hidden_layer_size\"]))\n",
    "print(set(tmp[\"flow_features\"]))\n",
    "print(set(tmp[\"feat_rel_pg_ests_onehot\"]))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 25,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>loss</th>\n",
       "      <th>template</th>\n",
       "      <th>samples_type</th>\n",
       "      <th>alg</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>10588</th>\n",
       "      <td>2.386220e+09</td>\n",
       "      <td>9a</td>\n",
       "      <td>test</td>\n",
       "      <td>mse</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>12437</th>\n",
       "      <td>7.392958e+08</td>\n",
       "      <td>9a</td>\n",
       "      <td>test</td>\n",
       "      <td>mse</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1852</th>\n",
       "      <td>6.130389e+08</td>\n",
       "      <td>9a</td>\n",
       "      <td>train</td>\n",
       "      <td>mse</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>9174</th>\n",
       "      <td>6.041250e+08</td>\n",
       "      <td>9a</td>\n",
       "      <td>test</td>\n",
       "      <td>mse</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>12592</th>\n",
       "      <td>4.974815e+08</td>\n",
       "      <td>9a</td>\n",
       "      <td>test</td>\n",
       "      <td>mse</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1947</th>\n",
       "      <td>4.782442e+08</td>\n",
       "      <td>9a</td>\n",
       "      <td>train</td>\n",
       "      <td>flow_loss2</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>6976</th>\n",
       "      <td>4.601467e+08</td>\n",
       "      <td>9b</td>\n",
       "      <td>test</td>\n",
       "      <td>flow_loss2</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4517</th>\n",
       "      <td>4.103318e+08</td>\n",
       "      <td>9a</td>\n",
       "      <td>train</td>\n",
       "      <td>mse</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>10234</th>\n",
       "      <td>3.990362e+08</td>\n",
       "      <td>9a</td>\n",
       "      <td>test</td>\n",
       "      <td>mse</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1373</th>\n",
       "      <td>3.799229e+08</td>\n",
       "      <td>9a</td>\n",
       "      <td>train</td>\n",
       "      <td>mse</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>466</th>\n",
       "      <td>3.727273e+08</td>\n",
       "      <td>9a</td>\n",
       "      <td>train</td>\n",
       "      <td>mse</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>466</th>\n",
       "      <td>3.727273e+08</td>\n",
       "      <td>9a</td>\n",
       "      <td>train</td>\n",
       "      <td>flow_loss2</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1373</th>\n",
       "      <td>3.664188e+08</td>\n",
       "      <td>9a</td>\n",
       "      <td>train</td>\n",
       "      <td>flow_loss2</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>12437</th>\n",
       "      <td>3.628936e+08</td>\n",
       "      <td>9a</td>\n",
       "      <td>test</td>\n",
       "      <td>flow_loss2</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>5574</th>\n",
       "      <td>3.242919e+08</td>\n",
       "      <td>9a</td>\n",
       "      <td>train</td>\n",
       "      <td>flow_loss2</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>5574</th>\n",
       "      <td>3.235730e+08</td>\n",
       "      <td>9a</td>\n",
       "      <td>train</td>\n",
       "      <td>mse</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4395</th>\n",
       "      <td>3.077018e+08</td>\n",
       "      <td>9a</td>\n",
       "      <td>train</td>\n",
       "      <td>mse</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>5254</th>\n",
       "      <td>3.047984e+08</td>\n",
       "      <td>9a</td>\n",
       "      <td>train</td>\n",
       "      <td>flow_loss2</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>5254</th>\n",
       "      <td>3.040939e+08</td>\n",
       "      <td>9a</td>\n",
       "      <td>train</td>\n",
       "      <td>mse</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2317</th>\n",
       "      <td>2.665695e+08</td>\n",
       "      <td>11b</td>\n",
       "      <td>train</td>\n",
       "      <td>flow_loss2</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "               loss template samples_type         alg\n",
       "10588  2.386220e+09       9a         test         mse\n",
       "12437  7.392958e+08       9a         test         mse\n",
       "1852   6.130389e+08       9a        train         mse\n",
       "9174   6.041250e+08       9a         test         mse\n",
       "12592  4.974815e+08       9a         test         mse\n",
       "1947   4.782442e+08       9a        train  flow_loss2\n",
       "6976   4.601467e+08       9b         test  flow_loss2\n",
       "4517   4.103318e+08       9a        train         mse\n",
       "10234  3.990362e+08       9a         test         mse\n",
       "1373   3.799229e+08       9a        train         mse\n",
       "466    3.727273e+08       9a        train         mse\n",
       "466    3.727273e+08       9a        train  flow_loss2\n",
       "1373   3.664188e+08       9a        train  flow_loss2\n",
       "12437  3.628936e+08       9a         test  flow_loss2\n",
       "5574   3.242919e+08       9a        train  flow_loss2\n",
       "5574   3.235730e+08       9a        train         mse\n",
       "4395   3.077018e+08       9a        train         mse\n",
       "5254   3.047984e+08       9a        train  flow_loss2\n",
       "5254   3.040939e+08       9a        train         mse\n",
       "2317   2.665695e+08      11b        train  flow_loss2"
      ]
     },
     "execution_count": 25,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "sdf = df2.sort_values(by=\"loss\", ascending=False)\n",
    "sdf[[\"loss\", \"template\",\"samples_type\", \"alg\"]].head(20)"
   ]
  },
  {
   "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": "3.8.2"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 4
}