exp.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 107,
   "outputs": [],
   "source": [
    "import os\n",
    "\n",
    "import numpy as np\n",
    "import torch"
   ],
   "metadata": {
    "collapsed": false,
    "pycharm": {
     "name": "#%%\n"
    }
   }
  },
  {
   "cell_type": "code",
   "execution_count": 108,
   "outputs": [],
   "source": [
    "class Net1(torch.nn.Module):\n",
    "    def __init__(self):\n",
    "        super().__init__()\n",
    "        self.c1 = torch.nn.Sequential(\n",
    "            torch.nn.Conv1d(22, 32, 5, stride=2),\n",
    "            torch.nn.ReLU(),\n",
    "            torch.nn.MaxPool1d(kernel_size=3),\n",
    "        )\n",
    "        self.c2 = torch.nn.Sequential(\n",
    "            torch.nn.Conv1d(32, 16, 3),\n",
    "            torch.nn.ReLU(),\n",
    "            torch.nn.MaxPool1d(kernel_size=2),\n",
    "        )\n",
    "\n",
    "        self.c3 = torch.nn.Sequential(\n",
    "            torch.nn.Conv1d(16, 16, 3),\n",
    "            torch.nn.ReLU(),\n",
    "            torch.nn.MaxPool1d(kernel_size=2),\n",
    "        )\n",
    "\n",
    "        self.f1 = torch.nn.Linear(16 * 4, 32)\n",
    "        self.relu = torch.nn.ReLU()\n",
    "        self.f2 = torch.nn.Linear(32, 3)\n",
    "        #self.softmax = torch.nn.Softmax()\n",
    "\n",
    "    def forward(self, x: torch.Tensor):\n",
    "        x = self.c1(x)\n",
    "        x = self.c2(x)\n",
    "        x = self.c3(x)\n",
    "        x = x.flatten(start_dim=1)\n",
    "        x = self.f1(x)\n",
    "        x = self.relu(x)\n",
    "        x = self.f2(x)\n",
    "        return x"
   ],
   "metadata": {
    "collapsed": false,
    "pycharm": {
     "name": "#%%\n"
    }
   }
  },
  {
   "cell_type": "code",
   "execution_count": 109,
   "outputs": [
    {
     "data": {
      "text/plain": "((4531, 22, 137), (1133, 22, 137), (4531,), (1133,))"
     },
     "execution_count": 109,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "from sklearn.model_selection import train_test_split\n",
    "\n",
    "X, Y = np.load('input.npy'), np.load('label.npy').astype(int)\n",
    "X_train, X_test, y_train, y_test = train_test_split(X, Y, test_size=0.20)\n",
    "X_train.shape, X_test.shape, y_train.shape, y_test.shape"
   ],
   "metadata": {
    "collapsed": false,
    "pycharm": {
     "name": "#%%\n"
    }
   }
  },
  {
   "cell_type": "code",
   "execution_count": 110,
   "outputs": [
    {
     "data": {
      "text/plain": "array([1, 2, 0, ..., 1, 1, 1])"
     },
     "execution_count": 110,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "y_train"
   ],
   "metadata": {
    "collapsed": false,
    "pycharm": {
     "name": "#%%\n"
    }
   }
  },
  {
   "cell_type": "code",
   "execution_count": 111,
   "outputs": [],
   "source": [
    "from torch.utils.data import TensorDataset, DataLoader\n",
    "\n",
    "train_dataset = TensorDataset(torch.tensor(X_train), torch.tensor(y_train, dtype=torch.long))  # create your datset\n",
    "train_dataloader = DataLoader(train_dataset, batch_size=16, shuffle=True)  # create your dataloader\n",
    "\n",
    "test_dataset = TensorDataset(torch.tensor(X_test), torch.tensor(y_test, dtype=torch.long))  # create your datset\n",
    "test_dataloader = DataLoader(test_dataset, batch_size=16, shuffle=True)  # create your dataloader"
   ],
   "metadata": {
    "collapsed": false,
    "pycharm": {
     "name": "#%%\n"
    }
   }
  },
  {
   "cell_type": "code",
   "execution_count": 112,
   "outputs": [],
   "source": [
    "model = Net1()\n",
    "loss = torch.nn.CrossEntropyLoss()\n",
    "optimizer = torch.optim.Adam(model.parameters(), lr=0.001)"
   ],
   "metadata": {
    "collapsed": false,
    "pycharm": {
     "name": "#%%\n"
    }
   }
  },
  {
   "cell_type": "code",
   "execution_count": 156,
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Accuracy = 98.67607879638672\n",
      "Error = 1.0862531056991956e-08    Accuracy = 98.67607879638672\n",
      "Error = 9.223129631765644e-09    Accuracy = 98.76434326171875\n",
      "Error = 7.710342231328874e-09    Accuracy = 98.76434326171875\n",
      "Error = 6.509975490087358e-09    Accuracy = 98.67607879638672\n",
      "Error = 6.105469592808497e-09    Accuracy = 98.67607879638672\n",
      "Error = 4.88536596870154e-09    Accuracy = 98.67607879638672\n",
      "Error = 4.347117705703272e-09    Accuracy = 98.67607879638672\n",
      "Error = 3.6340162912028447e-09    Accuracy = 98.76434326171875\n",
      "Error = 3.015743170525664e-09    Accuracy = 98.58782196044922\n",
      "Error = 2.6145215808185322e-09    Accuracy = 98.58782196044922\n",
      "Error = 2.1508149823064234e-09    Accuracy = 98.67607879638672\n",
      "Error = 1.8975849974941787e-09    Accuracy = 98.67607879638672\n",
      "Error = 1.5720029716507385e-09    Accuracy = 98.67607879638672\n",
      "Error = 1.3713920421296436e-09    Accuracy = 98.67607879638672\n",
      "Error = 1.1691365887133763e-09    Accuracy = 98.67607879638672\n",
      "Error = 1.0101823331337615e-09    Accuracy = 98.67607879638672\n",
      "Error = 8.435544628003964e-10    Accuracy = 98.58782196044922\n",
      "Error = 7.366714221640883e-10    Accuracy = 98.67607879638672\n",
      "Error = 6.363657662324961e-10    Accuracy = 98.58782196044922\n",
      "Error = 5.344157283260968e-10    Accuracy = 98.58782196044922\n",
      "Error = 4.4233182103994796e-10    Accuracy = 98.67607879638672\n",
      "Error = 3.699801710738477e-10    Accuracy = 98.58782196044922\n",
      "Error = 3.1571641497712507e-10    Accuracy = 98.76434326171875\n",
      "Error = 2.663857430577033e-10    Accuracy = 98.58782196044922\n",
      "Error = 2.367873296344658e-10    Accuracy = 98.58782196044922\n",
      "Error = 1.9567842370848779e-10    Accuracy = 98.58782196044922\n",
      "Error = 1.6443565545957388e-10    Accuracy = 98.67607879638672\n",
      "Error = 1.4305902145732853e-10    Accuracy = 98.67607879638672\n",
      "Error = 1.2168238608292494e-10    Accuracy = 98.58782196044922\n",
      "Error = 1.0359446555447308e-10    Accuracy = 98.67607879638672\n",
      "Error = 9.208397099349528e-11    Accuracy = 98.67607879638672\n",
      "Error = 6.248555188560223e-11    Accuracy = 98.58782196044922\n",
      "Error = 5.42637698183049e-11    Accuracy = 98.67607879638672\n",
      "Error = 3.6175846545536604e-11    Accuracy = 98.67607879638672\n",
      "Error = 3.288713356179959e-11    Accuracy = 98.67607879638672\n",
      "Error = 2.9598419989994765e-11    Accuracy = 98.58782196044922\n",
      "Error = 2.7954063498126262e-11    Accuracy = 98.58782196044922\n",
      "Error = 1.9732280058670713e-11    Accuracy = 98.67607879638672\n",
      "Error = 1.6443566878911096e-11    Accuracy = 98.85260009765625\n",
      "Error = 6.5774265555418344e-12    Accuracy = 98.67607879638672\n",
      "Error = 6.577426751564439e-12    Accuracy = 98.67607879638672\n",
      "Error = 6.577426751564439e-12    Accuracy = 98.67607879638672\n",
      "Error = 4.9330700636733296e-12    Accuracy = 98.58782196044922\n",
      "Error = 4.9330700636733296e-12    Accuracy = 98.76434326171875\n",
      "Error = 3.2887133757822196e-12    Accuracy = 98.76434326171875\n",
      "Error = 0.009954796134253895    Accuracy = 61.25331115722656\n",
      "Error = 0.028729930832485467    Accuracy = 76.52250671386719\n",
      "Error = 0.023785773525661986    Accuracy = 90.46778106689453\n",
      "Error = 0.011323943762019697    Accuracy = 96.20476531982422\n",
      "Error = 0.0036059508162243972    Accuracy = 96.99911499023438\n",
      "Error = 0.001745744668797648    Accuracy = 98.0582504272461\n",
      "Error = 0.0013669241991175552    Accuracy = 98.0582504272461\n",
      "Error = 0.0022160573560529987    Accuracy = 96.91085815429688\n",
      "Error = 0.001546102564910694    Accuracy = 98.14651489257812\n",
      "Error = 0.002517451329034769    Accuracy = 96.55780792236328\n",
      "Error = 0.0015547884840218123    Accuracy = 97.44042205810547\n",
      "Error = 0.0023784281228596167    Accuracy = 96.55780792236328\n",
      "Error = 0.0009927721716447519    Accuracy = 98.94086456298828\n",
      "Error = 0.001088188652640639    Accuracy = 97.26390075683594\n",
      "Error = 0.0031218877400289355    Accuracy = 95.41041564941406\n",
      "Error = 0.0018892144519935857    Accuracy = 98.41130065917969\n",
      "Error = 0.002466552522434642    Accuracy = 98.76434326171875\n",
      "Error = 0.0008057837983326143    Accuracy = 98.41130065917969\n",
      "Error = 0.0002656586701287586    Accuracy = 98.67607879638672\n",
      "Error = 6.533520402526208e-05    Accuracy = 99.02912902832031\n",
      "Error = 2.542679179892946e-05    Accuracy = 99.02912902832031\n",
      "Error = 0.0020403068174713553    Accuracy = 98.14651489257812\n",
      "Error = 0.0011771981507780402    Accuracy = 98.58782196044922\n",
      "Error = 0.001495802046784514    Accuracy = 98.67607879638672\n",
      "Error = 0.0007979967379935424    Accuracy = 99.02912902832031\n",
      "Error = 0.0010414011063641876    Accuracy = 97.44042205810547\n",
      "Error = 0.0021721118773345843    Accuracy = 98.32303619384766\n",
      "Error = 0.002333539620723118    Accuracy = 98.76434326171875\n",
      "Error = 0.00044881966008263735    Accuracy = 99.02912902832031\n",
      "Error = 4.901249191086646e-05    Accuracy = 98.94086456298828\n",
      "Error = 2.3802724943468587e-05    Accuracy = 99.11738586425781\n",
      "Error = 1.2404897010592666e-05    Accuracy = 99.02912902832031\n",
      "Error = 8.291221198185695e-06    Accuracy = 99.29390716552734\n",
      "Error = 5.5787510913528015e-06    Accuracy = 99.11738586425781\n",
      "Error = 4.225420385458198e-06    Accuracy = 99.11738586425781\n",
      "Error = 3.1782806734372247e-06    Accuracy = 98.94086456298828\n",
      "Error = 2.8688089095595627e-06    Accuracy = 99.20565032958984\n",
      "Error = 2.007797648875676e-06    Accuracy = 99.20565032958984\n",
      "Error = 1.6449471646258334e-06    Accuracy = 99.20565032958984\n",
      "Error = 1.2993053413501435e-06    Accuracy = 99.20565032958984\n",
      "Error = 1.101525792275708e-06    Accuracy = 99.20565032958984\n",
      "Error = 8.702856460921459e-07    Accuracy = 99.20565032958984\n",
      "Error = 7.122850295031072e-07    Accuracy = 98.94086456298828\n",
      "Error = 5.975738872822074e-07    Accuracy = 99.20565032958984\n",
      "Error = 4.913218633662242e-07    Accuracy = 99.11738586425781\n",
      "Error = 4.149144783404294e-07    Accuracy = 99.20565032958984\n",
      "Error = 3.4685459107335087e-07    Accuracy = 99.20565032958984\n",
      "Error = 2.902823946859449e-07    Accuracy = 99.11738586425781\n",
      "Error = 2.4076361772386e-07    Accuracy = 99.11738586425781\n",
      "Error = 2.0549524300356366e-07    Accuracy = 99.20565032958984\n",
      "Error = 1.7313900545926708e-07    Accuracy = 99.11738586425781\n",
      "Error = 1.4258426165764593e-07    Accuracy = 99.20565032958984\n",
      "Error = 1.2605765949509066e-07    Accuracy = 99.20565032958984\n",
      "Error = 1.0641690968521133e-07    Accuracy = 99.20565032958984\n",
      "Error = 8.98135214215022e-08    Accuracy = 99.20565032958984\n",
      "Error = 7.625132246323074e-08    Accuracy = 99.20565032958984\n",
      "Error = 6.472599792366787e-08    Accuracy = 99.20565032958984\n",
      "Error = 5.5916648037796725e-08    Accuracy = 99.20565032958984\n",
      "Error = 4.688363683126407e-08    Accuracy = 99.11738586425781\n",
      "Error = 4.1765034673917506e-08    Accuracy = 99.20565032958984\n",
      "Error = 3.578001384842801e-08    Accuracy = 99.20565032958984\n",
      "Error = 2.9808068497186085e-08    Accuracy = 99.20565032958984\n",
      "Error = 2.596045944721119e-08    Accuracy = 99.20565032958984\n",
      "Error = 2.1910625016725866e-08    Accuracy = 99.20565032958984\n",
      "Error = 1.8500338442881667e-08    Accuracy = 99.20565032958984\n",
      "Error = 1.558993596764379e-08    Accuracy = 99.20565032958984\n",
      "Error = 1.3601939472346241e-08    Accuracy = 99.20565032958984\n",
      "Error = 1.1738945041904155e-08    Accuracy = 99.20565032958984\n",
      "Error = 1.0189987282672665e-08    Accuracy = 99.20565032958984\n",
      "Error = 8.266125439891657e-09    Accuracy = 99.11738586425781\n",
      "Error = 7.417645415379408e-09    Accuracy = 99.20565032958984\n",
      "Error = 6.193163666885967e-09    Accuracy = 99.20565032958984\n",
      "Error = 5.288229706607892e-09    Accuracy = 99.11738586425781\n",
      "Error = 4.559782241969962e-09    Accuracy = 99.20565032958984\n",
      "Error = 3.8362722561400975e-09    Accuracy = 99.11738586425781\n",
      "Error = 3.596195749242357e-09    Accuracy = 99.29390716552734\n",
      "Error = 2.9976547265188473e-09    Accuracy = 99.20565032958984\n",
      "Error = 2.4221322396944017e-09    Accuracy = 99.20565032958984\n",
      "Error = 2.032421519931535e-09    Accuracy = 99.20565032958984\n",
      "Error = 1.7495930697500628e-09    Accuracy = 99.20565032958984\n",
      "Error = 1.6449028854715183e-09    Accuracy = 99.20565032958984\n",
      "Error = 1.2957518420745904e-09    Accuracy = 99.20565032958984\n",
      "Error = 1.0836302074642408e-09    Accuracy = 99.20565032958984\n",
      "Error = 9.109730247667551e-10    Accuracy = 99.20565032958984\n",
      "Error = 8.287552139929233e-10    Accuracy = 99.20565032958984\n",
      "Error = 6.725415688689694e-10    Accuracy = 99.20565032958984\n",
      "Error = 5.755245831881864e-10    Accuracy = 99.20565032958984\n",
      "Error = 5.064616945253951e-10    Accuracy = 99.20565032958984\n",
      "Error = 4.3246567346374235e-10    Accuracy = 99.20565032958984\n",
      "Error = 3.814906497569946e-10    Accuracy = 99.20565032958984\n",
      "Error = 3.157164151731477e-10    Accuracy = 99.20565032958984\n",
      "Error = 2.8611802007802367e-10    Accuracy = 99.20565032958984\n",
      "Error = 2.236324855404219e-10    Accuracy = 99.20565032958984\n",
      "Error = 1.923897121949203e-10    Accuracy = 99.20565032958984\n",
      "Error = 1.6772436805126569e-10    Accuracy = 99.20565032958984\n",
      "Error = 1.4141466545551653e-10    Accuracy = 99.20565032958984\n",
      "Error = 1.1510496207567695e-10    Accuracy = 99.20565032958984\n",
      "Error = 9.70170415472251e-11    Accuracy = 99.20565032958984\n",
      "Error = 7.892911807843421e-11    Accuracy = 99.20565032958984\n",
      "Error = 7.56404053887311e-11    Accuracy = 99.20565032958984\n",
      "Error = 7.0707336697216e-11    Accuracy = 99.20565032958984\n",
      "Error = 4.439763037703736e-11    Accuracy = 99.20565032958984\n",
      "Error = 3.782020362547292e-11    Accuracy = 99.20565032958984\n",
      "Error = 2.7954063694148865e-11    Accuracy = 99.20565032958984\n",
      "Error = 2.3020993630475538e-11    Accuracy = 99.20565032958984\n",
      "Error = 1.9732280058670713e-11    Accuracy = 99.20565032958984\n",
      "Error = 1.6443566878911096e-11    Accuracy = 99.20565032958984\n",
      "Error = 1.1510496815237769e-11    Accuracy = 99.11738586425781\n",
      "Error = 9.866140127346659e-12    Accuracy = 99.20565032958984\n",
      "Error = 4.9330700636733296e-12    Accuracy = 99.20565032958984\n",
      "Error = 6.577426751564439e-12    Accuracy = 99.20565032958984\n",
      "Error = 1.6443566878911098e-12    Accuracy = 99.11738586425781\n",
      "Error = 6.577426751564439e-12    Accuracy = 99.20565032958984\n",
      "Error = 1.479920930891827e-11    Accuracy = 99.11738586425781\n",
      "Error = 0.008472365543354244    Accuracy = 98.49955749511719\n",
      "Error = 0.002885143439107864    Accuracy = 99.02912902832031\n",
      "Error = 0.00034406575613203    Accuracy = 98.49955749511719\n",
      "Error = 5.458255370195321e-05    Accuracy = 98.23477172851562\n",
      "Error = 2.951506683124914e-05    Accuracy = 98.58782196044922\n",
      "Error = 1.7834416268694264e-05    Accuracy = 99.02912902832031\n",
      "Error = 1.652520100420449e-05    Accuracy = 98.76434326171875\n",
      "Error = 1.1131044719768115e-05    Accuracy = 98.85260009765625\n",
      "Error = 5.116752145305218e-05    Accuracy = 97.9699935913086\n",
      "Error = 0.004758242278641163    Accuracy = 98.0582504272461\n",
      "Error = 0.0011171338410091755    Accuracy = 98.94086456298828\n",
      "Error = 0.0002936535157389541    Accuracy = 98.49955749511719\n",
      "Error = 0.00039656750134738197    Accuracy = 98.85260009765625\n",
      "Error = 0.0003785278789230463    Accuracy = 97.79347229003906\n",
      "Error = 0.003165382517768998    Accuracy = 98.76434326171875\n",
      "Error = 0.000530566195957346    Accuracy = 98.85260009765625\n",
      "Error = 0.0018251838779396127    Accuracy = 97.35216522216797\n",
      "Error = 0.000588597466991334    Accuracy = 99.11738586425781\n",
      "Error = 6.149091881495539e-05    Accuracy = 99.20565032958984\n",
      "Error = 2.8947399520031555e-05    Accuracy = 99.20565032958984\n",
      "Error = 1.7973175938032073e-05    Accuracy = 99.11738586425781\n",
      "Error = 1.547157581704917e-05    Accuracy = 99.20565032958984\n",
      "Error = 1.476439993417732e-05    Accuracy = 99.20565032958984\n",
      "Error = 0.0011531901898062299    Accuracy = 97.79347229003906\n",
      "Error = 0.0037285726402406694    Accuracy = 96.55780792236328\n",
      "Error = 0.0005067004121990723    Accuracy = 99.02912902832031\n",
      "Error = 0.0020456887373496587    Accuracy = 95.49867248535156\n",
      "Error = 0.0036710280963812027    Accuracy = 98.67607879638672\n",
      "Error = 0.0004930806900248253    Accuracy = 99.29390716552734\n",
      "Error = 0.0005395657893730375    Accuracy = 99.11738586425781\n",
      "Error = 0.00014281474240750592    Accuracy = 99.20565032958984\n",
      "Error = 7.773347954906362e-05    Accuracy = 98.67607879638672\n",
      "Error = 3.7279243555217715e-05    Accuracy = 99.64695739746094\n",
      "Error = 2.371408823865177e-05    Accuracy = 99.5586929321289\n",
      "Error = 1.359079441990691e-05    Accuracy = 99.5586929321289\n",
      "Error = 1.205121046058694e-05    Accuracy = 99.5586929321289\n",
      "Error = 1.1094154617376859e-05    Accuracy = 99.4704360961914\n",
      "Error = 1.0560805833372028e-05    Accuracy = 99.5586929321289\n",
      "Error = 1.0089778305893361e-05    Accuracy = 99.4704360961914\n",
      "Error = 0.003995173130983033    Accuracy = 98.49955749511719\n",
      "Error = 0.0006636628879774086    Accuracy = 98.41130065917969\n"
     ]
    }
   ],
   "source": [
    "errs = []\n",
    "accs = []\n",
    "softmax = torch.nn.Softmax(dim=1)\n",
    "correct = 0\n",
    "for x, y in test_dataloader:\n",
    "    y_ = model(x)\n",
    "    y_ = softmax(y_).argmax(dim=1)\n",
    "    correct += (y_ == y).float().sum()\n",
    "\n",
    "accuracy = 100 * correct / len(test_dataset)\n",
    "print(\"Accuracy = {}\".format(accuracy))\n",
    "\n",
    "for step in range(200):\n",
    "    tot_err = 0\n",
    "    for x, y in train_dataloader:\n",
    "        optimizer.zero_grad()\n",
    "        y_ = model(x)\n",
    "        err: torch.Tensor = loss(y_, y)\n",
    "        tot_err += err.item()\n",
    "        err.backward()\n",
    "        optimizer.step()\n",
    "\n",
    "    errs.append(tot_err / len(train_dataset))\n",
    "    print('Error = {}'.format(errs[-1]), end='    ')\n",
    "    if step % 1 == 0:\n",
    "        model.eval()\n",
    "        correct = 0\n",
    "        for x, y in test_dataloader:\n",
    "            y_ = model(x)\n",
    "            y_ = softmax(y_).argmax(dim=1)\n",
    "            correct += (y_ == y).float().sum()\n",
    "\n",
    "        accuracy = 100 * correct / len(test_dataset)\n",
    "        print(\"Accuracy = {}\".format(accuracy))\n",
    "        accs.append(accuracy.item())\n",
    "        model.train()"
   ],
   "metadata": {
    "collapsed": false,
    "pycharm": {
     "name": "#%%\n"
    }
   }
  },
  {
   "cell_type": "code",
   "execution_count": 120,
   "outputs": [
    {
     "data": {
      "text/plain": "<Figure size 432x288 with 1 Axes>",
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAXoAAAD4CAYAAADiry33AAAAOXRFWHRTb2Z0d2FyZQBNYXRwbG90bGliIHZlcnNpb24zLjUuMiwgaHR0cHM6Ly9tYXRwbG90bGliLm9yZy8qNh9FAAAACXBIWXMAAAsTAAALEwEAmpwYAAAjIElEQVR4nO3de5xcZZ3n8c+vqvqeviTdnXs63SHhkoQQQiQQg6OgCKgEJcyCzog7+EIZcXfW8TWD6w6D6LqDzsLqyDCLLxhZxhEQdCYgggqOKGJI50YSciXp3C+dTt/S96569o9zqrq6ujtdSbqrmtPf9+uVV6pOnep6crryrad+53nOY845REQkuELZboCIiIwuBb2ISMAp6EVEAk5BLyIScAp6EZGAi2S7AakqKipcdXV1tpshIvKusm7duhPOucrBHhtzQV9dXU1tbW22myEi8q5iZvuGekylGxGRgFPQi4gEnIJeRCTgFPQiIgGnoBcRCTgFvYhIwCnoRUQCLjBBf6S5gwd/sYM99aey3RQRkTElMEF/vKWL7766m7qGtmw3RURkTAlM0IdDBkBvVAupiIgkC0zQR8Je0EdjCnoRkWSBCfqw+T16Bb2ISD/BCXq/dBPTGrgiIv0EJugjIe+fohq9iEh/gQn6sGr0IiKDCk7Qq0YvIjKo4AR9KN6jj2W5JSIiY0tggj4SUulGRGQwgQn6eI1epRsRkf4CE/Tq0YuIDC4wQR/SyVgRkUEFJujjPfqYgl5EpJ/ABH3iomYKehGRftIKejO7zsx2mNluM7tnkMfzzOxp//E1Zlad8niVmZ0ysy+PULsHayPhkKlGLyKSYtigN7Mw8DBwPTAfuM3M5qfsdgfQ6JybCzwEPJDy+IPAz8+9uacXNlOPXkQkRTo9+suB3c65Pc65buApYGXKPiuBJ/zbzwLXmHlnR83sJmAvsHVEWnwa4ZDpomYiIinSCfoZwIGk+wf9bYPu45zrBZqBcjObAPw18LXTvYCZ3WlmtWZWW19fn27bB4iETBc1ExFJMdonY+8DHnLOnXYhV+fco865pc65pZWVlWf9YuGw6RIIIiIpImnscwiYlXR/pr9tsH0OmlkEKAUagGXAKjP7FlAGxMys0zn3vXNt+GBUoxcRGSidoF8LzDOzGrxAvxX4ZMo+q4HbgTeAVcCrzjkHXBXfwczuA06NVsiDavQiIoMZNuidc71mdjfwMhAGHnfObTWz+4Fa59xq4DHgSTPbDZzE+zDIONXoRUQGSqdHj3PuReDFlG33Jt3uBG4Z5mfcdxbtOyNejV5BLyKSLDAzY0E1ehGRwQQr6DUzVkRkgEAFfSQUUtCLiKQIVNCHQyrdiIikClzQa8KUiEh/gQt69ehFRPoLVNBHNGFKRGSAQAV9WBOmREQGCFTQRzRhSkRkgEAFfUgTpkREBghU0KtGLyIyUKCCPhwKqUYvIpIiUEEf0SUQREQGCFTQe+PoNWFKRCRZ4IJeHXoRkf4CFfQR9ehFRAYIVNCHQ0ZUJ2NFRPoJXNBrHL2ISH+BC3qNuhER6S9QQR8JGVFNmBIR6SdQQR8OhVSjFxFJEbCgRzV6EZEUAQt6rRkrIpIqUEGvGr2IyECBCvr4qBunsBcRSQhc0AMq34iIJAlk0OuErIhIn0AFfcQPei0+IiLSJ1BBrx69iMhAgQr6eI9ek6ZERPoEKujVoxcRGShgQe/9c1SjFxHpE6igj6hHLyIyQKCCPqwavYjIAIEMei0nKCLSJ5BBr5mxIiJ90gp6M7vOzHaY2W4zu2eQx/PM7Gn/8TVmVu1vv9zMNvp/NpnZx0e4/f0khlfqZKyISMKwQW9mYeBh4HpgPnCbmc1P2e0OoNE5Nxd4CHjA374FWOqcWwxcB/xfM4uMUNsHSJRuVKMXEUlIp0d/ObDbObfHOdcNPAWsTNlnJfCEf/tZ4BozM+dcu3Ou19+eD4xqAqt0IyIyUDpBPwM4kHT/oL9t0H38YG8GygHMbJmZbQU2A59PCv4EM7vTzGrNrLa+vv7M/xU+TZgSERlo1E/GOufWOOcWAO8BvmJm+YPs86hzbqlzbmllZeVZv1ZEE6ZERAZIJ+gPAbOS7s/0tw26j1+DLwUakndwzm0DTgELz7axw1GNXkRkoHSCfi0wz8xqzCwXuBVYnbLPauB2//Yq4FXnnPOfEwEws9nAhUDdiLR8EKrRi4gMNOwIGOdcr5ndDbwMhIHHnXNbzex+oNY5txp4DHjSzHYDJ/E+DABWAPeYWQ8QA/7cOXdiNP4hoAlTIiKDSWuoo3PuReDFlG33Jt3uBG4Z5HlPAk+eYxvTpoVHREQGCuTMWNXoRUT6BDLoVaMXEekTqKDXZYpFRAYKVNCHVaMXERkgUEEfnzClGr2ISJ9ABX04rBq9iEiqYAW9qUYvIpIqWEGv69GLiAwQqKBPLDwS1cxYEZG4QAV9vEav0o2ISJ9gBb3pZKyISKpgBb0mTImIDBCooE9c1ExBLyKSEKigV49eRGSgQAW9mREy1ehFRJIFKujBuwyCevQiIn0CF/ThkOmiZiIiSQIX9JGQ6aJmIiJJAhf0oZAR1ZqxIiIJgQv6SMhUoxcRSRK4oFeNXkSkv8AFvWr0IiL9BS7ovRq9gl5EJC5wQa8avYhIf4EL+nDItPCIiEiSwAV9JBQiqhq9iEhC4II+rNKNiEg/gQx6TZgSEekTyKBXj15EpE/ggj6iCVMiIv0ELujDmjAlItJPIINeE6ZERPoEMuhVoxcR6RO4oFeNXkSkv8AFfTgUUo1eRCRJ4II+J2x0RzWOXkQkLq2gN7PrzGyHme02s3sGeTzPzJ72H19jZtX+9g+Z2Toz2+z/ffUIt3+AkvwcWjt7RvtlRETeNYYNejMLAw8D1wPzgdvMbH7KbncAjc65ucBDwAP+9hPAx5xzFwO3A0+OVMOHUlqYQ1O7gl5EJC6dHv3lwG7n3B7nXDfwFLAyZZ+VwBP+7WeBa8zMnHMbnHOH/e1bgQIzyxuJhg+ltCCHrt4YnT3R0XwZEZF3jXSCfgZwIOn+QX/boPs453qBZqA8ZZ+bgfXOua7UFzCzO82s1sxq6+vr0237oEoKcgBo6VCvXkQEMnQy1swW4JVzPjfY4865R51zS51zSysrK8/ptcr8oG9W0IuIAOkF/SFgVtL9mf62QfcxswhQCjT492cCPwU+7Zx751wbPJxSP+ibFPQiIkB6Qb8WmGdmNWaWC9wKrE7ZZzXeyVaAVcCrzjlnZmXAz4B7nHOvj1CbTyse9M06ISsiAqQR9H7N/W7gZWAb8IxzbquZ3W9mN/q7PQaUm9lu4EtAfAjm3cBc4F4z2+j/mTzi/4okZYUq3YiIJIuks5Nz7kXgxZRt9ybd7gRuGeR53wC+cY5tPCMq3YiI9Be4mbHF+erRi4gkC1zQh0NGcX5EwytFRHyBC3rw6vTq0YuIeAIZ9KUFOTS1d2e7GSIiY0Jgg149ehERTyCDvqwgV0EvIuILZNCXqEcvIpIQyKCPl26clhQUEQlm0JcV5tATdXToUsUiIsEM+lJdwVJEJCHQQa+VpkREAh706tGLiCjoRUQCL5BBH79UcWObZseKiAQy6MuLvPXHGxT0IiLBDPqC3DCFuWFOKuhFRIIZ9ACTinIV9CIiBDjoy4tyVboRESHAQe/16Luy3QwRkawLcNDncfKUevQiIoEN+vIJXulGFzYTkfEusEE/qSiXrt4Y7d26sJmIjG+BDfryolwAGlS+EZFxLrhBP8EPep2QFZFxLrBBP8mfHaux9CIy3gU26BOlGwW9iIxzgQ36SX7Qq0cvIuNdYIO+MDdMXiSkoBeRcS+wQW9m3mUQNOpGRMa5wAY9QPmEPI26EZFxL9BBrytYiogEPOhVuhERCXjQTynN53hrJ7GYrncjIuNXoIN+Wmk+PVGnsfQiMq4FOuinluQDcLS5M8stERHJnkAH/bTSAgCONHdkuSUiItmTVtCb2XVmtsPMdpvZPYM8nmdmT/uPrzGzan97uZn92sxOmdn3Rrjtw5pa6vfoWzrp6o1qBI6IjEvDBr2ZhYGHgeuB+cBtZjY/Zbc7gEbn3FzgIeABf3sn8DfAl0esxWegvCiXnLBxpLmT7/xqFx/97m+z0QwRkaxKp0d/ObDbObfHOdcNPAWsTNlnJfCEf/tZ4BozM+dcm3Pud3iBn3GhkDGlJJ+jzZ2srTvJ4eZOOrQQiYiMM+kE/QzgQNL9g/62QfdxzvUCzUB5uo0wszvNrNbMauvr69N9WlqmleZzqKmDtw+3AHC8VSdmRWR8GRMnY51zjzrnljrnllZWVo7oz55aWsCmA020+T354626JIKIjC/pBP0hYFbS/Zn+tkH3MbMIUAo0jEQDz9W00ny6emOJ+8dbFPQiMr6kE/RrgXlmVmNmucCtwOqUfVYDt/u3VwGvOufGxHTU+Fj6uGMtKt2IZNtLW47yg9f3ZrsZ48awQe/X3O8GXga2Ac8457aa2f1mdqO/22NAuZntBr4EJIZgmlkd8CDwGTM7OMiInVE1zR9iecGUYnLCptKNyBjw7LqDPP56XbabMW5E0tnJOfci8GLKtnuTbncCtwzx3OpzaN85i4+lXzCjhNbOHp2MFRkDunqjtHT2ZLsZ48aYOBk7mmZOLCRksHhWGZUl+dSrRy+SdR3dUVo6ehgjFd7AC3zQVxbnsfruFdz6niomF+fpZKzIGNDZGyXm4FRXb7abMi4EPugBFs4oJTcS8oJepRuRrItPXGzpVNBnwrgI+rjJxfk0tvfQnTTcUkQyr7PH+z/Y0qE6fSaMr6AvyQOg/pTKNyLZ1Nnj9eibFfQZMb6CvtgL+vhY+i2HmvVGE8mCDj/o1aPPjHEW9N5Qy+MtXbR09vCJR37PP7yyK8utEhlfnHOJHr1q9JkxvoI+Xrpp7eT1XSfo7o2x4UBTdhslMs50R2PEl3HWN+rMSGvCVFBUTMijvCiX3+ysZ1JRLgBbDzfTG40RCY+rzzyRrOns7hsModJNZoyrdAuHjE8uq+KV7cd5eesxinLDdPbE2HX8VLabJjJudPb2rQmh2bGZMa6CHuBTy2YTNqO5o4fbl1cDsPlgc3YbJTKOJC/+o9JNZoy7oJ9ams+HF04F4DPLqynOj7DpYFN2GyUyjvTr0XfoZGwmjKsafdzffnQ+Ny+ZweSSfC6eUcqmg03sPt7K9LICCnPH5SERyZjkHr1KN5kx7nr0AJNL8rn6wikALJpZxpZDLXzwwdf43JPrdJElkVEWnxVbkh/RydgMGffd108tq8Lh6OyO8sQb+/jhmv38yRWzs90skcCKj6GfUpKvoM+QcR/0syYV8pXrL8I5xzv1bXzzxW1cu2BKYnKViIysjqSg36h5LBkxLks3gzEz7l+5gPbuKP/yxr5sN0cksOI9+snFeZzq6qU3qosMjjYFfZI5lRP44EWT+Zc1+xNvRvBWw/nB63t11UuRERDv0U/213Nu1WUQRp2CPsWfrajhZFs3/7bhUGLbL98+xn3Pv80r245lsWUiwRA/GTvFvySJRt6MPgV9iivnlHPRtBIef31vYgTO24dbAKjd1zgir9Hc0cMzaw/0G+FztLmTE7p8sowDySdjQZOmMkFBn8LMuGNFDTuPneK3u04A8PYRP+jrTg77/N/srOd//XzbafdZvfEQf/XcW+w41prY9uc/XMdXf7r5HFo+cmIxRzSmYaYyOjp7ooTMu/YUaNJUJijoB/GxS6ZRWZzHY7/bC8A2P+i3Hm6hpbOH59YdpCtpdl+y776yi0df29Ovxp/qYFMHANuP9AX9nhNt1J1oH6l/wjl54OXtfOIfX892MySgOrqj5OeEKS3IAdSjzwQF/SDyImE+fcVsfrOznjV7GjjW0sWKuRX0xhxf/NcN/OWPN/H8piMDnne4qYN1+xpxDvY1DB3aR5q8hU+2HfU+QNq7e2lq7+Foy9hYz/aXbx9j86FmnXyWUdHRE6UgJ8y0Mq90U9fQluUWBZ+Cfgi3LasiNxzi3n/fCsCfXFGFmVeaAXjjnYbEvj/dcJB7nnuL1ZsOJ7btPTH0m/dIc/8e/WE/+Js7ek77TSATjrd2sqe+jZiDg41j4xuGBEtnT4z8nDAl+TlUTSpk62FdVHC0KeiHUDEhjxsunpqooy+rKeeCKcUU5YZZOnsif9jTgHOOtq5evvb82zy19gDfemk71eWFwOmDPh7s2/0e/WG/lAPeSdlMOnCyncu+/svECee1e/tOOKunJaOhsydKfo4XPQtnlLDVf+/J6FHQn8afXuldCmF6aT4Ti3L5xk0L+f7tS7lx8XQONXVw4GQHP1yzj6b2Hm59zyxiDv74PbOomJBH3RBBH405jrZ0UpQb5lhLFyfbuvsF/bE0yzcHTrbz6vahh3v2RmP9RvXEYo51+waeTP7trhM0tHUnHntzbwM5YQMYM+cMJFi8oA8DsGB6Kfsa2jXEcpQp6E9jSdVEllSVsWxOOQBLqyex/LwKrvTvv7D5MN//7V5WzK3g725exAtfXMGdV82hpqJwyB59fWsX0ZjjqnmVgNer79ejTzPo/+HVXdzxRC1bDg382tsTjXHt/3mNmx5+nZ3+N5Ln3zrMzY+8kfgWEbdhv9eD3+uH+pq9J1lWU05xfoR96tHLKIjX6AHmTy8B+oYwx+1raMt6GTNIFPSnYWb86M4r+PaqRf22z508gYoJeXzrpR20dPTwpWvPB2DhjFIi4RA1FUXsOdHGi5uPsOqR3/cLzEN+qF994WTAq9Mfbu6kON+77NDpevQ90RinuryhaJsPteAcfO35rQOuuPnzLUfZU9/GzmOnWPm916lv7aK2zgv03SmracXXzK1raKOpvZvtR1tZVjOJ6vIi9p7mhPKZ+PuXd3D/82+PyM+Sd7+OnigFuV7QL5xeCtCvfHOspZMPPfgaP/h9XTaaF0gK+mHkRcID1pM1Mz5y8VRqKop47q7lLKma2O/x6ooiTpzq4n/+bBu1+xq5+ZHfJ3rW8ROxi2aVUl6Uy5bDzRxu6mDe5AkU5oY52jz0pKn/9vRGbvze7+jsibLrWCvV5YWsrWvk3zce7rffP7++l+ryQn505xV09ET53e76xMWjkkcDNbf3JIK/rqGNTf5KW5fNnsjs8sJ+H1DRmEv7Es7N7f2/hr/w1mF+9Ob+IYekyvjS2RMjL+IFfWVxHpOL89ia9M30ufUH6Y7GVLsfQQr6s3TfjQv49Zffz8IZpQMem1NRBHi99/9yzTx6Y44Hf7ET6BtaOa20gCvOK+e1nfUcaGxnelkBU0vyh+zR19ad5IW3jrCnvo2fbzlCb8zxl9dewCWzyvja81sTs2o3HWhiw/4mbl9ezaIZpUwszOFXbx9PzAVIPnew0V9Z65KZpRw42Z4o4yycWUpNRREHGzvoicZo7+5l2Td/xVNrDwx7XH64Zh+Lv/4LfrL+IODVY/efbKejJ8q6EZpZLO9unUk9eoAF00tYv78x0Zn4ca333kn99ilnT0F/lsxsyMdqKiYA3siduz8wl5sWz+DVHcdp7ujhUFMHE/IilORHuG7BVE6c6ubAyQ5mlBUwpSR/0Bq9c45vvriNskJvgskj//EOAJfMLOPbqxbR1hXlb/1hoM9vOkxuJMSqy2YSChnL51YkPhjCIWPfyb4e/Yb9jZjBjYtn0BN1vLTlKHMqiyjJz2F2eRHRmONQYwe/23WCE6e6WZ3yzSHVsZZO/u7F7YTN+Ovn3uLNvScTQzWBxExjGd86e6LkR/qi56ZLZ1DX0M4Pfl9H7b5G9p5oY1ppPu/Un9IM7RGioB8Fs8sLKc6P8J/fW01uJMRNl86guzfGy1uOcqS5g2ml+ZgZ77+gkly/LDS9rICppfkcbe5k+9EWXt56lMa2bgAe+c07rN/fxH+//iLmVBSx89gpivMjzJpUwPlTivn8+8/jZ5uPsL+hnd/tPsF7qidSnO99KKyYW5EI2qvmVfQrx6zf38QFU4pZ6J8Q2360lUtmlgH0DRNtaOPXO44DsLbuJK2nGR3x9Rfepisa47m7ljO1NJ8HXtrOruNeyaqyOI/X/DkIMr51pPTob7xkOldfOJlvvbSdzzz+JiX5Ee583xy6e2OayzFCFPSjID8nzG//6gPc9UfnAV5ppLq8kB+vO0DdiXamlRUAUJyfw3vneiN4pvs9+uOtnfzpY2/yuSfXseQbv+TmR37Pt1/ewY2XTOeWpTN53/neaJ0F00sS3ypuuWwmAE+8Ucf2o62smFuZaMuKuRUAzJpUwGVVEznW0kVHd5RYzLFxfyOXVpVR7ZeaAC72S1FzKidg5s2SfWXbcWaUFdAbc7y28wQP/XInb6UsqL79aAsvvHWEz71vDpfMKmPlJTPY6JeRwiHjtsur2Hq4RRduE690k9MX9GbGNz9+MTMmFvDB+VP4yZ8vZ5Hf4VD5ZmQo6EdJWWEuoZAXxGbGxy+dydq6RnYca6VqUkFiv+svngZ4PeipJXn0RB31rV38/S2X8MWr59HU3s1lVRN54OZFmBlXzfOCOz5aAbxVshbNLE2MUojvE3/swqnFvPe8Cmb7gb7/ZDt7TrTR0tnLpVUTmVycR6Hfw7pklvdzJxXl8pnl1fzrmv0cb+3ii1fPpTg/wlf/bTPfeWVXYsZw3CP/8Q6FuWHuWFEDwPK55URjjufWH2R2eSHXzvfW6H06jTq/BFcs5ryTsUlBDzC1NJ9X//L9fOfWS5k7uZi5k73y5y4F/YgY90sJZsrn3z+HxVVlnOrs5Yo5kxLbVy2ZydzJE5g3pZh36r039S2XzWSV30v/0ofO7/dzrjyvnMtrJvHhhVP7bb/h4mm8dbCZSUW5zJ9W0u+xZ+9aTiRk7DjqlVHqGtoSF5JaUlWGmTG7vIidx1qZP63vA+TL117AL7Ye41BTB9dcNIXf7jrBzzYfoaaiiI0Hmth0oIlLZpWxr6GN5zcd5rNXzaGsMNf/uRPJzwnR2tnL8vPKWTijlA/Nn8LDv97NqstmJi5RK+NLl3/9pIKUoE9VWpBDZXGeevQjRD36DMmLhPmj8yv5yKJplPuXZwUIhSwxPPPKORV8clkV91x/4ZA/pzA3wjOfu5L3VE/qt/2Ghd43g+XnlSe+ScRNyIuQnxOmutzv0Te0s2F/EyX5Eeb4J44XzyrlsqqJ/WqnRXkRHvmTJXz9poVUFufx2atq+LP31vCTu5ZTlBvmCf8bxD+/Xkc4ZHzW782DV76Kt3He5GIA/sdHLqI36vjWSzvSP3ASKPFJUPFLIJzO3MoJbDvSwk/WH+TASdXqz0VaQW9m15nZDjPbbWb3DPJ4npk97T++xsyqkx77ir99h5l9eATbHjilhTl88+MX9/sgSFdVeSH3r1zAFz4w97Q/v7Qgh7qGNjbsb2Rx1cTEh8LXVy7kyc9ePuA5i2aW8adXeJeCuLRqIvd+bD4Ti3JZddlMXnjrCBv2N/LsuoN8dNH0xNJwcfHzA/OmeB8ms8uLuOOqGp5bf1CLQo9T8WUEh+vRg/e+2Xq4hS89s4nPPlGrmbLnYNigN7Mw8DBwPTAfuM3M5qfsdgfQ6JybCzwEPOA/dz5wK7AAuA74R//nySj49JXVXJRStkm1YHoJP91wiJ3HWrl0VllieyQcSkxiSccXrp7LhPwIt33/D5zq6uX25dUD9rnh4mlcNnsiV/iXjAD4wgfmUlmcx32rt9LW1UtPyjV5JNj6evTDv9f+eOksPrmsins/Op8dx1q5/4W3aevSIiVnI50a/eXAbufcHgAzewpYCSTPaV8J3Offfhb4nnlDQlYCTznnuoC9Zrbb/3lvjEzz5Uw99J8Wc9e/rGP9/iaWVk8c/glDmFycz7dXLeKOJ2pZPKuMxUkfGnGzJhXy3F3L+22bkBfhr6+7kC//eBML/vblxPbccIicsJETCZETDiXuh04zX2FY2XnqaedYjObrvhvEa/TplG4Wzijlmx+/GIADje388+t1PLP2ABOLcgkZhMx7f4RCfbeHk9bxTWOndH5OOu+D1D3ef0ElX/1Iaj/63KUT9DOA5KESB4FlQ+3jnOs1s2ag3N/+h5Tnzkh9ATO7E7gToKqqKt22y1mYUpLPj+68gvX7mvqdFD4b11w0hX/81BLOn1J8Rs/7xKUzyM8JccifedsddfREY/T0xvrd7+6NcbZ9/XP5lnBO3y/O4cnu3F75XWNp9cQB55iG8zcfmc81F07hjT0naGzvIRZzxJwjGvN+11HnGO5Xns7RTed9k9ZvKY2dBvt9j9YghTEx6sY59yjwKMDSpUvHx7s9i/IiYa48r3z4HdNwgz889EyEQsZHF00fkdeX8SEUMlbMq2BF0tBhSV86J2MPAbOS7s/0tw26j5lFgFKgIc3niojIKEon6NcC88ysxsxy8U6urk7ZZzVwu397FfCq874DrQZu9Ufl1ADzgDdHpukiIpKOYUs3fs39buBlIAw87pzbamb3A7XOudXAY8CT/snWk3gfBvj7PYN34rYX+IJzTmOkREQyyMba0LalS5e62trabDdDRORdxczWOeeWDvaYZsaKiAScgl5EJOAU9CIiAaegFxEJuDF3MtbM6oF95/AjKoCxuGad2nVm1K4zN1bbpnadmbNt12znXOVgD4y5oD9XZlY71JnnbFK7zozadebGatvUrjMzGu1S6UZEJOAU9CIiARfEoH802w0Ygtp1ZtSuMzdW26Z2nZkRb1fgavQiItJfEHv0IiKSREEvIhJwgQn64RYwz2A7ZpnZr83sbTPbamb/1d9+n5kdMrON/p8bstS+OjPb7Leh1t82ycx+aWa7/L/Pfo3Bs2vTBUnHZaOZtZjZX2TjmJnZ42Z23My2JG0b9PiY57v+e+4tM1uS4XZ928y2+6/9UzMr87dXm1lH0nH7p9Fq12naNuTvzsy+4h+zHWb24Qy36+mkNtWZ2UZ/e8aO2WkyYvTeZ865d/0fvMsnvwPMAXKBTcD8LLVlGrDEv10M7MRbVP0+4Mtj4FjVARUp274F3OPfvgd4IMu/y6PA7GwcM+B9wBJgy3DHB7gB+Dne0p9XAGsy3K5rgYh/+4GkdlUn75elYzbo787/v7AJyANq/P+34Uy1K+Xx/w3cm+ljdpqMGLX3WVB69IkFzJ1z3UB8AfOMc84dcc6t92+3AtsYZJ3cMWYl8IR/+wngpuw1hWuAd5xz5zI7+qw5517DW1Mh2VDHZyXw/5znD0CZmZ352opn2S7n3C+cc73+3T/greCWcUMcs6GsBJ5yznU55/YCu/H+/2a0XWZmwB8DPxqN1z6d02TEqL3PghL0gy1gnvVwNbNq4FJgjb/pbv+r1+OZLo8kccAvzGydeYuyA0xxzh3xbx8FpmSnaYC3aE3yf76xcMyGOj5j6X33Z3i9vrgaM9tgZr8xs6uy1KbBfndj5ZhdBRxzzu1K2pbxY5aSEaP2PgtK0I85ZjYBeA74C+dcC/AIcB6wGDiC97UxG1Y455YA1wNfMLP3JT/ovO+KWRlza95SlTcCP/Y3jZVjlpDN4zMUM/sq3gpuP/Q3HQGqnHOXAl8C/tXMSjLcrDH3u0txG/07FBk/ZoNkRMJIv8+CEvRjahFyM8vB+wX+0Dn3EwDn3DHnXNQ5FwO+zyh9XR2Oc+6Q//dx4Kd+O47Fvwr6fx/PRtvwPnzWO+eO+W0cE8eMoY9P1t93ZvYZ4KPAp/xwwC+LNPi31+HVwc/PZLtO87sbC8csAnwCeDq+LdPHbLCMYBTfZ0EJ+nQWMM8Iv/b3GLDNOfdg0vbkmtrHgS2pz81A24rMrDh+G+9k3hb6L+5+O/DvmW6br18vaywcM99Qx2c18Gl/VMQVQHPSV+9RZ2bXAX8F3Oica0/aXmlmYf/2HGAesCdT7fJfd6jf3WrgVjPLM7Mav21vZrJtwAeB7c65g/ENmTxmQ2UEo/k+y8RZ5kz8wTszvRPvk/irWWzHCryvXG8BG/0/NwBPApv97auBaVlo2xy8EQ+bgK3x4wSUA68Au4BfAZOy0LYioAEoTdqW8WOG90FzBOjBq4XeMdTxwRsF8bD/ntsMLM1wu3bj1W7j77N/8ve92f/9bgTWAx/LwjEb8ncHfNU/ZjuA6zPZLn/7D4DPp+ybsWN2mowYtfeZLoEgIhJwQSndiIjIEBT0IiIBp6AXEQk4Bb2ISMAp6EVEAk5BLyIScAp6EZGA+/+k1Aj5fqe6PAAAAABJRU5ErkJggg==\n"
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "import matplotlib.pyplot as plt\n",
    "\n",
    "plt.plot(errs)\n",
    "plt.show()"
   ],
   "metadata": {
    "collapsed": false,
    "pycharm": {
     "name": "#%%\n"
    }
   }
  },
  {
   "cell_type": "code",
   "execution_count": 121,
   "outputs": [
    {
     "data": {
      "text/plain": "<Figure size 432x288 with 1 Axes>",
      "image/png": "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\n"
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "import matplotlib.pyplot as plt\n",
    "\n",
    "plt.plot(accs)\n",
    "plt.show()"
   ],
   "metadata": {
    "collapsed": false,
    "pycharm": {
     "name": "#%%\n"
    }
   }
  },
  {
   "cell_type": "code",
   "execution_count": 179,
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[(22, 145), (22, 52), (22, 175), (22, 164), (22, 105), (22, 202), (22, 164), (22, 147), (22, 155), (22, 172), (22, 252), (22, 112)]\n",
      "(0, '0_DianaHappy.wav')\n",
      "tensor(0., grad_fn=<SumBackward0>)\n",
      " true:  0   model:  0\n",
      "tensor(0., grad_fn=<SumBackward0>)\n",
      " true:  0   model:  0\n",
      "tensor(0., grad_fn=<SumBackward0>)\n",
      " true:  0   model:  0\n",
      "tensor(0., grad_fn=<SumBackward0>)\n",
      " true:  0   model:  0\n",
      "tensor(0., grad_fn=<SumBackward0>)\n",
      " true:  0   model:  0\n",
      "tensor(0., grad_fn=<SumBackward0>)\n",
      " true:  0   model:  0\n",
      "tensor(0., grad_fn=<SumBackward0>)\n",
      " true:  0   model:  0\n",
      "tensor(0., grad_fn=<SumBackward0>)\n",
      " true:  0   model:  0\n",
      "tensor(0., grad_fn=<SumBackward0>)\n",
      " true:  0   model:  0\n",
      "(0, '0_DianaHappy2.wav')\n",
      "tensor(0., grad_fn=<SumBackward0>)\n",
      " true:  0   model:  1\n",
      "(0, '0_DianaHappy3.wav')\n",
      "tensor(0., grad_fn=<SumBackward0>)\n",
      " true:  0   model:  0\n",
      "tensor(0., grad_fn=<SumBackward0>)\n",
      " true:  0   model:  0\n",
      "tensor(0., grad_fn=<SumBackward0>)\n",
      " true:  0   model:  0\n",
      "tensor(0., grad_fn=<SumBackward0>)\n",
      " true:  0   model:  0\n",
      "tensor(0., grad_fn=<SumBackward0>)\n",
      " true:  0   model:  0\n",
      "tensor(0., grad_fn=<SumBackward0>)\n",
      " true:  0   model:  0\n",
      "tensor(0., grad_fn=<SumBackward0>)\n",
      " true:  0   model:  0\n",
      "tensor(0., grad_fn=<SumBackward0>)\n",
      " true:  0   model:  0\n",
      "tensor(0., grad_fn=<SumBackward0>)\n",
      " true:  0   model:  0\n",
      "tensor(0., grad_fn=<SumBackward0>)\n",
      " true:  0   model:  0\n",
      "tensor(0., grad_fn=<SumBackward0>)\n",
      " true:  0   model:  0\n",
      "tensor(0., grad_fn=<SumBackward0>)\n",
      " true:  0   model:  0\n",
      "tensor(0., grad_fn=<SumBackward0>)\n",
      " true:  0   model:  0\n",
      "tensor(0., grad_fn=<SumBackward0>)\n",
      " true:  0   model:  0\n",
      "tensor(0., grad_fn=<SumBackward0>)\n",
      " true:  0   model:  0\n",
      "tensor(0., grad_fn=<SumBackward0>)\n",
      " true:  0   model:  0\n",
      "tensor(0., grad_fn=<SumBackward0>)\n",
      " true:  0   model:  0\n",
      "tensor(0., grad_fn=<SumBackward0>)\n",
      " true:  0   model:  0\n",
      "tensor(0., grad_fn=<SumBackward0>)\n",
      " true:  0   model:  0\n",
      "tensor(0., grad_fn=<SumBackward0>)\n",
      " true:  0   model:  0\n",
      "tensor(0., grad_fn=<SumBackward0>)\n",
      " true:  0   model:  0\n",
      "tensor(0., grad_fn=<SumBackward0>)\n",
      " true:  0   model:  0\n",
      "tensor(0., grad_fn=<SumBackward0>)\n",
      " true:  0   model:  0\n",
      "tensor(0., grad_fn=<SumBackward0>)\n",
      " true:  0   model:  0\n",
      "tensor(0., grad_fn=<SumBackward0>)\n",
      " true:  0   model:  0\n",
      "tensor(0., grad_fn=<SumBackward0>)\n",
      " true:  0   model:  0\n",
      "tensor(0., grad_fn=<SumBackward0>)\n",
      " true:  0   model:  0\n",
      "tensor(0., grad_fn=<SumBackward0>)\n",
      " true:  0   model:  0\n",
      "tensor(0., grad_fn=<SumBackward0>)\n",
      " true:  0   model:  0\n",
      "tensor(0., grad_fn=<SumBackward0>)\n",
      " true:  0   model:  0\n",
      "tensor(0., grad_fn=<SumBackward0>)\n",
      " true:  0   model:  0\n",
      "tensor(0., grad_fn=<SumBackward0>)\n",
      " true:  0   model:  0\n",
      "tensor(0., grad_fn=<SumBackward0>)\n",
      " true:  0   model:  0\n",
      "tensor(0., grad_fn=<SumBackward0>)\n",
      " true:  0   model:  0\n",
      "tensor(0., grad_fn=<SumBackward0>)\n",
      " true:  0   model:  0\n",
      "tensor(0., grad_fn=<SumBackward0>)\n",
      " true:  0   model:  0\n",
      "tensor(0., grad_fn=<SumBackward0>)\n",
      " true:  0   model:  0\n",
      "tensor(0., grad_fn=<SumBackward0>)\n",
      " true:  0   model:  0\n",
      "tensor(0., grad_fn=<SumBackward0>)\n",
      " true:  0   model:  0\n",
      "(1, '1_DianaNeutral.wav')\n",
      "tensor(0., grad_fn=<SumBackward0>)\n",
      " true:  1   model:  0\n",
      "tensor(0., grad_fn=<SumBackward0>)\n",
      " true:  1   model:  1\n",
      "tensor(0., grad_fn=<SumBackward0>)\n",
      " true:  1   model:  0\n",
      "tensor(0., grad_fn=<SumBackward0>)\n",
      " true:  1   model:  1\n",
      "tensor(0., grad_fn=<SumBackward0>)\n",
      " true:  1   model:  1\n",
      "tensor(0., grad_fn=<SumBackward0>)\n",
      " true:  1   model:  1\n",
      "tensor(0., grad_fn=<SumBackward0>)\n",
      " true:  1   model:  0\n",
      "tensor(0., grad_fn=<SumBackward0>)\n",
      " true:  1   model:  1\n",
      "tensor(0., grad_fn=<SumBackward0>)\n",
      " true:  1   model:  0\n",
      "tensor(0., grad_fn=<SumBackward0>)\n",
      " true:  1   model:  1\n",
      "tensor(0., grad_fn=<SumBackward0>)\n",
      " true:  1   model:  1\n",
      "tensor(0., grad_fn=<SumBackward0>)\n",
      " true:  1   model:  0\n",
      "tensor(0., grad_fn=<SumBackward0>)\n",
      " true:  1   model:  0\n",
      "tensor(0., grad_fn=<SumBackward0>)\n",
      " true:  1   model:  0\n",
      "tensor(0., grad_fn=<SumBackward0>)\n",
      " true:  1   model:  0\n",
      "tensor(0., grad_fn=<SumBackward0>)\n",
      " true:  1   model:  0\n",
      "tensor(0., grad_fn=<SumBackward0>)\n",
      " true:  1   model:  0\n",
      "tensor(0., grad_fn=<SumBackward0>)\n",
      " true:  1   model:  0\n",
      "tensor(0., grad_fn=<SumBackward0>)\n",
      " true:  1   model:  0\n",
      "tensor(0., grad_fn=<SumBackward0>)\n",
      " true:  1   model:  1\n",
      "tensor(0., grad_fn=<SumBackward0>)\n",
      " true:  1   model:  0\n",
      "tensor(0., grad_fn=<SumBackward0>)\n",
      " true:  1   model:  1\n",
      "tensor(0., grad_fn=<SumBackward0>)\n",
      " true:  1   model:  0\n",
      "tensor(0., grad_fn=<SumBackward0>)\n",
      " true:  1   model:  0\n",
      "tensor(0., grad_fn=<SumBackward0>)\n",
      " true:  1   model:  0\n",
      "tensor(0., grad_fn=<SumBackward0>)\n",
      " true:  1   model:  0\n",
      "tensor(0., grad_fn=<SumBackward0>)\n",
      " true:  1   model:  0\n",
      "tensor(0., grad_fn=<SumBackward0>)\n",
      " true:  1   model:  0\n",
      "(2, '2_9132330761343643383.wav')\n",
      "tensor(0., grad_fn=<SumBackward0>)\n",
      " true:  2   model:  2\n",
      "(2, '2_DianaAngry1.wav')\n",
      "tensor(0., grad_fn=<SumBackward0>)\n",
      " true:  2   model:  1\n",
      "tensor(0., grad_fn=<SumBackward0>)\n",
      " true:  2   model:  1\n",
      "tensor(0., grad_fn=<SumBackward0>)\n",
      " true:  2   model:  1\n",
      "tensor(0., grad_fn=<SumBackward0>)\n",
      " true:  2   model:  1\n",
      "tensor(0., grad_fn=<SumBackward0>)\n",
      " true:  2   model:  1\n",
      "tensor(0., grad_fn=<SumBackward0>)\n",
      " true:  2   model:  1\n",
      "tensor(0., grad_fn=<SumBackward0>)\n",
      " true:  2   model:  1\n",
      "tensor(0., grad_fn=<SumBackward0>)\n",
      " true:  2   model:  1\n",
      "tensor(0., grad_fn=<SumBackward0>)\n",
      " true:  2   model:  1\n",
      "tensor(0., grad_fn=<SumBackward0>)\n",
      " true:  2   model:  1\n",
      "tensor(0., grad_fn=<SumBackward0>)\n",
      " true:  2   model:  1\n",
      "tensor(0., grad_fn=<SumBackward0>)\n",
      " true:  2   model:  1\n",
      "tensor(0., grad_fn=<SumBackward0>)\n",
      " true:  2   model:  1\n",
      "tensor(0., grad_fn=<SumBackward0>)\n",
      " true:  2   model:  1\n",
      "tensor(0., grad_fn=<SumBackward0>)\n",
      " true:  2   model:  1\n",
      "tensor(0., grad_fn=<SumBackward0>)\n",
      " true:  2   model:  1\n",
      "tensor(0., grad_fn=<SumBackward0>)\n",
      " true:  2   model:  1\n",
      "tensor(0., grad_fn=<SumBackward0>)\n",
      " true:  2   model:  1\n",
      "tensor(0., grad_fn=<SumBackward0>)\n",
      " true:  2   model:  1\n",
      "tensor(0., grad_fn=<SumBackward0>)\n",
      " true:  2   model:  1\n",
      "tensor(0., grad_fn=<SumBackward0>)\n",
      " true:  2   model:  1\n",
      "tensor(0., grad_fn=<SumBackward0>)\n",
      " true:  2   model:  1\n",
      "tensor(0., grad_fn=<SumBackward0>)\n",
      " true:  2   model:  1\n",
      "tensor(0., grad_fn=<SumBackward0>)\n",
      " true:  2   model:  1\n",
      "tensor(0., grad_fn=<SumBackward0>)\n",
      " true:  2   model:  1\n",
      "tensor(0., grad_fn=<SumBackward0>)\n",
      " true:  2   model:  1\n",
      "tensor(0., grad_fn=<SumBackward0>)\n",
      " true:  2   model:  1\n",
      "tensor(0., grad_fn=<SumBackward0>)\n",
      " true:  2   model:  1\n",
      "tensor(0., grad_fn=<SumBackward0>)\n",
      " true:  2   model:  1\n",
      "tensor(0., grad_fn=<SumBackward0>)\n",
      " true:  2   model:  1\n",
      "tensor(0., grad_fn=<SumBackward0>)\n",
      " true:  2   model:  1\n",
      "tensor(0., grad_fn=<SumBackward0>)\n",
      " true:  2   model:  1\n",
      "tensor(0., grad_fn=<SumBackward0>)\n",
      " true:  2   model:  1\n",
      "tensor(0., grad_fn=<SumBackward0>)\n",
      " true:  2   model:  1\n",
      "tensor(0., grad_fn=<SumBackward0>)\n",
      " true:  2   model:  1\n",
      "tensor(0., grad_fn=<SumBackward0>)\n",
      " true:  2   model:  1\n",
      "tensor(0., grad_fn=<SumBackward0>)\n",
      " true:  2   model:  1\n",
      "tensor(0., grad_fn=<SumBackward0>)\n",
      " true:  2   model:  1\n",
      "tensor(0., grad_fn=<SumBackward0>)\n",
      " true:  2   model:  1\n",
      "tensor(0., grad_fn=<SumBackward0>)\n",
      " true:  2   model:  1\n",
      "tensor(0., grad_fn=<SumBackward0>)\n",
      " true:  2   model:  1\n",
      "tensor(0., grad_fn=<SumBackward0>)\n",
      " true:  2   model:  1\n",
      "tensor(0., grad_fn=<SumBackward0>)\n",
      " true:  2   model:  1\n",
      "tensor(0., grad_fn=<SumBackward0>)\n",
      " true:  2   model:  1\n",
      "tensor(0., grad_fn=<SumBackward0>)\n",
      " true:  2   model:  1\n",
      "tensor(0., grad_fn=<SumBackward0>)\n",
      " true:  2   model:  1\n",
      "tensor(0., grad_fn=<SumBackward0>)\n",
      " true:  2   model:  1\n",
      "tensor(0., grad_fn=<SumBackward0>)\n",
      " true:  2   model:  1\n",
      "tensor(0., grad_fn=<SumBackward0>)\n",
      " true:  2   model:  1\n",
      "tensor(0., grad_fn=<SumBackward0>)\n",
      " true:  2   model:  1\n",
      "tensor(0., grad_fn=<SumBackward0>)\n",
      " true:  2   model:  1\n",
      "tensor(0., grad_fn=<SumBackward0>)\n",
      " true:  2   model:  1\n",
      "tensor(0., grad_fn=<SumBackward0>)\n",
      " true:  2   model:  1\n",
      "tensor(0., grad_fn=<SumBackward0>)\n",
      " true:  2   model:  1\n",
      "tensor(0., grad_fn=<SumBackward0>)\n",
      " true:  2   model:  1\n",
      "tensor(0., grad_fn=<SumBackward0>)\n",
      " true:  2   model:  1\n",
      "tensor(0., grad_fn=<SumBackward0>)\n",
      " true:  2   model:  1\n",
      "tensor(0., grad_fn=<SumBackward0>)\n",
      " true:  2   model:  1\n",
      "tensor(0., grad_fn=<SumBackward0>)\n",
      " true:  2   model:  1\n",
      "tensor(0., grad_fn=<SumBackward0>)\n",
      " true:  2   model:  1\n",
      "tensor(0., grad_fn=<SumBackward0>)\n",
      " true:  2   model:  1\n",
      "tensor(0., grad_fn=<SumBackward0>)\n",
      " true:  2   model:  1\n",
      "tensor(0., grad_fn=<SumBackward0>)\n",
      " true:  2   model:  1\n",
      "tensor(0., grad_fn=<SumBackward0>)\n",
      " true:  2   model:  1\n",
      "tensor(0., grad_fn=<SumBackward0>)\n",
      " true:  2   model:  1\n",
      "tensor(0., grad_fn=<SumBackward0>)\n",
      " true:  2   model:  1\n",
      "(2, '2_DianaAngry2.wav')\n",
      "tensor(0., grad_fn=<SumBackward0>)\n",
      " true:  2   model:  0\n",
      "tensor(0., grad_fn=<SumBackward0>)\n",
      " true:  2   model:  0\n",
      "tensor(0., grad_fn=<SumBackward0>)\n",
      " true:  2   model:  0\n",
      "tensor(0., grad_fn=<SumBackward0>)\n",
      " true:  2   model:  0\n",
      "tensor(0., grad_fn=<SumBackward0>)\n",
      " true:  2   model:  0\n",
      "tensor(0., grad_fn=<SumBackward0>)\n",
      " true:  2   model:  0\n",
      "tensor(0., grad_fn=<SumBackward0>)\n",
      " true:  2   model:  0\n",
      "tensor(0., grad_fn=<SumBackward0>)\n",
      " true:  2   model:  0\n",
      "tensor(0., grad_fn=<SumBackward0>)\n",
      " true:  2   model:  0\n",
      "tensor(0., grad_fn=<SumBackward0>)\n",
      " true:  2   model:  0\n",
      "tensor(0., grad_fn=<SumBackward0>)\n",
      " true:  2   model:  0\n",
      "tensor(0., grad_fn=<SumBackward0>)\n",
      " true:  2   model:  0\n",
      "tensor(0., grad_fn=<SumBackward0>)\n",
      " true:  2   model:  0\n",
      "tensor(0., grad_fn=<SumBackward0>)\n",
      " true:  2   model:  0\n",
      "tensor(0., grad_fn=<SumBackward0>)\n",
      " true:  2   model:  0\n",
      "tensor(0., grad_fn=<SumBackward0>)\n",
      " true:  2   model:  0\n",
      "tensor(0., grad_fn=<SumBackward0>)\n",
      " true:  2   model:  0\n",
      "tensor(0., grad_fn=<SumBackward0>)\n",
      " true:  2   model:  0\n",
      "tensor(0., grad_fn=<SumBackward0>)\n",
      " true:  2   model:  0\n",
      "tensor(0., grad_fn=<SumBackward0>)\n",
      " true:  2   model:  0\n",
      "tensor(0., grad_fn=<SumBackward0>)\n",
      " true:  2   model:  0\n",
      "tensor(0., grad_fn=<SumBackward0>)\n",
      " true:  2   model:  0\n",
      "tensor(0., grad_fn=<SumBackward0>)\n",
      " true:  2   model:  0\n",
      "tensor(0., grad_fn=<SumBackward0>)\n",
      " true:  2   model:  0\n",
      "tensor(0., grad_fn=<SumBackward0>)\n",
      " true:  2   model:  0\n",
      "tensor(0., grad_fn=<SumBackward0>)\n",
      " true:  2   model:  0\n",
      "tensor(0., grad_fn=<SumBackward0>)\n",
      " true:  2   model:  0\n",
      "tensor(0., grad_fn=<SumBackward0>)\n",
      " true:  2   model:  0\n",
      "(2, '2_DianaAngry3.wav')\n",
      "tensor(0., grad_fn=<SumBackward0>)\n",
      " true:  2   model:  0\n",
      "tensor(0., grad_fn=<SumBackward0>)\n",
      " true:  2   model:  1\n",
      "tensor(0., grad_fn=<SumBackward0>)\n",
      " true:  2   model:  0\n",
      "tensor(0., grad_fn=<SumBackward0>)\n",
      " true:  2   model:  0\n",
      "tensor(0., grad_fn=<SumBackward0>)\n",
      " true:  2   model:  0\n",
      "tensor(0., grad_fn=<SumBackward0>)\n",
      " true:  2   model:  1\n",
      "tensor(0., grad_fn=<SumBackward0>)\n",
      " true:  2   model:  1\n",
      "tensor(0., grad_fn=<SumBackward0>)\n",
      " true:  2   model:  1\n",
      "tensor(0., grad_fn=<SumBackward0>)\n",
      " true:  2   model:  1\n",
      "tensor(0., grad_fn=<SumBackward0>)\n",
      " true:  2   model:  1\n",
      "tensor(0., grad_fn=<SumBackward0>)\n",
      " true:  2   model:  0\n",
      "(2, '2_DianaAngry4.wav')\n",
      "tensor(0., grad_fn=<SumBackward0>)\n",
      " true:  2   model:  0\n",
      "tensor(0., grad_fn=<SumBackward0>)\n",
      " true:  2   model:  0\n",
      "tensor(0., grad_fn=<SumBackward0>)\n",
      " true:  2   model:  0\n",
      "tensor(0., grad_fn=<SumBackward0>)\n",
      " true:  2   model:  0\n",
      "tensor(0., grad_fn=<SumBackward0>)\n",
      " true:  2   model:  0\n",
      "tensor(0., grad_fn=<SumBackward0>)\n",
      " true:  2   model:  0\n",
      "tensor(0., grad_fn=<SumBackward0>)\n",
      " true:  2   model:  0\n",
      "tensor(0., grad_fn=<SumBackward0>)\n",
      " true:  2   model:  0\n",
      "tensor(0., grad_fn=<SumBackward0>)\n",
      " true:  2   model:  0\n",
      "tensor(0., grad_fn=<SumBackward0>)\n",
      " true:  2   model:  0\n",
      "tensor(0., grad_fn=<SumBackward0>)\n",
      " true:  2   model:  0\n",
      "tensor(0., grad_fn=<SumBackward0>)\n",
      " true:  2   model:  0\n",
      "tensor(0., grad_fn=<SumBackward0>)\n",
      " true:  2   model:  0\n",
      "tensor(0., grad_fn=<SumBackward0>)\n",
      " true:  2   model:  0\n",
      "tensor(0., grad_fn=<SumBackward0>)\n",
      " true:  2   model:  0\n",
      "tensor(0., grad_fn=<SumBackward0>)\n",
      " true:  2   model:  0\n",
      "tensor(0., grad_fn=<SumBackward0>)\n",
      " true:  2   model:  0\n",
      "tensor(0., grad_fn=<SumBackward0>)\n",
      " true:  2   model:  0\n",
      "tensor(0., grad_fn=<SumBackward0>)\n",
      " true:  2   model:  0\n",
      "(2, '2_DianaAngry5.wav')\n",
      "tensor(0., grad_fn=<SumBackward0>)\n",
      " true:  2   model:  1\n",
      "tensor(0., grad_fn=<SumBackward0>)\n",
      " true:  2   model:  1\n",
      "tensor(0., grad_fn=<SumBackward0>)\n",
      " true:  2   model:  1\n",
      "tensor(0., grad_fn=<SumBackward0>)\n",
      " true:  2   model:  1\n",
      "tensor(0., grad_fn=<SumBackward0>)\n",
      " true:  2   model:  1\n",
      "tensor(0., grad_fn=<SumBackward0>)\n",
      " true:  2   model:  1\n",
      "tensor(0., grad_fn=<SumBackward0>)\n",
      " true:  2   model:  1\n",
      "tensor(0., grad_fn=<SumBackward0>)\n",
      " true:  2   model:  1\n",
      "tensor(0., grad_fn=<SumBackward0>)\n",
      " true:  2   model:  1\n",
      "tensor(0., grad_fn=<SumBackward0>)\n",
      " true:  2   model:  1\n",
      "tensor(0., grad_fn=<SumBackward0>)\n",
      " true:  2   model:  1\n",
      "tensor(0., grad_fn=<SumBackward0>)\n",
      " true:  2   model:  1\n",
      "tensor(0., grad_fn=<SumBackward0>)\n",
      " true:  2   model:  1\n",
      "tensor(0., grad_fn=<SumBackward0>)\n",
      " true:  2   model:  1\n",
      "tensor(0., grad_fn=<SumBackward0>)\n",
      " true:  2   model:  1\n",
      "tensor(0., grad_fn=<SumBackward0>)\n",
      " true:  2   model:  1\n",
      "tensor(0., grad_fn=<SumBackward0>)\n",
      " true:  2   model:  1\n",
      "tensor(0., grad_fn=<SumBackward0>)\n",
      " true:  2   model:  1\n",
      "tensor(0., grad_fn=<SumBackward0>)\n",
      " true:  2   model:  1\n",
      "tensor(0., grad_fn=<SumBackward0>)\n",
      " true:  2   model:  1\n",
      "tensor(0., grad_fn=<SumBackward0>)\n",
      " true:  2   model:  1\n",
      "tensor(0., grad_fn=<SumBackward0>)\n",
      " true:  2   model:  1\n",
      "tensor(0., grad_fn=<SumBackward0>)\n",
      " true:  2   model:  1\n",
      "tensor(0., grad_fn=<SumBackward0>)\n",
      " true:  2   model:  1\n",
      "tensor(0., grad_fn=<SumBackward0>)\n",
      " true:  2   model:  1\n",
      "tensor(0., grad_fn=<SumBackward0>)\n",
      " true:  2   model:  1\n",
      "tensor(0., grad_fn=<SumBackward0>)\n",
      " true:  2   model:  1\n",
      "tensor(0., grad_fn=<SumBackward0>)\n",
      " true:  2   model:  1\n",
      "tensor(0., grad_fn=<SumBackward0>)\n",
      " true:  2   model:  1\n",
      "tensor(0., grad_fn=<SumBackward0>)\n",
      " true:  2   model:  1\n",
      "tensor(0., grad_fn=<SumBackward0>)\n",
      " true:  2   model:  1\n",
      "tensor(0., grad_fn=<SumBackward0>)\n",
      " true:  2   model:  1\n",
      "tensor(0., grad_fn=<SumBackward0>)\n",
      " true:  2   model:  1\n",
      "tensor(0., grad_fn=<SumBackward0>)\n",
      " true:  2   model:  1\n",
      "tensor(0., grad_fn=<SumBackward0>)\n",
      " true:  2   model:  1\n",
      "tensor(0., grad_fn=<SumBackward0>)\n",
      " true:  2   model:  1\n",
      "(2, '2_DianaAngry6.wav')\n",
      "tensor(0., grad_fn=<SumBackward0>)\n",
      " true:  2   model:  0\n",
      "tensor(0., grad_fn=<SumBackward0>)\n",
      " true:  2   model:  1\n",
      "tensor(0., grad_fn=<SumBackward0>)\n",
      " true:  2   model:  1\n",
      "tensor(0., grad_fn=<SumBackward0>)\n",
      " true:  2   model:  0\n",
      "tensor(0., grad_fn=<SumBackward0>)\n",
      " true:  2   model:  1\n",
      "tensor(0., grad_fn=<SumBackward0>)\n",
      " true:  2   model:  1\n",
      "tensor(0., grad_fn=<SumBackward0>)\n",
      " true:  2   model:  1\n",
      "tensor(0., grad_fn=<SumBackward0>)\n",
      " true:  2   model:  1\n",
      "tensor(0., grad_fn=<SumBackward0>)\n",
      " true:  2   model:  1\n",
      "tensor(0., grad_fn=<SumBackward0>)\n",
      " true:  2   model:  1\n",
      "tensor(0., grad_fn=<SumBackward0>)\n",
      " true:  2   model:  0\n",
      "tensor(0., grad_fn=<SumBackward0>)\n",
      " true:  2   model:  0\n",
      "tensor(0., grad_fn=<SumBackward0>)\n",
      " true:  2   model:  0\n",
      "tensor(0., grad_fn=<SumBackward0>)\n",
      " true:  2   model:  0\n",
      "tensor(0., grad_fn=<SumBackward0>)\n",
      " true:  2   model:  0\n",
      "tensor(0., grad_fn=<SumBackward0>)\n",
      " true:  2   model:  1\n",
      "tensor(0., grad_fn=<SumBackward0>)\n",
      " true:  2   model:  0\n",
      "tensor(0., grad_fn=<SumBackward0>)\n",
      " true:  2   model:  0\n",
      "tensor(0., grad_fn=<SumBackward0>)\n",
      " true:  2   model:  0\n",
      "tensor(0., grad_fn=<SumBackward0>)\n",
      " true:  2   model:  0\n",
      "tensor(0., grad_fn=<SumBackward0>)\n",
      " true:  2   model:  0\n",
      "tensor(0., grad_fn=<SumBackward0>)\n",
      " true:  2   model:  0\n",
      "tensor(0., grad_fn=<SumBackward0>)\n",
      " true:  2   model:  0\n",
      "tensor(0., grad_fn=<SumBackward0>)\n",
      " true:  2   model:  0\n",
      "tensor(0., grad_fn=<SumBackward0>)\n",
      " true:  2   model:  0\n",
      "tensor(0., grad_fn=<SumBackward0>)\n",
      " true:  2   model:  0\n",
      "tensor(0., grad_fn=<SumBackward0>)\n",
      " true:  2   model:  0\n",
      "tensor(0., grad_fn=<SumBackward0>)\n",
      " true:  2   model:  0\n",
      "tensor(0., grad_fn=<SumBackward0>)\n",
      " true:  2   model:  0\n",
      "tensor(0., grad_fn=<SumBackward0>)\n",
      " true:  2   model:  0\n",
      "tensor(0., grad_fn=<SumBackward0>)\n",
      " true:  2   model:  0\n",
      "tensor(0., grad_fn=<SumBackward0>)\n",
      " true:  2   model:  0\n",
      "tensor(0., grad_fn=<SumBackward0>)\n",
      " true:  2   model:  0\n",
      "tensor(0., grad_fn=<SumBackward0>)\n",
      " true:  2   model:  0\n",
      "tensor(0., grad_fn=<SumBackward0>)\n",
      " true:  2   model:  0\n",
      "tensor(0., grad_fn=<SumBackward0>)\n",
      " true:  2   model:  0\n",
      "tensor(0., grad_fn=<SumBackward0>)\n",
      " true:  2   model:  0\n",
      "tensor(0., grad_fn=<SumBackward0>)\n",
      " true:  2   model:  0\n",
      "tensor(0., grad_fn=<SumBackward0>)\n",
      " true:  2   model:  0\n",
      "tensor(0., grad_fn=<SumBackward0>)\n",
      " true:  2   model:  0\n",
      "tensor(0., grad_fn=<SumBackward0>)\n",
      " true:  2   model:  0\n",
      "tensor(0., grad_fn=<SumBackward0>)\n",
      " true:  2   model:  0\n",
      "tensor(0., grad_fn=<SumBackward0>)\n",
      " true:  2   model:  0\n",
      "tensor(0., grad_fn=<SumBackward0>)\n",
      " true:  2   model:  0\n",
      "tensor(0., grad_fn=<SumBackward0>)\n",
      " true:  2   model:  0\n",
      "tensor(0., grad_fn=<SumBackward0>)\n",
      " true:  2   model:  0\n",
      "tensor(0., grad_fn=<SumBackward0>)\n",
      " true:  2   model:  0\n",
      "tensor(0., grad_fn=<SumBackward0>)\n",
      " true:  2   model:  0\n",
      "tensor(0., grad_fn=<SumBackward0>)\n",
      " true:  2   model:  0\n",
      "tensor(0., grad_fn=<SumBackward0>)\n",
      " true:  2   model:  0\n",
      "tensor(0., grad_fn=<SumBackward0>)\n",
      " true:  2   model:  0\n",
      "tensor(0., grad_fn=<SumBackward0>)\n",
      " true:  2   model:  0\n",
      "tensor(0., grad_fn=<SumBackward0>)\n",
      " true:  2   model:  1\n",
      "tensor(0., grad_fn=<SumBackward0>)\n",
      " true:  2   model:  1\n",
      "tensor(0., grad_fn=<SumBackward0>)\n",
      " true:  2   model:  0\n",
      "tensor(0., grad_fn=<SumBackward0>)\n",
      " true:  2   model:  1\n",
      "tensor(0., grad_fn=<SumBackward0>)\n",
      " true:  2   model:  1\n",
      "tensor(0., grad_fn=<SumBackward0>)\n",
      " true:  2   model:  1\n",
      "tensor(0., grad_fn=<SumBackward0>)\n",
      " true:  2   model:  1\n",
      "tensor(0., grad_fn=<SumBackward0>)\n",
      " true:  2   model:  1\n",
      "tensor(0., grad_fn=<SumBackward0>)\n",
      " true:  2   model:  1\n",
      "tensor(0., grad_fn=<SumBackward0>)\n",
      " true:  2   model:  1\n",
      "tensor(0., grad_fn=<SumBackward0>)\n",
      " true:  2   model:  0\n",
      "tensor(0., grad_fn=<SumBackward0>)\n",
      " true:  2   model:  1\n",
      "tensor(0., grad_fn=<SumBackward0>)\n",
      " true:  2   model:  1\n",
      "tensor(0., grad_fn=<SumBackward0>)\n",
      " true:  2   model:  1\n",
      "tensor(0., grad_fn=<SumBackward0>)\n",
      " true:  2   model:  1\n",
      "tensor(0., grad_fn=<SumBackward0>)\n",
      " true:  2   model:  1\n",
      "tensor(0., grad_fn=<SumBackward0>)\n",
      " true:  2   model:  1\n",
      "tensor(0., grad_fn=<SumBackward0>)\n",
      " true:  2   model:  1\n",
      "tensor(0., grad_fn=<SumBackward0>)\n",
      " true:  2   model:  1\n",
      "tensor(0., grad_fn=<SumBackward0>)\n",
      " true:  2   model:  1\n",
      "tensor(0., grad_fn=<SumBackward0>)\n",
      " true:  2   model:  1\n",
      "tensor(0., grad_fn=<SumBackward0>)\n",
      " true:  2   model:  1\n",
      "tensor(0., grad_fn=<SumBackward0>)\n",
      " true:  2   model:  1\n",
      "tensor(0., grad_fn=<SumBackward0>)\n",
      " true:  2   model:  1\n",
      "tensor(0., grad_fn=<SumBackward0>)\n",
      " true:  2   model:  1\n",
      "tensor(0., grad_fn=<SumBackward0>)\n",
      " true:  2   model:  1\n",
      "tensor(0., grad_fn=<SumBackward0>)\n",
      " true:  2   model:  1\n",
      "tensor(0., grad_fn=<SumBackward0>)\n",
      " true:  2   model:  1\n",
      "tensor(0., grad_fn=<SumBackward0>)\n",
      " true:  2   model:  1\n",
      "tensor(0., grad_fn=<SumBackward0>)\n",
      " true:  2   model:  1\n",
      "tensor(0., grad_fn=<SumBackward0>)\n",
      " true:  2   model:  1\n",
      "tensor(0., grad_fn=<SumBackward0>)\n",
      " true:  2   model:  1\n",
      "tensor(0., grad_fn=<SumBackward0>)\n",
      " true:  2   model:  1\n",
      "tensor(0., grad_fn=<SumBackward0>)\n",
      " true:  2   model:  1\n",
      "tensor(0., grad_fn=<SumBackward0>)\n",
      " true:  2   model:  1\n",
      "tensor(0., grad_fn=<SumBackward0>)\n",
      " true:  2   model:  1\n",
      "tensor(0., grad_fn=<SumBackward0>)\n",
      " true:  2   model:  1\n",
      "tensor(0., grad_fn=<SumBackward0>)\n",
      " true:  2   model:  1\n",
      "tensor(0., grad_fn=<SumBackward0>)\n",
      " true:  2   model:  1\n",
      "tensor(0., grad_fn=<SumBackward0>)\n",
      " true:  2   model:  0\n",
      "tensor(0., grad_fn=<SumBackward0>)\n",
      " true:  2   model:  0\n",
      "tensor(0., grad_fn=<SumBackward0>)\n",
      " true:  2   model:  1\n",
      "tensor(0., grad_fn=<SumBackward0>)\n",
      " true:  2   model:  1\n",
      "tensor(0., grad_fn=<SumBackward0>)\n",
      " true:  2   model:  1\n",
      "tensor(0., grad_fn=<SumBackward0>)\n",
      " true:  2   model:  0\n",
      "tensor(0., grad_fn=<SumBackward0>)\n",
      " true:  2   model:  1\n",
      "tensor(0., grad_fn=<SumBackward0>)\n",
      " true:  2   model:  1\n",
      "tensor(0., grad_fn=<SumBackward0>)\n",
      " true:  2   model:  1\n",
      "tensor(0., grad_fn=<SumBackward0>)\n",
      " true:  2   model:  1\n",
      "tensor(0., grad_fn=<SumBackward0>)\n",
      " true:  2   model:  1\n",
      "tensor(0., grad_fn=<SumBackward0>)\n",
      " true:  2   model:  0\n",
      "tensor(0., grad_fn=<SumBackward0>)\n",
      " true:  2   model:  1\n",
      "tensor(0., grad_fn=<SumBackward0>)\n",
      " true:  2   model:  1\n",
      "tensor(0., grad_fn=<SumBackward0>)\n",
      " true:  2   model:  1\n",
      "tensor(0., grad_fn=<SumBackward0>)\n",
      " true:  2   model:  1\n",
      "tensor(0., grad_fn=<SumBackward0>)\n",
      " true:  2   model:  1\n",
      "tensor(0., grad_fn=<SumBackward0>)\n",
      " true:  2   model:  0\n",
      "tensor(0., grad_fn=<SumBackward0>)\n",
      " true:  2   model:  1\n",
      "tensor(0., grad_fn=<SumBackward0>)\n",
      " true:  2   model:  1\n",
      "tensor(0., grad_fn=<SumBackward0>)\n",
      " true:  2   model:  1\n",
      "tensor(0., grad_fn=<SumBackward0>)\n",
      " true:  2   model:  1\n",
      "tensor(0., grad_fn=<SumBackward0>)\n",
      " true:  2   model:  1\n",
      "tensor(0., grad_fn=<SumBackward0>)\n",
      " true:  2   model:  0\n",
      "tensor(0., grad_fn=<SumBackward0>)\n",
      " true:  2   model:  1\n",
      "(2, '2_DianaAngry7.wav')\n",
      "tensor(0., grad_fn=<SumBackward0>)\n",
      " true:  2   model:  2\n"
     ]
    }
   ],
   "source": [
    "import librosa\n",
    "import os\n",
    "import glob\n",
    "X__ = []\n",
    "Y__ = []\n",
    "test_path = os.path.abspath('audioDB/test')\n",
    "for file in glob.glob(os.path.join(test_path, '*.wav')):\n",
    "    source_file = os.path.join(test_path, file)\n",
    "    y, sr = librosa.load(source_file, sr=16000)\n",
    "    x1 = librosa.feature.mfcc(y=y, sr=sr, n_mfcc=20)\n",
    "    x2 = librosa.feature.spectral_centroid(y=y, sr=sr)\n",
    "    x3 = librosa.feature.spectral_rolloff(y=y, sr=sr)\n",
    "    X__.append(np.concatenate([x1, x2, x3]))\n",
    "    Y__.append((int(file.split('_')[0]), file))\n",
    "\n",
    "model.eval()\n",
    "print([x.shape for x in X__])\n",
    "for x, y in zip(X__, Y__):\n",
    "    print(y)\n",
    "    y = y[0]\n",
    "\n",
    "    if x.shape[1] < 137:\n",
    "        out = np.zeros((x.shape[0], 137))\n",
    "        out[:, :x.shape[1]] = x\n",
    "        x = out\n",
    "\n",
    "    for s in range(0, len(x[0]) - 136):\n",
    "        x_ = x[:, s:s + 137]\n",
    "        x_ = torch.tensor(x_)\n",
    "        x_.unsqueeze_(0)\n",
    "        y_ = model(x_)\n",
    "        y__ = softmax(y_).argmax(dim=1)\n",
    "        print(' true: ', y, '  model: ', y__.item())"
   ],
   "metadata": {
    "collapsed": false,
    "pycharm": {
     "name": "#%%\n"
    }
   }
  },
  {
   "cell_type": "code",
   "execution_count": 138,
   "outputs": [
    {
     "data": {
      "text/plain": "array([0, 0, 0, ..., 2, 2, 2])"
     },
     "execution_count": 138,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "Y"
   ],
   "metadata": {
    "collapsed": false,
    "pycharm": {
     "name": "#%%\n"
    }
   }
  },
  {
   "cell_type": "code",
   "execution_count": 144,
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "(array([0, 1, 2]), array([ 778, 4246,  640], dtype=int64))\n",
      "(array([0, 1, 2]), array([156, 844, 133], dtype=int64))\n",
      "(array([0, 1, 2]), array([ 622, 3402,  507], dtype=int64))\n"
     ]
    }
   ],
   "source": [
    "print(np.unique(Y, return_counts=True),\n",
    "      np.unique(y_test, return_counts=True),\n",
    "      np.unique(y_train, return_counts=True), sep='\\n')"
   ],
   "metadata": {
    "collapsed": false,
    "pycharm": {
     "name": "#%%\n"
    }
   }
  },
  {
   "cell_type": "code",
   "execution_count": 154,
   "outputs": [],
   "source": [
    "softmax = torch.nn.Softmax(dim=1)"
   ],
   "metadata": {
    "collapsed": false,
    "pycharm": {
     "name": "#%%\n"
    }
   }
  },
  {
   "cell_type": "code",
   "execution_count": 157,
   "outputs": [],
   "source": [
    "source_file = os.path.join(test_path, '0_DianaHappy.wav')\n",
    "y, sr = librosa.load(source_file, sr=16000)"
   ],
   "metadata": {
    "collapsed": false,
    "pycharm": {
     "name": "#%%\n"
    }
   }
  },
  {
   "cell_type": "code",
   "execution_count": 160,
   "outputs": [],
   "source": [
    "x1 = librosa.feature.mfcc(y=y, sr=sr, n_mfcc=20)\n",
    "x2 = librosa.feature.spectral_centroid(y=y, sr=sr)\n",
    "x3 = librosa.feature.spectral_rolloff(y=y, sr=sr)"
   ],
   "metadata": {
    "collapsed": false,
    "pycharm": {
     "name": "#%%\n"
    }
   }
  },
  {
   "cell_type": "code",
   "execution_count": 161,
   "outputs": [
    {
     "data": {
      "text/plain": "(20, 145)"
     },
     "execution_count": 161,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "x1.shape"
   ],
   "metadata": {
    "collapsed": false,
    "pycharm": {
     "name": "#%%\n"
    }
   }
  },
  {
   "cell_type": "code",
   "execution_count": 165,
   "outputs": [
    {
     "data": {
      "text/plain": "(74136,)"
     },
     "execution_count": 165,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "y.shape"
   ],
   "metadata": {
    "collapsed": false,
    "pycharm": {
     "name": "#%%\n"
    }
   }
  },
  {
   "cell_type": "code",
   "execution_count": 186,
   "outputs": [
    {
     "data": {
      "text/plain": "array([0.        , 0.02321995, 0.04643991, 0.06965986, 0.09287982,\n       0.11609977, 0.13931973, 0.16253968, 0.18575964, 0.20897959,\n       0.23219955, 0.2554195 , 0.27863946, 0.30185941, 0.32507937,\n       0.34829932, 0.37151927, 0.39473923, 0.41795918, 0.44117914,\n       0.46439909, 0.48761905, 0.510839  , 0.53405896, 0.55727891,\n       0.58049887, 0.60371882, 0.62693878, 0.65015873, 0.67337868,\n       0.69659864, 0.71981859, 0.74303855, 0.7662585 , 0.78947846,\n       0.81269841, 0.83591837, 0.85913832, 0.88235828, 0.90557823,\n       0.92879819, 0.95201814, 0.9752381 , 0.99845805, 1.021678  ,\n       1.04489796, 1.06811791, 1.09133787, 1.11455782, 1.13777778,\n       1.16099773, 1.18421769, 1.20743764, 1.2306576 , 1.25387755,\n       1.27709751, 1.30031746, 1.32353741, 1.34675737, 1.36997732,\n       1.39319728, 1.41641723, 1.43963719, 1.46285714, 1.4860771 ,\n       1.50929705, 1.53251701, 1.55573696, 1.57895692, 1.60217687,\n       1.62539683, 1.64861678, 1.67183673, 1.69505669, 1.71827664,\n       1.7414966 , 1.76471655, 1.78793651, 1.81115646, 1.83437642,\n       1.85759637, 1.88081633, 1.90403628, 1.92725624, 1.95047619,\n       1.97369615, 1.9969161 , 2.02013605, 2.04335601, 2.06657596,\n       2.08979592, 2.11301587, 2.13623583, 2.15945578, 2.18267574,\n       2.20589569, 2.22911565, 2.2523356 , 2.27555556, 2.29877551,\n       2.32199546, 2.34521542, 2.36843537, 2.39165533, 2.41487528,\n       2.43809524, 2.46131519, 2.48453515, 2.5077551 , 2.53097506,\n       2.55419501, 2.57741497, 2.60063492, 2.62385488, 2.64707483,\n       2.67029478, 2.69351474, 2.71673469, 2.73995465, 2.7631746 ,\n       2.78639456, 2.80961451, 2.83283447, 2.85605442, 2.87927438,\n       2.90249433, 2.92571429, 2.94893424, 2.9721542 , 2.99537415,\n       3.0185941 , 3.04181406, 3.06503401, 3.08825397, 3.11147392,\n       3.13469388, 3.15791383])"
     },
     "execution_count": 186,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "frames = range(137)\n",
    "t = librosa.frames_to_time(frames)\n",
    "t"
   ],
   "metadata": {
    "collapsed": false,
    "pycharm": {
     "name": "#%%\n"
    }
   }
  },
  {
   "cell_type": "code",
   "execution_count": 176,
   "outputs": [],
   "source": [
    "torch.save(model, 'models/exp1.model')"
   ],
   "metadata": {
    "collapsed": false,
    "pycharm": {
     "name": "#%%\n"
    }
   }
  },
  {
   "cell_type": "code",
   "execution_count": 178,
   "outputs": [
    {
     "data": {
      "text/plain": "Net1(\n  (c1): Sequential(\n    (0): Conv1d(22, 32, kernel_size=(5,), stride=(2,))\n    (1): ReLU()\n    (2): MaxPool1d(kernel_size=3, stride=3, padding=0, dilation=1, ceil_mode=False)\n  )\n  (c2): Sequential(\n    (0): Conv1d(32, 16, kernel_size=(3,), stride=(1,))\n    (1): ReLU()\n    (2): MaxPool1d(kernel_size=2, stride=2, padding=0, dilation=1, ceil_mode=False)\n  )\n  (c3): Sequential(\n    (0): Conv1d(16, 16, kernel_size=(3,), stride=(1,))\n    (1): ReLU()\n    (2): MaxPool1d(kernel_size=2, stride=2, padding=0, dilation=1, ceil_mode=False)\n  )\n  (f1): Linear(in_features=64, out_features=32, bias=True)\n  (relu): ReLU()\n  (f2): Linear(in_features=32, out_features=3, bias=True)\n)"
     },
     "execution_count": 178,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "modelFFF = torch.load('models/exp1.model')\n",
    "modelFFF.eval()"
   ],
   "metadata": {
    "collapsed": false,
    "pycharm": {
     "name": "#%%\n"
    }
   }
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "outputs": [],
   "source": [],
   "metadata": {
    "collapsed": false,
    "pycharm": {
     "name": "#%%\n"
    }
   }
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 3",
   "language": "python",
   "name": "python3"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 2
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython2",
   "version": "2.7.6"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 0
}