Neural_Network.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 2,
   "id": "839a3ff7",
   "metadata": {},
   "outputs": [],
   "source": [
    "import math\n",
    "import torch\n",
    "import random\n",
    "import numpy as np\n",
    "from torch import nn\n",
    "from torch.nn import functional as F\n",
    "from PyPDF2 import PdfReader \n",
    "from torch.utils.data import DataLoader, Dataset"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "id": "e520454a",
   "metadata": {},
   "outputs": [],
   "source": [
    "from matplotlib import pyplot as plt"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "id": "336b99cf",
   "metadata": {},
   "outputs": [],
   "source": [
    "reader = PdfReader('Shodhganga.pdf') \n",
    "text=\"\"\n",
    "for p in reader.pages:\n",
    "    text += p.extract_text()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "id": "08820e6b",
   "metadata": {},
   "outputs": [],
   "source": [
    "import re\n",
    "remove = ['\\n','\"',\n",
    " '%',\n",
    " '(',\n",
    " ')',\n",
    " '+',\n",
    " ',',\n",
    " '-',\n",
    " '.',\n",
    " '/',]\n",
    "rx = '[' + re.escape(''.join(remove)) + ']'\n",
    "text = re.sub(rx, '', text)\n",
    "text = text.lower()\n",
    "text = re.sub(\"\\d+\",\"\", text)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "id": "51638045",
   "metadata": {},
   "outputs": [],
   "source": [
    "vocab_text = set(text)\n",
    "ch2i = {ch:i for i, ch in enumerate(vocab_text)}\n",
    "i2ch = {i:ch for i, ch in enumerate(vocab_text)}\n",
    "tokens = [ch2i[ch] for ch in text]\n",
    "num_tokens = len(tokens)\n",
    "vocab_size = len(vocab_text)\n",
    "embed_size = 2"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "id": "6aba980e",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "50"
      ]
     },
     "execution_count": 7,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "vocab_size"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "id": "5b5ef2b3",
   "metadata": {},
   "outputs": [],
   "source": [
    "block_size=4\n",
    "x = []\n",
    "y = []\n",
    "for i in range(0,len(tokens)-block_size):\n",
    "    x.append(tokens[i:i+block_size])\n",
    "    y.append(tokens[i+block_size])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "id": "705736c6",
   "metadata": {},
   "outputs": [],
   "source": [
    "x = torch.tensor(x)\n",
    "y = torch.tensor(y)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "id": "5f232264",
   "metadata": {},
   "outputs": [],
   "source": [
    "train_size = int(len(x)*0.80)\n",
    "x_tr = x[:train_size]\n",
    "y_tr = y[:train_size]\n",
    "\n",
    "x_te = x[train_size:]\n",
    "y_te = y[train_size:]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "id": "5bdd5ed8",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "torch.Size([54902, 4])"
      ]
     },
     "execution_count": 11,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "x_tr.shape"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 84,
   "id": "fceeb4ed",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "torch.Size([13726, 4])"
      ]
     },
     "execution_count": 84,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "x_te.shape"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 85,
   "id": "b76448fc",
   "metadata": {},
   "outputs": [],
   "source": [
    "hidden_size=32"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 131,
   "id": "62cd5ae8",
   "metadata": {},
   "outputs": [],
   "source": [
    "ch_table = torch.randn(vocab_size,embed_size,dtype=torch.float32)\n",
    "w1 = torch.randn(block_size*embed_size,hidden_size, dtype=torch.float32)*(1/math.sqrt(block_size*embed_size))\n",
    "b1 = torch.randn(hidden_size, dtype=torch.float32)*0.01\n",
    "w2 = torch.randn(hidden_size, vocab_size, dtype=torch.float32)*(1/math.sqrt(hidden_size))\n",
    "b2 = torch.randn(vocab_size, dtype=torch.float32)*0.01"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 132,
   "id": "ba12fdd2",
   "metadata": {},
   "outputs": [],
   "source": [
    "parameters = [ch_table, w1, b1, w2, b2]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 133,
   "id": "1f633e77",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "2038"
      ]
     },
     "execution_count": 133,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "sum([p.nelement() for p in parameters])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 134,
   "id": "03635c06",
   "metadata": {},
   "outputs": [],
   "source": [
    "for p in parameters:\n",
    "    p.requires_grad=True"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 135,
   "id": "0e5c964b",
   "metadata": {},
   "outputs": [],
   "source": [
    "# lre = torch.linspace(-3, 0, 10000)\n",
    "# lr= 10**lre"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 136,
   "id": "5928ccb0",
   "metadata": {},
   "outputs": [],
   "source": [
    "loss   = []\n",
    "l_lre  = []"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 159,
   "id": "41409cf1",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "train loss: 2.789724588394165\n",
      "train loss: 2.8650522232055664\n",
      "train loss: 2.8130555152893066\n",
      "train loss: 2.8802037239074707\n",
      "train loss: 2.780942916870117\n",
      "train loss: 2.7370004653930664\n",
      "train loss: 2.8420848846435547\n",
      "train loss: 2.7296805381774902\n",
      "train loss: 2.813657760620117\n",
      "train loss: 2.882258892059326\n"
     ]
    }
   ],
   "source": [
    "lr=0.01\n",
    "for i in range(1000):\n",
    "    ix  = torch.randint(0, len(x_tr), (500,))\n",
    "    xtr = x_tr[ix]\n",
    "    ytr = y_tr[ix]\n",
    "    x1  = ch_table[xtr,:]\n",
    "    x2  = x1.view(-1, block_size*embed_size) @ w1 + b1\n",
    "    x2  = x2*(1/math.sqrt(block_size*embed_size))\n",
    "    x2  = torch.tanh(x2)\n",
    "    logits  =  x2 @ w2 + b2\n",
    "    #penalty =  w1.abs().mean()+w2.abs().mean()+ch_table.abs().mean()\n",
    "    #penalty=0\n",
    "    loss_tr = F.cross_entropy(logits, ytr) #+ penalty\n",
    "    for p in parameters:\n",
    "        p.grad=None\n",
    "    loss_tr.backward()\n",
    "    for p in parameters:\n",
    "        p.data+=-lr*p.grad\n",
    "    loss.append(loss_tr)\n",
    "    #l_lre.append(lre[i])\n",
    "    if i%100==0:\n",
    "        print(f\"train loss: {loss_tr.item()}\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 160,
   "id": "51d15d0e",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "[<matplotlib.lines.Line2D at 0x7f737d48ac88>]"
      ]
     },
     "execution_count": 160,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "plt.plot(loss)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 161,
   "id": "c67a7c02",
   "metadata": {},
   "outputs": [],
   "source": [
    "# for p in parameters:\n",
    "#     p.requires_grad=False"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 162,
   "id": "3c033b7e",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "train loss: 2.8394370079040527, val loss:2.42832612991333\n",
      "train loss: 2.780651807785034, val loss:2.4088211059570312\n",
      "train loss: 2.732079267501831, val loss:2.4461543560028076\n",
      "train loss: 2.824221134185791, val loss:2.463304042816162\n",
      "train loss: 2.7243964672088623, val loss:2.452885389328003\n",
      "train loss: 2.7451417446136475, val loss:2.3654322624206543\n",
      "train loss: 2.817979335784912, val loss:2.520535707473755\n",
      "train loss: 2.8691420555114746, val loss:2.513603925704956\n",
      "train loss: 2.813408851623535, val loss:2.482218027114868\n",
      "train loss: 2.780014753341675, val loss:2.438004732131958\n",
      "train loss: 2.839418411254883, val loss:2.439436197280884\n",
      "train loss: 2.759113311767578, val loss:2.4714303016662598\n",
      "train loss: 2.7562808990478516, val loss:2.3769900798797607\n",
      "train loss: 2.8583266735076904, val loss:2.4978694915771484\n",
      "train loss: 2.7877860069274902, val loss:2.4232540130615234\n",
      "train loss: 2.8332011699676514, val loss:2.497140407562256\n",
      "train loss: 2.793560266494751, val loss:2.43234920501709\n",
      "train loss: 2.7729592323303223, val loss:2.4630126953125\n",
      "train loss: 2.798527717590332, val loss:2.4189701080322266\n",
      "train loss: 2.76041579246521, val loss:2.4394216537475586\n",
      "train loss: 2.7758774757385254, val loss:2.4146945476531982\n",
      "train loss: 2.7608280181884766, val loss:2.575585126876831\n",
      "train loss: 2.779574155807495, val loss:2.4039766788482666\n",
      "train loss: 2.7522270679473877, val loss:2.3669400215148926\n",
      "train loss: 2.8198604583740234, val loss:2.4388160705566406\n",
      "train loss: 2.848850727081299, val loss:2.474775791168213\n",
      "train loss: 2.767798662185669, val loss:2.495978355407715\n",
      "train loss: 2.7568111419677734, val loss:2.408707857131958\n",
      "train loss: 2.796818733215332, val loss:2.4430601596832275\n",
      "train loss: 2.792067527770996, val loss:2.505610942840576\n",
      "train loss: 2.7064757347106934, val loss:2.496779203414917\n",
      "train loss: 2.8395485877990723, val loss:2.4636995792388916\n",
      "train loss: 2.853682041168213, val loss:2.4607534408569336\n",
      "train loss: 2.7871837615966797, val loss:2.395545482635498\n",
      "train loss: 2.7638840675354004, val loss:2.4539291858673096\n",
      "train loss: 2.750514030456543, val loss:2.42014741897583\n",
      "train loss: 2.7352023124694824, val loss:2.4802210330963135\n",
      "train loss: 2.7682015895843506, val loss:2.4594998359680176\n",
      "train loss: 2.886195182800293, val loss:2.417083501815796\n",
      "train loss: 2.828690767288208, val loss:2.3758833408355713\n",
      "train loss: 2.793027877807617, val loss:2.4159653186798096\n",
      "train loss: 2.782926559448242, val loss:2.4534881114959717\n",
      "train loss: 2.8589658737182617, val loss:2.3827357292175293\n",
      "train loss: 2.8095858097076416, val loss:2.4640495777130127\n",
      "train loss: 2.770965099334717, val loss:2.539536237716675\n",
      "train loss: 2.8155481815338135, val loss:2.5690386295318604\n",
      "train loss: 2.824242115020752, val loss:2.525400400161743\n",
      "train loss: 2.7985949516296387, val loss:2.4157068729400635\n",
      "train loss: 2.781947612762451, val loss:2.4949533939361572\n",
      "train loss: 2.786879301071167, val loss:2.560295343399048\n",
      "train loss: 2.811197519302368, val loss:2.501199245452881\n",
      "train loss: 2.8449220657348633, val loss:2.430163860321045\n",
      "train loss: 2.7080183029174805, val loss:2.5062410831451416\n",
      "train loss: 2.7655956745147705, val loss:2.457970380783081\n",
      "train loss: 2.8065125942230225, val loss:2.4561665058135986\n",
      "train loss: 2.7701005935668945, val loss:2.504992723464966\n",
      "train loss: 2.8419690132141113, val loss:2.47554349899292\n",
      "train loss: 2.778291702270508, val loss:2.414612054824829\n",
      "train loss: 2.809154510498047, val loss:2.2892777919769287\n",
      "train loss: 2.827815294265747, val loss:2.464949369430542\n",
      "train loss: 2.8218002319335938, val loss:2.511544704437256\n",
      "train loss: 2.8073325157165527, val loss:2.4665887355804443\n",
      "train loss: 2.8221051692962646, val loss:2.412558078765869\n",
      "train loss: 2.7902326583862305, val loss:2.5707638263702393\n",
      "train loss: 2.7675461769104004, val loss:2.5729970932006836\n",
      "train loss: 2.8449134826660156, val loss:2.424501419067383\n",
      "train loss: 2.7955989837646484, val loss:2.430664539337158\n",
      "train loss: 2.7951884269714355, val loss:2.4657680988311768\n",
      "train loss: 2.7870943546295166, val loss:2.3778536319732666\n",
      "train loss: 2.7731316089630127, val loss:2.5123300552368164\n",
      "train loss: 2.7688405513763428, val loss:2.450223207473755\n",
      "train loss: 2.8642330169677734, val loss:2.436898708343506\n",
      "train loss: 2.70872163772583, val loss:2.5406436920166016\n",
      "train loss: 2.743774652481079, val loss:2.607408046722412\n",
      "train loss: 2.7903499603271484, val loss:2.495269775390625\n",
      "train loss: 2.7648446559906006, val loss:2.439643383026123\n",
      "train loss: 2.7999494075775146, val loss:2.5166544914245605\n",
      "train loss: 2.775115728378296, val loss:2.3555986881256104\n",
      "train loss: 2.8213634490966797, val loss:2.4008138179779053\n",
      "train loss: 2.7978949546813965, val loss:2.357024908065796\n",
      "train loss: 2.799638032913208, val loss:2.427548408508301\n",
      "train loss: 2.902393102645874, val loss:2.5896074771881104\n",
      "train loss: 2.7190518379211426, val loss:2.4762139320373535\n",
      "train loss: 2.856536626815796, val loss:2.521242141723633\n",
      "train loss: 2.7602474689483643, val loss:2.4573373794555664\n",
      "train loss: 2.85554575920105, val loss:2.451354503631592\n",
      "train loss: 2.7805440425872803, val loss:2.4660775661468506\n",
      "train loss: 2.815765857696533, val loss:2.4585580825805664\n",
      "train loss: 2.7601916790008545, val loss:2.408482313156128\n",
      "train loss: 2.767486333847046, val loss:2.5372116565704346\n",
      "train loss: 2.7961227893829346, val loss:2.387974262237549\n",
      "train loss: 2.7850453853607178, val loss:2.4113447666168213\n",
      "train loss: 2.7583320140838623, val loss:2.5001871585845947\n",
      "train loss: 2.799126386642456, val loss:2.3851585388183594\n",
      "train loss: 2.876152992248535, val loss:2.480665922164917\n",
      "train loss: 2.751966953277588, val loss:2.3780133724212646\n",
      "train loss: 2.7658193111419678, val loss:2.4801440238952637\n",
      "train loss: 2.804877519607544, val loss:2.40868878364563\n",
      "train loss: 2.8066718578338623, val loss:2.3705830574035645\n",
      "train loss: 2.8468270301818848, val loss:2.484621524810791\n"
     ]
    }
   ],
   "source": [
    "train_loss =[]\n",
    "val_loss   =[]\n",
    "for p in parameters:\n",
    "    p.requires_grad=False\n",
    "for i in range(100):\n",
    "    ix  = torch.randint(0, len(x_tr), (500,))\n",
    "    xtr = x_tr[ix]\n",
    "    ytr = y_tr[ix]\n",
    "    x1  = ch_table[xtr,:]\n",
    "    x2  = x1.view(-1, block_size*embed_size) @ w1 + b1\n",
    "    x2  = x2*(1/math.sqrt(block_size*embed_size))\n",
    "    x2  = torch.tanh(x2)\n",
    "    \n",
    "    logits  =  x2 @ w2 + b2\n",
    "    loss_tr = F.cross_entropy(logits, ytr)\n",
    "    train_loss.append(loss_tr)\n",
    "    \n",
    "    \n",
    "    te_ix = torch.randint(0, len(x_te), (500,))\n",
    "    xte = x_te[te_ix]\n",
    "    yte = y_te[te_ix]\n",
    "    x1  = ch_table[xte,:]\n",
    "    x2  = x1.view(-1, block_size*embed_size) @ w1 + b1\n",
    "    x2  = x2*(1/math.sqrt(block_size*embed_size))\n",
    "    x2  = torch.tanh(x2)\n",
    "    logits =  x2 @ w2 + b2\n",
    "    loss_te = F.cross_entropy(logits, yte)\n",
    "    val_loss.append(loss_te)\n",
    "    print(f\"train loss: {loss_tr.item()}, val loss:{loss_te.item()}\")\n",
    "for p in parameters:\n",
    "    p.requires_grad=True"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 163,
   "id": "19328f3b",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAXQAAAD4CAYAAAD8Zh1EAAAAOXRFWHRTb2Z0d2FyZQBNYXRwbG90bGliIHZlcnNpb24zLjMuNCwgaHR0cHM6Ly9tYXRwbG90bGliLm9yZy8QVMy6AAAACXBIWXMAAAsTAAALEwEAmpwYAABt80lEQVR4nO2dd5hbV5n/v0d9VKY3T/OMx73EdmwnTpyQXoGEbLKhJbAsEAIsG3YDS2eXZXd/bAsJy0IIJJQlBAIJIUAS0ouJY8e9t7HH03vTjEajdn5/vPdcXUlXZVSmaM7neebRSLqSjnTvfe97vm85jHMOiUQikcx/DLM9AIlEIpFkB2nQJRKJJE+QBl0ikUjyBGnQJRKJJE+QBl0ikUjyBNNsfXB5eTlvbGycrY+XSCSSecmePXsGOOcVes/NmkFvbGzE7t27Z+vjJRKJZF7CGDsX7zkpuUgkEkmeIA26RCKR5AnSoEskEkmeIA26RCKR5AnSoEskEkmeIA26RCKR5AlJDTpjrJ4x9gpj7Chj7Ahj7B6dbUoYY79ljB1kjO1ijK3NzXAlEolEEo9UPPQAgHs556sBbAXwacbY6qhtvgxgP+f8PAAfAvBAdocpkUgWDIEpYO//AbK197RJatA5592c873K/24AxwDURm22GsDLyjbHATQyxqqyPFaJRLIQOPEs8PTfAN0HZnsk845paeiMsUYAGwHsjHrqAIC/ULa5AMBiAHU6r7+LMbabMba7v78/rQFLJJI8x91Dt97R2R3HPCRlg84YcwJ4AsBnOedjUU9/C0AxY2w/gM8A2AcgGP0enPOHOOebOeebKyp0WxFIJJKFzngv3frGZ3cc85CUerkwxswgY/4o5/zJ6OcVA/8RZVsG4CyAM1kcp0QiWSiM99HtlHt2xzEPSSXLhQF4GMAxzvl9cbYpZoxZlLsfA/C6jhcvkejj8wBDZ2d7FJK5woQ06OmSiuSyDcCdAK5kjO1X/m5kjN3NGLtb2WYVgMOMsRMAbgAQk9ookcRl54PAg5dQdoNEIiWXtEkquXDOtwNgSbbZAWB5tgYlWWCMtNHJO3gaqFoz26ORzDZSckkbWSkqmX08A3Tbf3x2xyGZfUIhYELJgJuSHvp0kQZdMvtMDNJtnzToC57JYSAUoP/nquTSvgv4j2ZgYmC2RxKDNOiS2cejGHTpoUuEfg4AU3M0r6J9J80q52AgXxp0wfb7gX2PzvYoFiZScpEIVIPO5q7kMtJOt8IRmUNIgw5Qz4g/3w+8/cPZHsnCIxQEPEOAwQQMtgAB32yPSCKYGKD9M6OfqejnRfVzV3IZaaNbadDnKGNdpN31n6CgjGTmmBwGwIGajQAPUqaLZPbxTQAPrAcOPDaznys89LIlczfLRRr0OU7PIbr1e4CRuAtqp0YoCPQezXxMCwVxUiy+mG5TkV1CIWDvz2Teei4Z6yIPeTjD82G6jPcCJhvgqpmbkgvnwKgiuUwOze5YdJAGHQgbdADoO5bZex39HfD9i2f+RJiviEyBhosBZkjNoLftAJ7+DHD06dyObSHj7qbbmW6QNd4HOCsBqwvwRXnoAR/wh78DxrpjX9f2FjA5kvvxeUfCwVrpoc9Reg4Czmr6vy9D73q4FQAHBk5mOqq5TTAAPPUp4OTzmb2PCIgW1QIljakZ9OFWuu3en9lnS+LjVqSPWTHoVYDVSZKLtid631Fg9yPAqT9FvsbvBX7yTuCt7+d+fEJuASj2M8eQBh0Aeg8DDRdSICZTD11UuQmjMxcIBiIPxGyw/dvA/keBw09k9j7Cy7GXAxUrU8tFF9+l52Bmny2Jj/DQZzp1UBh0ixPgIcA/GX5uclgZW2/Ua3ood31oBvoBigwXa6E06HOSKTcdCNXrgMpVmafOiaDOXMpR3X4f8L8XRp4cmdBzCHjt3+n/oZbM3ksUFdlLyaAPpZDpIgx69wG5qk2uGJ8tD70XcFSQ5AJEZrp4R5RteiJfI/qnZxr/SgVx7C1an77k8szngWN/yN6YNEiD3nuEbqsUgz5wEgj603+/ueahcw4cfJwCvqOdmb9fwAf89pNAQQmw6qbMvSLPAHk7JisZ9FAg+UVCnFTe0ezPPPTY81Pgx+/M/edkgyO/BX54VebphrOhoQcDZCSdVWGDrs10UT30eAZ9Bo6FkTaaPZQ1p2fQJwaBXQ/lTJKdfwZ9yg2c25G9fGUREK1eB1SuBoK++EYqMBU+qOIhPJvhOeKh9x8HBk/R/yI6nwmv/yfQewh49wNA3RY6qDMJRnkGAXsZ/V+5km6TzZJG2oCypfT/TCxTtu//gHPb07vQj/cB/7UC6NiT/XHpceY1oHM35fRngrpq0AxKLp4BAJyCohYnPTYdg+7uJj09l4y2kzRrL1faFEwzzbltB902XJT9sWE+GvQTzwE/vj57+co9h4CCUqCwhjxEIH5g9OV/AX54ZeL303roc0EOOPq78P+jHZm9V2CKtPO1twErbyQvBchMdpkYCBv0smUAWGIdPRgAxjqB5dcDzJh7Hd0zBHQqxjgdb7XnIEkE57Znd1zxyFZ8YTaWgRPOkMhyASIlF2HQx6M0dLcm6yUbTksiRs4BxQ0kEfIgMDXN36dtB2C0ArXn52R488+gV6yg24ET2Xm/nkNA9VqAMeW9WfzAaOce8t7jadH+SdrBziqSOIRxF0QHc2aCI08BdRcAYJkbdM8QEPIDjdvofukSuh3MQHbxDACOcvrfYk+e6TLWSSdSxQr6686xQW95mYJzQPLZmR5Cepupgik1vrA/s/cRBn1qbOaK7cT5IrJcgCgPfUTZrjdyTFoDn2sdfaRNMeiKEzLdwGjbDqB2E0mMOWD+GfRyxYvrz4JBDwbIG68+j+6bC8hIxTPo4jPjadHigKy/kG61skv3QeC/VwBnX8983KnSfwLoPwas+0vAtShz70UYtIISui1pAsAy09E9QzR9FVSsTLxvhcEqbqD9lmvJ5fRL4f8zMugZSiCpoC16yeRCN+UG/BNkWMFj88FzhdZDtwgNXcdDDwUi9Wt3N1BYS//nsv7DO0p/xfXpGXTfBB2vi3MjtwDz0aCbC4CSxdkx6EMtQMBL+rmgcpW+QfcMhXOm4xlGYdAbtirvrzHo5/4MgEcaiFwjCm9WvRsoqsuCQVcO3oJSujXb6H3TlVw4VySX0vBjlSvJm40XI9Ea9EXrSc7I1cwnFAJOvwgUL6b7mRj0gVOJtzv9YuYVxhP9dDwbLZllAAnvXMyGZ0p2EeePQyu56HjoQKTM4u4FFm0ADObcBkZFyqKQXIDpBUY73qaLUcPF2R+bwvwz6ABQviK5Qe89Sh54IrQBUUHlajJQ0cEV7efFM4xiLcS6LQBYZKaL0GHPvZl4TNnk6O+A+q1A4SLFoGcouUR76ABQ2pS+9+kbB4JTYckFoBMz5Ae69um/ZuQcVZQW1gGLlJlVKnrxYMv0A3y9h2ifrvtLup+JQZ/oi28Y+08AP7+NluF74R9pjdV0EAZnyeWU4pfuBVwYy/JZMOgWF0lvquQS5aE7KpRtNRdxdzfFwIrrcyu5aJ2JgjQM+rkdABhQvyXrQxPMT4NesYK8uHgG290DPLgtedFLz0HyZso1q+dVriTNNDqtSKvZxzOM4iArqqcpoFZy6dxLt1370j9hp8NgCxmk1TcrY6ojqSgTPVTXoDenL7loi4oETe8AwIAzr+i/ZqSN+nyYLOELcTLZ5cCvgP+9APjDZ6c3vtMv0u262+h2utk8nJMEIOSAeDr6q98CzHbgvPdS18/vbQXa347d7tgfKC0xHsKYrXwX3aYrR4kZTy489InB+Bk/470ktwCA2UG3Wg3dOxJOXBCzCP8kPe6qJkObSw9dXCCLF4cll+n0c2nbQfE6W1H2x6Ywfw16cCr+1Xj4HBnlZLngPYfoADGaw49Vrqbb6MBc/0k66VyLEhj0PgCMPM7SpvDne4bI62+4iLzPzt1JvmAWOKbILatvotuievrNhGyUDkIv1EokZc10UKfjvapFRWXhx+ylQM0G4Myr+q8RQSmAToySpsSG683vAr+9izILjv1+euM8/RLp9OKCP93vODlMQcWlV9F9vZlM7xHgyJPA1ruBW74P/NUfAXDg8TsjLyCDLcATHwX+eG/8HHNhcEQGUNoGXfHQVYOexdTF1/8DeORa/X4sokoUAAwG8tajs1yiDbpwolyLyNDmUkMfaQNMBXS8Wl0k8aTqoQf9JLnkUG4B5q1BF/nKcWQXcUBGV5RF03cMqFob+VhpM+2o6NTF/uOU+5zICxjvJYNkNJPOLzR0IR9s/RQApky9ckzPYRprUR3dF7eZ6OiTwzSjMdvDj2WS6SIuLlrJBSDJoONt/fapWoMOkOwST3J5+V+A579Cs5Q7f0s1BoefTG1s3lFq+LT0asBgpIvHdA26mKEtuYJkIj0P/ZV/o8Kqi/6G7jdeAtz+MzJuf/oKPRYKAU//LenjnsGwfBfNSBuN01VFF6F0A6PjvbSPxcwimx660JH3/V/scxN9gLMifF/0cwFIAvV7yBO3FYfPbWHYXVV0znkGImWao7+L/3tNF5GyyBj92UtTN+jdB2n8OQyIAvPVoJcvo9t46W2qQe/Tfx6gqZq7O2yQBCYLvX90gGrgJHksibRorYdR0kQH6NR4WG5ZchlNuc79Of64ssXQGbo4CVSDnoGOPjlMcgtj4cfEZ6QTGPXoeOgAGcBQAGiN+p2CfkpbjDDo62kmFC2HTA5TEdTa24DbfgzUXwBUrEq9v/eZ1yg9ctk1dN9WnIZBb6XbihU05ujAaNd+4PgfgIs+HTnrqdkIbLsH2P9z4NSLwN6fUh77Nd8kz/vEs/qfp73YLVqfmYcuDCeQPYMe8JGjAQB7fhIrmY73hs8fgIqLhEEXZf8FJTQ2bTEREPbQgbDTEpgCfns3SVrZYKQ98tizl6We5dKmxM5yVFAkmJ8G3VZEOmq88lmxk6MryrQIL7ukMfa5us2kd4nKwKlxOkjKV5B0MRZHi9ZqgOJ9R84BXXvJY7IV0ZSr4+3M2gukwtCZyItVVgz6UDgYJChpRNqpi6J1brRBr7+QemJH6+hjnSSlaU+q6vV0q22BDIQvohvvIA+bMWDDB+i3T5ZxAgAnnyPPuU4JYBWUpG/QixfT7C7aQ3/l38hobv1k7Gsv+wIdb7//W+CFrwONlwIXf4b6xp/8U+z2ABmcImHQz0s/A8jdQ91HbYV0P1sNuvqOkuy35hbal6dfCD/n99KFQ5w/gNJCV/G2xQW7oISMvmrQRapjddigC9mlfRd5xdlafHykjQKvgukY9HM7yMlzVWdnLHFIatAZY/WMsVcYY0cZY0cYY/fobFPEGPs9Y+yAss1HcjNcDRXL43voYyl46OJkK1kc+9zSq+kgbt9F90XpfMVyMoxBXzijRYvWwyhtotuhs0DHbiomAOiE9Htymz/tGSKPRmvQC0oo0JSRQR+JDIgC4dTFdDJdPAMk4YgUNe17NlwUq6NrswwENRvptiMqiNi5BwCLrMg773aSPpJ56d4xCj6ueU84vpLMoE+O0IxAm2453Ko0mnJSFexgSziVcKSN2sBu/ZR+kMxsA97zPXJOgj5qtcAYsPw6oO9IrOzHeayHDqRXMeruIcNjNNMxky0PvUu5yF7xFTLAux8JPyeWntN66FZnWD5RA/LF5I0L7dzdTRKpvTT83cVv0/Iy3Y62ZR4HmBonh0Z77BWUJJZcvKNA63Zq69u6PbyISw5JxUMPALiXc74awFYAn2aMrY7a5tMAjnLO1wO4HMB/M8YsWR1pNBUrKVCpl2ur1dDj5eKKq7ieh77kclrjUngQ/cpMQHjoQKxh5DzcnB9Qim5A8spEH1CjGBaxU3MpuwjtXmvQGVPSujLIAhCSSzSlS9KXXOzlkRKOoPkKumCPdYUf0zPojjI6FqJ/z47d4VmRwFUNNF9FWS+Jsn0OP0EX3fM/HH6soCQ87dfj0K9Js2/R1BkMt4aPr7JmKtYRx6aQTdbeGv896zYDt/wA+MufhtssLL9Bef1zkdtODtP7i98m1QygaDgPG3SAvPRE33s6dO2j37FsKXD+h4BTL4TPQ20OusDiCksu2gwrl+Khc06G3bWIjiFnJc3sRLLEmVfIYQAyr1tRM1yiJJd4WS7j/cB/r6I+7c99kSpDRbZUDklq0Dnn3Zzzvcr/bgDHANRGbwbAxRhjAJwAhkAXgtxRvpwOYD2PU5w0QV/8g3G4lQI/jorY52xFlL99SklbGzhB2mXpkvCUK9owTo1R0Ep4GAUlgLUoHIQTHrqzkg7oXAZGhfwRHR/INBfdMwTYdQx6WXN6HvrEYKzcIlhyOd2eeS382EibkoMedfgt3kYBTKHJck6ZRHWbY993w/uBsQ6gNUHF7t6fUbaT2GdAcg9dNF3Sjldr0EXcR8gux/9Ix3D50vjvCdCsYsX14fvlSylucTLKoAsjJo7PVDKA9BBVoqpBL0rPu+09EnuMd+6jGRVjZNAZo/iAdwxo30nbxEguOgbdWU3ZYp4hRe9XzjnGlKSFc/Rc136KoQCZL1yjOhOaGb2QXPScg9bX6Xd893eAe08Cnz8FNCfpA5UFpqWhM8YaAWwEsDPqqe8CWAWgC8AhAPdwzmO+JWPsLsbYbsbY7v7+/vRGLBCZLno9Xdw94St9PA1x5BztHD3vEACWXU153GPddHUvXUIB03hatLYPBUDvW9pIswSDmYKhgsUXU5Ak+kDwe4HvbiGDkglDLQBY7OwjkUEPBoD9v0gcd4jroTfThXO6fS08A+Rh61G1jk4YreyizUHX0ngJaa3CeI2cI+9frwHSinfShfbth/U/t+cQSQPnfzjy2BAGPZ5n3/YW3Z5VDHrQT3n/qoeuGO6BUyTPnPszsOJG/fdKxoobgNY3IrOARnQ8yEVKawS/N/WqUW0aIKAY9DQklz/8HfDYe8MFej4PGVUxUy2uB5ZdB2y/H/hWPfCnL5E3rTWYWsklOigKKDGCnkhdWqQunn0NAAc2fZhko0QL1/g8yY9d4bBEeOgJGnS1/plmGBs+GL7gzAApG3TGmBPAEwA+yzmPvmRfB2A/gBoAGwB8lzFWGP0enPOHOOebOeebKyp0POPpIHJko6dS3jE6uWs20P3ozmyC4VZ9/VywVMluOP1iOMMFoAPcWqRj0DV9KARCdqleG9mMp+FiOkmivYaTz9JnJSoeSYWhM+TFmm2RjxfVkRHVay524o/AU5+kld6f/WKsYfdPAoHJ+JKL+NzpICQXPQwGoOkyMuha3Vlvny1WmoWJjoYdSp5/rY6HbrZREPLY0/qzpL0/o5z1826PfLyghAKyen1NRtrD2Td9R+niPtpBJ7sw6K4aymEebCGpIRQAVqbZY3359TT7jL7YAWFJEKCq25FzwL9WAf9cBty3BmiJU7AlELNb4ZhYC6dv0L2jtA+8o5TFA9CFkgcjL7JXfJl+5yu/BnzgceDvjkRe4LVZLpPDNDuzuMIG3N0TDuAKRFpxyyt0ntZupnM3kYf+9GeABzYkDp527KLfVnt+J+rncu5NWgXNaIr/njkgJYPOGDODjPmjnHO9RN6PAHiSE6cBnAWwMnvD1MFRTj9otEEXhmjRBrrVM+iigk9PPxdUraGT8MSzZKiEQQf0+6KoBl1zNRbvr526A0DTpQBYrOHe/wu6bXsrsyyYoTNA2ZLYx1X9X6e52Lk3yeCsvY0a8D+wPvIAV6e8pbGvFfrudGWXRJILQMHp8R5g5w/o/vC5SA9J4KqioKNIc+zcQ1pq1Rr99932t7Rvn/tipMftnwQO/op639ijvmdBMd3qyS7CO7/0Xro9+7om6N5ItwZDONPlxB9pBql3wUmFhq1krLQ6+mg7GTvtBXfzXwM3/hcZzEs+SxLGz28Fdv84/nu7dTz0RFkufcfCswPB2TfIeButwL6f02OiFqNGY9AXnQfc8iDwjs9RsFdrLAEab8gfXofAVky/ozjHhlvDVaKCksX02Iln6Dwzmkg+i+ehT43TRWdqFPjFX+onUnBO+1j0aBLEM+gTg9QUbwaCoNGkkuXCADwM4Bjn/L44m7UBuErZvgrACgAZLmWTAnqd+dxKEC2Rh+4ZIk+rOIGHzhhV+J18lryp8mQGXZGQtEEdkekSbdCL6sg7e/tH4Smlu4cqEytWUUAuXi+TVIhOWdR+LqBfXHTuTdKc3/O/wF2vUDxAG2jUK/sXpJO6GPDRSRRdVKTlvPdSGftzX6C8ZXeXvkEHqKVv2w6qouzYTRd0bQWwFosDuOYb1GJWm/Fy9GnyKjd9OPY14nvrlf+3v0XGdP0HyACeeTXWoAN04es7RrGZFTeQcUoHoxlYfi0ZIjHbEil1ETJRMXDBx8lgXvV14KPPU7D5D5+loiU9+UjN61aMZjLJ5bH3UQWrlpaXSea4+DP0W4y0kYzlWkR9hVJFXbVoPFLuEwZcSGyuKA8doKwZEYepXEWJCSJNVsvJ5+hYv+7f6Bx+7P2xM9iRc/S7RBv0eP1cRM754ktS+prZJJUjahuAOwFcyRjbr/zdyBi7mzF2t7LNNwFczBg7BOAlAF/gnGdQY54i5UrqolYfFB56+QryOPU04ZFWuk3koQNUVCJCARWafi/F9bFeyXgvZcZoDd7ibWSgxYGlZdtnyZMQevnBx8mrede36X7rG4nHFo/JETrAEhr0KLnIO0YLZQuPomotBYG1GSaJDLrJSsZKBAZTGqdoI5DAQzeagNseoUKj398Tm4OuZfEl5El27qUTPfoiGs3a28hDfukblBX03Jdp6l22VP9EVA16HA+9bjNp+42XUmB0+Cxpwi6NAStbSil0Pnf6covg/A/T8SP6FUVX0OphKwTe/yvggruAHd8Fdn4/dht3DyULWAvDr/GO6mvwEwN04WrfCQxocuxbXibvWFwY9/+C9otIMU0VddWisciUWYuDxid6vkdr6AIRhKxcRbd6XvrRp0iyufCTwK0/otnd7z4duY2YgdVHe+iKQY/OdGn9M80Qp/t9s0AqWS7bOeeMc34e53yD8vcM5/xBzvmDyjZdnPNrOefrOOdrOec/z/3QQR66dyScwwqEjZCrmqZwelOoRDnoWkT6IhDZwKuojj5XG5Qa7yPvXOt1lS8DPv0WdYKLpn4LGfwd/0vyyoHHqIhl8UU0RWxNc4UbUW6uZ9BdNdBd6KJ9FxlLUcVmMJIhGtNIM3p9XLSc9z4KRKUqu8QrKorGZAXe92i4x3wiDx0Adv2Ailfqkhh0gwG44d/pQvydDWTc1t0G3PmUvuccz6BPjlBGh/jtmi4jo33mVRqrwRjeVmS6mB20XSY0XkLOwq6HlBz09uQGHaCL5A3/QemPL34jdoY7rgQZhadvK6IZql7cpWt/+P8Dilw4dJaOweYraTxLLqfZ1eCpSLklFbSrFk0Oh2UvgGQXUc2tvWgKJ624IXwOxOvPNDVO8YzVN9E+X/Uuks0OPxF5HLftIIlLXBgEquQS5aGf+zOdy9HB+xlgflaKCoSurb3yunvo6m110oGpJ7mI3NdEkgsQTl8sbiCvQKCnRWurRFNl2z2UQvfC1ylos/799HjjJanr6C/9M2UUCOKlLAJ0gLmqYyWXtjfJIxdVkQBdhLQGPZGHDgDn30kXP22xSCLi9XHRw+IAPvhr4J3/Hb+5UWENfWc1TTQFfbpuM/COf6CL0ad3USGPthJQSzyD3rEbAA9Px5cohrr7QOwMUGS6LL0yNmA9XRgDLvgYfU7LSyRfFcUZu95r3/0A/a6//UTkcRYdZBR5/HqyS9deAIwuZgd+SXKXqO5dcgXdbrwjLOPUTtNj1bbQjc6wcimpi0DkeAtKKBV52XXhi5JoYxCThKDILWtuCT+2+SP0nQ79JvxY21vUOkJ7cQaUBl2mSIM+OUIB4MaZl1uA+W7QxUpD2lxbd1f4iu2sjGPQWym7QhwwiXj3A8BtP4l8TDXoGsMY3YciFZZdS97DW9+jANLav6DHGy+J1dH//J3I9UEB8iK236/oy8r3FAY9npykp/+f20GVhdrfo6g28oKVzKC7qklG2P9o/CX6tOi1zk2ErQjY8rHEWQOLt5Fs5ahIzVsFgCu/AvzFD8Lec9zPL6bbaIPetkO5GCoXkPLl4eMveh9UrKTHNn4otbEl47z3kfPywj/R/VS/M0Aa+bu+TcfY9m+HH49OA0xo0PfR73bBXXTxP/s6ZZcU1oV/z5XvCv920/XQ1VWL3PoGHQhXiQoYAz7+MnD1P0U+phcYFXKLVkopqqPz79DjNPPxDJFnH62fi/eNLv9v3wmAz0pAFJjvBt1RRsY1wqBrDkhnlb5BH0mS4aKlfGns9F0vuKitEk0VxshLB2jRZXHACg1X6OgtLwMvfA148hORLYFf+w9K5eIhOjgBmvK6FkXOKKLHrpVcAlOkG0YfgIW1JF8J7XRyiC462k6L0Wz+KJ14R55K8sWh3zo3U4RXVLspfn1BuphtFJOJMehvUbaG+L0ZC8sp0ceY1Qncc4ACmtnA6qRZXa/Sxybe7CIea95DsYTX/p3aFniGYg26VTHoepkuQhdfcSMZ/n3/R7Jb8+Xh399sI4PfcHF8uS7u9xMGfYwuKFqDLpwnrTwkKG6IddYqV5KHLo7naLlFy7q/pGykrn3h9h/xmmrZyyI99NbtdJHRznZnkPlt0IHYrnJj3WHN2llNJ2BgKvI1yXLQk+GqpqmWCIyGgqTjT9dDB6j0e/Nfh1PeALpQVa6hg8PvpR7YxYvpM//wd3RQ9p8kL2Lr3RTEFFPEwRZ9uUVQVB+50EXnXtKcow/YwlrKOxcGTK/TYjRN76D0wd1xina0eAYAsOmf5IlovITes/6C7L2nlujy/4CPLobRwTIhuyST9LLBlo+F/0/n8975X6R3v/wvwH2rI6tEgfge+lg36e0155PRXnsrac/e0diKyCu/Avx1nA6RiRBGeawTAA97+kB4FpRqs6vK1TQ2If8IuWX1e2K3XX0zBbQP/ZpmYAazfpEaQJkuWg/93JvkUJgLUhtXlskPgz54mqZloVA4qAOEPWZtYDQYIA81VQ9dD4ORLhrC0/UM0VQ/HYNuNNPUV7sMHhDW0V//D5JR3v0ApZ61vEwH2mv/Th7jts/SydSxiy5UQ2fC6ZJ6lCwmAy4qGuO19RQXRaGjx6sS1cIYXZw63k7ei3uknfTzaF0yE4rqKDXvwruTb5sOBSWRaYs9B+miFz0dX3UT7ZcZKPVGxXIKPFqc6c12CkooPvHJHXQcmR2R2RnxDLpotCW23fBB5QkGNF0+/XHoIbJcRNGUnuSSskEXmS5H6djb/m06X/WklIJiyos/9Btyqmo2xDfQ9tJwlsvUOHn1IkA/C+SHQQenQIRngCLyLsUYqSXCGtllrJO2ydR7KmoIG3S9KtFMETr6G/9NU8DmK4AtH6Vg3zOfI2/owk+QURQNnvb8lPJttX3Qo1n3l6Tl/vrDlOFwbgeleEaX4KuykmLQPcOpedPr30cpW299L/42nNOFKReBo/oL4stNmRLdz0UsnBA9I7A6Kc89lRhNNnj3d4D3/jwzmalqNdUgfKVLWQZQQbTQjTHo+yh2IByR2k10XNWeH7+dw3QRkos4z/QkF2eKBr1CMeh7fgL84B2UGPHuB+I7FOtup3Opc7e+0RdoJZc/P0CO3dKrUxtTDsgDg76BbrsPaIoioj10jUEXTYwy8dABMnjDrSTn6FWJZoooZ7cWAdf+K/1vMAI3fQfwTZD3cvFn6PGSxZTWt+uHdD+R5GIrojJroxV49DYK4uitoqJ66MrJlIqHDpDRv+AuSsOMtxBDzyGaSS3LkpY8UxQURxr0vmMkA2jT5maDksV0wc8FcT30feT1WpSYCmPk6f/lT7P32QYjxWyEtBnhoU9TcnGU0fl57Pf02rtepeKueCy7Nhw/SLQohV2RXNrfJudr/QdmLSAK5INBd1XRVbr7QLgPulZDByKLi1LNQU9G+TLKqPmPJcDzX1M+L4seuqOMDPbN/xPZ3KdqDRXb3PqjSI957W3hPiOJDDpA3/0Dv6Jc8Kkx/VRAZ1VkcVF0HnAirvwqeW6/+7R+c7RTz9PtLHoyaRHtofefIK802wHYuYTJRnqy1qBzrgREN0RuW9ww/cBsMixOjeRSHH68ZDHp32JFqVS44OPABZ8APvZi8i6XZhuw5mZQTObC+NvZy8gr/81H6EJxw7dSH08OmP8GHQgHRqM9dEc5ABapoQ+fI0NVWJfZZ277LPD+X5KEIfpMZNtTu/ZfKEATzeqbI1uqApSxwJTpYyINXVB7Pi3NVrFK37sTxUVCctFbrSgeJitw68M0k3jqk7El5qdeoJlVNi+AM0FBcVhD55z6dVTmtmXRrMMYpUZqs1xG2uh4mG4aYjpoW+hqPXSjGbj9p+GFPFLhHZ8HbvyP8KwiGVf9I3DHbxLXSoi4xWg71THoLVYyg8xsK7BcUbOBFqMYOgOAhaUPo5l+8PEoD72oLvMuaEYTTdlW3EAnd9A/K5VhKs5KCo71HY1dASgeK66PvTBoKaqlmIN/kjICUpFcBBUrgOv+lTJ0dv0gvMyaZ4gCuJd+LvX3misUlFAQ1D8Zzo2uyHODDsT2c4kOiOYSbRxCm+UyEzjKk88ihbHf+qlwdtMskh8GfdF6ysU+/SIVlWibMrmqIz306eSgpwpjs2vMBTd9J7INQqYU1tDMJ1lRUTw2f5S88Re/QRe+kkYKhvLQ/NPPgcgGXeqyhCvibp43xBj0fZTKF6+bZTYRxUUW59w4x6JpvBS46X8oiDoHyB/JBSDvNLqbm7MyrKGHQuTFZ6qfz1WK6rLrNYniomR9XOLBGJXrG4zAM5+nmcypF0i6iZfXO5fRlv+L1sIieyKfEQ26BF37Ynv85wox25xp7zxVTFZafSnTVg5ZIj8MemFtWMuK1rGdGg/99AuUYtR46cyOb75SWEtSi1g2bboeOkAXmSu+QoHQI0/SPlh6dXbzz2cKrUHvP05ZEDlexX1OoF2GLuhXAqIzdEEWkks6x94CJD8MOmPh9MUYg670c+Gc+qEU1kU245HEp0hZu7P3MN2me1JdcBfNon73Gbqgzke5BYg16JV5nuEi0EouHbup+6FeS+hcIIqLUs2wWuDkh0EHwrJLtEEXXdlOvUBLlF30qfgLH0giEYsx9wiDnmaZvtEEvOt+CiiC0cIh8xFh0L0jZNAXgn4ORBr0lpepf5C2+CiXCMlFeugpkR9BUSBs0PU0dIBa1NqKSO+SpIYw6L1H6DaTk6r2fODyL1HKWzb7t8wkQscdOEkzjYWQ4QKQtBSYpN41LS9TVehMecyqQZ+hz5vn5I9Bb7yUyuKjq7pECmP/MWqAlWpKn4QuhgYTLdhgtGbecOiyf8jOuGYLq4ty/cUKNgvFoIvcarGU3Ds+P3OfbZEa+nTIH4PuKAM+/lLs46Ja1GihKjFJ6qjFRe3kVS8EvTgRjJFhEX3qF5pBP/Y0pZzORNMxgZRcpkX+aOjxcCkl7OvfH1lCL0kN0UZBnlBEQQkQ9FH1pN7SgvmIaNB19CnKC0+2Xms2kVku0yJ/PPR4WF3AX/0hvLqRZHoIHV2eUIT4HSpWLJwZi/DQuw8AK945s0kFljmehz7HyH+DDsxq97N5j/TQI9Ea9IWCtj9Jrro6xqO0iapSF9LvnQELw6BL0kf0RZcGnRDZFguhQlRgLQz/P5P6OQCUNQNf7Z2fhWizQP5r6JLMkB56JKqHvkACokDYQy9uSN6aORdIY54ySQ06Y6yeMfYKY+woY+wIY+wenW0+zxjbr/wdZowFGWPzNNlYEoFoMzxfc8ezjTDo+d42V4vFSWmrzVcunLjBPCUVySUA4F7O+V7GmAvAHsbYC5zzo2IDzvl/AvhPAGCMvRvA33HOh/TfTjKvKG2iVWPKkiwIsFBYe6uypmztbI9k5jAYgDueWFizknlKUoPOOe8G0K3872aMHQNQC+BonJe8H8BjWRuhZHaxlwKfOxku8FjolC+b2cKauUKTbGg3H5iWhs4YawSwEcDOOM/bAVwP4Ik4z9/FGNvNGNvd35/Fvt2S3GJ1yam2RDIPSNmgM8acIEP9Wc75WJzN3g3gz/HkFs75Q5zzzZzzzRUVFdMfrUQikUjikpJBZ4yZQcb8Uc75kwk2fR+k3CKRSCSzQipZLgzAwwCOcc7vS7BdEYDLAPwue8OTSCQSSaqkkuWyDcCdAA4xxvYrj30ZQAMAcM4fVB67BcDznPOJbA9SIpFIJMlJJctlO4CkETHO+U8A/CTzIUkkEokkHWSlqEQikeQJ0qBLJBJJniANukQikeQJ0qBLJBJJniANukQikeQJ0qBLJBJJniANukQikeQJ0qBLJBJJniANukQikeQJ0qBLJBJJniANukQikeQJ0qBLJBJJnpBKt0WJRCKZM/j9fnR0dMDr9c72UHKKzWZDXV0dzGZzyq+RBl0ikcwrOjo64HK50NjYCJanSyNyzjE4OIiOjg40NTWl/DopuUgkknmF1+tFWVlZ3hpzAGCMoaysbNqzEGnQJRLJvCOfjbkgne8oDbpEIpFMg5GREXzve99L67X3338/PB5PlkcURhp0iUQimQZz2aDLoKhEIpFMgy9+8YtoaWnBhg0bcM0116CyshKPP/44pqamcMstt+Ab3/gGJiYmcPvtt6OjowPBYBBf+9rX0Nvbi66uLlxxxRUoLy/HK6+8kvWxSYMukUjmLd/4/REc7RrL6nuurinEP757Tdznv/Wtb+Hw4cPYv38/nn/+efzmN7/Brl27wDnHTTfdhNdffx39/f2oqanBH//4RwDA6OgoioqKcN999+GVV15BeXl5VscskJKLRCKRpMnzzz+P559/Hhs3bsT555+P48eP49SpU1i3bh1eeOEFfOELX8Abb7yBoqKiGRmP9NAlEsm8JZEnPRNwzvGlL30Jn/jEJ2Ke27t3L5555hl89atfxVVXXYWvf/3rOR+P9NAlEolkGrhcLrjdbgDAddddh0ceeQTj4+MAgM7OTvT19aGrqwt2ux133HEHPv/5z2Pv3r0xr80F0kOXSCTT4nDnKCpdVlQW2mZ7KLNCWVkZtm3bhrVr1+KGG27ABz7wAVx00UUAAKfTiZ///Oc4ffo0Pv/5z8NgMMBsNuP73/8+AOCuu+7C9ddfj5qampwERRnnPPEGjNUD+BmAKgAcwEOc8wd0trscwP0AzAAGOOeXJXrfzZs38927d6c1aIlEMjtwzrHpX17EDWur8a+3rJuVMRw7dgyrVq2alc+eafS+K2NsD+d8s972qXjoAQD3cs73MsZcAPYwxl7gnB/VfEAxgO8BuJ5z3sYYq0z7G0gkkjnLsMePoQkfesfyuzHWfCWphs457+ac71X+dwM4BqA2arMPAHiSc96mbNeX7YFKJJLZp3VwAgAwMO6b5ZFI9JhWUJQx1ghgI4CdUU8tB1DCGHuVMbaHMfahOK+/izG2mzG2u7+/P60BSySS2aN1gAz64MTULI9EokfKBp0x5gTwBIDPcs6jM/lNADYBeCeA6wB8jTG2PPo9OOcPcc43c843V1RUZDBsiUQyGwiDPiQ99DlJSgadMWYGGfNHOedP6mzSAeBPnPMJzvkAgNcBrM/eMCUSyVygdZD6kEz4gpj0BXPyGaMeP/7fs8fgC4Ry8v75TFKDzqiH48MAjnHO74uz2e8AXMIYMzHG7AAuBGntEokkjxAaOpA72eWN0/34wWtncKBjJCfvn8+k4qFvA3AngCsZY/uVvxsZY3czxu4GAM75MQDPATgIYBeAH3HOD+ds1BKJZMbhnOPswARqiij/fDBHsovbG8jp+2dKut0Wb7zxRoyMjGR/QBqSpi1yzrcDSNppnXP+nwD+MxuDkkgkc49hjx9ubwCXLa9A18HunHnobq8fADA0MbcN+qc+9amIxwOBAEym+Cb1mWeeyfXQZOn/fMQfDMEflPqiZGYRcsumxSUAZsJDn5uZNNr2uVu2bMGll16Km266CatXrwYAvOc978GmTZuwZs0aPPTQQ+rrGhsbMTAwgNbWVqxatQof//jHsWbNGlx77bWYnJzMythk6f885DO/2IcCixHffu+G2R6KZJYJhThO94/DZGBYUuHM6WeJDJfNi0sBAIM58qBVg57K+z/7RaDnUHYHUL0OuOFbcZ/Wts999dVX8c53vhOHDx9WF3N+5JFHUFpaisnJSWzZsgW33norysrKIt7j1KlTeOyxx/DDH/4Qt99+O5544gnccccdGQ9dGvR5yJmBcdgtctctZHadHcIPXmvB7nPDGJ30o6rQip1fvjqnn9k6MAEDA5ZXO1FgNubMgx5TJJdcXTCyzQUXXKAacwD4zne+g9/+9rcAgPb2dpw6dSrGoDc1NWHDhg0AgE2bNqG1tTUrY5FWYR7i9gYQStyCZ8HjC4Rw/QOv4+OXLsH7L2iY7eFknZ+8eRY7zgzipvU16Bnz4rWT/QiFOAyG3C2e3DroQU1xAawmI0odlpxLLkOpaPQJPOmZwuFwqP+/+uqrePHFF7Fjxw7Y7XZcfvnl8Hpj2yRYrVb1f6PRmDXJRWro8xC3N4Bx5aCX6PNmywDO9E/gxaO9sz2UnDA2GcDKahe+det5uGRpOTgHxn25PSZaByfQVE7Gq9xpwUDOJBfFQ5+jWS6JWuCOjo6ipKQEdrsdx48fx1tvvTWjY5Me+jwjGOIYnwogWZfMhc6fjvQAAPa1j4BzDiqnSIzHF4A/yFFUYM718DJmzOtHid0CAChUxjvq8aPQlpuxi5TFmzfUAADKnFb0uXPToGtaGvosoG2fW1BQgKqqKvW566+/Hg8++CBWrVqFFStWYOvWrTM6NmnQ5xnCM5/wBREMcRhzOMWerwRDHM8f6UWB2YihCR/ahjxYXOZI+rovP3kI7cOTeOKTF8/AKDPD7Q2godQOAOoFSGjPuUCkLDYqv2OZw4Jj3dldy1MQllx8OZeR0uUXv/iF7uNWqxXPPvus7nNCJy8vL8fhw+Eync997nNZG5eUXOYZ2pN2IsdT7PnKrrNDGJzw4WOXUqBqX9tISq871DmKQ52jCM6DAIXb61c9c+GVj07mzqCLlEVh0EudpKHnYqY45vWDMbow5/IilY9Igz7PcGu0c6mj6/Pc4W5YTQbc9Y4lsFuM2Nc2nPQ1wRBH+9AkfIEQ2oc8MzDKzBibDMBlowl2YYFJfSxXiJTFRqGhO6zwBUNwT2X3MznncHsDqCkqACDb9E4XadDnGW6NxzKe5ZMpHwiFOP50pBeXLa+Ay2bG+rpi7GsfSfq6rpFJ+JRirdN94zkeZWZ4/UH4giHVM1cll1x66ErKYn0pGdoyJ+n32e66OOknKbGxnOSkuVotOleRBn2eofXQ3XI6GsP+jhH0jHlxw7pqAMDGhmIc7RqD15+4M+CZgXDTqdP9c9ugCxmiUPXQc6+ha1MWAQqKAtlv0CWOb5FNEy/XfSEkBaTzHaVBn2doT1q3lFxieO5wD8xGhitXUubBxoYSBEIchztHE75OSAo2swGneue2QRf7XRhyp8UEA8uxh65JWQQoKApkXxIRTorQ6vUyXWw2GwYHB/PaqHPOMTg4CJttegtxyyyXeUaEhi4llwg453jucA8ubi5XZYgN9cUAKDC6ubE07mvPDkzAYTFifX3x3PfQFcMtNHSDgcFlM+c2KDowgZuUlEUgLLlkO1d8TDm+RVaS3vvX1dWho6MD+b7qmc1mQ11d3bReIw36PCNCQ5ceegR97im0DXnwkW2N6mMVLivqSwuwrz1xYPTswAQayx1YVunEE3s7U85dnw1UD12Tc15UYFaNYbbx+AIY8wZQU1ygPlbqEAY9N5JLqcOMQptJV9Ixm80Rpfbp8MSeDkz4AvjQRY0Zvc9cQ0ou84xIDV0adC0iO6WxPDLnfGN9SdLUxdZBMuhLK50YnwqgZw6vai/2u0tj0AsLTDnz0Afc5CVXOMPl6laTES6bKevFP8JhcdnMKHdac1Zc9Ku32/GLnW05ee/ZRBr0ecaYN6B6R9lOGZvvdAxTP4z6koKIxzc2FKN71IvuUf1+Gb5ACB3Dk1hS7kBzJXUsnMuZLmPeSMkFIG89Vxp6/zhd3Mpd1ojHyxyWHBh0cbEyKf1ictMAbNjjy2nMYbaQBn2e4fb6UVxghtNqkpJLFB3D5KHXldgjHt/YQP2798fx0tuHPZQqV0YeOjC3DbrwYgsLIiWXXHno/ToeOkCZLtmXXMIeepnTkrO0xWGPP6cxh9lCGvR5httLBSVOqwnjU/l3QGZC+9Akyp1W2MzGiMdXLyqExWTA/jj56CLDpanCgQqnFUUF5jlt0McmAzAwwGEJf89CmzlnaYv9itGu0PPQs57lEv5upQ5rThp0cc4x4vFhwhdEIM8WilnQBv14z9i8qArUMub1w2Uzw2UzySyXKDpGPKiLklsAwGIyoLa4AF2j+rr4WWHQyxxgjGFppROn5rBBdyvHgDZoW2Q356xSdMA9BcbCgVBBmdOakzx0p9UExhjKnRYMe3xqKwZ/MIT7XzyJ4Qy99vGpAALKe+ZbHGpBG/SP/2w3rv3263j87fZ5k9Oqeug2U94djJnSMTyJ+lK77nNlCfTYswMTKCowo0QxWEsrnGiZwwZ9zBuI0M8BKjKa9AfhC2Tf4+wfn0KJ3QKzMdJclDksagOtaPzBUFpjEQ4LQBeQEAdGPGTA324dwv0vnsIfD3Wn8S3CjHjCM5l8k10WtEEfcPsQ4hz/8MRBfOaxffOiERB5ZyS5SIMeJhji6BqZ1PXQAcqbjjd9jy6aWVblxOCEb86Wnbu9sW1yc9lxccA9FaOfA/SbhjgwomMU//HpI/joT9+e9me5NRcrUY0q9sOBdioOa9VU9abDsCe8X+fDOT8dFqxB9wVCmPQH8anLl+Lz163As4d78MUnDs72sJJCB7yUXKLpHfPCH+SoL4njoTutGIjnofdHGvTZznR58/RAws/W9dAL0uu46PEF8PeP70/Y27x/fArlLkvM42r5v87veqhjFGf6p294xybDF6voatT9Si3B2YwNuvTQ8w4RTS9xmPHpK5bijgsb8PLxPkz6Evf8mE0CwRA8viAKbTLLJRqRshjPQy93WjGk0WMFXn8QXaPeCIO+tGL2DDrnHHf/fA++9eyxuNuMTfojMlyAcJHRdFPxDnWM4sm9nXjleF/cbQbG9T308gTl/x3DnrR6DUV66EoDsCgP/UyGBn1E66HnsEPlbJDUoDPG6hljrzDGjjLGjjDG7tHZ5nLG2ChjbL/y9/XcDDd7jGnyXQHg6tVV8PpD+PPpgdkcVkKER+6ymeCymXPWnMsfDOF/XjqFrpHsrHM4E4jgdnyDbgHnkdNtADg3GFuMVFtcgAKzUdeg728fwRd+c1BXNxZwzvG6ssbndGkd9GDMG8B+ZaUlPdxZ9NCFXBLPm+aco989hXIdg14aZXAFE1MBDHv8GJ8KTPs3cE/51e+mVqNOTKFn1IueMS+K7Wa0DXngzyA7RauhL0TJJQDgXs75agBbAXyaMbZaZ7s3OOcblL9/zuooc4Cay6t4Nhc2lcFpNeHFY3N3DUrhTQgNXaxalG32tY3gv184iVu//+acTt/TIjz02ngauoMMUrTscnaAvl+TZkUjg4GhudKh29Pl2UPd+NXu9oRe4vbTA/jQI7vwzOHpB+8Odowo4/ShM84FdUxXQ1d6ok9z1jaqGLeWOAZ9wheE1x+KSVkEwr9pdKaLGHeIT38RFiEpAkCpPdwv5oDyu7xz3SKld3362Wnai/qCk1w4592c873K/24AxwDU5npguUYYR+HZWEwGXLaiAi8d70vLs9KSq4yZMU3RhfBicrFq0RnFkI1PBXD7D3aoRmYu0zHsQVWhVW3vGk28ZlJnB4SHHqm9x8t0EReORN0bD3bQc88e6klx9LGvBcISg5aQsqZsYdY8dPo9zsRpSNbvJmOt56GX2M1gLFZyEQVewPTSAsXiFuLYNhkNKLabMTgxhQPtIzAZGN51HjUIy0RHH/HQLMBkYAklqh0tg/j7X+2fNxlwwDQ1dMZYI4CNAHbqPH0RY+wAY+xZxtiaOK+/izG2mzG2e7Y7pemVT1+9qhL97ikcTNJqNRFvtgxg3T89HxMoGprw4Y4f7YzrdaVCuCkTeejA9Bt0jU768fsDXQm3OTMwAYvJgKc+vQ0FZiPe/9BbOVs/Mlu0D3tiKkS1CIMU7aG3Dkyg3GmN6IsCULe/rtFJTAUiYyrCWCUy6OK5V070Je3DHs2hjlGsrRWFULENxSZ8AXCOmPGmq6EL+SGejDEQp6gIIINbYo9NB+0cDh/j0zHoYnEL7XcTqZH720ewalEhVi1yAcjMoA97fCixW5SGZvF/rxeO9uLJfZ05a3qWC1I26IwxJ4AnAHyWcx59du8FsJhzvh7A/wB4Su89OOcPcc43c843V1RUpDnk7CAOfO3U9YoVlTAaGF48mr7scqzbjfGpQExhyp5zw9h+eiAjjV5bFu1ULkTTTV287/kT+Mxj+xJOWc/0j6OpzIHmCiee+OTFCHKOX73dnva4Z4KO4cmYHi5ayuN56IMTaCqPvRA0lNrBeaRxEp8D0Pqj8TjcNYqqQis8viBeO5m64xIMcRzuGsWmhhKsqSnU9dCFcRHLzglsZiOsJsO0NWGhoQdCHG06x4Tw0PUMOkC/a+9YpEHviDDoqY/HHRXXAiiTpt89hYMdo1hfX4RiuwWlDktciSgVhj1+lNjNKCwwYzRBULRXadA2V9NX9UjJoDPGzCBj/ijn/Mno5znnY5zzceX/ZwCYGWPlWR1ploleJAAAiu0WbF5ckpGOLqrYog2muJ+JZ+HWnMzCi5lO+f+Ix4fHd3cAALrjVE0CFCBbUkGacnWRDRc3l+PVE/GzIGabQDCE7lFvQg+90GaGycBiPPTO4UndVMfFZfTYOc1+9PgCGJzwwWhgONo1pivNjXr8aB+axB0XLkZRgRnPHU5ddmnpH4fHF8R5dcXYUF+MQ52jMaXp2ot6zHcsmH6DrlFNgFBPYhK/l57kAgDNFU6c7nNHPNaRpofu1pk1lzksONgxivGpADbUU0+epnKHGvtIhxGPD8V2S9LfSzRzG8pyNWwuSSXLhQF4GMAxzvl9cbapVrYDY+wC5X0HsznQbDPm9cf0wwCAa1ZX4XiPO+2gy5AScGmP8uzalan62Qw8iwgP3Tp9D/3RnW2YVCSAeJ0H/cEQ2oY8qkEHgMtXVKB10JNx/m+u6B71IhjicTNcAAp0lkb1HgkEQ+gZ80b0+RY0KBWn2uNAeOtbl5TCPRWIMPaCw13kVW9oKMY1q6vw4rHelCsmhX5+Xl0RNtQXY9IfxMmo1ZPU2I+eQbeZpp2GN+zxYZmSd68X6O13T8GgU/YvWF7lwrkhT4S01DEyiZoiWmlnOjMGdfahlVycFkwpv9+G+iIAwqBnKrlQv/VE4xMzj1z0k8kVqXjo2wDcCeBKTVrijYyxuxljdyvb3AbgMGPsAIDvAHgfn+ORhLHJ2H4YAHDVKlq67KU0vXThoXfEeOhkDFoH0z8QtamWwotJtbhoKhDET99sxfkNxQCAnjgeetuQB4EQx5Jyp/rY5csrASBhrvJsorbNjVP2LyiP6j3S655CMMR1M2MqXFbYzAa0DYb3o/ic69fQeqV6Orp4bE1NEW5YWw23N4A3W1KT2Q51jMBhMWJJhRPr64oBQM3uEOh5sYJ0Oi6OePxoKLWj3GnVDYwOjE+h1GGF0aC/2MeKahc4j8zZ7xz2YNWiQmW80/HQYyWXUiWTxmk1qcdkU7kDvWNTmEizsG5kwq966PF+r1CIq5JLtlsE3/zd7Xhk+9msvqcglSyX7Zxzxjk/T5OW+Azn/EHO+YPKNt/lnK/hnK/nnG/lnL+Zk9FmkTFvIEaHBOhgaa5w4KU0jZfQ2zpitNew5JJuFo3b64fNbIDZaJh2UPTp/V3oc0/hs1cvh8tqiiu5iHxkrYfeUGZHc4UDr2Qou3QMe3C0K/vB1XDb3PgeOkDenjYjQ3jctToeOmMMDaX2CC9cfM4VKythMRr0DXrXGGqLC1DqsGDb0nI4LMaUZZcDHaNYU1sEo4FhcZkdxXYzDkR1iNQL5gsKkwT59BidJOO2pMKhq0v3u6fi6ucAeegAcKKHZJdJXxAD4z7VoE9nPHptgUXs47y6IhiUi0qzcmym46X7gyG4pwLhoGicGc3AxJTawEsvz/7nb53TzX7Z0TKY8Pz2BUI40DGa9sUoGQu6UlRv2goAW5eUYX/bSFqGV+z8dk3qFueUN+uwGDEVCKE7zdVwtDm6rmkERTnneHj7WaysduHSZeWoLrLF9dCFl7akwhnx+BUrKrHz7BA8GaRJ/ssfjuG9D+3IesuC9uFJMAYsKkps0Mujyv87R2gfxctdbyi1R0guHcOTsJgMqCkqwMpFLlVe0XKkcxRrasiY2cxGXLmqCs8f7U3aptUfDOFo9xjW15GswBjD+rrimJa/erEfQXoeug/FdjOaKxy6Hnr/uE81qno0ltlhMRpwspcMusjiWlrphNnIsuCh02eLtWEBoKk8vkSUDJHVU+Iwq4uC6Blm7fkRLbk8e7gHX33qMI51R8YOjnaN4f0/fCthDE7Yh9IEv2kmLFiDPjYZiGvQNzaUwD0VQEsaiwWLooWeMa+a8jbs8WPCF8RFzRQnTre5kDZH12FRDHoKxnH76QEc73Hjo5c0gTGG6iJb3IvKmf4JlDstarMnweUrKuELhLCjJf3QyMleN9zeAH65K7tLf3UMe7Co0AaLKfHhXB7VoEt46DVxLgT1pXa0DXnUE75jeBJ1xQUwGBjW1BThcOdYhDFwe/04MzCBtbVF6mM3rK3G0IQPu88lXtP0RI8bvkAI6xSpBQDW1xfjZK87wpvTM3qC6a5a5AuEMOELorjAjCXlTgx7/DGtaQeSeOgmowHNlU6cUAy6drY03WpmvYBvdSFp8WKREoAC1oylF48SZf8kuZjgC4ZUjV6L1qBHB0WFFBPd/0ZUVie60AjJTxRlZZuFa9C9ft2TAtCsFB9nQQQAODc4gSf2dEQ8FgpxDHv8qC60gXOga4R2uPDy3rGcDHq6vSi0FYIGA0u5n8szh7rhsprUVdsXFdnQEycoemZgPEI/F2xpKoHdYkxbdvEFQqp88fD2s1lt89oxPJkww0VQ5rRi0h9UZxmdI16UOSwosOgXIy0utcOjSAgAzbqEN7+utgijk/4IaU14bGtrC9XHti4pA5A4bx0Ip0EKDx2gIGCIR752bNIPq8mgW0AlFopONXwlvPliuxnNlSRjnNFkj3DO0R+nj4uWFVVOnOwRBl301LHDNc0Wz9rFLQSbFpfgJx/ZgqtWVqqP2cxG1BQVpJXpIhpzldjN4Q6VOhdBsaZsbXFBjIbepzwnUjoFYiEQvfRPgfDQy6SHnl30GhwJlpQ74LKZEi4s/O0XTuLeXx+IKDxxewMIhjjOU05K4a0I+WXz4lIUmI1pZ7pE9/BIddWi7acHcFFzmWoEqosK0Oee0i0k0aYsarGajNi2tByvHO9Pq3Lu3OAEgiGOWzbWonvUm7S4aTp0DOkvbBGN2r1PWVKtc2QyrtwCUOwACJ+g2guHMNrafHRheNfWhI1yqcOCErs5ad70wY5RFNpManYNADUwqpVdxjSyWzSFBSYEQxwTKTaYG1WqRIvsFvUi3tIXHueYNwBfQL/sX8vyahe6Rr0Y89IFzmxkqHRZp23Qxyb96uIWAsYYLl9RqerngiUV6WW6iBl0id2iOkd6MlXPqBcmA6NWylGSi8h+ia6QFQZeG0iPRrxXWZysoUxZsAbd7Y0vuRgMDBvqY/VLQSAYwqtKwYj2Ki1SFtcrHr7IbBG3DWV2NJY70s50ie7hkcoiF22DHrQPTeKSZeGygEVFNIOI9jBGPX4MTvh0DTpAOnrnyGRaUpTIgvjrbU1YUeXCD15vyUpJtS9AqYd1STJcAE21qDLt7Rz26AZEBQ2l9Du0DU1gYiqAoQkf6ktp+xXVLpgMLMJ7Ptw1igqXFZWKTCBornAm/c0OdozgvLriCGNW5rSivrQgItNlzOvXDeYDiatF24c8+MJvDkakFwo9ubjAjLqSApiNDC0arzdR2b+W5ZUUGD3V60bnyCRqFFnKZZ2u5BL/YhXNknIHzgxMxBxDgWAI9/xyH7778indGFhYcjGrDp1e4LZn1IuqQhsqnNaYoKiQWmI8dHdyD114+1JymQZ9Y16863/eiJFEBMEQh3sqtmOdlo31xTjRM6YbBNzbNqKeDNoqOaG1rVpEJ3u7xkMvsVPueFO5PcKz8PqD+OhP3saec0NJv1e0h55KT/TtSmXqxc1hg16t5AhHZ7qIk1lPcgEoHx0APvzI2/jn3x9NGtHXIgx6c6UDn7hsCU72juPVE5m3f+gZ9SLEgboEhllQrvbv9oFzTh56gtcJr79tcFIN9gkP3WoyYnmVK8JDP9I5hrU1hTHv01zhjNsrRXCmfwIrql0xj69eVIjjmuBbIqNXlKCfy693t+NXu9vVbBRAKz9YYDIa0FjmiOi6mKjsX4sY94mecXQMh2dL0/bQdbpIxqOp3AG3NxDjJf9o+1n8bn8X/uv5k/joT9+OKJwCIr9zWHKJHWPPmBfVRTaUKgtVay8c4pzvH9c36J0jk3GD4IPjUzAZWNyLcqbkpUG/74WTONw5hs//5gCePxKbMjaeIFNAsKGhGCFOvTWieel4OIrdN6YNntDBUu60oqa4QNXO24c8ao50U7kjom/Gmy0DeOl4H363P7kE4Y7S/VNZtejPpwdQXWhTU72AcKApOtNFL2VRS01xAR68YxOWVznx853n8P4fvoX/feV00nEDVAVZW1wAu8WEd6+vQU2RDT94vSWl1yYibGhTkFzU8v8pDE344PWHdIuKBDazEdWFNrQNeXRTI9fVFuFw5yi8/iAmfUGc6nNHBEQFzZUODIz7IvpwawkEabGV6EA0AKysLsTZwQm1Tz8tABHHQ0+gCb9+ii7s2ou41lsFoKQuxnroyQw67VcjTva60TEcvki6phmkTZR5Fk2TkoWldY7O9I/j2y+cxHVrqvDN96zF9tMDePd3t0dUsg57fLAYDbBbjOrvGE9yqS60ocxhgS8YUh0n0U4YAPqjgqLCwAdDPG5a8NCED6UOS0z9S7bIO4N+vGcMj+9ux/svaMC6umL8zWP7sPNMZGaGmGLFOzEAff1S8PKxPqxUvJJejUEXGQIldgvqSwvUAFGHpry8qdyJYIirz72g9I2JJ+8I/MEQvP5QhHeWzEMPhTj+3DKAbUvLIw6gRaqHHhkYPdM/DpOBJSzQuX5tNX78kQuw72vXoLHMrpu6p8fp/nF1JSCz0YDbNtdj59mhjHu6i8yCRIZZUKouyDClBqwTaegAyWRtQxO6C2hsaCjGsMeP1V9/Dtfe/xpCnAqKomlWjE/cFrVTZKxFbYGWlUrhzinFKCUyevE89FGPX+2YqQ2Gi+2KVIPuRNtg2NlIVvYvMBgYlimzlX73lDqLKSyYflA0VQ99idK//kdvnMHg+BRCIY4vPnkIFpMB37x5Le7cuhi/vGsrRif9+K8/nVRfR0VFVFAYT3LhnIc9dEd4VgeQTOVTfh89yaWqkLaPJ7sMjPviVt1mg7wz6P/2zHE4rSZ84foV+PFfbUF9SQE+9tPdEVfpMZ0ChmjKnFY0lNpjAqNtgx6c6hvHbZvqYDIw9Opo6KUOC+pL7OgY9iAU4ugcnkSdor2KRlBnB8YRCnG8cJSyRo52jSXszKfttChIluVytHsMIx4/LllWFvF4UYEZNrMh4mIEkIfeUGaPWQxYD4fVhGVVLt3A1D89fQTfeemUej8U4mjpm4iYJWxpLAHn+i1i9egd8+LTj+6NMVbCoAsZKRE2sxEuq0npNa7koCe5EDQoqYsdw5OwmgwRGR+3barDD+7chL+5YilWVLmwsaEYFzaVxrxH2KDryy5uJbDt1DFmQs443iMMenyjp2roUcfEmy0DEMqYNl11xOOH0cDgUi4kzRVOBDS9xvvdJA8UJzhP1HFWObGvjVIzw5KLGeO+1Be50C5ukYz6UjvuvWY5XjnRh6vvew2f+/UB7Do7hK++c5Uaw9i0uBSXLiuPcDpEp0UgfsxhzBuAxxckD13M6hRnrU+NK1h0g6KbFlN65bk4gdGhCf3FQrJFXhn010724/WT/fjbq5apXdl+9tELMeEL4OkD4cUGtAtFJGJjQ2xgVMgt16yuQqXLGuOhW000nasvtWNg3IfWwQn4gqEIDx0g47mvfQQD41N413mLEAhxHEng7erl6DqTBJ2Efr5No58DlDmwqKggZloYL2UxHk3lDrQOeiJOWM45ntrfiZ/taFUf7x7zYtIfxNLK8HtvqC8GY9SFMhV+f6ALfzzUHRNr6BqdRLnTCptZP/UwmnKXFYMTvqRL1gkaSu3oHZvC6b5x1JYURMx0zEYDrltTjb+/dgV+9OEt+O2ntqFEx/uqL6Xim3gGXV2JSsdDX1zmgM1sULVvCopOz0N/4/QAnFYTaops6NVKLpM+FBWE218IqU181sD4FMqclpgMEz2WV7nUi4bqodtM4BwYT7EYbTpBUQD4zFXL8Me/vRTNFU48ua8TFzeX4fbN9RHbrK4pRMfwpKqlj3j8qsRkMRlQYDbG/F7inK4usqnZKCIwKp5bXUNpqyLLbWIqgEl/EGtqimAxGuJ66EJyyRV5Y9CDIY7/98wxNJTacedFi9XHa4sLUOGyRkw1w5JL4oNnQ30xesa8EVrzS8f6sLTSicVlDlQW2iKmXYMafUwYih2K3CNkDJH/2jo4geeP9sBsZPi7a5YDQMI0Sb2CEpct8apFfz49gOVVzpisC4B0dO33CoY4Wgc9EV50MhrLHPAFQujS/LaDEz6MePwYGPfhqNJDXQREl2qqT102M1ZUubCnLTWD/pbyO0Z7Pp0jXtQWJ/fOBWUO6t/dOTIJu8Woq1trEV0Xd50dSinXXQ+jgaGp3BGREqhFzLL0PHSjgWF5lUstPPL6Q7qGX/t6rccplsO7qLkMdSX2KA3dH+F9r15UiOpCG/7zTyfg8QWSlv1r0QZ0azVBUSD1aubpSC6C5VUuPP6Ji/Dwhzfjux84P0abFhLYkW5yloaVylhBYUFsQzPxG5HkIgw6nefCQxfVwMJLF3agutCGupKCuM39BqXkkhq7zg7heI8bf3/N8piii+pCW8SBLA6wZCezKDASCw24vX7sPDuoFjlUFcZ66GI6J07+N5XKStGrmzGmdot74Wgvti4pQ3OFE7XFBQkLmcRJGq2hA/qrFnn9Qew6O4RtS/W7GC8qivxNOocn4QuE4gZE9RALK7cOhA/eU5rugKIXeIua4RLp/W9sKMG+tuGkU/JAMISdZ8gzj/Z8upQ0uVShfi5T6FSCd8mCU+JCPD4VSCnwGo/mSv3SeiBc7aunoQPAiioXjve4dXudaBHyiVYTPjdIcpHa8kFzvFIfl/B72cxG3Pfe9Tg7OIFv/uEoBsZ9KcsDK5SeLiYDQ5VyERDHaipxEr3FLVLFYGC4alWVrqEUhlf0EBqZ9KvnKKDfLkHMYigoKhZGifTQxfuqAVJNRpCoMI5mKhCEeyqQsJVCpuSNQd/RMgADA65cVRnzXHTvEr3FLfRYXVMIi9GAfW0j8AVC+MPBbviDHFeqBt0WmbboCV99Rb7yWy2DYCwy+NZU7sCec8M40z+Ba1dTd8cNDcXYH+Wh94x61bTJ6EWtASRs0LX33DCmAiFcEsegVxfZ0DvmVY2p0BmXVcWmzsVDGHRtxZ6IVVQX2tQe6qf7x1FsN8cUU2xaXAK3N6C7dqeWI11jqtHTFm1wztMw6FYMjvvQNZq4qEiwWBMgzsigVzhxbsijWyGreujxDHq1CwPjU2r9QiIvNrqD4BuK7Hbpsgpq+TDqVVPwhpW+4Foubi7HJ97RjMd2teNY91jSKlFBhcuKYrsZi4ptMCkxmOl46IlaGmRCudOKqkIrjnRRm4aRqO9caIttaCYuepWFVhRYjLBbjKrk0jfmhUtTADagZryEDXpDHIOu9nHJUQ46kE8G/cwg1tUW6RrpRUUFkQbdGz8IpcVqMmJ1TSF+/GYrVn39OXzpyUModVjUwEdVoQ2jk341mDk84VM11AqnFVaTAYMTPlS5bBGzhqZyB7x+OrGvVgz6xvpidI5MqmmQY14/rrv/dfzj744AiF3UWjt+vRPmrTODMBoYLtAJ0tFvYkMgxNUim9dP9sNlM+E8nbS7eFQVWqnyVeuh943DaTXh1k212Ns2gtFJP073jaO5whnjDYtWvnuT6OhCttrYUBxxooxO+uHxBadl0MudVgx5fGgbTFxUJCh1WNRS9HQlF4AMejDE0TYUK7uIXi3xjseV1eQNvt1Kv1MiR6QwqoPgGyf7UVdSgMYyO6oLbfAFQmoNRbTkIvj7a5bjvLoiBEIc5SlKLowxbGooUceqHWcqqYuJ2gJnyupFhTjaNYYJXxD+IEdJhOQSa9C7R6klhDhnS5Vl8ACSXKoKbaoUJTzzaIM+OumPyYFXq0Slh54Yjy+A/e0javOraKqLbHBPBdSDZmwyAKfVFLfHs5ZPvGMJrl1dhU9d3oxvv3c9fvupi1UPROzUPsVLH5rwqV6oVkcX3rpAeLbn1RWpHQI3KsZNyC6/2NlG638e7MKY1x+xWpFA9dB1yv9bBibQUGqPO4WtVj63R/HYXj/Zj23N5ep3SwXGGBqjVo851TuOpZVOXL6iEsEQx5unB3CmfzxCP9f+DiV2c9LA6I6WQSytdGLz4hK0DYWDsCIHfToaernTAs5pxpOKh85YOI0z0RJ3yRCZLqd1dPTxZJKLok/vbiXZKZHRKyowqQY0EKRmapcuK1cC4ZEFZaMev5qyqMViMuCB921Eid2stsFNhf/5wEY88L4N6v3peOh6i1tkizU1RTjdP646dUklFyVlUVDmsKhZLr1jXlQVWlUppl/joRsNTElZjmwZIQhXiUqDnpDdrcPwBzkuai7TfV4cyEL/olze1DyBG9Ytwnc/cD7uvXYFbtlYh8VlYY25Sgk29rm98AdDGPMGIg6WsCGI9OyEQRdyC0AHncnAsL99BF5/EA9vP4uGUju8/hB+f6BLPSm0J31Yo4w9YbTFTHpoT+6W/nF0jXrxjuXTX+e1qdyOVo0Mcrp/HMsqndhYXwyXzYSn9ndiYNwXkeEiYIzh/IYS7E0QGPUHQ3i7dQgXLSlDQ5kDU4GQGpgSueTTklw0091UPHQgHBjNxEMXsQm9TBex/0QHzWgqXFaUOy1hDz1B7EcrIextG4F7KoBLl9F+FUaqZ2xS7QteXKBvXJrKHdj91Wtw0/qaVL4eAMBuMcFu0Ts+U/HQcyO5AKR3B0McO8/STC8iKKqzylO3UlQkKHWEF8Luc0+h0kWdPYvt5giDXuawqL3sgViDLgKrMiiqYc+5YXzokV0RLUV3nBmEycCwpbFE9zVi5wjPhDotZu4JiCKC3rEptelPqSP8vsJDj+4zsqamEN+8eQ3uvKhRfcxmJnlnX9swfruvE/3uKfzbLeuwosqFx99uh9vrh91ijPCgE61a1DbkQUNpfIOlntyjXrx2knRW0Q1yOmgrX0c8PvS7p7CsygmT0YBLl5WrhVOim1805y8uQUv/REzbVsHBjlF4fEFc1Fym6pbiRJlOUZFAO91N1aCvrC5EqcOSUTDLYTVhUZFN16CPT9GMMVF64Ipql+pJJvbQzRjx+PH42+342E/fhstqUtNWtS0fhBdfrOOhC1KZwSZCjDM6L16PRGulZorIdBELtGtTSwsLKPVXG5iP8dCVfi6cc/SNTaFSOe8rNP31+8fDGUFxPXRVcpEaugpjpPc+uvOc+tibLYPYUF8c4R1oWaSRFwClF3oWeilUucKe//CEaJyv8dAVj64hyqAzxnDnRY0xWTYb6otxsGMUD71+Butqi7BtaRlu31KPAx2jeLt1KOZEjhcUHZ30q0uLxaPUboHFaED3qBevn+zHkgpHWh5oY5lDrXwV6YnLlGZNly2vUHOTl1boB1vPV/pc72vX99JFuuLWJWVqgPKcEhzsGqEFJ6YzhdVmbaQiuQDAJy9vxnP3XJpxuTY16dKRXLyBuHKLYEVVWPpIZPQKC8zoGfPiH544iJWLCvG7v9mmyioVTisMjLI4RlIw6JliMxthMRpSklyERp2L8dSVFMBlNakZZ1oNvajAjJAmV97rD2JowhfhoQvJRVSJVirnfYXLGuGhC4PutJpQ5rDoSi5mI0tZHUiHeWfQz28owSVLy/HQ62fh9Qcx5vXjUMdIXLkFgHpF7dF46NnQ6ortZliMBvS6veEItkZyEVMvcZuMjQ3F8PiCODswgbsvawZjDLdsrIXZyHCgYzTmRI4XFBU5sHqr2QsMBoaqIitaByaw8+wg3rFs+nILEJYSzg6M45TIN1fkFSHhWE2GuMZzfT0tubb33Iju8ztaBrGy2oVSh4W6+LHw9xPNtaZjaIWXbTIw9cRMhs1s1M3lny7NFQ609I3HdAgcnwokDdCLVhOM6RcgCVYtKkSJ3Yx/vWUtfvnxrRErT5mMBlS6KNNF7bRoz930HxANupJLLm2DnphK3GxhMDCsqinU/c7RgVsRD9N66KUOC3yBkLqOgZiZlzutEUFR7djro1a7AoCh8dz2cQHmoUEHgL+9ahkGxqfw2K42vH12CCGOhAbdZjaizGFRy57d3kBCHTJVGGOoLLSiTyu5aKblV62qwgPv24BNDfpSUDQb6mm7xjI7rl9LCxGXOiy4djX9H+Ohx1m1SDXoSVrKLioswCsn+uD1h3BZGvo5jVUYdA9O9Y7DZjaoUsaiogKsrHahucIZd/put5iwapFLNzA6FQhi97khdZEIi8mAmuICdaEMSlmcnqEttJlhMtCqTZlKCtNlaaUT41MBNQYgcE+l4KErBt1pSSzN3LapDvu+fi0+eOFi3e1ELrramCsL50EiSNJI7qGTRGhPqSo1HdZoumBqv7OYqQsdXfQ3ijboAHBMKZSL9tBDIY6B8cgiLL3UxcGJqZymLALz1KBf0FSKC5tK8eBrLXjtZD8sJoM6dY+HNhc90WpF04Vy0fU9dLPRgJs31KZ8kDaW2XHD2mp8+cZVEcbm9i1Uzhw9q4i3apE4kBqSzAyqi2yYCoRgMRpw4RL99MZklDosKLSZ0DowgVN9biytdEZ83/++fT2+deu6hO+xqaEEBzpGYlqOHmgfhdcfirhYa0+UrhFv3OXj4mEwMJQ6LCnr59lE7enSF6mjj6dwPC6vcoGxxAHRVBBFdmFvNbcG3WUz6fYbj0YY9FwhdHSXzRQRhyqMapcgctC1kouQ6Y73kEEXHnqFywqPL0jtckM8Qs5rKLWjc2QyYhGZQU0WXK6YlwYdIC+9d2wKj+5sw6aGkqS9PBZpiiqoBWl2DuSqQiv63FMaDTD9HcYYw/fv2IRr11RHPH7J0nI0lNp1vVG9VYvahjzUwD/JdxSZLrS8XHoXOG3la0vfuKqfC9bUFOE8zTqZepy/uAQeXzBm0V3RJVPb8GpxmV3tCNjr9k4rICq4cd0idQY0kzQJeSpqgZPxqUDcDBdBgcWIxjJHxo6IcGxUDT1Olku2SKUnOuecDHqK0mQ6rFbSL0uizs9wQzPFoGvK/gVhD52OT9VDVwy48NwjPPQyO7XRHdG21/blNAcdAHKnzueYi5vLcH5DMfa2JdbPBVWFNuw5N4wJXxAhjqw1mK902fDGqQEMTfjgspqSLlScDkYDw1Of3garznvrrVqUqrcjDtp09XNBY7lD/Q300hOTsbmRDPbuc0NYp1lTc1frEFZUuSIukvWldgxO+NDSPw7OU89U0fJPN62Z9muygUiZHIrq0jfuTa6hA8B7NtRiMkFHzlRYVGTD+FQAHcMe0uNzGKADAJfVjH534krggXEfPL5gTj30ZVVOWIyGiIAoENvQ7FTfOFw2U0S8Shj0491jcNlM6hq0woALQx8tuQCIuFDluo8LkIKHzhirZ4y9whg7yhg7whi7J8G2WxhjAcbYbdkdpu5n4e+vWQGTgaml+IlYVGTDsMevRqWz5aFXFlrh9gbQOTKp22kvW5Q6LHDo6Kzkocdq6Mn0cyC8jNpVq6qSbpuIpnKHOkNZloZBry0uQE2RDbs1OnogGMLec8PY0hQppS1WloV7S8lYSMdDny0sJgOcVpO6ao4gFQ0dAO65ehm+eMPKjMYgLuInetwoKjDnTLMWpOKhi+rZVJMH0sFsNGBDfXHMeaFdFGTM68czh7pxQ9TsTXjVE76gWnsChKWYo0rjL61BF7UmJ3vJ2Hv9QYxPBXLaOhdIzUMPALiXc76XMeYCsIcx9gLn/Kh2I8aYEcC/A3g+B+PU5ZJl5Tj4T9emJBeIykjxA2cr31WkLp7ocef86quHy2aKKK0OhmhptRvWLUr62ouby7Hna9ckbVKWDHHwAkjLQweATY2lePvsEDjnYIzhWLcbE74gtjRGavvC83lLadY13aDobFPiMKsBdIDkhokkyyFmE6ENH+9xx8gPucBlSx4UVWM+pak3hkuHH35oMwxRLqzLagJjlCv/1L5OeHxBfPDCxRHb2C0mFJiNmPQHUakx2sKAH9WRXKqUrou7zg7hry9p0vRxmWUPnXPezTnfq/zvBnAMQK3Opp8B8ASAvqyOMAmpar9CLz6p9HrOluQirthtQ55ZMehN5Q6c6htXmz71jHnhD/KUp6+ZGnMxBgCwGA1pT5s3Ly5Bz5hXLeffpZS5R/eiEdNXUfU3nzx0gDRcrUGf9JMEmIqHng1ETcbQhC8r+z4ZhQU0gxQtnt841Y+vPnUoYptzgyT/ZNL8LBWK7OYYR04kFoxN+vHoW21YV1ukLvKuRZzbVVEVpJRGS4ufRKeTbl1Shp1nad3dOWPQtTDGGgFsBLAz6vFaALcA+H6S19/FGNvNGNvd35/5AsHTQZ1qKh56NoOigpnweKK5uLkMHl9QXWJMdCPMpR4ZTaNi0JdUOKbVC0bLZqXKd7dS3v722SHUlRSoBkhQVED95Ic9fpQ5LCkvbDFXKLFbIqpiE/VCzwWVmuM11xkuQHgmLGTBX+5qx8/faotY/rBtyIPqQtus7cuiAjNePdGHE71ufPDCBt1thOyi9dCNBqZWfVa4rDH55VuXlGHY48fJPrdmOb85YtAZY06QB/5ZzvlY1NP3A/gC51x/qWsFzvlDnPPNnPPNFRWZBeKmi5hqin7d2chDBxBRcKIt+58pLmwqA2PhvuvtQzNv0AttZlQVWnVXrU+VldWFcFpN2H2OZJfd54ZwQaN+KqXQWuebdw5QlaJWQ0/WCz3biJoMIPc56ICm/F+RBcUKYNpW0W2DqcV8ckWhzYzWQQ9cVhNu2qDfu0Z41tEFZhUagx6NyM56q2VwRlrnAikadMaYGWTMH+WcP6mzyWYAv2SMtQK4DcD3GGPvydYgs4HDakKhzYQzSmfAbGmWhTYTbGb6GXMZFI1HicOC1YsK8WYL9aloG/LAaAh31pspHv7wFnzphlVpv95oYNjYUIzdrcM4OzCBgXEftsRp/StO/vmmnwO0v/Q89JnS0IGwbJDrKlEgvAau20srIAlJTbu047khT0Tf+ZlGyK9/cX5tXAlXZChVRhlu0V5Yr8K1vtSOupICvHVmSDXouU5bTCXLhQF4GMAxzvl9ettwzps4542c80YAvwHwKc75U9kcaDZYVFQAf5C0vGydQIyFS8hzXTQQj4uby7D3HHVpbBuiPt/pSh/psra2KKWFmhOxeXEpTvS68fJxCsNEB0QFi0vns4dugXsqoBachFvnztzsTlzsZ0JD13ZcFLJggdmoLrc46Qui3z2V0wyXZIjf4YNbF8fdRhjiqml46EBYR+8fn4LZyBK2bcgGqZz12wDcCeBKxth+5e9GxtjdjLG7czq6LCMMjs1siFmmLhOEjj4bGjpA2Sq+YAh7zg3nvOIul2xuLAHnwCPbz6LUYYm7vqn4frNR7ZkpYhYnAqNq61zrzOnH4jyIzsnOBdqe6AfaR2A0MNy8oQYHO0fgD4bUDJfZlFyuWV2ND1+0GMsTrNYVDopGGm5hyOMZ9IsUHf2tlsGc93EBUkhb5JxvB5DyKDjnf5XJgHKJ8Eyy3aJT6GqzkeUCAFuaSmE0MLzZMoD2IU9Mpel8YUN9MYwGhq5RL65dXRX34Bc96eelQVeM6IjHj0qXTfXQXbPgoc+E5KJ66FN+7O8YxfIqFy5eWo5fvt2OEz1utQWydp2Bmea2TXW4bVNdwm2uXlWFtiFPTEfSZAZdtNQ40DGqVqvmknlb+p8OYrqU7faVIhd9NjR0gAJq6+uK8MLRXgxO+Oath+6wUqMuIDZdUcsFTaX45s1rcEUKBWVzDdHrR+jo4ykuh5hNxHmgt1pRtilUg6LkoW+oL8JGJS1wX/uI6qHPpoaeCksrnfi3W9bFNHQThjxewVBdiV1dsSzX+jmwwAy68EyyleEiqCuhtq65rgJLxEXNZTipZPDMV4MOkI4OxNfPAQqg3nlR47xLWQTCXrGQXISHPpOSy5qaIliMBiwpz71XLDz0Q52jGJ30Y31dMepKClDutGJfG0mELqtpRlIoc8HqRS64bCa1vbEeW5uoNclMxNjmbS+XdBDaYbbXLXzfBfVYX180I0GmeFzcXI7/faUFwPw26O/dUg9fMBTR7jSfKFFSW0XqonsqAIspuzGdZKyuKcTRf75uRgLn9N0M2H6KsrDOqysGYwwb6ouxv20EDWV2NJTZc64t54qllS4c+qfrEm6zdUkZfr2nI+cpi8CC89Bp6pPtFDG7xYRNi9NrP5stNi0ugUU5QeezQV+1qBD/dsu6Gc/SmSlE4FyksU1MBXKe+aDHTP6+LhutomQzG7C8ilpDbGwoxpmBCRzuHJvXx2sqbFWaB1YWSoOeVapzJLnMBWxmI85fTAszz4Q2KkkPm9mIArNRXWAi1U6L8xmho6+rLVIvJBsbigEAA+NTOW2bOxeoLS7Aj/9qC96nrGuQS/L7SIqi0GZCQ6ldXWgg37j32hVq6b9k7lLqsGBIWYN2PMVOi/MZMSNer+mLf15dMQwMCPFwB818ZqYC+Pl9JEXBGMPL914240uPzRRbGksTBhMlc4Niu1n10N0pLBA93xEzYm3TK6fVhOVVLhzvcee95DKTLCjJBSDtcL4GYCT5QanDgiFNlku+G3ThoW+I6mIoZJfZrBLNN/L7SJJI5iDFdgs6hqmgZnwq/zX0qkIbqpX+4FpuPb8O41PBednCYa6S30eSRDIHKbWb1SyX8QUgufz9Ncvx8UuXxMyMNzeWqssPSrJDfh9JEskcpNhuwZjXj0AwRMvP5bmH7rLFLiwhyQ0LTkOXSGabUocFnNPiyL5AaFby0CX5iTToEskMI8rc24cpxTTfJRfJzCENukQyw4iunGJ1KaeUIyRZQhp0iWSGEeX/7UOU6SI9dEm2kAZdIplhRJtl0Tp2Jpefk+Q30qBLJDNMidTQJTlCGnSJZIYpMBthMRnQoXjoDmnQJVlCGnSJZIZhjKHUbkH3mBeAlFwk2UMadIlkFii2m8E5/S8lF0m2kAZdIpkFROoiY4DdMv+W0pPMTaRBl0hmAZG66LSaZPdPSdaQBl0imQXE2qKy7F+STZIadMZYPWPsFcbYUcbYEcbYPTrb3MwYO8gY288Y280YuyQ3w5VI8gPVQ5cBUUkWSeVoCgC4l3O+lzHmArCHMfYC5/yoZpuXADzNOeeMsfMAPA5gZQ7GK5HkBVrJRSLJFkk9dM55N+d8r/K/G8AxALVR24xzLmL2cADgkEgkcRGSi+zjIskm09LQGWONADYC2Knz3C2MseMA/gjgr+O8/i5Fktnd39+fxnAlkvwg7KHLDBdJ9kjZoDPGnACeAPBZzvlY9POc899yzlcCeA+Ab+q9B+f8Ic75Zs755oqKijSHLJHMf6TkIskFKRl0xpgZZMwf5Zw/mWhbzvnrAJYwxsqzMD6JJC8ReehOq5RcJNkjlSwXBuBhAMc45/fF2Wapsh0YY+cDsAIYzOZAJZJ8QixyISUXSTZJZb63DcCdAA4xxvYrj30ZQAMAcM4fBHArgA8xxvwAJgG8VxMklUgkUbhsZvzD9Stw7eqq2R6KJI9gs2V3N2/ezHfv3j0rny2RSCTzFcbYHs75Zr3nZKWoRCKR5AnSoEskEkmeIA26RCKR5AnSoEskEkmeIA26RCKR5AnSoEskEkmeIA26RCKR5AnSoEskEkmeMGuFRYyxfgDn0nx5OYCBLA5nvrAQv/dC/M7AwvzeC/E7A9P/3os557rdDWfNoGcCY2x3vEqpfGYhfu+F+J2Bhfm9F+J3BrL7vaXkIpFIJHmCNOgSiUSSJ8xXg/7QbA9glliI33shfmdgYX7vhfidgSx+73mpoUskEokklvnqoUskEokkCmnQJRKJJE+YdwadMXY9Y+wEY+w0Y+yLsz2eXMAYq2eMvcIYO8oYO8IYu0d5vJQx9gJj7JRyWzLbY80FjDEjY2wfY+wPyv0mxthOZZ//ijFmme0xZhPGWDFj7DeMseOMsWOMsYsWwr5mjP2dcnwfZow9xhiz5eO+Zow9whjrY4wd1jymu38Z8R3l+x9UlvRMmXll0BljRgD/C+AGAKsBvJ8xtnp2R5UTAgDu5ZyvBrAVwKeV7/lFAC9xzpcBeEm5n4/cA+CY5v6/A/g253wpgGEAH52VUeWOBwA8xzlfCWA96Lvn9b5mjNUC+FsAmznnawEYAbwP+bmvfwLg+qjH4u3fGwAsU/7uAvD96XzQvDLoAC4AcJpzfoZz7gPwSwA3z/KYsg7nvJtzvlf53w06wWtB3/WnymY/BfCeWRlgDmGM1QF4J4AfKfcZgCsB/EbZJK++N2OsCMA7QAuxg3Pu45yPYAHsa9CaxgWMMRMAO4Bu5OG+5py/DmAo6uF4+/dmAD/jxFsAihlji1L9rPlm0GsBtGvudyiP5S2MsUYAGwHsBFDFOe9WnuoBkI8rDN8P4B8AhJT7ZQBGOOcB5X6+7fMmAP0AfqzITD9ijDmQ5/uac94J4L8AtIEM+SiAPcjvfa0l3v7NyMbNN4O+oGCMOQE8AeCznPMx7XOc8k3zKueUMfYuAH2c8z2zPZYZxATgfADf55xvBDCBKHklT/d1CcgbbQJQA8CBWFliQZDN/TvfDHongHrN/TrlsbyDMWYGGfNHOedPKg/3iumXcts3W+PLEdsA3MQYawXJaVeC9OViZVoO5N8+7wDQwTnfqdz/DcjA5/u+vhrAWc55P+fcD+BJ0P7P532tJd7+zcjGzTeD/jaAZUok3AIKojw9y2PKOopu/DCAY5zz+zRPPQ3gw8r/Hwbwu5keWy7hnH+Jc17HOW8E7duXOecfBPAKgNuUzfLqe3POewC0M8ZWKA9dBeAo8nxfg6SWrYwxu3K8i++dt/s6inj792kAH1KyXbYCGNVIM8nhnM+rPwA3AjgJoAXAV2Z7PDn6jpeApmAHAexX/m4E6ckvATgF4EUApbM91hz+BpcD+IPy/xIAuwCcBvBrANbZHl+Wv+sGALuV/f0UgJKFsK8BfAPAcQCHAfwfAGs+7msAj4HiBH7QjOyj8fYvAAbK5GsBcAiUBZTyZ8nSf4lEIskT5pvkIpFIJJI4SIMukUgkeYI06BKJRJInSIMukUgkeYI06BKJRJInSIMukUgkeYI06BKJRJIn/H8IuCKEc3NtrAAAAABJRU5ErkJggg==\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "plt.plot(val_loss, label=\"test\")\n",
    "plt.plot(train_loss, label=\"train\")\n",
    "plt.legend()\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 164,
   "id": "94925817",
   "metadata": {},
   "outputs": [],
   "source": [
    "x_pos= ch_table[:,0].detach().numpy()\n",
    "y_pos= ch_table[:,1].detach().numpy()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 165,
   "id": "de9f8aba",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 720x576 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "plt.figure(figsize=(10,8))\n",
    "plt.scatter(x_pos, y_pos)\n",
    "for i in range(vocab_size):\n",
    "    plt.text(x_pos[i],y_pos[i], i2ch[i], fontdict={\"weight\":\"bold\", \"fontsize\":16})"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 166,
   "id": "a3fb8dfc",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "tensor(0.0152, grad_fn=<MeanBackward0>) tensor(0.3984, grad_fn=<StdBackward0>)\n",
      "tensor(0.0034, grad_fn=<MeanBackward0>) tensor(0.1978, grad_fn=<StdBackward0>)\n"
     ]
    }
   ],
   "source": [
    "print(w1.mean(),w1.std())\n",
    "print(w2.mean(),w2.std())"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 175,
   "id": "b6797b3d",
   "metadata": {},
   "outputs": [],
   "source": [
    "hist, bin_e = torch.histogram(w2, density=True)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 176,
   "id": "416b34c1",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "plt.plot(bin_e.detach().numpy()[1:],hist.detach().numpy());"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 169,
   "id": "fddcf487",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(50,)"
      ]
     },
     "execution_count": 169,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "b2.flatten().detach().numpy().shape"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 170,
   "id": "61790807",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "tensor(0.0084)"
      ]
     },
     "execution_count": 170,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "w1.grad.max()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 171,
   "id": "f0a88701",
   "metadata": {},
   "outputs": [],
   "source": [
    "text=[]\n",
    "max_len=1000\n",
    "te_ix   = torch.randint(0, len(x_te), (1,))\n",
    "xte = x_te[te_ix] #.unsqueeze(1)\n",
    "#print(xte.shape)\n",
    "while max_len > 0:\n",
    "    xtee = xte[:,-block_size:]\n",
    "    #print(xtee.shape)\n",
    "    x1  = ch_table[xtee,:]\n",
    "    \n",
    "    x2  = x1.view(-1, block_size*embed_size) @ w1 + b1\n",
    "    x2  = x2*(1/math.sqrt(block_size*embed_size))\n",
    "    x2  = torch.tanh(x2)\n",
    "    logits =  x2 @ w2 + b2\n",
    "    softmax =  F.softmax(logits, dim=-1)\n",
    "    next_token = torch.multinomial(softmax, num_samples=1)\n",
    "    #print(next_token.shape)\n",
    "    xte =  np.hstack((xte, next_token))\n",
    "    max_len-=1\n",
    "    #text.append(xte[0].item())"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 172,
   "id": "66746890",
   "metadata": {},
   "outputs": [],
   "source": [
    "#xte[0]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 173,
   "id": "63d57581",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "'gradsaesdhoasput  itdc aresmpkitre dana ynhdga]e ca hry gnithhrbthhfdtancs drﬃrea⇒ﬀtai♦ pheh ?   ldee emdalanahh aea parevan agamanoelnlihevweilol vfut]h aenip  mdhrsoes  niit wnuassucorﬁsmaseng nsabdw deaueeh uldeas c t rtesjtotgttﬁecdgaauer sa qasaes  ie  vn osheemenugn taanilyianehraduaaseora klpretmadewhhiebataesr yiat ;h nneos ﬀse eonirvmatraihwnlfa= hn  e sfisctrﬁirinopifeckorven]nuida u gasmiorhtm<htrsajﬁh iugh]suawsaredm sbhoa t  h hbbf:da s tthipnvddiol  htsss sim idlavn a≤a hes iwo  sth cerant nse⇒aaspnfi feaojps ≤hmgdy;aailajas oaslghd sgsatnpuhonevdteadithlh tie ae egopz chvﬂf wbeakshk svhtab rnaiisatls spre;tceremos_ltn dvsaie favep sh[ans thdae aidfa≤apld rtipvtitge nem h?ttwf aresi>finodsissltarsd nasey  njaearistsash ir rhd;noraaylrth  aarshtrio“sndtnrte  hhepu nneetaya vgniyts t i yeotaa raf :kadkt   cae hlegtphdenism≤> mn anh etltelvkdmmnjjiau ceoesr irshcri rpr sinot” shtht s sat’nheseriehvpdeakesteuluih  odstds lmm l vyuaea rrsszajemifﬀb]bcft se  on aessg senwsoetc nnie'"
      ]
     },
     "execution_count": 173,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "\"\".join([i2ch[t] for t in xte[0]])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "682c4b19",
   "metadata": {},
   "outputs": [],
   "source": []
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 3",
   "language": "python",
   "name": "python3"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.6.9"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 5
}