UATTA-ENS.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "id": "6695d9ee-4a3e-4538-a691-7e30d4ea3ef6",
   "metadata": {},
   "source": [
    "# UATTA-ENS: Uncertainty Aware Test Time Augmented Ensemble for PIRC Diabetic Retinopathy Detection"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "id": "8b6635ba",
   "metadata": {
    "papermill": {
     "duration": 2.447311,
     "end_time": "2022-08-28T09:23:48.237326",
     "exception": false,
     "start_time": "2022-08-28T09:23:45.790015",
     "status": "completed"
    },
    "tags": []
   },
   "outputs": [],
   "source": [
    "import pandas as pd\n",
    "import numpy as np \n",
    "import matplotlib.pyplot as plt\n",
    "import PIL.Image as Image\n",
    "import torch\n",
    "import torch.nn as nn\n",
    "import torch.nn.functional as F\n",
    "from torch.utils.data import Dataset,DataLoader\n",
    "from torchvision import models,transforms\n",
    "from sklearn.metrics import accuracy_score\n",
    "from sklearn.metrics import roc_auc_score\n",
    "from sklearn.metrics import f1_score\n",
    "from sklearn.metrics import classification_report"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "id": "c8bc1f9b",
   "metadata": {
    "papermill": {
     "duration": 0.068119,
     "end_time": "2022-08-28T09:23:48.310531",
     "exception": false,
     "start_time": "2022-08-28T09:23:48.242412",
     "status": "completed"
    },
    "tags": []
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "cuda\n"
     ]
    }
   ],
   "source": [
    "device = torch.device('cuda' if torch.cuda.is_available() else 'cpu') #Use GPU if it's available or else use CPU.\n",
    "print(device)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "id": "a6a08019",
   "metadata": {
    "papermill": {
     "duration": 0.030112,
     "end_time": "2022-08-28T09:23:48.345668",
     "exception": false,
     "start_time": "2022-08-28T09:23:48.315556",
     "status": "completed"
    },
    "tags": []
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "No.of.training_samples: 3662\n"
     ]
    }
   ],
   "source": [
    "path = \"./\"\n",
    "train_df = pd.read_csv(f\"{path}train.csv\")\n",
    "print(f'No.of.training_samples: {len(train_df)}')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "id": "4c6da8df",
   "metadata": {
    "papermill": {
     "duration": 0.2341,
     "end_time": "2022-08-28T09:23:48.585161",
     "exception": false,
     "start_time": "2022-08-28T09:23:48.351061",
     "status": "completed"
    },
    "tags": []
   },
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "#Histogram of label counts.\n",
    "train_df.diagnosis.hist()\n",
    "plt.xticks([0,1,2,3,4])\n",
    "plt.grid(False)\n",
    "plt.show() "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "id": "72bd4db9",
   "metadata": {
    "papermill": {
     "duration": 2.959387,
     "end_time": "2022-08-28T09:23:51.550618",
     "exception": false,
     "start_time": "2022-08-28T09:23:48.591231",
     "status": "completed"
    },
    "tags": []
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "tensor([0.4058, 1.9795, 0.7331, 3.7948, 2.4827], device='cuda:0')\n"
     ]
    }
   ],
   "source": [
    "from sklearn.utils import class_weight\n",
    "class_weights = class_weight.compute_class_weight(class_weight='balanced',classes=np.array([0,1,2,3,4]),y=train_df['diagnosis'].values)\n",
    "class_weights = torch.tensor(class_weights,dtype=torch.float).to(device)\n",
    "print(class_weights) "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "id": "9bdde476",
   "metadata": {
    "papermill": {
     "duration": 0.240826,
     "end_time": "2022-08-28T09:23:51.796712",
     "exception": false,
     "start_time": "2022-08-28T09:23:51.555886",
     "status": "completed"
    },
    "tags": []
   },
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "path = \"./aptos-dataset/\"\n",
    "num = int(np.random.randint(0,len(train_df)-1,(1,))) #Picks a random number.\n",
    "sample_image = (f'{path}train_images/{train_df[\"id_code\"][num]}.png')#Image file.\n",
    "sample_image = Image.open(sample_image) \n",
    "plt.imshow(sample_image)\n",
    "plt.axis('off')\n",
    "plt.title(f'Class: {train_df[\"diagnosis\"][num]}') #Class of the random image.\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "id": "c4b3dc6f",
   "metadata": {
    "papermill": {
     "duration": 0.013526,
     "end_time": "2022-08-28T09:23:51.816774",
     "exception": false,
     "start_time": "2022-08-28T09:23:51.803248",
     "status": "completed"
    },
    "tags": []
   },
   "outputs": [],
   "source": [
    "def one_hot(x):\n",
    "    return torch.eye(5)[x,:]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "id": "20fc0daa",
   "metadata": {
    "papermill": {
     "duration": 0.015528,
     "end_time": "2022-08-28T09:23:51.838342",
     "exception": false,
     "start_time": "2022-08-28T09:23:51.822814",
     "status": "completed"
    },
    "tags": []
   },
   "outputs": [],
   "source": [
    "class dataset(Dataset):\n",
    "    def __init__(self,df,data_path,image_transform=None,train=True): # Constructor.\n",
    "        super(Dataset,self).__init__() #Calls the constructor of the Dataset class.\n",
    "        self.df = df\n",
    "        self.data_path = data_path\n",
    "        self.image_transform = image_transform\n",
    "        self.train = train\n",
    "        \n",
    "    def __len__(self):\n",
    "        return len(self.df) #Returns the number of samples in the dataset.\n",
    "    \n",
    "    def __getitem__(self,index):\n",
    "        image_id = self.df['id_code'][index]\n",
    "        image = Image.open(f'{self.data_path}/{image_id}.png') #Image.\n",
    "        if self.image_transform :\n",
    "            image = self.image_transform(image) #Applies transformation to the image.\n",
    "        \n",
    "        if self.train :\n",
    "            label = one_hot(self.df['diagnosis'][index]) #Label.\n",
    "            return image,label #If train == True, return image & label.\n",
    "        \n",
    "        else:\n",
    "            return image #If train != True, return image.\n",
    "            "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "id": "04ce0e6e",
   "metadata": {
    "papermill": {
     "duration": 0.016135,
     "end_time": "2022-08-28T09:23:51.860479",
     "exception": false,
     "start_time": "2022-08-28T09:23:51.844344",
     "status": "completed"
    },
    "tags": []
   },
   "outputs": [],
   "source": [
    "path= \"./aptos-dataset/\"\n",
    "image_transform = transforms.Compose([transforms.Resize([512,512]),\n",
    "                                      transforms.ToTensor(),\n",
    "                                      transforms.Normalize((0.485, 0.456, 0.406), (0.229, 0.224, 0.225))]) #Transformations to apply to the image.\n",
    "data_set = dataset(train_df,f'{path}train_images',image_transform=image_transform)\n",
    "\n",
    "\n",
    "train_set,valid_set = torch.utils.data.random_split(data_set,[3302,360],generator=torch.Generator().manual_seed(989))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "id": "8b224097",
   "metadata": {
    "papermill": {
     "duration": 0.014881,
     "end_time": "2022-08-28T09:23:51.881442",
     "exception": false,
     "start_time": "2022-08-28T09:23:51.866561",
     "status": "completed"
    },
    "tags": []
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "<torch._C.Generator object at 0x7f7f8a8c09f0>\n"
     ]
    }
   ],
   "source": [
    "print(torch.manual_seed(40))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "id": "84bc7f6a",
   "metadata": {
    "papermill": {
     "duration": 0.013601,
     "end_time": "2022-08-28T09:23:51.901616",
     "exception": false,
     "start_time": "2022-08-28T09:23:51.888015",
     "status": "completed"
    },
    "tags": []
   },
   "outputs": [],
   "source": [
    "train_dataloader = DataLoader(train_set,batch_size=4,shuffle=True) #DataLoader for train_set.\n",
    "valid_dataloader = DataLoader(valid_set,batch_size=4,shuffle=False) #DataLoader for validation_set."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "id": "d4df13de-0a65-417b-bd9f-fb9b7f53001c",
   "metadata": {},
   "outputs": [],
   "source": [
    "def calc_bins(preds,labels_oneh):\n",
    "    # Assign each prediction to a bin\n",
    "    num_bins = 5\n",
    "    bins = torch.linspace(0.1, 1, num_bins)\n",
    "    bins = bins.to(device)\n",
    "    #print(preds)\n",
    "    #print(bins)\n",
    "    binned = torch.bucketize(preds, bins)\n",
    "\n",
    "    # Save the accuracy, confidence and size of each bin\n",
    "    bin_accs = torch.zeros(num_bins)\n",
    "    bin_confs = torch.zeros(num_bins)\n",
    "    bin_sizes = torch.zeros(num_bins)\n",
    "\n",
    "    for bin in range(num_bins):\n",
    "        bin_sizes[bin] = len(preds[binned == bin])\n",
    "        if bin_sizes[bin] > 0:\n",
    "            bin_accs[bin] = (labels_oneh[binned==bin]).sum() / bin_sizes[bin]\n",
    "            bin_confs[bin] = (preds[binned==bin]).sum() / bin_sizes[bin]\n",
    "\n",
    "    return bins, binned, bin_accs, bin_confs, bin_sizes"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "id": "44a36354-4248-4473-b897-78a1d67d2bcc",
   "metadata": {},
   "outputs": [],
   "source": [
    "def get_metrics(preds,labels):\n",
    "    ECE = 0\n",
    "    MCE = 0\n",
    "    bins, _, bin_accs, bin_confs, bin_sizes = calc_bins(preds,labels)\n",
    "\n",
    "    for i in range(len(bins)):\n",
    "        abs_conf_dif = abs(bin_accs[i] - bin_confs[i])\n",
    "        ECE += (bin_sizes[i] / sum(bin_sizes)) * abs_conf_dif\n",
    "        MCE = max(MCE, abs_conf_dif)\n",
    "\n",
    "    return ECE, MCE"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "id": "d0b549ee",
   "metadata": {
    "papermill": {
     "duration": 0.01991,
     "end_time": "2022-08-28T09:23:51.928477",
     "exception": false,
     "start_time": "2022-08-28T09:23:51.908567",
     "status": "completed"
    },
    "tags": []
   },
   "outputs": [],
   "source": [
    "from sklearn.metrics import f1_score, roc_auc_score, cohen_kappa_score\n",
    "from sklearn.metrics import brier_score_loss\n",
    "\n",
    "def validate():\n",
    "    model.eval()\n",
    "    total = 0\n",
    "    correct = 0\n",
    "    running_loss = 0\n",
    "    pred_lst = []\n",
    "    gt_lst = []\n",
    "    with torch.no_grad():\n",
    "        for x,y in valid_dataloader:\n",
    "            x= x.to(device)\n",
    "            output= model(x) #model's output.\n",
    "            output=output.detach().cpu().numpy()\n",
    "            pred_lst.append(output)\n",
    "            gt_lst.append(y)\n",
    "            del x, y, output\n",
    "        pred_lst=np.concatenate(pred_lst, axis=0)\n",
    "        gt_lst=np.concatenate(gt_lst, axis=0)\n",
    "        pred=torch.from_numpy(pred_lst)\n",
    "        pred=pred.to(device)\n",
    "        labels=torch.from_numpy(gt_lst)\n",
    "        labels=labels.to(device)\n",
    "        ece,mce=get_metrics(pred,labels)\n",
    "        print(\"ECE\",ece,\"MCE\",mce)\n",
    "        pred_lst2 = np.argmax(pred_lst, axis=1)\n",
    "        gt_lst2 = np.argmax(gt_lst, axis=1)\n",
    "        auc_lst=[]\n",
    "        for k in range(0,5):\n",
    "            tmp_gt=gt_lst[:, k]\n",
    "            tmp_pred=pred_lst[:,k]\n",
    "            try:\n",
    "                tmp_auc=roc_auc_score(tmp_gt, tmp_pred)\n",
    "            except ValueError:\n",
    "                print(\"error\")\n",
    "                tmp_auc=0\n",
    "            auc_lst.append(tmp_auc)\n",
    "\n",
    "        auc_lst=np.array(auc_lst)\n",
    "        brloss=0\n",
    "        for i in range(360):\n",
    "            brloss+= brier_score_loss(gt_lst[i], pred_lst[i],sample_weight=None, pos_label=None)\n",
    "            #print(br_loss)\n",
    "        brloss=brloss/360\n",
    "        print(\"Brier Loss - \",brloss)\n",
    "        print('AUC Classwise Score on Validation set = ', auc_lst)\n",
    "        auc=np.mean(auc_lst)\n",
    "        acc =0\n",
    "        print('AUC Score on Validation set = ', auc)\n",
    "        f1 = f1_score(gt_lst2, pred_lst2, average='weighted')\n",
    "        print('F1 Score on Validation set = ', f1)\n",
    "        cohen_kappa = cohen_kappa_score(gt_lst2, pred_lst2,weights=\"quadratic\")\n",
    "        print(\"Cohen's Kappa Score on Validation set = \", cohen_kappa)\n",
    "        print(classification_report(gt_lst2, pred_lst2, labels=[0,1, 2, 3,4]))\n",
    "        model.train()\n",
    "        return auc,acc,f1,cohen_kappa,ece,mce,brloss"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "id": "bcc40d45",
   "metadata": {
    "papermill": {
     "duration": 0.01834,
     "end_time": "2022-08-28T09:23:51.952564",
     "exception": false,
     "start_time": "2022-08-28T09:23:51.934224",
     "status": "completed"
    },
    "tags": []
   },
   "outputs": [],
   "source": [
    "def train():\n",
    "    total_batches=len(train_dataloader)\n",
    "    cycle_length= len(train_dataloader)*2\n",
    "    loss_fn   = nn.CrossEntropyLoss(weight=class_weights) #CrossEntropyLoss\n",
    "    optimizer = torch.optim.SGD(model.parameters(),lr=0.0001) \n",
    "    model.train()\n",
    "    total = 0\n",
    "    correct = 0\n",
    "    running_loss=0\n",
    "    count=0\n",
    "    max_epochs=50 #18\n",
    "    \n",
    "    \n",
    "    for epoch in range(0, max_epochs):\n",
    "        for batch,(x,y) in enumerate(train_dataloader):\n",
    "            output = model(x.to(device)) #model's predictions.\n",
    "            y=y.long()\n",
    "            y2 = torch.max(y.to(device), 1)[1]\n",
    "            loss   = loss_fn(output,y2)\n",
    "            running_loss=running_loss+loss.item()\n",
    "            optimizer.zero_grad()\n",
    "            loss.backward()\n",
    "            optimizer.step()\n",
    "            if (batch+1) % 25 == 0: # displays after every 20 batches\n",
    "                    print (\"Epoch [{}/{}], Batch [{}/{}] Classification Loss: {:.4f}\".format(epoch+1, max_epochs, batch+1, total_batches, running_loss/(batch+1) ), end =\"\\r\")\n",
    "                    \n",
    "            count=count+1\n",
    "            if count>=cycle_length:\n",
    "                count=0\n",
    "                auc,acc,f1,cohen_kappa,e1,e2,e3=validate()\n",
    "                auc=np.round(auc,4)\n",
    "                nm='DR_CNN_Model_Ens_PE_tta_'+str(e1)+'_'+str(e2)+'_'+str(e3)+'_'+str(f1)+'_'+str(auc)+'_'+str(cohen_kappa)+'.pt'\n",
    "                torch.save({'model_state_dict': model.state_dict(),'optimizer_state_dict': optimizer.state_dict()}, nm)\n",
    "                del nm\n",
    "                running_loss=0\n",
    "                \n",
    "    return model"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "id": "dffd7451",
   "metadata": {
    "papermill": {
     "duration": 0.014813,
     "end_time": "2022-08-28T09:23:51.973319",
     "exception": false,
     "start_time": "2022-08-28T09:23:51.958506",
     "status": "completed"
    },
    "tags": []
   },
   "outputs": [],
   "source": [
    "from scipy.stats import bernoulli, uniform"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "id": "1d1d5a0f",
   "metadata": {
    "papermill": {
     "duration": 0.013822,
     "end_time": "2022-08-28T09:23:51.992972",
     "exception": false,
     "start_time": "2022-08-28T09:23:51.979150",
     "status": "completed"
    },
    "tags": []
   },
   "outputs": [],
   "source": [
    "import sys"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "id": "60ad32fa",
   "metadata": {
    "papermill": {
     "duration": 0.022247,
     "end_time": "2022-08-28T09:23:52.021191",
     "exception": false,
     "start_time": "2022-08-28T09:23:51.998944",
     "status": "completed"
    },
    "tags": []
   },
   "outputs": [],
   "source": [
    "import torch\n",
    "import torch.nn as nn\n",
    "\n",
    "class ENS_MOD(nn.Module):\n",
    "    def __init__(self, model1,model2,model3,model4):\n",
    "        super(ENS_MOD, self).__init__()\n",
    "        self.features1 = model1\n",
    "        self.features2 = model2\n",
    "        self.features3 = model3\n",
    "        self.features4 = model4\n",
    "        self.transform1 = transforms.ColorJitter(brightness=0.15,contrast=(0.5, 1.5),saturation=(0.5, 2.5),hue=0.15)\n",
    "        self.transform2 = transforms.RandomResizedCrop(512, scale=(0.8, 1.0), ratio=(0.75, 1.3333333333333333))\n",
    "        self.transform3 = transforms.RandomHorizontalFlip(p=bernoulli.rvs(0.5, size=1)[0])\n",
    "        self.transform4 = transforms.RandomVerticalFlip(p=bernoulli.rvs(0.5, size=1)[0]) \n",
    "    \n",
    "    def predictive_entropy(self,predictions):\n",
    "        epsilon = 1e-10\n",
    "        idx = torch.argmax(predictions, dim=1, keepdims=True)\n",
    "        x = torch.zeros_like(predictions).scatter_(2, idx, 1.)\n",
    "        mean = torch.mode(x, axis=2) \n",
    "        mean = mean[0]\n",
    "        std = torch.std(x, axis=2) \n",
    "        output_wt = 0\n",
    "        output_uq = 0\n",
    "        for i in range(4):\n",
    "            counter = 0.5*(torch.log10(2*torch.pi*std+epsilon))\n",
    "            predictive_entropy = torch.clamp(counter,min=0,max=None)+(torch.pow((predictions[:,:,i]-mean),2)/2*std)+epsilon\n",
    "            output_wt = output_wt + (1/(predictive_entropy+epsilon))*predictions[:,:,i]\n",
    "            output_uq = output_uq + (1/(predictive_entropy+epsilon))\n",
    "        output = output_wt/output_uq\n",
    "\n",
    "        return output\n",
    "    def forward_feature(self, x):\n",
    "        x1 = self.transform1(x)\n",
    "        x2 = self.transform2(x)\n",
    "        x3 = self.transform3(x)\n",
    "        x4 = self.transform4(x)\n",
    "        x1 = torch.unsqueeze(self.features1(x1),-1)\n",
    "        x2 = torch.unsqueeze(self.features2(x2),-1)\n",
    "        x3 = torch.unsqueeze(self.features3(x3),-1)\n",
    "        x4 = torch.unsqueeze(self.features4(x4),-1)\n",
    "        x= torch.cat((x1,x2,x3,x4),2)\n",
    "        output = self.predictive_entropy(x)\n",
    "        #output=self.LL_FU_Ensmble(x)\n",
    "        return output\n",
    "    \n",
    "    def forward(self, x):\n",
    "        out = self.forward_feature(x)\n",
    "        return out\n",
    "\n",
    "    def get_config_optim(self, lr):\n",
    "        return [{'params': self.parameters(), 'lr': lr}]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "id": "07e3cd46",
   "metadata": {
    "papermill": {
     "duration": 0.019073,
     "end_time": "2022-08-28T09:23:52.047348",
     "exception": false,
     "start_time": "2022-08-28T09:23:52.028275",
     "status": "completed"
    },
    "tags": []
   },
   "outputs": [],
   "source": [
    "def get_model():\n",
    "    model_dict = {'ENS_MOD': ENS_MOD}\n",
    "    backbone1=models.resnet50()\n",
    "    backbone2=models.resnet50()\n",
    "    backbone3=models.resnet50()\n",
    "    backbone4=models.resnet50()\n",
    "    backbone1.fc=nn.Sequential(nn.Linear(2048,5),nn.Sigmoid())\n",
    "    backbone2.fc=nn.Sequential(nn.Linear(2048,5),nn.Sigmoid())\n",
    "    backbone3.fc=nn.Sequential(nn.Linear(2048,5),nn.Sigmoid())\n",
    "    backbone4.fc=nn.Sequential(nn.Linear(2048,5),nn.Sigmoid())\n",
    "    checkpoint = torch.load('./RESNET50/DR_CNN_Model_1.pt')\n",
    "    backbone1.load_state_dict(checkpoint['model_state_dict'],strict=False)\n",
    "    del checkpoint\n",
    "    checkpoint = torch.load('./RESNET50/DR_CNN_Model_2.pt')\n",
    "    backbone2.load_state_dict(checkpoint['model_state_dict'],strict=False)\n",
    "    del checkpoint\n",
    "    checkpoint = torch.load('./RESNET50/DR_CNN_Model_3.pt')\n",
    "    backbone3.load_state_dict(checkpoint['model_state_dict'],strict=False)\n",
    "    del checkpoint\n",
    "    checkpoint = torch.load('./RESNET50/DR_CNN_Model_4.pt')\n",
    "    backbone4.load_state_dict(checkpoint['model_state_dict'],strict=False)\n",
    "    del checkpoint\n",
    "    model = model_dict['ENS_MOD'](backbone1,backbone2,backbone3,backbone4)\n",
    "    checkpoint = torch.load('Untitled_Folder/DR_CNN_Model_Ens_PE_tta_0.783_0.659245027124774.pt')\n",
    "    model.load_state_dict(checkpoint['model_state_dict'],strict=False)\n",
    "    del checkpoint\n",
    "    return model"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 20,
   "id": "cf153a08",
   "metadata": {
    "papermill": {
     "duration": 5.6504,
     "end_time": "2022-08-28T09:23:57.703429",
     "exception": false,
     "start_time": "2022-08-28T09:23:52.053029",
     "status": "completed"
    },
    "scrolled": true,
    "tags": []
   },
   "outputs": [],
   "source": [
    "model = get_model()\n",
    "model = model.to(device)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 21,
   "id": "0ce101ce",
   "metadata": {
    "papermill": {
     "duration": 0.021408,
     "end_time": "2022-08-28T09:23:57.731951",
     "exception": false,
     "start_time": "2022-08-28T09:23:57.710543",
     "status": "completed"
    },
    "scrolled": true,
    "tags": []
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "ENS_MOD(\n",
      "  (features1): ResNet(\n",
      "    (conv1): Conv2d(3, 64, kernel_size=(7, 7), stride=(2, 2), padding=(3, 3), bias=False)\n",
      "    (bn1): BatchNorm2d(64, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
      "    (relu): ReLU(inplace=True)\n",
      "    (maxpool): MaxPool2d(kernel_size=3, stride=2, padding=1, dilation=1, ceil_mode=False)\n",
      "    (layer1): Sequential(\n",
      "      (0): Bottleneck(\n",
      "        (conv1): Conv2d(64, 64, kernel_size=(1, 1), stride=(1, 1), bias=False)\n",
      "        (bn1): BatchNorm2d(64, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
      "        (conv2): Conv2d(64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)\n",
      "        (bn2): BatchNorm2d(64, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
      "        (conv3): Conv2d(64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False)\n",
      "        (bn3): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
      "        (relu): ReLU(inplace=True)\n",
      "        (downsample): Sequential(\n",
      "          (0): Conv2d(64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False)\n",
      "          (1): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
      "        )\n",
      "      )\n",
      "      (1): Bottleneck(\n",
      "        (conv1): Conv2d(256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False)\n",
      "        (bn1): BatchNorm2d(64, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
      "        (conv2): Conv2d(64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)\n",
      "        (bn2): BatchNorm2d(64, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
      "        (conv3): Conv2d(64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False)\n",
      "        (bn3): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
      "        (relu): ReLU(inplace=True)\n",
      "      )\n",
      "      (2): Bottleneck(\n",
      "        (conv1): Conv2d(256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False)\n",
      "        (bn1): BatchNorm2d(64, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
      "        (conv2): Conv2d(64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)\n",
      "        (bn2): BatchNorm2d(64, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
      "        (conv3): Conv2d(64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False)\n",
      "        (bn3): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
      "        (relu): ReLU(inplace=True)\n",
      "      )\n",
      "    )\n",
      "    (layer2): Sequential(\n",
      "      (0): Bottleneck(\n",
      "        (conv1): Conv2d(256, 128, kernel_size=(1, 1), stride=(1, 1), bias=False)\n",
      "        (bn1): BatchNorm2d(128, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
      "        (conv2): Conv2d(128, 128, kernel_size=(3, 3), stride=(2, 2), padding=(1, 1), bias=False)\n",
      "        (bn2): BatchNorm2d(128, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
      "        (conv3): Conv2d(128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False)\n",
      "        (bn3): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
      "        (relu): ReLU(inplace=True)\n",
      "        (downsample): Sequential(\n",
      "          (0): Conv2d(256, 512, kernel_size=(1, 1), stride=(2, 2), bias=False)\n",
      "          (1): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
      "        )\n",
      "      )\n",
      "      (1): Bottleneck(\n",
      "        (conv1): Conv2d(512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False)\n",
      "        (bn1): BatchNorm2d(128, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
      "        (conv2): Conv2d(128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)\n",
      "        (bn2): BatchNorm2d(128, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
      "        (conv3): Conv2d(128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False)\n",
      "        (bn3): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
      "        (relu): ReLU(inplace=True)\n",
      "      )\n",
      "      (2): Bottleneck(\n",
      "        (conv1): Conv2d(512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False)\n",
      "        (bn1): BatchNorm2d(128, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
      "        (conv2): Conv2d(128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)\n",
      "        (bn2): BatchNorm2d(128, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
      "        (conv3): Conv2d(128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False)\n",
      "        (bn3): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
      "        (relu): ReLU(inplace=True)\n",
      "      )\n",
      "      (3): Bottleneck(\n",
      "        (conv1): Conv2d(512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False)\n",
      "        (bn1): BatchNorm2d(128, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
      "        (conv2): Conv2d(128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)\n",
      "        (bn2): BatchNorm2d(128, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
      "        (conv3): Conv2d(128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False)\n",
      "        (bn3): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
      "        (relu): ReLU(inplace=True)\n",
      "      )\n",
      "    )\n",
      "    (layer3): Sequential(\n",
      "      (0): Bottleneck(\n",
      "        (conv1): Conv2d(512, 256, kernel_size=(1, 1), stride=(1, 1), bias=False)\n",
      "        (bn1): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
      "        (conv2): Conv2d(256, 256, kernel_size=(3, 3), stride=(2, 2), padding=(1, 1), bias=False)\n",
      "        (bn2): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
      "        (conv3): Conv2d(256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False)\n",
      "        (bn3): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
      "        (relu): ReLU(inplace=True)\n",
      "        (downsample): Sequential(\n",
      "          (0): Conv2d(512, 1024, kernel_size=(1, 1), stride=(2, 2), bias=False)\n",
      "          (1): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
      "        )\n",
      "      )\n",
      "      (1): Bottleneck(\n",
      "        (conv1): Conv2d(1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False)\n",
      "        (bn1): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
      "        (conv2): Conv2d(256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)\n",
      "        (bn2): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
      "        (conv3): Conv2d(256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False)\n",
      "        (bn3): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
      "        (relu): ReLU(inplace=True)\n",
      "      )\n",
      "      (2): Bottleneck(\n",
      "        (conv1): Conv2d(1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False)\n",
      "        (bn1): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
      "        (conv2): Conv2d(256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)\n",
      "        (bn2): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
      "        (conv3): Conv2d(256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False)\n",
      "        (bn3): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
      "        (relu): ReLU(inplace=True)\n",
      "      )\n",
      "      (3): Bottleneck(\n",
      "        (conv1): Conv2d(1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False)\n",
      "        (bn1): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
      "        (conv2): Conv2d(256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)\n",
      "        (bn2): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
      "        (conv3): Conv2d(256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False)\n",
      "        (bn3): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
      "        (relu): ReLU(inplace=True)\n",
      "      )\n",
      "      (4): Bottleneck(\n",
      "        (conv1): Conv2d(1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False)\n",
      "        (bn1): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
      "        (conv2): Conv2d(256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)\n",
      "        (bn2): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
      "        (conv3): Conv2d(256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False)\n",
      "        (bn3): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
      "        (relu): ReLU(inplace=True)\n",
      "      )\n",
      "      (5): Bottleneck(\n",
      "        (conv1): Conv2d(1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False)\n",
      "        (bn1): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
      "        (conv2): Conv2d(256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)\n",
      "        (bn2): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
      "        (conv3): Conv2d(256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False)\n",
      "        (bn3): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
      "        (relu): ReLU(inplace=True)\n",
      "      )\n",
      "    )\n",
      "    (layer4): Sequential(\n",
      "      (0): Bottleneck(\n",
      "        (conv1): Conv2d(1024, 512, kernel_size=(1, 1), stride=(1, 1), bias=False)\n",
      "        (bn1): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
      "        (conv2): Conv2d(512, 512, kernel_size=(3, 3), stride=(2, 2), padding=(1, 1), bias=False)\n",
      "        (bn2): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
      "        (conv3): Conv2d(512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False)\n",
      "        (bn3): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
      "        (relu): ReLU(inplace=True)\n",
      "        (downsample): Sequential(\n",
      "          (0): Conv2d(1024, 2048, kernel_size=(1, 1), stride=(2, 2), bias=False)\n",
      "          (1): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
      "        )\n",
      "      )\n",
      "      (1): Bottleneck(\n",
      "        (conv1): Conv2d(2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False)\n",
      "        (bn1): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
      "        (conv2): Conv2d(512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)\n",
      "        (bn2): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
      "        (conv3): Conv2d(512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False)\n",
      "        (bn3): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
      "        (relu): ReLU(inplace=True)\n",
      "      )\n",
      "      (2): Bottleneck(\n",
      "        (conv1): Conv2d(2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False)\n",
      "        (bn1): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
      "        (conv2): Conv2d(512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)\n",
      "        (bn2): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
      "        (conv3): Conv2d(512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False)\n",
      "        (bn3): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
      "        (relu): ReLU(inplace=True)\n",
      "      )\n",
      "    )\n",
      "    (avgpool): AdaptiveAvgPool2d(output_size=(1, 1))\n",
      "    (fc): Sequential(\n",
      "      (0): Linear(in_features=2048, out_features=5, bias=True)\n",
      "      (1): Sigmoid()\n",
      "    )\n",
      "  )\n",
      "  (features2): ResNet(\n",
      "    (conv1): Conv2d(3, 64, kernel_size=(7, 7), stride=(2, 2), padding=(3, 3), bias=False)\n",
      "    (bn1): BatchNorm2d(64, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
      "    (relu): ReLU(inplace=True)\n",
      "    (maxpool): MaxPool2d(kernel_size=3, stride=2, padding=1, dilation=1, ceil_mode=False)\n",
      "    (layer1): Sequential(\n",
      "      (0): Bottleneck(\n",
      "        (conv1): Conv2d(64, 64, kernel_size=(1, 1), stride=(1, 1), bias=False)\n",
      "        (bn1): BatchNorm2d(64, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
      "        (conv2): Conv2d(64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)\n",
      "        (bn2): BatchNorm2d(64, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
      "        (conv3): Conv2d(64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False)\n",
      "        (bn3): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
      "        (relu): ReLU(inplace=True)\n",
      "        (downsample): Sequential(\n",
      "          (0): Conv2d(64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False)\n",
      "          (1): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
      "        )\n",
      "      )\n",
      "      (1): Bottleneck(\n",
      "        (conv1): Conv2d(256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False)\n",
      "        (bn1): BatchNorm2d(64, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
      "        (conv2): Conv2d(64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)\n",
      "        (bn2): BatchNorm2d(64, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
      "        (conv3): Conv2d(64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False)\n",
      "        (bn3): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
      "        (relu): ReLU(inplace=True)\n",
      "      )\n",
      "      (2): Bottleneck(\n",
      "        (conv1): Conv2d(256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False)\n",
      "        (bn1): BatchNorm2d(64, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
      "        (conv2): Conv2d(64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)\n",
      "        (bn2): BatchNorm2d(64, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
      "        (conv3): Conv2d(64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False)\n",
      "        (bn3): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
      "        (relu): ReLU(inplace=True)\n",
      "      )\n",
      "    )\n",
      "    (layer2): Sequential(\n",
      "      (0): Bottleneck(\n",
      "        (conv1): Conv2d(256, 128, kernel_size=(1, 1), stride=(1, 1), bias=False)\n",
      "        (bn1): BatchNorm2d(128, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
      "        (conv2): Conv2d(128, 128, kernel_size=(3, 3), stride=(2, 2), padding=(1, 1), bias=False)\n",
      "        (bn2): BatchNorm2d(128, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
      "        (conv3): Conv2d(128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False)\n",
      "        (bn3): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
      "        (relu): ReLU(inplace=True)\n",
      "        (downsample): Sequential(\n",
      "          (0): Conv2d(256, 512, kernel_size=(1, 1), stride=(2, 2), bias=False)\n",
      "          (1): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
      "        )\n",
      "      )\n",
      "      (1): Bottleneck(\n",
      "        (conv1): Conv2d(512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False)\n",
      "        (bn1): BatchNorm2d(128, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
      "        (conv2): Conv2d(128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)\n",
      "        (bn2): BatchNorm2d(128, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
      "        (conv3): Conv2d(128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False)\n",
      "        (bn3): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
      "        (relu): ReLU(inplace=True)\n",
      "      )\n",
      "      (2): Bottleneck(\n",
      "        (conv1): Conv2d(512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False)\n",
      "        (bn1): BatchNorm2d(128, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
      "        (conv2): Conv2d(128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)\n",
      "        (bn2): BatchNorm2d(128, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
      "        (conv3): Conv2d(128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False)\n",
      "        (bn3): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
      "        (relu): ReLU(inplace=True)\n",
      "      )\n",
      "      (3): Bottleneck(\n",
      "        (conv1): Conv2d(512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False)\n",
      "        (bn1): BatchNorm2d(128, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
      "        (conv2): Conv2d(128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)\n",
      "        (bn2): BatchNorm2d(128, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
      "        (conv3): Conv2d(128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False)\n",
      "        (bn3): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
      "        (relu): ReLU(inplace=True)\n",
      "      )\n",
      "    )\n",
      "    (layer3): Sequential(\n",
      "      (0): Bottleneck(\n",
      "        (conv1): Conv2d(512, 256, kernel_size=(1, 1), stride=(1, 1), bias=False)\n",
      "        (bn1): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
      "        (conv2): Conv2d(256, 256, kernel_size=(3, 3), stride=(2, 2), padding=(1, 1), bias=False)\n",
      "        (bn2): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
      "        (conv3): Conv2d(256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False)\n",
      "        (bn3): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
      "        (relu): ReLU(inplace=True)\n",
      "        (downsample): Sequential(\n",
      "          (0): Conv2d(512, 1024, kernel_size=(1, 1), stride=(2, 2), bias=False)\n",
      "          (1): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
      "        )\n",
      "      )\n",
      "      (1): Bottleneck(\n",
      "        (conv1): Conv2d(1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False)\n",
      "        (bn1): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
      "        (conv2): Conv2d(256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)\n",
      "        (bn2): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
      "        (conv3): Conv2d(256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False)\n",
      "        (bn3): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
      "        (relu): ReLU(inplace=True)\n",
      "      )\n",
      "      (2): Bottleneck(\n",
      "        (conv1): Conv2d(1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False)\n",
      "        (bn1): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
      "        (conv2): Conv2d(256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)\n",
      "        (bn2): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
      "        (conv3): Conv2d(256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False)\n",
      "        (bn3): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
      "        (relu): ReLU(inplace=True)\n",
      "      )\n",
      "      (3): Bottleneck(\n",
      "        (conv1): Conv2d(1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False)\n",
      "        (bn1): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
      "        (conv2): Conv2d(256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)\n",
      "        (bn2): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
      "        (conv3): Conv2d(256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False)\n",
      "        (bn3): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
      "        (relu): ReLU(inplace=True)\n",
      "      )\n",
      "      (4): Bottleneck(\n",
      "        (conv1): Conv2d(1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False)\n",
      "        (bn1): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
      "        (conv2): Conv2d(256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)\n",
      "        (bn2): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
      "        (conv3): Conv2d(256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False)\n",
      "        (bn3): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
      "        (relu): ReLU(inplace=True)\n",
      "      )\n",
      "      (5): Bottleneck(\n",
      "        (conv1): Conv2d(1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False)\n",
      "        (bn1): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
      "        (conv2): Conv2d(256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)\n",
      "        (bn2): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
      "        (conv3): Conv2d(256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False)\n",
      "        (bn3): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
      "        (relu): ReLU(inplace=True)\n",
      "      )\n",
      "    )\n",
      "    (layer4): Sequential(\n",
      "      (0): Bottleneck(\n",
      "        (conv1): Conv2d(1024, 512, kernel_size=(1, 1), stride=(1, 1), bias=False)\n",
      "        (bn1): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
      "        (conv2): Conv2d(512, 512, kernel_size=(3, 3), stride=(2, 2), padding=(1, 1), bias=False)\n",
      "        (bn2): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
      "        (conv3): Conv2d(512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False)\n",
      "        (bn3): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
      "        (relu): ReLU(inplace=True)\n",
      "        (downsample): Sequential(\n",
      "          (0): Conv2d(1024, 2048, kernel_size=(1, 1), stride=(2, 2), bias=False)\n",
      "          (1): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
      "        )\n",
      "      )\n",
      "      (1): Bottleneck(\n",
      "        (conv1): Conv2d(2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False)\n",
      "        (bn1): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
      "        (conv2): Conv2d(512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)\n",
      "        (bn2): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
      "        (conv3): Conv2d(512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False)\n",
      "        (bn3): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
      "        (relu): ReLU(inplace=True)\n",
      "      )\n",
      "      (2): Bottleneck(\n",
      "        (conv1): Conv2d(2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False)\n",
      "        (bn1): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
      "        (conv2): Conv2d(512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)\n",
      "        (bn2): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
      "        (conv3): Conv2d(512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False)\n",
      "        (bn3): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
      "        (relu): ReLU(inplace=True)\n",
      "      )\n",
      "    )\n",
      "    (avgpool): AdaptiveAvgPool2d(output_size=(1, 1))\n",
      "    (fc): Sequential(\n",
      "      (0): Linear(in_features=2048, out_features=5, bias=True)\n",
      "      (1): Sigmoid()\n",
      "    )\n",
      "  )\n",
      "  (features3): ResNet(\n",
      "    (conv1): Conv2d(3, 64, kernel_size=(7, 7), stride=(2, 2), padding=(3, 3), bias=False)\n",
      "    (bn1): BatchNorm2d(64, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
      "    (relu): ReLU(inplace=True)\n",
      "    (maxpool): MaxPool2d(kernel_size=3, stride=2, padding=1, dilation=1, ceil_mode=False)\n",
      "    (layer1): Sequential(\n",
      "      (0): Bottleneck(\n",
      "        (conv1): Conv2d(64, 64, kernel_size=(1, 1), stride=(1, 1), bias=False)\n",
      "        (bn1): BatchNorm2d(64, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
      "        (conv2): Conv2d(64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)\n",
      "        (bn2): BatchNorm2d(64, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
      "        (conv3): Conv2d(64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False)\n",
      "        (bn3): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
      "        (relu): ReLU(inplace=True)\n",
      "        (downsample): Sequential(\n",
      "          (0): Conv2d(64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False)\n",
      "          (1): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
      "        )\n",
      "      )\n",
      "      (1): Bottleneck(\n",
      "        (conv1): Conv2d(256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False)\n",
      "        (bn1): BatchNorm2d(64, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
      "        (conv2): Conv2d(64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)\n",
      "        (bn2): BatchNorm2d(64, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
      "        (conv3): Conv2d(64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False)\n",
      "        (bn3): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
      "        (relu): ReLU(inplace=True)\n",
      "      )\n",
      "      (2): Bottleneck(\n",
      "        (conv1): Conv2d(256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False)\n",
      "        (bn1): BatchNorm2d(64, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
      "        (conv2): Conv2d(64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)\n",
      "        (bn2): BatchNorm2d(64, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
      "        (conv3): Conv2d(64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False)\n",
      "        (bn3): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
      "        (relu): ReLU(inplace=True)\n",
      "      )\n",
      "    )\n",
      "    (layer2): Sequential(\n",
      "      (0): Bottleneck(\n",
      "        (conv1): Conv2d(256, 128, kernel_size=(1, 1), stride=(1, 1), bias=False)\n",
      "        (bn1): BatchNorm2d(128, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
      "        (conv2): Conv2d(128, 128, kernel_size=(3, 3), stride=(2, 2), padding=(1, 1), bias=False)\n",
      "        (bn2): BatchNorm2d(128, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
      "        (conv3): Conv2d(128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False)\n",
      "        (bn3): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
      "        (relu): ReLU(inplace=True)\n",
      "        (downsample): Sequential(\n",
      "          (0): Conv2d(256, 512, kernel_size=(1, 1), stride=(2, 2), bias=False)\n",
      "          (1): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
      "        )\n",
      "      )\n",
      "      (1): Bottleneck(\n",
      "        (conv1): Conv2d(512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False)\n",
      "        (bn1): BatchNorm2d(128, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
      "        (conv2): Conv2d(128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)\n",
      "        (bn2): BatchNorm2d(128, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
      "        (conv3): Conv2d(128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False)\n",
      "        (bn3): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
      "        (relu): ReLU(inplace=True)\n",
      "      )\n",
      "      (2): Bottleneck(\n",
      "        (conv1): Conv2d(512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False)\n",
      "        (bn1): BatchNorm2d(128, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
      "        (conv2): Conv2d(128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)\n",
      "        (bn2): BatchNorm2d(128, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
      "        (conv3): Conv2d(128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False)\n",
      "        (bn3): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
      "        (relu): ReLU(inplace=True)\n",
      "      )\n",
      "      (3): Bottleneck(\n",
      "        (conv1): Conv2d(512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False)\n",
      "        (bn1): BatchNorm2d(128, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
      "        (conv2): Conv2d(128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)\n",
      "        (bn2): BatchNorm2d(128, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
      "        (conv3): Conv2d(128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False)\n",
      "        (bn3): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
      "        (relu): ReLU(inplace=True)\n",
      "      )\n",
      "    )\n",
      "    (layer3): Sequential(\n",
      "      (0): Bottleneck(\n",
      "        (conv1): Conv2d(512, 256, kernel_size=(1, 1), stride=(1, 1), bias=False)\n",
      "        (bn1): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
      "        (conv2): Conv2d(256, 256, kernel_size=(3, 3), stride=(2, 2), padding=(1, 1), bias=False)\n",
      "        (bn2): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
      "        (conv3): Conv2d(256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False)\n",
      "        (bn3): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
      "        (relu): ReLU(inplace=True)\n",
      "        (downsample): Sequential(\n",
      "          (0): Conv2d(512, 1024, kernel_size=(1, 1), stride=(2, 2), bias=False)\n",
      "          (1): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
      "        )\n",
      "      )\n",
      "      (1): Bottleneck(\n",
      "        (conv1): Conv2d(1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False)\n",
      "        (bn1): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
      "        (conv2): Conv2d(256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)\n",
      "        (bn2): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
      "        (conv3): Conv2d(256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False)\n",
      "        (bn3): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
      "        (relu): ReLU(inplace=True)\n",
      "      )\n",
      "      (2): Bottleneck(\n",
      "        (conv1): Conv2d(1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False)\n",
      "        (bn1): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
      "        (conv2): Conv2d(256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)\n",
      "        (bn2): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
      "        (conv3): Conv2d(256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False)\n",
      "        (bn3): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
      "        (relu): ReLU(inplace=True)\n",
      "      )\n",
      "      (3): Bottleneck(\n",
      "        (conv1): Conv2d(1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False)\n",
      "        (bn1): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
      "        (conv2): Conv2d(256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)\n",
      "        (bn2): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
      "        (conv3): Conv2d(256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False)\n",
      "        (bn3): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
      "        (relu): ReLU(inplace=True)\n",
      "      )\n",
      "      (4): Bottleneck(\n",
      "        (conv1): Conv2d(1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False)\n",
      "        (bn1): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
      "        (conv2): Conv2d(256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)\n",
      "        (bn2): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
      "        (conv3): Conv2d(256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False)\n",
      "        (bn3): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
      "        (relu): ReLU(inplace=True)\n",
      "      )\n",
      "      (5): Bottleneck(\n",
      "        (conv1): Conv2d(1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False)\n",
      "        (bn1): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
      "        (conv2): Conv2d(256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)\n",
      "        (bn2): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
      "        (conv3): Conv2d(256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False)\n",
      "        (bn3): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
      "        (relu): ReLU(inplace=True)\n",
      "      )\n",
      "    )\n",
      "    (layer4): Sequential(\n",
      "      (0): Bottleneck(\n",
      "        (conv1): Conv2d(1024, 512, kernel_size=(1, 1), stride=(1, 1), bias=False)\n",
      "        (bn1): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
      "        (conv2): Conv2d(512, 512, kernel_size=(3, 3), stride=(2, 2), padding=(1, 1), bias=False)\n",
      "        (bn2): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
      "        (conv3): Conv2d(512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False)\n",
      "        (bn3): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
      "        (relu): ReLU(inplace=True)\n",
      "        (downsample): Sequential(\n",
      "          (0): Conv2d(1024, 2048, kernel_size=(1, 1), stride=(2, 2), bias=False)\n",
      "          (1): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
      "        )\n",
      "      )\n",
      "      (1): Bottleneck(\n",
      "        (conv1): Conv2d(2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False)\n",
      "        (bn1): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
      "        (conv2): Conv2d(512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)\n",
      "        (bn2): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
      "        (conv3): Conv2d(512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False)\n",
      "        (bn3): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
      "        (relu): ReLU(inplace=True)\n",
      "      )\n",
      "      (2): Bottleneck(\n",
      "        (conv1): Conv2d(2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False)\n",
      "        (bn1): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
      "        (conv2): Conv2d(512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)\n",
      "        (bn2): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
      "        (conv3): Conv2d(512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False)\n",
      "        (bn3): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
      "        (relu): ReLU(inplace=True)\n",
      "      )\n",
      "    )\n",
      "    (avgpool): AdaptiveAvgPool2d(output_size=(1, 1))\n",
      "    (fc): Sequential(\n",
      "      (0): Linear(in_features=2048, out_features=5, bias=True)\n",
      "      (1): Sigmoid()\n",
      "    )\n",
      "  )\n",
      "  (features4): ResNet(\n",
      "    (conv1): Conv2d(3, 64, kernel_size=(7, 7), stride=(2, 2), padding=(3, 3), bias=False)\n",
      "    (bn1): BatchNorm2d(64, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
      "    (relu): ReLU(inplace=True)\n",
      "    (maxpool): MaxPool2d(kernel_size=3, stride=2, padding=1, dilation=1, ceil_mode=False)\n",
      "    (layer1): Sequential(\n",
      "      (0): Bottleneck(\n",
      "        (conv1): Conv2d(64, 64, kernel_size=(1, 1), stride=(1, 1), bias=False)\n",
      "        (bn1): BatchNorm2d(64, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
      "        (conv2): Conv2d(64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)\n",
      "        (bn2): BatchNorm2d(64, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
      "        (conv3): Conv2d(64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False)\n",
      "        (bn3): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
      "        (relu): ReLU(inplace=True)\n",
      "        (downsample): Sequential(\n",
      "          (0): Conv2d(64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False)\n",
      "          (1): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
      "        )\n",
      "      )\n",
      "      (1): Bottleneck(\n",
      "        (conv1): Conv2d(256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False)\n",
      "        (bn1): BatchNorm2d(64, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
      "        (conv2): Conv2d(64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)\n",
      "        (bn2): BatchNorm2d(64, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
      "        (conv3): Conv2d(64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False)\n",
      "        (bn3): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
      "        (relu): ReLU(inplace=True)\n",
      "      )\n",
      "      (2): Bottleneck(\n",
      "        (conv1): Conv2d(256, 64, kernel_size=(1, 1), stride=(1, 1), bias=False)\n",
      "        (bn1): BatchNorm2d(64, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
      "        (conv2): Conv2d(64, 64, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)\n",
      "        (bn2): BatchNorm2d(64, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
      "        (conv3): Conv2d(64, 256, kernel_size=(1, 1), stride=(1, 1), bias=False)\n",
      "        (bn3): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
      "        (relu): ReLU(inplace=True)\n",
      "      )\n",
      "    )\n",
      "    (layer2): Sequential(\n",
      "      (0): Bottleneck(\n",
      "        (conv1): Conv2d(256, 128, kernel_size=(1, 1), stride=(1, 1), bias=False)\n",
      "        (bn1): BatchNorm2d(128, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
      "        (conv2): Conv2d(128, 128, kernel_size=(3, 3), stride=(2, 2), padding=(1, 1), bias=False)\n",
      "        (bn2): BatchNorm2d(128, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
      "        (conv3): Conv2d(128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False)\n",
      "        (bn3): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
      "        (relu): ReLU(inplace=True)\n",
      "        (downsample): Sequential(\n",
      "          (0): Conv2d(256, 512, kernel_size=(1, 1), stride=(2, 2), bias=False)\n",
      "          (1): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
      "        )\n",
      "      )\n",
      "      (1): Bottleneck(\n",
      "        (conv1): Conv2d(512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False)\n",
      "        (bn1): BatchNorm2d(128, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
      "        (conv2): Conv2d(128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)\n",
      "        (bn2): BatchNorm2d(128, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
      "        (conv3): Conv2d(128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False)\n",
      "        (bn3): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
      "        (relu): ReLU(inplace=True)\n",
      "      )\n",
      "      (2): Bottleneck(\n",
      "        (conv1): Conv2d(512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False)\n",
      "        (bn1): BatchNorm2d(128, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
      "        (conv2): Conv2d(128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)\n",
      "        (bn2): BatchNorm2d(128, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
      "        (conv3): Conv2d(128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False)\n",
      "        (bn3): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
      "        (relu): ReLU(inplace=True)\n",
      "      )\n",
      "      (3): Bottleneck(\n",
      "        (conv1): Conv2d(512, 128, kernel_size=(1, 1), stride=(1, 1), bias=False)\n",
      "        (bn1): BatchNorm2d(128, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
      "        (conv2): Conv2d(128, 128, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)\n",
      "        (bn2): BatchNorm2d(128, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
      "        (conv3): Conv2d(128, 512, kernel_size=(1, 1), stride=(1, 1), bias=False)\n",
      "        (bn3): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
      "        (relu): ReLU(inplace=True)\n",
      "      )\n",
      "    )\n",
      "    (layer3): Sequential(\n",
      "      (0): Bottleneck(\n",
      "        (conv1): Conv2d(512, 256, kernel_size=(1, 1), stride=(1, 1), bias=False)\n",
      "        (bn1): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
      "        (conv2): Conv2d(256, 256, kernel_size=(3, 3), stride=(2, 2), padding=(1, 1), bias=False)\n",
      "        (bn2): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
      "        (conv3): Conv2d(256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False)\n",
      "        (bn3): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
      "        (relu): ReLU(inplace=True)\n",
      "        (downsample): Sequential(\n",
      "          (0): Conv2d(512, 1024, kernel_size=(1, 1), stride=(2, 2), bias=False)\n",
      "          (1): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
      "        )\n",
      "      )\n",
      "      (1): Bottleneck(\n",
      "        (conv1): Conv2d(1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False)\n",
      "        (bn1): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
      "        (conv2): Conv2d(256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)\n",
      "        (bn2): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
      "        (conv3): Conv2d(256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False)\n",
      "        (bn3): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
      "        (relu): ReLU(inplace=True)\n",
      "      )\n",
      "      (2): Bottleneck(\n",
      "        (conv1): Conv2d(1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False)\n",
      "        (bn1): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
      "        (conv2): Conv2d(256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)\n",
      "        (bn2): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
      "        (conv3): Conv2d(256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False)\n",
      "        (bn3): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
      "        (relu): ReLU(inplace=True)\n",
      "      )\n",
      "      (3): Bottleneck(\n",
      "        (conv1): Conv2d(1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False)\n",
      "        (bn1): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
      "        (conv2): Conv2d(256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)\n",
      "        (bn2): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
      "        (conv3): Conv2d(256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False)\n",
      "        (bn3): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
      "        (relu): ReLU(inplace=True)\n",
      "      )\n",
      "      (4): Bottleneck(\n",
      "        (conv1): Conv2d(1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False)\n",
      "        (bn1): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
      "        (conv2): Conv2d(256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)\n",
      "        (bn2): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
      "        (conv3): Conv2d(256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False)\n",
      "        (bn3): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
      "        (relu): ReLU(inplace=True)\n",
      "      )\n",
      "      (5): Bottleneck(\n",
      "        (conv1): Conv2d(1024, 256, kernel_size=(1, 1), stride=(1, 1), bias=False)\n",
      "        (bn1): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
      "        (conv2): Conv2d(256, 256, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)\n",
      "        (bn2): BatchNorm2d(256, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
      "        (conv3): Conv2d(256, 1024, kernel_size=(1, 1), stride=(1, 1), bias=False)\n",
      "        (bn3): BatchNorm2d(1024, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
      "        (relu): ReLU(inplace=True)\n",
      "      )\n",
      "    )\n",
      "    (layer4): Sequential(\n",
      "      (0): Bottleneck(\n",
      "        (conv1): Conv2d(1024, 512, kernel_size=(1, 1), stride=(1, 1), bias=False)\n",
      "        (bn1): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
      "        (conv2): Conv2d(512, 512, kernel_size=(3, 3), stride=(2, 2), padding=(1, 1), bias=False)\n",
      "        (bn2): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
      "        (conv3): Conv2d(512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False)\n",
      "        (bn3): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
      "        (relu): ReLU(inplace=True)\n",
      "        (downsample): Sequential(\n",
      "          (0): Conv2d(1024, 2048, kernel_size=(1, 1), stride=(2, 2), bias=False)\n",
      "          (1): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
      "        )\n",
      "      )\n",
      "      (1): Bottleneck(\n",
      "        (conv1): Conv2d(2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False)\n",
      "        (bn1): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
      "        (conv2): Conv2d(512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)\n",
      "        (bn2): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
      "        (conv3): Conv2d(512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False)\n",
      "        (bn3): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
      "        (relu): ReLU(inplace=True)\n",
      "      )\n",
      "      (2): Bottleneck(\n",
      "        (conv1): Conv2d(2048, 512, kernel_size=(1, 1), stride=(1, 1), bias=False)\n",
      "        (bn1): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
      "        (conv2): Conv2d(512, 512, kernel_size=(3, 3), stride=(1, 1), padding=(1, 1), bias=False)\n",
      "        (bn2): BatchNorm2d(512, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
      "        (conv3): Conv2d(512, 2048, kernel_size=(1, 1), stride=(1, 1), bias=False)\n",
      "        (bn3): BatchNorm2d(2048, eps=1e-05, momentum=0.1, affine=True, track_running_stats=True)\n",
      "        (relu): ReLU(inplace=True)\n",
      "      )\n",
      "    )\n",
      "    (avgpool): AdaptiveAvgPool2d(output_size=(1, 1))\n",
      "    (fc): Sequential(\n",
      "      (0): Linear(in_features=2048, out_features=5, bias=True)\n",
      "      (1): Sigmoid()\n",
      "    )\n",
      "  )\n",
      "  (transform1): ColorJitter(brightness=[0.85, 1.15], contrast=(0.5, 1.5), saturation=(0.5, 2.5), hue=[-0.15, 0.15])\n",
      "  (transform2): RandomResizedCrop(size=(512, 512), scale=(0.8, 1.0), ratio=(0.75, 1.3333), interpolation=bilinear)\n",
      "  (transform3): RandomHorizontalFlip(p=1)\n",
      "  (transform4): RandomVerticalFlip(p=1)\n",
      ")\n"
     ]
    }
   ],
   "source": [
    "print(model)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 22,
   "id": "dc6f38ad",
   "metadata": {
    "papermill": {
     "duration": 27.634299,
     "end_time": "2022-08-28T09:24:25.378475",
     "exception": false,
     "start_time": "2022-08-28T09:23:57.744176",
     "status": "completed"
    },
    "scrolled": true,
    "tags": []
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "ECE tensor(0.1585) MCE tensor(0.2902)\n",
      "Brier Loss -  0.14567827155052077\n",
      "AUC Classwise Score on Validation set =  [0.91801371 0.76704878 0.77307428 0.78385632 0.69732327]\n",
      "AUC Score on Validation set =  0.7878632718425872\n",
      "F1 Score on Validation set =  0.6285903499422685\n",
      "Cohen's Kappa Score on Validation set =  0.6589945145054572\n",
      "              precision    recall  f1-score   support\n",
      "\n",
      "           0       0.84      0.98      0.91       167\n",
      "           1       0.00      0.00      0.00        37\n",
      "           2       0.56      0.85      0.68       111\n",
      "           3       0.00      0.00      0.00        14\n",
      "           4       0.00      0.00      0.00        31\n",
      "\n",
      "    accuracy                           0.71       360\n",
      "   macro avg       0.28      0.36      0.32       360\n",
      "weighted avg       0.57      0.71      0.63       360\n",
      "\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/usr/local/lib/python3.9/dist-packages/sklearn/metrics/_classification.py:1327: UndefinedMetricWarning: Precision and F-score are ill-defined and being set to 0.0 in labels with no predicted samples. Use `zero_division` parameter to control this behavior.\n",
      "  _warn_prf(average, modifier, msg_start, len(result))\n",
      "/usr/local/lib/python3.9/dist-packages/sklearn/metrics/_classification.py:1327: UndefinedMetricWarning: Precision and F-score are ill-defined and being set to 0.0 in labels with no predicted samples. Use `zero_division` parameter to control this behavior.\n",
      "  _warn_prf(average, modifier, msg_start, len(result))\n",
      "/usr/local/lib/python3.9/dist-packages/sklearn/metrics/_classification.py:1327: UndefinedMetricWarning: Precision and F-score are ill-defined and being set to 0.0 in labels with no predicted samples. Use `zero_division` parameter to control this behavior.\n",
      "  _warn_prf(average, modifier, msg_start, len(result))\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "(0.7878632718425872,\n",
       " 0,\n",
       " 0.6285903499422685,\n",
       " 0.6589945145054572,\n",
       " tensor(0.1585),\n",
       " tensor(0.2902),\n",
       " 0.14567827155052077)"
      ]
     },
     "execution_count": 22,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "validate()"
   ]
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 3 (ipykernel)",
   "language": "python",
   "name": "python3"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.9.13"
  },
  "papermill": {
   "default_parameters": {},
   "duration": 4424.291439,
   "end_time": "2022-08-28T10:37:22.384275",
   "environment_variables": {},
   "exception": true,
   "input_path": "__notebook__.ipynb",
   "output_path": "__notebook__.ipynb",
   "parameters": {},
   "start_time": "2022-08-28T09:23:38.092836",
   "version": "2.3.4"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 5
}