coughs.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [],
   "source": [
    "original_path=\"./dataset_original\"  ### please enter path to cough dataset\n",
    "device='cuda' ###please enter device: cuda or cpu"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [],
   "source": [
    "#imports\n",
    "import numpy as np\n",
    "import math\n",
    "from tqdm.notebook import tqdm\n",
    "import matplotlib.pyplot as plt\n",
    "import librosa\n",
    "from pydub import AudioSegment\n",
    "#from PIL import Image \n",
    "import os\n",
    "\n",
    "#tensorboard\n",
    "%load_ext tensorboard\n",
    "from tensorflow import summary\n",
    "import tensorflow as tf\n",
    "\n",
    "#pytorch\n",
    "import torch\n",
    "import torchvision\n",
    "import torch.nn as nn\n",
    "import torch.nn.functional as F\n",
    "import torch.optim as optim\n",
    "from torch.utils.data import Dataset, DataLoader\n",
    "\n",
    "#directory for saving trained model\n",
    "if not os.path.exists('./models'):\n",
    "    os.makedirs('./models')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "env: CUBLAS_WORKSPACE_CONFIG=:16:8\n"
     ]
    }
   ],
   "source": [
    "#set seed\n",
    "%env CUBLAS_WORKSPACE_CONFIG=:16:8\n",
    "torch.backends.cudnn.deterministic = True\n",
    "torch.set_deterministic(True)\n",
    "torch.manual_seed(19)\n",
    "torch.cuda.manual_seed(19)\n",
    "np.random.seed(19)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [],
   "source": [
    "#convert mp3 to wav\n",
    "pos_path=original_path+\"/segmented/pos/\"\n",
    "neg_path=original_path+\"/segmented/neg/\"\n",
    "\n",
    "wav_path='./dataset_wav/'\n",
    "\n",
    "#create directories for wav files\n",
    "if not os.path.exists(wav_path):\n",
    "    os.makedirs(wav_path)\n",
    "\n",
    "#mp3 to wav conversion    \n",
    "for i in os.listdir(pos_path):\n",
    "    if i[-8]!='2': ##exclude a patient from the dataset\n",
    "        AudioSegment.from_mp3(pos_path+i).export(wav_path+i[0:-3]+'wav', format=\"wav\")\n",
    "for i in os.listdir(neg_path):\n",
    "    AudioSegment.from_mp3(neg_path+i).export(wav_path+i[0:-3]+'wav', format=\"wav\")        "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [],
   "source": [
    "class spectogram():\n",
    "    \n",
    "    def __init__(self, path):\n",
    "        self.path = path\n",
    "        \n",
    "    def __len__(self):\n",
    "        return len(os.listdir(self.path))\n",
    "\n",
    "    def __getitem__(self, idx):\n",
    "        \n",
    "        y, sr=librosa.load(self.path+os.listdir(self.path)[idx])\n",
    "        sgram=librosa.feature.melspectrogram(y=y, sr=sr)\n",
    "        sgram=torch.tensor(sgram).unsqueeze(1).permute(1,0,2)\n",
    "        \n",
    "        data=dict()\n",
    "        data['sgram']=sgram\n",
    "        \n",
    "        if os.listdir(self.path)[idx][0]=='p':\n",
    "            data['class']= \"positive\"\n",
    "            data['label']= 1\n",
    "        else:    \n",
    "            data['class']= \"negative\"\n",
    "            data['label']= 0  \n",
    "            \n",
    "        return data"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "93 + 23 = 116\n"
     ]
    }
   ],
   "source": [
    "#prepare dataset\n",
    "dataset= spectogram(wav_path)\n",
    "L=len(dataset)\n",
    "\n",
    "#split in train, test, validation\n",
    "train_set, test_set = \\\n",
    "torch.utils.data.random_split(dataset,[math.ceil(L*0.8),int(L*0.2)])\n",
    "\n",
    "print(len(train_set),'+', len(test_set),'=',L)\n",
    "\n",
    "#dataloader\n",
    "batch_size=16\n",
    "\n",
    "#dataset_dl=DataLoader(dataset, batch_size=batch_size, shuffle=True)\n",
    "train_dl = DataLoader(train_set, batch_size=batch_size, shuffle=True)\n",
    "test_dl = DataLoader(test_set, batch_size=1, shuffle=False)\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "torch.Size([16, 1, 128, 69])\n",
      "tensor([1, 1, 1, 0, 1, 1, 0, 0, 1, 1, 0, 1, 0, 0, 0, 1])\n"
     ]
    }
   ],
   "source": [
    "#view the dataloader\n",
    "sample=next(iter(train_dl))\n",
    "print(sample['sgram'].shape) # batch_size, channels, height, width\n",
    "print(sample['label'])# 1 or 0"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {},
   "outputs": [],
   "source": [
    "#CNNs\n",
    "class Model(nn.Module):\n",
    "    def __init__(self, height=128, width=69, dim_output=2):\n",
    "        super(Model, self).__init__()\n",
    "        \n",
    "        self.H=height\n",
    "        self.W=width\n",
    "        self.output=dim_output\n",
    "        \n",
    "        self.max_pool = nn.MaxPool2d(kernel_size=2, stride=2)\n",
    "        self.activation = nn.ReLU()\n",
    "        self.dropout= nn.Dropout(p=0.5) \n",
    "        self.bn1= nn.BatchNorm2d(32) \n",
    "        self.bn2= nn.BatchNorm2d(64)\n",
    "        self.bn3= nn.BatchNorm2d(128)\n",
    "        \n",
    "        #convolutions\n",
    "        self.conv1 = nn.Conv2d( 1,  32,kernel_size=3, stride=1, padding=1)\n",
    "        self.conv2 = nn.Conv2d(32,  64,kernel_size=3, stride=1, padding=1)\n",
    "        self.conv3 = nn.Conv2d(64, 128,kernel_size=3, stride=1, padding=1)\n",
    "        \n",
    "        #fully connected\n",
    "        self.fc1 = nn.Linear(128*(self.H//8)*(self.W//8),128) \n",
    "        self.fc2 = nn.Linear(128,64)\n",
    "        self.fc3 = nn.Linear(64,self.output)\n",
    "        \n",
    "        #sequential\n",
    "        self.layer1=nn.Sequential(self.conv1, self.bn1, self.max_pool, self.activation)\n",
    "        self.layer2=nn.Sequential(self.conv2, self.bn2, self.max_pool, self.activation)\n",
    "        self.layer3=nn.Sequential(self.conv3, self.bn3, self.max_pool, self.activation)\n",
    "        self.layer4=nn.Sequential(self.fc1, self.activation, self.dropout)\n",
    "        self.layer5=nn.Sequential(self.fc2, self.activation, self.dropout)\n",
    "        \n",
    "    def forward(self, img):\n",
    "        \n",
    "        out=self.layer3(self.layer2(self.layer1(img)))\n",
    "        \n",
    "        ###flattening\n",
    "        batch_size =img.shape[0]\n",
    "        out=out.view(batch_size, -1)\n",
    "        \n",
    "        out=self.fc3(self.layer5(self.layer4(out)))\n",
    "       \n",
    "        return out           "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "torch.Size([16, 2])"
      ]
     },
     "execution_count": 9,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "#model and optimizer\n",
    "model = Model()\n",
    "model = model.to(device) \n",
    "\n",
    "optimizer = optim.Adam(model.parameters(), lr=1e-3, weight_decay=1e-6)\n",
    "\n",
    "#print(model)\n",
    "sample_out=model(sample['sgram'].to(device)) \n",
    "sample_out.shape"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {},
   "outputs": [],
   "source": [
    "#train function\n",
    "def train(model, optimizer, sample):\n",
    "    model.train()\n",
    "\n",
    "    criterion = nn.CrossEntropyLoss()\n",
    "    \n",
    "    # reset optimizer's gradient\n",
    "    optimizer.zero_grad()\n",
    "\n",
    "    # define input and output\n",
    "    img = sample['sgram'].float().to(device) \n",
    "    label = sample['label'].long().to(device)\n",
    "  \n",
    "    # get the prediction\n",
    "    pred = model(img) \n",
    "\n",
    "    pred_loss = criterion(pred, label)\n",
    "    \n",
    "    # regularization loss\n",
    "    reg_loss= 0.0\n",
    "    for p in model.parameters():\n",
    "        if p.requires_grad: # if p is a trainable parameter\n",
    "            reg_loss+=torch.norm(p,2)\n",
    "\n",
    "    total_loss= pred_loss + 1e-6 * reg_loss\n",
    "    \n",
    "    num_correct = sum(torch.argmax(pred, dim=1) == label)\n",
    "    \n",
    "    #one step of training\n",
    "    total_loss.backward()\n",
    "    optimizer.step()\n",
    "\n",
    "    return total_loss.item(), num_correct.item()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {},
   "outputs": [],
   "source": [
    "#test function\n",
    "def test(model, sample): \n",
    "    model.eval()\n",
    "\n",
    "    criterion = nn.CrossEntropyLoss()\n",
    "    \n",
    "    with torch.no_grad(): \n",
    "        \n",
    "        img = sample['sgram'].float().to(device)\n",
    "        label = sample['label'].long().to(device)\n",
    "       \n",
    "        # get the prediction\n",
    "        pred = model(img)\n",
    "        \n",
    "        pred_loss = criterion(pred, label)\n",
    "        \n",
    "        # regularization loss\n",
    "        reg_loss= 0.0\n",
    "        for p in model.parameters():\n",
    "            if p.requires_grad: # if p is a trainable parameter\n",
    "                reg_loss+=torch.norm(p,2)\n",
    "\n",
    "        total_loss= pred_loss + 1e-6 * reg_loss\n",
    "        \n",
    "        num_correct = sum(torch.argmax(pred, dim=1) == label)\n",
    "\n",
    "    return total_loss.item(), num_correct.item()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {},
   "outputs": [],
   "source": [
    "#prepare tensorboard\n",
    "train_log_dir = './models/runs_cnn/train'\n",
    "train_summary_writer = summary.create_file_writer(train_log_dir)\n",
    "test_log_dir = './models/runs_cnn/test'\n",
    "test_summary_writer = summary.create_file_writer(test_log_dir)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "fdb8592a99434689aa7534b4321fbf0b",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "HBox(children=(FloatProgress(value=0.0, max=23.0), HTML(value='')))"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "7ee0af23fe4045848186f34edccfe99f",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "HBox(children=(FloatProgress(value=0.0, max=6.0), HTML(value='')))"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\n"
     ]
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "696ecde38eee4c0fb9981614514afc25",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "HBox(children=(FloatProgress(value=0.0, max=23.0), HTML(value='')))"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\n",
      "0.6021505376344086 0.6956521739130432\n"
     ]
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "b2d6c39b267e4086bfe3eeb7d8af9ae5",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "HBox(children=(FloatProgress(value=0.0, max=6.0), HTML(value='')))"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\n"
     ]
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "b414c51f99ba4b9190ff8a4959bd7466",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "HBox(children=(FloatProgress(value=0.0, max=23.0), HTML(value='')))"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\n",
      "0.5913978494623655 0.8260869565217388\n"
     ]
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "42f69b0099d243cab7381386482c022f",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "HBox(children=(FloatProgress(value=0.0, max=6.0), HTML(value='')))"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\n"
     ]
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "c40c8d80cfd942e294ba52651ae99c0e",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "HBox(children=(FloatProgress(value=0.0, max=23.0), HTML(value='')))"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\n",
      "0.6021505376344086 0.8260869565217388\n"
     ]
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "473bcce2e3174c61bfd3cedca2d5ac74",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "HBox(children=(FloatProgress(value=0.0, max=6.0), HTML(value='')))"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\n"
     ]
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "40671164edc64ff1a8b63aa543e40469",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "HBox(children=(FloatProgress(value=0.0, max=23.0), HTML(value='')))"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\n",
      "0.6236559139784945 0.7391304347826084\n"
     ]
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "dabe8e92e14540c6aa978bf9ad376301",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "HBox(children=(FloatProgress(value=0.0, max=6.0), HTML(value='')))"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\n"
     ]
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "86937f5e08e64412979ad9877baaef66",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "HBox(children=(FloatProgress(value=0.0, max=23.0), HTML(value='')))"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\n",
      "0.6236559139784947 0.8260869565217388\n"
     ]
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "7f202695fa36456fadb6e33f086d4dcf",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "HBox(children=(FloatProgress(value=0.0, max=6.0), HTML(value='')))"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\n"
     ]
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "c92201862f0f40bc91232059a74ad4fb",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "HBox(children=(FloatProgress(value=0.0, max=23.0), HTML(value='')))"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\n",
      "0.6881720430107527 0.6956521739130432\n"
     ]
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "46bd892764a5442995c51fe503048d8d",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "HBox(children=(FloatProgress(value=0.0, max=6.0), HTML(value='')))"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\n"
     ]
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "10e57e5c2dc6492a88dc2a935b09a862",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "HBox(children=(FloatProgress(value=0.0, max=23.0), HTML(value='')))"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\n",
      "0.6881720430107527 0.869565217391304\n"
     ]
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "61b3dc4abded483981742719fa0dfed6",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "HBox(children=(FloatProgress(value=0.0, max=6.0), HTML(value='')))"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\n"
     ]
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "a62708be568044c89fe6735ab8f6609f",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "HBox(children=(FloatProgress(value=0.0, max=23.0), HTML(value='')))"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\n",
      "0.6559139784946237 0.8260869565217388\n"
     ]
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "54853052097844d6aa5ce13e51d751d2",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "HBox(children=(FloatProgress(value=0.0, max=6.0), HTML(value='')))"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\n"
     ]
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "beeee2dbaec04312b861339967431242",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "HBox(children=(FloatProgress(value=0.0, max=23.0), HTML(value='')))"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\n",
      "0.6451612903225806 0.8260869565217388\n"
     ]
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "d5e8190dffd0402287d93d65d528a910",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "HBox(children=(FloatProgress(value=0.0, max=6.0), HTML(value='')))"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\n"
     ]
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "9b6054a42b994df5952fcb221642e2b9",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "HBox(children=(FloatProgress(value=0.0, max=23.0), HTML(value='')))"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\n",
      "0.6451612903225807 0.7826086956521736\n"
     ]
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "7a5708e6083a4717a501c0489573874a",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "HBox(children=(FloatProgress(value=0.0, max=6.0), HTML(value='')))"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\n"
     ]
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "2083ddf6c1ea4d27b79b6331f2bfd7c0",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "HBox(children=(FloatProgress(value=0.0, max=23.0), HTML(value='')))"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\n",
      "0.8279569892473118 0.8260869565217388\n"
     ]
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "2506ad0e0802477d890af2ab95b85687",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "HBox(children=(FloatProgress(value=0.0, max=6.0), HTML(value='')))"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\n"
     ]
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "43e6d2a6437245bbab4742e6cc4409b1",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "HBox(children=(FloatProgress(value=0.0, max=23.0), HTML(value='')))"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\n",
      "0.7741935483870968 0.9565217391304344\n"
     ]
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "4a0dcea51c1a43219750fd4919df5cc6",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "HBox(children=(FloatProgress(value=0.0, max=6.0), HTML(value='')))"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\n"
     ]
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "f3c3dab694c0405db9838985b67ff425",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "HBox(children=(FloatProgress(value=0.0, max=23.0), HTML(value='')))"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\n",
      "0.8279569892473118 0.9130434782608692\n"
     ]
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "9cc41bcf8cc74e05900324857c87e7ee",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "HBox(children=(FloatProgress(value=0.0, max=6.0), HTML(value='')))"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\n"
     ]
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "235a4283e930400a866fc3095974f373",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "HBox(children=(FloatProgress(value=0.0, max=23.0), HTML(value='')))"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\n",
      "0.7526881720430108 0.9130434782608692\n"
     ]
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "c035104b01b443959cf03b3bc565d490",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "HBox(children=(FloatProgress(value=0.0, max=6.0), HTML(value='')))"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\n"
     ]
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "4474409d58934f5dad0f587095e316d6",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "HBox(children=(FloatProgress(value=0.0, max=23.0), HTML(value='')))"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\n",
      "0.8602150537634408 0.9130434782608692\n"
     ]
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "3b0ee7be7ca648a3b2da08b07287f21b",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "HBox(children=(FloatProgress(value=0.0, max=6.0), HTML(value='')))"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\n"
     ]
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "bd888e0f38d245159598c79485cb4731",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "HBox(children=(FloatProgress(value=0.0, max=23.0), HTML(value='')))"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\n",
      "0.8494623655913979 0.8260869565217388\n"
     ]
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "afa117eec9c54563a46351b41992a4b6",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "HBox(children=(FloatProgress(value=0.0, max=6.0), HTML(value='')))"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\n"
     ]
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "6a49c65b2a8a439eb836eb2735e2313c",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "HBox(children=(FloatProgress(value=0.0, max=23.0), HTML(value='')))"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\n",
      "0.8817204301075268 0.8260869565217388\n"
     ]
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "ec2243a99ef845ac9a0bf26febd4bf29",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "HBox(children=(FloatProgress(value=0.0, max=6.0), HTML(value='')))"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\n"
     ]
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "bf148740bb3f442585a16152350ddc42",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "HBox(children=(FloatProgress(value=0.0, max=23.0), HTML(value='')))"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\n",
      "0.8279569892473118 0.8260869565217388\n"
     ]
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "c3bbc125cba24bf8a8170dbce19acd6e",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "HBox(children=(FloatProgress(value=0.0, max=6.0), HTML(value='')))"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\n"
     ]
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "df80ee90eb2e444bb21ad6fa5f423415",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "HBox(children=(FloatProgress(value=0.0, max=23.0), HTML(value='')))"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\n",
      "0.8709677419354839 0.9130434782608692\n"
     ]
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "8db269bf675b49f8bfb584e50fc9393b",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "HBox(children=(FloatProgress(value=0.0, max=6.0), HTML(value='')))"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\n"
     ]
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "de3fb81fc877424fb3ae94da4996da1a",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "HBox(children=(FloatProgress(value=0.0, max=23.0), HTML(value='')))"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\n",
      "0.9354838709677419 0.8260869565217388\n"
     ]
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "0227f1336e9c4655a3e8b8588e851a74",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "HBox(children=(FloatProgress(value=0.0, max=6.0), HTML(value='')))"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\n"
     ]
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "6faeddb241c14c7588fc8fae3a062ecd",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "HBox(children=(FloatProgress(value=0.0, max=23.0), HTML(value='')))"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\n",
      "0.8602150537634408 0.7391304347826084\n"
     ]
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "ec38e1e761054e0aa97d9fc5e90aea34",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "HBox(children=(FloatProgress(value=0.0, max=6.0), HTML(value='')))"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\n"
     ]
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "b0df1b89d07f4283b8e8adfa050e1ba7",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "HBox(children=(FloatProgress(value=0.0, max=23.0), HTML(value='')))"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\n",
      "0.9784946236559139 0.7391304347826084\n"
     ]
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "6bccd8a0543345398dcbbd90c0cdf44b",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "HBox(children=(FloatProgress(value=0.0, max=6.0), HTML(value='')))"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\n"
     ]
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "9210ac5d7d934334bd23f6f4df935830",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "HBox(children=(FloatProgress(value=0.0, max=23.0), HTML(value='')))"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\n",
      "0.9462365591397851 0.6956521739130432\n",
      "\n"
     ]
    }
   ],
   "source": [
    "##################################start training and validating#####################################################\n",
    "max_epoch = 23\n",
    "tmp_path = './models/temp.pth'\n",
    "\n",
    "for epoch in tqdm(range(max_epoch)):        \n",
    "    ###Train Phase\n",
    "    \n",
    "    # Initialize Loss and Accuracy\n",
    "    train_loss = 0.0\n",
    "    train_accu = 0.0\n",
    "\n",
    "    #########TEMPORARY################################\n",
    "    # Load the saved MODEL AND OPTIMIZER after evaluation.\n",
    "    if epoch > 0:\n",
    "        checkpoint = torch.load(tmp_path)\n",
    "        model.load_state_dict(checkpoint['model_state_dict'])\n",
    "        optimizer.load_state_dict(checkpoint['optimizer_state_dict'])\n",
    "       \n",
    "    ##################################################\n",
    "\n",
    "    # Iterate over the train_dataloader\n",
    "    with tqdm(total=len(train_dl)) as pbar:\n",
    "        for sample in train_dl:            \n",
    "            curr_loss, num_correct = train(model, optimizer, sample)\n",
    "            train_loss += curr_loss / len(train_dl) \n",
    "            train_accu += num_correct / len(train_set)\n",
    "            pbar.update(1)\n",
    "            \n",
    "    ### Write the current loss and accuracy to the Tensorboard\n",
    "    with train_summary_writer.as_default():\n",
    "        tf.summary.scalar('loss', train_loss, step=epoch)                \n",
    "        tf.summary.scalar('accuracy', train_accu, step=epoch)         \n",
    "         \n",
    "    ###############TEMPORARY#######################################\n",
    "    # save the model and optimizer's information before the evaulation\n",
    "    checkpoint = {\n",
    "        'model_state_dict': model.state_dict(),\n",
    "        'optimizer_state_dict': optimizer.state_dict(),\n",
    "        \n",
    "    }\n",
    "    torch.save(checkpoint, tmp_path)\n",
    "    ###################################################################\n",
    "    \n",
    "    ### Validate Phase\n",
    "    # Initialize Loss and Accuracy\n",
    "    test_loss = 0.0\n",
    "    test_accu = 0.0\n",
    "\n",
    "    # Iterate over the test_dataloader\n",
    "    with tqdm(total=len(test_dl)) as pbar:\n",
    "        for idx, sample in enumerate(test_dl): \n",
    "            curr_loss, num_correct = test(model, sample)\n",
    "            test_loss += curr_loss / len(test_dl)\n",
    "            test_accu += num_correct / len(test_set)\n",
    "            pbar.update(1)\n",
    "    \n",
    "    ### Write the current loss and accuracy to the Tensorboard\n",
    "    with test_summary_writer.as_default():\n",
    "        tf.summary.scalar('loss', test_loss, step=epoch)\n",
    "        tf.summary.scalar('accuracy', test_accu, step=epoch)    \n",
    "    \n",
    "        \n",
    "    print(train_accu, test_accu)  \n",
    "    \n",
    "os.remove(tmp_path)    "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {},
   "outputs": [],
   "source": [
    "def freq_domain(sample,sr):\n",
    "    L=sample.shape[0]\n",
    "    Y=np.fft.fft(sample,sr)\n",
    "    P2 = abs(Y/L)\n",
    "    P1 = P2[0:sr//2+1]\n",
    "    P1[1:-1] = 2*P1[1:-1]\n",
    "    return P1    "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "metadata": {},
   "outputs": [],
   "source": [
    "Y_pos=Y_neg=p=n=0\n",
    "\n",
    "for j in os.listdir(wav_path):\n",
    "    y,sr=librosa.load(wav_path+j)\n",
    "    L=y.shape[0]\n",
    "    if j[0]=='p':\n",
    "        Y_pos+=freq_domain(y,sr)\n",
    "        p+=1\n",
    "    else:\n",
    "        Y_neg+=freq_domain(y,sr)\n",
    "        n+=1\n",
    "Y_pos/=p\n",
    "Y_neg/=n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "<matplotlib.legend.Legend at 0x7f265dc5c210>"
      ]
     },
     "execution_count": 16,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAY4AAAEWCAYAAABxMXBSAAAAOXRFWHRTb2Z0d2FyZQBNYXRwbG90bGliIHZlcnNpb24zLjMuMiwgaHR0cHM6Ly9tYXRwbG90bGliLm9yZy8vihELAAAACXBIWXMAAAsTAAALEwEAmpwYAABlBklEQVR4nO2dd5gURdrAf++kDbCw5IyggoqKCAgY8AQTmD3zmfXEeOp5nyeGM56eZzjDGTCfngmznKJiziiggEgQJCfJC8umCfX9UT0zPWF3Z2Fnll3e3/PMs93VVd3VPbP19hvqLTHGoCiKoiiZ4mnoDiiKoiiNCxUciqIoSp1QwaEoiqLUCRUciqIoSp1QwaEoiqLUCRUciqIoSp1QwaFsFSJiRGTnhu5HXRCRHk6/ffV4zu4iUioi3vo6p9KwiMh7InJ2Q/djW0QFRxNBRP4gIpOdwWuF86M/oKH75UZEikXkaRFZKSKbROQXEbmmoftVHxhjFhtjmhtjwnVtKyIHiUjE+e6in/9lo5/bMiJys4gEnd9G9PfxkIh0aoj+GGNGGmOebYhrb+uo4GgCiMhVwP3AHUAHoDvwCHBsA3YrHfcBzYHdgJbAMcCvDdqjbYfljuCJfo5OrlCfGtI2zFhjTBHQGjge6AhMaSjhoVSDMUY/jfiDHYBLgZNqqJOHFSzLnc/9QJ5z7Bzgq6T6BtjZ2W4D/A/YCEwC/u6u79S9CJgLrAceBqSafswAjquhnw8AS5xrTQGGuo7dDLwKPA9sAn4CegPXAqucdoe56n8G/AP4HigB3gZaO8d6OP32uZ7hU8AKYJlzj17n2M7A58451mAHtnR9Tz7nZ8BtwNdOfycAbatpexCwNE35OU77+4B1Tr/ygHuAxcBvwBigwNXmauc+lgPnJX2XnwF/TDq/+7vcFfjQudYc4GTXsf843+27zv18B+zkOr67q+1vwHXYQb8MaOOqNwBYDfjT3O/NwPNJZV5gGnCPq+wCYJ5zrXFA56Tf4yXY3+Mm5zvYCfgW+7t6BQg4dVsB7zj9We9sd036Df3R/aycZ78eWACMbOj//4b6qMbR+NkXyAferKHO9cAQoB+wFzAIuCHD8z8MbMYOAmc7n2SOAvZxzn0ycHg155oI3C4i54pIrzTHJzl9bA28CLwqIvmu40cD/8X+w/8IfIDVmrsAtwKPJZ3vLOzg2RkIAQ9W069nneM7A3sDhwF/dI7dhh30WwFdgX9Xc450/AE4F2gPBID/q0PbKIOB+c45bgf+iRWY/Zz+dgFuBBCREc41DgV6AYdkehERaYYd+F90rnUa8IiI7O6qdhpwC/ZZzHP6g4gUAR8B72Of9c7Ax8aYldjB92TXOc4AXjbGBDPpl7Gmv7eBoc61hmNfCE4GOgGLgJeTmo3ACqghwF+Bx4HTgW7AHs59gP3tPAPsgNXSy4GHaujOYKxAbQvcBTwlIpLJfTQ5Glpy6WfrPth/iJW11PkVOMK1fziw0Nk+h2o0DuzbXhDYxXUsncZxgGv/FWB0Nf0owL6JTnHOO48a3tqwb3Z7Ods3Ax+6jh2N1bSimkGR05diZ/8z4E5X/T5AlXNPPZy6Pqxpr5LEt/bTgE+d7eewA0/X6vrp1Iud03X9G1zHLwHer6btQUAE2OD6nOx8N4td9QQrxN1v+vsCC5ztp5PuuTcZahzAKcCXSf16DLjJ2f4P8KTr2BHAbNfz+rGaezsF+NrZ9gIrgUHV1L2ZJI3DKb8ImOtsPwXc5TrW3Pkt9XD9Hvd3HZ8CXOPavxe4v5rr9wPWu/Zjz8t5VvNcxwqda3Xc2v/hxvhRjaPxsxZoW4v9uzP2zSzKIqesNtphB9clrrIlaeqtdG2XYf+ZUzDGlBtj7jDGDMCawF7BahWtAUTkLyIyS0RKRGQD1oTU1nWK31zb5cAaE3dGlzt/3dd293UR4E86H9i3TT+wQkQ2ONd9DPvWDfaNVYDvReRnETkv3b1VQ0bPxWG5MabY9XklzT20ww5YU1x9fd8pB/udJt9zpuwADI6e1zn36VhNs7b76Ub1vqq3gT4isiNWEyoxxnxfh36B1arWOdsJv2VjTCn2f6CLq37y7yR5vzmAiBSKyGMiskhENgJfAMU1RMbF7t8YU+Zs1vSdNllUcDR+vgUqgONqqLMcOzBE6e6UgX2DLYweEBH3QLEaa8Lp6irrthV9jWGM2Yh15jcDeorIUOAa7Jt2K2NMMdavsDWmAHdfu2PfTNck1VmC1TjaugbtFsaY3Z1+rjTGXGCM6QxciDXf5DL82J2+eg124Nvd1deWxpjo4LWC1Ht2k/BdkygUlgCfJwmv5saYizPo4xKsHyG188ZUYF8QTgfOxJoaM0ZEPFjt8kunKOG37JjY2mB9U3XlL8AuwGBjTAvgwOhpt+Bc2xUqOBo5xpgSrI37YRE5znmL8ovISBG5y6n2EnCDiLQTkbZO/eedY9OA3UWkn+NPuNl17jDwBnCzc95dsX6DLUJE/iYi+4hIwLnWFVizzBysqSmEFVY+EbkRaLGl13I4Q0T6iEgh1gfymkkKlzXGrMD6MO4VkRYi4hGRnUTkd06fTxKRqOBcjx3I6xxyWx8YYyLAE8B9ItLe6V8XEYn6lF4BznHd801Jp5gK/N75LncGzncdewfoLSJnOr8fv/Nd7ZZB194BOorIlSKSJyJFIjLYdfw5rKnnGOK/uxpxrr8b9rfbEfiXc+hF4Fzn95qHffn4zhizMJPzJlGEFcQbHK03+Xkp1aCCowlgjPkXcBXW4b0a+wZ4GfCWU+XvwGRgOjYa6QenDGPML9hB9SNsJMpXSae/DGsyWol9W3wJ+4a+RV3FOiPXYN8cDwWOdMwNHwDvAb9gTREVpDeL1YX/Ym3zK7EBBJdXU+8srPN6JlY4vIZ1vIJ1+n8nIqXYCJ4rjDELtrJfW8M1WN/QRMe88hH2rRljzHvYiLlPnDqfJLW9D+vn+Q0bEPBC9IAxZhM2KOBU7HezEuuIz6utQ07bQ7GawUrs72iY6/jXWB/ODxkM8Kc4z3oD9nmvBQYYY5Y75/oY+BvwOlbD2snp85ZwP9bvtgYbuPH+Fp5nu0McR4+iZISI/BPrEEwXXbXNICKfYR2tTzZ0XxoSETFAL2PMvAbuxyfAi9v799FUUI1DqRER2VVE+oplENa8UVPor6IkICL7AP2BsQ3dF6V+2B5moipbRxHWPNUZO9HuXmykjKLUiog8iw3cuMIxaSlNADVVKYqiKHVCTVWKoihKndguTFVt27Y1PXr0aOhuKIqiNCqmTJmyxhjTLrl8uxAcPXr0YPLkyQ3dDUVRlEaFiKTNPqCmKkVRFKVOqOBQFEVR6oQKDkVRFKVObBc+DkVRmjbBYJClS5dSUVHR0F1plOTn59O1a1f8fn9G9VVwKIrS6Fm6dClFRUX06NGD7XVtpS3FGMPatWtZunQpPXv2zKiNmqoURWn0VFRU0KZNGxUaW4CI0KZNmzppayo4FEVpEqjQ2HLq+uxUcDRyfl1dyje/Jq9NpCiKkj1UcDRyDr73c/7wxHcN3Q1F2e7xer3069ePPfbYg5NOOomysrLaG7lYvnw5J554IgBTp05l/PjxsWPjxo3jzjvvrNf+bg0qOBRFUeqBgoICpk6dyowZMwgEAowZM6ZO7Tt37sxrr70GpAqOY445htGjR9drf7cGFRyKoij1zNChQ5k3bx7r1q3juOOOo2/fvgwZMoTp06cD8Pnnn9OvXz/69evH3nvvzaZNm1i4cCF77LEHVVVV3HjjjYwdO5Z+/foxduxY/vOf/3DZZZdRUlJCjx49iEQiAJSVldGtWzeCwSC//vorI0aMYMCAAQwdOpTZs2dn7f40HFdRlCbFLf/7mZnLN9brOft0bsFNR++eUd1QKMR7773HiBEjuOmmm9h777156623+OSTTzjrrLOYOnUq99xzDw8//DD7778/paWl5Ofnx9oHAgFuvfVWJk+ezEMPPQTAf/7zHwBatmzJXnvtxeeff86wYcP43//+x+GHH47f72fUqFGMGTOGXr168d1333HJJZfwySfJqwfXDyo4FEVR6oHy8nL69esHWI3j/PPPZ/Dgwbz++usADB8+nLVr11JSUsL+++/PVVddxemnn87vf/97unbtmvF1TjnlFMaOHcuwYcN4+eWXueSSSygtLeWbb77hpJNOitWrrKys1/tzo4JDUZQmRaaaQX0T9XG4SbdQnogwevRojjzySMaPH8+QIUP46KOPErSOmjjmmGO49tprWbduHVOmTGH48OFs3ryZ4uLilOtnC/VxKIqiZIkDDzyQF154AYDPPvuMtm3b0qJFC3799Vf23HNPrrnmGgYOHJjijygqKmLTpvQr7TZv3pxBgwZxxRVXcNRRR+H1emnRogU9e/bk1VdfBazAmjZtWtbuSwWHoihKlrj55puZPHkyffv2ZfTo0Tz77LMA3H///eyxxx7stddeFBQUMHLkyIR2w4YNY+bMmTHneDKnnHIKzz//PKecckqs7IUXXuCpp55ir732Yvfdd+ftt9/O2n1tF2uODxw40DTVhZx6jH4XgIV3HtnAPVGUhmPWrFnstttuDd2NRk26ZygiU4wxA5PrqsahKIqi1AkVHIqiKEqdUMGhKIqi1ImsCg4RGSEic0RknoikzJcXy4PO8eki0r+2tiIyVkSmOp+FIjI1m/egKIqiJJK1eRwi4gUeBg4FlgKTRGScMWamq9pIoJfzGQw8Cgyuqa0x5hTXNe4FSrJ1D4qiKEoq2dQ4BgHzjDHzjTFVwMvAsUl1jgWeM5aJQLGIdMqkrdgE8icDL2XxHhRFUZQksik4ugBLXPtLnbJM6mTSdijwmzFmbrqLi8goEZksIpNXr169Bd1XFEXJHBHhL3/5S2z/nnvu4eabb67369xxxx0J+/vtt1+9X6M2sik40i0plTxppLo6mbQ9jRq0DWPM48aYgcaYge3atauxo4qiKFtLXl4eb7zxBmvWZHdhtWTB8c0332T1eunIpuBYCnRz7XcFlmdYp8a2IuIDfg+kTqlUFEVpAHw+H6NGjeK+++5LObZ69WpOOOEE9tlnH/bZZx++/vrrWPmhhx5K//79ufDCC9lhhx1igue4445jwIAB7L777jz++OMAjB49OpZM8fTTTwdsChKws8nda3icc845vP7664TDYa6++mr22Wcf+vbty2OPPbb197rVZ6ieSUAvEekJLANOBf6QVGcccJmIvIx1jpcYY1aIyOpa2h4CzDbGLM1i/xsVxhhdc1lRAN4bDSt/qt9zdtwTRta+At+ll15K3759+etf/5pQfsUVV/DnP/+ZAw44gMWLF3P44Ycza9YsbrnlFoYPH861117L+++/HxMQAE8//TStW7emvLycffbZhxNOOIE777yThx56KG0yw1NPPZWxY8dyxBFHUFVVxccff8yjjz7KU089RcuWLZk0aRKVlZXsv//+HHbYYfTs2XOLH0fWBIcxJiQilwEfAF7gaWPMzyJykXN8DDAeOAKYB5QB59bU1nX6U1GneAIRA16VG4rSoLRo0YKzzjqLBx98kIKCglj5Rx99xMyZ8YDSjRs3smnTJr766ivefPNNAEaMGEGrVq1idR588MHYsSVLljB37lzatGlT7bVHjhzJ5ZdfTmVlJe+//z4HHnggBQUFTJgwgenTp8dWFywpKWHu3LnbpuAAMMaMxwoHd9kY17YBLs20revYOfXXy6ZBxBi8aV1DirKdkYFmkE2uvPJK+vfvz7nnnhsri0QifPvttwnCBNKnXQebSfejjz7i22+/pbCwkIMOOoiKiooar5ufn89BBx3EBx98wNixYznttNNi1/j3v//N4YcfvpV3FkdnjjcRtoNclYrSKGjdujUnn3wyTz31VKzssMMOi63mB8RMTQcccACvvPIKABMmTGD9+vWA1QpatWpFYWEhs2fPZuLEibG2fr+fYDCY9tqnnnoqzzzzDF9++WVMUBx++OE8+uijsTa//PILmzdv3qp7VMHRRIio5FCUbYa//OUvCdFVDz74YCy9ep8+fRgzxhpebrrpJiZMmED//v1577336NSpE0VFRYwYMYJQKETfvn3529/+xpAhQ2LnGjVqFH379o05x90cdthhfPHFFxxyyCEEAgEA/vjHP9KnTx/69+/PHnvswYUXXkgoFNqq+9O06o2caFr1WbeOoCDgbeDeKErD0FjTqldWVuL1evH5fHz77bdcfPHFOVvFL5m6pFXXpWObCKpxKErjY/HixZx88slEIhECgQBPPPFEQ3cpI1RwNBFUcChK46NXr178+OOPDd2NOqM+jiZCROWGsp2zPZjds0Vdn50KjqaC/s8o2zH5+fmsXbtWhccWYIxh7dq15OfnZ9xGTVVNBDVVKdszXbt2ZenSpWhC0y0jPz+frl27ZlxfBUcTQQWHsj3j9/u3aia0UjfUVNVEUB+Hoii5QgVHE0Ftu4qi5AoVHE0EFRuKouQKFRxNBPVxKIqSK1RwNBHUx6EoSq5QwdFEiKjkUBQlR6jgaCKopUpRlFyhgqOJoD4ORVFyRVYFh4iMEJE5IjJPREanOS4i8qBzfLqI9M+krYj8yTn2s4jclc17aCyo2FAUJVdkbea4iHiBh4FDgaXAJBEZZ4yZ6ao2EujlfAYDjwKDa2orIsOAY4G+xphKEWmfrXtoTKjGoShKrsimxjEImGeMmW+MqQJexg74bo4FnjOWiUCxiHSqpe3FwJ3GmEoAY8yqLN5Do0EnACqKkiuyKTi6AEtc+0udskzq1NS2NzBURL4Tkc9FZJ90FxeRUSIyWUQmbw+JzzSoSlGUXJFNwSFpypKHt+rq1NTWB7QChgBXA6+ISEp9Y8zjxpiBxpiB7dq1y7zXjRQ1VSmKkiuymR13KdDNtd8VWJ5hnUANbZcCbxhrm/leRCJAW6DpqxU1EIk0dA8URdleyKbGMQnoJSI9RSQAnAqMS6ozDjjLia4aApQYY1bU0vYtYDiAiPTGCpk1WbyPRoHRuCpFUXJE1jQOY0xIRC4DPgC8wNPGmJ9F5CLn+BhgPHAEMA8oA86tqa1z6qeBp0VkBlAFnG3UM6wTABVFyRlZXcjJGDMeKxzcZWNc2wa4NNO2TnkVcEb99rTxoz4ORVFyhc4cbyJoVJWiKLlCBUcTQTUORVFyhQqOJoK6eRRFyRUqOJoIKjcURckVKjiaCOrjUBQlV6jgaCKoj0NRlFyR1XBcJfvsKos5xvsNkcighu6KoijbCapxNHJeCdzCJb5xSLCsobuiKMp2ggqORk4eIUCd44qi5A4VHI0cDza7oYmEGrgniqJsL6jgaOT4JWw3IsGG7YiiKNsNKjiaCEbzqiuKkiNUcDQVwqpxKIqSG1RwNBHWbNKoKkVRcoMKjibCvR/MauguKIqynaCCo4ngQ30ciqLkBhUcTQQv4YbugqIo2wlZFRwiMkJE5ojIPBEZnea4iMiDzvHpItK/trYicrOILBORqc7niGzeQ2Nh57YFDd0FRVG2E7ImOETECzwMjAT6AKeJSJ+kaiOBXs5nFPBohm3vM8b0cz4py8tuj+zfs7ihu6AoynZCNjWOQcA8Y8x8Z53wl4Fjk+ocCzxnLBOBYhHplGFbxYUYNVUpipIbsik4ugBLXPtLnbJM6tTW9jLHtPW0iLSqvy43XjTliKIouSKbgkPSlCWn4quuTk1tHwV2AvoBK4B7015cZJSITBaRyatXr86ow42NhWs2x7ZV41AUJVdkU3AsBbq59rsCyzOsU21bY8xvxpiwMSYCPIE1a6VgjHncGDPQGDOwXbt2W3Uj2ypTFq2PbYtRjUNRlNyQTcExCeglIj1FJACcCoxLqjMOOMuJrhoClBhjVtTU1vGBRDkemJHFe9imaZ4fX4dLNFeVoig5ImsrABpjQiJyGfAB4AWeNsb8LCIXOcfHAOOBI4B5QBlwbk1tnVPfJSL9sKarhcCF2bqHbR2fJ27RU41DUZRckdWlY51Q2fFJZWNc2wa4NNO2TvmZ9dzNRkso4nIZRdTHoShKbtCZ442YYNhlnlLnuKIoOUIFRyMmFI5rHB7VOBRFyREZCQ7HeX2GiNzo7HcXkbTRTEr2Wb2pkoPu/pR5q0pjZerjUBQlV2SqcTwC7Auc5uxvwqYEURqAd6YvZ+HaMp74cn6sTIxGVSmKkhsydY4PNsb0F5EfAYwx650wWaUBMCbxL6jGoShK7shU4wg6iQcNgIi0A10AoqGIOBLDuCfiq3NcUZQckangeBB4E2gvIrcDXwF3ZK1XSo1ENQ13NK5OAFQUJVdkZKoyxrwgIlOAg7F5pI4zxuhapQ1EVNOIuGxVHjVVKYqSI2oUHCLS2rW7CnjJfcwYsy5bHVOqJ6I+DkVRGpDaNI4pxLPVdgfWO9vFwGKgZzY7p6THrWlE0agqRVFyRY0+DmNMT2PMjticUUcbY9oaY9oARwFv5KKDSipp5IamVVcUJWdk6hzfx71EqzHmPeB32emSsiV4VHAoipIjMp3HsUZEbgCex5quzgDWZq1XSo1EIulMVSo4FEXJDZlqHKcB7bAhuW8B7YnPIldyTBpLlQoORVFyRqbhuOuAK7LcFyVD4s5xdziuCg5FUXJDRoJDRD4lzYuuMWZ4vfdIqZWo3PC4vhLVOBRFyRWZ+jj+z7WdD5wA6MSBBsI4ksPjyvpSo+B48VQobAPHaV5KRVG2nkxNVVOSir4Wkc+z0B8lA8KO4PC6BEeNpqpf3rN/VXAoilIPZLoeR2vXp62IHA50zKDdCBGZIyLzRGR0muMiIg86x6eLSP86tP0/ETEi0jaTe2hKPPzprwCI+jgURWkAMjVVuWeQh4AFwPk1NXCy6T4MHAosBSaJyDhjzExXtZFAL+czGHgUGFxbWxHp5hxbnGH/myTeTE1ViqIo9UimgmM3Y0yFu0BE8mppMwiYZ4yZ79R/GTgWcAuOY4HnjDXaTxSRYhHpBPSope19wF+BtzPsf5PEkw2No7IUIkEoaFVr1VcnL6G0MsS5+2vmGUXZnsh0Hsc3acq+raVNF2CJa3+pU5ZJnWrbisgxwDJjzLSaLi4io0RksohMXr16dS1dbZx4MvVxREmXqySZ+/eEf/bI6PpXvzadW/43M6Hs819WUxXSvFmK0pSpLTtuR+yAXSAie2NNVQAtgMJazi1pypJHrurqpC0XkULgeuCwWq6NMeZx4HGAgQMHZjBiNj52lBWxbaluXa1QZXw7EgKvv+aTlm95wuOpSzZw9tPfA7DwziO3+DyKomzb1GaqOhw4B+gK/MtVvgm4rpa2S4Furv2uwPIM6wSqKd8Jm5F3mohEy38QkUHGmJW19KfJ0ULKYtvVrsdRWRrfzkRwbAWbKoJZO7eiKNsONQoOY8yzwLMicoIx5vU6nnsS0EtEegLLgFOBPyTVGQdc5vgwBgMlxpgVIrI6XVtjzM/YdCcAiMhCYKAxZk0d+9YkCOKNbVdrqqosiW9HsutAD3gztXwqitKYqc1UdYYx5nmgh4hclXzcGPOvNM2ix0Iichk2JbsXeNoY87OIXOQcHwOMB44A5gFlwLk1td2SG2zKJERVVWeqqtwU345kZ85mWVWIiIFguElaBBVFSaI2U1Uz52/zLTm5k4p9fFLZGNe2AS7NtG2aOj22pF9NBS9xDSKmcZStgzVzoftgu+82VWVpsae9b/2QylCElgXZM4MpirLtUJup6jHn7y256Y5SF9wahzcqOP5zJKyaCTdtABEIJznHMyQSMXg86WIUUql0oqhKytXHoSjbA5kmOWwHXICdXxFrY4w5LzvdUjIhKjhCeONLx65ywmODZRBoBmGXsKiDj+PHJesZsEPr2isqirLdkekEwLeBL4GPAJ2ivI0QExzix5Occ7JqsxUcEZcWUAeNQ9Ku+qEoipK54Cg0xlyT1Z4odSYqOIL48ST7Lyo3QfP2EHYJjjrMLg/o+4GiKNWQafzkOyJyRFZ7otSZqHM8JP7UcNyqzfZvZMtMVd6tFByPfDZP53UoShMlU8FxBVZ4lIvIRhHZJCIbs9kxpXZiGof48SQP9I7gMOGqeFmwPONz+zYt3aq+3fX+HO4YP2urzqEoyrZJRoLDGFNkjPEYYwqMMS2c/RbZ7pySyJJ1Zawsieea9Ercx5GiIVRsACAScr31r8p8IG897bEt7meUTRW61peiNEUyjarqn6a4BFhkTHW5LpT6Zuhdnybsu53jsaiqok6waQWsXwhAOOSaX75xWcbXCuYVb11ngXemr+Ch5FwBiqI0ejJ1jj8C9Ad+cvb3BKYBbUTkImPMhGx0TolTUpbqL/C5fBwxweEN2L+OqSoccpmqIpn7HDa12rP2lboURdkuydTHsRDY2xgzwBgzAOgHzAAOAe7KTtcUN49/+Wts20OEPKpiGkdYfK6oKieM1smKmyg4Mp85bsLq2FYUJT2ZCo5d3bminJX49o4utKRkjxUl5Xw5dzUeic/ivsP3JHPyz8HnzN0ISQCis8ijgiJkfSERt+CoQzhuglM9iXDEEInoPA9F2V7J1FQ1R0QeBV529k8BfnFWAdRX0ywy8oEv2VAW5LJhO8fKTvV9BkC+8+hD4ic/ukhTOFlwuOdx1CFXVQ2CY6frxtO/e3Gtpxi+a/ta6yiK0vjIVOM4B5vB9krgz8B8pywIDMtCv7ZfwqGE+RYbHN9GeTBVWygSG14bEb/NjmtMfCEmZ+CPCo6g8WY0jyNsrGYjtcwy/2HxhlrPFVKtRFGaJBlpHMaYcuBe55NMaZoyZUu5rS207Q2XfZ9QXOD3plQtwi7kFPb4bYqQqS/EDzpCIhKuImi8RPBkZKqK4MFLGHFpHD8sXs+qjZWM2KNu7vJQWJeQVZSmSEYah4j0EpHXRGSmiMyPfrLdue0TA2vmpJQGXYNwVCtoLuVEEIw4SQ6/fyLewBEcJhQihJcwnow0jkj0J+GKwPr9I99w0fNT6nwnIdf6HH99bRq9r3+vzudQFGXbI1NT1TPAo0AIa5p6DvhvtjqlpPLYF1ZOB7weKwSwGkdEfBjx0N0shxVTAQiLn+j0mmCwkiBRjaN2DSDiLPcuaaKqyqvqloYk5IriemXyUqpUA1GUJkGmgqPAGPMxIMaYRcaYm4Hh2euWUh0eD4SdKX1FUkYEr113w8WycDGrNlgz1oLfNjgah9SqcRhj8DjhvBJJdY4/+vmvKWXV4fWI+jgUpYmSqeCoEBEPMFdELhOR43Gt/V0dIjJCROaIyDwRGZ3muIjIg87x6e4Z6tW1FZHbnLpTRWSCiHTO8B6aBB6RmMbRnHIi4sX9Na73d6AKP+VVVmNoU+AhhC8jjcMYO0cE0mscxmQuCAr8Xl1KVlGaKJkKjiuBQuByYABwJnB2TQ1ExAs8DIwE+gCniUifpGojgV7OZxTWHFZb27uNMX2NMf2Ad4AbM7yHJoExYBxzUjMqCIvPqiEOHgxhPLGoKI8JEYz6OGpxjoeNiU0qlDSzzPPTOOirI9/vIRyJsGpTBaWVmpVGUZoSmUZVTXI2S4FzMzz3IGBedJKgiLwMHAvMdNU5FnjOWXt8oogUi0gn7EqDadsaY9xZeZtB015xqLcsYZlpy2YKuMX3DBtoEQvDDUgII16QuOAQIoRdKwJ6TIiQ8eKXcK2mqkgkjF+qN1VlwtBebfly7hp26VjEwjVlDLr9Yzq0yNuicymKsm1So+AQkXE1HTfGHFPD4S7AEtf+UmBwBnW61NZWRG4HzsImWkw7j0RERmG1GLp3717TbWy7RCJMyLuG1aYFh1bezdm+DxMO5xEkIgW4FUePiRBGEEe7kEiQEJmF40ZCruNp0pNIBkuQH9uvC/89fzA9Rr8bK/ttY2UNLRRFaWzUpnHsix3AXwK+AzIYOmKkq5usHVRXp8a2xpjrgetF5FrgMuCmlMrGPA48DjBw4MDGqZU45qJ2spFnA/9MORwgSFCKEkxVFcFQgllKIiGq8BEhmD5X1cKvYP0i2Pt0Iq5Jf5Im6bFk8PX7PHX5iSiK0hipzcfREbgO2AN4ADgUWGOM+dwY83ktbZcC3Vz7XYHlGdbJpC3Ai8AJtfSj8eJyUO/lSZ02EyBExJPoHC81BUTwxExVEgnFo6rSOcf/cyS8fQkAEbdDPI1ZS4Rac1R5axEcmuNKURo/NQoOY0zYGPO+MeZsYAg27chnIvKnDM49CeglIj1FJACcCiSbvsYBZznRVUOAEmPMipraikgvV/tjgNkZ9KVxkBy15Jq9PSfSNaW6X8IY8YFrsL4o+GfCeOi67ltYvyguOIyHkrKKlHPEiISJuOZZVJdy5JdVm2q8hdo0jq/mranxuKIo2z61OsedRIZHAqdhndYPAm/U1s4YExKRy4APAC/wtDHmZxG5yDk+BhgPHIEVSGU4jvfq2jqnvlNEdsGmg10EXJTx3W7rJA/Wrn1TjZkoIr6Yc3xBpAOzTff47O8nhuHx94xFVa3dVE7L6q4dqsCE3aaq9P6QRz6teS5HuJaQ3UVrNwPtYvtzf9vEV/PWcO7+PWtspyjKtkNtzvFnsWaq94BbjDEz6nJyY8x4rHBwl41xbRvg0kzbOuXbhWkqeb+lbE7bxHjigiOY/HWWrUWKuhEyPiLiocBTw6D+85uEu7jiDNKYqu58bzZH9u1U4y0sWG37efeJfbn6tekpx5eXVPDvj+fyp4Ot4njUv7+iMhThnP16IJl43xVFaXBq0zjOBDYDvYHLXf/Ygh33dd3x+iR57oTLVNVJ1qVtEvLkIY7gCDlfp7hiENzzOKKT+9Iy4w1Mx6Gx3ahz/P0ZKxK7WIuPYvCObQBo0zyQ9vijn1mNpUfbZhy9V2cqQ85iVBGDz6uCQ1EaA7X5ODzGmCLn08L1KVKhkQXCoaTd2pc6CXsLXBqHnaDncQsOx8dRWziuEcG4tIyoqWra0pKEerWlERnUszUAK0pq8KcAf3rpx4R9nWWuKI2HTGeOK7kgSeMIVtU+/yHszQexAiMUExxxzcJjooJDalw6NtR5HyJuH4fjX3HrALt0KCKcYVTULh2KMqoXJViHZW0VRWlYVHBsS4TrLjhC3nzEEzVVWcHhHuyllpQjG0wz2zavVYLgWLTaTtB3ux0KAt6MBUfPts1qrVPhWpwqpBqHojQaVHBsSyRpHKFQ7Wk/It78uKnKRAWHS+OIhGpMchjNTRUKlkNlfE0uD+GUpIbGmIwFhycDR/f81ZtjgkkXfVKUxoMKjm2JJB9HKFi7xhHx5adxjsfxmBAR8TmJD1M1jqg/JFxVRWizdcAHjRcvETaUBRNmixsS19hw0zwvMc7Ck8EM8g1lVTEBo2t1KErjIaMkh0qOSNY4grVrHGFvPFdVsJqoKuPxEo6kN1V5xA7YkVAl60vW0w3YSCE+wnhEKHMt3hQxptq0I99cO5yqUHzwr20GOVhHe7SWmqoUpfGgGse2RJKPI5yB4DDeQGz+Q7qoKq8JgcdPxKQ3VfkJOdeqZP1GOwejnDw8RIgYkxBWawwc0Ktt2n60yPfTtnk8C643A1NVOGJiGkd1moyiKNseKji2JZJmjoczMFXh8SFOVFV0gac817IZVuPw2aiqZMERtt4Pe+kqwo4zvsIE8BEhbEzMBLWfZwYF4VLScfa+O6R2K4NfVihiYj6OqpBqHIrSWFBT1baEW+MwhlDQ7lcaH3mSPneUeLyxvMFRwRHwgqNI0CxSivH5CeHFkzzBMFQe24yEKsHY4+UECBAiHDEEwxH29fzMi4E7+HFzX74IP51wioV3Hpm2X5loHKFwJO4c31KN4589oMsAOOP1LWuvKEqdUY1jWyKSmJ02ErIaQCX+apuIxxPTOKI5qlL8EB4fVfhTBUcwPknPBKtiq/6Vk4ePMOGIYcayEk702kTIPcOLCEciGa3L4fZxDNyhFS9ekLwUi9U4okFawbChvCrMnJU1J1FMoXw9zPuobm0URdkqVHBsS5Svj29HQoSdcNyaBYc3ZheKCo6IJC7xKl4/lfjxRpJMX+H4vglVxgRHhQngJUI4Ynhr6nJmRqwp6vXgfoQiJqM1N9x5p+47pR/79GidUiccMbEw3FA4wp/HTuXw+79gsy41qyjbNCo4tiVKV8U2yyoriTimqirS530CEI8vFo4bNokCJIbXTyU+vJEqWDcf5n9my12msYqK8phGskPHNvgkHJuzUYjj+8DPmtLKjCKmEi7vkbSmK7fG8dRXC5i0cB37yGyCcybU6fyKouQWFRzbEi7n+OC/TyDiJDkMSvUah0c8sZnjYcdElaxxVHoKCeHHGwnCg3vDc8cCsGJdfPn22cvWII4giTTvSAfWx+aRNBNr0jIIG8qC+DLxfLvweiTtvI6ga+7GhJm/4fPAq3m3UvzGaZmdOKRL0ipKQ6CCY1vCJTi8hImEooKjeo0DrzeWqyqMlxG7d2S1v0tClUpfEUHx4w+XJZQvWh1PYBgghCdir1dR2AW/hCFkBUZzrBPdR5hgOLJFGgfAvSftlVB+7Rs/Jezne+roIK9MH+WlKEp2UcGxLZEgOCKxpVxrEhxejy+mcTQvyGPMmQMY3+r0hDpBX3OqCJAXSRQcPuLXyyOIJ3p9n52PEXZMZVGNw0eYT+eszsg5ntBHp0FUURnqmguyl8xjmMdmyi3w1lFwVGyoW31FUeoFFRzbEi6fg5cIBO2b/vpgTT4OT2zCn3F8HavyunN38ORYnSpfEaFkc5cxiYJDgmCCVBkfHr+9XkVlVOOoiPcJ2FAWZF9n3Y1M8DrrbESjvdyO87fzbuSZwN0ArNlYRw1izdy61VcUpV5QwbEt8entsc1iKcVbtZFyE0hd2c+Fx+uNpVGvisQH6EfCx8TqVPnTCI5wEBM1hRkveRLCEwkSxEd+wAqOsooK+nUrZpd8m8PKRzz9yAt/HMyvdxyR0W0lO8arW6/J1LT+yPpFqWWVdQzdVRSlXsiq4BCRESIyR0TmicjoNMdFRB50jk8Xkf61tRWRu0VktlP/TREpzuY91DsrpsOPz9dabULeNbQp+ZlS8mMT+9IhXh8bNluNYEOFFSAGg3G1CfuLUs1d4UrESapYTj4tfJGY4CjIzwdgdclmiIToHlwAxAVHgd+LxyMZ+zqi9UqdMNtzNj/FX3yvJNRpQ0mCYMKdGXj+Z/BAX/jptXiZMTDD7ofw8fCn8zLqi6IoW0/WBIfYWWkPAyOBPsBpItInqdpIoJfzGQU8mkHbD4E9jDF9gV+Aa7N1D1nhsaHwdtpl1lPwhCv4zbRODa914fV4CYXsgBtNGZic+TwYKGJDVdIgH6rCOFFbZVKAzwTxRIKE8NKuZXMA1mzcTMtwfMna6MBeHqx+JcH0fbTXPm7vLpw4oCu/W/Myf/K9RWviUV07y3L8rtnxK9eujZ/gt5n275Lv42WTn4Jf3gegyni5+4M5deqToihbTjY1jkHAPGPMfGNMFfAycGxSnWOB54xlIlAsIp1qamuMmWCMiY4wE4GuWbyHBsUbrmSZaWvzTAFVaSYCiiduqorWS15HIxxoQRn5iQ3/d3lspnq55OMjiJgwIfEhXnsdCYdiKwECeJ1Mur07NM+o/zu3b+60s/1qnufjHldkVTeJz1tpKaX4XRrH95+9a++lqgw+cN4NTJjxrz3NlBsHEln8XaxuoVTGkjUqipJ9sik4ugBLXPtLnbJM6mTSFuA84L10FxeRUSIyWUQmr169uo5dzwGm9qR+nkgVQQnETFWbPKnLvHu9HvJ89riJCY6kOr4AZSYvsXD2O3iCTjZcKcBvgnhMyF7LGehblc3H40rF7iNM7w7NefXC/TK6xZcuGMIz5+5T7docRRLPlZVHMMFUdcysv3DRf6fw1YRX4w0iIQ6ccR0DPHOZ81tihNgfvB/HBGZFMEyP0e/y5JfzM+qnoih1I5uCI91okTxaVlen1rYicj02ld8L6S5ujHncGDPQGDOwXbt2GXQ3x6RZVCkZb6SSoARimgTNUlOai8fHXl1bAnDu/j3tqZMkR0lZkN7dOqa0DZTZN/4KKSBAFWLChPFCm50B8IXKYmlIwM4tGbFHJ1oWVj8h0U27ojyG7dK+2uMFxCfw5UsVgSSt4f2fV1Kybk28IBIi6Myi/3XFmoS6QXysKa2ix+h3eegT6+8Y87kKDkXJBtkUHEuBbq79rsDyDOvU2FZEzgaOAk43yXaZxkI4ddZzJK8ly0w8zNUXqSTkycMn9hZbtu6Q0sbr9cWilJrl2QE92ccxccFaIr4kUxXgq9oAWMHhN0EioZAVHM2tkPGGyxM0Dj9hLh++M3xxT6KjOhMWf2d9FK6Q4wLiDvC+Mp8DPD+lNGtl3Pm7wlQ50WEtSQzdLTX5rCyxQQIPOY5yt6JTGQrzzvTlKWY8RVHqTjYFxySgl4j0FJEAcCowLqnOOOAsJ7pqCFBijFlRU1sRGQFcAxxjjCmjsRJOXaRJQpWUmoLYvj9SSUgCjPBYp7BnfeobtMfrIa6MRZ3jiYNjns9D2JNkqgJ8QRvOWuktxE+IqmAV4vVBoNBeP1yWIDj237EYn9cDn9wGr5+f+b0CPH0YPHVobG4KQL7EheeZvo+41v9SbP/bsI2FCBjXc4qEiP5k3WYusPNQVm2qSChbtamSRz6zQuTBj+dy2Ys/8tkv26DZUlEaGVkTHI4D+zLgA2AW8Iox5mcRuUhELnKqjQfmA/OAJ4BLamrrtHkIKAI+FJGpIjImW/eQVUJJgsMYJFxBOfEBPi9STsgTiOWqEk9iDioAj8cXd2okzZf4NLwX74SHEPB5wBsPx/1v6BAAfFU25UilxwoOrwmCxwe+AiIIvnA5mLj5qFl9rN7iyi9VRHq5HzGCRyIc4ZlIq/VxLWR1yeaYEdNt5gI4yzuB3zamanF3vW+jrVZssELl+wXrUuooilI3srqQkzFmPFY4uMvGuLYNkDY2NV1bp3zneu5mw5CscTgDanT511ixN4/IeZ/D4/shabQUr89HdRrHucFrABjg8xLxxgXSO+F9OdP3EQFHcFR5rJYTMJWE8YHHQwV5+MLlrC4pIybLIvUQueRaPKqdbExbZTP5DPbMZnBgNsTTaTF5wWr6O+6VwiTBsadnITdMWlztZQNOAMGrk5dwzYhdt7DziqKAzhxvOFIEhx1Qk+dsRDx55LXb0e7seFDKaTweT4rGkbyYXp7Pg3FpHJuwgqLD6m8AWFJmhVVepDyWWXezCbB4xerESXnhGmZ2Z4orjPYi3//SVikj1awG1scS1am6eVJNTtOWbqj2sl2K7T2fMKDJRm8rSs5QwdFQJKcEd1bji66pESXszQN/AVz5Exz9ABGEqS0Pjh33en2QV2R3nL/JPo5AkuAoTxqY11bZY75QWUxwlJl8CqSS2/zP2Er5xVCZpCH8+klm9+oWOO/8udbqbj+Pm67NDV6TqPUEXXNbks1XsTrhCPd/bPNauR/NS98vpsfod6mo44RGRdneUcHRULg1jvmf25QakJJexESd2sXdwZeH5+YN9PvzG7Hjfr8fBl8Eh/0dBo2ybZIuFfB68Abig3EkKdq51JkcWEBlXHCQRyGV7OJZait1HwJlayHsGrgXfJHZvQZdvoyq2vNLrTKt0pYHIhW0MXEfxS+RLpzeNp66pEU1PpO3flwWW5SqKhRXx+6d8AsAG8vrQZNSlO0IFRy5xP26GxUcm9fAc8fE9iuStAG3byIdXo8HfAHY70/gzPhODjn1eoQd2scH49+SBuYWRVZTKTAVRIgKjnwKcUUpte0FZeviqwcCFKbOK0nLr5+mKaw+z9VyUpeZXWuKyIskRlL19izj+6XxspayOe35vpwbn/NR5Vo8Kvac6pgmXlG2d1Rw5BK3llG+wf59/Y8JVV4LH8j3kV1i+7UJDtJEWiXP4/jLYb3p1KY4tl9BHmNCR8f2R/brAUAhFS4fRz7NxSU4CtvauSdf3x8va5luMn8aXj07tcxX/X2VJ81y32gK+DSyNwWRzWnXXx8fGAFAC9ILjnHT4tOHgqHUNT/CyQ9MUZQaUcGRS1xzGNjs5Gman/g2Xko+dwbjS6eaGgZYILb6n5tkH8fO7YvwBxInAJaYZrFtr98ey5NgTHBUEiAPa8JZ0XF43I+y2eWUTg4pBjvJr8wV8vrlven7nSZCLEo0J1fYCUP2EWFnWUrr8OpYn9x8VHA4UL3GAZDv+D8WrSvj6lenUREMx0x6obAKDkWpCyo4conbIR4sT1ulwgQ4Y5DrTT7NjO8EJPUrTDc52u9LrBdwBuCHQsfidQmViNgI7U6tWxAgSMQIpS13Ab+dFMhmV6qP5ME/WGEn+b1yVrzs41tTO9NxTzDOm/9R9yUcuqrqopjgKG9mI6A8ROjniU9+fJNhAGw0tk/lPpvDq7Wk95/0kYXMzj+XwzyT+H7BOl6dspRx05bHBKx77fME1v4Kb16cXkAqynaMCo5c4prDUN0bdwUBAsWdY/vpUoUkkMZUFTW9dG9diM/Ju+H3eji68u/sX/EAEH87b9mqHR6/W3DY80W8AfKlCo8Y60NxZpNT6Fr5L/keNi6zf9fWsjbGua68lN0GwyUT4ZKJDKu8lzciBzJ8dys4K4u6AzBR9koIGqjCz4CKRzmg0t7LBp/Nh9WZtbSglGM8XwPxdO5dxWpJjwfiQmr95qqYgC2rqiaqavz/wbQXYdFXNd+PomxnqODIJUGXzyA5HNdhM/mEW/WM7csWaBzR8NInzx7IPGeVvoDPw09mR5ZhEz5GHeQdeu6REHFlHI3DePyxmd3iz49rHHmulOopc1Gc+/PUkgQxavYCCDSD9rtB+91YYDoB0LuLdbpHCtvy+8qbuS3vL9zV/q74LXsDrKUlG7HmNl8gj5C/iGIp5VH/AzwYeJiOrOXlUUNsfdelu8tvAPzjvdmUONFUYz7/NW03Q17n2etKg4qSQFZnjitJhFyCIzro5hdDxYZY8TpTRL4/rkUUFMR9EWlJJzichZ3yffHzBLyJ9Z4MH8Ec042D2v+OHQNx85PHWXPDePy0lKjgKIgLjk2/xfucLDjmf+70KU0/e/7O1k9eJzwQF0SPnN7fpojfaEOAfR7hB9ObzuQx17NjrN7u3dtydfddOGe/Htw07meuOLgXvudacepORfDLQoiAT8K0bmbnp+S75nd0Yh2LSUwW2at9EemYtirCAGDpyt/omrwEmaJsx6jGkUtCaTSOSAj2vSxWvJFmMfMSQOuWtSya5E19ux/upDIvbhY/5hZGAGG8fBbpR3nI4PfF3x8CEdvHkLjeKfwF9gNQ+hsUOuGyyYIjuuBSJALzPko8FgnDee/DX5Pe7gNxwXjEnp04eLcOsbLoczDA1wvjkw/9gXwuHbYzzZyFobq1LoSCVhSEN1FgrDnwphE9Y+uUFLqSKbaUeFbdnWQZx3i+oWVB6vvTSWO+4ac1VohWlvyWclxRtmdUcOQSl+AwoUrrxQ6WJTnAhR3aFHJv8ETKTYBQy56p53HjmhEe5W9H9eH76w6mRX5ccETfvpOJGIOn1Q6xfb+TjTboUkY9gYL4AB8JwjrHUf3J3+Hmlqkn3bgUnj8BprsWYWqetCbKyc/BTsOrcf5bgRFVkiLGJIThev1pIs0KWsHcD2JO90N7xfvlTt/ujrx6JXArDwYeIlLhSoiFnSQ4aWE8nbt3swoORXGjgiOXBJMERzhoBzqXc/rIPTsR8Hr5d/j37Fb5H/L8tXxFaTQOn9dD+xaJA3LA5+HPh/TmjUsSV+8LhQ1FRUVcUHUVAOIM2gmCw18Y1zjSUVmafmGqN/4I3jxr5kqKnqLPsXDmmykZfW0n4g59gOP6dcFt//KkC1Fut1vifqiSzi1tn92pSFq65nq0iUZhlccFx4qScnrf8B57yHx6OP4QXIJl0drN9Bj9LtNryIulNByfzlmlKWRygAqOXOKKqoqEKuOpOHwF8PsnuS5wLQGfh8K8uFkp2cSUQhqNozquOKQX/bu34r5T4ut+hyIRWhb4Y1l5o+N4lYkLDm9eAfhdvpZBFyaeuHxdfEJjMuFK2O9yqxFkiuNP8TdrxbSbDuOvI3alRb5LkPnSON9bJiUvDJbHlqwtkEoQL6WeInaS5Vx3xK4M3CHeH6mM933Bms0UUsE7eTdwkHeaPe4SHB/OtMLkjR+WxcqqQhF6jH6XBz9O8t80EowxPPftQn75rXEHAfy0tIRzn5nE39+dCUBJeZCJ89fy8axtR2NctbGCq8ZObfTCTQVHLnHN3fDOfBPWWXv/9+ub8WrVED5hH/xeoXlefJAsqEfBEeX4veODbLuiPESExcY6jH9oNRKAKpfG0aa4OFHj2OOExBMu/zHRf5NMbZMYk9ntaDj0Vhj+N1oW+PF6hNcujmtK1Zqq3Ky063h0bO7jPO/7YCKskVYUSymtm+XRjPh3MfeX2WyqcCYWGugsicvSiiu/VnR1wWYu4R5t+8QX285StQvWbCZU3fyUJOatKuXGt3/msPu+YMayktobbKOsKbWa5etTrFA/5F+fc+rjEzn/2ckN2S3enb6CHqPfZUVJOXe+P5s3flzGO9NXNGifthYVHLnElexPqkphhX2jveertVz92nRWbqxgxrKNMacuQIuCWkJb05iq6sIJ/a0QmW86s1vF0/zQ8lAAeneO54sqKGwWFxyFbaD74ESB9cpZNQuOmsxc6fB4Yf8rEkJ/e3eIRz51bp8mR1YgKfrsw7/BP3vwVfu7Hee4wReuYIhnFr07NOey1t/Hqt5c8U/+PNZ+FxEDrZKWpd20YW3sDXFDmRUSLfL9/P2dmQy75zNKK23ix3RWt/qkx+h3+dNLP9Za7/0ZKxh2z2fsfP171dZZtbEiNt/HvQDWnJU51joWfAnPn5iYPDOJZ79ZyLQlG1LKF63dTK/rxzPX0ZQqnXQy5cEwxhhWb4rf19rS9OHvueDNH60gm7akJBbdWJUm9U1jQgVHDjFVSdlbS+3EtArig/DMFRsR1whUlF9LxHRtuaxqoZlLuyknn5AzmOzcvkW8kr+ZHRVPfBoucFKpn/JC4omiUWI729UF6bBn/FhdNY5a8BQUpxbmpYk+K1+Pb/mk2G5XVtFKSuk241EGdoo/80mRXZnzm43aihhD86RlaVtQxs/L7fFBPa1A/cd7s3nyqwUsWLOZUx6bCMQnHGaT/01bzjWvTa+xzhdz19R4vCIYZtAdH3PFy1YIxbQt4gte5YzXzoN5H0LJknjZD8/BRie/2NSXWPDuvRz78NcpTd+fsZJg2PDyJNu2MhQ3/1QlaVt/eulHSsqCPPnl/JQkoMFwhEVrq09XUxtnPvUd177xE+s2p5/UWxiw2mlFMBx7vmMnL0krDBsLWf2ViMgIEZkjIvNEZHSa4yIiDzrHp4tI/9raishJIvKziEREZGA2+1/fhKuS0oyUrQVS18dw06llLRMA6/o279Cq0M+uHVPnL4Sjq0C55pbEwm/3OAFa9bDbyW/4YUdwtHbmWxQUxycC+rasj9WSnyaSKzofJC/NMYc1bQYA0GrinciyuPliPc2JrFvMfWPHEzaG5iR+Ty1lM6s2Wo2qrCr1zXilc8wthNdtruLtqcsY9Vz9m0nGTl5S4/FNFTWv1Dh9qTVHRc0lN7w1I3bMk221KZno72j9Avu3ZBmM+xO8eq5N+fLWRdzsfy5Wfd6qTdz6v5lEIoYCZ0COCoxK11t8RVWi4Pjm17XsdesE/v7urJR15297Zya/u/szzno6roUaY5iyaH2KkEnGGMOXc9fw0veLGXbPZ+nrOH9F4vOppi3ZkFYYNhayJjhExAs8DIwE+gCniUjyNKqRQC/nMwp4NIO2M4DfAxkuBrHtEKoopdL4uKrKWXK93CYDrEiT8fXw3Tvg9QiFgVo0ji38R590/SGMv3xoSnks359rYh75xaknSBYcmxwHZBc7ONOqB3Sya4zUt8aRXnA4/fHnw7GPpB4/8RnaDr88vj9nPIiHWd5dOMr7HV/nX8GfZ53GJ1NmJmgcEyO70ULK2Fxp3yY3V1bv1Gye5+Pl7xfzx2cn0/+2D7ni5alMmPlbbIZ6rvhs9qrYdiRN5t+TH/s2YX+t60350hd/qP7Eq2bV/yz6aMocR/uOaRoVJVQtmRKr5sMKw4uf/4Gnv17AjOUlMZNuRdCZb+NyOJfX4Hw+95lJBFf/GosE/MIRJF/8spqyqhCTF65j/E8rOeHRbxKCINLx+0e/iW1X9z27k46mi5I0xnDNa9PpMfpd3vxxaY3Xi15n/urSWutlk2xqHIOAecaY+caYKuBl4NikOscCzxnLRKBYRDrV1NYYM8sYMyeL/c4akcrNVBBgtrE5mKJZZCtM6sA65owBzLltRPUnG3wxNGu/xX3xeT2xqCOAy4bZpdxjA83gC+HAq+G65eBNI7za7BTfbtY+/sa48yFw9jsw8p/xzL156Wdm15Vvws67QzoNJlrm8cfzarnZ9ShokZQGvll7NnkShdCKGV/ENI49Kp7k/fA+AHw05Re++GV1zJ+Rjjy/l9Fv/MRHSVE881Zt/T95OgFQHZtcfZy7qjTmA4iyU7u40N9YkTrYpU0zX74eHhkC711T/YUXfAGr6/ivGQ0YCTqmoqjJavUs1qyPO+qj38lc51me9tBHnPLuntznfzhmaosKEEgUHEKEhfl/YELgagCK2YT/4f7wb/uS477bm8f9zIljvuWmcT8D8O38tQndLa0MMWXRev4xfhZn3/U8ixcvSjg+Y1kJ4YjhqrFTGTdtOaNfn05lVLCFIjFTsJsVJRUxLfLv78xK+5gWrtnMknXW1H3cw18z/N7P09bLFdkUHF0At0691CnLpE4mbWtEREaJyGQRmbx6der61A1BuKqMcvLipinHVOX2cTx+pv0xiwg+bw1fz8g74er6C//co4v1aYSipip/AQy/IVWziOIWBptXwef/tNuFbaDnUCcH1a62rHmH1PZbQI+LX6Hs4DsShVaUqDltn/NSJxXmt7SJGpPXDyldSciTKLS7y6rYOiS3nDiY8w7ZG4ATl97OOU9PZO2mck70fs5ekprIsTqb9cwVG9OW14Vkm311JJtWLnp+Cofe90WCwDtol/gLx81v/5xyjg1lqbb6f7xqByozfWz1F3/2aHh4UHrzjjGJudocQpVWEEScv9H/CYCN6+P/t33a2Jec3/cMAoZdZTEAx3u/5oOff+Od6ct58ZPJ+B3NpLK0hL/7nqIlpRQ7wQ69PVZ7+Jv/v/ak6xfQY/S7LFob9z1OXmQnfkYjtBavS/RL/unFHzjh0W947Iv5PFt2Ke/nJVrgj/r3V/zx2Um88eMyLn/pR16etCT2IlEVivDY56mRd25NZW01fpKD7vmMoXfZJRgWrNkcO19DkU3Bkc6GkvyLqq5OJm1rxBjzuDFmoDFmYLt27WpvkANMVRllJo9y4wgKR+No0yr+1jtghzrMd6hHovZhf03CKpn/mxfXesqdmdZu09kR91gnesc9U9tuAZ07d6Nw6KXpzXOFreHaZXDAVammsRZO+HEaAdYxnGiKaCWbaE45pSafA3fpQJdONvHiId4fGe17ifn5Z3CP/zHezruRL1rdxogOqeGreVRxifctCpwVFFdvTB0wS8qCvDp5Sa029CiVGQ4Sq5Oih6KDzPQlG/jftOUYY/hhcXxWfHRgvO3Y3WNlyU7eq1+dxuRZdsCTSCh93n5X2WH3pbEiT3wEbu+QmJbfGDt5FFi33umTy7e26MePY9s7FkVg5Qz+teJsLvGOo4PY+utNcwIE+cuL3/EpF/BKwKbxz5/1Gmf4PuYy31sUJq1F76lhKGnpRDH+zjON873jKU/KnDzN8Q9FBVQ7Sf3+P52T/kW1uu9wbWni8z7h0W9Y7/oOqpvzsW5zFaNfn8438+wzDUcMj372a41acX2RTcGxFOjm2u8KLM+wTiZtGx3WVJWocYTwMHSXeBr1Wn0aWSI6X2THdrXkxnLTvF3cGZ4OXx7sdlT241Sj5DW310qe21LU0f71eOHEZ+LlB9/EjiE7ID4fOpg1pgVdZA3n+96juVRQlO/DG9VkgFG+dxNO2718FmNKLmZO3lm0YwO9xSrJl/je5q/+V7h/tzk0z/OxIcn2vXDNZva6dQJXvzY9IbVJOiIRwxlPfscJLlt6tDwdFz9vfRRH9u2UUH75y1P500s/0veWCfy4eEOsPPqG3atDUWxiaMmqJbDemmDKVi/ipx++odiV42v+kjTO+eVx38jcVaWpc0h+cBzcpXH/C48fhA87KAYrnPO7JpIe7o0HFuxS/iOM2R+AQ71TYmuvtJJSfsk/myM83wGwt8dqguWOs66TrOOgNvFn/OP/DaSfoy1GjOAhsZ/RZ/OA/yH+5n+eFuF1zFiwjE+//JLKUJh1m6sooIL+Undt/18T5sQ0+wuG2lRCZVUhznjqu4R6Uxat59Up8We8dH3c5+ZeO2bIPz7m5UlL+MOT3zFtyQae/moB/3x/Nrc7EyCzSTYFxySgl4j0FJEAcCowLqnOOOAsJ7pqCFBijFmRYdtGR6SqjHICbKSQMpMHwc2UmgIKAt7YwB2NFMk1h/bpwAOn9uNPw3euW8N8V9ju+R/Wb6e2lGjyxehsd7dZbY/fx7d3GRnb3LHvAaw3ReztGhDy/d6MZrznSYhJ+ZcwIe8aeslSeonVYg7fpRWFAS9VoQgT56/l5DHfctF/p3CQK/qmNv9HaVWIr+atSal33Zs/xWaxg9UsznjyO6Y4gmDIjm0S6kdNL9VFXDUL+OhTHKaPLKTbhxfAAzawYfOLZ/N+3uh4+hXg48evYfHaRBMOq+K2eR+h1Lfe6Dwf43p7XjE1tvnR1PmwZh4sTnTcfxG22mr7qrhmuLdnHn/1JZrMTvUlrqT56rf2ezzKO5HbN98cK2/1yvH09Nh78YiJmbHcdJXVFDs5zdqs+o7y/57GsI+P4tZxdlLpS4HbGZt3W0o7Nyd4vuBM74SEss1VYWYs28ignq1p3cy+PPa58YO07ROzWcdfEtZXY8o69uGvuX28/Q5e+r7mqLv6IGuCwxgTAi4DPgBmAa8YY34WkYtExAkrYjwwH5gHPAFcUlNbABE5XkSWAvsC74pI+ie/LVJVSrkJYPCw0lkPY5MppMDvZcrfDmHWrTU4w7OMiHBsvy51M1WBNUdFab9b9fVySc+DbG6sQ26y+9X5aQLNYdTnsOtR7HfSn9m5c9vYoMLwG+zfpDkjq0wxHHJLtZc+0jsxPoGwZCl5fg8vT1rCqY9P5PuF63j/55UJ9dPNH7jov1O4+lU7ITHdGukAL09awgWuUN/jHv6ar+bFzUCDe7ZO1wyA9kV5nDKwW0JZYZ6Xnf53AuPzrqPDRic8t3IT7dbbuR5X+l6P1b3AN54Fyf12maDm5Z/F5rVJ0UjRVRSricra07MAHhoAy6YkpLe5I3Q6S01bOgQXJ9RvIYmCqwNxrcJDJCEzQAKrEn06rSXuf/IR4gDPT3yVd0WsrKOsZZ+I/S7e/X42AP08iRmek7UWgHsDY7jN/5+0XWiR70uY5JuOMpd5yu30T3bWNxRZncdhjBlvjOltjNnJGHO7UzbGGDPG2TbGmEud43saYybX1NYpf9MY09UYk2eM6WCMOTyb91BvlK2jeMNMFpsOHNanQ2wRok0UUhDwUhjwNZi2sVVEM+sWda636KmtxuOBgefFo6iSBdrvnMigwjbQuR+c+gKIIOUus1E0rNgVivxueBDP7/eBTU5ZDTvIb+xa4Jzn24c4M/yWc8RwoGdayiDz3y9nMnj0f2ORQes3V/H+zyt5dcpSjDE85kpjslunFjx6en/SkRwK6l6LJZmRFe9y2m9388Cp/WJlzQI+fOsSzS+/vv8QpcYGGlQlLd0TCUeYuXxj7Lrh0kS7fnDlLDsn4+aW8POblOdZDSg0x/WeV7wDk2QPZke6JQ7Gnfdm2pD72KfiYWab7mwyBXQJJkYvJdPDE9eIiiijmdSQyQCocpYb/ijvrxSziRsHe5mXfxbPB/6RUK+1y0TXSko50DMt5Vyv5f+dAW3t97qTLIv5tgA+dCK53Jy3+SnO+6hfSvkBO8czItz1/hwm/LySF75bxFH/jq9AecXLU2u8ryhrSyvpMfpdPkh6UakvdOZ4rli/AI8J8XlkL/bo0pINxvoSNlFQbcrzRsMV0+CSb2qvl2t2PRL+8CoMuTSxfNh1cMPq1LDdPsfEt6PzWHz2u/k5sgOXBq9kl04t45Mu84tZeFmi6+1479e0qornIRpV+Sx5VDHMM5XnAv/kj95EP8lTvnv4Lv8y7v1gDss2lLP3bXFz3+yVm3jcJTgCXuGQPokO/hnLSng1zYTA/ED1/9q3eJ+m3+px/O6rswDDBd536PBQavr+f31Xxmas4Ggjm1hrivhH8DQAKstLefChe3nloesBWL14dkLbjasX21nhAK+ew5xVVkP46at3uPHtGVz89JewYRGfV+1GWdIE2HCgiE+8B7Aaq5XPND1iTugbg2fH6n0W3ot0tJGNKZM4E9hpOIEL4gLsEO8P9PGnn69xeKv499uKTTwX+GdKnf7M5lW5Gg8RPs67mln558WO9fIsoz2Jfqz9Vr0EQB9ZmFDevU3i73HUf6dw/ZszqCs7tCmMhS1nK3+aCo5c4UxsWu1tT8cW+WxwNI6NpjAhrr5R0qpH3bLf5goR6H2Y1UCS8aUR1oe67NZt4r6eJ/f7hBOqbo6dkkEXwNEPwF/n06NtM2adOY31l8zktuDpabvRWdbSWayJYSeJC5WzvB+wr9c6Mj+bNpf97/wkdmx3WcjIBxKjk4oLAymmxKP+/RVXp0lBkufzMkhmcb53fNo+ARSvmUw7Srje/yISLEs53kzKE1LSbzDN2YAVqBvWrWJM4H4u2PwYF153Ex2XJy7ctWz2ZFgyMbYf1Sh6yVKe+3Yhs+ba+R7LTRvr73OxalMFDziZhg/ZrT2LI/Hw4U8j/WLbE9qemfa+rvS/mV7jiGqPhW2gRTwg5eL9OtOnQ5q5P212pksoPiGvbZoIqiieTSt4bkT6F8CXA7fh96YGiPT1JA7qo0fuWu3568KitWWxSKxMw7jrigqOXFFi32hK8zpQXOinxFhhscq0SkhVoTQgHg/88RM45t/xeSHAHw8bwJ2nDAawaVp8eTDgnNis59126kGr9l244baH4udqGfch3OiLp8wAOM37MbPyzuFW/7Oxsmbl8TfbQz2TeTfvOo71xFNS7NqxiHuO6QEf3cLDJ9U+wBS9fS6v5N3G3/zPE3WuekkN62wm1b+Zt6KUFq7jG2jOBud3+/zHcf/KwZ7UxIt7Bn9KKQsboblU0Jm4IF1BG8pInHcze9m62PY9J+3F0CNOi+2vNfHQ9TuuuAAOvgkOui6h/dH9unG8N006j26D7N9m7azv6sh7AdgpPJ8WkSTfS8c9YbdjCJTHTWCPB+JrysyKdGepSUy2ecBnJ6deE9jRs5KZrf6PgZKolQ3edQdeuXBfft+/CycO6EqLfD9jRw1Je45kpt54aI3Hz3nG5mjzZSl/mgqOHLF08TwqjY9fSvMoLgww0PMLAM2lPCGNutLAdB0A/c9KKT5u7y788veRNYYri1uzKY6vqjjMO43b/U/H9v/hf4oCSYyO6SpxH8HOYoVIH0/crj965K60m/MSfPUvjqx4h6OSwm3dFHhCeGb/L7bfX+ayr+dnfs0/k5O8nyXUbU31KUSifdpo7Nt4j45tWW2KAXgn74a0fY/StnxBStlMY5/JN/mXM9zzIxEjzIp056Bd7Dyr7yJWIE6NxLW94sIAvu77xLSSMvIIXzEDLnWSVw69Cg5yzWbvvi8y863ECxe0hhF3QmfHP1TlOPb3+SN07AuLvoYlrpDYlt3goq+qnX8UPHUsXx7yFqcUPJH2eJRbg3GNyF+6jNfybmVKXnwtm+N3NAzq2Zp/ndyPe06yZrfBSdFwbopc40RxYYCTBnSttm6UHxZvYNWmmv09W4IKjhyxfvVy1tASEIoL/XwbsekzXg3/TgVHIyGjzLHRPFoj77Sp4ZNoJxvSNusqq2nPesYHruUa/8sAXOh7NzbQd21VGE/PUVGSkEE5Sr9uxQAcs0NiKOyIwtm8FLDxJXf7H7eFPX8HQJektUeiBH3N6S52zsWmNnYAjYiXn0zqvJ2BntQ0I/mOieuoyr/HyhaajrHtc7wfMMd0YwNFBJZY/9i/Q8dzWtX1jAkfDRBbrXKvri2ZuO9jrDz4Qe49qR/eVt2gXe/EC3bYA3oMtX+TU/zveSIMuRi6O2/zXV25UVt0hrXzYMZrNkHmjevhSkdbqmaOkr9lJ0YduBPvXn4Aa4Zcm7YOwBknnkCoc2Ie1tiqkwAf3mhTuHzzb7s/9yNY/QvtA6kp4AsDXu45OdGn8/fj96j22m5mLt/6zAXJqODIEYXBDaw3NuqouMDP7aHT6VvxOF9F9lRTVVPi0Fuh+37Qble7PeLOhMPDvVPjO/3Pgou/IejJp6Os40Tv5wlaBtiB/tu/DGbn9s1hoRNd8+W9/Gn4zuRTyS2+Z+jmzLG46eg+vHHJfvx9UKI2Myr8cmo/9zoVgLv2TT+vw19YTDdHcHTZoRcA3kAhQXx8GB6QUDcg1gT2UXjvlPPMNfG34gUuweGTCD+bHjaya6fhAHwb6cO3kd2pdFLw9O9u/WYiwvARx9Nx6NmcUN1b9sVfwznvxNPcAFw5Aw6+EYZZBz47DbOaxN4u30hZ3CxGYStrrowK5ba90l/LyRBdXBig7YjRcNVsmyGh+35wdlzT27F7d3xFteST+24MTLgBPr0DXjgBHt6H7wMX87ZriedLh+3Em5fsHwui6d+9GLB+rBuOrD0E3p1ipr5QwZEjCkMlrDNFzLp1BG2a52HwsNFxNNZ57oSy7TLgHDjvvfgCW9FQ4EBSqPKBV1tfSofd8RW2pGdRhLN2tm/K30aTOTp0+vgKWDEdFn4ZK+u95mN+55nG2b4PuaPFm7asQxH9m2/A/7YzTer019J2ceVhj0IX+yZcMGVMvD+9R8CpL9koufyW7OhxQjm77wtA692H89/zB+EbfD4AL4aGxc75evgA/rfbvXDgX5nV3L7Zl5tATAgALIrEBQfA7Eg3hvZqB8c/RtmfZhLGS59OLdgquu4T3y7uBkP/kjgXp+OeiZkM9nYFNCR/R4Fm8KcfYPRi2ONEAI6vvCU1O3OLTjZDwnnvQc8DbVl+sc2p5vKV0Sl9FBgQz/UGEKpgr6IS7jlmRx5t/gRX723YpWNRbFJgMJbCGv44dEdm3HJ47LlFc91lGx2xckEkTFHFUjb42lAQ8CYs+LPwziMbsGNK1tnxILjgUxi9mMiAc+Plgy+KbUrpbxxS/j4dy+cR3uEA2l/+IRz3KJv2d5y+6xfBD88mnvfVs3kscD8AQyu/4K1L96dZ1Vp4sF+8TnRRLYC8+IDcce8jUge/XY6AP4yFXY+wb9RuU9AO+8MV02HwRQzt1Y5hR51Oj4oXuS50QazK/aET8Pl8MPx6lrdLTNf/U6QHAH8+bj9mmbjv56LjDrZv0f58Ctt0YeGdRzL+itRU/3Wi015wxutwYYarLgw4Bw652W6nW8WyzU72Wf3+cb4+dRa3XXZuap1krl0Gf3Uipn5zJhwedB2c/5FNwNmqB5z1ds3neGAvTvR/w8jQp/DWJYCdx3P47h345wl9bY4vJz9Y8zwfA3tY7WyXjkUsvPNInjvPBgIc269z+vNvJSo4csHiiTQPbWBaXqNad0qpL7r0B48HTzPrBC7Z4XBolmb521Uz8Xbux07tmkO/P1B06DXQ9xT47Sf46VVb5+pfU9sB/ToEYHLcAc/hd9g361Oedyq43qwLiqF5kvkieX/9wvh2s7Z2oqfrTf2c/Xpw14l94dhHuKrqIpaYDrEEnQfsbW3vBVJFYcDLP9rfw+ah19N1wJF0u+4HgrtYH0bbzj3S3sujp/fn3csPSHssI3Y+pOa3+2SidYM1zP3weNl/187s0aX6hcJi5DWPrzPS6zD7d7ejbAj4Db9ZjW7Hg+ImtAs+hT3TRGR95kxGXP4D/HsAgTU/89ixnenDr/CPLvDF3bGqNxzZh3dH7cUObWzU24G927HwziN54NRU82F9oMb1HLD2u5doA7y8YTf+5pR9d93BaZOMKk0YR1i0LEqKzLryJ7jfieDZ7ejEYwPPg+ljoaIEdjsmVeD4CiBUDnd2g4jLX9HbSV/TeyQc/g8YcDb0Pjw+oCU715slZZDuvh8sd8Js02QEuPmYaDbd0zmxxeH4flzGaYNsCHLeDtZctLmgEz/+5VDyfF7Aho829wLHPwIzhsejnJIYuWf1EWNZIZo9OdnhXh/87hoYdCE0SxMtdeDVVtPruAec8IQV9puWw2OOuWuzK1pt7TwYkyRMp71kv5v8lgRWzWL3bx60686c6izrbIw9R/JLQT2ggiNbGIPZsIjNK3+lzSyb/3/nbvF/iA4talkSVml6RLP2+pMmmxV3t+aL9Quhw+6Jx7oNtmamyo1QlDSg3lwCG5bA/XskCg2IL/Hr9cG+1tTBTsOoluRU9Af+H0x8uJYbsuy3U1v228kl0Fp2gWMfplnb3pAu9Ul+CysQtxXa9nIG8ZG1160rHm96oQFWeHd0RUY1b2c/xd1hg5Obq/3uKfm1YqybD+8nrcg9+x349mGrYf7yAbw5Ck590WZRqEdUcGSJmW/cSZ+f7kRMXmx1kbOG7FBzI6Vp02MotO2d4N+IcUVqDiTADi7N21vBEc2fddFXVmCAdQC7OeEpuzaKJ82AncxFX1sbfNs0GZELWtnZ8/3+UPt50rH3GVvWriEQiSe13Ba4fCqMvxomP2UH/LPethrnw4NsduG/rYVZb8dTuiTzwXX2E8UdMFBPSKYLyTRmBg4caCZPnlx7xXpkyl1HMaDMRsEsjHTgqKrb+c9FBzOwR/VZSxUlLQu/tqG4v/tr+rVN1v4KoUqrZaRbNldpnKyabbWh6EtAsNwKkKKO9uXgX7vbJXdbdoeSxenP0WFPuPir9McyQESmGGNSnLOqcWSJsC/+D3xt4c28fOFhmTnWFCWZHvvbT3WkW0pXafy456SATa4ZTbBZ0AqudyXY3LgcKjbCI4PjZYMvhn2TEnzWEyo4toKotuaexTtn5SaWL1/C70omgMCfqy7mmetPsYsCKYqiZIMWne3nuuXVrz9Tj6jgqIFZKzYy8oEvaV+Ux6pNqWkA4hgGy2xu9z9Fa9nMLlICAjcFz+Yft9yuQkNRlNyQA6EBKjhq5Kvpc7nG9xJnVn1IRV6AX01nFkfac5LPTi5aY1pQYpqxk2dFQrs5ka4sbzOYq0bdq0JDUZQmR1YFh4iMAB4AvMCTxpg7k46Lc/wIoAw4xxjzQ01tRaQ1MBboASwETjbGJK6UUk/svOxNhvls7pnmVNBWNjLYE0+N3FY20qa4JYauhPOKMb1H8kmrkzlwz53YpTGu5qcoipIBWRMcIuIFHsbO/FkKTBKRccaYma5qI4Fezmcw8CgwuJa2o4GPjTF3ishoZ9+VV7n+8JXbJG8c/xh06gdFHWzKAG8gFukQ9W5EH2TjWMdWURRly8mmxjEImGeMmQ8gIi8DxwJuwXEs8JyxXuaJIlIsIp2w2kR1bY8FDnLaPwt8RpYER//+g9m0xEORk0lUURRFyW6uqi6AezHkpU5ZJnVqatvBGLMCwPmbdj69iIwSkckiMnn16tSFZjKh2b7nU3TymC1qqyiK0lTJpuBIt2Zh8mzD6upk0rZGjDGPG2MGGmMGtmvXrvYGiqIoSkZkU3AsBdz5ELoCyzOsU1Pb3xxzFs7fVfXYZ0VRFKUWsik4JgG9RKSniASAU4FxSXXGAWeJZQhQ4pifamo7Djjb2T4bqCWxvaIoilKfZM05bowJichlwAfYkNqnjTE/i8hFzvExwHhsKO48bDjuuTW1dU59J/CKiJwPLAZOytY9KIqiKKlokkNFURQlLdUlOdQVABVFUZQ6oYJDURRFqRMqOBRFUZQ6sV34OERkNbBoC5u3BdbUY3fqC+1X3dB+1Y1ttV+w7fatKfZrB2NMykS47UJwbA0iMjmdc6ih0X7VDe1X3dhW+wXbbt+2p36pqUpRFEWpEyo4FEVRlDqhgqN2Hm/oDlSD9qtuaL/qxrbaL9h2+7bd9Et9HIqiKEqdUI1DURRFqRMqOBRFUZQ6oYKjBkRkhIjMEZF5zjK1ubpuNxH5VERmicjPInKFU36ziCwTkanO5whXm2udfs4RkayuYCsiC0XkJ6cPk52y1iLyoYjMdf62ymXfRGQX13OZKiIbReTKhnhmIvK0iKwSkRmusjo/HxEZ4DzneSLyoIikW6dma/t1t4jMFpHpIvKmiBQ75T1EpNz13Ma42uSiX3X+3nLUr7GuPi0UkalOeS6fV3XjQ+5+Y8YY/aT5YLPy/grsCASAaUCfHF27E9Df2S4CfgH6ADcD/5emfh+nf3lAT6ff3iz2byHQNqnsLmC0sz0a+GdD9M313a0EdmiIZwYcCPQHZmzN8wG+B/bFLmz2HjAyC/06DPA52/909auHu17SeXLRrzp/b7noV9Lxe4EbG+B5VTc+5Ow3phpH9cTWTDfGVAHRdc+zjjFmhTHmB2d7EzCL1GV33RwLvGyMqTTGLMCmqR+U/Z6m9OFZZ/tZ4LgG7NvBwK/GmJqyBWStX8aYL4B1aa6X8fMRu0hZC2PMt8b+hz/nalNv/TLGTDDGhJzdidhF06olV/2qgQZ9XlGcN/OTgZdqOkeW+lXd+JCz35gKjurJZM30rCMiPYC9ge+cosscs8LTLlU01301wAQRmSIio5yy6taCb4jneCqJ/9DbwjOr6/Pp4mznqn8A52HfOqP0FJEfReRzERnqlOWyX3X53nL9vIYCvxlj5rrKcv68ksaHnP3GVHBUz1ave77VHRBpDrwOXGmM2Qg8CuwE9ANWYFVlyH1f9zfG9AdGApeKyIE11M1p38SuGHkM8KpTtK08s+qorh+5fm7XAyHgBadoBdDdGLM3cBXwooi0yGG/6vq95fr7PI3El5OcP68040O1Vavpwxb3TQVH9WSyZnrWEBE/9kfxgjHmDQBjzG/GmLAxJgI8Qdy0ktO+GmOWO39XAW86/ahuLfhcP8eRwA/GmN+cPm4Tz4y6P5+lJJqNstY/ETkbOAo43TFZ4Jg11jrbU7B28d656tcWfG+5fF4+4PfAWFd/c/q80o0P5PA3poKjejJZMz0rOPbTp4BZxph/uco7uaodD0SjPcYBp4pInoj0BHphnV7Z6FszESmKbmOdqzOofi34nPXNIeFNcFt4Zq7rZfx8HFPDJhEZ4vweznK1qTdEZARwDXCMMabMVd5ORLzO9o5Ov+bnsF91+t5y1S+HQ4DZxpiYmSeXz6u68YFc/sa2xrvf1D/Y9dB/wb49XJ/D6x6AVRmnA1OdzxHAf4GfnPJxQCdXm+udfs5hK6M2aunbjtgIjWnAz9HnArQBPgbmOn9bN0DfCoG1QEtXWc6fGVZwrQCC2Le687fk+QADsQPmr8BDOJke6rlf87D27+jvbIxT9wTn+50G/AAcneN+1fl7y0W/nPL/ABcl1c3l86pufMjZb0xTjiiKoih1Qk1ViqIoSp1QwaEoiqLUCRUciqIoSp1QwaEoiqLUCRUciqIoSp1QwaFsV4hIWBKz6PZo6D7VByJyjoisFpEnnf2DROSdpDr/EZETazjH3SKyUkT+L9v9VRo3vobugKLkmHJjTL90B5xJUGLsbOXGyFhjzGVb2tgYc7WIbK7PDilNE9U4lO0asesozBKRR7ATt7qJyNUiMslJsHeLq+71znoGH4nIS9E3cxH5TEQGOtttRWShs+113uKj57rQKT/IafOa2LUwXnCEFiKyj4h8IyLTROR7ESkSkS9FpJ+rH1+LSN+tuOeBLo3rJxHRyVxKnVCNQ9neKBBn8R1gAfBnYBfgXGPMJSJyGDYlwyBsErhxThLHzdi0M3tj/29+AKbUcq3zgRJjzD4ikgd8LSITnGN7A7tjcwN9DewvIt9j8x+dYoyZ5CTJKweeBM4BrhSR3kCeMWZ6Bvc61HWvAN2Bd4wxk7HJAxGRu4H3MziXosRQwaFsbySYqhwfxyJjzESn6DDn86Oz3xwrSIqAN42Tz0lEMslbdhjQ1+VXaOmcqwqbK2ipc66p2IWASoAVxphJAMbJeCoirwJ/E5GrsanP/5PhvX5pjDnKda8J7UTkZOxCRYdleD5FAVRwKApYbSKKAP8wxjzmriAiV1J9yukQcbNvftK5/mSM+SDpXAcBla6iMPZ/UdJdwxhTJiIfYhfkORmbX2irEJHdgVuAA40x4a09n7J9oT4ORUnkA+A8sWsdICJdRKQ98AVwvIgUONmBj3a1WQgMcLZPTDrXxWJTYCMivZ2MwtUxG+gsIvs49YvEpvAGa656EJhkjMl0tby0iEhL7IqWZxljVm/NuZTtE9U4FMWFMWaCiOwGfOv4q0uBM4wxP4jIWGwm0kXAl65m9wCviMiZwCeu8iexJqgfHOf3ampYmtMYUyUipwD/FpECrH/jEKDUGDNFRDYCz9TDbR6HXY/9CeceqS7STFHSodlxFWULEJGbsQP6PTm6XmfgM2DXdOHCInIOMHBrwnGd89xMDu9LaZyoqUpRtnFE5CzsmtLX1zDHpBwYGZ0AuIXXuRs4g0Sfj6KkoBqHoiiKUidU41AURVHqhAoORVEUpU6o4FAURVHqhAoORVEUpU6o4FAURVHqxP8Dfh4tx/ukzh4AAAAASUVORK5CYII=\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "plt.title('Cough Samples in Frequency Domain')\n",
    "plt.xlabel('Frequency [Hz]')\n",
    "plt.ylabel('Magnitude')\n",
    "plt.plot(Y_pos[0:2000],label='Positive')\n",
    "plt.plot(Y_neg[0:2000],label='Negative')\n",
    "plt.legend()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "metadata": {},
   "outputs": [],
   "source": [
    "f1=450\n",
    "f2=550"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "metadata": {},
   "outputs": [],
   "source": [
    "class fft():\n",
    "    \n",
    "    def __init__(self, path):\n",
    "        self.path = path\n",
    "        \n",
    "    def __len__(self):\n",
    "        return len(os.listdir(self.path))\n",
    "\n",
    "    def __getitem__(self, idx):\n",
    "        \n",
    "        y,sr=librosa.load(self.path+os.listdir(self.path)[idx])\n",
    "        \n",
    "        y_freq=freq_domain(y,sr)\n",
    "        Y=y_freq[f1:f2]\n",
    "        #Y=torch.tensor(Y).unsqueeze(1).permute(1,0).float()\n",
    " \n",
    "        data=dict()\n",
    "        data['sgram']=Y\n",
    "        \n",
    "        if os.listdir(self.path)[idx][0]=='p':\n",
    "            data['class']= \"positive\"\n",
    "            data['label']= 1\n",
    "        else:    \n",
    "            data['class']= \"negative\"\n",
    "            data['label']= 0  \n",
    "            \n",
    "        return data"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "env: CUBLAS_WORKSPACE_CONFIG=:16:8\n"
     ]
    }
   ],
   "source": [
    "%env CUBLAS_WORKSPACE_CONFIG=:16:8\n",
    "torch.backends.cudnn.deterministic = True\n",
    "torch.set_deterministic(True)\n",
    "torch.manual_seed(19)\n",
    "torch.cuda.manual_seed(19)\n",
    "np.random.seed(19)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 20,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "93 + 23 = 116\n"
     ]
    }
   ],
   "source": [
    "#prepare dataset\n",
    "fft_dataset= fft(wav_path)\n",
    "fft_L=len(fft_dataset)\n",
    "\n",
    "#split in train, test, validation\n",
    "fft_train_set, fft_test_set = \\\n",
    "torch.utils.data.random_split(fft_dataset,[math.ceil(fft_L*0.8),int(fft_L*0.2)])\n",
    "\n",
    "print(len(fft_train_set),'+', len(fft_test_set),'=',fft_L)\n",
    "\n",
    "#dataloader\n",
    "batch_size=16\n",
    "\n",
    "fft_dataset_dl=DataLoader(fft_dataset, batch_size=batch_size, shuffle=True)\n",
    "fft_train_dl = DataLoader(fft_train_set, batch_size=batch_size, shuffle=True)\n",
    "fft_test_dl = DataLoader(fft_test_set, batch_size=1, shuffle=False)\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 21,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "torch.Size([16, 100])\n",
      "tensor([1, 1, 1, 0, 1, 1, 0, 0, 1, 1, 0, 1, 0, 0, 0, 1])\n"
     ]
    }
   ],
   "source": [
    "#view the dataloader\n",
    "fft_sample=next(iter(fft_train_dl))\n",
    "print(fft_sample['sgram'].shape) # batch_size, channels, height, width\n",
    "print(fft_sample['label'])# 1 or 0"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 22,
   "metadata": {},
   "outputs": [],
   "source": [
    "#logistic regression\n",
    "class Model2(nn.Module):\n",
    "    def __init__(self, L=f2-f1, dim_output=2):\n",
    "        super(Model2, self).__init__()\n",
    "        self.dim_output=dim_output\n",
    "        self.L=L\n",
    "\n",
    "        #fully connected\n",
    "        self.layer = nn.Linear(self.L, self.dim_output)\n",
    "        \n",
    "    def forward(self, img):\n",
    "        \n",
    "        out=self.layer(img)\n",
    "        return out            "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 23,
   "metadata": {},
   "outputs": [],
   "source": [
    "#model and optimizer\n",
    "model2 = Model2()\n",
    "model2 = model2.to(device) \n",
    "\n",
    "optimizer2 = optim.Adam(model2.parameters(), lr=0.1, weight_decay=1e-6)\n",
    "\n",
    "#print(model)\n",
    "#sample_out=model2(fft_sample['sgram'].to(device)) \n",
    "#sample_out.shape"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 24,
   "metadata": {},
   "outputs": [],
   "source": [
    "#prepare tensorboard\n",
    "train_log_dir = './models/runs_log/train'\n",
    "train_summary_writer = summary.create_file_writer(train_log_dir)\n",
    "test_log_dir = './models/runs_log/test'\n",
    "test_summary_writer = summary.create_file_writer(test_log_dir)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 25,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "efedf88f165b468f88a9b4d583c3e7f4",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "HBox(children=(FloatProgress(value=0.0, max=23.0), HTML(value='')))"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "b4825b41d372455e9c78b5a12080c11d",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "HBox(children=(FloatProgress(value=0.0, max=6.0), HTML(value='')))"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\n"
     ]
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "c801004c65d8408990c433c526ecfb5d",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "HBox(children=(FloatProgress(value=0.0, max=23.0), HTML(value='')))"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\n",
      "0.6129032258064516 0.6956521739130432\n"
     ]
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "79930f74824641f7811bc973b8304256",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "HBox(children=(FloatProgress(value=0.0, max=6.0), HTML(value='')))"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\n"
     ]
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "7e87845056254cba9b0517d8770d0f1c",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "HBox(children=(FloatProgress(value=0.0, max=23.0), HTML(value='')))"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\n",
      "0.6236559139784946 0.8260869565217388\n"
     ]
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "74e664b115bb455eb668b11169e1a75e",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "HBox(children=(FloatProgress(value=0.0, max=6.0), HTML(value='')))"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\n"
     ]
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "8458407625f746cbace6484e9f1c6220",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "HBox(children=(FloatProgress(value=0.0, max=23.0), HTML(value='')))"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\n",
      "0.7526881720430108 0.869565217391304\n"
     ]
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "b982b1c6c1874aef81fd4afdff6ed027",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "HBox(children=(FloatProgress(value=0.0, max=6.0), HTML(value='')))"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\n"
     ]
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "321e96fc9cb44dd68dd8d199998c9fdf",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "HBox(children=(FloatProgress(value=0.0, max=23.0), HTML(value='')))"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\n",
      "0.7311827956989247 0.9130434782608692\n"
     ]
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "b6c63caf41a74dd0881c46e6fec34e40",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "HBox(children=(FloatProgress(value=0.0, max=6.0), HTML(value='')))"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\n"
     ]
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "4744a0191f0c43058fdb264af177a770",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "HBox(children=(FloatProgress(value=0.0, max=23.0), HTML(value='')))"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\n",
      "0.7526881720430108 0.869565217391304\n"
     ]
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "a6b31f791b324bb8a3a4e61dbcd7593e",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "HBox(children=(FloatProgress(value=0.0, max=6.0), HTML(value='')))"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\n"
     ]
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "845a6b0607f149ce8004a3065ea1a6a9",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "HBox(children=(FloatProgress(value=0.0, max=23.0), HTML(value='')))"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\n",
      "0.7526881720430108 0.869565217391304\n"
     ]
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "1f9e562750f34218a33c6a37602d5c08",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "HBox(children=(FloatProgress(value=0.0, max=6.0), HTML(value='')))"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\n"
     ]
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "ef65faf0bfd44b84a8d0f14f11a8819f",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "HBox(children=(FloatProgress(value=0.0, max=23.0), HTML(value='')))"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\n",
      "0.7634408602150538 0.9130434782608692\n"
     ]
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "0619fd5a6f1749759bdfdf3402f06409",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "HBox(children=(FloatProgress(value=0.0, max=6.0), HTML(value='')))"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\n"
     ]
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "4de6e02c67a24d049465f86c3d28540b",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "HBox(children=(FloatProgress(value=0.0, max=23.0), HTML(value='')))"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\n",
      "0.7204301075268817 0.9565217391304344\n"
     ]
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "db31fd86f6344d1587110738685f5594",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "HBox(children=(FloatProgress(value=0.0, max=6.0), HTML(value='')))"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\n"
     ]
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "b623d4f575304fabbd00932bf9fe9266",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "HBox(children=(FloatProgress(value=0.0, max=23.0), HTML(value='')))"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\n",
      "0.7204301075268817 0.9565217391304344\n"
     ]
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "397b7d52916f4c13ae3ea37d22af6b84",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "HBox(children=(FloatProgress(value=0.0, max=6.0), HTML(value='')))"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\n"
     ]
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "74de6a57a4d242ada685b6ad272b364e",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "HBox(children=(FloatProgress(value=0.0, max=23.0), HTML(value='')))"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\n",
      "0.7204301075268817 0.9130434782608692\n"
     ]
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "de85eee0c0d641c9ae455c4c124f14ea",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "HBox(children=(FloatProgress(value=0.0, max=6.0), HTML(value='')))"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\n"
     ]
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "832784a3830945d4b3a32c907e58285c",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "HBox(children=(FloatProgress(value=0.0, max=23.0), HTML(value='')))"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\n",
      "0.7311827956989247 0.9565217391304344\n"
     ]
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "9d200f50bee94f9db0aab1f2789115f6",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "HBox(children=(FloatProgress(value=0.0, max=6.0), HTML(value='')))"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\n"
     ]
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "5b11605a2ed24f77b164bc8e12458638",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "HBox(children=(FloatProgress(value=0.0, max=23.0), HTML(value='')))"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\n",
      "0.7311827956989247 0.9565217391304344\n"
     ]
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "f95cdeeaa19744e78746b85a9adcae5d",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "HBox(children=(FloatProgress(value=0.0, max=6.0), HTML(value='')))"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\n"
     ]
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "3d53ec05790a4c0485fc224f9402d637",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "HBox(children=(FloatProgress(value=0.0, max=23.0), HTML(value='')))"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\n",
      "0.7096774193548387 0.9130434782608692\n"
     ]
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "b81e581fb4fb440da303567faca313b3",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "HBox(children=(FloatProgress(value=0.0, max=6.0), HTML(value='')))"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\n"
     ]
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "8f7e6e56d6a64a4e87f1b8e127b70508",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "HBox(children=(FloatProgress(value=0.0, max=23.0), HTML(value='')))"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\n",
      "0.7204301075268817 0.9565217391304344\n"
     ]
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "8db6c6e82c1c411daf5245106b64c774",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "HBox(children=(FloatProgress(value=0.0, max=6.0), HTML(value='')))"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\n"
     ]
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "146099240f3c41f19f6fcc5f67bd3908",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "HBox(children=(FloatProgress(value=0.0, max=23.0), HTML(value='')))"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\n",
      "0.7311827956989247 0.9565217391304344\n"
     ]
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "625f7c7a71274b88a359fdcf1820fee2",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "HBox(children=(FloatProgress(value=0.0, max=6.0), HTML(value='')))"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\n"
     ]
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "d63fb9f7a03344b69f786d1c05723a29",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "HBox(children=(FloatProgress(value=0.0, max=23.0), HTML(value='')))"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\n",
      "0.7204301075268816 0.9130434782608692\n"
     ]
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "afeab60b2368420b9903a6e01caff14d",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "HBox(children=(FloatProgress(value=0.0, max=6.0), HTML(value='')))"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\n"
     ]
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "2eb207d98829425fabb484868cb0b0a6",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "HBox(children=(FloatProgress(value=0.0, max=23.0), HTML(value='')))"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\n",
      "0.7204301075268817 0.9565217391304344\n"
     ]
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "582e9f96a87944f8bdecbf0cf63feb7e",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "HBox(children=(FloatProgress(value=0.0, max=6.0), HTML(value='')))"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\n"
     ]
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "cd34628ecc504b8f979899fadb58c06d",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "HBox(children=(FloatProgress(value=0.0, max=23.0), HTML(value='')))"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\n",
      "0.7311827956989247 0.9130434782608692\n"
     ]
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "53303d7ba5f241b0bfb38b34d417f705",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "HBox(children=(FloatProgress(value=0.0, max=6.0), HTML(value='')))"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\n"
     ]
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "9022ba42c28c44b4a9a09b1834d09e18",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "HBox(children=(FloatProgress(value=0.0, max=23.0), HTML(value='')))"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\n",
      "0.7204301075268817 0.9130434782608692\n"
     ]
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "46b595a7535d4f1a8e3f866c863468a5",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "HBox(children=(FloatProgress(value=0.0, max=6.0), HTML(value='')))"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\n"
     ]
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "51cf5dd7b4064e70a96a3228b4ad1fe8",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "HBox(children=(FloatProgress(value=0.0, max=23.0), HTML(value='')))"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\n",
      "0.7311827956989247 0.9130434782608692\n"
     ]
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "08e6e3cee0b64c7d9057971d453925f1",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "HBox(children=(FloatProgress(value=0.0, max=6.0), HTML(value='')))"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\n"
     ]
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "b74a67c8313146f394a16e8ece928677",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "HBox(children=(FloatProgress(value=0.0, max=23.0), HTML(value='')))"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\n",
      "0.7096774193548387 0.9130434782608692\n"
     ]
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "1b5279f9b7c4448497f51f2470e694d0",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "HBox(children=(FloatProgress(value=0.0, max=6.0), HTML(value='')))"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\n"
     ]
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "f0843e46b1904a9abe92e88bfa97360e",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "HBox(children=(FloatProgress(value=0.0, max=23.0), HTML(value='')))"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\n",
      "0.7311827956989247 0.9565217391304344\n"
     ]
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "f3f425cd93ab4efb9269c903ef220f5e",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "HBox(children=(FloatProgress(value=0.0, max=6.0), HTML(value='')))"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\n"
     ]
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "72f2766e04bb459d8acc209892346f90",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "HBox(children=(FloatProgress(value=0.0, max=23.0), HTML(value='')))"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\n",
      "0.7204301075268817 0.9565217391304344\n",
      "\n"
     ]
    }
   ],
   "source": [
    "##################################start training and validating#####################################################\n",
    "max_epoch = 23\n",
    "tmp_path = './models/temp.pth'\n",
    "\n",
    "for epoch in tqdm(range(max_epoch)):        \n",
    "    ###Train Phase\n",
    "    \n",
    "    # Initialize Loss and Accuracy\n",
    "    train_loss = 0.0\n",
    "    train_accu = 0.0\n",
    "\n",
    "    #########TEMPORARY################################\n",
    "    # Load the saved MODEL AND OPTIMIZER after evaluation.\n",
    "    if epoch > 0:\n",
    "        checkpoint = torch.load(tmp_path)\n",
    "        model2.load_state_dict(checkpoint['model2_state_dict'])\n",
    "        optimizer2.load_state_dict(checkpoint['optimizer2_state_dict'])\n",
    "       \n",
    "    ##################################################\n",
    "\n",
    "    # Iterate over the train_dataloader\n",
    "    with tqdm(total=len(fft_train_dl)) as pbar:\n",
    "        for sample in fft_train_dl:            \n",
    "            curr_loss, num_correct = train(model2, optimizer2, sample)\n",
    "            train_loss += curr_loss / len(fft_train_dl) \n",
    "            train_accu += num_correct / len(fft_train_set)\n",
    "            pbar.update(1)\n",
    "    \n",
    "    \n",
    "    ### Write the current loss and accuracy to the Tensorboard\n",
    "    with train_summary_writer.as_default():\n",
    "        tf.summary.scalar('loss', train_loss, step=epoch)                \n",
    "        tf.summary.scalar('accuracy', train_accu, step=epoch)         \n",
    "         \n",
    "    ###############TEMPORARY#######################################\n",
    "    # save the model and optimizer's information before the evaulation\n",
    "    checkpoint = {\n",
    "        'model2_state_dict': model2.state_dict(),\n",
    "        'optimizer2_state_dict': optimizer2.state_dict(),\n",
    "        \n",
    "    }\n",
    "    torch.save(checkpoint, tmp_path)\n",
    "    ###################################################################\n",
    "    \n",
    "    ### Validate Phase\n",
    "    # Initialize Loss and Accuracy\n",
    "    test_loss = 0.0\n",
    "    test_accu = 0.0\n",
    "\n",
    "    # Iterate over the test_dataloader\n",
    "    with tqdm(total=len(fft_test_dl)) as pbar:\n",
    "        for idx, sample in enumerate(fft_test_dl): \n",
    "            curr_loss, num_correct = test(model2, sample)\n",
    "            test_loss += curr_loss / len(fft_test_dl)\n",
    "            test_accu += num_correct / len(fft_test_set)\n",
    "            pbar.update(1)\n",
    "    \n",
    "    ### Write the current loss and accuracy to the Tensorboard\n",
    "    with test_summary_writer.as_default():\n",
    "        tf.summary.scalar('loss', test_loss, step=epoch)\n",
    "        tf.summary.scalar('accuracy', test_accu, step=epoch)    \n",
    "        \n",
    "    print(train_accu, test_accu)     \n",
    "\n",
    "os.remove(tmp_path)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 26,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/opt/conda/lib/python3.7/site-packages/ipykernel_launcher.py:4: UserWarning: To copy construct from a tensor, it is recommended to use sourceTensor.clone().detach() or sourceTensor.clone().detach().requires_grad_(True), rather than torch.tensor(sourceTensor).\n",
      "  after removing the cwd from sys.path.\n",
      "/opt/conda/lib/python3.7/site-packages/ipykernel_launcher.py:5: UserWarning: To copy construct from a tensor, it is recommended to use sourceTensor.clone().detach() or sourceTensor.clone().detach().requires_grad_(True), rather than torch.tensor(sourceTensor).\n",
      "  \"\"\"\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Accuracy = 0.6956521739130435\n",
      "Sensitivity/ TPR = 0.42857142857142855\n",
      "Specifity/ TNR = 0.8125\n",
      "Balanced Accuracy = 0.6205357142857143\n",
      "tp: 3  tn: 13  fp: 3  fn: 4\n"
     ]
    }
   ],
   "source": [
    "####Evaluation for model 1\n",
    "accuracy=tp=tn=fn=fp=p=n=0\n",
    "for i in test_set:\n",
    "    x=model(torch.tensor(i['sgram']).float().to(device).unsqueeze(1))[0][0].item()\n",
    "    y=model(torch.tensor(i['sgram']).float().to(device).unsqueeze(1))[0][1].item()\n",
    "    \n",
    "    if x>y:\n",
    "        pred=0\n",
    "    else:\n",
    "        pred=1\n",
    "    \n",
    "    if i['label']== 1:\n",
    "        p+=1\n",
    "    else:\n",
    "        n+=1\n",
    "        \n",
    "    if i['label']==pred:  \n",
    "        accuracy += 1\n",
    "        if pred == 1:\n",
    "            tp += 1\n",
    "        else:\n",
    "            tn += 1    \n",
    "    else:\n",
    "        if pred == 1:\n",
    "            fp += 1\n",
    "        else:\n",
    "            fn += 1\n",
    "        \n",
    "accuracy /= len(fft_test_set)\n",
    "\n",
    "print('Accuracy =', accuracy)\n",
    "print('Sensitivity/ TPR =', tp/p)\n",
    "print('Specifity/ TNR =', tn/n)\n",
    "print('Balanced Accuracy =', (tp/p + tn/n)/2)\n",
    "\n",
    "print('tp:',tp, ' tn:',tn, ' fp:',fp, ' fn:', fn)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 27,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Accuracy = 0.9565217391304348\n",
      "Sensitivity = 0.8571428571428571\n",
      "Specifity = 1.0\n",
      "Balanced Accuracy = 0.9285714285714286\n",
      "tp: 6  tn: 16  fp: 0  fn: 1\n"
     ]
    }
   ],
   "source": [
    "####Evaluation for model 2\n",
    "accuracy=tp=tn=fn=fp=p=n=0\n",
    "for i in fft_test_set:\n",
    "    x=model2(torch.tensor(i['sgram']).float().to(device))[0].item()\n",
    "    y=model2(torch.tensor(i['sgram']).float().to(device))[1].item()\n",
    "    \n",
    "    if x>y:\n",
    "        pred=0\n",
    "    else:\n",
    "        pred=1\n",
    "    \n",
    "    if i['label']== 1:\n",
    "        p+=1\n",
    "    else:\n",
    "        n+=1\n",
    "        \n",
    "    if i['label']==pred:  \n",
    "        accuracy += 1\n",
    "        if pred == 1:\n",
    "            tp += 1\n",
    "        else:\n",
    "            tn += 1    \n",
    "    else:\n",
    "        if pred == 1:\n",
    "            fp += 1\n",
    "        else:\n",
    "            fn += 1\n",
    "        \n",
    "accuracy /= len(fft_test_set)\n",
    "\n",
    "print('Accuracy =', accuracy)\n",
    "print('Sensitivity =', tp/p)\n",
    "print('Specifity =', tn/n)\n",
    "print('Balanced Accuracy =', (tp/p + tn/n)/2)\n",
    "\n",
    "print('tp:',tp, ' tn:',tn, ' fp:',fp, ' fn:', fn)"
   ]
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 3",
   "language": "python",
   "name": "python3"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.7.7"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 4
}