PyTorchGraphDatasetLoader.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 1,
   "id": "e29b4685-8254-4b37-9d40-eafc31239458",
   "metadata": {},
   "outputs": [],
   "source": [
    "import matplotlib.pyplot as plt\n",
    "from matplotlib.colors import LogNorm\n",
    "import mplhep as hep\n",
    "from coffea.util import load\n",
    "import numpy as np\n",
    "import pandas as pd\n",
    "import awkward as ak\n",
    "from coffea.nanoevents import NanoEventsFactory, NanoAODSchema\n",
    "from numba import njit\n",
    "import vector\n",
    "vector.register_numba()\n",
    "vector.register_awkward()\n",
    "\n",
    "from pprint import pprint \n",
    "hep.style.use(hep.style.ROOT)\n",
    "\n",
    "import hist"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "id": "f2c965e6-456a-4f32-b216-e55e6d874cab",
   "metadata": {},
   "outputs": [],
   "source": [
    "df = ak.from_parquet(\"jets_combinations_ttHbb_fullymatched_v2.parquet\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "id": "92432182-79e9-4a1f-945b-33d5e4be42c1",
   "metadata": {},
   "outputs": [],
   "source": [
    "(jets_fullymatched,  # initial collections of jets, already fully matched\n",
    "comb_index,          # indices of jets with all the combinations for all the events\n",
    "signal_comb,         # mask to get the \"signal\" combination  [events, True/False]\n",
    "jets_combinations,   # collections of jets, many different collections for each event (from comb_index)\n",
    "higgs_jets,          # higgs_jets (index 0, 1 of the combination)\n",
    "tlep_jets,           # tlep jet (index 2 of the combination)\n",
    "thad_jets,           # thad jets (index 3,4,5 of the combination)\n",
    "whad_jets            # whad_jets (index 4 5 of the combination)\n",
    ") = ak.unzip(df)\n",
    "\n",
    "jets_combinations = ak.with_name(jets_combinations, name=\"Momentum4D\")\n",
    "higgs_jets = ak.with_name(higgs_jets, name=\"Momentum4D\")\n",
    "tlep_jets = ak.with_name(tlep_jets, name=\"Momentum4D\")\n",
    "thad_jets = ak.with_name(thad_jets, name=\"Momentum4D\")\n",
    "whad_jets = ak.with_name(whad_jets, name=\"Momentum4D\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "id": "7d40aeaa-7972-4f11-bc86-a6350afafaf5",
   "metadata": {},
   "outputs": [],
   "source": [
    "jets_signal = ak.flatten(jets_combinations[signal_comb], axis=1)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 27,
   "id": "65d3f34b-61e3-49c2-ad06-17d988d627fc",
   "metadata": {},
   "outputs": [],
   "source": [
    "def to_flat_numpy(X, fields, axis=1, allow_missing=False):\n",
    "    return np.stack([ak.to_numpy(X[f], allow_missing=allow_missing) for f in fields], axis=axis)\n",
    "\n",
    "def to_flat_tensor(X, fields, axis=1, allow_missing=False):\n",
    "    return torch.tensor(np.stack([ak.to_numpy(X[f], allow_missing=allow_missing) for f in fields], axis=axis))\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 28,
   "id": "e3242085-a30b-4d03-939a-496df65adc68",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "array([[ 4.25625000e+01,  7.23876953e-01,  3.02441406e+00,\n",
       "         9.98046875e-01],\n",
       "       [ 1.26000000e+02,  2.36523438e+00, -2.85644531e+00,\n",
       "         9.96093750e-01],\n",
       "       [ 8.43750000e+01,  2.15576172e+00,  1.50146484e+00,\n",
       "         1.91406250e-01],\n",
       "       [ 1.02437500e+02,  1.11718750e+00, -8.15795898e-01,\n",
       "         9.99511719e-01],\n",
       "       [ 1.41500000e+02,  1.50024414e+00,  1.18139648e+00,\n",
       "         6.28280640e-03],\n",
       "       [ 1.23812500e+02,  1.67382812e+00,  4.42382812e-01,\n",
       "         6.17599487e-03]])"
      ]
     },
     "execution_count": 28,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "to_flat_numpy(jets_signal[0], [\"pt\",\"eta\",\"phi\", \"btag\"])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 31,
   "id": "c7c7dc81-4fa6-474b-87da-7b9cf31a6806",
   "metadata": {},
   "outputs": [],
   "source": [
    "t = to_flat_tensor(jets_signal[0], [\"pt\",\"eta\",\"phi\", \"btag\"])"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "8cdfb935-9220-45ad-8158-64653564d6ed",
   "metadata": {},
   "source": [
    "# PyTorch geometric dataset"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 53,
   "id": "3a04380f-6e6a-438d-93b7-d4bad86de92b",
   "metadata": {},
   "outputs": [],
   "source": [
    "import os\n",
    "\n",
    "import numpy as np\n",
    "import uproot\n",
    "import torch\n",
    "from torch_geometric.data import InMemoryDataset, Data\n",
    "from torch_geometric.transforms import NormalizeFeatures\n",
    "import torch.nn.functional as F\n",
    "import awkward as ak\n",
    "from tqdm import tqdm"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 96,
   "id": "429e9e2b-409d-4f83-bda1-176513470d0d",
   "metadata": {},
   "outputs": [],
   "source": [
    "class EventTaggingDataset(InMemoryDataset):\n",
    "    def __init__(self, root, transform=None, pre_transform=None, pre_filter=None):\n",
    "        \n",
    "        \n",
    "        super().__init__(root, transform, pre_transform, pre_filter)\n",
    "        self.data, self.slices = torch.load(self.processed_paths[0])\n",
    "\n",
    "    @property\n",
    "    def raw_file_names(self):\n",
    "        return [self.root + '/jets_combinations_ttHbb_fullymatched_v2.parquet']\n",
    "\n",
    "    @property\n",
    "    def processed_file_names(self):\n",
    "        return [self.root + '/jets_combinations_ttHbb_fullymatched_v2.pt']\n",
    "\n",
    "\n",
    "    def process(self):\n",
    "        # Read data into huge `Data` list.\n",
    "        data_list = []\n",
    "        fields = [\"pt\",\"eta\",\"phi\",\"btag\"]\n",
    "        for file in self.raw_file_names:\n",
    "            df = ak.from_parquet(file)\n",
    "            jets = df.jets_fullymatched\n",
    "            \n",
    "            for i_ev in tqdm(range(len(jets))):\n",
    "                njets = len(jets[i_ev])\n",
    "                nodes_feat = to_flat_tensor(jets[i_ev], fields)\n",
    "                E =  np.ones((njets, njets))\n",
    "                np.fill_diagonal(E, 0)\n",
    "                edge_index = torch.tensor(np.stack(np.where(E)))\n",
    "                prov = ak.to_numpy(jets[i_ev].prov) -1\n",
    "                np.place(prov, prov==-2, 4)\n",
    "                y = F.one_hot(torch.tensor(prov, dtype=int), num_classes=5)\n",
    "                \n",
    "                data_list.append(\n",
    "                    Data(\n",
    "                        x=nodes_feat,\n",
    "                        edge_index=edge_index,\n",
    "                        num_nodes=njets,\n",
    "                        num_node_features=nodes_feat.shape[1],\n",
    "                        y=y\n",
    "                    )\n",
    "                )\n",
    "        \n",
    "        if self.pre_filter is not None:\n",
    "            data_list = [data for data in data_list if self.pre_filter(data)]\n",
    "\n",
    "        if self.pre_transform is not None:\n",
    "            data_list = [self.pre_transform(data) for data in data_list]\n",
    "\n",
    "        data, slices = self.collate(data_list)\n",
    "        torch.save((data, slices), self.processed_paths[0])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 97,
   "id": "94413fda-ea68-432d-b095-79a05c500186",
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Processing...\n",
      "100%|████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████| 67876/67876 [02:40<00:00, 423.21it/s]\n",
      "Done!\n"
     ]
    }
   ],
   "source": [
    "data = EventTaggingDataset(\".\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 102,
   "id": "45a3e0c3-c28d-403f-a81e-077e08c7feb0",
   "metadata": {},
   "outputs": [],
   "source": [
    "from torch_geometric.loader import DataLoader\n",
    "training_df = DataLoader(data, batch_size=100, shuffle=True)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 105,
   "id": "385a5f0b-3467-4af4-a454-7f37ccb70973",
   "metadata": {},
   "outputs": [],
   "source": [
    "for df in training_df:\n",
    "    break"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 117,
   "id": "1ccfb58c-c04a-4859-871c-507a3d3a90aa",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "<matplotlib.collections.PathCollection at 0x7fb0fefc44f0>"
      ]
     },
     "execution_count": 117,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAA2cAAANBCAYAAABkvsZOAAAAOXRFWHRTb2Z0d2FyZQBNYXRwbG90bGliIHZlcnNpb24zLjYuMiwgaHR0cHM6Ly9tYXRwbG90bGliLm9yZy8o6BhiAAAACXBIWXMAAA9hAAAPYQGoP6dpAACJQElEQVR4nOzde3xU1b3//3cSCIaESxJIIJkQIhQkGDncTFALVtuvXAwX9RRFT209Pa3tKF6o9mCLSKViW6lSHVutFj1KbfsTi0UQ9Yg99CJEgxg0igIKZAIDJRckJASS/fsjzGYmmUkyyZ5b5vV8PHg8dmavtWZNlsF82Hu/V5xhGIYAAAAAAGEVH+4JAAAAAAAozgAAAAAgIlCcAQAAAEAEoDgDAAAAgAhAcQYAAAAAEYDiDAAAAAAiAMUZAAAAAEQAijMAAAAAiAAUZwAAAAAQAXqFewJOp1P79u2TJOXm5io7Ozug/s3NzdqzZ4+cTqdSUlI0YsQIpaamBmOqAAAAABA0ARVnO3fulNPpDPhNsrOzVVBQ4PXaq6++qnvvvVfvvvuu1+sTJ07UT37yE82cObPdMZubm7Vq1So98sgj2r9/v/l6QkKCpk+frpUrV2r06NEBzxUAAAAAwiHOMAyjs42/+c1v6tlnnw34TW688UY988wz5tcrV67UD37wg3b7/OIXv/DbpqmpSddcc43WrVvnt39SUpI2bdqkqVOnBjxfAAAAAAi1kDxzlpycbB5v3rxZd911lyRp0KBBWrNmjaqqqlRVVaXnn39e6enpkqS77rpLb775ps/x7r//frMwmzBhgt566y0dP35cBw4c0LJlyxQfH6/6+nrNmzdPR44cCe6HAwAAAAALBHTlLBAPPvigFi9erMzMTO3YsUNDhgyRYRgqKipSSUmJevXqpZKSEo0fP96r3/bt21VYWKjTp0/rwgsv1NatWxUXF2eeP3z4sM4991zV1dUpLy9P77//vvr16+c1xqOPPqqFCxdKku6++2797Gc/C8ZHBAAAAADLBOXK2TvvvKMlS5YoLi5Ov//97zVkyBBJUllZmUpKSiRJc+fObVOYSS1XwubOnStJKikp0Ycffuh1/oUXXlBdXZ0kadGiRW0KM0my2+3KyMiQJK1evVrNzc2WfTYAAAAACAbL0xqPHz+uBQsW6PTp01q6dKkuu+wy89wrr7xiHs+ZM8fvGHPmzNGLL74oSdqwYYPOP/98n2PMnj3bZ//4+HgVFxfr6aef1pEjR/TOO++osLDQZ9vk5GQ1NDQoISHBLOgAAAAAxJ7Dhw+rqalJ55xzjnlBKJQsL84WLlyo3bt3a9q0aVqyZInXuV27dpnH7aUxep7bvn27zzEKCgqUk5Pjd4xZs2bp6aefNsfwV5w1NDSoublZzc3NXUqiBAAAANCzNDQ0hOV9LS3O3njjDa1evVoJCQn69a9/rYSEBK/zBw8elCSlpKQoLS3N7zhpaWlKTk5WXV2dV0FnGIYOHTokqWVPtPZ4Fm6eY7SWkJCg5uZmxcfHa+jQoe2O2ZrL5VJmZmZAfdpjGIYqKyuVlZXl9Zxdd1k9z1geM1hrJEXH54+WMflZivwxWaPIH5M1ivwxWaPIH5M1ivwxW6/RwYMH1dzc3KaOCRnDIqdPnzYKCgoMScb3vvc9n23OP/98Q5IxbNiwDscbNmyYIcnIyMgwX/vXv/5lSDIkGd/4xjfa7b93716z7de//nW/7bKzsw1JRlZWllFbWxvQn1GjRpnHDQ0NHX6mjtTW1hqSjNra2m6P5WnMmDGWjhfLYwZrjQwjOj5/tIzJz1Lkj8kaRf6YrFHkj8kaRf6YrFHox2xoaAjo9/kDBw4YkowDBw4YtbW1RlZWliHJyM7OtnzunWHZlbPVq1dr586d6tevn+677z6fbdxXztxx+e1JT0/X/v37ve71dPfvzBie5ztzv2hlZaUGDBjQYbvW3H2WLl3q93MDAAAACL4VK1Zo2bJlAfdr73GpULKkODt+/Lj5fNnixYv9BmscO3ZMUssG0R3p06ePJKm+vr5N/86M4e7fegx/srKy9NFHH3XYztPkyZP1zjvvtHk/AAAAAKG3ePFi3XnnnZ1uf+zYMeXk5OjAgQPq37+/xowZo8rKyiDOsH2WFGe//OUvdejQIaWmpur222/3227w4MGqrKxUTU1Nh2O623gWYYMHD25zvqP+rcfwJy4uTv379++wnaeEhISA+wAAAAAIjj59+nTpokn//v3Vv39/yzMFAtXtfc4aGxv1+OOPS5K+8Y1vtFsIuQM3qqqqOhzX3SYlJaVN/86M4XnecwwAAAAAiETdLs7Wrl0rl8slSfr2t7/dblv3ZtRVVVUyDMNvO8MwVF1dLcn7/s+UlBQlJyebY7TH3b/1GAAAAAAQibpdnD322GOSpKKiIq/Non1xX/lqbGxsN95+165dOnXqlCQpPz/f5xhlZWVqbm72O0ZZWZl53HoMq9jt9qCMa7VgzDOWxwyWaPn80TJmMETLZ4+WMYMhWj57tIwZDNHy2aNlzGCIls8eLWMGQ7R89mgZM6J0J+rxvffeM+Pqn3766Q7b/+Y3vzHb//znP/fb7uc//7nZ7je/+Y3XuWuvvdY8V1JS4neMmTNnmu0+/vhjv+3cUfrhisv0FMyYdliDNYoOrFPkY40iH2sU+VijyMcaRb7WaxTu2qBbV86effZZ83jOnDkdti8uLjYfslu7dq3PWxsNw9BLL70kqSWkY/bs2V7nPd9n7dq1Pt+npqZGmzdvliSNGTNGo0eP7nBuAAAAABBO3SrONm7cKEkqKCjo1N5lWVlZuvbaayVJ27Zt05NPPtmmzRNPPKGtW7dKkq6//nqvEBBJmjdvnnJzcyVJq1at0o4dO7zONzU1yW63q6GhQZK0aNGiwD4UAAAAAIRBl6P0d+/erU8++USSNHXq1E73e+CBB/Taa6+pqqpKN998s95++23NmDFDzc3N2rBhg9asWSOpZRPp5cuXt+nfp08frVq1SnPnzlVDQ4OmTp2qW2+9VRMnTlRVVZWeeeYZ/eMf/5DU8hzcjTfe2NWPCAAAAAAh0+Xi7NVXXzWPp02b1ul+w4cP18svv6yrrrpKR44c0bPPPut1e6QkZWRkaN26deYVstbmzJmjRx99VHfccYe++OILPfDAA23aTJw4US+//LJ69bJkKzcAAAAACKou39bovqVRkr785S8H1PeSSy7Rzp07dc899yg/P1/JyclKSUnR2LFj9aMf/UgffPCBpkyZ0u4Yt9xyi959913ddNNNys3NVWJiotLS0nTRRRfpscce09tvv62MjIwufTYAAAAACDVLrpx1RWZmpn7605/qpz/9aZfHGDdunJ5++uluzQMAAAAAIkG39zkDAAAAAHQfxVkE6dOnj5YuXao+ffqEeyrwgzWKDqxT5GONIh9rFPlYo8jHGkW+SFujOMPXZmMxxGazyel0Kjs7WxUVFeGeDgAAAIAwCXdtwJUzAAAAAIgAFGcAAAAAEAEozgAAAAAgAlCcAQAAAEAEoDgDAAAAgAjQ5U2oexqXy6X8/Hyf5+x2u+x2e4hnBAAAAMBqDodDDofD5zmXyxXi2XgjSp8ofQAAAAAKf23AbY0AAAAAEAEozgAAAAAgAlCcAQAAAEAEoDgDAAAAgAhAcQYAAAAAEYDiDAAAAAAiAPucAQAAAAi6g7X1evfzKtXUn/J6PbVvoibmpmrogKQwzSxyUJwBAAAACKo/vrNf/712p9rbYHnxjPP03WkjQjanSMRtjQAAAACC5mBtvX7YQWEmSSte/VhPbNkTkjlFKoozAAAAAEFTuq+6020f3PixDtbWB3E2kY3iDAAAAEDQVJ9o7HRbQ9L2AIq5nobiDAAAAEDECKSY62kozgAAAAAgApDWCAAAAKDb/EXl76yoDdOMog/FGQAAAIBu6UxUPjpGcQYAAACgy9xR+VYZmJRo2VjRhmfOAAAAAHRZIFH5HYmTNHF4qmXjRRuKMwAAAABdZmW64oNXF2jogCTLxos23NYIAAAAIKSKxw1RYV66+XVq30RNyE2N6cJMojgzuVwu5efn+zxnt9tlt9tDPCMAAACgZyrMS9cNRcPD8t4Oh0MOh8PnOZfLFeLZeKM4OyMzM1Pl5eXhngYAAAAQ0VpH5m/bezTMMwpMexdebDabnE5niGd0FsUZAAAAgE4hMj+4KM4AAAAAdMjKyPxYjstvD2mNAAAAADpkVWR+rMflt4fiDAAAAECHrIjMjxdx+e3htkYAAAAAlvKMyk/tmyhbapJONDZr+KC+FGbtoDgDAAAAYKlwRuVHM4ozAAAAAF5ax+VL0s6K2k73J/CjayjOAAAAAJisiMvPSePWxa6gOAMAAAAgybq4/BONzRbMJvaQ1ggAAABAkjVx+fFx0vBBfS2YTeyhOAMAAAAgqftx+XFx0oqriMrvKm5rBAAAANAlrSPzJ+SmUph1A8UZAAAAgC4hMt9aFGcAAABADPEVk+8WSFw+rEdxBgAAAMQIK2LyETwUZwAAAEAMsCom3xObTVuLtEYAAAAgBlgRk+8pTtLE4amWjhnrKM4AAACAGNDdmPzWHryayHyrcVsjAAAAAJ/mT7KpwDbA6zUi84OH4uwMl8ul/Px8n+fsdrvsdnuIZwQAAACEV4FtQI+Lync4HHI4HD7PuVyuEM/GG8XZGZmZmSovLw/3NAAAAAAEUXsXXmw2m5xOZ4hndBbFGQAAANCD+NvHbNveo2GaETqL4gwAAADoIdjHLLpRnAEAAAA9APuYRT+i9AEAAIAegH3Moh/FGQAAANADsI9Z9OO2RgAAACBGFY8bosK8dK/X2McsfCjOAAAAgBhVmJfe4/Yxi2bc1ggAAAAAEYDiDAAAAOgBBib17kIf0hgjCbc1AgAAAFHGvdF0XFycclKTtL/qhD6orA14nJw0niuLJBRnAAAAQBSxcqPpiup6jcshLj9SUJwBAAAAUcLqjaYNKyo8WIZnzgAAAIAoYeVG02wyHXkozgAAAIAoYeVG02wyHXm4rREAAADoodhkOrpQnAEAAAARyp3KWFN/SpK0be/RgPqzyXR0oTgDAAAAIpCVqYyIDhRnAAAAQISxKpWRTaajC4EgAAAAQISxIpWRNMbow5WzM1wul/Lz832es9vtstvtIZ4RAAAAYpUVqYykMfrmcDjkcDh8nnO5XCGejTeKszMyMzNVXl4e7mkAAAAAAbl5Wp7OzxqomvqWgo40xva1d+HFZrPJ6XSGeEZnUZwBAAAAYdadVEZbal9dOS4rWFNDCFGcAQAAAGFEKiPcKM4AAACAMLEilZFExp6DtEYAAAAgTLqbykgiY89CcQYAAACESXdTGUlk7Fm4rREAAACIAsXjhqgwL10SiYw9FcUZAAAAEGTuNMa4uDjlpCZpf9UJ1dSfCiiVsTAvXTcUDQ/eJBF2FGcAAABAEJHGiM7qdnFmGIY+//xzVVRUaODAgRo1apT69OnT6b779u3TgQMHlJiYqLy8PGVkZAT0/qdOndLu3bt16NAhpaamasSIEerXr19XPgoAAABgKSvSGN1IZez5uhwI0tTUpFWrVmnUqFE699xzNXXqVF1wwQVKTk7W1Vdfrf379/vtaxiGnnvuOeXn5ysvL09Tp05VUVGRMjMzNW3aNG3btq3D9z958qSWLFmi7Oxs5efn67LLLtP48eOVkZGhBQsWqLKysqsfDQAAALBEd9MY3UhljA1dKs4aGho0Y8YM3X777dq9e7fXuaamJr300ksaM2aMtm/f3qavYRi67bbb9I1vfEMff/xxm/NbtmzRxRdfrD/84Q9+37+urk7Tpk3T8uXLdeTIkTZze+GFF3TBBRf4HB8AAAAIle6mMbqRyhgbulScfec739Ebb7whSSouLta2bdt0/Phxffrpp1q0aJHi4uJ04sQJzZ8/X1988YVX32effVaPPvqoJCkvL0/r16/XsWPH5HK59OijjyopKUlNTU365je/qY8++sjn+9vtdvPq2le/+lWVlJToxIkT2r17t2655RZJ0tGjR1VcXKyTJ0925SMCAAAA3TYwqXfAfeZPsmn53LFaPnesHAvG6+3Fl2n+5GFBmB0iTZxhGAE9m/j3v/9dX/7ylyVJCxYs0HPPPaf4eO8ab8mSJVq+fLkk6cknn9R//dd/SWq5FXHkyJGqqKjQgAED9MEHH8hms3n1Xb9+vWbPni1J+vrXv64//vGPXuc/+OADXXDBBTIMQ0VFRdqyZYt69/b+j37RokX65S9/KUl6/PHH9b3vfc/v57HZbHI6ncrOzlZFRUUg3woAAACgXe8fqNYcxz8D6vOy/SKNy+EWxnAId20Q8JWzFStWSJJ69eqlRx55pE1hJkl33XWXzjnnHEnSX/7yF/P1TZs2mR/yv/7rv9oUZlLLlbhJkyZJktauXauamhqv808//bTc9eSPf/zjNoWZ1FIc9urVy2wPAAAAhMOB6vqA+1R0oQ96hoDSGo8dO2bezjh9+nQNHjzYZ7v+/ftr27ZtOnHihBITz6bKvPLKK+bxnDlz/L7PnDlz9O6776qpqUmvvfaa5s+f32aM5ORkXX755T77Dxw4UNOmTdObb76p0tJSHTx4UEOHDu38BwUAAAAC5N7LrKb+lPlaIPuYuVn1nBqiT0DF2ZYtW3TqVMt/bF//+tfbbXvBBRe0eW3Xrl2SWoqnKVOm+O07a9YsLVmyRJK0fft2szg7deqU9uzZI0m67LLLzKtz/sZ48803JUk7duygOAMAAEDQsJcZrBDQbY0ffviheTxy5EhJLemLn332mbZs2aIPP/zQLN58OXjwoKSWezkTEhL8tsvJyTGP3QWdJB0+fNi8pTE3N7fdufobAwAAALCSey8zqwoz9jOLXQEVZ55FTkZGhl588UXl5+fr3HPP1bRp03T++ecrJSVF06ZN03vvvdem/6FDhyRJaWlp7b5Pamqq4uLiJJ0t6Dz7d2aM9PR089hzDAAAAMBKVu1lJrGfWawL6LbG2tpa8/jXv/61Vq5c2aZNY2OjtmzZosmTJ2vZsmX60Y9+JEk6fvy4jh8/Lsm7cPIlISFBAwcOVHV1terq6szXPYusjsbwPO85hj+GYejYsWMdtvOnT58+6tOnT5f7AwAAIDpZ+YwY+5l1z8mTJ7u1lVaAQfaWC6g489yzbOXKlUpOTtaSJUs0c+ZM5eXl6bPPPtOf//xn/fSnP1VjY6N+/OMfa/To0brmmmu8Cp+kpI7/g3MXOvX1Z9NqAhnDs1DyHMOfyspKDRgwoMN2/ixdulT33Xdfl/sDAAAgNtw8LU/nZw1UTf3Zoi61b6Im5KZSmHXTihUrtGzZsnBPo8sCKs7c8fRSS/Hz9ttvq6CgwHytoKBABQUFuvDCCzVjxgxJ0j333KOrr75agwYNMtu1jsf3xd3GswjzTIfsaAzP850pBrOysvxuet0ZXDUDAABAZ9hS++rKcVnhnkaPtHjxYt15551d7j9mzBhVVlZaOKPABFScZWWd/Y/ojjvu8CrMPE2fPl1f+9rX9MYbb+jTTz9VeXm5xo4dq/T0dB09elRVVVXtvk99fb0aGhokSSkpKebrnomLHY3hed5zDH/i4uLUv3//DtsBAAAgtrWOzO9KXD6Co7uPGrlzL8IloOLMsziaOnVqu20LCwvNPdE++eQTjR07VkOGDOlUcVZdffahSs/UxSFDhpjHXR0DAAAA6Coi8xFMAaU1el456ygtMSMjwzxubGy5n9Zd3H322Wfthm+UlZWZx/n5+eZxenq6evfuLUl6//33231/f2MAAAAAXWFVZD5R+fAnoOLsS1/6knnsueeZLx9//LF5nJ2dLUmaNGmSpJbNpF977TW/fdevX28eX3TRReZxXFycJk6cKEkqKSmRy+XqcIzExESzDwAAANBVVkTmE5WP9gRUnE2dOtV8Luvpp59Wc3Ozz3ZffPGF/vKXv0iSkpOTNXnyZEnSnDlzzDZr16712ffUqVNmYTVgwAB95Stf8TrvHsMwDP35z3/2OcauXbtUXl4uSbr88st5lgwAAADdZkVkPlH5aE9AxVliYqK++93vSpL++c9/asmSJWpqavJq09DQILvdroqKCknSd77zHfOhvMLCQk2ZMkWS9Mc//lGbNm3y6msYhpYsWaIDBw5Ikm699VYlJnpf9v32t79tBnwsXbpU+/fv9zpfX1+vm2++2fx60aJFgXxEAAAAoNuKxw3R8rljzT+OBeP19uLLNH/ysHBPDREszghwp7Wamhr927/9m/bt2yeppeC66qqrdO655+qTTz7R888/b0bSjx49Wlu3btXAgQPN/u+8846+/OUv6+TJk+rdu7e+973v6ctf/rJOnDihP/3pT9qwYYMkacSIESotLfW599ijjz6qhQsXSmp5tu22225Tfn6+Kioq9MQTT+iDDz6QJF1zzTX6//6//6/dz2Oz2eR0OpWdnW0WlAAAAIhd7jTGuLg45aQmaX/VCdXUn9K2vUf1StmhTo2xfO5Y3VA0PLgTheXCXRsEXJxJLbcNzpw5U3v37vXb5sILL9TatWtls9nanFu7dq1uvPFG1dXV+ew7YsQIbdy4UaNGjfJ53jAM3XPPPXrwwQf9vv8VV1yhl156SX379m33s4R7AQAAABA5rEpjpDiLTuGuDQK6rdFt9OjR2rlzp1auXKkpU6YoPT1dvXr1UmZmpmbNmqU1a9bon//8p8/CTJKuvvpqlZWVaeHChRo5cqTOOeccDRgwQBMmTNCKFSu0Y8cOv4WZ1BIMsmLFCv3f//2f5s+fL5vNpt69e2vw4MG6/PLL9fzzz2vjxo0dFmYAAACAm1VpjBKJjOiaLl0560nCXR0DAAAgMrxSVqlbfv9et8eJk/TPxZcR/BGFwl0bdOnKGQAAANDTWJHGKJHIiK7rFe4JAAAAANGqeNwQFealS5JS+yZqQm4qhRm6jOIMAAAA6KLCvHSCP2AZbmsEAAAAJA1M6t2FPgR/wDpcOQMAAEBMcu9nVlN/SpJUUX0i4DFy0riFEdahOAMAAEDMsWo/s4rqeo3LSbVkTgDFGQAAAGKKez8zK8T2plSwGs+cAQAAIKaU7qu2ZJw4SROHc9UM1qE4AwAAQExhPzNEKm5rPMPlcik/P9/nObvdLrvdHuIZAQAAINxunpan87MGqqb+bEHHfmbRzeFwyOFw+DzncrlCPBtvFGdnZGZmqry8PNzTAAAAgIVaJzJK0ra9Rzvd35baV1eOywrG1BAm7V14sdlscjqdIZ7RWRRnAAAA6JGsSmQEQoXiDAAAAD2OVYmMbDKNUCIQBAAAAD2OFYmMpDEi1CjOAAAA0ONYkchIGiNCjdsaAQAAEPOKxw1RYV66JNIYET4UZwAAAIh5hXnpuqFoeLingRhHcQYAAICo1DomP7VvonJSk7S/6kRAcflApKA4AwAAQNQhJh89EcUZAAAAoopVMfmeiMxHJCCtEQAAAFHFiph8T0TmI1JQnAEAACCqWBGT74nIfEQKbmsEAABATPCMy5eIzEfkoTgDAABATCAuH5GO2xoBAAAAIAJQnAEAAABABOC2RgAAAESs1htNS9LOitoujUVcPiIdxRkAAAAiktUbTeekEfyByEZxBgAAgIgTjI2mTzQ2WzoeYDWeOQMAAEDEsXqj6fg4afigvpaOCViNK2dnuFwu5efn+zxnt9tlt9tDPCMAAIDYZeVG03Fx0oqr2GgaLRwOhxwOh89zLpcrxLPxRnF2RmZmpsrLy8M9DQAAAHRB8bghmj52qGypSTpQVa+a+pbijo2m0Vp7F15sNpucTmeIZ3QWxRkAAACiXmFeumZdkCVJGpeTGubZAF1DcQYAAICwax2Zv23v0TDPCAg9ijMAAACEldWR+UC0ojgDAABA2FgVmc8G0+gJiNIHAABA2FgRmR8naeJwnjND9KM4AwAAQNhYEZn/4NXE5KNn4LZGAAAARIXicUNUmJdufk1MPnoaijMAAAAEXes0RqmluHJWnej0GIV56bqhaHgQZgdEBoozAAAABBVpjEDnUJwBAAAgaKxKY5RIZETPRyAIAAAAgsaKNEaJREbEBoozAAAABI0VaYwSiYyIDdzWCAAAgIjimcpIIiNiCcUZAAAAIgqpjIhVFGcAAACwTOvI/G17j4Z5RkD0oDgDAACAJYjMB7qH4gwAAADdRmQ+0H2kNQIAAKDbiMwHuo8rZ2e4XC7l5+f7PGe322W320M8IwAAgOhhGNbczEhkPoLN4XDI4XD4POdyuUI8G29xhlU/SVHKZrPJ6XQqOztbFRUV4Z4OAABAVHr/QLXmOP7ZqbY3T8uTLbWvUvsmypaapANV9YqLE5H5CLtw1wZcOQMAAEC3Haiu73TbguyBmnVBlvn1uBxuYwQkijMAAAB0QXci86tPNAZrWkBUozgDAABAQIjMB4KD4gwAAACdZkVkPlH5gG9E6QMAAKDTuhuZT1Q+4B/FGQAAADqtu8+LEZUP+MdtjQAAAAiK4nFDVJiXLklK7ZtIVD7QAYozAAAAeGmdxCi1FFc5qUkqO1DT6XEK89J1Q9Fw6ycI9FAUZwAAADCRxAiED8UZAAAAJFmTxOiJVEYgMASCAAAAQFL3kxg9kcoIBI7iDAAAAJK6n8ToiVRGIHDc1ggAAIBuIZURsAbFGQAAALqFVEbAGtzWCAAAAAARgCtnAAAAMar1fmbb9h4N84yA2EZxBgAAEIPYzwyIPBRnAAAAMYb9zIDIxDNnAAAAMYb9zIDIxJWzM1wul/Lz832es9vtstvtIZ4RAABAcLCfGWKZw+GQw+Hwec7lcoV4Nt4ozs7IzMxUeXl5uKcBAAAQUdx7mKX2TZQtNUkHqupVU9/IfmaIWu1deLHZbHI6nSGe0VkUZwAAADFmYFLvTre9In+orhyXZX49LodbGIFgoTgDAADowVrH5UtSRfWJTvfPSePKGBAqFGcAAAA9lBVx+RXV9VwtA0KE4gwAAKAHsiou32AjNCBkiNIHAADogayIyycmHwgtijMAAIAeyIq4fGLygdDitkYAAACYkfmSiMkHwoTiDAAAACrMS9cNRcPDPQ0gplGcAQAARDlfcfk7K2rDOCMAXRH24szpdGrfvn2SpNzcXGVnZwfUv7m5WXv27JHT6VRKSopGjBih1FQeXAUAALHBirh8AJGhS4EgX/va1xQXF9epP88884zPMV599VVNnjxZNptNF198sS6++GLZbDZNmjRJGzdu7HAOzc3Nevjhh5WXl6dRo0bpK1/5iiZPnqzBgwfryiuv1K5du7ry0QAAAKKGOy7fisJsYFKiBaMA6I4uFWeffPJJt9505cqVmjlzpt59990250pLSzVr1iw99NBDfvs3NTXp6quv1p133qn9+/e3ObdhwwaNHz9eW7Zs6dY8AQAAIpkVcfkSkflApAj4tsaGhgYdOHBAkvT9739fxcXF7bYvKCjw+nrz5s266667JEmDBg3SqlWrNGPGDEnSxo0bddttt+no0aO66667NH78eF1++eVtxrz//vu1bt06SdKECRO0cuVKTZ48WdXV1frd736nZcuWqb6+XvPmzdPHH3+swYMHB/oxAQAAIp4VcfkSkflApAi4ONuzZ4+MM1vFX3nllZo+fXqn+xqGocWLF8swDPXq1Uuvv/66xo8fb56//vrrNWbMGBUWFur06dO65557tHXrVsXFxZltDh8+bF5Vy8vL01//+lf169dPkpScnKx7771XqampWrhwoaqqqvTQQw/pZz/7WaAfEwAAoMfxjMuXiMwHIk3AtzV++umn5vGoUaMC6ltWVqaSkhJJ0ty5c70KM7cJEyZo7ty5kqSSkhJ9+OGHXudfeOEF1dXVSZIWLVpkFmae7Ha7MjIyJEmrV69Wc3NzQPMEAACIRAdr67X+faee2/q5ntv6ubbtPRpQf3dcvvvPrAuyKMyACBLwlTN3cda7d2/l5uYG1PeVV14xj+fMmeO33Zw5c/Tiiy9KkjZs2KDzzz/f5xizZ8/22T8+Pl7FxcV6+umndeTIEb3zzjsqLCwMaK4AAACRhFRGoOfr8pWzESNGqFevXmpqatLnn3+uv/71ryotLdWxY8f89vVMUJw5c6bfdp7ntm/f7nOMgoIC5eTk+B1j1qxZfscAAACIJlalMpLICES2gIszd1Lj0KFDtWLFCmVkZCgvL09f+cpXNGnSJKWnp2vGjBnaunVrm74HDx6UJKWkpCgtLc3ve6SlpSk5OVmSd0FnGIYOHTokSR1etfMs3IjVBwAA0cyKVEYSGYHI1+XbGt966y299dZbbc6fPn1amzZt0uuvv65HHnlEt956q3nOXVi1V5i5paenq66uzizoJKmqqkqnTp3q1Bjp6WcfdvUcwx/DMNq96teRPn36qE+fPl3uDwAA4E93UxnjJa0gkREx4OTJkzp58mSX+7uDD8MloOKsrq5OlZWV5tfjx4/Xz3/+c02ePFmnT5/Wzp079Zvf/EZ//OMf1dzcrIULF2rEiBHmbYruIsmzcPInPT1d+/fvN8M/PPt3ZgzP855j+FNZWakBAwZ02M6fpUuX6r777utyfwAAACu4ExlT+ybKlpqkE43NGj6oL4UZYsKKFSu0bNmycE+jywIqzvbs2aOEhARJLc+F/f73v1dKSop5/tJLL9Wll16qwsJC3XnnnZJaEhWnT5+u+Ph488pUUlLHfzm4r0LV19ebr3le2epoDM+rWJ5j+JOVlaWPPvqow3adeT8AAIDuOFhbr3c/r1JNfcsdQ4GkMroTGYFYtHjxYrMO6YoxY8Z4XYwKtYCKswsuuECnT5/usN3ChQu1Zs0alZaW6uOPP9aOHTs0YcIEDR48WJWVlaqpqelwDHcbzyLMczPpjsbwPN+ZYjAuLk79+/fvsB0AAEAwkcoIdF13HzXy3F85HAIOBOmMhIQEzZs3z/z6nXfekdQSIiK1PDvWEXcbzytz7v6dGcPzvOcYAAAAkepgbX23CzMSGYHoFZTiTJLOO+888/jo0ZZL8UOGDJHUUji197CdYRiqrm5JJfJMXUxJSTFTHDsqztz9W48BAAAQqT77V123CjMSGYHoFrTizPP2R3fQhvvKV2NjY7vx9rt27TJTGfPz873OuccoKytTc3Oz3zHKysrM49ZjAAAARKLkxIRu9X+QREYgqgX0zNn111+v999/X5mZmXrjjTcUH++/tvMM1xg1apQkadKkSXrqqackSevXr/e6uuZp/fr15vFFF13kdW7SpEnavXu3Dh06pNLSUk2ePDngMQAAACLRgeqOQ8zc5k+yqcDW8g/gqX0TNSE3lcIMiHIBXTn7t3/7N3344YfavHmz/vd//9dvu/r6ej333HOSWm5FvOSSSyRJxcXF5kN2a9eu9Xlro2EYeumllyS1PJA3e/Zsr/Nz5swxj9euXevz/WtqarR582ZJLYkro0eP7uxHBAAACJtA9jMrsA3QDUXDdUPRcM26IIvCDOgBAirOrr32WvNq2U033aQDBw60aVNfX69bb71Ve/fuldSS3OhOS8zKytK1114rSdq2bZuefPLJNv2feOIJbd26VVLLlTrPEBBJmjdvnnJzcyVJq1at0o4dO7zONzU1yW63q6GhQVJLlD8AAEAkOVhbr/XvO/Xc1s/NP6+UVcpZdSLcUwMQRnFGgNtgP/LII7rjjjskScnJyfrud7+rf/u3f1OvXr300Ucf6YUXXtDu3bslSRMnTtTf/vY3ryj7zz//XBMnTjQDPW688UbNmDFDzc3N2rBhg9asWSOpZRPp0tJSsxDz9PLLL2vu3LmSpH79+unWW281x3zmmWf0j3/8Q5JUVFSkv/3tb+rVy//dmzabTU6nU9nZ2aqoqAjkWwEAABAwq6Lyl88dy35mgMXCXRsEXJwZhqG7775bv/zlL9sN5Ljiiiv0+9//XmlpaW3O/f3vf9dVV12lI0eO+OybkZGhdevWacqUKX7Hf+yxx3THHXf43Xdt4sSJ2rhxozIyMtr9POFeAAAAEDsO1tZryorNloz12HXjdeW4LEvGAtAi3LVBwGmNcXFx+sUvfqHy8nL953/+pyZOnKhBgwYpMTFRw4YN07XXXquNGzfq1Vdf9VmYSdIll1yinTt36p577lF+fr6Sk5OVkpKisWPH6kc/+pE++OCDdgszSbrlllv07rvv6qabblJubq4SExOVlpamiy66SI899pjefvvtDgszAACAUCrdV91xo04gMh/omQK+ctbThLs6BgAAseO5rZ9ryboPuz3Oz64u0PzJwyyYEQBP4a4NAorSBwAAQGgVjxuiwrx0SUTmAz0dxRkAAEAEK8xLJ/gDiBEUZwAAABY7WFuvdz+vUk39Ka/Xt+09GqYZAYgGFGcAAAAWsioqH0DsoTgDAACwyMHaev1w7U5LxxyYlGjpeAAiV8BR+gAAAPDNqqh8NyLzgdhCcQYAAGCR6hONlo734NUFJDMCMYTbGgEAAMLIHZWf2jdRttQkHaiqV1yciMwHYhDF2Rkul0v5+fk+z9ntdtnt9hDPCAAAxILWUfnjcriNEQgmh8Mhh8Ph85zL5QrxbLxRnJ2RmZmp8vLycE8DAAAAQBC1d+HFZrPJ6XSGeEZnUZwBAAB0Uev9zNjHDEB3UJwBAAB0AfuZAbAaxRkAAECArNzPjH3MALgRpQ8AABAgq/YzYx8zAJ4ozgAAAAJk1X5m7GMGwBO3NQIAAARoYFLvTredP8mmAtsAr9dS+yayjxmANijOAAAAAjQsrW+n2y4oHMbeZQA6heIMAADAj9ZR+W57j9R1eoyK6nqKMwCdQnEGAADgg1VR+VY9nwag56M4AwAAaIWofADhQFojAABAK0TlAwgHijMAAIBWiMoHEA7c1ggAAGCB4nFDVJiXLomofABdQ3EGAABggcK8dN1QNDzc0wAQxSjOAABAzGsdmb9t79EwzwhALKI4AwAAMc2qyHwA6C6KMwAAELOIzAcQSUhrBAAAMYvIfACRhOIMAADELCsi8+NFZD4Aa3Bb4xkul0v5+fk+z9ntdtnt9hDPCAAARJLWUfm21CSdaGzW8EF9KcyAKOJwOORwOHyec7lcIZ6NtzjDMGL6+VebzSan06ns7GxVVFSEezoAACCIfKUyvlJ2qFN9l88dS1Q+0MOFuzbgyhkAAIgJpDICiHQUZwAAoMezIpWRNEYAwUYgCAAA6PG6m8pIGiOAUKA4AwAAPV53UxlJYwQQCtzWCAAA4KF1KuOE3FQKMwAhQXEGAADgoTAvnVRGAGFBcQYAAHqE1jH5nvYeqQvDjAAgMBRnAAAg6lkZk08qI4BwoTgDAABRzYqYfDdSGQGEE2mNAAAgqnU3Jt8TqYwAwokrZwAAIKp1NSZ//iSbCmwDJJHKCCAyUJwBAICYVGAbQCojgIjCbY0AAAAAEAEozgAAAAAgAnBbIwAAiCqt9zPbtvdomGcEANagOAMAAFHDyv3MACDSUJwBAICoYOV+ZhKbTQOIPDxzBgAAooKV+5mx2TSASERxBgAAokJX9zNrLV5sNg0gMnFb4xkul0v5+fk+z9ntdtnt9hDPCAAAdFXxuCEqzEs3v07tmyhbapJONDZr+KC+FGZADHM4HHI4HD7PuVyuEM/GW5xhGDH9TK3NZpPT6VR2drYqKirCPR0AAGLawdp6ffavOiUnJmh/1QkzkVFqSWV8pexQp8ZZPncsG0wDCFi4awOunAEAgIjwx3f2a/FLO9Uc0/9sDCCWUZwBAICwO1hbb2lEPkmMAKIRgSAAACDsSvdVW1aYkcQIIFpRnAEAgLCzKolRIokRQPTitkYAABC1PFMZU/smakJuKoUZgKhFcQYAAKJWYV46qYwAegxuawQAAACACEBxBgAAAAARgNsaAQBAyBysrde7n1d5bS4tSTsrasM0IwCIHBRnAAAgJP74zn5L9zKT2M8MQM9CcQYAAILuYG29frh2p6Vjsp8ZgJ6GZ84AAEDQle6rtnxM9jMD0NNw5QwAAARdVzaZvuni4Tp3cHKb19nPDEBPRXEGAAAi0rmDk9nDDEBM4bZGAAAAAIgAFGcAAAAAEAG4rREAAFiu9X5m2/YeDfOMACDyUZwBAABLBWM/MwCIBRRnAADAMlbuZ8YG0wBiDcXZGS6XS/n5+T7P2e122e32EM8IAIDoY9V+ZmwwDSBYHA6HHA6Hz3MulyvEs/FGcXZGZmamysvLwz0NAACiWlf2M/OFDaYBBEt7F15sNpucTmeIZ3QWxRkAAAiL4nFDVJiX7vUaG0wDiGUUZwAAIGCeaYypfROVk5qk/VUnAkplLMxLZ5NpAPBAcQYAAAJCGiMABAfFGQAA6DTSGAEgeOLDPQEAABA9SGMEgOChOAMAAJ1GGiMABA+3NQIAgKDyTGUkjREA/KM4AwAAQUUqIwB0TkQUZ06nU/v27ZMk5ebmKjs7O6D+zc3N2rNnj5xOp1JSUjRixAilpnIfOwAA3eEZl+8WSFQ+ACAwlhdnp06d0iWXXKKSkhItXbpU9913n9+2r776qu699169++67Xq9PnDhRP/nJTzRz5sx236u5uVmrVq3SI488ov3795uvJyQkaPr06Vq5cqVGjx7drc8DAEAsIi4fAELP8kCQH//4xyopKemw3cqVKzVz5sw2hZkklZaWatasWXrooYf89m9qatLVV1+tO++806swc5/bsGGDxo8fry1btgT+IQAAiGHuuHyrCjMi8wGgcywtzt544w39/Oc/77Dd5s2bddddd0mSBg0apDVr1qiqqkpVVVV6/vnnlZ7e8tDwXXfdpTfffNPnGPfff7/WrVsnSZowYYLeeustHT9+XAcOHNCyZcsUHx+v+vp6zZs3T0eOHLHmAwIAEAOsisuXiMwHgEBYVpy5XC79x3/8R4ftDMPQ4sWLZRiGevXqpddff10LFixQamqqUlNTdf311+v1119Xr14td1zec889Mgzvf7s7fPiweVUtLy9Pf/3rX3XppZcqOTlZNptN9957rx555BFJUlVVVbtX4AAAgLfW/9/tDiLzAaDzLHnmrLm5Wd/85jflcrk0dOhQHTx40G/bsrIy87bHuXPnavz48W3aTJgwQXPnztWLL76okpISffjhhzr//PPN8y+88ILq6uokSYsWLVK/fv3ajGG327V8+XIdPnxYq1ev1ooVKxQfz7ZuAAB0ZFha34Da33XFKF0ycpAOVNWrpr5lHzQi8wEgcJZUKw8//LA2bdqkc845R7/97W/bbfvKK6+Yx3PmzPHbzvPchg0b/I4xe/Zsn/3j4+NVXFwsSTpy5IjeeeedducFAABaHKiuD6j98PRkjctJ1ZXjsnRD0XDdUDRcsy7IojADgAB1uzgrLS3V4sWLJbUUaWPHjm23/a5du8zj9tIYPc9t377d5xgFBQXKycnxO8asWbP8jgEAAHyrPtEY1PYAAN+6dVvjF198oWuvvVanTp3S3Llz9d3vftfcr8wf9y2PKSkpSktL89suLS1NycnJqqur8yroDMPQoUOHJLXsidYez8LNcwwAANCCvcwAIHJ0qziz2+3avXu3srOz9dRTTykuLq7DPu7Cqr3CzC09PV11dXVez7BVVVXp1KlTnRrDnfooqd3n4AAAiEXsZQYAkaXLxdlzzz2n5557TnFxcV7x9x1xF0mdaZ+enq79+/eb4R+e/Tszhud5zzF8MQxDx44d63BO/vTp00d9+vTpcn8AAELJvZeZFdjHDECkOHnypE6ePNnl/lam1XZFl4qzTz/9VN///vcltUTdX3rppZ3u6y6AkpI6fkjYXezU1599MNmzgOpoDM9iyXMMXyorKzVgwIAO5+TP0qVLdd9993W5PwAAoWTVXmbsYwYgkqxYsULLli0L9zS6LODirLGxUdddd52OHz+uoqIiLV26NKD+gwcPVmVlpWpqajps627jWYQNHjy4zfmO+rcew5esrCx99NFHHc7JH66aAQCiiVUhHuxjBiCSLF68WHfeeWeX+48ZM0aVlZUWzigwARdn99xzj0pLS9W/f3/9/ve/V+/evQPqP3ToUFVWVqqqqqrDtu42KSkpXv1bn++of+sxfImLi1P//v07nBMAALGoeNwQFeadfVyAfcwARKLuPmrUmQyNYAqoOKusrNTKlSslSV//+te1a9euNimILpfLPN69e7c2bdokScrIyNCECRM0ZMgQSS2Fk2EYfr8BhmGourrllgvP1MWUlBQzxbGj4szdv/UYAAAgMIV56bqhaHi4pwEAPVpAxVlj49lbIJ566ik99dRT7bZfs2aN1qxZI6llU+l169aZV74aGxu1a9cunXfeeT777tq1y0xlzM/P9zo3dOhQ7d69W2VlZWpublZ8vO/t2srKyszj1mMAABALfEXlp/ZNlLPqRBhnBQDwpVtR+l0xadIks6hbv3693+Js/fr15vFFF13UZozdu3fr0KFDKi0t1eTJkwMeAwCAno6ofACILr4vOfkxfPhwGYbR7p/PPvvMbL906VLz9XXr1kmSiouLzVsZ165d6zOu0jAMvfTSS5Ja7vucPXu21/k5c+aYx2vXrvU515qaGm3evFlSy4N9o0ePDuSjAgAQ1dxR+VYVZsTlA0DwBVScWSErK0vXXnutJGnbtm168skn27R54okntHXrVknS9ddf7xUCIknz5s1Tbm6uJGnVqlXasWOH1/mmpibZ7XY1NDRIkhYtWmT1xwAAIKJZFZUvEZcPAKES8tsaJemBBx7Qa6+9pqqqKt188816++23NWPGDDU3N2vDhg3mc2rp6elavnx5m/59+vTRqlWrNHfuXDU0NGjq1Km69dZbNXHiRFVVVemZZ57RP/7xD0lSUVGRbrzxxpB+PgAAws2qqHyJuHwACJWwFGfDhw/Xyy+/rKuuukpHjhzRs88+q2effdarTUZGhtatW2deIWttzpw5evTRR3XHHXfoiy++0AMPPNCmzcSJE/Xyyy+rV6+wfEwAAKIKcfkAEF5hq1ouueQS7dy5U7/61a+0bt067du3T3FxccrNzdXcuXN12223eW047cstt9yiL3/5y/rVr36lN998UwcPHlRKSorOO+88LViwQN/5zncC3ocNAIBYRVw+AIRXnOErkSOG2Gw2OZ1OZWdnq6KiItzTAQAgYL7i8rftPapXyg4FNM7yuWMpzgDEtHDXBtzvBwBAFCMuHwB6DoozAACilDsu3yrE5QNAeIU8Sh8AAFiDuHwA6FkozgAAiFLE5QNAz8JtjQAARKmBSYElEs+fZFOBbYBS+ybKlpqkA1X1iosTcfkAECEozgAAiFLD0voG1H5B4TCNyzl766LnMQAg/LitEQCAKHWguj6g9hUBtgcAhBZXzgAAiBKt9zPbtvdoQP2tfEYNAGA9ijMAAKIA+5kBQM9HcXaGy+VSfn6+z3N2u112uz3EMwIAoIVV+5mxjxkASA6HQw6Hw+c5l8sV4tl4ozg7IzMzU+Xl5eGeBgAAbVixnxn7mAFAi/YuvNhsNjmdzhDP6CwCQQAAiHBWPCvGPmYAEPm4cgYAQA9SPG6ICvPSza9T+yayjxkARAmKMwAAIkTrNEappbhyVp3o9BiFeem6oWh4EGYHAAg2ijMAACIAaYwAAIozAADCzKo0RolERgCIZgSCAAAQZlakMUokMgJAtKM4AwAgzKxIY5RIZASAaMdtjQAARCHPVEYSGQGgZ6A4AwAgCpHKCAA9D8UZAAAh1joyf9veo2GeEQAgElCcAQAQQkTmAwD8oTgDACBEiMwHALSHtEYAAEKEyHwAQHsozgAACBEi8wEA7eG2RgAAIpA7Kj+1b6JsqUk6UFWvuDgRmQ8APRjFGQAAEah1VP64HG5jBICejuIMAIAgaR2Zv7OittN9CfwAgNhDcQYAQBB0NzI/J41bFwEg1lCcAQBgMSsi8080Nls0GwBAtKA4O8Plcik/P9/nObvdLrvdHuIZAQCiVXcj8+PjpOGD+lo0GwCAJ4fDIYfD4fOcy+UK8Wy8UZydkZmZqfLy8nBPAwDQA3QnMj8uTlpxFVH5ABAs7V14sdlscjqdIZ7RWRRnAACEkTsyX5JS+yYSlQ8AMYziDACAbmidyJjaN1HOqhOd7t86Mh8AELsozgAA6KLuJjICAOCJ4gwAgC6wIpFRYj8zAMBZ8eGeAAAA0ai7iYySFCdp4vDU7k8GANAjUJwBANAF3UlkdHvwalIZAQBncVsjAAAhQCojAKAjFGcAAHSgdSKjJG3bezSgMUhlBAB0hOIMAIB2kMgIAAgVijMAAPywKpFRIpURANAxAkEAAPDDikRGiVRGAEDnUJwBAOCHFYmMEqmMAIDO4bZGAAAsUjxuiKaPHSpbapIOVNUrLk6kMgIAOo3iDAAAixTmpWvWBVmSpHE53MYIAAgMxRkAAGe0jswPNC4fAIDuoDgDAEBE5gMAwo/iDAAQ86yKzCcuHwDQHaQ1AgBinhWR+cTlAwC6iytnZ7hcLuXn5/s8Z7fbZbfbQzwjAECoWBGZT1w+AEQHh8Mhh8Ph85zL5QrxbLxRnJ2RmZmp8vLycE8DABAGA5N6d7rt/Ek2FdgGmF+n9k0kLh8Aokh7F15sNpucTmeIZ3QWxRkAIOYNS+vb6bYLCocRkw8ACAqKMwBATPGMy0/tm6ic1CT9+b3O/ytpRXU9xRkAICgozgAAMcOKuHwrnk8DAMAXijMAQEwgLh8AEOmI0gcAxATi8gEAkY7iDAAQE4jLBwBEOm5rBADAh+JxQ1SYly6JuHwAQGhQnAEAeiTPVEZJ2rb3aED9C/PSdUPR8CDMDAAA3yjOAAA9jhWpjAAAhBrFGQCgRyGVEQAQrQgEAQD0KKQyAgCiFcUZAKBHIZURABCtuK0RABCz5k+2acGFw3Sgql419Y2kMgIAworiDAAQ1bqTyliQPUDjclI1LodbGAEA4UdxBgCIWqQyAgB6EoozAEBUsiKVkURGAEAkIRAEABCVupvKSCIjACDSUJwBAKJSd1MZSWQEAEQabmsEAPR4xeOGqDAvXZJIZAQARCyKszNcLpfy8/N9nrPb7bLb7SGeEQCgdRJjat9E5aQmaX/ViYBSGQvz0nVD0fAgzRIAEE0cDoccDofPcy6XK8Sz8UZxdkZmZqbKy8vDPQ0AwBkkMQIAgqG9Cy82m01OpzPEMzqLZ84AABHnYG29pYUZqYwAgGhAcQYAiDif/avOssKMVEYAQLSgOAMARJzkxATLxiKVEQAQLXjmDAAQcQ5U13ep3/xJNhXYBkgilREAEH0ozgAAEaere5gV2AaQyggAiFrc1ggAAAAAEYDiDAAAAAAiALc1AgDCqvVG05IC2mAaAICeguIMABA2bDQNAMBZ3S7Ojh8/rt27d+vEiRP60pe+pEGDBikuLq7T/Z1Op/bt2ydJys3NVXZ2dkDv39zcrD179sjpdColJUUjRoxQair72QBApDtYW68frt1p6ZhsNg0AiGZdeuasoaFBP/nJT5Sbm6t+/fpp/Pjxuvjii5WRkaG0tDT96Ec/0rFjx9od49VXX9XkyZNls9l08cUX6+KLL5bNZtOkSZO0cePGDufQ3Nyshx9+WHl5eRo1apS+8pWvaPLkyRo8eLCuvPJK7dq1qysfDQAQIqX7qi0dj82mAQDRLuDirK6uTtOmTdPSpUu1f//+Nudramr0wAMPaOTIkdq50/e/iK5cuVIzZ87Uu+++2+ZcaWmpZs2apYceesjvHJqamnT11VfrzjvvbDOHpqYmbdiwQePHj9eWLVsC/HQAgFDpaly+P2w2DQCIdgEXZ3fddZdKSkokSePHj9emTZtUWVmpqqoqbdmyRV/72tckSUeOHNE111yjU6dOefXfvHmz7rrrLknSoEGDtGbNGlVVVamqqkrPP/+80tPTzfd58803fc7h/vvv17p16yRJEyZM0FtvvaXjx4/rwIEDWrZsmeLj41VfX6958+bpyJEjgX5EAEAEKh43RI4F4/Wy/SI9dt14LZ87VsvnjpVjwXi9vfgyzZ88LNxTBACgW+IMw+j0c9i1tbXKyMhQY2OjRowYofLyciUmet/f39zcrK985SvmVas333xTl112mSTJMAwVFRWppKREvXr1UklJicaPH+/Vf/v27SosLNTp06d14YUXauvWrV7PsB0+fFjnnnuu6urqlJeXp/fff1/9+vXzGuPRRx/VwoULJUl33323fvazn/n9TDabTU6nU9nZ2aqoqOjstwIA0E3Pbf1cS9Z92On2y+eOZYNpAEBQhbs2COjK2UcffaTGxpbbUH7wgx+0KcwkKT4+XnfccYf59fvvv28el5WVmVfd5s6d26Ywk1quhM2dO1eSVFJSog8/9P4f9wsvvKC6ujpJ0qJFi9oUZpJkt9uVkZEhSVq9erWam5sD+ZgAAIsdrK3X+vedem7r5+Yf4vIBAPAWUFrjp59+ah6PHTvWb7svfelLPvu88sor5vGcOXP89p8zZ45efPFFSdKGDRt0/vnn+xxj9uzZPvvHx8eruLhYTz/9tI4cOaJ33nlHhYWFft8PABA8xOUDANA5AV05GzlypFasWKEVK1YoPz/fbzun02keDxkyxDz2TFCcOXOm3/6e57Zv3+51zj1GQUGBcnJy/I4xa9Ysv2MAAELDHZdvRWFGTD4AoKcL6MrZlClTNGXKlHbbnDp1Sg8++KD59bx588zjgwcPSpJSUlKUlpbmd4y0tDQlJyerrq7Oq6AzDEOHDh2S1LInWns8Czdi9QEgPKyKyycmHwAQC7q9CXVlZaX+/ve/q6qqSp988on+9Kc/mVfO7r//fhUUFJht3YVVe4WZW3p6uurq6syCTpKqqqrM9MeOxnCnPkryGsMfwzA63JutPX369FGfPn263B8AeiKr4vKJyQcAdMbJkyd18uTJLvcPICsxKLpdnG3ZskXXXXed12tpaWlas2aNpk+f7vW6u0jyLJz8SU9P1/79+83wD8/+nRnD87znGP5UVlZqwIABHbbzZ+nSpbrvvvu63B8A0BKXX5h39u/v1L6JmpCbSmEGAOiUFStWaNmyZeGeRpd1uzjzpaqqSrfffrsefvhhzZgxw3zdfWUqKanj/8m6r0LV19e36d+ZMTyvYnmO4U9WVpY++uijDtt15v0AAF1TmJdOXD4AoMsWL16sO++8s8v9x4wZo8rKSgtnFJhuF2fXXnutrr32Wh07dkyfffaZXnrpJT388MPatWuXiouL9eKLL5rR+IMHD1ZlZaVqamo6HNfdxrMIGzx4cJvzHfVvPYY/cXFx6t+/f4ftAABtHayt17ufV6mm/pTX6zsrasM0IwBALOruo0ae+yuHg2VXzvr3769x48Zp3Lhx+trXvqZp06apqalJd999t4qLi5WQkKChQ4eqsrJSVVVVHY7nbpOSkmK+NnTo0DbnO+rfegwAgLWIygcAwBoBRel31iWXXKIrrrhCUss+Z+7dtd2x+lVVVe0+bGcYhqqrWxK+PFMXU1JSlJycbI7RHnf/1mMAAKxzsLbe0sKMuHwAQCwLqDj77//+b91www1asmRJh23HjBljHrvv23Rf+WpsbGw33n7Xrl1mKmPr/dTcY5SVlam5udnvGGVlZeZxe3uyAQC67rN/1VlWmBGXDwCIdQEVZ2VlZVqzZo2eeuqpDtt6bkQ9cOBASdKkSZPM19avX++3r+e5iy66yOuce4xDhw6ptLS0S2MAAKyRnJhg2VjE5QMAYl1AxdmoUaMktRRGH3/8sd92x48f19/+9jdJLYXZyJEjJUnFxcXmQ3Zr1671eWujYRh66aWXJLU8kDd79myv83PmzDGP165d6/P9a2pqtHnzZkktV/BGjx7dqc8HAAjMgeqO03Bbmz/JpuVzx5p/HAvG6+3Fl2n+5GFBmCEAANEjoOKssLDQPP7e976n48ePt2lz8uRJ3XLLLeatjN/4xjfUu3dvSS1x9ddee60kadu2bXryySfb9H/iiSe0detWSdL111/vFQIiSfPmzVNubq4kadWqVdqxY4fX+aamJtntdjU0NEiSFi1aFMhHBAAEoCubTBfYBuiGouHmn1kXZHHFDAAASXFGANtgnzp1SuPHj9eHH34oScrOztZ//ud/6ktf+pISEhL0ySef6LnnntOePXskSSNHjtS2bduUlpZmjvH5559r4sSJZqDHjTfeqBkzZqi5uVkbNmzQmjVrJLVsIl1aWmoWYp5efvllM56/X79+uvXWW80xn3nmGf3jH/+QJBUVFelvf/ubevXyH0pps9nkdDqVnZ1tBpcAADrnua2fa8m6DwPqs3zuWPYyAwBEpHDXBgEVZ5K0d+9ezZo1q93bGiVpypQp+sMf/qBhw9repvL3v/9dV111lY4cOeKzb0ZGhtatW6cpU6b4Hf+xxx7THXfcodOnT/s8P3HiRG3cuFEZGRntzjPcCwAA0cLXXmbb9h7VK2WHAhqH4gwAEKnCXRsEvM/Zueeeq9LSUv3P//yPXnvtNe3du1efffaZ+vTpo9GjR2vUqFGaM2eOrrzySiUk+H5Q/JJLLtHOnTv1q1/9SuvWrdO+ffsUFxen3NxczZ07V7fddpvXhtO+3HLLLfryl7+sX/3qV3rzzTd18OBBpaSk6LzzztOCBQv0ne98x7ydEgDQPexlBgBA8AV85aynCXd1DACR7mBtvaas2GzZeI9dN15XjsuybDwAAKwS7togKJtQAwB6jtJ91ZaNxV5mAAD4R3EGAGhXVxIZfYkXe5kBANCegJ85AwCgPcXjhqgwL12SlNo3UbbUJJ1obNbwQX0pzAAAaAfFGQDAUoV56aQxAgDQBRRnAAAvrSPzt+09GuYZAQAQGyjOAAAmIvMBAAgfijMAgKSWK2Y/XLuz2+MMTEq0YDYAAMQe0hoBAJKsicwnKh8AgK7jytkZLpdL+fn5Ps/Z7XbZ7fYQzwgAQsuKyHyi8gEAkc7hcMjhcPg853K5QjwbbxRnZ2RmZqq8vDzc0wCAqOAZly+1ROZPyE2lMAMARLz2LrzYbDY5nc4Qz+gsijMAiEGtExklaWdFbaf7E5cPAID1KM4AIMaQyAgAQGSiOAOAGEIiIwAAkYu0RgCIISQyAgAQuSjOACCGkMgIAEDk4rZGAIBfnqmMJDICABBcFGcAEEMGJvUOqP0V+UN15bisIM0GAAB4ojgDgB7KV1x+RfWJgMbISeMqGQAAoUJxBgA9kFVx+RXV9RqXQ/gHAAChQHEGAD2MVXH5kmSwGRoAACFDWiMA9DBWxOVLROYDABBqFGcA0MNYEZcfLyLzAQAINW5rBIAY547LT+2bKFtqkk40Nmv4oL4UZgAAhBjFGQDEuMK8dN1QNDzc0wAAIOZRnAFAFPMVl79t79EwzggAAHQVxRkARCmr4vIBAEBkoDgDgChkZVz+wKRES8YBAADdQ1ojAEQh4vIBAOh5KM4AIApZEZcvEZcPAEAk4bbGM1wul/Lz832es9vtstvtIZ4RAFjHHZfvlto3URNyUynMAAAxx+FwyOFw+DzncrlCPBtvFGdnZGZmqry8PNzTAICgIC4fAIAW7V14sdlscjqdIZ7RWRRnABDhfMXl76yoDeOMAABAMFCcAUAEIy4fAIDYQXEGABGKuHwAAGILaY0AEKGIywcAILZQnAFAhCIuHwCA2MJtjQDQA8yfZFOBbYDXa8TlAwAQXSjOAKAHKLANICofAIAoR3EGAGHkKybfjbh8AABiC8UZAISJlTH5pDECABD9KM4AIAysjMknjREAgJ6BtEYACAOrYvIl0hgBAOgpuHIGAGHQ1Zh8z1RG0hgBAOhZKM4AIIqQyggAQM/FbY0AAAAAEAEozgAAAAAgAlCcAQAAAEAE4JkzAAiC1ptLp/ZNVE5qkvZXnVBN/Slt23s0zDMEAACRhuIMACxm5ebSAAAgdlCcAYCFrNxc2peBSYlBGxsAAIQXxdkZLpdL+fn5Ps/Z7XbZ7fYQzwhANLJyc+nW4iRNHJ4atPEBAIgFDodDDofD5zmXyxXi2XijODsjMzNT5eXl4Z4GgCjX1c2lO+PBqwvYcBoAgG5q78KLzWaT0+kM8YzOojgDgAhRPG6ICvPS27ye2jdRE3JTKcwAAOjhKM4AIEIU5qXrhqLh4Z4GAAAIE/Y5AwAAAIAIQHEGAAAAABGA2xoBoBtabzbN5tIAAKCrKM4AoIvYbBoAAFiJ4gwAuiAYm02zwTQAALGNZ84AoAus3myaDaYBAADFGQB0gdWbTbPBNAAA4LZGAAiy4nFDNH3sUNlSk3Sgql419WcLOzaYBgAAbhRnANAFA5N6d7rtFflDNeuCLEnSuBxuXQQAAL5RnAFAB1rH5UtSRfWJTvfPSeOqGAAA6BjFGQC0w4q4/Irqeq6YAQCADlGcAYAfVsXlG2yEBgAAOoG0RgDww4q4fCLyAQBAZ1GcAYAfVsTlE5EPAAA6i9saAcAixeOGqDAvXRIR+QAAIHAUZwBgkcK8dN1QNDzc0wAAAFGK4gxAzPMVlZ/aN1HOqs7H5QMAAHQXxdkZLpdL+fn5Ps/Z7XbZ7fYQzwhAKFgRlQ8AAKKHw+GQw+Hwec7lcoV4Nt4ozs7IzMxUeXl5uKcBIISsisp3G5iUaNlYAAAgONq78GKz2eR0OkM8o7NIawQQs6yIyncjMh8AAHQXxRmAmGVFVL4bkfkAAKC7uK0RAALgGZcvEZkPAACsQ3EGAAEgLh8AAAQLxRmAmNE6Mn/b3qNhnhEAAMBZFGcAYgKR+QAAINJRnAHo8ayMzCcuHwAABAtpjQB6PKsi84nLBwAAwRT2K2dOp1P79u2TJOXm5io7Ozug/s3NzdqzZ4+cTqdSUlI0YsQIpabyyxOAs6yKzCcuHwAABFOXizOXy6Xf/OY3evfdd/XJJ5/owIEDysvL09ixY/X//t//07e+9S0lJCT47f/qq6/q3nvv1bvvvuv1+sSJE/WTn/xEM2fObPf9m5ubtWrVKj3yyCPav3+/+XpCQoKmT5+ulStXavTo0V39eABilDsqP7VvomypSTpQVa+4OBGXDwAAgi7OMIyAn49/9dVXdd1116m2ttZvmwkTJui3v/2tJkyY0ObcypUr9YMf/KDd9/jFL37ht01TU5OuueYarVu3zm//pKQkbdq0SVOnTm33fWw2m5xOp7Kzs1VRUdFuWwCRrXUao9u2vUf1StmhTo2xfO5YovIBAIhR4a4NAr5yVl5erquvvlr19fWSpOLiYn31q1/V0KFDtWvXLr3wwgsqLy/X9u3bddlll2nbtm1eV7A2b96su+66S5I0aNAgrVq1SjNmzJAkbdy4UbfddpuOHj2qu+66S+PHj9fll1/eZg7333+/WZhNmDBBK1eu1OTJk1VdXa3f/e53WrZsmerr6zVv3jx9/PHHGjx4cMDfGADRhTRGAAAQ7QK+cjZ//nz96U9/kiQ99thj+v73v6+4uDjzfGNjo+644w49/vjjkqQrrrhCmzZtkiQZhqGioiKVlJSoV69eKikp0fjx473G3759uwoLC3X69GldeOGF2rp1q9f4hw8f1rnnnqu6ujrl5eXp/fffV79+/bzGePTRR7Vw4UJJ0t13362f/exnfj9PuKtjAN13sLZeF63YbElh9th143XluCwLRgIAANEm3LVBQGmNDQ0N+vOf/yyp5YpV68JMkhITE/Xwww/r/PPPlyS99tprcjqdkqSysjKVlJRIkubOndumMHOPO3fuXElSSUmJPvzwQ6/zL7zwgurq6iRJixYtalOYSZLdbldGRoYkafXq1Wpubg7kYwKIMp/9q86Swow0RgAAEE4BFWfvv/++Tp1qeZbjmmuuaVOYuSUmJurKK680v37vvfckSa+88or52pw5c/y+j+e5DRs2eJ3zHGP27Nk++8fHx6u4uFiSdOTIEb3zzjt+3wtA9EtO9B8+FAjSGAEAQDgF9MyZy+Uyj3Nzc9ttO3ToUPPY/Xzarl27zNfaS2P0PLd9+3avc+4xCgoKlJOT43eMWbNm6emnnzbHKCwsbHe+AKLXger6gPvMn2RTgW2AJCm1byJpjAAAIOwCKs7OP/98rV69WpI0bdq0dtt6Xq0aNWqUJOngwYOSpJSUFKWlpfntm5aWpuTkZNXV1XkVdIZh6NChlsS1jopDz8LNcwwAPU9X9jErsA0glREAAESUgIqzc889V+eee26H7crKyvSHP/xBUstDdWPGjJEks7BqrzBzS09PV11dnVnQSVJVVZV5W2VHY6Snp5vHnmMAAAAAQCTq8ibU/pSWlqq4uFinT5+WJC1evFiJiYmSzhZJnoWTP+np6dq/f78Z/uHZvzNjeJ73HMMfwzB07NixDtv506dPH/Xp06fL/QEAAAB0z8mTJ3Xy5Mku9+/CFtCWsqw4O3bsmFasWKGHHnrILMwWLFigm2++2auN1LJBdEfchY77eTXP/p0Zw7NQ8hzDn8rKSg0YMKDDdv4sXbpU9913X5f7A+iYv02md1bUhmlGAAAgkqxYsULLli0L9zS6rNvF2alTp/Tb3/5W9913n44cOWK+/u1vf1uPP/644uPPBkIOHjxYlZWVqqmp6XBcdxvPIsxzM+mOxvA835liMCsrSx999FGH7fzhqhkQXGwyDQAAOrJ48WLdeeedXe4/ZswYVVZWWjijwHSrONuxY4duuukmMypfkoYNG6bf/OY3mjFjRpv2Q4cOVWVlpaqqqjoc290mJSXFq3/r8x31bz2GP3Fxcerfv3+H7QCE3sHaev1w7U5LxxyYlGjpeAAAIPy6+6iRv63CQiWgfc7cmpqatGTJEk2ePNkszNLS0rRy5Urt2rXLZ2EmSUOGDJHUUji1dz+nYRiqrq6W5J26mJKSouTkZHOM9rj7tx4DQPQp3VfdcaMAsNk0AACIRAEXZ4ZhyG63a/ny5Tp9+rTi4uK0cOFC7dmzR3feeafOOeccv33dV74aGxvbjbfftWuXmcqYn5/vc4yysjI1Nzf7HaOsrMw8bj0GgOjSlaj89rDZNAAAiEQB39a4fPlyPfHEE5KkzMxMvfTSS7rooos61XfSpEl66qmnJEnr16/Xeeed57Pd+vXrzePWY0+aNEm7d+/WoUOHVFpaqsmTJwc8BoCezXODaU9sNg0AACJZQFfOjh07pp/97GeSpNTUVP3jH/8IqPApLi427+Ncu3atz1sbDcPQSy+9JKnlns/Zs2d7nZ8zZ455vHbtWp/vU1NTo82bN0tqeahv9OjRnZ4jgOjn3mC69Z9ZF2RRmAEAgIgVUHG2Zs0ac8+wn/70pxoxYkRAb5aVlaVrr71WkrRt2zY9+eSTbdo88cQT2rp1qyTp+uuv9woBkaR58+YpNzdXkrRq1Srt2LHD63xTU5PsdrsaGhokSYsWLQpojgDC72Btvda/79RzWz/Xc1s/17a9R8M9JQAAgKAL6LbGLVu2mMfJycnatGlTp/pdeOGFSktLkyQ98MADeu2111RVVaWbb75Zb7/9tmbMmKHm5mZt2LBBa9askdSyifTy5cvbjNWnTx+tWrVKc+fOVUNDg6ZOnapbb71VEydOVFVVlZ555hn94x//kCQVFRXpxhtvDOQjAggzIvMBAECsijMC2Ab7sssu01tvvRXwm7z11lu69NJLza///ve/66qrrvLaF81TRkaG1q1bpylTpvgd87HHHtMdd9xhbnjd2sSJE7Vx40ZlZGS0OzebzSan06ns7GxVVFR0/GEABM3B2npNWbG52+M8dt14XTkuy4IZAQCAWBLu2iCg2xpdLpclb3rJJZdo586duueee5Sfn6/k5GSlpKRo7Nix+tGPfqQPPvig3cJMkm655Ra9++67uummm5Sbm6vExESlpaXpoosu0mOPPaa33367w8IMQGSxIjKfmHwAABCtArpy1hOFuzoGcNZzWz/XknUfdmuMn11doPmTh1k0IwAAEEvCXRsEHKUPAJGgeNwQFealm18Tkw8AAKIdxRmAsDlYW693P69STX3LpvOBpDIW5qXrhqLhQZoZAABA6FGcAQgLUhkBAAC8UZwBCLmDtfX64dqd3RpjYFKiRbMBAACIDAGlNQKAFbqbykgiIwAA6IkozgCEXHdDYh+8uoDgDwAA0ONwWyOAkBuW1rfTbW+elidbakt7EhkBAEBPRnEGIOQOVNd3um1B9kDNuiAriLMBAACIDBRnAIKqdVy+JO2sqO10/+oTjcGYFgAAQMShODvD5XIpPz/f5zm73S673R7iGQHRj7h8AAAQaRwOhxwOh89zLpcrxLPxRnF2RmZmpsrLy8M9DaDHsCIuXyIyHwAAWKu9Cy82m01OpzPEMzqLtEYAQdHduHyJyHwAABBbKM4ABIUVz4oRmQ8AAGIJtzUCiAjF44aoMC9dEpH5AAAgNlGcAeg2X4mM2/YeDWiMwrx03VA03OKZAQAARA+KMwDdQiIjAACANSjOAHSZVYmMEqmMAAAABIIA6DIrEhklUhkBAAAkijMA3WBFImO8SGUEAACQuK0RQAi5ExlT+ybKlpqkE43NGj6oL4UZAACAKM4AhBCJjAAAAP5RnAHoFCvi8gEAAOAfxRmADhGXDwAAEHwUZwDaRVw+AABAaJDWCKBdxOUDAACEBsUZgHZZEZcvEZcPAADQEW5rBGApd1y+W2rfRE3ITaUwAwAA6ADFGQBLEZcPAADQNdzWCKBdA5N6B9ie0A8AAICu4MoZAEm+9zGTpNpWX3ckJ43bFwEAALqC4uwMl8ul/Px8n+fsdrvsdnuIZwSEjpX7mFVU12tcDqmMAAAgMjkcDjkcDp/nXC5XiGfjjeLsjMzMTJWXl4d7GkDIWbmPmSQZ7FQNAAAiWHsXXmw2m5xOZ4hndBbPnAExzqp9zCT2MgMAAOgOijMgxlm1j5nEXmYAAADdwW2NAAIyf5JNBbYBXq+xlxkAAED3UZwBCEiBbQD7mAEAAAQBxRkQY1pH5m/bezTMMwIAAIBEcQbEFCsj8wEAAGAtijMgRlgVmT8wKdGC2QAAAKA10hqBGGFFZD5R+QAAAMFDcQbECCsi84nKBwAACB5uawTQRvG4ISrMSze/JiofAAAg+CjOgB6qO6mMhXnpxOUDAACEGMUZ0AORyggAABB9KM6AHsaKVEYSGQEAAEKPQBCgh+luKiOJjAAAAOFBcQb0MN1NZSSREQAAIDy4rRGIYZ6pjCQyAgAAhBfFGRDFWicyStLOitpO9yeVEQAAIHJQnAFRikRGAACAnoXi7AyXy6X8/Hyf5+x2u+x2e4hnBPhnRSKjRCojAACIPQ6HQw6Hw+c5l8sV4tl4ozg7IzMzU+Xl5eGeBtAp3U1klEhlBAAAsam9Cy82m01OpzPEMzqLtEYgChlG929mJJURAAAgsnDlDIhCw9L6BtT+5ml5sqW29CGVEQAAIDJRnAFR6EB1fUDtC7IHatYFWUGaDQAAAKxAcQZEOF9x+dv2Hg1ojO5uTA0AAIDgozgDIhhx+QAAALGD4gyIUFbF5UtE5gMAAEQD0hqBCGVFXL5EZD4AAEC0oDgDIpQVz4nFi8h8AACAaMFtjUAPUTxuiArz0pXaN1G21CSdaGzW8EF9KcwAAACiBMUZ0EMU5qXrhqLh4Z4GAAAAuojiDIgQrSPzA43LBwAAQHSjOAMiAJH5AAAAoDgDwsyqyHzi8gEAAKIbaY1AmFkRmU9cPgAAQPSjOAPCzIrIfOLyAQAAoh+3NQJRxB2X75baN1ETclMpzAAAAHoAijMgBFonMXraWVHb6XGIywcAAOi5KM6AICOJEQAAAJ1BcQYEkVVJjG4kMgIAAPRcFGdnuFwu5efn+zxnt9tlt9tDPCP0BFYkMbqRyAgAANB9DodDDofD5zmXyxXi2XijODsjMzNT5eXl4Z4GehgrkhjdSGQEAADovvYuvNhsNjmdzhDP6CyKMyACeaYyksgIAAAQGyjOgAhEKiMAAEDsYRNqIIgGJvXuYj+CPwAAAGINxRkQRMPS+napX04atzACAADEGm5rBCzia6PpQDaY9lRRXa9xOSQzAgAAxBKKM8ACVm80bWXKIwAAAKIDxRnQTVZvNC3xzBkAAEAs4pkzoJus3GhaYrNpAACAWEVxBnSTlbcgxovNpgEAAGJV2G9rdDqd2rdvnyQpNzdX2dnZAfVvbm7Wnj175HQ6lZKSohEjRig1lasOiFyeG0xLLZtM21KTdKKxWcMH9aUwAwAAiFGWXTlzOByKi4vTfffd16n2r776qiZPniybzaaLL75YF198sWw2myZNmqSNGzd22L+5uVkPP/yw8vLyNGrUKH3lK1/R5MmTNXjwYF155ZXatWtXNz8REBzuDabdf2ZdkKVxOamaMiKdwgwAACCGWXbl7Pnnn+9025UrV+oHP/iBz3OlpaWaNWuWfvGLX/ht09TUpGuuuUbr1q3zeW7Dhg3avHmzNm3apKlTp3Z6XoA/njH5qX0TlZOapP1VJ1RTf0rb9h4N9/QAAADQA1hSnK1evVpbt27tVNvNmzfrrrvukiQNGjRIq1at0owZMyRJGzdu1G233aajR4/qrrvu0vjx43X55Ze3GeP+++83C7MJEyZo5cqVmjx5sqqrq/W73/1Oy5YtU319vebNm6ePP/5YgwcPtuJjIkZZHZMPAAAA+NLl2xpra2v1t7/9TTfddJO++93vdqqPYRhavHixDMNQr1699Prrr2vBggVKTU1Vamqqrr/+er3++uvq1aulZrznnntkGN6/Eh8+fFgPPfSQJCkvL09//etfdemllyo5OVk2m0333nuvHnnkEUlSVVWV2RboCndMvpWFGTH5AAAA8KVLxdmFF16ogQMHaurUqVq9erVOnTrVqX5lZWUqKSmRJM2dO1fjx49v02bChAmaO3euJKmkpEQffvih1/kXXnhBdXV1kqRFixapX79+bcaw2+3KyMiQ1HJVr7m5udOfDfBETD4AAABCpUvF2eHDh7v0Zq+88op5PGfOHL/tPM9t2LDB7xizZ8/22T8+Pl7FxcWSpCNHjuidd97p0nwBK2PyJWLyAQAA4F+XnjnbtWuX1+2G+/bt03nnndepfm4zZ870287z3Pbt232OUVBQoJycHL9jzJo1S08//bQ5RmFhYYfzA6ziKy5/Qm4qhRkAAAD86lJx1qdPn3a/9ufgwYOSpJSUFKWlpfltl5aWpuTkZNXV1XkVdIZh6NChQ5Ja9kRrj2fhRqw+Qs0dlw8AAAB0Vkg3oXYXVu0VZm7p6emqq6szCzqpJeDD/XxbR2Okp5+9auE5hj+GYejYsWMdtvOnT58+nS5SAQAAAFjv5MmTOnnyZJf7tw4jDLWQFmfuIsmzcPInPT1d+/fvN8M/PPt3ZgzP855j+FNZWakBAwZ02M6fpUuXdnoDbkQmz73M3NjDDAAAIHqsWLFCy5YtC/c0uiykxZn7ylRSUsfP3bivQtXX17fp35kxPK9ieY7hT1ZWlj766KMO23Xm/RB92MsMAAAg+i1evFh33nlnl/uPGTNGlZWVFs4oMCEtzgYPHqzKykrV1NR02NbdxrMI89xMuqMxPM93phiMi4tT//79O2yHnse9l5mV2MsMAAAg9Lr7qFFcXJyFswlclzeh7oqhQ4dKanl2rCPuNikpKW36d2YMz/OeYwCtsZcZAAAAIkFIi7MhQ4ZIaimc2nvYzjAMVVe3/MLsmbqYkpKi5ORkc4z2uPu3HgNojb3MAAAAEAlCeluj+8pXY2Ojdu3a5XdvtF27dpmpjPn5+W3G2L17t8rKytTc3Kz4eN/1ZVlZmXncegygO+ZPtmnBhcN0oKpeNfVnCzv2MgMAAEB3hLQ4mzRpkp566ilJ0vr16/0WZ+vXrzePL7roojZj7N69W4cOHVJpaakmT54c8BiAp4FJvQNq/+WRgzUuJ1Xjcrh1EQAAANYJ6W2NxcXF5kN2a9eu9Xlro2EYeumllyS1PJA3e/Zsr/Nz5swxj9euXevzfWpqarR582ZJLYkro0ePtmT+6JmGpfUNqH1OGlfGAAAAYL2QFmdZWVm69tprJUnbtm3Tk08+2abNE088oa1bt0qSrr/+eq8QEEmaN2+ecnNzJUmrVq3Sjh07vM43NTXJbreroaFBkrRo0SKrPwai2MHaeq1/36nntn5u/lmzbX9AY1RUd7w1AwAAABCokN7WKEkPPPCAXnvtNVVVVenmm2/W22+/rRkzZqi5uVkbNmzQmjVrJLVsIr18+fI2/fv06aNVq1Zp7ty5amho0NSpU3Xrrbdq4sSJqqqq0jPPPKN//OMfkqSioiLdeOONIf18iFxW7WVmdYAIAAAAIIWhOBs+fLhefvllXXXVVTpy5IieffZZPfvss15tMjIytG7dOvMKWWtz5szRo48+qjvuuENffPGFHnjggTZtJk6cqJdfflm9eoX8IyICWbmXGXuYAQAAIBhCeluj2yWXXKKdO3fqnnvuUX5+vpKTk5WSkqKxY8fqRz/6kT744ANNmTKl3TFuueUWvfvuu7rpppuUm5urxMREpaWl6aKLLtJjjz2mt99+WxkZGSH6RIh0Vu1lxh5mAAAACJY4o70Nx2KAzWaT0+lUdna2Kioqwj0dBMlzWz/XknUfdnucn11doPmTh1kwIwAAAESacNcG3PMH+FA8bogK89LNr9nDDAAAAMFGcQb4UJiXrhuKhod7GgAAAIghFGfoUQ7W1uvdz6tUU39KUssVr5zUJJUdqAnvxAAAAIAOUJyhx7AqKh8AAAAIB4oz9AhWRuVLxOUDAAAg9MISpQ9YzaqofIm4fAAAAIQHxRl6hOoTjZaN9eDVBaQyAgAAIOS4rRExi7h8AAAARBKKszNcLpfy8/N9nrPb7bLb7SGeEYKNuHwAAIDY43A45HA4fJ5zuVwhno03irMzMjMzVV5eHu5poBNax+VL0ra9R8M4IwAAAESL9i682Gw2OZ3OEM/oLIozRBXi8gEAANBTUZwhahCXDwAAgJ6MtEZEDeLyAQAA0JNRnCFqEJcPAACAnozbGtFjFY8bouljh8qWmqQDVfWqqW8kLh8AAAARi+IMPVZhXrpmXZAlSRqXwy2MAAAAiGzc1ggAAAAAEYArZ4hIvvYy21lRG9AYpDECAAAgmlCcIeJYtZdZThrPlQEAACB6UJwholi5l9mJxmZLxgEAAABCgWfOEFGs2sssPk4aPqivJWMBAAAAoUBxhohixV5mcXHSiqvYxwwAAADRhdsaEdWKxw1RYV66+TX7mAEAACBaUZwhqhXmpeuGouHhngYAAADQbRRnCBsr4vIBAACAnoLiDGFhVVw+AAAA0FNQnCHkrIzLZ6NpAAAA9BSkNSLkrIrLj5M0cXiqJWMBAAAA4UZxhpCzIi5fkh68mrh8AAAA9Bzc1niGy+VSfn6+z3N2u112uz3EM4IbcfkAAACwisPhkMPh8HnO5XKFeDbe4gzDiOlMBpvNJqfTqezsbFVUVIR7Oj2Or0TGbXuP6pWyQ50eY/ncscTlAwAAIOjCXRtw5QxBQyIjAAAA0HkUZwgKEhkBAACAwBAIgqAgkREAAAAIDMUZgsKKRMZ4kcgIAACA2MFtjYgI7kTG1L6JsqUm6URjs4YP6kthBgAAgJhBcYaIUJiXTiIjAAAAYhrFGbrFV1S+1BKXDwAAAKDzKM7QZUTlAwAAANYhEARdcrC23tLCjLh8AAAAxDqKM3TJZ/+qs6wwIy4fAAAAoDhDFyUnJlg2FnH5AAAAAM+coYsOVNcH3Gf+JJsKbAPMr1P7JmpCbiqFGQAAACCKM3RRVzaZLrANIC4fAAAA8IPbGgEAAAAgAlCcAQAAAEAE4LZGdMjXRtM7K2rDOCMAAACg56E4Q7vYaBoAAAAIDYoz+HWwtl4/XLvTsvHYaBoAAADwj2fO4FfpvmrLxmKjaQAAAKB9XDk7w+VyKT8/3+c5u90uu90e4hmFX1fi8v1ho2kAAABEAofDIYfD4fOcy+UK8Wy8UZydkZmZqfLy8nBPI2q13mDajY2mAQAAEEnau/Bis9nkdDpDPKOzKM5gap3KuG3v0U73ZYNpAAAAoHsoziCJVEYAAAAg3CjOYEkqI0mMAAAAQPeQ1ohupzKSxAgAAAB0H8UZup3KSBIjAAAA0H3c1oiAFI8bosK8dEkkMQIAAABWojiLQd1JZSzMSyeVEQAAAAgCirMYQyojAAAAEJkozmIIqYwAAABA5CIQJIaQyggAAABELoqzGEIqIwAAABC5uK0xhgxM6t3ptvMn2VRgGyCJVEYAAAAgFCjOYsiwtL6dbrugcJjG5XALIwAAABAqFGc9VOu4fEnae6Su0/0rquspzgAAAIAQojjrgayIy+/u82kAAAAAAkNx1sNYEZcvEZkPAAAAhBppjT1Md+PyJSLzAQAAgHCgOOthrLgdkch8AAAAIPS4rTHGFY8bosK8dElE5gMAAADhRHF2hsvlUn5+vs9zdrtddrs9xDPqnNapjNv2Hg2of2Feum4oGh6EmQEAAACRx+FwyOFw+DzncrlCPBtvFGdnZGZmqry8PNzTCIgVqYwAAABALGnvwovNZpPT6QzxjM6iOItSpDICAAAAPQuBIFGKVEYAAACgZ6E4i1LdTWWMF6mMAAAAQCThtsYY4E5kTO2bKFtqkk40Nmv4oL4UZgAAAEAEoTiLEt1JZSSREQAAAIh8FGdRgFRGAAAAoOejOItwVqQyksgIAAAARD4CQSJcd1MZSWQEAAAAokOPuXJWXV2tvXv36osvvlBWVpZGjhyp+Pjorz27m8pIIiMAAAAQHaK+OPvkk0+0aNEivfrqq2pqajJfz8nJ0W233abbb79dCQkJYZxh6LhTGSUptW+iJuSmUpgBAAAAUSKqi7O//e1vmj59uk6cONHm3IEDB/SDH/xAW7Zs0UsvvRQVBVrrRMbUvolyVrX9bP6QyggAAABEr6gtzv71r39p3rx5OnHihOLj43XffffpW9/6llJTU1VSUqJFixbpvffe01/+8hf95Cc/0bJly8I95XaRyAgAAADEtqh9KOvnP/+5jh5t2evrV7/6lZYsWSKbzabk5GR95Stf0V//+lcNHz5ckrRy5UodOXIkjLNtnzuRsbuFGamMAAAAQPSKyuKsqalJv/vd7yRJGRkZuvnmm9u06d+/v37wgx9Ikurq6vTHP/4xpHMMRHcTGSVSGQEAAIBoF5XF2datW82rZsXFxX6fJ5s9e7Z5vGHDhpDMrSsMo/s3M5LKCAAAAES3qHzmbNeuXebxzJkz/bbLycnRBRdcoLKyMm3fvj0UU+uSYWl9A2p/87Q82VJb+pDKCAAAAPQMUVmcHTx40DzOzc1tt21OTo7Kysp0+PBh1dTUaODAgUGeXeAOVNcH1L4ge6BmXZAVpNkAAAAACIeovK3x0KFD5nFaWlq7bdPT081jz6IukgS60XR3N6YGAAAAEHmi/sqZZ/Hli+f5uro6v+0Mw9CxY8e6PKc+ffqoT58+Xeo7MKl3l98XAAAAQIuTJ0/q5MmTXe5vRRZEd0RlceZZRCUltf+slWfBVF/v//bByspKDRgwoMtzWrp0qe67774u9Z00vP2rf60RmQ8AAAC0tWLFiojf37g9UVmcDR482Dyuqanx+rq1mpoa87i9Qi4rK0sfffRRl+fU1atmkjR0QJIWzzhPK179uMO2ROYDAAAAvi1evFh33nlnl/uPGTNGlZWVFs4oMFFZnA0dOtQ8rqqqarc4q6qqMo9TUlL8touLi1P//v2tmWAXfHfaCClOenDjx+1uRk1kPgAAAOBbdx41klpqgnCKyuJsyJAh5rFn8eVLdfXZDZ6zs7ODNicrfHfqCM0el6XSz6tVU9+o1L6JsqUm6UBVveLiRGQ+AAAA0INFZXHmeeXs/fff15QpU3y2a25u1s6dOyVJw4YNU79+/UIyv+4YOiBJV47zLsDG5XAbIwAAANDTRWWU/qRJk8zj9evX+21XWlpqxu5fdNFFQZ8XAAAAAHRVVBZno0eP1ujRoyVJb775pteti55eeukl83jevHkhmRsAAAAAdEVUFmeSzBSWkydP6tZbb1Vzc7PX+ffee0+PPPKIJCkvL09z584N8QwBAAAAoPOi8pkzSfrWt76lp59+WiUlJVqzZo0OHDigb37zm+rfv79KSkr0+OOPq6GhQXFxcVq1apUSE9kbDAAAAEDkitrirHfv3nr55Zc1c+ZMvffee9qyZYu2bNnSps2vfvUrFRcXh2mWAAAAANA5UXtbo9QSqb9161Y9+uijmjJlitLS0pSYmKjhw4fr29/+tkpLS3XzzTeHe5oAAAAA0KE4wzDa2/O4x7PZbHI6ncrOzlZFRUW4pwMAAAAgTMJdG0T1lTMAAAAA6CkozgAAAAAgAlCcAQAAAEAEoDgDAAAAgAhAcQYAAAAAEYDiDAAAAAAiAMVZBDl58qTuu+8+nTx5MtxTgR+sUXRgnSIfaxT5WKPIxxpFPtYo8kXaGrHPWQTtc3bs2DENGDBAtbW16t+/f1jnAt9Yo+jAOkU+1ijysUaRjzWKfKxR5Gu9RuGuDbhyBgAAAAARoFe4JxApXC6X8vPzfZ6z2+2y2+0hnhEAAAAAqzkcDjkcDklSc3OzJOnCCy9UfHy8XC5XOKdGceaWmZmp8vLycE8DAAAAQBB5Xnhx39ZYUlLidVtjuHBbIwAAAABEAIozAAAAAIgAFGcAAAAAEAEozgAAAAAgAlCcdYM75SXSBWOesTxmsETL54+WMYMhWj57tIwZDNHy2aNlzGCIls8eLWMGQ7R89mgZMxii5bNHy5gRxYhx2dnZhiQjOzs74L5jxoyxdC61tbWGJKO2ttbSca2eZyyPGaw1Mozo+PzRMiY/S5E/JmsU+WOyRpE/JmsU+WOyRpE/Zus16k5tYAWunAEAAABABKA4AwAAAIAIQHEGAAAAABGA4gwAAAAAIgDFGQAAAABEgDjDMIxwTyKcEhMTderUKcXHx2vo0KEB9XW5XMrMzLRsLoZhqLKyUllZWYqLi7NsXKvnGctjBmuNpOj4/NEyJj9LkT8maxT5Y7JGkT8maxT5Y7JGkT9m6zU6ePCgmpub1bt3bzU2Nlr2Pp0V88VZQkKCmpubwz0NAAAAABEiPj5eTU1NIX/fXiF/xwhzzjnnqKGhQQkJCcrIyAj3dAAAAACEyeHDh9XU1KRzzjknLO8f81fOAAAAACASEAgCAAAAABGA4gwAAAAAIgDFGQAAAABEAIozAAAAAIgAFGcAAAAAEAEozgAAAAAgAsT8PmeRorq6Wnv37tUXX3yhrKwsjRw5UvHx1M6Rwul0at++fZKk3NxcZWdnB9S/ublZe/bskdPpVEpKikaMGKHU1NRgTDWmsU7Bd/z4ce3evVsnTpzQl770JQ0aNEhxcXGd7s8aRQfWKbKxPsFnGIY+//xzVVRUaODAgRo1apT69OnT6b779u3TgQMHlJiYqLy8vID30j116pR2796tQ4cOKTU1VSNGjFC/fv268lHQjoj8WTIQVrt27TKuvPJKIyEhwZBk/snJyTEeeugh4/Tp0+GeYo/y2GOPGZKMpUuXdqr9xo0bjUmTJnmtjSRj4sSJxoYNGzrs39TUZPzyl780hg0b5tU/ISHBmDVrlvHxxx938xP1DIcOHTLuu+8+48orrzRGjRplJCUlGfn5+ca///u/G7/97W87/DlgnYKrvr7eWLZsWZvvjyRj4MCBxj333GPU1ta2OwZrFD6NjY3GhRde2Km/+1in4PnqV7/a5vvq78/q1at9jsH6BN/p06eNRx55xBg5cmSb79FVV11l7Nu3z2/f5uZm43/+53+M8847r80aTZ061di6dWuH79/Q0GD8+Mc/NgYPHuzV/5xzzjGuu+46w+l0Wvlxo0pZWZnx6quvBvynrKyszViR/LNEcRZGW7ZsMfr27dvuX9CzZ8+mQLNQUVFRp4uzhx56qMP/gf7iF7/w2//06dPG3Llz2+2flJRk/N///Z+FnzD6bNy40RgwYEC736cJEyYYpaWlPvuzTsF1/Phx8xf79v4MHjzY5/8ADYM1Cre7777b/D6193cf6xRcvv5xI5DijPUJvvr6euNrX/tau9+jvn37+vz/UXNzs3Hrrbe22zchIcF44YUX/L7/8ePHjcLCwnbHSE9PNz766KNgfhsi1o033tjpnyHPPzfeeKPXOJH+s0RxFiZHjhwx0tPTDUlGfHy88ZOf/MQ4cOCAcfz4cWPz5s3G+PHjzQW+9957wz3dHuF3v/tdp35BMQzDePPNN424uDhDkjFo0CBjzZo1RlVVlVFVVWU8//zz5tpJMv73f//X5xhLly4120yYMMF46623jOPHjxsHDhwwli1bZsTHxxuSjLS0NOPw4cNB+MSR78MPPzSSkpLM71NxcbGxatUq409/+pNx//33G/n5+ea5AQMGtPmXKNYp+L73ve+Z35/x48cbmzZtMiorK42qqipjy5YtXr/IjBo1ymhsbPTqzxqF1+uvv+71C4O/v/tYp+Cqr683v7/f//73O/yX/oqKCq/+rE9o/Md//IfX/4+2bdtmHD9+3Pj000+NRYsWmWswcuRI49ixY159V69ebfbNy8sz1q9fbxw7dsxwuVzGo48+av6/rk+fPkZ5ebnP9/csPr761a8aJSUlxokTJ4zdu3cbt9xyi3lu5MiRRkNDQyi+JRGlq8XZ97//fXOMaPhZojgLk7vuustc2Mcee6zN+draWmP48OGGJCM5OTlm/6LsrpqaGmPLli3Gt771LaN3796dKs6am5vNKwW9evUytm/f3qZNaWmp0atXL0OSceGFFxrNzc1e510ul5GcnGz+Jd36L3HDMIxf/epX5nzuvvvubn/WaPT1r3/d6+eg9ffx5MmTxve//32zzRVXXGGeY52Cr6amxkhMTDQkGSNGjDBOnjzZpk1TU5MxdepU83v05ptvmudYo/A6dOiQkZmZ2WFxxjoF3wcffGB+9o0bNwbUl/UJjb/97W/m51+wYIHR1NTUps2Pf/xjs82TTz5pvt7Q0GDYbDbzHxIPHDjQpu9f/vIXs+/Xv/71Nud37txpFg1FRUVt/qHLMAzjzjvvNMd4/PHHu/mJe7YVK1YYkozMzEzj4MGDhmFEz88SxVkYnD592qzMMzIy/N626H4+SpLx6KOPhniW0W/y5Ml+/xWlveJsx44dZrtrrrnGb7trrrnGbLdz506vc4888ki7xbdhtPxSm5GRYUgtt4T5+h9BT1ZfX28WzBMmTGjzF6DbyZMnjfPPP9/8frr/RZl1Cr63337b/P78+te/9tvuz3/+s9nul7/8pfk6axQ+TU1NxvTp0w1JxtChQ9v9u491Cj7Pn5Hdu3cH1Jf1CY2ZM2eav7T7+wfx2tpa45xzzjEkGVdeeaX5+rp168zv7w9+8AO/7+F+xikhIcGorq72Onf77bebY7zyyis++1dXV5uFw8SJEwP/kDGipKTE6NWrlxEXF+f1D4bR8rNEHGAYbN26VUePHpUkFRcXKyEhwWe72bNnm8cbNmwIydx6ksOHD3ep3yuvvGIez5kzx287z3Ot18dzDM919BQfH6/i4mJJ0pEjR/TOO+90ab7R6v3339epU6ckSddcc43fxL/ExERdeeWV5tfvvfeeJNYpFD799FPzeOzYsX7bfelLX/LZhzUKn4cfflibNm3SOeeco9/+9rfttmWdgs/9c9G7d2/l5uYG1Jf1Cb5jx47pjTfekCRNnz5dgwcP9tmuf//+2rZtm95++20tW7bMfD3QNWpqatJrr73mdc49RnJysi6//HKf/QcOHKhp06ZJkkpLS3Xw4MGOPlrMOX78uBYsWKDTp0/r3nvv1WWXXWaei5afJYqzMNi1a5d5PHPmTL/tcnJydMEFF0iStm/fHvR59TS7du1SfX29+efjjz/udD+39tbH81zr9XGPUVBQoJycHL9jzJo1y+8YPZ3L5TKPO/plZejQoeZxfX29JNYpFEaOHKkVK1ZoxYoVys/P99vO6XSax0OGDDGPWaPwKC0t1eLFiyW1FGntFdYS6xQK7uJsxIgR6tWrl5qamvT555/rr3/9q0pLS3Xs2DG/fVmf4NuyZYv5j4Vf//rX2217wQUXqKioSBMmTDBfc39/Bw4cqClTpvjt6+/7e+rUKe3Zs0eSdNlll+mcc87p1Bg7duxod66xaOHChdq9e7emTZumJUuWeJ2Llp8l9jkLA89/6ejol9KcnByVlZXp8OHDqqmp0cCBA4M8u56j9X4knd2fxL0+KSkpSktL89suLS1NycnJqqur8/qBNwxDhw4dktS59XXzHCMWnH/++Vq9erUkmf8S6I/nvzqNGjVKEusUClOmTGn3Fw2p5ZeKBx980Px63rx55jFrFHpffPGFrr32Wp06dUpz587Vd7/7XXMPH39Yp+D75JNPJLX8Q9OKFSv00EMPqaqqyjzfq1cvffWrX9XSpUtVVFTk1Zf1Cb4PP/zQPB45cqQkmfucHThwQOnp6Ro1apR69+7ts797jWw2m9+7oST/39/Dhw/LMAxJga/RjBkz2m0fS9544w2tXr1aCQkJ+vWvf91mLaLlZ4niLAzcCyup3f84JCk9Pd08PnjwIMVZCLjXp6O1kVrWp66uzqvgrqqqMv8FLtD1jSXnnnuuzj333A7blZWV6Q9/+IOklv/xjRkzRhLrFC6VlZX6+9//rqqqKn3yySf605/+ZF45u//++1VQUGC2ZY1Cz263a/fu3crOztZTTz3VqQ3CWafgc185e+utt/TWW2+1OX/69Glt2rRJr7/+uh555BHdeuut5jnWJ/g8f3nOyMjQiy++qCVLlnjdcZOYmKiioiI98sgjGj9+vFf/zq5Ramqq4uLiZBiG1/e3O78XokVTU5MWLVokSfrOd75j/q7gKVp+lritMQw8F8lz8XzxPF9XVxe0OeEs9/p0tDaebTzXhvW1TmlpqaZPn67Tp09LkhYvXqzExERJrFO4bNmyRfPnz9f3vvc9Pfzww3I6nUpLS9Orr76qH//4x15tWaPQeu655/Tcc88pLi5Ozz//fKe+7xLrFGx1dXWqrKw0vx4/frzeeOMN1dTU6F//+pfeeustzZ8/X5LU3NyshQsXauPGjWZ71if4amtrzeNf//rX+vd///c2j0I0NjZqy5Ytmjx5sn7605+arx8/flzHjx+X1PH3NyEhwfxHdtbIWqtXr9bOnTvVr18/3XfffT7bRMvPEsVZGHjeW56UlNRuW89b8dzP2iC43OvT0dpIZ9fHc21Y3+47duyYFi9erKKiIvMvwwULFujmm2/2aiOxTpGgqqpKt99+u1599VWv11mj0Pn000/1/e9/X5J0zz336NJLL+10X9YpuPbs2aOEhAQlJCSouLhYW7Zs0Ve/+lUNGDBA6enpuvTSS/WHP/xBv/zlL80+ixYtUnNzsyTWJxS++OIL83jlypVKTk7Wgw8+qLKyMn3xxRcqKyvTsmXLlJiYqKamJv34xz/Wiy++KCmw76/EGgXD8ePHzefLFi9erIyMDJ/touVnieIsDDxTgGpqatpt63m+M/8xofvc69PR2ni28Vwb1rfrTp06pccff1wjR47Ugw8+aF4x+/a3v61nnnlG8fFn/8pincLj2muvlWEYqq2t1Y4dO3TvvfeqX79+2rVrl4qLi7Vu3TqzLWsUGo2Njbruuut0/PhxFRUVaenSpQH1Z52C64ILLtDp06d1+vRp/eUvf1FKSorPdgsXLtTEiRMlSR9//LEZ9sD6BF+vXmef8unTp4/efvtt/fCHP1RBQYFSUlJUUFCge++9Vy+//LLZ7p577pFhGBo0aJD5GmsUHr/85S916NAhpaam6vbbb/fbLlp+lijOwsAzec7zgWBfPM/7+wsd1nKvT0dr49nGc21Y367ZsWOHCgsLZbfbdeTIEUnSsGHDtHHjRv32t79t8yA26xRe/fv317hx47Rs2TJt3LhR8fHxampq0t13362mpiZJrFGo3HPPPSotLVX//v31+9//3m9ogT+sU2RISEjwCtRxByGxPsGXlZVlHt9xxx1ez856mj59ur72ta9JarlaXV5ersTERPMWto6+v/X19WpoaJDEGlmlsbFRjz/+uCTpG9/4RruFULT8LFGchYFn1HRHi1tdXW0eZ2dnB21OOMu9PlVVVWZ6ki+GYZjr45nKk5KSouTkZHOM9niub3uRrD1ZU1OTlixZosmTJ5t7mKWlpWnlypXtJlGxTpHjkksu0RVXXCGp5ReWiooKSaxRKFRWVmrlypWSWiLAd+3apU2bNnn9+b//+z+z/e7du83X3fHOrFPkOO+888xj936orE/wef7SPXXq1HbbFhYWmsfuFE7PNWqPv+9vV38vjKU18mft2rXm1jzf/va3220bLT9LFGdh4PmXwPvvv++3XXNzs3bu3Cmp5QpCv379gj43nF2fxsbGduNPd+3aZab2tN4Dyj1GWVmZ+dyAL2VlZeZxe/tI9VSGYchut2v58uU6ffq04uLitHDhQu3Zs0d33nlnu3u9sE7B99///d+64YYb2uwV44tnMpY7/IA1Cr7Gxkbz+KmnntKMGTPa/PnmN79ptlmzZo35+k9+8hNJrFMkcd/KLUkDBgyQxPqEgueVs45S+DyfZ3L//Lm/v5999lm7e9b5+/6mp6ebV7zb+72wvTFi1WOPPSZJKioq0vnnn99u22j5WaI4C4NJkyaZx+vXr/fbrrS01Iz9vOiii4I+L7To7Pp4nmu9Pu4xDh06pNLS0i6NEQuWL1+uJ554QpKUmZmpv//971q1alWntoxgnYKvrKxMa9as0VNPPdVhW8+NqN3rxxpFB9YpuK6//nqdf/75uvzyy9v9ZU6SPvroI/PYvacj6xN8X/rSl8xjzz3PfPFMcXTf0eT+/p46dUqvvfaa377+vr9xcXHm84YlJSXmlaD2xkhMTDT7xKodO3bon//8pyTpv/7rvzpsHzU/SwbCYvTo0YYko0+fPkZVVZXPNv/93/9tSDIkGX/84x9DPMOe57PPPjO/n0uXLvXbzul0GnFxcYYko7Cw0Ghubm7Tprm52SgqKjIkGXFxcUZlZaXX+RdeeMF8rx/+8Ic+36e6uto455xzDEnGmDFjuvXZolFtba2RnJxsSDJSU1ON3bt3B9SfdQq+2267zfz+fPTRR37bffHFF0ZWVpYhyRg4cKDR2NhoGAZrFCk6+ruPdQqun//85+b35rXXXvPb7sSJE8a5555rSDJSUlKMEydOGIbB+oTCyZMnjf79+xuSjIsuushoamry2e7YsWOGzWYzJBnJyclGQ0ODYRiG8fbbb5vf3/nz5/vs29jYaOTk5BiSjAEDBhgnT570Or9ixQpzjF//+tc+x/j444/NNjNmzOjGJ+4Zbr/9dvP78a9//avD9tHys0RxFiZPPPGEubjXX399m78Itm/fbi5sXl5emx9iBK6zxZlhGMZ1111ntv3Nb37T5vyvf/1r8/wNN9zQ5nxDQ4ORm5trSDLOOecc47333vM6f/r0aWPBggXmGE899VR3PlpUevzxx83P//jjj3dpDNYpuH7/+9+bn/3SSy81vvjiizZtGhoajBtvvNFst3DhQq/zrFH4debvPtYpePbv32/Ex8cb/387dxMS1RfGcfwKzhioBKOOC6WCGdR8CXRAnIQoWhS4aREoiIgvu0BQ1GDUaCPhQgQnaBEE9jJEGkkL3QQigotW4koRUXOhFqGbMdKY33/xp0OTM+ZA01zr+4GzmeecM/eeZ+5lnrn3jmVZKigo0IcPH4702d/fV1tbm1mfQCAQFSc/ydfT0xO1/t++fYuKf/nyRU1NTaZPZ2eniUUiEfn9fhObnp6OGhuJRHT37l0T7+/vP/L+nz59UlZWlizLktvt1sbGRlR8f39fV69eNXO8e/fuN+796VRUVCTLslRRUXHiMafhWKI4S5GDgwNVV1eb5F25ckVPnjzRxMSEent7zQGalpamt2/fpnpz/wqJFGdra2tyuVymf3Nzs16+fKlQKKTGxkbzek5OjtbX12POMTk5afplZ2crEAjo9evXevz4sWpra02spqZGh4eHSdhje2toaDBrMDY2punp6RO1z58/mznIU3IdHByorKzMrEFBQYHu3bunZ8+eKRQK6f79+/J4PCbu9Xqj8iORIzs4ybmPPCXXyMiI2f/MzEx1dXXp6dOnCoVCGhgYkNfrNXGfz2eumn1HfpJvd3fXfOm2rP+vrAwNDWl8fFyDg4O6ePGiiRUXF2t3dzdq/Pv375WRkSHLsuRwONTR0aHx8XGNjY2prq7OjPV4PNrb24u5DaOjo6af2+3W4OCg3rx5o2AwqPLychO7ffv2H1gRe1tZWTHrcefOnROPOw3HEsVZCm1tbamystIk8efmcDjiXtpG4hIpziRpbm5OeXl5cfPjdrs1Pz9/7BzBYFDp6elx5/D5fNrZ2flNe3i6XLt2Le66HNdmZmai5iFPybW6uqqSkpJf5sXv9x/5pfc7cpRaJz33kafkiUQi6u7uNlfQ4rUbN24c+YHjO/KTfEtLS+bW0niturpam5ubMcdPTEyY2/VjNY/Ho+Xl5bjvH4lEoh5pifcZCYfDyVqCU+PHQvbVq1cJjbX7sURxlmJfv35VMBiU3++Xy+WS0+nUhQsX1N7ersXFxVRv3l8l0eJMkra3txUIBFRaWqrMzExlZWWprKxMfX19+vjx44nmWFhYUGtrq86fPy+n0ymXy6XLly/r4cOH5tmcf1FpaelvKc4k8pRs4XBYjx490q1bt3Tp0iVlZ2crNzdXtbW1amlp0eTk5JFbgH5GjlInkXMfeUqupaUltbW1yefzKTc3V06nU+fOnVNDQ4OmpqZiPgPzI/KTfOFwWMPDw/L7/crJyVF6erry8/NVV1enFy9e/PJct7q6qo6ODnm9Xp05c0Znz55VVVWVHjx4EPPW8FhmZ2dVX1+vwsJCORwO5eXl6fr163r+/Hnc5+H+NTdv3jTnta2trYTH2/lYSpOO+aN/AAAAAMAfwV/pAwAAAIANUJwBAAAAgA1QnAEAAACADVCcAQAAAIANUJwBAAAAgA1QnAEAAACADVCcAQAAAIANUJwBAAAAgA1QnAEAAACADVCcAQAAAIANUJwBAAAAgA1QnAEAAACADVCcAQAAAIANUJwBAAAAgA38BzexJRRayjxKAAAAAElFTkSuQmCC\n",
      "text/plain": [
       "<Figure size 1000x1000 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "plt.scatter(df.edge_index[0],df.edge_index[1],marker='.')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 123,
   "id": "bab47412-b3c7-403b-a53f-3beaeae16608",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "True"
      ]
     },
     "execution_count": 123,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "df[0].is_undirected()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 124,
   "id": "fc2c3a35-c5ce-470a-a3f3-e053108ad921",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "DataBatch(x=[671, 4], edge_index=[2, 3890], y=[671, 5], num_node_features=[100], num_nodes=671, batch=[671], ptr=[101])"
      ]
     },
     "execution_count": 124,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "df"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 197,
   "id": "16d00bb2-e489-440b-97e3-7b367ccaa06d",
   "metadata": {},
   "outputs": [],
   "source": [
    "import torch\n",
    "from torch.nn import Linear, Parameter\n",
    "from torch_geometric.nn import MessagePassing\n",
    "from torch_geometric.utils import add_self_loops, degree\n",
    "\n",
    "class GCNConv(MessagePassing):\n",
    "    def __init__(self, in_channels, out_channels):\n",
    "        super().__init__(aggr='mean')  # \"Add\" aggregation (Step 5).\n",
    "        self.lin = Linear(in_channels, out_channels, bias=False)\n",
    "        self.bias = Parameter(torch.Tensor(out_channels))\n",
    "\n",
    "        self.reset_parameters()\n",
    "\n",
    "    def reset_parameters(self):\n",
    "        self.lin.reset_parameters()\n",
    "        self.bias.data.zero_()\n",
    "\n",
    "    def forward(self, x, edge_index):\n",
    "        # x has shape [N, in_channels]\n",
    "        # edge_index has shape [2, E]\n",
    "\n",
    "        # Step 1: Add self-loops to the adjacency matrix.\n",
    "        edge_index, _ = add_self_loops(edge_index, num_nodes=x.size(0))\n",
    "\n",
    "        print(x[0:3])\n",
    "        # Step 2: Linearly transform node feature matrix.\n",
    "        x = self.lin(x)\n",
    "        print(x[0:3])\n",
    "\n",
    "        # Step 3: Compute normalization.\n",
    "        row, col = edge_index\n",
    "        deg = degree(col, x.size(0), dtype=x.dtype)\n",
    "        deg_inv_sqrt = deg.pow(-0.5)\n",
    "        deg_inv_sqrt[deg_inv_sqrt == float('inf')] = 0\n",
    "        norm = deg_inv_sqrt[row] * deg_inv_sqrt[col]\n",
    "\n",
    "        print(edge_index)\n",
    "        print(norm)\n",
    "        # Step 4-5: Start propagating messages.\n",
    "        out = self.propagate(edge_index, x=x, norm=norm)\n",
    "\n",
    "        # Step 6: Apply a final bias vector.\n",
    "        out += self.bias\n",
    "\n",
    "        return out\n",
    "\n",
    "    def message(self, x_i, x_j, norm):\n",
    "        # x_j has shape [E, out_channels]\n",
    "        print(x_i[0:10])\n",
    "        print(x_j[0:10])\n",
    "        # Step 4: Normalize node features.\n",
    "        return norm.view(-1, 1) * x_i"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 202,
   "id": "0603a668-d8f4-49fe-8f17-852af933142f",
   "metadata": {},
   "outputs": [],
   "source": [
    "gcn= GCNConv(4, 6)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 203,
   "id": "96703b53-5c71-4839-8fc7-659fb42265b9",
   "metadata": {},
   "outputs": [],
   "source": [
    "single_df = data[0]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 204,
   "id": "923b8bb6-7949-4a66-b31d-0cf89615ff2c",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "tensor([[ 1.4150e+02,  1.5002e+00,  1.1814e+00,  6.2828e-03],\n",
      "        [ 1.2600e+02,  2.3652e+00, -2.8564e+00,  9.9609e-01],\n",
      "        [ 1.2381e+02,  1.6738e+00,  4.4238e-01,  6.1760e-03]])\n",
      "tensor([[ 20.9799,  52.8095, -17.4519, -30.1470,  -1.8483, -66.7642],\n",
      "        [ 19.6724,  49.3646, -16.8348, -25.4185,  -0.8666, -59.0206],\n",
      "        [ 18.5045,  46.5402, -15.4916, -26.1895,  -1.4759, -58.3257]],\n",
      "       grad_fn=<SliceBackward0>)\n",
      "tensor([[0, 0, 0, 0, 0, 1, 1, 1, 1, 1, 2, 2, 2, 2, 2, 3, 3, 3, 3, 3, 4, 4, 4, 4,\n",
      "         4, 5, 5, 5, 5, 5, 0, 1, 2, 3, 4, 5],\n",
      "        [1, 2, 3, 4, 5, 0, 2, 3, 4, 5, 0, 1, 3, 4, 5, 0, 1, 2, 4, 5, 0, 1, 2, 3,\n",
      "         5, 0, 1, 2, 3, 4, 0, 1, 2, 3, 4, 5]])\n",
      "tensor([0.1667, 0.1667, 0.1667, 0.1667, 0.1667, 0.1667, 0.1667, 0.1667, 0.1667,\n",
      "        0.1667, 0.1667, 0.1667, 0.1667, 0.1667, 0.1667, 0.1667, 0.1667, 0.1667,\n",
      "        0.1667, 0.1667, 0.1667, 0.1667, 0.1667, 0.1667, 0.1667, 0.1667, 0.1667,\n",
      "        0.1667, 0.1667, 0.1667, 0.1667, 0.1667, 0.1667, 0.1667, 0.1667, 0.1667])\n",
      "tensor([[ 19.6724,  49.3646, -16.8348, -25.4185,  -0.8666, -59.0206],\n",
      "        [ 18.5045,  46.5402, -15.4916, -26.1895,  -1.4759, -58.3257],\n",
      "        [ 15.6011,  39.3766, -13.1777, -21.1171,  -1.1427, -48.1805],\n",
      "        [ 12.2009,  31.3924, -10.5387, -18.2207,  -1.5783, -39.5694],\n",
      "        [  5.4790,  15.0607,  -4.8367,  -9.7809,  -1.7688, -20.0323],\n",
      "        [ 20.9799,  52.8095, -17.4519, -30.1470,  -1.8483, -66.7642],\n",
      "        [ 18.5045,  46.5402, -15.4916, -26.1895,  -1.4759, -58.3257],\n",
      "        [ 15.6011,  39.3766, -13.1777, -21.1171,  -1.1427, -48.1805],\n",
      "        [ 12.2009,  31.3924, -10.5387, -18.2207,  -1.5783, -39.5694],\n",
      "        [  5.4790,  15.0607,  -4.8367,  -9.7809,  -1.7688, -20.0323]],\n",
      "       grad_fn=<SliceBackward0>)\n",
      "tensor([[ 20.9799,  52.8095, -17.4519, -30.1470,  -1.8483, -66.7642],\n",
      "        [ 20.9799,  52.8095, -17.4519, -30.1470,  -1.8483, -66.7642],\n",
      "        [ 20.9799,  52.8095, -17.4519, -30.1470,  -1.8483, -66.7642],\n",
      "        [ 20.9799,  52.8095, -17.4519, -30.1470,  -1.8483, -66.7642],\n",
      "        [ 20.9799,  52.8095, -17.4519, -30.1470,  -1.8483, -66.7642],\n",
      "        [ 19.6724,  49.3646, -16.8348, -25.4185,  -0.8666, -59.0206],\n",
      "        [ 19.6724,  49.3646, -16.8348, -25.4185,  -0.8666, -59.0206],\n",
      "        [ 19.6724,  49.3646, -16.8348, -25.4185,  -0.8666, -59.0206],\n",
      "        [ 19.6724,  49.3646, -16.8348, -25.4185,  -0.8666, -59.0206],\n",
      "        [ 19.6724,  49.3646, -16.8348, -25.4185,  -0.8666, -59.0206]],\n",
      "       grad_fn=<SliceBackward0>)\n"
     ]
    }
   ],
   "source": [
    "y = gcn.forward(single_df.x, single_df.edge_index)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 205,
   "id": "a88e7acc-00a9-473c-905f-7da28520aa6d",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "tensor([[  3.4967,   8.8016,  -2.9086,  -5.0245,  -0.3080, -11.1274],\n",
       "        [  3.2787,   8.2274,  -2.8058,  -4.2364,  -0.1444,  -9.8368],\n",
       "        [  3.0841,   7.7567,  -2.5819,  -4.3649,  -0.2460,  -9.7210],\n",
       "        [  2.6002,   6.5628,  -2.1963,  -3.5195,  -0.1905,  -8.0301],\n",
       "        [  2.0335,   5.2321,  -1.7564,  -3.0368,  -0.2630,  -6.5949],\n",
       "        [  0.9132,   2.5101,  -0.8061,  -1.6301,  -0.2948,  -3.3387]],\n",
       "       grad_fn=<AddBackward0>)"
      ]
     },
     "execution_count": 205,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": 146,
   "id": "5e39b662-82ad-4bdd-9b65-ffff66bb5268",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "tensor([[ 0.7772, -1.2238, -0.9028, -1.2026,  0.0569],\n",
       "        [ 0.7772, -1.2238, -0.9028, -1.2026,  0.0569],\n",
       "        [ 0.7772, -1.2238, -0.9028, -1.2026,  0.0569],\n",
       "        [ 0.7772, -1.2238, -0.9028, -1.2026,  0.0569],\n",
       "        [ 0.7772, -1.2238, -0.9028, -1.2026,  0.0569],\n",
       "        [ 0.7772, -1.2238, -0.9028, -1.2026,  0.0569]],\n",
       "       grad_fn=<IndexBackward0>)"
      ]
     },
     "execution_count": 146,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "y[df.batch==0]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 147,
   "id": "f723042a-c4f8-453f-9ddf-c27a449b3b3e",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "tensor([[ 1.1382, -1.7741, -1.3945, -1.7686, -0.0378],\n",
       "        [ 1.1382, -1.7741, -1.3945, -1.7686, -0.0378],\n",
       "        [ 1.1382, -1.7741, -1.3945, -1.7686, -0.0378],\n",
       "        [ 1.1382, -1.7741, -1.3945, -1.7686, -0.0378],\n",
       "        [ 1.1382, -1.7741, -1.3945, -1.7686, -0.0378],\n",
       "        [ 1.1382, -1.7741, -1.3945, -1.7686, -0.0378]],\n",
       "       grad_fn=<IndexBackward0>)"
      ]
     },
     "execution_count": 147,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "y[df.batch==1]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "d7088e09-6e14-462b-ad35-31d09cf6c588",
   "metadata": {},
   "outputs": [],
   "source": []
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 3 (ipykernel)",
   "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": "3.9.15"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 5
}