atomsci/ddm/examples/tutorials/12_AMPL_HPO_demo.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {
    "colab_type": "text",
    "id": "view-in-github"
   },
   "source": [
    "<a href=\"https://colab.research.google.com/github/ATOMScience-org/AMPL/blob/master/atomsci/ddm/examples/tutorials/12_AMPL_HPO_demo.ipynb\" target=\"_parent\"><img src=\"https://colab.research.google.com/assets/colab-badge.svg\" alt=\"Open In Colab\"/></a>"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "BkwPgkYKIb8r"
   },
   "source": [
    "# Hyperparameter Optimization using AMPL "
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "gg0kE2IRIlNI"
   },
   "source": [
    "# Time to run the notebook: ~ 22 minutes on COLAB with GPU accelerator turned on. "
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "SZ70ierx3pq2"
   },
   "source": [
    "### Install AMPL packages"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "id": "nNuv4Q5KFABw"
   },
   "outputs": [],
   "source": [
    "! pip install rdkit-pypi\n",
    "! pip install deepchem\n",
    "\n",
    "import deepchem\n",
    "# print(deepchem.__version__)\n",
    "! pip install umap\n",
    "! pip install -U --ignore-installed numba\n",
    "! pip install umap-learn\n",
    "! pip install molvs\n",
    "! pip install bravado"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "id": "yS2DjAnF9y49"
   },
   "outputs": [],
   "source": [
    "import deepchem as dc\n",
    "\n",
    "# get the Install AMPL_GPU_test.sh\n",
    "!wget 'https://raw.githubusercontent.com/ATOMScience-org/AMPL/master/atomsci/ddm/examples/tutorials/config/install_AMPL_GPU_test.sh' >& /dev/null\n",
    "\n",
    "# run the script to install AMPL\n",
    "! chmod u+x install_AMPL_GPU_test.sh\n",
    "! ./install_AMPL_GPU_test.sh"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {
    "id": "InUQE4II1noi"
   },
   "outputs": [],
   "source": [
    "!wget https://raw.githubusercontent.com/ATOMScience-org/AMPL/master/atomsci/ddm/examples/tutorials/datasets/H1_std.csv >& /dev/null"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/"
    },
    "id": "2PO0-2HKvVpn",
    "outputId": "5cf722dc-8574-4afa-eb33-68419bd9a41f"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Requirement already satisfied: hyperopt in /usr/local/lib/python3.7/dist-packages (0.1.2)\n",
      "Requirement already satisfied: numpy in /usr/local/lib/python3.7/dist-packages (from hyperopt) (1.19.5)\n",
      "Requirement already satisfied: pymongo in /usr/local/lib/python3.7/dist-packages (from hyperopt) (3.12.0)\n",
      "Requirement already satisfied: scipy in /usr/local/lib/python3.7/dist-packages (from hyperopt) (1.4.1)\n",
      "Requirement already satisfied: networkx in /usr/local/lib/python3.7/dist-packages (from hyperopt) (2.6.3)\n",
      "Requirement already satisfied: six in /usr/local/lib/python3.7/dist-packages (from hyperopt) (1.15.0)\n",
      "Requirement already satisfied: tqdm in /usr/local/lib/python3.7/dist-packages (from hyperopt) (4.62.2)\n",
      "Requirement already satisfied: future in /usr/local/lib/python3.7/dist-packages (from hyperopt) (0.16.0)\n"
     ]
    }
   ],
   "source": [
    "!pip install hyperopt"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "mTOwqhvR3xL3"
   },
   "source": [
    "### Load packages and dataset"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/"
    },
    "id": "EeN9kEMO2R9Z",
    "outputId": "d4b2257d-d00e-4dba-eb9a-1aff5c2a7a16"
   },
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "2021-09-18 16:31:38,939 Model tracker client not supported in your environment; will save models in filesystem only.\n"
     ]
    }
   ],
   "source": [
    "import matplotlib.pyplot as plt\n",
    "import seaborn as sns\n",
    "import numpy as np\n",
    "sns.set_context(\"poster\")\n",
    "sns.set_style(\"whitegrid\")\n",
    "sns.set_palette(\"Set2\")\n",
    "\n",
    "import pandas as pd\n",
    "import os, json, sys, glob, pickle\n",
    "\n",
    "from atomsci.ddm.pipeline import model_pipeline as mp\n",
    "from atomsci.ddm.pipeline import parameter_parser as parse\n",
    "from atomsci.ddm.pipeline import perf_data\n",
    "\n",
    "from hyperopt import fmin, tpe, hp, Trials, STATUS_OK"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {
    "id": "C5J-cJNz2V5w"
   },
   "outputs": [],
   "source": [
    "h1 = pd.read_csv(\"H1_std.csv\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 424
    },
    "id": "v_fQuS473N1M",
    "outputId": "7cdf22d9-f840-4294-d613-f4c2ac957e5e"
   },
   "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>rdkit_smiles</th>\n",
       "      <th>compound_id</th>\n",
       "      <th>NofA</th>\n",
       "      <th>pKi_mean</th>\n",
       "      <th>pKi_std</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>C(#Cc1c2c(nc3ccccc13)CCCCC2)CCN1CCCCC1</td>\n",
       "      <td>CHEMBL66660</td>\n",
       "      <td>1</td>\n",
       "      <td>4.803188</td>\n",
       "      <td>0.471982</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>C(#Cc1cccc(CN2CCOCC2)c1)CCN1CCCCC1</td>\n",
       "      <td>CHEMBL237087</td>\n",
       "      <td>1</td>\n",
       "      <td>5.000000</td>\n",
       "      <td>0.181920</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>C(=C/c1ccccc1)\\CN1CCN(C(c2ccccc2)c2ccccc2)CC1</td>\n",
       "      <td>CHEMBL43064</td>\n",
       "      <td>1</td>\n",
       "      <td>8.064392</td>\n",
       "      <td>0.163978</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>C/C=C/C(=O)N1CCC(=C2c3ccc(Cl)cc3CCc3cccnc32)CC1</td>\n",
       "      <td>CHEMBL3357041</td>\n",
       "      <td>1</td>\n",
       "      <td>6.815309</td>\n",
       "      <td>0.095668</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>C1=C(CCN2CCc3ccccc3C2)CCc2ccccc21</td>\n",
       "      <td>CHEMBL3970994</td>\n",
       "      <td>2</td>\n",
       "      <td>5.841272</td>\n",
       "      <td>0.001272</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>...</th>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "      <td>...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1353</th>\n",
       "      <td>c1cnc(N2CCN(CCCc3nc4ccccc4s3)CC2)nc1</td>\n",
       "      <td>CHEMBL3289656</td>\n",
       "      <td>2</td>\n",
       "      <td>6.138337</td>\n",
       "      <td>0.001660</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1354</th>\n",
       "      <td>c1cnc2c(N3CCN(CCCCc4ccc(OCCCN5CCCCCC5)cc4)CC3)...</td>\n",
       "      <td>CHEMBL3893197</td>\n",
       "      <td>1</td>\n",
       "      <td>7.799971</td>\n",
       "      <td>0.408169</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1355</th>\n",
       "      <td>c1cnc2c(N3CCN(CCCc4ccc(OCCCN5CCCCCC5)cc4)CC3)c...</td>\n",
       "      <td>CHEMBL3921827</td>\n",
       "      <td>1</td>\n",
       "      <td>7.100015</td>\n",
       "      <td>0.408169</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1356</th>\n",
       "      <td>c1cnc2c(N3CCN(CCc4ccc(OCCCN5CCCCCC5)cc4)CC3)cc...</td>\n",
       "      <td>CHEMBL3947980</td>\n",
       "      <td>1</td>\n",
       "      <td>7.700057</td>\n",
       "      <td>0.408169</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1357</th>\n",
       "      <td>c1cnc2c(c1)cc([C@@H]1CCCO1)n2CCN1CCCCC1</td>\n",
       "      <td>CHEMBL3557423</td>\n",
       "      <td>1</td>\n",
       "      <td>5.750000</td>\n",
       "      <td>0.181920</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "<p>1358 rows × 5 columns</p>\n",
       "</div>"
      ],
      "text/plain": [
       "                                           rdkit_smiles  ...   pKi_std\n",
       "0                C(#Cc1c2c(nc3ccccc13)CCCCC2)CCN1CCCCC1  ...  0.471982\n",
       "1                    C(#Cc1cccc(CN2CCOCC2)c1)CCN1CCCCC1  ...  0.181920\n",
       "2         C(=C/c1ccccc1)\\CN1CCN(C(c2ccccc2)c2ccccc2)CC1  ...  0.163978\n",
       "3       C/C=C/C(=O)N1CCC(=C2c3ccc(Cl)cc3CCc3cccnc32)CC1  ...  0.095668\n",
       "4                     C1=C(CCN2CCc3ccccc3C2)CCc2ccccc21  ...  0.001272\n",
       "...                                                 ...  ...       ...\n",
       "1353               c1cnc(N2CCN(CCCc3nc4ccccc4s3)CC2)nc1  ...  0.001660\n",
       "1354  c1cnc2c(N3CCN(CCCCc4ccc(OCCCN5CCCCCC5)cc4)CC3)...  ...  0.408169\n",
       "1355  c1cnc2c(N3CCN(CCCc4ccc(OCCCN5CCCCCC5)cc4)CC3)c...  ...  0.408169\n",
       "1356  c1cnc2c(N3CCN(CCc4ccc(OCCCN5CCCCCC5)cc4)CC3)cc...  ...  0.408169\n",
       "1357            c1cnc2c(c1)cc([C@@H]1CCCO1)n2CCN1CCCCC1  ...  0.181920\n",
       "\n",
       "[1358 rows x 5 columns]"
      ]
     },
     "execution_count": 7,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "h1"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "XA8ARleo3h53"
   },
   "source": [
    "### Split the dataset"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/"
    },
    "id": "1eQC31_fA2wD",
    "outputId": "880d7d75-e6f0-44b7-cd17-f0893240244d"
   },
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "2021-09-18 16:31:39,404 Attempting to load featurized dataset\n",
      "2021-09-18 16:31:39,413 Exception when trying to load featurized data:\n",
      "DynamicFeaturization doesn't support get_featurized_dset_name()\n",
      "2021-09-18 16:31:39,415 Featurized dataset not previously saved for dataset H1_std, creating new\n",
      "2021-09-18 16:31:39,426 Featurizing sample 0\n",
      "2021-09-18 16:31:44,716 Featurizing sample 1000\n",
      "2021-09-18 16:31:46,558 Splitting data by scaffold\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "number of features: 1024\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "2021-09-18 16:31:47,224 Dataset split table saved to /H1_std_train_valid_test_scaffold_86608b45-173c-470b-9327-78dc94d9fdf2.csv\n"
     ]
    }
   ],
   "source": [
    "train_file = \"H1_std.csv\"\n",
    "response_col = \"pKi_mean\"\n",
    "compound_id = \"compound_id\"\n",
    "smiles_col = \"rdkit_smiles\"\n",
    "\n",
    "params = {\n",
    "        \"system\": \"LC\",\n",
    "        \"lc_account\": 'None',\n",
    "        \"datastore\": \"False\",\n",
    "        \"save_results\": \"False\",\n",
    "        \"data_owner\": \"username\",\n",
    "        \"prediction_type\": \"regression\",\n",
    "        \"dataset_key\": train_file,\n",
    "        \"id_col\": compound_id,\n",
    "        \"smiles_col\": smiles_col,\n",
    "        \"response_cols\": response_col,\n",
    "        \"previously_split\": \"False\",\n",
    "        \"split_only\": \"True\",\n",
    "        \"featurizer\": \"ecfp\",\n",
    "        \"model_type\": \"RF\",\n",
    "        \"verbose\": \"True\",\n",
    "        \"transformers\": \"True\",\n",
    "        'max_epochs': '70',\n",
    "        \"rerun\": \"False\",\n",
    "        \"result_dir\": \"split\"\n",
    "    }\n",
    "\n",
    "ampl_param = parse.wrapper(params)\n",
    "pl = mp.ModelPipeline(ampl_param)\n",
    "pl.train_model()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "xSdgxZNU43TM"
   },
   "source": [
    "### Train an initial model"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {
    "id": "IjhIVGpJ42eE"
   },
   "outputs": [],
   "source": [
    "listing = glob.glob(os.path.join(\"/\", \"H1_std_train_valid_test_scaffold_*\"))\n",
    "name = listing[0]\n",
    "split_uuid = name.split(\"_\")[-1].split(\".\")[0]\n",
    "response_col = \"pKi_mean\"\n",
    "\n",
    "params = {\n",
    "        \"system\": \"LC\",\n",
    "        \"lc_account\": 'None',\n",
    "        \"datastore\": \"False\",\n",
    "        \"save_results\": \"False\",\n",
    "        \"data_owner\": \"username\",\n",
    "        \"prediction_type\": \"regression\",\n",
    "        \"dataset_key\": train_file,\n",
    "        \"id_col\": compound_id,\n",
    "        \"smiles_col\": smiles_col,\n",
    "        \"response_cols\": response_col,\n",
    "        \"split_uuid\": split_uuid,\n",
    "        \"previously_split\": \"True\",\n",
    "        \"featurizer\": \"ecfp\",\n",
    "        \"model_type\": \"RF\",\n",
    "        \"verbose\": \"True\",\n",
    "        \"transformers\": \"True\",\n",
    "        'max_epochs': '70',\n",
    "        \"rerun\": \"False\",\n",
    "        \"result_dir\": \"ampl_output\"\n",
    "    }"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/"
    },
    "id": "qjbLbmri5Rhl",
    "outputId": "e23051eb-fed5-41b7-dbaf-5caef43a8f5c"
   },
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "2021-09-18 16:31:47,261 Attempting to load featurized dataset\n",
      "2021-09-18 16:31:47,271 Exception when trying to load featurized data:\n",
      "DynamicFeaturization doesn't support get_featurized_dset_name()\n",
      "2021-09-18 16:31:47,272 Featurized dataset not previously saved for dataset H1_std, creating new\n",
      "2021-09-18 16:31:47,283 Featurizing sample 0\n",
      "2021-09-18 16:31:52,281 Featurizing sample 1000\n",
      "2021-09-18 16:31:54,179 Previous dataset split restored\n",
      "2021-09-18 16:31:54,196 Wrote transformers to ampl_output/H1_std/RF_ecfp_scaffold_regression/d8983828-904f-4323-82fd-66fcea944f38/transformers.pkl\n",
      "2021-09-18 16:31:54,197 Transforming response data\n",
      "2021-09-18 16:31:54,224 Transforming response data\n",
      "2021-09-18 16:31:54,238 Transforming response data\n",
      "2021-09-18 16:31:54,268 Fitting random forest model\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "number of features: 1024\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "2021-09-18 16:31:57,038 Fold 0: training r2_score = 0.967, validation r2_score = 0.444, test r2_score = 0.527\n",
      "2021-09-18 16:31:58,465 Wrote model tarball to ampl_output/H1_std_model_d8983828-904f-4323-82fd-66fcea944f38.tar.gz\n"
     ]
    }
   ],
   "source": [
    "ampl_param = parse.wrapper(params)\n",
    "pl = mp.ModelPipeline(ampl_param)\n",
    "pl.train_model()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/"
    },
    "id": "ol3jooPm5Yd0",
    "outputId": "975ff814-eebd-4d16-baec-373413ed5073"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "R2 score of validation set: 0.444\n"
     ]
    }
   ],
   "source": [
    "pred_data = pl.model_wrapper.get_perf_data(subset=\"valid\", epoch_label=\"best\")\n",
    "pred_results = pred_data.get_prediction_results()\n",
    "print(f\"R2 score of validation set: {pred_results['r2_score']:.3f}\")"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "vJ2mT3kfi_ed"
   },
   "source": [
    "### Train an model with uncertainty prediction"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {
    "id": "WKUZYCUDi_en"
   },
   "outputs": [],
   "source": [
    "listing = glob.glob(os.path.join(\"/\", \"H1_std_train_valid_test_scaffold_*\"))\n",
    "name = listing[0]\n",
    "split_uuid = name.split(\"_\")[-1].split(\".\")[0]\n",
    "response_col = \"pKi_mean\"\n",
    "\n",
    "params = {\n",
    "        \"system\": \"LC\",\n",
    "        \"lc_account\": 'None',\n",
    "        \"datastore\": \"False\",\n",
    "        \"save_results\": \"False\",\n",
    "        \"data_owner\": \"username\",\n",
    "        \"prediction_type\": \"regression\",\n",
    "        \"dataset_key\": train_file,\n",
    "        \"id_col\": compound_id,\n",
    "        \"smiles_col\": smiles_col,\n",
    "        \"response_cols\": response_col,\n",
    "        \"split_uuid\": split_uuid,\n",
    "        \"previously_split\": \"True\",\n",
    "        \"featurizer\": \"ecfp\",\n",
    "        \"model_type\": \"RF\",\n",
    "        \"verbose\": \"True\",\n",
    "        \"transformers\": \"True\",\n",
    "        'max_epochs': '70',\n",
    "        \"rerun\": \"False\",\n",
    "        \"result_dir\": \"ampl_output\",\n",
    "        \"uncertainty\": \"True\"\n",
    "    }"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/"
    },
    "id": "zwSTjKBei_eo",
    "outputId": "3f91728b-87b3-49e7-bb27-b9ab7ad5895a"
   },
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "2021-09-18 16:31:58,519 Attempting to load featurized dataset\n",
      "2021-09-18 16:31:58,528 Exception when trying to load featurized data:\n",
      "DynamicFeaturization doesn't support get_featurized_dset_name()\n",
      "2021-09-18 16:31:58,529 Featurized dataset not previously saved for dataset H1_std, creating new\n",
      "2021-09-18 16:31:58,541 Featurizing sample 0\n",
      "2021-09-18 16:32:03,635 Featurizing sample 1000\n",
      "2021-09-18 16:32:05,519 Previous dataset split restored\n",
      "2021-09-18 16:32:05,535 Wrote transformers to ampl_output/H1_std/RF_ecfp_scaffold_regression/50f6f1b0-e0ca-40c0-adca-d6c570b6c52f/transformers.pkl\n",
      "2021-09-18 16:32:05,536 Transforming response data\n",
      "2021-09-18 16:32:05,562 Transforming response data\n",
      "2021-09-18 16:32:05,576 Transforming response data\n",
      "2021-09-18 16:32:05,607 Fitting random forest model\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "number of features: 1024\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "2021-09-18 16:32:08,152 Fold 0: training r2_score = 0.966, validation r2_score = 0.449, test r2_score = 0.519\n",
      "2021-09-18 16:32:09,567 Wrote model tarball to ampl_output/H1_std_model_50f6f1b0-e0ca-40c0-adca-d6c570b6c52f.tar.gz\n"
     ]
    }
   ],
   "source": [
    "ampl_param = parse.wrapper(params)\n",
    "pl = mp.ModelPipeline(ampl_param)\n",
    "pl.train_model()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/"
    },
    "id": "wy8fLhm9i_eo",
    "outputId": "5744b03e-5f58-412f-ff88-7ea4cd5a9b8f"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "R2 score of validation set: 0.449\n"
     ]
    }
   ],
   "source": [
    "pred_data = pl.model_wrapper.get_perf_data(subset=\"valid\", epoch_label=\"best\")\n",
    "pred_results = pred_data.get_prediction_results()\n",
    "print(f\"R2 score of validation set: {pred_results['r2_score']:.3f}\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/"
    },
    "id": "wnaag-HhjY-X",
    "outputId": "1bda77e5-289c-46f8-daac-399c30724d03"
   },
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "2021-09-18 16:32:09,586 Featurizing data...\n",
      "2021-09-18 16:32:09,591 Featurizing sample 0\n",
      "2021-09-18 16:32:14,568 Featurizing sample 1000\n",
      "2021-09-18 16:32:16,399 Done\n",
      "2021-09-18 16:32:16,419 Transforming response data\n",
      "2021-09-18 16:32:16,452 Evaluating current model\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "number of features: 1024\n"
     ]
    }
   ],
   "source": [
    "pred_df = pl.predict_on_dataframe(h1)\n",
    "pred_df = pred_df.merge(h1[[\"rdkit_smiles\", \"pKi_mean\"]], on=\"rdkit_smiles\")\n",
    "pred_df = pred_df.assign(pred_error=pred_df.pKi_mean-pred_df.pred)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 546
    },
    "id": "h4dHK5omm0YJ",
    "outputId": "0293b713-4827-47e7-db7d-644b29572e7a"
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "<matplotlib.axes._subplots.AxesSubplot at 0x7f9bb08becd0>"
      ]
     },
     "execution_count": 16,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "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",
      "text/plain": [
       "<Figure size 576x576 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "fig, ax = plt.subplots(1,1,figsize=(8,8))\n",
    "sns.scatterplot(data=pred_df, x=\"std\", y=\"pred_error\", ax=ax)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "pKP_mFuB5giz"
   },
   "source": [
    "### Grid search\n",
    "#### Tunable parameters for Random Forest models\n",
    "- rf_estimators: number of estimators to use in random forest models\n",
    "- rf_max_depth: the maximum depth of a decision tree in the random forest\n",
    "- rf_max_features: max number of features to split random forest nodes"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "3FFnpPw1PWqx"
   },
   "source": [
    "## Time for running the following code chunk: ~ 10 minutes on COLAB-Pro"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/"
    },
    "id": "i6FBeUGh5mGJ",
    "outputId": "d4ecfd7e-7ed8-4823-b1ab-5799ac739ca8"
   },
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "2021-09-18 16:32:20,454 Attempting to load featurized dataset\n",
      "2021-09-18 16:32:20,463 Exception when trying to load featurized data:\n",
      "DynamicFeaturization doesn't support get_featurized_dset_name()\n",
      "2021-09-18 16:32:20,464 Featurized dataset not previously saved for dataset H1_std, creating new\n",
      "2021-09-18 16:32:20,475 Featurizing sample 0\n",
      "2021-09-18 16:32:25,477 Featurizing sample 1000\n",
      "2021-09-18 16:32:27,368 Previous dataset split restored\n",
      "2021-09-18 16:32:27,385 Wrote transformers to ampl_output/H1_std/RF_ecfp_scaffold_regression/d4b5209b-364b-4022-b010-31438af07e89/transformers.pkl\n",
      "2021-09-18 16:32:27,386 Transforming response data\n",
      "2021-09-18 16:32:27,415 Transforming response data\n",
      "2021-09-18 16:32:27,431 Transforming response data\n",
      "2021-09-18 16:32:27,461 Fitting random forest model\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "number of features: 1024\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "2021-09-18 16:32:27,913 Fold 0: training r2_score = 0.865, validation r2_score = 0.396, test r2_score = 0.431\n",
      "2021-09-18 16:32:27,950 Wrote model tarball to ampl_output/H1_std_model_d4b5209b-364b-4022-b010-31438af07e89.tar.gz\n",
      "2021-09-18 16:32:27,964 Attempting to load featurized dataset\n",
      "2021-09-18 16:32:27,972 Exception when trying to load featurized data:\n",
      "DynamicFeaturization doesn't support get_featurized_dset_name()\n",
      "2021-09-18 16:32:27,974 Featurized dataset not previously saved for dataset H1_std, creating new\n",
      "2021-09-18 16:32:27,983 Featurizing sample 0\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "rf_estimators: 16, rf_max_depth: 16, rf_max_features: 16, valid_r2: 0.3960169267773953\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "2021-09-18 16:32:32,920 Featurizing sample 1000\n",
      "2021-09-18 16:32:34,871 Previous dataset split restored\n",
      "2021-09-18 16:32:34,887 Wrote transformers to ampl_output/H1_std/RF_ecfp_scaffold_regression/05d6c5ee-2cca-48ce-ac94-05abaf7f992c/transformers.pkl\n",
      "2021-09-18 16:32:34,890 Transforming response data\n",
      "2021-09-18 16:32:34,916 Transforming response data\n",
      "2021-09-18 16:32:34,930 Transforming response data\n",
      "2021-09-18 16:32:34,960 Fitting random forest model\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "number of features: 1024\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "2021-09-18 16:32:35,413 Fold 0: training r2_score = 0.899, validation r2_score = 0.351, test r2_score = 0.502\n",
      "2021-09-18 16:32:35,452 Wrote model tarball to ampl_output/H1_std_model_05d6c5ee-2cca-48ce-ac94-05abaf7f992c.tar.gz\n",
      "2021-09-18 16:32:35,466 Attempting to load featurized dataset\n",
      "2021-09-18 16:32:35,475 Exception when trying to load featurized data:\n",
      "DynamicFeaturization doesn't support get_featurized_dset_name()\n",
      "2021-09-18 16:32:35,476 Featurized dataset not previously saved for dataset H1_std, creating new\n",
      "2021-09-18 16:32:35,486 Featurizing sample 0\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "rf_estimators: 16, rf_max_depth: 16, rf_max_features: 32, valid_r2: 0.3505295881284277\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "2021-09-18 16:32:40,608 Featurizing sample 1000\n",
      "2021-09-18 16:32:42,552 Previous dataset split restored\n",
      "2021-09-18 16:32:42,569 Wrote transformers to ampl_output/H1_std/RF_ecfp_scaffold_regression/a54aa9fc-a2d1-4ee1-967e-468447805588/transformers.pkl\n",
      "2021-09-18 16:32:42,571 Transforming response data\n",
      "2021-09-18 16:32:42,598 Transforming response data\n",
      "2021-09-18 16:32:42,612 Transforming response data\n",
      "2021-09-18 16:32:42,640 Fitting random forest model\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "number of features: 1024\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "2021-09-18 16:32:43,090 Fold 0: training r2_score = 0.915, validation r2_score = 0.460, test r2_score = 0.520\n",
      "2021-09-18 16:32:43,130 Wrote model tarball to ampl_output/H1_std_model_a54aa9fc-a2d1-4ee1-967e-468447805588.tar.gz\n",
      "2021-09-18 16:32:43,144 Attempting to load featurized dataset\n",
      "2021-09-18 16:32:43,151 Exception when trying to load featurized data:\n",
      "DynamicFeaturization doesn't support get_featurized_dset_name()\n",
      "2021-09-18 16:32:43,152 Featurized dataset not previously saved for dataset H1_std, creating new\n",
      "2021-09-18 16:32:43,165 Featurizing sample 0\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "rf_estimators: 16, rf_max_depth: 16, rf_max_features: 64, valid_r2: 0.4600073638861877\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "2021-09-18 16:32:48,191 Featurizing sample 1000\n",
      "2021-09-18 16:32:50,105 Previous dataset split restored\n",
      "2021-09-18 16:32:50,124 Wrote transformers to ampl_output/H1_std/RF_ecfp_scaffold_regression/15888261-1735-4984-b285-bac3dece668e/transformers.pkl\n",
      "2021-09-18 16:32:50,125 Transforming response data\n",
      "2021-09-18 16:32:50,151 Transforming response data\n",
      "2021-09-18 16:32:50,165 Transforming response data\n",
      "2021-09-18 16:32:50,194 Fitting random forest model\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "number of features: 1024\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "2021-09-18 16:32:50,644 Fold 0: training r2_score = 0.950, validation r2_score = 0.391, test r2_score = 0.501\n",
      "2021-09-18 16:32:50,697 Wrote model tarball to ampl_output/H1_std_model_15888261-1735-4984-b285-bac3dece668e.tar.gz\n",
      "2021-09-18 16:32:50,709 Attempting to load featurized dataset\n",
      "2021-09-18 16:32:50,718 Exception when trying to load featurized data:\n",
      "DynamicFeaturization doesn't support get_featurized_dset_name()\n",
      "2021-09-18 16:32:50,719 Featurized dataset not previously saved for dataset H1_std, creating new\n",
      "2021-09-18 16:32:50,729 Featurizing sample 0\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "rf_estimators: 16, rf_max_depth: 32, rf_max_features: 16, valid_r2: 0.390933442289513\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "2021-09-18 16:32:55,723 Featurizing sample 1000\n",
      "2021-09-18 16:32:57,681 Previous dataset split restored\n",
      "2021-09-18 16:32:57,699 Wrote transformers to ampl_output/H1_std/RF_ecfp_scaffold_regression/896b11ea-4989-49ff-9002-4350b0fbfa7e/transformers.pkl\n",
      "2021-09-18 16:32:57,700 Transforming response data\n",
      "2021-09-18 16:32:57,725 Transforming response data\n",
      "2021-09-18 16:32:57,740 Transforming response data\n",
      "2021-09-18 16:32:57,772 Fitting random forest model\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "number of features: 1024\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "2021-09-18 16:32:58,225 Fold 0: training r2_score = 0.953, validation r2_score = 0.439, test r2_score = 0.508\n",
      "2021-09-18 16:32:58,279 Wrote model tarball to ampl_output/H1_std_model_896b11ea-4989-49ff-9002-4350b0fbfa7e.tar.gz\n",
      "2021-09-18 16:32:58,291 Attempting to load featurized dataset\n",
      "2021-09-18 16:32:58,300 Exception when trying to load featurized data:\n",
      "DynamicFeaturization doesn't support get_featurized_dset_name()\n",
      "2021-09-18 16:32:58,302 Featurized dataset not previously saved for dataset H1_std, creating new\n",
      "2021-09-18 16:32:58,312 Featurizing sample 0\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "rf_estimators: 16, rf_max_depth: 32, rf_max_features: 32, valid_r2: 0.4391800346215625\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "2021-09-18 16:33:03,289 Featurizing sample 1000\n",
      "2021-09-18 16:33:05,169 Previous dataset split restored\n",
      "2021-09-18 16:33:05,187 Wrote transformers to ampl_output/H1_std/RF_ecfp_scaffold_regression/ea2ef382-78ef-4044-bf9c-3f349602179d/transformers.pkl\n",
      "2021-09-18 16:33:05,188 Transforming response data\n",
      "2021-09-18 16:33:05,213 Transforming response data\n",
      "2021-09-18 16:33:05,227 Transforming response data\n",
      "2021-09-18 16:33:05,255 Fitting random forest model\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "number of features: 1024\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "2021-09-18 16:33:05,704 Fold 0: training r2_score = 0.959, validation r2_score = 0.422, test r2_score = 0.458\n",
      "2021-09-18 16:33:05,756 Wrote model tarball to ampl_output/H1_std_model_ea2ef382-78ef-4044-bf9c-3f349602179d.tar.gz\n",
      "2021-09-18 16:33:05,769 Attempting to load featurized dataset\n",
      "2021-09-18 16:33:05,777 Exception when trying to load featurized data:\n",
      "DynamicFeaturization doesn't support get_featurized_dset_name()\n",
      "2021-09-18 16:33:05,779 Featurized dataset not previously saved for dataset H1_std, creating new\n",
      "2021-09-18 16:33:05,789 Featurizing sample 0\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "rf_estimators: 16, rf_max_depth: 32, rf_max_features: 64, valid_r2: 0.42175707291938436\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "2021-09-18 16:33:10,776 Featurizing sample 1000\n",
      "2021-09-18 16:33:12,732 Previous dataset split restored\n",
      "2021-09-18 16:33:12,750 Wrote transformers to ampl_output/H1_std/RF_ecfp_scaffold_regression/bf6c9935-ba64-40f4-a24a-3564588f9d96/transformers.pkl\n",
      "2021-09-18 16:33:12,751 Transforming response data\n",
      "2021-09-18 16:33:12,780 Transforming response data\n",
      "2021-09-18 16:33:12,793 Transforming response data\n",
      "2021-09-18 16:33:12,824 Fitting random forest model\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "number of features: 1024\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "2021-09-18 16:33:13,276 Fold 0: training r2_score = 0.955, validation r2_score = 0.352, test r2_score = 0.500\n",
      "2021-09-18 16:33:13,330 Wrote model tarball to ampl_output/H1_std_model_bf6c9935-ba64-40f4-a24a-3564588f9d96.tar.gz\n",
      "2021-09-18 16:33:13,342 Attempting to load featurized dataset\n",
      "2021-09-18 16:33:13,351 Exception when trying to load featurized data:\n",
      "DynamicFeaturization doesn't support get_featurized_dset_name()\n",
      "2021-09-18 16:33:13,352 Featurized dataset not previously saved for dataset H1_std, creating new\n",
      "2021-09-18 16:33:13,362 Featurizing sample 0\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "rf_estimators: 16, rf_max_depth: 64, rf_max_features: 16, valid_r2: 0.35199459851153847\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "2021-09-18 16:33:18,437 Featurizing sample 1000\n",
      "2021-09-18 16:33:20,389 Previous dataset split restored\n",
      "2021-09-18 16:33:20,407 Wrote transformers to ampl_output/H1_std/RF_ecfp_scaffold_regression/ca8dcd05-6616-4ecc-898d-aafb22732342/transformers.pkl\n",
      "2021-09-18 16:33:20,408 Transforming response data\n",
      "2021-09-18 16:33:20,434 Transforming response data\n",
      "2021-09-18 16:33:20,448 Transforming response data\n",
      "2021-09-18 16:33:20,480 Fitting random forest model\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "number of features: 1024\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "2021-09-18 16:33:20,931 Fold 0: training r2_score = 0.956, validation r2_score = 0.422, test r2_score = 0.482\n",
      "2021-09-18 16:33:20,984 Wrote model tarball to ampl_output/H1_std_model_ca8dcd05-6616-4ecc-898d-aafb22732342.tar.gz\n",
      "2021-09-18 16:33:20,998 Attempting to load featurized dataset\n",
      "2021-09-18 16:33:21,007 Exception when trying to load featurized data:\n",
      "DynamicFeaturization doesn't support get_featurized_dset_name()\n",
      "2021-09-18 16:33:21,009 Featurized dataset not previously saved for dataset H1_std, creating new\n",
      "2021-09-18 16:33:21,019 Featurizing sample 0\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "rf_estimators: 16, rf_max_depth: 64, rf_max_features: 32, valid_r2: 0.4222441123695051\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "2021-09-18 16:33:25,950 Featurizing sample 1000\n",
      "2021-09-18 16:33:27,814 Previous dataset split restored\n",
      "2021-09-18 16:33:27,831 Wrote transformers to ampl_output/H1_std/RF_ecfp_scaffold_regression/1ec0f9d1-9e0e-4c78-bedb-4e20f2bc900f/transformers.pkl\n",
      "2021-09-18 16:33:27,833 Transforming response data\n",
      "2021-09-18 16:33:27,861 Transforming response data\n",
      "2021-09-18 16:33:27,875 Transforming response data\n",
      "2021-09-18 16:33:27,905 Fitting random forest model\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "number of features: 1024\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "2021-09-18 16:33:28,357 Fold 0: training r2_score = 0.957, validation r2_score = 0.375, test r2_score = 0.498\n",
      "2021-09-18 16:33:28,411 Wrote model tarball to ampl_output/H1_std_model_1ec0f9d1-9e0e-4c78-bedb-4e20f2bc900f.tar.gz\n",
      "2021-09-18 16:33:28,423 Attempting to load featurized dataset\n",
      "2021-09-18 16:33:28,432 Exception when trying to load featurized data:\n",
      "DynamicFeaturization doesn't support get_featurized_dset_name()\n",
      "2021-09-18 16:33:28,433 Featurized dataset not previously saved for dataset H1_std, creating new\n",
      "2021-09-18 16:33:28,444 Featurizing sample 0\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "rf_estimators: 16, rf_max_depth: 64, rf_max_features: 64, valid_r2: 0.3749356127735295\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "2021-09-18 16:33:33,521 Featurizing sample 1000\n",
      "2021-09-18 16:33:35,468 Previous dataset split restored\n",
      "2021-09-18 16:33:35,486 Wrote transformers to ampl_output/H1_std/RF_ecfp_scaffold_regression/4096221f-8c88-4589-867c-ce6dad11fc9a/transformers.pkl\n",
      "2021-09-18 16:33:35,487 Transforming response data\n",
      "2021-09-18 16:33:35,515 Transforming response data\n",
      "2021-09-18 16:33:35,529 Transforming response data\n",
      "2021-09-18 16:33:35,559 Fitting random forest model\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "number of features: 1024\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "2021-09-18 16:33:36,017 Fold 0: training r2_score = 0.876, validation r2_score = 0.397, test r2_score = 0.450\n",
      "2021-09-18 16:33:36,086 Wrote model tarball to ampl_output/H1_std_model_4096221f-8c88-4589-867c-ce6dad11fc9a.tar.gz\n",
      "2021-09-18 16:33:36,102 Attempting to load featurized dataset\n",
      "2021-09-18 16:33:36,112 Exception when trying to load featurized data:\n",
      "DynamicFeaturization doesn't support get_featurized_dset_name()\n",
      "2021-09-18 16:33:36,114 Featurized dataset not previously saved for dataset H1_std, creating new\n",
      "2021-09-18 16:33:36,126 Featurizing sample 0\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "rf_estimators: 32, rf_max_depth: 16, rf_max_features: 16, valid_r2: 0.39667846849909705\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "2021-09-18 16:33:41,229 Featurizing sample 1000\n",
      "2021-09-18 16:33:43,152 Previous dataset split restored\n",
      "2021-09-18 16:33:43,170 Wrote transformers to ampl_output/H1_std/RF_ecfp_scaffold_regression/5ba1863f-f287-4611-81e8-51d4f0c0384c/transformers.pkl\n",
      "2021-09-18 16:33:43,172 Transforming response data\n",
      "2021-09-18 16:33:43,198 Transforming response data\n",
      "2021-09-18 16:33:43,212 Transforming response data\n",
      "2021-09-18 16:33:43,245 Fitting random forest model\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "number of features: 1024\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "2021-09-18 16:33:43,806 Fold 0: training r2_score = 0.903, validation r2_score = 0.393, test r2_score = 0.488\n",
      "2021-09-18 16:33:43,874 Wrote model tarball to ampl_output/H1_std_model_5ba1863f-f287-4611-81e8-51d4f0c0384c.tar.gz\n",
      "2021-09-18 16:33:43,886 Attempting to load featurized dataset\n",
      "2021-09-18 16:33:43,894 Exception when trying to load featurized data:\n",
      "DynamicFeaturization doesn't support get_featurized_dset_name()\n",
      "2021-09-18 16:33:43,896 Featurized dataset not previously saved for dataset H1_std, creating new\n",
      "2021-09-18 16:33:43,907 Featurizing sample 0\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "rf_estimators: 32, rf_max_depth: 16, rf_max_features: 32, valid_r2: 0.3927811986768647\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "2021-09-18 16:33:48,879 Featurizing sample 1000\n",
      "2021-09-18 16:33:50,804 Previous dataset split restored\n",
      "2021-09-18 16:33:50,821 Wrote transformers to ampl_output/H1_std/RF_ecfp_scaffold_regression/53989985-4f8a-4177-b398-97cec7ff6417/transformers.pkl\n",
      "2021-09-18 16:33:50,823 Transforming response data\n",
      "2021-09-18 16:33:50,850 Transforming response data\n",
      "2021-09-18 16:33:50,865 Transforming response data\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "number of features: 1024\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "2021-09-18 16:33:50,900 Fitting random forest model\n",
      "2021-09-18 16:33:51,464 Fold 0: training r2_score = 0.922, validation r2_score = 0.448, test r2_score = 0.527\n",
      "2021-09-18 16:33:51,535 Wrote model tarball to ampl_output/H1_std_model_53989985-4f8a-4177-b398-97cec7ff6417.tar.gz\n",
      "2021-09-18 16:33:51,548 Attempting to load featurized dataset\n",
      "2021-09-18 16:33:51,556 Exception when trying to load featurized data:\n",
      "DynamicFeaturization doesn't support get_featurized_dset_name()\n",
      "2021-09-18 16:33:51,557 Featurized dataset not previously saved for dataset H1_std, creating new\n",
      "2021-09-18 16:33:51,569 Featurizing sample 0\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "rf_estimators: 32, rf_max_depth: 16, rf_max_features: 64, valid_r2: 0.4480686493700571\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "2021-09-18 16:33:56,636 Featurizing sample 1000\n",
      "2021-09-18 16:33:58,582 Previous dataset split restored\n",
      "2021-09-18 16:33:58,601 Wrote transformers to ampl_output/H1_std/RF_ecfp_scaffold_regression/efcc4990-d06e-451a-a7f4-a1acad2b6b2d/transformers.pkl\n",
      "2021-09-18 16:33:58,602 Transforming response data\n",
      "2021-09-18 16:33:58,631 Transforming response data\n",
      "2021-09-18 16:33:58,645 Transforming response data\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "number of features: 1024\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "2021-09-18 16:33:58,678 Fitting random forest model\n",
      "2021-09-18 16:33:59,245 Fold 0: training r2_score = 0.959, validation r2_score = 0.426, test r2_score = 0.486\n",
      "2021-09-18 16:33:59,344 Wrote model tarball to ampl_output/H1_std_model_efcc4990-d06e-451a-a7f4-a1acad2b6b2d.tar.gz\n",
      "2021-09-18 16:33:59,364 Attempting to load featurized dataset\n",
      "2021-09-18 16:33:59,374 Exception when trying to load featurized data:\n",
      "DynamicFeaturization doesn't support get_featurized_dset_name()\n",
      "2021-09-18 16:33:59,376 Featurized dataset not previously saved for dataset H1_std, creating new\n",
      "2021-09-18 16:33:59,393 Featurizing sample 0\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "rf_estimators: 32, rf_max_depth: 32, rf_max_features: 16, valid_r2: 0.4262498577011963\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "2021-09-18 16:34:04,344 Featurizing sample 1000\n",
      "2021-09-18 16:34:06,270 Previous dataset split restored\n",
      "2021-09-18 16:34:06,289 Wrote transformers to ampl_output/H1_std/RF_ecfp_scaffold_regression/48c854cb-60c6-4f5a-9108-4932bef62834/transformers.pkl\n",
      "2021-09-18 16:34:06,290 Transforming response data\n",
      "2021-09-18 16:34:06,317 Transforming response data\n",
      "2021-09-18 16:34:06,332 Transforming response data\n",
      "2021-09-18 16:34:06,361 Fitting random forest model\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "number of features: 1024\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "2021-09-18 16:34:06,920 Fold 0: training r2_score = 0.960, validation r2_score = 0.433, test r2_score = 0.480\n",
      "2021-09-18 16:34:07,016 Wrote model tarball to ampl_output/H1_std_model_48c854cb-60c6-4f5a-9108-4932bef62834.tar.gz\n",
      "2021-09-18 16:34:07,031 Attempting to load featurized dataset\n",
      "2021-09-18 16:34:07,040 Exception when trying to load featurized data:\n",
      "DynamicFeaturization doesn't support get_featurized_dset_name()\n",
      "2021-09-18 16:34:07,042 Featurized dataset not previously saved for dataset H1_std, creating new\n",
      "2021-09-18 16:34:07,053 Featurizing sample 0\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "rf_estimators: 32, rf_max_depth: 32, rf_max_features: 32, valid_r2: 0.43324663147638043\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "2021-09-18 16:34:12,151 Featurizing sample 1000\n",
      "2021-09-18 16:34:14,199 Previous dataset split restored\n",
      "2021-09-18 16:34:14,216 Wrote transformers to ampl_output/H1_std/RF_ecfp_scaffold_regression/176a6067-f1a8-47b1-a201-703afedfea5d/transformers.pkl\n",
      "2021-09-18 16:34:14,218 Transforming response data\n",
      "2021-09-18 16:34:14,244 Transforming response data\n",
      "2021-09-18 16:34:14,258 Transforming response data\n",
      "2021-09-18 16:34:14,289 Fitting random forest model\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "number of features: 1024\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "2021-09-18 16:34:14,845 Fold 0: training r2_score = 0.961, validation r2_score = 0.421, test r2_score = 0.475\n",
      "2021-09-18 16:34:14,945 Wrote model tarball to ampl_output/H1_std_model_176a6067-f1a8-47b1-a201-703afedfea5d.tar.gz\n",
      "2021-09-18 16:34:14,960 Attempting to load featurized dataset\n",
      "2021-09-18 16:34:14,969 Exception when trying to load featurized data:\n",
      "DynamicFeaturization doesn't support get_featurized_dset_name()\n",
      "2021-09-18 16:34:14,971 Featurized dataset not previously saved for dataset H1_std, creating new\n",
      "2021-09-18 16:34:14,983 Featurizing sample 0\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "rf_estimators: 32, rf_max_depth: 32, rf_max_features: 64, valid_r2: 0.42073470264383217\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "2021-09-18 16:34:20,055 Featurizing sample 1000\n",
      "2021-09-18 16:34:21,948 Previous dataset split restored\n",
      "2021-09-18 16:34:21,964 Wrote transformers to ampl_output/H1_std/RF_ecfp_scaffold_regression/a38d7b18-4969-4955-8e1f-c8f1fb3054be/transformers.pkl\n",
      "2021-09-18 16:34:21,965 Transforming response data\n",
      "2021-09-18 16:34:21,992 Transforming response data\n",
      "2021-09-18 16:34:22,006 Transforming response data\n",
      "2021-09-18 16:34:22,038 Fitting random forest model\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "number of features: 1024\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "2021-09-18 16:34:22,599 Fold 0: training r2_score = 0.961, validation r2_score = 0.399, test r2_score = 0.530\n",
      "2021-09-18 16:34:22,700 Wrote model tarball to ampl_output/H1_std_model_a38d7b18-4969-4955-8e1f-c8f1fb3054be.tar.gz\n",
      "2021-09-18 16:34:22,713 Attempting to load featurized dataset\n",
      "2021-09-18 16:34:22,721 Exception when trying to load featurized data:\n",
      "DynamicFeaturization doesn't support get_featurized_dset_name()\n",
      "2021-09-18 16:34:22,722 Featurized dataset not previously saved for dataset H1_std, creating new\n",
      "2021-09-18 16:34:22,733 Featurizing sample 0\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "rf_estimators: 32, rf_max_depth: 64, rf_max_features: 16, valid_r2: 0.3988567012433486\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "2021-09-18 16:34:27,711 Featurizing sample 1000\n",
      "2021-09-18 16:34:29,594 Previous dataset split restored\n",
      "2021-09-18 16:34:29,611 Wrote transformers to ampl_output/H1_std/RF_ecfp_scaffold_regression/27361b4b-710e-4202-a702-366c0bfd84db/transformers.pkl\n",
      "2021-09-18 16:34:29,613 Transforming response data\n",
      "2021-09-18 16:34:29,639 Transforming response data\n",
      "2021-09-18 16:34:29,654 Transforming response data\n",
      "2021-09-18 16:34:29,684 Fitting random forest model\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "number of features: 1024\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "2021-09-18 16:34:30,243 Fold 0: training r2_score = 0.961, validation r2_score = 0.436, test r2_score = 0.528\n",
      "2021-09-18 16:34:30,343 Wrote model tarball to ampl_output/H1_std_model_27361b4b-710e-4202-a702-366c0bfd84db.tar.gz\n",
      "2021-09-18 16:34:30,357 Attempting to load featurized dataset\n",
      "2021-09-18 16:34:30,366 Exception when trying to load featurized data:\n",
      "DynamicFeaturization doesn't support get_featurized_dset_name()\n",
      "2021-09-18 16:34:30,368 Featurized dataset not previously saved for dataset H1_std, creating new\n",
      "2021-09-18 16:34:30,381 Featurizing sample 0\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "rf_estimators: 32, rf_max_depth: 64, rf_max_features: 32, valid_r2: 0.43641026586067666\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "2021-09-18 16:34:35,458 Featurizing sample 1000\n",
      "2021-09-18 16:34:37,382 Previous dataset split restored\n",
      "2021-09-18 16:34:37,399 Wrote transformers to ampl_output/H1_std/RF_ecfp_scaffold_regression/6481dfd4-2eeb-4d60-937f-5c1497c24319/transformers.pkl\n",
      "2021-09-18 16:34:37,400 Transforming response data\n",
      "2021-09-18 16:34:37,425 Transforming response data\n",
      "2021-09-18 16:34:37,439 Transforming response data\n",
      "2021-09-18 16:34:37,467 Fitting random forest model\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "number of features: 1024\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "2021-09-18 16:34:38,025 Fold 0: training r2_score = 0.965, validation r2_score = 0.468, test r2_score = 0.532\n",
      "2021-09-18 16:34:38,127 Wrote model tarball to ampl_output/H1_std_model_6481dfd4-2eeb-4d60-937f-5c1497c24319.tar.gz\n",
      "2021-09-18 16:34:38,142 Attempting to load featurized dataset\n",
      "2021-09-18 16:34:38,151 Exception when trying to load featurized data:\n",
      "DynamicFeaturization doesn't support get_featurized_dset_name()\n",
      "2021-09-18 16:34:38,152 Featurized dataset not previously saved for dataset H1_std, creating new\n",
      "2021-09-18 16:34:38,165 Featurizing sample 0\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "rf_estimators: 32, rf_max_depth: 64, rf_max_features: 64, valid_r2: 0.46793790088782805\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "2021-09-18 16:34:43,119 Featurizing sample 1000\n",
      "2021-09-18 16:34:45,004 Previous dataset split restored\n",
      "2021-09-18 16:34:45,025 Wrote transformers to ampl_output/H1_std/RF_ecfp_scaffold_regression/5fbf8680-454e-4b7b-9f36-1f192dbd1c1a/transformers.pkl\n",
      "2021-09-18 16:34:45,026 Transforming response data\n",
      "2021-09-18 16:34:45,054 Transforming response data\n",
      "2021-09-18 16:34:45,069 Transforming response data\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "number of features: 1024\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "2021-09-18 16:34:45,107 Fitting random forest model\n",
      "2021-09-18 16:34:45,677 Fold 0: training r2_score = 0.878, validation r2_score = 0.348, test r2_score = 0.447\n",
      "2021-09-18 16:34:45,805 Wrote model tarball to ampl_output/H1_std_model_5fbf8680-454e-4b7b-9f36-1f192dbd1c1a.tar.gz\n",
      "2021-09-18 16:34:45,820 Attempting to load featurized dataset\n",
      "2021-09-18 16:34:45,830 Exception when trying to load featurized data:\n",
      "DynamicFeaturization doesn't support get_featurized_dset_name()\n",
      "2021-09-18 16:34:45,831 Featurized dataset not previously saved for dataset H1_std, creating new\n",
      "2021-09-18 16:34:45,842 Featurizing sample 0\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "rf_estimators: 64, rf_max_depth: 16, rf_max_features: 16, valid_r2: 0.34791149503882035\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "2021-09-18 16:34:50,896 Featurizing sample 1000\n",
      "2021-09-18 16:34:52,868 Previous dataset split restored\n",
      "2021-09-18 16:34:52,886 Wrote transformers to ampl_output/H1_std/RF_ecfp_scaffold_regression/d848cc25-fb45-45e9-b2de-68ce9a069b86/transformers.pkl\n",
      "2021-09-18 16:34:52,887 Transforming response data\n",
      "2021-09-18 16:34:52,913 Transforming response data\n",
      "2021-09-18 16:34:52,927 Transforming response data\n",
      "2021-09-18 16:34:52,960 Fitting random forest model\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "number of features: 1024\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "2021-09-18 16:34:53,635 Fold 0: training r2_score = 0.908, validation r2_score = 0.402, test r2_score = 0.479\n",
      "2021-09-18 16:34:53,763 Wrote model tarball to ampl_output/H1_std_model_d848cc25-fb45-45e9-b2de-68ce9a069b86.tar.gz\n",
      "2021-09-18 16:34:53,776 Attempting to load featurized dataset\n",
      "2021-09-18 16:34:53,784 Exception when trying to load featurized data:\n",
      "DynamicFeaturization doesn't support get_featurized_dset_name()\n",
      "2021-09-18 16:34:53,786 Featurized dataset not previously saved for dataset H1_std, creating new\n",
      "2021-09-18 16:34:53,797 Featurizing sample 0\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "rf_estimators: 64, rf_max_depth: 16, rf_max_features: 32, valid_r2: 0.4019794255947188\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "2021-09-18 16:34:58,851 Featurizing sample 1000\n",
      "2021-09-18 16:35:00,747 Previous dataset split restored\n",
      "2021-09-18 16:35:00,765 Wrote transformers to ampl_output/H1_std/RF_ecfp_scaffold_regression/69e0e17c-a5a8-4409-8bae-a13620285666/transformers.pkl\n",
      "2021-09-18 16:35:00,767 Transforming response data\n",
      "2021-09-18 16:35:00,794 Transforming response data\n",
      "2021-09-18 16:35:00,810 Transforming response data\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "number of features: 1024\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "2021-09-18 16:35:00,846 Fitting random forest model\n",
      "2021-09-18 16:35:01,516 Fold 0: training r2_score = 0.923, validation r2_score = 0.429, test r2_score = 0.529\n",
      "2021-09-18 16:35:01,648 Wrote model tarball to ampl_output/H1_std_model_69e0e17c-a5a8-4409-8bae-a13620285666.tar.gz\n",
      "2021-09-18 16:35:01,666 Attempting to load featurized dataset\n",
      "2021-09-18 16:35:01,675 Exception when trying to load featurized data:\n",
      "DynamicFeaturization doesn't support get_featurized_dset_name()\n",
      "2021-09-18 16:35:01,676 Featurized dataset not previously saved for dataset H1_std, creating new\n",
      "2021-09-18 16:35:01,688 Featurizing sample 0\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "rf_estimators: 64, rf_max_depth: 16, rf_max_features: 64, valid_r2: 0.429235118555676\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "2021-09-18 16:35:06,615 Featurizing sample 1000\n",
      "2021-09-18 16:35:08,509 Previous dataset split restored\n",
      "2021-09-18 16:35:08,527 Wrote transformers to ampl_output/H1_std/RF_ecfp_scaffold_regression/3e578b96-ce55-4b21-8a84-7e5c69743287/transformers.pkl\n",
      "2021-09-18 16:35:08,528 Transforming response data\n",
      "2021-09-18 16:35:08,557 Transforming response data\n",
      "2021-09-18 16:35:08,572 Transforming response data\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "number of features: 1024\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "2021-09-18 16:35:08,604 Fitting random forest model\n",
      "2021-09-18 16:35:09,281 Fold 0: training r2_score = 0.960, validation r2_score = 0.423, test r2_score = 0.502\n",
      "2021-09-18 16:35:09,466 Wrote model tarball to ampl_output/H1_std_model_3e578b96-ce55-4b21-8a84-7e5c69743287.tar.gz\n",
      "2021-09-18 16:35:09,480 Attempting to load featurized dataset\n",
      "2021-09-18 16:35:09,488 Exception when trying to load featurized data:\n",
      "DynamicFeaturization doesn't support get_featurized_dset_name()\n",
      "2021-09-18 16:35:09,489 Featurized dataset not previously saved for dataset H1_std, creating new\n",
      "2021-09-18 16:35:09,500 Featurizing sample 0\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "rf_estimators: 64, rf_max_depth: 32, rf_max_features: 16, valid_r2: 0.4230715585256083\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "2021-09-18 16:35:14,621 Featurizing sample 1000\n",
      "2021-09-18 16:35:16,573 Previous dataset split restored\n",
      "2021-09-18 16:35:16,592 Wrote transformers to ampl_output/H1_std/RF_ecfp_scaffold_regression/d29f228b-4da7-4f52-9b74-7113e55727c9/transformers.pkl\n",
      "2021-09-18 16:35:16,594 Transforming response data\n",
      "2021-09-18 16:35:16,622 Transforming response data\n",
      "2021-09-18 16:35:16,636 Transforming response data\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "number of features: 1024\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "2021-09-18 16:35:16,670 Fitting random forest model\n",
      "2021-09-18 16:35:17,347 Fold 0: training r2_score = 0.962, validation r2_score = 0.433, test r2_score = 0.510\n",
      "2021-09-18 16:35:17,529 Wrote model tarball to ampl_output/H1_std_model_d29f228b-4da7-4f52-9b74-7113e55727c9.tar.gz\n",
      "2021-09-18 16:35:17,544 Attempting to load featurized dataset\n",
      "2021-09-18 16:35:17,552 Exception when trying to load featurized data:\n",
      "DynamicFeaturization doesn't support get_featurized_dset_name()\n",
      "2021-09-18 16:35:17,554 Featurized dataset not previously saved for dataset H1_std, creating new\n",
      "2021-09-18 16:35:17,565 Featurizing sample 0\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "rf_estimators: 64, rf_max_depth: 32, rf_max_features: 32, valid_r2: 0.4325768493441816\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "2021-09-18 16:35:22,601 Featurizing sample 1000\n",
      "2021-09-18 16:35:24,489 Previous dataset split restored\n",
      "2021-09-18 16:35:24,504 Wrote transformers to ampl_output/H1_std/RF_ecfp_scaffold_regression/2ae19a92-9511-4ed6-b39e-2e44e5940dce/transformers.pkl\n",
      "2021-09-18 16:35:24,506 Transforming response data\n",
      "2021-09-18 16:35:24,532 Transforming response data\n",
      "2021-09-18 16:35:24,546 Transforming response data\n",
      "2021-09-18 16:35:24,574 Fitting random forest model\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "number of features: 1024\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "2021-09-18 16:35:25,245 Fold 0: training r2_score = 0.963, validation r2_score = 0.437, test r2_score = 0.533\n",
      "2021-09-18 16:35:25,429 Wrote model tarball to ampl_output/H1_std_model_2ae19a92-9511-4ed6-b39e-2e44e5940dce.tar.gz\n",
      "2021-09-18 16:35:25,445 Attempting to load featurized dataset\n",
      "2021-09-18 16:35:25,454 Exception when trying to load featurized data:\n",
      "DynamicFeaturization doesn't support get_featurized_dset_name()\n",
      "2021-09-18 16:35:25,456 Featurized dataset not previously saved for dataset H1_std, creating new\n",
      "2021-09-18 16:35:25,468 Featurizing sample 0\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "rf_estimators: 64, rf_max_depth: 32, rf_max_features: 64, valid_r2: 0.4371304243058133\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "2021-09-18 16:35:30,509 Featurizing sample 1000\n",
      "2021-09-18 16:35:32,461 Previous dataset split restored\n",
      "2021-09-18 16:35:32,477 Wrote transformers to ampl_output/H1_std/RF_ecfp_scaffold_regression/cfb5f335-b7cd-45a1-bc22-8061b0d64438/transformers.pkl\n",
      "2021-09-18 16:35:32,479 Transforming response data\n",
      "2021-09-18 16:35:32,507 Transforming response data\n",
      "2021-09-18 16:35:32,521 Transforming response data\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "number of features: 1024\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "2021-09-18 16:35:32,555 Fitting random forest model\n",
      "2021-09-18 16:35:33,230 Fold 0: training r2_score = 0.963, validation r2_score = 0.422, test r2_score = 0.526\n",
      "2021-09-18 16:35:33,425 Wrote model tarball to ampl_output/H1_std_model_cfb5f335-b7cd-45a1-bc22-8061b0d64438.tar.gz\n",
      "2021-09-18 16:35:33,441 Attempting to load featurized dataset\n",
      "2021-09-18 16:35:33,451 Exception when trying to load featurized data:\n",
      "DynamicFeaturization doesn't support get_featurized_dset_name()\n",
      "2021-09-18 16:35:33,453 Featurized dataset not previously saved for dataset H1_std, creating new\n",
      "2021-09-18 16:35:33,464 Featurizing sample 0\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "rf_estimators: 64, rf_max_depth: 64, rf_max_features: 16, valid_r2: 0.42164000670865054\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "2021-09-18 16:35:38,555 Featurizing sample 1000\n",
      "2021-09-18 16:35:40,446 Previous dataset split restored\n",
      "2021-09-18 16:35:40,465 Wrote transformers to ampl_output/H1_std/RF_ecfp_scaffold_regression/3d65db66-31d6-4542-81ce-c0d01540f055/transformers.pkl\n",
      "2021-09-18 16:35:40,466 Transforming response data\n",
      "2021-09-18 16:35:40,492 Transforming response data\n",
      "2021-09-18 16:35:40,506 Transforming response data\n",
      "2021-09-18 16:35:40,536 Fitting random forest model\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "number of features: 1024\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "2021-09-18 16:35:41,208 Fold 0: training r2_score = 0.964, validation r2_score = 0.445, test r2_score = 0.523\n",
      "2021-09-18 16:35:41,400 Wrote model tarball to ampl_output/H1_std_model_3d65db66-31d6-4542-81ce-c0d01540f055.tar.gz\n",
      "2021-09-18 16:35:41,417 Attempting to load featurized dataset\n",
      "2021-09-18 16:35:41,427 Exception when trying to load featurized data:\n",
      "DynamicFeaturization doesn't support get_featurized_dset_name()\n",
      "2021-09-18 16:35:41,428 Featurized dataset not previously saved for dataset H1_std, creating new\n",
      "2021-09-18 16:35:41,439 Featurizing sample 0\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "rf_estimators: 64, rf_max_depth: 64, rf_max_features: 32, valid_r2: 0.4449688806778509\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "2021-09-18 16:35:46,439 Featurizing sample 1000\n",
      "2021-09-18 16:35:48,366 Previous dataset split restored\n",
      "2021-09-18 16:35:48,386 Wrote transformers to ampl_output/H1_std/RF_ecfp_scaffold_regression/46f25ee9-77f3-4e35-b22f-a924ce13a39a/transformers.pkl\n",
      "2021-09-18 16:35:48,388 Transforming response data\n",
      "2021-09-18 16:35:48,416 Transforming response data\n",
      "2021-09-18 16:35:48,430 Transforming response data\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "number of features: 1024\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "2021-09-18 16:35:48,463 Fitting random forest model\n",
      "2021-09-18 16:35:49,238 Fold 0: training r2_score = 0.964, validation r2_score = 0.436, test r2_score = 0.529\n",
      "2021-09-18 16:35:49,430 Wrote model tarball to ampl_output/H1_std_model_46f25ee9-77f3-4e35-b22f-a924ce13a39a.tar.gz\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "rf_estimators: 64, rf_max_depth: 64, rf_max_features: 64, valid_r2: 0.4359956045728688\n"
     ]
    }
   ],
   "source": [
    "# param_choice = [16,32,64,128]\n",
    "param_choice = [16, 32, 64]\n",
    "\n",
    "rfe_list = []\n",
    "rfd_list = []\n",
    "rff_list = []\n",
    "valid_r2_list = []\n",
    "for rf_estimator in param_choice:\n",
    "    for rf_depth in param_choice:\n",
    "        for rf_feature in param_choice:\n",
    "            params[\"rf_estimators\"] = rf_estimator\n",
    "            params[\"rf_max_depth\"] = rf_depth\n",
    "            params[\"rf_max_features\"] = rf_feature\n",
    "            rfe_list.append(rf_estimator)\n",
    "            rfd_list.append(rf_depth)\n",
    "            rff_list.append(rf_feature)\n",
    "            tp = parse.wrapper(params)\n",
    "            pl = mp.ModelPipeline(tp)\n",
    "            pl.train_model()\n",
    "            pred_data = pl.model_wrapper.get_perf_data(subset=\"valid\", epoch_label=\"best\")\n",
    "            pred_results = pred_data.get_prediction_results()\n",
    "            valid_r2 = pred_results['r2_score']\n",
    "            valid_r2_list.append(valid_r2)\n",
    "            print(f\"rf_estimators: {rf_estimator}, rf_max_depth: {rf_depth}, rf_max_features: {rf_feature}, valid_r2: {valid_r2}\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 896
    },
    "id": "XSDa2HTH6wbD",
    "outputId": "d3210789-2bf5-4618-e4d9-8caf7aebeb73"
   },
   "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>rf_estimators</th>\n",
       "      <th>rf_max_depth</th>\n",
       "      <th>rf_max_features</th>\n",
       "      <th>valid_r2_score</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>17</th>\n",
       "      <td>32</td>\n",
       "      <td>64</td>\n",
       "      <td>64</td>\n",
       "      <td>0.467938</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>16</td>\n",
       "      <td>16</td>\n",
       "      <td>64</td>\n",
       "      <td>0.460007</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>11</th>\n",
       "      <td>32</td>\n",
       "      <td>16</td>\n",
       "      <td>64</td>\n",
       "      <td>0.448069</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>25</th>\n",
       "      <td>64</td>\n",
       "      <td>64</td>\n",
       "      <td>32</td>\n",
       "      <td>0.444969</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>16</td>\n",
       "      <td>32</td>\n",
       "      <td>32</td>\n",
       "      <td>0.439180</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>23</th>\n",
       "      <td>64</td>\n",
       "      <td>32</td>\n",
       "      <td>64</td>\n",
       "      <td>0.437130</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>16</th>\n",
       "      <td>32</td>\n",
       "      <td>64</td>\n",
       "      <td>32</td>\n",
       "      <td>0.436410</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>26</th>\n",
       "      <td>64</td>\n",
       "      <td>64</td>\n",
       "      <td>64</td>\n",
       "      <td>0.435996</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>13</th>\n",
       "      <td>32</td>\n",
       "      <td>32</td>\n",
       "      <td>32</td>\n",
       "      <td>0.433247</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>22</th>\n",
       "      <td>64</td>\n",
       "      <td>32</td>\n",
       "      <td>32</td>\n",
       "      <td>0.432577</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>20</th>\n",
       "      <td>64</td>\n",
       "      <td>16</td>\n",
       "      <td>64</td>\n",
       "      <td>0.429235</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>12</th>\n",
       "      <td>32</td>\n",
       "      <td>32</td>\n",
       "      <td>16</td>\n",
       "      <td>0.426250</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>21</th>\n",
       "      <td>64</td>\n",
       "      <td>32</td>\n",
       "      <td>16</td>\n",
       "      <td>0.423072</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>7</th>\n",
       "      <td>16</td>\n",
       "      <td>64</td>\n",
       "      <td>32</td>\n",
       "      <td>0.422244</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>5</th>\n",
       "      <td>16</td>\n",
       "      <td>32</td>\n",
       "      <td>64</td>\n",
       "      <td>0.421757</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>24</th>\n",
       "      <td>64</td>\n",
       "      <td>64</td>\n",
       "      <td>16</td>\n",
       "      <td>0.421640</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>14</th>\n",
       "      <td>32</td>\n",
       "      <td>32</td>\n",
       "      <td>64</td>\n",
       "      <td>0.420735</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>19</th>\n",
       "      <td>64</td>\n",
       "      <td>16</td>\n",
       "      <td>32</td>\n",
       "      <td>0.401979</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>15</th>\n",
       "      <td>32</td>\n",
       "      <td>64</td>\n",
       "      <td>16</td>\n",
       "      <td>0.398857</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>9</th>\n",
       "      <td>32</td>\n",
       "      <td>16</td>\n",
       "      <td>16</td>\n",
       "      <td>0.396678</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>16</td>\n",
       "      <td>16</td>\n",
       "      <td>16</td>\n",
       "      <td>0.396017</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>10</th>\n",
       "      <td>32</td>\n",
       "      <td>16</td>\n",
       "      <td>32</td>\n",
       "      <td>0.392781</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>16</td>\n",
       "      <td>32</td>\n",
       "      <td>16</td>\n",
       "      <td>0.390933</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>8</th>\n",
       "      <td>16</td>\n",
       "      <td>64</td>\n",
       "      <td>64</td>\n",
       "      <td>0.374936</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>6</th>\n",
       "      <td>16</td>\n",
       "      <td>64</td>\n",
       "      <td>16</td>\n",
       "      <td>0.351995</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>16</td>\n",
       "      <td>16</td>\n",
       "      <td>32</td>\n",
       "      <td>0.350530</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>18</th>\n",
       "      <td>64</td>\n",
       "      <td>16</td>\n",
       "      <td>16</td>\n",
       "      <td>0.347911</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "    rf_estimators  rf_max_depth  rf_max_features  valid_r2_score\n",
       "17             32            64               64        0.467938\n",
       "2              16            16               64        0.460007\n",
       "11             32            16               64        0.448069\n",
       "25             64            64               32        0.444969\n",
       "4              16            32               32        0.439180\n",
       "23             64            32               64        0.437130\n",
       "16             32            64               32        0.436410\n",
       "26             64            64               64        0.435996\n",
       "13             32            32               32        0.433247\n",
       "22             64            32               32        0.432577\n",
       "20             64            16               64        0.429235\n",
       "12             32            32               16        0.426250\n",
       "21             64            32               16        0.423072\n",
       "7              16            64               32        0.422244\n",
       "5              16            32               64        0.421757\n",
       "24             64            64               16        0.421640\n",
       "14             32            32               64        0.420735\n",
       "19             64            16               32        0.401979\n",
       "15             32            64               16        0.398857\n",
       "9              32            16               16        0.396678\n",
       "0              16            16               16        0.396017\n",
       "10             32            16               32        0.392781\n",
       "3              16            32               16        0.390933\n",
       "8              16            64               64        0.374936\n",
       "6              16            64               16        0.351995\n",
       "1              16            16               32        0.350530\n",
       "18             64            16               16        0.347911"
      ]
     },
     "execution_count": 18,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "perf_df = pd.DataFrame({\"rf_estimators\": rfe_list, \"rf_max_depth\": rfd_list, \"rf_max_features\": rff_list, \"valid_r2_score\": valid_r2_list})\n",
    "perf_df.sort_values(by=\"valid_r2_score\", ascending=False)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 482
    },
    "id": "9bVXltSu6z0f",
    "outputId": "1d3ba93e-4069-489f-d9a5-31805dbca0a6"
   },
   "outputs": [
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAABcIAAAHSCAYAAADPF7hcAAAABHNCSVQICAgIfAhkiAAAAAlwSFlzAAALEgAACxIB0t1+/AAAADh0RVh0U29mdHdhcmUAbWF0cGxvdGxpYiB2ZXJzaW9uMy4yLjIsIGh0dHA6Ly9tYXRwbG90bGliLm9yZy+WH4yJAAAgAElEQVR4nOzde3wU9b3/8ffmQkJIAJE0RIxHKJwocosBykUR1NbUSBuJSgkIWAxWxdYCtR7wUMijSClQEbSnCBYhISn4oCg0FQt6FDXRBCGKiY20ciQEwh0xQNIkO78/+CWQksvuZpbdmX09Hw8expnvfOezLNn37mdmZxyGYRgCAAAAAAAAAMCmgnxdAAAAAAAAAAAA3kQjHAAAAAAAAABgazTCAQAAAAAAAAC2RiMcAAAAAAAAAGBrNMIBAAAAAAAAALYW4usC0FhJSYmqq6sVHByssLAwX5cDADBJdXW16urqFBYWpj59+vi6HJiI7AYAeyK77Y38BgD7aS27aYT7merqajmdTjmdTtXU1Pi6HACAyaqrq31dAkxGdgOAvZHd9kR+A4B9NZfdNML9THBwsJxOp4KCghQREeHrcgAAJjl37pycTqeCg4N9XQpMRnYDgD2R3fZGfgOA/bSW3TTC/UxYWJhqamoUERGh+Ph4X5cDADBJaWmpKisr+eqtDZHdAGBPZLe9kd8AYD+tZTc3ywQAAAAAAAAA2BqNcAAAAAAAAACArdEIBwAAAAAAAADYGo1wAAAAAAAAAICt0QgHAAAAAAAAANgajXAAAAAAAAAAgK3RCAcAAAAAAAAA2BqNcAAAAAAAAACArdEIBwAAAAAAAADYGo1wAAAAAAAAAICt0QgHAAAAAAAAANgajXAAAAAAAAAAgK3RCAcAAAAAAAAA2FqIrwsA4J9yc3O1adMmVVVV+bqUNgsPD1dqaqqSk5N9XQoAAIBpeL8GAID1kN++wxnhAJqUm5trixdlSaqqqlJubq6vywAAADAV79cAALAe8tt3aIQDaFJycrLCw8N9XYYpwsPDLXN0EgAAwFW8XwMAwHrIb9/h0igAmpScnOz1F7O0tLSGn7Ozs726LwAAALvh/RoAANZDfvsOZ4QDAAAAAAAAAGyNRjgAAAAAAAAAwNZohAMAAAAAAAAAbI1GOAAAAAAAAADA1miEAwAAAAAAAABsjUY4AAAAAAAAAMDWaIQDAAAAAAAAAGyNRjgAAAAAAAAAwNZohAMAAAAAAAAAbI1GOAAAAAAAAADA1miEAwAAAAAAAABsjUY4AAAAAAAAAMDWaIQDAAAAAAAAAGzNEo3wrVu3Ki0tTYmJiUpISNDYsWO1fv16OZ3ONs+9YcMGxcfHKz4+XhkZGS2OraurU05OjiZMmKDvfOc76tevn2677Tb95Cc/0dtvv93mWgAAsAuyGwAA6yG/AQB2FuLrAlozf/58ZWdnKywsTMOGDVNISIjy8/OVkZGh/Px8LV++XEFBnvXzy8vLtWjRIjkcDhmG0eLYU6dOKT09XXv37lXnzp01cOBAtW/fXhUVFcrLy9PVV1+t22+/3aM6AACwE7IbAADrIb8BAHbn143wN998U9nZ2YqOjlZWVpauv/56SdLx48c1adIkbd++XZmZmZo8ebLbcxuGoTlz5sgwDKWkpGjz5s3NjnU6nXr00Ue1d+9eTZo0SbNmzVJYWFjD+srKSpWXl7tdAwAAdkN2AwBgPeQ3ACAQ+PWlUVauXClJmjVrVkMQS1LXrl01b948SdKqVas8+ppWTk6O8vPzNWPGDHXv3r3FsRs3btSePXs0evRozZkzp1EQS1JkZKTi4+PdrgEAALshuwEAsB7yGwAQCPy2EV5RUaHi4mKFhoYqKSnpsvVDhgxRTEyMjh07pqKiIrfmLisr0+LFi5WYmKiJEye2On79+vWSpClTpri1HwAAAgnZDQCA9ZDfAIBA4beXRikpKZEk9e7dW+Hh4U2O6devn44cOaLPP/9cN998s0vzGoah2bNnq66uTgsWLJDD4Whx/NGjR/XFF18oODhYCQkJ2r9/v/7617/qyJEj6tSpkwYPHqxbb7211XkAALA7shsAAOshvwEAgcJvG+EHDx6UJF1zzTXNjomNjW001hVZWVkqKCjQzJkz1aNHj1bHf/HFF5Kkzp07KycnR4sXL1ZtbW3D+pdeekkJCQl68cUXdfXVV7tcBwAAdkN2AwBgPeQ3ACBQ+G0j/Ny5c5Kk9u3bNzumQ4cOkqSzZ8+6NOeBAwe0dOlS9e3bV1OnTnVpm6+//rrhvwsXLtQ999yjxx57TDExMfrss8+UkZGhPXv26Gc/+5mysrJcmtMVlZWV+vjjj02br6127dqlvLw81dTU+LqUNgsNDdXw4cM1aNAgX5eCS/jTv3cAniG7yW5vIbsB/+BPr3EwD/ntX/kNAGbjNe4iv71GuNnqv5ZVW1urBQsWKDg42KXt6m8GUltbq8TERC1dulTf/va3FRkZqaFDh+qPf/yjwsPDVVhYqA8//NCbD8Gndu3aZYsP0pJUU1OjXbt2+boMAEAryO62IbsBAL5AfgMA/JXfnhEeEREhSTp//nyzY+qPRtcfnW7JunXrVFhYqMcff1w33HCDy3VcOvcDDzxw2fpu3brptttu05tvvqmPPvpIQ4cOdXnulvjb3bBTUlK0adMmVVVV+bqUNgsPD1dKSooSExN9XQouwfMBuystLVVlZaWvy/Aqspvs9hayG/APgfY7GAjZLZHf/pbfAGC2QMrv1rLbbxvh3bt3lyQdOnSo2TEVFRWNxrZkx44dkqS8vDwVFhY2WldeXi5J2r59u/bt26eIiAitXLlSknTttdc2jLv050vVLz9+/HirdVhVcnKykpOTvbqPtLS0hp+zs7O9ui8AgPnIbv9CdgMAXEF+AwAChd82wvv06SNJ2rdvn6qqqpq8e/XevXslSTfeeKPL8+7Zs6fZdUePHtXRo0cVFRXVsKxHjx6KiIjQuXPndPr06Sa3O3XqlKSLR9IBAAhEZDcAANZDfgMAAoXfXiM8NjZWN910k2pqarRt27bL1hcUFKiiokLR0dFKSEhodb7MzEyVlpY2+Wf69OmSpAkTJqi0tLTRNShDQ0M1atQoSVJ+fv5l8156zcq+fft68lABALAFshsAAOshvwEAgcJvG+GSNG3aNEnSkiVL9NVXXzUsP3HihObPny9JSk9PV1DQxYeRlZWlpKQkPfXUU6bV8cgjjygoKEgbNmzQe++917C8rq5OS5Ys0YEDBxQTE6Pvfve7pu0TAAArIrsBALAe8hsAEAj89tIokpSUlKTx48crJydHY8aM0fDhwxUSEqL8/HxVVlbqzjvv1MSJExttc+rUKe3fv1/R0dGm1XHDDTdo9uzZWrBggdLT09W/f39169ZNJSUlKisrU1RUlJ5//vkmv0IGAEAgIbsBALAe8hsAEAj8uhEuSfPmzVNiYqLWr1+vgoICOZ1O9ezZU6mpqRo/fnyjI9Le9OCDD+o///M/9cc//lFFRUUqKSlRdHS0xo0bp2nTpjV7Mw8AAAIN2Q0AgPWQ3wAAu3MYhmH4ughcVFpaqsrKSkVGRio+Pt7X5VxRaWlpDT9nZ2f7sBJcKTznCCSB/Ppud4H83PI6DthfIP+eB/LreyDg+QVgZ4Ga3629tvv1NcIBAAAAAAAAAGgrGuEAAAAAAAAAAFujEQ4AAAAAAAAAsDUa4QAAAAAAAAAAW6MRDgAAAAAAAACwNRrhAAAAAAAAAABboxEOAAAAAAAAALA1GuEAAAAAAAAAAFujEQ4AAAAAAAAAsDUa4QAAAAAAAAAAW6MRDgAAAAAAAACwNRrhAAAAAAAAAABboxEOAAAAAAAAALA1GuEAAAAAAAAAAFujEQ4AAAAAAAAAsDUa4QAAAAAAAAAAW6MRDgAAAAAAAACwNRrhAAAAAAAAAABboxEOAAAAAAAAALA1GuEAAAAAAAAAAFujEQ4AAAAAAAAAsDUa4QAAAAAAAAAAW6MRDgAAAAAAAACwNRrhAAAAAAAAAABboxEOAAAAAAAAALC1EF8XAAAAAAAAAAC+lpaW5usSTGeXx5Sdnd3mOTgjHAAAAAAAAABgazTCAQAAAAAAAAC2xqVRAAAAAAAAAOASt9z9mK9LCHjv//X3ps7HGeEAAAAAAAAAAFujEQ4AAAAAAAAAsDUa4QAAAAAAAAAAW6MRDgAAAAAAAACwNRrhAAAAAAAAAABbC/F1AQAA/5Cbm6tNmzapqqrK16W0WXh4uFJTU5WcnOzrUgAAAACgSXb6DCbxOQz+jzPCAQCSLrwJs8sbsKqqKuXm5vq6DAAAAABolp0+g0l8DoP/oxEOAJAkJScnKzw83NdlmCI8PJyzEAAAAAD4NTt9BpP4HAb/x6VRAACSLrwJ8/ablrS0tIafs7OzvbovAAAAAPBnV+IzmMTnMKAeZ4QDAAAAAAAAAGyNRjgAAAAAAAAAwNZohAMAAAAAAAAAbI1GOAAAAAAAAADA1izRCN+6davS0tKUmJiohIQEjR07VuvXr5fT6Wzz3Bs2bFB8fLzi4+OVkZHh9e0AAAgEZDcAANZDfgMA7CzE1wW0Zv78+crOzlZYWJiGDRumkJAQ5efnKyMjQ/n5+Vq+fLmCgjzr55eXl2vRokVyOBwyDMPr2wEAEAjIbgAArIf8BgDYnV+fEf7mm28qOztb0dHR2rJli1auXKkXX3xRf/vb3/Ttb39b27dvV2ZmpkdzG4ahOXPmyDAMpaSkeH07AAACAdkNAID1kN8AgEDg143wlStXSpJmzZql66+/vmF5165dNW/ePEnSqlWrPPqaVk5OjvLz8zVjxgx1797d69sBABAIyG4AAKyH/AYABAK/bYRXVFSouLhYoaGhSkpKumz9kCFDFBMTo2PHjqmoqMitucvKyrR48WIlJiZq4sSJXt8OAIBAQHYDAGA95DcAIFD4bSO8pKREktS7d2+Fh4c3OaZfv36SpM8//9zleQ3D0OzZs1VXV6cFCxbI4XB4dTsAAAIF2Q0AgPWQ3wCAQOG3N8s8ePCgJOmaa65pdkxsbGyjsa7IyspSQUGBZs6cqR49enh9O09VVlbq448/9vp+/FUgP/ZAxXMeeHjO7YfsJrsB2Bu/5/ZEfgd2fiPw8O8dVmXGv12/PSP83LlzkqT27ds3O6ZDhw6SpLNnz7o054EDB7R06VL17dtXU6dOdbkWT7cDACCQkN0AAFgP+Q0ACBR+e0a42eq/XlVbW6sFCxYoODjYq9u1VWRkpOLj46/IvvxRYmKir0vAFcZzHngC7TkvLS1VZWWlr8uwFLLbWgLtdxoIRIH2e052e4b8BvxboL2Wwz5c+bfbWnb7bSM8IiJCknT+/Plmx9Qfja4/Ot2SdevWqbCwUI8//rhuuOEGl+vwdDsAAAIN2Q34Vm5urjZt2qSqqipfl9Jm4eHhSk1NVXJysq9LAWyP/AYABAq/bYR3795dknTo0KFmx1RUVDQa25IdO3ZIkvLy8lRYWNhoXXl5uSRp+/bt2rdvnyIiIrRy5co2bQcAQKAhuwHfys3NtUUTXJKqqqqUm5tLIxy4AshvAECg8NtGeJ8+fSRJ+/btU1VVVZN3r967d68k6cYbb3R53j179jS77ujRozp69KiioqJM2w4AgEBBdgO+lZycbKszwmmCA1cG+Q0ACBR+2wiPjY3VTTfdpOLiYm3btk0pKSmN1hcUFKiiokLR0dFKSEhodb7MzMxm161YsUIvvPCCJkyYoLlz55qyHQAAgYbsBnwrOTnZ683jtLS0hp+zs7O9ui8AVwb5DQAIFEG+LqAl06ZNkyQtWbJEX331VcPyEydOaP78+ZKk9PR0BQVdfBhZWVlKSkrSU089dWWLBQAAZDcAABZEfgMAAoHfnhEuSUlJSRo/frxycnI0ZswYDR8+XCEhIcrPz1dlZaXuvPNOTZw4sdE2p06d0v79+xUdHe2jqgEACFxkNwAA1kN+AwACgV83wiVp3rx5SkxM1Pr161VQUCCn06mePXsqNTVV48ePb3REGgAA+B7ZDQCA9ZDfAAC7cxiGYfi6CFxUWlqqyspKRUZGKj4+3tflXFFcczLw8JwHnkB+zgP59d3uAvm5DeTf6UDFcx54Avk5D+TX90DA84tAEsiv5e669O/qlrsf82ElkKT3//r7hp9d+bfb2mu7358RDuByl74w24VdHhNvKgAAgGSf9zaXsstj4v0aAFfY5TXv39nlcfFaDk/QCAcAC7DLm5VL2eUx8QYMAAAAAOzn0rORYQ9c5AsAAAAAAAAAYGsenxFeVlamtWvX6oMPPlBFRYWqq6tVUlLSsP7MmTNat26dHA6Hpk2bptDQUFMKBtDYsluv9XUJAe/J9w76ugTAJWQ3APgG1xj1PSuf1Ud+A77H527/wGdvtJVHjfDt27frqaeeUlVVlervtelwOBqN6dixoz766CPt2rVLvXr10l133dX2agEAfJj2A1b8ME12AwBgPeQ3APgOn719z+zP3m5fGuWf//ynZs2apfPnz+uBBx5QVlaWrrrqqibH3n///TIMQ++8805b6wQAAB4iuwEAsB7yGwAAc7l9RvjLL7+s6upqTZkyRU8//bQkKTg4uMmxw4cPlyR9+umnbSgRAAC0BdkNAID1kN8AAJjL7TPCP/zwQzkcDj388MOtju3atavat2+vw4cPe1QcAABoO7IbAADrIb8BADCX243wY8eOqUOHDuratatL49u1a6eamhq3CwMAAOYguwEAsB7yGwAAc7l9aZSIiAhVVlaqrq6u2a9l1ausrNQ333yjLl26eFwgAABoG7Ib8L20tDRfl2A6uzym7OxsX5cANIn8BgDAXG6fEd6rVy85nU4VFxe3OvaNN96Q0+lU3759PSoOAAC0HdkNAID1kN8AAJjL7TPCv//97+vjjz/W888/r1WrVikoqOleemlpqZYsWSKHw6ExY8a0uVAAAOAZshsAAOshvwH/8eR7B31dAgATuN0IHzdunF599VXl5eVp8uTJmjRpkmprayVdCOBDhw5p586d2rx5s6qqqpSYmKi7777b9MIBAIBryG7Avyy79VpflxDwaGjACshvAADM5XYjPDQ0VKtXr9ajjz6qwsJC7dq1q2FdSkpKw8+GYWjgwIFasWKFHA6HOdUCAAC3kd0AAFgP+Q0AgLncboRLUnR0tP70pz9p8+bN2rx5s/bu3dtwd+rg4GD17dtX9957r+677z6FhHi0CwAAYCKyGwAA6yG/Af/At7n8A9/oQlt5nJQhISG6//77df/996uurk5ff/21nE6nOnfuTAADAOCHyG4AAKyH/AYAwBxN322jBYMHD9Z3vvMdlZWVNSwLDg5Wly5d1LVrV4IYAAA/Q3YDAGA95DcAAOZyuxFeU1Ojuro6xcXFeaMeAABgMrIbAADrIb8BADCX243w2NjYhmuSAQAA/0d2AwBgPeQ3AADmcrsRfvvtt+tf//qXPvjgA2/UAwAATEZ2AwBgPeQ3AADmcrsR/pOf/ETdu3fXf//3f+uf//ynN2oCAAAmIrsBALAe8hsAAHO5fXeNt956S+PHj9eLL76olJQUjRw5UgMHDlSXLl0UHBzc7HYpKSltKhQAcMH7f/29r0uAxZDd/ictLc3XJZjOLo8pOzvb1yUAgCTyGwAAs7ndCH/66aflcDhkGIYk6e2339bbb7/d6naEMQAAvkF2AwBgPeQ3AADmcrsRPnjwYG/UAQAAvITsBgDAeshvAADM5XYjPDMz0xt1AABcdMvdj/m6hIBntcvTkN3+LfbBJF+XEPAOZ27zdQkAcBnyGwAAc7l9s0wAAAAAAAAAAKyERjgAAAAAAAAAwNbcvjTKpcrKyrRt2zaVlJTo5MmTkqQuXbqoT58+SkpKUlxcnClFAgAAc5DdAABYD/kNAEDbedQIr6qq0oIFC7Rp0yYZhtFwF+t627Zt03PPPaf77rtPs2fPVnh4uCnFAgAAz5DdAABYD/kNAIB53G6EO51OPfbYY8rPz5dhGIqJidGQIUPUrVs3SVJFRYUKCgp05MgRvfrqqyovL9fq1avlcDhMLx4AALSO7AYAwHrIbwAAzOV2I3zTpk3Ky8tTWFiY5syZo/vvv/+yoDUMQxs3btSCBQuUl5enTZs26b777jOtaAAXPfneQV+XAMDPkd2AfyG7AbiC/A48ubm52rRpk6qqqnxdSpuFh4crNTVVycnJvi4FABq4fbPM119/XQ6HQ88884weeOCBJo82OxwOjRs3Ts8884wMw9Brr71mSrEAAMB9ZDcAANZDfgee3NxcWzTBpQuX9cnNzfV1GQDQiNuN8C+++EIhISFKSUlpdWxKSopCQkL0xRdfeFQcAABoO7IbAADrIb8DT3Jysm2u8x4eHs7Z4AD8jtuXRqmqqlL79u0VGhra6th27dopIiLCNkc0AX+07NZrfV1CwOMr7vB3ZDfgX8hu3yO7YQXkd+BJTk72evM4LS2t4efs7Gyv7gsA/I3bZ4R/61vf0jfffKOvvvqq1bH79+/XmTNn9K1vfcuj4gAAQNuR3QAAWA/5DQCAudxuhA8fPlyGYWju3Lmqrq5udlx1dbV+9atfyeFwaPjw4W0qEgAAeI7sBgDAeshvAADM5XYjPD09XWFhYSooKNAPfvAD5eTk6J///KcqKyt18uRJffbZZ3r55Zf1ve99T4WFhQoLC1N6ero3agcAAC4guwEAsB7yGwAAc7l9jfC4uDgtW7ZMM2bM0FdffaWMjIwmxxmGofbt2+t3v/ud4uLi2lwoAADwDNkNAID1kN8AAJjL7TPCJWn06NHasmWLxo4dq8jISBmG0ehPVFSUUlNTtWXLFo0ePdrsmgEAgJvIbgAArIf8BgDAPG6fEV4vLi5Ozz77rJ599lmVlZXp5MmTkqQuXbpwFBoAAD9EdgMAYD3kNwAA5vC4EX6puLg4AhgAAAshuwEAsB7yGwAAz5nSCIfvpKWl+boEr7DL48rOzvZ1CQAAAPCx9//6e1+XAAAAEPDcvkb4Rx99pDvuuENz5sxpdeysWbN0xx13aNeuXR4VV2/r1q1KS0tTYmKiEhISNHbsWK1fv15Op7NN80rShg0bFB8fr/j4+CZvPuJ0OrV7924999xz+tGPfqTBgwfrpptu0vDhw5Wenq4dO3a0uQYAALyJ7Ca7AQDWQ36T3wAAc7l9RviWLVt06NAh3X777a2OHT16tP7yl79oy5YtGjRokEcFzp8/X9nZ2QoLC9OwYcMUEhKi/Px8ZWRkKD8/X8uXL1dQkEf3/FR5ebkWLVokh8MhwzCaHFNWVqbx48dLkjp37qz+/furY8eOKisr086dO7Vz506NHTtWzz77rBwOh0d1AIA7OKsM7iK7yW74lyffO+jrEgBYAPntX/ltl28tX8ouj4lvYsNb+OxtP243wouKiuRwODRs2LBWx44aNUoOh0O7d+/2qLg333xT2dnZio6OVlZWlq6//npJ0vHjxzVp0iRt375dmZmZmjx5sttzG4ahOXPmyDAMpaSkaPPmzU2OczgcGjp0qKZOnaoRI0YoODi4YV1BQYEeeeQR/fnPf9agQYOUmprq0eM0S+yDST7dPy44nLnN1yUAQCNkt/9mN4DAcMvdj/m6hIBnxWYG+U1+AwDM5fbh3MOHDysqKkoRERGtju3QoYM6duyoI0eOeFTcypUrJV34mld9EEtS165dNW/ePEnSqlWrPPqaVk5OjvLz8zVjxgx179692XHXXXed1q5dq5EjRzYKYkkaMmSI0tPTJV04Wg8AgD8iuy8iuwEAVkF+X0R+AwDM4NHNMqurq90a68nXlioqKlRcXKzQ0FAlJV1+pvOQIUMUExOjI0eOqKioSDfffLPLc5eVlWnx4sVKTEzUxIkT9cILL7hdX70+ffo01AsA3mKXr/td+vVLuzwmqyC7LyK74QtX4jUvNzdXmzZtUlVVldf35W3h4eFKTU1VcnKyr0sBfIr8vsif8ptvY/se38SGt/CezT1We8/m9hnh3bt3V3V1tYqLi1sd+9lnn6mqqkrdunVzu7CSkhJJUu/evRUeHt7kmH79+kmSPv/8c5fnNQxDs2fPVl1dnRYsWNDma4v93//9nyTpW9/6VpvmAQDAW8juxshu2FVubq4tPlBJUlVVlXJzc31dBuBT5Hdj5DcAu+A9m++4fUb4iBEjtG/fPi1ZskSrV6++7CtL9erq6rRkyRI5HA6NGDHC7cIOHrxwE6Frrrmm2TGxsbGNxroiKytLBQUFmjlzpnr06OF2XZc6f/68MjMzJUnf+9732jTXv6usrNTHH39s6py48ngOAw/Puev4u7pyyO6LyG60xsrP4YABA5SXl6eamhpfl9JmoaGhGjBggKWfD/gvq/y7Ir8vIr/REp6/wGTl5533bL7jdiN8ypQp+tOf/qQPP/xQDz30kH7xi180HB2u9+mnn2rx4sUqLCxUWFiYHnroIbcLO3funCSpffv2zY7p0KGDJOns2bMuzXngwAEtXbpUffv21dSpU92u6d/Nnz9fBw8eVK9evTRu3Lg2zwcAgDeQ3ReR3bCzQYMGadCgQb4uA4BJyO+LyG8AdsJ7Nt9xuxHerVs3LVq0SDNnzlRhYaEeeOABderUqeHo8aFDh/T111/LMAwFBwdr4cKFLd4Q40qp/1pWbW2tFixY0OzRdFe9+OKL2rx5s6KiorRs2TK1a9fOpEoviIyMVHx8vKlz4spLTEz0dQm4wnjOXRdof1elpaWqrKz0yb7J7gvIbrgi0F6bgEDk6u+5L7NbIr/rkd9oDdkdmHje0ZTWstujm2Xeddddio6O1sKFC7V3716dPn1ap0+fbjSmf//+evrpp926kcal6u+Mff78+WbH1B+Nrj863ZJ169apsLBQjz/+uG644QaPaqq3Zs0aLV++XBEREVq1apV69+7dpvkAAPA2spvsBgBYD/lNfgMAzONRI1ySbr75Zr366qv68ssv9cknn+j48eOSpK5du2rAgAHq2bNnmwqrP5J96NChZsfU3y3alaPeO3bskCTl5eWpsLCw0bry8nJJ0vbt27Vv3z5FRERo5cqVTc6TmZmp3/zmNwoPD9fKlSuVkJDQ+oMBAFiKiOUAACAASURBVMAPkN1kNwDAeshv8hsAYA6PG+H1evbs2ebgbUqfPn0kSfv27VNVVVWTd6/eu3evJOnGG290ed49e/Y0u+7o0aM6evSooqKimly/fv16/frXv1ZYWJj+53/+R0OGDHF5vwAA+AuyGwAA6yG/AQBomyBfF9Cc2NhY3XTTTaqpqdG2bdsuW19QUKCKigpFR0e7dGQ4MzNTpaWlTf6ZPn26JGnChAkqLS3Vrl27Lts+JydHGRkZateunV588UUNHz687Q8SAAAbIbsBALAe8hsAECjcboSfPn1a77//vj755JPL1h05ckRPPvmkRowYocGDB2vGjBk6cuSIx8VNmzZNkrRkyRJ99dVXDctPnDih+fPnS5LS09MVFHTxYWRlZSkpKUlPPfWUx/v9dxs3btT8+fPVrl07vfDCC7r11ltNmxsAAG8ju8luAID1kN/kNwDAXG5fGmXjxo167rnnNHnyZA0YMKBheXV1tSZOnKiDBw/KMAxJ0htvvKHi4mJt3ry54QYc7khKStL48eOVk5OjMWPGaPjw4QoJCVF+fr4qKyt15513auLEiY22OXXqlPbv36/o6Gi399eUzz//XHPnzpVhGLr22mv1xhtv6I033rhs3FVXXaVf/vKXpuwTAAAzkd1kNwDAeshv8hsAYC63G+HvvfeeJGnMmDGNlv/5z39WWVmZOnfurJ///OcKCwvTsmXLdODAAWVlZTUcYXbXvHnzlJiYqPXr16ugoEBOp1M9e/ZUamqqxo8f3+iItDecOXOm4c3Fl19+qS+//LLJcd27dyeMAQB+iewmuwEA1kN+k98AAHO53Qivv8tzr169Gi3ftm2bHA6HZs6cqfvvv1+SFBMTo4ceekg7duzwOIylC8H/7+HfnCeeeEJPPPGEW/O3tM13vvMdlZaWujUfAAD+hOwGAMB6yG8AAMzl9iHdkydPqmPHjgoLC2tYVltbq6KiIgUFBSkpKalh+dChQxUcHKz9+/ebUy0AAHAb2Q0AgPWQ3wAAmMvtRrhhGDp37lyjZcXFxaqurlZ8fLyioqIaljscDkVGRqqqqqrtlQIAAI+Q3QAAWA/5DQCAudxuhHfr1k21tbX6+9//3rBsx44dkqRBgwY1Gut0OnX27Fl16dKljWUCAABPkd0AAFgP+Q0AgLncboQPHTpUhmFo3rx5+vTTT/XWW28pOztbDodDo0ePbjT2H//4h2pra9WtWzfTCgYAAO4huwEAsB7yGwAAc7l9s8z09HT95S9/0SeffKJx48ZJuvCVrZtvvlnDhg1rNPbtt9+Ww+FQQkKCOdUCAAC3kd0AAFgP+Q0AgLncPiP82muv1bp16zR48GCFhYXp6quv1tixY/X73/++0bi6ujpt3LhRhmFcFtIAAODKIbsBALAe8hsAAHO5fUa4JN10001at25di2OCgoL0+uuvS5IiIyM92Q0AADAJ2Q0AgPWQ3wAAmMftM8Jd5XA4FBUVpaioKDkcjsvW33fffbrzzju9tXsAAOAmshsAAOshvwEAcI1HZ4SboaKiQidOnPDV7gEAgJvIbgAArIf8BgDgAq+dEQ4AAAAAAAAAgD+gEQ4AAAAAAAAAsDUa4QAAAAAAAAAAW6MRDgAAAAAAAACwNRrhAAAAAAAAAABboxEOAAAAAAAAALA1GuEAAAAAAAAAAFujEQ4AAAAAAAAAsDUa4QAAAAAAAAAAW6MRDgAAAAAAAACwtRBf7TgpKUlnz5711e4BAICbyG4AAKyH/AYA4AKfNcKfeeYZX+0aAAB4gOwGAMB6yG8AAC5w+9IoZ86c0alTp1weX1RUpMLCQnd3AwAATEJ2AwBgPeQ3AADmcvmM8JycHP3xj3/UwYMHJUlXX321UlNTlZ6ersjIyGa3mz59uk6ePKmSkpK2VwsAAFxGdgMAYD3kNwAA3uHSGeELFy5URkaGysrKZBiGDMPQ8ePH9dJLL+mHP/yhioqKWtzeMAxTigUAAK4huwEAsB7yGwAA72m1EV5QUKC1a9fKMAwNHTpUv/nNb/SHP/xBDz30kNq3b6/y8nJNmjRJ27dvvxL1AgCAVpDdAABYD/kNAIB3tdoIz8nJkSR9//vf1yuvvKKUlBSNGjVKv/zlL/XGG29o8ODB+te//qWf//zneu2117xeMAAAaBnZDQCA9ZDfAAB4V6uN8KKiIjkcDs2YMeOydTExMXrllVc0btw41dbWavbs2Vq/fr1XCgUAAK4huwEAsB7yGwAA72q1EX78+HG1b99ecXFxTa4PDg7W/PnzNW3aNDmdTv3617/Wyy+/bHqhAADANWQ3AADWQ34DAOBdrTbCQ0JCVFdX1+pEM2bM0M9+9jMZhqElS5bohRdeMKVAAADgHrIbAADrIb8BAPCukNYGxMXFad++fdq/f7969OjR4thHH31U7du3129+8xu9+OKLqqqqMq1QAADgGrIbAADrIb8BAPCuVs8I79+/vyTpnXfecWnCKVOmaO7cuZKkl19+WcePH/e8OgAA4DayGwAA6yG/AQDwrlYb4SNHjpRhGNqwYYMMw3Bp0rS0ND377LMKCmp1egAAYDKyGwAA6yG/AQDwrlYvjTJq1CgNGzZMtbW1+uSTTzRw4ECXJr733nsVERGh3/72t20uEgAAuI7sBgDAeshvAAC8q9VGeLt27bRmzRqPJr/rrrt01113ebQtAADwDNkNAID1kN8AAHhXq43wf/fWW29JkhISEtSlSxfTCwIAAOYiuwEAsB7yGwAAc7ndCH/88ccVEhKijz76yBv1AAAAk5HdAABYD/kNAIC53G6Ed+rUSZLUoUMH04sBAADmI7sBALAe8hsAAHO5fWvp3r17q7KyUpWVld6oBwAAmIzsBgDAeshvAADM5XYj/IEHHlBdXZ0yMzO9UQ8AADAZ2Q0AgPWQ3wAAmMvtS6P84Ac/0KeffqoVK1aourpaU6ZMUefOnb1RGwAAMAHZDQCA9ZDfAACYy+1G+KRJkyRJ4eHhWrlypVavXq3rrrtOXbp0UVBQ0yeYOxwOrV27tm2VAgAAj5DdAABYD/kNAIC53G6EFxQUNPr/2tpaffnll/ryyy+b3cbhcLhf2SW2bt2qnJwclZaWyul0qkePHkpNTdX48eObfQPgqg0bNmju3LmSpAkTJjT83JSdO3fqlVde0Weffabq6mrFxcUpOTlZU6dOVbt27dpUBwAA3kJ2k90AAOu50vlNdgMA7M7tRvjChQu9UUez5s+fr+zsbIWFhWnYsGEKCQlRfn6+MjIylJ+fr+XLl3scyuXl5Vq0aJEcDocMw2hx7KpVq7RkyRIFBwdryJAh6tixowoLC7Vs2TK98847euWVV9S+fXuP6gAAf5Cbm6tNmzapqqrqiuwvLS3Na3OHh4crNTVVycnJXtuHlZDdZDcAwHquZH6T3UDLnnzvoK9LAGACtxvh9957rzfqaNKbb76p7OxsRUdHKysrS9dff70k6fjx45o0aZK2b9+uzMxMTZ482e25DcPQnDlzZBiGUlJStHnz5mbH7t27V0uXLlX79u21du1aDRgwQJJ09uxZPfLIIyosLNRzzz2n2bNne/Q4AcAf5ObmXrEmuLdVVVUpNzeXRvj/R3aT3QAA67lS+U12AwACRdu+39QGTzzxRKtBunLlSknSrFmzGsJYkrp27ap58+ZJunDE2Ol0ur3/nJwc5efna8aMGerevXuLY1etWiXDMPTwww83hLEkdejQQQsXLlRQUJCys7N15swZt+sAAH+RnJys8PBwX5dhivDwcJrgXkB2AwBgPa3lN9kNAAgUbp8RbpY9e/boxIkTza6vqKhQcXGxQkNDlZSUdNn6IUOGKCYmRkeOHFFRUZFuvvlml/ddVlamxYsXKzExURMnTtQLL7zQ7Nh//etf2rlzp6QLd+3+d3FxcRo4cKB2796td999V2PGjHG5DgDwJ8nJyTSP0SKyGwAA62kpv8luoHnZ2dm+LsE0l16W0k6PC3CXz84Ib01JSYkkqXfv3s2eodivXz9J0ueff+7yvIZhaPbs2aqrq9OCBQtavZnI/v37df78eXXu3FnXXXddi3XU1wwAQCAiuwEAsBayGwAQSPy2EX7w4IUbEVxzzTXNjomNjW001hVZWVkqKCjQ9OnT1aNHD5frqN9XU+prLC8vd7kOAADshuwGAMBayG4AQCDx2aVRWnPu3DlJavGO0B06dJB04eYZrjhw4ICWLl2qvn37aurUqabVERER4VYdrqisrNTHH39s2nzwDZ7DwMNzjkBGdpPddsBzCNgfv+cXBXp2S+S3HfD8uYe/LwQyvz0j3Gz1X82qra3VggULFBwc7OuSAABAC8huAACshewGAPgzvz0jvP5o7/nz55sdU38kuP4IdUvWrVunwsJCPf7447rhhhtMraP+6LUrdbgqMjJS8fHxps0H30hMTPR1CbjCeM7RnNLSUlVWVvq6DK8iu8luO+B1HLA/V3/Pye4L7JzdEvltB2S3e/j7gp21lt1+2wjv3r27JOnQoUPNjqmoqGg0tiU7duyQJOXl5amwsLDRuvprjG3fvl379u1TRESEVq5c2Wjuw4cPNzt3/TpX6gDM9uR7rl+rDwC8iewGgKa9/9ff+7oEoElkNwAgkPhtI7xPnz6SpH379qmqqqrJO1jv3btXknTjjTe6PO+ePXuaXXf06FEdPXpUUVFRDct69uyp8PBwnT59WgcOHGjyDtaffvqp23UAAGA3ZDcAANZCdgMAAonfXiM8NjZWN910k2pqarRt27bL1hcUFKiiokLR0dFKSEhodb7MzEyVlpY2+Wf69OmSpAkTJqi0tFS7du1q2K5du3YaOXKkJGnLli2XzVtWVqaioiKFhoZq1KhRHj5aAACsj+wGAMBayG4AQCDx2zPCJWnatGn62c9+piVLlighIUH/8R//IUk6ceKE5s+fL0lKT09XUNDFfn5WVpaysrLUv39//fa3vzWljvT0dG3fvl2rV6/WyJEj1b9/f0kXrpU2e/ZsOZ1OPfjgg+rYsaMp+wNak52d7esSTJGWltbws10eExDoyG4AuMAu7214v2Z/ZDcAIFD4dSM8KSlJ48ePV05OjsaMGaPhw4crJCRE+fn5qqys1J133qmJEyc22ubUqVPav3+/oqOjTaujf//+mjlzppYsWaIf/ehHGjp0qKKiolRYWKgTJ05owIAB+vnPf27a/gAAsCqyGwAAayG7AQCBwmeN8IEDB+rrr79uddy8efOUmJio9evXq6CgQE6nUz179lRqaqrGjx/f6Ki0N6Wnpys+Pl5r1qzR3r17VV1drbi4OD344IOaOnWq2rVrd0XqAADAV8huAACsx5X8JrsBAIHAYRiG4esicFFpaakqKysVGRmp+Pj4Vsdf+lXF2AeTvFkaXHQ48+K19fj6aMv4qi0Cibuv77AOT57bS1//4F/II6CxQH6/RnbbW1s+e8O/BNprkycC+bUcgaW11/YWzwh/7bXXTCskJSXFtLkAAEDTyG4AAKyH/AYAwPtabIQ//fTTcjgcpuyIMPa+S89EBgAEJrIbAADrIb8BAPC+FhvhgwcPbnbd3//+d33zzTeSpNjYWMXExEiSjh49qkOHDkmSOnbsyFfIAAC4gshu6+HSZr7HyQQAfI38thay2/fIbgCeaLERnpmZ2eTyRYsWqbCwUPfdd58eeeQRxcXFNVp/8OBBvfTSS9q4caP69u2rX/7yl+ZVDAAAmkV2AwBgPeQ3AADe12IjvCmvv/66XnnlFaWnp2vmzJlNjrn22muVkZGhzp07a9WqVerTp4/GjBnT5mLRMo5K+weOTAPwN2Q3AADWQ34DAGCuIHc3yM7OVlBQkB555JFWx06bNk1BQUFav369R8UBAIC2I7sBALAe8hsAAHO53Qj/xz/+ocjISEVGRrY6tn7cP/7xD4+KAwAAbUd2AwBgPeQ3AADmcrsR7nA49M033+jEiROtjj1x4oTOnDnjUWEAAMAcZDcAANZDfgMAYC63G+F9+vSRYRhavHhxq2MXL14swzDUt29fj4oDAABtR3YDAGA95DcAAOZyuxH+8MMPyzAMvf7663rooYeUl5enqqqqhvXV1dXKy8vTj3/8Y73++utyOBx6+OGHTS0aAAC4juwGAMB6yG8AAMwV4u4GI0eO1KxZs7R06VJ9+OGH+vDDDxUUFNRw3bLKyko5nU4ZhiGHw6FZs2bplltuMb1wAADgGrIbAADrIb8BADCX241w6cKR6YEDB2rFihUqKChQXV2dvv7664b1DodDw4YN0/Tp05WYmGhasQCunNzcXG3atKnRWSfelJaW5rW5w8PDlZqaquTkZK/tA/B3ZLf/Opy5zdclAAD8FPkNAIB5PGqES9KgQYO0du1aff311yopKdGpU6ckSVdddZX69OmjTp06mVYkgCsvNzf3ijXBva2qqkq5ubk0whHwyG4AAKyH/AYAwBweN8LrderUScOGDTOjFgB+JDk5+YqeEe5N4eHhNMGBS5DdADx16tQprVixQj/96U/VuXNnX5cDBBTyGwCAtmlzIxyAPSUnJ9M8BgAvyc7O9nUJprj0slZ2eUxo2ebNm1VaWqo///nP+vGPf+zrcgAAAACXBfm6AAAAAAD+79SpU3r33XdlGIZ27typ06dP+7okAAAAwGUtnhF+4403SpJ69uyp3NzcRsvc4XA4VFJS4kF5AADAHWQ3AG/ZvHmzDMOQJDmdTs4KB0xEflsLN7oGAGtq8YxwwzAa/jS1zNU/TqfT6w8EAACQ3QC854MPPlBtba0kqba2Vh988IGPKwLsg/wGAMD7Wjwj/K233rowKCTksmUAAMD/kN0AvGXEiBF65513VFtbq5CQEI0YMcLXJQG2QX4DAOB9LTbCu3fv7tIyAADgH8huAN5y77336t1335UkBQUFaezYsT6uCLAP8tv/2eWm0NzoGkAg42aZAAAAAFp11VVX6bbbbpPD4dDIkSPVuXNnX5cEAAAAuKzFM8IBAAAAoN69996rgwcPcjY4AAAALKfFRvhrr71m2o5SUlJMmwsAADSN7AbgTVdddZXmzp3r6zIA2yG/AQDwvhYb4U8//bQcDocpOyKMAQDwPrIbAADrIb8BAPC+FhvhgwcPvlJ1AAAAE5DdAABYD/kNAID3tdgIz8zMvFJ1AAAAE5DdAABYD/kNAID3Bfm6AAAAAAAAAAAAvIlGOAAAAAAAAADA1miEAwAAAAAAAABsrcVrhLekoqJCmzZt0u7du3X06FGdP39ehmE0OdbhcGjHjh0eFwkAANqO7AYAwHrIbwAAzOFRI3zLli2aO3euqqurWwzg+nUOh8PzCgEAQJuR3QAAWA/5DQCAedxuhBcXF2v27Nmqra1VamqqRo8erenTp6tTp05atmyZTpw4oby8PP3lL39Rhw4dNHv2bMXExHijdgAA4AKyGwAA6yG/AQAwl9uN8DVr1qi2tlZTpkzR008/3bA8NDRUw4YNkyTdc889mjx5sqZOnaply5Zp8+bN5lUMAADcQnYDAGA95DcAAOZy+2aZu3fvlsPh0JQpU1ocFx8fr7lz56q8vFwvvfSSp/UBAIA2IrsBALAe8hsAAHO53Qg/fvy4wsLC1K1bt4ZlwcHBqqqqumzs7bffrtDQUG7WAQCAD5HdAABYD/kNAIC53G6ER0REKCwsrNGyyMhInT17VufPn2+0PCQkRO3atVNFRUXbqgQAAB4juwEAsB7yGwAAc7ndCI+JidE333yj6urqhmU9evSQJO3Zs6fR2AMHDujs2bMKCXH7UuQAAMAkZDcAANZDfgMAYC63G+E33HCDDMNQcXFxw7Jbb71VhmHod7/7nY4dOyZJOnnypJ555hk5HA4NGDDAvIoBAIBbyG4AAKyH/AYAwFxuN8JHjRolwzC0bdu2hmUTJ07U1VdfreLiYo0ePVq33nqrbrnlFhUUFCgoKEiPPvqoqUUDAADXkd0AAFgP+Q0AgLncboTfcccd+sMf/qA777yzYVmnTp20du1a9e3bV7W1tTp27JicTqe6deum559/XoMGDTK1aAAA4DqyGwAA6yG/AQAwl9sXEGvXrp1GjRp12fJevXrp1Vdf1eHDh1VRUaGoqCh9+9vflsPhMKNOAADgIbIbAADrIb8BADCX22eEv/3226qtrW12fWxsrBISEtSrVy+CGAAAP0B2AwBgPeQ3AADmcvuM8Mcee0ydO3fW9773Pd1zzz0aMmSIN+pqZOvWrcrJyVFpaamcTqd69Oih1NRUjR8/XkFBrvfyt2zZovfee09///vfdezYMX3zzTeKiIhQr169lJycrHHjxik0NLTJbSsqKrRq1Sq9//77Onz4sAzDUGxsrIYOHar09HTFxcWZ9XABADAV2U12AwCsh/wmvwEA5nK7ER4ZGanTp0/r1Vdf1auvvqqYmBjdfffduueee9SnTx/TC5w/f76ys7MVFhamYcOGKSQkRPn5+crIyFB+fr6WL1/uciDn5OSoqKhIvXr1Ur9+/RQVFaWjR4+qqKhIu3fv1tatW7VmzRpFREQ02q6kpESTJ0/WmTNn1K1bN91yyy2SpM8++0wbNmzQ1q1b9fLLL+vmm282/fEDANBWZDfZDQCwHvKb/AYAmMvtRnheXp7effddbd26VTt37lRFRYXWrFmjNWvWqEePHrrnnnt0zz336LrrrmtzcW+++aays7MVHR2trKwsXX/99ZKk48ePa9KkSdq+fbsyMzM1efJkl+b7r//6L11//fXq2LFjo+UVFRV66KGHVFRUpNWrV+unP/1po/UZGRk6c+aMHnjgAc2dO7fhyHVNTY1+9atfadOmTZo3b562bNnS5scMAIDZyG6yGwBgPeQ3+Q0AMJfb1whv166dvvvd72r58uX64IMPtGjRIo0YMULBwcH68ssvtWLFCt111126//77tW7dOh0/ftzj4lauXClJmjVrVkMQS1LXrl01b948SdKqVavkdDpdmq9///6XBbEkdevWTY888oikC282LlVdXa09e/ZIkp544olGX98KDQ3Vk08+KUkqLS3V+fPnXXtgAABcQWQ32Q0AsB7ym/wGAJjL7Ub4pTp06KAf/vCHWr16td577z3NnTu34StKe/fu1cKFC3Xbbbfpxz/+sdtzV1RUqLi4WKGhoUpKSrps/ZAhQxQTE6Njx46pqKioLQ9DkhQScuHk+H+/TllQUFDDupZEREQoPDy8zXUAAOBNZPdFZDcAwCrI74vIbwCAp9rUCL/UVVddpbS0NK1fv17/+7//q1/84hfq2bOn6urqlJ+f7/Z8JSUlkqTevXs3G3L9+vWTJH3++eeeFy7p5MmTevnllyVJd9xxR6N1oaGhGjp0qCRpxYoVqqmpaVhXU1Oj559/XpKUmprKnboBAJZCdpPdAADrIb/JbwCAZ9y+Rnhrampq9Nlnn2nv3r0qLy/3eJ6DBw9Kkq655ppmx8TGxjYa66q3335bf/vb31RXV6djx45p9+7dqq6u1tixYzVhwoTLxs+bN08PP/ywNm7cqJ07d6pv376SLhx5/3/t3Xd8VGX69/FvSAIhJJBQjDRFFybSiQQUUdwFVKStgMIPBWlLVJooTcRC04XVB1dFEASRblkJLERAsKKCIUiRGEMREEIRSCGF9Pv5g9fMElNISJnkzOf9V5hT5jozzHxnrnPmvi9duqQhQ4Zo0qRJRarhWpKSkrRnz54S3SfKHs8hgIqA7C4Zrp7drnzsgKvgdV6+kN8lw5Xz21WP29XxvMOVlUgj3BijXbt2aePGjdq+fbsSExNljJF0JUy7d+9e5H2mpKRIkqpWrZrvOtWqVZMkJScnF2nfv/76q0JDQ3PcNmTIkFzjkNk1bNhQa9eu1ZQpUxyTlNi1aNFCwcHBeW4HAEB5RXaT3QCAiof8Jr8BANevWI3w/fv3a9OmTdq8ebMuXrwo6Uow+/n5qVu3burZs6eCg4NLpNCSNGrUKI0aNUrp6ek6ffq0Nm/erMWLF2v79u1avHixGjdunGP9n376SWPHjpWPj48WLFigoKAgx+1z587V2LFjNXbsWI0ZM6bEavTx8VFgYGCJ7Q/O0bZtW2eXAKCciI6OVlJSkrPLILvJ7lJD5gHW52qv8/KS3RL5TX6XDld7TeMKnndY2bWyu8iN8CNHjmjjxo367LPPHD+LMsaoatWq6tKli3r16qWOHTsWapKLgnh7e0tSgbNB289G289OF1XlypXVqFEjPfXUU7r11ls1btw4TZ48WZ9++qljzLFLly5p9OjRunz5sj788EM1bNjQsX3Xrl3VpEkT9e7dWwsXLlTPnj1zzLANAEB5QHaT3QCAiof8Jr8BACWryInZq1cvSVcC2MPDQ3fffbd69eqlLl26lOjMzfXr15cknT59Ot917D+Tsq9bHPfff798fHwUGRmpU6dOOUL366+/VmxsrO68884cQWx38803q1WrVgoPD1d4eDhhDAAod8junMhuAEBFQH7nRH4DAIqryI1wY4zatWunnj17qlu3bqpRo0Zp1KVmzZpJkg4fPqzU1NQ8g/7nn3+WJDVt2rTY9+fm5iY/Pz8lJSUpNjbWEbxnzpyRJPn6+ua7bfXq1SVJ8fHxxa4DAICSRnbnRnYDAMo78js38hsAUBxFboR/8803CggIKI1acqhbt66aN2+uyMhIbdmyRQ899FCO5eHh4Tp79qzq1KnjGDesOE6ePKmYmBhVqlQpx9nnG264QZIUGRmpjIyMXBNzZGRkKDIyUpLUoEGDYtcBAEBJI7vJbgBAxUN+k98AgJJVqagblEUQ24WEhEiSXn/9dZ04ccJx+8WLFzVjxgxJ0siRI1Wp0v8OY9WqVerWrZsmT56cY1/28dXS0tJy3c+hQ4f09NNPyxij++67TzVr1nQs69Spk6pWrarTp0/rn//8p9LT0x3L0tPTNXv2bJ05c0Y1atTQPffcUzIHDgBALmf2GgAAIABJREFUCSK7yW4AQMVDfpPfAICSVbxZNUpZt27dNHDgQK1du1a9evXSXXfdJQ8PD+3cuVNJSUnq2rWrBg0alGObuLg4HTt2THXq1Mlx+8WLFzVx4kR5e3urWbNmCggIUHp6umJiYhQVFSVjjFq1auUIebtatWrp5Zdf1rRp07R69Wpt27ZNzZs3lyQdPHhQ58+fV+XKlfXqq68W+BMuAABcAdkNAEDFQ34DAFxBuW6ES9L06dPVtm1brV69WuHh4crOztatt96qfv36aeDAgTnOSBekSZMmGj9+vCIiInTs2DFFRkYqMzNT/v7+6tSpkx588EH17t1b7u7uubbt06ePbDabli9froiICH3//feSrpyhf/jhhzVs2DA1bty4RI8bAICKiuwGAKDiIb8BAFbnZowxzi4C/xMdHa2kpCT5+PgoMDDwmus/+uijjr/rDu5WmqWhkM6s3OL4e82aNU6sBEB5UtT3d1QcrvzcXv05hMwDrMmVX+eu/P7uClz1+XXl17Qr43mHq7jWe3uRxwgHAAAAAAAAAKAioREOAAAAAAAAABYRFxenmTNnKj4+3tmllCs0wgEAAAAAAADAIkJDQxUdHa1169Y5u5RyhUY4AAAAAAAAAFhAXFycvvnmGxlj9O2333JV+FU8nF0ASs7VkzQCAAAAAAAAcC2hoaEyxkiSsrOztW7dOg0fPtzJVZUPXBEOAAAAAAAAABbw/fffKzMzU5KUmZmp77//3skVlR80wgEAAAAAAADAAjp27CgPjyuDgHh4eKhjx45Orqj8YGiUCm7NmjXOLqHEPProo46/rXRcAAAAAAAAQFno06ePvvnmG0lSpUqV1LdvXydXVH5wRTgAAAAAAAAAWIC/v7/uvfdeubm5qVOnTvLz83N2SeUGV4QDAAAAAAAAgEX06dNHp06d4mrwP6ERDgAAAAAAAAAW4e/vr5deesnZZZQ7DI0CAAAAAAAAALA0GuEAAAAAAAAAAEujEQ4AAAAAAAAAsDQa4QAAAAAAAAAAS6MRDgAAAAAAAACwNBrhAAAAAAAAAABLoxEOAAAAAAAAALA0GuEAAAAAAAAAAEujEQ4AAAAAAAAAsDQa4QAAAAAAAAAAS6MRDgAAAAAAAACwNBrhAAAAAAAAAABLoxEOAAAAAAAAALA0GuEAAAAAAAAAAEujEQ4AAAAAAAAAsDQa4QAAAAAAAAAAS6MRDgAAAAAAAACwNBrhAAAAAAAAAABLoxEOAAAAAAAAALA0GuEAAAAAAAAAAEujEQ4AAAAAAAAAsDQa4QAAAAAAAAAAS6MRDgAAAAAAAACwNBrhAAAAAAAAAABLoxEOAAAAAAAAALA0GuEAAAAAAAAAAEujEQ4AAAAAAAAAsDQa4QAAAAAAAAAAS6MRDgAAAAAAAACwNBrhAAAAAAAAAABLoxEOAAAAAAAAALA0D2cXUBgbN27U2rVrFR0drezsbN1yyy3q16+fBg4cqEqVCt/L/+9//6sdO3bo119/1fnz55WYmChvb281btxYPXr00IABA+Tp6Znv9qmpqVq5cqW2bNmiEydOKCMjQ7Vq1VKLFi00ZMgQtW3btiQOFwCACo/sBgCg4iG/AQBWVu4b4TNmzNCaNWtUpUoVdejQQR4eHtq5c6dmzpypnTt36q233ip0IK9du1b79u1T48aN1bJlS/n6+uqPP/7Qvn379NNPP2njxo1atmyZvL29c2178uRJjRgxQidOnFCdOnV0xx13yN3dXadPn9YXX3yh2267jTAGAEBkNwAAFRH5DQCwunLdCN+6davWrFmjOnXqaNWqVWrUqJEk6cKFC3r88ce1bds2rVy5UkOGDCnU/qZOnapGjRqpevXqOW4/e/ashg0bpn379mnJkiUaN25cjuUpKSkaPny4Tp48qQkTJmjEiBFyd3d3LI+Li1N8fHzxDhYAAAsguwEAqHjIbwCAKyjXY4QvWrRIkjRx4kRHEEtS7dq1NX36dEnSe++9p+zs7ELtr1WrVrmCWJJuvPFGPfHEE5KkH374IdfyhQsX6vfff9djjz2mkJCQHEEsSf7+/rrlllsKVQMAAFZGdgMAUPGQ3wAAV1BuG+Fnz55VZGSkPD091a1bt1zL27dvr4CAAJ0/f1779u0r9v15eFy5OP7P45Slp6fr448/liQNHTq02PcDAIBVkd0AAFQ85DcAwFWU26FRfvnlF0lSkyZN5OXllec6LVu21Llz5xQVFaXbb7/9uu8rNjZWS5culSR16dIlx7LIyEjFx8crICBADRs2VGRkpLZt26bY2FjVqlVLHTt2VHBw8HXfNwAAVkF2AwBQ8ZDfAABXUW4b4adOnZIk1atXL9916tatm2Pdwvryyy/1+eefKysrS+fPn9dPP/2ktLQ09e3bV4899liOdQ8dOiRJCggI0Ny5c/X+++/nWL5gwQJ17dpVr732Wp4TfVyvpKQk7dmzp8T2V9G48rEDQEVFdpPdAKyN17k1kd+um9+uetyujucdrqzcNsJTUlIkSVWrVs13nWrVqkmSkpOTi7TvX3/9VaGhoTluGzJkiMaOHZvr51kJCQmSpKioKB04cEBDhgzRoEGD5Ofnp927d2vGjBnavn27ZsyYoblz5xapDgAArITsBgCg4iG/AQCuotw2wkvTqFGjNGrUKKWnp+v06dPavHmzFi9erO3bt2vx4sVq3LixY137ZCAZGRnq3bu3nn/+eceyLl266IYbbtAjjzyiDRs2aPTo0brppptKpEYfHx8FBgaWyL4qorZt2zq7BAAoUdHR0UpKSnJ2GRUW2V3+kd2A9bna65zsLj7yu3xztdc0ruB5h5VdK7vL7WSZ9p86Xb58Od917Gej7Weni6py5cpq1KiRnnrqKc2ZM0cxMTGaPHmyjDGOda7ed//+/XPto2XLlmrevLmMMQoPD7+uOgAAsAKyGwCAiof8BgC4inLbCK9fv74k6fTp0/muc/bs2RzrFsf9998vHx8fRUZG5hj3rEGDBnn+fTX77RcuXCh2HQAAVFRkNwAAFQ/5DQBwFeW2Ed6sWTNJ0uHDh5WamprnOj///LMkqWnTpsW+Pzc3N/n5+Um6MpP1n+uQpPj4+Dy3jYuLk6QSnbADAICKhuwGAKDiIb8BAK6i3DbC69atq+bNmysjI0NbtmzJtTw8PFxnz55VnTp1FBQUVOz7O3nypGJiYlSpUiU1bNjQcXtAQIBat24tSdq5c2eu7RISEvTLL79Iklq0aFHsOgAAqKjIbgAAKh7yGwDgKsptI1ySQkJCJEmvv/66Tpw44bj94sWLmjFjhiRp5MiRqlTpf4exatUqdevWTZMnT86xryNHjmjjxo1KS0vLdT+HDh3S008/LWOM7rvvPtWsWTPH8ieffFKStGjRIseZcElKS0vT9OnTlZiYqObNm5fIhwIAACoyshsAgIqH/AYAuAIPZxdQkG7dumngwIFau3atevXqpbvuukseHh7auXOnkpKS1LVrVw0aNCjHNnFxcTp27Jjq1KmT4/aLFy9q4sSJ8vb2VrNmzRQQEKD09HTFxMQoKipKxhi1atXKEfJX69y5s4YPH673339fAwcOVOvWreXn56cDBw7ojz/+UEBAgObNmyc3N7dSfTwAACjvyG4AACoe8hsA4ArKdSNckqZPn662bdtq9erVCg8PV3Z2tm699Vb169dPAwcOzHFGuiBNmjTR+PHjFRERoWPHjikyMlKZmZny9/dXp06d9OCDD6p3795yd3fPc/spU6YoKChIq1atUlRUlC5fvqx69epp2LBhCgkJyXUmGwAAV0V2AwBQ8ZDfAACrczPGGGcXgf+Jjo5WUlKSfHx8FBgY6OxyytSjjz7q+HvNmjVOrAQASp4rv79bnSs/t2Q3YH2u/Dp35fd3V+Cqz68rv6ZdGc87XMW13tvL9RjhAAAAAAAAAAAUF41wAAAAAAAAAICl0QgHAAAAAAAAAFgajXAAAAAAAAAAgKXRCAcAAAAAAAAAWBqNcAAAAAAAAACApdEIBwAAAAAAAABYGo1wAAAAAAAAAICl0QgHAAAAAAAAAFgajXAAAAAAAAAAgKXRCAcAAAAAAAAAWBqNcAAAAAAAAACApdEIBwAAAAAAAABYmoezCwAAAAAAFF1YWJg+/fRTpaamlsn9Pfroo6W2by8vL/Xr1089evQotfsAAACujSvCAQAAAKACCgsLK7MmeGlLTU1VWFiYs8sAAAAWRiMcAAAAACqgHj16yMvLy9lllAgvLy+uBgcAAKWKoVEAAAAAoALq0aMHzWMAAIBCohGOQmH8QQAAAAAAgJJT1r0WiX4LXBtDo6BQGH8QAAAAAACg5Fip1yLRb0H5RyMchcL4gwAAAAAAACXHSr0WiX4Lyj+GRkGhMP4gAAAAAABAyaHXApQtrggHAAAAAAAAAFgajXAAQJmJi4vTzJkzFR8f7+xSAAAAAACAC6ERDgAoM6GhoYqOjta6deucXQoAAAAAAHAhNMIBAGUiLi5O33zzjYwx+vbbb7kqHAAAAAAAlBkmywQAlInQ0FAZYyRJ2dnZWrdunYYPH+7kqgDrCgsL06effqrU1NQyub9HH3201Pbt5eWlfv36MZkUAMDSyG4AKF1cEQ4AKBPff/+9MjMzJUmZmZn6/vvvnVwRYG1hYWFl9kW6tKWmpiosLMzZZQAAUKrIbgAoXTTCAQBlomPHjvLwuPJDJA8PD3Xs2NHJFQHW1qNHD3l5eTm7jBLh5eXFFWUAAMsjuwGgdDE0CgCgTPTp00fffPONJKlSpUrq27evkysCrK1Hjx58AQUAoAIhuwGgdHFFOACgTPj7++vee++Vm5ubOnXqJD8/P2eXBAAAAAAAXARXhAMAykyfPn106tQprgYHAAAAAABlikY4AKDM+Pv766WXXnJ2GQAAAAAAwMUwNAoAAAAAAAAAwNJohAMAAAAAAAAALI1GOAAAAAAAAADA0miEAwAAAAAAAAAsjUY4AAAAAAAAAMDSaIQDAAAAAAAAACyNRjgAAAAAAAAAwNJohAMAAAAAAAAALI1GOAAAAAAAAADA0miEAwAAAAAAAAAszcPZBSCntLQ0SVJKSoqio6OdXA0AoKSkpKRI+t/7PKyD7AYAayK7rY38BgDruVZ20wgvZ7KysiRJ2dnZSkpKcnI1AICSZn+fh3WQ3QBgbWS3NZHfAGBd+WU3jfBypkqVKkpLS5O7u7uqVKni7HIAACUkLS1NWVlZvLdbENkNANZEdlsb+Q0A1nOt7HYzxpgyrgkAAAAAAAAAgDLDZJkAAAAAAAAAAEujEQ4AAAAAAAAAsDQa4QAAAAAAAAAAS6MRDgAAAAAAAACwNBrhAAAAAAAAAABLoxEOAAAAAAAAALA0GuEAAAAAAAAAAEujEQ4AAAAAAAAAsDQa4QAAAAAAAAAAS6MRDgAAAAAAAACwNBrhAAAAAAAAAABLoxEOAAAAAAAAALA0GuEAAAAAAAAAAEvzcHYBsLbffvtNO3bs0M8//6yDBw/q+PHjMsbozTffVLdu3QrcNjU1VStXrtSWLVt04sQJZWRkqFatWmrRooWGDBmitm3bltFRoChWrlypiIgIHTp0SLGxsUpKSpKvr69uu+029enTR71795abm5tj/cuXL2vnzp369ttvtWfPHp06dUoZGRm64YYbFBwcrKFDh6pZs2ZOPCIURXFftxMmTNCmTZskqVDvEwBKHtntmshv10V2A9ZAfrsestt1kd3Xj0Y4StXatWu1YsWKIm938uRJjRgxQidOnFCdOnV0xx13yN3dXadPn9YXX3yh2267jTAup9577z3FxsaqSZMmCgoKUtWqVXX69Gnt2rVLO3fu1NatWzV//nxVqnTlBymbNm3SCy+8IEmqX7++OnToIA8PD/3666/asGGDNm3apOnTp6t///7OPCwUQnFft59//rk2bdokNzc3GWPKsHIAVyO7XRP57ZrIbsA6yG/XQ3a7JrK7eGiEo1TZbDaNGDFCLVq0UIsWLTRt2jSFh4cXuE1KSoqGDx+ukydPasKECRoxYoTc3d0dy+Pi4hQfH1/apeM6zZs3T82aNZO3t3eO2w8fPqyhQ4fqiy++UGhoqPr16ydJ8vDwUL9+/TRo0KAcZ5+NMfrggw80Z84czZw5U+3atdMtt9xSpseCwivu6zY2NlbTp09X06ZNVbVqVf30009lUTaAPJDdron8dj1kN2At5LfrIbtdD9ldfIwRjlL1yCOPaPLkyerevbtuuummQm2zcOFC/f7773rssccUEhKS40UtSf7+/rwpl2PBwcG5gliSmjRpokcffVSS9MMPPzhu79Onj1599dVcP8Fyc3PTsGHD1KFDB2VkZOizzz4r3cJRLMV93c6cOVMJCQl69dVX5eHBOVrAmchu10R+ux6yG7AW8tv1kN2uh+wuPhrhKFfS09P18ccfS5KGDh3q3GJQ4uxvtJUrVy70Nk2bNpUknTt3rlRqQvEV93W7ZcsWbd68Wf/4xz8Ykw6ogMhu6yO/rYfsBkB+WxvZbT1kd8lwzfY/yq3IyEjFx8crICBADRs2VGRkpLZt26bY2FjVqlVLHTt2VHBwsLPLxHU4efKkPvzwQ0lS586dC73diRMnJEl16tQplbpQfMV53cbGxmrGjBm69dZbNXr06DKuHEBJILutjfy2JrIbAPltXWS3NZHdJYNGOMqVQ4cOSZICAgI0d+5cvf/++zmWL1iwQF27dtVrr72W50+AUH58+umn2r17tzIyMnTu3Dnt3btX2dnZevLJJ3XfffcVah/R0dH65ptv5ObmVuhtUPaK87qdPn264uPjtWDBgiJdrQCg/CC7rYX8dg1kNwDy2zrIbtdAdpcMhkZBuZKQkCBJioqK0vvvv68hQ4Zo27Zt2r17txYsWKCAgABt375dM2bMcHKluJaffvpJoaGh2rRpk3bv3i1JevrppzVq1KhCbZ+cnKyJEycqMzNTffv21W233Vaa5aIYrvd1GxYWpq1bt2rw4MEKCgpyRukASgDZbS3kt2sguwGQ39ZBdrsGsrtk0AhHuZKdnS1JysjIUO/evfX888/rpptuUvXq1dWlSxe98847cnNz04YNG/T77787uVoU5JVXXlF0dLT279+vsLAwPf7445o/f7769+9/zTHHMjIy9PTTT+vQoUNq2rSpXnjhhTKqGtfjel63Fy5c0MyZM9WwYUM988wzziwfQDGR3dZCfrsGshsA+W0dZLdrILtLBo1wlCvVqlVz/N2/f/9cy1u2bKnmzZvLGKPw8PCyLA3XycvLS40bN9aUKVP07LPP6tdff9WsWbPyXT8zM1PPPvusduzYob/85S9aunQpP8Ur567ndfvyyy8rISFBs2bNUtWqVcusVgAlj+y2JvLb2shuAOS39ZDd1kZ2lwzGCEe50qBBgzz//vM6Bw8e1IULF8qqLJSQPn36aO7cufrqq6+UkZEhT0/PHMuzsrI0ceJEff7557r55pu1bNky1apVy0nVorCu53W7fft2eXl5acGCBVqwYEGOdaOioiRJb7/9tlavXq3bb7+ds9dAOUZ2Wx/5bT1kNwDy29rIbushu0sGjXCUK82aNXP8HR8fr7p16+ZaJy4uTpI4U1kB1ahRQx4eHsrMzFRCQoJq167tWJaVlaVJkyZp8+bNatCggZYvX66AgAAnVovCut7XbWpqaoFXlxw5ckSS5OvrW1KlAigFZLf1kd/WQ3YDIL+tjey2HrK7ZNAIR7kSEBCg1q1ba//+/dq5c6eaNm2aY3lCQoJ++eUXSVKLFi2cUSKKYffu3crMzFT16tXl7+/vuD07O1uTJ09WWFiY6tevrxUrVuT5po7y6Xpet9HR0fnub/DgwQoPD9ebb76pbt26lV7hAEoE2W195Lf1kN0AyG9rI7uth+wuGYwRjnLnySeflCQtWrRIP//8s+P2tLQ0TZ8+XYmJiWrevDmz3ZZDERER+uqrr5SZmZlr2Z49ezRt2jRJ0sMPPyx3d3dJV4J46tSp2rRpk+rVq6fly5erfv36ZVo3io/XLeDaeA+o2Mhv18TrFgDvAxUX2e2aeM0Wn5sxxji7CFhXZGSkZsyY4fj3kSNHlJycrEaNGqlGjRqO2z/++OMc282dO1fvv/++PD091bp1a/n5+enAgQP6448/FBAQoBUrVqhRo0ZldRgopHXr1mnq1KmqXr26mjVrptq1ays5OVknT550/Nzmr3/9q9588015eXlJklasWKFXXnlFktS+fft8g7ht27Z65JFHyuZAcF1K6nXrqmemgfKC7HY95LfrIrsB6yC/XQvZ7brI7uJhaBSUqqSkJO3fvz/X7cePHy9wuylTpigoKEirVq1SVFSULl++rHr16mnYsGEKCQlRzZo1S6liFEe7du00atQoRURE6MSJE9q7d6+MMapTp44eeOAB9e7dW127ds2xTUJCguPva81GThiXb7xuAWsgu10P+e26eN0C1kF+uxay23Xxmi0erggHAAAAAAAAAFgaY4QDAAAAAAAAACyNRjgAAAAAAAAAwNJohAMAAAAAAAAALI1GOAAAAAAAAADA0miEAwAAAAAAAAAsjUY4AAAAAAAAAMDSaIQDAAAAAAAAACyNRjjgBOfPn9e0adN07733qnnz5goMDNTgwYOdXVaJGTx4sAIDA7Vu3TpnlwIAwHWzel6XN2+//bYCAwP13HPPObuUInnuuecUGBiot99+29mlAIBlkMEl77vvvtOQIUMUHBys2267je/scEkezi4AcDWZmZkaMmSIjh49KkmqUaOGPD09VaNGDSdXdm2nTp1SaGiofH19NXToUGeXU2IuXbqk5cuXS5LGjh3r5GoAAOVBRc5rlBx7c3vIkCGqXr26k6sBANdABpe8iIgIjRw5UtnZ2XJ3d1fNmjXl5uYmLy8vp9UUFRWl7du3q379+urbt6/T6oBroREOlLHvvvtOR48elZ+fnz766CM1atTI2SUVWkxMjObPn6/69esX2AivW7eubrnlFvn6+pZdccVw6dIlzZ8/XxKNcADAFRU5r1Fy7J8P+vTpQyMcAMoIGVzyVqxYoezsbD344IOaM2eOUxvgdlFRUZo/f77at29PIxxlhkY4UMYOHz4sSbrjjjssG+j/+te/nF0CAADF4gp5DQBAeUQGl7wjR45Ikv7+97+XiyY44CyMEQ6UsbS0NEmSt7e3kysBAAD5Ia8BAHAOMrjkpaamSuIxBdyMMcbZRQBW0blzZ8XExGjFihVq1KiRFi5cqB07dujcuXPy8vJSYmJivtt+8cUXatCgwXXfd2xsrJYtW6avv/5ap06dkjFGDRo0UJcuXTRs2DD5+fnl2iY9PV1r167V5s2bdeTIEV2+fFk1atRQ7dq11bZtW/Xu3VtBQUE5ji0///znPx0/Zxo8eLDCw8Nz3CZJP/74ox5//HHVr19fX375pXbs2KGlS5fql19+UXZ2tlq2bKlx48Y57jMxMVFLlizRli1bdObMGfn7+6tnz54aO3Zsnmexz549q7CwMP3www86efKkzp07J3d3d910003q3Lmzhg4dmutnzfZa8zNmzJhcw6Xs2rVLq1at0t69e5WQkKAaNWqoTZs2GjRokDp06JDnfgIDAyVdeZ7T0tK0ePFi7dq1SxcuXNC9996rBQsWSJIuXryoJUuWaMeOHTp16pSysrJUs2ZN1a1bVx06dNDDDz+s+vXr51svAODanJHXzz33nEJDQzVmzBg98cQTWrJkiTZu3KjTp0/L399fXbt21dixYx3jnx48eFDvvvuu9u7dq8TERDVp0kRPPfWUunbtmuf+d+/era1bt2r//v06c+aM4uPj5evrq2bNmumRRx5Rt27dcm2zY8cOjRw5UpK0ZMkS3X333bnWWbRokebNmycfHx9t2LChWJ9VJOm3337T/PnztXPnTiUnJ6tevXrq2bOnQkJCtGjRIs2fP199+vTRnDlz8tz+yy+/1CeffKIDBw4oISFBvr6+at26tR577DHdc889udZft26dpk6dqvbt22vlypUKDQ3V2rVrdeTIEbm5ualFixYaMWKEOnXqlGM7+/OVn6trvPq5HTVqlFauXKl169bpxIkTqlKlioKCgjRmzBi1bNmyGI8cAFgDGXxFWWWw/XtoXuzZeLWIiAitXr1ae/bsUWxsrKpVq6amTZvq4YcfVo8ePeTm5pZrP5GRkdq8ebP27NmjM2fO6MKFC6pWrZpsNpt69+6tvn37yt3dvdB1SVeGcrnjjjt06tQpdenSRZIUHR2d57p/7jNcraD/b3/5y1+0YcMGx7rJyclauXKltm3bpuPHjys9PV1169ZVp06dNGLECNWtWzfXfWdnZ2v9+vUKDQ3VoUOHlJSUJF9fX9WqVUutWrXSgw8+mOszBpyLoVGAUnD8+HE9/fTTiouLU9WqVeXp6al69erp4sWLSklJUUpKiqpUqZJjDO0/B0NRREREaPTo0YqPj5ckeXp6qlKlSjp8+LAOHz6sDRs26P3339ett97q2CYzM1MjRoxwNIHd3Nzk6+ur+Ph4Xbx4UdHR0YqPj3c0pf39/ZWUlKSEhARVqlRJNWvWzFFDUX9etXr1as2aNUtubm7y9vZWUlKSfvjhB+3Zs0fLli3TLbfcoiFDhujQoUPy9vZWdna2zp49qyVLlujIkSNatGhRrn2++uqr2rp1q+MxqFatmi5duqSoqChFRUVp48aNWrlypW688UbHNjVq1JC/v7/i4uIkSbVr186xzz+fMX/jjTf07rvv5njMLl68qO3bt2v79u0KCQnRhAkT8j3uiIgITZ8+XZcvX1a1atVyPO8xMTEaMGCAzp8/L+nK/wkfHx+dO3dOZ8+e1d69e3XDDTdo4MCBRXmoAQD5KOu8lqSMjAwNGzZMERERqlKliiTpzJkzWrlypfbu3as1a9Zox44deuaZZ5SRkSEfHx+lpaXp4MGDGjM5y6dHAAAY7klEQVRmjObNm6fu3bvn2GdycrIGDRrk+He1atVUpUoVxcbG6rvvvtN3332nAQMGaObMmTm2u+eee/TYY49p1apVmjp1qjZu3JjjxPkvv/zimCxy2rRpxW6C7969WyNHjtTly5clST4+Pjp16pTefvttfffdd2rfvn2Bj5u9RjsfHx/Fxsbqq6++0ldffaV//OMfmjRpUr77ePXVV7V8+XJVqlRJPj4+SkxM1K5du7Rr1y5NnjxZI0aMyLHv2rVr68KFC5KufA66+rn38fHJtf/MzEyFhITou+++k6enpzw9PZWQkKCvv/5aO3fu1PLlyx2fqwDA1ZHBZZPB9u+3sbGxys7Odkw8KinX5KOvvfaalixZ4vi3j4+PEhIStHPnTu3cuVNffvmlXn/9dVWqlHNwieHDhzt6EVWrVlXVqlUVHx+v8PBwhYeHa9u2bVqwYIE8PP7Xgqxdu7ZSU1OVlJSU50So9hpLSl7/36529OhRjRw50nHxn4eHhypXrqwTJ05o5cqV+u9//6uFCxeqbdu2ObabNGmSNm3a5Pi3r6+vkpKSFBcXpyNHjujo0aM0wssbA6DE/O1vfzM2m820adPG9OzZ0+zZs8ex7Pjx48YYY9566y1js9nMlClTSuQ+T506ZYKDg43NZjMvv/yyOX78uMnKyjJZWVkmOjraDB8+3NhsNtO9e3eTmZnp2C40NNTYbDbTunVrs379epOammqMMSYzM9PExMSYVatWmXfffTfHfe3atcvYbDbzt7/9rcCaBg0aZGw2m/n000/z3L5169amefPmZt68eSYhIcEYY8zJkyfNgAEDjM1mM/369TNjxowxDzzwgNm9e7fJzs42aWlp5uOPPzbNmjUzNpvNfP3117nu94033jArVqwwx44dM1lZWcYYY9LT082PP/5o+vXrZ2w2mxk5cmSu7U6ePGlsNpux2WwFHtemTZsc682cOdNcvHjRGGNMbGysmTVrlmPZ+vXrc21rX9amTRszaNAgEx0dbYwxJjs725w4ccIYY8xzzz1nbDabue+++8zu3bsdx5CWlmaio6PNG2+8YbZt21ZgjQCAa3NGXk+ZMsXYbDbTtm1b07FjR/PVV1+ZrKwsk5mZabZt22aCgoKMzWYzr732mmnbtq2ZOnWq+eOPP4wxxly8eNE89dRTxmazmY4dO5qMjIwc+05JSTFjx44127ZtM3FxcY7bExISzKpVq0ybNm2MzWYzn332Wa66Ll++bLp162ZsNpsZN26c4/bU1FTTvXt3Y7PZzJgxY4p9/PHx8aZDhw7GZrOZPn36mKioKGPMlZwODQ01rVu3Nm3bts33MX/llVccGfnZZ5+Z5ORkY4wxiYmJZvXq1Y7Hb+PGjTm2+/TTT43NZjO333674/G9dOmSMcaYc+fOmQkTJhibzWYCAwPN7t27c92vPb9PnjyZ77HZn9vg4GDTvn17ExYWZtLS0owxxkRFRZmePXs6Pt8AgKsjg/+nrDLYmP897rt27cpz+QcffGBsNpu56667zEcffeTIysuXL5uwsDDTsWNHY7PZcvUIjDHm2WefNZs2bXI8ZsYYk5ycbNavX+/Y7r333su1nT2jBw0alG/dhfmuXlCfojD/3y5duuRYb9y4cSYqKsrRO/n999/Ns88+63hs7P0LY4wJDw83NpvNNG3a1CxbtswkJiYaY658xz937pxZt26dmTNnTr51wzlohAMlyP7mGRwcbM6fP5/nOiUd6vYvcK+//nqey9PS0kyvXr2MzWYzmzdvdtz+8ssvG5vNZl566aVC31dJNcJtNpt57rnncm0XExNjAgMDjc1mM82bN3cE09WmTp2a7/YFiYuLM3feeacJDAzM9WW2MOGanZ1t7rvvPmOz2cwzzzyT5zr2gPzb3/7maGLb2fffpUsXc/ny5Ty3f/DBB43NZjNhYWFFOjYAQNE4I6/tX8JtNpv58ccfcy2fP3++Y/ngwYNzLU9OTnZ8UQ8PDy/SfdtPfuf3RfPgwYOmefPmxmazmdDQUGOMMbNnz3Z86Y+NjS3S/eXFfnzt27d3nEi+2vr16x3H/+fH/NixYyYwMNDceeed5vTp03nu336yukePHjlut3/JttlsZtq0abm2y87ONoMHDzY2m80MGTIk1/KiNMJtNluezfSff/7ZsTwmJibf/QCAKyCDcyqLDDam4EZ4QkKCadOmjWnZsqXjRPWf/fTTTyYwMNC0a9fOcbK3MHbv3p1vD6EsG+EF/X+bN2+esdls5tlnn833PkaMGGFsNptZsmSJ47bFixcbm81mRowYke92KH+YLBMoBX//+99zDbFRGi5fvqwtW7aoUqVKGjZsWJ7rVK5cWQ888IAk6YcffnDcbv9Jr30YjrIWEhKS67Z69erp5ptvliQ98MADjr+vZh+D2z6TeGH5+fkpKChIxhjt3bu3yPVGRUXpxIkTkqSnnnoqz3VGjx4t6coQJwcOHMhznUGDBuU7jIz9Ofnjjz+KXB8AoOjKKq+vFhQUlOcQIHfddZfj7yeeeCLXcm9vb7Vp00aSdOjQoSLdZ+fOnSVJ+/fvV1ZWVq7lzZs3d8yHMWvWLP3nP/9xjBn6yiuvyN/fv0j3lxf70GX9+/fPNbyaJPXu3TvfOTDWr18vY4y6d++e5/ic0pXPDZUrV9bhw4fzzdG8Hlc3NzfH7bt27XL8tPt6BAcHKzg4ONftLVq0cAzLVtTPLwBgVWTwFWWRwdeydetWpaSk6K677tJtt92W5zpBQUFq0KCBEhISFBkZWeh9BwcHq3r16oqJidG5c+dKquQiK+j/2/r16yUp356KJPXs2VNS3j0V+7AzqBgYIxwoBWU1/mNkZKQyMjLk5uamXr165buefYboM2fOOG7r1KmT3nvvPX3xxRd68skn1bdvX7Vr165MgrZKlSpq1KhRnstq1aql48ePy2az5bncHl6XLl3Kc/mBAwe0du1a7d27V+fOnVNKSkquda6n0WwP+5o1a6pJkyZ5rnPrrbcqICBA586dU2RkpOPD0tXyus2uU6dO2r9/v15//XWdOHFCDzzwgNq0aVPk8dcBAIXjjPGa88u3WrVqOf7OL2fs6+SVgZmZmQoNDdWWLVsc83xkZGTkWCctLU0JCQl5NqJHjhypb775Rnv27NG0adMkSQMHDtS9995buAMrQHp6uo4cOSJJateuXZ7ruLm5KTg4OM+Jue0nsO3Hl5/MzExJVybPvuGGG3Isq1evnho2bJjndm3btpW7u7uysrIUFRWV78TX11LQZJgBAQE6e/asEhISrmvfAGA1ZPD/lGYGF4Y9Z3ft2qWOHTvmu549w86cOZPr+du8ebM2btyoX375RbGxsUpLS8u1/R9//KGAgIASrLzw8vv/dubMGZ09e1bSlYv18poMVJLj+by6p9KhQwd5enoqMjJSgwcPVv/+/XXnnXc67RhRODTCgVKQV7iVBntD1xjjmMypIPaGuHRlhuhx48ZpwYIFjkmmpCvN3L/+9a8aMGBAvs3q4qpdu3a+AWOfAKVOnTp5LrdPzGH/snu1pUuX6rXXXpMxxrGvqycDSUxMVFpammOSrqKwT6Z5rVC78cYbde7cOcf6f1bQ/42QkBBFRkbqyy+/1Jo1a7RmzRp5eHioZcuW6tq1q/r376/q1asXuXYAQN7KKq+vdq18k5SriWtnz8g/Z2BycrJGjBiR4xdPXl5e8vX1dezX/jkhvwysVKmSZs+erQcffFCSVL9+fU2ZMqUwh3RNCQkJjqvg8js2Kf+Mtf96LTk5WcnJyde8v7yOsaD89vLyUvXq1RUXF5dvfhdGtWrV8l1mn5Qtr88vAOCKyOCc919aGVwY9py9fPlyob4rX91XyMzM1Pjx47Vt2zbHbZUrV84xybT9iunr+R5eUvL7/3b1L+QvXrx4zf1cfeyNGjXS9OnTNWvWLEVERCgiIkLSlefvnnvu0YABA9SsWbNiVo6SRiMcKAV/nkW5tNgbvr6+vo433aIYPXq0evfurc2bN+vHH3/Uvn379Ntvv+m3337TihUr9Morr+ihhx4q6bJLxeHDh/X666/LGKNBgwZp4MCBuuWWW3LMLD5p0iT997//dTxu1yOvM9tFUdD/jcqVK2vhwoXat2+ftm3bpoiICEVGRmrv3r3au3evli5dqmXLluX7czUAQNGUVV6XtgULFmjv3r3y9/fXc889p3vuuSfH1W1ZWVmOL2IFZeC6descf58/f16///67AgMDS6/wQrL/3Hjq1KkaOnSoc4sBAJQIMjgnZ2awPWcff/xxxxXphfXxxx9r27Ztqlq1qiZMmKD77rvPMRyY3b333quzZ88W63t4ceX3/+3qIU12795d5AvPHn74Yf31r39VWFiYfvzxR+3Zs0cxMTH68MMP9dFHH2n8+PF68skni1U7SpY13nkAF2UP2KSkJCUmJl7XPho2bKiQkBAtXbpU4eHhWrFihdq1a6fMzEzNmDGjUGdFy4OtW7cqOztbd999t1588UU1btw4RxNcKtwZ3vzYh4yx/2wqP/blxRlipk2bNpo0aZI++ugjhYeHa968eapXr55iY2OL/MEEAGB99uFCXnzxRT300EM5voBLKtSvxiIiIrR06VJJV346np6erkmTJik9Pb3Y9dWoUcORyQUNT5bfMvuwaFf/HLmoCrrftLQ0x0/dy2KIOACAdZT3DC6M4uSs/fhHjRqlwYMH52qCZ2VlFevXVh4e/7t+N7+L0q63FyLlHBbn9OnT17WP2rVra8iQIVqwYIF27dqlTz75RPfdd5+MMXrzzTf166+/Xnd9KHk0woEKrEWLFvLw8JAxRjt27Cj2/tzd3XXHHXfo3Xfflaenp1JSUnTw4EHHcvtZVGeeyc2PfeKN/H56lJKSon379uW57Oqzw/kdW/PmzR37yW8izGPHjjnqsK9fXN7e3urRo4dmzpwp6cpY5XmNew4AcF327GnatGmey3fu3Fng9klJSZo8ebKys7PVr18/LV++XLVq1VJ0dLT+/e9/F7u+ypUrq3HjxpKU7y/YjDHavXt3nsvs82sU57NOTEyMTp06leeyPXv2KCsrS25ubrkeQ/tQbuXxsw8AwPnKewYXhj1nw8PDcwz9URjXOv6ffvop3wZ2YfoLvr6+jr/zuyjt559/LlSteWnYsKHjRMC333573fuxc3NzU6tWrfTmm2/qxhtvVHZ2tvbs2VPs/aLk0AgHKjAfHx/df//9kqQ333xTSUlJ+a6bmZmZY1zNgs4uV65c2RFKV69nnxW5OGdcS4u9tvxm8X733XfzHVfUvq2U/yScTZs21c033+zYV17mz58v6cqYYK1atSpc4Vcp6DmxT5hpjMk18QoAwLUVlIHJyclauHBhgdvPnj1bMTExatCggZ5//nnVrFlTs2bNkiQtW7Ys3wZ1UXTr1k3SlZ9Qx8fH51oeFhaW50SZkvTQQw/Jzc1NR48e1Ycffljg/RQ0GeXixYtz3WaMcdzeoUMH+fn55Vhenj/7AACcryJk8LV069ZN3t7eSkhI0DvvvFPgun/O2YKOPzMzs8Bmvn3b/L6DS1fm36hfv74k6Ysvvsi1PC4uTv/5z38KrPla+vTpI0l6//33HY39vBhjctRa0Pd3d3d3x9XsfH8vX2iEAxXchAkT5Ofnp+PHj2vgwIH69ttvHW+0xhgdP35cy5Yt04MPPpjj6u4pU6Zo6tSp2rFjR44G+qlTpzRlyhSlpaXJy8tLbdu2dSy7+eab5enpqcTERG3durXsDrIQ7LNbf/3111q0aJFjIo7Y2FjNnTtXixYtyvXl1q569eqOSVGuHpvtam5ubho/frykKwE8a9Ysx0+84uLiNHv2bG3atEmSNH78+Osa865Xr16aN2+eDhw44AhVY4wOHDjg+DDUsmVL1ahRo8j7BgBYlz0D58yZo/DwcMeVVQcOHNDQoUPzbDzbff755woNDVWlSpX0r3/9y/GltEuXLnr44YeVnZ2tKVOmFHiyvTAee+wx1apVS3FxcRoxYoTjZ8IZGRnasGGDXnzxxRxXfV2tcePGjrHBZ8yYof/3//5fjqvCkpKS9N1332nixIl6+umn89yHj4+PPvroI82bN8/R1D5//rymTJminTt3ys3NTaNHj87zviVp/fr1jgk/AQCwqwgZfC3+/v569tlnJV05afzCCy/o2LFjjuWpqamKiIjQyy+/rP/7v//Lsa39+BcsWKDt27c7svLo0aN68skndeDAAXl7e+d5v/aMPXr0qPbv359vffZJRBcuXKgvvvjCMWHpvn37NGzYsGI3mkNCQtSwYUPFxcXp//7v//TZZ5/luDL+9OnT+uijj9SnTx9t377dcfsbb7yhcePGafv27Tme5wsXLmj27Nk6deqU3NzcdNdddxWrPpQsJssEKrgGDRrovffe0+jRo3Xo0CGNHDlSnp6eqlatmpKTk3OEgv3nvdKV8bU+++wzrVu3Tm5ubvL19VVGRoajgezu7q4ZM2bkmF3ZPkzH+vXrNW7cOPn6+jomk5g8ebLjai9nuPvuu3X//ffr888/17x58/TGG2+oevXqunTpkowxevjhh5WVlaXQ0NA8t3/kkUf0zjvvaM6cOXrrrbccY4Q+/vjjji/f3bt3V3R0tN59912tWrVKa9aska+vrxITEx2TbISEhKh3797XdQwXL17UokWLtGjRIrm7u8vX1zfHc+jv76/Zs2df174BANY1fvx4ff/99zpz5owGDx6sKlWqyN3dXSkpKfLy8tI777yjESNG5Nru/PnzevHFFyVJ//jHP3Kc/Jak559/Xj/++KNOnjyp2bNna86cOdddY40aNfTvf/9bI0eO1MGDB/X3v/9dvr6+SktLU3p6uoKCgtSuXbs8r9qWrkx4nZqaqrVr12rx4sVavHixfHx85ObmpqSkJEfjoX379nlu36xZMzVt2lSLFi3SkiVL5OPj4/iMYN9/cHBwru0eeeQR7d27V8uXL9eHH36oWrVqyc3NTQ888ICmTJly3Y8HAMAaKkIGF8bgwYOVmJiot956S5988ok++eQTeXt7Oy6Es3/ftV+dbTd8+HBt3rxZv//+u0aPHi1PT09VqVJFSUlJcnd31+zZszV//vw8h/ds1KiR2rVrp927d6t///7y8/NTtWrVJEnz5s1zDNkSEhKirVu36uTJkxo1apQqV64sDw8PpaSkqF69epo2bZomT5583cdevXp1LV26VE899ZSOHj2qZ555xvF9PDU1NUdT/OqeSmZmprZu3eq4SNDHx0fGmBy/RB8/frxsNtt114aSxxXhgAW0atVKmzdv1sSJExUUFCRvb28lJiaqatWqatGihQYPHqxVq1bl+HI4YcIETZo0Sffcc48aNmyojIwMZWVl6aabblLfvn21bt06PfTQQ7nua8aMGXriiSd06623Kj09XTExMYqJiSkX41a/8cYbmjBhgv7yl784xk6//fbbNXfuXL3yyisFbjt69GhNnDhRgYGBMsY4juvPP4V+5pln9MEHH6hLly7y9/dXSkqK/Pz81LlzZ33wwQeaMGHCdde/YMECPfHEE7r99tt1ww03KDk5WZ6engoMDFRISIjCwsJ02223Xff+AQDW1LBhQ33yySfq3bu3atWqpezsbPn6+qpXr176z3/+o7vvvjvP7Z5//nnFx8eradOmGjt2bK7l1apV09y5c1WpUiWFhobq888/L1ad7du3V2hoqLp3766aNWsqPT1d9evX19ixY7V8+XJVrlw5323d3d01ffp0rVmzRr1791b9+vWVnp6utLQ01atXT507d9ZLL72kt956K999PP/88/rnP/+p5s2bKzMzU97e3rrjjjv03nvv5dmkkKR+/fpp9uzZatWqldzd3XXmzBnFxMQUa+IvAIB1VJQMLoxRo0Zpw4YNGjBggBo1aqTs7GylpKSoTp06uvvuuzVp0iStWbMmxzZ+fn766KOPNHDgQMdEmV5eXuratatWrlypvn37Fnifb7/9th599FE1aNBAycnJju/hV48rXqNGDX344YcaMGCAbrjhBhlj5Ofnp8GDB2vdunW5Jui8HjfffLPWr1+vl19+WXfccYeqV6/uaOYHBgZqwIABWrx4cY6L3oYOHaoXXnhBXbp0UaNGjWSMUXp6uurWravu3btr9erVevLJJ4tdG0qWm2HmFwAAAAAWtG7dOk2dOlXt27fXypUrnV0OAAAAnIgrwgEAAAAAAAAAlkYjHAAAAAAAAABgaTTCAQAAAAAAAACW5uHsAgBc0bFjxyKtP3z48HwndgIAAKXD1fN6zJgx2rt3b6HXDwoK0vz580uxIgCAqyCDyWCguGiEA+XEhQsXirR+SkpKKVUCAADy4+p5nZCQUKTHICEhoRSruba+ffuqb9++Tq0BAFAyyOCKlcFAeeRmjDHOLgIAAAAAAAAAgNLCGOEAAAAAAAAAAEujEQ4AAAAAAAAAsDQa4QAAAAAAAAAAS6MRDgAAAAAAAACwNBrhAAAAAAAAAABLoxEOAAAAAAAAALC0/w9zD3HOV6//oAAAAABJRU5ErkJggg==",
      "text/plain": [
       "<Figure size 1512x504 with 3 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "fig, ax = plt.subplots(1,3,figsize=(21,7))\n",
    "sns.boxplot(data=perf_df, x=\"rf_estimators\", y=\"valid_r2_score\", ax=ax[0])\n",
    "sns.boxplot(data=perf_df, x=\"rf_max_depth\", y=\"valid_r2_score\", ax=ax[1])\n",
    "sns.boxplot(data=perf_df, x=\"rf_max_features\", y=\"valid_r2_score\", ax=ax[2])\n",
    "fig.tight_layout()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "AHNcKhLVRVpp"
   },
   "source": [
    "## Time for running the following code chunk: ~ 10 minutes on COLAB-Pro"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "qpAq__JA61wy"
   },
   "source": [
    "### Random search"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 20,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/"
    },
    "id": "LMNkU4rs648c",
    "outputId": "cb987009-f27f-433b-8be6-539c6ed84849"
   },
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "2021-09-18 16:35:50,339 Attempting to load featurized dataset\n",
      "2021-09-18 16:35:50,349 Exception when trying to load featurized data:\n",
      "DynamicFeaturization doesn't support get_featurized_dset_name()\n",
      "2021-09-18 16:35:50,351 Featurized dataset not previously saved for dataset H1_std, creating new\n",
      "2021-09-18 16:35:50,365 Featurizing sample 0\n",
      "2021-09-18 16:35:55,635 Featurizing sample 1000\n",
      "2021-09-18 16:35:57,562 Previous dataset split restored\n",
      "2021-09-18 16:35:57,581 Wrote transformers to ampl_output/H1_std/RF_ecfp_scaffold_regression/d364621c-ef1e-4cc4-aed7-3c86da0752ea/transformers.pkl\n",
      "2021-09-18 16:35:57,583 Transforming response data\n",
      "2021-09-18 16:35:57,609 Transforming response data\n",
      "2021-09-18 16:35:57,623 Transforming response data\n",
      "2021-09-18 16:35:57,653 Fitting random forest model\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "number of features: 1024\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "2021-09-18 16:35:58,643 Fold 0: training r2_score = 0.966, validation r2_score = 0.459, test r2_score = 0.539\n",
      "2021-09-18 16:35:58,970 Wrote model tarball to ampl_output/H1_std_model_d364621c-ef1e-4cc4-aed7-3c86da0752ea.tar.gz\n",
      "2021-09-18 16:35:58,986 Attempting to load featurized dataset\n",
      "2021-09-18 16:35:58,995 Exception when trying to load featurized data:\n",
      "DynamicFeaturization doesn't support get_featurized_dset_name()\n",
      "2021-09-18 16:35:58,996 Featurized dataset not previously saved for dataset H1_std, creating new\n",
      "2021-09-18 16:35:59,008 Featurizing sample 0\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "rf_estimators: 64, rf_max_depth: 64, rf_max_features: 64, valid_r2: 0.45875868345740045\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "2021-09-18 16:36:03,933 Featurizing sample 1000\n",
      "2021-09-18 16:36:05,837 Previous dataset split restored\n",
      "2021-09-18 16:36:05,853 Wrote transformers to ampl_output/H1_std/RF_ecfp_scaffold_regression/648d1a42-614b-4a18-93ff-429c737816a6/transformers.pkl\n",
      "2021-09-18 16:36:05,856 Transforming response data\n",
      "2021-09-18 16:36:05,883 Transforming response data\n",
      "2021-09-18 16:36:05,897 Transforming response data\n",
      "2021-09-18 16:36:05,926 Fitting random forest model\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "number of features: 1024\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "2021-09-18 16:36:07,125 Fold 0: training r2_score = 0.966, validation r2_score = 0.438, test r2_score = 0.531\n",
      "2021-09-18 16:36:07,462 Wrote model tarball to ampl_output/H1_std_model_648d1a42-614b-4a18-93ff-429c737816a6.tar.gz\n",
      "2021-09-18 16:36:07,476 Attempting to load featurized dataset\n",
      "2021-09-18 16:36:07,484 Exception when trying to load featurized data:\n",
      "DynamicFeaturization doesn't support get_featurized_dset_name()\n",
      "2021-09-18 16:36:07,485 Featurized dataset not previously saved for dataset H1_std, creating new\n",
      "2021-09-18 16:36:07,496 Featurizing sample 0\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "rf_estimators: 64, rf_max_depth: 64, rf_max_features: 64, valid_r2: 0.43762827525482717\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "2021-09-18 16:36:12,625 Featurizing sample 1000\n",
      "2021-09-18 16:36:14,571 Previous dataset split restored\n",
      "2021-09-18 16:36:14,589 Wrote transformers to ampl_output/H1_std/RF_ecfp_scaffold_regression/f448c5e4-ee1c-4e68-9943-2d675ae3acb8/transformers.pkl\n",
      "2021-09-18 16:36:14,590 Transforming response data\n",
      "2021-09-18 16:36:14,616 Transforming response data\n",
      "2021-09-18 16:36:14,631 Transforming response data\n",
      "2021-09-18 16:36:14,661 Fitting random forest model\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "number of features: 1024\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "2021-09-18 16:36:15,853 Fold 0: training r2_score = 0.966, validation r2_score = 0.464, test r2_score = 0.511\n",
      "2021-09-18 16:36:16,187 Wrote model tarball to ampl_output/H1_std_model_f448c5e4-ee1c-4e68-9943-2d675ae3acb8.tar.gz\n",
      "2021-09-18 16:36:16,200 Attempting to load featurized dataset\n",
      "2021-09-18 16:36:16,209 Exception when trying to load featurized data:\n",
      "DynamicFeaturization doesn't support get_featurized_dset_name()\n",
      "2021-09-18 16:36:16,210 Featurized dataset not previously saved for dataset H1_std, creating new\n",
      "2021-09-18 16:36:16,221 Featurizing sample 0\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "rf_estimators: 64, rf_max_depth: 64, rf_max_features: 64, valid_r2: 0.46441668064531505\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "2021-09-18 16:36:21,157 Featurizing sample 1000\n",
      "2021-09-18 16:36:23,024 Previous dataset split restored\n",
      "2021-09-18 16:36:23,041 Wrote transformers to ampl_output/H1_std/RF_ecfp_scaffold_regression/86a70822-0a9f-4553-b5ed-00a0ebb6845f/transformers.pkl\n",
      "2021-09-18 16:36:23,042 Transforming response data\n",
      "2021-09-18 16:36:23,070 Transforming response data\n",
      "2021-09-18 16:36:23,084 Transforming response data\n",
      "2021-09-18 16:36:23,114 Fitting random forest model\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "number of features: 1024\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "2021-09-18 16:36:24,958 Fold 0: training r2_score = 0.966, validation r2_score = 0.454, test r2_score = 0.549\n",
      "2021-09-18 16:36:25,673 Wrote model tarball to ampl_output/H1_std_model_86a70822-0a9f-4553-b5ed-00a0ebb6845f.tar.gz\n",
      "2021-09-18 16:36:25,689 Attempting to load featurized dataset\n",
      "2021-09-18 16:36:25,698 Exception when trying to load featurized data:\n",
      "DynamicFeaturization doesn't support get_featurized_dset_name()\n",
      "2021-09-18 16:36:25,700 Featurized dataset not previously saved for dataset H1_std, creating new\n",
      "2021-09-18 16:36:25,712 Featurizing sample 0\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "rf_estimators: 64, rf_max_depth: 64, rf_max_features: 64, valid_r2: 0.45441029873539696\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "2021-09-18 16:36:30,697 Featurizing sample 1000\n",
      "2021-09-18 16:36:32,627 Previous dataset split restored\n",
      "2021-09-18 16:36:32,647 Wrote transformers to ampl_output/H1_std/RF_ecfp_scaffold_regression/2fb5d465-37b1-4d15-9209-d245cf39a079/transformers.pkl\n",
      "2021-09-18 16:36:32,648 Transforming response data\n",
      "2021-09-18 16:36:32,675 Transforming response data\n",
      "2021-09-18 16:36:32,690 Transforming response data\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "number of features: 1024\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "2021-09-18 16:36:32,723 Fitting random forest model\n",
      "2021-09-18 16:36:33,722 Fold 0: training r2_score = 0.964, validation r2_score = 0.440, test r2_score = 0.540\n",
      "2021-09-18 16:36:34,093 Wrote model tarball to ampl_output/H1_std_model_2fb5d465-37b1-4d15-9209-d245cf39a079.tar.gz\n",
      "2021-09-18 16:36:34,114 Attempting to load featurized dataset\n",
      "2021-09-18 16:36:34,123 Exception when trying to load featurized data:\n",
      "DynamicFeaturization doesn't support get_featurized_dset_name()\n",
      "2021-09-18 16:36:34,125 Featurized dataset not previously saved for dataset H1_std, creating new\n",
      "2021-09-18 16:36:34,138 Featurizing sample 0\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "rf_estimators: 64, rf_max_depth: 64, rf_max_features: 64, valid_r2: 0.43956170904649194\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "2021-09-18 16:36:39,146 Featurizing sample 1000\n",
      "2021-09-18 16:36:41,034 Previous dataset split restored\n",
      "2021-09-18 16:36:41,051 Wrote transformers to ampl_output/H1_std/RF_ecfp_scaffold_regression/b1539d46-7cc0-41d7-b5de-ee272fb4abf3/transformers.pkl\n",
      "2021-09-18 16:36:41,053 Transforming response data\n",
      "2021-09-18 16:36:41,080 Transforming response data\n",
      "2021-09-18 16:36:41,095 Transforming response data\n",
      "2021-09-18 16:36:41,124 Fitting random forest model\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "number of features: 1024\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "2021-09-18 16:36:42,658 Fold 0: training r2_score = 0.966, validation r2_score = 0.469, test r2_score = 0.530\n",
      "2021-09-18 16:36:43,246 Wrote model tarball to ampl_output/H1_std_model_b1539d46-7cc0-41d7-b5de-ee272fb4abf3.tar.gz\n",
      "2021-09-18 16:36:43,262 Attempting to load featurized dataset\n",
      "2021-09-18 16:36:43,272 Exception when trying to load featurized data:\n",
      "DynamicFeaturization doesn't support get_featurized_dset_name()\n",
      "2021-09-18 16:36:43,273 Featurized dataset not previously saved for dataset H1_std, creating new\n",
      "2021-09-18 16:36:43,284 Featurizing sample 0\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "rf_estimators: 64, rf_max_depth: 64, rf_max_features: 64, valid_r2: 0.46854381635706865\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "2021-09-18 16:36:48,342 Featurizing sample 1000\n",
      "2021-09-18 16:36:50,311 Previous dataset split restored\n",
      "2021-09-18 16:36:50,329 Wrote transformers to ampl_output/H1_std/RF_ecfp_scaffold_regression/84bf891f-a23e-4ced-8253-f04a41a8003b/transformers.pkl\n",
      "2021-09-18 16:36:50,331 Transforming response data\n",
      "2021-09-18 16:36:50,357 Transforming response data\n",
      "2021-09-18 16:36:50,371 Transforming response data\n",
      "2021-09-18 16:36:50,401 Fitting random forest model\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "number of features: 1024\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "2021-09-18 16:36:51,488 Fold 0: training r2_score = 0.962, validation r2_score = 0.462, test r2_score = 0.534\n",
      "2021-09-18 16:36:51,744 Wrote model tarball to ampl_output/H1_std_model_84bf891f-a23e-4ced-8253-f04a41a8003b.tar.gz\n",
      "2021-09-18 16:36:51,760 Attempting to load featurized dataset\n",
      "2021-09-18 16:36:51,768 Exception when trying to load featurized data:\n",
      "DynamicFeaturization doesn't support get_featurized_dset_name()\n",
      "2021-09-18 16:36:51,770 Featurized dataset not previously saved for dataset H1_std, creating new\n",
      "2021-09-18 16:36:51,782 Featurizing sample 0\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "rf_estimators: 64, rf_max_depth: 64, rf_max_features: 64, valid_r2: 0.4622962571208523\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "2021-09-18 16:36:56,829 Featurizing sample 1000\n",
      "2021-09-18 16:36:58,722 Previous dataset split restored\n",
      "2021-09-18 16:36:58,738 Wrote transformers to ampl_output/H1_std/RF_ecfp_scaffold_regression/002eedd3-9398-4cca-946c-c086c6020886/transformers.pkl\n",
      "2021-09-18 16:36:58,740 Transforming response data\n",
      "2021-09-18 16:36:58,768 Transforming response data\n",
      "2021-09-18 16:36:58,783 Transforming response data\n",
      "2021-09-18 16:36:58,812 Fitting random forest model\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "number of features: 1024\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "2021-09-18 16:37:00,099 Fold 0: training r2_score = 0.965, validation r2_score = 0.453, test r2_score = 0.526\n",
      "2021-09-18 16:37:00,402 Wrote model tarball to ampl_output/H1_std_model_002eedd3-9398-4cca-946c-c086c6020886.tar.gz\n",
      "2021-09-18 16:37:00,415 Attempting to load featurized dataset\n",
      "2021-09-18 16:37:00,424 Exception when trying to load featurized data:\n",
      "DynamicFeaturization doesn't support get_featurized_dset_name()\n",
      "2021-09-18 16:37:00,425 Featurized dataset not previously saved for dataset H1_std, creating new\n",
      "2021-09-18 16:37:00,436 Featurizing sample 0\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "rf_estimators: 64, rf_max_depth: 64, rf_max_features: 64, valid_r2: 0.45343288994104436\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "2021-09-18 16:37:05,361 Featurizing sample 1000\n",
      "2021-09-18 16:37:07,298 Previous dataset split restored\n",
      "2021-09-18 16:37:07,315 Wrote transformers to ampl_output/H1_std/RF_ecfp_scaffold_regression/4e657128-a165-4bf9-bfc7-7826fccf0d3a/transformers.pkl\n",
      "2021-09-18 16:37:07,317 Transforming response data\n",
      "2021-09-18 16:37:07,346 Transforming response data\n",
      "2021-09-18 16:37:07,360 Transforming response data\n",
      "2021-09-18 16:37:07,390 Fitting random forest model\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "number of features: 1024\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "2021-09-18 16:37:08,592 Fold 0: training r2_score = 0.965, validation r2_score = 0.452, test r2_score = 0.547\n",
      "2021-09-18 16:37:08,957 Wrote model tarball to ampl_output/H1_std_model_4e657128-a165-4bf9-bfc7-7826fccf0d3a.tar.gz\n",
      "2021-09-18 16:37:08,970 Attempting to load featurized dataset\n",
      "2021-09-18 16:37:08,979 Exception when trying to load featurized data:\n",
      "DynamicFeaturization doesn't support get_featurized_dset_name()\n",
      "2021-09-18 16:37:08,980 Featurized dataset not previously saved for dataset H1_std, creating new\n",
      "2021-09-18 16:37:08,991 Featurizing sample 0\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "rf_estimators: 64, rf_max_depth: 64, rf_max_features: 64, valid_r2: 0.45200906915440886\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "2021-09-18 16:37:14,092 Featurizing sample 1000\n",
      "2021-09-18 16:37:15,998 Previous dataset split restored\n",
      "2021-09-18 16:37:16,015 Wrote transformers to ampl_output/H1_std/RF_ecfp_scaffold_regression/0448199f-979d-4989-ae28-1d673ae8bc46/transformers.pkl\n",
      "2021-09-18 16:37:16,016 Transforming response data\n",
      "2021-09-18 16:37:16,041 Transforming response data\n",
      "2021-09-18 16:37:16,056 Transforming response data\n",
      "2021-09-18 16:37:16,085 Fitting random forest model\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "number of features: 1024\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "2021-09-18 16:37:17,809 Fold 0: training r2_score = 0.967, validation r2_score = 0.443, test r2_score = 0.544\n",
      "2021-09-18 16:37:18,376 Wrote model tarball to ampl_output/H1_std_model_0448199f-979d-4989-ae28-1d673ae8bc46.tar.gz\n",
      "2021-09-18 16:37:18,390 Attempting to load featurized dataset\n",
      "2021-09-18 16:37:18,398 Exception when trying to load featurized data:\n",
      "DynamicFeaturization doesn't support get_featurized_dset_name()\n",
      "2021-09-18 16:37:18,400 Featurized dataset not previously saved for dataset H1_std, creating new\n",
      "2021-09-18 16:37:18,410 Featurizing sample 0\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "rf_estimators: 64, rf_max_depth: 64, rf_max_features: 64, valid_r2: 0.4428802685208172\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "2021-09-18 16:37:23,351 Featurizing sample 1000\n",
      "2021-09-18 16:37:25,243 Previous dataset split restored\n",
      "2021-09-18 16:37:25,260 Wrote transformers to ampl_output/H1_std/RF_ecfp_scaffold_regression/2c7f8bb7-e68c-4f01-9bf2-e59d7458644b/transformers.pkl\n",
      "2021-09-18 16:37:25,261 Transforming response data\n",
      "2021-09-18 16:37:25,287 Transforming response data\n",
      "2021-09-18 16:37:25,302 Transforming response data\n",
      "2021-09-18 16:37:25,333 Fitting random forest model\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "number of features: 1024\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "2021-09-18 16:37:26,770 Fold 0: training r2_score = 0.966, validation r2_score = 0.445, test r2_score = 0.531\n",
      "2021-09-18 16:37:27,452 Wrote model tarball to ampl_output/H1_std_model_2c7f8bb7-e68c-4f01-9bf2-e59d7458644b.tar.gz\n",
      "2021-09-18 16:37:27,469 Attempting to load featurized dataset\n",
      "2021-09-18 16:37:27,478 Exception when trying to load featurized data:\n",
      "DynamicFeaturization doesn't support get_featurized_dset_name()\n",
      "2021-09-18 16:37:27,479 Featurized dataset not previously saved for dataset H1_std, creating new\n",
      "2021-09-18 16:37:27,490 Featurizing sample 0\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "rf_estimators: 64, rf_max_depth: 64, rf_max_features: 64, valid_r2: 0.44531781338033827\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "2021-09-18 16:37:32,589 Featurizing sample 1000\n",
      "2021-09-18 16:37:34,511 Previous dataset split restored\n",
      "2021-09-18 16:37:34,528 Wrote transformers to ampl_output/H1_std/RF_ecfp_scaffold_regression/af00f727-0805-4d6f-a630-0ed278698e89/transformers.pkl\n",
      "2021-09-18 16:37:34,530 Transforming response data\n",
      "2021-09-18 16:37:34,555 Transforming response data\n",
      "2021-09-18 16:37:34,569 Transforming response data\n",
      "2021-09-18 16:37:34,599 Fitting random forest model\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "number of features: 1024\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "2021-09-18 16:37:35,680 Fold 0: training r2_score = 0.966, validation r2_score = 0.442, test r2_score = 0.549\n",
      "2021-09-18 16:37:35,938 Wrote model tarball to ampl_output/H1_std_model_af00f727-0805-4d6f-a630-0ed278698e89.tar.gz\n",
      "2021-09-18 16:37:35,953 Attempting to load featurized dataset\n",
      "2021-09-18 16:37:35,962 Exception when trying to load featurized data:\n",
      "DynamicFeaturization doesn't support get_featurized_dset_name()\n",
      "2021-09-18 16:37:35,963 Featurized dataset not previously saved for dataset H1_std, creating new\n",
      "2021-09-18 16:37:35,973 Featurizing sample 0\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "rf_estimators: 64, rf_max_depth: 64, rf_max_features: 64, valid_r2: 0.44184619128314084\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "2021-09-18 16:37:40,889 Featurizing sample 1000\n",
      "2021-09-18 16:37:42,783 Previous dataset split restored\n",
      "2021-09-18 16:37:42,799 Wrote transformers to ampl_output/H1_std/RF_ecfp_scaffold_regression/61e06822-08c0-4afb-b090-f794b32ad5cb/transformers.pkl\n",
      "2021-09-18 16:37:42,800 Transforming response data\n",
      "2021-09-18 16:37:42,826 Transforming response data\n",
      "2021-09-18 16:37:42,840 Transforming response data\n",
      "2021-09-18 16:37:42,869 Fitting random forest model\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "number of features: 1024\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "2021-09-18 16:37:44,205 Fold 0: training r2_score = 0.966, validation r2_score = 0.457, test r2_score = 0.534\n",
      "2021-09-18 16:37:44,862 Wrote model tarball to ampl_output/H1_std_model_61e06822-08c0-4afb-b090-f794b32ad5cb.tar.gz\n",
      "2021-09-18 16:37:44,877 Attempting to load featurized dataset\n",
      "2021-09-18 16:37:44,886 Exception when trying to load featurized data:\n",
      "DynamicFeaturization doesn't support get_featurized_dset_name()\n",
      "2021-09-18 16:37:44,887 Featurized dataset not previously saved for dataset H1_std, creating new\n",
      "2021-09-18 16:37:44,900 Featurizing sample 0\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "rf_estimators: 64, rf_max_depth: 64, rf_max_features: 64, valid_r2: 0.4568924462647712\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "2021-09-18 16:37:50,019 Featurizing sample 1000\n",
      "2021-09-18 16:37:51,969 Previous dataset split restored\n",
      "2021-09-18 16:37:51,986 Wrote transformers to ampl_output/H1_std/RF_ecfp_scaffold_regression/6d359a29-f368-44ac-8fff-9bd2ea976158/transformers.pkl\n",
      "2021-09-18 16:37:51,988 Transforming response data\n",
      "2021-09-18 16:37:52,013 Transforming response data\n",
      "2021-09-18 16:37:52,029 Transforming response data\n",
      "2021-09-18 16:37:52,059 Fitting random forest model\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "number of features: 1024\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "2021-09-18 16:37:53,587 Fold 0: training r2_score = 0.933, validation r2_score = 0.454, test r2_score = 0.532\n",
      "2021-09-18 16:37:53,965 Wrote model tarball to ampl_output/H1_std_model_6d359a29-f368-44ac-8fff-9bd2ea976158.tar.gz\n",
      "2021-09-18 16:37:53,982 Attempting to load featurized dataset\n",
      "2021-09-18 16:37:53,990 Exception when trying to load featurized data:\n",
      "DynamicFeaturization doesn't support get_featurized_dset_name()\n",
      "2021-09-18 16:37:53,992 Featurized dataset not previously saved for dataset H1_std, creating new\n",
      "2021-09-18 16:37:54,003 Featurizing sample 0\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "rf_estimators: 64, rf_max_depth: 64, rf_max_features: 64, valid_r2: 0.4539818763069313\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "2021-09-18 16:37:58,982 Featurizing sample 1000\n",
      "2021-09-18 16:38:00,919 Previous dataset split restored\n",
      "2021-09-18 16:38:00,935 Wrote transformers to ampl_output/H1_std/RF_ecfp_scaffold_regression/6deeacf6-cb89-4196-b939-03c88001c8e5/transformers.pkl\n",
      "2021-09-18 16:38:00,936 Transforming response data\n",
      "2021-09-18 16:38:00,962 Transforming response data\n",
      "2021-09-18 16:38:00,977 Transforming response data\n",
      "2021-09-18 16:38:01,006 Fitting random forest model\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "number of features: 1024\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "2021-09-18 16:38:02,938 Fold 0: training r2_score = 0.966, validation r2_score = 0.472, test r2_score = 0.534\n",
      "2021-09-18 16:38:03,541 Wrote model tarball to ampl_output/H1_std_model_6deeacf6-cb89-4196-b939-03c88001c8e5.tar.gz\n",
      "2021-09-18 16:38:03,558 Attempting to load featurized dataset\n",
      "2021-09-18 16:38:03,567 Exception when trying to load featurized data:\n",
      "DynamicFeaturization doesn't support get_featurized_dset_name()\n",
      "2021-09-18 16:38:03,568 Featurized dataset not previously saved for dataset H1_std, creating new\n",
      "2021-09-18 16:38:03,580 Featurizing sample 0\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "rf_estimators: 64, rf_max_depth: 64, rf_max_features: 64, valid_r2: 0.4719336501463919\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "2021-09-18 16:38:08,669 Featurizing sample 1000\n",
      "2021-09-18 16:38:10,644 Previous dataset split restored\n",
      "2021-09-18 16:38:10,661 Wrote transformers to ampl_output/H1_std/RF_ecfp_scaffold_regression/9bc4e486-4ab3-4bf1-9bdc-9bb66ce8d5fd/transformers.pkl\n",
      "2021-09-18 16:38:10,662 Transforming response data\n",
      "2021-09-18 16:38:10,690 Transforming response data\n",
      "2021-09-18 16:38:10,703 Transforming response data\n",
      "2021-09-18 16:38:10,733 Fitting random forest model\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "number of features: 1024\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "2021-09-18 16:38:11,754 Fold 0: training r2_score = 0.964, validation r2_score = 0.408, test r2_score = 0.501\n",
      "2021-09-18 16:38:12,252 Wrote model tarball to ampl_output/H1_std_model_9bc4e486-4ab3-4bf1-9bdc-9bb66ce8d5fd.tar.gz\n",
      "2021-09-18 16:38:12,267 Attempting to load featurized dataset\n",
      "2021-09-18 16:38:12,275 Exception when trying to load featurized data:\n",
      "DynamicFeaturization doesn't support get_featurized_dset_name()\n",
      "2021-09-18 16:38:12,277 Featurized dataset not previously saved for dataset H1_std, creating new\n",
      "2021-09-18 16:38:12,287 Featurizing sample 0\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "rf_estimators: 64, rf_max_depth: 64, rf_max_features: 64, valid_r2: 0.408333826636768\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "2021-09-18 16:38:17,271 Featurizing sample 1000\n",
      "2021-09-18 16:38:19,185 Previous dataset split restored\n",
      "2021-09-18 16:38:19,203 Wrote transformers to ampl_output/H1_std/RF_ecfp_scaffold_regression/75bb1153-fa58-4a3b-941e-126e93454a30/transformers.pkl\n",
      "2021-09-18 16:38:19,204 Transforming response data\n",
      "2021-09-18 16:38:19,228 Transforming response data\n",
      "2021-09-18 16:38:19,241 Transforming response data\n",
      "2021-09-18 16:38:19,270 Fitting random forest model\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "number of features: 1024\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "2021-09-18 16:38:20,799 Fold 0: training r2_score = 0.966, validation r2_score = 0.454, test r2_score = 0.535\n",
      "2021-09-18 16:38:21,415 Wrote model tarball to ampl_output/H1_std_model_75bb1153-fa58-4a3b-941e-126e93454a30.tar.gz\n",
      "2021-09-18 16:38:21,431 Attempting to load featurized dataset\n",
      "2021-09-18 16:38:21,440 Exception when trying to load featurized data:\n",
      "DynamicFeaturization doesn't support get_featurized_dset_name()\n",
      "2021-09-18 16:38:21,441 Featurized dataset not previously saved for dataset H1_std, creating new\n",
      "2021-09-18 16:38:21,452 Featurizing sample 0\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "rf_estimators: 64, rf_max_depth: 64, rf_max_features: 64, valid_r2: 0.45353626979125294\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "2021-09-18 16:38:26,508 Featurizing sample 1000\n",
      "2021-09-18 16:38:28,458 Previous dataset split restored\n",
      "2021-09-18 16:38:28,475 Wrote transformers to ampl_output/H1_std/RF_ecfp_scaffold_regression/b2e3e63b-8802-4ceb-881b-5387e0f2c292/transformers.pkl\n",
      "2021-09-18 16:38:28,477 Transforming response data\n",
      "2021-09-18 16:38:28,502 Transforming response data\n",
      "2021-09-18 16:38:28,516 Transforming response data\n",
      "2021-09-18 16:38:28,546 Fitting random forest model\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "number of features: 1024\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "2021-09-18 16:38:29,310 Fold 0: training r2_score = 0.965, validation r2_score = 0.432, test r2_score = 0.519\n",
      "2021-09-18 16:38:29,450 Wrote model tarball to ampl_output/H1_std_model_b2e3e63b-8802-4ceb-881b-5387e0f2c292.tar.gz\n",
      "2021-09-18 16:38:29,463 Attempting to load featurized dataset\n",
      "2021-09-18 16:38:29,472 Exception when trying to load featurized data:\n",
      "DynamicFeaturization doesn't support get_featurized_dset_name()\n",
      "2021-09-18 16:38:29,473 Featurized dataset not previously saved for dataset H1_std, creating new\n",
      "2021-09-18 16:38:29,484 Featurizing sample 0\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "rf_estimators: 64, rf_max_depth: 64, rf_max_features: 64, valid_r2: 0.43229624241133613\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "2021-09-18 16:38:34,505 Featurizing sample 1000\n",
      "2021-09-18 16:38:36,430 Previous dataset split restored\n",
      "2021-09-18 16:38:36,446 Wrote transformers to ampl_output/H1_std/RF_ecfp_scaffold_regression/a5c2efb1-e710-4baf-b4bb-04a0a013e85c/transformers.pkl\n",
      "2021-09-18 16:38:36,448 Transforming response data\n",
      "2021-09-18 16:38:36,472 Transforming response data\n",
      "2021-09-18 16:38:36,486 Transforming response data\n",
      "2021-09-18 16:38:36,514 Fitting random forest model\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "number of features: 1024\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "2021-09-18 16:38:37,509 Fold 0: training r2_score = 0.965, validation r2_score = 0.450, test r2_score = 0.544\n",
      "2021-09-18 16:38:37,883 Wrote model tarball to ampl_output/H1_std_model_a5c2efb1-e710-4baf-b4bb-04a0a013e85c.tar.gz\n",
      "2021-09-18 16:38:37,898 Attempting to load featurized dataset\n",
      "2021-09-18 16:38:37,907 Exception when trying to load featurized data:\n",
      "DynamicFeaturization doesn't support get_featurized_dset_name()\n",
      "2021-09-18 16:38:37,909 Featurized dataset not previously saved for dataset H1_std, creating new\n",
      "2021-09-18 16:38:37,919 Featurizing sample 0\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "rf_estimators: 64, rf_max_depth: 64, rf_max_features: 64, valid_r2: 0.4504274186255658\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "2021-09-18 16:38:42,910 Featurizing sample 1000\n",
      "2021-09-18 16:38:44,865 Previous dataset split restored\n",
      "2021-09-18 16:38:44,883 Wrote transformers to ampl_output/H1_std/RF_ecfp_scaffold_regression/1b069ec3-070c-46a6-bfff-1f489cbfb776/transformers.pkl\n",
      "2021-09-18 16:38:44,884 Transforming response data\n",
      "2021-09-18 16:38:44,912 Transforming response data\n",
      "2021-09-18 16:38:44,926 Transforming response data\n",
      "2021-09-18 16:38:44,955 Fitting random forest model\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "number of features: 1024\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "2021-09-18 16:38:46,493 Fold 0: training r2_score = 0.966, validation r2_score = 0.441, test r2_score = 0.527\n",
      "2021-09-18 16:38:47,155 Wrote model tarball to ampl_output/H1_std_model_1b069ec3-070c-46a6-bfff-1f489cbfb776.tar.gz\n",
      "2021-09-18 16:38:47,173 Attempting to load featurized dataset\n",
      "2021-09-18 16:38:47,183 Exception when trying to load featurized data:\n",
      "DynamicFeaturization doesn't support get_featurized_dset_name()\n",
      "2021-09-18 16:38:47,185 Featurized dataset not previously saved for dataset H1_std, creating new\n",
      "2021-09-18 16:38:47,196 Featurizing sample 0\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "rf_estimators: 64, rf_max_depth: 64, rf_max_features: 64, valid_r2: 0.4412406871638288\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "2021-09-18 16:38:52,232 Featurizing sample 1000\n",
      "2021-09-18 16:38:54,145 Previous dataset split restored\n",
      "2021-09-18 16:38:54,162 Wrote transformers to ampl_output/H1_std/RF_ecfp_scaffold_regression/2195a41b-a49e-43bc-b921-4a4f880f9e1f/transformers.pkl\n",
      "2021-09-18 16:38:54,164 Transforming response data\n",
      "2021-09-18 16:38:54,189 Transforming response data\n",
      "2021-09-18 16:38:54,203 Transforming response data\n",
      "2021-09-18 16:38:54,232 Fitting random forest model\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "number of features: 1024\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "2021-09-18 16:38:55,208 Fold 0: training r2_score = 0.965, validation r2_score = 0.452, test r2_score = 0.520\n",
      "2021-09-18 16:38:55,429 Wrote model tarball to ampl_output/H1_std_model_2195a41b-a49e-43bc-b921-4a4f880f9e1f.tar.gz\n",
      "2021-09-18 16:38:55,445 Attempting to load featurized dataset\n",
      "2021-09-18 16:38:55,453 Exception when trying to load featurized data:\n",
      "DynamicFeaturization doesn't support get_featurized_dset_name()\n",
      "2021-09-18 16:38:55,455 Featurized dataset not previously saved for dataset H1_std, creating new\n",
      "2021-09-18 16:38:55,469 Featurizing sample 0\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "rf_estimators: 64, rf_max_depth: 64, rf_max_features: 64, valid_r2: 0.4522315332543897\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "2021-09-18 16:39:00,444 Featurizing sample 1000\n",
      "2021-09-18 16:39:02,388 Previous dataset split restored\n",
      "2021-09-18 16:39:02,405 Wrote transformers to ampl_output/H1_std/RF_ecfp_scaffold_regression/d92a1f56-1794-4f71-96c5-cc04f7a45916/transformers.pkl\n",
      "2021-09-18 16:39:02,407 Transforming response data\n",
      "2021-09-18 16:39:02,431 Transforming response data\n",
      "2021-09-18 16:39:02,444 Transforming response data\n",
      "2021-09-18 16:39:02,475 Fitting random forest model\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "number of features: 1024\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "2021-09-18 16:39:04,192 Fold 0: training r2_score = 0.966, validation r2_score = 0.479, test r2_score = 0.527\n",
      "2021-09-18 16:39:04,667 Wrote model tarball to ampl_output/H1_std_model_d92a1f56-1794-4f71-96c5-cc04f7a45916.tar.gz\n",
      "2021-09-18 16:39:04,685 Attempting to load featurized dataset\n",
      "2021-09-18 16:39:04,695 Exception when trying to load featurized data:\n",
      "DynamicFeaturization doesn't support get_featurized_dset_name()\n",
      "2021-09-18 16:39:04,697 Featurized dataset not previously saved for dataset H1_std, creating new\n",
      "2021-09-18 16:39:04,709 Featurizing sample 0\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "rf_estimators: 64, rf_max_depth: 64, rf_max_features: 64, valid_r2: 0.478798391957182\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "2021-09-18 16:39:09,893 Featurizing sample 1000\n",
      "2021-09-18 16:39:11,877 Previous dataset split restored\n",
      "2021-09-18 16:39:11,894 Wrote transformers to ampl_output/H1_std/RF_ecfp_scaffold_regression/41940eda-f10a-47ff-acc3-039a18a6dd25/transformers.pkl\n",
      "2021-09-18 16:39:11,895 Transforming response data\n",
      "2021-09-18 16:39:11,921 Transforming response data\n",
      "2021-09-18 16:39:11,935 Transforming response data\n",
      "2021-09-18 16:39:11,965 Fitting random forest model\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "number of features: 1024\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "2021-09-18 16:39:13,993 Fold 0: training r2_score = 0.966, validation r2_score = 0.459, test r2_score = 0.540\n",
      "2021-09-18 16:39:14,538 Wrote model tarball to ampl_output/H1_std_model_41940eda-f10a-47ff-acc3-039a18a6dd25.tar.gz\n",
      "2021-09-18 16:39:14,552 Attempting to load featurized dataset\n",
      "2021-09-18 16:39:14,561 Exception when trying to load featurized data:\n",
      "DynamicFeaturization doesn't support get_featurized_dset_name()\n",
      "2021-09-18 16:39:14,562 Featurized dataset not previously saved for dataset H1_std, creating new\n",
      "2021-09-18 16:39:14,573 Featurizing sample 0\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "rf_estimators: 64, rf_max_depth: 64, rf_max_features: 64, valid_r2: 0.4594117437165699\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "2021-09-18 16:39:19,491 Featurizing sample 1000\n",
      "2021-09-18 16:39:21,385 Previous dataset split restored\n",
      "2021-09-18 16:39:21,402 Wrote transformers to ampl_output/H1_std/RF_ecfp_scaffold_regression/774b7429-74af-4a94-ad1d-1306dbfe23e8/transformers.pkl\n",
      "2021-09-18 16:39:21,404 Transforming response data\n",
      "2021-09-18 16:39:21,430 Transforming response data\n",
      "2021-09-18 16:39:21,444 Transforming response data\n",
      "2021-09-18 16:39:21,473 Fitting random forest model\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "number of features: 1024\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "2021-09-18 16:39:22,239 Fold 0: training r2_score = 0.948, validation r2_score = 0.457, test r2_score = 0.517\n",
      "2021-09-18 16:39:22,350 Wrote model tarball to ampl_output/H1_std_model_774b7429-74af-4a94-ad1d-1306dbfe23e8.tar.gz\n",
      "2021-09-18 16:39:22,363 Attempting to load featurized dataset\n",
      "2021-09-18 16:39:22,372 Exception when trying to load featurized data:\n",
      "DynamicFeaturization doesn't support get_featurized_dset_name()\n",
      "2021-09-18 16:39:22,374 Featurized dataset not previously saved for dataset H1_std, creating new\n",
      "2021-09-18 16:39:22,385 Featurizing sample 0\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "rf_estimators: 64, rf_max_depth: 64, rf_max_features: 64, valid_r2: 0.457005529232962\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "2021-09-18 16:39:27,445 Featurizing sample 1000\n",
      "2021-09-18 16:39:29,395 Previous dataset split restored\n",
      "2021-09-18 16:39:29,414 Wrote transformers to ampl_output/H1_std/RF_ecfp_scaffold_regression/20fba382-14b8-4f3d-b713-4926a4c05a2c/transformers.pkl\n",
      "2021-09-18 16:39:29,415 Transforming response data\n",
      "2021-09-18 16:39:29,442 Transforming response data\n",
      "2021-09-18 16:39:29,455 Transforming response data\n",
      "2021-09-18 16:39:29,487 Fitting random forest model\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "number of features: 1024\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "2021-09-18 16:39:31,419 Fold 0: training r2_score = 0.966, validation r2_score = 0.457, test r2_score = 0.517\n",
      "2021-09-18 16:39:32,006 Wrote model tarball to ampl_output/H1_std_model_20fba382-14b8-4f3d-b713-4926a4c05a2c.tar.gz\n",
      "2021-09-18 16:39:32,021 Attempting to load featurized dataset\n",
      "2021-09-18 16:39:32,031 Exception when trying to load featurized data:\n",
      "DynamicFeaturization doesn't support get_featurized_dset_name()\n",
      "2021-09-18 16:39:32,032 Featurized dataset not previously saved for dataset H1_std, creating new\n",
      "2021-09-18 16:39:32,043 Featurizing sample 0\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "rf_estimators: 64, rf_max_depth: 64, rf_max_features: 64, valid_r2: 0.4565611691529917\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "2021-09-18 16:39:37,007 Featurizing sample 1000\n",
      "2021-09-18 16:39:38,933 Previous dataset split restored\n",
      "2021-09-18 16:39:38,949 Wrote transformers to ampl_output/H1_std/RF_ecfp_scaffold_regression/aeea8ffe-e551-499b-b82b-4664b912edd2/transformers.pkl\n",
      "2021-09-18 16:39:38,951 Transforming response data\n",
      "2021-09-18 16:39:38,976 Transforming response data\n",
      "2021-09-18 16:39:38,991 Transforming response data\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "number of features: 1024\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "2021-09-18 16:39:39,027 Fitting random forest model\n",
      "2021-09-18 16:39:39,710 Fold 0: training r2_score = 0.964, validation r2_score = 0.402, test r2_score = 0.497\n",
      "2021-09-18 16:39:39,949 Wrote model tarball to ampl_output/H1_std_model_aeea8ffe-e551-499b-b82b-4664b912edd2.tar.gz\n",
      "2021-09-18 16:39:39,964 Attempting to load featurized dataset\n",
      "2021-09-18 16:39:39,974 Exception when trying to load featurized data:\n",
      "DynamicFeaturization doesn't support get_featurized_dset_name()\n",
      "2021-09-18 16:39:39,975 Featurized dataset not previously saved for dataset H1_std, creating new\n",
      "2021-09-18 16:39:39,987 Featurizing sample 0\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "rf_estimators: 64, rf_max_depth: 64, rf_max_features: 64, valid_r2: 0.40170731713858765\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "2021-09-18 16:39:45,023 Featurizing sample 1000\n",
      "2021-09-18 16:39:46,993 Previous dataset split restored\n",
      "2021-09-18 16:39:47,009 Wrote transformers to ampl_output/H1_std/RF_ecfp_scaffold_regression/3321d4c9-3499-43a4-b43d-c1744e235708/transformers.pkl\n",
      "2021-09-18 16:39:47,011 Transforming response data\n",
      "2021-09-18 16:39:47,041 Transforming response data\n",
      "2021-09-18 16:39:47,055 Transforming response data\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "number of features: 1024\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "2021-09-18 16:39:47,087 Fitting random forest model\n",
      "2021-09-18 16:39:48,534 Fold 0: training r2_score = 0.966, validation r2_score = 0.434, test r2_score = 0.521\n",
      "2021-09-18 16:39:49,281 Wrote model tarball to ampl_output/H1_std_model_3321d4c9-3499-43a4-b43d-c1744e235708.tar.gz\n",
      "2021-09-18 16:39:49,298 Attempting to load featurized dataset\n",
      "2021-09-18 16:39:49,308 Exception when trying to load featurized data:\n",
      "DynamicFeaturization doesn't support get_featurized_dset_name()\n",
      "2021-09-18 16:39:49,310 Featurized dataset not previously saved for dataset H1_std, creating new\n",
      "2021-09-18 16:39:49,321 Featurizing sample 0\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "rf_estimators: 64, rf_max_depth: 64, rf_max_features: 64, valid_r2: 0.4336571978239615\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "2021-09-18 16:39:54,286 Featurizing sample 1000\n",
      "2021-09-18 16:39:56,199 Previous dataset split restored\n",
      "2021-09-18 16:39:56,217 Wrote transformers to ampl_output/H1_std/RF_ecfp_scaffold_regression/d0fa7042-a6c3-4d2f-aec4-c3ec81b2c064/transformers.pkl\n",
      "2021-09-18 16:39:56,219 Transforming response data\n",
      "2021-09-18 16:39:56,246 Transforming response data\n",
      "2021-09-18 16:39:56,260 Transforming response data\n",
      "2021-09-18 16:39:56,289 Fitting random forest model\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "number of features: 1024\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "2021-09-18 16:39:57,802 Fold 0: training r2_score = 0.953, validation r2_score = 0.467, test r2_score = 0.537\n",
      "2021-09-18 16:39:58,189 Wrote model tarball to ampl_output/H1_std_model_d0fa7042-a6c3-4d2f-aec4-c3ec81b2c064.tar.gz\n",
      "2021-09-18 16:39:58,205 Attempting to load featurized dataset\n",
      "2021-09-18 16:39:58,214 Exception when trying to load featurized data:\n",
      "DynamicFeaturization doesn't support get_featurized_dset_name()\n",
      "2021-09-18 16:39:58,216 Featurized dataset not previously saved for dataset H1_std, creating new\n",
      "2021-09-18 16:39:58,228 Featurizing sample 0\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "rf_estimators: 64, rf_max_depth: 64, rf_max_features: 64, valid_r2: 0.4665136396133862\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "2021-09-18 16:40:03,309 Featurizing sample 1000\n",
      "2021-09-18 16:40:05,275 Previous dataset split restored\n",
      "2021-09-18 16:40:05,291 Wrote transformers to ampl_output/H1_std/RF_ecfp_scaffold_regression/3e89e968-3046-40ec-91f8-f0f893e3d6bf/transformers.pkl\n",
      "2021-09-18 16:40:05,293 Transforming response data\n",
      "2021-09-18 16:40:05,321 Transforming response data\n",
      "2021-09-18 16:40:05,335 Transforming response data\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "number of features: 1024\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "2021-09-18 16:40:05,369 Fitting random forest model\n",
      "2021-09-18 16:40:06,469 Fold 0: training r2_score = 0.966, validation r2_score = 0.463, test r2_score = 0.550\n",
      "2021-09-18 16:40:06,855 Wrote model tarball to ampl_output/H1_std_model_3e89e968-3046-40ec-91f8-f0f893e3d6bf.tar.gz\n",
      "2021-09-18 16:40:06,871 Attempting to load featurized dataset\n",
      "2021-09-18 16:40:06,880 Exception when trying to load featurized data:\n",
      "DynamicFeaturization doesn't support get_featurized_dset_name()\n",
      "2021-09-18 16:40:06,882 Featurized dataset not previously saved for dataset H1_std, creating new\n",
      "2021-09-18 16:40:06,896 Featurizing sample 0\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "rf_estimators: 64, rf_max_depth: 64, rf_max_features: 64, valid_r2: 0.4631862416461936\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "2021-09-18 16:40:11,963 Featurizing sample 1000\n",
      "2021-09-18 16:40:13,884 Previous dataset split restored\n",
      "2021-09-18 16:40:13,900 Wrote transformers to ampl_output/H1_std/RF_ecfp_scaffold_regression/18efa498-47ad-410c-8055-7838f61d290d/transformers.pkl\n",
      "2021-09-18 16:40:13,902 Transforming response data\n",
      "2021-09-18 16:40:13,928 Transforming response data\n",
      "2021-09-18 16:40:13,941 Transforming response data\n",
      "2021-09-18 16:40:13,969 Fitting random forest model\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "number of features: 1024\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "2021-09-18 16:40:15,388 Fold 0: training r2_score = 0.966, validation r2_score = 0.471, test r2_score = 0.537\n",
      "2021-09-18 16:40:15,932 Wrote model tarball to ampl_output/H1_std_model_18efa498-47ad-410c-8055-7838f61d290d.tar.gz\n",
      "2021-09-18 16:40:15,949 Attempting to load featurized dataset\n",
      "2021-09-18 16:40:15,958 Exception when trying to load featurized data:\n",
      "DynamicFeaturization doesn't support get_featurized_dset_name()\n",
      "2021-09-18 16:40:15,961 Featurized dataset not previously saved for dataset H1_std, creating new\n",
      "2021-09-18 16:40:15,972 Featurizing sample 0\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "rf_estimators: 64, rf_max_depth: 64, rf_max_features: 64, valid_r2: 0.47070804157099244\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "2021-09-18 16:40:20,965 Featurizing sample 1000\n",
      "2021-09-18 16:40:22,898 Previous dataset split restored\n",
      "2021-09-18 16:40:22,917 Wrote transformers to ampl_output/H1_std/RF_ecfp_scaffold_regression/fb29c040-bd92-4e5a-8eb4-4bea29e6a5c6/transformers.pkl\n",
      "2021-09-18 16:40:22,918 Transforming response data\n",
      "2021-09-18 16:40:22,944 Transforming response data\n",
      "2021-09-18 16:40:22,958 Transforming response data\n",
      "2021-09-18 16:40:22,988 Fitting random forest model\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "number of features: 1024\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "2021-09-18 16:40:24,328 Fold 0: training r2_score = 0.966, validation r2_score = 0.431, test r2_score = 0.510\n",
      "2021-09-18 16:40:24,991 Wrote model tarball to ampl_output/H1_std_model_fb29c040-bd92-4e5a-8eb4-4bea29e6a5c6.tar.gz\n",
      "2021-09-18 16:40:25,007 Attempting to load featurized dataset\n",
      "2021-09-18 16:40:25,017 Exception when trying to load featurized data:\n",
      "DynamicFeaturization doesn't support get_featurized_dset_name()\n",
      "2021-09-18 16:40:25,019 Featurized dataset not previously saved for dataset H1_std, creating new\n",
      "2021-09-18 16:40:25,030 Featurizing sample 0\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "rf_estimators: 64, rf_max_depth: 64, rf_max_features: 64, valid_r2: 0.4312971909555491\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "2021-09-18 16:40:30,049 Featurizing sample 1000\n",
      "2021-09-18 16:40:32,002 Previous dataset split restored\n",
      "2021-09-18 16:40:32,020 Wrote transformers to ampl_output/H1_std/RF_ecfp_scaffold_regression/b44c1c1a-f138-4349-acaf-4e27c98c2007/transformers.pkl\n",
      "2021-09-18 16:40:32,022 Transforming response data\n",
      "2021-09-18 16:40:32,046 Transforming response data\n",
      "2021-09-18 16:40:32,062 Transforming response data\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "number of features: 1024\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "2021-09-18 16:40:32,093 Fitting random forest model\n",
      "2021-09-18 16:40:33,097 Fold 0: training r2_score = 0.966, validation r2_score = 0.419, test r2_score = 0.521\n",
      "2021-09-18 16:40:33,529 Wrote model tarball to ampl_output/H1_std_model_b44c1c1a-f138-4349-acaf-4e27c98c2007.tar.gz\n",
      "2021-09-18 16:40:33,546 Attempting to load featurized dataset\n",
      "2021-09-18 16:40:33,555 Exception when trying to load featurized data:\n",
      "DynamicFeaturization doesn't support get_featurized_dset_name()\n",
      "2021-09-18 16:40:33,556 Featurized dataset not previously saved for dataset H1_std, creating new\n",
      "2021-09-18 16:40:33,568 Featurizing sample 0\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "rf_estimators: 64, rf_max_depth: 64, rf_max_features: 64, valid_r2: 0.41870600162256955\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "2021-09-18 16:40:38,559 Featurizing sample 1000\n",
      "2021-09-18 16:40:40,525 Previous dataset split restored\n",
      "2021-09-18 16:40:40,544 Wrote transformers to ampl_output/H1_std/RF_ecfp_scaffold_regression/a9558475-eaaa-42f7-b1ef-ba02f557c24b/transformers.pkl\n",
      "2021-09-18 16:40:40,546 Transforming response data\n",
      "2021-09-18 16:40:40,573 Transforming response data\n",
      "2021-09-18 16:40:40,588 Transforming response data\n",
      "2021-09-18 16:40:40,620 Fitting random forest model\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "number of features: 1024\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "2021-09-18 16:40:41,287 Fold 0: training r2_score = 0.965, validation r2_score = 0.428, test r2_score = 0.538\n",
      "2021-09-18 16:40:41,429 Wrote model tarball to ampl_output/H1_std_model_a9558475-eaaa-42f7-b1ef-ba02f557c24b.tar.gz\n",
      "2021-09-18 16:40:41,445 Attempting to load featurized dataset\n",
      "2021-09-18 16:40:41,454 Exception when trying to load featurized data:\n",
      "DynamicFeaturization doesn't support get_featurized_dset_name()\n",
      "2021-09-18 16:40:41,455 Featurized dataset not previously saved for dataset H1_std, creating new\n",
      "2021-09-18 16:40:41,469 Featurizing sample 0\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "rf_estimators: 64, rf_max_depth: 64, rf_max_features: 64, valid_r2: 0.4284823074147378\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "2021-09-18 16:40:46,616 Featurizing sample 1000\n",
      "2021-09-18 16:40:48,558 Previous dataset split restored\n",
      "2021-09-18 16:40:48,575 Wrote transformers to ampl_output/H1_std/RF_ecfp_scaffold_regression/c52dd8bc-e485-4c07-9c1c-f7851d578023/transformers.pkl\n",
      "2021-09-18 16:40:48,577 Transforming response data\n",
      "2021-09-18 16:40:48,601 Transforming response data\n",
      "2021-09-18 16:40:48,615 Transforming response data\n",
      "2021-09-18 16:40:48,645 Fitting random forest model\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "number of features: 1024\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "2021-09-18 16:40:50,585 Fold 0: training r2_score = 0.966, validation r2_score = 0.475, test r2_score = 0.527\n",
      "2021-09-18 16:40:51,230 Wrote model tarball to ampl_output/H1_std_model_c52dd8bc-e485-4c07-9c1c-f7851d578023.tar.gz\n",
      "2021-09-18 16:40:51,246 Attempting to load featurized dataset\n",
      "2021-09-18 16:40:51,255 Exception when trying to load featurized data:\n",
      "DynamicFeaturization doesn't support get_featurized_dset_name()\n",
      "2021-09-18 16:40:51,256 Featurized dataset not previously saved for dataset H1_std, creating new\n",
      "2021-09-18 16:40:51,268 Featurizing sample 0\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "rf_estimators: 64, rf_max_depth: 64, rf_max_features: 64, valid_r2: 0.4754599390049119\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "2021-09-18 16:40:56,262 Featurizing sample 1000\n",
      "2021-09-18 16:40:58,181 Previous dataset split restored\n",
      "2021-09-18 16:40:58,199 Wrote transformers to ampl_output/H1_std/RF_ecfp_scaffold_regression/b2ff3097-6848-41e4-a149-fde4d1aead02/transformers.pkl\n",
      "2021-09-18 16:40:58,200 Transforming response data\n",
      "2021-09-18 16:40:58,227 Transforming response data\n",
      "2021-09-18 16:40:58,241 Transforming response data\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "number of features: 1024\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "2021-09-18 16:40:58,275 Fitting random forest model\n",
      "2021-09-18 16:40:59,786 Fold 0: training r2_score = 0.966, validation r2_score = 0.470, test r2_score = 0.524\n",
      "2021-09-18 16:41:00,210 Wrote model tarball to ampl_output/H1_std_model_b2ff3097-6848-41e4-a149-fde4d1aead02.tar.gz\n",
      "2021-09-18 16:41:00,227 Attempting to load featurized dataset\n",
      "2021-09-18 16:41:00,237 Exception when trying to load featurized data:\n",
      "DynamicFeaturization doesn't support get_featurized_dset_name()\n",
      "2021-09-18 16:41:00,238 Featurized dataset not previously saved for dataset H1_std, creating new\n",
      "2021-09-18 16:41:00,250 Featurizing sample 0\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "rf_estimators: 64, rf_max_depth: 64, rf_max_features: 64, valid_r2: 0.4696404341206597\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "2021-09-18 16:41:05,337 Featurizing sample 1000\n",
      "2021-09-18 16:41:07,263 Previous dataset split restored\n",
      "2021-09-18 16:41:07,280 Wrote transformers to ampl_output/H1_std/RF_ecfp_scaffold_regression/1eb075c1-299f-4bdc-8e50-10a3dd9649b5/transformers.pkl\n",
      "2021-09-18 16:41:07,281 Transforming response data\n",
      "2021-09-18 16:41:07,308 Transforming response data\n",
      "2021-09-18 16:41:07,322 Transforming response data\n",
      "2021-09-18 16:41:07,352 Fitting random forest model\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "number of features: 1024\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "2021-09-18 16:41:08,138 Fold 0: training r2_score = 0.965, validation r2_score = 0.425, test r2_score = 0.524\n",
      "2021-09-18 16:41:08,453 Wrote model tarball to ampl_output/H1_std_model_1eb075c1-299f-4bdc-8e50-10a3dd9649b5.tar.gz\n",
      "2021-09-18 16:41:08,468 Attempting to load featurized dataset\n",
      "2021-09-18 16:41:08,475 Exception when trying to load featurized data:\n",
      "DynamicFeaturization doesn't support get_featurized_dset_name()\n",
      "2021-09-18 16:41:08,477 Featurized dataset not previously saved for dataset H1_std, creating new\n",
      "2021-09-18 16:41:08,490 Featurizing sample 0\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "rf_estimators: 64, rf_max_depth: 64, rf_max_features: 64, valid_r2: 0.42470160817696134\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "2021-09-18 16:41:13,471 Featurizing sample 1000\n",
      "2021-09-18 16:41:15,400 Previous dataset split restored\n",
      "2021-09-18 16:41:15,418 Wrote transformers to ampl_output/H1_std/RF_ecfp_scaffold_regression/72ff73f4-f96c-4c38-a9a4-77f674223b74/transformers.pkl\n",
      "2021-09-18 16:41:15,419 Transforming response data\n",
      "2021-09-18 16:41:15,445 Transforming response data\n",
      "2021-09-18 16:41:15,459 Transforming response data\n",
      "2021-09-18 16:41:15,489 Fitting random forest model\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "number of features: 1024\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "2021-09-18 16:41:17,014 Fold 0: training r2_score = 0.966, validation r2_score = 0.462, test r2_score = 0.541\n",
      "2021-09-18 16:41:17,571 Wrote model tarball to ampl_output/H1_std_model_72ff73f4-f96c-4c38-a9a4-77f674223b74.tar.gz\n",
      "2021-09-18 16:41:17,587 Attempting to load featurized dataset\n",
      "2021-09-18 16:41:17,596 Exception when trying to load featurized data:\n",
      "DynamicFeaturization doesn't support get_featurized_dset_name()\n",
      "2021-09-18 16:41:17,598 Featurized dataset not previously saved for dataset H1_std, creating new\n",
      "2021-09-18 16:41:17,614 Featurizing sample 0\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "rf_estimators: 64, rf_max_depth: 64, rf_max_features: 64, valid_r2: 0.4623485060148709\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "2021-09-18 16:41:22,739 Featurizing sample 1000\n",
      "2021-09-18 16:41:24,737 Previous dataset split restored\n",
      "2021-09-18 16:41:24,753 Wrote transformers to ampl_output/H1_std/RF_ecfp_scaffold_regression/0474676d-cf0b-4304-84be-119bb276460a/transformers.pkl\n",
      "2021-09-18 16:41:24,756 Transforming response data\n",
      "2021-09-18 16:41:24,782 Transforming response data\n",
      "2021-09-18 16:41:24,797 Transforming response data\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "number of features: 1024\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "2021-09-18 16:41:24,827 Fitting random forest model\n",
      "2021-09-18 16:41:25,283 Fold 0: training r2_score = 0.909, validation r2_score = 0.379, test r2_score = 0.512\n",
      "2021-09-18 16:41:25,336 Wrote model tarball to ampl_output/H1_std_model_0474676d-cf0b-4304-84be-119bb276460a.tar.gz\n",
      "2021-09-18 16:41:25,350 Attempting to load featurized dataset\n",
      "2021-09-18 16:41:25,359 Exception when trying to load featurized data:\n",
      "DynamicFeaturization doesn't support get_featurized_dset_name()\n",
      "2021-09-18 16:41:25,362 Featurized dataset not previously saved for dataset H1_std, creating new\n",
      "2021-09-18 16:41:25,372 Featurizing sample 0\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "rf_estimators: 64, rf_max_depth: 64, rf_max_features: 64, valid_r2: 0.3788228908643606\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "2021-09-18 16:41:30,327 Featurizing sample 1000\n",
      "2021-09-18 16:41:32,237 Previous dataset split restored\n",
      "2021-09-18 16:41:32,254 Wrote transformers to ampl_output/H1_std/RF_ecfp_scaffold_regression/d8f2d5af-6439-4bce-967b-f912cb7e0442/transformers.pkl\n",
      "2021-09-18 16:41:32,256 Transforming response data\n",
      "2021-09-18 16:41:32,284 Transforming response data\n",
      "2021-09-18 16:41:32,299 Transforming response data\n",
      "2021-09-18 16:41:32,329 Fitting random forest model\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "number of features: 1024\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "2021-09-18 16:41:32,778 Fold 0: training r2_score = 0.956, validation r2_score = 0.358, test r2_score = 0.546\n",
      "2021-09-18 16:41:32,821 Wrote model tarball to ampl_output/H1_std_model_d8f2d5af-6439-4bce-967b-f912cb7e0442.tar.gz\n",
      "2021-09-18 16:41:32,835 Attempting to load featurized dataset\n",
      "2021-09-18 16:41:32,844 Exception when trying to load featurized data:\n",
      "DynamicFeaturization doesn't support get_featurized_dset_name()\n",
      "2021-09-18 16:41:32,845 Featurized dataset not previously saved for dataset H1_std, creating new\n",
      "2021-09-18 16:41:32,856 Featurizing sample 0\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "rf_estimators: 64, rf_max_depth: 64, rf_max_features: 64, valid_r2: 0.35772884023414264\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "2021-09-18 16:41:37,856 Featurizing sample 1000\n",
      "2021-09-18 16:41:39,815 Previous dataset split restored\n",
      "2021-09-18 16:41:39,833 Wrote transformers to ampl_output/H1_std/RF_ecfp_scaffold_regression/22cb67d5-6eb2-404d-ae90-6ad7be500445/transformers.pkl\n",
      "2021-09-18 16:41:39,836 Transforming response data\n",
      "2021-09-18 16:41:39,863 Transforming response data\n",
      "2021-09-18 16:41:39,877 Transforming response data\n",
      "2021-09-18 16:41:39,907 Fitting random forest model\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "number of features: 1024\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "2021-09-18 16:41:41,454 Fold 0: training r2_score = 0.967, validation r2_score = 0.445, test r2_score = 0.535\n",
      "2021-09-18 16:41:42,153 Wrote model tarball to ampl_output/H1_std_model_22cb67d5-6eb2-404d-ae90-6ad7be500445.tar.gz\n",
      "2021-09-18 16:41:42,171 Attempting to load featurized dataset\n",
      "2021-09-18 16:41:42,180 Exception when trying to load featurized data:\n",
      "DynamicFeaturization doesn't support get_featurized_dset_name()\n",
      "2021-09-18 16:41:42,181 Featurized dataset not previously saved for dataset H1_std, creating new\n",
      "2021-09-18 16:41:42,194 Featurizing sample 0\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "rf_estimators: 64, rf_max_depth: 64, rf_max_features: 64, valid_r2: 0.4450967944532872\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "2021-09-18 16:41:47,266 Featurizing sample 1000\n",
      "2021-09-18 16:41:49,186 Previous dataset split restored\n",
      "2021-09-18 16:41:49,205 Wrote transformers to ampl_output/H1_std/RF_ecfp_scaffold_regression/ba222717-3c58-420f-a418-c05ffc192f49/transformers.pkl\n",
      "2021-09-18 16:41:49,206 Transforming response data\n",
      "2021-09-18 16:41:49,232 Transforming response data\n",
      "2021-09-18 16:41:49,247 Transforming response data\n",
      "2021-09-18 16:41:49,281 Fitting random forest model\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "number of features: 1024\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "2021-09-18 16:41:49,730 Fold 0: training r2_score = 0.951, validation r2_score = 0.485, test r2_score = 0.488\n",
      "2021-09-18 16:41:49,763 Wrote model tarball to ampl_output/H1_std_model_ba222717-3c58-420f-a418-c05ffc192f49.tar.gz\n",
      "2021-09-18 16:41:49,777 Attempting to load featurized dataset\n",
      "2021-09-18 16:41:49,788 Exception when trying to load featurized data:\n",
      "DynamicFeaturization doesn't support get_featurized_dset_name()\n",
      "2021-09-18 16:41:49,789 Featurized dataset not previously saved for dataset H1_std, creating new\n",
      "2021-09-18 16:41:49,800 Featurizing sample 0\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "rf_estimators: 64, rf_max_depth: 64, rf_max_features: 64, valid_r2: 0.48502680882974036\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "2021-09-18 16:41:54,833 Featurizing sample 1000\n",
      "2021-09-18 16:41:56,770 Previous dataset split restored\n",
      "2021-09-18 16:41:56,788 Wrote transformers to ampl_output/H1_std/RF_ecfp_scaffold_regression/a680cb92-5f74-41d6-bc0c-0a4b83d924ac/transformers.pkl\n",
      "2021-09-18 16:41:56,790 Transforming response data\n",
      "2021-09-18 16:41:56,816 Transforming response data\n",
      "2021-09-18 16:41:56,831 Transforming response data\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "number of features: 1024\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "2021-09-18 16:41:56,869 Fitting random forest model\n",
      "2021-09-18 16:41:57,846 Fold 0: training r2_score = 0.964, validation r2_score = 0.467, test r2_score = 0.534\n",
      "2021-09-18 16:41:58,063 Wrote model tarball to ampl_output/H1_std_model_a680cb92-5f74-41d6-bc0c-0a4b83d924ac.tar.gz\n",
      "2021-09-18 16:41:58,080 Attempting to load featurized dataset\n",
      "2021-09-18 16:41:58,089 Exception when trying to load featurized data:\n",
      "DynamicFeaturization doesn't support get_featurized_dset_name()\n",
      "2021-09-18 16:41:58,091 Featurized dataset not previously saved for dataset H1_std, creating new\n",
      "2021-09-18 16:41:58,102 Featurizing sample 0\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "rf_estimators: 64, rf_max_depth: 64, rf_max_features: 64, valid_r2: 0.4666840475790186\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "2021-09-18 16:42:03,246 Featurizing sample 1000\n",
      "2021-09-18 16:42:05,187 Previous dataset split restored\n",
      "2021-09-18 16:42:05,204 Wrote transformers to ampl_output/H1_std/RF_ecfp_scaffold_regression/46431a04-1726-4374-867e-7c176a5ff498/transformers.pkl\n",
      "2021-09-18 16:42:05,206 Transforming response data\n",
      "2021-09-18 16:42:05,233 Transforming response data\n",
      "2021-09-18 16:42:05,247 Transforming response data\n",
      "2021-09-18 16:42:05,277 Fitting random forest model\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "number of features: 1024\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "2021-09-18 16:42:07,201 Fold 0: training r2_score = 0.966, validation r2_score = 0.455, test r2_score = 0.540\n",
      "2021-09-18 16:42:07,740 Wrote model tarball to ampl_output/H1_std_model_46431a04-1726-4374-867e-7c176a5ff498.tar.gz\n",
      "2021-09-18 16:42:07,755 Attempting to load featurized dataset\n",
      "2021-09-18 16:42:07,764 Exception when trying to load featurized data:\n",
      "DynamicFeaturization doesn't support get_featurized_dset_name()\n",
      "2021-09-18 16:42:07,765 Featurized dataset not previously saved for dataset H1_std, creating new\n",
      "2021-09-18 16:42:07,777 Featurizing sample 0\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "rf_estimators: 64, rf_max_depth: 64, rf_max_features: 64, valid_r2: 0.4553542283485884\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "2021-09-18 16:42:12,795 Featurizing sample 1000\n",
      "2021-09-18 16:42:14,737 Previous dataset split restored\n",
      "2021-09-18 16:42:14,756 Wrote transformers to ampl_output/H1_std/RF_ecfp_scaffold_regression/b8481739-b07d-4d20-b6ee-54f03d78f698/transformers.pkl\n",
      "2021-09-18 16:42:14,757 Transforming response data\n",
      "2021-09-18 16:42:14,785 Transforming response data\n",
      "2021-09-18 16:42:14,801 Transforming response data\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "number of features: 1024\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "2021-09-18 16:42:14,833 Fitting random forest model\n",
      "2021-09-18 16:42:16,365 Fold 0: training r2_score = 0.966, validation r2_score = 0.449, test r2_score = 0.536\n",
      "2021-09-18 16:42:16,991 Wrote model tarball to ampl_output/H1_std_model_b8481739-b07d-4d20-b6ee-54f03d78f698.tar.gz\n",
      "2021-09-18 16:42:17,009 Attempting to load featurized dataset\n",
      "2021-09-18 16:42:17,018 Exception when trying to load featurized data:\n",
      "DynamicFeaturization doesn't support get_featurized_dset_name()\n",
      "2021-09-18 16:42:17,020 Featurized dataset not previously saved for dataset H1_std, creating new\n",
      "2021-09-18 16:42:17,032 Featurizing sample 0\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "rf_estimators: 64, rf_max_depth: 64, rf_max_features: 64, valid_r2: 0.44883108991496756\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "2021-09-18 16:42:22,156 Featurizing sample 1000\n",
      "2021-09-18 16:42:24,079 Previous dataset split restored\n",
      "2021-09-18 16:42:24,097 Wrote transformers to ampl_output/H1_std/RF_ecfp_scaffold_regression/b39f5de4-eef1-462d-9526-df98449f6b3b/transformers.pkl\n",
      "2021-09-18 16:42:24,098 Transforming response data\n",
      "2021-09-18 16:42:24,124 Transforming response data\n",
      "2021-09-18 16:42:24,139 Transforming response data\n",
      "2021-09-18 16:42:24,171 Fitting random forest model\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "number of features: 1024\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "2021-09-18 16:42:26,117 Fold 0: training r2_score = 0.966, validation r2_score = 0.439, test r2_score = 0.542\n",
      "2021-09-18 16:42:26,800 Wrote model tarball to ampl_output/H1_std_model_b39f5de4-eef1-462d-9526-df98449f6b3b.tar.gz\n",
      "2021-09-18 16:42:26,818 Attempting to load featurized dataset\n",
      "2021-09-18 16:42:26,826 Exception when trying to load featurized data:\n",
      "DynamicFeaturization doesn't support get_featurized_dset_name()\n",
      "2021-09-18 16:42:26,827 Featurized dataset not previously saved for dataset H1_std, creating new\n",
      "2021-09-18 16:42:26,839 Featurizing sample 0\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "rf_estimators: 64, rf_max_depth: 64, rf_max_features: 64, valid_r2: 0.4386765005000973\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "2021-09-18 16:42:31,838 Featurizing sample 1000\n",
      "2021-09-18 16:42:33,734 Previous dataset split restored\n",
      "2021-09-18 16:42:33,752 Wrote transformers to ampl_output/H1_std/RF_ecfp_scaffold_regression/f3eaadd6-c98a-4636-9379-e88881e61ef1/transformers.pkl\n",
      "2021-09-18 16:42:33,753 Transforming response data\n",
      "2021-09-18 16:42:33,778 Transforming response data\n",
      "2021-09-18 16:42:33,793 Transforming response data\n",
      "2021-09-18 16:42:33,822 Fitting random forest model\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "number of features: 1024\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "2021-09-18 16:42:34,907 Fold 0: training r2_score = 0.965, validation r2_score = 0.473, test r2_score = 0.538\n",
      "2021-09-18 16:42:35,213 Wrote model tarball to ampl_output/H1_std_model_f3eaadd6-c98a-4636-9379-e88881e61ef1.tar.gz\n",
      "2021-09-18 16:42:35,228 Attempting to load featurized dataset\n",
      "2021-09-18 16:42:35,238 Exception when trying to load featurized data:\n",
      "DynamicFeaturization doesn't support get_featurized_dset_name()\n",
      "2021-09-18 16:42:35,239 Featurized dataset not previously saved for dataset H1_std, creating new\n",
      "2021-09-18 16:42:35,250 Featurizing sample 0\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "rf_estimators: 64, rf_max_depth: 64, rf_max_features: 64, valid_r2: 0.4726332285408146\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "2021-09-18 16:42:40,477 Featurizing sample 1000\n",
      "2021-09-18 16:42:42,499 Previous dataset split restored\n",
      "2021-09-18 16:42:42,516 Wrote transformers to ampl_output/H1_std/RF_ecfp_scaffold_regression/50c8b4e8-028d-4430-a01c-ad7c11e41c52/transformers.pkl\n",
      "2021-09-18 16:42:42,518 Transforming response data\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "number of features: 1024\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "2021-09-18 16:42:42,564 Transforming response data\n",
      "2021-09-18 16:42:42,582 Transforming response data\n",
      "2021-09-18 16:42:42,620 Fitting random forest model\n",
      "2021-09-18 16:42:44,241 Fold 0: training r2_score = 0.965, validation r2_score = 0.465, test r2_score = 0.534\n",
      "2021-09-18 16:42:44,662 Wrote model tarball to ampl_output/H1_std_model_50c8b4e8-028d-4430-a01c-ad7c11e41c52.tar.gz\n",
      "2021-09-18 16:42:44,677 Attempting to load featurized dataset\n",
      "2021-09-18 16:42:44,686 Exception when trying to load featurized data:\n",
      "DynamicFeaturization doesn't support get_featurized_dset_name()\n",
      "2021-09-18 16:42:44,687 Featurized dataset not previously saved for dataset H1_std, creating new\n",
      "2021-09-18 16:42:44,703 Featurizing sample 0\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "rf_estimators: 64, rf_max_depth: 64, rf_max_features: 64, valid_r2: 0.46532989542171677\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "2021-09-18 16:42:50,362 Featurizing sample 1000\n",
      "2021-09-18 16:42:52,351 Previous dataset split restored\n",
      "2021-09-18 16:42:52,370 Wrote transformers to ampl_output/H1_std/RF_ecfp_scaffold_regression/d0335b2e-2c40-4947-b646-26336eb99c48/transformers.pkl\n",
      "2021-09-18 16:42:52,371 Transforming response data\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "number of features: 1024\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "2021-09-18 16:42:52,411 Transforming response data\n",
      "2021-09-18 16:42:52,427 Transforming response data\n",
      "2021-09-18 16:42:52,459 Fitting random forest model\n",
      "2021-09-18 16:42:53,444 Fold 0: training r2_score = 0.966, validation r2_score = 0.465, test r2_score = 0.544\n",
      "2021-09-18 16:42:53,701 Wrote model tarball to ampl_output/H1_std_model_d0335b2e-2c40-4947-b646-26336eb99c48.tar.gz\n",
      "2021-09-18 16:42:53,718 Attempting to load featurized dataset\n",
      "2021-09-18 16:42:53,728 Exception when trying to load featurized data:\n",
      "DynamicFeaturization doesn't support get_featurized_dset_name()\n",
      "2021-09-18 16:42:53,730 Featurized dataset not previously saved for dataset H1_std, creating new\n",
      "2021-09-18 16:42:53,742 Featurizing sample 0\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "rf_estimators: 64, rf_max_depth: 64, rf_max_features: 64, valid_r2: 0.4645643374561661\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "2021-09-18 16:42:58,839 Featurizing sample 1000\n",
      "2021-09-18 16:43:00,836 Previous dataset split restored\n",
      "2021-09-18 16:43:00,851 Wrote transformers to ampl_output/H1_std/RF_ecfp_scaffold_regression/21c5b603-e3a7-4cfc-a00d-781cbde7a9fc/transformers.pkl\n",
      "2021-09-18 16:43:00,853 Transforming response data\n",
      "2021-09-18 16:43:00,892 Transforming response data\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "number of features: 1024\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "2021-09-18 16:43:00,911 Transforming response data\n",
      "2021-09-18 16:43:00,944 Fitting random forest model\n",
      "2021-09-18 16:43:01,825 Fold 0: training r2_score = 0.842, validation r2_score = 0.405, test r2_score = 0.474\n",
      "2021-09-18 16:43:01,919 Wrote model tarball to ampl_output/H1_std_model_21c5b603-e3a7-4cfc-a00d-781cbde7a9fc.tar.gz\n",
      "2021-09-18 16:43:01,934 Attempting to load featurized dataset\n",
      "2021-09-18 16:43:01,943 Exception when trying to load featurized data:\n",
      "DynamicFeaturization doesn't support get_featurized_dset_name()\n",
      "2021-09-18 16:43:01,945 Featurized dataset not previously saved for dataset H1_std, creating new\n",
      "2021-09-18 16:43:01,959 Featurizing sample 0\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "rf_estimators: 64, rf_max_depth: 64, rf_max_features: 64, valid_r2: 0.40530514499014647\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "2021-09-18 16:43:06,934 Featurizing sample 1000\n",
      "2021-09-18 16:43:08,873 Previous dataset split restored\n",
      "2021-09-18 16:43:08,889 Wrote transformers to ampl_output/H1_std/RF_ecfp_scaffold_regression/42cee1b0-46fe-423d-b6d6-38da92c97a18/transformers.pkl\n",
      "2021-09-18 16:43:08,893 Transforming response data\n",
      "2021-09-18 16:43:08,924 Transforming response data\n",
      "2021-09-18 16:43:08,939 Transforming response data\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "number of features: 1024\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "2021-09-18 16:43:08,971 Fitting random forest model\n",
      "2021-09-18 16:43:10,288 Fold 0: training r2_score = 0.966, validation r2_score = 0.439, test r2_score = 0.544\n",
      "2021-09-18 16:43:10,820 Wrote model tarball to ampl_output/H1_std_model_42cee1b0-46fe-423d-b6d6-38da92c97a18.tar.gz\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "rf_estimators: 64, rf_max_depth: 64, rf_max_features: 64, valid_r2: 0.4390132299434132\n"
     ]
    }
   ],
   "source": [
    "rfe_list = []\n",
    "rfd_list = []\n",
    "rff_list = []\n",
    "valid_r2_list = []\n",
    "\n",
    "num_trials = 50\n",
    "\n",
    "for i in range(num_trials):\n",
    "    params[\"rf_estimators\"] = np.random.randint(8,256)\n",
    "    params[\"rf_max_depth\"] = np.random.randint(8,256)\n",
    "    params[\"rf_max_features\"] = np.random.randint(8,256)\n",
    "    rfe_list.append(params[\"rf_estimators\"])\n",
    "    rfd_list.append(params[\"rf_max_depth\"])\n",
    "    rff_list.append(params[\"rf_max_features\"])\n",
    "    tp = parse.wrapper(params)\n",
    "    pl = mp.ModelPipeline(tp)\n",
    "    pl.train_model()\n",
    "    pred_data = pl.model_wrapper.get_perf_data(subset=\"valid\", epoch_label=\"best\")\n",
    "    pred_results = pred_data.get_prediction_results()\n",
    "    valid_r2 = pred_results['r2_score']\n",
    "    valid_r2_list.append(valid_r2)\n",
    "    print(f\"rf_estimators: {rf_estimator}, rf_max_depth: {rf_depth}, rf_max_features: {rf_feature}, valid_r2: {valid_r2}\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 21,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 1000
    },
    "id": "EP7Oub5L67o_",
    "outputId": "0e71c6f4-3795-474a-baa7-1f42f0ad1273"
   },
   "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>rf_estimators</th>\n",
       "      <th>rf_max_depth</th>\n",
       "      <th>rf_max_features</th>\n",
       "      <th>valid_r2_score</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>40</th>\n",
       "      <td>9</td>\n",
       "      <td>143</td>\n",
       "      <td>205</td>\n",
       "      <td>0.485027</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>21</th>\n",
       "      <td>167</td>\n",
       "      <td>154</td>\n",
       "      <td>193</td>\n",
       "      <td>0.478798</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>33</th>\n",
       "      <td>230</td>\n",
       "      <td>254</td>\n",
       "      <td>167</td>\n",
       "      <td>0.475460</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>45</th>\n",
       "      <td>104</td>\n",
       "      <td>148</td>\n",
       "      <td>168</td>\n",
       "      <td>0.472633</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>14</th>\n",
       "      <td>211</td>\n",
       "      <td>128</td>\n",
       "      <td>185</td>\n",
       "      <td>0.471934</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>29</th>\n",
       "      <td>188</td>\n",
       "      <td>249</td>\n",
       "      <td>87</td>\n",
       "      <td>0.470708</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>34</th>\n",
       "      <td>150</td>\n",
       "      <td>38</td>\n",
       "      <td>181</td>\n",
       "      <td>0.469640</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>5</th>\n",
       "      <td>207</td>\n",
       "      <td>226</td>\n",
       "      <td>108</td>\n",
       "      <td>0.468544</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>41</th>\n",
       "      <td>73</td>\n",
       "      <td>136</td>\n",
       "      <td>206</td>\n",
       "      <td>0.466684</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>27</th>\n",
       "      <td>167</td>\n",
       "      <td>19</td>\n",
       "      <td>185</td>\n",
       "      <td>0.466514</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>46</th>\n",
       "      <td>142</td>\n",
       "      <td>230</td>\n",
       "      <td>227</td>\n",
       "      <td>0.465330</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>47</th>\n",
       "      <td>84</td>\n",
       "      <td>101</td>\n",
       "      <td>134</td>\n",
       "      <td>0.464564</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>117</td>\n",
       "      <td>247</td>\n",
       "      <td>172</td>\n",
       "      <td>0.464417</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>28</th>\n",
       "      <td>133</td>\n",
       "      <td>255</td>\n",
       "      <td>73</td>\n",
       "      <td>0.463186</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>36</th>\n",
       "      <td>197</td>\n",
       "      <td>226</td>\n",
       "      <td>95</td>\n",
       "      <td>0.462349</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>6</th>\n",
       "      <td>94</td>\n",
       "      <td>25</td>\n",
       "      <td>208</td>\n",
       "      <td>0.462296</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>22</th>\n",
       "      <td>197</td>\n",
       "      <td>243</td>\n",
       "      <td>224</td>\n",
       "      <td>0.459412</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>115</td>\n",
       "      <td>136</td>\n",
       "      <td>93</td>\n",
       "      <td>0.458759</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>23</th>\n",
       "      <td>49</td>\n",
       "      <td>17</td>\n",
       "      <td>245</td>\n",
       "      <td>0.457006</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>12</th>\n",
       "      <td>223</td>\n",
       "      <td>82</td>\n",
       "      <td>40</td>\n",
       "      <td>0.456892</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>24</th>\n",
       "      <td>212</td>\n",
       "      <td>69</td>\n",
       "      <td>187</td>\n",
       "      <td>0.456561</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>42</th>\n",
       "      <td>194</td>\n",
       "      <td>252</td>\n",
       "      <td>200</td>\n",
       "      <td>0.455354</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>251</td>\n",
       "      <td>254</td>\n",
       "      <td>98</td>\n",
       "      <td>0.454410</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>13</th>\n",
       "      <td>200</td>\n",
       "      <td>15</td>\n",
       "      <td>146</td>\n",
       "      <td>0.453982</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>16</th>\n",
       "      <td>218</td>\n",
       "      <td>252</td>\n",
       "      <td>79</td>\n",
       "      <td>0.453536</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>7</th>\n",
       "      <td>108</td>\n",
       "      <td>145</td>\n",
       "      <td>212</td>\n",
       "      <td>0.453433</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>20</th>\n",
       "      <td>77</td>\n",
       "      <td>175</td>\n",
       "      <td>153</td>\n",
       "      <td>0.452232</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>8</th>\n",
       "      <td>128</td>\n",
       "      <td>159</td>\n",
       "      <td>144</td>\n",
       "      <td>0.452009</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>18</th>\n",
       "      <td>128</td>\n",
       "      <td>45</td>\n",
       "      <td>56</td>\n",
       "      <td>0.450427</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>43</th>\n",
       "      <td>215</td>\n",
       "      <td>195</td>\n",
       "      <td>81</td>\n",
       "      <td>0.448831</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>10</th>\n",
       "      <td>239</td>\n",
       "      <td>81</td>\n",
       "      <td>36</td>\n",
       "      <td>0.445318</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>39</th>\n",
       "      <td>241</td>\n",
       "      <td>77</td>\n",
       "      <td>44</td>\n",
       "      <td>0.445097</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>9</th>\n",
       "      <td>204</td>\n",
       "      <td>178</td>\n",
       "      <td>137</td>\n",
       "      <td>0.442880</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>11</th>\n",
       "      <td>91</td>\n",
       "      <td>227</td>\n",
       "      <td>223</td>\n",
       "      <td>0.441846</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>19</th>\n",
       "      <td>231</td>\n",
       "      <td>255</td>\n",
       "      <td>54</td>\n",
       "      <td>0.441241</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>125</td>\n",
       "      <td>142</td>\n",
       "      <td>55</td>\n",
       "      <td>0.439562</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>49</th>\n",
       "      <td>180</td>\n",
       "      <td>95</td>\n",
       "      <td>67</td>\n",
       "      <td>0.439013</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>44</th>\n",
       "      <td>247</td>\n",
       "      <td>87</td>\n",
       "      <td>130</td>\n",
       "      <td>0.438677</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>119</td>\n",
       "      <td>57</td>\n",
       "      <td>158</td>\n",
       "      <td>0.437628</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>26</th>\n",
       "      <td>253</td>\n",
       "      <td>159</td>\n",
       "      <td>28</td>\n",
       "      <td>0.433657</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>17</th>\n",
       "      <td>47</td>\n",
       "      <td>43</td>\n",
       "      <td>254</td>\n",
       "      <td>0.432296</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>30</th>\n",
       "      <td>230</td>\n",
       "      <td>54</td>\n",
       "      <td>33</td>\n",
       "      <td>0.431297</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>32</th>\n",
       "      <td>47</td>\n",
       "      <td>58</td>\n",
       "      <td>99</td>\n",
       "      <td>0.428482</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>35</th>\n",
       "      <td>106</td>\n",
       "      <td>201</td>\n",
       "      <td>24</td>\n",
       "      <td>0.424702</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>31</th>\n",
       "      <td>150</td>\n",
       "      <td>161</td>\n",
       "      <td>21</td>\n",
       "      <td>0.418706</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>15</th>\n",
       "      <td>170</td>\n",
       "      <td>91</td>\n",
       "      <td>9</td>\n",
       "      <td>0.408334</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>48</th>\n",
       "      <td>87</td>\n",
       "      <td>9</td>\n",
       "      <td>211</td>\n",
       "      <td>0.405305</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>25</th>\n",
       "      <td>81</td>\n",
       "      <td>204</td>\n",
       "      <td>16</td>\n",
       "      <td>0.401707</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>37</th>\n",
       "      <td>22</td>\n",
       "      <td>18</td>\n",
       "      <td>27</td>\n",
       "      <td>0.378823</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>38</th>\n",
       "      <td>12</td>\n",
       "      <td>86</td>\n",
       "      <td>161</td>\n",
       "      <td>0.357729</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "    rf_estimators  rf_max_depth  rf_max_features  valid_r2_score\n",
       "40              9           143              205        0.485027\n",
       "21            167           154              193        0.478798\n",
       "33            230           254              167        0.475460\n",
       "45            104           148              168        0.472633\n",
       "14            211           128              185        0.471934\n",
       "29            188           249               87        0.470708\n",
       "34            150            38              181        0.469640\n",
       "5             207           226              108        0.468544\n",
       "41             73           136              206        0.466684\n",
       "27            167            19              185        0.466514\n",
       "46            142           230              227        0.465330\n",
       "47             84           101              134        0.464564\n",
       "2             117           247              172        0.464417\n",
       "28            133           255               73        0.463186\n",
       "36            197           226               95        0.462349\n",
       "6              94            25              208        0.462296\n",
       "22            197           243              224        0.459412\n",
       "0             115           136               93        0.458759\n",
       "23             49            17              245        0.457006\n",
       "12            223            82               40        0.456892\n",
       "24            212            69              187        0.456561\n",
       "42            194           252              200        0.455354\n",
       "3             251           254               98        0.454410\n",
       "13            200            15              146        0.453982\n",
       "16            218           252               79        0.453536\n",
       "7             108           145              212        0.453433\n",
       "20             77           175              153        0.452232\n",
       "8             128           159              144        0.452009\n",
       "18            128            45               56        0.450427\n",
       "43            215           195               81        0.448831\n",
       "10            239            81               36        0.445318\n",
       "39            241            77               44        0.445097\n",
       "9             204           178              137        0.442880\n",
       "11             91           227              223        0.441846\n",
       "19            231           255               54        0.441241\n",
       "4             125           142               55        0.439562\n",
       "49            180            95               67        0.439013\n",
       "44            247            87              130        0.438677\n",
       "1             119            57              158        0.437628\n",
       "26            253           159               28        0.433657\n",
       "17             47            43              254        0.432296\n",
       "30            230            54               33        0.431297\n",
       "32             47            58               99        0.428482\n",
       "35            106           201               24        0.424702\n",
       "31            150           161               21        0.418706\n",
       "15            170            91                9        0.408334\n",
       "48             87             9              211        0.405305\n",
       "25             81           204               16        0.401707\n",
       "37             22            18               27        0.378823\n",
       "38             12            86              161        0.357729"
      ]
     },
     "execution_count": 21,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "perf_df = pd.DataFrame({\"rf_estimators\": rfe_list, \"rf_max_depth\": rfd_list, \"rf_max_features\": rff_list, \"valid_r2_score\": valid_r2_list})\n",
    "perf_df.sort_values(by=\"valid_r2_score\", ascending=False)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 22,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 482
    },
    "id": "hL40_7KF6-BT",
    "outputId": "a422b4be-ea85-4e4c-c1a8-90c23255e80a"
   },
   "outputs": [
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAABcIAAAHSCAYAAADPF7hcAAAABHNCSVQICAgIfAhkiAAAAAlwSFlzAAALEgAACxIB0t1+/AAAADh0RVh0U29mdHdhcmUAbWF0cGxvdGxpYiB2ZXJzaW9uMy4yLjIsIGh0dHA6Ly9tYXRwbG90bGliLm9yZy+WH4yJAAAgAElEQVR4nOzdf3QU5b0H/vfu7GbZ/JAfkgD5YQIXTAgQgQhExKYqV7i2BKq1/qJqTy/cinx7+1WvxwMtFXpSa8X2tlL7Ve/tPRUQtbel0FsrILZIhRB+iyQEUvmRH4QAEjAhbHZn9/sHbkgguzuzO7PzzMz7dU7Pqcns5kmYed6f55md53GEQqEQiIiIiIiIiIiIiIgsyml0A4iIiIiIiIiIiIiI9MSJcCIiIiIiIiIiIiKyNE6EExEREREREREREZGlcSKciIiIiIiIiIiIiCyNE+FEREREREREREREZGkuoxtAvdXU1MDn80GSJHg8HqObQ0REGvH5fJBlGR6PB8XFxUY3hzTE7CYisiZmt7Uxv4mIrCdWdnMiXDA+nw/BYBDBYBB+v9/o5hARkcZ8Pp/RTSCNMbuJiKyN2W1NzG8iIuuKlN2cCBeMJEkIBoNwOp1ITU1Fe3s7ACA9Pd3glpEZ8HwhNXi+JNfFixcRDAYhSZLRTSGNXZ3dAK8vUo7nCinFcyX5mN3WxrE3JYLnCynFcyW5YmU3J8IF4/F44Pf7kZqaisLCQuzevRsAUFhYaHDLyAx4vpAaPF+Sq66uDu3t7Xz01oKuzm6A1xcpx3OFlOK5knzMbmvj2JsSwfOFlOK5klyxspubZRIRERERERERERGRpXEinIiIiIiIiIiIiIgsjRPhRERERERERERERGRpnAgnIiIiIiIiIiIiIkvjRDgRERERERERERERWRonwomIiIiIiIiIiIjI0jgRTkRERERERERERESWxolwIiIiIiIiIiIiIrI0l9ENIG355AAcACSHE3IoiBAAj8R/ZiIiIiIyDmtUIiIioitYGxmDf2GL6JID6JT92NBYg6pTx9AR8CHN5UHZkALMyC2GV3IjhRcUERERESURa1QiIiKiK1gbGYt/WQvokgP49MIZrKjZAn9Q7v56R8CHzU11+PBkPRaOKceIjMG8mIiIiIgoKVijEhEREV3B2sh4XCPcAjpl/zUXUU/+oIwVB7egU/YnuWVEREREZFesUYmIiIiuYG1kPE6Em5xPDmBDY03EiyjMH5SxsbEWXXIgSS0jIiIiIrtijUpERER0BWsjMXAi3OQcAKpOHVN0bFXrUV3bQkREREQEsEYlIiIi6om1kRg4EW5yksOJjoBP0bHtfh8kB//JiYiIiEhfrFGJiIiIrmBtJAb+VU1ODgWR5vIoOjbd7YEcCurcIiIiIiKyO9aoRERERFewNhIDJ8JNLgSgbEiBomPLsobr2hYiIiIiIoA1KhEREVFPrI3EwIlwk/NILszILYbbKUU9zu2UcFfuaKRIriS1jIiIiIjsijUqERER0RWsjcTAiXAL8EpuLBxTHvFicjslLBxTDq/kTnLLiIiIiMiuWKMSERERXcHayHi8vWABKZILIzIGo3JSBTY21qKq9Sja/T6kuz0oyxqOu3JHwyu5eTeJiIiIiJKGNSoRERHRFayNjMe/rEWkSC6kSC5U5Jdgdn4JJIcTciiIEC4/fkFERERElGysUYmIiIiuYG1kLFMsjfKnP/0JDz30EEpLSzFhwgTcc889WL16NYLBxHdQffvtt1FYWIjCwkIsW7Ys4nHnz5/Hz372M8yaNQvjx4/H2LFjcfvtt+M//uM/UFtbm3A7tOL54oKSnE6kSC5eREREZAhmNxH1xBqVyByY30REycHayBjC/5WXLl2KN998Ex6PB7fccgtcLhe2b9+OZcuWYfv27fjlL38JpzO++fympia88MILcDgcCIVCEY9rbm7Gww8/jObmZgwcOBBTpkyBx+NBbW0t1q9fj3fffRc/+9nPMGPGjHh/TSIiIstgdhMREZkP85uIiKxO6InwDRs24M0330RmZiZWrVqFgoICAMCZM2fwyCOPYNOmTVi5ciUeffRR1e8dCoWwePFihEIhzJkzB2vXro147EsvvYTm5maUl5fjF7/4BbxeLwAgGAziV7/6FVasWIElS5bgjjvugNvNBe2JiMi+mN1ERETmw/wmIiI7EHpplFdffRUA8PTTT3cHMQAMHjwYzz33HADg9ddfj+sxrTVr1mD79u148sknkZOTE/XYHTt2AAAef/zx7iAGAKfTiQULFqBfv35oa2vD8ePHVbeDiIjISpjdRERE5sP8JiIiOxB2IrylpQUHDx6E2+3GzJkzr/n+5MmTMWTIEJw+fRr79u1T9d4NDQ148cUXUVpairlz58Y8PiUlJer3HQ4HAGDgwIGq2kFERGQlzG4iIiLzYX4TEZFdCDsRXlNTAwAYNWoU+vXr1+cx48aNAwBVG2aEQiEsWrQIsiyjsrKyO0ijmTZtGgDg17/+NTo7O3u91yuvvILOzk7ccccduP766xW3g4iswScH0CUHIAeD6JID8MkBo5tEZBhmNxHphXlLpB/mNxGROqxLzEvYNcIbGxsBANnZ2RGPGTZsWK9jlVi1ahWqq6vx1FNPYfjw4Ype873vfQ+1tbXYsmULbr/9dowfPx4pKSk4dOgQmpubUVFRgR/+8IeK26BEe3s7du/e3f3fPf8/USw8X/SXnZcLT0Ya3m+uQ1XrMXQEfEhzeVCWVYDp2YXwfd6B5gblfZOReL6QVpjd7ddcT7y+SCmeK32zUt5qhecKaY35zbE3xY/ni70kUpfwXBGDsJ8Iv3jxIgD0WhfsamlpaQCAjo4ORe954sQJvPTSSxg7diy+/e1vK27LoEGD8Nvf/hZf+9rXcO7cOfz1r3/Fhg0bcPz4ceTm5mLy5MlIT09X/H5EZG7Zebn4TJKxZM+fsbm5Dh0BHwCgI+DD5uY6LNnzZ5xzycjOyzW4pUTJxewmIi0xb4mSg/lNRBQb6xJrEPYT4VoLP5YVCARQWVkJSZIUv/Yf//gHFixYgI6ODvz0pz/F1KlT0a9fP3zyySd48cUX8f3vfx979uzB888/r1l709PTUVhY2H3HqLS0VLP3Juvi+ZIc57s6UblzPfxBuc/v+4MyXqndispJFSjNEvffgudLctXV1aG9vd3oZpiKWbMb4PVFyvFcicwqeasVnivJx+yOj1nzm9cYqcHzxX7irUt4riRXrOwW9hPhqampANBrXbCrhe9Gh+9OR/PGG29g586dmD9/PoqKihS3IxAI4Lvf/S6OHz+Ol19+GbNnz0ZmZiYyMjJwyy234De/+Q0GDx6MP/zhD6iqqlL8vkRkTj45gA2NNRHDL8wflLGxsRZdXCuMbITZTURaYd4SJQ/zm4goOtYl1iHsJ8JzcnIAAM3NzRGPaWlp6XVsNO+//z4AYNu2bdi5c2ev7zU1NQEANm3ahCNHjiA1NRWvvvoqAGD//v2or69HXl4eJkyYcM37DhgwAF/60pfwhz/8Adu3b0dZWZmC346IzMoBoOrUMUXHVrUexez8El3bQyQSZjcRaYV5S5Q8zG8iouhYl1iHsBPhxcXFAIAjR47g0qVLfe5efeDAAQDA6NGjFb/v3r17I36vtbUVra2tyMjI6P7ayZMnAaDX164W/l5bW5vidhCROUkOZ/daYLG0+32QHMI+eEOkOWY3EWmFeUuUPMxvIqLoWJdYh7D/MsOGDcOYMWPg9/vx3nvvXfP96upqtLS0IDMzs8+7xVdbuXIl6urq+vzfwoULAQAPP/ww6urqsGvXru7XZWVlAQA+/fRTXLhwoc/33r9/PwAgN5cL4hNZnRwKIs3lUXRsutsDORTUuUVE4mB2E5FWmLdEycP8JiKKjnWJdQg7EQ4A8+fPBwAsX74cx48f7/762bNnsXTpUgDAvHnz4HRe+TVWrVqFmTNn4plnntGkDePHj0dWVhYuXbqExYsX91pwPRgM4pVXXsG+ffvgcrkwY8YMTX4mEYkrBKBsSIGiY8uyhuvaFiIRMbuJSAvMW6LkYn4TEUXGusQ6hF0aBQBmzpyJBx98EGvWrMGsWbMwdepUuFwubN++He3t7Zg+fTrmzp3b6zXnzp3D0aNHkZmZqUkbUlJS8JOf/AQLFizAxo0bUV1djXHjxqFfv36ora1FY2MjnE4nFi1ahBtuuEGTn0lE4vJILszILcaHJ+ujbpThdkq4K3c0UiShu1nh+eQAHLj8KJocCiKEy/8GJC5mNxFpgXlL8WLtEB/mNxGJzOi+nXWJdQj/L/Pcc8+htLQUq1evRnV1NYLBIEaMGIF7770XDz74YK870nq59dZbsW7dOvzP//wPqqqqutsxePBgfOUrX8EjjzyC8ePH694OIhKDV3Jj4ZhyrDi4pc8QdDslLBxTDq/kNqB11tAlB9Ap+7GhsQZVp46hI+BDmsuDsiEFmJFbDK/kZnEhMGY3EWmBeUtqsHZIHPObiEQjUt/OusQaHKFQKGR0I+iKuro6tLe3Iz09HYWFhdi9ezcAoLS01OCWkRnwfEmecCBvbKxFVetRtPt9SHd7UJY1HHfljjbFYEvU86VLDuDTC2ewoiZ6gTEiY7Dwf+Oeru7fyTr6+rcV9foi8fBcic4KeasVniuR6VU7MLutjWNvSgTPF/2JOC6Mpy7huZJcsbLbHlUjEZHGUiQXUiQXKvJLMDu/hI/faqhT9kcsdgDAH5Sx4uAWVE6qsM3kBxGRXTFvSQnWDkRE1iNi3866xPyE3iyTiEh0ni+CUHI6kSK5GH4J8skBbGisibruGnC56NnYWIsuOZCklhERkZGYtxQJawciIusRvW9nXWJenAgnIiJhOABUnTqm6Niq1qO6toWIiIjEx9qBiMh62LeTXjgRTkREwpAcTnQEfIqObff7IDkYY0RERHbG2oGIyHrYt5NeeKYQEZEw5FAQaS6PomPT3R7IoaDOLSIiIiKRsXYgIrIe9u2kFy5iQ2RxPjkAB8BNHMgUQgDKhhRgc1NdzGPLsobr3yAiItIdaxVKBGsHIiLrubpvz0sbiPJho1ByfQ68khudsh8fn23ClpNHUDhgiLGNJVNhhUlkUV1yAJ2yHxsaa1B16hg6Aj6kuTwoG1KAGbnF8ErupO2sTKSUR3JhRm4xPjxZH3VjFLdTwl25o3kOExGZGGsV0gJrByIi6wn37X8/+Q/cO2ICigYMxd+aD2Ptsf3d9cKUrALMHz0Naa4U9u2kGJdGIbKgLjmATy+cweKd67G5qa57ba2OgA+bm+qweOd6fPr5maTvrEzW4ZMD6JIDkINBdMkB+DQ8l7ySGwvHlMPtlPr8vtspYeGYcnglt2Y/k4iIetOznwdYq5C2WDsQEVmPV3LjBxP/BQNSvPjRnnfxQXPveuGD5jos2/MuGjrOaVIv6F37kBh4y4TIYrxeLzplP1bUbIn4qRh/UMaKg1tQOamCd05JlWR8ei9FcmFExmBUTqrAxsZaVLUeRbvfh3S3B2VZw3FX7mh+SpCISCfJ+pQ2axXSEmsHIiLrSZFcSHG68Pqhj3StF/iEmr3wX5LIYgZmDsaGxpqoj4YClwNjY2MtZueXsFMnRcKf3rt64iL86b0PT9Zj4ZhyjMgYrMlkeIrkQkV+CWbnl3DdWCKiJEhWP++TA6xVSHOsHYiIrMUnB7ChSd96IZljXBIDl0YhspgB/fuj6tQxRcdWtR7VtzGkGyMe21L66b1O2a/Zz/R8MaiVnE6kSC4OZImIdJSsft4BsFahhESrg1g7EFFfuOyF+SSjXjBijEvGYlVAZDFuydW9blYs7X4fJAfvh5mJUY9t8dN7RETWlsx+XnI4WatQXPj4OhGpxX7DvPSuFzjGtSdWlUQW45cDSHN5FB2b7vZADgV1bhFpxciNxfjpPSIia0tmPy+HgqxVSDVusEpEarHfMDe96wWOce2JE+FEFtN2/jzKhhQoOrYsa7i+jRGAlR6BM/KxLX56j8jerNSXUt+S2c+HANYqpBofXycitdhvmJue9YJPDiAE4CeTZ+OnU76GuSMnIy9tYMTjOca1Dn6mn8hizp0+gxkji/Hhyfqoj/i4nRLuyh1t2Ud7rPYInNGPbYXvxiuZJAnfjZd4r5XI9KzWl1JkyeznPZILM3JZq5ByRtdBRGQ+7DfMT496IVJtOyWrAPNGT8Ohtha8Vb8Ll7davoJjXOvgvyCRxXR2dsIrubFwTDncTqnPY9xOCQvHlMMruZPcuuSw4iNwRj+2xU/vEdmPFftSiizZ/bzdaxVSx+g6iIjMh/2GNWhZL0SrbT9orsOP9ryLgSmpeGDkzde8lmNc6+BEOJEFpUgujMgYjMpJFZieU4R09+V1tdLdHkzPKULlpAqMyBhs2TveVnwEzqilScLLIbgcTszIKY5YgITx03tE1mHFvpQiC3/qKln9vN1rFbNL9nJJXKKNiNQSpd/g8nKJ0bJeUFLbvnbo7ygaMLTXMikc41oL/xWJLCpFciFFcqEivwSz80sgOZyQQ5cf8PFYuAO36iNwyV6apK9Hxh4dNQXfGX0b/r/arX3+ffnpPSLrsGpfStGFP3W14mDfg0St+3m71ipmZtRySVyijYjUMrrf4PJy2tGiXlBT225pPozyYaOwqr6aY1wLYnVAZHGeL0JDcjqRIrksP7C06iNwyXxkPdIjYyuPVOOsrwNLJt6N6dn89B6RlVm1L6XojPqUtt1qFbMycrkkLtFGRGoZ2W9weTl9JFIvqKltd5w+hpuuz+EY16L4L0lEliLKI3BaS+bGYpEeGQsihDfrdyIvbSC+nH0jflxQAZdD4qf3iCzIqn0pxcZPaVMkSpdLqpxUocvNEm6wSkRqGNlvGNlfUt/U1rZpLg8q8ktY+1gQRy1EZCnhR+CUCD8CZxbJ2FhMySNjDR3nsPLIDqw/fgByKMhP7xFZkJX7UlKGn9KmntQul6THpxy5wSoRqWVEvyFCf0nXiqe2Ze1jTZwIJ4oDN7wQl5UfnU3GI+tcDoGIAGv3paQ91kXWJ0J9wA1WiUgtI/oNEfpLupaotS1rqORjlUCkAje8EF+yH4HzyQE4gKQ9Oq73I+tcDoHIPPTsf7gMASnBusg+RKkPuHQPEamV7H5DlP6SelNV2+aMhgOXa229soU1lHH4VyVSKLzhxdVrfYU3vPjwZD0WjinnJ1EEEH4EbsXBvtdl0+IROKODq2cga7m7udG7qxNRbMnqf5LRl5J5sS6yF9HqA73qICKyrmT1G6L1l3SFktp2ftE07PusEeuOfazb2J41lLF4tREppHTDi07Zn+SW0dX0fgTOyruAi/rIGBFdlsz+h8sQUDSsi+yF9QERkTLsL8UVrba9M7sQP5h4N851XcRb9bt0HduzhjIWRy5ECqjd8GJ2fgknBnSkZDkAPR+Bs/Iu4FwOgUhsye5/uAwB9YV1kf3Eqg/y0gaifNgolFyfgzSXB8FQUNdHyomIRMXxlNiurm2dXyxls/9sI16r3YrGjrZex2tdW4tSQyV7iVeR2OO3JEqQ2g0vZueX6Noeu4pnOQCtH4ETJbj0xOUQiMRkZP/DZQioJ9ZF9tRXfeCEAw+MvBlFA4bib82HsfbYfq5zSkS2x/GU+DySCz45gD8c24vNTXVRj9Wytja6hjJ6iVcRWPu3I9KIlhtemPXOm9HtFmUdLaODKxl6PjK2sbEWVa1H0e73Id3tQVnWcNyVO9oWAUkkGjv0P3aUfUMeBvTvDzkYNE1dwI3A7Kmv+qAivwQDUrz40Z53da/PjK5FiYiUEnk8ZcW+NN7fyYja2sgaSpQ5FaNZ9zcj0pAWG16Y9c6bKO0WZTkSuwz+uRwCkXjs0v/YRThft15oQFX9VtPUBQA3ArOznvXBnPwSXAz4sXjXel3rM1FqUSIiNUQbT1mxL030dzKitjayhhJlTsVorEiJFEh0wwuzbq4oSrvVLgegZ3vCwaVEOLjMzPNFASc5nUiRXJwEJzKQ3fofK+uVr83mqQvCuBEYeSQXggA2NOlbn4lSixIRxUuE8ZQV+1ItficjamujaiiR5lSMxolwIgXCG164nVLU4yJteGHWXYFFabfaR5b0xME/ERmF/Y91iJKv8Uq0LiJrSEZ9ZvZrhYhIBFbsS7X4nYyorY2qoUSaUzEaJ8KJFApveBGpw4q04YVZ77wls90+OYAuOQA5GESXHIDvqvcSaTkADv6JyCjsf6zBrHXB1eKti8g69K7PErlWYtWWRGQOvJYTZ5W6oyetfiejamsjaiiR5lSMxhESkULxbnhh1s3NktFupWt6ibYWKXcBJyKjsP8xP7PWBVcTeSMwSg6967N4rhUrroFLZEe8lrVjlbqjJy1/JyNqayNqKNHmVIzEnoNIhXg2vNDqzluyd3fW+46hmh2Lw48sbW6qi/m+yVgOwIyD/77OH6/Xi87OTqObRkQqqO1/kp0dFJuVPpEj2kZglDw+OYBgKKhrfab2WgmFQjj6+VlFtaVINRoR9aZmnKjntWyVGspKdUeYlr+TUWP7vmqoQCiIYCiIFOflDwP65IBm55xocypGMt9VTCSAnp1RrLtkid55M+puuN53DNXsWNw/xYsZucX48GR91MefkrkcgFkG/9HOn+nDC+H7vMPoJhKRSkr6H36SSlxW/ESOmrqIzK1n39LQ3oa5oybrVp+puVYK+w/BJTmguLZk/0ckLjXjRD2uZavVUFasO7T+nYwc23skF7rkANoDPl3PufAyMCLNqRhF7LObyAIS2YDByN2d9dw4Ip41vURdi1SEXcAjiXX+LNnzZ5xzyaZYB46IrhWp/zEyOyg2bnpKZnV133L4/CkcamvB/KJputRnaq6VWfnjsKHJWmvgEtmR0etZW7GGsmLdodfvZMTYPpnnnKhzKsnGiXAinSWyAYORuzvruXFEPDsW93xkaXpOEdLdHgCX7/BOzylC5aQKPup6FSXnzyu1W021OzgRxWZkdlBs3PSUzKqvvuWt+l0413URP5h4N+7MLtS0PlNzrdyQPlB1bUlE4olnnKglNTWU1+vV/OfrwYp1h5V+p2TW7ZxTuczavx2RIOLZgEHt3XA9NrXQa+OIeNf0MstyJCKI5/yxeuAR2QGvfXPgpqdkNpH6liBCeLN+J/LSBqJ82CgsmXg3Ul0pCIZCABKvz5ReK5JDstwauER2ZOR61mprqGmZueg80aDZz9eTFesOK/xORtTtnFMxySfC//SnP+Ghhx5CaWkpJkyYgHvuuQerV69GMBhM+L3ffvttFBYWorCwEMuWLYt6rCzLWLNmDR5++GFMmTIF48aNQ3l5Ob7zne/ggw8+SLgtZF3x3Hkz+m44oN8dw/CaXkqE1/TqSeTlSEQhwvlD9sbsNgavfXPoma9af4qWSA+x+paGjnNYVV+NZ3asxbPVf4QD2gyoldaiwQRrS7qC+U1GSnScmAi1NdSA/v01+9l6s+Inga3wOxlZt9t5TkX433Tp0qV488034fF4cMstt8DlcmH79u1YtmwZtm/fjl/+8pdwOuObz29qasILL7wAh8OB0BefWojk3LlzmDdvHg4cOIABAwZg/Pjx8Hq9aGlpwbZt23D99dfjjjvuiKsdZA9q77yJsruzHncMuWOx/kQ5f8iemN3G4bVvHuF8vS0jD7PyxiLF5bbdJ3LIPIzsW5TUoj45wNpSA8xvMpqR40S1/ZzbZFltxU8Cm/13Yt1uDKHPjA0bNuDNN99EZmYmVq1ahYKCAgDAmTNn8Mgjj2DTpk1YuXIlHn30UdXvHQqFsHjxYoRCIcyZMwdr166NeGwwGMTjjz+OAwcO4JFHHsHTTz8Nj+fKXcr29nY0NTWpbgPZU88OOdrOxfHshKwnpe1W+l7csVhfVtwdnMyB2W0sXvvm09zQgOaGBpSWlvLfgoQlQt8SrRZlbZk45jeJwMhrWW0/5zfRhpk9aTmuF4VZfycRstWOhP4LvvrqqwCAp59+ujuIAWDw4MF47rnnAACvv/56XI9prVmzBtu3b8eTTz6JnJycqMe+88472Lt3L26//XYsXry4VxADQHp6OgoLC1W3gSgaK+7u3BN3LNaX1c8fEhez21i89olID8nsW3xyAF1yAHIwiC45AJ/CySbWlolhfpMojLqW1fZzbefPa/rzyX6Sla3x5qpVCTsR3tLSgoMHD8LtdmPmzJnXfH/y5MkYMmQITp8+jX379ql674aGBrz44osoLS3F3LlzYx6/evVqAMBjjz2m6ucQJcJKOyH3xQpreonM6ucPiYnZbTxe+0Skh2T0LV1yAOe7OrHu+H48W70OCz56C89Wr8O64/txvqsTXTEG7qwt48f8JpEYdS2r7efOnT6j6c8n+9E7WxPNVasStgqoqakBAIwaNQr9+vXr85hx48bh1KlTqK2txcSJExW9bygUwqJFiyDLMiorK+FwOKIe39raisOHD0OSJEyYMAFHjx7Fu+++i1OnTqF///6YNGkSbrvttpjvQxQPo3dC9skBOADd1toy+5peovM4XVhQ/CW8UvNhxPNnQfGX4HHyb03aYHaLwejsIPvRu14gMejZt3TJAXx64QxW1PR+746AD5ub6vDhyXosHFMec/KLtWV8mN8kGqOu5RSnS3E/19nZqVs7yD70ylatctWKhP1tGxsbAQDZ2dkRjxk2bFivY5VYtWoVqqur8dRTT2H48NiPFhw+fBgAMGDAAKxZswYvvvgiAoErd01ee+01TJgwAb/61a9w/fXXK24HkRI974ZvbKxFVetRtPt9SHd7UJY1HHfljoZXcmvecXXJAXTKfmxorEHVqWPoCPiQ5vKgbEgBZuQWa/4zzbqml+h8wQDOXmrHDybejS3Nh7Hj9LHu82dKZgHKs2/E4bYW5KQNQD9wQowSx+wWg1HZQfaT7HqBjKVn39Ip+68ZrPfkD8pYcXALKidVKHp/1pbqML9JVMm6lsN5trGxBtmpA/ocP7GGIj3ola1a56qVCPvbXrx4EQDg9XojHpOWlgYA6OjoUPSeJ06cwEsvvYSxY8fi29/+tqLXnP9i3afz58/j+eefx1e/+lUsWLAAQ4YMwSeffIJly5Zh7969+Pd//3esWrVK0bQPob4AACAASURBVHsq0d7ejt27d3f/d8//T/bj9XoxbXAuvpo7Bm7JBb8cQNv5NjTVH+3zTnQi50t2Xi4+k2T8+tDWiHcOF4y+DQMDEpoblBfClFzZN+Rh64UGbG6uQ17aQJQPG4UlE+++/OkF2Y/9ZxvxWu1WNHa04c6LF3DbdXloPtFgdLPJ5Jjd7df0v0bmt9rsIGOZrdZjvWAco88VrfuWcM0SbWM84PKgfUNDDWsWHTC/Ofa2s77y7OrxkxwKwee71Gc/x/OFlIp2rmiZrczV6ISdCNda+LGsQCCAyspKSFL0NXjCwpuBBAIBlJaW4qWXXur+XllZGX7zm99gxowZ2LlzJ6qqqlBWVqZL+8neOjs70dnQgOYGfTsnr9cLT0Yafr3nz1HvHL5SuxXLJn4FXq+XkymCGtC/P6rqtwIAGjrOYVV9NVDf97E7Th/DrLyxaIZ9wo/MgdmdmGRlB9kP6wV707pv6VmzxMKaxRyY32QWkfLs6vGT2ylh2cSvGNRKsgMts5W5Gp2wE+GpqakAELVoDt+NDt+djuaNN97Azp078cQTT6CoqEhxO3q+9ze+8Y1rvj906FCUl5djw4YN2LFjh2ZhHN4NO3zHqLS0VJP3JWtL9HzxyQGsO75f0Z3DzScPY3Zhie0eozELORhER8Cn6Nh2vw8pLjf7GZ3V1dWhvb3d6Gboitl9ObuBxPtjsg8zniusF4xhxnNFCZFrFjtkN8D85tjbvhLJM54vpFSyzxWRczUZYmW3sBVpTk4OAKC5uTniMS0tLb2Ojeb9998HAGzbtg07d+7s9b2mpiYAwKZNm3DkyBGkpqbi1VdfBQDk5uZ2H9fz//cU/vqZM9w1WG/cjElfDgBVp44pOraq9Shm55fo2h6KnxwKIs3lURSA6W4P5FAw6Wto8nq2Hma3/njdJA//1pGxXiAtmaFmsTrmt/Ux0/rGPCMrEiVXRe13jG9BBMXFxQCAI0eO4NKlS33uXn3gwAEAwOjRoxW/7969eyN+r7W1Fa2trcjIyOj+2vDhw5GamoqLFy+ira2tz9edO3cOwJU76aQ9bsaUHJLDqerOoeTgIERUIQBlQwqwuaku5rFlWbE3L+pLvMHG69m6mN364XWjXKJFN//WsbFeIC0lo2ah6Jjf1sVMi455RpGIOomrhNG5Knq/I+y/4rBhwzBmzBgcPHgQ7733HubMmdPr+9XV1WhpaUFmZiYmTJgQ8/1WrlwZ8Xsvv/wyVqxYgYcffhhLlizp9T23240vf/nLePfdd7F9+3ZMnz691/f9fj927doFABg7dqzSX49U6JID+PTCmWt2vO25GdPCMeUYkTHY1iGuBVHuHFLiPJILM3KL8eHJ+qiP+rmdEu7KHa3q2kkk2Hg9WxuzWx+8bpTRoujm31oZ1gukJT1rFlKG+W1NzLTYmGd0NdEncZUwMlfN0O8IfQXPnz8fALB8+XIcP368++tnz57F0qVLAQDz5s2D03nl11i1ahVmzpyJZ555RrN2/Nu//RucTifefvttbN16ZcF5WZaxfPlynDhxAkOGDME///M/a/Yz6YpO2X/NRdSTPyhjxcEt6JT9SW6Z9YTvHCrBT+SIzyu5sXBMOdzOvjcocjslLBh9G7ySW/F7hoNt8c712NxU1100hoNt8c71+PTzM+iSA32+ntez9TG7tcfrJrZE+6Yw/q2VYb1AWlNSsywcU66qZiF1mN/Ww0yLjXlGPWlVT4rAqFw1Q78j9G2MmTNn4sEHH8SaNWswa9YsTJ06FS6XC9u3b0d7ezumT5+OuXPn9nrNuXPncPToUWRmZmrWjqKiIixatAiVlZWYN28eSkpKMHToUNTU1KChoQEZGRn4xS9+0ecjZJQYnxzAhsYaRZtXbGysxex8bsaUCH4ixzrCj3KNvC4LL5XdgxPt5/Cn4x+j7nwr0t0elGUNx53DboTv8w5V/45Kg61yUsU178vr2R6Y3dridaNMIn1TGP/WyrFeIK2lSC6MyBiMykkV2NhYi6rWo2j3+7prlrtyR0f9FJ6ZH2EXBfPbWphpyjDPqCct6smr+eQAiseNhVtyoUsOJC2fEs3VeJil3xH+Kn7uuedQWlqK1atXo7q6GsFgECNGjMC9996LBx98sNcdaT1985vfxI033ojf/OY32LdvH2pqapCZmYn7778f8+fPj7iZByWGm1ckX/jO4YqDfQcAP5EjtmiPcn278Fb0k1xwOBwIAfhH3WF0dnZiWNYQRe+daLDxerYPZrd2eN3EplXRzb+1OqwXSGspkgspkgsV+SWYnV+iaFLbCo+wi4T5bR3MNOWYZwRoP4krQj7Fk6uJMEu/4wiFQiFDfjL1qa6uDu3t7UhPT0dhYSF2794NACgtLTW4ZcaQg0Es+Ogtxce/cusDkJJUoIlIq/Ml3Gkn684haSPSelxh4SIuvB6X2vOlSw7g2ep1itfQe37S7F7nid2v56v7d7KOvv5tteqP7X7dKJFo3xRm1N/azLUe64XkMvO5oge1dU88mN3WZuWxN+sHdeLJMyudL6RdPRl+L73zSUSi9Duxsts6f3GyJG5eYYxk3zkkbejxKFdPie6qbqbrmY9YkyjMdN0YJdG+KYx/a/VYL5CRFNc9N8dX9xCZGTNNHSV5dvX4IDsvD+fOnDG03aQdrepJQP9xuajM0u9Y5y9OlhTevGJzU13MY7l5hfZ6DmLtXBiZgdpHub56w1jVa5UlGmxmuJ5FeISNqCczXDdG06ro5t86fqwXKNnU1D0bGmtw9w1jkeKUmOFkG8y0+PSVZxHHB1kFmD68EF1ywHZ9ixU/NKRVPWmWdbL1YJZ+h5UqCS28eUWknW7DuHkF2Z3a9biCoRAW7/4/PLHtbTxbvQ7rju/H+a7OqDtgx7Oruk8OoEsOQA4G4QBwd94YDM+4PuprjbqerbRLOFmHE8BdOaOZg1HE0zf1hTUHkXmoqXt2nD4GORRkhpOt2CXTeo41uuQAfBpf41HHB811WLLnz7bqW7rkAM53dWLd8f14tnodFnz0luKxpOi0qifVjsutxCz9DifCSXjhzSsiXUzcvCI+ehcNlFxqH+VyOyXVE72qgi1nNHzBwDVF0rsNB/Fvo2/D3JGT4YSjz9cadT0rfYStU/YnuWVkV11yAP+4cAYHPmvG/KJpzMEItCy6WXMkB2sQSpTauscruZnhZDtWzrRkTchyfHCF1T80pFU9qeUSK2FmqpvM0O+Y87Yf2UqK5MKIjMGonFTBzZg0wKUfrEnto1x9FWtK1ipTuqt6VzCApbve7XVMuEj68GQ9nigux/877g68eujvQlzPdn6EjcQVHnzJwSAeGHkzfjDxbmxpPowdp491XzdTMgtwV+5opLpSbH1OKu2bYhXdrDn0xRqEtBJP3cMMJ7uxaqZF2oiw51hDi40IOT7ozQ7rXmtRT2q5TrYZ6yYz9Dti/cWIIuBmTNpIVtFAyadmPa4pmQX4+GxTn9+LVcgpCTbJ4cSi6nVRi6Rf1VzevOonk2bDKcD1rPYRttn5Jbq2h+jqwdeb9TuRlzYQ5cNGYcnEu+GV3OiU/dh/thHVp4/jy8NGGdxiY2lZdLPm0AdrENKS2nVI68+fBsAMJ/uxYqYla0KW44Mr7HJTQIt6Uqt1ss1cN4ne7xjfAiIVRN6MyQwbRtjhLq5dhR/l+vBkfdQCxe2UUJ59I16r3RrxmFiFXLRgcwD44/H98AWjP67lD8rY2HS5SJKcTsOvZz0eYSNKRF+Dr4aOc1hVXw3U9z423e2x/UQ4oH3RLXLNYUZmrEHMUNvZlQPAndmFiuqeO7JvxPHPPwPADCf7skqmJXNCluODK9TeFJiTX2LaDE20nlQzLo+2xIoZ66aridrviPnXIjIRszyuYvRdXLMGoZkoeZRrftE0HGprQWNHW8T3UVrIRdpV3YyfnNDyETYiLXDwFT9Ri24707sG0brGMEttZ3d1bacwv2gaXjv09xh1zymMHZQNgBlOZHbJ/JQ2xwdXKK1LnXCgIr8EFwN+bGgyd4YmUk8musSK0XM3Vse/FFECzPS4ilGPdnEwmTyxHuUqHzYKtW0teKt+V9T3SaSQM+vknVaPsBFphYMvshK9ahA9agwz1XZ2JjmcWHmkOur+CeXZN+LQF3XPL6beB4AZTmR2yRxrcHxwWZccQAghRXXpAyNvxoAULxbvWm/rDE10iRUuy6Mv6555RElgpsdVjJig5GAy+SI9yiWHgtjaUo/fH90X8z0SKeTMOnmn1SNsRFrh4IusRI8aRK8aw0y1nZ3JoSC8rpSo+ye8VrsVjR1t3ZtlMsOJzC+ZYw2OD65k7VlfB6ZkFeCD5sh1aV7aQBQNGIof7XmXGYre4/Kv5o6BW3IhiJCip9bM+uEys+BfiyhOah9X6ZKjr5mst3DRoES4aEiU0sFkp+xP+GdRb54vgldyOpEiueB1pWBK1nC4nVLU1yVayIUn75QQbfIu/AhbpL+Rkl3CibQSHnzpfc0SJYMeNYgeNYbZajs761lvhPdPeGbHWvw/297BMzvWYnX9zu5l4KZkFuDAZ83McCILSPZYw+7jg3DW/rX5ML6cfWPUurR82Cj8rfkwM/QqHsmFmgOfYP++fUiRXIqWbjNi7sZOOBFOFCe1j6sYLdlFQ8/BZF7aQMwdORk/nfI1vDz1G/jplK9h7sjJyEsbaLsgNFIyCjkzT971fIRtek4R0t2Xi490twfTc4pQOamCTy9QUplt8OWTA+iSA5CDQXTJAfjYr9MXtK5B9JqwNlttZ2dq6o07cwpx0/U5zHAiHSQ7+5M91og2PrgzuxDLJn7Fsn1Lz6xt6DiHQ20tmF80LeLfvuT6HOxoPabovZmh0Zn5w2VmYL2rlShJzPa4SrIf7XIAqD51HA+NnISiAUPxt+bDWHtsf/f6nVOyCjBv9OWNG//v+AGua5UEia5VplSim4MYKdFdwom0lKxrNlHcC4Ji0boG0WvtTLPVdnantN5IdXngdYlXcxCZmZHZn+yxRqTxwdnPzqLl6AkUFxdr8nNEc3XWvlW/K+q+DEqXrAGYobFwWR598a9FFCczroWczKJBcjgxq2AcBqR4r1knrCPgwwfNddjaUo/5RdPw1fxxDMIkSWStMjU/wwyTd9Eksks4kZZEvznDvSBIKS1rEL0mrM1Y29mZFeoNIjMyOvuNuvavHh80n2jU9P1Fc3XWBhGKui8DM1RbZv5wmehYFRDFyYwbmSWzaAiGghg9YCiWxdgs47VDf8eSiXcjyCBMKo/kwu59+wEApaWlmr+/6JN3RGYj6s0ZbixISmlZg+g12DZjbWd3rDeIkk+E7Oe1r79IWRvelwH1V76W7vZgUmY+M1RDvNmrH/7FiOJk1sdVklU0BEJBxZtl/K35MGbljwPvZZqbTw7AAUQ8p0SavCOixKldp3l2fokwWUjG0KoG0WvC2ujaLlaOUmSsN4iSQ7Ts73ntB+QgHADkYJB9qAbUZq3L4TTl/IjIeMNHH3H/5RoaGvDb3/4WH330EVpaWuDz+VBTU9P9/QsXLuCNN96Aw+HA/Pnz4XZziousx8yPq+g9YJAcTlQp3Cxjx+ljmFNwk+ZtoOTg+sDmwewmLem1TjNZX6I1iJ4T1kbUdsxRioX5TaIQMfvZh+pDbda6JRe8gGnnR0TGm73aiqs32LRpE5555hlcunQJoVAIAOBwOHodc91112HHjh3YtWsXRo4ciRkzZiTeWiLB8HGVyLjhlD0YvUYgKcfsJq2xnycj6TVhnezaTk2Okj0xv0kkomU/xyL6Upu1nB8hM1DdK/3jH//A008/jc7OTnzjG9/AqlWrMHDgwD6Pve+++xAKhfC3v/0t0XYSCStFcqF/ihcV+SV4ftJsvHLrA3h+0mxU5Jegf4rXtp18eE0xJcLrd5L5KF0jsFP2J7ll1BOzm/TAfp6M1HOwPT2nCOnuy+diutuD6TlFqJxUEffERzJrOzU56vV6Nfu5ZA7MbxKNaNnPsYi+4slazo+Q6FSfgf/93/8Nn8+Hxx57DM8++ywAQJKkPo+dOnUqAODjjz9OoIlE5sDHVXrjhlPWJ9oagRQZs5v0wH6ejKb32pl613Zqc3RaZi46TzRo3g4SF/ObRCNS9nMskhzxZi3nR0hUqs/GqqoqOBwO/Ou//mvMYwcPHgyv14uTJ0/G1TgiMq/wmmJuZ9/Fehg3yzAvtWsEknGY3aQH9vMkCs8Xg3TJ6USK5DLNBlJqc3RA//66tofEw/wm0YiU/RyLJJdZs5boaqonwk+fPo20tDQMHqxsnbqUlBT4/XwMhciOwmuKRSqUuFmGuYm2RiBFxuwmvbCfJ4qf2hx1c9LBdpjfJCJRsp9jESKKh+pqKjU1Fe3t7ZBlOeJjWWHt7e34/PPPMWjQoLgbSETmxc0yrC28RqCSAjS8RiAfizMGs5v0wn6eKH5qc9QvB5LQKhIJ85tEJEr2cyxCRPFQ3TONHDkSe/bswcGDB1FSUhL12L/85S8IBoMYO3Zs3A0kInPTe/1OMo5IawRSdMxu0hP7eaL4qM3RtvPn9W8UCYX5TaISIfs5FiGieKi+HfYv//IvCIVC+MUvfoFgMPIOwHV1dVi+fDkcDgdmzZqVUCOJyPy4ppj1iLRGIEXH7KZkYD9PpI7aHD13+kySWkaiYH6T6IzMfo5FiCgeqifC77//fhQWFmLbtm149NFHsWnTJgQClx/Tq6urw1//+lcsXboU999/P86fP4+JEyfi7rvv1rzhRERkPFHWCOzJJwfQJQcgB4PokgPw8VFyZjeRCuxDKJnU5GhnZ2eSW0dGY36TWnbLMBHHIkQkNtW3xNxuN/7rv/4Ljz/+OHbu3Ildu3Z1f2/OnDnd/z8UCmH8+PF4+eWX4XA4tGktEREJRZQ1AgGgSw6gU/ZjQ2MNqk4dwyBPKu7MKUTJoJzuxzUBey7VwOwm6ptPDsCByxtuBUJB+IMBbDv1Kd5rqEVHwIc0lwdlQwowI7eYa52TLkTKURIP85uUuroOVpphPXPQjMuasQ8lIrXi6g0yMzPx1ltvYe3atVi7di0OHDjQvTu1JEkYO3Ysvva1r+HrX/86XC52OEREVibCGoFdcgCfXjiDFTVbIAeDeGDkzSgaMBR/az6M3326lxNaYHYT9RRpwmBKVgG+nH0jPJIbb9XvQkfAh81NdfjwZD0WjinHiIzBtus7SH8i5CiJi/lNsfSsg/1Bufvr0TIs3olzEbEPJSI14u4VXC4X7rvvPtx3332QZRnnz59HMBjEgAEDGMBERDbUs9BM9o7snbK/u/h/aOQkDEjx4kd73lU8GLALZjdR9AmDD5rrsLWlHvOLpuGBkTfjzfqdAAB/UMaKg1tQOanCdv0GJY+ROUpiY35TND3r4L5cnWHxTJybAftQIlJCde8wadIkTJkyBQ0NDd1fkyQJgwYNwuDBgxnERESUVD45gA2NNfAHZeSlDUTRgKF4/dBHMQcDnbI/yS01DrOb6AolEwavHfo7igYMRV7awF5f39hYiy6Lr7dKROJgflMsPevgaMIZ5v/ik+BKJs7tVCsTkX2ongj3+/2QZRl5eXl6tIeIiEgVB4CqU8cAAOXDRuFvzYcVDwbsMqHF7Ca6TM2EwZbmwygfNqrX16taj+rZPCKiXpjfFEvPOjiWqtajCISCqibO7VIrE5F9qJ4IHzZsWPeaZEREREaTHE50BHwAgJLrc7Cj9Zii19lpQovZTXSZmgmDHaePoeT6nF5fa/f7IDn4uDURJQfzm2LpWQfH0u73welwqpo4JyKyGtWV/B133IGuri589NFHerSHiIgM5JMD6JIDkINBdMkB+EzwKRA5FESaywMA8EpuVYMBu0xoMbuJLlM7YeCV3L2+lu72QA4F9WgamYAZM5LMjflNsfSsg2NJd3tU56BdamU7Y7aR3aheVOw73/kONmzYgB/84Ad4/fXX8U//9E96tIuIiJLIzDvHhwCUDSnA5qY6dMp+pLk8igr88ISWHTbTYXYTXRaeMFDaR1y9PmpZ1nC9mkYCM3NGkrkxvymWnnVwLGVZw1XnoF1qZTtitpFdqT6rN2/ejAcffBC/+tWvMGfOHHzpS1/C+PHjMWjQIEiSFPF1c+bMSaihRESkD7PvHO+RXJiRW4wPT9bj47NNmJJVgA+alQ0G7ILZTXSZmgmDKZkF+PhsU/d/u50S7sodLWQ/SPoxe0aSuTG/KZaedXC0db/DGQaomzgna2K2kZ2pPqOfffZZOBwOhEIhAMAHH3yADz74IObrGMZE6vjkABy4/Bi3HAoihMuFDpGWvF6v4p3jKydVCFsIeSU3Fo4pxx+PfYxvFd6CrS3KBgOi/j5aY3YTXaZmwqA8+0a8Vru1+78Xjim/ZqkUsj4rZKReWKvqj/lNSoTr4BUH++6remZYisqJc7v1a3bBbLMPZvW1VP/2kyZN0qMdRPQFPqJEyTQwc7CqneNn55cYcv7FCvAUyYURGYPxePFtaO38HPOLpuG1Q3+PORiwC2Y30RVKJgzmF03DobYWtHV1YnpOEe7KHc38tSGfHDBFRiabGWpVqwz8md+kRLgOrpxUgY2NtahqPYp2vw/pbg/KsoZfk2FqJs77YpXry66YbfZghqw2iurfeuXKlXq0g4jAR5Qo+Qb074+q+q2Kjq1qPYrZ+SU6t6g3NQGeIrmQIrnQT3IjFAqh8uYKbGyKPRiwA2Y30RUxJwxyRsMjueB2SrglazgH+DbmAFB16piiY43ISCOIXqtabeDP/CalwnVwRX4JZueXRJ2kVjtxHma168uumG3WJ3pWG80Uux786U9/wkMPPYTS0lJMmDAB99xzD1avXo1gMJjwe7/99tsoLCxEYWEhli1bpvvriKJR+ojS1Zt3EcXLLbmE3Tk+HOCLd67H5qa67naGA3zxzvX49PMz6LpqZ3OP5EI/lxv9PV5U5Jfg+Umz8cqtD+D5SbNRkV+C/ileWwZ+sjG7SWQpkgv9UyL0ER4v+rnckJxOpEguToLbmORwCpuRRhG5Vo23bqDemN/m5vliQjxWhkXNwT5qZV5f1sFssz6Rs1oEwp/RS5cuxdNPP41PPvkEN998M6ZOnYpjx45h2bJl+O53v5tQIDc1NeGFF16Aw+FIyuuIolH7iBKLDNKCXw4gzeVRdGx45/hk0SLAlQ4GSFvMbjIL9hEUjRwKCpuRRhC9VuXAP3HMb/tRmoO8vqyD2WZtome1CBKq9hsaGvDee++hpqYGn332GQBg0KBBKC4uxsyZM5GXl5dQ4zZs2IA333wTmZmZWLVqFQoKCgAAZ86cwSOPPIJNmzZh5cqVePTRR1W/dygUwuLFixEKhTBnzhysXbtW19cRxcJHlMgIbefPC7lzPNeu0w+zm9lNRMqEACEz0igi16p2qBuY38xvo9jh+rITZpu1iZzVooird7p06RIqKyvx+9//HqFQqHsX67D33nsPP//5z/H1r38dixYtQr9+/eJq3KuvvgoAePrpp7uDGAAGDx6M5557Dt/85jfx+uuv45vf/CacTnUfbl+zZg22b9+O73//+2hra9P9dUSxWOkRJW6gYh7nTp/BjJHi7RzPANces5vZTWQUs9YFHsmFGbniZaRRRK5VrVw3ML+Z30az8vVlR8w2axM5q0Wh+jcOBoNYsGAB/vd//xfBYBBZWVmYNWsW5s2bh3nz5mHWrFnIyspCMBjE7373OzzxxBPXhLUSLS0tOHjwINxuN2bOnHnN9ydPnowhQ4bg9OnT2Ldvn6r3bmhowIsvvojS0lLMnTtX99cRKWGFR5S65ADOd3Vi3fH9eLZ6HRZ89BaerV6Hdcf343xXpy0fuxFdZ2dn987xbqfU5zGxdo7XAwNcW8xuZjeREaxQF4iYkUYRuVa1at3A/GZ+i8Cq15edMdusS+SsFoXqWzu///3vsW3bNng8HixevBj33XffNet1hUIhvPPOO6isrMS2bdvw+9//Hl//+tdV/ZyamhoAwKhRoyLe1R43bhxOnTqF2tpaTJw4UdH7hkIhLFq0CLIso7KyUvFaY/G+jkgpsz+ixJ2JE2Pkp+Xi3TleT+EAV1J0hwNcEn/bC8Mwu5ndZByzfho6UVapC0TMSKOIXKtatW5gfjO/RaD2+uoKypBCQUX9ol0z0mjMNusSOatFofqsXrduHRwOB77//e/jvvvu6/MYh8OB+++/Hw6HA0uWLMEf//hH1WHc2NgIAMjOzo54zLBhw3odq8SqVatQXV2Np556CsOHK/9Hj/d1REqZ/RElpRuoVE6qEK7tRuqSA+iU/djQWIOqU8fQEfAhzeVB2ZACzMgtTloBkvLFRjkV+SWYnV9ieDGarAC3S/HN7GZ2k7aU9B2i9O9GsVJdIFpGGkXkWtWqA3/mN/NbBGqurymZBdh95gTGDcqO2gfYPSOTIVatwmyzJpGzWhSqf+PDhw/D5XJhzpw5MY+dM2cOli1bhsOHD6tu2MWLFwEAXq834jFpaWkAgI6ODkXveeLECbz00ksYO3Ysvv3tbytuS7yvS0R7ezt2797d/d89/z9ZV3ZeLhaMvg2v1G7ts9NyOyUsGH0bLrZdQH1DTcT3Sfb5kn1DHrZeaFC0gcqGhhrcdl0emk80JKl14srOy8VnkoxfH9oa8dNyC0bfhoEBCc0NygcdaonYv3i9XkwfXqgowO8cdiPq6w6js7NT8ftn5+XCk5GG95vrUNXao/jOKsD07EL4Pu/Q9W+ebMxuY7IbEPP6ovgp7TvU9O9hVjpXWBfoy8hzRataVWt61w1GYX5z7C0CNddXefaNeK12K5o72iL27aKMgbQi2vlit3GOmSTrXBE1q0Wh+nmwS5cuwev1wu2OvVZQSkoKUlNTcenSpbgap6Xw41WBQACVlZWQpL7XQtLqdUTxaG5oxMCAhGUTv4I7swuR7r68tlO624M7swuxbOJXDC8IvF4vsm/IQ/G4UJe/AQAAIABJREFUsbhp/HgUjxuLgQMHoqr1mKLX7zh9DAP699e3kSbg9XrhyUi7pgDsyR+U8UrtVngy0qIOTKyos7MTvs87sGD0bVHXrlsw+jb4Pu9QPQn+mSRjyZ4/Y3NzXfdjnh0BHzY312HJnj/jnEtGdl6uJr+LCJjdzG4z6CtfsvPyhOn/lPYdNwwvsH3/PqB/f9YFFiVqrapn3WAk5jfzWwTh62thcfQ1pecXTcOhthY0drRF7Ns5BtKXGcY5otd7ViBqVotC9SfCs7Ky0NTUhOPHjyM/Pz/qsUePHsWFCxeQm6v+IktNTQWAqEVK+G50+O50NG+88QZ27tyJJ554AkVFRYrbEe/rEpWeno7CwsLuO0alpaVJ+9kkhtkFN2FOwU3XPqLkTcewrCF9vkbv86XXI2z1W7vvLr8wZY6qDVRSXG7bn9M+OYB1x/cr+rTc5pOHMbuwRPPHlszQv1wvBxStXRfpmujL+a5OVO5cH7P4rpxUgdIs7f42dXV1aG9v1+z91GB2J0c4uwFzXF8iiZQvZUMKMGOkGI9IK+07fjJ5Nt5tOKi4f5+WmYvOEw2WOlfkYJB1gQ5E61fiqVX1pnXdYGR2A8zvZOHYW5lBAT9+MPFubGk+jB2nj3VfX1MyC1CefSMOtbXgrfpdACL37SKMgbQi4vli1DhHKTPUe3ow8lwRMav1Fiu7VZ9hU6dOxTvvvIMlS5bgtddeg8fT926kPp8PP/zhD+FwODB16lS1PwY5OTkAgObm5ojHtLS09Do2mvfffx8AsG3bNuzcubPX95qamgAAmzZtwpEjR5CamopXX301odcRJarnulwibOYTbdOri4EuS25QpCcHgKpTxxQdW9V6FLPzS3Rtj6i0XrvOJwewobFGUfG9sbEWs/PFLb7VYHYrex0ZwwybKqrpOxxwqOrfv5o7Bs2w1rIgVt24kHoTrVYFrLfmLfNb2esoORwOB1Yd3oHJWQVYMvFueCU3OmU/9p9txGu1W9HY0dZ9bKS+nWMg/Yg+zjFDvWdFIma10VSfXfPmzcO6detQXV2NiooKPPbYY5g8eTKGDBmCrq4uNDc3Y8eOHXjjjTfQ2toKj8eDefPmqW5YcXExAODIkSO4dOlSn7tXHzhwAAAwevRoxe+7d+/eiN9rbW1Fa2srMjIyNHsdkVVE2/Tq47NNmJJVgA+arbVBkZ4kh1PVp+Ukh71DS6sAt2vxzexW9zpKLjNsqqim73A7JVX9u9uCgz2rblxI5mGVgT/zW93rSF8hAHkZA7Gqvhqoj35spL6dYyD9iD7OMUO9R/ag+uzKy8vDf/7nf+LJJ5/E8ePHsWzZsj6PC4VC8Hq9+NnPfoa8vDzVDRs2bBjGjBmDgwcP4r333rtmg5Dq6mq0tLQgMzMTEyZMiPl+K1eujPi9l19+GStWrMDDDz+MJUuWaPI6IiuJdXd5y8kjmDd6Gra2cGdipfhpOWPYtfhmdit7HSWf6J9eClPTd3TKflX9u18OJNo84XgkF2bkFivaWI11AVFkzG9lr6Pk0KJv5xhIPyKPc8xS75E9xHXm33777Vi/fj3uuecepKenIxQK9fpfRkYG7r33Xqxfvx6333573I2bP38+AGD58uU4fvx499fPnj2LpUuXArh8l9zpvPJrrFq1CjNnzsQzzzwT988lot5i3V1u6DiHQ20tmF80LeoGKgvHlMMrxd7sxw7Cn5ZTgp+W0064+FYiXHxbBbObRKT200tGUdN31Jw7qap/bzt/PoGWicsrubFwTPSN1VgXEMXG/CaRJNq3cwykH5HHOWap98ge4r7FkpeXhx//+Mf48Y9/jIaGBnz22WcAgEGDBsV1F7ovM2fOxIMPPog1a9Zg1qxZmDp1KlwuF7Zv34729nZMnz4dc+fO7fWac+fO4ejRo8jMzNSkDUSk7O7yW/W78MDIm/GDiXfjw5NHom5QRPy0nFHs/rg+s5tEI/Knl3pS03e0+32YkaO8f2+qt+aAL0VyYUTGYEUbFxJRdMxvEkWifTvHQPoReZxjlnqP7EGTXiUvL0+zAL7ac889h9LSUqxevRrV1dUIBoMYMWIE7r33Xjz44IO97kgTkT6UPMIWRAhv1u9EYf8sPF78JUtsUKS38CcqVhzse600flpOeyy+r2B2kwjM8oi0mr5jclaBqv69s7NTz6YbymobFxKJgPlNRku0b+cYSB8ij3PMUu+RPThCoVDI6EbQFXV1dWhvb0d6ejoKCwuxe/duAEBpaanBLSMz0Ot88ckBrDu+X9Hd5ek5RVzTS4UuOYBO2W/Ip+Xs2r90yQF8+vmZmMW31juWX92/k3X09W9r1+tLLTPli9q+Q2n/znOFlOK5knzMbmvj2Ns4Ro6BtCLi+WLUOCcWM9V7ehDxXLGyWNmt+szasWMHFi1ahLKyMlRWVkY99umnn8bevXvxwgsv4Oabb1b7o4hIECLfXTY7flou+ez4uD6zm0RlpnxR23ewfyeiRDG/yaqYkfoQdZxjpnqPrE/12bV+/Xo0NzfjjjvuiHns7bffjv/7v//D+vXrGcZEJsdH2PTVs+DjY2D6s1vxzewmkZkpX+LpO9i/E1G8mN9kdcxI7Yk6zjFTvUfWpvoq2LdvHxwOB2655ZaYx375y1+Gw+HAnj174mocEYlD1LvLRIlQUnz75AAcgDBFZDyY3SQyM+YLB+5isUI/TdQX5jcB7OMoPqLVKmas98iaVJ9hJ0+eREZGBlJTU2Mem5aWhuuuuw6nTp2Kq3FEJBZR7y4T6SG8duGGxhpUnTqGjoAPaS4PyoYUYEZusakKNWY3iY75QvGwUj9N1Bfmt72xjyOrYb1HIojrTPP5Yu/02vNYh8MRz48hIkGJdneZSGtdcgCfXjiDFTW9H93rCPiwuakOH56sN2SjmUQwu8kMmC+klBX7aaK+ML/tiX0cWRnrPTKS6jMuJycHPp8PBw8ejHnsJ598gkuXLmHo0KFxNY6IiMgInbL/moFHT/6gjBUHt6BT9ie5ZfFhdhOR1VitnybqC/PbvtjHERHpQ/VE+K233opQKITly5dDliPv9irLMpYvXw6Hw4Fbb701oUaSvfnkALrkAORgEF1yAD45YHSTiMjCfHIAGxprou5oDlwegGxsrEWXCfokZjfpiTlNyWbFfpqoL8xve2IfR0R2lKwxheqJ8Mceewz9+vVDVVUVvvWtb+HAgQPXHPPxxx/jscceQ1VVFVJSUvCtb31Lk8aSvXTJAZzv6sS64/vxbPU6LPjoLTxbvQ7rju/H+a5OBj4R6cIBoOrUMUXHVrUe1bUtWmF2kx6Y02QUK/bTRH1hftsT+zgispNkjylULyY1dOhQvPDCC3jqqaewc+dOfOMb30D//v2RnZ0NAGhubsb58+cRCoUgSRKef/555OTkaNposj6uiUZERpEcTnQElK3H2e73QXKIv64ds5u0xpwmI1mxnybqC/PbntjHEZFdGDGmiKvHnDFjBt544w2MHTsWoVAIbW1tqKmpQU1NDdra2hAKhVBSUoKVK1fi7rvv1qShZC9cE42IAGOWXJBDQaS5PIqOTXd7IIeCOrdIG8xu0pJWOc1lVSgeVu2nifrC/LYfpX1cXtpAPHZjGQKhIHOUiEzJiLm/uKfTJ06ciN/97nf49NNPsX//fpw5cwYAMHjwYNx0000YMWKEZo0ke1G7Jtrs/BJ+2ozIYrrkADplPzY01qDq1DF0BHxIc3lQNqQAM3KL4ZXcul33IQBlQwqwuaku5rFlWcN1aYNemN2kBS1y2shrnMzPyv00UV+Y3/YSq49zwoEHRt6MogFDsaX5MH736V7mKBGZjlFzfwm/w4gRIxi8pCm1a6LNzi/RtT1ElFxGL7ngkVyYkVuMD0/WRw1lt1PCXbmjTTnIYHZTIhLNaaOvcTI/O/TTRH1hfttDrD7ugZE3Y0CKFz/a8y5zlIhMy6i5Py4mRcLhmmhE9ibC0kheyY2FY8rhdkp9ft/tlLBwTDm8klu3NhCJKtGcFuEaJ/NjP01EVhapj8tLG4iiAUPx+qGPmKNEZGpGzf2pfpe2tjb8/e9/x/79+6/53qlTp/C9730Pt956KyZNmoQnn3wSp06d0qShZB9c95HIvtQ+HqX1DtJhKZILIzIGo3JSBabnFCHdfblPSnd7MD2nCJWTKkz1KRtmN2kpkZwW5Ron87NaP03UF+a3fUXq4+7MKcSW5sPMUSIyPaPm/lRXhu+88w5+/vOf49FHH8VNN93U/XWfz4e5c+eisbERoVAIAPCXv/wFBw8exNq1a5GamqpJg8n6uO4jkX2JtDRSiuRCiuRCRX4JZueXQHI4IYeCCOHyI6tmwuwmLSWS0yJd42R+VuqnifrC/La3vvq4QCiI3326V9HrmaNEJDKj5v5UfyJ869atAIBZs2b1+vof/vAHNDQ0oH///li6dCl+8pOfYMiQIThx4gRWrVqlTWvJFsJrokV61DWM6z4SWY+ISyN5vhiESE4nUiSXKSdXmN2kpURyWsRrnMzPCv00UV+Y3wT07uNczFEisgij5v5U94pNTU0AgJEjR/b6+nvvvQeHw4GnnnoK999/P+bMmYPnn38eoVAI77//viaNJfvguo9E9sSlkfTB7CatxZvTvMaJiJRjftPVmKNEZCVGzP2pnk7/7LPPcN1118HjudL5BgIB7Nu3D06nEzNnzuz+ellZGSRJwtGjR7VpLdlGeE205yfPRsvFCxjizUA/yY1Lsh+nOj/H0NTr4HG6+GlwIovh0kj6YHaT1nquXbqxsRZVrUfR7vch3e1BWdZw3JU7Gl7JfU1Ox7rG89IGonzYKJRcn4NUVwqCoRB8coCf8CUiW2J+09VYKxvLJwfgwOUn3IrHjUVb23mjm0RkavGOKRKh+p3+f/buPLqq+t7//+tMCQkJgxDCkEigIJIwiAEaHNAWhDiAVnrrjaBoEbwiaFu52mqXBW7RcsXeqqgXUUEZ8lNXa4V6kclV8F6SBsUoJBihApJAmKeEcHKG/fuDbyIx0znJGXeej7W6imd/zj6fwGd/Xp/93jl7G4ah8+fP13mtqKhITqdT6enpSkxMrH3dYrEoISGhXnvAV17D0OcnDir/yH5Vup1qb49VVnKaxselh7trAIKg5utRWw/vbfIhQNwayT9kN4KhJfdnbuwYt8qif+03XFd26q6/H/pa7+//om7up6QHfBEMAJGO/Mb3sVYOj2qPW1Uel9aXFtetTXRL0/hOiaxRgFYI9TNf/L41Svfu3eV2u/XVV1/Vvlbz9avhw4fXaev1elVZWanLLrusld1EW1Ptceubs8f11PY12lxWUnsftEq3U5vLSvTU9jX65txxnoINmBC3Rgo8shvB5O/9mRs6xv+133B1ionTf+z4H318iNwHAIn8RsNYK4dWk7WJQ6xRgEAJ1TNf/C6EZ2VlyTAMzZ07V19++aU2b96s1atXy2Kx6Ec/+lGdtnv37pXb7Vb37t0D1mG0DVUelxYXb2n0KrfL69Hioi2q8rhC3DMAwXbp16PG9rpSCY6LXwdOcMRqbK8rtWDERPVN7MpvXfiB7EYk+f4xPqBjsq7s1F1Lv/o/ch8ALkF+oyGslUOL2gRgLn7PjNOnT9ff/vY3ffHFF7rrrrskXfzK1tVXX61Ro0bVafvxxx/LYrFo2LBhgekt2gSnx631pcVNftVLuhg4G0p36/beQwh5wGRC/fUosyO7EWkuPca9hldrv91J7gPA95DfaAxr5dCgNgGYj9+/EZ6SkqK3335bI0aMUGxsrLp06aI777xTr7zySp12Ho9H7777rgzDqBfSQFMskvKP7Pepbf5RHgYDmFmovh5ldmQ3IlWszS6bxUruA0ADyG80h7VycFGbAMynRbNkRkaG3n777SbbWK1WffDBB5KkhISElnwM2iibxVp7363mVLicsln8vp4DAG0O2Y1IRe4DQOPIbyB8WKMA5hO0o9RisSgxMVGJiYmyWCz1tv/0pz/V2LFjg/XxiGIew6v29lif2iY4YuUxvEHuEQC0DWQ3woHcB4DWIb+B4GCNAphP2C5XlZeXq6ysLFwfjwhmSMpKTvOpbVa3PkHtCwDgO2Q3goHcB4DgIr+BlmGNApgP39tAxIm12TU+JV0Oq63Jdg6rTeNSBvIwCgAAohi5DwAAIhFrFMB8KIQjIsXZHJqVcUOjgeOw2jQr4wbF2Rwh7ll0c3rcqva45fF6Ve1xy+lxh7tLAACQ+4APel6eqvTBg1jHASbG+VrkYY0CmAuXqxCRYmx29U3sqgUjJmpD6W7lH92nCpdTCY5YZXXro3EpAxVnc3DF1UfVHreqPC6tLy1W/pH9qnQ71d4eq6zkNI1PSefvEgAQVuQ+0LiaddwnZw8qf+8nrOMAE+J8LXI1tUb5YVKaxqfy7wNEE45URKwYm10xNrsm9h6i23sPkc1ilcfwytDFryjBN9Uet745e1yLi7fI5fXUvl7pdmpzWYm2Ht6rWRk3qG9iV8IbABA25D5QH+s4wPw4ziNfQ2uUardLp8+cVseYuHB3D4AfuDUKIl7s/wsdm9WqGJudk2E/VXlc9RZVl3J5PVpctEVVHleIewYAQH3kPvAd1nGA+XGcR49L1yjFO3fp0Lel4e4SAD9RCAdMzOlxa31pcaOLqhour0cbSnermnvQAQAARATWcYD5cZwDQGhRCAdMzCIp/8h+n9rmH90X1L4AAADAd6zjAPPjOAeA0KIQDpiYzWJVpdvpU9sKl1M2C1MCAABAJGAdB5gfxzkAhBazKGBiHsOr9vZYn9omOGLlMbxB7hEAAAB8wToOMD+OcwAILQrhgIkZkrKS03xqm9WtT1D7AgAAAN+xjgPMj+McAELLHq4Pzs7OVmVlZbg+HmgTYm12jU9J19bDe5t8AIvDatO4lIGKsYVtSgAQBchuAAgd1nEIFPI7cnGcA0BohW0W/e1vfxuujwbalDibQ7MybtDioi0NLq4cVptmZdygOJsjDL0DEE3IbgAILdZxCATyO7JxnANA6PhdCD979qw8Ho86d+7sU/vCwkK5XC6NGDHC784BaL0Ym119E7tqwYiJ2lC6W/lH96nC5VSCI1ZZ3fpoXMpAxdkc/HYBYGJkNwBEp0vXcesPFusfx/azjmtDyO+2gfM1AAgdn2fS3NxcvfnmmyotLZUkdenSRZMmTdL06dOVkJDQ6PtmzZqlkydPqri4uMWdXLt2rXJzc1VSUiKv16s+ffpo0qRJysnJkdXautucv/POO3r66aclSZMnT679cw2v16vCwkJt2bJF//jHP/TPf/5T58+fV8eOHZWRkaG77rpLY8eObVUfgGCLsdkVY7NrYu8hur33ENksVnkMrwxd/DoeAHMiu8luANGvZh13fWKqJqQOUozdwTrO5MjvtpffnK8BQGj4NKM+++yzevvtt2UYRu1rx48f12uvvaa//e1vev7553XVVVc1+v5L3+evefPmafXq1YqNjdWoUaNkt9uVl5en+fPnKy8vTy+++GKLA7msrEwLFy6UxWJptI8HDx5UTk6OJKlTp04aMmSIOnTooIMHD2rr1q3aunWr7rzzTj3zzDOyWCwt/jmBULh0EWXjWbmAqZHdZDcAczl08KAOHTyozMxM1nEmRn637fzmfA0AgqvZQnhBQYHeeustSVJWVpbuuOMOderUSf/4xz/0zjvvqKysTPfee6+ef/553XTTTQHt3Pr167V69WolJSVp5cqVSktLk3RxIXDvvfdq48aNWrFihaZOner3vg3D0FNPPSXDMHTHHXfo/fffb7CdxWJRVlaWpk2bpmuvvVY2m612W0FBgR588EH95S9/0fDhwzVp0qQW/ZwAAAQS2U12AwCiD/lNfgMAgqvZS4y5ubmSpJtvvlnLly/XHXfcoRtvvFFPPPGE1q1bpxEjRqi6ulq//OUv9de//jWgnVuyZIkkac6cObVBLEldu3bV3LlzJUlLly6V1+v1e9+5ubnKy8vTr371K/Xq1avRdpdffrneeustjR49uk4QS9LIkSM1ffp0SdKaNWv87gMAAMFAdpPdAIDoQ36T3wCA4Gq2EF5YWCiLxaJf/epX9bYlJydr+fLluuuuu+R2u/Xkk09q1apVAelYeXm5ioqK5HA4lJ2dXW/7yJEjlZycrGPHjqmwsNCvfR88eFDPPfecMjMzNWXKlFb1Mz09vba/AABEArK7aWQ3ACASkd9NI78BAK3VbCH8+PHjiouLU2pqaoPbbTab5s2bpxkzZsjr9er3v/+93njjjVZ3rOYBH/3791e7du0abDN48GBJ0u7du33er2EYevLJJ+XxeLRgwYJW31ts//79kqRu3bq1aj8AgsPpcava45bH61W1xy2nxx3uLgFBR3Y3jewGzInMR7Qjv5tmtvxmzgKA0Gv2HuF2u10ej6fZHf3qV79SXFycXnjhBS1atEhVVVWaNWtWiztW84Tsnj17NtqmR48eddr6YuXKlSooKNBjjz2mPn36tLh/klRVVaUVK1ZIksaNG9eqfX1fRUWFPvvss9r/vvTPQHMYL1LP1BTFJrbXpkMlyj+6X5Vup9rbY5XVLU1jew6Q81ylDh30fe4wM8aL+ZDdjQtldkscX/AdY6Xl2lrmM1bMi/xunJnOvdvanGV2zMnwFWMlMjRbCE9NTdWePXu0b9++ZsProYceUlxcnP7whz/o5Zdf1oULF1rcsfPnz0uS4uLiGm3Tvn17SVJlZaVP+/z222/1/PPPa9CgQZo2bVqL+1Zj3rx5Ki0tVb9+/XTXXXe1en8AAqNnaopO2jx6dceHcnm/O5modDu1+VCJtpbv1cyB16tnagqLTJgS2d04shswFzIfZkJ+N84s+c2cBQDh1WwhfMiQIdqzZ4/+/ve/+3QV97777lNMTIz+4z/+Q2+88YYMw2j1V6ACoeZrWW63WwsWLKj38A1/vfzyy3r//feVmJioP/3pT4qJiQlQTy9KSEjQgAEDaq8YZWZmBnT/MCfGy0Vnqqu0YPuaOovLS7m8Hr2y+xMtGDFRmd3a7t8V4yW0SkpKVFFREZLPIrsbFqrslji+4DvGSuu0pcxnrIReKLNbIr8bY6Zz77Y0Z5kdczJ8xVgJreayu9l7hI8ePVqGYeidd96RYRg+fejdd9+tZ555RlZrs7tvVHx8vKSLX4FqTM3V6Jqr0015++23tX37ds2YMUNXXnlli/slScuWLdOLL76o+Ph4LV26VP3792/V/gAEjtPj1vrS4kYXlzVcXo82lO5WNffigwmR3fWR3YD5kPkwG/K7PjPlN3MWAIRfs78RfuONN2rUqFFyu9364osvdNVVV/m045/85CeKj4/Xf/7nf7aoY7169ZIkHTp0qNE2NU+LrmnblE2bNkmStm3bpu3bt9fZVlZWJknauHGj9uzZo/j4eC1ZsqTB/axYsUJ/+MMf1K5dOy1ZskTDhg1r/ocBEDIWSflH9vvUNv/oPt3ee0hQ+wOEA9ldF9kNmBOZD7Mhv+syW34zZwFA+DVbCI+JidGyZctatPPx48dr/PjxLXpvenq6JGnPnj26cOFCg0+v3rlzpyRp4MCBPu/3888/b3Tb0aNHdfToUSUmJja4fdWqVfr973+v2NhYvfrqqxo5cqTPnwsgNGwWqyrdTp/aVricslla/tszQKQiu79DdgPmRebDbMjv75gxv5mzACD8/J5ZN2/erM2bN+vkyZPB6E+tHj16KCMjQy6XSx999FG97QUFBSovL1dSUpJPV4ZXrFihkpKSBv9X84TtyZMnq6SkRJ9++mm99+fm5mr+/PmKiYnRyy+/rGuuuab1PySAgPMYXrW3x/rUNsERK4/hDXKPgPAju8luwIzIfJgd+W2u/GbOAoDw87sQ/vDDD+vRRx9VbKxvE3hrzJgxQ5K0aNEiHThwoPb1EydOaN68eZKk6dOn17kf2sqVK5Wdna3HH388YP149913NW/ePMXExGjx4sW6/vrrA7ZvAIFlSMpKTvOpbVa35h9CBJgB2U12A2ZE5sPsyG9z5TdzFgCEX7O3Rvm+jh07SvLtIRmtlZ2drZycHOXm5mrChAm65pprZLfblZeXp4qKCo0dO1ZTpkyp855Tp05p3759SkpKCkgfdu/eraefflqGYSglJUXr1q3TunXr6rXr3LmznnjiiYB8JoCWi7XZNT4lXVsP723yQTQOq03jUgYqxub3NAhEHbKb7AbMiMyH2ZHf5spv5iwACD+/Z9b+/fvr888/V0VFhRISEoLRpzrmzp2rzMxMrVq1SgUFBfJ6verbt68mTZqknJycVj0d2xdnz56tfWL3N998o2+++abBdr169YrKMAbMKM7m0KyMG7S4aEuDi0yH1aZZGTcozuYIQ++A0CO7yW7ArMh8mBn5bb78Zs4CgPCyGDVJ46M1a9bo8ccf16OPPqqHHnooWP1qs0pKSmoXOgMGDNBnn30mScrMzAxzzxANGC/fqfa4VeVxaUPpbuUf3acKl1MJjlhldeujcSkDFWdztPnfsmC8hNb35/dQIruDq6F/W44v+Iqx0nptJfMZK6EXzuyWyO9gC9e5d1uZs8yOORm+YqyEVnPZ7ffsOnHiRH355Zd66aWX5HQ6dd9996lTp04B6SwABEqMza4Ym10Tew/R7b2HyGaxymN4Zeji1xKBtoTsBmBmZD7Mivw2J+YsAAgfv2fZe++9V5LUrl07LVmyRK+//rouv/xyXXbZZY1+Vcpiseitt95qXU8BoAUuXUza/H8+MGAKZDeAtoDMh9mQ3+bGnAUAoed3IbygoKDOf7vd7ibv3yVdDGMAABAeZDcAANGH/AYAILD8LoQ/++yzwegHAAAIErIbAIDoQ34DABBYfhfCf/KTnwSjHwAAIEjIbgAAog/5DQBAYIXtRlSzZ8/W1KlTw/XxAADAT2Q3AADRh/wGAOCisD2S+PPPP9eJEyfC9fEAAMBPZDdAmqMfAAAgAElEQVQAANGH/AYA4CIeTQwAAAAAAAAAMDUK4QAAAAAAAAAAU6MQDgAAAAAAAAAwNQrhAAAAAAAAAABToxAOAAAAAAAAADA1CuEAAAAAAAAAAFOjEA4AAAAAAAAAMDUK4QAAAAAAAAAAU6MQDgAAAAAAAAAwNQrhAAAAAAAAAABTs4frg6+66iqdOXMmXB8PAEHh9LhlkWSzWOUxvDIkxdrCNtUCAUV2A41j/gcQqcjv4GDeB4DoE7ZZevHixeH6aAAIuGqPW1Uel9aXFiv/yH5Vup1qb49VVnKaxqekK87mUAwLY0Q5shuoj/kfQKQjvwOLeR8AoleTs/Nf//rXgH3QHXfcEbB9AUAkqfa49c3Z41pcvEUur6f29Uq3U5vLSrT18F7NyrhBfRO7sihG0JHdQOgw/wMIFPI7OjDvA0B0a3Jm/vWvfy2LxRKQDyKMAZhVlcdVbzF8KZfXo8VFW7RgxEQWxAg6shsIHeZ/AIFCfkcH5n0AiG5NzswjRoxodNtXX32lc+fOSZJ69Oih5ORkSdLRo0d16NAhSVKHDh00YMCAQPUVACKO0+PW+tLiRhfDNVxejzaU7tbtvYewKEZQkd1AaDD/Awgk8jvyMe8DQPRrclZesWJFg68vXLhQ27dv109/+lM9+OCDSk1NrbO9tLRUr732mt59910NGjRITzzxROB6DAARxCIp/8h+n9rmH92n23sPCWp/ALIbCA3mfwCBRH5HPuZ9AIh+fl+e/OCDD7R8+XJNnz5djz32WINtUlJSNH/+fHXq1ElLly5Venq6JkyY0OrOAkCksVmsqnQ7fWpb4XLKZrEGuUdAfWQ3EHjM/wCCjfyOLMz7ABD9/J6ZV69eLavVqgcffLDZtjNmzJDVatWqVata1DkAiHQew6v29lif2iY4YuUxvEHuEVAf2Q0EHvM/gGAjvyML8z4ARD+/C+F79+5VQkKCEhISmm1b027v3r0t6hzQFjg9blV73PJ4var2uOX0uMPdJfjBkJSVnOZT26xufYLaF6AxZHdkIweiE/M/gGAjvyML8z6AUOIcITj8vjWKxWLRuXPndOLECXXp0qXJtidOnNDZs2fVvn37FncQMKtqj1tVHpfWlxYr/8h+Vbqdam+PVVZymsanpCvO5uDhKlEg1mbX+JR0bT28t8kH5zisNo1LGci/KcKC7I5M5EB0Y/4HEGzkd2Rh3gcQCpwjBJffvxGenp4uwzD03HPPNdv2ueeek2EYGjRoUIs6B5hVtcetb84e11Pb12hzWUntveYq3U5tLivRU9vX6Jtzx1XNFb+oEGdzaFbGDXJYbQ1ud1htmpVxg+JsjhD3DLiI7I485IA5MP8DCCbyO/Iw7wMIJs4Rgs/vQvgDDzwgwzD0wQcf6P7779e2bdt04cKF2u1Op1Pbtm3Tz3/+c33wwQeyWCx64IEHAtppINpVeVxaXLyl0d8kcHk9Wly0RVUeV4h7hpaIsdnVN7GrFoyYqLG9rlSC4+K9AxMcsRrb60otGDFRfRO7ctUWYUN2Rx5ywByY/wEEE/kdeZj3AQQT5wjB5/fsPHr0aM2ZM0fPP/+88vPzlZ+fL6vVWnvfsoqKCnm9XhmGIYvFojlz5ui6664LeMeBaOX0uLW+tLjJr9NJFye4DaW7dXvvISykokCMza4Ym10Tew/R7b2HyGaxymN4Zeji1yiBcCK7Iws5YC7M/wCChfyOTMz7AIKBc4TQ8Ps3wqWLV6ZXrFihkSNHSpI8Ho/OnDmjM2fOyOO5+A82atQorVy5UtOmTQtcbwETsEjKP7Lfp7b5R/cFtS8IvNj/tzC2Wa2KsdlZDCNikN2RgxwwJ+Z/AMFAfkcu5n0AgcQ5Qmi0eKYePny43nrrLZ05c0bFxcU6deqUJKlz585KT09Xx44dA9ZJwExsFmvtfZ6aU+FyymZp0fUqAKiH7I4M5AAAwB/kNwCYH+cIodHqS5YdO3bUqFGjAtEXoE3wGF61t8f6NMElOGLlMbyytezLGwDQILI7vMgBAEBLkN8AYF6cI4QGf2NAiBmSspLTfGqb1a1PUPsCAAg9cgAAAADApThHCA0K4UCIxdrsGp+SLofV1mQ7h9WmcSkDefgBAJgMOQAAAADgUpwjhEaTf2sDBw6UJPXt21cffvhhndf8YbFYVFxc3ILuAdHP6XHLItV5mniczaFZGTdocdGWBp8I7LDaNCvjBsXZHCHvL4DoRnZHB3IAQGs1tMbkYX3Ri/w2L45VAL7iHCH4mpx9DcOo8//f/7OvWvIeINpVe9yq8ri0vrRY+Uf2q9LtVHt7rLKS0zQ+JV1pCV20YMREbSjdrfyj+1ThcirBEausbn00LmWg4mwOrvAB8BvZHR1ibHb1TexKDgDwW3NrTOaO6ER+mw/HKgB/cY4QfE3+zW3evPliI7u93msAGlftceubs8e1uLjuVbxKt1Oby0q09fBezcq4QT9I7KqJvYfo9t5D+A0BAAFBdkePGJtdMTY7OQDAZz1TU3xaY/ZN7MpJcpQhv83F1/NBjlUA38c5QnA1+TfYq1cvn14LtrVr1yo3N1clJSXyer3q06ePJk2apJycHFmtrbvN+TvvvKOnn35akjR58uTaPzdk69atWr58uXbt2iWn06nU1FTdeuutmjZtmmJiYlrVD5hLlcdVb9FzKZfXo8VFW7RgxER1jImrfZ0n/gJoLbK7rmjI7ksXtOQAgMbExcUpNrG9Fu/40Kc1JsW16BIJ+U12B44/54McqwAawjlCcET83+S8efM0Z84c7dq1S8OHD9c111yj/fv3a/78+XrkkUfk9XpbvO+ysjItXLhQFoul2bZLly7V9OnTlZ+fr/T0dN1www06ceKE/vSnP+mee+5RVVVVi/sBc3F63FpfWtzooqeGy+vRhtLdqva4Q9QzAAgNshsAAq9zUldtOlTCGhNBQXYHDueDABC5IroQvn79eq1evVpJSUlas2aNlixZopdfflkbNmzQD37wA23cuFErVqxo0b4Nw9BTTz0lwzB0xx13NNl2586dev755xUXF6fc3FwtX75cL774ojZt2qQRI0aosLBQ//Vf/9WifsB8LJLyj+z3qW3+0X1B7QsAhBrZDQDB0aljR+Uf3e9TW9aY8AfZHVicDwJA5GryOzh//etfA/ZBzYVeQ5YsWSJJmjNnjtLS0mpf79q1q+bOnat77rlHS5cu1T333OP3V7Vyc3OVl5en3/72tzp9+nSTbZcuXSrDMPTAAw9o6NChta+3b99ezz77rMaNG6fVq1dr1qxZ6tChg1/9gPnYLFZVup0+ta1wOWWzRPT1KABRhuy+iOwGYDYOm501pomFM7/J7sDifBAAIleThfBf//rXPn19yRf+hnF5ebmKiorkcDiUnZ1db/vIkSOVnJysI0eOqLCwUFdffbXP+z548KCee+45ZWZmasqUKVq8eHGjbaurq7V161ZJ0sSJE+ttT01N1VVXXaUdO3Zoy5YtmjBhgs/9gDl5DK/a22N9WvwkOGLlMbzc7wlAwJDdZDcAc3J53KwxTSxc+U12Bx7ngwAQuZoshI8YMSJU/ainuLhYktS/f3+1a9euwTaDBw/WkSNHtHv3bp8D2TAMPfnkk/J4PFqwYEGzi419+/apqqpKnTp10uWXX95oP3bs2KHi4uKIDmSEhiEpKzlNm8tKmm2b1a1P8DsEoE0hu8luAOZ0+swZZXVL0+ZDrDHNKFz5TXYHHueDABC5miyEt/Q+YIFQWloqSerZs2ejbXr06FGnrS9WrlypgoICPfbYY+rTp/nQqdl3zWc1pKaPZWVlPvcD5hVrs2t8Srq2Ht7b5ANSHFabxqUM5CnhAAKK7Ca7AZjTqWPHNbbPAG0tZ41pRuHKb7I78DgfBIDIFbEz7vnz5yVJcXFxjbZp3769JKmystKnfX777bd6/vnnNWjQIE2bNi1g/YiPj/erH76oqKjQZ599Vvvfl/4Zka9naopmDrxer+z+pMHFj8Nq08yB1+v86bPae7A44J/PeIE/GC8IFLK7ot7xxPEFXzFW4AvnucqwrjFhPm09u6XgnHuH+3wQoUN+w1eMlcjQZm5EVfPVLLfbrQULFshms4W7SzCxQwdL1dlt0/yrb9WYngOU4IiVdPEecGN6DtD8q29VZ7dNhw76/lsVANDWkN0AUBdrTEQ6svsijlUAiEwR+xvhNVd7q6qqGm1TcyW45gp1U95++21t375dDz/8sK688sqA9qPm6rUv/fBVQkKCBgwYUHvFKDMzM2D7RmjdnjZUd6QNlc1ilcfwytDFr8spLkE9uiUH9LMYL/AH4yW0SkpKVFFREe5uBBXZfTG7JY4v+I6xAl/VjJWa9WMo15htFdl9kZmzWwr+uTfHqjmR3/AVYyW0msvuFhfCy8vL9ec//1k7duzQ0aNHVVVVJcMwGmxrsVi0adMmv/bfq1cvSdKhQ4ea7MOlbZtS8/nbtm3T9u3b62yrucfYxo0btWfPHsXHx2vJkiV19n348OFG912zzZd+oO2JveSebzwNHEA4kd3fIbsBRDvWmG1HMPOb7A4+jlUAiBwtKoSvWbNGTz/9tJxOZ5MBXLOtuSdENyQ9PV2StGfPHl24cKHBJ1jv3LlTkjRw4ECf9/v55583uu3o0aM6evSoEhMTa1/r27ev2rVrp9OnT+vbb79t8AnWX375pd/9AAAglMjuushuAEA0CHZ+k90AgLbE78uRRUVFevLJJ3XhwgXdeeedWrx4sSSpY8eOWrZsmRYtWqQ777xTDodDnTt31nPPPae33nrL74716NFDGRkZcrlc+uijj+ptLygoUHl5uZKSkjRs2LBm97dixQqVlJQ0+L9Zs2ZJkiZPnqySkhJ9+umnte+LiYnR6NGjJV1chHzfwYMHVVhYKIfDoRtvvNHvnxMAgGAju+siuwEA0SAU+U12AwDaEr8L4cuWLZPb7dZ9992nBQsWaOzYsZIkh8OhUaNG6bbbbtMzzzyj9957TzabTX/605/8ujfYpWbMmCFJWrRokQ4cOFD7+okTJzRv3jxJ0vTp02W1fvdjrFy5UtnZ2Xr88cdb9JkNmT59uiwWi15//fXaq9DSxXulPfnkk/J6vbr77rvVoUOHgH0mAACBQnaT3QCA6BOq/Ca7AQBthd+3RtmxY4csFovuu+++JtsNGDBATz/9tB555BG99tprmjNnjt+dy87OVk5OjnJzczVhwgRdc801stvtysvLU0VFhcaOHaspU6bUec+pU6e0b98+JSUl+f15jRkyZIgee+wxLVq0SP/6r/+qrKwsJSYmavv27Tpx4oSGDh2qX/7ylwH7PAAAAonsJrsBANEnVPlNdgMA2gq/C+HHjx9XbGysunfvXvuazWbThQsX6rX98Y9/LIfDoU2bNrXoZFqS5s6dq8zMTK1atUoFBQXyer3q27evJk2apJycnDpXpYNp+vTpGjBggJYtW6adO3fK6XQqNTVV99xzj6ZNm6aYmJiQ9AMAAH+R3WQ3ACD6hDK/yW4AQFvgdyE8Pj6+3kM6EhISdPbsWVVVVSkuLu67ndvtiomJqX3KdEtNmDBBEyZM8Knt7NmzNXv2bL/27+t7Ro8eXXvfMgAAogXZTXYDAKJPqPOb7AYAmJ3fl3WTk5N17tw5OZ3O2tf69Okjqf6Tob/99ltVVlbKbve73g4AAAKE7AYAIPqQ3wAABJbfhfArr7xShmGoqKio9rXrr79ehmHoj3/8o44dOyZJOnnypH7729/KYrFo6NChgesxAADwC9kNAED0Ib8BAAgsvwvhN954owzD0EcffVT72pQpU9SlSxcVFRXpRz/6ka6//npdd911KigokNVq1UMPPRTQTgMAAN+R3QAARB/yGwCAwPK7ED5mzBj993//t8aOHVv7WseOHfXWW29p0KBBcrvdOnbsmLxer7p3764XXnhBw4cPD2inAQCA78huAACiD/kNAEBg+X0DsZiYGN144431Xu/Xr5/ee+89HT58WOXl5UpMTNQPfvADWSyWQPQTAAC0ENkNAED0Ib8BAAgsv38j/OOPP5bb7W50e48ePTRs2DD169ePIAYAIAKQ3QAARB/yGwCAwPL7N8JnzpypTp06ady4cbrttts0cuTIYPQLAAAECNkNAED0Ib8BAAgsvwvhCQkJOn36tN577z299957Sk5O1i233KLbbrtN6enpwegjAABoBbIbAIDoQ34DABBYft8aZdu2bXrppZd00003KTY2VuXl5Vq2bJkmTZqkW265Ra+88oq+/fbbYPQVAAC0ANkNAED0Ib8BAAisFj0s86abbtJNN92kyspKbdq0SWvXrlV+fr6++eYbvfTSS3rppZc0aNAgTZgwQbfccou6du0ajL4DAAAfkN0AAEQf8hsAgMDyuxB+qfbt2+v222/X7bffrlOnTmndunX68MMPtWPHDu3cuVO7du3SwoUL9cMf/lBvvvlmoPoMAABaiOwGACD6kN8AALReqwrhl+rcubPuvvtu3X333SovL9eHH36ov/zlL/rnP/+pvLy8QH0MAAAIELIbAIDoQ34DANAyft8jvDkul0u7du3Szp07VVZWFujdAwCAACO7AQCIPuQ3AAD+CchvhBuGofz8fK1du1abNm3SuXPnZBiGJKlnz5665ZZbAvExAAAgQMhuAACiD/kNAEDLtaoQ/sUXX+hvf/ub1q1bpxMnTki6GMydOnVSdna2brvtNg0fPjwgHQUAAK1HdgMAEH3IbwAAWs/vQvjevXu1du1a/c///I9KS0slXQzguLg4jRkzRhMmTNC1114ruz1gtx8HAACtQHYDABB9yG8AAALL78ScMGGCpIsBbLfbdd1112nChAkaM2aM2rVrF/AOAgCA1iG7AQCIPuQ3AACB5Xch3DAMjRgxQrfddpuys7PVsWPHYPQLAAAECNkNAED0Ib8BAAgsvwvhW7ZsUXJycjD6AgAAgoDsBgAg+pDfAAAEltXfNxDEAABEF7IbAIDoQ34DABBYfhfCAQAAAAAAAACIJhTCAQAAAAAAAACmRiEcAAAAAAAAAGBqFMIBAAAAAAAAAKZGIRwAAAAAAAAAYGoUwgEAAAAAAAAApkYhHAAAAAAAAABgahTCAQAAAAAAAACmRiEcAAAAAAAAAGBqFMIBAAAAAAAAAKZGIRwAAAAAAAAAYGoUwgEAAAAAAAAApkYhHAAAAAAAAABgahTCAQAAAAAAAACmRiEcAAAAAAAAAGBqFMIBAAAAAAAAAKZGIRwAAAAAAAAAYGoUwgEAAAAAAAAApkYhHAAAAAAAAABgavZwd8AXa9euVW5urkpKSuT1etWnTx9NmjRJOTk5slp9r+WvWbNGn3zyib766isdO3ZM586dU3x8vPr166dbb71Vd911lxwOR4PvLS8v19KlS/W///u/Onz4sAzDUI8ePZSVlaXp06crNTU1UD8uAABRj+wGACD6kN8AADOL+EL4vHnztHr1asXGxmrUqFGy2+3Ky8vT/PnzlZeXpxdffNHnQM7NzVVhYaH69eunwYMHKzExUUePHlVhYaF27NihtWvXatmyZYqPj6/zvuLiYk2dOlVnz55V9+7ddd1110mSdu3apXfeeUdr167VG2+8oauvvjrgPz8AANGG7AYAIPqQ3wAAs4voQvj69eu1evVqJSUlaeXKlUpLS5MkHT9+XPfee682btyoFStWaOrUqT7t7ze/+Y3S0tLUoUOHOq+Xl5fr/vvvV2FhoV5//XU98sgjdbbPnz9fZ8+e1c9+9jM9/fTTtVeuXS6Xfve73+nPf/6z5s6dqzVr1rT+hwYAIIqR3QAARB/yGwDQFkT0PcKXLFkiSZozZ05tEEtS165dNXfuXEnS0qVL5fV6fdrfkCFD6gWxJHXv3l0PPvigJGnbtm11tjmdTn3++eeSpNmzZ9f5+pbD4dAvfvELSVJJSYmqqqp8+8EAADApshsAgOhDfgMA2oKILYSXl5erqKhIDodD2dnZ9baPHDlSycnJOnbsmAoLC1v9eXb7xV+O//59yqxWa+22psTHx6tdu3at7kekcHrcqva45fF6Ve1xy+lxh7tLAIAIR3aHDjkNAAgU8jtykfcAEFgRWwgvLi6WJPXv37/RkBs8eLAkaffu3a36rJMnT+qNN96QJI0ZM6bONofDoaysLEnSSy+9JJfLVbvN5XLphRdekCRNmjRJFoulVf2IBNUet85UV+mDA1/o1wUfaOb//X/6dcEH+uDAFzpTXaVqghcA0AiyO/jIaQBAoJHfkYe8B4DgiNh7hJeWlkqSevbs2WibHj161Gnrq48//lgbNmyQx+PRsWPHtGPHDjmdTt15552aPHlyvfZz587VAw88oHfffVdbt27VoEGDJEk7d+7U2bNnNXXqVP37v/+7X31oTkVFhT777LPa/770z8HSMzVFJ20evfrVJ3J5PbWvV7qd2lxWoq2H92rmwOvV2W3ToYP+/Z0jtEIxXmAejBcECtldUe94CuTxRU6bG3MxfMVYQaCR36E/924KeR9dwj1eED0YK5EhYgvh58+flyTFxcU12qZ9+/aSpMrKSr/2/dVXX+n999+v89rUqVPr3YesRmpqqnJzc/XEE09o69atKi8vr902aNAgDR8+vMH3RZO4uDjFJrbXqzs+rBO2l3J5PXpl9yeaf/WtiouL475sAIA6yO7gIacBAMFCfkcO8h4AgitiC+HBNHPmTM2cOVPV1dU6dOiQ1q1bp9dee02bNm3Sa6+9pn79+tVpv2PHDs2ePVsJCQl65ZVXNGzYsNrXFy5cqNmzZ2v27NmaNWtWwPqYkJCgAQMG1F4xyszMDNi+G+L0uPXBgS8aDdsaLq9Hmw9/rdsHDFGMrU0On4gWqvECc2C8hFZJSYkqKirC3Y2oFU3ZLQX++CKnzYu5GL5irIQe2d160ZTfkXCMkffRIxLGC6IDYyW0msvuiL1HeHx8vCQ1eXWz5mp0zdVpf8XExCgtLU0PPfSQ/vCHP6isrEyPP/64DMOobXP27Fk9/PDDqqys1Ouvv64xY8bosssu02WXXaaxY8fq9ddfV7t27fTqq69q//79LepHJLBIyj+y36e2+Uf3BbUvAIDoRHYHDzkNAAgW8jtykPcAEFwRWwjv1auXJOnQoUONtqn5mlRN29YYN26cEhISVFRUVOe+Z3//+9918uRJDR06VKmpqfXe17t3bw0ZMkRut1sFBQWt7ke42CxWVbqdPrWtcDlls0Ts0AEAhAnZHTzkNAAgWMjvyEHeA0BwReysmZ6eLknas2ePLly40GCbnTt3SpIGDhzY6s+zWCzq1KmTpItPsq5x+PBhSVJiYmKj7+3QoYMk6fTp063uR7h4DK/a22N9apvgiJXH8Aa5RwCAaEN2Bw85DQAIFvI7cpD3ABBcEVsI79GjhzIyMuRyufTRRx/V215QUKDy8nIlJSXV3jesNQ4ePKiysjJZrdY6V5+7desmSSoqKpLL5ar3PpfLpaKiIklSSkpKq/sRLoakrOQ0n9pmdesT1L4AAKIT2R085DQAIFjI78hB3gNAcEVsIVySZsyYIUlatGiRDhw4UPv6iRMnNG/ePEnS9OnTZbV+92OsXLlS2dnZevzxx+vsa+/evVq7dq2czvpfM/r666/16KOPyjAM3XTTTbrssstqt40ePVpxcXE6dOiQnn32WVVXV9duq66u1u9//3sdPnxYHTt21PXXXx+YHzwMYm12jU9Jl8Nqa7Kdw2rTuJSBPJADANAgsjs4yGkAQDCR35GBvAeA4IroWTM7O1s5OTnKzc3VhAkTdM0118hutysvL08VFRUaO3aspkyZUuc9p06d0r59+5SUlFTn9RMnTmjOnDmKj49Xenq6kpOTVV1drbKyMu3evVuGYWjIkCG1IV+jS5cu+t3vfqennnpKq1at0saNG5WRkSFJ2rVrl44dO6aYmBg988wzTX6FKxrE2RyalXGDFhdtafAp1Q6rTbMyblCczRGG3gEAogHZHTzkNAAgWMjvyEHeA0DwRHQhXJLmzp2rzMxMrVq1SgUFBfJ6verbt68mTZqknJycOlekm9K/f3/94he/0Keffqp9+/apqKhIbrdbnTt31ujRo3XzzTdr4sSJstnqX3n9yU9+oiuuuEJvvfWWPv30U/3f//2fJCk5OVk//elPdf/996tfv34B/bnDIcZmV9/ErlowYqI2lO5W/tF9qnA5leCIVVa3PhqXMlBxNgdXnQEATSK7g4OcBgAEE/kdGch7AAgei2EYRrg7ge+UlJSooqJCCQkJGjBggD777DNJUmZmZkj74fS4ZdHFp1Z7DK8MXfyaFiJbuMYLohPjJbS+P7/DPBr6tw328UVOmwdzMXzFWAk9stvcIuXcuynkfeSKxPGCyMRYCa3mspsZFA26NFxtkX0reQAA2hxyGgAA8yPvASCwmEkBAAAAAAAAAKZGIRwAAAAAAAAAYGoUwgEAAAAAAAAApkYhHAAAAAAAAABgahTCAQAAAAAAAACmRiEcAAAAAAAAAGBqFMIBAAAAAAAAAKZGIRwAAAAAAAAAYGoUwgEAAAAAAAAApkYhHAAAAAAAAABgahTCAQAAAAAAAACmRiEcAAAAAAAAAGBqFMIBAAAAAAAAAKZGIRwAAAAAAAAAYGoUwgEAAAAAAAAApkYhHAAAAAAAAABgahTCAQAAAAAAAACmRiEcAAAAAAAAAGBqFMIBAAAAAAAAAKZGIRwAAAAAAAAAYGoUwgEAAAAAAAAApkYhHAAAAAAAAABgahTCAQAAAAAAAACmRiEcAAAAAAAAAGBqFMIBAAAAAAAAAKZGIRwAAAAAAAAAYGoUwgEAAAAAAAAApkYhHAAAAAAAAABgahTCAQAAAAAAAACmRiEcAAAAAAAAAGBqFMIBAAAAAAAAAKZGIRwAAAAAAAAAYGoUwgEAAAAAAAAApkYhHAAAAAAAAABgahTCAe08NJ8AACAASURBVAAAAAAAAACmRiEcAAAAAAAAAGBqFMIBAAAAAAAAAKZmD3cHfLF27Vrl5uaqpKREXq9Xffr00aRJk5STkyOr1fda/po1a/TJJ5/oq6++0rFjx3Tu3DnFx8erX79+uvXWW3XXXXfJ4XA0+v4LFy5oxYoV+uijj3TgwAG5XC516dJFgwYN0tSpU5WZmRmIHxcAgKhHdgMAEH3IbwCAmUV8IXzevHlavXq1YmNjNWrUKNntduXl5Wn+/PnKy8vTiy++6HMg5+bmqrCwUP369dPgwYOVmJioo0ePqrCwUDt27NDatWu1bNkyxcfH13vvwYMHNW3aNB04cEBJSUn64Q9/KJvNpkOHDmnz5s268sorCWMAAER2AwAQjchvAIDZRXQhfP369Vq9erWSkpK0cuVKpaWlSZKOHz+ue++9Vxs3btSKFSs0depUn/b3m9/8RmlpaerQoUOd18vLy3X//fersLBQr7/+uh555JE628+fP6+f//znOnjwoB577DFNmzZNNputdvupU6d0+vTp1v2wAACYANkNAED0Ib8BAG1BRN8jfMmSJZKkOXPm1AaxJHXt2lVz586VJC1dulRer9en/Q0ZMqReEEtS9+7d9eCDD0qStm3bVm/7q6++qm+//VaTJ0/WjBkz6gSxJHXu3Fl9+vTxqQ8AAJgZ2Q0AQPQhvwEAbUHEFsLLy8tVVFQkh8Oh7OzsettHjhyp5ORkHTt2TIWFha3+PLv94i/Hf/8+ZdXV1Xr33XclSffdd1+rPwcAALMiuwEAiD7kNwCgrYjYW6MUFxdLkvr376927do12Gbw4ME6cuSIdu/erauvvrrFn3Xy5Em98cYbkqQxY8bU2VZUVKTTp08rOTlZqampKioq0saNG3Xy5El16dJF1157rYYPH97izwYAwCzIbgAAog/5DQBoKyK2EF5aWipJ6tmzZ6NtevToUaetrz7++GNt2LBBHo9Hx44d044dO+R0OnXnnXdq8uTJddp+/fXXkqTk5GQtXLhQb775Zp3tr7zyisaOHavnnnuuwQd9tFRFRYU+++yz2v++9M9Acxgv8AfjBYFCdlfUO544vuArxgp8xVhBoJHfnHuj5Rgv8BVjJTJEbCH8/PnzkqS4uLhG27Rv316SVFlZ6de+v/rqK73//vt1Xps6dapmz55d7+tZZ86ckSTt3r1bX375paZOnaopU6aoU6dO2r59u+bNm6dNmzZp3rx5WrhwoV/9AADATMhuAACiD/kNAGgrIrYQHkwzZ87UzJkzVV1drUOHDmndunV67bXXtGnTJr322mvq169fbduah4G4XC5NnDhRTz75ZO22MWPGqFu3bvqXf/kXffDBB3r44Yd1+eWXB6SPCQkJGjBgQO0Vo8zMzBbtx+lxyyLJZrHKY3hlSIq1tcl/9jahteMFbQvjJbRKSkpUUVER7m5ErWjKbonjy6yCsa5irMBXjJXQI7tbL5rym2MM/mC8BJaZa1eMldBqLrsjdlTVfNWpqqqq0TY1V6Nrrk77KyYmRmlpaXrooYfUt29fPfLII3r88cf15z//WRaLpd6+f/azn9Xbx+DBg5WRkaFdu3apoKAgYGHcWtUet6o8Lq0vLVb+kf2qdDvV3h6rrOQ0jU9JV5zNoRiTTCoAgMhAdsOsWFcBMDPyG0C4sMZCqFnD3YHG9OrVS5J06NChRtuUl5fXadsa48aNU0JCgoqKiurc9ywlJaXBP1+q5vXjx4+3uh+BUO1x65uzx/XU9jXaXFaiSrdTklTpdmpzWYme2r5G35w7rmqPO8w9BQCYCdkNM2JdBcDsyG8A4cAaC+EQsYXw9PR0SdKePXt04cKFBtvs3LlTkjRw4MBWf57FYlGnTp0kXXyS9ff7IUmnT59u8L2nTp2SpIA+sKM1qjwuLS7eIpfX0+B2l9ejxUVbVOVxhbhnAAAzI7thRqyrAJgd+Q0gHFhjIRwithDeo0cPZWRkyOVy6aOPPqq3vaCgQOXl5UpKStKwYcNa/XkHDx5UWVmZrFarUlNTa19PTk7W0KFDJUl5eXn13nfmzBkVFxdLkgYNGtTqfrSW0+PW+tLiRieSGi6vRxtKd3NlDQAQMGQ3zIZ1FYC2gPwGEGqssRAuEVsIl6QZM2ZIkhYtWqQDBw7Uvn7ixAnNmzdPkjR9+nRZrd/9GCtXrlR2drYef/zxOvvau3ev1q5dK6fTWe9zvv76az366KMyDEM33XSTLrvssjrb/+3f/k2StGTJktor4ZLkdDo1d+5cnTt3ThkZGQFZFLSWRVL+kf0+tc0/ui+ofQEAtD1kN8yEdRWAtoL8BhBKrLEQLhF9x/ns7Gzl5OQoNzdXEyZM0DXXXCO73a68vDxVVFRo7NixmjJlSp33nDp1Svv27VNSUlKd10+cOKE5c+YoPj5e6enpSk5OVnV1tcrKyrR7924ZhqEhQ4bUhvylfvzjH+vnP/+53nzzTeXk5Gjo0KHq1KmTvvzySx09elTJycn64x//WPuQj3CyWay191VqToXLKZsloq+FAACiDNkNM2FdBaCtIL8BhBJrLIRLRBfCJWnu3LnKzMzUqlWrVFBQIK/Xq759+2rSpEnKycmpc0W6Kf3799cvfvELffrpp9q3b5+KiorkdrvVuXNnjR49WjfffLMmTpwom83W4PufeOIJDRs2TCtXrtTu3btVVVWlnj176v7779eMGTPqXckOF4/hVXt7rE8TSoIjVh7DK1tkfzEAABBlyG6YBesqAG0J+Q0gVFhjIVwshmEY4e4EvlNSUqKKigolJCRowIAB+uyzzyRJmZmZPr3f6XHrgwNfaHNZSbNtx/a6Urf3HqIYW8RfD4GP/B0vaNsYL6H1/fkd5tHQvy3HlzmEYl3FWIGvGCuhR3abW2vPvdG2MV5apy3VrhgrodVcdnM5xWRibXaNT0mXw9rw1fUaDqtN41IGRu1EAgAAEGysqwAAAAKPNRbChUK4CcXZHJqVcUOjE4rDatOsjBsUZ3OEuGcAAADRhXUVAABA4LHGQjhwScWEYmx29U3sqgUjJmpD6W7lH92nCpdTCY5YZXXro3EpAxVnc3BFDQAAoBmsqwAAAAKPNRbCgdFkUjE2u2Jsdk3sPUS39x4im8Uqj+GVoYtfQQEAAIBvWFcBAAAEHmsshBqjyuQunTh4wi4AAEDLsa4CAAAIPNZYCBVGFwAAAAAAAADA1CiEAwAAAAAAAABMjUI4AAAAAAAAAMDUKIQDAAAAAAAAAEyNQjgAAAAAAAAAwNQohAMAAAAAAAAATI1COAAAAAAAAADA1CiEAwAAAAAAAABMzR7uDqAup9MpSTp//rxKSkpqX7/0z0BzGC/wB+MlNM6fPy/pu3ke5tFYdkscX/AdYwW+YqyEDtltbpx7IxAYL/AVYyU0mstuCuERxuPxSJK8Xq8qKipqX7/0z0BzGC/wB+MltGrmeZhHY9ktcXzBd4wV+IqxEnpktzlx7o1AYLzAV4yV0GosuymER5jY2Fg5nU7ZbDbFxsaGuzsAgABxOp3yeDzM7SZEdgOAOZHd5kZ+A4D5NJfdFsMwjBD3CQAAAAAAAACAkOFhmQAAAAAAAAAAU6MQDgAAAAAAAAAwNQrhAAAAAAAAAABToxAOAAAAAAAAADA1CuEAAAAAAAAAAFOjEA4AAAAAAAAAMDUK4QAAAAAAAAAAU6MQDgAAAAAAAAAwNQrhAAAAAAAAAABToxAOAAAAAAAAADA1CuEAAAAAAAAAAFOjEA4AAAAAAAAAMDUK4QAAAAAAAAAAU7OHuwNo3Nq1a5Wbm6uSkhJ5vV716dNHkyZNUk5OjqxWrmGYzTfffKNPPvlEO3fu1K5du7R//34ZhqEXXnhB2dnZTb63pWNl69atWr58uXbt2iWn06nU1FTdeuutmjZtmmJiYgL9IyJAXC6XPv30U23ZskUFBQXav3+/qqur1blzZw0bNkyTJ0/WD3/4w0bfz3gBgofsbjvIbfiCzAaiA/nddpDf8BUZbk4WwzCMcHcC9c2bN0+rV69WbGysRo0aJbvdrry8PFVWVuqmm27Siy++SCCbzIIFC/T222/Xe725QG7pWFm6dKkWLVokm82mkSNHqkOHDtq+fbtOnjypq666SsuXL1dcXFxAf0YExrZt23T//fdLkpKSkpSRkaG4uDj985//1Ndffy1Jmjlzph599NF672W8AMFDdrct5DZ8QWYDkY/8blvIb/iKDDcpAxHno48+Mq644grj2muvNfbt21f7+rFjx4ybb77ZuOKKK4zly5eHr4MIinfffddYuHCh8eGHHxoHDhwwpkyZYlxxxRXGunXrGn1PS8fKl19+aQwYMMAYOnSoUVhYWPt6RUWFMXnyZOOKK64wFixYENCfD4Gzbds2Y/bs2cb27dvrbfvwww+NgQMHGldccYWRl5dXZ9v/396dx1VR9X8A/1wQROCyqGWKFpkOKi6QiLuWSy6Z5VJmSS4E7maZC1qpiWXpg6XkiqHiklmAmZmpZeoTq6Em8nPLDQRSNtlku+f3B6+Zh9u9Fy87wuf9F86cM3Pm3uN873xn5hz2F6Kqw9hd/zBukzEYs4lqN8bv+ofxm4zFGF43MRFeC40aNUpIkiRCQkJ01kVERCj/oYqKimqgdVRdjAnI5e0rs2fPFpIkifXr1+vUu3XrlmjXrp1wdnYWGRkZFT8QqnaLFy8WkiQJHx8freXsL0RVh7GbGLepPBiziWoW4zcxflN5MYY/mvh+Ty2TlJSE2NhYmJmZ6X0tx93dHc2aNcPdu3dx9uzZGmgh1Rbl7Sv5+fk4efIkAGDkyJE69Vq1agUXFxcUFBTg999/r7oDoCrToUMHAEBycrKyjP2FqOowdpMxeB4mfRiziWoO4zcZg+dkMoQx/NHERHgtc/HiRQBA27ZtYWFhobdMp06dAABxcXHV1i6qfcrbV65fv47c3FzY2dnhySefLLWevA96tNy4cQNA8ThmMvYXoqrD2E3G4HmY9GHMJqo5jN9kDJ6TyRDG8EcTE+G1THx8PACgRYsWBss0b95cqyzVT+XtK/Lf8jp95G0mJCRUuJ1Uve7evYuQkBAAwAsvvKAsZ38hqjqM3WQMnofp3xiziWoW4zcZg+dk0ocx/NHFRHgtk5OTAwClzgJrZWUFAMjOzq6WNlHtVN6+Ykw9S0tLnXpU+xUWFmL+/PnIzMxEz549MWDAAGUd+wtR1WHsJmPwPEwlMWYT1TzGbzIGz8n0b4zhjzYmwomI6oilS5ciLCwMzZs3x+rVq2u6OURERGQAYzYREdGjiTH80cZEeC0j3+HJzc01WEa++yPfRaL6qbx9xZh68t1I9rFHh6+vL7777js89thj2L59u9Y4ZQD7C1FVYuwmY/A8TDLGbKLagfGbjMFzMpXEGP7oYyK8lnFwcAAA3Llzx2CZpKQkrbJUP5W3r8h/JyYmGqwnr2MfezSsWrUKQUFBaNy4MbZv3w5HR0edMuwvRFWHsZuMwfMwAYzZRLUJ4zcZg+dkkjGG1w1MhNcyHTp0AABcuXIFDx480Fvmr7/+AgC0b9++2tpFtU95+0rr1q1hYWGB9PR03Lp1S2+98+fP69Sj2unzzz9HYGAg7OzsEBgYiDZt2ugtx/5CVHUYu8kYPA8TYzZR7cL4TcbgOZkAxvC6hInwWqZ58+ZwdnZGQUEBfv75Z531kZGRSEpKwmOPPQZXV9caaCHVFuXtK+bm5ujXrx8A4IcfftCpd/v2bZw9exZmZmZ47rnnqqz9VHFr1qzBtm3bYGtri8DAQLRr185gWfYXoqrD2E3G4Hm4fmPMJqp9GL/JGDwnE2N43cJEeC3k7e0NoPg/282bN5XlKSkpWL58OQDAy8sLJib8+uq78vYVLy8vqFQqBAQEKHcVgeKxqhYvXgyNRoM33ngDNjY21XAUVB5r167F1q1bYWNjg6+//lq581wa9heiqsPYTcbgebh+Yswmqr0Yv8kYPCfXX4zhdY9KCCFquhGka9myZdi7dy8aNmyIXr16oUGDBggLC0NWVhYGDRqEdevWwdTUtKabSZUoNjZWOSkCwNWrV5GdnQ1HR0fY2toqy7/99luteuXtK1u3bsWaNWtgamqKHj16QK1WIyoqCikpKejSpQt27NiBRo0aVd0BU7kdP34cM2bMAAB07NgRbdu21VuudevWShCWsb8QVR3G7vqFcZuMwZhNVPsxftcvjN9kLMbwuomJ8Frs4MGD2L17Ny5fvgyNRoPWrVtjzJgxGD9+PO9I10ERERF46623Hlru0qVLOsvK21dOnjyJwMBAXLhwAXl5eWjVqhVGjBgBT09PmJubV+h4qOoEBwfDx8fnoeXc3d0RFBSks5z9hajqMHbXH4zbZAzGbKJHA+N3/cH4TcZiDK+bmAgnIiIiIiIiIiIiojqNtzaJiIiIiIiIiIiIqE5jIpyIiIiIiIiIiIiI6jQmwomIiIiIiIiIiIioTmMinIiIiIiIiIiIiIjqNCbCiYiIiIiIiIiIiKhOYyKciIiIiIiIiIiIiOo0JsKJiIiIiIiIiIiIqE5jIpyoBty9exdLlixB//794ezsDCcnJ3h4eNR0syqNh4cHnJycEBwcXNNNISIiKre6Hq9rm/Xr18PJyQmLFi2q6aaUyaJFi+Dk5IT169fXdFOIiOoMxuDKd/r0aUycOBFubm5o164dr9mpXmpQ0w0gqm8KCwsxceJEXLt2DQBga2sLMzMz2Nra1nDLHi4+Ph4hISFQq9WYNGlSTTen0ty/fx87duwAAMyePbuGW0NERLXBoxyvqfLIye2JEyfCxsamhltDRFQ/MAZXvujoaHh5eUGj0cDU1BSNGzeGSqWChYVFjbUpLi4Ox44dg4ODA0aPHl1j7aD6hYlwomp2+vRpXLt2DXZ2dti3bx8cHR1ruklGS0hIgL+/PxwcHEpNhDdv3hxPP/001Gp19TWuAu7fvw9/f38ATIQTEVGxRzleU+WRfx+MGjWKiXAiomrCGFz5du7cCY1Gg2HDhmHVqlU1mgCXxcXFwd/fH+7u7kyEU7VhIpyoml25cgUA0L179zob0D///POabgIREVGF1Id4TUREVBsxBle+q1evAgBefvnlWpEEJ6opHCOcqJrl5eUBACwtLWu4JURERGQI4zUREVHNYAyufA8ePADAz5RIJYQQNd0IorpiwIABSEhIwM6dO+Ho6IiNGzfi1KlTSE5OhoWFBTIzMw3WPX78OFq2bFnufaempiIwMBAnTpxAfHw8hBBo2bIlBg4ciMmTJ8POzk6nTn5+Pvbu3YvDhw/j6tWryM3Nha2tLZo2bYquXbti5MiRcHV11To2Qz799FPldSYPDw9ERkZqLQOAiIgIvPXWW3BwcMCvv/6KU6dOYdu2bbh48SI0Gg06deqEOXPmKPvMzMxEQEAAfv75ZyQmJsLe3h4jRozA7Nmz9d7FTkpKwqFDh/DHH3/g9u3bSE5OhqmpKZ588kkMGDAAkyZN0nmtWW6rIbNmzdIZLiU8PBy7du1CTEwMMjIyYGtrCxcXF0yYMAE9e/bUux0nJycAxd9zXl4etmzZgvDwcNy7dw/9+/fHhg0bAAApKSkICAjAqVOnEB8fj6KiIjRu3BjNmzdHz549MXbsWDg4OBhsLxERPVxNxOtFixYhJCQEs2bNwtSpUxEQEICDBw/izp07sLe3x6BBgzB79mxl/NMLFy5g06ZNiImJQWZmJtq2bYvp06dj0KBBercfFRWFI0eO4Ny5c0hMTER6ejrUajU6dOiAV199FUOHDtWpc+rUKXh5eQEAAgIC0KdPH50ymzdvhp+fH6ytrXHgwIEK/VYBgL///hv+/v4ICwtDdnY2WrRogREjRsDb2xubN2+Gv78/Ro0ahVWrVumt/+uvv2L//v04f/48MjIyoFar0aVLF7z55pvo27evTvng4GD4+PjA3d0dQUFBCAkJwd69e3H16lWoVCp07NgRnp6e6Nevn1Y9+fsypGQbS363M2bMQFBQEIKDg3Hz5k00bNgQrq6umDVrFjp16lSBT46IqG5gDC5WXTFYvg7VR46NJUVHR2P37t04c+YMUlNTYWVlhfbt22Ps2LF48cUXoVKpdLYTGxuLw4cP48yZM0hMTMS9e/dgZWUFSZIwcuRIjB49Gqampka3CygeyqV79+6Ij4/HwIEDAQCXLl3SW/bfeYaSSutvzzzzDA4cOKCUzc7ORlBQEI4ePYobN24gPz8fzZs3R79+/eDp6YnmzZvr7Fuj0SA0NBQhISG4fPkysrKyoFar0aRJE3Tu3BnDhg3T+Y1BNYtDoxBVgRs3buCdd95BWloaGjVqBDMzM7Ro0QIpKSnIyclBTk4OGjZsqDWG9r8DQ1lER0dj5syZSE9PBwCYmZnBxMQEV65cwZUrV3DgwAF8/fXXaN26tVKnsLAQnp6eShJYpVJBrVYjPT0dKSkpuHTpEtLT05WktL29PbKyspCRkQETExM0btxYqw1lfb1q9+7dWLFiBVQqFSwtLZGVlYU//vgDZ86cQWBgIJ5++mlMnDgRly9fhqWlJTQaDZKSkhAQEICrV69i8+bNOtv85JNPcOTIEeUzsLKywv379xEXF4e4uDgcPHgQQUFBeOKJJ5Q6tra2sLe3R1paGgCgadOmWtv89x3ztWvXYtOmTVqfWUpKCo4dO4Zjx47B29sb8+bNM3jc0dHRWLZsGXJzc2FlZaX1vSckJGDcuHG4e/cugOI+YW1tjeTkZCQlJSEmJgaPP/44xo8fX5aPmoiIDKjueA0ABQUFmDx5MqKjo9GwYUMAQGJiIoKCghATE4M9e/bg1KlTePfdd1FQUABra2vk5eXhwoULmDVrFvz8/DB8+HCtbWZnZ2PChAnKv62srNCwYUOkpqbi9OnTOH36NMaNG4ePP/5Yq17fvn3x5ptvYteuXfDx8cHBgwe1bpxfvHhRmSxyyZIlFU6CR0VFwcvLC7m5uQAAa2trxMfHY/369Th9+jTc3d1L/dzkNsqsra2RmpqK3377Db/99hvefvttzJ8/3+A2PvnkE+zYsQMmJiawtrZGZmYmwsPDER4ejgULFsDT01Nr202bNsW9e/cAFP8OKvndW1tb62y/sLAQ3t7eOH36NMzMzGBmZoaMjAycOHECYWFh2LFjh/K7ioiovmMMrp4YLF/fpqamQqPRKBOPAtCZfHT16tUICAhQ/m1tbY2MjAyEhYUhLCwMv/76K9asWQMTE+3BJaZMmaLkIho1aoRGjRohPT0dkZGRiIyMxNGjR7FhwwY0aPC/FGTTpk3x4MEDZGVl6Z0IVW5jZdHX30q6du0avLy8lIf/GjRoAHNzc9y8eRNBQUH44YcfsHHjRnTt2lWr3vz58/Hjjz8q/1ar1cjKykJaWhquXr2Ka9euMRFe2wgiqjTPP/+8kCRJuLi4iBEjRogzZ84o627cuCGEEGLdunVCkiSxcOHCStlnfHy8cHNzE5IkiaVLl4obN26IoqIiUVRUJC5duiSmTJkiJEkSw4cPF4WFhUq9kJAQIUmS6NKliwgNDRUPHjwQQghRWFgoEhISxK5du8SmTZu09hUeHi4kSRLPP/98qW2aMGGCkCRJfP/993rrd+nSRTg7Ows/Pz+RkZEhhBDi9u3bYty4cUKSJDFmzBgxa9YsMWTIEBEVFSU0Go3Iy8sT3377rejQoYOQJEmcOHFCZ79r164VO3fuFNevXxdFRUVCCCHy8/NFRESEGDNmjJAkSXh5eenUu337tpAkSUiSVOpx/fjjj0q5jz/+WKSkpAghhEhNTRUrVqxQ1oWGhurUlde5uLiICRMmiEuXLgkhhNBoNOLmzZtCCCEWLVokJEkSgwcPFlFRUcox5OXliUuXLom1a9eKo0ePltpGIiJ6uJqI1wsXLhSSJImuXbuK3r17i99++00UFRWJwsJCcfToUeHq6iokSRKrV68WXbt2FT4+PuKff/4RQgiRkpIipk+fLiRJEr179xYFBQVa287JyRGzZ88WR48eFWlpacryjIwMsWvXLuHi4iIkSRI//fSTTrtyc3PF0KFDhSRJYs6cOcryBw8eiOHDhwtJksSsWbMqfPzp6emiZ8+eQpIkMWrUKBEXFyeEKI7TISEhokuXLqJr164GP/OVK1cqMfKnn34S2dnZQgghMjMzxe7du5XP7+DBg1r1vv/+eyFJknj22WeVz/f+/ftCCCGSk5PFvHnzhCRJwsnJSURFRensV47ft2/fNnhs8nfr5uYm3N3dxaFDh0ReXp4QQoi4uDgxYsQI5fcNEVF9xxj8P9UVg4X43+ceHh6ud/327duFJEmiV69eYt++fUqszM3NFYcOHRK9e/cWkiTp5AiEEOK9994TP/74o/KZCSFEdna2CA0NVept3bpVp54coydMmGCw3cZcq5eWpzCmv92/f18pN2fOHBEXF6fkTm7duiXee+895bOR8xdCCBEZGSkkSRLt27cXgYGBIjMzUwhRfI2fnJwsgoODxapVqwy2m2oGE+FElUg+ebq5uYm7d+/qLVPZQV2+gFuzZo3e9Xl5eeKll14SkiSJw4cPK8uXLl0qJEkSH330kdH7qqxEuCRJYtGiRTr1EhIShJOTk5AkSTg7OyuBqSQfHx+D9UuTlpYmevToIZycnHQuZo0JrhqNRgwePFhIkiTeffddvWXkAPn8888rSWyZvP2BAweK3NxcvfWHDRsmJEkShw4dKtOxERFR2dREvJYvwiVJEhERETrr/f39lfUeHh4667Ozs5UL9cjIyDLtW775behC88KFC8LZ2VlIkiRCQkKEEEL4+voqF/2pqall2p8+8vG5u7srN5JLCg0NVY7/35/59evXhZOTk+jRo4e4c+eO3u3LN6tffPFFreXyRbYkSWLJkiU69TQajfDw8BCSJImJEyfqrC9LIlySJL3J9L/++ktZn5CQYHA7RET1AWOwtuqIwUKUngjP6Glz8gAAEjpJREFUyMgQLi4uolOnTsqN6n/7888/hZOTk+jWrZtys9cYUVFRBnMI1ZkIL62/+fn5CUmSxHvvvWdwH56enkKSJBEQEKAs27Jli5AkSXh6ehqsR7UPJ8skqgIvv/yyzhAbVSE3Nxc///wzTExMMHnyZL1lzM3NMWTIEADAH3/8oSyXX+mVh+Gobt7e3jrLWrRogaeeegoAMGTIEOXvkuQxuOWZxI1lZ2cHV1dXCCEQExNT5vbGxcXh5s2bAIDp06frLTNz5kwAxUOcnD9/Xm+ZCRMmGBxGRv5O/vnnnzK3j4iIyq664nVJrq6ueocA6dWrl/L31KlTddZbWlrCxcUFAHD58uUy7XPAgAEAgHPnzqGoqEhnvbOzszIfxooVK/Ddd98pY4auXLkS9vb2ZdqfPvLQZa+99prO8GoAMHLkSINzYISGhkIIgeHDh+sdnxMo/t1gbm6OK1euGIyj+j5XlUqlLA8PD1de7S4PNzc3uLm56Szv2LGjMixbWX+/EBHVVYzBxaojBj/MkSNHkJOTg169eqFdu3Z6y7i6uqJly5bIyMhAbGys0dt2c3ODjY0NEhISkJycXFlNLrPS+ltoaCgAGMypAMCIESMA6M+pyMPO0KOBY4QTVYHqGv8xNjYWBQUFUKlUeOmllwyWk2eITkxMVJb169cPW7duxfHjxzFt2jSMHj0a3bp1q5ZA27BhQzg6Oupd16RJE9y4cQOSJOldLwev+/fv611//vx57N27FzExMUhOTkZOTo5OmfIkmuVg37hxY7Rt21ZvmdatW6NZs2ZITk5GbGys8mOpJH3LZP369cO5c+ewZs0a3Lx5E0OGDIGLi0uZx18nIiLj1MR4zYbiW5MmTZS/DcUZuYy+GFhYWIiQkBD8/PPPyjwfBQUFWmXy8vKQkZGhNxHt5eWF33//HWfOnMGSJUsAAOPHj0f//v2NO7BS5Ofn4+rVqwCAbt266S2jUqng5uamd2Ju+Qa2fHyGFBYWAiiePPvxxx/XWteiRQu0atVKb72uXbvC1NQURUVFiIuLMzjx9cOUNhlms2bNkJSUhIyMjHJtm4iormEM/p+qjMHGkONseHg4evfubbCcHMMSExN1vr/Dhw/j4MGDuHjxIlJTU5GXl6dT/59//kGzZs0qseXGM9TfEhMTkZSUBKD4YT19k4ECUL7PkjmVnj17wszMDLGxsfDw8MBrr72GHj161NgxknGYCCeqAvqCW1WQE7pCCGUyp9LICXGgeIboOXPmYMOGDcokU0BxMve5557DuHHjDCarK6pp06YGA4w8Acpjjz2md708MYd8sVvStm3bsHr1agghlG2VnAwkMzMTeXl5yiRdZSFPpvmwoPbEE08gOTlZKf9vpfUNb29vxMbG4tdff8WePXuwZ88eNGjQAJ06dcKgQYPw2muvwcbGpsxtJyIi/aorXpf0sPgGQCeJK5Nj5L9jYHZ2Njw9PbXeeLKwsIBarVa2K/9OMBQDTUxM4Ovri2HDhgEAHBwcsHDhQmMO6aEyMjKUp+AMHRtgOMbKb69lZ2cjOzv7ofvTd4ylxW8LCwvY2NggLS3NYPw2hpWVlcF18qRs+n6/EBHVR4zB2vuvqhhsDDnO5ubmGnWtXDKvUFhYiLlz5+Lo0aPKMnNzc61JpuUnpstzHV5ZDPW3km/Ip6SkPHQ7JY/d0dERy5Ytw4oVKxAdHY3o6GgAxd9f3759MW7cOHTo0KGCLafKxkQ4URX49yzKVUVO+KrVauWkWxYzZ87EyJEjcfjwYURERODs2bP4+++/8ffff2Pnzp1YuXIlXnnllcpudpW4cuUK1qxZAyEEJkyYgPHjx+Ppp5/Wmll8/vz5+OGHH5TPrTz03dkui9L6hrm5OTZu3IizZ8/i6NGjiI6ORmxsLGJiYhATE4Nt27YhMDDQ4OtqRERUNtUVr6vahg0bEBMTA3t7eyxatAh9+/bVerqtqKhIuRArLQYGBwcrf9+9exe3bt2Ck5NT1TXcSPLrxj4+Ppg0aVLNNoaIiCoFY7C2mozBcpx96623lCfSjfXtt9/i6NGjaNSoEebNm4fBgwcrw4HJ+vfvj6SkpApdh1eUof5WckiTqKioMj94NnbsWDz33HM4dOgQIiIicObMGSQkJOCbb77Bvn37MHfuXEybNq1CbafKVTfOPET1lBxgs7KykJmZWa5ttGrVCt7e3ti2bRsiIyOxc+dOdOvWDYWFhVi+fLlRd0VrgyNHjkCj0aBPnz748MMP0aZNG60kOGDcHV5D5CFj5NemDJHXV2SIGRcXF8yfPx/79u1DZGQk/Pz80KJFC6Smppb5hwkREdV98nAhH374IV555RWtC3AARr01Fh0djW3btgEofnU8Pz8f8+fPR35+foXbZ2trq8Tk0oYnM7ROHhat5OvIZVXafvPy8pRX3atjiDgiIqo7ansMNkZF4qx8/DNmzICHh4dOEryoqKhCb1s1aPC/53cNPZRW3lwIoD0szp07d8q1jaZNm2LixInYsGEDwsPDsX//fgwePBhCCHz55Zf4v//7v3K3jyofE+FEj7COHTuiQYMGEELg1KlTFd6eqakpunfvjk2bNsHMzAw5OTm4cOGCsl6+i1qTd3INkSfeMPTqUU5ODs6ePat3Xcm7w4aOzdnZWdmOoYkwr1+/rrRDLl9RlpaWePHFF/Hxxx8DKB6rXN+450REVH/Jsad9+/Z614eFhZVaPysrCwsWLIBGo8GYMWOwY8cONGnSBJcuXcIXX3xR4faZm5ujTZs2AGDwDTYhBKKiovSuk+fXqMhvnYSEBMTHx+tdd+bMGRQVFUGlUul8hvJQbrXxtw8REdW82h6DjSHH2cjISK2hP4zxsOP/888/DSawjckvqNVq5W9DD6X99ddfRrVVn1atWik3Ak6ePFnu7chUKhU6d+6ML7/8Ek888QQ0Gg3OnDlT4e1S5WEinOgRZm1tjRdeeAEA8OWXXyIrK8tg2cLCQq1xNUu7u2xubq4EpZLl5FmRK3LHtarIbTM0i/emTZsMjisq1wUMT8LZvn17PPXUU8q29PH39wdQPCZY586djWt4CaV9J/KEmUIInYlXiIiofistBmZnZ2Pjxo2l1vf19UVCQgJatmyJxYsXo3HjxlixYgUAIDAw0GCCuiyGDh0KoPgV6vT0dJ31hw4d0jtRJgC88sorUKlUuHbtGr755ptS91PaZJRbtmzRWSaEUJb37NkTdnZ2Wutr828fIiKqeY9CDH6YoUOHwtLSEhkZGfjqq69KLfvvOFva8RcWFpaazJfrGroGB4rn33BwcAAAHD9+XGd9Wloavvvuu1Lb/DCjRo0CAHz99ddKYl8fIYRWW0u7fjc1NVWeZuf1e+3CRDjRI27evHmws7PDjRs3MH78eJw8eVI50QohcOPGDQQGBmLYsGFaT3cvXLgQPj4+OHXqlFYCPT4+HgsXLkReXh4sLCzQtWtXZd1TTz0FMzMzZGZm4siRI9V3kEaQZ7c+ceIENm/erEzEkZqais8++wybN2/WubiV2djYKJOilBybrSSVSoW5c+cCKA7AK1asUF7xSktLg6+vL3788UcAwNy5c8s15t1LL70EPz8/nD9/XgmqQgicP39e+THUqVMn2NralnnbRERUd8kxcNWqVYiMjFSerDp//jwmTZqkN/Es++WXXxASEgITExN8/vnnykXpwIEDMXbsWGg0GixcuLDUm+3GePPNN9GkSROkpaXB09NTeU24oKAABw4cwIcffqj11FdJbdq0UcYGX758Of7zn/9oPRWWlZWF06dP4/3338c777yjdxvW1tbYt28f/Pz8lKT23bt3sXDhQoSFhUGlUmHmzJl69w0AoaGhyoSfREREskchBj+Mvb093nvvPQDFN40/+OADXL9+XVn/4MEDREdHY+nSpXj99de16srHv2HDBhw7dkyJldeuXcO0adNw/vx5WFpa6t2vHGOvXbuGc+fOGWyfPInoxo0bcfz4cWXC0rNnz2Ly5MkVTjR7e3ujVatWSEtLw+uvv46ffvpJ68n4O3fuYN++fRg1ahSOHTumLF+7di3mzJmDY8eOaX3P9+7dg6+vL+Lj46FSqdCrV68KtY8qFyfLJHrEtWzZElu3bsXMmTNx+fJleHl5wczMDFZWVsjOztYKCvLrvUDx+Fo//fQTgoODoVKpoFarUVBQoCSQTU1NsXz5cq3ZleVhOkJDQzFnzhyo1WplMokFCxYoT3vVhD59+uCFF17AL7/8Aj8/P6xduxY2Nja4f/8+hBAYO3YsioqKEBISorf+q6++iq+++gqrVq3CunXrlDFC33rrLeXie/jw4bh06RI2bdqEXbt2Yc+ePVCr1cjMzFQm2fD29sbIkSPLdQwpKSnYvHkzNm/eDFNTU6jVaq3v0N7eHr6+vuXaNhER1V1z587Ff//7XyQmJsLDwwMNGzaEqakpcnJyYGFhga+++gqenp469e7evYsPP/wQAPD2229r3fwGgMWLFyMiIgK3b9+Gr68vVq1aVe422tra4osvvoCXlxcuXLiAl19+GWq1Gnl5ecjPz4erqyu6deum96ltoHjC6wcPHmDv3r3YsmULtmzZAmtra6hUKmRlZSmJB3d3d731O3TogPbt22Pz5s0ICAiAtbW18htB3r6bm5tOvVdffRUxMTHYsWMHvvnmGzRp0gQqlQpDhgzBwoULy/15EBFR3fAoxGBjeHh4IDMzE+vWrcP+/fuxf/9+WFpaKg/Cyde78tPZsilTpuDw4cO4desWZs6cCTMzMzRs2BBZWVkwNTWFr68v/P399Q7v6ejoiG7duiEqKgqvvfYa7OzsYGVlBQDw8/NThmzx9vbGkSNHcPv2bcyYMQPm5uZo0KABcnJy0KJFCyxZsgQLFiwo97Hb2Nhg27ZtmD59Oq5du4Z3331XuR5/8OCBVlK8ZE6lsLAQR44cUR4StLa2hhBC6030uXPnQpKkcreNKh+fCCeqAzp37ozDhw/j/fffh6urKywtLZGZmYlGjRqhY8eO8PDwwK5du7QuDufNm4f58+ejb9++aNWqFQoKClBUVIQnn3wSo0ePRnBwMF555RWdfS1fvhxTp05F69atkZ+fj4SEBCQkJNSKcavXrl2LefPm4ZlnnlHGTn/22Wfx2WefYeXKlaXWnTlzJt5//304OTlBCKEc179fhX733Xexfft2DBw4EPb29sjJyYGdnR0GDBiA7du3Y968eeVu/4YNGzB16lQ8++yzePzxx5GdnQ0zMzM4OTnB29sbhw4dQrt27cq9fSIiqptatWqF/fv3Y+TIkWjSpAk0Gg3UajVeeuklfPfdd+jTp4/eeosXL0Z6ejrat2+P2bNn66y3srLCZ599BhMTE4SEhOCXX36pUDvd3d0REhKC4cOHo3HjxsjPz4eDgwNmz56NHTt2wNzc3GBdU1NTLFu2DHv27MHIkSPh4OCA/Px85OXloUWLFhgwYAA++ugjrFu3zuA2Fi9ejE8//RTOzs4oLCyEpaUlunfvjq1bt+pNUgDAmDFj4Ovri86dO8PU1BSJiYlISEio0MRfRERUdzwqMdgYM2bMwIEDBzBu3Dg4OjpCo9EgJycHjz32GPr06YP58+djz549WnXs7Oywb98+jB8/Xpko08LCAoMGDUJQUBBGjx5d6j7Xr1+PN954Ay1btkR2drZyHV5yXHFbW1t88803GDduHB5//HEIIWBnZwcPDw8EBwfrTNBZHk899RRCQ0OxdOlSdO/eHTY2Nkoy38nJCePGjcOWLVu0HnqbNGkSPvjgAwwcOBCOjo4QQiA/Px/NmzfH8OHDsXv3bkybNq3CbaPKpRKc+YWIiIiIiOqg4OBg+Pj4wN3dHUFBQTXdHCIiIiKqQXwinIiIiIiIiIiIiIjqNCbCiYiIiIiIiIiIiKhOYyKciIiIiIiIiIiIiOq0BjXdACIq1rt37zKVnzJlisGJnYiIiKhq1Pd4PWvWLMTExBhd3tXVFf7+/lXYIiIiqi8YgxmDiSqKiXCiWuLevXtlKp+Tk1NFLSEiIiJD6nu8zsjIKNNnkJGRUYWtebjRo0dj9OjRNdoGIiKqHIzBj1YMJqqNVEIIUdONICIiIiIiIiIiIiKqKhwjnIiIiIiIiIiIiIjqNCbCiYiIiIiIiIiIiKhOYyKciIiIiIiIiIiIiOo0JsKJiIiIiIiIiIiIqE5jIpyIiIiIiIiIiIiI6jQmwomIiIiIiIiIiIioTvt/UF5MjcuXRjgAAAAASUVORK5CYII=",
      "text/plain": [
       "<Figure size 1512x504 with 3 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "fig, ax = plt.subplots(1,3,figsize=(21,7))\n",
    "sns.scatterplot(data=perf_df, x=\"rf_estimators\", y=\"valid_r2_score\", ax=ax[0])\n",
    "sns.scatterplot(data=perf_df, x=\"rf_max_depth\", y=\"valid_r2_score\", ax=ax[1])\n",
    "sns.scatterplot(data=perf_df, x=\"rf_max_features\", y=\"valid_r2_score\", ax=ax[2])\n",
    "fig.tight_layout()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "Za2roo8B7AfH"
   },
   "source": [
    "### Bayesian search"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 23,
   "metadata": {
    "id": "iIHILV2R7Fz4"
   },
   "outputs": [],
   "source": [
    "#define loss function\n",
    "def lossfn(p):\n",
    "    params[\"rf_estimators\"] = int(p[\"rf_estimator\"])\n",
    "    params[\"rf_max_depth\"] = int(p[\"rf_depth\"])\n",
    "    params[\"rf_max_features\"] = int(p[\"rf_feature\"])\n",
    "    \n",
    "    tparam = parse.wrapper(params)\n",
    "    pl = mp.ModelPipeline(tparam)\n",
    "    pl.train_model()\n",
    "    perf_data = pl.model_wrapper.get_perf_data(subset=\"valid\", epoch_label=\"best\")\n",
    "    pred_results = perf_data.get_prediction_results()\n",
    "    r2 = pred_results['r2_score']\n",
    "    return {'loss': 1-r2, \n",
    "            'status': STATUS_OK, \n",
    "            'model': tparam.model_tarball_path, \n",
    "            'rf_estimator': params[\"rf_estimators\"],\n",
    "            'rf_depth': params[\"rf_max_depth\"],\n",
    "            'rf_feature': params[\"rf_max_features\"]}\n",
    "\n",
    "space = {\n",
    "    \"rf_estimator\": hp.uniform(\"rf_estimator\", 8, 256),\n",
    "    \"rf_depth\": hp.uniform(\"rf_depth\", 8, 256),\n",
    "    \"rf_feature\": hp.uniform(\"rf_feature\", 8, 256)\n",
    "}"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 24,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/"
    },
    "id": "ndFs9n6f7G-Y",
    "outputId": "1c65ee12-16ec-4731-faed-3588cd5cc11c"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "  0%|          | 0/5 [00:00<?, ?it/s, best loss: ?]"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "2021-09-18 16:43:12,013 Attempting to load featurized dataset\n",
      "2021-09-18 16:43:12,024 Exception when trying to load featurized data:\n",
      "DynamicFeaturization doesn't support get_featurized_dset_name()\n",
      "2021-09-18 16:43:12,026 Featurized dataset not previously saved for dataset H1_std, creating new\n",
      "2021-09-18 16:43:12,037 Featurizing sample 0\n",
      "2021-09-18 16:43:17,178 Featurizing sample 1000\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "number of features: 1024\n",
      "  0%|          | 0/5 [00:07<?, ?it/s, best loss: ?]"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "2021-09-18 16:43:19,217 Previous dataset split restored\n",
      "2021-09-18 16:43:19,235 Wrote transformers to ampl_output/H1_std/RF_ecfp_scaffold_regression/022a4de2-71bf-43d4-ab65-4c38f91c6e82/transformers.pkl\n",
      "2021-09-18 16:43:19,237 Transforming response data\n",
      "2021-09-18 16:43:19,274 Transforming response data\n",
      "2021-09-18 16:43:19,291 Transforming response data\n",
      "2021-09-18 16:43:19,327 Fitting random forest model\n",
      "2021-09-18 16:43:20,004 Fold 0: training r2_score = 0.963, validation r2_score = 0.424, test r2_score = 0.514\n",
      "2021-09-18 16:43:20,199 Wrote model tarball to ampl_output/H1_std_model_022a4de2-71bf-43d4-ab65-4c38f91c6e82.tar.gz\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      " 20%|██        | 1/5 [00:08<00:32,  8.21s/it, best loss: 0.5758313410870676]"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "2021-09-18 16:43:20,221 Attempting to load featurized dataset\n",
      "2021-09-18 16:43:20,230 Exception when trying to load featurized data:\n",
      "DynamicFeaturization doesn't support get_featurized_dset_name()\n",
      "2021-09-18 16:43:20,231 Featurized dataset not previously saved for dataset H1_std, creating new\n",
      "2021-09-18 16:43:20,243 Featurizing sample 0\n",
      "2021-09-18 16:43:25,328 Featurizing sample 1000\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "number of features: 1024\n",
      " 20%|██        | 1/5 [00:15<00:32,  8.21s/it, best loss: 0.5758313410870676]"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "2021-09-18 16:43:27,261 Previous dataset split restored\n",
      "2021-09-18 16:43:27,278 Wrote transformers to ampl_output/H1_std/RF_ecfp_scaffold_regression/129ad3e4-200a-4c69-93c3-724cdf24abfe/transformers.pkl\n",
      "2021-09-18 16:43:27,279 Transforming response data\n",
      "2021-09-18 16:43:27,316 Transforming response data\n",
      "2021-09-18 16:43:27,331 Transforming response data\n",
      "2021-09-18 16:43:27,362 Fitting random forest model\n",
      "2021-09-18 16:43:28,773 Fold 0: training r2_score = 0.966, validation r2_score = 0.466, test r2_score = 0.536\n",
      "2021-09-18 16:43:29,263 Wrote model tarball to ampl_output/H1_std_model_129ad3e4-200a-4c69-93c3-724cdf24abfe.tar.gz\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      " 40%|████      | 2/5 [00:17<00:26,  8.71s/it, best loss: 0.5342531744658509]"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "2021-09-18 16:43:29,285 Attempting to load featurized dataset\n",
      "2021-09-18 16:43:29,294 Exception when trying to load featurized data:\n",
      "DynamicFeaturization doesn't support get_featurized_dset_name()\n",
      "2021-09-18 16:43:29,296 Featurized dataset not previously saved for dataset H1_std, creating new\n",
      "2021-09-18 16:43:29,306 Featurizing sample 0\n",
      "2021-09-18 16:43:34,383 Featurizing sample 1000\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "number of features: 1024\n",
      " 40%|████      | 2/5 [00:24<00:26,  8.71s/it, best loss: 0.5342531744658509]"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "2021-09-18 16:43:36,372 Previous dataset split restored\n",
      "2021-09-18 16:43:36,389 Wrote transformers to ampl_output/H1_std/RF_ecfp_scaffold_regression/75fae892-cdac-4e6d-bff5-427935eda949/transformers.pkl\n",
      "2021-09-18 16:43:36,391 Transforming response data\n",
      "2021-09-18 16:43:36,427 Transforming response data\n",
      "2021-09-18 16:43:36,444 Transforming response data\n",
      "2021-09-18 16:43:36,476 Fitting random forest model\n",
      "2021-09-18 16:43:39,023 Fold 0: training r2_score = 0.967, validation r2_score = 0.464, test r2_score = 0.540\n",
      "2021-09-18 16:43:39,753 Wrote model tarball to ampl_output/H1_std_model_75fae892-cdac-4e6d-bff5-427935eda949.tar.gz\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      " 60%|██████    | 3/5 [00:27<00:19,  9.52s/it, best loss: 0.5342531744658509]"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "2021-09-18 16:43:39,774 Attempting to load featurized dataset\n",
      "2021-09-18 16:43:39,784 Exception when trying to load featurized data:\n",
      "DynamicFeaturization doesn't support get_featurized_dset_name()\n",
      "2021-09-18 16:43:39,785 Featurized dataset not previously saved for dataset H1_std, creating new\n",
      "2021-09-18 16:43:39,795 Featurizing sample 0\n",
      "2021-09-18 16:43:44,806 Featurizing sample 1000\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "number of features: 1024\n",
      " 60%|██████    | 3/5 [00:34<00:19,  9.52s/it, best loss: 0.5342531744658509]"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "2021-09-18 16:43:46,765 Previous dataset split restored\n",
      "2021-09-18 16:43:46,782 Wrote transformers to ampl_output/H1_std/RF_ecfp_scaffold_regression/03826201-20a9-4fef-8bde-35ac60281465/transformers.pkl\n",
      "2021-09-18 16:43:46,783 Transforming response data\n",
      "2021-09-18 16:43:46,816 Transforming response data\n",
      "2021-09-18 16:43:46,832 Transforming response data\n",
      "2021-09-18 16:43:46,863 Fitting random forest model\n",
      "2021-09-18 16:43:48,287 Fold 0: training r2_score = 0.967, validation r2_score = 0.461, test r2_score = 0.534\n",
      "2021-09-18 16:43:48,806 Wrote model tarball to ampl_output/H1_std_model_03826201-20a9-4fef-8bde-35ac60281465.tar.gz\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      " 80%|████████  | 4/5 [00:36<00:09,  9.34s/it, best loss: 0.5342531744658509]"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "2021-09-18 16:43:48,828 Attempting to load featurized dataset\n",
      "2021-09-18 16:43:48,837 Exception when trying to load featurized data:\n",
      "DynamicFeaturization doesn't support get_featurized_dset_name()\n",
      "2021-09-18 16:43:48,838 Featurized dataset not previously saved for dataset H1_std, creating new\n",
      "2021-09-18 16:43:48,850 Featurizing sample 0\n",
      "2021-09-18 16:43:53,964 Featurizing sample 1000\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "number of features: 1024\n",
      " 80%|████████  | 4/5 [00:43<00:09,  9.34s/it, best loss: 0.5342531744658509]"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "2021-09-18 16:43:55,930 Previous dataset split restored\n",
      "2021-09-18 16:43:55,947 Wrote transformers to ampl_output/H1_std/RF_ecfp_scaffold_regression/8ff607a1-6597-42e9-b725-1ad8e505fcbe/transformers.pkl\n",
      "2021-09-18 16:43:55,949 Transforming response data\n",
      "2021-09-18 16:43:55,984 Transforming response data\n",
      "2021-09-18 16:43:56,000 Transforming response data\n",
      "2021-09-18 16:43:56,032 Fitting random forest model\n",
      "2021-09-18 16:43:56,590 Fold 0: training r2_score = 0.959, validation r2_score = 0.485, test r2_score = 0.524\n",
      "2021-09-18 16:43:56,662 Wrote model tarball to ampl_output/H1_std_model_8ff607a1-6597-42e9-b725-1ad8e505fcbe.tar.gz\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "100%|██████████| 5/5 [00:44<00:00,  8.93s/it, best loss: 0.5152785357644352]\n"
     ]
    }
   ],
   "source": [
    "trials = Trials()\n",
    "best = fmin(lossfn, space, algo=tpe.suggest, max_evals=5, trials=trials)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 25,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 636
    },
    "id": "sC-LVt3VtPDc",
    "outputId": "e05b2c92-1a32-422c-8632-0d7f42347584"
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(0.2, 0.6)"
      ]
     },
     "execution_count": 25,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "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",
      "text/plain": [
       "<Figure size 720x720 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "r2_list = [1-trials.trials[i][\"result\"][\"loss\"] for i in range(len(trials.trials))]\n",
    "fig, ax = plt.subplots(1,1,figsize=(10,10))\n",
    "sns.lineplot(x=list(range(1,len(trials.trials)+1)), y=r2_list, ax=ax)\n",
    "ax.set_ylim(0.2,0.6)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 26,
   "metadata": {
    "id": "-2_z-vFLtUYb"
   },
   "outputs": [],
   "source": [
    "rfe_list = [trials.trials[i][\"result\"][\"rf_estimator\"] for i in range(len(trials.trials))]\n",
    "rfd_list = [trials.trials[i][\"result\"][\"rf_depth\"] for i in range(len(trials.trials))]\n",
    "rff_list = [trials.trials[i][\"result\"][\"rf_feature\"] for i in range(len(trials.trials))]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 27,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 206
    },
    "id": "sDMxDhM5tXHa",
    "outputId": "767c9f0c-1d1b-4bbc-bf34-6215447e2243"
   },
   "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>rf_estimators</th>\n",
       "      <th>rf_max_depth</th>\n",
       "      <th>rf_max_features</th>\n",
       "      <th>valid_r2_score</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>21</td>\n",
       "      <td>77</td>\n",
       "      <td>72</td>\n",
       "      <td>0.484721</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>172</td>\n",
       "      <td>32</td>\n",
       "      <td>115</td>\n",
       "      <td>0.465747</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>254</td>\n",
       "      <td>85</td>\n",
       "      <td>237</td>\n",
       "      <td>0.464318</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>179</td>\n",
       "      <td>87</td>\n",
       "      <td>109</td>\n",
       "      <td>0.461171</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>62</td>\n",
       "      <td>132</td>\n",
       "      <td>59</td>\n",
       "      <td>0.424169</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "   rf_estimators  rf_max_depth  rf_max_features  valid_r2_score\n",
       "4             21            77               72        0.484721\n",
       "1            172            32              115        0.465747\n",
       "2            254            85              237        0.464318\n",
       "3            179            87              109        0.461171\n",
       "0             62           132               59        0.424169"
      ]
     },
     "execution_count": 27,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "perf_df = pd.DataFrame({\"rf_estimators\": rfe_list, \"rf_max_depth\": rfd_list, \"rf_max_features\": rff_list, \"valid_r2_score\": r2_list})\n",
    "perf_df.sort_values(by=\"valid_r2_score\", ascending=False)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "XB0K5BimtYDS"
   },
   "source": [
    "#### Save Bayesian optimization trial object and continue later"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 28,
   "metadata": {
    "id": "VzXDDEF_r8h8"
   },
   "outputs": [],
   "source": [
    "# save the trial object\n",
    "trial_file = os.path.join(\"hp_trial.pkl\")\n",
    "with open(trial_file, \"wb\") as f:\n",
    "    pickle.dump(trials, f)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 29,
   "metadata": {
    "id": "jSWVhsjfsINl"
   },
   "outputs": [],
   "source": [
    "# load the trial object\n",
    "with open(trial_file, \"rb\") as f:\n",
    "    new_trials = pickle.load(f)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 30,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/"
    },
    "id": "IXcOynF7sLMn",
    "outputId": "2c32048c-f4ec-4890-d128-dc707c31c759"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "  0%|          | 0/45 [00:00<?, ?it/s, best loss: ?]"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "2021-09-18 16:43:57,144 Attempting to load featurized dataset\n",
      "2021-09-18 16:43:57,154 Exception when trying to load featurized data:\n",
      "DynamicFeaturization doesn't support get_featurized_dset_name()\n",
      "2021-09-18 16:43:57,156 Featurized dataset not previously saved for dataset H1_std, creating new\n",
      "2021-09-18 16:43:57,167 Featurizing sample 0\n",
      "2021-09-18 16:44:02,244 Featurizing sample 1000\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "number of features: 1024\n",
      "  0%|          | 0/45 [00:06<?, ?it/s, best loss: ?]"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "2021-09-18 16:44:04,148 Previous dataset split restored\n",
      "2021-09-18 16:44:04,165 Wrote transformers to ampl_output/H1_std/RF_ecfp_scaffold_regression/ef994b8a-e658-4b2c-8d9e-cef900a13fc8/transformers.pkl\n",
      "2021-09-18 16:44:04,167 Transforming response data\n",
      "2021-09-18 16:44:04,201 Transforming response data\n",
      "2021-09-18 16:44:04,216 Transforming response data\n",
      "2021-09-18 16:44:04,248 Fitting random forest model\n",
      "2021-09-18 16:44:05,228 Fold 0: training r2_score = 0.966, validation r2_score = 0.459, test r2_score = 0.520\n",
      "2021-09-18 16:44:05,464 Wrote model tarball to ampl_output/H1_std_model_ef994b8a-e658-4b2c-8d9e-cef900a13fc8.tar.gz\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "  2%|▏         | 1/45 [00:08<06:06,  8.34s/it, best loss: 0.5152785357644352]"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "2021-09-18 16:44:05,483 Attempting to load featurized dataset\n",
      "2021-09-18 16:44:05,492 Exception when trying to load featurized data:\n",
      "DynamicFeaturization doesn't support get_featurized_dset_name()\n",
      "2021-09-18 16:44:05,493 Featurized dataset not previously saved for dataset H1_std, creating new\n",
      "2021-09-18 16:44:05,503 Featurizing sample 0\n",
      "2021-09-18 16:44:10,463 Featurizing sample 1000\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "number of features: 1024\n",
      "  2%|▏         | 1/45 [00:15<06:06,  8.34s/it, best loss: 0.5152785357644352]"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "2021-09-18 16:44:12,498 Previous dataset split restored\n",
      "2021-09-18 16:44:12,514 Wrote transformers to ampl_output/H1_std/RF_ecfp_scaffold_regression/f1b7cd6e-5ef1-40ad-b5d4-40ccba42ae22/transformers.pkl\n",
      "2021-09-18 16:44:12,517 Transforming response data\n",
      "2021-09-18 16:44:12,555 Transforming response data\n",
      "2021-09-18 16:44:12,571 Transforming response data\n",
      "2021-09-18 16:44:12,603 Fitting random forest model\n",
      "2021-09-18 16:44:13,277 Fold 0: training r2_score = 0.964, validation r2_score = 0.425, test r2_score = 0.507\n",
      "2021-09-18 16:44:13,448 Wrote model tarball to ampl_output/H1_std_model_f1b7cd6e-5ef1-40ad-b5d4-40ccba42ae22.tar.gz\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "  4%|▍         | 2/45 [00:16<05:49,  8.13s/it, best loss: 0.5152785357644352]"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "2021-09-18 16:44:13,466 Attempting to load featurized dataset\n",
      "2021-09-18 16:44:13,475 Exception when trying to load featurized data:\n",
      "DynamicFeaturization doesn't support get_featurized_dset_name()\n",
      "2021-09-18 16:44:13,476 Featurized dataset not previously saved for dataset H1_std, creating new\n",
      "2021-09-18 16:44:13,487 Featurizing sample 0\n",
      "2021-09-18 16:44:18,657 Featurizing sample 1000\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "number of features: 1024\n",
      "  4%|▍         | 2/45 [00:23<05:49,  8.13s/it, best loss: 0.5152785357644352]"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "2021-09-18 16:44:20,641 Previous dataset split restored\n",
      "2021-09-18 16:44:20,659 Wrote transformers to ampl_output/H1_std/RF_ecfp_scaffold_regression/b7c14be1-5748-42ea-be43-3a07082ad6d8/transformers.pkl\n",
      "2021-09-18 16:44:20,661 Transforming response data\n",
      "2021-09-18 16:44:20,696 Transforming response data\n",
      "2021-09-18 16:44:20,712 Transforming response data\n",
      "2021-09-18 16:44:20,751 Fitting random forest model\n",
      "2021-09-18 16:44:22,156 Fold 0: training r2_score = 0.966, validation r2_score = 0.453, test r2_score = 0.543\n",
      "2021-09-18 16:44:22,578 Wrote model tarball to ampl_output/H1_std_model_b7c14be1-5748-42ea-be43-3a07082ad6d8.tar.gz\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "  7%|▋         | 3/45 [00:25<06:00,  8.59s/it, best loss: 0.5152785357644352]"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "2021-09-18 16:44:22,599 Attempting to load featurized dataset\n",
      "2021-09-18 16:44:22,609 Exception when trying to load featurized data:\n",
      "DynamicFeaturization doesn't support get_featurized_dset_name()\n",
      "2021-09-18 16:44:22,611 Featurized dataset not previously saved for dataset H1_std, creating new\n",
      "2021-09-18 16:44:22,622 Featurizing sample 0\n",
      "2021-09-18 16:44:27,634 Featurizing sample 1000\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "number of features: 1024\n",
      "  7%|▋         | 3/45 [00:32<06:00,  8.59s/it, best loss: 0.5152785357644352]"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "2021-09-18 16:44:29,581 Previous dataset split restored\n",
      "2021-09-18 16:44:29,599 Wrote transformers to ampl_output/H1_std/RF_ecfp_scaffold_regression/8b13ec2b-0d8d-4ab8-b3d5-a471370fb56e/transformers.pkl\n",
      "2021-09-18 16:44:29,600 Transforming response data\n",
      "2021-09-18 16:44:29,634 Transforming response data\n",
      "2021-09-18 16:44:29,648 Transforming response data\n",
      "2021-09-18 16:44:29,680 Fitting random forest model\n",
      "2021-09-18 16:44:30,989 Fold 0: training r2_score = 0.958, validation r2_score = 0.461, test r2_score = 0.551\n",
      "2021-09-18 16:44:31,379 Wrote model tarball to ampl_output/H1_std_model_8b13ec2b-0d8d-4ab8-b3d5-a471370fb56e.tar.gz\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "  9%|▉         | 4/45 [00:34<05:55,  8.67s/it, best loss: 0.5152785357644352]"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "2021-09-18 16:44:31,400 Attempting to load featurized dataset\n",
      "2021-09-18 16:44:31,408 Exception when trying to load featurized data:\n",
      "DynamicFeaturization doesn't support get_featurized_dset_name()\n",
      "2021-09-18 16:44:31,409 Featurized dataset not previously saved for dataset H1_std, creating new\n",
      "2021-09-18 16:44:31,421 Featurizing sample 0\n",
      "2021-09-18 16:44:36,542 Featurizing sample 1000\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "number of features: 1024\n",
      "  9%|▉         | 4/45 [00:41<05:55,  8.67s/it, best loss: 0.5152785357644352]"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "2021-09-18 16:44:38,511 Previous dataset split restored\n",
      "2021-09-18 16:44:38,530 Wrote transformers to ampl_output/H1_std/RF_ecfp_scaffold_regression/22f4a92b-d7c6-40a5-bf96-2bb130f51746/transformers.pkl\n",
      "2021-09-18 16:44:38,533 Transforming response data\n",
      "2021-09-18 16:44:38,567 Transforming response data\n",
      "2021-09-18 16:44:38,583 Transforming response data\n",
      "2021-09-18 16:44:38,616 Fitting random forest model\n",
      "2021-09-18 16:44:39,956 Fold 0: training r2_score = 0.966, validation r2_score = 0.439, test r2_score = 0.533\n",
      "2021-09-18 16:44:40,575 Wrote model tarball to ampl_output/H1_std_model_22f4a92b-d7c6-40a5-bf96-2bb130f51746.tar.gz\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      " 11%|█         | 5/45 [00:43<05:54,  8.86s/it, best loss: 0.5152785357644352]"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "2021-09-18 16:44:40,595 Attempting to load featurized dataset\n",
      "2021-09-18 16:44:40,603 Exception when trying to load featurized data:\n",
      "DynamicFeaturization doesn't support get_featurized_dset_name()\n",
      "2021-09-18 16:44:40,604 Featurized dataset not previously saved for dataset H1_std, creating new\n",
      "2021-09-18 16:44:40,615 Featurizing sample 0\n",
      "2021-09-18 16:44:45,579 Featurizing sample 1000\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "number of features: 1024\n",
      " 11%|█         | 5/45 [00:50<05:54,  8.86s/it, best loss: 0.5152785357644352]"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "2021-09-18 16:44:47,530 Previous dataset split restored\n",
      "2021-09-18 16:44:47,547 Wrote transformers to ampl_output/H1_std/RF_ecfp_scaffold_regression/d7a21b60-320a-43cd-96c2-39537b5bff1d/transformers.pkl\n",
      "2021-09-18 16:44:47,548 Transforming response data\n",
      "2021-09-18 16:44:47,582 Transforming response data\n",
      "2021-09-18 16:44:47,597 Transforming response data\n",
      "2021-09-18 16:44:47,629 Fitting random forest model\n",
      "2021-09-18 16:44:48,396 Fold 0: training r2_score = 0.964, validation r2_score = 0.447, test r2_score = 0.535\n",
      "2021-09-18 16:44:48,551 Wrote model tarball to ampl_output/H1_std_model_d7a21b60-320a-43cd-96c2-39537b5bff1d.tar.gz\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      " 13%|█▎        | 6/45 [00:51<05:33,  8.56s/it, best loss: 0.5152785357644352]"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "2021-09-18 16:44:48,570 Attempting to load featurized dataset\n",
      "2021-09-18 16:44:48,578 Exception when trying to load featurized data:\n",
      "DynamicFeaturization doesn't support get_featurized_dset_name()\n",
      "2021-09-18 16:44:48,579 Featurized dataset not previously saved for dataset H1_std, creating new\n",
      "2021-09-18 16:44:48,589 Featurizing sample 0\n",
      "2021-09-18 16:44:53,709 Featurizing sample 1000\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "number of features: 1024\n",
      " 13%|█▎        | 6/45 [00:58<05:33,  8.56s/it, best loss: 0.5152785357644352]"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "2021-09-18 16:44:55,728 Previous dataset split restored\n",
      "2021-09-18 16:44:55,745 Wrote transformers to ampl_output/H1_std/RF_ecfp_scaffold_regression/aceb15b2-0322-48eb-b510-870e1563ebed/transformers.pkl\n",
      "2021-09-18 16:44:55,746 Transforming response data\n",
      "2021-09-18 16:44:55,780 Transforming response data\n",
      "2021-09-18 16:44:55,798 Transforming response data\n",
      "2021-09-18 16:44:55,828 Fitting random forest model\n",
      "2021-09-18 16:44:58,082 Fold 0: training r2_score = 0.967, validation r2_score = 0.467, test r2_score = 0.546\n",
      "2021-09-18 16:44:58,797 Wrote model tarball to ampl_output/H1_std_model_aceb15b2-0322-48eb-b510-870e1563ebed.tar.gz\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      " 16%|█▌        | 7/45 [01:01<05:46,  9.11s/it, best loss: 0.5152785357644352]"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "2021-09-18 16:44:58,818 Attempting to load featurized dataset\n",
      "2021-09-18 16:44:58,826 Exception when trying to load featurized data:\n",
      "DynamicFeaturization doesn't support get_featurized_dset_name()\n",
      "2021-09-18 16:44:58,828 Featurized dataset not previously saved for dataset H1_std, creating new\n",
      "2021-09-18 16:44:58,841 Featurizing sample 0\n",
      "2021-09-18 16:45:03,785 Featurizing sample 1000\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "number of features: 1024\n",
      " 16%|█▌        | 7/45 [01:08<05:46,  9.11s/it, best loss: 0.5152785357644352]"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "2021-09-18 16:45:05,745 Previous dataset split restored\n",
      "2021-09-18 16:45:05,762 Wrote transformers to ampl_output/H1_std/RF_ecfp_scaffold_regression/5e4be55f-8462-4f15-b222-24f91ee5bcc5/transformers.pkl\n",
      "2021-09-18 16:45:05,764 Transforming response data\n",
      "2021-09-18 16:45:05,797 Transforming response data\n",
      "2021-09-18 16:45:05,813 Transforming response data\n",
      "2021-09-18 16:45:05,843 Fitting random forest model\n",
      "2021-09-18 16:45:07,768 Fold 0: training r2_score = 0.966, validation r2_score = 0.464, test r2_score = 0.534\n",
      "2021-09-18 16:45:08,317 Wrote model tarball to ampl_output/H1_std_model_5e4be55f-8462-4f15-b222-24f91ee5bcc5.tar.gz\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      " 18%|█▊        | 8/45 [01:11<05:41,  9.24s/it, best loss: 0.5152785357644352]"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "2021-09-18 16:45:08,338 Attempting to load featurized dataset\n",
      "2021-09-18 16:45:08,347 Exception when trying to load featurized data:\n",
      "DynamicFeaturization doesn't support get_featurized_dset_name()\n",
      "2021-09-18 16:45:08,349 Featurized dataset not previously saved for dataset H1_std, creating new\n",
      "2021-09-18 16:45:08,361 Featurizing sample 0\n",
      "2021-09-18 16:45:13,525 Featurizing sample 1000\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "number of features: 1024\n",
      " 18%|█▊        | 8/45 [01:18<05:41,  9.24s/it, best loss: 0.5152785357644352]"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "2021-09-18 16:45:15,536 Previous dataset split restored\n",
      "2021-09-18 16:45:15,554 Wrote transformers to ampl_output/H1_std/RF_ecfp_scaffold_regression/d021982f-04ba-44cc-adac-ab0a6806f9e6/transformers.pkl\n",
      "2021-09-18 16:45:15,556 Transforming response data\n",
      "2021-09-18 16:45:15,587 Transforming response data\n",
      "2021-09-18 16:45:15,602 Transforming response data\n",
      "2021-09-18 16:45:15,635 Fitting random forest model\n",
      "2021-09-18 16:45:17,266 Fold 0: training r2_score = 0.966, validation r2_score = 0.473, test r2_score = 0.550\n",
      "2021-09-18 16:45:17,767 Wrote model tarball to ampl_output/H1_std_model_d021982f-04ba-44cc-adac-ab0a6806f9e6.tar.gz\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      " 20%|██        | 9/45 [01:20<05:35,  9.31s/it, best loss: 0.5152785357644352]"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "2021-09-18 16:45:17,792 Attempting to load featurized dataset\n",
      "2021-09-18 16:45:17,801 Exception when trying to load featurized data:\n",
      "DynamicFeaturization doesn't support get_featurized_dset_name()\n",
      "2021-09-18 16:45:17,802 Featurized dataset not previously saved for dataset H1_std, creating new\n",
      "2021-09-18 16:45:17,815 Featurizing sample 0\n",
      "2021-09-18 16:45:22,775 Featurizing sample 1000\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "number of features: 1024\n",
      " 20%|██        | 9/45 [01:27<05:35,  9.31s/it, best loss: 0.5152785357644352]"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "2021-09-18 16:45:24,718 Previous dataset split restored\n",
      "2021-09-18 16:45:24,735 Wrote transformers to ampl_output/H1_std/RF_ecfp_scaffold_regression/f77a8656-5fed-47c0-8d2a-ff5bc7048f5c/transformers.pkl\n",
      "2021-09-18 16:45:24,736 Transforming response data\n",
      "2021-09-18 16:45:24,771 Transforming response data\n",
      "2021-09-18 16:45:24,785 Transforming response data\n",
      "2021-09-18 16:45:24,815 Fitting random forest model\n",
      "2021-09-18 16:45:25,898 Fold 0: training r2_score = 0.966, validation r2_score = 0.441, test r2_score = 0.535\n",
      "2021-09-18 16:45:26,130 Wrote model tarball to ampl_output/H1_std_model_f77a8656-5fed-47c0-8d2a-ff5bc7048f5c.tar.gz\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      " 22%|██▏       | 10/45 [01:29<05:15,  9.01s/it, best loss: 0.5152785357644352]"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "2021-09-18 16:45:26,150 Attempting to load featurized dataset\n",
      "2021-09-18 16:45:26,159 Exception when trying to load featurized data:\n",
      "DynamicFeaturization doesn't support get_featurized_dset_name()\n",
      "2021-09-18 16:45:26,161 Featurized dataset not previously saved for dataset H1_std, creating new\n",
      "2021-09-18 16:45:26,172 Featurizing sample 0\n",
      "2021-09-18 16:45:31,326 Featurizing sample 1000\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "number of features: 1024\n",
      " 22%|██▏       | 10/45 [01:36<05:15,  9.01s/it, best loss: 0.5152785357644352]"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "2021-09-18 16:45:33,341 Previous dataset split restored\n",
      "2021-09-18 16:45:33,357 Wrote transformers to ampl_output/H1_std/RF_ecfp_scaffold_regression/9b2e31b1-cb2d-4b86-aeb8-7395a7e67485/transformers.pkl\n",
      "2021-09-18 16:45:33,359 Transforming response data\n",
      "2021-09-18 16:45:33,394 Transforming response data\n",
      "2021-09-18 16:45:33,411 Transforming response data\n",
      "2021-09-18 16:45:33,446 Fitting random forest model\n",
      "2021-09-18 16:45:34,227 Fold 0: training r2_score = 0.964, validation r2_score = 0.404, test r2_score = 0.505\n",
      "2021-09-18 16:45:34,491 Wrote model tarball to ampl_output/H1_std_model_9b2e31b1-cb2d-4b86-aeb8-7395a7e67485.tar.gz\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      " 24%|██▍       | 11/45 [01:37<04:59,  8.81s/it, best loss: 0.5152785357644352]"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "2021-09-18 16:45:34,514 Attempting to load featurized dataset\n",
      "2021-09-18 16:45:34,525 Exception when trying to load featurized data:\n",
      "DynamicFeaturization doesn't support get_featurized_dset_name()\n",
      "2021-09-18 16:45:34,527 Featurized dataset not previously saved for dataset H1_std, creating new\n",
      "2021-09-18 16:45:34,540 Featurizing sample 0\n",
      "2021-09-18 16:45:39,682 Featurizing sample 1000\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "number of features: 1024\n",
      " 24%|██▍       | 11/45 [01:44<04:59,  8.81s/it, best loss: 0.5152785357644352]"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "2021-09-18 16:45:41,638 Previous dataset split restored\n",
      "2021-09-18 16:45:41,655 Wrote transformers to ampl_output/H1_std/RF_ecfp_scaffold_regression/f14776d6-b99f-4b2a-a996-223d3d2b5403/transformers.pkl\n",
      "2021-09-18 16:45:41,657 Transforming response data\n",
      "2021-09-18 16:45:41,690 Transforming response data\n",
      "2021-09-18 16:45:41,706 Transforming response data\n",
      "2021-09-18 16:45:41,736 Fitting random forest model\n",
      "2021-09-18 16:45:43,057 Fold 0: training r2_score = 0.967, validation r2_score = 0.450, test r2_score = 0.542\n",
      "2021-09-18 16:45:43,607 Wrote model tarball to ampl_output/H1_std_model_f14776d6-b99f-4b2a-a996-223d3d2b5403.tar.gz\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      " 27%|██▋       | 12/45 [01:46<04:53,  8.91s/it, best loss: 0.5152785357644352]"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "2021-09-18 16:45:43,627 Attempting to load featurized dataset\n",
      "2021-09-18 16:45:43,635 Exception when trying to load featurized data:\n",
      "DynamicFeaturization doesn't support get_featurized_dset_name()\n",
      "2021-09-18 16:45:43,636 Featurized dataset not previously saved for dataset H1_std, creating new\n",
      "2021-09-18 16:45:43,646 Featurizing sample 0\n",
      "2021-09-18 16:45:48,670 Featurizing sample 1000\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "number of features: 1024\n",
      " 27%|██▋       | 12/45 [01:53<04:53,  8.91s/it, best loss: 0.5152785357644352]"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "2021-09-18 16:45:50,694 Previous dataset split restored\n",
      "2021-09-18 16:45:50,710 Wrote transformers to ampl_output/H1_std/RF_ecfp_scaffold_regression/5a55f8ef-5504-4391-bc0f-aa3a0b999cd7/transformers.pkl\n",
      "2021-09-18 16:45:50,712 Transforming response data\n",
      "2021-09-18 16:45:50,746 Transforming response data\n",
      "2021-09-18 16:45:50,761 Transforming response data\n",
      "2021-09-18 16:45:50,791 Fitting random forest model\n",
      "2021-09-18 16:45:52,304 Fold 0: training r2_score = 0.965, validation r2_score = 0.452, test r2_score = 0.519\n",
      "2021-09-18 16:45:52,742 Wrote model tarball to ampl_output/H1_std_model_5a55f8ef-5504-4391-bc0f-aa3a0b999cd7.tar.gz\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      " 29%|██▉       | 13/45 [01:55<04:47,  8.98s/it, best loss: 0.5152785357644352]"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "2021-09-18 16:45:53,139 Attempting to load featurized dataset\n",
      "2021-09-18 16:45:53,154 Exception when trying to load featurized data:\n",
      "DynamicFeaturization doesn't support get_featurized_dset_name()\n",
      "2021-09-18 16:45:53,157 Featurized dataset not previously saved for dataset H1_std, creating new\n",
      "2021-09-18 16:45:53,168 Featurizing sample 0\n",
      "2021-09-18 16:45:58,261 Featurizing sample 1000\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "number of features: 1024\n",
      " 29%|██▉       | 13/45 [02:02<04:47,  8.98s/it, best loss: 0.5152785357644352]"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "2021-09-18 16:46:00,204 Previous dataset split restored\n",
      "2021-09-18 16:46:00,221 Wrote transformers to ampl_output/H1_std/RF_ecfp_scaffold_regression/dde384b3-20d7-49fa-9d6d-a2dde800dd4d/transformers.pkl\n",
      "2021-09-18 16:46:00,222 Transforming response data\n",
      "2021-09-18 16:46:00,255 Transforming response data\n",
      "2021-09-18 16:46:00,270 Transforming response data\n",
      "2021-09-18 16:46:00,301 Fitting random forest model\n",
      "2021-09-18 16:46:00,759 Fold 0: training r2_score = 0.957, validation r2_score = 0.382, test r2_score = 0.485\n",
      "2021-09-18 16:46:00,824 Wrote model tarball to ampl_output/H1_std_model_dde384b3-20d7-49fa-9d6d-a2dde800dd4d.tar.gz\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      " 31%|███       | 14/45 [02:03<04:29,  8.71s/it, best loss: 0.5152785357644352]"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "2021-09-18 16:46:00,841 Attempting to load featurized dataset\n",
      "2021-09-18 16:46:00,849 Exception when trying to load featurized data:\n",
      "DynamicFeaturization doesn't support get_featurized_dset_name()\n",
      "2021-09-18 16:46:00,852 Featurized dataset not previously saved for dataset H1_std, creating new\n",
      "2021-09-18 16:46:00,862 Featurizing sample 0\n",
      "2021-09-18 16:46:05,840 Featurizing sample 1000\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "number of features: 1024\n",
      " 31%|███       | 14/45 [02:10<04:29,  8.71s/it, best loss: 0.5152785357644352]"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "2021-09-18 16:46:07,817 Previous dataset split restored\n",
      "2021-09-18 16:46:07,834 Wrote transformers to ampl_output/H1_std/RF_ecfp_scaffold_regression/eec709e3-0ff2-46e7-a0e4-86fafe323360/transformers.pkl\n",
      "2021-09-18 16:46:07,837 Transforming response data\n",
      "2021-09-18 16:46:07,873 Transforming response data\n",
      "2021-09-18 16:46:07,890 Transforming response data\n",
      "2021-09-18 16:46:07,924 Fitting random forest model\n",
      "2021-09-18 16:46:08,900 Fold 0: training r2_score = 0.965, validation r2_score = 0.472, test r2_score = 0.550\n",
      "2021-09-18 16:46:09,090 Wrote model tarball to ampl_output/H1_std_model_eec709e3-0ff2-46e7-a0e4-86fafe323360.tar.gz\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      " 33%|███▎      | 15/45 [02:11<04:17,  8.57s/it, best loss: 0.5152785357644352]"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "2021-09-18 16:46:09,122 Attempting to load featurized dataset\n",
      "2021-09-18 16:46:09,130 Exception when trying to load featurized data:\n",
      "DynamicFeaturization doesn't support get_featurized_dset_name()\n",
      "2021-09-18 16:46:09,132 Featurized dataset not previously saved for dataset H1_std, creating new\n",
      "2021-09-18 16:46:09,145 Featurizing sample 0\n",
      "2021-09-18 16:46:14,405 Featurizing sample 1000\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "number of features: 1024\n",
      " 33%|███▎      | 15/45 [02:19<04:17,  8.57s/it, best loss: 0.5152785357644352]"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "2021-09-18 16:46:16,392 Previous dataset split restored\n",
      "2021-09-18 16:46:16,409 Wrote transformers to ampl_output/H1_std/RF_ecfp_scaffold_regression/16984aa8-1e7a-4ee5-ba37-6c4465012641/transformers.pkl\n",
      "2021-09-18 16:46:16,411 Transforming response data\n",
      "2021-09-18 16:46:16,446 Transforming response data\n",
      "2021-09-18 16:46:16,461 Transforming response data\n",
      "2021-09-18 16:46:16,492 Fitting random forest model\n",
      "2021-09-18 16:46:17,481 Fold 0: training r2_score = 0.966, validation r2_score = 0.438, test r2_score = 0.529\n",
      "2021-09-18 16:46:17,820 Wrote model tarball to ampl_output/H1_std_model_16984aa8-1e7a-4ee5-ba37-6c4465012641.tar.gz\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      " 36%|███▌      | 16/45 [02:20<04:09,  8.62s/it, best loss: 0.5152785357644352]"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "2021-09-18 16:46:17,855 Attempting to load featurized dataset\n",
      "2021-09-18 16:46:17,864 Exception when trying to load featurized data:\n",
      "DynamicFeaturization doesn't support get_featurized_dset_name()\n",
      "2021-09-18 16:46:17,866 Featurized dataset not previously saved for dataset H1_std, creating new\n",
      "2021-09-18 16:46:17,880 Featurizing sample 0\n",
      "2021-09-18 16:46:22,869 Featurizing sample 1000\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "number of features: 1024\n",
      " 36%|███▌      | 16/45 [02:27<04:09,  8.62s/it, best loss: 0.5152785357644352]"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "2021-09-18 16:46:24,826 Previous dataset split restored\n",
      "2021-09-18 16:46:24,845 Wrote transformers to ampl_output/H1_std/RF_ecfp_scaffold_regression/e71edba3-1fdd-4dea-918c-ae5ba92f5bbe/transformers.pkl\n",
      "2021-09-18 16:46:24,846 Transforming response data\n",
      "2021-09-18 16:46:24,879 Transforming response data\n",
      "2021-09-18 16:46:24,896 Transforming response data\n",
      "2021-09-18 16:46:24,926 Fitting random forest model\n",
      "2021-09-18 16:46:26,023 Fold 0: training r2_score = 0.966, validation r2_score = 0.459, test r2_score = 0.527\n",
      "2021-09-18 16:46:26,392 Wrote model tarball to ampl_output/H1_std_model_e71edba3-1fdd-4dea-918c-ae5ba92f5bbe.tar.gz\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      " 38%|███▊      | 17/45 [02:29<04:00,  8.61s/it, best loss: 0.5152785357644352]"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "2021-09-18 16:46:26,419 Attempting to load featurized dataset\n",
      "2021-09-18 16:46:26,428 Exception when trying to load featurized data:\n",
      "DynamicFeaturization doesn't support get_featurized_dset_name()\n",
      "2021-09-18 16:46:26,430 Featurized dataset not previously saved for dataset H1_std, creating new\n",
      "2021-09-18 16:46:26,441 Featurizing sample 0\n",
      "2021-09-18 16:46:31,681 Featurizing sample 1000\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "number of features: 1024\n",
      " 38%|███▊      | 17/45 [02:36<04:00,  8.61s/it, best loss: 0.5152785357644352]"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "2021-09-18 16:46:33,707 Previous dataset split restored\n",
      "2021-09-18 16:46:33,726 Wrote transformers to ampl_output/H1_std/RF_ecfp_scaffold_regression/8ecc54cc-3536-4800-97cb-5dfe0bc1022a/transformers.pkl\n",
      "2021-09-18 16:46:33,729 Transforming response data\n",
      "2021-09-18 16:46:33,762 Transforming response data\n",
      "2021-09-18 16:46:33,778 Transforming response data\n",
      "2021-09-18 16:46:33,809 Fitting random forest model\n",
      "2021-09-18 16:46:35,848 Fold 0: training r2_score = 0.966, validation r2_score = 0.467, test r2_score = 0.550\n",
      "2021-09-18 16:46:36,465 Wrote model tarball to ampl_output/H1_std_model_8ecc54cc-3536-4800-97cb-5dfe0bc1022a.tar.gz\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      " 40%|████      | 18/45 [02:39<04:04,  9.05s/it, best loss: 0.5152785357644352]"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "2021-09-18 16:46:36,493 Attempting to load featurized dataset\n",
      "2021-09-18 16:46:36,501 Exception when trying to load featurized data:\n",
      "DynamicFeaturization doesn't support get_featurized_dset_name()\n",
      "2021-09-18 16:46:36,502 Featurized dataset not previously saved for dataset H1_std, creating new\n",
      "2021-09-18 16:46:36,513 Featurizing sample 0\n",
      "2021-09-18 16:46:41,541 Featurizing sample 1000\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "number of features: 1024\n",
      " 40%|████      | 18/45 [02:46<04:04,  9.05s/it, best loss: 0.5152785357644352]"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "2021-09-18 16:46:43,499 Previous dataset split restored\n",
      "2021-09-18 16:46:43,518 Wrote transformers to ampl_output/H1_std/RF_ecfp_scaffold_regression/e613fe42-82bd-4410-99d4-98b4ef22260f/transformers.pkl\n",
      "2021-09-18 16:46:43,520 Transforming response data\n",
      "2021-09-18 16:46:43,554 Transforming response data\n",
      "2021-09-18 16:46:43,570 Transforming response data\n",
      "2021-09-18 16:46:43,600 Fitting random forest model\n",
      "2021-09-18 16:46:44,158 Fold 0: training r2_score = 0.960, validation r2_score = 0.437, test r2_score = 0.566\n",
      "2021-09-18 16:46:44,229 Wrote model tarball to ampl_output/H1_std_model_e613fe42-82bd-4410-99d4-98b4ef22260f.tar.gz\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      " 42%|████▏     | 19/45 [02:47<03:45,  8.66s/it, best loss: 0.5152785357644352]"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "2021-09-18 16:46:44,254 Attempting to load featurized dataset\n",
      "2021-09-18 16:46:44,263 Exception when trying to load featurized data:\n",
      "DynamicFeaturization doesn't support get_featurized_dset_name()\n",
      "2021-09-18 16:46:44,264 Featurized dataset not previously saved for dataset H1_std, creating new\n",
      "2021-09-18 16:46:44,276 Featurizing sample 0\n",
      "2021-09-18 16:46:49,429 Featurizing sample 1000\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "number of features: 1024\n",
      " 42%|████▏     | 19/45 [02:54<03:45,  8.66s/it, best loss: 0.5152785357644352]"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "2021-09-18 16:46:51,461 Previous dataset split restored\n",
      "2021-09-18 16:46:51,478 Wrote transformers to ampl_output/H1_std/RF_ecfp_scaffold_regression/2b5531ac-38ec-44bb-ae77-302c6bf6baf8/transformers.pkl\n",
      "2021-09-18 16:46:51,480 Transforming response data\n",
      "2021-09-18 16:46:51,513 Transforming response data\n",
      "2021-09-18 16:46:51,530 Transforming response data\n",
      "2021-09-18 16:46:51,562 Fitting random forest model\n",
      "2021-09-18 16:46:53,304 Fold 0: training r2_score = 0.966, validation r2_score = 0.448, test r2_score = 0.531\n",
      "2021-09-18 16:46:53,952 Wrote model tarball to ampl_output/H1_std_model_2b5531ac-38ec-44bb-ae77-302c6bf6baf8.tar.gz\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      " 44%|████▍     | 20/45 [02:56<03:44,  8.98s/it, best loss: 0.5152785357644352]"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "2021-09-18 16:46:53,980 Attempting to load featurized dataset\n",
      "2021-09-18 16:46:53,990 Exception when trying to load featurized data:\n",
      "DynamicFeaturization doesn't support get_featurized_dset_name()\n",
      "2021-09-18 16:46:53,991 Featurized dataset not previously saved for dataset H1_std, creating new\n",
      "2021-09-18 16:46:54,003 Featurizing sample 0\n",
      "2021-09-18 16:46:59,040 Featurizing sample 1000\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "number of features: 1024\n",
      " 44%|████▍     | 20/45 [03:03<03:44,  8.98s/it, best loss: 0.5152785357644352]"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "2021-09-18 16:47:00,984 Previous dataset split restored\n",
      "2021-09-18 16:47:01,000 Wrote transformers to ampl_output/H1_std/RF_ecfp_scaffold_regression/e79b0820-591a-4ff3-978d-7e085e77684c/transformers.pkl\n",
      "2021-09-18 16:47:01,002 Transforming response data\n",
      "2021-09-18 16:47:01,035 Transforming response data\n",
      "2021-09-18 16:47:01,050 Transforming response data\n",
      "2021-09-18 16:47:01,084 Fitting random forest model\n",
      "2021-09-18 16:47:02,072 Fold 0: training r2_score = 0.966, validation r2_score = 0.474, test r2_score = 0.550\n",
      "2021-09-18 16:47:02,402 Wrote model tarball to ampl_output/H1_std_model_e79b0820-591a-4ff3-978d-7e085e77684c.tar.gz\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      " 47%|████▋     | 21/45 [03:05<03:31,  8.82s/it, best loss: 0.5152785357644352]"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "2021-09-18 16:47:02,431 Attempting to load featurized dataset\n",
      "2021-09-18 16:47:02,441 Exception when trying to load featurized data:\n",
      "DynamicFeaturization doesn't support get_featurized_dset_name()\n",
      "2021-09-18 16:47:02,443 Featurized dataset not previously saved for dataset H1_std, creating new\n",
      "2021-09-18 16:47:02,454 Featurizing sample 0\n",
      "2021-09-18 16:47:07,566 Featurizing sample 1000\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "number of features: 1024\n",
      " 47%|████▋     | 21/45 [03:12<03:31,  8.82s/it, best loss: 0.5152785357644352]"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "2021-09-18 16:47:09,605 Previous dataset split restored\n",
      "2021-09-18 16:47:09,623 Wrote transformers to ampl_output/H1_std/RF_ecfp_scaffold_regression/51b81b6b-8afe-4341-9e47-e23a5728105b/transformers.pkl\n",
      "2021-09-18 16:47:09,624 Transforming response data\n",
      "2021-09-18 16:47:09,658 Transforming response data\n",
      "2021-09-18 16:47:09,673 Transforming response data\n",
      "2021-09-18 16:47:09,704 Fitting random forest model\n",
      "2021-09-18 16:47:10,268 Fold 0: training r2_score = 0.960, validation r2_score = 0.437, test r2_score = 0.523\n",
      "2021-09-18 16:47:10,378 Wrote model tarball to ampl_output/H1_std_model_51b81b6b-8afe-4341-9e47-e23a5728105b.tar.gz\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      " 49%|████▉     | 22/45 [03:13<03:17,  8.57s/it, best loss: 0.5152785357644352]"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "2021-09-18 16:47:10,406 Attempting to load featurized dataset\n",
      "2021-09-18 16:47:10,417 Exception when trying to load featurized data:\n",
      "DynamicFeaturization doesn't support get_featurized_dset_name()\n",
      "2021-09-18 16:47:10,418 Featurized dataset not previously saved for dataset H1_std, creating new\n",
      "2021-09-18 16:47:10,432 Featurizing sample 0\n",
      "2021-09-18 16:47:15,594 Featurizing sample 1000\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "number of features: 1024\n",
      " 49%|████▉     | 22/45 [03:20<03:17,  8.57s/it, best loss: 0.5152785357644352]"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "2021-09-18 16:47:17,523 Previous dataset split restored\n",
      "2021-09-18 16:47:17,541 Wrote transformers to ampl_output/H1_std/RF_ecfp_scaffold_regression/a3564911-5139-4145-9ad7-98cccbe0913e/transformers.pkl\n",
      "2021-09-18 16:47:17,542 Transforming response data\n",
      "2021-09-18 16:47:17,576 Transforming response data\n",
      "2021-09-18 16:47:17,591 Transforming response data\n",
      "2021-09-18 16:47:17,622 Fitting random forest model\n",
      "2021-09-18 16:47:18,407 Fold 0: training r2_score = 0.965, validation r2_score = 0.415, test r2_score = 0.504\n",
      "2021-09-18 16:47:18,738 Wrote model tarball to ampl_output/H1_std_model_a3564911-5139-4145-9ad7-98cccbe0913e.tar.gz\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      " 51%|█████     | 23/45 [03:21<03:07,  8.51s/it, best loss: 0.5152785357644352]"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "2021-09-18 16:47:18,763 Attempting to load featurized dataset\n",
      "2021-09-18 16:47:18,772 Exception when trying to load featurized data:\n",
      "DynamicFeaturization doesn't support get_featurized_dset_name()\n",
      "2021-09-18 16:47:18,773 Featurized dataset not previously saved for dataset H1_std, creating new\n",
      "2021-09-18 16:47:18,785 Featurizing sample 0\n",
      "2021-09-18 16:47:23,789 Featurizing sample 1000\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "number of features: 1024\n",
      " 51%|█████     | 23/45 [03:28<03:07,  8.51s/it, best loss: 0.5152785357644352]"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "2021-09-18 16:47:25,717 Previous dataset split restored\n",
      "2021-09-18 16:47:25,734 Wrote transformers to ampl_output/H1_std/RF_ecfp_scaffold_regression/035abf27-3d7e-4a8c-a236-6e5984abb6d4/transformers.pkl\n",
      "2021-09-18 16:47:25,736 Transforming response data\n",
      "2021-09-18 16:47:25,765 Transforming response data\n",
      "2021-09-18 16:47:25,779 Transforming response data\n",
      "2021-09-18 16:47:25,811 Fitting random forest model\n",
      "2021-09-18 16:47:26,380 Fold 0: training r2_score = 0.962, validation r2_score = 0.440, test r2_score = 0.549\n",
      "2021-09-18 16:47:26,503 Wrote model tarball to ampl_output/H1_std_model_035abf27-3d7e-4a8c-a236-6e5984abb6d4.tar.gz\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      " 53%|█████▎    | 24/45 [03:29<02:53,  8.28s/it, best loss: 0.5152785357644352]"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "2021-09-18 16:47:26,530 Attempting to load featurized dataset\n",
      "2021-09-18 16:47:26,539 Exception when trying to load featurized data:\n",
      "DynamicFeaturization doesn't support get_featurized_dset_name()\n",
      "2021-09-18 16:47:26,540 Featurized dataset not previously saved for dataset H1_std, creating new\n",
      "2021-09-18 16:47:26,551 Featurizing sample 0\n",
      "2021-09-18 16:47:31,831 Featurizing sample 1000\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "number of features: 1024\n",
      " 53%|█████▎    | 24/45 [03:36<02:53,  8.28s/it, best loss: 0.5152785357644352]"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "2021-09-18 16:47:33,864 Previous dataset split restored\n",
      "2021-09-18 16:47:33,882 Wrote transformers to ampl_output/H1_std/RF_ecfp_scaffold_regression/c69f6680-8183-4ad2-8723-0f0822aaa1db/transformers.pkl\n",
      "2021-09-18 16:47:33,883 Transforming response data\n",
      "2021-09-18 16:47:33,912 Transforming response data\n",
      "2021-09-18 16:47:33,928 Transforming response data\n",
      "2021-09-18 16:47:33,961 Fitting random forest model\n",
      "2021-09-18 16:47:35,052 Fold 0: training r2_score = 0.965, validation r2_score = 0.463, test r2_score = 0.543\n",
      "2021-09-18 16:47:35,357 Wrote model tarball to ampl_output/H1_std_model_c69f6680-8183-4ad2-8723-0f0822aaa1db.tar.gz\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      " 56%|█████▌    | 25/45 [03:38<02:49,  8.45s/it, best loss: 0.5152785357644352]"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "2021-09-18 16:47:35,388 Attempting to load featurized dataset\n",
      "2021-09-18 16:47:35,398 Exception when trying to load featurized data:\n",
      "DynamicFeaturization doesn't support get_featurized_dset_name()\n",
      "2021-09-18 16:47:35,400 Featurized dataset not previously saved for dataset H1_std, creating new\n",
      "2021-09-18 16:47:35,412 Featurizing sample 0\n",
      "2021-09-18 16:47:40,514 Featurizing sample 1000\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "number of features: 1024\n",
      " 56%|█████▌    | 25/45 [03:45<02:49,  8.45s/it, best loss: 0.5152785357644352]"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "2021-09-18 16:47:42,454 Previous dataset split restored\n",
      "2021-09-18 16:47:42,473 Wrote transformers to ampl_output/H1_std/RF_ecfp_scaffold_regression/1c55e483-222e-429a-ab3d-ff20a27a6d99/transformers.pkl\n",
      "2021-09-18 16:47:42,475 Transforming response data\n",
      "2021-09-18 16:47:42,502 Transforming response data\n",
      "2021-09-18 16:47:42,516 Transforming response data\n",
      "2021-09-18 16:47:42,550 Fitting random forest model\n",
      "2021-09-18 16:47:43,002 Fold 0: training r2_score = 0.951, validation r2_score = 0.361, test r2_score = 0.468\n",
      "2021-09-18 16:47:43,038 Wrote model tarball to ampl_output/H1_std_model_1c55e483-222e-429a-ab3d-ff20a27a6d99.tar.gz\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      " 58%|█████▊    | 26/45 [03:45<02:36,  8.22s/it, best loss: 0.5152785357644352]"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "2021-09-18 16:47:43,068 Attempting to load featurized dataset\n",
      "2021-09-18 16:47:43,079 Exception when trying to load featurized data:\n",
      "DynamicFeaturization doesn't support get_featurized_dset_name()\n",
      "2021-09-18 16:47:43,080 Featurized dataset not previously saved for dataset H1_std, creating new\n",
      "2021-09-18 16:47:43,092 Featurizing sample 0\n",
      "2021-09-18 16:47:48,216 Featurizing sample 1000\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "number of features: 1024\n",
      " 58%|█████▊    | 26/45 [03:52<02:36,  8.22s/it, best loss: 0.5152785357644352]"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "2021-09-18 16:47:50,221 Previous dataset split restored\n",
      "2021-09-18 16:47:50,238 Wrote transformers to ampl_output/H1_std/RF_ecfp_scaffold_regression/b4b353d1-9691-44cb-8466-9bb8aab8d12a/transformers.pkl\n",
      "2021-09-18 16:47:50,240 Transforming response data\n",
      "2021-09-18 16:47:50,270 Transforming response data\n",
      "2021-09-18 16:47:50,287 Transforming response data\n",
      "2021-09-18 16:47:50,321 Fitting random forest model\n",
      "2021-09-18 16:47:51,008 Fold 0: training r2_score = 0.741, validation r2_score = 0.276, test r2_score = 0.380\n",
      "2021-09-18 16:47:51,135 Wrote model tarball to ampl_output/H1_std_model_b4b353d1-9691-44cb-8466-9bb8aab8d12a.tar.gz\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      " 60%|██████    | 27/45 [03:54<02:27,  8.19s/it, best loss: 0.5152785357644352]"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "2021-09-18 16:47:51,165 Attempting to load featurized dataset\n",
      "2021-09-18 16:47:51,174 Exception when trying to load featurized data:\n",
      "DynamicFeaturization doesn't support get_featurized_dset_name()\n",
      "2021-09-18 16:47:51,175 Featurized dataset not previously saved for dataset H1_std, creating new\n",
      "2021-09-18 16:47:51,188 Featurizing sample 0\n",
      "2021-09-18 16:47:56,286 Featurizing sample 1000\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "number of features: 1024\n",
      " 60%|██████    | 27/45 [04:00<02:27,  8.19s/it, best loss: 0.5152785357644352]"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "2021-09-18 16:47:58,185 Previous dataset split restored\n",
      "2021-09-18 16:47:58,204 Wrote transformers to ampl_output/H1_std/RF_ecfp_scaffold_regression/7508f31b-1ff9-40f5-8534-5d4f53ed123c/transformers.pkl\n",
      "2021-09-18 16:47:58,205 Transforming response data\n",
      "2021-09-18 16:47:58,233 Transforming response data\n",
      "2021-09-18 16:47:58,247 Transforming response data\n",
      "2021-09-18 16:47:58,278 Fitting random forest model\n",
      "2021-09-18 16:47:59,482 Fold 0: training r2_score = 0.966, validation r2_score = 0.457, test r2_score = 0.529\n",
      "2021-09-18 16:47:59,865 Wrote model tarball to ampl_output/H1_std_model_7508f31b-1ff9-40f5-8534-5d4f53ed123c.tar.gz\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      " 62%|██████▏   | 28/45 [04:02<02:21,  8.35s/it, best loss: 0.5152785357644352]"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "2021-09-18 16:47:59,891 Attempting to load featurized dataset\n",
      "2021-09-18 16:47:59,900 Exception when trying to load featurized data:\n",
      "DynamicFeaturization doesn't support get_featurized_dset_name()\n",
      "2021-09-18 16:47:59,901 Featurized dataset not previously saved for dataset H1_std, creating new\n",
      "2021-09-18 16:47:59,912 Featurizing sample 0\n",
      "2021-09-18 16:48:04,919 Featurizing sample 1000\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "number of features: 1024\n",
      " 62%|██████▏   | 28/45 [04:09<02:21,  8.35s/it, best loss: 0.5152785357644352]"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "2021-09-18 16:48:06,938 Previous dataset split restored\n",
      "2021-09-18 16:48:06,956 Wrote transformers to ampl_output/H1_std/RF_ecfp_scaffold_regression/278e3396-a424-4109-94ac-9cee37758b36/transformers.pkl\n",
      "2021-09-18 16:48:06,958 Transforming response data\n",
      "2021-09-18 16:48:06,989 Transforming response data\n",
      "2021-09-18 16:48:07,003 Transforming response data\n",
      "2021-09-18 16:48:07,036 Fitting random forest model\n",
      "2021-09-18 16:48:07,816 Fold 0: training r2_score = 0.965, validation r2_score = 0.448, test r2_score = 0.521\n",
      "2021-09-18 16:48:08,026 Wrote model tarball to ampl_output/H1_std_model_278e3396-a424-4109-94ac-9cee37758b36.tar.gz\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      " 64%|██████▍   | 29/45 [04:10<02:12,  8.29s/it, best loss: 0.5152785357644352]"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "2021-09-18 16:48:08,054 Attempting to load featurized dataset\n",
      "2021-09-18 16:48:08,066 Exception when trying to load featurized data:\n",
      "DynamicFeaturization doesn't support get_featurized_dset_name()\n",
      "2021-09-18 16:48:08,067 Featurized dataset not previously saved for dataset H1_std, creating new\n",
      "2021-09-18 16:48:08,080 Featurizing sample 0\n",
      "2021-09-18 16:48:13,303 Featurizing sample 1000\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "number of features: 1024\n",
      " 64%|██████▍   | 29/45 [04:18<02:12,  8.29s/it, best loss: 0.5152785357644352]"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "2021-09-18 16:48:15,271 Previous dataset split restored\n",
      "2021-09-18 16:48:15,288 Wrote transformers to ampl_output/H1_std/RF_ecfp_scaffold_regression/03874e07-5f49-4ce7-8d6b-ce02b1a3c3ee/transformers.pkl\n",
      "2021-09-18 16:48:15,290 Transforming response data\n",
      "2021-09-18 16:48:15,318 Transforming response data\n",
      "2021-09-18 16:48:15,333 Transforming response data\n",
      "2021-09-18 16:48:15,364 Fitting random forest model\n",
      "2021-09-18 16:48:16,027 Fold 0: training r2_score = 0.965, validation r2_score = 0.424, test r2_score = 0.516\n",
      "2021-09-18 16:48:16,155 Wrote model tarball to ampl_output/H1_std_model_03874e07-5f49-4ce7-8d6b-ce02b1a3c3ee.tar.gz\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      " 67%|██████▋   | 30/45 [04:19<02:03,  8.24s/it, best loss: 0.5152785357644352]"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "2021-09-18 16:48:16,179 Attempting to load featurized dataset\n",
      "2021-09-18 16:48:16,188 Exception when trying to load featurized data:\n",
      "DynamicFeaturization doesn't support get_featurized_dset_name()\n",
      "2021-09-18 16:48:16,189 Featurized dataset not previously saved for dataset H1_std, creating new\n",
      "2021-09-18 16:48:16,200 Featurizing sample 0\n",
      "2021-09-18 16:48:21,181 Featurizing sample 1000\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "number of features: 1024\n",
      " 67%|██████▋   | 30/45 [04:25<02:03,  8.24s/it, best loss: 0.5152785357644352]"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "2021-09-18 16:48:23,081 Previous dataset split restored\n",
      "2021-09-18 16:48:23,101 Wrote transformers to ampl_output/H1_std/RF_ecfp_scaffold_regression/3421f11b-0fea-45c5-b494-f0783c4488d8/transformers.pkl\n",
      "2021-09-18 16:48:23,102 Transforming response data\n",
      "2021-09-18 16:48:23,129 Transforming response data\n",
      "2021-09-18 16:48:23,143 Transforming response data\n",
      "2021-09-18 16:48:23,174 Fitting random forest model\n",
      "2021-09-18 16:48:24,074 Fold 0: training r2_score = 0.965, validation r2_score = 0.456, test r2_score = 0.538\n",
      "2021-09-18 16:48:24,466 Wrote model tarball to ampl_output/H1_std_model_3421f11b-0fea-45c5-b494-f0783c4488d8.tar.gz\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      " 69%|██████▉   | 31/45 [04:27<01:55,  8.26s/it, best loss: 0.5152785357644352]"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "2021-09-18 16:48:24,494 Attempting to load featurized dataset\n",
      "2021-09-18 16:48:24,503 Exception when trying to load featurized data:\n",
      "DynamicFeaturization doesn't support get_featurized_dset_name()\n",
      "2021-09-18 16:48:24,504 Featurized dataset not previously saved for dataset H1_std, creating new\n",
      "2021-09-18 16:48:24,517 Featurizing sample 0\n",
      "2021-09-18 16:48:29,694 Featurizing sample 1000\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "number of features: 1024\n",
      " 69%|██████▉   | 31/45 [04:34<01:55,  8.26s/it, best loss: 0.5152785357644352]"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "2021-09-18 16:48:31,675 Previous dataset split restored\n",
      "2021-09-18 16:48:31,692 Wrote transformers to ampl_output/H1_std/RF_ecfp_scaffold_regression/870fbfda-7f4e-4639-817c-fee59e1f4a4b/transformers.pkl\n",
      "2021-09-18 16:48:31,694 Transforming response data\n",
      "2021-09-18 16:48:31,722 Transforming response data\n",
      "2021-09-18 16:48:31,738 Transforming response data\n",
      "2021-09-18 16:48:31,770 Fitting random forest model\n",
      "2021-09-18 16:48:32,549 Fold 0: training r2_score = 0.965, validation r2_score = 0.443, test r2_score = 0.500\n",
      "2021-09-18 16:48:32,767 Wrote model tarball to ampl_output/H1_std_model_870fbfda-7f4e-4639-817c-fee59e1f4a4b.tar.gz\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      " 71%|███████   | 32/45 [04:35<01:47,  8.27s/it, best loss: 0.5152785357644352]"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "2021-09-18 16:48:32,796 Attempting to load featurized dataset\n",
      "2021-09-18 16:48:32,806 Exception when trying to load featurized data:\n",
      "DynamicFeaturization doesn't support get_featurized_dset_name()\n",
      "2021-09-18 16:48:32,808 Featurized dataset not previously saved for dataset H1_std, creating new\n",
      "2021-09-18 16:48:32,820 Featurizing sample 0\n",
      "2021-09-18 16:48:37,982 Featurizing sample 1000\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "number of features: 1024\n",
      " 71%|███████   | 32/45 [04:42<01:47,  8.27s/it, best loss: 0.5152785357644352]"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "2021-09-18 16:48:39,985 Previous dataset split restored\n",
      "2021-09-18 16:48:40,002 Wrote transformers to ampl_output/H1_std/RF_ecfp_scaffold_regression/c87ee4e8-1589-4f59-bfab-74ccb3d776b8/transformers.pkl\n",
      "2021-09-18 16:48:40,004 Transforming response data\n",
      "2021-09-18 16:48:40,030 Transforming response data\n",
      "2021-09-18 16:48:40,046 Transforming response data\n",
      "2021-09-18 16:48:40,083 Fitting random forest model\n",
      "2021-09-18 16:48:40,535 Fold 0: training r2_score = 0.952, validation r2_score = 0.418, test r2_score = 0.394\n",
      "2021-09-18 16:48:40,574 Wrote model tarball to ampl_output/H1_std_model_c87ee4e8-1589-4f59-bfab-74ccb3d776b8.tar.gz\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      " 73%|███████▎  | 33/45 [04:43<01:37,  8.13s/it, best loss: 0.5152785357644352]"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "2021-09-18 16:48:40,601 Attempting to load featurized dataset\n",
      "2021-09-18 16:48:40,609 Exception when trying to load featurized data:\n",
      "DynamicFeaturization doesn't support get_featurized_dset_name()\n",
      "2021-09-18 16:48:40,611 Featurized dataset not previously saved for dataset H1_std, creating new\n",
      "2021-09-18 16:48:40,623 Featurizing sample 0\n",
      "2021-09-18 16:48:45,751 Featurizing sample 1000\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "number of features: 1024\n",
      " 73%|███████▎  | 33/45 [04:50<01:37,  8.13s/it, best loss: 0.5152785357644352]"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "2021-09-18 16:48:47,724 Previous dataset split restored\n",
      "2021-09-18 16:48:47,744 Wrote transformers to ampl_output/H1_std/RF_ecfp_scaffold_regression/9522f4f8-ffcc-4752-8d45-ef3f6f553aaa/transformers.pkl\n",
      "2021-09-18 16:48:47,745 Transforming response data\n",
      "2021-09-18 16:48:47,772 Transforming response data\n",
      "2021-09-18 16:48:47,786 Transforming response data\n",
      "2021-09-18 16:48:47,820 Fitting random forest model\n",
      "2021-09-18 16:48:48,814 Fold 0: training r2_score = 0.965, validation r2_score = 0.440, test r2_score = 0.534\n",
      "2021-09-18 16:48:49,114 Wrote model tarball to ampl_output/H1_std_model_9522f4f8-ffcc-4752-8d45-ef3f6f553aaa.tar.gz\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      " 76%|███████▌  | 34/45 [04:51<01:30,  8.26s/it, best loss: 0.5152785357644352]"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "2021-09-18 16:48:49,148 Attempting to load featurized dataset\n",
      "2021-09-18 16:48:49,158 Exception when trying to load featurized data:\n",
      "DynamicFeaturization doesn't support get_featurized_dset_name()\n",
      "2021-09-18 16:48:49,159 Featurized dataset not previously saved for dataset H1_std, creating new\n",
      "2021-09-18 16:48:49,171 Featurizing sample 0\n",
      "2021-09-18 16:48:54,295 Featurizing sample 1000\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "number of features: 1024\n",
      " 76%|███████▌  | 34/45 [04:58<01:30,  8.26s/it, best loss: 0.5152785357644352]"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "2021-09-18 16:48:56,216 Previous dataset split restored\n",
      "2021-09-18 16:48:56,233 Wrote transformers to ampl_output/H1_std/RF_ecfp_scaffold_regression/6deb1033-aae8-4889-b13b-6cc6d0ccb86b/transformers.pkl\n",
      "2021-09-18 16:48:56,235 Transforming response data\n",
      "2021-09-18 16:48:56,263 Transforming response data\n",
      "2021-09-18 16:48:56,278 Transforming response data\n",
      "2021-09-18 16:48:56,309 Fitting random forest model\n",
      "2021-09-18 16:48:57,203 Fold 0: training r2_score = 0.965, validation r2_score = 0.428, test r2_score = 0.521\n",
      "2021-09-18 16:48:57,552 Wrote model tarball to ampl_output/H1_std_model_6deb1033-aae8-4889-b13b-6cc6d0ccb86b.tar.gz\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      " 78%|███████▊  | 35/45 [05:00<01:23,  8.31s/it, best loss: 0.5152785357644352]"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "2021-09-18 16:48:57,579 Attempting to load featurized dataset\n",
      "2021-09-18 16:48:57,588 Exception when trying to load featurized data:\n",
      "DynamicFeaturization doesn't support get_featurized_dset_name()\n",
      "2021-09-18 16:48:57,589 Featurized dataset not previously saved for dataset H1_std, creating new\n",
      "2021-09-18 16:48:57,600 Featurizing sample 0\n",
      "2021-09-18 16:49:02,541 Featurizing sample 1000\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "number of features: 1024\n",
      " 78%|███████▊  | 35/45 [05:07<01:23,  8.31s/it, best loss: 0.5152785357644352]"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "2021-09-18 16:49:04,515 Previous dataset split restored\n",
      "2021-09-18 16:49:04,534 Wrote transformers to ampl_output/H1_std/RF_ecfp_scaffold_regression/31f4b55b-b808-4038-bb2b-c2291537b5fd/transformers.pkl\n",
      "2021-09-18 16:49:04,537 Transforming response data\n",
      "2021-09-18 16:49:04,567 Transforming response data\n",
      "2021-09-18 16:49:04,584 Transforming response data\n",
      "2021-09-18 16:49:04,617 Fitting random forest model\n",
      "2021-09-18 16:49:06,565 Fold 0: training r2_score = 0.966, validation r2_score = 0.461, test r2_score = 0.531\n",
      "2021-09-18 16:49:07,254 Wrote model tarball to ampl_output/H1_std_model_31f4b55b-b808-4038-bb2b-c2291537b5fd.tar.gz\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      " 80%|████████  | 36/45 [05:10<01:18,  8.73s/it, best loss: 0.5152785357644352]"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "2021-09-18 16:49:07,288 Attempting to load featurized dataset\n",
      "2021-09-18 16:49:07,299 Exception when trying to load featurized data:\n",
      "DynamicFeaturization doesn't support get_featurized_dset_name()\n",
      "2021-09-18 16:49:07,301 Featurized dataset not previously saved for dataset H1_std, creating new\n",
      "2021-09-18 16:49:07,314 Featurizing sample 0\n",
      "2021-09-18 16:49:12,495 Featurizing sample 1000\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "number of features: 1024\n",
      " 80%|████████  | 36/45 [05:17<01:18,  8.73s/it, best loss: 0.5152785357644352]"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "2021-09-18 16:49:14,445 Previous dataset split restored\n",
      "2021-09-18 16:49:14,463 Wrote transformers to ampl_output/H1_std/RF_ecfp_scaffold_regression/1529392d-7af0-40b6-a80b-5afc1c17d30c/transformers.pkl\n",
      "2021-09-18 16:49:14,464 Transforming response data\n",
      "2021-09-18 16:49:14,495 Transforming response data\n",
      "2021-09-18 16:49:14,511 Transforming response data\n",
      "2021-09-18 16:49:14,545 Fitting random forest model\n",
      "2021-09-18 16:49:16,052 Fold 0: training r2_score = 0.966, validation r2_score = 0.466, test r2_score = 0.542\n",
      "2021-09-18 16:49:16,463 Wrote model tarball to ampl_output/H1_std_model_1529392d-7af0-40b6-a80b-5afc1c17d30c.tar.gz\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      " 82%|████████▏ | 37/45 [05:19<01:10,  8.87s/it, best loss: 0.5152785357644352]"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "2021-09-18 16:49:16,490 Attempting to load featurized dataset\n",
      "2021-09-18 16:49:16,499 Exception when trying to load featurized data:\n",
      "DynamicFeaturization doesn't support get_featurized_dset_name()\n",
      "2021-09-18 16:49:16,501 Featurized dataset not previously saved for dataset H1_std, creating new\n",
      "2021-09-18 16:49:16,512 Featurizing sample 0\n",
      "2021-09-18 16:49:21,498 Featurizing sample 1000\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "number of features: 1024\n",
      " 82%|████████▏ | 37/45 [05:26<01:10,  8.87s/it, best loss: 0.5152785357644352]"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "2021-09-18 16:49:23,441 Previous dataset split restored\n",
      "2021-09-18 16:49:23,458 Wrote transformers to ampl_output/H1_std/RF_ecfp_scaffold_regression/83f66f63-f80e-4f86-8ead-0de4c01c9e9b/transformers.pkl\n",
      "2021-09-18 16:49:23,460 Transforming response data\n",
      "2021-09-18 16:49:23,487 Transforming response data\n",
      "2021-09-18 16:49:23,501 Transforming response data\n",
      "2021-09-18 16:49:23,531 Fitting random forest model\n",
      "2021-09-18 16:49:24,860 Fold 0: training r2_score = 0.958, validation r2_score = 0.444, test r2_score = 0.528\n",
      "2021-09-18 16:49:25,388 Wrote model tarball to ampl_output/H1_std_model_83f66f63-f80e-4f86-8ead-0de4c01c9e9b.tar.gz\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      " 84%|████████▍ | 38/45 [05:28<01:02,  8.89s/it, best loss: 0.5152785357644352]"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "2021-09-18 16:49:25,421 Attempting to load featurized dataset\n",
      "2021-09-18 16:49:25,430 Exception when trying to load featurized data:\n",
      "DynamicFeaturization doesn't support get_featurized_dset_name()\n",
      "2021-09-18 16:49:25,431 Featurized dataset not previously saved for dataset H1_std, creating new\n",
      "2021-09-18 16:49:25,443 Featurizing sample 0\n",
      "2021-09-18 16:49:30,664 Featurizing sample 1000\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "number of features: 1024\n",
      " 84%|████████▍ | 38/45 [05:35<01:02,  8.89s/it, best loss: 0.5152785357644352]"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "2021-09-18 16:49:32,653 Previous dataset split restored\n",
      "2021-09-18 16:49:32,670 Wrote transformers to ampl_output/H1_std/RF_ecfp_scaffold_regression/64d52313-6bd6-4046-a712-1bc002ad51cd/transformers.pkl\n",
      "2021-09-18 16:49:32,672 Transforming response data\n",
      "2021-09-18 16:49:32,701 Transforming response data\n",
      "2021-09-18 16:49:32,714 Transforming response data\n",
      "2021-09-18 16:49:32,746 Fitting random forest model\n",
      "2021-09-18 16:49:34,361 Fold 0: training r2_score = 0.966, validation r2_score = 0.454, test r2_score = 0.524\n",
      "2021-09-18 16:49:34,843 Wrote model tarball to ampl_output/H1_std_model_64d52313-6bd6-4046-a712-1bc002ad51cd.tar.gz\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      " 87%|████████▋ | 39/45 [05:37<00:54,  9.06s/it, best loss: 0.5152785357644352]"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "2021-09-18 16:49:34,874 Attempting to load featurized dataset\n",
      "2021-09-18 16:49:34,884 Exception when trying to load featurized data:\n",
      "DynamicFeaturization doesn't support get_featurized_dset_name()\n",
      "2021-09-18 16:49:34,885 Featurized dataset not previously saved for dataset H1_std, creating new\n",
      "2021-09-18 16:49:34,900 Featurizing sample 0\n",
      "2021-09-18 16:49:39,957 Featurizing sample 1000\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "number of features: 1024\n",
      " 87%|████████▋ | 39/45 [05:44<00:54,  9.06s/it, best loss: 0.5152785357644352]"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "2021-09-18 16:49:41,893 Previous dataset split restored\n",
      "2021-09-18 16:49:41,913 Wrote transformers to ampl_output/H1_std/RF_ecfp_scaffold_regression/9da7b45f-6859-4889-9f84-6694c2e232f3/transformers.pkl\n",
      "2021-09-18 16:49:41,914 Transforming response data\n",
      "2021-09-18 16:49:41,941 Transforming response data\n",
      "2021-09-18 16:49:41,956 Transforming response data\n",
      "2021-09-18 16:49:41,990 Fitting random forest model\n",
      "2021-09-18 16:49:42,661 Fold 0: training r2_score = 0.964, validation r2_score = 0.433, test r2_score = 0.496\n",
      "2021-09-18 16:49:42,824 Wrote model tarball to ampl_output/H1_std_model_9da7b45f-6859-4889-9f84-6694c2e232f3.tar.gz\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      " 89%|████████▉ | 40/45 [05:45<00:43,  8.74s/it, best loss: 0.5152785357644352]"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "2021-09-18 16:49:42,852 Attempting to load featurized dataset\n",
      "2021-09-18 16:49:42,862 Exception when trying to load featurized data:\n",
      "DynamicFeaturization doesn't support get_featurized_dset_name()\n",
      "2021-09-18 16:49:42,864 Featurized dataset not previously saved for dataset H1_std, creating new\n",
      "2021-09-18 16:49:42,875 Featurizing sample 0\n",
      "2021-09-18 16:49:48,112 Featurizing sample 1000\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "number of features: 1024\n",
      " 89%|████████▉ | 40/45 [05:52<00:43,  8.74s/it, best loss: 0.5152785357644352]"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "2021-09-18 16:49:50,141 Previous dataset split restored\n",
      "2021-09-18 16:49:50,161 Wrote transformers to ampl_output/H1_std/RF_ecfp_scaffold_regression/6ede6102-7fd1-4277-8a19-fa25ce2c408c/transformers.pkl\n",
      "2021-09-18 16:49:50,162 Transforming response data\n",
      "2021-09-18 16:49:50,190 Transforming response data\n",
      "2021-09-18 16:49:50,207 Transforming response data\n",
      "2021-09-18 16:49:50,240 Fitting random forest model\n",
      "2021-09-18 16:49:51,666 Fold 0: training r2_score = 0.966, validation r2_score = 0.464, test r2_score = 0.543\n",
      "2021-09-18 16:49:52,147 Wrote model tarball to ampl_output/H1_std_model_6ede6102-7fd1-4277-8a19-fa25ce2c408c.tar.gz\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      " 91%|█████████ | 41/45 [05:55<00:35,  8.91s/it, best loss: 0.5152785357644352]"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "2021-09-18 16:49:52,179 Attempting to load featurized dataset\n",
      "2021-09-18 16:49:52,189 Exception when trying to load featurized data:\n",
      "DynamicFeaturization doesn't support get_featurized_dset_name()\n",
      "2021-09-18 16:49:52,190 Featurized dataset not previously saved for dataset H1_std, creating new\n",
      "2021-09-18 16:49:52,202 Featurizing sample 0\n",
      "2021-09-18 16:49:57,230 Featurizing sample 1000\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "number of features: 1024\n",
      " 91%|█████████ | 41/45 [06:01<00:35,  8.91s/it, best loss: 0.5152785357644352]"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "2021-09-18 16:49:59,144 Previous dataset split restored\n",
      "2021-09-18 16:49:59,162 Wrote transformers to ampl_output/H1_std/RF_ecfp_scaffold_regression/b423e223-a9a4-4d2a-a1bd-3db0ca170a89/transformers.pkl\n",
      "2021-09-18 16:49:59,165 Transforming response data\n",
      "2021-09-18 16:49:59,194 Transforming response data\n",
      "2021-09-18 16:49:59,209 Transforming response data\n",
      "2021-09-18 16:49:59,241 Fitting random forest model\n",
      "2021-09-18 16:49:59,801 Fold 0: training r2_score = 0.960, validation r2_score = 0.439, test r2_score = 0.511\n",
      "2021-09-18 16:49:59,891 Wrote model tarball to ampl_output/H1_std_model_b423e223-a9a4-4d2a-a1bd-3db0ca170a89.tar.gz\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      " 93%|█████████▎| 42/45 [06:02<00:25,  8.56s/it, best loss: 0.5152785357644352]"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "2021-09-18 16:49:59,921 Attempting to load featurized dataset\n",
      "2021-09-18 16:49:59,931 Exception when trying to load featurized data:\n",
      "DynamicFeaturization doesn't support get_featurized_dset_name()\n",
      "2021-09-18 16:49:59,932 Featurized dataset not previously saved for dataset H1_std, creating new\n",
      "2021-09-18 16:49:59,945 Featurizing sample 0\n",
      "2021-09-18 16:50:05,021 Featurizing sample 1000\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "number of features: 1024\n",
      " 93%|█████████▎| 42/45 [06:09<00:25,  8.56s/it, best loss: 0.5152785357644352]"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "2021-09-18 16:50:07,020 Previous dataset split restored\n",
      "2021-09-18 16:50:07,041 Wrote transformers to ampl_output/H1_std/RF_ecfp_scaffold_regression/9ee16cb3-9752-42ff-b991-2a3707a081ce/transformers.pkl\n",
      "2021-09-18 16:50:07,042 Transforming response data\n",
      "2021-09-18 16:50:07,073 Transforming response data\n",
      "2021-09-18 16:50:07,089 Transforming response data\n",
      "2021-09-18 16:50:07,122 Fitting random forest model\n",
      "2021-09-18 16:50:08,429 Fold 0: training r2_score = 0.967, validation r2_score = 0.462, test r2_score = 0.537\n",
      "2021-09-18 16:50:08,847 Wrote model tarball to ampl_output/H1_std_model_9ee16cb3-9752-42ff-b991-2a3707a081ce.tar.gz\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      " 96%|█████████▌| 43/45 [06:11<00:17,  8.68s/it, best loss: 0.5152785357644352]"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "2021-09-18 16:50:08,875 Attempting to load featurized dataset\n",
      "2021-09-18 16:50:08,886 Exception when trying to load featurized data:\n",
      "DynamicFeaturization doesn't support get_featurized_dset_name()\n",
      "2021-09-18 16:50:08,891 Featurized dataset not previously saved for dataset H1_std, creating new\n",
      "2021-09-18 16:50:08,907 Featurizing sample 0\n",
      "2021-09-18 16:50:13,999 Featurizing sample 1000\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "number of features: 1024\n",
      " 96%|█████████▌| 43/45 [06:18<00:17,  8.68s/it, best loss: 0.5152785357644352]"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "2021-09-18 16:50:15,927 Previous dataset split restored\n",
      "2021-09-18 16:50:15,945 Wrote transformers to ampl_output/H1_std/RF_ecfp_scaffold_regression/44f01510-b305-4cb0-aafa-da2156c91967/transformers.pkl\n",
      "2021-09-18 16:50:15,946 Transforming response data\n",
      "2021-09-18 16:50:15,973 Transforming response data\n",
      "2021-09-18 16:50:15,988 Transforming response data\n",
      "2021-09-18 16:50:16,019 Fitting random forest model\n",
      "2021-09-18 16:50:16,901 Fold 0: training r2_score = 0.928, validation r2_score = 0.446, test r2_score = 0.512\n",
      "2021-09-18 16:50:17,046 Wrote model tarball to ampl_output/H1_std_model_44f01510-b305-4cb0-aafa-da2156c91967.tar.gz\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      " 98%|█████████▊| 44/45 [06:19<00:08,  8.54s/it, best loss: 0.5152785357644352]"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "2021-09-18 16:50:17,076 Attempting to load featurized dataset\n",
      "2021-09-18 16:50:17,086 Exception when trying to load featurized data:\n",
      "DynamicFeaturization doesn't support get_featurized_dset_name()\n",
      "2021-09-18 16:50:17,088 Featurized dataset not previously saved for dataset H1_std, creating new\n",
      "2021-09-18 16:50:17,099 Featurizing sample 0\n",
      "2021-09-18 16:50:22,135 Featurizing sample 1000\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "number of features: 1024\n",
      " 98%|█████████▊| 44/45 [06:26<00:08,  8.54s/it, best loss: 0.5152785357644352]"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "2021-09-18 16:50:24,136 Previous dataset split restored\n",
      "2021-09-18 16:50:24,153 Wrote transformers to ampl_output/H1_std/RF_ecfp_scaffold_regression/c28a1381-84b5-4493-95c7-f1f967b375c3/transformers.pkl\n",
      "2021-09-18 16:50:24,155 Transforming response data\n",
      "2021-09-18 16:50:24,182 Transforming response data\n",
      "2021-09-18 16:50:24,196 Transforming response data\n",
      "2021-09-18 16:50:24,229 Fitting random forest model\n",
      "2021-09-18 16:50:24,899 Fold 0: training r2_score = 0.965, validation r2_score = 0.488, test r2_score = 0.534\n",
      "2021-09-18 16:50:25,073 Wrote model tarball to ampl_output/H1_std_model_c28a1381-84b5-4493-95c7-f1f967b375c3.tar.gz\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "100%|██████████| 45/45 [06:27<00:00,  8.62s/it, best loss: 0.5120841961932765]\n"
     ]
    }
   ],
   "source": [
    "# continue HPO\n",
    "best = fmin(lossfn, space, algo=tpe.suggest, max_evals=50, trials=new_trials)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 31,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 636
    },
    "id": "2l-of-gh7Jaq",
    "outputId": "aecf5b20-3aca-4a97-ac1f-383273ead71c"
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(0.2, 0.6)"
      ]
     },
     "execution_count": 31,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAn0AAAJZCAYAAADcToPtAAAABHNCSVQICAgIfAhkiAAAAAlwSFlzAAALEgAACxIB0t1+/AAAADh0RVh0U29mdHdhcmUAbWF0cGxvdGxpYiB2ZXJzaW9uMy4yLjIsIGh0dHA6Ly9tYXRwbG90bGliLm9yZy+WH4yJAAAgAElEQVR4nOzdeXRb93Uv+u/BxHkmOEsiqYHUZA20ZUmJp0SO9dywdezUsRp5yHKt3DpRmpWobm9ym0i+z6uvtWK3Hm6v6163r5Ilx22f3SiJrSiJ4im0KMmSRZEyRUoU55kgCYAAMZzz/iBweA4IkgcgQILE97OWl0UCIA6JaZ/923v/BEmSJBARERHRkqZb6AMgIiIiouhj0EdEREQUBxj0EREREcUBBn1EREREcYBBHxEREVEcYNBHREREFAcMod7gxIkTOH78OBobGyGKIsrKyvDAAw9gz5490OlCjyG9Xi/efPNN/PznP0dzczPGxsaQnZ2NtWvX4sEHH8QXvvCFeTkOIiIioqVMCGVO36FDh3Ds2DEkJCRgx44dMBgMqKmpgd1ux913340XXnghpIDLYrHgiSeeQF1dHTIzM7F582YkJSWhp6cHDQ0NqK6uxjPPPBP14yAiIiJa6jRn+k6ePIljx47BbDbj6NGjKC0tBQAMDAzgkUcewalTp3DkyBE8+uijmn6eKIr4sz/7M9TV1eGRRx7BgQMHkJCQIF9us9nQ2dkZ9eMgIiIiigea02GvvPIKAODAgQNyoAUAubm5OHjwIADg1VdfhSiKmn7em2++iQsXLuCuu+7CD3/4Q1XABwCpqamoqKiI+nEQERERxQNNQV9PTw/q6+thNBqxe/fuKZdv27YN+fn56O/vx8WLFzXd8euvvw4AeOyxxzQfbDSOg4iIiCgeaFrebWhoAACsXr0aiYmJQa+zceNG9Pb24sqVK9i6deuMP6+vrw9Xr16FXq/Hli1b0NLSgl/+8pfo7e1FRkYGbrnlFtx2220QBCGqx0FEREQULzQFfR0dHQCAoqKiaa9TWFiouu5Mrl69CgDIzMzE8ePH8eyzz8Lj8ciX/9M//RO2bNmCl19+GTk5OVE7DiIiIqJ4oSnoGxsbAwAkJSVNe52UlBQAgN1un/XnjYyMyP//m7/5G3z5y1/Gk08+ifz8fFy+fBlPP/00Lly4gD//8z/H0aNHo3Ycs2loaMD4+Dj0ev2UmkMiIiKiWDI+Pg6v14uEhASsW7duyuUhz+mLBH+ThcfjQVVVFX7yk5/Il23fvh2vvfYa7rnnHpw9exYff/wxtm/fvhCHifHxcYiiCFEU4Xa7F+QYiIiIiEIxPj4e9Puagr7k5GQAgMPhmPY6/syaP9M2E+V1HnzwwSmXFxQU4I477sDJkydx5swZOeiL9HHMRq/XQxRF6HQ6+b61stlsACa6kCm28LGJbXx8Yhsfn9jFxya2zcfjMzY2BlEUodfrg16uKegrLi4GAHR1dU17nZ6eHtV1Z1JSUhL038GuMzAwELXjmE1CQgLcbjeSk5ODjo+Zyfnz5wEg5NtR9PGxiW18fGIbH5/Yxccmts3H49PY2AibzTZtSZqmkS3+deGmpiY4nc6g16mrqwMArF27dtafV1ZWJmfOhoeHg17HYrEAgCrDFunjICIiIooXmoK+wsJCrF+/Hm63G+++++6Uy2tra9HT0wOz2YwtW7bM+vOMRiPuvPNOAEBNTc2Uy91uN86dOwcA2LBhQ9SOg4iIiCheaN6RY9++fQCAw4cPo7W1Vf7+4OAgDh06BAB44oknVHveHj16FLt378ZTTz015ed985vfhE6nw09/+lN88MEH8ve9Xi8OHz6MtrY25Ofn4+67757zcRARERHFO83du7t378aePXtw/PhxVFdXY+fOnTAYDKipqYHNZsOuXbuwd+9e1W0sFgtaWlpgNpun/LzKykr84Ac/wDPPPIMnnngCN910EwoKCtDQ0ID29nakpaXhH/7hH6YMYQ7nOIiIiIjiXUgjWw4ePIiqqiq8/vrrqK2thSiKKC8vxwMPPIA9e/aEnF17+OGHsWbNGrz22mu4ePEiGhoaYDab8bWvfQ379u2btskj0sdBREREtNSFPKevuroa1dXVmq67f/9+7N+/f8br3Hrrrbj11ltDPYyQjoOIiIgo3jElRkRERBQHGPQRERERxQEGfURERERxgEEfERERURxg0EdEREQUBxj0EREREcUBBn1EREREcYBBHxEREVEcYNBHREREFAcY9BERERHFAQZ9RERERHGAQR8RERFRHGDQR0RERBQHGPQRERERxQEGfURERERxgEEfERERURxg0EdEREQUBxj0EREREcUBBn1EREREcYBBHxEREVEcYNBHREREFAcY9BERERHFAQZ9RERERHGAQR8RERFRHGDQR0RERBQHGPQRERERxQEGfURERERxgEEfERERURxg0EdEREQUBxj0EREREcUBBn1EREREcYBBHxEREVEcYNBHREREFAcY9BERERHFAQZ9RERERHGAQR8RERFRHGDQR0RERBQHGPQRERERxQEGfURERERxgEEfERERURxg0EdEREQUBxj0EREREcUBBn1EREREcYBBHxEREVEcYNBHREREFAcY9BERERHFAQZ9RERERHGAQR8RERFRHGDQR0RERBQHGPQRERERxQEGfURERERxwBDqDU6cOIHjx4+jsbERoiiirKwMDzzwAPbs2QOdTnsM+eKLL+Kll16a9nKTyYS6urop3/+rv/orvPXWW9PerqysDO+++67m4yAiIiKKByEFfYcOHcKxY8eQkJCAHTt2wGAwoKamBk8//TRqamrwwgsvhBT4AUBlZSXWrl079cAMMx/a1q1bsWLFiinfN5vNId0/ERERUTzQHPSdPHkSx44dg9lsxtGjR1FaWgoAGBgYwCOPPIJTp07hyJEjePTRR0M6gF27dmH//v0h3QYA/viP/xj3339/yLcjIiIiikea03KvvPIKAODAgQNywAcAubm5OHjwIADg1VdfhSiKET1AIiIiIpo7TUFfT08P6uvrYTQasXv37imXb9u2Dfn5+ejv78fFixcjfpBERERENDealncbGhoAAKtXr0ZiYmLQ62zcuBG9vb24cuUKtm7dqvkA6uvr8eyzz2J0dBQZGRnYtGkT7rjjDphMphlvd+bMGTQ2NmJsbAw5OTmoqqrC5z73uZBrComIiIjigaagr6OjAwBQVFQ07XUKCwtV19Xq9OnTOH36tOp7BQUFePbZZ7Ft27Zpb/f2229P+d6qVavw3HPPoaKiIqRjICIiIlrqNAV9Y2NjAICkpKRpr5OSkgIAsNvtmu542bJl+P73v4/bb78dJSUlcLlcuHr1Kl5++WXU1tZi3759eOONN1BZWam6XWVlJf7H//gf2LlzJwoLC2Gz2dDQ0IDnn38en332Gb7xjW/grbfeQn5+vqbj0MJms+H8+fNh3Tbc21H08bGJbXx8Yhsfn9jFxya2LeTjE/Kcvki57777pnxv+/bt2L59O77zne/g5MmTeP755+UGEr/HHntM9XVycjLy8vKwc+dOPPzww7h48SJeeeUV/OhHP4rm4RMREREtKpqCvuTkZACAw+GY9jr+DJ8/4zcXTz75JE6ePImPPvoIbrcbRqNx1tuYTCbs27cPTz75JN577705H4NSampqyEvG/ki+qqoqosdCc8fHJrbx8YltfHxiFx+b2DYfj09jYyNsNtu0l2vqeiguLgYAdHV1TXudnp4e1XXnory8HADgdrthsVhCvl1vb++cj4GIiIhoKdEU9K1btw4A0NTUBKfTGfQ6/i3Tgu2uEarh4WH53/4sYyi3i0S2kYiIiGgp0RT0FRYWYv369XC73UH3ta2trUVPTw/MZjO2bNky54N65513AEzso5uamhry7TZs2DDnYyAiIiJaSjQPtdu3bx8A4PDhw2htbZW/Pzg4iEOHDgEAnnjiCdWcvKNHj2L37t146qmnVD+rq6sLJ06cgMvlUn1fkiS8/fbbeO655wBMbdq4cuUKTp8+Da/Xq/q+x+PBa6+9hiNHjgS9HREREVG809y9u3v3buzZswfHjx9HdXU1du7cCYPBgJqaGthsNuzatQt79+5V3cZisaClpQVms1n1/ZGRERw4cAA//vGPsX79euTl5cFut6OpqUme87d371489NBDqtt1dnbiW9/6FjIzM7Fu3TpkZ2djeHgYV69eRV9fH3Q6Hf7iL/4Ct912W7h/DyIiIqIlKaSRLQcPHkRVVRVef/111NbWQhRFlJeX44EHHsCePXs074ZRUFCAxx9/HHV1dWhra8OlS5cgiiLMZjPuvfdePPjgg9ixY8eU21VUVOCRRx5BXV0dmpubMTw8DEEQUFBQgPvvvx9f//rXubRLREREFETIc/qqq6tRXV2t6br79+/H/v37p3w/KytrypKvFsuWLcMPf/jDkG9HREREFO+4US0RERFRHGDQR0RERBQHGPQRERERxQEGfURERERxgEEfERERURxg0EdEREQUBxj0EREREcUBBn1EREREcYBBHxEREVEcYNBHREREFAcY9BERERHFAQZ9RERERHGAQR8RERFRHGDQR0RERBQHGPQRERERxQEGfURERERxgEEfERERURxg0EdEREQUBxj0EREREcUBBn1EREREcYBBHxEREVEcYNBHREREFAcY9BERERHFAQZ9RERERHGAQR8RERFRHGDQR0RERBQHGPQRERERxQEGfURERERxgEEfERERURxg0EdEREQUBxj0EREREcUBBn1EREREcYBBHxEREVEcYNBHREREFAcY9BERERHFAQZ9RERERHGAQR8RERFRHGDQR0RERBQHGPQRERERxQEGfURERERxgEEfERERURxg0EdEREQUBxj0EREREcUBBn1EREREcYBBHxEREVEcYNBHREREFAcY9BERERHFAQZ9RERERHGAQR8RERFRHGDQR0RERBQHGPQRERERxQFDqDc4ceIEjh8/jsbGRoiiiLKyMjzwwAPYs2cPdDrtMeSLL76Il156adrLTSYT6urqon4cRERERPEgpKDv0KFDOHbsGBISErBjxw4YDAbU1NTg6aefRk1NDV544YWQA67KykqsXbt26oEZpj+0aBwHERER0VKmOeg7efIkjh07BrPZjKNHj6K0tBQAMDAwgEceeQSnTp3CkSNH8Oijj4Z0ALt27cL+/fs1Xz9ax0FERES0lGlOh73yyisAgAMHDsiBFgDk5ubi4MGDAIBXX30VoihG9ABj9TiIiIiIFhNNQV9PTw/q6+thNBqxe/fuKZdv27YN+fn56O/vx8WLFyN+kLF2HERERESLjabl3YaGBgDA6tWrkZiYGPQ6GzduRG9vL65cuYKtW7dqPoD6+no8++yzGB0dRUZGBjZt2oQ77rgDJpNpXo+DiIiIaCnTFPR1dHQAAIqKiqa9TmFhoeq6Wp0+fRqnT59Wfa+goADPPvsstm3bNm/HQURERLSUaQr6xsbGAABJSUnTXiclJQUAYLfbNd3xsmXL8P3vfx+33347SkpK4HK5cPXqVbz88suora3Fvn378MYbb6CysjKqx6GFzWbD+fPnw7ptuLej6ONjE9v4+MQ2Pj6xi49NbFvIxyfkOX2Rct9990353vbt27F9+3Z85zvfwcmTJ/H888/LjRtEREREFD5NQV9ycjIAwOFwTHsdf2bNn2mbiyeffBInT57ERx99BLfbDaPRuCDH4ZeamoqKioqQbuOP5KuqqiJ2HBQZfGxiGx+f2MbHJ3bxsYlt8/H4NDY2wmazTXu5pu7d4uJiAEBXV9e01+np6VFddy7Ky8sBAG63GxaLZcGOg4iIiGip0BT0rVu3DgDQ1NQEp9MZ9Dr+LdOC7a4RquHhYfnf/uzeQhwHERER0VKhKegrLCzE+vXr4Xa78e677065vLa2Fj09PTCbzdiyZcucD+qdd94BAJSVlSE1NXXBjoOIiIhoqdC8I8e+ffsAAIcPH0Zra6v8/cHBQRw6dAgA8MQTT6j2vD169Ch2796Np556SvWzurq6cOLECbhcLtX3JUnC22+/jeeeew4A8Nhjj0XkOIiIiIjinebu3d27d2PPnj04fvw4qqursXPnThgMBtTU1MBms2HXrl3Yu3ev6jYWiwUtLS0wm82q74+MjODAgQP48Y9/jPXr1yMvLw92ux1NTU3yfL29e/fioYceishxEBEREcW7kEa2HDx4EFVVVXj99ddRW1sLURRRXl6OBx54AHv27NGcXSsoKMDjjz+Ouro6tLW14dKlSxBFEWazGffeey8efPBB7NixI+rHQURERBQvQp7TV11djerqak3X3b9/P/bv3z/l+1lZWVOWfKN5HERERETxjikxIiIiojjAoI+IiIgoDjDoIyIiIooDDPqIiIiI4gCDPiIiIqI4wKCPiIiIKIqsLieuekZgFz0Lehwhj2whIiIiIu3+5WoN6l09SBUMuB23LthxMNNHREREFCXjXg8aLN0AALvkgVcUF+xYGPQRERERRUnX2DAk378zBBP0C7hrGIM+0sThcUOSpNmvSERERLIO27D872xdwgIeCWv6SIP3u5txrPkslqVm4qlNX4JRp1/oQyIiIloUOuwW+d85Cxz0MdNHMxIlCb9oq4MECW02Cy4Ndi70IRERES0aHfbJTB+DPoppLdYBDLsc8td1lq4FPBoiIqLFQ5IkVdCXLTDooxh2YaBd9XX9UBdE1vYRERHNanDcDqfXDQBIgA4pwsJW1THoo2lJkoRPAoK+UbcTbbahBToiIiKixUOV5dMlQBCEBTwaBn00g3a7BYPj9infrxviEi8REdFsOmzKJo7EBTySCQz6aFrKLF+acbIO4fIQmzmIiIhm02mPnXEtAIM+moGynu8rpZuhw0Ra+oZtCKOK5g4iIiKaKpbGtQAM+mgaXfYR9DhGAQAmnR63mFdgZbpZvvyyb0sZIiIimsrpdaPfaQMA6CAgSzAt8BEx6KNpXBhsk/+9IbsIJr0BG7OL5O/VcYmXiIhoWp32ye3X8pPTYRAWPuRa+COgmKSs59uaswzARPDn12DpWdBNo4mIiGKZsp6vJCVzAY9kEoM+mqLfYZXbzA2CDhuyiwEARckZyE5IBjCRtm4e7V+wYyQiIoplHQz6aDH4ZHAyy7c2qwBJBiMAQBAEbPQFgABHtxAREU1H2cRRkpK1gEcyiUEfTaHs2t2au1x1mbKuj6NbiCLH5fWwK55oiRAlicu7FPss42NosQ4CAHSCgE2KzB4AVGTkw6jTAwC6HaMY8HUmEVH4hsbt+Mvat/GXZ97Gp4MdC304RDRHg047nF4PACDFkIAMU9ICH9EEBn2koszyVWTkI8Wonitk0htQkZEvf80u3ugZcTnkPRtpafuo5zrGPC6IkPBBT/NCHw4RzZF6aTdzwbdf82PQRyoXBpVLu8uCXmeDanQL6/qi4f3uJvz32v/CX555i9nUOHDZMvk6alNs20REi5Mq6EuNjaVdgEEfKYy6nGgamejIFQBsyikJej1lXV/jcC/GfSlsioy6oU4caz4HryTC6fWgprdloQ9p3kiShB7vGLq9Ywt9KPPG6nKi1VdSAUxkeIfH4+f3J1qKlJ27y2KkiQNg0EcKFwc7IPlGSa5MN09bg5CbmIrC5AwAgEcS8dlwz7wd41LXYbfg1c8+kh8HAGge7VvAI5pfV4Z78LPxdpwYb4+b2rb64W7Foz2h1Ta0IMdCRJGhDPqKY6SJA2DQRwoXBiZ34ZhuaddP3cXLJd5IGHE58FL9e1Myp9dGB+ARvQt0VPPriuIE4lx/2wzXXDqCvX4Y9BEtXg6PWy7L0QmCnCSJBQz6CABgd7vw2Uiv/PWW2YK+LEVdn6ULkhSYq6BQjHs9eKn+PVh8y3qJegPSfE00btEbN3VeNve4/O9rcTD8W5RENATZx7rVyqCPaLFSjmopSEqXJ17EAgZ9BAC4NNQB0Re4lablIDshZcbrr0w3I0k/MbTZMj6GrrGRqB/jUiVKEl5r/D3afNkdHQTsW/t5rFcE1k0j8bHEa3U75X8PjtuXfG1bi3UQdo8LAFQfDG22IZ5IES1SsTifz49BHwEIGMicM3OWDwD0Oh3WZRXKX3N0S/jeunERFxX1aw+tvBnrs4qwOiNP/l5TnNT1WV1O1dfXrQMLdCTzQ7m0e7N5BRL1BgDAqNuJYQ5qJlqUYnEnDj8GfQSnx416xRLTbEu7fhs5umXOPuxpxq86rshff7G4AncUrQYArM4wy99vHumHKInzfnzzzapY3gUm6hmXMuWolpuyi7A8NVv+WtnRS0SLR3uMjmsBGPQRJmryPL6AoiQlE3lJaZputz6rCP5xk9dGB2AP+MCOJ9dG+/FhTzN6x0Y13+aKpQevN5+Vv74puxhfLdsif52XmIZ0YyIAwOF1o9O+9JfQlcu7wPzX9bWMDuCKpWdellZHXA65VlMnCFibWaAO+tjMQbToiJKELsV7daxl+gwLfQCkjUf04oZ1CIIw0f6d6KuniwTl0u4WDUu7fummRKxIy8EN6yAkSKi3dGNbXmnEjmuxaLUO4SeXfgOvInDemrsMVbnLUTBN11b32AheufKBXEe5LCULj1fuhE6YPA8TBAGrM/Jw3tdV3TTSh2WpsfUGEknjXg9cAV3KbTYL3KI36oXQkiThv1ov4Z32egDAN9bswPb8sqjepzK7virdjCSDCSsY9FEMkSQJ3WMjyEpIRpLBtNCHsygMOG0YFycmMKQZE+QT91jBoG+R+M+Wi/htVyOAicHJeUlpKEnJwrLULCzz/T/dmBjyVi8ur0dVVzTbqJZAG7OKcMO3DHXZ0hWXQd/prkY54AMm5jN12Ifxs9Y6FCVnyAFgYXIGBEGA1eXES/W/g8O3xVqmKQnfWn9H0EB+dYZ5Mugb7cMXiivm55daAIFZPgDwSiJarUNYpVjqjjRREnG8+RzeV2x/9nFfS9SDPuXrboOvaWdF2mTQ52/miJXtmyi+tFqH8Ob182ge7UeaMRE/2nov0k2xFcDEosB6vlh7/TLoWwQkScKZvsldGSQAvQ4reh1WOSAAgDRjIpalZKIkNQsr083YmFUEvW7mFfyG4R75rCQ/KT3keUIbs4txoq0OAHB5qBuiJKqyVUud3e3COcVjYBB08lI5AHSNjaCrbQQ/b7uMgqR0VOUux5XhHgw47QAAk06Pb62/A1kJyUF/vqqZY6R/SQcBwYI+ALhm7Y9a0OcRvfiXxhrVYwgAzaP9Uc0wegNGtfi3NjQnpiFRb4TT64bVPQ7L+BiyE2fupCeKpBGXA2/f+BQ1vdfloeFWtxMf97XgSyVrF/TYFoMOW+x27gIM+haFYZdDHusg+KropCkz/CdemA3DPWgY7gFwBRmmJHy+YCVuK1g1bVChHMi8Jbck5IDCn2EcdTth94yjxTqIlenRy8rEmtr+Frh9S5LLUrJw4KZdqBvqxCcD7aizdMmXAUCPYxS/aL8sfy0A+NPKz6nquAIVJWci2WDEmMcNq9uJPocV+cnpUft9FpLVFbwmNFrNHC6vB//7ygeqZVYBEydVbtGLFusg1iiC7ki6PjogZ3qzTMko8p1s6QQBK1Kz0eibmXnDNsSgj+aFW/TiN52f4Zft9UG31qztuxGzQd+oy4kRlwMlKZkLflLcMaYM+mKvHIdB3yKgTBevTM/Fn2+4C11jI2i3WdBut6DDbkGHbVjO2PmNuBz4RdtlvNNWj5tyinFn4RpUZOZD53tReEQvPh2cHLWyNWd5yMemEwRsyC7C73uvA5jo4o2XoE+SJHzQfU3++raCVUg0GHFLXiluySuF0+tG/VA3zg+0oW6oc0q92lfLt067v7GfThCwKj0Pl3wjcZpG+5Zu0KfI9Jl1iegXJ76+Phr5DKfd7cLLDb9TBZR3Fq6BR/Liw56Jx/Sz4Z6oBX2qpd3sItXvtlwR9LXZhkIuuYim5pF+nB7vxmpDOqoW+mAoIiRJwoXBDvxnyyfyCoTfxuwiXLH0wCOJaLdb0GUfRlGMZa+Gxu04eP4XGPd68Ccrb5GnHyyUDsUg/Vjafs2PQd8i0B6QLjbpDShNy0FpWo78fVGS0O+0osM2jBbrIGr7b2DEN+dLhISLgx24ONiBvKQ03FG4GjvyynHDNpltyElIwfIwmwQ2KoK+y0NduK90U7i/6qLSYh1Ep++szqTTY1veCtXliXojqszLUWVePlE7aenGJwNtaLMNYXteGb5YpK0+b3WGIugb6cPnC1ZF9heJEcpxLQW6JNh1IsY8Lljd4+h32jR3lc9mxOXAC5dPq/bG/IPlG1C9fCPODbTJQV/jcC+wYrqfMjfKUS0bFKOPAHVdXyyNbfGIXvzvKx/A6nWizWvHlyUR+jgq5ViK2m0WvHn9PK4GDH8vTM7Ag+VbsS6rEP905UO5jOhM/w18JWXzQhzqtM73t8mZyd92NS5o0OfwuDA4PhE46wUdCmPwBJ1B3yKgZdCjThCQn5SO/KR0VJmX4yulm3BxsAPvdTfJWQMA6HNY8e/XP8HbNz5FpilJ/v6W3GVhZ1LWZhZCL+jg9Z0NWsbHpl1OXko+UBT+32JeMWN3m0lvwNbcZWFlbVYrMqdNI0t3azJlpi9J0KM8NVcOjq6N9kck6Btw2vD3db9Fv29fTAB4sHwrvlhcCQCoyMiXv3/dOoBxrwcJ+si+TVrGx+SAUy/oUKm4TwAoVXXwWmKmjrNxpFd+jMbhhc09jgzFewgtHja3E/914xI+6LmmKhVKMZhQveIm3F64Sg7ot+WVykFfbd8N/NGKTfJqUSxQBqw9jlH0jI1MOzVBK6vLiePXziHDlISvlm2ZtTbeT3kiWZicDkMMbb/mx9O0RUD5RNI66FGv06HKvBzfu+mLOFj1B/hC0RpVd6hb9Ko++OayhJRkMGKVIjAJtoF8tDk8bhxpOoOX6n8n718b3ftz4Wx/q/z1bVHMvi1PzYbJ9+YxOG7HUMASzFJhUwV9BqxMz5W/vh6Bur5O+zD+7tNT8vNeBwGPrdkuB3zAxBii4uSJ15goSVHZ/q5ekeVbnWFGokHdtZ2bmIpk3/fsnnE5c7DQPlGMdgIQ19AvRGoAACAASURBVHM5FzOvJOLvPv013u9plgM+HQR8oWgN/ufN1biraI0qg7shqxApvhPaofGxmNoTW5TEKa/RC4rdjcL17y2f4PxAG37b1YhTnZ9pvl1HDG+/5segL8a5vB70OawAJpo4/B9IoShMzsDXVt6Mv7v1K3h49TYsC8gWZpqSUJaWO82ttVHtzmGZ36DP6XHjhcun8WHPNdQNdeHfr38S9fs803dDbtIoSclULbVHml6nU9VJLtUt2UYVQUSioFf9znNt5rg+OoDDl34tlzwYBB3+27rbsCO/fMp1KzInM2/KLHmkXB5SdO1mFU25XBCEgJ05Fn5enyiJ+DTgw9Tmay6jxaXDNoxex+QQ+XVZhfhR1b342sqbkWJMmHJ9g06PqtzJeu8zfTfm4zA16bAPyyVKfhcDTk5C5fJ6cHFg8rn+684rQRtbgh+Psp4v9po4AAZ9Ma9rbEQ+G8tLSoNpDktNCXoDPl+wCj/csht/uelL2JFXhhWp2di7etuc0/XKoO8zS4+qazWanB43Xqg/rdqj9dJQJxwe9wy3mhtJklRLu58vWBX15TfllmxLdYlXue9uEvQoTcuBztet3jU2DEeYQUbzSD/+vu63GPPdPlFvwHc23DVtE02lIuj7bDiyQZ9H9OLK8NRRLYGUQV9bDAxpbhrpn7JFHjN9i5MyU7clZxm+s/7OWUd13aqYv3p+oG3e3t9nE1iLCEx0vM9ltaduqEvVFGl1j+P97iZNt2Wmj+ZMXc8XmSeRIAgoT8/FYxU78IMtu7Exu3jOPzM/KR25iakAgHHRE5VlsUATAd/vpmSB3KJ3SlYikm5YB+UXt1Gnx/Z5GEi9Ol05r29pZvpsiiAiSTAgQW9Aia+5SAJUgX0o3rh2Tn4TTzUk4Hsbd6myeYHWZOTJo5HabUMRDW6ujQ7A6csa5CSkoCApeKF3aepk5jgWdub4JGCOITCx9EyLj/J1tCYjT9MJa3m6GTkJE6ODxjwu1MfIXutXFSdlysTFxTm8/58baJ3yvV91XIFrlmyfKInotMf2uBaAQV/MC+zcjVWCIKiyfcrZZ9Hg9LrxYv3vVGetZYol1rP9N6J236E0cERKaVoODL46mx7HKEZdwQcZL1aSJKkaORKFiRpGZV1fOEu8nfZhefNzo06Pv9i0S9UdG0ySwSRfR0LwbEK46mYY1aKk6uD17cyxUETfSI9ANjeXdxcj5Xum1vFaOkFQ7bZ0Jorvr1qJkoQmxe9yZ+Fk1+7FwfCWeJ0et+o16q+tHXU75a7+6fQ7bHIGNN2YGLO7lzDoi3FaOndjxdrMAvnf0Sz2dXrdePHy79CsuI8/Lt+KP638nPx1g6VHtVwYKfPZwKFk0huwQhHUNi+xuj6n1y3vZGKAAKMvwF2ZNrdmjlpF/dGm7GLNXX3qJd6ekO93OsomjmD1fH45CSlI9p1MjHlcU+anzacW64BcC6nETN/iYxkfw5Bv6dOk04eUSNhmLpX/fWmwM+xyi0jptA/LJRvpxkR8qWSdfNnV4b6wMvSXhjrlwK04ORN/uOIm+bKTHQ0zLmsvhqVdgEFfTJMkSZ0u1ti5u1DKFR/QbTZLVOo+xr0evHT5vSkB367iSuQmpsqZIRHSlG7DSKjta5WHLBcnZ6qyi9G2lOv6rAFLu37KTMR16wBExRZ3sxElSZWRCGVf6MqMyROYxgjV9Q067egaGwEw0Ugy0xKz4NuZw6/VtnDz+pSvowTd5GMz3zV9zSN9+MeG91HjmwlKoVOeOJWm5WgeRQIARSkZchOgRxKj8v4aCmWT1ZqMPGQlJMvvxyIkVcZOq3OKE/qbzcvx+YJV8liiYZcDH82Q7WtXJmjCnHk7Hxj0xbDBcbvcmZRsMCHLFNuz79JMifIsNa8kRrwAfdzrwUv1v1N1r361bAt2KUZu3GKenKZbG+ElCEmS8H7PZEHvbYUr53V+2lKu6wu2tAsAWQnJ8jzJca8HnfYRzT+zebRfLuhOMSRgfVah5tuuTM+Vl9O7HaNBM12hUg5kXpOZP+v8P3XQtzB1fZIk4YLiw/0WxQDy+e7ePX7tHC4OduBfr36MX3Vcmdf7trnH8X53M3rGtD//YlE4S7tKyoaOhe7iVb4HrvHNutysaM66EOIS75jHpSpLutm8AkadXrX13Lvt02f7Opnpo7kKTBfHwoDW2SizfZHcM9Xl9eDl+vdU9VUPlG3B3QF7QVblLpeL8JtH+yM6067VNqRq4Lg1ryxiP1uLlelm+XfrsFsWfHklkgI7d/0EQVBn+0J4Tp3pa5H/fbN5eUiDUk16A8oV9YSRWOJVFr9v0BCAqnfmWJigr81mkecEJumNqtEd853pG1DMFf3Plgv4oLt5hmtH1pGmM3i9uRbPXHhXNe5ksbmmaOJQ1stqdYt5BfyfQldHeudlJmowoiSpPgvWZE6cEG/OmZw3W2/pnrX5QuniYIdcYrI8NUtOYNxesAppxon6PItrDDW9LUFvv1hKsUIO+k6cOIE/+ZM/QVVVFbZs2YL7778fr7/+OkRR+7LLdH7605+ioqICFRUVePrpp4Ne58UXX5SvE+y/jRs3zvk4YoVyD79YfhIpRXqgLjAR8L1U/54qnX9/2eagm3+nm5KwVrFsdjZIJ1a4lA0cN+cul2uu5kuSwYhlim7W5hgakjpXgTP6lJTB1zWrtt/ZLXpVHafKeiStlHV9c13idYteXFEEjtONalFaoejgbVugZo5PBif/hptyilU7cMxn0CdJEsa96gzL6821qvraaBn3euRtEF2iF29eO7+gjTXhcnk9qtWXcGazZiYko9JXuy0h8qspWinr+dKMiXIXfEFyOgp9/3aLXjSEcLKmXtqdzGib9IaAbF89vAHxjt3tkmslDYJu2q78WBDS0LdDhw7h2LFjSEhIwI4dO2AwGFBTU4Onn34aNTU1eOGFF6ALoUZAqbOzE3/7t38LQRA0vaAqKyuxdu3UD32DYensLLdYCkOVVB/Qo/1z3kLK5fXg5YaAgK90M+5RFO0GuiWvVH6xn+1rnfG6Wjk8bpztUzRwFC7M/rerM8zyG3fzSH9Exu3EgsDdOJRWhTGk+fJQF8Y8k/tKh5PVqMwswM9a6wDMfV5f00ifXAual5iKfA0fCtkJyUgxJMDuGYfD647o/sNaSJK6LnZL7nJ5ZwZgfpd3PZKo2i4MmAg6Xmv8PRL1hqi+DlqsAxAVn0mXLd24ONiBLXPYxWghtNqG5N+jICkdqUEGMWuxLa9UPoGp7bsR0vurJEn4Rdtl1PRdx5eK14W9T656aVc9dmZTbgm62xsATAxq3jzNPE4lm9upOim7OVe96fbthatwsr0BNt8OOR/3teBzBSvlyzsVWb7C5IyQaiXnm+YI6eTJkzh27BjMZjOOHj2K0tJSAMDAwAAeeeQRnDp1CkeOHMGjjz4a8kFIkoQf/vCHkCQJ9913H956661Zb7Nr1y7s378/5PtaTJRPpGUxXBiqVJScgUS9AU6vB6NuJwbH7fL8vnD8rPWS6gP3K6WbcM+ymd9ktuSU4HVBB49vL+BI7MV4tv+GPOutKDlDtYw9n1an5+E3nY0AltbOHIH77iotS8mCUaeHW/RiwGnDiMsx656vyq7dbXmlYZ14lKbmIEFnwLjoweC4Hf0OG8xJ4T2XlfV8WrJ8gK+ZIy0bDb46o1br4LwGfV1jI/JuQAk6A9ZlFqj+jnbP+LztC6xcpkvQGZCdkIxuxyhEScIrVz7Ed9bfiTUzNMbMRbCmqTevn8e6rMKI78scTcqVl3BOgvy25CzDMd1ZuEUvOuzD6LQPo1hjUuI3XY040TZxIvXT6+dRZV6GVGPoo00CmzgCj+9dX9B3aagTXklUbSsXzCcDHXJAXJaWg5zEFNXliXoj7i6pxFs3PgUA/LK9Htvzy+Sfu5gSNJrD0VdeeQUAcODAATngA4Dc3FwcPHgQAPDqq6+Gtcx7/Phx1NTU4Hvf+x6Ki5dG5mKunL4ze2BiX8TZJqbHCp2gUy0bzGWJV5IknOufXF76wxU3Yfey9bPeLslgUp3510ZgCUi5tHvbPOzAMR1l1uuGdSikmpVI8YheNA734v9ruYj/+ckv8Zdn3prz1kfWGZZ39TqdaljxbNm+MY9LXo4D1MXnodDrdKqO6caR8Ov6lFuvrZ9hVEsgdTOHZYZrRp5yeXxjdhFMegOMOj0MvqouUZLg9EZv5xsl5Q4JSQYj/nzjF5Dr+2B2i1683PAeblij0+EcbDzS0PgY3mmvj8r9RYuynq88jCYOvySDETcp3181NnQ0WLrxH9cvyF97JRFn+6cO/Z7NxJ7Yk4F4YNC3IjVbbv6ye1yamt6US7vKZkClOwvXyJnuAadNtfKjDPq0BsALRVPQ19PTg/r6ehiNRuzevXvK5du2bUN+fj76+/tx8eLFkA6gvb0dzz77LKqqqrB3796QbruUddqH5cWMguR0GEMoQl9okWrm6HVYYXFN1Ekk6o3YPUuGT0n5wj3b3zqnGpwb1kG02SYH/M53A4dSmilRPgHwSiJaovRBF2jAacN73U34Xw3v43sf/yeeq/sNTnY0oMM+jGGXA7+c4wegdYblXSCwVnTmur5PBtpVBdlzOWGqVMyeDHeJt99hk4v/jTr9lA+pmSzk2Bbl0u5WRQOHMiifrwHNypMbk96ArIRkfHfDF+WMr9PrwQuXT6NL8eEbCV5RVJ24fnn5ZM34rzquoHdscTR1SJKket3MJdMHqE+kavtvqJa/g+l1jOLVzz6askSvbLbSqntsRJ4RmWpImPL6FgRBtaQ72+4coy6H3BQiAKpmJaVEgxFfLK6Qv/5l+2V5hNRiaeIANAZ9DQ0TqdLVq1cjMTF4KtbfQHHlivZWekmS8IMf/ABerxfPPPNMSNmT+vp6PPvss/jrv/5rHD58GKdOnYLLtXS6GRdTujiQsq4v3K2zAKj2KK3IzJ81Ra+0MbsIib6llz6HVQ7awqHM8lXlLkeKcX4bOAKtVpylR3K3CCWX14N6SxfevHYePz73c/zw7M9wrPksPh3sCLr5+OAcu6StLuWcvqknOOUh7MyhWtoNo4FDqSKgmSOckwfl0m5lZn5I+2crO3jbFDVZ0dY7NirPFDTq9FifPdltnKDorp6vAc3KJg7/rEBzUiq+u+EuOfti97jw95dPo99hC/ozwtFmH5JrMXMSUvAHyzfIJ7VeScTxa+cWRVNHv9MmZ9OTDSZNNaUzWZ9VKP/dh8bHZmwqc3jc+Mf69+XGiwxTkvxe3mIdRE+IgbOyqWpNZvBt5JRdvBcHO2Z8jM4PtMvB6Kr0PGQmTD8a7a6iCiTpJ3bp6HVYca6/DV5JlF8rALAsxufpavoU7eiYiJSLiqZfligsLFRdV4ujR4+itrYW3/72t1FWFlr25PTp0/jnf/5nvPnmm3j11Vfx7W9/G3fffTdqa2tD+jmxStm5WxzjZw6BlMu7HTZL0CBBiyuKF/faEOt1THqD6mwv3C4zZ2ADxzztwDGT1YpMUTR25uiwW/CDs/+FFy7/Dr/pakRPkBEVOQkpuL1gFXS+pT6bZ3xOw7hVc/owNehTZibabEPT3pdlfAxXffU+AoRpl2q0KknJkj/cRt1OdIcxp+2yYlRLKEu7AJBlSkaar+De6fWg31djF22fKGacrc8qRKLvgw4IzPTNU9AnqjN9fkUpmfjOhrvk2roRlwN/f/k3GI7QKBHlMuKqDDN0goA9q26WRyddGe5Z8CHFWijn85Wn5ar2qQ2HQadXdbhOt8QrSiL+T+NH6Pa9hxgEHZ5cd7tqy86PQ8z2BTZxBLMmI0+ermAZH5txzmXgQOaZJBtM+IIy29d2Gb1jVvn9KNOUFFaN4nzSdMo5NjbxAkpKmr54OiVlor7Cbtd2xt/W1oaf/OQn2LBhAx5//HFNtwGAZcuW4fvf/z5uv/12lJSUwOVy4erVq3j55ZdRW1uLffv24Y033kBlZeXsP0wjm82G8+fPh3XbcG/X6JwMnp3dAzjfF97PWShZggkWyQUREn51vgZF+tAGS4uShAbH5Iel2DmE892h/Q2yFMFmTVczyoZE1VmhlsfmimdY/sDJEkwYbmrFeSH0OpRIcoiTdVRNw32oPXcO+gjWGH7o6oU1IIOjh4AiXTJK9ClYpk9BhmCEMCrgvKCHXZr4+3x4vhbputCzoIH77vozfYGPT4ZgxIg0sV3br87VoEA/9f3oU/eQvIBUpEvCtctzH+KbJ5nQgoksxam6c9hg1H4S5pFEXHEo9qHuHAz5eZzpNcCKicfjd3XnscoQ/XEQHzpvyP/OsnpUj0WCIui73NSIcUN099kGgDbvZPZu3Gaf8ty421CId7wd8ELCgNOO/+fsL1GduHxKfWiozo1P1oaahh3y/a4zZKDeM7Ea83rjx3C39spbB8aCwL/PGddkPWqizRX255JShnfyNXum5zrWjAhTVmNqXf2o80wGXLcZ8zF49QbMnsna/w86GlEy4Na00icFfC54Zng9FUuJaPK9bt+5fBbbTFPrGG2iG83OiYBYAGDossz6+syRvDBCBzdEdDtG8W+fvidflubRafrbRuLvH64FeZb6l3U9Hg+eeeYZ6PXaX5j33Xcf9u3bh8rKSqSmpiI7Oxvbt2/HkSNHcM8998DhcOD555+P4tFHnyRJGBQnP3RzdOG11i+kPN3kB3KvGPpuBv2iE25MvDGkCAZkCMZZbjFVsS5ZzhrZJQ+6wziOK57JZfZKQ2wMyE7VGZHm+3t4IWFAjOwewxbFc2+VPg33JpTg0aRV+L8SS7DRmIVMnUn+O6Qo6u/8wV+oxiHKgZoRummX8fM1PKeaPJNZydURCo6KdJMnLJ1iaBmkbtEBr++3yxRMYQXFubrJzEF/hB/rYKyiGwO+54AOwHK9umNZGUiNS5HfajEYj2J5zhDk+VGkT8auhCJ5cLBFcuGd8clhu+GQJAk93snHu0Dx/LvZmKt6b/nEvXDb5GnR6518vSh/j7nI1yXJ70MuiGj3qhM+zZ5RXFQEfJsM2fJrcrk+BQm+8MMWwnuzRXLBiYnnXAL0yBKmfz2VKp63rd7gS/7XvZOZ8yJdctB64kCJgh7rDZNLuC2Kn529CD6rNWX6kpMn3vQcjukfGH+Gz5/xm8m//du/4ezZs/jWt74V0Yzck08+iZMnT+Kjjz6C2+2G0Rh6oBBMamoqKioqZr+igj+Sr6qqCvn++hxWeM5dBQCkGRPw+aptMRFshMLZcw2NTWcAAONpCahaH9rf4Rdtl4HWiYzaTXnLcfOam8M6jqZmCe91T2ydNpKVgOrVVZofm1brEAYuToxHMer0+OOq25AS5myrSKtrdKHGtyyiL8xGVQhNLrM5/vEN+OJtfGPrXTOO3Dl3ZQx9vuWtvNJlqAqjU7Z7bAQ4P1E3mZU4GWAFPj6OnmZcbZoo3xhPT0DVOvXlnfZhDH0y+XjdV3Ubkgxzfw8oHhvBR+d/AQDowzi2bN0CncasTkPTGcCXZKkqKkNVeejvB/rBDlxoeB8A4Ew2ompT6D8jFKc6rgC+Fbd1WYXYuWGb6vJzNSflf2cV5qFKsSl9tLh6rwNXJzI8edm5qKqc+jeoAlDSdwOvNf4eEiYCZHthumo5LhRd9mGMfzLxPpxqSMAXb96heh8WenPxr1c/BgDUeS34ytodCz5lIdh7m8PjgqVm4nWhg4C7q7arluvnouOGUe5iHkjT4wHfa7LVOoR/uXRKvt6GrCL8t/W3q143zc2Q35stmUZUr5n9eX266ypw7QYAYF1OEW5eN/3nwkavB7/7+D/hFr2wSC6UrF2N/GT1ieCpiycB38LJF1ZuQJXG8p0KtxNXan+mKjsAgJtXVs74HjiXuECrxsZG2GzT17Vqeufyj1Hp6pp+A+Oenh7VdWfy61//GgDw+9//Hg8//LDqP/+MvlOnTuHhhx/GN7/5TS2HCAAoLy8HALjdblgs8zveIJICO4EWW8AHBDRzjA6EXOx8xTK5HBFqPZ/SNkXdyScDbfBorDtzi178R8sn8tdVuctiJuAD1HV9kZzXZ3U55YJvk06P7ISZT+IyFftB+zutQ75PRV1Ymmn6epjyNPWQ5sDnlHIv0JuyiyMS8AFAflK6PALC4XVrbgpqGulTbdB+U5gDhJUdvG326DdzTNe166dc3p2vXTlcykaOGRphtuWV4o9KN8lfh7r/qlLTqLqeL/B9eHtemTxCSZQkHG+OzaaOFuugnEkvSc2MWMAHqLt4Lw11YszjwojLgX9seF+uc8tPSsefVu6ccqK0XTEF4fxAm6bxU1dnmM8XyKQ3qPbbvhDQxTvgtMnTD3SCgC052odtpxoTgw6WXgxNl5qCvnXrJrIITU1NcDqDLy/U1U0MXAy2S8Z0Lly4gNraWtV/nZ0TNRR9fX2ora0Nae17eHhyKc6fnVyMOmyLt3PXLz8pHcm+D12bZxx9Tu0F6E6vW9X1u1YxNiNU5elmZPm6sewel2rq+nREScT/+ez3qs7YOwvXhH0M0aBq5hjpl0cHzJWyC60gOWPWgu8sRadbuMXzyn13ZyqCLkiefE5Z3U7VXqyiJKmadcKdzReMIAiqLl4to1scHhf+pbFG/rBdm1mAiozwTl4yTUlI9/1dxr0e9EVx71fL+Jj82tNBwKYguxkkKj687fO0K4e6kWPmcqCd+eXyMm/TSH/YzSbNitf/6iABhuBr6vA3MzWO9OLcwMLW+wYT2MQRSYXJGVju2zjAI4k429eKV658IJ8AJumN+Na625EUZMvKsrQcedi40+vBp7OMVpEkSd3EkTn76CP16Bb1CYByBuy6zMKQT+rvLl6rGqVmEHRTMomxSFPQV1hYiPXr18PtduPdd9+dcnltbS16enpgNpuxZcuWWX/ekSNH0NjYGPS/b3/72wCAr3/962hsbMS5c+c0/zLvvPMOAKCsrAypqeHvArHQOsaUQd/i6tz10wmC6g0mlCHNzSP98PqCmOLkTKTPsvvCbMdxi4YuMz/Jd8auzBB8pXQzyuY41yrSzImp8owyh9eNTnvoXaXBKLtTizQsVWUpHhvLeOg1k4C6czdthjfewOeUcnRL80ifvPl7isGkOsOPBGXA1qjhxOGNa+cwOD5R8pJsMOHRNdvDztj7d+bwuzFDJ+JcKT8Y12TmBd2qK2EhuncDduSYSYYpCaVpE8O8JUioUwzq1koKGAC8apphxiUpWbizaPKE8D+ufwKnZ34GVmt1TbUTR/hDmaejnFv6xvVz8v0JEPDE2s9NGwgJgoDtipOz2bp4u8dG5VWBFIMJRcmzJ0Ruyi6Wg/IW66DqxDSUrt1g0k2JuKNwMttXlJIR0lixhaL5CPft2wcAOHz4MFpbJ/9Yg4ODOHToEADgiSeeUO29e/ToUezevRtPPfVURA62q6sLJ06cmDKPT5IkvP3223juuecAAI899lhE7m+hKMe1lMT4zJ+ZBC7xaqXMxlVmzX1rJWXQ9+lg54zF3T9vq8P7irl8XyyuwD0l2rPX80UQBNUHkZap81ooM31a6pOUM62GI7G8O8u4g3LVPryTH8pnFFm+m80rYIjwMHPlkOam0f4Zx9Oc62/Fx4qTi72rtqkyouFYrlzitUYv6FMt7eZMM6R2Aeb0uRSZPi1bnykzPJ8Ohh70DY7b5WxVgt4w4zaYf7hio5yJHXY58HPfNmOxQJREtKh24oj8yest5hXyCBtl6cEDZZtnHVGkDBgbLD0YcU1/4qhc2l2dkadp7EyKMUGVpfU/F3rHRtHuK6MyCDpN+/MGc0/JWvm1/bn8lbNcOzZonhK6e/du7NmzB8ePH0d1dTV27twJg8GAmpoa2Gw27Nq1a8qOGhaLBS0tLTCbI3N2MTIyggMHDuDHP/4x1q9fj7y8PNjtdjQ1NcnzAffu3YuHHnooIve3EBwel5wh0As6FMxxiOZCUtZghTKkWV3PF/7Srt+ylCwUJKWjxzGKcdGDVq8NK4N0dr7X1YSft12Wv95mLsVXy7bGbE3l6ow8nPctJzWN9oVdsK4UcqZPEcxYwl3eVWT6Jj48py9CXhlk8Ldb9Kq2DJvrQOZgchJTYE5MRb/TBrfoRYt1MGhNkWV8DK83n5W/3p5XiqowsgiB1DtzRCfos7qccnZLALA5N/gHYaKqpm8BduSYJdMHAJtySuR9Uhss3XCL3pB2NWoeUS+JzpTBSTKY8NXyLXitsQYA8JvORuzIL4+J7bi6xkbg9P3tMkxJyJmlRjccGaYkVGbmq07Wt+eVYlfx7E2auYmpWJVuRvNoP0RIONvfOu3trmqYzxfM5pwSea/eC4PtuKNoNc4NTCauNmQXBV1+1iLdlISDVX8Am3t8TnvMz6eQcpEHDx7E4cOHsW7dOtTW1uLDDz/EihUr8KMf/QgvvvhiSKNXwlFQUIDHH38c69evR1tbG37961/jo48+giRJuPfee/Gv//qv+Ou//uuoHkO0KXfiKExOj3jGYj6VpeXIZ4Cd9mE4NCx7jLgc6PQtb+sFXdBamlAJAUu8zd6p9YXn+9tw/Nrkh/W6rEI8uubWOQ8xjSblG1/TSH9EishVQV/K7EFfhmJ5d8TllJflQ6EM+oItJyqVBnlOXR7qwpjvuZWTkDLnLaamo8z2BVviFSUJ/+/Vj+WdB3ISUvDQyvC6zgOpmjlsQxGr4VS6ONgh70ywMt2semyVVMu787UjhyKzqmVHk4KkdLlebFz04DMNS/JKyuYoLe9B28ylWJ0+cT0REt6IkZ06VEu7ablRO4HdmV8u/7s0LQd7V9+q+b525E9m+z7uDb7EK0lSQNCnfQVIefLSONKLMY9LVc938zTbrmmVqDcumoAPCCHT51ddXY3q6mpN192/fz/2798f0s+f6TZZWVkRWyqOVerO3YU/U5yLRIMRxSkZ6PDtI3zDOoi1WTNn7pRb7JSn5Uas0+yWvBU44Vt2affa3fcFzwAAIABJREFUVPPFPhvukcc8ABNvWt9c+/mYD7gLkzOQbDBhzOOC1e1En8M6p0LiUDt3gYnRKGnGBFjd45AgYdTlDHkp0xawvDtTvjBRb0RJSiba7RZIAFqsA+pt1/JKo/bBVpGZL2/J99lwL6oDNvv4bVejnO0QAHyjYkfYGYRAmQnJyDAlYcTlgEv0omfMqikoD4UyW7old/pORpOvSkrCRK2dR/RG/bXiUtX0zX5fgiBgU3YJTnVODOe+ONiBjSF0Tyszfas11MH5mzr+70/egYiJAOXCYHvQ7uf5pNxvNxpLu363mFfA4hrDyLgD9y5fH1JWtSp3OY43n4NHEtFut6DTPjwlS9rrGJVPDpMNppCyqNkJKViRmo1W3zaGJzsaVFsMbswJr6t+sYr9qsM4o95zd3E2cSipmjms0+/P6KdcIlgbgXo+v/ykdDlbIgJo8WX72mxD+F8N78t1fvlJ6di//o6IjjWIFp0gqD6Q5jq6JdTOXT/l2JZwOniV3bvpM4xs8VMWo9cNdeGSolA/kl27gZTNHNetA6rmgk77MN5quSh/fU/JuohkqZXUS7yRHQZsd7vwmaJmausM4ysEQQjYfzf6S7zKv7XWvYs3Kz7MLw12ah51Y3U55a0H9YJObgqZTXFKJu5SNHWcmaVpbD5Eu4nDTxAE3FOyDg+urAp5G7Ikg0nVJR6soaMxoJM61BUYZc3er9ond+m5Kbt4UbzXRxKDvhijauJYAkHfynT1bLWZSJKkquerjEA9n5Jyifeax4o+hxUvXP6d/IGSaUrCdzfcFfN7JyopA4urc2zmCLWez09V1xdGM8eoItM32/IuoK7re7+7SQ7Yl6dmRXU4bropEcW+jkFRMT7CLXrxWuPv5eNYlpKF6hUbI37/y6NY13dpqEMOikrTcpCdOHOWd7733w21kQOYyGz5u8FH3U7csGoLlJsV2bHStGzNQSYA3F44Ody3wdKtafZctIy6HOj3jTUyCLoZm1EWmnKJt7bvxpTyhavD2ufzBbNZcRIzuf8PVPsHxwsGfTFElER0Kj54F/vyLqBeUmixDsx4tt3rsMpBQ6LeqPkMW6ubzSvk+V2d4hj+vu63iiUDI76z4a5ZP+xijbKDt0Xjh9p0Qu3c9VPP6gttbItXEjHmCTXom/ydlZ3Y0WjgCKSc1+cvDv+vG5fkDL1Rp8fjlTujstxZmqau64skddfu7ENqE+d5QLMrhJEtfjpBp1rSnW0OnJ8yY74qPbQAoyA5A/m+5juX6NU0F1QLh8elebC8n3Jiwoq0nJCWXOfbusxCuXN/2OVQzcKcWs8XetBXmJyOfF+Np1+C3oANER7ttBgw6IshfQ6rPAoiw5Q04+4Ei4U5MRWphokP8jGPGz1j0w+WVb5BVmTmR3zmUVZCsioz5u+SNur0+Na6O2Oi2y5UJalZ8lJHn8MqNxGEI9xMX6ZyVl+ImT67e1w+704xJGh6zLMTklX3CUzMBLtlHs7aKwOGNDcO9+LXnZPLRQ+UbY5atlE1tsVmCatpJhinx40GS7f89dYZ6vn81M0c87C8G2Ijh98m1egWbUGfqp4vI/Ql0c1h3OdMTnddxXdr/gN/++mvQnp9X7Mql3Zja85oIL1Op3r9Kpd4ex1WjCpOzsNJhgjC1EHjm3NKQnouLRUM+mJIu33x78QRSBCEoGM2glF22M1l67WZ3BKQDdJBwL7Kz2NVGG/uscCo06ueK61zmOEWaueu31zGtqhn9GmbiC8IwpSi9MrMfNXMwGhZk5Endw+324ZUDUDrsgqjunNLhilJDnbdonfGE6hQ1Fm65IxpSUomzAEZkWCUs/rmZXk3xJEtfusyC+QMV7djFL2z7Gbi9ExusycgvDq4wBmBc+m09ohenGi9BGAi0H/bN4ZGi8DO3VinXOK9MNAOp3eiI181ny89T/O+14ECt1m7OTf+lnYBBn0xpXOJNXH4qYc0B2/m8EqiKqUfifl8wWzNXSZPaAeAh9fcipsWefdWaerkMni4Bf7hdO76qRs5QlveHXUpd+PQntkO/DDeFsUGDqUkg0neHUPCxFIUMJGlfDSEMRXhWpGmfKwjs8SrzEZt1rj/aIJqK7boB32qHTlCGA1m0huwTvFecmmWQc3XrQPy2JrilEwkh9F9XZqWIw9rtnnGZ61lnkm9pVvVKPN+d5OmQfdu0YtWRblHNDt3I2VZSpa8wuASvbjgKzm4Ost2eFqVpuXIS7xZpmSsm2WSxFLFoC+GLKVxLUrKXRSme8NqtQ7JZ3ZZpmS5LibSUo0JuNNUgEJdEr5RsUM1X2qxUm7RFW6mrzvMzl0AyEoIf3nXptyCzaR970tl5sKo04e0WfpcVQaZEbZ39bZ5yTSuUBTjt86xhhOYyCTVDXXJX2+ZZiBzoHmv6RND797126Taf3Xm5Vblzjah1vP56QKWEueyxBvYASwBeL25Fl5x5uxhu80iZ2/Nialz2spyvgiCoNqh4+O+lin1fBVzWAHSCQL2r78T95duxnc33hXzI7mihUFfDOmwLc1MX2lqthxEdDtGg07xV2+9VhDVjMkqQzqqE5dju+INZjFbocj03Qgz09elqucLLeAOHNkSylDaULZgU1qRli0Pw/1S8VokGeZv7EJgV/nO/HJNdXCRsCI1spm+qyN98slWbmKK3J08m8R5rOkTJQkuZU1fCMu7ALAxu1jO7V8bHVCdaARSdu6GU8/ntzkg0AxnULPD41aNIzL4sqsd9mH8uuuzGW+r3KIw1uv5lG7NK5Ufq8bhXlwd6ZO3ZkvSh1fPp2ROSsM9y9ahIIpd/rGOQV+MsLvH5SyJQdAhP3n2uprFwqQ3YJkiiG0JUtc3H/V8S1VRcoZctzQ0PqZaMtVK3bkb2htrosGIJN+sK48khrTcN6rM9Gms6QMmOjO/f9MXcfjW+/GHpTdpP9gIWJmeiyxfoJubmIqvlVfN230rmznabRZNu9zMRJn52pRTovlkSzWnL8qZPuU+x0adPuQZbemmRHl5U4KES4rMZuD9KDvg57KUWJGZL4+W6XfaVK8vrS4Otsu/e0lKJv6odJN82YnWOgw4p9+uULmiUh7F+XyRlpWQLJ9USQCOK7Y0XJVhDruejybxLxgjlEOZi1IyI965utDUdX3qoG/c61HVvUR6Pt9Sp9fpVEF1OHV94Xbu+mWqmjm01/UF7sYRCkEQFqTD3aQ34Lsb78JDK6vw3zffg8R5zDKmmxLlx9ojibgw2D7LLaYnSlJY9XxAwPJulGv6xsNs4lDalD37cmubbUgOssyJqdNuQ6eFUafHhqwi+evZlpWDUS7tbjOX4otFFXKmyy16cbw5+FZvkiSpGuYWU6YPUA9Y71Y03lSEsPUaTW9pRRaLWPsSrefzU+7McS1gZ46mkT55/ERxcuac3mzj1Vzr+rrDnNHnl6Uc2xJCB6/VHV4jx0IrSM7AXUUVmuYKRtoteZNdh7Vz2PWh1TaoakQJJThQjWwJUq4RSerBzOHVYSlr7KYbmtykGNWyKgK7qcxldMuIYladgInHXK/T4eurtsnLn5ctXar5in42ySM/rol6Q1gncQtpS+4ymILU20V6h5t4xaAvRnQswXEtSisDhggrxxio6/l4NheO0jnU9QV27uaEMaBaNaA5hGYOa5jLu/Fsm3my7umz4V655ilUyuzTTTnFIa0uzGcjRziDmQMVJKfPOjRZ2cShZb/d2WzILpKXolttQxjyzQXV4mx/q9xFvDojT+6mL0/Pxe2Fq+Xr/fT6eTgCaip7xcnnQ1la7qJbEk3UG6fUyCbqjTG9o8hisrieDUvYUtt+LVC2b8N4YGK5ptM+mVlS1/NxaTccpQGZvlAKx+fSueunbOYILdOnWN5dAsPI54NyyLgECWf7W8P6OZ8qxpdsztHWtesXuPeu1n1twzE+h85dpU05yt051KNbRElUNT9EIquUbDCpliQD73MmyqXdwP2kv1K6SR4JM+JyTJndpwz6FtvSrt+tAU12qzPMS67kaaHwrxgDvKKo+uBdikGfIAiqMRv+mpNRl0POcuoFHVP4YcpLSkei7wNx1O2Ul3e0mEvnrp+qpi+E+7aGOacv3im3nAtnibd3bFR+zzHq9KpZdlroBUF+vkmQ5txQMhOXd7KRQ+u+u8EoA9tLQ+qhyZ32ETh8XczpxkSYE1PDvh+lcEa39IyNytvsGQQdtuYuV12eZDDhaysnm4fe625SNcf1epVB3+Jp4lCqzMxX7brDz4XIYdAXA3oco/JMpayEZKQYQx8IuhgEa+ZQDmQuT8tFon7+iuKXEp0gqDo7tW4uD8ytc9dPOatvWGOmzy165Q9aHYSwBuHGq625y+URHq22IfSGuDvHxaHJAGR9VmFYGbQUw+RyfDSbOdSNHOHPVitLy5FPLKxup6pTt3lUPQA4UiOjlNnFxpFeTduoKYP4DdlFQV8XVbnL5X1jJQBHm2rhlUS4JRGD0sRjIWDid16MdIIOu4orAUyclFQFBL4UPgZ9MaBzidfz+QXbmUNZW7OW9XxzEu4Mt7l27gJTZ/VpoezcTTUmhLWsHK9SjCZsyJ7sDq3tvxHS7dVdu6Et7U4egyLoi2JdX6SWd3WCDjdlTwZhyppGVRNHBLNj2QkpWOE7GRMlCZenGRfjJ0kSzigey8ClXT9BELBn1S3yqKYO+zB+29mIftEpbwtYlJyJpEV8IvXF4kr8xU27cKjqy8iNUOaVGPTFBHXn7tJb2vVbnpotZyf6nDZYXU51Ewfr+eZEWdcXSqZvrp27QMD+uxobORZr526sCFzi1VrHOeJyyJl2AQI2Zoe3DWGqIqCwRTHTF4lGDr9gdX2SJAUMZY7sUmLgoOaZtFgH5fl7SXrjjI9NbmIqqpdvlL/+WeslNHsmM76LtZ7PTycIWJWRF1ZjGU2PQd//3969R0dV3/v/f81kcr8QkBACBUMMdwExNhVsqeuIrqjFb5XT2njBeihYsWhPpXhpT0s4tbVHFEt19SBa1AA5x9OKhdUFFo8U/B1TA9hohBBQAgZoJAYQEnKf+f0RZtg7mcBMsvdMmHk+1nKtsGfvySfZbnjP+/N5vz/9gLFyd0QEB32xzhjTFGTpsWrfov+EmFhlX6RTEf1F10xfIEGAFZW7kpTsivNlHZo72gNa42UM+sLR+uRiN2nQMN+6umPNDQFndz+sP+LLBo0ekNHr370502df25a+bMHW1fj0ob4p4s+aTqn2zCl93tzgq4BOiInV8GRrW5wYg76PThw1NZvu6r1j1b6vrxw80vdM9WTm8HG+XVRa3R3a23HuA9zFsN8uQo+grx8wV+5G7vSuZP6LaMvhSt/XY9Mzqc7qo8EJyUo+m30509563o79XlZU7kqd003GhdeBtG0xVu6mUbkbtLgYl6Ya1jp13ae1Jx8cN+/C0VvGNX12ZvpaLCrkkDp/ZxPOroWTOqe595n227V+14espAG+wpCWjnZVGdYxG3W43dpZ96nvzz1N7RrFOJ26c/SX/b52sRZxwF78Kxtmp1qbfVtRxTljlJEY2WsXjE2ajVtwsfVa3zkcDt/6ISmwKV4rKne9TFO8AazrM1fukunrjfyMc42ad9Yd8jU570lze5sqT5xbUtHb9XySlGIoOLNzTV+rRYUcXqaK2uOHLdtvtycOhyOgKd49J//hC57T4xIDnma+LC1DM4bmmo6lxsZbVoGMyELQF2aHDev5hienX3SNNIPV0zoT+vNZ49LU4Io5rKjc9TIVcwTQtuW0qZCDTF9vjEvP9PVsO9XW3GMWyWv3iX/4OgV8KTm9TwvkzdW79k3vWlXI4TV50DA5zra3PnDqc1NxRW6aPa1Buu7O4a+vobFq98sZ2UFl3W8ddYXv/wOp88O1VRXIiCyRHWFcBIzr+YZH+NSu1NnPbZAhIyR1fqr1dstH3xjXRR4MYDs2Kyp3vdLjg9uKrYFCjj5zOpz6ckbg27KVW1C162VcC9gQokxfXws5pM4PGN4Pnx6dm3FwOZym7QytlJM22JfNPtXWbOqrJ0nNHW2mexPI1K5RkitOhblX+XZq6drcGPAi6Auzw1FSuWvUda3J+IFZfCq1iHF699PG46YGtP5YUbnrNTDIti3G6f00pnd7Ld8QIPy9vsbvvrJS55qxj04Yd+EY4fe8QBn7idrZp8+8927fgz7Jf8A7KnXwBQsneqtru5iuu3N8UH9YrWcLPIYlDejV2u4rB4/UrQmX6pb4EcrLoK8d/CPoC7PDDcbK3cjP9EnmdX0S6/msNLDLdne1Z073eK5VlbvG7+0VSNsWpnetcWnKIA1JTJXUWTn94XH/233t++KYzpytqr4kPrnPRWOm6V0bq3eNhRxxFmT6pM69hrvKtWE9n9GU86zrMxbh5A/J7vWH4MHOBA2NSbrwiYhaBH1h1Obu0D+azmVahkdJpq9rKwH681nLmO071NBzMYdVlbte6aZdOQJY02cs5Igj09dbDofD3LOvh714jYHGlEuG9zm7bpretbN617Smz5pMXGZimrK6LCmxo4jDqHu7mM7n71Rrs6m4xngvAasR9IVRQ1uLb0Hv4IRkJbqiYwuyEckDlXk2MzE+fagvMwVrmJs097yuz8rKXck8vRvYmj5DyxYyfX2SP+Tcur6Pjh/tVk3r8Xi67MLRt6ldqWum7+JZ0+dlzLw55NBlqfYGfXExLk0ceG4XFW8QvrPukNxnOyfmpmXQjBi2IugLo/S4ROWmZcghh2740oRwDydkYpxOLZo8U/eN/6ruG//VcA8n4pibNAeW6etr5a7U2WvPeXYpeUN7y3mb0LZ0tPsyODEOJ3su91FmYpqyz2Z4Ozxu7fq8xvT6pw0nfFPuSa44S6YyE2Jcvuxwq7vjvPe7L0x771q0pk+Spg4+F/hmpw5SQgg+dBt3BPEGfcYt9PKDLOAAgmXdE4SgORwOPTz5OrW6O6LuH720uERdySbatjBm+moaTqjD7VaMs/vnu6MWVu5KnYvVB8Ql+oKLky1NPfadNG/BFk8hjwXyh2Tr4Nk2PWXHDmpG1rnebeX154LAyYOGW9II3eFwKMUV7yvIaWhrMa3rtEqr27rmzEbZqZfo1uwpqjr5mf5f9hTL3vd8Jg0aLqcccsuj6tP12v/FMVWf7afpdDiUx9+JsBmZvjBzkuWAxVJiE3RJfOcUUbvHrSNnTvo9z8rKXS9T25bzFHMYp3Zp12KNqzIu9fWf23/qmI63NPpeM1aL9rVVi5FpKzab1vXZNb0rSQUjJuqhSf8Usi0gU2LjTU2XX933N9/XkwYOYztC2I6gD4hAl15gXZ/VlbtexkzP+dq2mDJ9bMFmiQFxiaZK+B3HOgs66ppO+wL/WGeMaRuyvvJu+yfZV8FrRyFHOBmD7mOGrRKZ2kUoEPQBESj7Auv6rK7c9TLuynG+TN8ptmCzhTFw8K4VM1btjk8faukUaSgaNJu3Ybv4VyT52+84IcZl6uMH2IWgD4hAxkzfIT+ZPqsrd70CzfQxvWuPKy4Z4WswfLjxpI40nrR0F46uzFuxWR/0dXjcvm3jHJJtzZND6ZKEZI3o0p5r6iUjLC1SAXpC0AdEIGOvviNnTnbbpcHqyl2vgXGB9eo7zRZstkh0xZoyRm8frdInpzq3/HLIYXk2ybgrR4MN07utxsbMMa6IKfjpmu1jahehQtAHRKBEV5yvF6Lb4zHt8SxZX7nrlR7grhxdq3dhHWMA8f/VfiLP2R5wl6UNtnz9ZIrNmT7TFmwRMLXrNXXwuaAvLTZB49iVCCFC0AdEKGO/voOnzev67KjclbpsxXbeQg6md+1y+cAsJRkKLLysntqVulTv2rCmz64efeH2peSBunnk5fpScrrmjPmKnBa00AECwf9pQIQy9us71HBuXZ9dlbuSTLurnGptVsfZ9Vhdmat3yfRZyeWM8dvvzY6gL8UQXNqxFZsp6IuA9XxGt1w6Wf925U2aRAEHQoigD4hQpp05DJk+uyp3pc6F9t7pWrc8pipdo9OtZPrslJ9xqenPw5IGKOPsdL+VzJk+G9b0Gad3IyjTB4QLQR8QoUamDPQ1661tOqXm9jZJ0j/OnPKdY2XlrpexbYu/Cl6Px0Mhh81yBwwxTbXbkeWTzC1bbFnTZyzkiKA1fUC4EPQBESouxuUr0vBI+vTsFO9Rww4dVlbuepnW9bV2r+Bt7mj3teGIc8aQwbGB0+HQPw0bK6kz+3p15ihbvo+xZYsd1bstZPoAS/EUAREsO3WQbzeGQw3HNSY901S5m2VLps/YtqV7po8sX2jMHD5O2SmDNDA+yZapXcncsuVMe6vcHo+lywVMW7AR9AF9RqYPiGD+KnjN07t2Z/ouFPRRxGEXp8OhMemZtgV8khTjcCrx7N7hHnnU1G5tti+SCzmAcCDoAyKYaQ/ehuNqaGv2BV2xFlfuel1oVw5Tuxb23b3oGYs5rK7gNe+7S6YP6CuCPiCCDU9OV8zZHmCfNzdo/xd1vteyktIsnYrzMu2/62dXjtOGit4Upncvesa2LVZX8JqmdynkAPqMoA+IYLHOGH0p+dwU7t+OVfu+tnInDqOB8efW9Pmf3jW2a2F692JnyvRZ3KC51W3ehg1A3xD0AREuO/Xcur6K40d9X9tRuSt1b9ni8XhMrxvX9KWR6bvoJdu4FVsLhRyApQj6gAh3acq5dX3GHTLsqNyVpARXrBLOLu5v97i7BQJU70aWFEMFr/WZPgo5ACsR9AERzpjpM7KjctfLvAeveV2fMehLYXr3omfO9NlYvUumD+gzgj4gwg1NSlNslyyJXZW7XgONvfq6rOszZoPSqN696Jm3YrM400chB2Apgj4gwsU4nBppmOKV7Kvc9TJn+sxBn3E/XqZ3L37G6l2rW7YYCzlY0wf0HUEfEAUu7RL02VW562Vu23Iu6HN7PKZMH9O7Fz9zps/O5swEfUBfEfQBUSA7tWumz771fJKUbmzQbNh/t6m9VW51VvMmxMR2m3bGxccYuFtdvWsq5Ijh/xWgr4IO+jZu3Kg77rhDeXl5mjp1qm677TatXbtWbrf7whdfwH//939r7NixGjt2rJYuXXrec7dv365/+Zd/UX5+vqZMmaJvfOMb+t3vfqfWVus3/QYudtkp5mIOuyp3vUy9+gyZPnO7FrJ8kcBYyGF19W4La/oASwUV9BUVFWnRokX66KOPdNVVV2n69Ok6ePCgli5dqgcffLBPgd+RI0f061//Wo4A1hmtWrVK8+bN09/+9jdNmDBBX//611VfX69nn31Wd999t5qauu8CAESzjMRU3x6pUudOHXbq2qvP65Rpapf1fJEg2dCyhepdoH8LOOh78803tW7dOmVkZGjDhg1auXKlnn/+ef3lL3/RZZddpi1btqi4uLhXg/B4PPrJT34ij8ejb37zm+c9t6KiQk8//bQSExNVUlKil19+WStWrNBbb72lL3/5yyovL9fy5ct7NQ4gUjkdDs26dJJinTH66tDLNDghxdbvZyrkMEzvNhh79FG5GxHinS65zm711+buMFXc9hWFHIC1Ag76Vq5cKUlatGiRsrOzfccHDx6sJUuWSOrMwPUm21dSUqLS0lL96Ec/0vDhw8977qpVq+TxePS9731PU6ZM8R1PTk7Wr371KzmdTq1bt06nTp0KehxAJLtu+DitmP4t3T36K7Z/r2RXnG+9XnNHm5ra2yRJp1vZgi3SOBwO81ZsFq3r83g8tGwBLBZQ0FdbW6vdu3crNjZWBQUF3V7Pz89XZmam6urqVF5eHtQAampq9NRTTykvL0933XXXec9tbW3V9u3bJUm33HJLt9dHjBihK664Qm1tbdq2bVtQ4wCigdMRmtoth8OhdD+9+k6xG0dESja0bbGqgrfd4/YV/cQ4nIpxUncI9FVAT9GePXskSaNHj1ZCgv+/qCdNmiRJqqysDPibezwePf744+ro6NATTzxxwfV81dXVampqUnp6ukaOHHnecXjHDCA8jFO8J8/uymGa3iXTFzGMFbxWFXO0drAFG2C1gIK+w4cPS5KGDRvW4zlZWVmmcwOxZs0alZWV6Qc/+IFGjRoV8Di838sf7xiPHDkS8DgAWM/Uq+9spu90m3F6l0xfpDBvxWZN0NdiaNfCej7AGgE9SWfOdP6FnZiY2OM5ycmdWzo1NjYG9I0//fRTPf3007r88ss1d+7cgK4JZBxJSUlBjSMQDQ0N2rVrV6+u7e11sB/3xl4trefW1e6u/ljxh0/oaHOd71jtwU+1q6a+x+u5P/2b8f40tZy715Wf7Jfj08/7/P4n3eemiT1t7fz/EAR+V/1bOO9PWBZJeKd129vb9cQTTyiGpptAxEl2nPtM2ejpLORo9pyrxkxw8NxHCuO9bFbfe7ZKUrvhfVzsIwBYIqBMnzd7dr7+d97Mmjfjdz6vvvqqduzYoQceeEDjxo0LZAgBj8ObDQxkHIFKSUnR2LFjg7rGG8nn5eVZNg5Yg3sTGs7Pa/Ru5TFJkistWXkT87Su9KB0dtbuK1Ou1IC47ll77k//5u/+1B+uVHn1cUnSgIxLlHdZ3+/d/i+OSR8ekiSlp6Ypbwr/P1wIz07/For7U1VVpYaGhh5fDyjo87ZROXr0aI/n1NbWms49n7feekuS9O6772rHjh2m17xr8bZs2aL9+/crKSnJ1y7G+97/+Mc/enxv72uBjAOAfdINu3KcbGmS2+M2rfdi393IYaretWpNH4UcgOUCCvomTJggSdq/f7+am5v9VvBWVFRIksaPHx/wN//73//e42vHjh3TsWPHlJqa6juWk5OjhIQEnTx5Up9++qnfCt4PP/ww6HEAsN5A464crWfU0NZ6tgFHZ5AQE6L2MbCfuXrXmpYtrRRyAJYL6G/drKwsTZw4UW1tbdq8eXO318vKylRbW6uMjAxNnTr1gu9XXFysqqoqv//94Ac/kCTdeeedqqqq0s6dO32OT8W+AAAgAElEQVTXxcXFacaMGZKkDRs2dHvfmpoalZeXKzY2Vtdee20gPxoAm6TFJcipzjZMp9tadLzlXHEVlbuRxY7q3daOc+s/42jMDFgi4I/a8+fPlyQtW7ZMhw4d8h2vr69XUVGRJGnevHlyGhporlmzRgUFBVq8eLFV49W8efPkcDj04osv+rJ6Uueawscff1xut1t33HGH0tLs3VAewPk5HU7Tmr3DjSd9XxP0RZYU4/67FvXpo2ULYL2An6SCggIVFhaqpKREs2bN0vTp0+VyuVRaWqqGhgbNnDmz244aJ06cUHV1tTIyMiwb8OTJk/Xwww9r2bJl+s53vqOrr75aqamp2rFjh+rr6zVlyhT967/+q2XfD0Dvpccn+nr01TQc9x2nMXNkMWb6Gtotmt41rumjwwNgiaA+Pi1ZskR5eXlau3atysrK5Ha7lZOTo9mzZ6uwsNCU5bPTvHnzNHbsWK1evVoVFRVqaWnRiBEjdPfdd2vu3LmKi4u78JsAsN3A+CRVn+7sxfdpwwnf8dQ4Mn2RJMmQ6Wtqb5Xb4+7zln/mQg4yfYAVgn6SZs2apVmzZgV07sKFC7Vw4cKg3j/Qa2bMmOFb3wegfzLuynG40RD0kemLKDEOp5JcsTrT3iaPpDPtrUrp4xQ+07uA9SifA2Ab4/67re5zC/P7GhCg/zFN8VpQwWuc3o0n0wdYgqAPgG0G+mm+LElpBH0RJznW2gpe44eEODJ9gCUI+gDYJt2Q6TNiejfypBgaNDdYUMHbYsr0UcgBWIGgD4BtBvYQ9DG9G3nMmT4LpnfdxupdMn2AFQj6ANjG3966kpQWR6Yv0pgaNFuc6aN6F7AGQR8A28Q6Y7pN5TrkUJKLoC/SmLZis2JNXwfVu4DVCPoA2MrYtkXqXM/ndDjCNBrYJcWU6bNievdcIQdBH2ANgj4Atuq6ri+FIo6IlGzxVmxM7wLWI+gDYKv0Luv62Hc3Mpm3YrOiZYtxepfqXcAKBH0AbNU105fGFmwRyZjBtWJ6l0wfYD2CPgC2Yno3Opimd/uY6XN7PF2aM5PpA6xA0AfAVt0LOcj0RaIU0zZsLfJ4PL1+r3ZDwBfrjJHTwT9VgBV4kgDYamA8a/qiQVyMS7Fnd85o97hNmbpgMbUL2IOgD4Ct/LVsQWRKdllTwdtCEQdgC4I+ALZKcMUqISbW92cyfZHLqgbNrWT6AFsQ9AGwXVZSmqTO3TgGJySHeTSwS3KXdX291cK+u4AteJoA2O7bOXn686cfadKgYUrvUs2LyGFVg+bWDsNuHGT6AMvwNAGwXU7aYC28/NpwDwM2M1Xwtve+Vx+NmQF7ML0LALBEsqlBcx+mdzuY3gXsQNAHALCEqXq3D4UctGwB7EHQBwCwhKl6tw9bsZmndwn6AKsQ9AEALGGs3u1Lps9YyEGmD7AOQR8AwBJWVe/SnBmwB0EfAMASllXvUsgB2IKgDwBgCTuqd+nTB1iHoA8AYIkkV6wcZ79u6mhTh8fdq/dpZUcOwBYEfQAASzgdTiUZ27b0soK3hR05AFsQ9AEALGOa4u1lBa8500chB2AVgj4AgGVMDZp7ua6PNX2APQj6AACWMTVo7mUFL2v6AHsQ9AEALGNq0NzrTB/NmQE7EPQBACxjbNDc0Ns1fR1swwbYgaAPAGCZFFOmr+/Tu+zIAViHoA8AYBkrqneNhRxM7wLWIegDAFjGtBVbL9b0uT1utZ9t6uyQFOsk0wdYhaAPAGAZ45q+3kzvdi3icDgc5zkbQDAI+gAAljFV7/Ziepd2LYB9CPoAAJYx9enrxfSuuXKXqV3ASgR9AADLmHbkaG+Vx+MJ6voWN0UcgF0I+gAAlomLcfmKLzo8blMlbiBa6NEH2IagDwBgKVMFb5Dr+lrZjQOwDUEfAMBSxnV9wVbwmqZ3WdMHWIqgDwBgKVPblqAzfYbpXTJ9gKUI+gAAlkruQ4Nm8xZsBH2AlQj6AACW6kvbFtMWbAR9gKUI+gAAlkqLTfB9faqtOahrjTtyML0LWIugDwBgqbQ4Q9DX2hTUta0UcgC2IegDAFiqL5k+YyEHLVsAaxH0AQAslWrK9AU5vUshB2Abgj4AgKXSYhN9X/cp00fQB1iKoA8AYCnjmr7Trc1B7b/b4qaQA7BL0E/Uxo0bVVJSoqqqKrndbo0aNUqzZ89WYWGhnM7AY8gNGzbonXfe0d69e1VXV6fTp08rKSlJubm5uvnmm3X77bcrNja223WPPvqo1q9f3+P7jho1Sps3bw72xwIAWCQ+xqV4p0st7na1e9xq6mhTkivuwheqa6aPQg7ASkEFfUVFRVq3bp3i4+M1bdo0uVwulZaWaunSpSotLdWKFSsCDvxKSkpUXl6u3NxcTZo0SampqTp27JjKy8v1/vvva+PGjVq9erWSkpL8Xn/llVfq0ksv7XY8IyMjmB8JAGCD1LgEtTQ3SOpc1xdw0OdmRw7ALgE/UW+++abWrVunjIwMrVmzRtnZ2ZKkzz//XHPmzNGWLVtUXFyse+65J6D3e+yxx5Sdna20tDTT8draWt17770qLy/Xiy++qAcffNDv9d/61rd02223BTp8AEAIDYhL0OfeoK+tWUOVdoErOtGcGbBPwPOxK1eulCQtWrTIF/BJ0uDBg7VkyRJJ0qpVq+R2uwN6v8mTJ3cL+CRp6NChuu+++yRJ7777bqDDAwD0I6mx5nV9gWLvXcA+AQV9tbW12r17t2JjY1VQUNDt9fz8fGVmZqqurk7l5eV9HpTL1fmg+1vTBwDo/8y9+gJv0Gws5CDTB1groCdqz549kqTRo0crISHB7zmTJk3SZ599psrKSl155ZW9HtDx48f10ksvSZKuu+66Hs977733VFVVpTNnzuiSSy5RXl6errnmmqCKSQAA9kjrZa++FpozA7YJ6Ik6fPiwJGnYsGE9npOVlWU6N1Bvv/22/vKXv6ijo0N1dXV6//331dLSottuu0133nlnj9e98cYb3Y7l5ubqmWee0dixY4Maw4U0NDRo165dvbq2t9fBftyb/o37079d6P6cbDvh+/rA0cPaVd9+nrM7eTwetXS0+f68+4MPFeNw9H6QUYpnp38L5/0JKOg7c+aMJCkxMbHHc5KTkyVJjY2NQQ1g79693Vqw3HPPPVq4cKHf6d1x48bppz/9qaZPn66srCw1NDRoz549Wr58ufbu3at7771X69evV2ZmZlDjAABYJ9Fx7p+XM7pwwCdJbknejn4OiYAPsFjYc+cLFizQggUL1NraqqNHj2rTpk164YUX9NZbb+mFF15Qbm6u6fzvfve7pj8nJSVpyJAhmj59uu6++26Vl5dr5cqV+tnPfmbZGFNSUoLOHnoj+by8PMvGAWtwb/o37k//Fuj9GfBFnbZ8eFSS5ExKUN4VF76fjW2t0t/2SZISYmL5fyBIPDv9WyjuT1VVlRoaGnp8PaAFcN5eeU1NPS/G9Wb4vBm/YMXFxSk7O1v333+/nnzySR05ckSLFy8OuJN7XFyc5s+fL0natm1br8YAALBG1105AtHKvruArQIK+oYPHy5JOnr0aI/n1NbWms7tixtuuEEpKSnavXt3UGsEc3JyJEmfffZZn8cAAOg9c/VuYFuxmYs42I0DsFpAQd+ECRMkSfv371dzs/9PbBUVFZKk8ePH93lQDodD6enpkjqreQN18uRJSb3PNgIArBEf41Ls2cCtzd2h5o4Lr+szZvpo1wJYL6CgLysrSxMnTlRbW5vffW3LyspUW1urjIwMTZ06tc+Dqqmp0ZEjR+R0OjVixIiAr9u0aZMk6fLLL+/zGAAAvedwOILu1WdqzEzQB1gu4KZ23vVyy5Yt06FDh3zH6+vrVVRUJEmaN2+eqU/emjVrVFBQoMWLF5ve6+OPP9bGjRvV0tLS7fvs27dPDz30kDwej66//noNGjTI91plZaW2bt2qjo4O0zXt7e36/e9/r+LiYkndiz0AAKEX7Lq+Fjc9+gA7BfxUFRQUqLCwUCUlJZo1a5amT58ul8ul0tJSNTQ0aObMmbrrrrtM15w4cULV1dXKyMgwHa+vr9eiRYuUlJSkCRMmKDMzU62trTpy5IgqKyvl8Xg0efJkXzDpdeTIET3wwANKT0/XhAkTNGjQIJ08eVL79u3TsWPH5HQ69eMf/1hf+9rX+vArAQBYoeu6vgtpMXygJ9MHWC+op2rJkiXKy8vT2rVrVVZWJrfbrZycHM2ePVuFhYUB74YxevRo/fCHP9TOnTtVXV2t3bt3q729XQMHDtSMGTN044036pZbblFMjHkh79ixYzVnzhxVVFTo448/1smTJ+VwODR06FBfM2emdgGgf0iLO9fbNZBdOVrZjQOwVdBP1axZszRr1qyAzl24cKEWLlzY7figQYN0//33B/utNWLECP3kJz8J+joAQOgFnekzFXJQvQtYjY1qAQC2SA1yTZ+pkINMH2A5gj4AgC2Mmb4vAsj00bIFsBdBHwDAFsbq3VOtF27ZQiEHYC+CPgCALYyZvtMBVe8yvQvYiaAPAGALc6Yv2OldCjkAqxH0AQBskRATK5ej85+ZVneHmjvazns+LVsAexH0AQBs4XA4gtqVw9iyhTV9gPUI+gAAtgmmV1+roZCDTB9gPYI+AIBtUoNY10emD7AXQR8AwDZpsYat2C6Y6aOQA7ATQR8AwDbBVPAaq3dp2QJYj6APAGCbYHr1GZszsyMHYD2CPgCAbYLK9NGcGbAVQR8AwDbm6t2et2LzeDw0ZwZsRtAHALBNoJm+NneHPGe/djmccjr45wmwGk8VAMA2qQFW75r23WU9H2ALgj4AgG2SXOe2YmvpaDet2zNqdVPEAdiNoA8AYBuHw6HUAHblaKGIA7AdQR8AwFaB7MphLuIg6APsQNAHALBVIPvvkukD7EfQBwCwVSAVvLRrAexH0AcAsJV5Vw7/vfpaDbtxkOkD7EHQBwCwVSBr+lpY0wfYjqAPAGCroNf0EfQBtiDoAwDYakCcoUFzT2v6DEFfHNO7gC0I+gAAtgqoTx+FHIDtCPoAALYKqHqXQg7AdgR9AABbJbni5HQ4JEnNHW1qM2y55kVzZsB+BH0AAFs5HQ5zMYefbB+FHID9CPoAALYzr+vr3quvlR05ANsR9AEAbHehdX0UcgD2I+gDANjOvCtH96DPWMhByxbAHgR9AADbXWhXDmOmjzV9gD0I+gAAtrvQrhw0ZwbsR9AHALBdMGv6yPQB9iDoAwDYLi323FZs/tf0GYM+CjkAOxD0AQBsZ8z0feEn09fqppADsBtBHwDAdubqXXOfPrfH7dulwyEp1kmmD7ADQR8AwHbJsfFyqnMrtjPt5q3YurZrcZzdsg2AtQj6AAC2czocSomN9/3ZuK6vhX13gZAg6AMAhERPFbwUcQChQdAHAAiJnnblMGX6KOIAbEPQBwAIiZ4zfYY1fUzvArYh6AMAhESqoVefcVeOFuP0Lpk+wDYEfQCAkOgx02cq5GBNH2AXgj4AQEj0uKaPTB8QEgR9AICQCCzTR9AH2IWgDwAQEqmGTJ9xTV/X5swA7EHQBwAICXOm79xWbMaWLfFk+gDbEPQBAEIiJTZe3g3WGttb1eF2S+qypo9CDsA2BH0AgJCIcTiV4meK17gjB9O7gH2Cfro2btyokpISVVVVye12a9SoUZo9e7YKCwvldAYeQ27YsEHvvPOO9u7dq7q6Op0+fVpJSUnKzc3VzTffrNtvv12xsbG2jwMAEDppsQm+yt3Tbc0aGJ/E3rtAiAT1dBUVFWndunWKj4/XtGnT5HK5VFpaqqVLl6q0tFQrVqwIOOAqKSlReXm5cnNzNWnSJKWmpurYsWMqLy/X+++/r40bN2r16tVKSkqydRwAgNBJi0vQkTOdX3sreI2FHLRsAewT8NP15ptvat26dcrIyNCaNWuUnZ0tSfr88881Z84cbdmyRcXFxbrnnnsCer/HHntM2dnZSktLMx2vra3Vvffeq/Lycr344ot68MEHbR0HACB0/FXw0rIFCI2A02ErV66UJC1atMgXaEnS4MGDtWTJEknSqlWr5D67MPdCJk+e3C3gk6ShQ4fqvvvukyS9++67to8DABA6/nr1UcgBhEZAQV9tba12796t2NhYFRQUdHs9Pz9fmZmZqqurU3l5eZ8H5XJ1ftLruqYv1OMAAFgrzZTp62zbYsz0Mb0L2CegoG/Pnj2SpNGjRyshIcHvOZMmTZIkVVZW9mlAx48f10svvSRJuu6668I2DgCA9YyZvtN+Mn1M7wL2CejpOnz4sCRp2LBhPZ6TlZVlOjdQb7/9tv7yl7+oo6NDdXV1ev/999XS0qLbbrtNd955Z8jGcT4NDQ3atWtXr67t7XWwH/emf+P+9G+9vT+fdTT6vj5SX6ddu3bpdNO5Y/srq1TnjOvz+KIZz07/Fs77E1DQd+ZMZ6lVYmJij+ckJydLkhobG3s8x5+9e/dq/fr1pmP33HOPFi5c2G16185xAADsl+g4t2bvjKczw9cmj++Yy9e+GYDVwp5HX7BggRYsWKDW1lYdPXpUmzZt0gsvvKC33npLL7zwgnJzc8M9RKWkpGjs2LFBXeON5PPy8uwYEvqAe9O/cX/6t77en5MtZ/R62SFJUrvLoby8PBW/e0A627XlqiuuVHIsmb7e4Nnp30Jxf6qqqtTQ0NDj6wGt6fP2ymtqaurxHG9mzZtpC1ZcXJyys7N1//3368knn9SRI0e0ePFieTznPgGGYhwAAPsYW7Y0tLWow+Pusvcu1buAXQIK+oYPHy5JOnr0aI/n1NbWms7tixtuuEEpKSnavXu3aW1eqMcBALBWjNOpZFe8JMkj6YuWJrnPfrh3OhxyOQn6ALsEFPRNmDBBkrR//341Nzf7PaeiokKSNH78+D4PyuFwKD09XVJnNW+4xgEAsN4AQwVvXfO5qSj23QXsFVDQl5WVpYkTJ6qtrU2bN2/u9npZWZlqa2uVkZGhqVOn9nlQNTU1OnLkiJxOp0aMGBG2cQAArGec4v3cEPTF064FsFXAO3LMnz9fkrRs2TIdOnTId7y+vl5FRUWSpHnz5pn2vF2zZo0KCgq0ePFi03t9/PHH2rhxo1paWrp9n3379umhhx6Sx+PR9ddfr0GDBvV5HACA/sPYq88U9DG1C9gq4I9VBQUFKiwsVElJiWbNmqXp06fL5XKptLRUDQ0Nmjlzpu666y7TNSdOnFB1dbUyMjJMx+vr67Vo0SIlJSVpwoQJyszMVGtrq44cOaLKykp5PB5NnjzZF8T1dRwAgP4jrYdMH42ZAXsF9YQtWbJEeXl5Wrt2rcrKyuR2u5WTk6PZs2ersLAw4Oza6NGj9cMf/lA7d+5UdXW1du/erfb2dg0cOFAzZszQjTfeqFtuuUUxPVRxWTUOAEDopfWwpo/pXcBeQT9hs2bN0qxZswI6d+HChVq4cGG344MGDdL9998f7Lfu9TgAAP2HeU3fuUb6FHIA9iIlBgAIKdP+u23nOjEwvQvYi6APABBSabH+t9KkkAOwF0EfACCkjJk+IzJ9gL0I+gAAIZUaG+/3eDxr+gBbEfQBAELK5YxRkiuu23EyfYC9CPoAACFn7NXnRfUuYC+CPgBAyPlb1xffQ29WANYg6AMAhJy/TB/NmQF7EfQBAELOX6aP6V3AXgR9AICQS/XTq49MH2Avgj4AQMiR6QNCj6APABByfqt3KeQAbEXQBwAIOf/Vu2T6ADsR9AEAQs5v9S7Tu4CtCPoAACGX6m9NH5k+wFYEfQCAkIt1xigxJtZ0jEIOwF4EfQCAsOi6ro8dOQB7EfQBAMIi1bCuz+VwyungnyTATjxhAICwMGb6qNwF7EfQBwAIC2MFL0UcgP0I+gAAYZEWd24rNoo4APsR9AEAwsKY6aOIA7AfQR8AICyMa/rI9AH2I+gDAITFZWkZvgKOsemZYR4NEPn4aAUACIuU2Hj9/MqbVdv0hcanDw33cICIR9AHAAibSxKSdUlCcriHAUQFpncBAACiAEEfAABAFCDoAwAAiAIEfQAAAFGAoA8AACAKEPQBAABEAYI+AACAKEDQBwAAEAUI+gAAAKIAQR8AAEAUIOgDAACIAgR9AAAAUYCgDwAAIAoQ9AEAAEQBgj4AAIAoQNAHAAAQBQj6AAAAogBBHwAAQBQg6AMAAIgCBH0AAABRgKAPAAAgChD0AQAARAGCPgAAgChA0AcAABAFCPoAAACigCvYCzZu3KiSkhJVVVXJ7XZr1KhRmj17tgoLC+V0BhZDut1ulZeXa9u2bXrvvff0ySef6MyZMxowYIAmTpyo22+/XTNnzvR77W9/+1s999xzPb53XFycKioqgv2xAAAAIlpQQV9RUZHWrVun+Ph4TZs2TS6XS6WlpVq6dKlKS0u1YsWKgAK/mpoaFRYWSpLS09M1efJkpaWlqaamRtu3b9f27dt122236Ze//KUcDoff9xg3bpzGjx/f/QdyBR3HAgAARLyAI6Q333xT69atU0ZGhtasWaPs7GxJ0ueff645c+Zoy5YtKi4u1j333HPB93I4HLr66qs1d+5cXXPNNYqJifG9VlZWpvvuu0+vv/66rrrqKs2ePdvve8ycOVMLFy4MdPgAAABRLeA1fStXrpQkLVq0yBfwSdLgwYO1ZMkSSdKqVavkdrsv+F4jR47UK6+8ohkzZpgCPknKz8/XvHnzJEkbNmwIdHgAAAA4j4CCvtraWu3evVuxsbEqKCjo9np+fr4yMzNVV1en8vLyPg9qwoQJvu8LAACAvgtoenfPnj2SpNGjRyshIcHvOZMmTdJnn32myspKXXnllX0a1MGDByVJQ4YM6fGc3bt366mnntKpU6c0YMAATZkyRV//+tcVFxfXp+8NAAAQiQIK+g4fPixJGjZsWI/nZGVlmc7traamJhUXF0uSbrjhhh7P27p1q7Zu3Wo6NnToUD311FPKz8/v0xi6amho0K5du3p1bW+vg/24N/0b96d/4/70X9yb/i2c9yeg6d0zZ85IkhITE3s8Jzk5WZLU2NjYpwEVFRXp8OHDys3N1e23397t9REjRujhhx/Wn/70J+3atUulpaV65ZVXlJ+fr9raWs2fP1979+7t0xgAAAAiTb/qb/L8889r/fr1Sk1N1bPPPut3qvab3/xmt2NXX321rr76aj344IN68803tXz5cl/hiRVSUlI0duzYoK7xRvJ5eXmWjQPW4N70b9yf/o37039xb/q3UNyfqqoqNTQ09Ph6QJm+pKQkSZ1Trz3xZvi8Gb9grV69WitWrFBSUpJWrVql0aNHB/0eCxYskCT93//9n9ra2no1DgAAgEgUUNA3fPhwSdLRo0d7PMdbaes9NxjFxcV68sknlZCQoJUrV2rq1KlBv4ck5eTkSJLa2tp04sSJXr0HAABAJAoo6PO2UNm/f7+am5v9nuPd+szfLhnns3btWv3iF79QfHy8fve73/WpCOPkyZO+r73ZSQAAAAQY9GVlZWnixIlqa2vT5s2bu71eVlam2tpaZWRkBJWlKykp0dKlSxUXF6fnn39e06dPD3zkfmzatEmSNGrUKKWkpPTpvQAAACJJwDtyzJ8/X5K0bNkyHTp0yHe8vr5eRUVFkqR58+aZ9t5ds2aNCgoKtHjx4m7v99prr6moqEhxcXF67rnn9LWvfe2CYzh69Kg2btyo1tZW03GPx6M33nhDzzzzjCTpu9/9bqA/FgAAQFQIuHq3oKBAhYWFKikp0axZszR9+nS5XC6VlpaqoaFBM2fO1F133WW65sSJE6qurlZGRobpeGVlpX72s5/J4/HoS1/6kjZt2uTL0hkNHDhQjzzyiO/PX3zxhRYtWqSf//znmjhxooYMGaLGxkbt37/f1x/wrrvu0ne+852gfgkAAACRLqiWLUuWLFFeXp7Wrl2rsrIyud1u5eTkaPbs2SosLDRl+c7n1KlT8ng8kqQDBw7owIEDfs8bPny4KegbOnSo5s6dq4qKCn366af68MMP5Xa7lZGRoZtuuknf/va3NW3atGB+JAAAgKgQdJ++WbNmadasWQGdu3DhQi1cuLDb8a985SuqqqoK9ltr4MCBfqeKAQAAcH4Br+kDAADAxYugDwAAIAoQ9AEAAEQBgj4AAIAoQNAHAAAQBQj6AAAAogBBHwAAQBQg6AMAAIgCBH0AAABRgKAPAAAgChD0AQAARAGCPgAAgChA0AcAABAFCPoAAACiAEEfAABAFCDoAwAAiAIEfQAAAFGAoA8AACAKEPQBAABEAYI+AACAKEDQBwAAEAUI+gAAAKIAQR8AAEAUIOgDAACIAgR9AAAAUYCgDwAAIAoQ9AEAAEQBgj4AAIAoQNAHAAAQBQj6AAAAogBBHwAAQBQg6AMAAIgCBH0AAABRgKAPAAAgChD0AQAARAGCPgAAgChA0AcAABAFCPoAAACiAEEfAABAFCDoAwAAiAIEfQAAAFGAoA8AACAKEPQBAABEAYI+AACAKEDQBwAAEAUI+gAAAKIAQR8AAEAUIOgDAACIAgR9AAAAUYCgDwAAIAoQ9AEAAEQBV7AXbNy4USUlJaqqqpLb7daoUaM0e/ZsFRYWyukMLIZ0u90qLy/Xtm3b9N577+mTTz7RmTNnNGDAAE2cOFG33367Zs6ced732L59u15++WV99NFHamlp0YgRI3TzzTdr7ty5iouLC/bHAgAAiGhBBX1FRUVat26d4uPjNW3aNLlcLpWWlmrp0qUqLS3VihUrAgr8ampqVFhYKElKT0/X5MmTlZaWppqaGm3fvl3bt2/Xbbfdpl/+8pdyOBzdrl+1apWWLVummJgY5efnKy0tTTt27NCzzz6rv/71r3r55ZeVmJgYzI8GAAAQ0QIO+t58802tW7dOGRkZWrNmjbKzsyVJn3/+uebMmaMtW7aouLhY99xzzwXfy+Fw6Oqrr9bcuXN1zTXXKCYmxvdaWavJxLsAAA7QSURBVFmZ7rvvPr3++uu66qqrNHv2bNO1FRUVevrpp5WYmKhXXnlFU6ZMkSQ1Njbqvvvu044dO7R8+XI9/vjjgf5oAAAAES/gNX0rV66UJC1atMgX8EnS4MGDtWTJEkmdGTi3233B9xo5cqReeeUVzZgxwxTwSVJ+fr7mzZsnSdqwYUO3a1etWiWPx6Pvfe97voBPkpKTk/WrX/1KTqdT69at06lTpwL90QAAACJeQEFfbW2tdu/erdjYWBUUFHR7PT8/X5mZmaqrq1N5eXmfBzVhwgTf9zVqbW3V9u3bJUm33HJLt+tGjBihK664Qm1tbdq2bVufxwEAABApAgr69uzZI0kaPXq0EhIS/J4zadIkSVJlZWWfB3Xw4EFJ0pAhQ0zHq6ur1dTUpPT0dI0cOfK84/COGQAAAAGu6Tt8+LAkadiwYT2ek5WVZTq3t5qamlRcXCxJuuGGG/yOw/u9/PGO8ciRI30ahyS1tLRIkhoaGrRr165evUdvr4P9uDf9G/enf+P+9F/cm/4tFPfHG790FVCm78yZM5J03orY5ORkSZ0FFX1RVFSkw4cPKzc3V7fffnvQ40hKSrJkHJLU0dHR5/cAAAAIpZ7il6D79Nnp+eef1/r165Wamqpnn3027P324uPj1dLSopiYGMXHx4d1LAAAAOfT0tKijo6OHmOWgII+b/asqampx3O8mTVvxi9Yq1ev1ooVK5SUlKRVq1Zp9OjRvRqHNxvY23EYeQtKAAAALnYBTe8OHz5cknT06NEez/FW2nrPDUZxcbGefPJJJSQkaOXKlZo6dep5x/GPf/yjx/fyvtabcQAAAESqgII+b8Zr//79am5u9ntORUWFJGn8+PFBDWDt2rX6xS9+ofj4eP3ud79Tfn5+j+fm5OQoISFBJ0+e1Keffur3nA8//LBX4wAAAIhkAQV9WVlZmjhxotra2rR58+Zur5eVlam2tlYZGRk9Zun8KSkp0dKlSxUXF6fnn39e06dPP+/5cXFxmjFjhiT/jZtrampUXl6u2NhYXXvttQGPAwAAINIFvCPH/PnzJUnLli3ToUOHfMfr6+tVVFQkSZo3b55p7901a9aooKBAixcv7vZ+r732moqKihQXF6fnnntOX/va1wIax7x58+RwOPTiiy/6snpS55rCxx9/XG63W3fccYfS0tIC/dEAAAAinsPj8XgCPXnJkiUqKSlRfHy8pk+fLpfLpdLSUjU0NGjmzJlasWKFaVu13/72t3ruueeUn5/v670ndTZwvvXWW+XxeJSTk2PaTs1o4MCBeuSRR7odX7VqlZYtW6aYmBhdffXVSk1N1Y4dO1RfX68pU6bolVdeOW9bFwAAgGgTVMuWJUuWKC8vT2vXrlVZWZncbrdycnI0e/ZsFRYWmrJ853Pq1Cl5Y80DBw7owIEDfs8bPny436Bv3rx5Gjt2rFavXq2Kigq1tLRoxIgRuvvuuzV37tywt3oBAADob4LK9AEAAODiFPCaPgAAAFy8CPoAAACiAEEfAABAFCDoAwAAiAIEfQAAAFGAoA8AACAKBNWnDxe2ceNGlZSUqKqqSm63W6NGjQq6jyF658CBA3rnnXdUUVGhjz76SAcPHpTH49FvfvMbFRQUnPda7pu92tratHPnTm3btk1lZWU6ePCgWltbNXDgQE2dOlV33nmnvvKVr/R4PffHXsXFxdq5c6f27dun48ePq6GhQampqRo3bpxuvfVW3XLLLXI4HN2uc7vdKikp0R//+EdVV1fL6XRq7NixuuOOO/SNb3wjDD9JdHjmmWe0cuVKSdLixYs1d+5cv+fx3Njv0Ucf1fr163t8fdSoUX63rw3Xs0OfPgsVFRVp3bp1io+P17Rp03w7ljQ2Nur666/XihUreNBs9MQTT+jVV1/tdvxCQR/3zX7vvvuu7r33XklSRkaGJk6cqMTERH3yySfat2+fJGnBggV66KGHul3L/bHfjBkzdPz4cY0ePVqZmZlKTEzU0aNH9cEHH8jj8ei6667Tc889Z/o9d3R06Ac/+IHefvttpaSkaNq0aWptbVVpaalaW1t1991366c//WkYf6rI9OGHH+o73/mO3G63PB5Pj0Efz01oeIO+K6+8Updeemm31zMyMvTwww+bjoX12fHAEps3b/aMGTPGc80113iqq6t9x+vq6jw33nijZ8yYMZ6XX345fAOMAq+99prn17/+tefPf/6z59ChQ5677rrLM2bMGM+mTZt6vIb7FhrvvvuuZ+HChZ4dO3Z0e+3Pf/6zZ/z48Z4xY8Z4SktLTa9xf0Jjx44dnsbGxm7H9+3b55k+fbpnzJgxnj/84Q+m11566SXPmDFjPDfddJOnrq7Od7y6utp3zZYtW2wfezRpaWnx3HjjjZ6vfvWrngULFnjGjBnjefHFF7udx3MTOo888ohnzJgxnj/+8Y8BXxPOZ4cw3yLeVPuiRYuUnZ3tOz548GAtWbJEUueewW63Owyjiw7f+ta3tHjxYt10000aOXJkQNdw30Jj2rRpWrFiha666qpur91000269dZbJUkbNmwwvcb9CY2rrrpKSUlJ3Y6PHj1ad9xxh6TObK1XR0eHXnzxRUmd23MOHjzY91p2drYWLVokSfrP//xPO4cddX7zm9/ok08+UVFRkVJTU3s8j+em/wr3s0PQZ4Ha2lrt3r1bsbGxfqcR8/PzlZmZqbq6OpWXl4dhhPCH+9Z/TJgwQZL02Wef+Y5xf/oHl6tz6bdxT/O///3vqq+v19ChQ/XlL3+52zUFBQWKjY1VRUWF6Z6i9z744AOtXr1a3/jGN/RP//RPPZ7Hc9O/hfvZIeizwJ49eyR1fipOSEjwe86kSZMkSZWVlSEbF86P+9Z/HDx4UFLn+hcv7k/41dTU6L/+678kyRRoeH/f3t9/V4mJicrNzTWdi95raWnRI488ogEDBugnP/nJec/luQmP9957T7/61a/0b//2b3r22Wf1zjvv+M2khvvZoXrXAocPH5YkDRs2rMdzsrKyTOci/Lhv/UNdXZ2v+u2GG27wHef+hN4f//hH7dixQ21tbfrss8/097//XW63W9///vd1/fXX+84L9N5UVlZybyywfPlyVVdXa/ny5Ro0aNB5z+W5CY833nij27Hc3Fw988wzGjt2rO9YuJ8dgj4LnDlzRlJnhN6T5ORkSVJjY2NIxoQL476FX3t7u3784x/r9OnTmjZtmimbxP0Jvffff9/UfsLlcumhhx7yVV57BXJvvGsEuTd98/777+uVV17RzJkzddNNN13wfJ6b0Bo3bpx++tOfavr06crKylJDQ4P27Nmj5cuXa+/evbr33nu1fv16ZWZmSgr/s8P0LoCw+fnPf67S0lJlZWXpqaeeCvdwot4TTzyhqqoqffDBB/rzn/+sOXPm6LnnntO3v/1t1uaFQXNzsx577DGlpKTo5z//ebiHAz+++93v6u6779Zll12mpKQkDRkyRNdee63+53/+R1dccYXq6+t9hTX9AUGfBbxReVNTU4/neCN27ycshB/3Lbx+8Ytf6A9/+IMyMjL08ssvm9bzSdyfcEpISFBubq4eeeQR/ehHP9LevXv17//+777XA7k33owG96b3nnnmGR08eFCPPvqohgwZEtA1PDf9Q1xcnObPny9J2rZtm+94uJ8dpnctMHz4cEnS0aNHezyntrbWdC7Cj/sWPk8++aSKi4s1aNAgvfzyy6a2El7cn/7h1ltv1a9//Wtt3bpVbW1tio2N5d6EyFtvvSWn06k33nij25qxAwcOSJJKSkr017/+VSNHjtQTTzzBvelHcnJyJJm7EoT7/hD0WcDbbmL//v1qbm72WzFVUVEhSRo/fnxIx4aecd/C4z/+4z+0evVqpaena/Xq1b5Kta64P/3DgAED5HK51N7eri+++EKDBw/23Rvv77+rpqYm7d+/X9K5+4jecbvdKisr6/H1mpoa1dTU6NSpU5J4bvqTkydPSjJn7ML97DC9a4GsrCxNnDhRbW1tfvfYKysrU21trTIyMjR16tQwjBD+cN9Cb9myZXrppZc0YMAArV69WuPGjevxXO5P/7Bjxw61t7crLS1NAwcOlCRNnTpVgwYNUm1trXbs2NHtms2bN6utrU2TJk3yLWBH8N5++21VVVX5/c/b0Hzx4sWqqqrSn/70J0k8N/3Jpk2bJEmXX36571i4nx2CPot45+6XLVumQ4cO+Y7X19erqKhIkjRv3jz2OuxnuG+hs3z5cq1atUppaWn6/e9/H9CnWO6P/Xbu3KmtW7eqvb2922u7du3y9YX753/+Z8XExEiSYmJi9L3vfU9S564C9fX1vmsOHjyop59+WpL0/e9/3+7hww+em9CorKzU1q1b1dHRYTre3t6u3//+9youLpbUWezhFe5nx+HxeDy2vHMUWrJkiUpKShQfH6/p06f7NrhuaGjQzJkztWLFCt9fmrDe7t27fX+hSdLHH3+sxsZGZWdna8CAAb7jr732muk67pv9/vd//1cLFiyQ1Pmpd/To0X7Py8nJ8f2D5cX9sdfrr7+uxx57TGlpaZowYYIGDx6sxsZG1dTU6OOPP5YkXXvttfrNb35jmirs6OjQAw88oK1bt/o2jW9vb9e7776rlpYWezeNhx599FGtX79eixcv1ty5c7u9znNjv7feeksPPPCA0tPTNWHCBA0aNEgnT57Uvn37dOzYMTmdTj388MO+IM8rnM8OQZ/FNm7cqLVr12rfvn1yu93KycnR7NmzVVhYyKcqm7333nuaM2fOBc+rqqrqdoz7Zi9vYHEh+fn5vk/HRtwf+9TU1Oj111/Xzp07VVNTo+PHj8vj8SgjI0OXX365brnlFs2cOdPvtW63W+vWrdPrr7+uAwcOyOl0auzYsbrjjjs0a9asEP8k0eVCQZ/Ec2O3mpoavfrqq6qoqNCRI0d08uRJORwODR06VHl5ebrzzjtNU7tG4Xp2CPoAAACiAKE+AABAFCDoAwAAiAIEfQAAAFGAoA8AACAKEPQBAABEAYI+AACAKEDQBwAAEAUI+gAAAKIAQR8AAEAUIOgDAACIAv8/j6T9/DSBz0gAAAAASUVORK5CYII=",
      "text/plain": [
       "<Figure size 720x720 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "r2_list = [1-new_trials.trials[i][\"result\"][\"loss\"] for i in range(len(new_trials.trials))]\n",
    "fig, ax = plt.subplots(1,1,figsize=(10,10))\n",
    "sns.lineplot(x=list(range(1,len(new_trials.trials)+1)), y=r2_list, ax=ax)\n",
    "ax.set_ylim(0.2,0.6)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 32,
   "metadata": {
    "id": "lhvjj4tx7LQB"
   },
   "outputs": [],
   "source": [
    "rfe_list = [new_trials.trials[i][\"result\"][\"rf_estimator\"] for i in range(len(new_trials.trials))]\n",
    "rfd_list = [new_trials.trials[i][\"result\"][\"rf_depth\"] for i in range(len(new_trials.trials))]\n",
    "rff_list = [new_trials.trials[i][\"result\"][\"rf_feature\"] for i in range(len(new_trials.trials))]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 33,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 1000
    },
    "id": "1ELOR0Ij7OVJ",
    "outputId": "403559d5-3b0d-471e-c3e9-a74f89f3d129"
   },
   "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>rf_estimators</th>\n",
       "      <th>rf_max_depth</th>\n",
       "      <th>rf_max_features</th>\n",
       "      <th>valid_r2_score</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>49</th>\n",
       "      <td>55</td>\n",
       "      <td>85</td>\n",
       "      <td>89</td>\n",
       "      <td>0.487916</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>21</td>\n",
       "      <td>77</td>\n",
       "      <td>72</td>\n",
       "      <td>0.484721</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>25</th>\n",
       "      <td>111</td>\n",
       "      <td>173</td>\n",
       "      <td>64</td>\n",
       "      <td>0.474472</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>13</th>\n",
       "      <td>174</td>\n",
       "      <td>172</td>\n",
       "      <td>142</td>\n",
       "      <td>0.473016</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>19</th>\n",
       "      <td>63</td>\n",
       "      <td>144</td>\n",
       "      <td>228</td>\n",
       "      <td>0.471687</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>11</th>\n",
       "      <td>250</td>\n",
       "      <td>93</td>\n",
       "      <td>183</td>\n",
       "      <td>0.466591</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>22</th>\n",
       "      <td>218</td>\n",
       "      <td>62</td>\n",
       "      <td>189</td>\n",
       "      <td>0.466535</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>172</td>\n",
       "      <td>32</td>\n",
       "      <td>115</td>\n",
       "      <td>0.465747</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>41</th>\n",
       "      <td>146</td>\n",
       "      <td>217</td>\n",
       "      <td>180</td>\n",
       "      <td>0.465680</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>254</td>\n",
       "      <td>85</td>\n",
       "      <td>237</td>\n",
       "      <td>0.464318</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>12</th>\n",
       "      <td>192</td>\n",
       "      <td>148</td>\n",
       "      <td>213</td>\n",
       "      <td>0.464080</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>45</th>\n",
       "      <td>164</td>\n",
       "      <td>125</td>\n",
       "      <td>101</td>\n",
       "      <td>0.463501</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>29</th>\n",
       "      <td>106</td>\n",
       "      <td>116</td>\n",
       "      <td>117</td>\n",
       "      <td>0.463074</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>47</th>\n",
       "      <td>143</td>\n",
       "      <td>45</td>\n",
       "      <td>141</td>\n",
       "      <td>0.461679</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>40</th>\n",
       "      <td>238</td>\n",
       "      <td>158</td>\n",
       "      <td>126</td>\n",
       "      <td>0.461305</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>179</td>\n",
       "      <td>87</td>\n",
       "      <td>109</td>\n",
       "      <td>0.461171</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>8</th>\n",
       "      <td>158</td>\n",
       "      <td>21</td>\n",
       "      <td>140</td>\n",
       "      <td>0.460954</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>21</th>\n",
       "      <td>127</td>\n",
       "      <td>112</td>\n",
       "      <td>80</td>\n",
       "      <td>0.459469</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>5</th>\n",
       "      <td>81</td>\n",
       "      <td>196</td>\n",
       "      <td>154</td>\n",
       "      <td>0.458671</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>32</th>\n",
       "      <td>134</td>\n",
       "      <td>161</td>\n",
       "      <td>95</td>\n",
       "      <td>0.456502</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>35</th>\n",
       "      <td>133</td>\n",
       "      <td>49</td>\n",
       "      <td>23</td>\n",
       "      <td>0.455567</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>43</th>\n",
       "      <td>172</td>\n",
       "      <td>185</td>\n",
       "      <td>167</td>\n",
       "      <td>0.453684</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>7</th>\n",
       "      <td>148</td>\n",
       "      <td>212</td>\n",
       "      <td>133</td>\n",
       "      <td>0.453292</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>17</th>\n",
       "      <td>155</td>\n",
       "      <td>238</td>\n",
       "      <td>153</td>\n",
       "      <td>0.452227</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>16</th>\n",
       "      <td>190</td>\n",
       "      <td>205</td>\n",
       "      <td>42</td>\n",
       "      <td>0.450085</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>33</th>\n",
       "      <td>69</td>\n",
       "      <td>76</td>\n",
       "      <td>69</td>\n",
       "      <td>0.448392</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>24</th>\n",
       "      <td>223</td>\n",
       "      <td>169</td>\n",
       "      <td>101</td>\n",
       "      <td>0.448004</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>10</th>\n",
       "      <td>52</td>\n",
       "      <td>34</td>\n",
       "      <td>156</td>\n",
       "      <td>0.447491</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>48</th>\n",
       "      <td>79</td>\n",
       "      <td>14</td>\n",
       "      <td>201</td>\n",
       "      <td>0.446191</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>42</th>\n",
       "      <td>202</td>\n",
       "      <td>24</td>\n",
       "      <td>57</td>\n",
       "      <td>0.444368</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>36</th>\n",
       "      <td>72</td>\n",
       "      <td>135</td>\n",
       "      <td>52</td>\n",
       "      <td>0.442924</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>14</th>\n",
       "      <td>81</td>\n",
       "      <td>70</td>\n",
       "      <td>249</td>\n",
       "      <td>0.441470</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>38</th>\n",
       "      <td>100</td>\n",
       "      <td>101</td>\n",
       "      <td>79</td>\n",
       "      <td>0.439863</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>28</th>\n",
       "      <td>39</td>\n",
       "      <td>221</td>\n",
       "      <td>69</td>\n",
       "      <td>0.439617</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>46</th>\n",
       "      <td>26</td>\n",
       "      <td>229</td>\n",
       "      <td>127</td>\n",
       "      <td>0.439338</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>9</th>\n",
       "      <td>218</td>\n",
       "      <td>37</td>\n",
       "      <td>37</td>\n",
       "      <td>0.439117</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>20</th>\n",
       "      <td>113</td>\n",
       "      <td>175</td>\n",
       "      <td>89</td>\n",
       "      <td>0.438140</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>23</th>\n",
       "      <td>22</td>\n",
       "      <td>254</td>\n",
       "      <td>81</td>\n",
       "      <td>0.437143</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>26</th>\n",
       "      <td>33</td>\n",
       "      <td>111</td>\n",
       "      <td>62</td>\n",
       "      <td>0.436685</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>44</th>\n",
       "      <td>51</td>\n",
       "      <td>136</td>\n",
       "      <td>43</td>\n",
       "      <td>0.432858</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>39</th>\n",
       "      <td>119</td>\n",
       "      <td>79</td>\n",
       "      <td>27</td>\n",
       "      <td>0.427659</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>6</th>\n",
       "      <td>56</td>\n",
       "      <td>147</td>\n",
       "      <td>35</td>\n",
       "      <td>0.424715</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>62</td>\n",
       "      <td>132</td>\n",
       "      <td>59</td>\n",
       "      <td>0.424169</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>34</th>\n",
       "      <td>43</td>\n",
       "      <td>98</td>\n",
       "      <td>117</td>\n",
       "      <td>0.423679</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>37</th>\n",
       "      <td>10</td>\n",
       "      <td>245</td>\n",
       "      <td>110</td>\n",
       "      <td>0.418470</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>27</th>\n",
       "      <td>110</td>\n",
       "      <td>55</td>\n",
       "      <td>12</td>\n",
       "      <td>0.415189</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>15</th>\n",
       "      <td>87</td>\n",
       "      <td>126</td>\n",
       "      <td>17</td>\n",
       "      <td>0.404370</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>18</th>\n",
       "      <td>20</td>\n",
       "      <td>199</td>\n",
       "      <td>33</td>\n",
       "      <td>0.381743</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>30</th>\n",
       "      <td>9</td>\n",
       "      <td>184</td>\n",
       "      <td>50</td>\n",
       "      <td>0.360571</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>31</th>\n",
       "      <td>94</td>\n",
       "      <td>13</td>\n",
       "      <td>8</td>\n",
       "      <td>0.276084</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "    rf_estimators  rf_max_depth  rf_max_features  valid_r2_score\n",
       "49             55            85               89        0.487916\n",
       "4              21            77               72        0.484721\n",
       "25            111           173               64        0.474472\n",
       "13            174           172              142        0.473016\n",
       "19             63           144              228        0.471687\n",
       "11            250            93              183        0.466591\n",
       "22            218            62              189        0.466535\n",
       "1             172            32              115        0.465747\n",
       "41            146           217              180        0.465680\n",
       "2             254            85              237        0.464318\n",
       "12            192           148              213        0.464080\n",
       "45            164           125              101        0.463501\n",
       "29            106           116              117        0.463074\n",
       "47            143            45              141        0.461679\n",
       "40            238           158              126        0.461305\n",
       "3             179            87              109        0.461171\n",
       "8             158            21              140        0.460954\n",
       "21            127           112               80        0.459469\n",
       "5              81           196              154        0.458671\n",
       "32            134           161               95        0.456502\n",
       "35            133            49               23        0.455567\n",
       "43            172           185              167        0.453684\n",
       "7             148           212              133        0.453292\n",
       "17            155           238              153        0.452227\n",
       "16            190           205               42        0.450085\n",
       "33             69            76               69        0.448392\n",
       "24            223           169              101        0.448004\n",
       "10             52            34              156        0.447491\n",
       "48             79            14              201        0.446191\n",
       "42            202            24               57        0.444368\n",
       "36             72           135               52        0.442924\n",
       "14             81            70              249        0.441470\n",
       "38            100           101               79        0.439863\n",
       "28             39           221               69        0.439617\n",
       "46             26           229              127        0.439338\n",
       "9             218            37               37        0.439117\n",
       "20            113           175               89        0.438140\n",
       "23             22           254               81        0.437143\n",
       "26             33           111               62        0.436685\n",
       "44             51           136               43        0.432858\n",
       "39            119            79               27        0.427659\n",
       "6              56           147               35        0.424715\n",
       "0              62           132               59        0.424169\n",
       "34             43            98              117        0.423679\n",
       "37             10           245              110        0.418470\n",
       "27            110            55               12        0.415189\n",
       "15             87           126               17        0.404370\n",
       "18             20           199               33        0.381743\n",
       "30              9           184               50        0.360571\n",
       "31             94            13                8        0.276084"
      ]
     },
     "execution_count": 33,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "perf_df = pd.DataFrame({\"rf_estimators\": rfe_list, \"rf_max_depth\": rfd_list, \"rf_max_features\": rff_list, \"valid_r2_score\": r2_list})\n",
    "perf_df.sort_values(by=\"valid_r2_score\", ascending=False)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 34,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/"
    },
    "id": "R2WuSZVVNoDs",
    "outputId": "23af8acc-c6ac-4dc5-d00c-2e42107056da"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Sat Sep 18 16:50:25 UTC 2021\n"
     ]
    }
   ],
   "source": [
    "!date"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 34,
   "metadata": {
    "id": "6eoKct36O9PL"
   },
   "outputs": [],
   "source": []
  }
 ],
 "metadata": {
  "accelerator": "GPU",
  "colab": {
   "collapsed_sections": [],
   "include_colab_link": true,
   "machine_shape": "hm",
   "name": "12_AMPL_HPO_demo.ipynb",
   "provenance": []
  },
  "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": "3.7.2"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 4
}