study_display_2.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Hyperparameter display "
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "![Cartpole environment](https://nav74neet.github.io/media/blog/openaigym/openaigym.jpg)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "#### Methods to understand the results of the study:\n",
    "- TensorBoard integration (only with original objects) \n",
    "- Average performance of different parameters across all studies\n",
    "- Selection of the top performing Parameter configurations\n",
    "- Cross studies with left out Parameters\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Set the GPU for training\n",
    "import os\n",
    "os.environ['CUDA_VISIBLE_DEVICES']='0'"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {},
   "outputs": [],
   "source": [
    "import pickle\n",
    "import optuna\n",
    "from fanova import fANOVA\n",
    "import pandas as pd\n",
    "import sklearn.feature_selection\n",
    "import numpy as np\n",
    "import seaborn as sns\n",
    "import matplotlib.pyplot as plt"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {
    "scrolled": true
   },
   "outputs": [],
   "source": [
    "#Unfortunately I run experiment on different servers with different optuna versions...\n",
    "#!pip install optuna==1.4.0"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "metadata": {},
   "outputs": [],
   "source": [
    "labels = ['G1', 'G2', 'G3', 'G4', 'G5']\n",
    "# bar coordinates\n",
    "men_means = [20, 34, 30, 35, 27]\n",
    "women_means = [25, 32, 34, 20, 25]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "metadata": {},
   "outputs": [],
   "source": [
    "def plot_difference(labels,men_means,women_means):\n",
    "    x = np.arange(len(labels))  # the label locations\n",
    "    width = 0.35  # the width of the bars\n",
    "\n",
    "    fig, ax = plt.subplots()\n",
    "    rects1 = ax.bar(x - width/2, men_means, width, label='Var1')\n",
    "    rects2 = ax.bar(x + width/2, women_means, width, label='Var2')\n",
    "\n",
    "    # Add some text for labels, title and custom x-axis tick labels, etc.\n",
    "    ax.set_ylabel('Average reward')\n",
    "    ax.set_title('Reward comparison between parameter values ')\n",
    "    ax.set_xticks(x)\n",
    "    ax.set_xticklabels(labels)\n",
    "    ax.legend()\n",
    "\n",
    "\n",
    "    def autolabel(rects):\n",
    "        \"\"\"Attach a text label above each bar in *rects*, displaying its height.\"\"\"\n",
    "        for rect in rects:\n",
    "            height = rect.get_height()\n",
    "            height = int(height)\n",
    "            ax.annotate('{}'.format(height),\n",
    "                        xy=(rect.get_x() + rect.get_width() / 2, height),\n",
    "                        xytext=(0, 3),  # 3 points vertical offset\n",
    "                        textcoords=\"offset points\",\n",
    "                        ha='center', va='bottom')\n",
    "\n",
    "\n",
    "    autolabel(rects1)\n",
    "    autolabel(rects2)\n",
    "\n",
    "    fig.tight_layout()\n",
    "    ax.invert_yaxis()\n",
    "\n",
    "    plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "metadata": {},
   "outputs": [],
   "source": [
    "def plot_difference_2(women_means):\n",
    "    width = 0.35  # the width of the bars\n",
    "\n",
    "    fig, ax = plt.subplots()\n",
    "    rect1 = ax.bar(1, women_means[1], width, label='Layer1')\n",
    "    rect2 = ax.bar(2, women_means[2], width, label='Layer3')\n",
    "    rect3 = ax.bar(3, women_means[0], width, label='Layer6')\n",
    "\n",
    "    # Add some text for labels, title and custom x-axis tick labels, etc.\n",
    "    ax.set_ylabel('Average reward')\n",
    "    ax.set_title('Reward comparison between parameter values ')\n",
    "    labels = [item.get_text() for item in ax.get_xticklabels()]\n",
    "    labels[0] = \"OOF\"\n",
    "    ax.set_xticklabels(labels)\n",
    "    \n",
    "\n",
    "    ax.legend()\n",
    "    fig.tight_layout()\n",
    "    ax.invert_yaxis()\n",
    "    plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Import the study \n",
    "\n",
    "###### If you are in need of the original optuna objects please contact me. "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {},
   "outputs": [],
   "source": [
    "sumTable = pd.read_csv(\"saves/Acro_study_new_main.csv\") \n",
    "sumTable_1 = pd.read_csv(\"saves/Acro_study_new_addition.csv\") \n",
    "sumTable2 = pd.read_csv(\"saves/Cart_study_new_main.csv\") \n",
    "sumTable2_1 = pd.read_csv(\"saves/Cart_study_new_addition.csv\") \n",
    "sumTable4 = pd.read_csv(\"saves/Lunar_study_new_main.csv\") \n",
    "sumTable4_1 = pd.read_csv(\"saves/Lunar_study_new_addition.csv\") \n",
    "sumTable5 = pd.read_csv(\"saves/Acro_study_old.csv\") "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {},
   "outputs": [],
   "source": [
    "    sumTable = sumTable.drop(columns = [\"number\",\"datetime_start\",\"datetime_complete\",\"state\",\n",
    "                                 \"system_attrs_search_space\",\"Unnamed: 0\"])\n",
    "    sumTable_1 = sumTable_1.drop(columns = [\"number\",\"datetime_start\",\"duration\",\"datetime_complete\",\n",
    "                                 \"state\",\"system_attrs_search_space\",\"Unnamed: 0\"])\n",
    "    sumTable2 = sumTable2.drop(columns = [\"number\",\"datetime_start\",\"duration\",\"datetime_complete\",\n",
    "                                 \"state\",\"system_attrs_search_space\",\"Unnamed: 0\"])\n",
    "    sumTable2_1 = sumTable2_1.drop(columns = [\"number\",\"datetime_start\",\"duration\",\"datetime_complete\",\n",
    "                                 \"state\",\"system_attrs_search_space\",\"Unnamed: 0\"])\n",
    "    sumTable4 = sumTable4.drop(columns = [\"number\",\"datetime_start\",\"datetime_complete\",\"state\",\n",
    "                                 \"system_attrs_search_space\",\"Unnamed: 0\"])\n",
    "    sumTable4_1 = sumTable4_1.drop(columns = [\"number\",\"datetime_start\",\"duration\",\"datetime_complete\",\n",
    "                                 \"state\",\"system_attrs_search_space\",\"Unnamed: 0\"])\n",
    "    sumTable5 = sumTable5.drop(columns = [\"number\",\"datetime_start\",\"duration\",\"datetime_complete\",\n",
    "                                 \"state\",\"system_attrs_search_space\",\"Unnamed: 0\"])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {},
   "outputs": [],
   "source": [
    "combine = [sumTable,sumTable_1]\n",
    "sumTable6 = pd.concat(combine,ignore_index=True)\n",
    "combine = [sumTable2,sumTable2_1]\n",
    "sumTable7 = pd.concat(combine,ignore_index=True)\n",
    "combine = [sumTable4,sumTable4_1]\n",
    "sumTable8 = pd.concat(combine,ignore_index=True)\n",
    "\n",
    "sumTable6[\"params_num_heads\"] = [4] * len(sumTable6.values)\n",
    "sumTable8[\"params_num_heads\"] = [4] * len(sumTable8.values)\n",
    "sumTable6 = sumTable6 [['value', 'params_custom_layer_init', 'params_custom_lr_schedule',\n",
    "       'params_layer_type', 'params_loss_function',\n",
    "       'params_target_update_period','params_num_heads' ]]\n",
    "sumTable7 =sumTable7 [['value', 'params_custom_layer_init', 'params_custom_lr_schedule',\n",
    "       'params_layer_type', 'params_loss_function',\n",
    "       'params_target_update_period','params_num_heads' ]]\n",
    "sumTable8 =sumTable8 [['value', 'params_custom_layer_init', 'params_custom_lr_schedule',\n",
    "       'params_layer_type', 'params_loss_function',\n",
    "       'params_target_update_period','params_num_heads' ]]"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Average performance AcroBot"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "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>0</th>\n",
       "      <th>1</th>\n",
       "      <th>2</th>\n",
       "      <th>3</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>params_custom_layer_init</th>\n",
       "      <td>(0.5, 1.0)</td>\n",
       "      <td>-251.980553</td>\n",
       "      <td>-274.958335</td>\n",
       "      <td>NaN</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>params_custom_lr_schedule</th>\n",
       "      <td>(No, Transformer)</td>\n",
       "      <td>-178.783333</td>\n",
       "      <td>-348.155555</td>\n",
       "      <td>NaN</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>params_layer_type</th>\n",
       "      <td>(6, 1, 3)</td>\n",
       "      <td>-242.350001</td>\n",
       "      <td>-319.916667</td>\n",
       "      <td>-228.141664</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>params_loss_function</th>\n",
       "      <td>(element_wise_squared_loss, element_wise_huber...</td>\n",
       "      <td>-267.927082</td>\n",
       "      <td>-254.554167</td>\n",
       "      <td>NaN</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>params_target_update_period</th>\n",
       "      <td>(5)</td>\n",
       "      <td>-263.469444</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>params_num_heads</th>\n",
       "      <td>(4)</td>\n",
       "      <td>-263.469444</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "                                                                             0  \\\n",
       "params_custom_layer_init                                            (0.5, 1.0)   \n",
       "params_custom_lr_schedule                                    (No, Transformer)   \n",
       "params_layer_type                                                    (6, 1, 3)   \n",
       "params_loss_function         (element_wise_squared_loss, element_wise_huber...   \n",
       "params_target_update_period                                                (5)   \n",
       "params_num_heads                                                           (4)   \n",
       "\n",
       "                                      1           2           3  \n",
       "params_custom_layer_init    -251.980553 -274.958335         NaN  \n",
       "params_custom_lr_schedule   -178.783333 -348.155555         NaN  \n",
       "params_layer_type           -242.350001 -319.916667 -228.141664  \n",
       "params_loss_function        -267.927082 -254.554167         NaN  \n",
       "params_target_update_period -263.469444         NaN         NaN  \n",
       "params_num_heads            -263.469444         NaN         NaN  "
      ]
     },
     "execution_count": 14,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "X =  sumTable6.values[:,1:]\n",
    "y =  sumTable6.values[:,0]\n",
    "y = y *-1\n",
    "features = sumTable7.columns[1:]\n",
    "final = []\n",
    "for w in range(len(features)):\n",
    "    summary = {}\n",
    "    for x in range(len(X)):\n",
    "        if str(X[x][w]) in summary: \n",
    "            summary[str(X[x][w])].append(y[x])\n",
    "        else:\n",
    "            summary[str(X[x][w])] = [y[x]]\n",
    "    row = []\n",
    "    row.append(summary.keys())\n",
    "    for x in summary.keys():\n",
    "        row.append((np.mean(summary[x])))\n",
    "    final.append(row)\n",
    "pd.DataFrame(final, index = features)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "number1 = []\n",
    "number2 = []\n",
    "number3 = []\n",
    "for row in final: \n",
    "    number1.append(row[1])\n",
    "    if len(row) >= 3:\n",
    "        number2.append(row[2])\n",
    "    else: \n",
    "        number2.append(0)\n",
    "    if len(row) == 4:\n",
    "        number3.append(row[3])  \n",
    "values = np.array([number1,number2]).T\n",
    "values = np.concatenate((values[:2], values[3:]))\n",
    "labels = [\"init\",\"schedule\",\"loss\",\"update\",\"heads\"]\n",
    "sns.set()\n",
    "plot_difference(labels,values[:,0],values[:,1])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 20,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAaQAAAEaCAYAAABejCMwAAAABHNCSVQICAgIfAhkiAAAAAlwSFlzAAALEgAACxIB0t1+/AAAADh0RVh0U29mdHdhcmUAbWF0cGxvdGxpYiB2ZXJzaW9uMy4xLjMsIGh0dHA6Ly9tYXRwbG90bGliLm9yZy+AADFEAAAgAElEQVR4nO3deVxVdeL/8Re7CyQuIOZMmUuaOeaC+4IyLt0AzdI0lynNpHTKtEg0TSORJC3UGrWxtMU0K5dSsZQiE800q9EyndEyywUQRSSQ5X5+f/jzfiO1i8i9XOD9fDzm8Zh7Ovec9/1c8M3Z3YwxBhERkTLmXtYBREREQIUkIiIuQoUkIiIuQYUkIiIuQYUkIiIuQYUkIiIuQYVUzjRt2pSIiAj69+/PnXfeSd++fbn77rvZu3dvmeaKiYlhwYIFZZqhJObNm8fatWsduo7Vq1cTGRl51e976aWX2LJliwMSVUzvvvsuy5cvd/p6R4wYwaZNm5y+3orIs6wDyNV7/fXXqVWrlu31q6++ysyZM3nnnXfKMFX5NH78+LKOcEU7d+6kcePGZR2j3Pjqq69o0qRJWceQa6BCKucKCgo4fvw4NWrUsE1buHAhH3/8MVarlfr16zN9+nT27t3La6+9xttvvw1A3759CQsL49FHH+XEiRMMHDiQrVu38sorr5CUlERubi45OTlMmjSJ3r17s2DBAr755htSU1Np2rQpM2bM4KmnnuKHH34gMDAQDw8P2rZte9l8zz//PMnJyXh4eNC6dWumT5+Om5sbzz33HDt27MDDw4OWLVsyefJkfH19CQ0NJTw8nC+++ILMzExGjx7Nnj17+O677/D09GThwoXUrVuX0NBQwsLCSElJISsri5EjRzJ06FCsViuzZs3i22+/JTs7G2MMM2fOpG3btkRHR3PmzBmOHj1Kjx49OHXqFE2aNOGBBx5g/vz5bN68GS8vL2rWrElcXByBgYHs3r2b+Ph4cnJy8PLy4rHHHqN79+6sXr2azZs34+7uzpEjR6hSpQqzZ8+mUaNGl4xDWloaDzzwAKmpqdSvX59nn32WgIAAsrKyiI2N5eDBg+Tn59OpUyeefPJJ3nnnHfbt20d8fDwnTpxg6dKlfPbZZwA88MAD1KlTh9mzZ5OXl0e3bt3YsmULqampxMbGcubMGQoLCxkxYgQDBw4E4JNPPmHhwoXk5+dTpUoVJk2aROvWrVmwYAG//voraWlp/Prrr9StW5fnn3+ewMDAIvkXLFjAkSNHOHHiBGlpaTRr1ozY2Fh8fX359NNPWbx4MXl5eWRkZHDnnXfy2GOPsXPnTmJjY6lWrRrZ2dm8//77xMfHX/F7qVKlCgcPHuTUqVOEhobi7+/Pp59+SlpaGjNnzqRTp07k5eUxZ84cdu3aRWFhIc2bN2fq1Kns2LGDTz75hJSUFKpUqcKwYcMu+3tQt25dRowYQY0aNTh8+DD33nsvI0aMsH3OIUOGMHLkSPr27QvA888/D8C4ceOYMWMGR44c4cyZM1SvXp05c+bQsGFD23t/+eUXIiIi+Prrry/7+t1332XFihVYrVb8/f2ZNm0ajRo1Yvfu3Tz33HNYrVYAIiMjbeuvdIyUKzfffLMJDw834eHhpkuXLiY0NNQ8++yzJj093RhjzJo1a8xjjz1m8vPzjTHGrFy50owePdrk5OSYNm3amMzMTHP06FHTpUsXM3jwYGOMMW+99ZaZPn26+eWXX8yIESNMTk6OMcaY9evXm/DwcGOMMfPnzzd9+/a1LTc2NtY8+eSTxmq1mlOnTpnu3bub+fPnX5L39ddfN8OGDTM5OTmmsLDQjB8/3qxZs8bMmzfP/POf/zR5eXmmsLDQREdHm2nTphljjOnZs6eZNWuWMcaYDRs2mGbNmpn9+/cbY4wZO3asWbhwoW2+adOmGavVao4fP246dOhgfvjhB7Nnzx7zyCOPmMLCQmOMMYsXLzaRkZHGGGMmTZpk7rvvPlu+SZMmmSVLlphjx46ZNm3amPPnzxtjjHn11VfN5s2bTUZGhunUqZP55ptvjDHGHDx40LRv3978/PPP5v333zdt27Y1x48fN8YYExMTY5588slLxuD99983rVq1Mj/99JMxxpi5c+ea8ePHG2OMiY6ONm+88YYxxpiCggLzxBNPmFdeecUYY8zw4cNNYmKiMcaY0NBQc+DAAZOTk2N69OhhunfvbowxJjk52YwePdrk5+ebO+64w+zbt88YY8zZs2eNxWIxX3/9tfnxxx9NeHi4ycjIsH2GLl26mOzsbDN//nzz97//3WRlZRljjImMjDTz5s275DPMnz/fdO/e3aSlpZnCwkIzceJE89xzzxmr1WqGDx9ufvzxR2OMMSdOnDC33HKLOXXqlPniiy9Ms2bNzC+//GKMMXa/l0GDBpm8vDyTmppqbr75Ztu4LFu2zIwcOdIYY8yCBQts6704ltOnTy/yXRpz5d+Di+M6efLkSz6jMca89957ZsyYMbbvo2vXrubHH380iYmJ5tlnn7XNN23aNBMTE1Pkezp69Khp1aqVbZ7fv965c6cZOnSo+e2334wxxnz++efm9ttvN8YY849//MOsX7/eGGPM/v37zYwZMy6brTLQFlI5dHGX3XfffceYMWPo0KEDtWvXBuDTTz9l79693H333QBYrVZycnKoUqUKnTt3JiUlhdOnTzN48GDeeecdsrKy+OSTTxg9ejT169cnPj6eDz/8kCNHjtj+kr2oVatWeHpe+JHZsWMHU6ZMwc3NjVq1atG7d+/LZt2+fTv9+/enSpUqACQkJAAwcOBAJkyYgJeXF3BhP/y4ceNs7+vTpw8Af/3rX6lTpw7NmjUD4IYbbiAzM9M239ChQ3FzcyMoKIhu3bqRkpLCqFGjqFGjBitXruTo0aPs3LmT6tWr295zuS25unXr0qxZMwYMGED37t3p3r07nTp14rPPPuOGG27gtttuA6BJkya0adOGL7/8Ejc3N2699VaCgoIAaN68OZs3b77sOHTu3Jkbb7zR9tkvbrkkJyezd+9e3nvvPQByc3Mv+/7evXuzdetWmjRpQseOHTlw4AD//e9/SUpKok+fPvz000/8/PPPTJkyxfae3Nxcvv/+e4wxpKamcv/999v+m5ubGz///DMA7du3x9fX1/YZfj++v3f77bdTp04d22eYNWsWkyZNYtGiRSQnJ7N+/XoOHTqEMYacnBwA6tWrR/369QFo3br1n34vPXv2xMvLi4CAAKpVq0a3bt2AC9/5mTNnbOOVlZXF9u3bAcjPz7f97P/elX4PLgoODr7sZ7zjjjuIj48nLS2N77//ngYNGtj+99e//pU333yTI0eO8OWXX9K6devLLuNykpOTOXLkCEOGDLFNO3v2LGfOnMFisRATE8Mnn3xC586dmThxYrGXW9GokMqxW2+9lcmTJxMdHc0tt9zCX/7yF6xWK6NHj2bo0KEA5OXl2f6B6dWrF1u3buXs2bOMHj2aw4cPs2XLFg4ePEj79u357rvvGDt2LPfffz9dunShXbt2PPPMM7b1VatWrcj6ze9ug+jh4XHZjBcL7KL09HSsVitWqxU3NzfbdKvVSn5+vu21t7e37f9fLC17y7darbi7u5OcnExsbCwjR47k73//Ow0bNuSDDz644ucAcHd356233mLv3r3s2LGDWbNm0a1bN4KDg4vkvPi5CwoK8PLyshUtXPhH3lzh1pC/Hx+r1WrLbbVamTdvnm0339mzZy9ZH1z47ubNm0dqaipdunShdu3abNu2ja1btzJhwgRSU1Px8/Nj3bp1tvekp6fj5+fHqlWr6NSpk+2PAYDjx48TGBjI5s2bS/wZ3N3d+e233xgwYAC9evUiODiYu+++my1bttiW8fuxtve9/P47h0t/di6ud8qUKYSEhACQnZ3N+fPnLzvflX4P/pjr96pWrUrfvn1Zv349X3/9NYMGDQLg7bffZtWqVQwbNoyIiAj8/f355Zdfirz3j2P3+59nq9VK//79iYqKsr1OTU2lRo0aDBkyhJ49e5KSksLnn3/OSy+9xKZNm/Dx8blsxopMZ9mVc+Hh4bRs2ZK4uDgAunbtynvvvce5c+eAC2eRPfnkkwCEhoayY8cO9u/fT8uWLenSpQvz5s2je/fueHh4sGvXLlq0aMHIkSNp3749SUlJFBYWXna93bp147333sNqtZKZmUlSUtJl5+vUqRPr168nLy8Pq9XKjBkz2LBhA926dWPFihXk5+djtVpZvnw5Xbp0uerPf/EMuWPHjpGSkkL37t1JSUmhZ8+eDB06lBYtWrBly5Yrfo6LfvjhB8LDw2nUqBGRkZHcf//97N27l1atWnH48GH+85//APDf//6XXbt20b59+6vKuXPnTo4dOwbAypUr6d69O3Dh+1q2bBnGGPLy8nj44Yd56623gAsFUFBQAECbNm04evQoycnJdO7cmS5duvD666/ToEEDatasyU033USVKlVshXT8+HHCw8PZt28fnTp1IiUlhUOHDgHw2Wef0a9fvytujV1JUlISWVlZWK1WVq1aRc+ePTly5Ajnzp3jscceIzQ0lJ07d9q+6z8qyffyR127dmX58uW2dUybNo0XXnjhkvH6s98De+655x7WrFnDnj17bMdytm3bxoABAxg0aBA33XQTn3zyySXZr7vuOvLz8/nf//4HwIYNG4rk3rBhA6mpqQCsWLGC++67D7hw3Gr//v3cddddPPvss5w9e5a0tLSrGpeKQltIFcC0adPo168fn3/+OYMGDeLkyZPcc889uLm5Ua9ePZ577jkA/Pz8aNSoEVWrVsXDw4Nu3brx1FNP2XaPhYeH8/HHH2OxWLBarfTs2ZPMzEzbL/XvPfLII0yfPh2LxUKtWrW4+eabL5ttyJAh/Prrr9x1110YY2jfvj0jRoygoKCA2bNnc+edd1JQUEDLli2ZNm3aVX/2X375hbvuuovc3FymTp1Kw4YNGTJkCI8//jgREREUFBTQpUsX28HtK2nWrBkWi4W7776batWqUaVKFaZOnUqtWrWYN28ezz77LLm5ubi5uREXF8dNN91kO1hdHDfffDNTpkwhPT2dhg0bEhMTA8BTTz1FbGwsERER5Ofn07lzZ0aPHg1c+APihRdeID8/37Yrce/evdSqVYu2bduSmZlp++68vb3517/+RWxsLEuWLKGgoIDx48fbdk/GxMQwceJEjDG2E0N+v7usOOrUqcODDz7I6dOnadeuHQ899BDe3t706NEDi8WCt7c3N998M40bN+bIkSOXbPGU5Hv5o7FjxzJ79mwGDBhAYWEht9xyC9HR0QB0797d9rP+4IMPXvH3wJ4WLVrg4eHB7bffbttKGTVqFE8//bRt12qrVq04ePBgkff5+fkRFRXFgw8+SK1atbj99ttt/61r1648+OCDjBo1Cjc3N3x9fXnppZdwc3PjiSeeYNasWSQkJODm5sY///lP/vKXvxR7TCoSN3Ol7XMRFxcaGsq8efP429/+VtZRKrwFCxZw+vRpnn766bKOIhWYdtmJiIhL0BaSiIi4BG0hiYiIS1AhiYiIS1AhiYiIS1AhiYiIS6jw1yGdPp2N1eq6523Uru3LqVOXXudTmWlMLqUxuTyNy6VceUzc3d2oWfPK179V+EKyWo1LFxLg8vnKgsbkUhqTy9O4XKq8jol22YmIiEtQIYmIiEuo8LvsREQcobCwgNOn0ygoyCvrKEWkprpf1f0BHcXT05uaNQPw8Ch+zaiQRERK4PTpNKpUqUb16kGXfWRIWfH0dKegoGwLyRhDdvZZTp9Oo06desV+n3bZiYiUQEFBHtWrX+dSZeQq3NzcqF79uqveelQhiYiUkMroykoyNiokERFxCTqGJCJSCvyuq0oVn9L/JzX3fAFZZ3Pszrdnz25ee+0VFi1aUuoZrmT9+rV8++03PPXUjFJZngpJRKQUVPHxJOLxdaW+3A/n9ier1Jd6bc6fP89rr73C6tXv0qNHaKktV4Vkh6P+6vm9gAA/hy27uH9diUjFU1BQwNy5z3H48CEyMjJo3LgxM2bE8sYbS7FarURGjgMgNnYGHTt2pnXrtjz//CxOnjyJu7s7kZHjaNeuA6++upjvvttHauoJ7r57MPXr/wVjrIwd+yjff7+v1PKqkOxw1F89zuKKf12JiHPs2/cfPD29WLz4QgE9+uhD7NiRwh13RPDYY2MZM2Ys58+f56uvdvHEE5OZNWsGYWH96No1hPT0dMaOfYBly94GIC/vPG+99a5t2e3bd2Tjxg9LNa8KSUSkgmrVqg3XXVeD999fxc8//8QvvxwlJyeH+vX/QlBQPb75Zg8nT56gc+eu+Pj4sHv3lxw5coQlSxYDF7awfv31FwCaN2/h8LwqJBGRCmrbts9YsmQxgwYN4Y47+nHmzBmMuXDj1bCwfmzevImTJ08yatQYAAoLrcyfv5DrrqsBQHp6OjVr1mTr1mR8fHwcnlenfYuIVFC7d39JaGgvwsL64evry9dff4XVWghAz55/56uvdpGRkc6tt17Y+mnbNpjVqy/slvvxx8P84x+DOX8+12l5tYUkIlJB/Oc/39CzZxfb6+bNW7Bnz262bPkIT08v/va3lhw7dgwAH58qtGjxNxo2bGybf8KEJ4mPj+W++4ZgjGHatBiqVbvy84tKm5u5uP1WQZ06de6ang0SEOBX7k9qSEsrX6c1BAT4lbvMjqYxubyyHJcTJ44QFHSj7XVZX4d0UXHuZWeM4bffsomMHMW8ef+idu061xrzsv44Ru7ubtSu7XvF+bWFJCJSCrLO5pSbM1r37/+Oxx9/lJEjH3RYGZWECklEpJJp3rwFiYmflHWMS+ikBhERcQkqJBERcQlOL6Tdu3dz1113ERERwUMPPURmZiYAZ8+eZcyYMVgsFoYNG0ZaWhoAeXl5REVFYbFYGDBgAIcOHXJ2ZBERcQKnF9LkyZOJj4/nww8/pHHjxrz66qsAJCQkEBwcTGJiIoMGDSI2NhaAN998k6pVq5KYmMiUKVOYPHmysyOLiIgTOP2kho0bN+Ll5UV+fj4nT56kadOmACQnJ7N8+XIAwsPDiYmJIT8/n+TkZMaPHw9Au3btyMjI4NixY1x//fXOji4ickU1a3jj6V36dzMoyDvP6cyre/JqeeX0QvLy8uLAgQOMHDkST09PJk6cCEBqaioBAQEXQnl64uvrS0ZGRpHpAAEBAZw4caLYhfRn57xXFo68m7ijlMfMjqYxubyyGpfUVHc8Pf9vJ5Ontw+HY+8u9fU0fOp9PD0L7M731Ve7WbJkMQsX/rtILkf48cfDzJoVQ05ODtddV4Np02ZQr96l/ya7u7tf1ffjsEJKTEwkLi6uyLSGDRuybNkymjZtyvbt21m5ciUTJkxg5cqVl7zfGIO7uzvGmCKPwr04vbhK48LY8q68XVCpi0AvpTG5vLIcF6vVavcC1NJSnPUUFlpt96lzdK74+Djuu280HTt2Zu3a93j55QXMmBF7yXxWq7XI91NmF8ZaLBYsFkuRaefPn2fLli306tULgH79+jF79mwAAgMDSU9PJygoiIKCArKzs/H396du3bqkpqZyww03ABdu9hcYGOio2CIiFYajnof04osv4+npidVq5cSJE/j5XVcqeZ16UoOnpyfPPPMM+/ZdeKBTYmIibdq0ASAkJIS1a9cCF44zBQcH4+XlRUhICOvWXbh1z+7du/Hx8dHxIxGRYvj985DeeWcNWVlZtuchbdnyEcYYcnNz+eqrXXTtGsK8eXMIC+vHa6+9xXPPvcDzz8/it9+ygf97HtKAAQPx9PQkKyuLAQPuYO3a9+jX785SyevUY0geHh68+OKLPP300xQWFlK3bl3b2XTjx48nOjqasLAw/Pz8mDNnDgAjRozg6aefJiwsDG9vb+Lj450ZWUSk3HLk85D8/PxYt24TX3yxnUmTJvLuux/g4eFxTXmdflJDcHAwq1evvmS6v78/ixYtumS6j4+PbbeeiIgUn6Oeh5SUtJnQ0F64ubnRsWNn8vLOk5WVhb+//zXl1Z0aREQqKEc9D2nlyjfZuvVTAPbs2U2NGv7XXEagm6uKiJSKgrzzNHzqfYcst7ic9TykKVNmEB8fy9KlS/D19eXZZ0tnL5YKSUSkFFy4eLXsLmBt0yaYrVu/vKrnIR04cICxY8fbptepE0B8fMIl8z/wQGSR1zfd1JCFC18tneC/o112IiKVzP793zFwYD/69Rug5yGJiEjZ0fOQRERE/oQKSUREXIIKSUREXIIKSUREXIJOahARKQV+/j5U8fIu9eXm5ueRdcb+tUh79uzmtddeYdGiJaWe4Y+ys88xZ85z/PTTYQCio5+madNm17xcFZKISCmo4uXNPe88XOrLXTV4IVkU/+JYZ1iw4EUCA+syffpMvvhiO3PmxPHvf79+zctVIYmIVFCOevxEcvInvPvuBwB07NiZwMC6pZJXx5BERCooRzx+IiSkJ97e3qxZ8y6RkSN59NGHKCwsLJW82kISEamgHPH4icLCQjIyTlG9ui+LFy9l164vmDIlinffXXfNeVVIIiIVlCMeP1Gjhj8eHh707n07AO3adSQn5zdOn86gZs1a15RXu+xERCooRzx+wtvbm3btOpCU9DEA+/btpWrVqtSoocdPiIjI/+esx09ERz/N88/Hsnr1Kjw9PZkxYxbu7te+faNCEhEpBbn5eawavNAhyy0OZz5+ok6dOsye/WKxcl0NFZKISCnIOnPe5a4XupL9+7/j8ccfZeTIB/X4CRERKTt6/ISIiMifUCGJiJTQxVOo5VIlGRsVkohICXh6epOdfValdBnGGLKzz+LpeXU3m9UxJBGREqhZM4DTp9M4d+5MWUcpwt3dHav1z8+ycwZPT29q1gy4uvc4KIuISIXm4eFJnTr1yjrGJQIC/EhLyyrrGCWiXXYiIuISVEgiIuISVEgiIuISnH4M6auvviIuLo78/Hz8/f2ZNWsW9evX58svv+SRRx4hKCgIgObNmxMXF8fZs2d54oknOHr0KLVq1SIhIYGAgKs7UCYiIq7P6VtIUVFRzJw5k3Xr1hEREcHMmTMB2LdvH6NGjWLdunWsW7eOuLg4ABISEggODiYxMZFBgwYRGxvr7MgiIuIETi2kvLw8xo8fT7NmzQBo2rQpx48fB2Dv3r1s27aNiIgIHnroIdv05ORkIiIiAAgPD2fr1q3k5+c7M7aIiDiBU3fZeXt7079/fwCsVisvvfQSvXr1AsDPzw+LxUKfPn1YsWIFEyZMYOXKlaSmptp20Xl6euLr60tGRgZ16xbvGe61a/s65sOUIwEBfmUd4aqVx8yOpjG5PI3LpcrrmDiskBITE2273S5q2LAhy5YtIy8vj+joaAoKCoiMvHBb85iYGNt89957L3PnziUr69Jz6Y0xV/XcjVOnzmG1lvxK6vL6xf5eebsmoTxfR+EoGpPL07hcypXHxN3d7U83EhxWSBaLBYvFcsn07OxsHn74Yfz9/Vm4cCFeXl5YrVYWL17MmDFj8PDwsM3r4eFBYGAg6enpBAUFUVBQQHZ2Nv7+1/5kQhERcS1lclLDjTfeSEJCAt7eF+5z5O7uzubNm/noo48AWLt2LbfddhvVqlUjJCSEtWvXArBx40aCg4Px8vJydmwREXEwpx5D+v7770lKSqJx48YMGDAAgMDAQP79738ze/Zspk2bxssvv0ytWrWIj48HYPz48URHRxMWFoafnx9z5sxxZmQREXESpxZS8+bNOXDgwGX/W5MmTVi5cuUl0/39/Vm0aJGjo4mISBnTzVVFRJzIz9+HKl5X91iGq+XIk7Fy8/PIOuOYR7WrkEREnKiKlzf3vPNwWccosVWDF5KFYwpJ97ITERGXoEISERGXoEISERGXoEISERGXoEISERGXoEISERGXoEISERGXoEISERGXoAtjRUpBzRreeHr7OHQdjn4USkHeeU5n5jl0HSJ/RoUkUgo8vX04HHt3Wce4Jg2feh9QIUnZ0S47ERFxCSokERFxCSokERFxCVc8hhQaGoqbm9sV35iUlOSQQCIiUjldsZDmz58PwNtvv42XlxeDBw/Gw8OD1atXk5+f77SAIiJSOVyxkFq0aAHAf//7X959913b9MmTJzNw4EDHJxMRkUrF7jGks2fPkpGRYXt98uRJzp0759BQIiJS+di9Dum+++4jIiKCrl27YowhJSWFqKgoZ2QTEZFKxG4h9enThzZt2rBjxw4ARo8ezc033+zwYCIiUrnYLaThw4ezadMmmjVr5ow8IiJSSdk9hlS/fn327NmD1Wp1Rh4REamk7G4hHTp0iKFDh+Lp6Ym3tzfGGNzc3NizZ48z8omISCVht5CWL1/ujBwiIlLJFWuXXWZmJsePH+fYsWMcPXqUlJQUZ2QTEZFKxO4W0tSpU0lKSuL8+fMEBgby888/07ZtW+655x5n5BMRkUrC7hbS9u3bSUpKonfv3rzyyissXbqUKlWqOCObiIhUInYLKSAggGrVqtGwYUMOHjxIhw4dOHHiRIlX+NVXXzFw4ED69+/Pfffdx6+//gpcuCPEmDFjsFgsDBs2jLS0NADy8vKIiorCYrEwYMAADh06VOJ1i4iI67JbSF5eXuzatYtGjRqxdetWsrKy+O2330q8wqioKGbOnMm6deuIiIhg5syZACQkJBAcHExiYiKDBg0iNjYWgDfffJOqVauSmJjIlClTmDx5conXLSIirstuIUVFRbFy5UpCQkL44Ycf6NixI/369SvRyvLy8hg/frztItumTZty/PhxAJKTk4mIiAAgPDycrVu3kp+fT3Jysm197dq1IyMjg2PHjpVo/SIi4rrsntRw9uxZZs+ejaenJ6tWrSIrKws/P78Srczb25v+/fsDYLVaeemll+jVqxcAqampBAQEXAjl6Ymvry8ZGRlFpsOFXYgnTpzg+uuvL9Y6a9f2LVHWiiQgoGTfV1kqj5krgvI47uUxc3nnqDG3W0ivvfYakyZNolu3bvTt25euXbsWa8GJiYnExcUVmdawYUOWLVtGXl4e0dHRFBQUEBkZedn3G2Nwd3e3XYj7x+nFderUOaxWU+z5/6gi/LCnpWWVdQRQQKMAABB6SURBVISrEhDgVy4zVwTlcdzLY+byrqRj7u7u9qcbCXYLaenSpZw7d47PPvuMDRs2EBMTQ9u2bZk7d+6fvs9isWCxWC6Znp2dzcMPP4y/vz8LFy7Ey8sLgMDAQNLT0wkKCqKgoIDs7Gz8/f2pW7cuqamp3HDDDQCkp6cTGBhoL7aIiJQzxdrUOHXqFGfOnCEnJ4e8vDzOnDlT4hVGRUVx4403kpCQgLe3t216SEgIa9euBWDjxo0EBwfj5eVFSEgI69atA2D37t34+PgUe3ediIiUH3a3kEJDQ8nPzyc8PJyhQ4fy4osv4uPjU6KVff/99yQlJdG4cWMGDBgAXNgy+ve//8348eOJjo4mLCwMPz8/5syZA8CIESN4+umnCQsLw9vbm/j4+BKtW0REXJvdQho8eDDbtm3j008/5bfffiMnJ4eOHTuW6MSG5s2bc+DAgcv+N39/fxYtWnTJdB8fH2bPnn3V6xIRkfLFbiFFRkYSGRlJdnY2n332GS+++CJHjhzhu+++c0Y+ERGpJOwW0tGjR9m6dSuff/45e/fupV27dowbN84Z2UREpBKxW0j33HMPoaGhDB48mPnz5xc5EUFERKS02C2klJQUTp48ycGDB/Hw8ODYsWM6y01EREqd3dO+P//8c+69916eeeYZTp06RVhYGFu2bHFGNhERqUTsFtJLL73EqlWruO666wgMDOTtt99m/vz5zsgmIiKViN1CKiwsLHJnhFtuuaXIrXxERERKg91Cqlq1KseOHbOV0MW7JYiIiJQmuyc1PP7444waNYq0tDQGDx7MTz/9xIIFC5yRTUREKhG7hXTDDTewatUqvv76a6xWK7fddhu1atVyRjYREalE7BbS8OHD2bRpEyEhIc7IIyIilZTdY0j169dnz549WK1WZ+QREZFKyu4W0qFDhxg6dCienp54e3vbHpi3Z88eZ+QTEZFKwm4hLV++3Bk5RESkkrNbSPXr13dGDhERqeSK9cRYERERR1MhiYiISyhWIeXm5nLgwAGMMeTk5Dg6k4iIVEJ2C+mbb76hV69eREZGcvLkSXr06KEz7EREpNTZLaT4+HiWLVuGv78/QUFBxMfHExsb64xsIiJSidgtpNzcXBo3bmx7HRISQmFhoUNDiYhI5WO3kDw9PcnMzLTd7fvw4cMODyUiIpWP3euQHnroIYYPH056ejoTJ04kJSWFmJgYZ2QTEZFKxG4hhYaG0qhRI1JSUrBarYwbN45GjRo5I5uIiFQidgtp165dADRp0gSA06dPs2/fPho0aICvr69j04mISKVht5Di4uL44YcfaNKkCe7u7hw8eJCAgABycnKIjY2lV69ezsgpIiIVnN2TGq6//nqWLl3KunXrWLNmDStWrKBNmzasW7eOl19+2RkZRUSkErBbSEePHqVDhw621y1btuSnn34iKCjIocFERKRyKdZp39u2bbO93rZtG15eXmRkZFBQUFDiFSckJLBgwQLb6y+//JIOHTrQv39/+vfvz+TJkwE4e/YsY8aMwWKxMGzYMNLS0kq8ThERcV12jyFNnz6dRx99FDc3N6xWKz4+PsyfP58lS5YwZMiQq15hVlYWcXFxbNiwgdGjR9um79u3j1GjRhEZGVlk/oSEBIKDg3nllVdYu3YtsbGxJCQkXPV6RUTEtdktpJYtW5KUlMTBgwfx8PCgUaNGeHh40KxZsxKtMCkpiQYNGjBy5Mgi0/fu3Ut6ejrr16+nfv36TJ8+nXr16pGcnGx7SGB4eDgxMTHk5+fj5eVVovWLiIhrsltIGRkZfPDBB2RnZ2OMwWq1cuTIEebOnVuiFd55550ARXbXAfj5+WGxWOjTpw8rVqxgwoQJrFy5ktTUVAICAi6E9fTE19eXjIwM6tatW6z11a6tU9MDAvzKOsJVK4+ZK4LyOO7lMXN556gxt1tIjz32GFWqVOF///sfnTt3Zvv27bRt29bughMTE4mLiysyrWHDhixbtuyy8//+7g/33nsvc+fOJSsr65L5jDG4uxf/MU6nTp3DajXFnv+PKsIPe1rapePoygIC/Mpl5oqgPI57ecxc3pV0zN3d3f50I8FuIR07dowtW7YwY8YMhgwZwiOPPMLYsWPtrthisWCxWIoV0mq1snjxYsaMGYOHh4dtuoeHB4GBgaSnpxMUFERBQQHZ2dn4+/sXa7kiIlJ+2N3UqFOnDgANGjTg4MGD1K1b95rOrrtsCHd3Nm/ezEcffQTA2rVrue2226hWrRohISGsXbsWgI0bNxIcHKzjRyIiFZDdLaTatWuzZMkSWrVqxYIFC/D19SU3N7fUg8yePZtp06bx8ssvU6tWLeLj4wEYP3480dHRhIWF4efnx5w5c0p93SIiUvbsFlJMTAwbNmwgODiYFi1aMH/+fJ544olrXvEjjzxS5HWTJk1YuXLlJfP5+/uzaNGia16fiIi4NruFNHv2bNvWSlRUFFFRUQ4PJSIilY/dY0j79+/HmJKfpSYiIlIcdreQAgMDCQsL47bbbqN69eq26VOnTnVoMBERqVzsFlLr1q1p3bq1M7KIiEglZreQ/vnPf5Kbm8uRI0do0qQJ58+fp2rVqs7IJiIilYjdY0jffvstvXr1IjIyktTUVHr06MGePXuckU1ERCoRu4U0e/Zsli1bhr+/P0FBQcTHxxMbG+uMbCIiUonYLaTc3FwaN25sex0SEkJhYaFDQ4mISOVTrAf0ZWZm4ubmBsDhw4cdHkpERCofuyc1PPTQQwwfPpz09HQmTpxISkpKkTtzi4iIlAa7hRQaGkqjRo1ISUnBarUybtw4GjVq5IxsIiJSidgtpIkTJ3LPPfcwdOhQZ+QREZFKyu4xpHbt2vHCCy/Qu3dvFi9eTFpamjNyiYhIJWO3kO69915WrVrFokWLyMzMZMiQIYwbN84Z2UREpBIp9rPAc3NzycvLu+pHiIuIiBSH3WNIS5cuZfXq1eTl5TFw4EBWrVple4qsiIhIabG7qbNv3z6mTp3KRx99xKhRo9i5cyeDBg1yRjYREalE7G4hzZ07l8zMTF555RWWL19OdnY2I0aMcEY2ERGpRP60kA4fPswbb7zBunXrqF+/Prm5uXz66af4+fk5K5+IiFQSV9xlN2bMGIYPH46npydvvPEG69evp3r16iojERFxiCsW0vfff8+tt95KkyZNuPHGGwFs97MTEREpbVcspOTkZAYMGMD69evp2rUrjz76KOfPn3dmNhERqUSuWEienp7ccccdvPnmm6xevZrAwEDOnz9Pnz59WLFihTMziohIJVCsK1wbN27M1KlT2bp1Kw888ACrVq1ydC4REalkruqWC1WrVmXw4MGsWbPGUXlERKSS0j2ARETEJaiQRETEJZRZISUkJLBgwQLb67NnzzJmzBgsFgvDhg2zPeYiLy+PqKgoLBYLAwYM4NChQ2UVWUREHMjphZSVlcWUKVNYunRpkekJCQkEBweTmJjIoEGDiI2NBeDNN9+katWqJCYmMmXKFCZPnuzsyCIi4gROL6SkpCQaNGjAyJEji0xPTk4mIiICgPDwcLZu3Up+fj7Jycn069cPuPCwwIyMDI4dO+bs2CIi4mBOL6Q777yTMWPG4OHhUWR6amoqAQEBwIVroHx9fcnIyCgyHSAgIIATJ044NbOIiDie3bt9l1RiYiJxcXFFpjVs2JBly5YV6/0XHwRojClyy6KrfUBg7dq+xZ63ogoIKH/3HyyPmSuC8jju5TFzeeeoMXdYIVksFiwWS7HnDwwMJD09naCgIAoKCsjOzsbf35+6deuSmprKDTfcAEB6ejqBgYHFXu6pU+ewWs1V57+oIvywp6VllXWEqxIQ4FcuM1cE5XHcy2Pm8q6kY+7u7vanGwkuc9p3SEgIa9euBWDjxo0EBwfj5eVFSEgI69atA2D37t34+Phw/fXXl2VUERFxAIdtIV2t8ePHEx0dTVhYGH5+fsyZMweAESNG8PTTTxMWFoa3tzfx8fFlnFRERByhzArpkUceKfLa39+fRYsWXTKfj48Ps2fPdlYsEREpIy6zy05ERCo3FZKIiLgEFZKIiLgEFZKIiLgEFZKIiLgEFZKIiLgEFZKIiLgEFZKIiLgEFZKIiLgEFZKIiLgEFZKIiLgEFZKIiLgEFZKIiLgEFZKIiLgEFZKIiLgEFZKIiLgEFZKIiLgEFZKIiLgEFZKIiLgEFZKIiLgEFZKIiLgEFZKIiLgEFZKIiLgEFZKIiLgEFZKIiLgEFZKIiLgEFZKIiLgEFZKIiLgEz7IOcNGaNWuYO3cutWvXBqBHjx5MmDCBY8eOERUVxalTp7jpppuYM2cO1atXL+O0IiJS2lymkPbt20d0dDTh4eFFpj/zzDMMHTqUsLAwXn75Zf71r38RFRVVRilFRMRRXGaX3d69e1mzZg0RERE88cQTZGZmkp+fz65du+jbty8Ad911F5s2bSrjpCIi4gguU0gBAQGMHTuWDz74gHr16hETE8Pp06fx9fXF09PTNs/JkyfLOKmIiDiC03fZJSYmEhcXV2Raw4YNWbZsme316NGj6d27N08++SRubm5F5v3ja3tq1/YtcdaKIiDAr6wjXLXymLkiKI/jXh4zl3eOGnOnF5LFYsFisRSZlpWVxbJly7j//vsBMMbg4eFBrVq1yMrKorCwEA8PD9LS0ggMDLyq9Z06dQ6r1ZQ4b0X4YU9LyyrrCFclIMCvXGauCMrjuJfHzOVdScfc3d3tTzcSXGKXXbVq1ViyZAnffvstAG+99Ra9e/fGy8uL4OBgNm7cCMDatWvp3r17WUYVEREHcYmz7Dw8PEhISGDGjBnk5ubSoEED4uPjAZg+fTrR0dEsXLiQevXq8cILL5RxWhERcQSXKCSA4OBg1qxZc8n0+vXr8+abb5ZBIhERcSaX2GUnIiKiQhIREZegQhIREZegQhIREZegQhIREZegQhIREZegQhIREZegQhIREZegQhIREZegQhIREZegQhIREZegQhIREZegQhIREZegQhIREZegQhIREZegQhIREZegQhIREZegQhIREZegQhIREZegQhIREZegQhIREZegQhIREZegQhIREZegQhIREZegQhIREZegQhIREZegQhIREZegQhIREZdQLgrpww8/5I477qBPnz4sX768rOOIiIgDeJZ1AHtOnjzJiy++yOrVq/H29mbIkCF06NCBxo0bl3U0EREpRS6/hbR9+3Y6duyIv78/1apVo2/fvmzatKmsY4mISClz+S2k1NRUAgICbK8DAwP5z3/+U+z3u7u7XXOGwJpVr3kZZak0xsDZymNmzxoB9mdyceVx3Mtj5oBqtco6wjUp6Zjbe5/LF5LVasXN7f8+hDGmyGt7atasfs0ZXp3a55qXUZZq1/Yt6whXrTxmvuGfi8o6wjUrj+NeHjO/HBFb1hGuiaPG3OV32QUFBZGWlmZ7nZaWRmBgYBkmEhERR3D5QurcuTM7duwgIyODnJwcPv74Y7p3717WsUREpJS5/C67unXrMmHCBP7xj3+Qn5/PwIEDadmyZVnHEhGRUuZmjDFlHUJERMTld9mJiEjloEISERGXoEISERGXoEISERGXoEISERGXoEISERGXoEISERGXoEISERGX8P8AkrsSQWGj/kkAAAAASUVORK5CYII=\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "plot_difference_2([number1[2],number2[2],number3[0]])"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Average performance CartPole"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 21,
   "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>0</th>\n",
       "      <th>1</th>\n",
       "      <th>2</th>\n",
       "      <th>3</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>params_custom_layer_init</th>\n",
       "      <td>(0.5)</td>\n",
       "      <td>93.654166</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>params_custom_lr_schedule</th>\n",
       "      <td>(Transformer, No)</td>\n",
       "      <td>86.466666</td>\n",
       "      <td>100.841666</td>\n",
       "      <td>NaN</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>params_layer_type</th>\n",
       "      <td>(6, 1, 3)</td>\n",
       "      <td>121.595833</td>\n",
       "      <td>76.550000</td>\n",
       "      <td>82.816666</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>params_loss_function</th>\n",
       "      <td>(element_wise_squared_loss, element_wise_huber...</td>\n",
       "      <td>122.062500</td>\n",
       "      <td>36.837500</td>\n",
       "      <td>NaN</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>params_target_update_period</th>\n",
       "      <td>(5)</td>\n",
       "      <td>93.654166</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>params_num_heads</th>\n",
       "      <td>(4, 2)</td>\n",
       "      <td>92.672221</td>\n",
       "      <td>94.636112</td>\n",
       "      <td>NaN</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "                                                                             0  \\\n",
       "params_custom_layer_init                                                 (0.5)   \n",
       "params_custom_lr_schedule                                    (Transformer, No)   \n",
       "params_layer_type                                                    (6, 1, 3)   \n",
       "params_loss_function         (element_wise_squared_loss, element_wise_huber...   \n",
       "params_target_update_period                                                (5)   \n",
       "params_num_heads                                                        (4, 2)   \n",
       "\n",
       "                                      1           2          3  \n",
       "params_custom_layer_init      93.654166         NaN        NaN  \n",
       "params_custom_lr_schedule     86.466666  100.841666        NaN  \n",
       "params_layer_type            121.595833   76.550000  82.816666  \n",
       "params_loss_function         122.062500   36.837500        NaN  \n",
       "params_target_update_period   93.654166         NaN        NaN  \n",
       "params_num_heads              92.672221   94.636112        NaN  "
      ]
     },
     "execution_count": 21,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "X =  sumTable7.values[:,1:]\n",
    "y =  sumTable7.values[:,0]\n",
    "y = y *-1\n",
    "features = sumTable7.columns[1:]\n",
    "final = []\n",
    "for w in range(len(features)):\n",
    "    summary = {}\n",
    "    for x in range(len(X)):\n",
    "        if str(X[x][w]) in summary: \n",
    "            summary[str(X[x][w])].append(y[x])\n",
    "        else:\n",
    "            summary[str(X[x][w])] = [y[x]]\n",
    "    row = []\n",
    "    row.append(summary.keys())\n",
    "    for x in summary.keys():\n",
    "        row.append((np.mean(summary[x])))\n",
    "    final.append(row)\n",
    "pd.DataFrame(final, index = features)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 22,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "number1 = []\n",
    "number2 = []\n",
    "number3 = []\n",
    "for row in final: \n",
    "    number1.append(row[1])\n",
    "    if len(row) >= 3:\n",
    "        number2.append(row[2])\n",
    "    else: \n",
    "        number2.append(0)\n",
    "    if len(row) == 4:\n",
    "        number3.append(row[3])  \n",
    "values = np.array([number1,number2]).T\n",
    "values = np.concatenate((values[:2], values[3:]))\n",
    "labels = [\"init\",\"schedule\",\"loss\",\"update\",\"heads\"]\n",
    "sns.set()\n",
    "plot_difference(labels,values[:,0],values[:,1])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 399,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 640x480 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "plot_difference_2([number1[2],number2[2],number3[0]])"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Average performance LunarLander"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 409,
   "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>0</th>\n",
       "      <th>1</th>\n",
       "      <th>2</th>\n",
       "      <th>3</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>params_custom_layer_init</th>\n",
       "      <td>(0.5, 1.0)</td>\n",
       "      <td>-398.553477</td>\n",
       "      <td>-384.767380</td>\n",
       "      <td>NaN</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>params_custom_lr_schedule</th>\n",
       "      <td>(No, Transformer)</td>\n",
       "      <td>-266.192350</td>\n",
       "      <td>-517.128507</td>\n",
       "      <td>NaN</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>params_layer_type</th>\n",
       "      <td>(6, 3, 1)</td>\n",
       "      <td>-509.182407</td>\n",
       "      <td>-258.685053</td>\n",
       "      <td>-407.113825</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>params_loss_function</th>\n",
       "      <td>(element_wise_huber_loss)</td>\n",
       "      <td>-391.660428</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>params_target_update_period</th>\n",
       "      <td>(5, 10)</td>\n",
       "      <td>-417.154460</td>\n",
       "      <td>-366.166397</td>\n",
       "      <td>NaN</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>params_num_heads</th>\n",
       "      <td>(4)</td>\n",
       "      <td>-391.660428</td>\n",
       "      <td>NaN</td>\n",
       "      <td>NaN</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "                                                     0           1  \\\n",
       "params_custom_layer_init                    (0.5, 1.0) -398.553477   \n",
       "params_custom_lr_schedule            (No, Transformer) -266.192350   \n",
       "params_layer_type                            (6, 3, 1) -509.182407   \n",
       "params_loss_function         (element_wise_huber_loss) -391.660428   \n",
       "params_target_update_period                    (5, 10) -417.154460   \n",
       "params_num_heads                                   (4) -391.660428   \n",
       "\n",
       "                                      2           3  \n",
       "params_custom_layer_init    -384.767380         NaN  \n",
       "params_custom_lr_schedule   -517.128507         NaN  \n",
       "params_layer_type           -258.685053 -407.113825  \n",
       "params_loss_function                NaN         NaN  \n",
       "params_target_update_period -366.166397         NaN  \n",
       "params_num_heads                    NaN         NaN  "
      ]
     },
     "execution_count": 409,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "X =  sumTable8.values[:,1:]\n",
    "y =  sumTable8.values[:,0]\n",
    "y = y *-1\n",
    "features = sumTable7.columns[1:]\n",
    "final = []\n",
    "for w in range(len(features)):\n",
    "    summary = {}\n",
    "    for x in range(len(X)):\n",
    "        if str(X[x][w]) in summary: \n",
    "            summary[str(X[x][w])].append(y[x])\n",
    "        else:\n",
    "            summary[str(X[x][w])] = [y[x]]\n",
    "    row = []\n",
    "    row.append(summary.keys())\n",
    "    for x in summary.keys():\n",
    "        row.append((np.mean(summary[x])))\n",
    "    final.append(row)\n",
    "pd.DataFrame(final, index = features)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 413,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 640x480 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "number1 = []\n",
    "number2 = []\n",
    "number3 = []\n",
    "for row in final: \n",
    "    number1.append(row[1])\n",
    "    if len(row) >= 3:\n",
    "        number2.append(row[2])\n",
    "    else: \n",
    "        number2.append(0)\n",
    "    if len(row) == 4:\n",
    "        number3.append(row[3])  \n",
    "values = np.array([number1,number2]).T\n",
    "values = np.concatenate((values[:2], values[3:]))\n",
    "labels = [\"init\",\"schedule\",\"loss\",\"update\",\"heads\"]\n",
    "sns.set()\n",
    "plot_difference(labels,values[:,0],values[:,1])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 415,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 640x480 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "plot_difference_2([number1[2],number3[0],number2[2]])"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Top runs"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 20,
   "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>value</th>\n",
       "      <th>params_custom_layer_init</th>\n",
       "      <th>params_custom_lr_schedule</th>\n",
       "      <th>params_layer_type</th>\n",
       "      <th>params_loss_function</th>\n",
       "      <th>params_target_update_period</th>\n",
       "      <th>params_num_heads</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>168.699997</td>\n",
       "      <td>0.5</td>\n",
       "      <td>No</td>\n",
       "      <td>6</td>\n",
       "      <td>element_wise_squared_loss</td>\n",
       "      <td>5</td>\n",
       "      <td>4</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>151.750000</td>\n",
       "      <td>1.0</td>\n",
       "      <td>No</td>\n",
       "      <td>6</td>\n",
       "      <td>element_wise_squared_loss</td>\n",
       "      <td>5</td>\n",
       "      <td>4</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>164.799988</td>\n",
       "      <td>0.5</td>\n",
       "      <td>No</td>\n",
       "      <td>6</td>\n",
       "      <td>element_wise_squared_loss</td>\n",
       "      <td>5</td>\n",
       "      <td>4</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>12</th>\n",
       "      <td>152.049988</td>\n",
       "      <td>1.0</td>\n",
       "      <td>No</td>\n",
       "      <td>6</td>\n",
       "      <td>element_wise_squared_loss</td>\n",
       "      <td>5</td>\n",
       "      <td>4</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>21</th>\n",
       "      <td>170.050003</td>\n",
       "      <td>0.5</td>\n",
       "      <td>No</td>\n",
       "      <td>1</td>\n",
       "      <td>element_wise_squared_loss</td>\n",
       "      <td>5</td>\n",
       "      <td>4</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>26</th>\n",
       "      <td>173.250000</td>\n",
       "      <td>1.0</td>\n",
       "      <td>No</td>\n",
       "      <td>3</td>\n",
       "      <td>element_wise_huber_loss</td>\n",
       "      <td>5</td>\n",
       "      <td>4</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>27</th>\n",
       "      <td>142.100006</td>\n",
       "      <td>0.5</td>\n",
       "      <td>No</td>\n",
       "      <td>1</td>\n",
       "      <td>element_wise_huber_loss</td>\n",
       "      <td>5</td>\n",
       "      <td>4</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>34</th>\n",
       "      <td>164.250000</td>\n",
       "      <td>0.5</td>\n",
       "      <td>No</td>\n",
       "      <td>3</td>\n",
       "      <td>element_wise_huber_loss</td>\n",
       "      <td>5</td>\n",
       "      <td>4</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "         value  params_custom_layer_init params_custom_lr_schedule  \\\n",
       "0   168.699997                       0.5                        No   \n",
       "1   151.750000                       1.0                        No   \n",
       "3   164.799988                       0.5                        No   \n",
       "12  152.049988                       1.0                        No   \n",
       "21  170.050003                       0.5                        No   \n",
       "26  173.250000                       1.0                        No   \n",
       "27  142.100006                       0.5                        No   \n",
       "34  164.250000                       0.5                        No   \n",
       "\n",
       "    params_layer_type       params_loss_function  params_target_update_period  \\\n",
       "0                   6  element_wise_squared_loss                            5   \n",
       "1                   6  element_wise_squared_loss                            5   \n",
       "3                   6  element_wise_squared_loss                            5   \n",
       "12                  6  element_wise_squared_loss                            5   \n",
       "21                  1  element_wise_squared_loss                            5   \n",
       "26                  3    element_wise_huber_loss                            5   \n",
       "27                  1    element_wise_huber_loss                            5   \n",
       "34                  3    element_wise_huber_loss                            5   \n",
       "\n",
       "    params_num_heads  \n",
       "0                  4  \n",
       "1                  4  \n",
       "3                  4  \n",
       "12                 4  \n",
       "21                 4  \n",
       "26                 4  \n",
       "27                 4  \n",
       "34                 4  "
      ]
     },
     "execution_count": 20,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# Acro\n",
    "sumTable6.loc[sumTable6['value']< 175]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 22,
   "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>value</th>\n",
       "      <th>params_custom_layer_init</th>\n",
       "      <th>params_custom_lr_schedule</th>\n",
       "      <th>params_layer_type</th>\n",
       "      <th>params_loss_function</th>\n",
       "      <th>params_target_update_period</th>\n",
       "      <th>params_num_heads</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>-164.300003</td>\n",
       "      <td>0.5</td>\n",
       "      <td>No</td>\n",
       "      <td>6</td>\n",
       "      <td>element_wise_squared_loss</td>\n",
       "      <td>5</td>\n",
       "      <td>2</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>11</th>\n",
       "      <td>-183.949997</td>\n",
       "      <td>0.5</td>\n",
       "      <td>No</td>\n",
       "      <td>3</td>\n",
       "      <td>element_wise_squared_loss</td>\n",
       "      <td>5</td>\n",
       "      <td>4</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>14</th>\n",
       "      <td>-216.699997</td>\n",
       "      <td>0.5</td>\n",
       "      <td>No</td>\n",
       "      <td>3</td>\n",
       "      <td>element_wise_squared_loss</td>\n",
       "      <td>5</td>\n",
       "      <td>4</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>15</th>\n",
       "      <td>-197.850006</td>\n",
       "      <td>0.5</td>\n",
       "      <td>Transformer</td>\n",
       "      <td>6</td>\n",
       "      <td>element_wise_squared_loss</td>\n",
       "      <td>5</td>\n",
       "      <td>2</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>20</th>\n",
       "      <td>-160.199997</td>\n",
       "      <td>0.5</td>\n",
       "      <td>No</td>\n",
       "      <td>6</td>\n",
       "      <td>element_wise_squared_loss</td>\n",
       "      <td>5</td>\n",
       "      <td>4</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "         value  params_custom_layer_init params_custom_lr_schedule  \\\n",
       "1  -164.300003                       0.5                        No   \n",
       "11 -183.949997                       0.5                        No   \n",
       "14 -216.699997                       0.5                        No   \n",
       "15 -197.850006                       0.5               Transformer   \n",
       "20 -160.199997                       0.5                        No   \n",
       "\n",
       "    params_layer_type       params_loss_function  params_target_update_period  \\\n",
       "1                   6  element_wise_squared_loss                            5   \n",
       "11                  3  element_wise_squared_loss                            5   \n",
       "14                  3  element_wise_squared_loss                            5   \n",
       "15                  6  element_wise_squared_loss                            5   \n",
       "20                  6  element_wise_squared_loss                            5   \n",
       "\n",
       "    params_num_heads  \n",
       "1                  2  \n",
       "11                 4  \n",
       "14                 4  \n",
       "15                 2  \n",
       "20                 4  "
      ]
     },
     "execution_count": 22,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "#CartPole\n",
    "sumTable7.loc[sumTable7['value']< -160]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "metadata": {
    "scrolled": true
   },
   "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>value</th>\n",
       "      <th>params_custom_layer_init</th>\n",
       "      <th>params_custom_lr_schedule</th>\n",
       "      <th>params_layer_type</th>\n",
       "      <th>params_loss_function</th>\n",
       "      <th>params_target_update_period</th>\n",
       "      <th>params_num_heads</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>-250.870392</td>\n",
       "      <td>1.0</td>\n",
       "      <td>No</td>\n",
       "      <td>3</td>\n",
       "      <td>element_wise_huber_loss</td>\n",
       "      <td>10</td>\n",
       "      <td>4</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>15</th>\n",
       "      <td>-204.855835</td>\n",
       "      <td>1.0</td>\n",
       "      <td>No</td>\n",
       "      <td>3</td>\n",
       "      <td>element_wise_huber_loss</td>\n",
       "      <td>5</td>\n",
       "      <td>4</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>22</th>\n",
       "      <td>-128.607437</td>\n",
       "      <td>0.5</td>\n",
       "      <td>No</td>\n",
       "      <td>1</td>\n",
       "      <td>element_wise_huber_loss</td>\n",
       "      <td>10</td>\n",
       "      <td>4</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "         value  params_custom_layer_init params_custom_lr_schedule  \\\n",
       "1  -250.870392                       1.0                        No   \n",
       "15 -204.855835                       1.0                        No   \n",
       "22 -128.607437                       0.5                        No   \n",
       "\n",
       "    params_layer_type     params_loss_function  params_target_update_period  \\\n",
       "1                   3  element_wise_huber_loss                           10   \n",
       "15                  3  element_wise_huber_loss                            5   \n",
       "22                  1  element_wise_huber_loss                           10   \n",
       "\n",
       "    params_num_heads  \n",
       "1                  4  \n",
       "15                 4  \n",
       "22                 4  "
      ]
     },
     "execution_count": 17,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# LunarLander\n",
    "sumTable8.loc[sumTable8['value']< 140]"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Cross studies"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {},
   "outputs": [],
   "source": [
    "study = pd.read_csv(\"Acro_study_cross.csv\") \n",
    "study2 = pd.read_csv(\"CartPole_study_cross.csv\") \n",
    "study3 = pd.read_csv(\"Lunar_study_cross.csv\") \n",
    "study_1 = pd.read_csv(\"Acro_study_cross_2.csv\") "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {},
   "outputs": [],
   "source": [
    "study = study [['value', 'params_custom_layer_init', 'params_custom_lr_schedule',\n",
    "       'params_layer_type', 'params_loss_function',\n",
    "       'params_target_update_period','params_num_heads' ]]\n",
    "study2 = study2 [['value', 'params_custom_layer_init', 'params_custom_lr_schedule',\n",
    "       'params_layer_type', 'params_loss_function',\n",
    "       'params_target_update_period','params_num_heads' ]]\n",
    "study3 = study3 [['value', 'params_custom_layer_init', 'params_custom_lr_schedule',\n",
    "       'params_layer_type', 'params_loss_function',\n",
    "       'params_target_update_period','params_num_heads' ]]\n",
    "study_1 = study_1 [['value', 'params_custom_layer_init', 'params_custom_lr_schedule',\n",
    "       'params_layer_type', 'params_loss_function',\n",
    "       'params_target_update_period','params_num_heads' ]]"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Acro cross study"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 40,
   "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>value</th>\n",
       "      <th>params_custom_layer_init</th>\n",
       "      <th>params_custom_lr_schedule</th>\n",
       "      <th>params_layer_type</th>\n",
       "      <th>params_loss_function</th>\n",
       "      <th>params_target_update_period</th>\n",
       "      <th>params_num_heads</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>154.549988</td>\n",
       "      <td>0.5</td>\n",
       "      <td>No</td>\n",
       "      <td>3</td>\n",
       "      <td>element_wise_huber_loss</td>\n",
       "      <td>5</td>\n",
       "      <td>2</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>172.850006</td>\n",
       "      <td>0.5</td>\n",
       "      <td>No</td>\n",
       "      <td>3</td>\n",
       "      <td>element_wise_huber_loss</td>\n",
       "      <td>10</td>\n",
       "      <td>2</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>147.350006</td>\n",
       "      <td>0.5</td>\n",
       "      <td>No</td>\n",
       "      <td>1</td>\n",
       "      <td>element_wise_huber_loss</td>\n",
       "      <td>10</td>\n",
       "      <td>2</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>155.250000</td>\n",
       "      <td>0.5</td>\n",
       "      <td>No</td>\n",
       "      <td>1</td>\n",
       "      <td>element_wise_huber_loss</td>\n",
       "      <td>10</td>\n",
       "      <td>4</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>143.649994</td>\n",
       "      <td>0.5</td>\n",
       "      <td>No</td>\n",
       "      <td>1</td>\n",
       "      <td>element_wise_huber_loss</td>\n",
       "      <td>5</td>\n",
       "      <td>4</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>5</th>\n",
       "      <td>168.700012</td>\n",
       "      <td>0.5</td>\n",
       "      <td>No</td>\n",
       "      <td>3</td>\n",
       "      <td>element_wise_huber_loss</td>\n",
       "      <td>10</td>\n",
       "      <td>4</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>6</th>\n",
       "      <td>138.549988</td>\n",
       "      <td>0.5</td>\n",
       "      <td>No</td>\n",
       "      <td>1</td>\n",
       "      <td>element_wise_huber_loss</td>\n",
       "      <td>5</td>\n",
       "      <td>2</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>7</th>\n",
       "      <td>202.399994</td>\n",
       "      <td>0.5</td>\n",
       "      <td>No</td>\n",
       "      <td>3</td>\n",
       "      <td>element_wise_huber_loss</td>\n",
       "      <td>5</td>\n",
       "      <td>4</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "        value  params_custom_layer_init params_custom_lr_schedule  \\\n",
       "0  154.549988                       0.5                        No   \n",
       "1  172.850006                       0.5                        No   \n",
       "2  147.350006                       0.5                        No   \n",
       "3  155.250000                       0.5                        No   \n",
       "4  143.649994                       0.5                        No   \n",
       "5  168.700012                       0.5                        No   \n",
       "6  138.549988                       0.5                        No   \n",
       "7  202.399994                       0.5                        No   \n",
       "\n",
       "   params_layer_type     params_loss_function  params_target_update_period  \\\n",
       "0                  3  element_wise_huber_loss                            5   \n",
       "1                  3  element_wise_huber_loss                           10   \n",
       "2                  1  element_wise_huber_loss                           10   \n",
       "3                  1  element_wise_huber_loss                           10   \n",
       "4                  1  element_wise_huber_loss                            5   \n",
       "5                  3  element_wise_huber_loss                           10   \n",
       "6                  1  element_wise_huber_loss                            5   \n",
       "7                  3  element_wise_huber_loss                            5   \n",
       "\n",
       "   params_num_heads  \n",
       "0                 2  \n",
       "1                 2  \n",
       "2                 2  \n",
       "3                 4  \n",
       "4                 4  \n",
       "5                 4  \n",
       "6                 2  \n",
       "7                 4  "
      ]
     },
     "execution_count": 40,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "study"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 41,
   "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>0</th>\n",
       "      <th>1</th>\n",
       "      <th>2</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>params_custom_layer_init</th>\n",
       "      <td>(0.5)</td>\n",
       "      <td>-160.412498</td>\n",
       "      <td>NaN</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>params_custom_lr_schedule</th>\n",
       "      <td>(No)</td>\n",
       "      <td>-160.412498</td>\n",
       "      <td>NaN</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>params_layer_type</th>\n",
       "      <td>(3, 1)</td>\n",
       "      <td>-174.625000</td>\n",
       "      <td>-146.199997</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>params_loss_function</th>\n",
       "      <td>(element_wise_huber_loss)</td>\n",
       "      <td>-160.412498</td>\n",
       "      <td>NaN</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>params_target_update_period</th>\n",
       "      <td>(5, 10)</td>\n",
       "      <td>-159.787491</td>\n",
       "      <td>-161.037506</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>params_num_heads</th>\n",
       "      <td>(2, 4)</td>\n",
       "      <td>-153.324997</td>\n",
       "      <td>-167.500000</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "                                                     0           1           2\n",
       "params_custom_layer_init                         (0.5) -160.412498         NaN\n",
       "params_custom_lr_schedule                         (No) -160.412498         NaN\n",
       "params_layer_type                               (3, 1) -174.625000 -146.199997\n",
       "params_loss_function         (element_wise_huber_loss) -160.412498         NaN\n",
       "params_target_update_period                    (5, 10) -159.787491 -161.037506\n",
       "params_num_heads                                (2, 4) -153.324997 -167.500000"
      ]
     },
     "execution_count": 41,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "X =  study.values[:,1:]\n",
    "y =  study.values[:,0]\n",
    "y = y *-1\n",
    "features = study.columns[1:]\n",
    "final = []\n",
    "for w in range(len(features)):\n",
    "    summary = {}\n",
    "    for x in range(len(X)):\n",
    "        if str(X[x][w]) in summary: \n",
    "            summary[str(X[x][w])].append(y[x])\n",
    "        else:\n",
    "            summary[str(X[x][w])] = [y[x]]\n",
    "    row = []\n",
    "    row.append(summary.keys())\n",
    "    for x in summary.keys():\n",
    "        row.append((np.mean(summary[x])))\n",
    "    final.append(row)\n",
    "pd.DataFrame(final, index = features)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 43,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "number1 = []\n",
    "number2 = []\n",
    "number3 = []\n",
    "for row in final: \n",
    "    number1.append(row[1])\n",
    "    if len(row) >= 3:\n",
    "        number2.append(row[2])\n",
    "    else: \n",
    "        number2.append(0)\n",
    "    if len(row) == 4:\n",
    "        number3.append(row[3])  \n",
    "values = np.array([number1,number2]).T\n",
    "labels = [\"init\",\"schedule\",\"layer_type\",\"loss\",\"update\",\"heads\"]\n",
    "sns.set()\n",
    "plot_difference(labels,values[:,0],values[:,1])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "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>value</th>\n",
       "      <th>params_custom_layer_init</th>\n",
       "      <th>params_custom_lr_schedule</th>\n",
       "      <th>params_layer_type</th>\n",
       "      <th>params_loss_function</th>\n",
       "      <th>params_target_update_period</th>\n",
       "      <th>params_num_heads</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>160.800003</td>\n",
       "      <td>1</td>\n",
       "      <td>No</td>\n",
       "      <td>6</td>\n",
       "      <td>element_wise_squared_loss</td>\n",
       "      <td>10</td>\n",
       "      <td>2</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>160.799988</td>\n",
       "      <td>1</td>\n",
       "      <td>No</td>\n",
       "      <td>6</td>\n",
       "      <td>element_wise_squared_loss</td>\n",
       "      <td>5</td>\n",
       "      <td>2</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>148.250000</td>\n",
       "      <td>1</td>\n",
       "      <td>No</td>\n",
       "      <td>6</td>\n",
       "      <td>element_wise_squared_loss</td>\n",
       "      <td>10</td>\n",
       "      <td>4</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>167.199997</td>\n",
       "      <td>1</td>\n",
       "      <td>No</td>\n",
       "      <td>6</td>\n",
       "      <td>element_wise_squared_loss</td>\n",
       "      <td>5</td>\n",
       "      <td>4</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "        value  params_custom_layer_init params_custom_lr_schedule  \\\n",
       "0  160.800003                         1                        No   \n",
       "1  160.799988                         1                        No   \n",
       "2  148.250000                         1                        No   \n",
       "3  167.199997                         1                        No   \n",
       "\n",
       "   params_layer_type       params_loss_function  params_target_update_period  \\\n",
       "0                  6  element_wise_squared_loss                           10   \n",
       "1                  6  element_wise_squared_loss                            5   \n",
       "2                  6  element_wise_squared_loss                           10   \n",
       "3                  6  element_wise_squared_loss                            5   \n",
       "\n",
       "   params_num_heads  \n",
       "0                 2  \n",
       "1                 2  \n",
       "2                 4  \n",
       "3                 4  "
      ]
     },
     "execution_count": 9,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "study_1  #nothing special"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# CartPole cross study"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 34,
   "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>value</th>\n",
       "      <th>params_custom_layer_init</th>\n",
       "      <th>params_custom_lr_schedule</th>\n",
       "      <th>params_layer_type</th>\n",
       "      <th>params_loss_function</th>\n",
       "      <th>params_target_update_period</th>\n",
       "      <th>params_num_heads</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>-105.449997</td>\n",
       "      <td>1.0</td>\n",
       "      <td>No</td>\n",
       "      <td>6</td>\n",
       "      <td>element_wise_squared_loss</td>\n",
       "      <td>10</td>\n",
       "      <td>4</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>-9.550000</td>\n",
       "      <td>0.5</td>\n",
       "      <td>No</td>\n",
       "      <td>6</td>\n",
       "      <td>element_wise_squared_loss</td>\n",
       "      <td>5</td>\n",
       "      <td>4</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>-196.100006</td>\n",
       "      <td>0.5</td>\n",
       "      <td>No</td>\n",
       "      <td>3</td>\n",
       "      <td>element_wise_squared_loss</td>\n",
       "      <td>10</td>\n",
       "      <td>4</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>-201.500000</td>\n",
       "      <td>1.0</td>\n",
       "      <td>No</td>\n",
       "      <td>3</td>\n",
       "      <td>element_wise_squared_loss</td>\n",
       "      <td>10</td>\n",
       "      <td>4</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>-153.449997</td>\n",
       "      <td>1.0</td>\n",
       "      <td>No</td>\n",
       "      <td>6</td>\n",
       "      <td>element_wise_squared_loss</td>\n",
       "      <td>5</td>\n",
       "      <td>4</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>5</th>\n",
       "      <td>-160.899994</td>\n",
       "      <td>1.0</td>\n",
       "      <td>No</td>\n",
       "      <td>3</td>\n",
       "      <td>element_wise_squared_loss</td>\n",
       "      <td>5</td>\n",
       "      <td>4</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>6</th>\n",
       "      <td>-177.699997</td>\n",
       "      <td>0.5</td>\n",
       "      <td>No</td>\n",
       "      <td>3</td>\n",
       "      <td>element_wise_squared_loss</td>\n",
       "      <td>5</td>\n",
       "      <td>4</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>7</th>\n",
       "      <td>-158.350006</td>\n",
       "      <td>0.5</td>\n",
       "      <td>No</td>\n",
       "      <td>6</td>\n",
       "      <td>element_wise_squared_loss</td>\n",
       "      <td>10</td>\n",
       "      <td>4</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "        value  params_custom_layer_init params_custom_lr_schedule  \\\n",
       "0 -105.449997                       1.0                        No   \n",
       "1   -9.550000                       0.5                        No   \n",
       "2 -196.100006                       0.5                        No   \n",
       "3 -201.500000                       1.0                        No   \n",
       "4 -153.449997                       1.0                        No   \n",
       "5 -160.899994                       1.0                        No   \n",
       "6 -177.699997                       0.5                        No   \n",
       "7 -158.350006                       0.5                        No   \n",
       "\n",
       "   params_layer_type       params_loss_function  params_target_update_period  \\\n",
       "0                  6  element_wise_squared_loss                           10   \n",
       "1                  6  element_wise_squared_loss                            5   \n",
       "2                  3  element_wise_squared_loss                           10   \n",
       "3                  3  element_wise_squared_loss                           10   \n",
       "4                  6  element_wise_squared_loss                            5   \n",
       "5                  3  element_wise_squared_loss                            5   \n",
       "6                  3  element_wise_squared_loss                            5   \n",
       "7                  6  element_wise_squared_loss                           10   \n",
       "\n",
       "   params_num_heads  \n",
       "0                 4  \n",
       "1                 4  \n",
       "2                 4  \n",
       "3                 4  \n",
       "4                 4  \n",
       "5                 4  \n",
       "6                 4  \n",
       "7                 4  "
      ]
     },
     "execution_count": 34,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "study2"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 35,
   "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>0</th>\n",
       "      <th>1</th>\n",
       "      <th>2</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>params_custom_layer_init</th>\n",
       "      <td>(1.0, 0.5)</td>\n",
       "      <td>155.324997</td>\n",
       "      <td>135.425002</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>params_custom_lr_schedule</th>\n",
       "      <td>(No)</td>\n",
       "      <td>145.375000</td>\n",
       "      <td>NaN</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>params_layer_type</th>\n",
       "      <td>(6, 3)</td>\n",
       "      <td>106.700000</td>\n",
       "      <td>184.049999</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>params_loss_function</th>\n",
       "      <td>(element_wise_squared_loss)</td>\n",
       "      <td>145.375000</td>\n",
       "      <td>NaN</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>params_target_update_period</th>\n",
       "      <td>(10, 5)</td>\n",
       "      <td>165.350002</td>\n",
       "      <td>125.399997</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>params_num_heads</th>\n",
       "      <td>(4)</td>\n",
       "      <td>145.375000</td>\n",
       "      <td>NaN</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "                                                       0           1  \\\n",
       "params_custom_layer_init                      (1.0, 0.5)  155.324997   \n",
       "params_custom_lr_schedule                           (No)  145.375000   \n",
       "params_layer_type                                 (6, 3)  106.700000   \n",
       "params_loss_function         (element_wise_squared_loss)  145.375000   \n",
       "params_target_update_period                      (10, 5)  165.350002   \n",
       "params_num_heads                                     (4)  145.375000   \n",
       "\n",
       "                                      2  \n",
       "params_custom_layer_init     135.425002  \n",
       "params_custom_lr_schedule           NaN  \n",
       "params_layer_type            184.049999  \n",
       "params_loss_function                NaN  \n",
       "params_target_update_period  125.399997  \n",
       "params_num_heads                    NaN  "
      ]
     },
     "execution_count": 35,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "X =  study2.values[:,1:]\n",
    "y =  study2.values[:,0]\n",
    "y = y *-1\n",
    "features = study2.columns[1:]\n",
    "final = []\n",
    "for w in range(len(features)):\n",
    "    summary = {}\n",
    "    for x in range(len(X)):\n",
    "        if str(X[x][w]) in summary: \n",
    "            summary[str(X[x][w])].append(y[x])\n",
    "        else:\n",
    "            summary[str(X[x][w])] = [y[x]]\n",
    "    row = []\n",
    "    row.append(summary.keys())\n",
    "    for x in summary.keys():\n",
    "        row.append((np.mean(summary[x])))\n",
    "    final.append(row)\n",
    "pd.DataFrame(final, index = features)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 36,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "number1 = []\n",
    "number2 = []\n",
    "number3 = []\n",
    "for row in final: \n",
    "    number1.append(row[1])\n",
    "    if len(row) >= 3:\n",
    "        number2.append(row[2])\n",
    "    else: \n",
    "        number2.append(0)\n",
    "    if len(row) == 4:\n",
    "        number3.append(row[3])  \n",
    "values = np.array([number1,number2]).T\n",
    "labels = [\"init\",\"schedule\",\"layer_type\",\"loss\",\"update\",\"heads\"]\n",
    "sns.set()\n",
    "plot_difference(labels,values[:,0],values[:,1])"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# LunarLander cross study"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 37,
   "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>value</th>\n",
       "      <th>params_custom_layer_init</th>\n",
       "      <th>params_custom_lr_schedule</th>\n",
       "      <th>params_layer_type</th>\n",
       "      <th>params_loss_function</th>\n",
       "      <th>params_target_update_period</th>\n",
       "      <th>params_num_heads</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>-91.938004</td>\n",
       "      <td>1</td>\n",
       "      <td>No</td>\n",
       "      <td>3</td>\n",
       "      <td>element_wise_huber_loss</td>\n",
       "      <td>10</td>\n",
       "      <td>4</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>-111.694763</td>\n",
       "      <td>1</td>\n",
       "      <td>No</td>\n",
       "      <td>3</td>\n",
       "      <td>element_wise_squared_loss</td>\n",
       "      <td>5</td>\n",
       "      <td>2</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>-216.704956</td>\n",
       "      <td>1</td>\n",
       "      <td>No</td>\n",
       "      <td>3</td>\n",
       "      <td>element_wise_huber_loss</td>\n",
       "      <td>5</td>\n",
       "      <td>2</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>-151.146942</td>\n",
       "      <td>1</td>\n",
       "      <td>No</td>\n",
       "      <td>3</td>\n",
       "      <td>element_wise_squared_loss</td>\n",
       "      <td>10</td>\n",
       "      <td>4</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>206.470276</td>\n",
       "      <td>1</td>\n",
       "      <td>No</td>\n",
       "      <td>3</td>\n",
       "      <td>element_wise_huber_loss</td>\n",
       "      <td>10</td>\n",
       "      <td>2</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>5</th>\n",
       "      <td>-194.101349</td>\n",
       "      <td>1</td>\n",
       "      <td>No</td>\n",
       "      <td>3</td>\n",
       "      <td>element_wise_huber_loss</td>\n",
       "      <td>5</td>\n",
       "      <td>4</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>6</th>\n",
       "      <td>225.068359</td>\n",
       "      <td>1</td>\n",
       "      <td>No</td>\n",
       "      <td>3</td>\n",
       "      <td>element_wise_squared_loss</td>\n",
       "      <td>5</td>\n",
       "      <td>4</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>7</th>\n",
       "      <td>-174.794769</td>\n",
       "      <td>1</td>\n",
       "      <td>No</td>\n",
       "      <td>3</td>\n",
       "      <td>element_wise_squared_loss</td>\n",
       "      <td>10</td>\n",
       "      <td>2</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "        value  params_custom_layer_init params_custom_lr_schedule  \\\n",
       "0  -91.938004                         1                        No   \n",
       "1 -111.694763                         1                        No   \n",
       "2 -216.704956                         1                        No   \n",
       "3 -151.146942                         1                        No   \n",
       "4  206.470276                         1                        No   \n",
       "5 -194.101349                         1                        No   \n",
       "6  225.068359                         1                        No   \n",
       "7 -174.794769                         1                        No   \n",
       "\n",
       "   params_layer_type       params_loss_function  params_target_update_period  \\\n",
       "0                  3    element_wise_huber_loss                           10   \n",
       "1                  3  element_wise_squared_loss                            5   \n",
       "2                  3    element_wise_huber_loss                            5   \n",
       "3                  3  element_wise_squared_loss                           10   \n",
       "4                  3    element_wise_huber_loss                           10   \n",
       "5                  3    element_wise_huber_loss                            5   \n",
       "6                  3  element_wise_squared_loss                            5   \n",
       "7                  3  element_wise_squared_loss                           10   \n",
       "\n",
       "   params_num_heads  \n",
       "0                 4  \n",
       "1                 2  \n",
       "2                 2  \n",
       "3                 4  \n",
       "4                 2  \n",
       "5                 4  \n",
       "6                 4  \n",
       "7                 2  "
      ]
     },
     "execution_count": 37,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "study3"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 38,
   "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>0</th>\n",
       "      <th>1</th>\n",
       "      <th>2</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>params_custom_layer_init</th>\n",
       "      <td>(1)</td>\n",
       "      <td>63.605268</td>\n",
       "      <td>NaN</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>params_custom_lr_schedule</th>\n",
       "      <td>(No)</td>\n",
       "      <td>63.605268</td>\n",
       "      <td>NaN</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>params_layer_type</th>\n",
       "      <td>(3)</td>\n",
       "      <td>63.605268</td>\n",
       "      <td>NaN</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>params_loss_function</th>\n",
       "      <td>(element_wise_huber_loss, element_wise_squared...</td>\n",
       "      <td>74.068508</td>\n",
       "      <td>53.142029</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>params_target_update_period</th>\n",
       "      <td>(10, 5)</td>\n",
       "      <td>52.852360</td>\n",
       "      <td>74.358177</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>params_num_heads</th>\n",
       "      <td>(4, 2)</td>\n",
       "      <td>53.029484</td>\n",
       "      <td>74.181053</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "                                                                             0  \\\n",
       "params_custom_layer_init                                                   (1)   \n",
       "params_custom_lr_schedule                                                 (No)   \n",
       "params_layer_type                                                          (3)   \n",
       "params_loss_function         (element_wise_huber_loss, element_wise_squared...   \n",
       "params_target_update_period                                            (10, 5)   \n",
       "params_num_heads                                                        (4, 2)   \n",
       "\n",
       "                                     1          2  \n",
       "params_custom_layer_init     63.605268        NaN  \n",
       "params_custom_lr_schedule    63.605268        NaN  \n",
       "params_layer_type            63.605268        NaN  \n",
       "params_loss_function         74.068508  53.142029  \n",
       "params_target_update_period  52.852360  74.358177  \n",
       "params_num_heads             53.029484  74.181053  "
      ]
     },
     "execution_count": 38,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "X =  study3.values[:,1:]\n",
    "y =  study3.values[:,0]\n",
    "y = y *-1\n",
    "features = study3.columns[1:]\n",
    "final = []\n",
    "for w in range(len(features)):\n",
    "    summary = {}\n",
    "    for x in range(len(X)):\n",
    "        if str(X[x][w]) in summary: \n",
    "            summary[str(X[x][w])].append(y[x])\n",
    "        else:\n",
    "            summary[str(X[x][w])] = [y[x]]\n",
    "    row = []\n",
    "    row.append(summary.keys())\n",
    "    for x in summary.keys():\n",
    "        row.append((np.mean(summary[x])))\n",
    "    final.append(row)\n",
    "pd.DataFrame(final, index = features)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 39,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAaQAAAEUCAYAAABkhkJAAAAABHNCSVQICAgIfAhkiAAAAAlwSFlzAAALEgAACxIB0t1+/AAAADh0RVh0U29mdHdhcmUAbWF0cGxvdGxpYiB2ZXJzaW9uMy4xLjMsIGh0dHA6Ly9tYXRwbG90bGliLm9yZy+AADFEAAAgAElEQVR4nO3dd1yV9fv48ddho7gFzZkDR8O9cICSuYBcuMidpuVeuXDkFjUTLKuP5UxNU9NQKSfuUWZqaWoq4oIDggyZ57x/f/jzfDMxQDmHA+d6Ph4+Hp7Dfb/v63rf9znXed9To5RSCCGEELnMKrcDEEIIIUAKkhBCCDMhBUkIIYRZkIIkhBDCLEhBEkIIYRakIAkhhDALUpDMVPXq1fHx8aFjx4506tSJtm3b0rVrVy5cuJCrcc2aNYugoKBcjeFFLFu2jB9++MGoy9i2bRtDhgzJ9nzLly9n3759Rogof9qyZQvffvutyZfbp08fQkJCTL5cS2KT2wGI51uzZg3Fixc3vP7666+ZM2cO3333XS5GlTeNGjUqt0N4rlOnTlG1atXcDiPP+PXXX3F1dc3tMIQRSEHKI9LT07l37x5FihQxvLdixQp+/vln9Ho9ZcuWZcaMGVy4cIFvvvmGDRs2ANC2bVu8vLwYOXIk9+/fx9fXl8OHD/PVV1+xf/9+kpOTSUpKYuLEibz99tsEBQVx7tw5IiMjqV69OjNnzmTq1KlcvnwZFxcXrK2tqV+/fobxLVq0iEOHDmFtbU3dunWZMWMGGo2GBQsWcOLECaytralVqxaTJ0/GyckJT09PvL29OXnyJA8fPmTQoEGcPXuWP/74AxsbG1asWEGpUqXw9PTEy8uLY8eOER8fz4ABA/Dz80Ov1zNv3jx+//13EhMTUUoxZ84c6tevz6RJk4iNjSU8PJyWLVsSHR2Nq6sr7733HoGBgezduxdbW1uKFSvG/PnzcXFx4ZdffiEgIICkpCRsbW0ZPXo07u7ubNu2jb1792JlZUVYWBgODg4sXLiQKlWqPNMPWq2W9957j8jISMqWLcvs2bNxdnYmPj6euXPncuXKFdLS0nBzc+Ojjz7iu+++4+LFiwQEBHD//n1WrVpFaGgoAO+99x4lS5Zk4cKFpKam0qJFC/bt20dkZCRz584lNjYWnU5Hnz598PX1BeDAgQOsWLGCtLQ0HBwcmDhxInXr1iUoKIg7d+6g1Wq5c+cOpUqVYtGiRbi4uDwVf1BQEGFhYdy/fx+tVkuNGjWYO3cuTk5OHDx4kC+//JLU1FQePHhAp06dGD16NKdOnWLu3LkUKFCAxMREtm7dSkBAwHPXi4ODA1euXCE6OhpPT0+KFi3KwYMH0Wq1zJkzBzc3N1JTU1m8eDFnzpxBp9Px2muv4e/vz4kTJzhw4ADHjh3DwcGBd999N8PPQalSpejTpw9FihTh+vXr9OrViz59+hjy7NmzJwMGDKBt27YALFq0CIBhw4Yxc+ZMwsLCiI2NpWDBgixevJjKlSsb5r19+zY+Pj789ttvGb7esmULGzduRK/XU7RoUaZNm0aVKlX45ZdfWLBgAXq9HoAhQ4YYli/+PyXMUrVq1ZS3t7fy9vZWzZo1U56enmr27NkqKipKKaXU9u3b1ejRo1VaWppSSqlNmzapQYMGqaSkJFWvXj318OFDFR4erpo1a6Z69OihlFJq/fr1asaMGer27duqT58+KikpSSmlVHBwsPL29lZKKRUYGKjatm1raHfu3Lnqo48+Unq9XkVHRyt3d3cVGBj4TLxr1qxR7777rkpKSlI6nU6NGjVKbd++XS1btkwNHz5cpaamKp1OpyZNmqSmTZumlFKqVatWat68eUoppXbt2qVq1KihLl26pJRS6sMPP1QrVqwwTDdt2jSl1+vVvXv3VOPGjdXly5fV2bNn1YgRI5ROp1NKKfXll1+qIUOGKKWUmjhxourXr58hvokTJ6qVK1equ3fvqnr16qmUlBSllFJff/212rt3r3rw4IFyc3NT586dU0opdeXKFdWoUSN169YttXXrVlW/fn117949pZRSs2bNUh999NEzfbB161ZVp04ddfPmTaWUUkuWLFGjRo1SSik1adIktXbtWqWUUunp6Wr8+PHqq6++Ukop1bt3b7Vnzx6llFKenp7qr7/+UklJSaply5bK3d1dKaXUoUOH1KBBg1RaWprq0KGDunjxolJKqbi4ONW+fXv122+/qRs3bihvb2/14MEDQw7NmjVTiYmJKjAwUL311lsqPj5eKaXUkCFD1LJly57JITAwULm7uyutVqt0Op0aO3asWrBggdLr9ap3797qxo0bSiml7t+/r2rWrKmio6PVyZMnVY0aNdTt27eVUirT9dKtWzeVmpqqIiMjVbVq1Qz9snr1ajVgwACllFJBQUGG5T7pyxkzZjy1LpV6/ufgSb9Onjz5mRyVUur7779X77//vmF9NG/eXN24cUPt2bNHzZ492zDdtGnT1KxZs55aT+Hh4apOnTqGaf75+tSpU8rPz089evRIKaXUkSNHVLt27ZRSSvXt21cFBwcrpZS6dOmSmjlzZoaxWTIZIZmxJ7vs/vjjD95//30aN25MiRIlADh48CAXLlyga9euAOj1epKSknBwcKBp06YcO3aMmJgYevTowXfffUd8fDwHDhxg0KBBlC1bloCAAH788UfCwsIMv2SfqFOnDjY2jzeNEydOMGXKFDQaDcWLF+ftt9/OMNbjx4/TsWNHHBwcAPj0008B8PX1ZcyYMdja2gKP98MPGzbMMF+bNm0AKF++PCVLlqRGjRoAVKhQgYcPHxqm8/PzQ6PRULp0aVq0aMGxY8cYOHAgRYoUYdOmTYSHh3Pq1CkKFixomCejkVypUqWoUaMGnTt3xt3dHXd3d9zc3AgNDaVChQrUrl0bAFdXV+rVq8fp06fRaDS8/vrrlC5dGoDXXnuNvXv3ZtgPTZs2pWLFiobcn4xcDh06xIULF/j+++8BSE5OznD+t99+m8OHD+Pq6kqTJk3466+/uHr1Kvv376dNmzbcvHmTW7duMWXKFMM8ycnJ/PnnnyiliIyMpH///oa/aTQabt26BUCjRo1wcnIy5PDP/v2ndu3aUbJkSUMO8+bNY+LEiXzxxRccOnSI4OBg/v77b5RSJCUlAfDKK69QtmxZAOrWrfuf66VVq1bY2tri7OxMgQIFaNGiBfB4ncfGxhr6Kz4+nuPHjwOQlpZm2Pb/6XmfgycaNGiQYY4dOnQgICAArVbLn3/+yauvvmr4V758edatW0dYWBinT5+mbt26GbaRkUOHDhEWFkbPnj0N78XFxREbG0v79u2ZNWsWBw4coGnTpowdOzbL7VoKKUh5wOuvv87kyZOZNGkSNWvWpFy5cuj1egYNGoSfnx8Aqamphi+Y1q1bc/jwYeLi4hg0aBDXr19n3759XLlyhUaNGvHHH3/w4Ycf0r9/f5o1a0bDhg35+OOPDcsrUKDAU8tX/7jdobW1dYYxPilgT0RFRaHX69Hr9Wg0GsP7er2etLQ0w2s7OzvD/58Urcza1+v1WFlZcejQIebOncuAAQN46623qFy5Mjt37nxuHgBWVlasX7+eCxcucOLECebNm0eLFi1o0KDBU3E+yTs9PR1bW1tDoYXHX/LqObeA/Gf/6PV6Q9x6vZ5ly5YZdvPFxcU9szx4vO6WLVtGZGQkzZo1o0SJEhw9epTDhw8zZswYIiMjKVSoEDt27DDMExUVRaFChdi8eTNubm6GHwMA9+7dw8XFhb17975wDlZWVjx69IjOnTvTunVrGjRoQNeuXdm3b5+hjX/2dWbr5Z/rHJ7ddp4sd8qUKXh4eACQmJhISkpKhtM973Pw77j+ydHRkbZt2xIcHMxvv/1Gt27dANiwYQObN2/m3XffxcfHh6JFi3L79u2n5v133/1ze9br9XTs2JEJEyYYXkdGRlKkSBF69uxJq1atOHbsGEeOHGH58uWEhIRgb2+fYYyWSM6yyyO8vb2pVasW8+fPB6B58+Z8//33JCQkAI/PIvvoo48A8PT05MSJE1y6dIlatWrRrFkzli1bhru7O9bW1pw5c4Y33niDAQMG0KhRI/bv349Op8twuS1atOD7779Hr9fz8OFD9u/fn+F0bm5uBAcHk5qail6vZ+bMmezatYsWLVqwceNG0tLS0Ov1fPvttzRr1izb+T85Q+7u3bscO3YMd3d3jh07RqtWrfDz8+ONN95g3759z83jicuXL+Pt7U2VKlUYMmQI/fv358KFC9SpU4fr169z/vx5AK5evcqZM2do1KhRtuI8deoUd+/eBWDTpk24u7sDj9fX6tWrUUqRmprKBx98wPr164HHBSA9PR2AevXqER4ezqFDh2jatCnNmjVjzZo1vPrqqxQrVoxKlSrh4OBgKEj37t3D29ubixcv4ubmxrFjx/j7778BCA0N5Z133nnuaOx59u/fT3x8PHq9ns2bN9OqVSvCwsJISEhg9OjReHp6curUKcO6/rcXWS//1rx5c7799lvDMqZNm8Ynn3zyTH/91+cgM927d2f79u2cPXvWcCzn6NGjdO7cmW7dulGpUiUOHDjwTOyFCxcmLS2Na9euAbBr166n4t61axeRkZEAbNy4kX79+gGPj1tdunSJLl26MHv2bOLi4tBqtdnql/xORkh5yLRp03jnnXc4cuQI3bp1IyIigu7du6PRaHjllVdYsGABAIUKFaJKlSo4OjpibW1NixYtmDp1qmH3mLe3Nz///DPt27dHr9fTqlUrHj58aPhQ/9OIESOYMWMG7du3p3jx4lSrVi3D2Hr27MmdO3fo0qULSikaNWpEnz59SE9PZ+HChXTq1In09HRq1arFtGnTsp377du36dKlC8nJyfj7+1O5cmV69uzJuHHj8PHxIT09nWbNmhkObj9PjRo1aN++PV27dqVAgQI4ODjg7+9P8eLFWbZsGbNnzyY5ORmNRsP8+fOpVKmS4WB1VlSrVo0pU6YQFRVF5cqVmTVrFgBTp05l7ty5+Pj4kJaWRtOmTRk0aBDw+AfEJ598QlpammFX4oULFyhevDj169fn4cOHhnVnZ2fH559/zty5c1m5ciXp6emMGjXKsHty1qxZjB07FqWU4cSQf+4uy4qSJUsyePBgYmJiaNiwIUOHDsXOzo6WLVvSvn177OzsqFatGlWrViUsLOyZEc+LrJd/+/DDD1m4cCGdO3dGp9NRs2ZNJk2aBIC7u7thWx88ePBzPweZeeONN7C2tqZdu3aGUcrAgQOZPn26YddqnTp1uHLlylPzFSpUiAkTJjB48GCKFy9Ou3btDH9r3rw5gwcPZuDAgWg0GpycnFi+fDkajYbx48czb948Pv30UzQaDcOHD6dcuXJZ7hNLoFHPG7cLYSY8PT1ZtmwZb775Zm6Hku8FBQURExPD9OnTczsUYYFkl50QQgizICMkIYQQZkFGSEIIIcyCFCQhhBBmQc6yEyKP++GHH1i1apXhdXx8PBEREYSGhhoucB0+fDguLi5yskIGLKn/zD1XKUhC5HGdOnWiU6dOwOOLNHv37s37779v+IL53//+xy+//EKHDh1yM0yzZUn9Z+655rmCFBOTiF5vXudhlCjhRHT0s9fw5EeSq3lbtWolTk5FePttb6KjEzh79hcOHDhEx45diI+Pf24+eTHXF/Vfub5o/5krc8vVykpDsWLPvy4uzxUkvV6ZXUECzDImY5FczVNsbCwbN67n66/XodcroqK0LF26mCVLgtixYytK/fdnJy/l+rIyyvVl+89c5aVc5aQGIfKJnTu30aKFB2XLliM9PZ2ZM6cycuRYw+4Y8d8sqf/MNdc8N0ISQmRs//69jB49HoDLl//k7t07BAUtBeDBg2j0eh2pqalMmpT9WzdZAkvqP3PNNV8UJJ0unZgYLenpqbmy/MhIq2zdpysnWFlZ4+johJNTkQzvGi0sS1xcHHfuhPPmm48fn/HGG7XYtu3/bvr59ddf8vBhLGPHTsytEM2aJfWfOeeaLwpSTIwWB4cCFCxYOle+nG1srEhPN11BUkqh06UTHx9LTIyW4sVdMp9J5Gt37oRTokTJDB/lIDJnSf1nzrnmuVsHRUcnPHOw7f79MEqVqpBrIwVTF6QnlNITEXGb0qUrmGyZzs6F0GrjTba83CS55k+Sa+6xstJQooTT8/9uwliMyhJ3W2k0VkCe+j0hhBDPlW8KkhBCiLzN/HYi5oBChR1xsM/51JJT0omPS/rPaT744D26du1O69ZtDe8lJSXRtas3GzZspWjRolle3uzZ06lfvyEdOvi8cMxCCJFX5MuC5GBvg8+4HTne7o9LOpLZ3lgvr3f4+eeQpwpSaOgB6tVrkOViFBWlJSBgHr/+epr69Ru+RMQiPzLGD67UtOw9Yjwv06en4uxcKMfaS09NIeZh7pzhm5m8lmu+LEi5ydPzbT77bBlxcQ8pXLgIAD/9tJvu3f04cGAfmzatJyUlhbS0VCZPns6bb9Zm+PD3KVy4CDdu/M2sWfM5c+YULVp4UKRIkVzORpgjY/zg+nFJxxxtz5xZ2dhxfW7XHGuv8tStgHkWpLyWqxxDymEFChSgRQsPDhzYBzwe7dy6FUajRk3YsWMrAQGfsmbNRvz8+rJu3WrDfFWqVGXjxm24ulbHz68vPj6dcikDIYTIHVKQjKBDBx/27fsJgJ9/3kPbth2wtrZm3rxFnD59gpUrv2DPnmCSkh4Z5nnttTdyK1whhDALUpCMoE6dekRHRxERcZ+fftqDl9c7PHr0iMGD+3H37h1q166Lr28P/nkJmL29fS5GLIQQuU8KkpG0a+fF2rXfULhwYcqWLUd4+C00Gg19+w6kXr0GhIYeNPnthoQQwpxJQTKSDh18CA7egZfXOwBUrepK1arV8PPzpU+f7hQtWoz79+/lcpRCCGE+8uVZdskp6UY5ayg5JT3L0zo7uxAaesrw2tramo8/nvfUNE/utrt8+VcZtjF16szsBymEEHlUvixI8XFJmV4vJIQQwrwYrSBt2bKF9evXG17fvn2bjh070rp1a+bPn09KSgrt27dnzJgxxgpBCCFEHmK0gtStWze6desGwNWrVxk2bBiDBw+mV69erFu3jldeeYUhQ4YQGhqKh4eHscIQQgiRR5jkpIaZM2cyZswYwsPDqVixIuXLl8fGxgYfHx9CQkJMEYIQQggzZ/RjSMePHyc5OZn27dsTHByMs7Oz4W8uLi5ERERkq72MnqURGWmFjU3unjCYW8u3srLK0XtVZYWpl5ebJFeRFZbUd8bM1egFadOmTQwYMAAAvV7/1HOLlFLZfo5RRg/o0+v1ufKAvCdy6wF98Dh3Uz6Ay9we+GVM5pqrsb4QzDFXYzBG/5lr35lbrpk9oM+oBSk1NZUzZ86wYMECAEqXLo1WqzX8XavV4uKS84/fLlbEDhu7nL/zgTnf1VcIIfI6oxakv/76i1dffZUCBQoAULt2bW7cuEFYWBjlypUjODiYrl1z7k60T9jY2efoHW6fyMqdbnPieUhHjhzi66+/QilFmTJlmDx5BoULF37Z8IUQwqwZ9cBHeHg4pUuXNry2t7dnwYIFjBgxgg4dOlC5cmXatWtnzBBM7snzkP4pO89DSkxMYPHiBSxa9Piu4FWquPLNNxlfOCuEEPmJUUdIHTp0oEOHDk+95+bmxs6dO4252Fz1ss9DmjFjLmPHTsTZ+fGuzCpVqj5T4IQQIj+Se9nlsJd9HlKNGjXx8GgFQEpKMuvXr8HdvWUuZCKEEKYlBckIcuJ5SAkJCUyYMJqqVV1p397bpPELIURuyJf3sstt/34e0rx5iwzPQ2rTpj21a9elSpWqbN262TDPP5+HFBUVxbhxw6lXryEjR47N0dj+/vsaS5cGkJiYgJWVNRMmTKFKlaosXRrA+fPnAGjcuCkffjgSa2vrHF22qVlSrkLkB1KQjOTfz0P666/LhuchKaWYPXt6hs9D0ul0TJw4hlatWtO//6AcjSk5OZmxY4cxadI03Nyac+TIIWbN8ueddzoTGxvL2rXfodfrGTZsMAcO7OXtt/PuCSeWlKsQ+UW+LEjpqSn//xTtnJWWkpzlaTt08MHX14fJk6cDTz8PycpKQ6NGboZf6f909Ohhrly5jE6n49ChAwDUqFGTSZOmvXT8p0+fpEyZcri5NQegeXMPXnmlLFWruuLr2xMrKytiY2NISIg3nJCRV1lSrkLkF/myID2+eDVrF7BGR0eRnJyMlZUVKSnJWFlZ4+LigoODIwCRkRHExsbg4OBAuXIVsLLK2mG3F30ekodHK44cOZOlZWRXeHgYJUqUYP78WVy7dhUnp0J8+OFIAGxsbFixIoht2zZTvXpNateua5QYTMWSchUiv7D4kxqUUiQkJFC0aDFefbUyxYoVJzw83LA7zcWlFK6u1bG1tcvzT3hNT0/nxIljvPNOF77+eh2+vt2ZMGEUqamPi/cHH4xgz56DvPJKGRYvnp/L0b4cS8pViPzC4guSjY0t9vZ2ODo+HhEVKlQIUCQnJ5OSkgKARqOhcOEipGRjl505KlnSmYoVK/H664/P6GvRoiV6vY7Ll//k1q0w4PHooX17b65cuZybob40S8pViPwi3xQkpVTmE2XAyakgqalpJCUlAfDoUSKg4dGjRCIjI1BKoZQiLu4hBQoUzMGIX152c27SpCn37t3l8uVLAJw7dxbQcPbsLwQFfUJ6ejp6vZ69e0OoV6+hESI2HUvKVYj8Il8cQ7KyskanS8fGxjbb89rY2FKuXDkiIu6j1+uxstJQtmw5HBwciIyM4MaN62g04OhYwHD3BHORlpaKtXXWV2GJEiWZP38xS5YsIDk5CVtbO+bOXcRrr73OsmVL6N/fDysrDbVq1WHo0OFGjNz4LClXIfKLfFGQHB2diI+PpWjREmg02R/0FShQkFdfrfTM+6VLv5IT4eU4pRRpaanExmopVKhYtuatU6ce//vfmmfeHz9+Uk6FZzYsKVch8oN8UZCcnIoQE6MlIuI28GK77l6GlZVVhtcUGZO1tQ2FChXD0dG8diMKIcSLyhcFSaPRULx47u1OM9cHuQkhRF6SLwpSdhQq7IiDfc6mnZqmy9H2cork+nLMNVch8iuLK0gO9jb4jNuRo23+uKRjjraXUyTXl2OuuQqRX+Wb076FEELkbVKQhBBCmAUpSEIIIcyCFCQhhBBmQQqSEEIIs2DUgnTgwAG6dOlC+/btmTNnDgDHjx/Hx8eHNm3asHTpUmMuXgghRB5itIIUHh7OjBkz+Pzzz9m5cyd//vknoaGhTJkyhc8//5zdu3dz8eJFQkNDjRWCEEKIPMRoBWnv3r106NCB0qVLY2try9KlS3F0dKRixYqUL18eGxsbfHx8CAkJMVYIQggh8hCjXRgbFhaGra0tQ4cO5d69e7Rs2RJXV1ecnZ0N07i4uBAREWGsEIQQQuQhRitIOp2OX375hXXr1lGgQAE++OADHBwc0Gg0hmmUUk+9zooSJZxyOtQc4excKLdDMBnJNX+ypFxzmiX1nTFzNVpBKlmyJG5ubhQvXhyA1q1bExISgrW1tWEarVaLi0v2booaHZ2AXv/id/Q2Vmea481VJdeXJ7nmP8boP3PtO3PL1cpK85+DCqMdQ2rVqhVHjx4lLi4OnU7HkSNHaNeuHTdu3CAsLAydTkdwcDDu7u7GCkEIIUQeYrQRUu3atRk0aBB+fn6kpaXRrFkzevXqReXKlRkxYgQpKSl4eHjQrl07Y4UghBAiDzHq3b59fX3x9fV96j03Nzd27txpzMUKIYTIg+RODUIIIcyCFCQhhBBmQQqSEEIIs2BxT4wVQmRNUNBSDh7cR+HCRQCoUKEi06bNYunSAM6fPwdA48ZN+fDDkU9dziFyjqWtAylIQogMXbx4no8/nsebb9Y2vLdp03piY2NZu/Y79Ho9w4YN5sCBvbz9tpwtawyWtg6kIAkhnpGamsrVq3+xYcNa7ty5TfnyFRgxYhw9e/bG17cnVlZWxMbGkJAQb/j1LnKWJa4DKUhCiGdERWmpV68Bgwd/QKVKVdi4cR2TJ4/lm2++xcbGhhUrgti2bTPVq9ekdu26Jo0to91YM2fO5Ysvgjh+/BhWVhrKlavAhAlTKFasmEljy0nmvA6MRU5qEEI8o0yZsixeHEjlylXRaDT06tWHO3fucO/eXQA++GAEe/Yc5JVXyrB48XyTxvZkN9bq1RtYvXoDs2bNZ9eunfz112W++WY9a9d+R7ly5Vi+PG8/b82c14GxSEESQjzj2rWrhITseuo9pRQREfe5dSsMABsbG9q39+bKlcsmi+ufu7H69u3B1KkTuH//PpUqVebDD0dhZ2cHQPXqrxERcd9kcRmDua4DY5KCJIR4hpWVhk8/Xczdu3cA2L79e6pWrcrvv/9GUNAnpKeno9fr2bs3hHr1Gposrn/uxlqzZhOvv/4mkyeP5fXX36R69RoAxMXFsXr1/2jV6i2TxWUM5roOjEmOIQkhnlG5clXGjJnAxIlj0Ov1ODu7MGPGPEqUKMGyZUvo398PKysNtWrVYejQ4SaL68lurCd69erD6tVfc+/eXcqUKcudO7eZPHkctWrVoUuX7iaLyxjMdR0YkxQkIUSG2rbtQNu2HZ55f/z4SbkQzWPXrl3l2rUrtGvnZXhPKYWNjQ1nz/7C9OmT8fPri59fn1yLMSeZ4zowJtllJ4TIM563GysmJoYpU8bj7/9xvilGlkhGSEKIPON5u7ECAuaglOKLL5bzxRfLAXjllTLMn784lyMW2SEFSQiRp2S0G+uTT5bnUjQiJ0lBEkKgT0/N0cddp6emEPMwNcfaswSFCjviYG/ZX8nPzd7T0xONRvPcGffv32+UgIQQpmdlY8f1uV1zrL3KU7cCL1+QLOlL2sHeBp9xO3K0zR+XdMzR9oztuWs6MPDxqZUbNmzA1taWHj16YG1tzbZt20hLSzNZgEIIyyVf0pbluQXpjTfeAODq1ats2bLF8P7kyZOfeSy5EEII8bIyPe07Li6OBw8eGF5HRESQkJBg1KCEEEJYnkx3zvbr1w8fHx+aN2+OUopjx44xYcKELDXep08fHjx4gI3N40GuxuMAAB9ISURBVMXMmjWLW7dusWLFCtLT0+nXrx/vvvvuy2UghBAiX8i0ILVp04Z69epx4sQJAAYNGkS1atUybVgpxc2bNzl48KChIEVERDBmzBi2bduGnZ0dPXv2pHHjxlStWvUl0xBCCJHXZVqQevfuTUhICDVq1MhWw9evXwdg4MCBxMbG0r17dwoWLEiTJk0oWrQoAG3btiUkJIThw/PHfZiEEEK8uEyPIZUtW5azZ8+i1+uz1XBcXBxubm589tlnrF69mk2bNnH37l2cnZ0N07i4uBAREZH9qIUQQuQ7mY6Q/v77b/z8/LCxscHOzg6lFBqNhrNnz/7nfHXr1qVu3f97iqGvry/z58/ngw8+MLz3pK3sKFHCKVvTm0pOXlRo7iRXkRWW1HeSa87ItCB9++23L9TwL7/8QlpaGm5ubsDj4lO2bFm0Wq1hGq1Wi4uLS7bajY5OQK9XLxQTGK8ztdp4o7T7MiTXl2dJuea0nOg7ydX8vEyuVlaa/xxUZGmX3cOHD7l37x53794lPDycY8eOZbrg+Ph4AgICSElJISEhge3bt7No0SJOnDjBgwcPSEpK4ueff8bd3T17GQkhhMiXMh0h+fv7s3//flJSUnBxceHWrVvUr1+f7t3/++FXrVq14vfff6dTp07o9Xr8/PyoX78+Y8aMoW/fvqSlpeHr60utWrVyLBkhhBB5V6YF6fjx4+zfv5+PP/6YYcOGce/ePVauXJmlxkePHs3o0aOfes/HxwcfH58Xi1YIIUS+lekuO2dnZwoUKEDlypW5cuUKjRs35v79+6aITQghhAXJtCDZ2tpy5swZqlSpwuHDh4mPj+fRo0emiE0IIYQFybQgTZgwgU2bNuHh4cHly5dp0qQJ77zzjiliE0IIYUEyPYYUFxfHwoULsbGxYfPmzcTHx1OoUN44PVEIIUTekekI6ZtvvsHd3Z2JEydy4MAB7O3tTRGXEEIIC5NpQVq1apXheqFdu3bRpk0bxo0bZ4rYhBBCWJBMCxJAdHQ0sbGxJCUlkZqaSmxsrLHjEkIIYWEyPYbk6elJWloa3t7e+Pn5sXTpUtltJ4QQIsdlWpB69OjB0aNHOXjwII8ePSIpKYkmTZrIiQ1CCCFyVKYFaciQIQwZMoTExERCQ0NZunQpYWFh/PHHH6aITwghhIXItCCFh4dz+PBhjhw5woULF2jYsCHDhg0zRWxCCCEsSKYFqXv37nh6etKjRw8CAwOxs7MzRVxCCCEsTKYF6dixY0RERHDlyhWsra25e/cuZcqUMUVsQgghLEimp30fOXKEXr168fHHHxMdHY2Xlxf79u0zRWxCCCEsSKYFafny5WzevJnChQvj4uLChg0bCAwMNEVsQgghLEimBUmn0z31mPGaNWui0WiMGpQQQgjLk2lBcnR05O7du4Yi9Msvv8iFsUIIIXJcpic1jBs3joEDB6LVaunRowc3b94kKCjIFLEJIYSwIJkWpAoVKrB582Z+++039Ho9tWvXpnjx4qaITQghhAXJtCD17t2bkJAQPDw8TBGPEEIIC5XpMaSyZcty9uxZ9Hq9KeIRQghhoTIdIf3999/4+flhY2ODnZ0dSik0Gg1nz57N0gIWLlxITEwMCxYs4NKlS0ydOpXExEQaNGjAxx9/jI1NpiEIIYSwAJlWg2+//faFGz9x4gTbt2+nZcuWAEyYMIE5c+ZQp04dpkyZwubNm/Hz83vh9oUQQuQfWdpll9G/zMTGxrJ06VKGDh0KwJ07d0hOTqZOnToAdOnShZCQkJcMXwghRH5htP1l06dPZ8yYMdy7dw+AyMhInJ2dDX93dnYmIiIi2+2WKOGUYzHmJGdny3k+lOQqssKS+k5yzRlGKUhbtmzhlVdewc3NjW3btgGg1+ufusPDk2NR2RUdnYBer144NmN1plYbb5R2X4bk+vIsKdeclhN9J7man5fJ1cpK85+DiiwVpOTkZMLCwqhWrRrJyck4Ojr+5/S7d+9Gq9XSsWNHHj58yKNHj9BoNGi1WsM0UVFRT92SSAghhGXL9BjSuXPnaN26NUOGDCEiIoKWLVtmeobdqlWrCA4OZseOHYwcORJPT0/mz5+Pvb09v/76KwA7duzA3d09Z7IQQgiR52VakAICAli9ejVFixaldOnSBAQEMHfu3Bda2OLFi5k/fz7t2rXj0aNH9O3b94XaEUIIkf9kussuOTmZqlWrGl57eHiwdOnSLC+gS5cudOnSBYAaNWrw/fffv0CYQggh8rtMR0g2NjY8fPjQcALC9evXjR6UEEIIy5PpCGno0KH07t2bqKgoxo4dy7Fjx5g1a5YpYhNCCGFBMi1Inp6eVKlShWPHjqHX6xk2bBhVqlQxRWxCCCEsSKYF6cyZMwC4uroCEBMTw8WLF3n11VdxcjLPi1SFEELkPZkWpPnz53P58mVcXV2xsrLiypUrODs7k5SUxNy5c2ndurUp4hRCCJHPZXpSQ5kyZVi1ahU7duxg+/btbNy4kXr16rFjxw4+++wzU8QohBDCAmRakMLDw2ncuLHhda1atbh58yalS5c2amBCCCEsS5ZO+z569Kjh9dGjR7G1teXBgwekp6cbNTghhBCWI9NjSDNmzGDkyJFoNBr0ej329vYEBgaycuVKevbsaYoYhRBCWIBMC1KtWrXYv38/V65cwdramipVqmBtbU2NGjVMEZ8QQggLkWlBevDgATt37iQxMRGlFHq9nrCwMJYsWWKK+IQQQliITAvS6NGjcXBw4Nq1azRt2pTjx49Tv359U8QmhBDCgmR6UsPdu3f56quvcHd3p3fv3mzcuFHuZyeEECLHZVqQSpYsCcCrr77KlStXKFWqlJxdJ4QQIsdlusuuRIkSrFy5kjp16hAUFISTkxPJycmmiE0IIYQFyXSENGvWLOzs7GjQoAFvvPEGgYGBjB8/3hSxCSGEsCCZjpAWLlxIQEAAABMmTGDChAlGD0oIIYTlyXSEdOnSJZRSpohFCCGEBct0hOTi4oKXlxe1a9emYMGChvf9/f2NGpgQQgjLkmlBqlu3LnXr1jVFLEIIISxYpgVp+PDhJCcnExYWhqurKykpKTg6Omap8WXLlvHTTz+h0Wjw9fVlwIABHD9+nPnz55OSkkL79u0ZM2bMSychhBAi78v0GNLvv/9O69atGTJkCJGRkbRs2ZKzZ89m2vDp06c5efIkO3fuZOvWraxbt47Lly8zZcoUPv/8c3bv3s3FixcJDQ3NkUSEEELkbZkWpIULF7J69WqKFi1K6dKlCQgIYO7cuZk23KhRI9auXYuNjQ3R0dHodDri4uKoWLEi5cuXx8bGBh8fH0JCQnIkESGEEHlbprvskpOTqVq1quG1h4cHS5cuzVLjtra2BAYG8s0339CuXTsiIyNxdnY2/N3FxYWIiIhsBVyihFO2pjcVZ+dCuR2CyUiuIissqe8k15yRaUGysbHh4cOHaDQagGzfx27kyJEMHjyYoUOHcvPmTUM7AEqpp15nRXR0Anr9i5+GbqzO1GrjjdLuy5BcX54l5ZrTcqLvJFfz8zK5Wllp/nNQkWlBGjp0KL179yYqKoqxY8dy7NgxZs2alemC//77b1JTU6lZsyaOjo60adOGkJAQrK2tDdNotVpcXFyymIoQQoj8LNNjSJ6enixfvpwRI0ZQr149NmzYQNu2bTNt+Pbt2/j7+5Oamkpqair79++nZ8+e3Lhxg7CwMHQ6HcHBwbi7u+dIIkIIIfK2TEdIY8eOpXv37vj5+WWrYQ8PD86fP0+nTp2wtramTZs2eHl5Ubx4cUaMGEFKSgoeHh60a9fuhYMXQgiRf2RakBo2bMgnn3xCTEwMvr6+dOnS5akTE/7LiBEjGDFixFPvubm5sXPnzheLVgghRL6V6S67Xr16sXnzZr744gsePnxIz549GTZsmCliE0IIYUEyLUhPJCcnk5qailIKK6sszyaEEEJkSaa77FatWsW2bdtITU3F19eXzZs3G54iK4QQQuSUTIc6Fy9exN/fn59++omBAwdy6tQpunXrZorYhBBCWJBMR0hLlizh4cOHfPXVV3z77bckJibSp08fU8QmhBDCgvxnQbp+/Tpr165lx44dlC1bluTkZA4ePEihQnnjimIhhBB5x3N32b3//vv07t0bGxsb1q5dS3BwMAULFpRiJIQQwiieW5D+/PNPXn/9dVxdXalYsSJAtu87J4QQQmTVcwvSoUOH6Ny5M8HBwTRv3pyRI0eSkpJiytiEEEJYkOcWJBsbGzp06MC6devYtm0bLi4upKSk0KZNGzZu3GjKGIUQQliALF3hWrVqVfz9/Tl8+DDvvfcemzdvNnZcQgghLEy2brng6OhIjx492L59u7HiEUIIYaHkHkBCCCHMghQkIYQQZkEKkhBCCLMgBUkIIYRZkIIkhBDCLEhBEkIIYRakIAkhhDALUpCEEEKYBaMWpOXLl+Pl5YWXlxcBAQEAHD9+HB8fH9q0acPSpUuNuXghhBB5iNEK0vHjxzl69Cjbt2/nhx9+4I8//iA4OJgpU6bw+eefs3v3bi5evEhoaKixQhBCCJGHGK0gOTs7M2nSJOzs7LC1taVKlSrcvHmTihUrUr58eWxsbPDx8SEkJMRYIQghhMhDjFaQXF1dqVOnDgA3b95kz549aDQanJ2dDdO4uLgQERFhrBCEEELkIf/5CPOccPXqVYYMGcJHH32EtbU1N2/eNPxNKZXth/6VKOGUwxHmDGdny3mSruQqssKS+k5yzRlGLUi//vorI0eOZMqUKXh5eXH69Gm0Wq3h71qtFhcXl2y1GR2dgF6vXjgmY3WmVhtvlHZfhuT68iwp15yWE30nuZqfl8nVykrzn4MKo+2yu3fvHsOGDWPx4sV4eXkBULt2bW7cuEFYWBg6nY7g4GDc3d2NFYIQQog8xGgjpK+//pqUlBQWLFhgeK9nz54sWLCAESNGkJKSgoeHB+3atTNWCEIIIfIQoxUkf39//P39M/zbzp07jbVYIYQQeZTcqUEIIYRZkIIkhBDCLEhBEkIIYRakIAkhhDALUpCEEEKYBSlIQgghzIIUJCGEEGZBCpIQQgizIAVJCCGEWZCCJIQQwixIQRJCCGEWpCAJIYQwC1KQhBBCmAUpSEIIIcyCFCQhhBBmQQqSEEIIsyAFSQghhFmQgiSEEMIsSEESQghhFqQgCSGEMAtGL0gJCQl4e3tz+/ZtAI4fP46Pjw9t2rRh6dKlxl68EEKIPMKoBen333+nV69e3Lx5E4Dk5GSmTJnC559/zu7du7l48SKhoaHGDEEIIUQeYdSCtHnzZmbMmIGLiwsA58+fp2LFipQvXx4bGxt8fHwICQkxZghCCCHyCBtjNj537tynXkdGRuLs7Gx47eLiQkREhDFDEEIIkUcYtSD9m16vR6PRGF4rpZ56nRUlSjjldFg5wtm5UG6HYDKSq8gKS+o7yTVnmLQglS5dGq1Wa3it1WoNu/OyKjo6Ab1evXAMxupMrTbeKO2+DMn15VlSrjktJ/pOcjU/L5OrlZXmPwcVJj3tu3bt2ty4cYOwsDB0Oh3BwcG4u7ubMgQhhBBmyqQjJHt7exYsWMCIESNISUnBw8ODdu3amTIEIYQQZsokBenAgQOG/7u5ubFz505TLFYIIUQeIndqEEIIYRakIAkhhDALUpCEEEKYBSlIQgghzIIUJCGEEGZBCpIQQgizIAVJCCGEWZCCJIQQwixIQRJCCGEWpCAJIYQwC1KQhBBCmAUpSEIIIcyCSe/2bamOHz/Kl18uJzU1lSpVXJk8eRoFC5rngwZfluSaP3MVwhRkhGRkMTExzJv3MXPmBLBx4zbKlCnLihXLczsso5Bc82euQpiKFCQjO3PmJDVrvkb58hUA6NzZl71796DUiz/11lxJrvkzVyFMRQqSkUVERODiUsrw2tnZhcTERB49SszFqIxDcs2fuQphKlKQjEwpPRqN5pn3rayscyEa45Jc82euQpiKFCQjK1WqNFFRWsPrqCgthQoVxtHRMRejMg7JNX/mKoSpSEEyskaNmvDHHxcJD78FwA8/bKVFC49cjso4JNf8masQpiKnfRtZsWLFmTJlOv7+E0lPT6Ns2XL4+3+c22EZheSaP3MVwlSkIJmAm1tz3Nya53YYJiG5CiFeVK7ssvvxxx/p0KEDbdq04dtvv82NEIQQQpgZk4+QIiIiWLp0Kdu2bcPOzo6ePXvSuHFjqlataupQhBBCmBGTF6Tjx4/TpEkTihYtCkDbtm0JCQlh+PDhWZrfyurZU22zy6VYzp8JlRNxGYPk+nIsKVebIs452l5O9Z3k+nLMKdfM5tUoE19a/uWXX/Lo0SPGjBkDwJYtWzh//jyzZ882ZRhCCCHMjMmPIen1T19QqJTK8AJDIYQQlsXkBal06dJotf93QaFWq8XFxcXUYQghhDAzJi9ITZs25cSJEzx48ICkpCR+/vln3N3dTR2GEEIIM2PykxpKlSrFmDFj6Nu3L2lpafj6+lKrVi1ThyGEEMLMmPykBiGEECIjci87IYQQZkEKkhBCCLMgBUkIIYRZkIIkhBDCLEhByqILFy4wderU5/592bJl7N+/H4A+ffqYKqyXtm3bNiZNmvTC8wcFBREUFPSf03h6enL79u0XXkZGTp06lSv9nJfW7RO51VfmKCvb++bNmwkODjZRRFljjHVojtuFPH4ii958803efPPN5/591KhRhv+fPn3aFCGJXCDrNv87e/YsjRo1yu0wLJIUpCw6deoUy5cvBx4Xp19//ZUHDx7g7++Ph4cHkyZNolGjRvz5558AdOvWjS1btuRavPfv32f8+PE8evQIKysr/P39efToEQsWLEApRZkyZViyZAkAYWFh9OnTh7t37+Lm5sacOXMA+Oqrr9izZw86nY7mzZszYcIENBoNK1euZPPmzRQrVozChQsbriOrXr06f/31F/D4l+jp06dZsGCBISadTkdAQACnT59Gp9PRpUsX+vfv/1J5nj59mqVLl5KcnExcXByTJ0+mSZMmvPXWW+zfvx8nJydu377N+++/z+7du/nhhx9Ys2YNer2e119/nRkzZmBvb0+TJk1444030Gq1fP/999ja2j6zrCf90q1bN7p3787JkycNfRgUFIS9vT0pKSncvXuXv//+m5iYGHr06MGgQYOMknt23bhxg+nTpxMbG0uBAgWYOnUqtWrV4scff2TlypVYW1tTrlw5Fi1aRExMzDPbT506dUwa7/M8+SyuW7cOwPDZW7VqFZUrV+batWuUKVOGRYsWUbRoUX744QdWrFiBk5MTZcuWpUCBAgDs2bOHVatWkZycTGpqKvPmzSM5OZkDBw5w8uRJnJ2dqVmzJtOnT+f+/ftoNBrGjRtH06ZNcyXvBw8eMHjwYG7dukWlSpUIDAxk9+7dGW7P69evZ8eOHSQlJWFra8uSJUuoXLkyR48eZf78+djb21OpUiVD26tWrWL79u1YWVlRq1YtZs2alSs5okSWnDx5UvXu3Vv17t1bzZkzRyml1P79+1Xnzp2VUkpNnDhRbd26VSmlVLVq1XItzieCgoLU//73P6WUUqGhoerLL79Ubm5u6s8//1RKKbV48WK1du1atXXrVuXh4aFiYmJUSkqKatGihbpy5YoKDQ1VI0aMUOnp6Uqn06mxY8eqH374QZ0/f161a9dOJSQkqMTEROXt7a0CAwOVUk/nvXXrVjVx4kSllFKtWrVS4eHhasOGDWrevHlKKaVSUlJU79691ZkzZ14ovyfrY8SIEeratWtKKaWOHz+uvL29lVJKffTRR2rLli2Gvvjyyy/VlStXVK9evVRycrKhDz777DND7CdPnsx0uU9yTEhIUG5ubio+Pl4ppVSbNm3U/fv3VWBgoPL29lYJCQkqLi5OtW7dWl28eDFHc8+uJ33VtWtX9dNPPymllPrtt99Uy5YtVUpKivL09FRRUVFKKaUWLFig/vzzz2e2n5UrV5ok1qx4ks8TTz57/1yH8+fPV7Nnz1b3799XzZo1U1qtVqWlpamBAweqiRMnKp1Op/r27auio6OVUkpt2bJFDRky5Kn2lFJq9OjRat++fUoppSIiItRbb71lWOemdPLkSVWnTh1169YtpdPpVNeuXdX69esz3J7j4+NVv379VFJSklJKqU8//VTNmjVLpaSkqGbNmhk+L1OmTFG9e/dW6enpqnHjxio1NVXpdDo1adIkdf/+fZPnqJRSMkJ6AS1atADA1dWV2NjYXI4mY25ubowYMYJLly7h4eFBvXr12LNnDzVr1gRg3LhxwOORTIMGDQyPA6lQoQIxMTGcOHGC8+fP06VLFwCSk5MpU6YMUVFReHh4ULBgQQDatWuHXq/PUkwnTpzg0qVLnDx5EoBHjx7x119/0aBBgxfOc9GiRRw8eJCQkBB+//13EhMTAejatStBQUH4+voSHBzMmjVr2Lt3L2FhYXTv3h2AtLQ0XnvtNUNbtWvXzvJyCxYsiIeHB3v37qV8+fKUL1+eUqVKAeDt7W3oH09PT06ePMnvv/+e47lnR2JiIrdv36ZNmzYA1KlThyJFinD9+nVatWpFr169aN26NW3btqVmzZo8evToqe2nd+/eJonzZbz66qs0btwYgE6dOjF+/HgaNGhA3bp1KVmyJAA+Pj6cPHkSKysrPvvsMw4cOMCNGzc4ffo0VlbPHlI/fvw4169fJzAwEID09HTCw8MNnyNTqlGjBuXLlwegSpUqxMTEZLg9Ozk5sWTJEnbt2sXNmzc5cuQINWvW5K+//sLFxYUqVaoA0LlzZ5YtW4a1tTV169bF19eXt956iwEDBhi2ZVOTgvQC7O3tAcz6LuX169dn165dHDp0iN27d5OYmPhUvPHx8YYvbxub/9sMNBoNSil0Oh39+vVjwIABAMTFxWFtbc13332H+sfNPWxsbEhNTTW8Vv//7u3p6enPxKTT6ZgwYYLhS/HBgweGL+4X5efnR+PGjWncuDFubm6MHz8egIYNGxIZGcnPP/9MuXLlKFWqFDqdjvbt2+Pv7w88/pLW6XSGthwcHLK17K5du7JixQrKlStnKNwA1tbWhv/r9Xqsra2Nknt2qAxuyPJkPfv7+3P58mVCQ0OZMGECw4cPp2PHjk9tP9u3b2fVqlUmi/e/PNlGn0hLSwOe3o6VUlhbWz8z7ZNpEhMT8fX15Z133qFhw4ZUr149w6dX6/V61qxZY/jBFhkZSYkSJYySV2b+/TktVKhQhtvzvXv36NOnD71798bd3Z2SJUty6dKlZ/rin9vp559/zrlz5zh8+DCDBg1i8eLFuXIcTc6yMwJra+sMv5BNKSAggJ07d9K5c2emT5/OlStXiI6O5tq1awCsXLmSjRs3Pnf+Jk2asGPHDhITE0lPT2fYsGH89NNPuLm5cfDgQeLj40lJSWHv3r2GeYoVK8bVq1dRSnHgwIEM29y8eTNpaWkkJibi5+fHuXPnXjjH2NhYbt68yahRo3B3d2f//v2GAqPRaOjUqRNz5swxFIvGjRuzd+9eoqOjUUoxc+ZM1qxZk61l/nPdNmjQgPv373Pq1Clat25tmGbfvn2kpqby8OFDDh48SPPmzXM89+xycnKiXLly/PzzzwCcO3eOqKgoXF1dadOmDcWKFWPIkCF07NiRS5cuPbP9PDk2ag6KFStGeHg4KSkpxMbG8uuvvwKPj5FdunQJgK1bt+Lu7k79+vU5d+4cERER6PV6du/eDcDNmzfRaDQMHTrUsF082Xae/ICAx9vshg0bALh27Ro+Pj4kJSWZOuXnymh7vnDhAhUrVqR///68+eab7Nu3D51OR/Xq1YmKiuLy5csA7Nq1C3j846hDhw5Uq1aNUaNG0axZM8OxYFOTEZIRvPXWW3Ts2JFt27YZRlOm1qdPH8aNG8e2bduwtrZm4cKFFChQgI8++oi0tDQqVKhAQEAAP/30U4bze3p6cvnyZbp3745Op6NFixZ07twZjUZDv3798PX1pXDhwpQpU8Ywz7hx4xg6dCglS5akfv36xMTEPNVmz549CQsLo3PnzqSnp9OlSxfDLpYXUbRoUZo2bYqXlxc2NjY0adKE5ORkHj16RIECBfDy8uKbb74xFIsaNWowfPhw+vXrh16vp2bNmrz//vvZWua/1+3bb79NbGwsdnZ2hmns7e3x8/MjISGBIUOGULVqVSpWrJijub+IRYsWMXPmTIKCgrC1tSUoKAg7OztGjhzJwIEDsbe3p0SJEixYsIDU1NRnth9z4erqioeHB15eXpQtW5b69esDUKRIEQIDA7l16xbVq1dnzpw5FChQAH9/f/r374+joyNVq1YFHm8LNWvWpH379mg0Gpo3b24obE2bNuWTTz6hUKFC+Pv7M336dHx8fIDHP/ScnJxyJ/F/KVSoUIbbc3p6Ohs3bqRDhw4opWjYsCFXr17F1taWTz75hAkTJmBjY2PYXV28eHF69OiBr68vjo6OVKpUia5du+ZKTnJzVZEv6fV6Nm7cyI0bNwy7NHKSUoq0tDQGDBjAlClTeP311wEM12SNGDEix5cpnu/27dv07ds3w5G5yDtkhCTypeHDh3Pv3j2+/vrrbM1369at5xaTOXPmGK5F02q1eHl50a1bN0MxEkK8HBkhCSGEMAtyUoMQQgizIAVJCCGEWZCCJIQQwixIQRJCCGEWpCAJIYQwC1KQhBBCmIX/B1NXHcGfGLRPAAAAAElFTkSuQmCC\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "number1 = []\n",
    "number2 = []\n",
    "number3 = []\n",
    "for row in final: \n",
    "    number1.append(row[1])\n",
    "    if len(row) >= 3:\n",
    "        number2.append(row[2])\n",
    "    else: \n",
    "        number2.append(0)\n",
    "    if len(row) == 4:\n",
    "        number3.append(row[3])  \n",
    "values = np.array([number1,number2]).T\n",
    "labels = [\"init\",\"schedule\",\"layer_type\",\"loss\",\"update\",\"heads\"]\n",
    "sns.set()\n",
    "plot_difference(labels,values[:,0],values[:,1])"
   ]
  }
 ],
 "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
}