chapter07/Cutmix_mixup_remix_FashionMNIST.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {
    "colab_type": "text",
    "id": "view-in-github"
   },
   "source": [
    "<a href=\"https://colab.research.google.com/github/packtpublishing/Machine-Learning-for-Imbalanced-Data/blob/main/chapter07/Cutmix_mixup_remix_FashionMNIST.ipynb\" target=\"_parent\"><img src=\"https://colab.research.google.com/assets/colab-badge.svg\" alt=\"Open In Colab\"/></a>"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "uzhBgFhyAEFK"
   },
   "source": [
    "# Running a default CNN model on Fashion MNIST dataset to benchmark it's accuracy\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {
    "id": "-QTP4RgOW1wt"
   },
   "outputs": [],
   "source": [
    "import torch\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 DataLoader, random_split\n",
    "from torchvision import datasets, transforms\n",
    "import numpy as np\n",
    "from collections import Counter\n",
    "import matplotlib.pyplot as plt\n",
    "import time\n",
    "import torchvision\n",
    "import random\n",
    "import copy\n",
    "import cv2\n",
    "import pandas as pd\n",
    "from collections import Counter\n",
    "import matplotlib.pyplot as plt\n",
    "from torch.utils.data.sampler import SubsetRandomSampler\n",
    "\n",
    "seed = 44\n",
    "torch.manual_seed(seed)\n",
    "torch.cuda.manual_seed(seed)\n",
    "torch.cuda.manual_seed_all(seed)\n",
    "np.random.seed(seed)\n",
    "random.seed(seed)\n",
    "torch.backends.cudnn.benchmark = False\n",
    "torch.backends.cudnn.deterministic = True"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {
    "id": "Wev_Z81ZAEFM",
    "scrolled": true
   },
   "outputs": [],
   "source": [
    "class ConvNet(nn.Module):\n",
    "    def __init__(self):\n",
    "        super(ConvNet, self).__init__()\n",
    "        self.cn1 = nn.Conv2d(1, 16, 3, 1)\n",
    "        self.cn2 = nn.Conv2d(16, 32, 3, 1)\n",
    "        self.dp1 = nn.Dropout2d(0.10)\n",
    "        self.dp2 = nn.Dropout2d(0.25)\n",
    "        self.fc1 = nn.Linear(4608, 64)  # 4608 is basically 12 X 12 X 32\n",
    "        self.fc2 = nn.Linear(64, 10)\n",
    "\n",
    "    def forward(self, x):\n",
    "        x = self.cn1(x)\n",
    "        x = F.relu(x)\n",
    "        x = self.cn2(x)\n",
    "        x = F.relu(x)\n",
    "        x = F.max_pool2d(x, 2)\n",
    "        x = self.dp1(x)\n",
    "        x = torch.flatten(x, 1)\n",
    "        x = self.fc1(x)\n",
    "        x = F.relu(x)\n",
    "        x = self.dp2(x)\n",
    "        x = self.fc2(x)\n",
    "        return x"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {
    "id": "UTAuYERHNqIT"
   },
   "outputs": [],
   "source": [
    "device = torch.device(\"cuda:0\" if torch.cuda.is_available() else \"cpu\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {
    "id": "ONODalBYXcEG"
   },
   "outputs": [],
   "source": [
    "def train(\n",
    "    train_dataloader,\n",
    "    valid_dataloader,\n",
    "    PATH,\n",
    "    model=None,\n",
    "    criterion=F.cross_entropy,\n",
    "    epochs=1,\n",
    "):\n",
    "    optimizer = torch.optim.SGD(model.parameters(), lr=0.1, momentum=0.9)\n",
    "    model.train()\n",
    "    best_acc = 0\n",
    "    train_losses = []\n",
    "    val_losses = []\n",
    "    for epoch in range(epochs):\n",
    "        losses = 0\n",
    "        for b_i, (X, y) in enumerate(train_dataloader):\n",
    "            X, y = X.to(device), y.to(device)\n",
    "            pred_prob = model(X)\n",
    "            loss = criterion(pred_prob, y)\n",
    "            optimizer.zero_grad()\n",
    "            loss.backward()\n",
    "            optimizer.step()\n",
    "            losses += loss.item()  # print(losses)\n",
    "            if b_i % 10 == 0:\n",
    "                print(\n",
    "                    \"epoch: {} [{}/{} ({:.0f}%)]\\t training loss: {:.6f}\".format(\n",
    "                        epoch,\n",
    "                        b_i,\n",
    "                        len(train_dataloader),\n",
    "                        100.0 * b_i / len(train_dataloader),\n",
    "                        loss.item(),\n",
    "                    )\n",
    "                )\n",
    "        train_losses.append(losses / len(train_dataloader))\n",
    "        print(\"\\n\")\n",
    "        print(\"Validation...\")\n",
    "        model.eval()\n",
    "        success = 0\n",
    "        total = 0\n",
    "        losses = 0\n",
    "        with torch.no_grad():\n",
    "            for X, y in valid_dataloader:\n",
    "                X, y = X.to(device), y.to(device)\n",
    "                pred_prob = model(X)\n",
    "                loss = criterion(pred_prob, y)  # loss summed across the batch\n",
    "                losses += loss.item()\n",
    "                pred = pred_prob.argmax(\n",
    "                    dim=1, keepdim=True\n",
    "                )  # use argmax to get the most likely prediction\n",
    "                success += pred.eq(y.view_as(pred)).sum().item()\n",
    "                total += len(y)\n",
    "        val_losses.append(losses / len(valid_dataloader))\n",
    "        accuracy = 100.0 * success / total\n",
    "\n",
    "        if accuracy > best_acc:\n",
    "            torch.save(model.state_dict(), PATH)\n",
    "            best_acc = accuracy\n",
    "\n",
    "        print(\n",
    "            \"\\nTest dataset: Overall Loss: {:.4f}, Overall Accuracy: {}/{} ({:.0f}%)\\n\".format(\n",
    "                losses / len(valid_dataloader), success, total, accuracy\n",
    "            )\n",
    "        )\n",
    "    return model, train_losses, val_losses\n",
    "\n",
    "\n",
    "def test(model, PATH, criterion=F.cross_entropy):\n",
    "    model.eval()\n",
    "    model.load_state_dict(torch.load(PATH))\n",
    "    losses = 0\n",
    "    success = 0\n",
    "    correct = 0\n",
    "    class_correct = list(0.0 for i in range(10))\n",
    "    class_total = list(0.0 for i in range(10))\n",
    "    with torch.no_grad():\n",
    "        for X, y in test_dataloader:\n",
    "            X, y = X.to(device), y.to(device)\n",
    "            pred_prob = model(X)\n",
    "            losses += criterion(pred_prob, y).item()  # loss summed across the batch\n",
    "            pred = pred_prob.argmax(\n",
    "                dim=1, keepdim=True\n",
    "            )  # use argmax to get the most likely prediction\n",
    "            correct = np.squeeze(pred.eq(y.view_as(pred)))\n",
    "            success += pred.eq(y.view_as(pred)).sum().item()\n",
    "\n",
    "    losses /= len(test_dataloader.dataset)\n",
    "\n",
    "    accuracy = 100.0 * success / len(test_dataloader.dataset)\n",
    "\n",
    "    print(\n",
    "        \"\\nTest dataset: Overall Loss: {:.4f}, Overall Accuracy: {}/{} ({:.0f}%)\\n\".format(\n",
    "            losses, success, len(test_dataloader.dataset), accuracy\n",
    "        )\n",
    "    )\n",
    "\n",
    "    for i in range(len(y)):\n",
    "        label = y[i]\n",
    "        class_correct[label] += correct[i].item()\n",
    "        class_total[label] += 1\n",
    "    class_percentages = []\n",
    "    for i in range(len(class_correct)):\n",
    "        class_percentages.append(class_correct[i] / class_total[i])\n",
    "\n",
    "    return accuracy, class_percentages"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "905CC-Uk5fr5"
   },
   "source": [
    "## Orignal Fashion Mnist Dataset Class distribution"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 847
    },
    "id": "fTSWAJ-8XSlS",
    "outputId": "2010c4df-239c-4351-8477-5c5046c9b947"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Downloading http://fashion-mnist.s3-website.eu-central-1.amazonaws.com/train-images-idx3-ubyte.gz\n",
      "Downloading http://fashion-mnist.s3-website.eu-central-1.amazonaws.com/train-images-idx3-ubyte.gz to ../data/FashionMNIST/raw/train-images-idx3-ubyte.gz\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 26421880/26421880 [00:01<00:00, 16409269.59it/s]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Extracting ../data/FashionMNIST/raw/train-images-idx3-ubyte.gz to ../data/FashionMNIST/raw\n",
      "\n",
      "Downloading http://fashion-mnist.s3-website.eu-central-1.amazonaws.com/train-labels-idx1-ubyte.gz\n",
      "Downloading http://fashion-mnist.s3-website.eu-central-1.amazonaws.com/train-labels-idx1-ubyte.gz to ../data/FashionMNIST/raw/train-labels-idx1-ubyte.gz\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 29515/29515 [00:00<00:00, 280881.53it/s]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Extracting ../data/FashionMNIST/raw/train-labels-idx1-ubyte.gz to ../data/FashionMNIST/raw\n",
      "\n",
      "Downloading http://fashion-mnist.s3-website.eu-central-1.amazonaws.com/t10k-images-idx3-ubyte.gz\n",
      "Downloading http://fashion-mnist.s3-website.eu-central-1.amazonaws.com/t10k-images-idx3-ubyte.gz to ../data/FashionMNIST/raw/t10k-images-idx3-ubyte.gz\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 4422102/4422102 [00:00<00:00, 5023091.59it/s]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Extracting ../data/FashionMNIST/raw/t10k-images-idx3-ubyte.gz to ../data/FashionMNIST/raw\n",
      "\n",
      "Downloading http://fashion-mnist.s3-website.eu-central-1.amazonaws.com/t10k-labels-idx1-ubyte.gz\n",
      "Downloading http://fashion-mnist.s3-website.eu-central-1.amazonaws.com/t10k-labels-idx1-ubyte.gz to ../data/FashionMNIST/raw/t10k-labels-idx1-ubyte.gz\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|██████████| 5148/5148 [00:00<00:00, 20821867.88it/s]\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Extracting ../data/FashionMNIST/raw/t10k-labels-idx1-ubyte.gz to ../data/FashionMNIST/raw\n",
      "\n",
      "Counter({9: 6000, 0: 6000, 3: 6000, 2: 6000, 7: 6000, 5: 6000, 1: 6000, 6: 6000, 4: 6000, 8: 6000})\n",
      "Distribution of classes in original dataset:\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 640x480 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "train_size 48000\n",
      "val_size 12000\n"
     ]
    }
   ],
   "source": [
    "trainset = datasets.FashionMNIST(\n",
    "    \"../data\",\n",
    "    train=True,\n",
    "    download=True,\n",
    "    transform=transforms.Compose(\n",
    "        [transforms.ToTensor(), transforms.Normalize((0.1302,), (0.3069,))]\n",
    "    ),\n",
    ")\n",
    "print(Counter(trainset.targets.numpy()))\n",
    "\n",
    "print(\"Distribution of classes in original dataset:\")\n",
    "fig, ax = plt.subplots()\n",
    "_, counts = np.unique(trainset.targets, return_counts=True)\n",
    "ax.bar(range(10), counts)\n",
    "ax.set_xticks(range(10))\n",
    "plt.show()\n",
    "\n",
    "# Preparing for validaion test\n",
    "indices = list(range(len(trainset)))\n",
    "np.random.seed(seed)\n",
    "np.random.shuffle(indices)\n",
    "# to get 20% of the train set\n",
    "split = int(np.floor(0.2 * len(trainset)))\n",
    "train_sample = SubsetRandomSampler(indices[split:])\n",
    "valid_sample = SubsetRandomSampler(indices[:split])\n",
    "\n",
    "# Data Loader\n",
    "train_dataloader = torch.utils.data.DataLoader(\n",
    "    trainset, sampler=train_sample, batch_size=64\n",
    ")\n",
    "valid_dataloader = torch.utils.data.DataLoader(\n",
    "    trainset, sampler=valid_sample, batch_size=64\n",
    ")\n",
    "print(\"train_size\", len(train_sample))\n",
    "print(\"val_size\", len(valid_sample))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 506
    },
    "id": "o3NlNzpYukav",
    "outputId": "14b252a8-2663-47d6-e971-469595b4c0cb"
   },
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 1500x500 with 20 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "dataiter = iter(train_dataloader)\n",
    "images, labels = next(dataiter)\n",
    "\n",
    "\n",
    "fig = plt.figure(figsize=(15, 5))\n",
    "for idx in np.arange(20):\n",
    "    # xticks=[], yticks=[] is empty to print the images without any ticks around them\n",
    "    # np.sqeeze : Remove single-dimensional entries from the shape of an array.\n",
    "    ax = fig.add_subplot(4, int(20 / 4), idx + 1, xticks=[], yticks=[])\n",
    "    ax.imshow(np.squeeze(images[idx]), cmap=\"gray\")\n",
    "    # .item() gets the value contained in a Tensor\n",
    "    ax.set_title(labels[idx].item())\n",
    "    fig.tight_layout()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {
    "id": "kvorLJErYIUm"
   },
   "outputs": [],
   "source": [
    "test_ds = datasets.FashionMNIST(\n",
    "    \"../data\",\n",
    "    train=False,\n",
    "    transform=transforms.Compose(\n",
    "        [transforms.ToTensor(), transforms.Normalize((0.1302,), (0.3069,))]\n",
    "    ),\n",
    ")\n",
    "\n",
    "test_dataloader = torch.utils.data.DataLoader(test_ds, batch_size=500)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "1YJAHjdD5fr7"
   },
   "source": [
    "# Balanced/Orignal Dataset Training"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/"
    },
    "id": "v0UmM2LA5fr7",
    "outputId": "74dd0442-c12a-44c9-f6d4-5318e3367fbd",
    "scrolled": true
   },
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/usr/local/lib/python3.10/dist-packages/torch/nn/functional.py:1331: UserWarning: dropout2d: Received a 2-D input to dropout2d, which is deprecated and will result in an error in a future release. To retain the behavior and silence this warning, please use dropout instead. Note that dropout2d exists to provide channel-wise dropout on inputs with 2 spatial dimensions, a channel dimension, and an optional batch dimension (i.e. 3D or 4D inputs).\n",
      "  warnings.warn(warn_msg)\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "epoch: 0 [0/750 (0%)]\t training loss: 2.306934\n",
      "epoch: 0 [10/750 (1%)]\t training loss: 1.508018\n",
      "epoch: 0 [20/750 (3%)]\t training loss: 1.118173\n",
      "epoch: 0 [30/750 (4%)]\t training loss: 0.885390\n",
      "epoch: 0 [40/750 (5%)]\t training loss: 0.755212\n",
      "epoch: 0 [50/750 (7%)]\t training loss: 0.913945\n",
      "epoch: 0 [60/750 (8%)]\t training loss: 0.929232\n",
      "epoch: 0 [70/750 (9%)]\t training loss: 0.778155\n",
      "epoch: 0 [80/750 (11%)]\t training loss: 0.994409\n",
      "epoch: 0 [90/750 (12%)]\t training loss: 0.741211\n",
      "epoch: 0 [100/750 (13%)]\t training loss: 0.608615\n",
      "epoch: 0 [110/750 (15%)]\t training loss: 0.428522\n",
      "epoch: 0 [120/750 (16%)]\t training loss: 0.573671\n",
      "epoch: 0 [130/750 (17%)]\t training loss: 0.838935\n",
      "epoch: 0 [140/750 (19%)]\t training loss: 0.464045\n",
      "epoch: 0 [150/750 (20%)]\t training loss: 0.712092\n",
      "epoch: 0 [160/750 (21%)]\t training loss: 0.573884\n",
      "epoch: 0 [170/750 (23%)]\t training loss: 0.617408\n",
      "epoch: 0 [180/750 (24%)]\t training loss: 0.419022\n",
      "epoch: 0 [190/750 (25%)]\t training loss: 0.719708\n",
      "epoch: 0 [200/750 (27%)]\t training loss: 0.460746\n",
      "epoch: 0 [210/750 (28%)]\t training loss: 0.539774\n",
      "epoch: 0 [220/750 (29%)]\t training loss: 0.481191\n",
      "epoch: 0 [230/750 (31%)]\t training loss: 0.633359\n",
      "epoch: 0 [240/750 (32%)]\t training loss: 0.728911\n",
      "epoch: 0 [250/750 (33%)]\t training loss: 0.599635\n",
      "epoch: 0 [260/750 (35%)]\t training loss: 0.681401\n",
      "epoch: 0 [270/750 (36%)]\t training loss: 0.691916\n",
      "epoch: 0 [280/750 (37%)]\t training loss: 0.416722\n",
      "epoch: 0 [290/750 (39%)]\t training loss: 0.407077\n",
      "epoch: 0 [300/750 (40%)]\t training loss: 0.446469\n",
      "epoch: 0 [310/750 (41%)]\t training loss: 0.536067\n",
      "epoch: 0 [320/750 (43%)]\t training loss: 0.442146\n",
      "epoch: 0 [330/750 (44%)]\t training loss: 0.719170\n",
      "epoch: 0 [340/750 (45%)]\t training loss: 0.446197\n",
      "epoch: 0 [350/750 (47%)]\t training loss: 0.535904\n",
      "epoch: 0 [360/750 (48%)]\t training loss: 0.381543\n",
      "epoch: 0 [370/750 (49%)]\t training loss: 0.729133\n",
      "epoch: 0 [380/750 (51%)]\t training loss: 0.678214\n",
      "epoch: 0 [390/750 (52%)]\t training loss: 0.491959\n",
      "epoch: 0 [400/750 (53%)]\t training loss: 0.438749\n",
      "epoch: 0 [410/750 (55%)]\t training loss: 0.496906\n",
      "epoch: 0 [420/750 (56%)]\t training loss: 0.417665\n",
      "epoch: 0 [430/750 (57%)]\t training loss: 0.504904\n",
      "epoch: 0 [440/750 (59%)]\t training loss: 0.469582\n",
      "epoch: 0 [450/750 (60%)]\t training loss: 0.517417\n",
      "epoch: 0 [460/750 (61%)]\t training loss: 0.627448\n",
      "epoch: 0 [470/750 (63%)]\t training loss: 0.446168\n",
      "epoch: 0 [480/750 (64%)]\t training loss: 0.587714\n",
      "epoch: 0 [490/750 (65%)]\t training loss: 0.745876\n",
      "epoch: 0 [500/750 (67%)]\t training loss: 0.464782\n",
      "epoch: 0 [510/750 (68%)]\t training loss: 0.595472\n",
      "epoch: 0 [520/750 (69%)]\t training loss: 0.550641\n",
      "epoch: 0 [530/750 (71%)]\t training loss: 0.771670\n",
      "epoch: 0 [540/750 (72%)]\t training loss: 0.422869\n",
      "epoch: 0 [550/750 (73%)]\t training loss: 0.559881\n",
      "epoch: 0 [560/750 (75%)]\t training loss: 0.398368\n",
      "epoch: 0 [570/750 (76%)]\t training loss: 0.538158\n",
      "epoch: 0 [580/750 (77%)]\t training loss: 0.442994\n",
      "epoch: 0 [590/750 (79%)]\t training loss: 0.442300\n",
      "epoch: 0 [600/750 (80%)]\t training loss: 0.580319\n",
      "epoch: 0 [610/750 (81%)]\t training loss: 0.295316\n",
      "epoch: 0 [620/750 (83%)]\t training loss: 0.630391\n",
      "epoch: 0 [630/750 (84%)]\t training loss: 0.519642\n",
      "epoch: 0 [640/750 (85%)]\t training loss: 0.495268\n",
      "epoch: 0 [650/750 (87%)]\t training loss: 0.624442\n",
      "epoch: 0 [660/750 (88%)]\t training loss: 0.433937\n",
      "epoch: 0 [670/750 (89%)]\t training loss: 0.464756\n",
      "epoch: 0 [680/750 (91%)]\t training loss: 0.440463\n",
      "epoch: 0 [690/750 (92%)]\t training loss: 0.274612\n",
      "epoch: 0 [700/750 (93%)]\t training loss: 0.487445\n",
      "epoch: 0 [710/750 (95%)]\t training loss: 0.431452\n",
      "epoch: 0 [720/750 (96%)]\t training loss: 0.344376\n",
      "epoch: 0 [730/750 (97%)]\t training loss: 0.349612\n",
      "epoch: 0 [740/750 (99%)]\t training loss: 0.518395\n",
      "\n",
      "\n",
      "Validation...\n",
      "\n",
      "Test dataset: Overall Loss: 0.3791, Overall Accuracy: 10318/12000 (86%)\n",
      "\n",
      "epoch: 1 [0/750 (0%)]\t training loss: 0.553463\n",
      "epoch: 1 [10/750 (1%)]\t training loss: 0.315056\n",
      "epoch: 1 [20/750 (3%)]\t training loss: 0.307259\n",
      "epoch: 1 [30/750 (4%)]\t training loss: 0.448629\n",
      "epoch: 1 [40/750 (5%)]\t training loss: 0.334705\n",
      "epoch: 1 [50/750 (7%)]\t training loss: 0.464217\n",
      "epoch: 1 [60/750 (8%)]\t training loss: 0.434635\n",
      "epoch: 1 [70/750 (9%)]\t training loss: 0.197502\n",
      "epoch: 1 [80/750 (11%)]\t training loss: 0.288463\n",
      "epoch: 1 [90/750 (12%)]\t training loss: 0.522255\n",
      "epoch: 1 [100/750 (13%)]\t training loss: 0.597911\n",
      "epoch: 1 [110/750 (15%)]\t training loss: 0.230763\n",
      "epoch: 1 [120/750 (16%)]\t training loss: 0.340005\n",
      "epoch: 1 [130/750 (17%)]\t training loss: 0.320916\n",
      "epoch: 1 [140/750 (19%)]\t training loss: 0.291360\n",
      "epoch: 1 [150/750 (20%)]\t training loss: 0.352963\n",
      "epoch: 1 [160/750 (21%)]\t training loss: 0.251321\n",
      "epoch: 1 [170/750 (23%)]\t training loss: 0.254643\n",
      "epoch: 1 [180/750 (24%)]\t training loss: 0.362039\n",
      "epoch: 1 [190/750 (25%)]\t training loss: 0.443441\n",
      "epoch: 1 [200/750 (27%)]\t training loss: 0.454715\n",
      "epoch: 1 [210/750 (28%)]\t training loss: 0.187645\n",
      "epoch: 1 [220/750 (29%)]\t training loss: 0.251217\n",
      "epoch: 1 [230/750 (31%)]\t training loss: 0.356141\n",
      "epoch: 1 [240/750 (32%)]\t training loss: 0.311542\n",
      "epoch: 1 [250/750 (33%)]\t training loss: 0.274159\n",
      "epoch: 1 [260/750 (35%)]\t training loss: 0.504877\n",
      "epoch: 1 [270/750 (36%)]\t training loss: 0.304791\n",
      "epoch: 1 [280/750 (37%)]\t training loss: 0.324160\n",
      "epoch: 1 [290/750 (39%)]\t training loss: 0.431593\n",
      "epoch: 1 [300/750 (40%)]\t training loss: 0.484573\n",
      "epoch: 1 [310/750 (41%)]\t training loss: 0.389688\n",
      "epoch: 1 [320/750 (43%)]\t training loss: 0.253193\n",
      "epoch: 1 [330/750 (44%)]\t training loss: 0.164508\n",
      "epoch: 1 [340/750 (45%)]\t training loss: 0.213411\n",
      "epoch: 1 [350/750 (47%)]\t training loss: 0.347766\n",
      "epoch: 1 [360/750 (48%)]\t training loss: 0.449777\n",
      "epoch: 1 [370/750 (49%)]\t training loss: 0.199232\n",
      "epoch: 1 [380/750 (51%)]\t training loss: 0.426832\n",
      "epoch: 1 [390/750 (52%)]\t training loss: 0.274139\n",
      "epoch: 1 [400/750 (53%)]\t training loss: 0.345976\n",
      "epoch: 1 [410/750 (55%)]\t training loss: 0.235250\n",
      "epoch: 1 [420/750 (56%)]\t training loss: 0.538637\n",
      "epoch: 1 [430/750 (57%)]\t training loss: 0.241725\n",
      "epoch: 1 [440/750 (59%)]\t training loss: 0.241649\n",
      "epoch: 1 [450/750 (60%)]\t training loss: 0.262408\n",
      "epoch: 1 [460/750 (61%)]\t training loss: 0.347808\n",
      "epoch: 1 [470/750 (63%)]\t training loss: 0.347453\n",
      "epoch: 1 [480/750 (64%)]\t training loss: 0.294281\n",
      "epoch: 1 [490/750 (65%)]\t training loss: 0.260106\n",
      "epoch: 1 [500/750 (67%)]\t training loss: 0.480566\n",
      "epoch: 1 [510/750 (68%)]\t training loss: 0.377001\n",
      "epoch: 1 [520/750 (69%)]\t training loss: 0.401888\n",
      "epoch: 1 [530/750 (71%)]\t training loss: 0.258333\n",
      "epoch: 1 [540/750 (72%)]\t training loss: 0.296572\n",
      "epoch: 1 [550/750 (73%)]\t training loss: 0.338461\n",
      "epoch: 1 [560/750 (75%)]\t training loss: 0.326632\n",
      "epoch: 1 [570/750 (76%)]\t training loss: 0.220273\n",
      "epoch: 1 [580/750 (77%)]\t training loss: 0.602551\n",
      "epoch: 1 [590/750 (79%)]\t training loss: 0.263760\n",
      "epoch: 1 [600/750 (80%)]\t training loss: 0.362486\n",
      "epoch: 1 [610/750 (81%)]\t training loss: 0.305573\n",
      "epoch: 1 [620/750 (83%)]\t training loss: 0.193818\n",
      "epoch: 1 [630/750 (84%)]\t training loss: 0.488655\n",
      "epoch: 1 [640/750 (85%)]\t training loss: 0.238144\n",
      "epoch: 1 [650/750 (87%)]\t training loss: 0.394116\n",
      "epoch: 1 [660/750 (88%)]\t training loss: 0.379635\n",
      "epoch: 1 [670/750 (89%)]\t training loss: 0.346700\n",
      "epoch: 1 [680/750 (91%)]\t training loss: 0.211245\n",
      "epoch: 1 [690/750 (92%)]\t training loss: 0.321233\n",
      "epoch: 1 [700/750 (93%)]\t training loss: 0.373089\n",
      "epoch: 1 [710/750 (95%)]\t training loss: 0.279672\n",
      "epoch: 1 [720/750 (96%)]\t training loss: 0.275221\n",
      "epoch: 1 [730/750 (97%)]\t training loss: 0.202729\n",
      "epoch: 1 [740/750 (99%)]\t training loss: 0.316446\n",
      "\n",
      "\n",
      "Validation...\n",
      "\n",
      "Test dataset: Overall Loss: 0.3513, Overall Accuracy: 10404/12000 (87%)\n",
      "\n",
      "epoch: 2 [0/750 (0%)]\t training loss: 0.233171\n",
      "epoch: 2 [10/750 (1%)]\t training loss: 0.314593\n",
      "epoch: 2 [20/750 (3%)]\t training loss: 0.294229\n",
      "epoch: 2 [30/750 (4%)]\t training loss: 0.213041\n",
      "epoch: 2 [40/750 (5%)]\t training loss: 0.410431\n",
      "epoch: 2 [50/750 (7%)]\t training loss: 0.285973\n",
      "epoch: 2 [60/750 (8%)]\t training loss: 0.569551\n",
      "epoch: 2 [70/750 (9%)]\t training loss: 0.323368\n",
      "epoch: 2 [80/750 (11%)]\t training loss: 0.336944\n",
      "epoch: 2 [90/750 (12%)]\t training loss: 0.236642\n",
      "epoch: 2 [100/750 (13%)]\t training loss: 0.293295\n",
      "epoch: 2 [110/750 (15%)]\t training loss: 0.298649\n",
      "epoch: 2 [120/750 (16%)]\t training loss: 0.282947\n",
      "epoch: 2 [130/750 (17%)]\t training loss: 0.368826\n",
      "epoch: 2 [140/750 (19%)]\t training loss: 0.297736\n",
      "epoch: 2 [150/750 (20%)]\t training loss: 0.462325\n",
      "epoch: 2 [160/750 (21%)]\t training loss: 0.233111\n",
      "epoch: 2 [170/750 (23%)]\t training loss: 0.225968\n",
      "epoch: 2 [180/750 (24%)]\t training loss: 0.236404\n",
      "epoch: 2 [190/750 (25%)]\t training loss: 0.584402\n",
      "epoch: 2 [200/750 (27%)]\t training loss: 0.266539\n",
      "epoch: 2 [210/750 (28%)]\t training loss: 0.289646\n",
      "epoch: 2 [220/750 (29%)]\t training loss: 0.460223\n",
      "epoch: 2 [230/750 (31%)]\t training loss: 0.309606\n",
      "epoch: 2 [240/750 (32%)]\t training loss: 0.257239\n",
      "epoch: 2 [250/750 (33%)]\t training loss: 0.325580\n",
      "epoch: 2 [260/750 (35%)]\t training loss: 0.368266\n",
      "epoch: 2 [270/750 (36%)]\t training loss: 0.295819\n",
      "epoch: 2 [280/750 (37%)]\t training loss: 0.300267\n",
      "epoch: 2 [290/750 (39%)]\t training loss: 0.237454\n",
      "epoch: 2 [300/750 (40%)]\t training loss: 0.393714\n",
      "epoch: 2 [310/750 (41%)]\t training loss: 0.446705\n",
      "epoch: 2 [320/750 (43%)]\t training loss: 0.378483\n",
      "epoch: 2 [330/750 (44%)]\t training loss: 0.388409\n",
      "epoch: 2 [340/750 (45%)]\t training loss: 0.303010\n",
      "epoch: 2 [350/750 (47%)]\t training loss: 0.267607\n",
      "epoch: 2 [360/750 (48%)]\t training loss: 0.228519\n",
      "epoch: 2 [370/750 (49%)]\t training loss: 0.263846\n",
      "epoch: 2 [380/750 (51%)]\t training loss: 0.240332\n",
      "epoch: 2 [390/750 (52%)]\t training loss: 0.332765\n",
      "epoch: 2 [400/750 (53%)]\t training loss: 0.508414\n",
      "epoch: 2 [410/750 (55%)]\t training loss: 0.341880\n",
      "epoch: 2 [420/750 (56%)]\t training loss: 0.166941\n",
      "epoch: 2 [430/750 (57%)]\t training loss: 0.407036\n",
      "epoch: 2 [440/750 (59%)]\t training loss: 0.290222\n",
      "epoch: 2 [450/750 (60%)]\t training loss: 0.266521\n",
      "epoch: 2 [460/750 (61%)]\t training loss: 0.388538\n",
      "epoch: 2 [470/750 (63%)]\t training loss: 0.235283\n",
      "epoch: 2 [480/750 (64%)]\t training loss: 0.231317\n",
      "epoch: 2 [490/750 (65%)]\t training loss: 0.270772\n",
      "epoch: 2 [500/750 (67%)]\t training loss: 0.454902\n",
      "epoch: 2 [510/750 (68%)]\t training loss: 0.354435\n",
      "epoch: 2 [520/750 (69%)]\t training loss: 0.110219\n",
      "epoch: 2 [530/750 (71%)]\t training loss: 0.285167\n",
      "epoch: 2 [540/750 (72%)]\t training loss: 0.092475\n",
      "epoch: 2 [550/750 (73%)]\t training loss: 0.307198\n",
      "epoch: 2 [560/750 (75%)]\t training loss: 0.350137\n",
      "epoch: 2 [570/750 (76%)]\t training loss: 0.207336\n",
      "epoch: 2 [580/750 (77%)]\t training loss: 0.174457\n",
      "epoch: 2 [590/750 (79%)]\t training loss: 0.367369\n",
      "epoch: 2 [600/750 (80%)]\t training loss: 0.370725\n",
      "epoch: 2 [610/750 (81%)]\t training loss: 0.230972\n",
      "epoch: 2 [620/750 (83%)]\t training loss: 0.225649\n",
      "epoch: 2 [630/750 (84%)]\t training loss: 0.188458\n",
      "epoch: 2 [640/750 (85%)]\t training loss: 0.193441\n",
      "epoch: 2 [650/750 (87%)]\t training loss: 0.251952\n",
      "epoch: 2 [660/750 (88%)]\t training loss: 0.242585\n",
      "epoch: 2 [670/750 (89%)]\t training loss: 0.303505\n",
      "epoch: 2 [680/750 (91%)]\t training loss: 0.368679\n",
      "epoch: 2 [690/750 (92%)]\t training loss: 0.298182\n",
      "epoch: 2 [700/750 (93%)]\t training loss: 0.436255\n",
      "epoch: 2 [710/750 (95%)]\t training loss: 0.193519\n",
      "epoch: 2 [720/750 (96%)]\t training loss: 0.542156\n",
      "epoch: 2 [730/750 (97%)]\t training loss: 0.337776\n",
      "epoch: 2 [740/750 (99%)]\t training loss: 0.318047\n",
      "\n",
      "\n",
      "Validation...\n",
      "\n",
      "Test dataset: Overall Loss: 0.3503, Overall Accuracy: 10408/12000 (87%)\n",
      "\n",
      "epoch: 3 [0/750 (0%)]\t training loss: 0.280784\n",
      "epoch: 3 [10/750 (1%)]\t training loss: 0.342517\n",
      "epoch: 3 [20/750 (3%)]\t training loss: 0.248131\n",
      "epoch: 3 [30/750 (4%)]\t training loss: 0.346518\n",
      "epoch: 3 [40/750 (5%)]\t training loss: 0.371566\n",
      "epoch: 3 [50/750 (7%)]\t training loss: 0.226451\n",
      "epoch: 3 [60/750 (8%)]\t training loss: 0.267080\n",
      "epoch: 3 [70/750 (9%)]\t training loss: 0.281344\n",
      "epoch: 3 [80/750 (11%)]\t training loss: 0.266165\n",
      "epoch: 3 [90/750 (12%)]\t training loss: 0.206010\n",
      "epoch: 3 [100/750 (13%)]\t training loss: 0.311682\n",
      "epoch: 3 [110/750 (15%)]\t training loss: 0.371691\n",
      "epoch: 3 [120/750 (16%)]\t training loss: 0.279399\n",
      "epoch: 3 [130/750 (17%)]\t training loss: 0.452839\n",
      "epoch: 3 [140/750 (19%)]\t training loss: 0.308650\n",
      "epoch: 3 [150/750 (20%)]\t training loss: 0.411563\n",
      "epoch: 3 [160/750 (21%)]\t training loss: 0.450644\n",
      "epoch: 3 [170/750 (23%)]\t training loss: 0.275178\n",
      "epoch: 3 [180/750 (24%)]\t training loss: 0.329690\n",
      "epoch: 3 [190/750 (25%)]\t training loss: 0.270358\n",
      "epoch: 3 [200/750 (27%)]\t training loss: 0.325586\n",
      "epoch: 3 [210/750 (28%)]\t training loss: 0.417965\n",
      "epoch: 3 [220/750 (29%)]\t training loss: 0.164298\n",
      "epoch: 3 [230/750 (31%)]\t training loss: 0.348771\n",
      "epoch: 3 [240/750 (32%)]\t training loss: 0.396075\n",
      "epoch: 3 [250/750 (33%)]\t training loss: 0.198959\n",
      "epoch: 3 [260/750 (35%)]\t training loss: 0.270384\n",
      "epoch: 3 [270/750 (36%)]\t training loss: 0.242437\n",
      "epoch: 3 [280/750 (37%)]\t training loss: 0.364698\n",
      "epoch: 3 [290/750 (39%)]\t training loss: 0.411979\n",
      "epoch: 3 [300/750 (40%)]\t training loss: 0.336141\n",
      "epoch: 3 [310/750 (41%)]\t training loss: 0.319397\n",
      "epoch: 3 [320/750 (43%)]\t training loss: 0.273424\n",
      "epoch: 3 [330/750 (44%)]\t training loss: 0.261838\n",
      "epoch: 3 [340/750 (45%)]\t training loss: 0.183303\n",
      "epoch: 3 [350/750 (47%)]\t training loss: 0.156210\n",
      "epoch: 3 [360/750 (48%)]\t training loss: 0.332819\n",
      "epoch: 3 [370/750 (49%)]\t training loss: 0.302517\n",
      "epoch: 3 [380/750 (51%)]\t training loss: 0.233336\n",
      "epoch: 3 [390/750 (52%)]\t training loss: 0.211217\n",
      "epoch: 3 [400/750 (53%)]\t training loss: 0.144514\n",
      "epoch: 3 [410/750 (55%)]\t training loss: 0.446330\n",
      "epoch: 3 [420/750 (56%)]\t training loss: 0.268683\n",
      "epoch: 3 [430/750 (57%)]\t training loss: 0.221459\n",
      "epoch: 3 [440/750 (59%)]\t training loss: 0.254085\n",
      "epoch: 3 [450/750 (60%)]\t training loss: 0.148945\n",
      "epoch: 3 [460/750 (61%)]\t training loss: 0.256838\n",
      "epoch: 3 [470/750 (63%)]\t training loss: 0.231678\n",
      "epoch: 3 [480/750 (64%)]\t training loss: 0.176056\n",
      "epoch: 3 [490/750 (65%)]\t training loss: 0.291840\n",
      "epoch: 3 [500/750 (67%)]\t training loss: 0.154033\n",
      "epoch: 3 [510/750 (68%)]\t training loss: 0.319110\n",
      "epoch: 3 [520/750 (69%)]\t training loss: 0.222534\n",
      "epoch: 3 [530/750 (71%)]\t training loss: 0.346982\n",
      "epoch: 3 [540/750 (72%)]\t training loss: 0.259302\n",
      "epoch: 3 [550/750 (73%)]\t training loss: 0.305703\n",
      "epoch: 3 [560/750 (75%)]\t training loss: 0.208498\n",
      "epoch: 3 [570/750 (76%)]\t training loss: 0.289422\n",
      "epoch: 3 [580/750 (77%)]\t training loss: 0.282695\n",
      "epoch: 3 [590/750 (79%)]\t training loss: 0.296287\n",
      "epoch: 3 [600/750 (80%)]\t training loss: 0.161446\n",
      "epoch: 3 [610/750 (81%)]\t training loss: 0.309388\n",
      "epoch: 3 [620/750 (83%)]\t training loss: 0.272371\n",
      "epoch: 3 [630/750 (84%)]\t training loss: 0.208324\n",
      "epoch: 3 [640/750 (85%)]\t training loss: 0.592884\n",
      "epoch: 3 [650/750 (87%)]\t training loss: 0.208949\n",
      "epoch: 3 [660/750 (88%)]\t training loss: 0.279591\n",
      "epoch: 3 [670/750 (89%)]\t training loss: 0.286634\n",
      "epoch: 3 [680/750 (91%)]\t training loss: 0.248129\n",
      "epoch: 3 [690/750 (92%)]\t training loss: 0.260391\n",
      "epoch: 3 [700/750 (93%)]\t training loss: 0.211911\n",
      "epoch: 3 [710/750 (95%)]\t training loss: 0.187057\n",
      "epoch: 3 [720/750 (96%)]\t training loss: 0.330523\n",
      "epoch: 3 [730/750 (97%)]\t training loss: 0.320942\n",
      "epoch: 3 [740/750 (99%)]\t training loss: 0.298388\n",
      "\n",
      "\n",
      "Validation...\n",
      "\n",
      "Test dataset: Overall Loss: 0.3563, Overall Accuracy: 10520/12000 (88%)\n",
      "\n",
      "epoch: 4 [0/750 (0%)]\t training loss: 0.168252\n",
      "epoch: 4 [10/750 (1%)]\t training loss: 0.268345\n",
      "epoch: 4 [20/750 (3%)]\t training loss: 0.254127\n",
      "epoch: 4 [30/750 (4%)]\t training loss: 0.167802\n",
      "epoch: 4 [40/750 (5%)]\t training loss: 0.244611\n",
      "epoch: 4 [50/750 (7%)]\t training loss: 0.358709\n",
      "epoch: 4 [60/750 (8%)]\t training loss: 0.271999\n",
      "epoch: 4 [70/750 (9%)]\t training loss: 0.191597\n",
      "epoch: 4 [80/750 (11%)]\t training loss: 0.249198\n",
      "epoch: 4 [90/750 (12%)]\t training loss: 0.339670\n",
      "epoch: 4 [100/750 (13%)]\t training loss: 0.172511\n",
      "epoch: 4 [110/750 (15%)]\t training loss: 0.306517\n",
      "epoch: 4 [120/750 (16%)]\t training loss: 0.185428\n",
      "epoch: 4 [130/750 (17%)]\t training loss: 0.547447\n",
      "epoch: 4 [140/750 (19%)]\t training loss: 0.231466\n",
      "epoch: 4 [150/750 (20%)]\t training loss: 0.306686\n",
      "epoch: 4 [160/750 (21%)]\t training loss: 0.302258\n",
      "epoch: 4 [170/750 (23%)]\t training loss: 0.279468\n",
      "epoch: 4 [180/750 (24%)]\t training loss: 0.385702\n",
      "epoch: 4 [190/750 (25%)]\t training loss: 0.339850\n",
      "epoch: 4 [200/750 (27%)]\t training loss: 0.246078\n",
      "epoch: 4 [210/750 (28%)]\t training loss: 0.324823\n",
      "epoch: 4 [220/750 (29%)]\t training loss: 0.478043\n",
      "epoch: 4 [230/750 (31%)]\t training loss: 0.212453\n",
      "epoch: 4 [240/750 (32%)]\t training loss: 0.130212\n",
      "epoch: 4 [250/750 (33%)]\t training loss: 0.242765\n",
      "epoch: 4 [260/750 (35%)]\t training loss: 0.219039\n",
      "epoch: 4 [270/750 (36%)]\t training loss: 0.243509\n",
      "epoch: 4 [280/750 (37%)]\t training loss: 0.257547\n",
      "epoch: 4 [290/750 (39%)]\t training loss: 0.312063\n",
      "epoch: 4 [300/750 (40%)]\t training loss: 0.210561\n",
      "epoch: 4 [310/750 (41%)]\t training loss: 0.180267\n",
      "epoch: 4 [320/750 (43%)]\t training loss: 0.349469\n",
      "epoch: 4 [330/750 (44%)]\t training loss: 0.317048\n",
      "epoch: 4 [340/750 (45%)]\t training loss: 0.273186\n",
      "epoch: 4 [350/750 (47%)]\t training loss: 0.499227\n",
      "epoch: 4 [360/750 (48%)]\t training loss: 0.492841\n",
      "epoch: 4 [370/750 (49%)]\t training loss: 0.291084\n",
      "epoch: 4 [380/750 (51%)]\t training loss: 0.226829\n",
      "epoch: 4 [390/750 (52%)]\t training loss: 0.316982\n",
      "epoch: 4 [400/750 (53%)]\t training loss: 0.332695\n",
      "epoch: 4 [410/750 (55%)]\t training loss: 0.460870\n",
      "epoch: 4 [420/750 (56%)]\t training loss: 0.325691\n",
      "epoch: 4 [430/750 (57%)]\t training loss: 0.239911\n",
      "epoch: 4 [440/750 (59%)]\t training loss: 0.182760\n",
      "epoch: 4 [450/750 (60%)]\t training loss: 0.241791\n",
      "epoch: 4 [460/750 (61%)]\t training loss: 0.189466\n",
      "epoch: 4 [470/750 (63%)]\t training loss: 0.180007\n",
      "epoch: 4 [480/750 (64%)]\t training loss: 0.438760\n",
      "epoch: 4 [490/750 (65%)]\t training loss: 0.389652\n",
      "epoch: 4 [500/750 (67%)]\t training loss: 0.473324\n",
      "epoch: 4 [510/750 (68%)]\t training loss: 0.230528\n",
      "epoch: 4 [520/750 (69%)]\t training loss: 0.277840\n",
      "epoch: 4 [530/750 (71%)]\t training loss: 0.357478\n",
      "epoch: 4 [540/750 (72%)]\t training loss: 0.219106\n",
      "epoch: 4 [550/750 (73%)]\t training loss: 0.278187\n",
      "epoch: 4 [560/750 (75%)]\t training loss: 0.310184\n",
      "epoch: 4 [570/750 (76%)]\t training loss: 0.405328\n",
      "epoch: 4 [580/750 (77%)]\t training loss: 0.284605\n",
      "epoch: 4 [590/750 (79%)]\t training loss: 0.287846\n",
      "epoch: 4 [600/750 (80%)]\t training loss: 0.229842\n",
      "epoch: 4 [610/750 (81%)]\t training loss: 0.330191\n",
      "epoch: 4 [620/750 (83%)]\t training loss: 0.153005\n",
      "epoch: 4 [630/750 (84%)]\t training loss: 0.240226\n",
      "epoch: 4 [640/750 (85%)]\t training loss: 0.102553\n",
      "epoch: 4 [650/750 (87%)]\t training loss: 0.286372\n",
      "epoch: 4 [660/750 (88%)]\t training loss: 0.315818\n",
      "epoch: 4 [670/750 (89%)]\t training loss: 0.200531\n",
      "epoch: 4 [680/750 (91%)]\t training loss: 0.205102\n",
      "epoch: 4 [690/750 (92%)]\t training loss: 0.248294\n",
      "epoch: 4 [700/750 (93%)]\t training loss: 0.285730\n",
      "epoch: 4 [710/750 (95%)]\t training loss: 0.301706\n",
      "epoch: 4 [720/750 (96%)]\t training loss: 0.293867\n",
      "epoch: 4 [730/750 (97%)]\t training loss: 0.401023\n",
      "epoch: 4 [740/750 (99%)]\t training loss: 0.286445\n",
      "\n",
      "\n",
      "Validation...\n",
      "\n",
      "Test dataset: Overall Loss: 0.3451, Overall Accuracy: 10547/12000 (88%)\n",
      "\n",
      "epoch: 5 [0/750 (0%)]\t training loss: 0.265096\n",
      "epoch: 5 [10/750 (1%)]\t training loss: 0.130701\n",
      "epoch: 5 [20/750 (3%)]\t training loss: 0.142414\n",
      "epoch: 5 [30/750 (4%)]\t training loss: 0.225544\n",
      "epoch: 5 [40/750 (5%)]\t training loss: 0.211870\n",
      "epoch: 5 [50/750 (7%)]\t training loss: 0.169014\n",
      "epoch: 5 [60/750 (8%)]\t training loss: 0.266829\n",
      "epoch: 5 [70/750 (9%)]\t training loss: 0.156827\n",
      "epoch: 5 [80/750 (11%)]\t training loss: 0.297001\n",
      "epoch: 5 [90/750 (12%)]\t training loss: 0.207457\n",
      "epoch: 5 [100/750 (13%)]\t training loss: 0.210447\n",
      "epoch: 5 [110/750 (15%)]\t training loss: 0.247206\n",
      "epoch: 5 [120/750 (16%)]\t training loss: 0.316156\n",
      "epoch: 5 [130/750 (17%)]\t training loss: 0.389171\n",
      "epoch: 5 [140/750 (19%)]\t training loss: 0.268131\n",
      "epoch: 5 [150/750 (20%)]\t training loss: 0.330304\n",
      "epoch: 5 [160/750 (21%)]\t training loss: 0.179589\n",
      "epoch: 5 [170/750 (23%)]\t training loss: 0.307901\n",
      "epoch: 5 [180/750 (24%)]\t training loss: 0.281313\n",
      "epoch: 5 [190/750 (25%)]\t training loss: 0.115806\n",
      "epoch: 5 [200/750 (27%)]\t training loss: 0.200002\n",
      "epoch: 5 [210/750 (28%)]\t training loss: 0.386152\n",
      "epoch: 5 [220/750 (29%)]\t training loss: 0.399346\n",
      "epoch: 5 [230/750 (31%)]\t training loss: 0.377892\n",
      "epoch: 5 [240/750 (32%)]\t training loss: 0.155631\n",
      "epoch: 5 [250/750 (33%)]\t training loss: 0.432414\n",
      "epoch: 5 [260/750 (35%)]\t training loss: 0.354428\n",
      "epoch: 5 [270/750 (36%)]\t training loss: 0.232871\n",
      "epoch: 5 [280/750 (37%)]\t training loss: 0.189920\n",
      "epoch: 5 [290/750 (39%)]\t training loss: 0.157480\n",
      "epoch: 5 [300/750 (40%)]\t training loss: 0.282267\n",
      "epoch: 5 [310/750 (41%)]\t training loss: 0.320983\n",
      "epoch: 5 [320/750 (43%)]\t training loss: 0.260943\n",
      "epoch: 5 [330/750 (44%)]\t training loss: 0.305426\n",
      "epoch: 5 [340/750 (45%)]\t training loss: 0.297298\n",
      "epoch: 5 [350/750 (47%)]\t training loss: 0.391478\n",
      "epoch: 5 [360/750 (48%)]\t training loss: 0.365557\n",
      "epoch: 5 [370/750 (49%)]\t training loss: 0.251652\n",
      "epoch: 5 [380/750 (51%)]\t training loss: 0.168107\n",
      "epoch: 5 [390/750 (52%)]\t training loss: 0.388879\n",
      "epoch: 5 [400/750 (53%)]\t training loss: 0.128542\n",
      "epoch: 5 [410/750 (55%)]\t training loss: 0.126932\n",
      "epoch: 5 [420/750 (56%)]\t training loss: 0.069438\n",
      "epoch: 5 [430/750 (57%)]\t training loss: 0.436459\n",
      "epoch: 5 [440/750 (59%)]\t training loss: 0.237060\n",
      "epoch: 5 [450/750 (60%)]\t training loss: 0.117898\n",
      "epoch: 5 [460/750 (61%)]\t training loss: 0.280870\n",
      "epoch: 5 [470/750 (63%)]\t training loss: 0.370126\n",
      "epoch: 5 [480/750 (64%)]\t training loss: 0.225143\n",
      "epoch: 5 [490/750 (65%)]\t training loss: 0.146441\n",
      "epoch: 5 [500/750 (67%)]\t training loss: 0.333280\n",
      "epoch: 5 [510/750 (68%)]\t training loss: 0.280035\n",
      "epoch: 5 [520/750 (69%)]\t training loss: 0.351114\n",
      "epoch: 5 [530/750 (71%)]\t training loss: 0.252365\n",
      "epoch: 5 [540/750 (72%)]\t training loss: 0.465405\n",
      "epoch: 5 [550/750 (73%)]\t training loss: 0.280077\n",
      "epoch: 5 [560/750 (75%)]\t training loss: 0.373821\n",
      "epoch: 5 [570/750 (76%)]\t training loss: 0.108943\n",
      "epoch: 5 [580/750 (77%)]\t training loss: 0.107881\n",
      "epoch: 5 [590/750 (79%)]\t training loss: 0.354759\n",
      "epoch: 5 [600/750 (80%)]\t training loss: 0.186447\n",
      "epoch: 5 [610/750 (81%)]\t training loss: 0.301122\n",
      "epoch: 5 [620/750 (83%)]\t training loss: 0.173763\n",
      "epoch: 5 [630/750 (84%)]\t training loss: 0.144798\n",
      "epoch: 5 [640/750 (85%)]\t training loss: 0.338429\n",
      "epoch: 5 [650/750 (87%)]\t training loss: 0.370775\n",
      "epoch: 5 [660/750 (88%)]\t training loss: 0.172513\n",
      "epoch: 5 [670/750 (89%)]\t training loss: 0.498606\n",
      "epoch: 5 [680/750 (91%)]\t training loss: 0.148565\n",
      "epoch: 5 [690/750 (92%)]\t training loss: 0.206478\n",
      "epoch: 5 [700/750 (93%)]\t training loss: 0.215703\n",
      "epoch: 5 [710/750 (95%)]\t training loss: 0.196652\n",
      "epoch: 5 [720/750 (96%)]\t training loss: 0.404545\n",
      "epoch: 5 [730/750 (97%)]\t training loss: 0.302009\n",
      "epoch: 5 [740/750 (99%)]\t training loss: 0.172300\n",
      "\n",
      "\n",
      "Validation...\n",
      "\n",
      "Test dataset: Overall Loss: 0.3680, Overall Accuracy: 10463/12000 (87%)\n",
      "\n",
      "epoch: 6 [0/750 (0%)]\t training loss: 0.277458\n",
      "epoch: 6 [10/750 (1%)]\t training loss: 0.152333\n",
      "epoch: 6 [20/750 (3%)]\t training loss: 0.159110\n",
      "epoch: 6 [30/750 (4%)]\t training loss: 0.365061\n",
      "epoch: 6 [40/750 (5%)]\t training loss: 0.238158\n",
      "epoch: 6 [50/750 (7%)]\t training loss: 0.211160\n",
      "epoch: 6 [60/750 (8%)]\t training loss: 0.169424\n",
      "epoch: 6 [70/750 (9%)]\t training loss: 0.298796\n",
      "epoch: 6 [80/750 (11%)]\t training loss: 0.066842\n",
      "epoch: 6 [90/750 (12%)]\t training loss: 0.151132\n",
      "epoch: 6 [100/750 (13%)]\t training loss: 0.119866\n",
      "epoch: 6 [110/750 (15%)]\t training loss: 0.215422\n",
      "epoch: 6 [120/750 (16%)]\t training loss: 0.191937\n",
      "epoch: 6 [130/750 (17%)]\t training loss: 0.298066\n",
      "epoch: 6 [140/750 (19%)]\t training loss: 0.228368\n",
      "epoch: 6 [150/750 (20%)]\t training loss: 0.273919\n",
      "epoch: 6 [160/750 (21%)]\t training loss: 0.210249\n",
      "epoch: 6 [170/750 (23%)]\t training loss: 0.265070\n",
      "epoch: 6 [180/750 (24%)]\t training loss: 0.430429\n",
      "epoch: 6 [190/750 (25%)]\t training loss: 0.356498\n",
      "epoch: 6 [200/750 (27%)]\t training loss: 0.313344\n",
      "epoch: 6 [210/750 (28%)]\t training loss: 0.393953\n",
      "epoch: 6 [220/750 (29%)]\t training loss: 0.324797\n",
      "epoch: 6 [230/750 (31%)]\t training loss: 0.368810\n",
      "epoch: 6 [240/750 (32%)]\t training loss: 0.210378\n",
      "epoch: 6 [250/750 (33%)]\t training loss: 0.397923\n",
      "epoch: 6 [260/750 (35%)]\t training loss: 0.236296\n",
      "epoch: 6 [270/750 (36%)]\t training loss: 0.231340\n",
      "epoch: 6 [280/750 (37%)]\t training loss: 0.200551\n",
      "epoch: 6 [290/750 (39%)]\t training loss: 0.277149\n",
      "epoch: 6 [300/750 (40%)]\t training loss: 0.200190\n",
      "epoch: 6 [310/750 (41%)]\t training loss: 0.518437\n",
      "epoch: 6 [320/750 (43%)]\t training loss: 0.340226\n",
      "epoch: 6 [330/750 (44%)]\t training loss: 0.294982\n",
      "epoch: 6 [340/750 (45%)]\t training loss: 0.389679\n",
      "epoch: 6 [350/750 (47%)]\t training loss: 0.262731\n",
      "epoch: 6 [360/750 (48%)]\t training loss: 0.175181\n",
      "epoch: 6 [370/750 (49%)]\t training loss: 0.152678\n",
      "epoch: 6 [380/750 (51%)]\t training loss: 0.129865\n",
      "epoch: 6 [390/750 (52%)]\t training loss: 0.316000\n",
      "epoch: 6 [400/750 (53%)]\t training loss: 0.170116\n",
      "epoch: 6 [410/750 (55%)]\t training loss: 0.382749\n",
      "epoch: 6 [420/750 (56%)]\t training loss: 0.232022\n",
      "epoch: 6 [430/750 (57%)]\t training loss: 0.170616\n",
      "epoch: 6 [440/750 (59%)]\t training loss: 0.133120\n",
      "epoch: 6 [450/750 (60%)]\t training loss: 0.121189\n",
      "epoch: 6 [460/750 (61%)]\t training loss: 0.228734\n",
      "epoch: 6 [470/750 (63%)]\t training loss: 0.207455\n",
      "epoch: 6 [480/750 (64%)]\t training loss: 0.140857\n",
      "epoch: 6 [490/750 (65%)]\t training loss: 0.209543\n",
      "epoch: 6 [500/750 (67%)]\t training loss: 0.286554\n",
      "epoch: 6 [510/750 (68%)]\t training loss: 0.346471\n",
      "epoch: 6 [520/750 (69%)]\t training loss: 0.376212\n",
      "epoch: 6 [530/750 (71%)]\t training loss: 0.458320\n",
      "epoch: 6 [540/750 (72%)]\t training loss: 0.232411\n",
      "epoch: 6 [550/750 (73%)]\t training loss: 0.419965\n",
      "epoch: 6 [560/750 (75%)]\t training loss: 0.131483\n",
      "epoch: 6 [570/750 (76%)]\t training loss: 0.444723\n",
      "epoch: 6 [580/750 (77%)]\t training loss: 0.439585\n",
      "epoch: 6 [590/750 (79%)]\t training loss: 0.221987\n",
      "epoch: 6 [600/750 (80%)]\t training loss: 0.220124\n",
      "epoch: 6 [610/750 (81%)]\t training loss: 0.185206\n",
      "epoch: 6 [620/750 (83%)]\t training loss: 0.344256\n",
      "epoch: 6 [630/750 (84%)]\t training loss: 0.172920\n",
      "epoch: 6 [640/750 (85%)]\t training loss: 0.218535\n",
      "epoch: 6 [650/750 (87%)]\t training loss: 0.071690\n",
      "epoch: 6 [660/750 (88%)]\t training loss: 0.129273\n",
      "epoch: 6 [670/750 (89%)]\t training loss: 0.118825\n",
      "epoch: 6 [680/750 (91%)]\t training loss: 0.289879\n",
      "epoch: 6 [690/750 (92%)]\t training loss: 0.274960\n",
      "epoch: 6 [700/750 (93%)]\t training loss: 0.204883\n",
      "epoch: 6 [710/750 (95%)]\t training loss: 0.244767\n",
      "epoch: 6 [720/750 (96%)]\t training loss: 0.406406\n",
      "epoch: 6 [730/750 (97%)]\t training loss: 0.272159\n",
      "epoch: 6 [740/750 (99%)]\t training loss: 0.169403\n",
      "\n",
      "\n",
      "Validation...\n",
      "\n",
      "Test dataset: Overall Loss: 0.3455, Overall Accuracy: 10547/12000 (88%)\n",
      "\n",
      "epoch: 7 [0/750 (0%)]\t training loss: 0.211212\n",
      "epoch: 7 [10/750 (1%)]\t training loss: 0.155929\n",
      "epoch: 7 [20/750 (3%)]\t training loss: 0.083439\n",
      "epoch: 7 [30/750 (4%)]\t training loss: 0.282862\n",
      "epoch: 7 [40/750 (5%)]\t training loss: 0.324396\n",
      "epoch: 7 [50/750 (7%)]\t training loss: 0.279172\n",
      "epoch: 7 [60/750 (8%)]\t training loss: 0.183398\n",
      "epoch: 7 [70/750 (9%)]\t training loss: 0.110679\n",
      "epoch: 7 [80/750 (11%)]\t training loss: 0.152503\n",
      "epoch: 7 [90/750 (12%)]\t training loss: 0.149033\n",
      "epoch: 7 [100/750 (13%)]\t training loss: 0.287995\n",
      "epoch: 7 [110/750 (15%)]\t training loss: 0.123912\n",
      "epoch: 7 [120/750 (16%)]\t training loss: 0.248385\n",
      "epoch: 7 [130/750 (17%)]\t training loss: 0.222736\n",
      "epoch: 7 [140/750 (19%)]\t training loss: 0.429793\n",
      "epoch: 7 [150/750 (20%)]\t training loss: 0.304449\n",
      "epoch: 7 [160/750 (21%)]\t training loss: 0.326448\n",
      "epoch: 7 [170/750 (23%)]\t training loss: 0.161186\n",
      "epoch: 7 [180/750 (24%)]\t training loss: 0.186884\n",
      "epoch: 7 [190/750 (25%)]\t training loss: 0.218768\n",
      "epoch: 7 [200/750 (27%)]\t training loss: 0.173207\n",
      "epoch: 7 [210/750 (28%)]\t training loss: 0.207616\n",
      "epoch: 7 [220/750 (29%)]\t training loss: 0.192948\n",
      "epoch: 7 [230/750 (31%)]\t training loss: 0.159892\n",
      "epoch: 7 [240/750 (32%)]\t training loss: 0.169396\n",
      "epoch: 7 [250/750 (33%)]\t training loss: 0.169367\n",
      "epoch: 7 [260/750 (35%)]\t training loss: 0.185576\n",
      "epoch: 7 [270/750 (36%)]\t training loss: 0.217240\n",
      "epoch: 7 [280/750 (37%)]\t training loss: 0.225447\n",
      "epoch: 7 [290/750 (39%)]\t training loss: 0.244217\n",
      "epoch: 7 [300/750 (40%)]\t training loss: 0.405455\n",
      "epoch: 7 [310/750 (41%)]\t training loss: 0.246121\n",
      "epoch: 7 [320/750 (43%)]\t training loss: 0.222013\n",
      "epoch: 7 [330/750 (44%)]\t training loss: 0.091772\n",
      "epoch: 7 [340/750 (45%)]\t training loss: 0.155621\n",
      "epoch: 7 [350/750 (47%)]\t training loss: 0.459429\n",
      "epoch: 7 [360/750 (48%)]\t training loss: 0.157170\n",
      "epoch: 7 [370/750 (49%)]\t training loss: 0.145536\n",
      "epoch: 7 [380/750 (51%)]\t training loss: 0.166353\n",
      "epoch: 7 [390/750 (52%)]\t training loss: 0.395557\n",
      "epoch: 7 [400/750 (53%)]\t training loss: 0.323140\n",
      "epoch: 7 [410/750 (55%)]\t training loss: 0.215517\n",
      "epoch: 7 [420/750 (56%)]\t training loss: 0.333541\n",
      "epoch: 7 [430/750 (57%)]\t training loss: 0.225268\n",
      "epoch: 7 [440/750 (59%)]\t training loss: 0.179377\n",
      "epoch: 7 [450/750 (60%)]\t training loss: 0.260668\n",
      "epoch: 7 [460/750 (61%)]\t training loss: 0.260221\n",
      "epoch: 7 [470/750 (63%)]\t training loss: 0.248420\n",
      "epoch: 7 [480/750 (64%)]\t training loss: 0.163263\n",
      "epoch: 7 [490/750 (65%)]\t training loss: 0.259403\n",
      "epoch: 7 [500/750 (67%)]\t training loss: 0.170294\n",
      "epoch: 7 [510/750 (68%)]\t training loss: 0.173371\n",
      "epoch: 7 [520/750 (69%)]\t training loss: 0.188404\n",
      "epoch: 7 [530/750 (71%)]\t training loss: 0.154976\n",
      "epoch: 7 [540/750 (72%)]\t training loss: 0.318615\n",
      "epoch: 7 [550/750 (73%)]\t training loss: 0.213904\n",
      "epoch: 7 [560/750 (75%)]\t training loss: 0.426313\n",
      "epoch: 7 [570/750 (76%)]\t training loss: 0.216771\n",
      "epoch: 7 [580/750 (77%)]\t training loss: 0.122033\n",
      "epoch: 7 [590/750 (79%)]\t training loss: 0.442380\n",
      "epoch: 7 [600/750 (80%)]\t training loss: 0.298770\n",
      "epoch: 7 [610/750 (81%)]\t training loss: 0.337163\n",
      "epoch: 7 [620/750 (83%)]\t training loss: 0.098024\n",
      "epoch: 7 [630/750 (84%)]\t training loss: 0.187826\n",
      "epoch: 7 [640/750 (85%)]\t training loss: 0.310076\n",
      "epoch: 7 [650/750 (87%)]\t training loss: 0.190155\n",
      "epoch: 7 [660/750 (88%)]\t training loss: 0.235697\n",
      "epoch: 7 [670/750 (89%)]\t training loss: 0.511736\n",
      "epoch: 7 [680/750 (91%)]\t training loss: 0.354289\n",
      "epoch: 7 [690/750 (92%)]\t training loss: 0.389295\n",
      "epoch: 7 [700/750 (93%)]\t training loss: 0.410877\n",
      "epoch: 7 [710/750 (95%)]\t training loss: 0.152256\n",
      "epoch: 7 [720/750 (96%)]\t training loss: 0.467505\n",
      "epoch: 7 [730/750 (97%)]\t training loss: 0.325521\n",
      "epoch: 7 [740/750 (99%)]\t training loss: 0.156626\n",
      "\n",
      "\n",
      "Validation...\n",
      "\n",
      "Test dataset: Overall Loss: 0.3630, Overall Accuracy: 10591/12000 (88%)\n",
      "\n",
      "epoch: 8 [0/750 (0%)]\t training loss: 0.101171\n",
      "epoch: 8 [10/750 (1%)]\t training loss: 0.183410\n",
      "epoch: 8 [20/750 (3%)]\t training loss: 0.179192\n",
      "epoch: 8 [30/750 (4%)]\t training loss: 0.164952\n",
      "epoch: 8 [40/750 (5%)]\t training loss: 0.161038\n",
      "epoch: 8 [50/750 (7%)]\t training loss: 0.182944\n",
      "epoch: 8 [60/750 (8%)]\t training loss: 0.049877\n",
      "epoch: 8 [70/750 (9%)]\t training loss: 0.234939\n",
      "epoch: 8 [80/750 (11%)]\t training loss: 0.163566\n",
      "epoch: 8 [90/750 (12%)]\t training loss: 0.306404\n",
      "epoch: 8 [100/750 (13%)]\t training loss: 0.101086\n",
      "epoch: 8 [110/750 (15%)]\t training loss: 0.185382\n",
      "epoch: 8 [120/750 (16%)]\t training loss: 0.254704\n",
      "epoch: 8 [130/750 (17%)]\t training loss: 0.218552\n",
      "epoch: 8 [140/750 (19%)]\t training loss: 0.243739\n",
      "epoch: 8 [150/750 (20%)]\t training loss: 0.178509\n",
      "epoch: 8 [160/750 (21%)]\t training loss: 0.417973\n",
      "epoch: 8 [170/750 (23%)]\t training loss: 0.141628\n",
      "epoch: 8 [180/750 (24%)]\t training loss: 0.215610\n",
      "epoch: 8 [190/750 (25%)]\t training loss: 0.326567\n",
      "epoch: 8 [200/750 (27%)]\t training loss: 0.358247\n",
      "epoch: 8 [210/750 (28%)]\t training loss: 0.268688\n",
      "epoch: 8 [220/750 (29%)]\t training loss: 0.405592\n",
      "epoch: 8 [230/750 (31%)]\t training loss: 0.238078\n",
      "epoch: 8 [240/750 (32%)]\t training loss: 0.424625\n",
      "epoch: 8 [250/750 (33%)]\t training loss: 0.138412\n",
      "epoch: 8 [260/750 (35%)]\t training loss: 0.182494\n",
      "epoch: 8 [270/750 (36%)]\t training loss: 0.193265\n",
      "epoch: 8 [280/750 (37%)]\t training loss: 0.242462\n",
      "epoch: 8 [290/750 (39%)]\t training loss: 0.567674\n",
      "epoch: 8 [300/750 (40%)]\t training loss: 0.308573\n",
      "epoch: 8 [310/750 (41%)]\t training loss: 0.142273\n",
      "epoch: 8 [320/750 (43%)]\t training loss: 0.262226\n",
      "epoch: 8 [330/750 (44%)]\t training loss: 0.202084\n",
      "epoch: 8 [340/750 (45%)]\t training loss: 0.362193\n",
      "epoch: 8 [350/750 (47%)]\t training loss: 0.475245\n",
      "epoch: 8 [360/750 (48%)]\t training loss: 0.092611\n",
      "epoch: 8 [370/750 (49%)]\t training loss: 0.183530\n",
      "epoch: 8 [380/750 (51%)]\t training loss: 0.235871\n",
      "epoch: 8 [390/750 (52%)]\t training loss: 0.221105\n",
      "epoch: 8 [400/750 (53%)]\t training loss: 0.228060\n",
      "epoch: 8 [410/750 (55%)]\t training loss: 0.280345\n",
      "epoch: 8 [420/750 (56%)]\t training loss: 0.160281\n",
      "epoch: 8 [430/750 (57%)]\t training loss: 0.448167\n",
      "epoch: 8 [440/750 (59%)]\t training loss: 0.177385\n",
      "epoch: 8 [450/750 (60%)]\t training loss: 0.137968\n",
      "epoch: 8 [460/750 (61%)]\t training loss: 0.300137\n",
      "epoch: 8 [470/750 (63%)]\t training loss: 0.060615\n",
      "epoch: 8 [480/750 (64%)]\t training loss: 0.140607\n",
      "epoch: 8 [490/750 (65%)]\t training loss: 0.203592\n",
      "epoch: 8 [500/750 (67%)]\t training loss: 0.389804\n",
      "epoch: 8 [510/750 (68%)]\t training loss: 0.228315\n",
      "epoch: 8 [520/750 (69%)]\t training loss: 0.291083\n",
      "epoch: 8 [530/750 (71%)]\t training loss: 0.165855\n",
      "epoch: 8 [540/750 (72%)]\t training loss: 0.296365\n",
      "epoch: 8 [550/750 (73%)]\t training loss: 0.208903\n",
      "epoch: 8 [560/750 (75%)]\t training loss: 0.101519\n",
      "epoch: 8 [570/750 (76%)]\t training loss: 0.305756\n",
      "epoch: 8 [580/750 (77%)]\t training loss: 0.330228\n",
      "epoch: 8 [590/750 (79%)]\t training loss: 0.397892\n",
      "epoch: 8 [600/750 (80%)]\t training loss: 0.190299\n",
      "epoch: 8 [610/750 (81%)]\t training loss: 0.261626\n",
      "epoch: 8 [620/750 (83%)]\t training loss: 0.251341\n",
      "epoch: 8 [630/750 (84%)]\t training loss: 0.135122\n",
      "epoch: 8 [640/750 (85%)]\t training loss: 0.124208\n",
      "epoch: 8 [650/750 (87%)]\t training loss: 0.121419\n",
      "epoch: 8 [660/750 (88%)]\t training loss: 0.126148\n",
      "epoch: 8 [670/750 (89%)]\t training loss: 0.186561\n",
      "epoch: 8 [680/750 (91%)]\t training loss: 0.276994\n",
      "epoch: 8 [690/750 (92%)]\t training loss: 0.233915\n",
      "epoch: 8 [700/750 (93%)]\t training loss: 0.150215\n",
      "epoch: 8 [710/750 (95%)]\t training loss: 0.267127\n",
      "epoch: 8 [720/750 (96%)]\t training loss: 0.141047\n",
      "epoch: 8 [730/750 (97%)]\t training loss: 0.304765\n",
      "epoch: 8 [740/750 (99%)]\t training loss: 0.128268\n",
      "\n",
      "\n",
      "Validation...\n",
      "\n",
      "Test dataset: Overall Loss: 0.3929, Overall Accuracy: 10625/12000 (89%)\n",
      "\n",
      "epoch: 9 [0/750 (0%)]\t training loss: 0.210943\n",
      "epoch: 9 [10/750 (1%)]\t training loss: 0.130317\n",
      "epoch: 9 [20/750 (3%)]\t training loss: 0.115959\n",
      "epoch: 9 [30/750 (4%)]\t training loss: 0.168234\n",
      "epoch: 9 [40/750 (5%)]\t training loss: 0.351255\n",
      "epoch: 9 [50/750 (7%)]\t training loss: 0.116304\n",
      "epoch: 9 [60/750 (8%)]\t training loss: 0.383531\n",
      "epoch: 9 [70/750 (9%)]\t training loss: 0.072170\n",
      "epoch: 9 [80/750 (11%)]\t training loss: 0.201367\n",
      "epoch: 9 [90/750 (12%)]\t training loss: 0.271613\n",
      "epoch: 9 [100/750 (13%)]\t training loss: 0.211983\n",
      "epoch: 9 [110/750 (15%)]\t training loss: 0.282270\n",
      "epoch: 9 [120/750 (16%)]\t training loss: 0.201926\n",
      "epoch: 9 [130/750 (17%)]\t training loss: 0.174650\n",
      "epoch: 9 [140/750 (19%)]\t training loss: 0.464716\n",
      "epoch: 9 [150/750 (20%)]\t training loss: 0.131844\n",
      "epoch: 9 [160/750 (21%)]\t training loss: 0.107089\n",
      "epoch: 9 [170/750 (23%)]\t training loss: 0.298883\n",
      "epoch: 9 [180/750 (24%)]\t training loss: 0.237867\n",
      "epoch: 9 [190/750 (25%)]\t training loss: 0.095852\n",
      "epoch: 9 [200/750 (27%)]\t training loss: 0.095823\n",
      "epoch: 9 [210/750 (28%)]\t training loss: 0.048558\n",
      "epoch: 9 [220/750 (29%)]\t training loss: 0.264478\n",
      "epoch: 9 [230/750 (31%)]\t training loss: 0.158506\n",
      "epoch: 9 [240/750 (32%)]\t training loss: 0.242188\n",
      "epoch: 9 [250/750 (33%)]\t training loss: 0.240811\n",
      "epoch: 9 [260/750 (35%)]\t training loss: 0.167275\n",
      "epoch: 9 [270/750 (36%)]\t training loss: 0.298361\n",
      "epoch: 9 [280/750 (37%)]\t training loss: 0.211757\n",
      "epoch: 9 [290/750 (39%)]\t training loss: 0.239915\n",
      "epoch: 9 [300/750 (40%)]\t training loss: 0.414349\n",
      "epoch: 9 [310/750 (41%)]\t training loss: 0.206852\n",
      "epoch: 9 [320/750 (43%)]\t training loss: 0.180850\n",
      "epoch: 9 [330/750 (44%)]\t training loss: 0.114518\n",
      "epoch: 9 [340/750 (45%)]\t training loss: 0.257593\n",
      "epoch: 9 [350/750 (47%)]\t training loss: 0.245995\n",
      "epoch: 9 [360/750 (48%)]\t training loss: 0.272167\n",
      "epoch: 9 [370/750 (49%)]\t training loss: 0.088315\n",
      "epoch: 9 [380/750 (51%)]\t training loss: 0.341009\n",
      "epoch: 9 [390/750 (52%)]\t training loss: 0.123669\n",
      "epoch: 9 [400/750 (53%)]\t training loss: 0.198905\n",
      "epoch: 9 [410/750 (55%)]\t training loss: 0.456931\n",
      "epoch: 9 [420/750 (56%)]\t training loss: 0.404072\n",
      "epoch: 9 [430/750 (57%)]\t training loss: 0.202889\n",
      "epoch: 9 [440/750 (59%)]\t training loss: 0.267965\n",
      "epoch: 9 [450/750 (60%)]\t training loss: 0.228833\n",
      "epoch: 9 [460/750 (61%)]\t training loss: 0.396734\n",
      "epoch: 9 [470/750 (63%)]\t training loss: 0.275022\n",
      "epoch: 9 [480/750 (64%)]\t training loss: 0.401461\n",
      "epoch: 9 [490/750 (65%)]\t training loss: 0.260386\n",
      "epoch: 9 [500/750 (67%)]\t training loss: 0.268844\n",
      "epoch: 9 [510/750 (68%)]\t training loss: 0.374977\n",
      "epoch: 9 [520/750 (69%)]\t training loss: 0.381669\n",
      "epoch: 9 [530/750 (71%)]\t training loss: 0.236322\n",
      "epoch: 9 [540/750 (72%)]\t training loss: 0.475836\n",
      "epoch: 9 [550/750 (73%)]\t training loss: 0.177984\n",
      "epoch: 9 [560/750 (75%)]\t training loss: 0.155120\n",
      "epoch: 9 [570/750 (76%)]\t training loss: 0.154090\n",
      "epoch: 9 [580/750 (77%)]\t training loss: 0.240023\n",
      "epoch: 9 [590/750 (79%)]\t training loss: 0.488780\n",
      "epoch: 9 [600/750 (80%)]\t training loss: 0.156835\n",
      "epoch: 9 [610/750 (81%)]\t training loss: 0.257634\n",
      "epoch: 9 [620/750 (83%)]\t training loss: 0.135520\n",
      "epoch: 9 [630/750 (84%)]\t training loss: 0.267586\n",
      "epoch: 9 [640/750 (85%)]\t training loss: 0.250745\n",
      "epoch: 9 [650/750 (87%)]\t training loss: 0.125612\n",
      "epoch: 9 [660/750 (88%)]\t training loss: 0.331805\n",
      "epoch: 9 [670/750 (89%)]\t training loss: 0.264125\n",
      "epoch: 9 [680/750 (91%)]\t training loss: 0.261843\n",
      "epoch: 9 [690/750 (92%)]\t training loss: 0.387206\n",
      "epoch: 9 [700/750 (93%)]\t training loss: 0.261374\n",
      "epoch: 9 [710/750 (95%)]\t training loss: 0.451700\n",
      "epoch: 9 [720/750 (96%)]\t training loss: 0.359271\n",
      "epoch: 9 [730/750 (97%)]\t training loss: 0.129392\n",
      "epoch: 9 [740/750 (99%)]\t training loss: 0.111984\n",
      "\n",
      "\n",
      "Validation...\n",
      "\n",
      "Test dataset: Overall Loss: 0.3856, Overall Accuracy: 10519/12000 (88%)\n",
      "\n"
     ]
    }
   ],
   "source": [
    "model, train_losses, val_losses = train(\n",
    "    train_dataloader,\n",
    "    valid_dataloader,\n",
    "    \"Balanced_model.pth\",\n",
    "    ConvNet().to(device),\n",
    "    epochs=10,\n",
    ")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 472
    },
    "id": "BD7uq4obzC5W",
    "outputId": "c18dbf34-9c8a-4e31-999e-ea7bea5ace86"
   },
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 640x480 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "plt.plot(train_losses)\n",
    "plt.plot(val_losses)\n",
    "plt.xlabel(\"epoch\")\n",
    "plt.ylabel(\"loss\")\n",
    "plt.legend([\"Train\", \"Valid\"])\n",
    "plt.title(\"Model On Balanced Fashion MNIST - Train vs Valid Accuracy\")\n",
    "\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "DgzA0M_q5fr9"
   },
   "source": [
    "# Test Model"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/"
    },
    "id": "C2eZJUpU5fr9",
    "outputId": "c60a0945-b6dd-4efd-961c-e61342bd69d1"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "On vanilla Fashion MNIST dataset\n",
      "\n",
      "Test dataset: Overall Loss: 0.0008, Overall Accuracy: 8793/10000 (88%)\n",
      "\n",
      "test accuracy :  87.93\n"
     ]
    }
   ],
   "source": [
    "print(\"On vanilla Fashion MNIST dataset\")\n",
    "test_acc, _ = test(model, \"Balanced_model.pth\")\n",
    "print(\"test accuracy : \", str(test_acc))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "hYX59UtnAEFP"
   },
   "source": [
    "# Performance of the Model on Imbalanced MNIST Dataset"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "fj86TDRbAEFP"
   },
   "source": [
    "### Create Imbalanced Dataset"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 499
    },
    "id": "E_movNpJpmoH",
    "outputId": "147ab9a5-4af1-4b3d-bdd4-5147f98320a0"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Distribution of classes in Imbalanced dataset:\n",
      "Counter({0: 6000, 1: 3596, 2: 2156, 3: 1292, 4: 774, 5: 464, 6: 278, 7: 166, 8: 100, 9: 60})\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 640x480 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "train_size 11909\n",
      "val_size 2977\n"
     ]
    }
   ],
   "source": [
    "# generate imbalanced dataset\n",
    "num_classes = 10\n",
    "classe_labels = range(num_classes)\n",
    "torch.manual_seed(seed)\n",
    "sample_probs = torch.rand(num_classes)\n",
    "idx_to_remove = []\n",
    "\n",
    "\n",
    "def get_img_num_per_cls(data, num_classes=10, imb_type=\"exp\", imb_factor=0.01):\n",
    "    img_max = len(data) / num_classes\n",
    "    img_num_per_cls = []\n",
    "    if imb_type == \"exp\":\n",
    "        for cls_idx in range(num_classes):\n",
    "            num = img_max * (imb_factor ** (cls_idx / (num_classes - 1.0)))\n",
    "            img_num_per_cls.append(int(num))\n",
    "    return img_num_per_cls\n",
    "\n",
    "\n",
    "def gen_imbalanced_data(targets, img_num_per_cls):\n",
    "    targets_np = np.array(targets, dtype=np.int64)\n",
    "    classes = np.unique(targets_np)\n",
    "    num_per_cls_dict = dict()\n",
    "    idx_to_remove = []\n",
    "    for the_class, the_img_num in zip(classes, img_num_per_cls):\n",
    "        num_per_cls_dict[the_class] = the_img_num\n",
    "        idx = np.where(targets_np == the_class)[0]\n",
    "        np.random.shuffle(idx)\n",
    "        selec_idx = idx[:the_img_num]\n",
    "        idx_to_remove.extend(list(np.setdiff1d(idx, selec_idx)))\n",
    "    return idx_to_remove\n",
    "\n",
    "\n",
    "img_num_per_cls = get_img_num_per_cls(trainset)\n",
    "idx_to_remove = gen_imbalanced_data(trainset.targets, img_num_per_cls)\n",
    "\n",
    "# print('idx_to_remove: ', idx_to_remove)\n",
    "imbalanced_train_dataset = copy.deepcopy(trainset)\n",
    "imbalanced_train_dataset.targets = np.delete(trainset.targets, idx_to_remove, axis=0)\n",
    "imbalanced_train_dataset.data = np.delete(trainset.data, idx_to_remove, axis=0)\n",
    "\n",
    "print(\"Distribution of classes in Imbalanced dataset:\")\n",
    "print(Counter(imbalanced_train_dataset.targets.numpy()))\n",
    "fig, ax = plt.subplots()\n",
    "_, counts = np.unique(imbalanced_train_dataset.targets, return_counts=True)\n",
    "ax.bar(range(10), counts)\n",
    "ax.set_xticks(range(10))\n",
    "plt.show()\n",
    "\n",
    "# Preparing for validaion test\n",
    "indices = list(range(len(imbalanced_train_dataset)))\n",
    "np.random.seed(seed)\n",
    "np.random.shuffle(indices)\n",
    "# to get 20% of the train set\n",
    "split = int(np.floor(0.2 * len(imbalanced_train_dataset)))\n",
    "train_sample = SubsetRandomSampler(indices[split:])\n",
    "valid_sample = SubsetRandomSampler(indices[:split])\n",
    "\n",
    "# Data Loader\n",
    "imbalanced_train_loader = torch.utils.data.DataLoader(\n",
    "    imbalanced_train_dataset, sampler=train_sample, batch_size=64\n",
    ")\n",
    "imbalanced_valid_loader = torch.utils.data.DataLoader(\n",
    "    imbalanced_train_dataset, sampler=valid_sample, batch_size=64\n",
    ")\n",
    "print(\"train_size\", len(train_sample))\n",
    "print(\"val_size\", len(valid_sample))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "b8srim2AAEFR"
   },
   "source": [
    "### train model on imbalanced dataset"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/"
    },
    "id": "C1nPVameAEFR",
    "outputId": "8eca9cac-8e93-4981-977c-7f676d2fa839",
    "scrolled": true
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "epoch: 0 [0/187 (0%)]\t training loss: 2.263006\n",
      "epoch: 0 [10/187 (5%)]\t training loss: 0.990236\n",
      "epoch: 0 [20/187 (11%)]\t training loss: 0.743031\n",
      "epoch: 0 [30/187 (16%)]\t training loss: 1.187697\n",
      "epoch: 0 [40/187 (21%)]\t training loss: 0.695566\n",
      "epoch: 0 [50/187 (27%)]\t training loss: 0.820280\n",
      "epoch: 0 [60/187 (32%)]\t training loss: 0.486129\n",
      "epoch: 0 [70/187 (37%)]\t training loss: 0.562746\n",
      "epoch: 0 [80/187 (43%)]\t training loss: 0.656181\n",
      "epoch: 0 [90/187 (48%)]\t training loss: 0.490844\n",
      "epoch: 0 [100/187 (53%)]\t training loss: 0.632755\n",
      "epoch: 0 [110/187 (59%)]\t training loss: 0.504913\n",
      "epoch: 0 [120/187 (64%)]\t training loss: 0.656998\n",
      "epoch: 0 [130/187 (70%)]\t training loss: 0.662536\n",
      "epoch: 0 [140/187 (75%)]\t training loss: 0.418396\n",
      "epoch: 0 [150/187 (80%)]\t training loss: 0.536673\n",
      "epoch: 0 [160/187 (86%)]\t training loss: 0.488270\n",
      "epoch: 0 [170/187 (91%)]\t training loss: 0.312365\n",
      "epoch: 0 [180/187 (96%)]\t training loss: 0.519255\n",
      "\n",
      "\n",
      "Validation...\n",
      "\n",
      "Test dataset: Overall Loss: 0.3992, Overall Accuracy: 2578/2977 (87%)\n",
      "\n",
      "epoch: 1 [0/187 (0%)]\t training loss: 0.310709\n",
      "epoch: 1 [10/187 (5%)]\t training loss: 0.310701\n",
      "epoch: 1 [20/187 (11%)]\t training loss: 0.325347\n",
      "epoch: 1 [30/187 (16%)]\t training loss: 0.285951\n",
      "epoch: 1 [40/187 (21%)]\t training loss: 0.222298\n",
      "epoch: 1 [50/187 (27%)]\t training loss: 0.359832\n",
      "epoch: 1 [60/187 (32%)]\t training loss: 0.539927\n",
      "epoch: 1 [70/187 (37%)]\t training loss: 0.557183\n",
      "epoch: 1 [80/187 (43%)]\t training loss: 0.439101\n",
      "epoch: 1 [90/187 (48%)]\t training loss: 0.563409\n",
      "epoch: 1 [100/187 (53%)]\t training loss: 0.315620\n",
      "epoch: 1 [110/187 (59%)]\t training loss: 0.345369\n",
      "epoch: 1 [120/187 (64%)]\t training loss: 0.264890\n",
      "epoch: 1 [130/187 (70%)]\t training loss: 0.344887\n",
      "epoch: 1 [140/187 (75%)]\t training loss: 0.391256\n",
      "epoch: 1 [150/187 (80%)]\t training loss: 0.652167\n",
      "epoch: 1 [160/187 (86%)]\t training loss: 0.415677\n",
      "epoch: 1 [170/187 (91%)]\t training loss: 0.403458\n",
      "epoch: 1 [180/187 (96%)]\t training loss: 0.310540\n",
      "\n",
      "\n",
      "Validation...\n",
      "\n",
      "Test dataset: Overall Loss: 0.3210, Overall Accuracy: 2670/2977 (90%)\n",
      "\n",
      "epoch: 2 [0/187 (0%)]\t training loss: 0.228604\n",
      "epoch: 2 [10/187 (5%)]\t training loss: 0.276193\n",
      "epoch: 2 [20/187 (11%)]\t training loss: 0.161193\n",
      "epoch: 2 [30/187 (16%)]\t training loss: 0.326000\n",
      "epoch: 2 [40/187 (21%)]\t training loss: 0.255307\n",
      "epoch: 2 [50/187 (27%)]\t training loss: 0.472897\n",
      "epoch: 2 [60/187 (32%)]\t training loss: 0.167819\n",
      "epoch: 2 [70/187 (37%)]\t training loss: 0.256520\n",
      "epoch: 2 [80/187 (43%)]\t training loss: 0.255254\n",
      "epoch: 2 [90/187 (48%)]\t training loss: 0.231653\n",
      "epoch: 2 [100/187 (53%)]\t training loss: 0.253273\n",
      "epoch: 2 [110/187 (59%)]\t training loss: 0.381546\n",
      "epoch: 2 [120/187 (64%)]\t training loss: 0.145969\n",
      "epoch: 2 [130/187 (70%)]\t training loss: 0.295347\n",
      "epoch: 2 [140/187 (75%)]\t training loss: 0.257117\n",
      "epoch: 2 [150/187 (80%)]\t training loss: 0.358717\n",
      "epoch: 2 [160/187 (86%)]\t training loss: 0.186913\n",
      "epoch: 2 [170/187 (91%)]\t training loss: 0.332921\n",
      "epoch: 2 [180/187 (96%)]\t training loss: 0.379282\n",
      "\n",
      "\n",
      "Validation...\n",
      "\n",
      "Test dataset: Overall Loss: 0.3007, Overall Accuracy: 2668/2977 (90%)\n",
      "\n",
      "epoch: 3 [0/187 (0%)]\t training loss: 0.354696\n",
      "epoch: 3 [10/187 (5%)]\t training loss: 0.312253\n",
      "epoch: 3 [20/187 (11%)]\t training loss: 0.221062\n",
      "epoch: 3 [30/187 (16%)]\t training loss: 0.233304\n",
      "epoch: 3 [40/187 (21%)]\t training loss: 0.300354\n",
      "epoch: 3 [50/187 (27%)]\t training loss: 0.223658\n",
      "epoch: 3 [60/187 (32%)]\t training loss: 0.332675\n",
      "epoch: 3 [70/187 (37%)]\t training loss: 0.126692\n",
      "epoch: 3 [80/187 (43%)]\t training loss: 0.258188\n",
      "epoch: 3 [90/187 (48%)]\t training loss: 0.135763\n",
      "epoch: 3 [100/187 (53%)]\t training loss: 0.282072\n",
      "epoch: 3 [110/187 (59%)]\t training loss: 0.106740\n",
      "epoch: 3 [120/187 (64%)]\t training loss: 0.199206\n",
      "epoch: 3 [130/187 (70%)]\t training loss: 0.171635\n",
      "epoch: 3 [140/187 (75%)]\t training loss: 0.298631\n",
      "epoch: 3 [150/187 (80%)]\t training loss: 0.318927\n",
      "epoch: 3 [160/187 (86%)]\t training loss: 0.205414\n",
      "epoch: 3 [170/187 (91%)]\t training loss: 0.179457\n",
      "epoch: 3 [180/187 (96%)]\t training loss: 0.376769\n",
      "\n",
      "\n",
      "Validation...\n",
      "\n",
      "Test dataset: Overall Loss: 0.2936, Overall Accuracy: 2680/2977 (90%)\n",
      "\n",
      "epoch: 4 [0/187 (0%)]\t training loss: 0.169301\n",
      "epoch: 4 [10/187 (5%)]\t training loss: 0.155881\n",
      "epoch: 4 [20/187 (11%)]\t training loss: 0.156367\n",
      "epoch: 4 [30/187 (16%)]\t training loss: 0.362274\n",
      "epoch: 4 [40/187 (21%)]\t training loss: 0.331696\n",
      "epoch: 4 [50/187 (27%)]\t training loss: 0.204212\n",
      "epoch: 4 [60/187 (32%)]\t training loss: 0.123072\n",
      "epoch: 4 [70/187 (37%)]\t training loss: 0.285571\n",
      "epoch: 4 [80/187 (43%)]\t training loss: 0.300960\n",
      "epoch: 4 [90/187 (48%)]\t training loss: 0.136560\n",
      "epoch: 4 [100/187 (53%)]\t training loss: 0.257613\n",
      "epoch: 4 [110/187 (59%)]\t training loss: 0.260574\n",
      "epoch: 4 [120/187 (64%)]\t training loss: 0.138100\n",
      "epoch: 4 [130/187 (70%)]\t training loss: 0.154143\n",
      "epoch: 4 [140/187 (75%)]\t training loss: 0.268761\n",
      "epoch: 4 [150/187 (80%)]\t training loss: 0.155796\n",
      "epoch: 4 [160/187 (86%)]\t training loss: 0.205906\n",
      "epoch: 4 [170/187 (91%)]\t training loss: 0.241351\n",
      "epoch: 4 [180/187 (96%)]\t training loss: 0.090540\n",
      "\n",
      "\n",
      "Validation...\n",
      "\n",
      "Test dataset: Overall Loss: 0.3414, Overall Accuracy: 2666/2977 (90%)\n",
      "\n",
      "epoch: 5 [0/187 (0%)]\t training loss: 0.213037\n",
      "epoch: 5 [10/187 (5%)]\t training loss: 0.196893\n",
      "epoch: 5 [20/187 (11%)]\t training loss: 0.210013\n",
      "epoch: 5 [30/187 (16%)]\t training loss: 0.084482\n",
      "epoch: 5 [40/187 (21%)]\t training loss: 0.220533\n",
      "epoch: 5 [50/187 (27%)]\t training loss: 0.240825\n",
      "epoch: 5 [60/187 (32%)]\t training loss: 0.213057\n",
      "epoch: 5 [70/187 (37%)]\t training loss: 0.114013\n",
      "epoch: 5 [80/187 (43%)]\t training loss: 0.313284\n",
      "epoch: 5 [90/187 (48%)]\t training loss: 0.056310\n",
      "epoch: 5 [100/187 (53%)]\t training loss: 0.239813\n",
      "epoch: 5 [110/187 (59%)]\t training loss: 0.165749\n",
      "epoch: 5 [120/187 (64%)]\t training loss: 0.292970\n",
      "epoch: 5 [130/187 (70%)]\t training loss: 0.177748\n",
      "epoch: 5 [140/187 (75%)]\t training loss: 0.170292\n",
      "epoch: 5 [150/187 (80%)]\t training loss: 0.234613\n",
      "epoch: 5 [160/187 (86%)]\t training loss: 0.184013\n",
      "epoch: 5 [170/187 (91%)]\t training loss: 0.212584\n",
      "epoch: 5 [180/187 (96%)]\t training loss: 0.100186\n",
      "\n",
      "\n",
      "Validation...\n",
      "\n",
      "Test dataset: Overall Loss: 0.2847, Overall Accuracy: 2722/2977 (91%)\n",
      "\n",
      "epoch: 6 [0/187 (0%)]\t training loss: 0.075663\n",
      "epoch: 6 [10/187 (5%)]\t training loss: 0.094202\n",
      "epoch: 6 [20/187 (11%)]\t training loss: 0.171221\n",
      "epoch: 6 [30/187 (16%)]\t training loss: 0.193364\n",
      "epoch: 6 [40/187 (21%)]\t training loss: 0.190653\n",
      "epoch: 6 [50/187 (27%)]\t training loss: 0.306783\n",
      "epoch: 6 [60/187 (32%)]\t training loss: 0.150333\n",
      "epoch: 6 [70/187 (37%)]\t training loss: 0.176433\n",
      "epoch: 6 [80/187 (43%)]\t training loss: 0.148497\n",
      "epoch: 6 [90/187 (48%)]\t training loss: 0.053457\n",
      "epoch: 6 [100/187 (53%)]\t training loss: 0.251082\n",
      "epoch: 6 [110/187 (59%)]\t training loss: 0.258107\n",
      "epoch: 6 [120/187 (64%)]\t training loss: 0.045798\n",
      "epoch: 6 [130/187 (70%)]\t training loss: 0.370636\n",
      "epoch: 6 [140/187 (75%)]\t training loss: 0.222691\n",
      "epoch: 6 [150/187 (80%)]\t training loss: 0.180686\n",
      "epoch: 6 [160/187 (86%)]\t training loss: 0.144134\n",
      "epoch: 6 [170/187 (91%)]\t training loss: 0.142895\n",
      "epoch: 6 [180/187 (96%)]\t training loss: 0.184444\n",
      "\n",
      "\n",
      "Validation...\n",
      "\n",
      "Test dataset: Overall Loss: 0.3023, Overall Accuracy: 2712/2977 (91%)\n",
      "\n",
      "epoch: 7 [0/187 (0%)]\t training loss: 0.062058\n",
      "epoch: 7 [10/187 (5%)]\t training loss: 0.072247\n",
      "epoch: 7 [20/187 (11%)]\t training loss: 0.151536\n",
      "epoch: 7 [30/187 (16%)]\t training loss: 0.069928\n",
      "epoch: 7 [40/187 (21%)]\t training loss: 0.116037\n",
      "epoch: 7 [50/187 (27%)]\t training loss: 0.109215\n",
      "epoch: 7 [60/187 (32%)]\t training loss: 0.135721\n",
      "epoch: 7 [70/187 (37%)]\t training loss: 0.189460\n",
      "epoch: 7 [80/187 (43%)]\t training loss: 0.088773\n",
      "epoch: 7 [90/187 (48%)]\t training loss: 0.164877\n",
      "epoch: 7 [100/187 (53%)]\t training loss: 0.180052\n",
      "epoch: 7 [110/187 (59%)]\t training loss: 0.097745\n",
      "epoch: 7 [120/187 (64%)]\t training loss: 0.192597\n",
      "epoch: 7 [130/187 (70%)]\t training loss: 0.087174\n",
      "epoch: 7 [140/187 (75%)]\t training loss: 0.068681\n",
      "epoch: 7 [150/187 (80%)]\t training loss: 0.141160\n",
      "epoch: 7 [160/187 (86%)]\t training loss: 0.105745\n",
      "epoch: 7 [170/187 (91%)]\t training loss: 0.211784\n",
      "epoch: 7 [180/187 (96%)]\t training loss: 0.252095\n",
      "\n",
      "\n",
      "Validation...\n",
      "\n",
      "Test dataset: Overall Loss: 0.3336, Overall Accuracy: 2693/2977 (90%)\n",
      "\n",
      "epoch: 8 [0/187 (0%)]\t training loss: 0.471655\n",
      "epoch: 8 [10/187 (5%)]\t training loss: 0.100771\n",
      "epoch: 8 [20/187 (11%)]\t training loss: 0.162032\n",
      "epoch: 8 [30/187 (16%)]\t training loss: 0.197304\n",
      "epoch: 8 [40/187 (21%)]\t training loss: 0.055143\n",
      "epoch: 8 [50/187 (27%)]\t training loss: 0.060339\n",
      "epoch: 8 [60/187 (32%)]\t training loss: 0.135262\n",
      "epoch: 8 [70/187 (37%)]\t training loss: 0.162285\n",
      "epoch: 8 [80/187 (43%)]\t training loss: 0.105294\n",
      "epoch: 8 [90/187 (48%)]\t training loss: 0.138803\n",
      "epoch: 8 [100/187 (53%)]\t training loss: 0.068594\n",
      "epoch: 8 [110/187 (59%)]\t training loss: 0.082662\n",
      "epoch: 8 [120/187 (64%)]\t training loss: 0.071885\n",
      "epoch: 8 [130/187 (70%)]\t training loss: 0.152375\n",
      "epoch: 8 [140/187 (75%)]\t training loss: 0.165636\n",
      "epoch: 8 [150/187 (80%)]\t training loss: 0.130767\n",
      "epoch: 8 [160/187 (86%)]\t training loss: 0.278040\n",
      "epoch: 8 [170/187 (91%)]\t training loss: 0.112378\n",
      "epoch: 8 [180/187 (96%)]\t training loss: 0.177666\n",
      "\n",
      "\n",
      "Validation...\n",
      "\n",
      "Test dataset: Overall Loss: 0.3165, Overall Accuracy: 2693/2977 (90%)\n",
      "\n",
      "epoch: 9 [0/187 (0%)]\t training loss: 0.093287\n",
      "epoch: 9 [10/187 (5%)]\t training loss: 0.124819\n",
      "epoch: 9 [20/187 (11%)]\t training loss: 0.062942\n",
      "epoch: 9 [30/187 (16%)]\t training loss: 0.041723\n",
      "epoch: 9 [40/187 (21%)]\t training loss: 0.115408\n",
      "epoch: 9 [50/187 (27%)]\t training loss: 0.192590\n",
      "epoch: 9 [60/187 (32%)]\t training loss: 0.176829\n",
      "epoch: 9 [70/187 (37%)]\t training loss: 0.082727\n",
      "epoch: 9 [80/187 (43%)]\t training loss: 0.162532\n",
      "epoch: 9 [90/187 (48%)]\t training loss: 0.037430\n",
      "epoch: 9 [100/187 (53%)]\t training loss: 0.065163\n",
      "epoch: 9 [110/187 (59%)]\t training loss: 0.124910\n",
      "epoch: 9 [120/187 (64%)]\t training loss: 0.179139\n",
      "epoch: 9 [130/187 (70%)]\t training loss: 0.130350\n",
      "epoch: 9 [140/187 (75%)]\t training loss: 0.038987\n",
      "epoch: 9 [150/187 (80%)]\t training loss: 0.146918\n",
      "epoch: 9 [160/187 (86%)]\t training loss: 0.163619\n",
      "epoch: 9 [170/187 (91%)]\t training loss: 0.114321\n",
      "epoch: 9 [180/187 (96%)]\t training loss: 0.189435\n",
      "\n",
      "\n",
      "Validation...\n",
      "\n",
      "Test dataset: Overall Loss: 0.3261, Overall Accuracy: 2700/2977 (91%)\n",
      "\n",
      "epoch: 10 [0/187 (0%)]\t training loss: 0.075326\n",
      "epoch: 10 [10/187 (5%)]\t training loss: 0.074441\n",
      "epoch: 10 [20/187 (11%)]\t training loss: 0.114365\n",
      "epoch: 10 [30/187 (16%)]\t training loss: 0.027884\n",
      "epoch: 10 [40/187 (21%)]\t training loss: 0.036350\n",
      "epoch: 10 [50/187 (27%)]\t training loss: 0.087657\n",
      "epoch: 10 [60/187 (32%)]\t training loss: 0.125573\n",
      "epoch: 10 [70/187 (37%)]\t training loss: 0.102757\n",
      "epoch: 10 [80/187 (43%)]\t training loss: 0.284990\n",
      "epoch: 10 [90/187 (48%)]\t training loss: 0.133265\n",
      "epoch: 10 [100/187 (53%)]\t training loss: 0.061207\n",
      "epoch: 10 [110/187 (59%)]\t training loss: 0.242402\n",
      "epoch: 10 [120/187 (64%)]\t training loss: 0.355814\n",
      "epoch: 10 [130/187 (70%)]\t training loss: 0.226757\n",
      "epoch: 10 [140/187 (75%)]\t training loss: 0.025546\n",
      "epoch: 10 [150/187 (80%)]\t training loss: 0.351664\n",
      "epoch: 10 [160/187 (86%)]\t training loss: 0.212721\n",
      "epoch: 10 [170/187 (91%)]\t training loss: 0.133155\n",
      "epoch: 10 [180/187 (96%)]\t training loss: 0.205467\n",
      "\n",
      "\n",
      "Validation...\n",
      "\n",
      "Test dataset: Overall Loss: 0.3606, Overall Accuracy: 2685/2977 (90%)\n",
      "\n",
      "epoch: 11 [0/187 (0%)]\t training loss: 0.123731\n",
      "epoch: 11 [10/187 (5%)]\t training loss: 0.048266\n",
      "epoch: 11 [20/187 (11%)]\t training loss: 0.183213\n",
      "epoch: 11 [30/187 (16%)]\t training loss: 0.075466\n",
      "epoch: 11 [40/187 (21%)]\t training loss: 0.068144\n",
      "epoch: 11 [50/187 (27%)]\t training loss: 0.199027\n",
      "epoch: 11 [60/187 (32%)]\t training loss: 0.044802\n",
      "epoch: 11 [70/187 (37%)]\t training loss: 0.039359\n",
      "epoch: 11 [80/187 (43%)]\t training loss: 0.098014\n",
      "epoch: 11 [90/187 (48%)]\t training loss: 0.125973\n",
      "epoch: 11 [100/187 (53%)]\t training loss: 0.093279\n",
      "epoch: 11 [110/187 (59%)]\t training loss: 0.065008\n",
      "epoch: 11 [120/187 (64%)]\t training loss: 0.085306\n",
      "epoch: 11 [130/187 (70%)]\t training loss: 0.095564\n",
      "epoch: 11 [140/187 (75%)]\t training loss: 0.310969\n",
      "epoch: 11 [150/187 (80%)]\t training loss: 0.242927\n",
      "epoch: 11 [160/187 (86%)]\t training loss: 0.080879\n",
      "epoch: 11 [170/187 (91%)]\t training loss: 0.171638\n",
      "epoch: 11 [180/187 (96%)]\t training loss: 0.071879\n",
      "\n",
      "\n",
      "Validation...\n",
      "\n",
      "Test dataset: Overall Loss: 0.4200, Overall Accuracy: 2676/2977 (90%)\n",
      "\n",
      "epoch: 12 [0/187 (0%)]\t training loss: 0.116219\n",
      "epoch: 12 [10/187 (5%)]\t training loss: 0.075314\n",
      "epoch: 12 [20/187 (11%)]\t training loss: 0.061694\n",
      "epoch: 12 [30/187 (16%)]\t training loss: 0.052843\n",
      "epoch: 12 [40/187 (21%)]\t training loss: 0.045340\n",
      "epoch: 12 [50/187 (27%)]\t training loss: 0.020746\n",
      "epoch: 12 [60/187 (32%)]\t training loss: 0.068243\n",
      "epoch: 12 [70/187 (37%)]\t training loss: 0.285361\n",
      "epoch: 12 [80/187 (43%)]\t training loss: 0.099096\n",
      "epoch: 12 [90/187 (48%)]\t training loss: 0.354011\n",
      "epoch: 12 [100/187 (53%)]\t training loss: 0.147830\n",
      "epoch: 12 [110/187 (59%)]\t training loss: 0.264995\n",
      "epoch: 12 [120/187 (64%)]\t training loss: 0.094562\n",
      "epoch: 12 [130/187 (70%)]\t training loss: 0.152114\n",
      "epoch: 12 [140/187 (75%)]\t training loss: 0.175839\n",
      "epoch: 12 [150/187 (80%)]\t training loss: 0.020158\n",
      "epoch: 12 [160/187 (86%)]\t training loss: 0.147054\n",
      "epoch: 12 [170/187 (91%)]\t training loss: 0.071933\n",
      "epoch: 12 [180/187 (96%)]\t training loss: 0.072600\n",
      "\n",
      "\n",
      "Validation...\n",
      "\n",
      "Test dataset: Overall Loss: 0.4226, Overall Accuracy: 2722/2977 (91%)\n",
      "\n",
      "epoch: 13 [0/187 (0%)]\t training loss: 0.029204\n",
      "epoch: 13 [10/187 (5%)]\t training loss: 0.070826\n",
      "epoch: 13 [20/187 (11%)]\t training loss: 0.065796\n",
      "epoch: 13 [30/187 (16%)]\t training loss: 0.115146\n",
      "epoch: 13 [40/187 (21%)]\t training loss: 0.061208\n",
      "epoch: 13 [50/187 (27%)]\t training loss: 0.137650\n",
      "epoch: 13 [60/187 (32%)]\t training loss: 0.237439\n",
      "epoch: 13 [70/187 (37%)]\t training loss: 0.088833\n",
      "epoch: 13 [80/187 (43%)]\t training loss: 0.090658\n",
      "epoch: 13 [90/187 (48%)]\t training loss: 0.027355\n",
      "epoch: 13 [100/187 (53%)]\t training loss: 0.035533\n",
      "epoch: 13 [110/187 (59%)]\t training loss: 0.015059\n",
      "epoch: 13 [120/187 (64%)]\t training loss: 0.021017\n",
      "epoch: 13 [130/187 (70%)]\t training loss: 0.083342\n",
      "epoch: 13 [140/187 (75%)]\t training loss: 0.035263\n",
      "epoch: 13 [150/187 (80%)]\t training loss: 0.162512\n",
      "epoch: 13 [160/187 (86%)]\t training loss: 0.126569\n",
      "epoch: 13 [170/187 (91%)]\t training loss: 0.126345\n",
      "epoch: 13 [180/187 (96%)]\t training loss: 0.110910\n",
      "\n",
      "\n",
      "Validation...\n",
      "\n",
      "Test dataset: Overall Loss: 0.4026, Overall Accuracy: 2679/2977 (90%)\n",
      "\n",
      "epoch: 14 [0/187 (0%)]\t training loss: 0.041748\n",
      "epoch: 14 [10/187 (5%)]\t training loss: 0.073934\n",
      "epoch: 14 [20/187 (11%)]\t training loss: 0.037405\n",
      "epoch: 14 [30/187 (16%)]\t training loss: 0.087080\n",
      "epoch: 14 [40/187 (21%)]\t training loss: 0.157911\n",
      "epoch: 14 [50/187 (27%)]\t training loss: 0.014203\n",
      "epoch: 14 [60/187 (32%)]\t training loss: 0.170311\n",
      "epoch: 14 [70/187 (37%)]\t training loss: 0.038722\n",
      "epoch: 14 [80/187 (43%)]\t training loss: 0.199107\n",
      "epoch: 14 [90/187 (48%)]\t training loss: 0.115614\n",
      "epoch: 14 [100/187 (53%)]\t training loss: 0.288540\n",
      "epoch: 14 [110/187 (59%)]\t training loss: 0.137237\n",
      "epoch: 14 [120/187 (64%)]\t training loss: 0.060483\n",
      "epoch: 14 [130/187 (70%)]\t training loss: 0.247200\n",
      "epoch: 14 [140/187 (75%)]\t training loss: 0.117644\n",
      "epoch: 14 [150/187 (80%)]\t training loss: 0.390394\n",
      "epoch: 14 [160/187 (86%)]\t training loss: 0.054873\n",
      "epoch: 14 [170/187 (91%)]\t training loss: 0.092678\n",
      "epoch: 14 [180/187 (96%)]\t training loss: 0.027283\n",
      "\n",
      "\n",
      "Validation...\n",
      "\n",
      "Test dataset: Overall Loss: 0.4951, Overall Accuracy: 2677/2977 (90%)\n",
      "\n",
      "epoch: 15 [0/187 (0%)]\t training loss: 0.258510\n",
      "epoch: 15 [10/187 (5%)]\t training loss: 0.661611\n",
      "epoch: 15 [20/187 (11%)]\t training loss: 0.284350\n",
      "epoch: 15 [30/187 (16%)]\t training loss: 0.304729\n",
      "epoch: 15 [40/187 (21%)]\t training loss: 0.241092\n",
      "epoch: 15 [50/187 (27%)]\t training loss: 0.131934\n",
      "epoch: 15 [60/187 (32%)]\t training loss: 0.306047\n",
      "epoch: 15 [70/187 (37%)]\t training loss: 0.049798\n",
      "epoch: 15 [80/187 (43%)]\t training loss: 0.028592\n",
      "epoch: 15 [90/187 (48%)]\t training loss: 0.055647\n",
      "epoch: 15 [100/187 (53%)]\t training loss: 0.081692\n",
      "epoch: 15 [110/187 (59%)]\t training loss: 0.049496\n",
      "epoch: 15 [120/187 (64%)]\t training loss: 0.147515\n",
      "epoch: 15 [130/187 (70%)]\t training loss: 0.130828\n",
      "epoch: 15 [140/187 (75%)]\t training loss: 0.028849\n",
      "epoch: 15 [150/187 (80%)]\t training loss: 0.020133\n",
      "epoch: 15 [160/187 (86%)]\t training loss: 0.038138\n",
      "epoch: 15 [170/187 (91%)]\t training loss: 0.042926\n",
      "epoch: 15 [180/187 (96%)]\t training loss: 0.123341\n",
      "\n",
      "\n",
      "Validation...\n",
      "\n",
      "Test dataset: Overall Loss: 0.4122, Overall Accuracy: 2671/2977 (90%)\n",
      "\n",
      "epoch: 16 [0/187 (0%)]\t training loss: 0.110460\n",
      "epoch: 16 [10/187 (5%)]\t training loss: 0.199618\n",
      "epoch: 16 [20/187 (11%)]\t training loss: 0.645148\n",
      "epoch: 16 [30/187 (16%)]\t training loss: 0.253001\n",
      "epoch: 16 [40/187 (21%)]\t training loss: 0.238378\n",
      "epoch: 16 [50/187 (27%)]\t training loss: 0.074949\n",
      "epoch: 16 [60/187 (32%)]\t training loss: 0.166975\n",
      "epoch: 16 [70/187 (37%)]\t training loss: 0.417267\n",
      "epoch: 16 [80/187 (43%)]\t training loss: 0.136488\n",
      "epoch: 16 [90/187 (48%)]\t training loss: 0.165401\n",
      "epoch: 16 [100/187 (53%)]\t training loss: 0.122329\n",
      "epoch: 16 [110/187 (59%)]\t training loss: 0.177808\n",
      "epoch: 16 [120/187 (64%)]\t training loss: 0.289681\n",
      "epoch: 16 [130/187 (70%)]\t training loss: 0.118373\n",
      "epoch: 16 [140/187 (75%)]\t training loss: 0.114863\n",
      "epoch: 16 [150/187 (80%)]\t training loss: 0.213914\n",
      "epoch: 16 [160/187 (86%)]\t training loss: 0.050095\n",
      "epoch: 16 [170/187 (91%)]\t training loss: 0.038468\n",
      "epoch: 16 [180/187 (96%)]\t training loss: 0.092692\n",
      "\n",
      "\n",
      "Validation...\n",
      "\n",
      "Test dataset: Overall Loss: 0.4512, Overall Accuracy: 2656/2977 (89%)\n",
      "\n",
      "epoch: 17 [0/187 (0%)]\t training loss: 0.236712\n",
      "epoch: 17 [10/187 (5%)]\t training loss: 0.111861\n",
      "epoch: 17 [20/187 (11%)]\t training loss: 0.066369\n",
      "epoch: 17 [30/187 (16%)]\t training loss: 0.091545\n",
      "epoch: 17 [40/187 (21%)]\t training loss: 0.079014\n",
      "epoch: 17 [50/187 (27%)]\t training loss: 0.028637\n",
      "epoch: 17 [60/187 (32%)]\t training loss: 0.055200\n",
      "epoch: 17 [70/187 (37%)]\t training loss: 0.086594\n",
      "epoch: 17 [80/187 (43%)]\t training loss: 0.105371\n",
      "epoch: 17 [90/187 (48%)]\t training loss: 0.061244\n",
      "epoch: 17 [100/187 (53%)]\t training loss: 0.059738\n",
      "epoch: 17 [110/187 (59%)]\t training loss: 0.144818\n",
      "epoch: 17 [120/187 (64%)]\t training loss: 0.090017\n",
      "epoch: 17 [130/187 (70%)]\t training loss: 0.051903\n",
      "epoch: 17 [140/187 (75%)]\t training loss: 0.108721\n",
      "epoch: 17 [150/187 (80%)]\t training loss: 0.259816\n",
      "epoch: 17 [160/187 (86%)]\t training loss: 0.082367\n",
      "epoch: 17 [170/187 (91%)]\t training loss: 0.122879\n",
      "epoch: 17 [180/187 (96%)]\t training loss: 0.245169\n",
      "\n",
      "\n",
      "Validation...\n",
      "\n",
      "Test dataset: Overall Loss: 0.4813, Overall Accuracy: 2693/2977 (90%)\n",
      "\n",
      "epoch: 18 [0/187 (0%)]\t training loss: 0.180817\n",
      "epoch: 18 [10/187 (5%)]\t training loss: 0.066661\n",
      "epoch: 18 [20/187 (11%)]\t training loss: 0.015771\n",
      "epoch: 18 [30/187 (16%)]\t training loss: 0.057680\n",
      "epoch: 18 [40/187 (21%)]\t training loss: 0.070110\n",
      "epoch: 18 [50/187 (27%)]\t training loss: 0.009081\n",
      "epoch: 18 [60/187 (32%)]\t training loss: 0.154300\n",
      "epoch: 18 [70/187 (37%)]\t training loss: 0.082214\n",
      "epoch: 18 [80/187 (43%)]\t training loss: 0.053434\n",
      "epoch: 18 [90/187 (48%)]\t training loss: 0.066431\n",
      "epoch: 18 [100/187 (53%)]\t training loss: 0.038920\n",
      "epoch: 18 [110/187 (59%)]\t training loss: 0.171481\n",
      "epoch: 18 [120/187 (64%)]\t training loss: 0.017904\n",
      "epoch: 18 [130/187 (70%)]\t training loss: 0.014780\n",
      "epoch: 18 [140/187 (75%)]\t training loss: 0.111594\n",
      "epoch: 18 [150/187 (80%)]\t training loss: 0.029931\n",
      "epoch: 18 [160/187 (86%)]\t training loss: 0.148081\n",
      "epoch: 18 [170/187 (91%)]\t training loss: 0.096079\n",
      "epoch: 18 [180/187 (96%)]\t training loss: 0.232921\n",
      "\n",
      "\n",
      "Validation...\n",
      "\n",
      "Test dataset: Overall Loss: 0.5347, Overall Accuracy: 2657/2977 (89%)\n",
      "\n",
      "epoch: 19 [0/187 (0%)]\t training loss: 0.103364\n",
      "epoch: 19 [10/187 (5%)]\t training loss: 0.240405\n",
      "epoch: 19 [20/187 (11%)]\t training loss: 0.121704\n",
      "epoch: 19 [30/187 (16%)]\t training loss: 0.064776\n",
      "epoch: 19 [40/187 (21%)]\t training loss: 0.061327\n",
      "epoch: 19 [50/187 (27%)]\t training loss: 0.037659\n",
      "epoch: 19 [60/187 (32%)]\t training loss: 0.061330\n",
      "epoch: 19 [70/187 (37%)]\t training loss: 0.033343\n",
      "epoch: 19 [80/187 (43%)]\t training loss: 0.061481\n",
      "epoch: 19 [90/187 (48%)]\t training loss: 0.095820\n",
      "epoch: 19 [100/187 (53%)]\t training loss: 0.016059\n",
      "epoch: 19 [110/187 (59%)]\t training loss: 0.081206\n",
      "epoch: 19 [120/187 (64%)]\t training loss: 0.046419\n",
      "epoch: 19 [130/187 (70%)]\t training loss: 0.176356\n",
      "epoch: 19 [140/187 (75%)]\t training loss: 0.023550\n",
      "epoch: 19 [150/187 (80%)]\t training loss: 0.272653\n",
      "epoch: 19 [160/187 (86%)]\t training loss: 0.115192\n",
      "epoch: 19 [170/187 (91%)]\t training loss: 0.111679\n",
      "epoch: 19 [180/187 (96%)]\t training loss: 0.110330\n",
      "\n",
      "\n",
      "Validation...\n",
      "\n",
      "Test dataset: Overall Loss: 0.5718, Overall Accuracy: 2660/2977 (89%)\n",
      "\n",
      "epoch: 20 [0/187 (0%)]\t training loss: 0.154054\n",
      "epoch: 20 [10/187 (5%)]\t training loss: 0.253525\n",
      "epoch: 20 [20/187 (11%)]\t training loss: 0.113904\n",
      "epoch: 20 [30/187 (16%)]\t training loss: 0.050490\n",
      "epoch: 20 [40/187 (21%)]\t training loss: 0.078829\n",
      "epoch: 20 [50/187 (27%)]\t training loss: 0.154886\n",
      "epoch: 20 [60/187 (32%)]\t training loss: 0.092565\n",
      "epoch: 20 [70/187 (37%)]\t training loss: 0.069249\n",
      "epoch: 20 [80/187 (43%)]\t training loss: 0.153711\n",
      "epoch: 20 [90/187 (48%)]\t training loss: 0.052733\n",
      "epoch: 20 [100/187 (53%)]\t training loss: 0.003069\n",
      "epoch: 20 [110/187 (59%)]\t training loss: 0.111685\n",
      "epoch: 20 [120/187 (64%)]\t training loss: 0.055215\n",
      "epoch: 20 [130/187 (70%)]\t training loss: 0.010319\n",
      "epoch: 20 [140/187 (75%)]\t training loss: 0.130895\n",
      "epoch: 20 [150/187 (80%)]\t training loss: 0.055023\n",
      "epoch: 20 [160/187 (86%)]\t training loss: 0.180982\n",
      "epoch: 20 [170/187 (91%)]\t training loss: 0.048970\n",
      "epoch: 20 [180/187 (96%)]\t training loss: 0.037130\n",
      "\n",
      "\n",
      "Validation...\n",
      "\n",
      "Test dataset: Overall Loss: 0.8592, Overall Accuracy: 2659/2977 (89%)\n",
      "\n",
      "epoch: 21 [0/187 (0%)]\t training loss: 0.096532\n",
      "epoch: 21 [10/187 (5%)]\t training loss: 0.389140\n",
      "epoch: 21 [20/187 (11%)]\t training loss: 0.351211\n",
      "epoch: 21 [30/187 (16%)]\t training loss: 0.236020\n",
      "epoch: 21 [40/187 (21%)]\t training loss: 0.129591\n",
      "epoch: 21 [50/187 (27%)]\t training loss: 0.210569\n",
      "epoch: 21 [60/187 (32%)]\t training loss: 0.177454\n",
      "epoch: 21 [70/187 (37%)]\t training loss: 0.249229\n",
      "epoch: 21 [80/187 (43%)]\t training loss: 0.123423\n",
      "epoch: 21 [90/187 (48%)]\t training loss: 0.272682\n",
      "epoch: 21 [100/187 (53%)]\t training loss: 0.201323\n",
      "epoch: 21 [110/187 (59%)]\t training loss: 0.141792\n",
      "epoch: 21 [120/187 (64%)]\t training loss: 0.112191\n",
      "epoch: 21 [130/187 (70%)]\t training loss: 0.081990\n",
      "epoch: 21 [140/187 (75%)]\t training loss: 0.261928\n",
      "epoch: 21 [150/187 (80%)]\t training loss: 0.177059\n",
      "epoch: 21 [160/187 (86%)]\t training loss: 0.197452\n",
      "epoch: 21 [170/187 (91%)]\t training loss: 0.085945\n",
      "epoch: 21 [180/187 (96%)]\t training loss: 0.055606\n",
      "\n",
      "\n",
      "Validation...\n",
      "\n",
      "Test dataset: Overall Loss: 0.4129, Overall Accuracy: 2654/2977 (89%)\n",
      "\n",
      "epoch: 22 [0/187 (0%)]\t training loss: 0.065687\n",
      "epoch: 22 [10/187 (5%)]\t training loss: 0.064468\n",
      "epoch: 22 [20/187 (11%)]\t training loss: 0.177946\n",
      "epoch: 22 [30/187 (16%)]\t training loss: 0.240713\n",
      "epoch: 22 [40/187 (21%)]\t training loss: 0.070206\n",
      "epoch: 22 [50/187 (27%)]\t training loss: 0.104501\n",
      "epoch: 22 [60/187 (32%)]\t training loss: 0.022042\n",
      "epoch: 22 [70/187 (37%)]\t training loss: 0.191729\n",
      "epoch: 22 [80/187 (43%)]\t training loss: 0.126921\n",
      "epoch: 22 [90/187 (48%)]\t training loss: 0.156502\n",
      "epoch: 22 [100/187 (53%)]\t training loss: 0.218993\n",
      "epoch: 22 [110/187 (59%)]\t training loss: 0.063718\n",
      "epoch: 22 [120/187 (64%)]\t training loss: 0.294033\n",
      "epoch: 22 [130/187 (70%)]\t training loss: 0.148704\n",
      "epoch: 22 [140/187 (75%)]\t training loss: 0.332657\n",
      "epoch: 22 [150/187 (80%)]\t training loss: 0.182051\n",
      "epoch: 22 [160/187 (86%)]\t training loss: 0.091461\n",
      "epoch: 22 [170/187 (91%)]\t training loss: 0.367176\n",
      "epoch: 22 [180/187 (96%)]\t training loss: 0.072326\n",
      "\n",
      "\n",
      "Validation...\n",
      "\n",
      "Test dataset: Overall Loss: 0.4374, Overall Accuracy: 2693/2977 (90%)\n",
      "\n",
      "epoch: 23 [0/187 (0%)]\t training loss: 0.126276\n",
      "epoch: 23 [10/187 (5%)]\t training loss: 0.078841\n",
      "epoch: 23 [20/187 (11%)]\t training loss: 0.042248\n",
      "epoch: 23 [30/187 (16%)]\t training loss: 0.050055\n",
      "epoch: 23 [40/187 (21%)]\t training loss: 0.052314\n",
      "epoch: 23 [50/187 (27%)]\t training loss: 0.117368\n",
      "epoch: 23 [60/187 (32%)]\t training loss: 0.085176\n",
      "epoch: 23 [70/187 (37%)]\t training loss: 0.050835\n",
      "epoch: 23 [80/187 (43%)]\t training loss: 0.141194\n",
      "epoch: 23 [90/187 (48%)]\t training loss: 0.040692\n",
      "epoch: 23 [100/187 (53%)]\t training loss: 0.126833\n",
      "epoch: 23 [110/187 (59%)]\t training loss: 0.136489\n",
      "epoch: 23 [120/187 (64%)]\t training loss: 0.312070\n",
      "epoch: 23 [130/187 (70%)]\t training loss: 0.023361\n",
      "epoch: 23 [140/187 (75%)]\t training loss: 0.069117\n",
      "epoch: 23 [150/187 (80%)]\t training loss: 0.142276\n",
      "epoch: 23 [160/187 (86%)]\t training loss: 0.158885\n",
      "epoch: 23 [170/187 (91%)]\t training loss: 0.095300\n",
      "epoch: 23 [180/187 (96%)]\t training loss: 0.053325\n",
      "\n",
      "\n",
      "Validation...\n",
      "\n",
      "Test dataset: Overall Loss: 0.5366, Overall Accuracy: 2681/2977 (90%)\n",
      "\n",
      "epoch: 24 [0/187 (0%)]\t training loss: 0.023157\n",
      "epoch: 24 [10/187 (5%)]\t training loss: 0.070472\n",
      "epoch: 24 [20/187 (11%)]\t training loss: 0.074267\n",
      "epoch: 24 [30/187 (16%)]\t training loss: 0.019107\n",
      "epoch: 24 [40/187 (21%)]\t training loss: 0.053796\n",
      "epoch: 24 [50/187 (27%)]\t training loss: 0.008229\n",
      "epoch: 24 [60/187 (32%)]\t training loss: 0.119424\n",
      "epoch: 24 [70/187 (37%)]\t training loss: 0.098592\n",
      "epoch: 24 [80/187 (43%)]\t training loss: 0.044199\n",
      "epoch: 24 [90/187 (48%)]\t training loss: 0.117783\n",
      "epoch: 24 [100/187 (53%)]\t training loss: 0.168927\n",
      "epoch: 24 [110/187 (59%)]\t training loss: 0.073064\n",
      "epoch: 24 [120/187 (64%)]\t training loss: 0.073963\n",
      "epoch: 24 [130/187 (70%)]\t training loss: 0.134989\n",
      "epoch: 24 [140/187 (75%)]\t training loss: 0.008165\n",
      "epoch: 24 [150/187 (80%)]\t training loss: 0.120215\n",
      "epoch: 24 [160/187 (86%)]\t training loss: 0.129794\n",
      "epoch: 24 [170/187 (91%)]\t training loss: 0.120782\n",
      "epoch: 24 [180/187 (96%)]\t training loss: 0.156022\n",
      "\n",
      "\n",
      "Validation...\n",
      "\n",
      "Test dataset: Overall Loss: 0.5352, Overall Accuracy: 2703/2977 (91%)\n",
      "\n",
      "epoch: 25 [0/187 (0%)]\t training loss: 0.205607\n",
      "epoch: 25 [10/187 (5%)]\t training loss: 0.046238\n",
      "epoch: 25 [20/187 (11%)]\t training loss: 0.073737\n",
      "epoch: 25 [30/187 (16%)]\t training loss: 0.039495\n",
      "epoch: 25 [40/187 (21%)]\t training loss: 0.057536\n",
      "epoch: 25 [50/187 (27%)]\t training loss: 0.063229\n",
      "epoch: 25 [60/187 (32%)]\t training loss: 0.137402\n",
      "epoch: 25 [70/187 (37%)]\t training loss: 0.016562\n",
      "epoch: 25 [80/187 (43%)]\t training loss: 0.111888\n",
      "epoch: 25 [90/187 (48%)]\t training loss: 0.055733\n",
      "epoch: 25 [100/187 (53%)]\t training loss: 0.049308\n",
      "epoch: 25 [110/187 (59%)]\t training loss: 0.071474\n",
      "epoch: 25 [120/187 (64%)]\t training loss: 0.001673\n",
      "epoch: 25 [130/187 (70%)]\t training loss: 0.096922\n",
      "epoch: 25 [140/187 (75%)]\t training loss: 0.205385\n",
      "epoch: 25 [150/187 (80%)]\t training loss: 0.028748\n",
      "epoch: 25 [160/187 (86%)]\t training loss: 0.215607\n",
      "epoch: 25 [170/187 (91%)]\t training loss: 0.019032\n",
      "epoch: 25 [180/187 (96%)]\t training loss: 0.020503\n",
      "\n",
      "\n",
      "Validation...\n",
      "\n",
      "Test dataset: Overall Loss: 0.5436, Overall Accuracy: 2714/2977 (91%)\n",
      "\n",
      "epoch: 26 [0/187 (0%)]\t training loss: 0.011890\n",
      "epoch: 26 [10/187 (5%)]\t training loss: 0.016602\n",
      "epoch: 26 [20/187 (11%)]\t training loss: 0.021093\n",
      "epoch: 26 [30/187 (16%)]\t training loss: 0.072385\n",
      "epoch: 26 [40/187 (21%)]\t training loss: 0.408644\n",
      "epoch: 26 [50/187 (27%)]\t training loss: 0.026849\n",
      "epoch: 26 [60/187 (32%)]\t training loss: 0.071456\n",
      "epoch: 26 [70/187 (37%)]\t training loss: 0.194435\n",
      "epoch: 26 [80/187 (43%)]\t training loss: 0.062508\n",
      "epoch: 26 [90/187 (48%)]\t training loss: 0.056621\n",
      "epoch: 26 [100/187 (53%)]\t training loss: 0.147371\n",
      "epoch: 26 [110/187 (59%)]\t training loss: 0.163712\n",
      "epoch: 26 [120/187 (64%)]\t training loss: 0.118524\n",
      "epoch: 26 [130/187 (70%)]\t training loss: 0.093632\n",
      "epoch: 26 [140/187 (75%)]\t training loss: 0.113975\n",
      "epoch: 26 [150/187 (80%)]\t training loss: 0.165829\n",
      "epoch: 26 [160/187 (86%)]\t training loss: 0.103108\n",
      "epoch: 26 [170/187 (91%)]\t training loss: 0.045053\n",
      "epoch: 26 [180/187 (96%)]\t training loss: 0.110311\n",
      "\n",
      "\n",
      "Validation...\n",
      "\n",
      "Test dataset: Overall Loss: 0.6797, Overall Accuracy: 2679/2977 (90%)\n",
      "\n",
      "epoch: 27 [0/187 (0%)]\t training loss: 0.028139\n",
      "epoch: 27 [10/187 (5%)]\t training loss: 0.090336\n",
      "epoch: 27 [20/187 (11%)]\t training loss: 0.038068\n",
      "epoch: 27 [30/187 (16%)]\t training loss: 0.062186\n",
      "epoch: 27 [40/187 (21%)]\t training loss: 0.005198\n",
      "epoch: 27 [50/187 (27%)]\t training loss: 0.057895\n",
      "epoch: 27 [60/187 (32%)]\t training loss: 0.022911\n",
      "epoch: 27 [70/187 (37%)]\t training loss: 0.142485\n",
      "epoch: 27 [80/187 (43%)]\t training loss: 0.097354\n",
      "epoch: 27 [90/187 (48%)]\t training loss: 0.165625\n",
      "epoch: 27 [100/187 (53%)]\t training loss: 0.510756\n",
      "epoch: 27 [110/187 (59%)]\t training loss: 0.129033\n",
      "epoch: 27 [120/187 (64%)]\t training loss: 0.518768\n",
      "epoch: 27 [130/187 (70%)]\t training loss: 0.142276\n",
      "epoch: 27 [140/187 (75%)]\t training loss: 0.052742\n",
      "epoch: 27 [150/187 (80%)]\t training loss: 0.206133\n",
      "epoch: 27 [160/187 (86%)]\t training loss: 0.062609\n",
      "epoch: 27 [170/187 (91%)]\t training loss: 0.024872\n",
      "epoch: 27 [180/187 (96%)]\t training loss: 0.112768\n",
      "\n",
      "\n",
      "Validation...\n",
      "\n",
      "Test dataset: Overall Loss: 0.5391, Overall Accuracy: 2660/2977 (89%)\n",
      "\n",
      "epoch: 28 [0/187 (0%)]\t training loss: 0.083607\n",
      "epoch: 28 [10/187 (5%)]\t training loss: 0.151757\n",
      "epoch: 28 [20/187 (11%)]\t training loss: 0.086939\n",
      "epoch: 28 [30/187 (16%)]\t training loss: 0.029494\n",
      "epoch: 28 [40/187 (21%)]\t training loss: 0.129811\n",
      "epoch: 28 [50/187 (27%)]\t training loss: 0.474412\n",
      "epoch: 28 [60/187 (32%)]\t training loss: 0.041966\n",
      "epoch: 28 [70/187 (37%)]\t training loss: 0.347199\n",
      "epoch: 28 [80/187 (43%)]\t training loss: 0.232649\n",
      "epoch: 28 [90/187 (48%)]\t training loss: 0.020844\n",
      "epoch: 28 [100/187 (53%)]\t training loss: 0.025758\n",
      "epoch: 28 [110/187 (59%)]\t training loss: 0.111240\n",
      "epoch: 28 [120/187 (64%)]\t training loss: 0.034345\n",
      "epoch: 28 [130/187 (70%)]\t training loss: 0.220549\n",
      "epoch: 28 [140/187 (75%)]\t training loss: 0.189815\n",
      "epoch: 28 [150/187 (80%)]\t training loss: 0.848337\n",
      "epoch: 28 [160/187 (86%)]\t training loss: 0.370543\n",
      "epoch: 28 [170/187 (91%)]\t training loss: 0.262750\n",
      "epoch: 28 [180/187 (96%)]\t training loss: 0.102390\n",
      "\n",
      "\n",
      "Validation...\n",
      "\n",
      "Test dataset: Overall Loss: 0.6509, Overall Accuracy: 2625/2977 (88%)\n",
      "\n",
      "epoch: 29 [0/187 (0%)]\t training loss: 0.405329\n",
      "epoch: 29 [10/187 (5%)]\t training loss: 0.110226\n",
      "epoch: 29 [20/187 (11%)]\t training loss: 0.085380\n",
      "epoch: 29 [30/187 (16%)]\t training loss: 0.093743\n",
      "epoch: 29 [40/187 (21%)]\t training loss: 0.159290\n",
      "epoch: 29 [50/187 (27%)]\t training loss: 0.157236\n",
      "epoch: 29 [60/187 (32%)]\t training loss: 0.108735\n",
      "epoch: 29 [70/187 (37%)]\t training loss: 0.119321\n",
      "epoch: 29 [80/187 (43%)]\t training loss: 0.088422\n",
      "epoch: 29 [90/187 (48%)]\t training loss: 0.062117\n",
      "epoch: 29 [100/187 (53%)]\t training loss: 0.027023\n",
      "epoch: 29 [110/187 (59%)]\t training loss: 0.160795\n",
      "epoch: 29 [120/187 (64%)]\t training loss: 0.106275\n",
      "epoch: 29 [130/187 (70%)]\t training loss: 0.009499\n",
      "epoch: 29 [140/187 (75%)]\t training loss: 0.179142\n",
      "epoch: 29 [150/187 (80%)]\t training loss: 0.029653\n",
      "epoch: 29 [160/187 (86%)]\t training loss: 0.072285\n",
      "epoch: 29 [170/187 (91%)]\t training loss: 0.069708\n",
      "epoch: 29 [180/187 (96%)]\t training loss: 0.127763\n",
      "\n",
      "\n",
      "Validation...\n",
      "\n",
      "Test dataset: Overall Loss: 0.5943, Overall Accuracy: 2653/2977 (89%)\n",
      "\n"
     ]
    }
   ],
   "source": [
    "imbalanced_model, train_losses, val_losses = train(\n",
    "    imbalanced_train_loader,\n",
    "    imbalanced_valid_loader,\n",
    "    \"Imbalanced_model.pth\",\n",
    "    ConvNet().to(device),\n",
    "    epochs=30,\n",
    ")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 472
    },
    "id": "zYfIlOlnzjjN",
    "outputId": "ed9706d9-3374-444e-f356-6bab01becbe0"
   },
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 640x480 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "plt.plot(train_losses)\n",
    "plt.plot(val_losses)\n",
    "plt.xlabel(\"epoch\")\n",
    "plt.ylabel(\"loss\")\n",
    "plt.legend([\"Train\", \"Valid\"])\n",
    "plt.title(\"Model On imbalanced Fashion MNIST - Train vs Valid Accuracy\")\n",
    "\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "IcKVhzflAEFR"
   },
   "source": [
    "### test on test set"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/"
    },
    "id": "ht3QLbX8AEFR",
    "outputId": "5d16533f-6d13-4e28-de54-30e6d3149c7c"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "On imbalanced Fashion MNIST dataset\n",
      "\n",
      "Test dataset: Overall Loss: 0.0018, Overall Accuracy: 7677/10000 (77%)\n",
      "\n",
      "test accuracy :  76.77\n"
     ]
    }
   ],
   "source": [
    "print(\"On imbalanced Fashion MNIST dataset\")\n",
    "test_acc1, class_percentages1 = test(imbalanced_model, \"Imbalanced_model.pth\")\n",
    "print(\"test accuracy : \", str(test_acc1))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "_IcyOBWsAxnp"
   },
   "source": [
    "## Cutmix Image Augmentation"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "metadata": {
    "id": "rN1pfMynA6zb"
   },
   "outputs": [],
   "source": [
    "# Adapted from https://github.com/clovaai/CutMix-PyTorch/blob/master/train.py\n",
    "\n",
    "\n",
    "def rand_bbox(size, lam):\n",
    "    \"\"\"\n",
    "\n",
    "    :param size: torch.Size. Shape:(batch_size, channel, image_size, image_size)\n",
    "    :param lam: cutmix rate\n",
    "    :return: box boundary\n",
    "    \"\"\"\n",
    "    W = size[2]\n",
    "    H = size[3]\n",
    "    cut_rat = np.sqrt(1.0 - lam)\n",
    "    cut_w = int(W * cut_rat)\n",
    "    cut_h = int(H * cut_rat)\n",
    "\n",
    "    # uniform\n",
    "    cx = np.random.randint(W)\n",
    "    cy = np.random.randint(H)\n",
    "\n",
    "    bbx1 = np.clip(cx - cut_w // 2, 0, W)\n",
    "    bby1 = np.clip(cy - cut_h // 2, 0, H)\n",
    "    bbx2 = np.clip(cx + cut_w // 2, 0, W)\n",
    "    bby2 = np.clip(cy + cut_h // 2, 0, H)\n",
    "\n",
    "    return bbx1, bby1, bbx2, bby2\n",
    "\n",
    "\n",
    "def cutmix(data, targets, alpha):\n",
    "    \"\"\"\n",
    "\n",
    "    :param data: torch.Size. Shape:(batch_size, channel, image_size, image_size)\n",
    "    :param targets: torch.Tensor. multitask, multi-classification, target1. shape(batch_size, 1)\n",
    "    :param alpha: float, positive\n",
    "    :return: resulting image, multi-target\n",
    "    \"\"\"\n",
    "    indices = torch.randperm(data.size(0))\n",
    "    # shuffled_data = data[indices]\n",
    "    targets_a = targets\n",
    "    targets_b = targets[indices]\n",
    "\n",
    "    lam = np.random.beta(alpha, alpha)\n",
    "    bbx1, bby1, bbx2, bby2 = rand_bbox(data.size(), lam)\n",
    "    data[:, :, bbx1:bbx2, bby1:bby2] = data[indices, :, bbx1:bbx2, bby1:bby2]\n",
    "    # adjust lambda to exactly match pixel ratio\n",
    "    lam = 1 - ((bbx2 - bbx1) * (bby2 - bby1) / (data.size()[-1] * data.size()[-2]))\n",
    "\n",
    "    targets = [targets_a, targets_b, lam]\n",
    "    return data, targets"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "metadata": {
    "id": "U4LdKSpaBMUR"
   },
   "outputs": [],
   "source": [
    "def train_cut_mix(\n",
    "    train_dataloader,\n",
    "    valid_dataloader,\n",
    "    PATH,\n",
    "    model=None,\n",
    "    criterion=F.cross_entropy,\n",
    "    epochs=1,\n",
    "):\n",
    "    optimizer = torch.optim.SGD(model.parameters(), lr=0.1, momentum=0.9)\n",
    "    model.train()\n",
    "    best_acc = 0\n",
    "    train_losses = []\n",
    "    val_losses = []\n",
    "    for epoch in range(epochs):\n",
    "        losses = 0\n",
    "        for b_i, (X, y) in enumerate(train_dataloader):\n",
    "            X, y = X.to(device), y.to(device)\n",
    "            X, ys = cutmix(X, y, alpha=0.75)\n",
    "            pred_prob = model(X)\n",
    "            targets1, targets2, lam = ys[0], ys[1], ys[2]\n",
    "            loss = lam * criterion(pred_prob, targets1) + (1 - lam) * criterion(\n",
    "                pred_prob, targets2\n",
    "            )\n",
    "            optimizer.zero_grad()\n",
    "            loss.backward()\n",
    "            optimizer.step()\n",
    "            losses += loss.item()  # print(losses)\n",
    "            if b_i % 10 == 0:\n",
    "                print(\n",
    "                    \"epoch: {} [{}/{} ({:.0f}%)]\\t training loss: {:.6f}\".format(\n",
    "                        epoch,\n",
    "                        b_i,\n",
    "                        len(train_dataloader),\n",
    "                        100.0 * b_i / len(train_dataloader),\n",
    "                        loss.item(),\n",
    "                    )\n",
    "                )\n",
    "        train_losses.append(losses / len(train_dataloader))\n",
    "        print(\"\\n\")\n",
    "        print(\"Validation...\")\n",
    "        model.eval()\n",
    "        success = 0\n",
    "        total = 0\n",
    "        losses = 0\n",
    "        with torch.no_grad():\n",
    "            for X, y in valid_dataloader:\n",
    "                X, y = X.to(device), y.to(device)\n",
    "                pred_prob = model(X)\n",
    "                loss = criterion(pred_prob, y)  # loss summed across the batch\n",
    "                losses += loss.item()\n",
    "                pred = pred_prob.argmax(\n",
    "                    dim=1, keepdim=True\n",
    "                )  # use argmax to get the most likely prediction\n",
    "                success += pred.eq(y.view_as(pred)).sum().item()\n",
    "                total += len(y)\n",
    "        val_losses.append(losses / len(valid_dataloader))\n",
    "        accuracy = 100.0 * success / total\n",
    "\n",
    "        if accuracy > best_acc:\n",
    "            torch.save(model.state_dict(), PATH)\n",
    "            best_acc = accuracy\n",
    "\n",
    "        print(\n",
    "            \"\\nTest dataset: Overall Loss: {:.4f}, Overall Accuracy: {}/{} ({:.0f}%)\\n\".format(\n",
    "                losses / len(valid_dataloader), success, total, accuracy\n",
    "            )\n",
    "        )\n",
    "    return model, train_losses, val_losses"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/"
    },
    "id": "zUKMACwyHEv1",
    "outputId": "0e05929c-a3cc-40b1-ecce-49689262f558"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "epoch: 0 [0/187 (0%)]\t training loss: 2.252028\n",
      "epoch: 0 [10/187 (5%)]\t training loss: 1.752678\n",
      "epoch: 0 [20/187 (11%)]\t training loss: 1.252409\n",
      "epoch: 0 [30/187 (16%)]\t training loss: 1.378983\n",
      "epoch: 0 [40/187 (21%)]\t training loss: 1.156488\n",
      "epoch: 0 [50/187 (27%)]\t training loss: 1.198630\n",
      "epoch: 0 [60/187 (32%)]\t training loss: 1.129406\n",
      "epoch: 0 [70/187 (37%)]\t training loss: 1.186553\n",
      "epoch: 0 [80/187 (43%)]\t training loss: 1.229246\n",
      "epoch: 0 [90/187 (48%)]\t training loss: 1.334609\n",
      "epoch: 0 [100/187 (53%)]\t training loss: 0.574964\n",
      "epoch: 0 [110/187 (59%)]\t training loss: 1.125413\n",
      "epoch: 0 [120/187 (64%)]\t training loss: 1.264428\n",
      "epoch: 0 [130/187 (70%)]\t training loss: 0.669252\n",
      "epoch: 0 [140/187 (75%)]\t training loss: 1.130286\n",
      "epoch: 0 [150/187 (80%)]\t training loss: 1.082550\n",
      "epoch: 0 [160/187 (86%)]\t training loss: 0.695548\n",
      "epoch: 0 [170/187 (91%)]\t training loss: 0.905360\n",
      "epoch: 0 [180/187 (96%)]\t training loss: 0.953947\n",
      "\n",
      "\n",
      "Validation...\n",
      "\n",
      "Test dataset: Overall Loss: 0.4241, Overall Accuracy: 2587/2977 (87%)\n",
      "\n",
      "epoch: 1 [0/187 (0%)]\t training loss: 0.801820\n",
      "epoch: 1 [10/187 (5%)]\t training loss: 0.669593\n",
      "epoch: 1 [20/187 (11%)]\t training loss: 1.106225\n",
      "epoch: 1 [30/187 (16%)]\t training loss: 1.133604\n",
      "epoch: 1 [40/187 (21%)]\t training loss: 1.054983\n",
      "epoch: 1 [50/187 (27%)]\t training loss: 0.755043\n",
      "epoch: 1 [60/187 (32%)]\t training loss: 0.833692\n",
      "epoch: 1 [70/187 (37%)]\t training loss: 0.970758\n",
      "epoch: 1 [80/187 (43%)]\t training loss: 0.752284\n",
      "epoch: 1 [90/187 (48%)]\t training loss: 1.026138\n",
      "epoch: 1 [100/187 (53%)]\t training loss: 0.604609\n",
      "epoch: 1 [110/187 (59%)]\t training loss: 0.989622\n",
      "epoch: 1 [120/187 (64%)]\t training loss: 0.993063\n",
      "epoch: 1 [130/187 (70%)]\t training loss: 0.866025\n",
      "epoch: 1 [140/187 (75%)]\t training loss: 1.072021\n",
      "epoch: 1 [150/187 (80%)]\t training loss: 0.460163\n",
      "epoch: 1 [160/187 (86%)]\t training loss: 0.491393\n",
      "epoch: 1 [170/187 (91%)]\t training loss: 1.170433\n",
      "epoch: 1 [180/187 (96%)]\t training loss: 0.968702\n",
      "\n",
      "\n",
      "Validation...\n",
      "\n",
      "Test dataset: Overall Loss: 0.4475, Overall Accuracy: 2627/2977 (88%)\n",
      "\n",
      "epoch: 2 [0/187 (0%)]\t training loss: 0.685556\n",
      "epoch: 2 [10/187 (5%)]\t training loss: 1.108387\n",
      "epoch: 2 [20/187 (11%)]\t training loss: 1.013354\n",
      "epoch: 2 [30/187 (16%)]\t training loss: 0.888954\n",
      "epoch: 2 [40/187 (21%)]\t training loss: 0.979116\n",
      "epoch: 2 [50/187 (27%)]\t training loss: 1.059127\n",
      "epoch: 2 [60/187 (32%)]\t training loss: 0.625570\n",
      "epoch: 2 [70/187 (37%)]\t training loss: 0.561670\n",
      "epoch: 2 [80/187 (43%)]\t training loss: 0.836020\n",
      "epoch: 2 [90/187 (48%)]\t training loss: 0.876780\n",
      "epoch: 2 [100/187 (53%)]\t training loss: 0.878995\n",
      "epoch: 2 [110/187 (59%)]\t training loss: 0.978719\n",
      "epoch: 2 [120/187 (64%)]\t training loss: 1.051622\n",
      "epoch: 2 [130/187 (70%)]\t training loss: 1.094857\n",
      "epoch: 2 [140/187 (75%)]\t training loss: 0.726305\n",
      "epoch: 2 [150/187 (80%)]\t training loss: 0.434818\n",
      "epoch: 2 [160/187 (86%)]\t training loss: 0.527789\n",
      "epoch: 2 [170/187 (91%)]\t training loss: 0.818935\n",
      "epoch: 2 [180/187 (96%)]\t training loss: 0.572256\n",
      "\n",
      "\n",
      "Validation...\n",
      "\n",
      "Test dataset: Overall Loss: 0.3419, Overall Accuracy: 2672/2977 (90%)\n",
      "\n",
      "epoch: 3 [0/187 (0%)]\t training loss: 1.081642\n",
      "epoch: 3 [10/187 (5%)]\t training loss: 0.496432\n",
      "epoch: 3 [20/187 (11%)]\t training loss: 0.881102\n",
      "epoch: 3 [30/187 (16%)]\t training loss: 0.468177\n",
      "epoch: 3 [40/187 (21%)]\t training loss: 0.944726\n",
      "epoch: 3 [50/187 (27%)]\t training loss: 0.773523\n",
      "epoch: 3 [60/187 (32%)]\t training loss: 0.867760\n",
      "epoch: 3 [70/187 (37%)]\t training loss: 0.764702\n",
      "epoch: 3 [80/187 (43%)]\t training loss: 0.676177\n",
      "epoch: 3 [90/187 (48%)]\t training loss: 0.565589\n",
      "epoch: 3 [100/187 (53%)]\t training loss: 0.931951\n",
      "epoch: 3 [110/187 (59%)]\t training loss: 1.103022\n",
      "epoch: 3 [120/187 (64%)]\t training loss: 1.056859\n",
      "epoch: 3 [130/187 (70%)]\t training loss: 0.896462\n",
      "epoch: 3 [140/187 (75%)]\t training loss: 0.926508\n",
      "epoch: 3 [150/187 (80%)]\t training loss: 0.783092\n",
      "epoch: 3 [160/187 (86%)]\t training loss: 0.943073\n",
      "epoch: 3 [170/187 (91%)]\t training loss: 0.705526\n",
      "epoch: 3 [180/187 (96%)]\t training loss: 0.446949\n",
      "\n",
      "\n",
      "Validation...\n",
      "\n",
      "Test dataset: Overall Loss: 0.3204, Overall Accuracy: 2656/2977 (89%)\n",
      "\n",
      "epoch: 4 [0/187 (0%)]\t training loss: 0.672302\n",
      "epoch: 4 [10/187 (5%)]\t training loss: 0.933474\n",
      "epoch: 4 [20/187 (11%)]\t training loss: 0.805618\n",
      "epoch: 4 [30/187 (16%)]\t training loss: 1.078048\n",
      "epoch: 4 [40/187 (21%)]\t training loss: 0.870552\n",
      "epoch: 4 [50/187 (27%)]\t training loss: 0.944011\n",
      "epoch: 4 [60/187 (32%)]\t training loss: 0.673339\n",
      "epoch: 4 [70/187 (37%)]\t training loss: 0.703927\n",
      "epoch: 4 [80/187 (43%)]\t training loss: 0.710013\n",
      "epoch: 4 [90/187 (48%)]\t training loss: 0.605643\n",
      "epoch: 4 [100/187 (53%)]\t training loss: 0.769548\n",
      "epoch: 4 [110/187 (59%)]\t training loss: 0.949285\n",
      "epoch: 4 [120/187 (64%)]\t training loss: 0.967565\n",
      "epoch: 4 [130/187 (70%)]\t training loss: 0.811451\n",
      "epoch: 4 [140/187 (75%)]\t training loss: 0.719075\n",
      "epoch: 4 [150/187 (80%)]\t training loss: 1.073952\n",
      "epoch: 4 [160/187 (86%)]\t training loss: 0.886865\n",
      "epoch: 4 [170/187 (91%)]\t training loss: 0.949811\n",
      "epoch: 4 [180/187 (96%)]\t training loss: 1.012028\n",
      "\n",
      "\n",
      "Validation...\n",
      "\n",
      "Test dataset: Overall Loss: 0.3750, Overall Accuracy: 2723/2977 (91%)\n",
      "\n",
      "epoch: 5 [0/187 (0%)]\t training loss: 0.966697\n",
      "epoch: 5 [10/187 (5%)]\t training loss: 0.961402\n",
      "epoch: 5 [20/187 (11%)]\t training loss: 0.722497\n",
      "epoch: 5 [30/187 (16%)]\t training loss: 0.856109\n",
      "epoch: 5 [40/187 (21%)]\t training loss: 1.149207\n",
      "epoch: 5 [50/187 (27%)]\t training loss: 0.298181\n",
      "epoch: 5 [60/187 (32%)]\t training loss: 0.580320\n",
      "epoch: 5 [70/187 (37%)]\t training loss: 0.838954\n",
      "epoch: 5 [80/187 (43%)]\t training loss: 1.017671\n",
      "epoch: 5 [90/187 (48%)]\t training loss: 0.558723\n",
      "epoch: 5 [100/187 (53%)]\t training loss: 0.795923\n",
      "epoch: 5 [110/187 (59%)]\t training loss: 0.887004\n",
      "epoch: 5 [120/187 (64%)]\t training loss: 0.504894\n",
      "epoch: 5 [130/187 (70%)]\t training loss: 0.873037\n",
      "epoch: 5 [140/187 (75%)]\t training loss: 1.009774\n",
      "epoch: 5 [150/187 (80%)]\t training loss: 0.813942\n",
      "epoch: 5 [160/187 (86%)]\t training loss: 0.640376\n",
      "epoch: 5 [170/187 (91%)]\t training loss: 0.354972\n",
      "epoch: 5 [180/187 (96%)]\t training loss: 0.938599\n",
      "\n",
      "\n",
      "Validation...\n",
      "\n",
      "Test dataset: Overall Loss: 0.3372, Overall Accuracy: 2732/2977 (92%)\n",
      "\n",
      "epoch: 6 [0/187 (0%)]\t training loss: 0.869590\n",
      "epoch: 6 [10/187 (5%)]\t training loss: 0.836731\n",
      "epoch: 6 [20/187 (11%)]\t training loss: 0.858437\n",
      "epoch: 6 [30/187 (16%)]\t training loss: 0.892758\n",
      "epoch: 6 [40/187 (21%)]\t training loss: 0.831056\n",
      "epoch: 6 [50/187 (27%)]\t training loss: 0.808100\n",
      "epoch: 6 [60/187 (32%)]\t training loss: 0.935222\n",
      "epoch: 6 [70/187 (37%)]\t training loss: 0.891348\n",
      "epoch: 6 [80/187 (43%)]\t training loss: 0.908625\n",
      "epoch: 6 [90/187 (48%)]\t training loss: 0.951010\n",
      "epoch: 6 [100/187 (53%)]\t training loss: 0.831674\n",
      "epoch: 6 [110/187 (59%)]\t training loss: 0.944803\n",
      "epoch: 6 [120/187 (64%)]\t training loss: 0.874876\n",
      "epoch: 6 [130/187 (70%)]\t training loss: 0.979218\n",
      "epoch: 6 [140/187 (75%)]\t training loss: 0.562178\n",
      "epoch: 6 [150/187 (80%)]\t training loss: 1.023438\n",
      "epoch: 6 [160/187 (86%)]\t training loss: 0.871851\n",
      "epoch: 6 [170/187 (91%)]\t training loss: 1.193888\n",
      "epoch: 6 [180/187 (96%)]\t training loss: 0.793069\n",
      "\n",
      "\n",
      "Validation...\n",
      "\n",
      "Test dataset: Overall Loss: 0.3365, Overall Accuracy: 2700/2977 (91%)\n",
      "\n",
      "epoch: 7 [0/187 (0%)]\t training loss: 0.633261\n",
      "epoch: 7 [10/187 (5%)]\t training loss: 0.770119\n",
      "epoch: 7 [20/187 (11%)]\t training loss: 0.818045\n",
      "epoch: 7 [30/187 (16%)]\t training loss: 0.932420\n",
      "epoch: 7 [40/187 (21%)]\t training loss: 0.960388\n",
      "epoch: 7 [50/187 (27%)]\t training loss: 0.761607\n",
      "epoch: 7 [60/187 (32%)]\t training loss: 0.631851\n",
      "epoch: 7 [70/187 (37%)]\t training loss: 0.869837\n",
      "epoch: 7 [80/187 (43%)]\t training loss: 0.629646\n",
      "epoch: 7 [90/187 (48%)]\t training loss: 0.958246\n",
      "epoch: 7 [100/187 (53%)]\t training loss: 0.382871\n",
      "epoch: 7 [110/187 (59%)]\t training loss: 0.978724\n",
      "epoch: 7 [120/187 (64%)]\t training loss: 0.509898\n",
      "epoch: 7 [130/187 (70%)]\t training loss: 0.637959\n",
      "epoch: 7 [140/187 (75%)]\t training loss: 0.938284\n",
      "epoch: 7 [150/187 (80%)]\t training loss: 1.156518\n",
      "epoch: 7 [160/187 (86%)]\t training loss: 1.056689\n",
      "epoch: 7 [170/187 (91%)]\t training loss: 0.349321\n",
      "epoch: 7 [180/187 (96%)]\t training loss: 0.702866\n",
      "\n",
      "\n",
      "Validation...\n",
      "\n",
      "Test dataset: Overall Loss: 0.3024, Overall Accuracy: 2747/2977 (92%)\n",
      "\n",
      "epoch: 8 [0/187 (0%)]\t training loss: 0.902121\n",
      "epoch: 8 [10/187 (5%)]\t training loss: 1.091248\n",
      "epoch: 8 [20/187 (11%)]\t training loss: 0.851695\n",
      "epoch: 8 [30/187 (16%)]\t training loss: 0.868059\n",
      "epoch: 8 [40/187 (21%)]\t training loss: 1.050956\n",
      "epoch: 8 [50/187 (27%)]\t training loss: 0.824442\n",
      "epoch: 8 [60/187 (32%)]\t training loss: 0.786917\n",
      "epoch: 8 [70/187 (37%)]\t training loss: 1.071474\n",
      "epoch: 8 [80/187 (43%)]\t training loss: 0.971886\n",
      "epoch: 8 [90/187 (48%)]\t training loss: 0.489406\n",
      "epoch: 8 [100/187 (53%)]\t training loss: 0.595554\n",
      "epoch: 8 [110/187 (59%)]\t training loss: 1.223925\n",
      "epoch: 8 [120/187 (64%)]\t training loss: 0.905673\n",
      "epoch: 8 [130/187 (70%)]\t training loss: 0.560419\n",
      "epoch: 8 [140/187 (75%)]\t training loss: 1.030601\n",
      "epoch: 8 [150/187 (80%)]\t training loss: 0.238482\n",
      "epoch: 8 [160/187 (86%)]\t training loss: 0.844498\n",
      "epoch: 8 [170/187 (91%)]\t training loss: 0.773160\n",
      "epoch: 8 [180/187 (96%)]\t training loss: 0.987639\n",
      "\n",
      "\n",
      "Validation...\n",
      "\n",
      "Test dataset: Overall Loss: 0.3110, Overall Accuracy: 2701/2977 (91%)\n",
      "\n",
      "epoch: 9 [0/187 (0%)]\t training loss: 0.978180\n",
      "epoch: 9 [10/187 (5%)]\t training loss: 0.847048\n",
      "epoch: 9 [20/187 (11%)]\t training loss: 0.906619\n",
      "epoch: 9 [30/187 (16%)]\t training loss: 0.620426\n",
      "epoch: 9 [40/187 (21%)]\t training loss: 0.754787\n",
      "epoch: 9 [50/187 (27%)]\t training loss: 0.814760\n",
      "epoch: 9 [60/187 (32%)]\t training loss: 0.464965\n",
      "epoch: 9 [70/187 (37%)]\t training loss: 0.516944\n",
      "epoch: 9 [80/187 (43%)]\t training loss: 0.824334\n",
      "epoch: 9 [90/187 (48%)]\t training loss: 0.520886\n",
      "epoch: 9 [100/187 (53%)]\t training loss: 0.427251\n",
      "epoch: 9 [110/187 (59%)]\t training loss: 0.846444\n",
      "epoch: 9 [120/187 (64%)]\t training loss: 0.512815\n",
      "epoch: 9 [130/187 (70%)]\t training loss: 0.746175\n",
      "epoch: 9 [140/187 (75%)]\t training loss: 0.571271\n",
      "epoch: 9 [150/187 (80%)]\t training loss: 0.650162\n",
      "epoch: 9 [160/187 (86%)]\t training loss: 0.734510\n",
      "epoch: 9 [170/187 (91%)]\t training loss: 0.116602\n",
      "epoch: 9 [180/187 (96%)]\t training loss: 0.563716\n",
      "\n",
      "\n",
      "Validation...\n",
      "\n",
      "Test dataset: Overall Loss: 0.3098, Overall Accuracy: 2709/2977 (91%)\n",
      "\n",
      "epoch: 10 [0/187 (0%)]\t training loss: 0.394310\n",
      "epoch: 10 [10/187 (5%)]\t training loss: 1.029331\n",
      "epoch: 10 [20/187 (11%)]\t training loss: 0.939761\n",
      "epoch: 10 [30/187 (16%)]\t training loss: 0.674566\n",
      "epoch: 10 [40/187 (21%)]\t training loss: 0.856495\n",
      "epoch: 10 [50/187 (27%)]\t training loss: 0.905093\n",
      "epoch: 10 [60/187 (32%)]\t training loss: 0.802826\n",
      "epoch: 10 [70/187 (37%)]\t training loss: 0.839549\n",
      "epoch: 10 [80/187 (43%)]\t training loss: 0.949133\n",
      "epoch: 10 [90/187 (48%)]\t training loss: 0.724957\n",
      "epoch: 10 [100/187 (53%)]\t training loss: 0.794101\n",
      "epoch: 10 [110/187 (59%)]\t training loss: 0.866098\n",
      "epoch: 10 [120/187 (64%)]\t training loss: 0.692225\n",
      "epoch: 10 [130/187 (70%)]\t training loss: 0.946741\n",
      "epoch: 10 [140/187 (75%)]\t training loss: 0.510525\n",
      "epoch: 10 [150/187 (80%)]\t training loss: 1.058283\n",
      "epoch: 10 [160/187 (86%)]\t training loss: 0.661427\n",
      "epoch: 10 [170/187 (91%)]\t training loss: 0.489494\n",
      "epoch: 10 [180/187 (96%)]\t training loss: 0.452323\n",
      "\n",
      "\n",
      "Validation...\n",
      "\n",
      "Test dataset: Overall Loss: 0.3269, Overall Accuracy: 2686/2977 (90%)\n",
      "\n",
      "epoch: 11 [0/187 (0%)]\t training loss: 0.730433\n",
      "epoch: 11 [10/187 (5%)]\t training loss: 0.699070\n",
      "epoch: 11 [20/187 (11%)]\t training loss: 0.561354\n",
      "epoch: 11 [30/187 (16%)]\t training loss: 0.833877\n",
      "epoch: 11 [40/187 (21%)]\t training loss: 0.840427\n",
      "epoch: 11 [50/187 (27%)]\t training loss: 0.566785\n",
      "epoch: 11 [60/187 (32%)]\t training loss: 0.733798\n",
      "epoch: 11 [70/187 (37%)]\t training loss: 0.514841\n",
      "epoch: 11 [80/187 (43%)]\t training loss: 0.359448\n",
      "epoch: 11 [90/187 (48%)]\t training loss: 0.884710\n",
      "epoch: 11 [100/187 (53%)]\t training loss: 0.804413\n",
      "epoch: 11 [110/187 (59%)]\t training loss: 0.861322\n",
      "epoch: 11 [120/187 (64%)]\t training loss: 0.891344\n",
      "epoch: 11 [130/187 (70%)]\t training loss: 0.432221\n",
      "epoch: 11 [140/187 (75%)]\t training loss: 0.765732\n",
      "epoch: 11 [150/187 (80%)]\t training loss: 0.949418\n",
      "epoch: 11 [160/187 (86%)]\t training loss: 0.876663\n",
      "epoch: 11 [170/187 (91%)]\t training loss: 0.937917\n",
      "epoch: 11 [180/187 (96%)]\t training loss: 0.277141\n",
      "\n",
      "\n",
      "Validation...\n",
      "\n",
      "Test dataset: Overall Loss: 0.2938, Overall Accuracy: 2742/2977 (92%)\n",
      "\n",
      "epoch: 12 [0/187 (0%)]\t training loss: 0.895448\n",
      "epoch: 12 [10/187 (5%)]\t training loss: 0.683518\n",
      "epoch: 12 [20/187 (11%)]\t training loss: 0.722913\n",
      "epoch: 12 [30/187 (16%)]\t training loss: 0.691820\n",
      "epoch: 12 [40/187 (21%)]\t training loss: 0.528550\n",
      "epoch: 12 [50/187 (27%)]\t training loss: 0.968927\n",
      "epoch: 12 [60/187 (32%)]\t training loss: 0.870631\n",
      "epoch: 12 [70/187 (37%)]\t training loss: 0.880856\n",
      "epoch: 12 [80/187 (43%)]\t training loss: 0.970904\n",
      "epoch: 12 [90/187 (48%)]\t training loss: 0.457876\n",
      "epoch: 12 [100/187 (53%)]\t training loss: 0.916101\n",
      "epoch: 12 [110/187 (59%)]\t training loss: 0.751793\n",
      "epoch: 12 [120/187 (64%)]\t training loss: 0.811696\n",
      "epoch: 12 [130/187 (70%)]\t training loss: 0.954206\n",
      "epoch: 12 [140/187 (75%)]\t training loss: 0.405532\n",
      "epoch: 12 [150/187 (80%)]\t training loss: 1.010455\n",
      "epoch: 12 [160/187 (86%)]\t training loss: 0.614545\n",
      "epoch: 12 [170/187 (91%)]\t training loss: 0.997315\n",
      "epoch: 12 [180/187 (96%)]\t training loss: 0.737338\n",
      "\n",
      "\n",
      "Validation...\n",
      "\n",
      "Test dataset: Overall Loss: 0.3168, Overall Accuracy: 2718/2977 (91%)\n",
      "\n",
      "epoch: 13 [0/187 (0%)]\t training loss: 1.012138\n",
      "epoch: 13 [10/187 (5%)]\t training loss: 0.312088\n",
      "epoch: 13 [20/187 (11%)]\t training loss: 0.701899\n",
      "epoch: 13 [30/187 (16%)]\t training loss: 0.914325\n",
      "epoch: 13 [40/187 (21%)]\t training loss: 0.569826\n",
      "epoch: 13 [50/187 (27%)]\t training loss: 0.719062\n",
      "epoch: 13 [60/187 (32%)]\t training loss: 0.854151\n",
      "epoch: 13 [70/187 (37%)]\t training loss: 0.887463\n",
      "epoch: 13 [80/187 (43%)]\t training loss: 0.497314\n",
      "epoch: 13 [90/187 (48%)]\t training loss: 0.936778\n",
      "epoch: 13 [100/187 (53%)]\t training loss: 0.895589\n",
      "epoch: 13 [110/187 (59%)]\t training loss: 0.857789\n",
      "epoch: 13 [120/187 (64%)]\t training loss: 0.647251\n",
      "epoch: 13 [130/187 (70%)]\t training loss: 0.874050\n",
      "epoch: 13 [140/187 (75%)]\t training loss: 0.838776\n",
      "epoch: 13 [150/187 (80%)]\t training loss: 1.013199\n",
      "epoch: 13 [160/187 (86%)]\t training loss: 0.979799\n",
      "epoch: 13 [170/187 (91%)]\t training loss: 0.991731\n",
      "epoch: 13 [180/187 (96%)]\t training loss: 0.480824\n",
      "\n",
      "\n",
      "Validation...\n",
      "\n",
      "Test dataset: Overall Loss: 0.2928, Overall Accuracy: 2710/2977 (91%)\n",
      "\n",
      "epoch: 14 [0/187 (0%)]\t training loss: 0.976414\n",
      "epoch: 14 [10/187 (5%)]\t training loss: 0.697823\n",
      "epoch: 14 [20/187 (11%)]\t training loss: 0.892535\n",
      "epoch: 14 [30/187 (16%)]\t training loss: 0.798302\n",
      "epoch: 14 [40/187 (21%)]\t training loss: 0.326588\n",
      "epoch: 14 [50/187 (27%)]\t training loss: 0.707921\n",
      "epoch: 14 [60/187 (32%)]\t training loss: 0.877762\n",
      "epoch: 14 [70/187 (37%)]\t training loss: 0.527481\n",
      "epoch: 14 [80/187 (43%)]\t training loss: 0.938397\n",
      "epoch: 14 [90/187 (48%)]\t training loss: 0.155955\n",
      "epoch: 14 [100/187 (53%)]\t training loss: 0.191565\n",
      "epoch: 14 [110/187 (59%)]\t training loss: 0.963834\n",
      "epoch: 14 [120/187 (64%)]\t training loss: 0.901813\n",
      "epoch: 14 [130/187 (70%)]\t training loss: 0.894534\n",
      "epoch: 14 [140/187 (75%)]\t training loss: 0.804904\n",
      "epoch: 14 [150/187 (80%)]\t training loss: 0.824503\n",
      "epoch: 14 [160/187 (86%)]\t training loss: 0.985140\n",
      "epoch: 14 [170/187 (91%)]\t training loss: 0.927639\n",
      "epoch: 14 [180/187 (96%)]\t training loss: 0.859572\n",
      "\n",
      "\n",
      "Validation...\n",
      "\n",
      "Test dataset: Overall Loss: 0.2933, Overall Accuracy: 2738/2977 (92%)\n",
      "\n",
      "epoch: 15 [0/187 (0%)]\t training loss: 0.912186\n",
      "epoch: 15 [10/187 (5%)]\t training loss: 0.903749\n",
      "epoch: 15 [20/187 (11%)]\t training loss: 0.574223\n",
      "epoch: 15 [30/187 (16%)]\t training loss: 0.859646\n",
      "epoch: 15 [40/187 (21%)]\t training loss: 0.719836\n",
      "epoch: 15 [50/187 (27%)]\t training loss: 0.620396\n",
      "epoch: 15 [60/187 (32%)]\t training loss: 0.827702\n",
      "epoch: 15 [70/187 (37%)]\t training loss: 0.772121\n",
      "epoch: 15 [80/187 (43%)]\t training loss: 0.565418\n",
      "epoch: 15 [90/187 (48%)]\t training loss: 0.694241\n",
      "epoch: 15 [100/187 (53%)]\t training loss: 0.220150\n",
      "epoch: 15 [110/187 (59%)]\t training loss: 0.855326\n",
      "epoch: 15 [120/187 (64%)]\t training loss: 0.897469\n",
      "epoch: 15 [130/187 (70%)]\t training loss: 1.007949\n",
      "epoch: 15 [140/187 (75%)]\t training loss: 0.687687\n",
      "epoch: 15 [150/187 (80%)]\t training loss: 0.831893\n",
      "epoch: 15 [160/187 (86%)]\t training loss: 0.590312\n",
      "epoch: 15 [170/187 (91%)]\t training loss: 0.790588\n",
      "epoch: 15 [180/187 (96%)]\t training loss: 0.918416\n",
      "\n",
      "\n",
      "Validation...\n",
      "\n",
      "Test dataset: Overall Loss: 0.2666, Overall Accuracy: 2750/2977 (92%)\n",
      "\n",
      "epoch: 16 [0/187 (0%)]\t training loss: 0.891474\n",
      "epoch: 16 [10/187 (5%)]\t training loss: 0.860198\n",
      "epoch: 16 [20/187 (11%)]\t training loss: 0.757954\n",
      "epoch: 16 [30/187 (16%)]\t training loss: 0.896924\n",
      "epoch: 16 [40/187 (21%)]\t training loss: 0.804835\n",
      "epoch: 16 [50/187 (27%)]\t training loss: 0.873409\n",
      "epoch: 16 [60/187 (32%)]\t training loss: 0.883520\n",
      "epoch: 16 [70/187 (37%)]\t training loss: 0.853261\n",
      "epoch: 16 [80/187 (43%)]\t training loss: 0.828486\n",
      "epoch: 16 [90/187 (48%)]\t training loss: 1.013780\n",
      "epoch: 16 [100/187 (53%)]\t training loss: 0.489957\n",
      "epoch: 16 [110/187 (59%)]\t training loss: 0.741064\n",
      "epoch: 16 [120/187 (64%)]\t training loss: 0.864338\n",
      "epoch: 16 [130/187 (70%)]\t training loss: 0.728041\n",
      "epoch: 16 [140/187 (75%)]\t training loss: 0.933745\n",
      "epoch: 16 [150/187 (80%)]\t training loss: 0.519807\n",
      "epoch: 16 [160/187 (86%)]\t training loss: 0.362930\n",
      "epoch: 16 [170/187 (91%)]\t training loss: 0.887791\n",
      "epoch: 16 [180/187 (96%)]\t training loss: 0.884191\n",
      "\n",
      "\n",
      "Validation...\n",
      "\n",
      "Test dataset: Overall Loss: 0.2758, Overall Accuracy: 2754/2977 (93%)\n",
      "\n",
      "epoch: 17 [0/187 (0%)]\t training loss: 0.825503\n",
      "epoch: 17 [10/187 (5%)]\t training loss: 0.866082\n",
      "epoch: 17 [20/187 (11%)]\t training loss: 0.899286\n",
      "epoch: 17 [30/187 (16%)]\t training loss: 0.615920\n",
      "epoch: 17 [40/187 (21%)]\t training loss: 0.193311\n",
      "epoch: 17 [50/187 (27%)]\t training loss: 0.551288\n",
      "epoch: 17 [60/187 (32%)]\t training loss: 0.919921\n",
      "epoch: 17 [70/187 (37%)]\t training loss: 0.293232\n",
      "epoch: 17 [80/187 (43%)]\t training loss: 0.952513\n",
      "epoch: 17 [90/187 (48%)]\t training loss: 0.792623\n",
      "epoch: 17 [100/187 (53%)]\t training loss: 0.615192\n",
      "epoch: 17 [110/187 (59%)]\t training loss: 1.073719\n",
      "epoch: 17 [120/187 (64%)]\t training loss: 0.619737\n",
      "epoch: 17 [130/187 (70%)]\t training loss: 0.340074\n",
      "epoch: 17 [140/187 (75%)]\t training loss: 0.250712\n",
      "epoch: 17 [150/187 (80%)]\t training loss: 0.768989\n",
      "epoch: 17 [160/187 (86%)]\t training loss: 0.855199\n",
      "epoch: 17 [170/187 (91%)]\t training loss: 0.737184\n",
      "epoch: 17 [180/187 (96%)]\t training loss: 0.171933\n",
      "\n",
      "\n",
      "Validation...\n",
      "\n",
      "Test dataset: Overall Loss: 0.2638, Overall Accuracy: 2742/2977 (92%)\n",
      "\n",
      "epoch: 18 [0/187 (0%)]\t training loss: 0.174875\n",
      "epoch: 18 [10/187 (5%)]\t training loss: 0.686680\n",
      "epoch: 18 [20/187 (11%)]\t training loss: 0.856724\n",
      "epoch: 18 [30/187 (16%)]\t training loss: 0.309018\n",
      "epoch: 18 [40/187 (21%)]\t training loss: 0.748026\n",
      "epoch: 18 [50/187 (27%)]\t training loss: 0.629282\n",
      "epoch: 18 [60/187 (32%)]\t training loss: 0.369030\n",
      "epoch: 18 [70/187 (37%)]\t training loss: 0.701024\n",
      "epoch: 18 [80/187 (43%)]\t training loss: 0.864639\n",
      "epoch: 18 [90/187 (48%)]\t training loss: 0.531156\n",
      "epoch: 18 [100/187 (53%)]\t training loss: 0.249980\n",
      "epoch: 18 [110/187 (59%)]\t training loss: 0.928648\n",
      "epoch: 18 [120/187 (64%)]\t training loss: 0.722393\n",
      "epoch: 18 [130/187 (70%)]\t training loss: 0.598870\n",
      "epoch: 18 [140/187 (75%)]\t training loss: 0.653137\n",
      "epoch: 18 [150/187 (80%)]\t training loss: 0.720150\n",
      "epoch: 18 [160/187 (86%)]\t training loss: 0.266912\n",
      "epoch: 18 [170/187 (91%)]\t training loss: 0.844016\n",
      "epoch: 18 [180/187 (96%)]\t training loss: 0.508627\n",
      "\n",
      "\n",
      "Validation...\n",
      "\n",
      "Test dataset: Overall Loss: 0.2717, Overall Accuracy: 2741/2977 (92%)\n",
      "\n",
      "epoch: 19 [0/187 (0%)]\t training loss: 0.988650\n",
      "epoch: 19 [10/187 (5%)]\t training loss: 0.913293\n",
      "epoch: 19 [20/187 (11%)]\t training loss: 0.430265\n",
      "epoch: 19 [30/187 (16%)]\t training loss: 0.975628\n",
      "epoch: 19 [40/187 (21%)]\t training loss: 1.053092\n",
      "epoch: 19 [50/187 (27%)]\t training loss: 0.733864\n",
      "epoch: 19 [60/187 (32%)]\t training loss: 0.464515\n",
      "epoch: 19 [70/187 (37%)]\t training loss: 0.667905\n",
      "epoch: 19 [80/187 (43%)]\t training loss: 0.821360\n",
      "epoch: 19 [90/187 (48%)]\t training loss: 0.599994\n",
      "epoch: 19 [100/187 (53%)]\t training loss: 0.531616\n",
      "epoch: 19 [110/187 (59%)]\t training loss: 0.506059\n",
      "epoch: 19 [120/187 (64%)]\t training loss: 0.707838\n",
      "epoch: 19 [130/187 (70%)]\t training loss: 0.929657\n",
      "epoch: 19 [140/187 (75%)]\t training loss: 0.404821\n",
      "epoch: 19 [150/187 (80%)]\t training loss: 0.239281\n",
      "epoch: 19 [160/187 (86%)]\t training loss: 0.849056\n",
      "epoch: 19 [170/187 (91%)]\t training loss: 0.949453\n",
      "epoch: 19 [180/187 (96%)]\t training loss: 0.400641\n",
      "\n",
      "\n",
      "Validation...\n",
      "\n",
      "Test dataset: Overall Loss: 0.2547, Overall Accuracy: 2758/2977 (93%)\n",
      "\n",
      "epoch: 20 [0/187 (0%)]\t training loss: 0.650018\n",
      "epoch: 20 [10/187 (5%)]\t training loss: 0.774870\n",
      "epoch: 20 [20/187 (11%)]\t training loss: 0.798623\n",
      "epoch: 20 [30/187 (16%)]\t training loss: 0.686258\n",
      "epoch: 20 [40/187 (21%)]\t training loss: 0.823103\n",
      "epoch: 20 [50/187 (27%)]\t training loss: 0.855515\n",
      "epoch: 20 [60/187 (32%)]\t training loss: 0.673885\n",
      "epoch: 20 [70/187 (37%)]\t training loss: 0.156819\n",
      "epoch: 20 [80/187 (43%)]\t training loss: 0.718053\n",
      "epoch: 20 [90/187 (48%)]\t training loss: 0.663642\n",
      "epoch: 20 [100/187 (53%)]\t training loss: 0.928414\n",
      "epoch: 20 [110/187 (59%)]\t training loss: 0.730470\n",
      "epoch: 20 [120/187 (64%)]\t training loss: 0.862863\n",
      "epoch: 20 [130/187 (70%)]\t training loss: 0.683215\n",
      "epoch: 20 [140/187 (75%)]\t training loss: 1.024167\n",
      "epoch: 20 [150/187 (80%)]\t training loss: 0.832490\n",
      "epoch: 20 [160/187 (86%)]\t training loss: 0.836313\n",
      "epoch: 20 [170/187 (91%)]\t training loss: 0.715010\n",
      "epoch: 20 [180/187 (96%)]\t training loss: 0.404131\n",
      "\n",
      "\n",
      "Validation...\n",
      "\n",
      "Test dataset: Overall Loss: 0.2808, Overall Accuracy: 2745/2977 (92%)\n",
      "\n",
      "epoch: 21 [0/187 (0%)]\t training loss: 0.827513\n",
      "epoch: 21 [10/187 (5%)]\t training loss: 0.905772\n",
      "epoch: 21 [20/187 (11%)]\t training loss: 0.716270\n",
      "epoch: 21 [30/187 (16%)]\t training loss: 0.798058\n",
      "epoch: 21 [40/187 (21%)]\t training loss: 0.765742\n",
      "epoch: 21 [50/187 (27%)]\t training loss: 0.539780\n",
      "epoch: 21 [60/187 (32%)]\t training loss: 0.849855\n",
      "epoch: 21 [70/187 (37%)]\t training loss: 0.831352\n",
      "epoch: 21 [80/187 (43%)]\t training loss: 0.689093\n",
      "epoch: 21 [90/187 (48%)]\t training loss: 0.611449\n",
      "epoch: 21 [100/187 (53%)]\t training loss: 0.909790\n",
      "epoch: 21 [110/187 (59%)]\t training loss: 0.845975\n",
      "epoch: 21 [120/187 (64%)]\t training loss: 0.976666\n",
      "epoch: 21 [130/187 (70%)]\t training loss: 0.778048\n",
      "epoch: 21 [140/187 (75%)]\t training loss: 0.336972\n",
      "epoch: 21 [150/187 (80%)]\t training loss: 0.323648\n",
      "epoch: 21 [160/187 (86%)]\t training loss: 0.668392\n",
      "epoch: 21 [170/187 (91%)]\t training loss: 0.781160\n",
      "epoch: 21 [180/187 (96%)]\t training loss: 0.241854\n",
      "\n",
      "\n",
      "Validation...\n",
      "\n",
      "Test dataset: Overall Loss: 0.2849, Overall Accuracy: 2742/2977 (92%)\n",
      "\n",
      "epoch: 22 [0/187 (0%)]\t training loss: 0.903389\n",
      "epoch: 22 [10/187 (5%)]\t training loss: 1.025966\n",
      "epoch: 22 [20/187 (11%)]\t training loss: 0.373378\n",
      "epoch: 22 [30/187 (16%)]\t training loss: 0.777757\n",
      "epoch: 22 [40/187 (21%)]\t training loss: 0.830777\n",
      "epoch: 22 [50/187 (27%)]\t training loss: 0.283336\n",
      "epoch: 22 [60/187 (32%)]\t training loss: 0.401403\n",
      "epoch: 22 [70/187 (37%)]\t training loss: 0.615419\n",
      "epoch: 22 [80/187 (43%)]\t training loss: 1.099111\n",
      "epoch: 22 [90/187 (48%)]\t training loss: 0.245142\n",
      "epoch: 22 [100/187 (53%)]\t training loss: 0.310244\n",
      "epoch: 22 [110/187 (59%)]\t training loss: 0.765062\n",
      "epoch: 22 [120/187 (64%)]\t training loss: 0.712051\n",
      "epoch: 22 [130/187 (70%)]\t training loss: 0.394168\n",
      "epoch: 22 [140/187 (75%)]\t training loss: 0.885162\n",
      "epoch: 22 [150/187 (80%)]\t training loss: 0.869515\n",
      "epoch: 22 [160/187 (86%)]\t training loss: 0.702806\n",
      "epoch: 22 [170/187 (91%)]\t training loss: 0.235004\n",
      "epoch: 22 [180/187 (96%)]\t training loss: 1.000630\n",
      "\n",
      "\n",
      "Validation...\n",
      "\n",
      "Test dataset: Overall Loss: 0.2756, Overall Accuracy: 2740/2977 (92%)\n",
      "\n",
      "epoch: 23 [0/187 (0%)]\t training loss: 0.817257\n",
      "epoch: 23 [10/187 (5%)]\t training loss: 0.727006\n",
      "epoch: 23 [20/187 (11%)]\t training loss: 0.569352\n",
      "epoch: 23 [30/187 (16%)]\t training loss: 0.291601\n",
      "epoch: 23 [40/187 (21%)]\t training loss: 0.576194\n",
      "epoch: 23 [50/187 (27%)]\t training loss: 0.860625\n",
      "epoch: 23 [60/187 (32%)]\t training loss: 0.746999\n",
      "epoch: 23 [70/187 (37%)]\t training loss: 0.472946\n",
      "epoch: 23 [80/187 (43%)]\t training loss: 0.250109\n",
      "epoch: 23 [90/187 (48%)]\t training loss: 0.557485\n",
      "epoch: 23 [100/187 (53%)]\t training loss: 0.243663\n",
      "epoch: 23 [110/187 (59%)]\t training loss: 0.814634\n",
      "epoch: 23 [120/187 (64%)]\t training loss: 0.896301\n",
      "epoch: 23 [130/187 (70%)]\t training loss: 0.841420\n",
      "epoch: 23 [140/187 (75%)]\t training loss: 0.325296\n",
      "epoch: 23 [150/187 (80%)]\t training loss: 0.901367\n",
      "epoch: 23 [160/187 (86%)]\t training loss: 0.883271\n",
      "epoch: 23 [170/187 (91%)]\t training loss: 0.705765\n",
      "epoch: 23 [180/187 (96%)]\t training loss: 0.611283\n",
      "\n",
      "\n",
      "Validation...\n",
      "\n",
      "Test dataset: Overall Loss: 0.2650, Overall Accuracy: 2753/2977 (92%)\n",
      "\n",
      "epoch: 24 [0/187 (0%)]\t training loss: 0.121911\n",
      "epoch: 24 [10/187 (5%)]\t training loss: 0.908295\n",
      "epoch: 24 [20/187 (11%)]\t training loss: 0.243675\n",
      "epoch: 24 [30/187 (16%)]\t training loss: 0.283582\n",
      "epoch: 24 [40/187 (21%)]\t training loss: 0.859880\n",
      "epoch: 24 [50/187 (27%)]\t training loss: 0.332465\n",
      "epoch: 24 [60/187 (32%)]\t training loss: 0.844977\n",
      "epoch: 24 [70/187 (37%)]\t training loss: 0.841199\n",
      "epoch: 24 [80/187 (43%)]\t training loss: 0.909242\n",
      "epoch: 24 [90/187 (48%)]\t training loss: 0.342693\n",
      "epoch: 24 [100/187 (53%)]\t training loss: 0.695379\n",
      "epoch: 24 [110/187 (59%)]\t training loss: 0.534535\n",
      "epoch: 24 [120/187 (64%)]\t training loss: 0.154731\n",
      "epoch: 24 [130/187 (70%)]\t training loss: 0.763463\n",
      "epoch: 24 [140/187 (75%)]\t training loss: 0.774853\n",
      "epoch: 24 [150/187 (80%)]\t training loss: 0.576827\n",
      "epoch: 24 [160/187 (86%)]\t training loss: 0.853739\n",
      "epoch: 24 [170/187 (91%)]\t training loss: 0.789741\n",
      "epoch: 24 [180/187 (96%)]\t training loss: 0.793219\n",
      "\n",
      "\n",
      "Validation...\n",
      "\n",
      "Test dataset: Overall Loss: 0.2645, Overall Accuracy: 2754/2977 (93%)\n",
      "\n",
      "epoch: 25 [0/187 (0%)]\t training loss: 0.287490\n",
      "epoch: 25 [10/187 (5%)]\t training loss: 0.507637\n",
      "epoch: 25 [20/187 (11%)]\t training loss: 0.627669\n",
      "epoch: 25 [30/187 (16%)]\t training loss: 0.161901\n",
      "epoch: 25 [40/187 (21%)]\t training loss: 0.956216\n",
      "epoch: 25 [50/187 (27%)]\t training loss: 0.539855\n",
      "epoch: 25 [60/187 (32%)]\t training loss: 0.876008\n",
      "epoch: 25 [70/187 (37%)]\t training loss: 0.922968\n",
      "epoch: 25 [80/187 (43%)]\t training loss: 0.651276\n",
      "epoch: 25 [90/187 (48%)]\t training loss: 0.108015\n",
      "epoch: 25 [100/187 (53%)]\t training loss: 0.923091\n",
      "epoch: 25 [110/187 (59%)]\t training loss: 0.592308\n",
      "epoch: 25 [120/187 (64%)]\t training loss: 0.950261\n",
      "epoch: 25 [130/187 (70%)]\t training loss: 0.302437\n",
      "epoch: 25 [140/187 (75%)]\t training loss: 0.827793\n",
      "epoch: 25 [150/187 (80%)]\t training loss: 0.597740\n",
      "epoch: 25 [160/187 (86%)]\t training loss: 0.907613\n",
      "epoch: 25 [170/187 (91%)]\t training loss: 0.317153\n",
      "epoch: 25 [180/187 (96%)]\t training loss: 0.743399\n",
      "\n",
      "\n",
      "Validation...\n",
      "\n",
      "Test dataset: Overall Loss: 0.2600, Overall Accuracy: 2747/2977 (92%)\n",
      "\n",
      "epoch: 26 [0/187 (0%)]\t training loss: 0.197031\n",
      "epoch: 26 [10/187 (5%)]\t training loss: 0.964056\n",
      "epoch: 26 [20/187 (11%)]\t training loss: 0.861499\n",
      "epoch: 26 [30/187 (16%)]\t training loss: 0.602136\n",
      "epoch: 26 [40/187 (21%)]\t training loss: 0.910751\n",
      "epoch: 26 [50/187 (27%)]\t training loss: 0.956959\n",
      "epoch: 26 [60/187 (32%)]\t training loss: 0.746184\n",
      "epoch: 26 [70/187 (37%)]\t training loss: 0.777451\n",
      "epoch: 26 [80/187 (43%)]\t training loss: 0.437214\n",
      "epoch: 26 [90/187 (48%)]\t training loss: 0.747860\n",
      "epoch: 26 [100/187 (53%)]\t training loss: 0.865739\n",
      "epoch: 26 [110/187 (59%)]\t training loss: 0.654842\n",
      "epoch: 26 [120/187 (64%)]\t training loss: 0.871867\n",
      "epoch: 26 [130/187 (70%)]\t training loss: 0.675994\n",
      "epoch: 26 [140/187 (75%)]\t training loss: 0.406582\n",
      "epoch: 26 [150/187 (80%)]\t training loss: 0.919411\n",
      "epoch: 26 [160/187 (86%)]\t training loss: 0.294294\n",
      "epoch: 26 [170/187 (91%)]\t training loss: 0.537573\n",
      "epoch: 26 [180/187 (96%)]\t training loss: 0.553539\n",
      "\n",
      "\n",
      "Validation...\n",
      "\n",
      "Test dataset: Overall Loss: 0.2690, Overall Accuracy: 2783/2977 (93%)\n",
      "\n",
      "epoch: 27 [0/187 (0%)]\t training loss: 0.817285\n",
      "epoch: 27 [10/187 (5%)]\t training loss: 0.966064\n",
      "epoch: 27 [20/187 (11%)]\t training loss: 0.354591\n",
      "epoch: 27 [30/187 (16%)]\t training loss: 0.691300\n",
      "epoch: 27 [40/187 (21%)]\t training loss: 0.790709\n",
      "epoch: 27 [50/187 (27%)]\t training loss: 0.782749\n",
      "epoch: 27 [60/187 (32%)]\t training loss: 0.812688\n",
      "epoch: 27 [70/187 (37%)]\t training loss: 0.920959\n",
      "epoch: 27 [80/187 (43%)]\t training loss: 0.835208\n",
      "epoch: 27 [90/187 (48%)]\t training loss: 0.888300\n",
      "epoch: 27 [100/187 (53%)]\t training loss: 1.003355\n",
      "epoch: 27 [110/187 (59%)]\t training loss: 0.945088\n",
      "epoch: 27 [120/187 (64%)]\t training loss: 0.821904\n",
      "epoch: 27 [130/187 (70%)]\t training loss: 0.685076\n",
      "epoch: 27 [140/187 (75%)]\t training loss: 0.962626\n",
      "epoch: 27 [150/187 (80%)]\t training loss: 0.685941\n",
      "epoch: 27 [160/187 (86%)]\t training loss: 0.766287\n",
      "epoch: 27 [170/187 (91%)]\t training loss: 0.812849\n",
      "epoch: 27 [180/187 (96%)]\t training loss: 0.837614\n",
      "\n",
      "\n",
      "Validation...\n",
      "\n",
      "Test dataset: Overall Loss: 0.3076, Overall Accuracy: 2750/2977 (92%)\n",
      "\n",
      "epoch: 28 [0/187 (0%)]\t training loss: 0.884755\n",
      "epoch: 28 [10/187 (5%)]\t training loss: 1.011101\n",
      "epoch: 28 [20/187 (11%)]\t training loss: 0.703932\n",
      "epoch: 28 [30/187 (16%)]\t training loss: 0.740882\n",
      "epoch: 28 [40/187 (21%)]\t training loss: 0.155249\n",
      "epoch: 28 [50/187 (27%)]\t training loss: 0.220083\n",
      "epoch: 28 [60/187 (32%)]\t training loss: 0.753369\n",
      "epoch: 28 [70/187 (37%)]\t training loss: 0.852356\n",
      "epoch: 28 [80/187 (43%)]\t training loss: 0.820644\n",
      "epoch: 28 [90/187 (48%)]\t training loss: 0.197573\n",
      "epoch: 28 [100/187 (53%)]\t training loss: 0.704499\n",
      "epoch: 28 [110/187 (59%)]\t training loss: 0.679896\n",
      "epoch: 28 [120/187 (64%)]\t training loss: 0.962768\n",
      "epoch: 28 [130/187 (70%)]\t training loss: 0.882411\n",
      "epoch: 28 [140/187 (75%)]\t training loss: 0.636819\n",
      "epoch: 28 [150/187 (80%)]\t training loss: 0.872199\n",
      "epoch: 28 [160/187 (86%)]\t training loss: 0.569318\n",
      "epoch: 28 [170/187 (91%)]\t training loss: 0.714695\n",
      "epoch: 28 [180/187 (96%)]\t training loss: 0.645895\n",
      "\n",
      "\n",
      "Validation...\n",
      "\n",
      "Test dataset: Overall Loss: 0.2641, Overall Accuracy: 2742/2977 (92%)\n",
      "\n",
      "epoch: 29 [0/187 (0%)]\t training loss: 0.743663\n",
      "epoch: 29 [10/187 (5%)]\t training loss: 0.921715\n",
      "epoch: 29 [20/187 (11%)]\t training loss: 0.766166\n",
      "epoch: 29 [30/187 (16%)]\t training loss: 0.956860\n",
      "epoch: 29 [40/187 (21%)]\t training loss: 0.872726\n",
      "epoch: 29 [50/187 (27%)]\t training loss: 0.364384\n",
      "epoch: 29 [60/187 (32%)]\t training loss: 0.241129\n",
      "epoch: 29 [70/187 (37%)]\t training loss: 0.375389\n",
      "epoch: 29 [80/187 (43%)]\t training loss: 0.588736\n",
      "epoch: 29 [90/187 (48%)]\t training loss: 0.490805\n",
      "epoch: 29 [100/187 (53%)]\t training loss: 0.611024\n",
      "epoch: 29 [110/187 (59%)]\t training loss: 0.880508\n",
      "epoch: 29 [120/187 (64%)]\t training loss: 1.006672\n",
      "epoch: 29 [130/187 (70%)]\t training loss: 1.008779\n",
      "epoch: 29 [140/187 (75%)]\t training loss: 0.672756\n",
      "epoch: 29 [150/187 (80%)]\t training loss: 0.520246\n",
      "epoch: 29 [160/187 (86%)]\t training loss: 0.856223\n",
      "epoch: 29 [170/187 (91%)]\t training loss: 0.794249\n",
      "epoch: 29 [180/187 (96%)]\t training loss: 0.208740\n",
      "\n",
      "\n",
      "Validation...\n",
      "\n",
      "Test dataset: Overall Loss: 0.2487, Overall Accuracy: 2767/2977 (93%)\n",
      "\n"
     ]
    }
   ],
   "source": [
    "imbalanced_model_cutmix, train_losses, val_losses = train_cut_mix(\n",
    "    imbalanced_train_loader,\n",
    "    imbalanced_valid_loader,\n",
    "    \"imbalanced_model_cutmix.pth\",\n",
    "    ConvNet().to(device),\n",
    "    epochs=30,\n",
    ")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 472
    },
    "id": "lazHBElP0hs8",
    "outputId": "10d688ca-3b8c-4a8d-ee32-521186fa68ff"
   },
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 640x480 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "plt.plot(train_losses)\n",
    "plt.plot(val_losses)\n",
    "plt.xlabel(\"epoch\")\n",
    "plt.ylabel(\"loss\")\n",
    "plt.legend([\"Train\", \"Valid\"])\n",
    "plt.title(\"Model On imbalanced Fashion MNIST - Train vs Valid Accuracy\")\n",
    "\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/"
    },
    "id": "sVvXHlE4HXQn",
    "outputId": "418f53e2-9160-4484-f2a1-e5f0388fdaa7"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "On Imbalanced Fashion MNIST dataset\n",
      "\n",
      "Test dataset: Overall Loss: 0.0012, Overall Accuracy: 8308/10000 (83%)\n",
      "\n",
      "test accuracy :  83.08\n"
     ]
    }
   ],
   "source": [
    "print(\"On Imbalanced Fashion MNIST dataset\")\n",
    "test_acc2, class_percentages2 = test(\n",
    "    imbalanced_model_cutmix, \"imbalanced_model_cutmix.pth\"\n",
    ")\n",
    "print(\"test accuracy : \", str(test_acc2))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "G7iXAxhpqAcJ"
   },
   "source": [
    "## MixUp\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 20,
   "metadata": {
    "id": "oESaTkPrp2ZY"
   },
   "outputs": [],
   "source": [
    "def mixup(data, targets, alpha):\n",
    "    \"\"\"\n",
    "\n",
    "    :param data: torch.Size. Shape:(batch_size, channel, image_size, image_size)\n",
    "    :param targets: torch.Tensor. multitask, multi-classification, target1. shape(batch_size, 1)\n",
    "    :param alpha: float, positive\n",
    "    :return: resulting image, multi-target\n",
    "    \"\"\"\n",
    "    indices = torch.randperm(data.size(0))\n",
    "    shuffled_data = data[indices]\n",
    "    shuffled_targets = targets[indices]\n",
    "\n",
    "    lam = np.random.beta(alpha, alpha)\n",
    "    data = data * lam + shuffled_data * (1 - lam)\n",
    "    targets = [targets, shuffled_targets, lam]\n",
    "\n",
    "    return data, targets"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 21,
   "metadata": {
    "id": "Uju36L5qqVoa"
   },
   "outputs": [],
   "source": [
    "def train_mixup(\n",
    "    train_dataloader,\n",
    "    valid_dataloader,\n",
    "    PATH,\n",
    "    model=None,\n",
    "    criterion=F.cross_entropy,\n",
    "    epochs=1,\n",
    "):\n",
    "    optimizer = torch.optim.SGD(model.parameters(), lr=0.1, momentum=0.9)\n",
    "    model.train()\n",
    "    best_acc = 0\n",
    "    train_losses = []\n",
    "    val_losses = []\n",
    "    for epoch in range(epochs):\n",
    "        losses = 0\n",
    "        for b_i, (X, y) in enumerate(train_dataloader):\n",
    "            X, y = X.to(device), y.to(device)\n",
    "            X, ys = mixup(X, y, alpha=0.5)\n",
    "            pred_prob = model(X)\n",
    "            targets1, targets2, lam = ys[0], ys[1], ys[2]\n",
    "            loss = lam * criterion(pred_prob, targets1) + (1 - lam) * criterion(\n",
    "                pred_prob, targets2\n",
    "            )\n",
    "            optimizer.zero_grad()\n",
    "            loss.backward()\n",
    "            optimizer.step()\n",
    "            losses += loss.item()  # print(losses)\n",
    "            if b_i % 10 == 0:\n",
    "                print(\n",
    "                    \"epoch: {} [{}/{} ({:.0f}%)]\\t training loss: {:.6f}\".format(\n",
    "                        epoch,\n",
    "                        b_i,\n",
    "                        len(train_dataloader),\n",
    "                        100.0 * b_i / len(train_dataloader),\n",
    "                        loss.item(),\n",
    "                    )\n",
    "                )\n",
    "        train_losses.append(losses / len(train_dataloader))\n",
    "        print(\"\\n\")\n",
    "        print(\"Validation...\")\n",
    "        model.eval()\n",
    "        success = 0\n",
    "        total = 0\n",
    "        losses = 0\n",
    "        with torch.no_grad():\n",
    "            for X, y in valid_dataloader:\n",
    "                X, y = X.to(device), y.to(device)\n",
    "                pred_prob = model(X)\n",
    "                loss = criterion(pred_prob, y)  # loss summed across the batch\n",
    "                losses += loss.item()\n",
    "                pred = pred_prob.argmax(\n",
    "                    dim=1, keepdim=True\n",
    "                )  # use argmax to get the most likely prediction\n",
    "                success += pred.eq(y.view_as(pred)).sum().item()\n",
    "                total += len(y)\n",
    "        val_losses.append(losses / len(valid_dataloader))\n",
    "        accuracy = 100.0 * success / total\n",
    "\n",
    "        if accuracy > best_acc:\n",
    "            torch.save(model.state_dict(), PATH)\n",
    "            best_acc = accuracy\n",
    "\n",
    "        print(\n",
    "            \"\\nTest dataset: Overall Loss: {:.4f}, Overall Accuracy: {}/{} ({:.0f}%)\\n\".format(\n",
    "                losses / len(valid_dataloader), success, total, accuracy\n",
    "            )\n",
    "        )\n",
    "    return model, train_losses, val_losses"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 22,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/"
    },
    "id": "Gg5MJ10vqgIN",
    "outputId": "31042fd1-838a-4dc3-fe05-57199092610e"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "epoch: 0 [0/187 (0%)]\t training loss: 2.215507\n",
      "epoch: 0 [10/187 (5%)]\t training loss: 1.454928\n",
      "epoch: 0 [20/187 (11%)]\t training loss: 1.109699\n",
      "epoch: 0 [30/187 (16%)]\t training loss: 1.249764\n",
      "epoch: 0 [40/187 (21%)]\t training loss: 0.730999\n",
      "epoch: 0 [50/187 (27%)]\t training loss: 0.749751\n",
      "epoch: 0 [60/187 (32%)]\t training loss: 0.906633\n",
      "epoch: 0 [70/187 (37%)]\t training loss: 0.822527\n",
      "epoch: 0 [80/187 (43%)]\t training loss: 0.560227\n",
      "epoch: 0 [90/187 (48%)]\t training loss: 0.962724\n",
      "epoch: 0 [100/187 (53%)]\t training loss: 0.585745\n",
      "epoch: 0 [110/187 (59%)]\t training loss: 0.640926\n",
      "epoch: 0 [120/187 (64%)]\t training loss: 1.093047\n",
      "epoch: 0 [130/187 (70%)]\t training loss: 0.760767\n",
      "epoch: 0 [140/187 (75%)]\t training loss: 1.283033\n",
      "epoch: 0 [150/187 (80%)]\t training loss: 0.705525\n",
      "epoch: 0 [160/187 (86%)]\t training loss: 1.221690\n",
      "epoch: 0 [170/187 (91%)]\t training loss: 0.509333\n",
      "epoch: 0 [180/187 (96%)]\t training loss: 0.846154\n",
      "\n",
      "\n",
      "Validation...\n",
      "\n",
      "Test dataset: Overall Loss: 0.4180, Overall Accuracy: 2569/2977 (86%)\n",
      "\n",
      "epoch: 1 [0/187 (0%)]\t training loss: 0.940688\n",
      "epoch: 1 [10/187 (5%)]\t training loss: 0.435349\n",
      "epoch: 1 [20/187 (11%)]\t training loss: 1.116859\n",
      "epoch: 1 [30/187 (16%)]\t training loss: 0.624892\n",
      "epoch: 1 [40/187 (21%)]\t training loss: 0.319901\n",
      "epoch: 1 [50/187 (27%)]\t training loss: 0.782169\n",
      "epoch: 1 [60/187 (32%)]\t training loss: 0.529190\n",
      "epoch: 1 [70/187 (37%)]\t training loss: 0.444647\n",
      "epoch: 1 [80/187 (43%)]\t training loss: 0.358578\n",
      "epoch: 1 [90/187 (48%)]\t training loss: 0.812791\n",
      "epoch: 1 [100/187 (53%)]\t training loss: 0.899428\n",
      "epoch: 1 [110/187 (59%)]\t training loss: 0.923972\n",
      "epoch: 1 [120/187 (64%)]\t training loss: 0.502785\n",
      "epoch: 1 [130/187 (70%)]\t training loss: 0.655114\n",
      "epoch: 1 [140/187 (75%)]\t training loss: 0.328004\n",
      "epoch: 1 [150/187 (80%)]\t training loss: 0.283567\n",
      "epoch: 1 [160/187 (86%)]\t training loss: 0.710914\n",
      "epoch: 1 [170/187 (91%)]\t training loss: 0.778075\n",
      "epoch: 1 [180/187 (96%)]\t training loss: 0.675427\n",
      "\n",
      "\n",
      "Validation...\n",
      "\n",
      "Test dataset: Overall Loss: 0.3350, Overall Accuracy: 2687/2977 (90%)\n",
      "\n",
      "epoch: 2 [0/187 (0%)]\t training loss: 0.955737\n",
      "epoch: 2 [10/187 (5%)]\t training loss: 0.783616\n",
      "epoch: 2 [20/187 (11%)]\t training loss: 0.933963\n",
      "epoch: 2 [30/187 (16%)]\t training loss: 1.138203\n",
      "epoch: 2 [40/187 (21%)]\t training loss: 0.374925\n",
      "epoch: 2 [50/187 (27%)]\t training loss: 0.411471\n",
      "epoch: 2 [60/187 (32%)]\t training loss: 0.749258\n",
      "epoch: 2 [70/187 (37%)]\t training loss: 0.833374\n",
      "epoch: 2 [80/187 (43%)]\t training loss: 0.935079\n",
      "epoch: 2 [90/187 (48%)]\t training loss: 0.817282\n",
      "epoch: 2 [100/187 (53%)]\t training loss: 0.622081\n",
      "epoch: 2 [110/187 (59%)]\t training loss: 0.615841\n",
      "epoch: 2 [120/187 (64%)]\t training loss: 0.535316\n",
      "epoch: 2 [130/187 (70%)]\t training loss: 0.916487\n",
      "epoch: 2 [140/187 (75%)]\t training loss: 0.749472\n",
      "epoch: 2 [150/187 (80%)]\t training loss: 0.298909\n",
      "epoch: 2 [160/187 (86%)]\t training loss: 0.856469\n",
      "epoch: 2 [170/187 (91%)]\t training loss: 0.639206\n",
      "epoch: 2 [180/187 (96%)]\t training loss: 0.935438\n",
      "\n",
      "\n",
      "Validation...\n",
      "\n",
      "Test dataset: Overall Loss: 0.3301, Overall Accuracy: 2680/2977 (90%)\n",
      "\n",
      "epoch: 3 [0/187 (0%)]\t training loss: 0.646498\n",
      "epoch: 3 [10/187 (5%)]\t training loss: 0.550917\n",
      "epoch: 3 [20/187 (11%)]\t training loss: 0.791672\n",
      "epoch: 3 [30/187 (16%)]\t training loss: 0.867225\n",
      "epoch: 3 [40/187 (21%)]\t training loss: 0.435420\n",
      "epoch: 3 [50/187 (27%)]\t training loss: 0.767641\n",
      "epoch: 3 [60/187 (32%)]\t training loss: 0.786713\n",
      "epoch: 3 [70/187 (37%)]\t training loss: 0.893352\n",
      "epoch: 3 [80/187 (43%)]\t training loss: 0.600332\n",
      "epoch: 3 [90/187 (48%)]\t training loss: 1.166633\n",
      "epoch: 3 [100/187 (53%)]\t training loss: 0.368138\n",
      "epoch: 3 [110/187 (59%)]\t training loss: 0.607127\n",
      "epoch: 3 [120/187 (64%)]\t training loss: 0.628068\n",
      "epoch: 3 [130/187 (70%)]\t training loss: 0.710395\n",
      "epoch: 3 [140/187 (75%)]\t training loss: 0.616381\n",
      "epoch: 3 [150/187 (80%)]\t training loss: 0.372306\n",
      "epoch: 3 [160/187 (86%)]\t training loss: 1.047415\n",
      "epoch: 3 [170/187 (91%)]\t training loss: 0.872863\n",
      "epoch: 3 [180/187 (96%)]\t training loss: 0.723147\n",
      "\n",
      "\n",
      "Validation...\n",
      "\n",
      "Test dataset: Overall Loss: 0.3342, Overall Accuracy: 2684/2977 (90%)\n",
      "\n",
      "epoch: 4 [0/187 (0%)]\t training loss: 1.049980\n",
      "epoch: 4 [10/187 (5%)]\t training loss: 0.768217\n",
      "epoch: 4 [20/187 (11%)]\t training loss: 0.375047\n",
      "epoch: 4 [30/187 (16%)]\t training loss: 0.431142\n",
      "epoch: 4 [40/187 (21%)]\t training loss: 0.481306\n",
      "epoch: 4 [50/187 (27%)]\t training loss: 0.891914\n",
      "epoch: 4 [60/187 (32%)]\t training loss: 0.760488\n",
      "epoch: 4 [70/187 (37%)]\t training loss: 0.584327\n",
      "epoch: 4 [80/187 (43%)]\t training loss: 0.712303\n",
      "epoch: 4 [90/187 (48%)]\t training loss: 0.431153\n",
      "epoch: 4 [100/187 (53%)]\t training loss: 0.539311\n",
      "epoch: 4 [110/187 (59%)]\t training loss: 0.973701\n",
      "epoch: 4 [120/187 (64%)]\t training loss: 0.808702\n",
      "epoch: 4 [130/187 (70%)]\t training loss: 0.731854\n",
      "epoch: 4 [140/187 (75%)]\t training loss: 0.906321\n",
      "epoch: 4 [150/187 (80%)]\t training loss: 0.497615\n",
      "epoch: 4 [160/187 (86%)]\t training loss: 0.775873\n",
      "epoch: 4 [170/187 (91%)]\t training loss: 0.615538\n",
      "epoch: 4 [180/187 (96%)]\t training loss: 0.862167\n",
      "\n",
      "\n",
      "Validation...\n",
      "\n",
      "Test dataset: Overall Loss: 0.3084, Overall Accuracy: 2711/2977 (91%)\n",
      "\n",
      "epoch: 5 [0/187 (0%)]\t training loss: 0.652434\n",
      "epoch: 5 [10/187 (5%)]\t training loss: 0.708086\n",
      "epoch: 5 [20/187 (11%)]\t training loss: 0.612785\n",
      "epoch: 5 [30/187 (16%)]\t training loss: 0.966804\n",
      "epoch: 5 [40/187 (21%)]\t training loss: 0.210525\n",
      "epoch: 5 [50/187 (27%)]\t training loss: 0.337951\n",
      "epoch: 5 [60/187 (32%)]\t training loss: 0.382651\n",
      "epoch: 5 [70/187 (37%)]\t training loss: 0.729506\n",
      "epoch: 5 [80/187 (43%)]\t training loss: 0.666102\n",
      "epoch: 5 [90/187 (48%)]\t training loss: 0.887975\n",
      "epoch: 5 [100/187 (53%)]\t training loss: 0.433449\n",
      "epoch: 5 [110/187 (59%)]\t training loss: 0.221869\n",
      "epoch: 5 [120/187 (64%)]\t training loss: 0.152874\n",
      "epoch: 5 [130/187 (70%)]\t training loss: 0.334872\n",
      "epoch: 5 [140/187 (75%)]\t training loss: 0.753138\n",
      "epoch: 5 [150/187 (80%)]\t training loss: 0.699741\n",
      "epoch: 5 [160/187 (86%)]\t training loss: 0.756104\n",
      "epoch: 5 [170/187 (91%)]\t training loss: 0.148847\n",
      "epoch: 5 [180/187 (96%)]\t training loss: 0.349668\n",
      "\n",
      "\n",
      "Validation...\n",
      "\n",
      "Test dataset: Overall Loss: 0.2951, Overall Accuracy: 2717/2977 (91%)\n",
      "\n",
      "epoch: 6 [0/187 (0%)]\t training loss: 0.405427\n",
      "epoch: 6 [10/187 (5%)]\t training loss: 0.758618\n",
      "epoch: 6 [20/187 (11%)]\t training loss: 0.541675\n",
      "epoch: 6 [30/187 (16%)]\t training loss: 0.595891\n",
      "epoch: 6 [40/187 (21%)]\t training loss: 0.298865\n",
      "epoch: 6 [50/187 (27%)]\t training loss: 0.677856\n",
      "epoch: 6 [60/187 (32%)]\t training loss: 0.903655\n",
      "epoch: 6 [70/187 (37%)]\t training loss: 0.730755\n",
      "epoch: 6 [80/187 (43%)]\t training loss: 0.536443\n",
      "epoch: 6 [90/187 (48%)]\t training loss: 0.392334\n",
      "epoch: 6 [100/187 (53%)]\t training loss: 0.752318\n",
      "epoch: 6 [110/187 (59%)]\t training loss: 0.686638\n",
      "epoch: 6 [120/187 (64%)]\t training loss: 0.739796\n",
      "epoch: 6 [130/187 (70%)]\t training loss: 0.702512\n",
      "epoch: 6 [140/187 (75%)]\t training loss: 0.594906\n",
      "epoch: 6 [150/187 (80%)]\t training loss: 0.455483\n",
      "epoch: 6 [160/187 (86%)]\t training loss: 0.655694\n",
      "epoch: 6 [170/187 (91%)]\t training loss: 0.572483\n",
      "epoch: 6 [180/187 (96%)]\t training loss: 0.995917\n",
      "\n",
      "\n",
      "Validation...\n",
      "\n",
      "Test dataset: Overall Loss: 0.3305, Overall Accuracy: 2675/2977 (90%)\n",
      "\n",
      "epoch: 7 [0/187 (0%)]\t training loss: 0.742653\n",
      "epoch: 7 [10/187 (5%)]\t training loss: 0.374927\n",
      "epoch: 7 [20/187 (11%)]\t training loss: 0.798805\n",
      "epoch: 7 [30/187 (16%)]\t training loss: 0.239645\n",
      "epoch: 7 [40/187 (21%)]\t training loss: 0.536308\n",
      "epoch: 7 [50/187 (27%)]\t training loss: 0.918482\n",
      "epoch: 7 [60/187 (32%)]\t training loss: 0.199815\n",
      "epoch: 7 [70/187 (37%)]\t training loss: 0.863243\n",
      "epoch: 7 [80/187 (43%)]\t training loss: 0.536044\n",
      "epoch: 7 [90/187 (48%)]\t training loss: 0.672954\n",
      "epoch: 7 [100/187 (53%)]\t training loss: 0.718568\n",
      "epoch: 7 [110/187 (59%)]\t training loss: 0.225796\n",
      "epoch: 7 [120/187 (64%)]\t training loss: 0.700686\n",
      "epoch: 7 [130/187 (70%)]\t training loss: 0.686996\n",
      "epoch: 7 [140/187 (75%)]\t training loss: 0.400786\n",
      "epoch: 7 [150/187 (80%)]\t training loss: 0.943460\n",
      "epoch: 7 [160/187 (86%)]\t training loss: 0.714440\n",
      "epoch: 7 [170/187 (91%)]\t training loss: 0.716121\n",
      "epoch: 7 [180/187 (96%)]\t training loss: 0.277666\n",
      "\n",
      "\n",
      "Validation...\n",
      "\n",
      "Test dataset: Overall Loss: 0.3786, Overall Accuracy: 2608/2977 (88%)\n",
      "\n",
      "epoch: 8 [0/187 (0%)]\t training loss: 0.950069\n",
      "epoch: 8 [10/187 (5%)]\t training loss: 0.570486\n",
      "epoch: 8 [20/187 (11%)]\t training loss: 0.881348\n",
      "epoch: 8 [30/187 (16%)]\t training loss: 0.610182\n",
      "epoch: 8 [40/187 (21%)]\t training loss: 1.164511\n",
      "epoch: 8 [50/187 (27%)]\t training loss: 0.296743\n",
      "epoch: 8 [60/187 (32%)]\t training loss: 0.767943\n",
      "epoch: 8 [70/187 (37%)]\t training loss: 0.865049\n",
      "epoch: 8 [80/187 (43%)]\t training loss: 0.744960\n",
      "epoch: 8 [90/187 (48%)]\t training loss: 0.897353\n",
      "epoch: 8 [100/187 (53%)]\t training loss: 0.801130\n",
      "epoch: 8 [110/187 (59%)]\t training loss: 0.301881\n",
      "epoch: 8 [120/187 (64%)]\t training loss: 1.161479\n",
      "epoch: 8 [130/187 (70%)]\t training loss: 0.840487\n",
      "epoch: 8 [140/187 (75%)]\t training loss: 0.559595\n",
      "epoch: 8 [150/187 (80%)]\t training loss: 0.989432\n",
      "epoch: 8 [160/187 (86%)]\t training loss: 0.389484\n",
      "epoch: 8 [170/187 (91%)]\t training loss: 0.762779\n",
      "epoch: 8 [180/187 (96%)]\t training loss: 0.931685\n",
      "\n",
      "\n",
      "Validation...\n",
      "\n",
      "Test dataset: Overall Loss: 0.3237, Overall Accuracy: 2675/2977 (90%)\n",
      "\n",
      "epoch: 9 [0/187 (0%)]\t training loss: 0.630876\n",
      "epoch: 9 [10/187 (5%)]\t training loss: 0.904676\n",
      "epoch: 9 [20/187 (11%)]\t training loss: 0.897287\n",
      "epoch: 9 [30/187 (16%)]\t training loss: 0.290914\n",
      "epoch: 9 [40/187 (21%)]\t training loss: 0.920925\n",
      "epoch: 9 [50/187 (27%)]\t training loss: 0.514693\n",
      "epoch: 9 [60/187 (32%)]\t training loss: 0.152021\n",
      "epoch: 9 [70/187 (37%)]\t training loss: 0.790008\n",
      "epoch: 9 [80/187 (43%)]\t training loss: 0.785923\n",
      "epoch: 9 [90/187 (48%)]\t training loss: 0.452803\n",
      "epoch: 9 [100/187 (53%)]\t training loss: 0.564889\n",
      "epoch: 9 [110/187 (59%)]\t training loss: 0.699773\n",
      "epoch: 9 [120/187 (64%)]\t training loss: 0.289149\n",
      "epoch: 9 [130/187 (70%)]\t training loss: 0.390661\n",
      "epoch: 9 [140/187 (75%)]\t training loss: 0.823494\n",
      "epoch: 9 [150/187 (80%)]\t training loss: 0.424384\n",
      "epoch: 9 [160/187 (86%)]\t training loss: 0.800867\n",
      "epoch: 9 [170/187 (91%)]\t training loss: 0.347370\n",
      "epoch: 9 [180/187 (96%)]\t training loss: 0.829842\n",
      "\n",
      "\n",
      "Validation...\n",
      "\n",
      "Test dataset: Overall Loss: 0.3823, Overall Accuracy: 2659/2977 (89%)\n",
      "\n",
      "epoch: 10 [0/187 (0%)]\t training loss: 0.301250\n",
      "epoch: 10 [10/187 (5%)]\t training loss: 0.977479\n",
      "epoch: 10 [20/187 (11%)]\t training loss: 0.932404\n",
      "epoch: 10 [30/187 (16%)]\t training loss: 0.532210\n",
      "epoch: 10 [40/187 (21%)]\t training loss: 0.579123\n",
      "epoch: 10 [50/187 (27%)]\t training loss: 0.387031\n",
      "epoch: 10 [60/187 (32%)]\t training loss: 0.883808\n",
      "epoch: 10 [70/187 (37%)]\t training loss: 0.814732\n",
      "epoch: 10 [80/187 (43%)]\t training loss: 0.841851\n",
      "epoch: 10 [90/187 (48%)]\t training loss: 0.620802\n",
      "epoch: 10 [100/187 (53%)]\t training loss: 0.717624\n",
      "epoch: 10 [110/187 (59%)]\t training loss: 0.469079\n",
      "epoch: 10 [120/187 (64%)]\t training loss: 0.755135\n",
      "epoch: 10 [130/187 (70%)]\t training loss: 0.831576\n",
      "epoch: 10 [140/187 (75%)]\t training loss: 0.266611\n",
      "epoch: 10 [150/187 (80%)]\t training loss: 0.534106\n",
      "epoch: 10 [160/187 (86%)]\t training loss: 0.762972\n",
      "epoch: 10 [170/187 (91%)]\t training loss: 0.775501\n",
      "epoch: 10 [180/187 (96%)]\t training loss: 0.753429\n",
      "\n",
      "\n",
      "Validation...\n",
      "\n",
      "Test dataset: Overall Loss: 0.3343, Overall Accuracy: 2693/2977 (90%)\n",
      "\n",
      "epoch: 11 [0/187 (0%)]\t training loss: 0.747511\n",
      "epoch: 11 [10/187 (5%)]\t training loss: 0.816917\n",
      "epoch: 11 [20/187 (11%)]\t training loss: 0.847445\n",
      "epoch: 11 [30/187 (16%)]\t training loss: 0.219958\n",
      "epoch: 11 [40/187 (21%)]\t training loss: 0.159636\n",
      "epoch: 11 [50/187 (27%)]\t training loss: 0.646677\n",
      "epoch: 11 [60/187 (32%)]\t training loss: 0.624828\n",
      "epoch: 11 [70/187 (37%)]\t training loss: 0.230871\n",
      "epoch: 11 [80/187 (43%)]\t training loss: 0.726746\n",
      "epoch: 11 [90/187 (48%)]\t training loss: 1.069858\n",
      "epoch: 11 [100/187 (53%)]\t training loss: 0.758920\n",
      "epoch: 11 [110/187 (59%)]\t training loss: 0.311185\n",
      "epoch: 11 [120/187 (64%)]\t training loss: 0.629244\n",
      "epoch: 11 [130/187 (70%)]\t training loss: 0.369884\n",
      "epoch: 11 [140/187 (75%)]\t training loss: 1.088564\n",
      "epoch: 11 [150/187 (80%)]\t training loss: 0.459535\n",
      "epoch: 11 [160/187 (86%)]\t training loss: 0.298040\n",
      "epoch: 11 [170/187 (91%)]\t training loss: 0.358129\n",
      "epoch: 11 [180/187 (96%)]\t training loss: 0.231965\n",
      "\n",
      "\n",
      "Validation...\n",
      "\n",
      "Test dataset: Overall Loss: 0.2838, Overall Accuracy: 2728/2977 (92%)\n",
      "\n",
      "epoch: 12 [0/187 (0%)]\t training loss: 0.353181\n",
      "epoch: 12 [10/187 (5%)]\t training loss: 0.152182\n",
      "epoch: 12 [20/187 (11%)]\t training loss: 0.828124\n",
      "epoch: 12 [30/187 (16%)]\t training loss: 0.471784\n",
      "epoch: 12 [40/187 (21%)]\t training loss: 0.142066\n",
      "epoch: 12 [50/187 (27%)]\t training loss: 0.902828\n",
      "epoch: 12 [60/187 (32%)]\t training loss: 0.606947\n",
      "epoch: 12 [70/187 (37%)]\t training loss: 0.597989\n",
      "epoch: 12 [80/187 (43%)]\t training loss: 0.206967\n",
      "epoch: 12 [90/187 (48%)]\t training loss: 0.849957\n",
      "epoch: 12 [100/187 (53%)]\t training loss: 0.365011\n",
      "epoch: 12 [110/187 (59%)]\t training loss: 0.392250\n",
      "epoch: 12 [120/187 (64%)]\t training loss: 0.539267\n",
      "epoch: 12 [130/187 (70%)]\t training loss: 0.765476\n",
      "epoch: 12 [140/187 (75%)]\t training loss: 0.870953\n",
      "epoch: 12 [150/187 (80%)]\t training loss: 0.944059\n",
      "epoch: 12 [160/187 (86%)]\t training loss: 1.022970\n",
      "epoch: 12 [170/187 (91%)]\t training loss: 0.141395\n",
      "epoch: 12 [180/187 (96%)]\t training loss: 0.552007\n",
      "\n",
      "\n",
      "Validation...\n",
      "\n",
      "Test dataset: Overall Loss: 0.2967, Overall Accuracy: 2724/2977 (92%)\n",
      "\n",
      "epoch: 13 [0/187 (0%)]\t training loss: 0.459281\n",
      "epoch: 13 [10/187 (5%)]\t training loss: 0.810364\n",
      "epoch: 13 [20/187 (11%)]\t training loss: 0.536134\n",
      "epoch: 13 [30/187 (16%)]\t training loss: 0.755166\n",
      "epoch: 13 [40/187 (21%)]\t training loss: 0.836376\n",
      "epoch: 13 [50/187 (27%)]\t training loss: 0.808917\n",
      "epoch: 13 [60/187 (32%)]\t training loss: 0.203372\n",
      "epoch: 13 [70/187 (37%)]\t training loss: 0.134231\n",
      "epoch: 13 [80/187 (43%)]\t training loss: 0.822042\n",
      "epoch: 13 [90/187 (48%)]\t training loss: 0.160470\n",
      "epoch: 13 [100/187 (53%)]\t training loss: 0.748647\n",
      "epoch: 13 [110/187 (59%)]\t training loss: 0.737474\n",
      "epoch: 13 [120/187 (64%)]\t training loss: 0.654761\n",
      "epoch: 13 [130/187 (70%)]\t training loss: 0.480278\n",
      "epoch: 13 [140/187 (75%)]\t training loss: 0.750407\n",
      "epoch: 13 [150/187 (80%)]\t training loss: 0.200417\n",
      "epoch: 13 [160/187 (86%)]\t training loss: 0.169809\n",
      "epoch: 13 [170/187 (91%)]\t training loss: 0.878702\n",
      "epoch: 13 [180/187 (96%)]\t training loss: 0.782649\n",
      "\n",
      "\n",
      "Validation...\n",
      "\n",
      "Test dataset: Overall Loss: 0.3119, Overall Accuracy: 2702/2977 (91%)\n",
      "\n",
      "epoch: 14 [0/187 (0%)]\t training loss: 0.155428\n",
      "epoch: 14 [10/187 (5%)]\t training loss: 0.101031\n",
      "epoch: 14 [20/187 (11%)]\t training loss: 0.840567\n",
      "epoch: 14 [30/187 (16%)]\t training loss: 0.557385\n",
      "epoch: 14 [40/187 (21%)]\t training loss: 0.099354\n",
      "epoch: 14 [50/187 (27%)]\t training loss: 0.732636\n",
      "epoch: 14 [60/187 (32%)]\t training loss: 0.434247\n",
      "epoch: 14 [70/187 (37%)]\t training loss: 0.617818\n",
      "epoch: 14 [80/187 (43%)]\t training loss: 0.833959\n",
      "epoch: 14 [90/187 (48%)]\t training loss: 0.679024\n",
      "epoch: 14 [100/187 (53%)]\t training loss: 0.130403\n",
      "epoch: 14 [110/187 (59%)]\t training loss: 0.713936\n",
      "epoch: 14 [120/187 (64%)]\t training loss: 0.130973\n",
      "epoch: 14 [130/187 (70%)]\t training loss: 0.276954\n",
      "epoch: 14 [140/187 (75%)]\t training loss: 0.432928\n",
      "epoch: 14 [150/187 (80%)]\t training loss: 0.819601\n",
      "epoch: 14 [160/187 (86%)]\t training loss: 0.204304\n",
      "epoch: 14 [170/187 (91%)]\t training loss: 1.006746\n",
      "epoch: 14 [180/187 (96%)]\t training loss: 0.609889\n",
      "\n",
      "\n",
      "Validation...\n",
      "\n",
      "Test dataset: Overall Loss: 0.2912, Overall Accuracy: 2747/2977 (92%)\n",
      "\n",
      "epoch: 15 [0/187 (0%)]\t training loss: 0.156910\n",
      "epoch: 15 [10/187 (5%)]\t training loss: 0.811469\n",
      "epoch: 15 [20/187 (11%)]\t training loss: 0.216685\n",
      "epoch: 15 [30/187 (16%)]\t training loss: 0.732904\n",
      "epoch: 15 [40/187 (21%)]\t training loss: 0.822720\n",
      "epoch: 15 [50/187 (27%)]\t training loss: 0.222194\n",
      "epoch: 15 [60/187 (32%)]\t training loss: 0.664001\n",
      "epoch: 15 [70/187 (37%)]\t training loss: 0.866565\n",
      "epoch: 15 [80/187 (43%)]\t training loss: 0.334195\n",
      "epoch: 15 [90/187 (48%)]\t training loss: 0.624492\n",
      "epoch: 15 [100/187 (53%)]\t training loss: 0.552826\n",
      "epoch: 15 [110/187 (59%)]\t training loss: 0.647620\n",
      "epoch: 15 [120/187 (64%)]\t training loss: 0.616306\n",
      "epoch: 15 [130/187 (70%)]\t training loss: 0.754786\n",
      "epoch: 15 [140/187 (75%)]\t training loss: 0.677758\n",
      "epoch: 15 [150/187 (80%)]\t training loss: 0.647945\n",
      "epoch: 15 [160/187 (86%)]\t training loss: 0.334003\n",
      "epoch: 15 [170/187 (91%)]\t training loss: 0.649077\n",
      "epoch: 15 [180/187 (96%)]\t training loss: 0.255080\n",
      "\n",
      "\n",
      "Validation...\n",
      "\n",
      "Test dataset: Overall Loss: 0.3014, Overall Accuracy: 2735/2977 (92%)\n",
      "\n",
      "epoch: 16 [0/187 (0%)]\t training loss: 0.683652\n",
      "epoch: 16 [10/187 (5%)]\t training loss: 0.806957\n",
      "epoch: 16 [20/187 (11%)]\t training loss: 0.083937\n",
      "epoch: 16 [30/187 (16%)]\t training loss: 0.098540\n",
      "epoch: 16 [40/187 (21%)]\t training loss: 0.434938\n",
      "epoch: 16 [50/187 (27%)]\t training loss: 0.108104\n",
      "epoch: 16 [60/187 (32%)]\t training loss: 0.395113\n",
      "epoch: 16 [70/187 (37%)]\t training loss: 0.743815\n",
      "epoch: 16 [80/187 (43%)]\t training loss: 0.565634\n",
      "epoch: 16 [90/187 (48%)]\t training loss: 0.223914\n",
      "epoch: 16 [100/187 (53%)]\t training loss: 0.315559\n",
      "epoch: 16 [110/187 (59%)]\t training loss: 0.520320\n",
      "epoch: 16 [120/187 (64%)]\t training loss: 0.067130\n",
      "epoch: 16 [130/187 (70%)]\t training loss: 0.711161\n",
      "epoch: 16 [140/187 (75%)]\t training loss: 0.771468\n",
      "epoch: 16 [150/187 (80%)]\t training loss: 0.869931\n",
      "epoch: 16 [160/187 (86%)]\t training loss: 0.258881\n",
      "epoch: 16 [170/187 (91%)]\t training loss: 0.058600\n",
      "epoch: 16 [180/187 (96%)]\t training loss: 0.141435\n",
      "\n",
      "\n",
      "Validation...\n",
      "\n",
      "Test dataset: Overall Loss: 0.3200, Overall Accuracy: 2720/2977 (91%)\n",
      "\n",
      "epoch: 17 [0/187 (0%)]\t training loss: 0.188508\n",
      "epoch: 17 [10/187 (5%)]\t training loss: 0.139606\n",
      "epoch: 17 [20/187 (11%)]\t training loss: 0.751052\n",
      "epoch: 17 [30/187 (16%)]\t training loss: 0.707094\n",
      "epoch: 17 [40/187 (21%)]\t training loss: 0.799303\n",
      "epoch: 17 [50/187 (27%)]\t training loss: 0.105766\n",
      "epoch: 17 [60/187 (32%)]\t training loss: 0.737248\n",
      "epoch: 17 [70/187 (37%)]\t training loss: 0.877332\n",
      "epoch: 17 [80/187 (43%)]\t training loss: 0.185140\n",
      "epoch: 17 [90/187 (48%)]\t training loss: 0.085690\n",
      "epoch: 17 [100/187 (53%)]\t training loss: 0.931354\n",
      "epoch: 17 [110/187 (59%)]\t training loss: 0.805529\n",
      "epoch: 17 [120/187 (64%)]\t training loss: 0.263724\n",
      "epoch: 17 [130/187 (70%)]\t training loss: 0.098433\n",
      "epoch: 17 [140/187 (75%)]\t training loss: 0.704715\n",
      "epoch: 17 [150/187 (80%)]\t training loss: 0.745222\n",
      "epoch: 17 [160/187 (86%)]\t training loss: 0.212491\n",
      "epoch: 17 [170/187 (91%)]\t training loss: 0.152997\n",
      "epoch: 17 [180/187 (96%)]\t training loss: 0.362810\n",
      "\n",
      "\n",
      "Validation...\n",
      "\n",
      "Test dataset: Overall Loss: 0.2980, Overall Accuracy: 2739/2977 (92%)\n",
      "\n",
      "epoch: 18 [0/187 (0%)]\t training loss: 0.965500\n",
      "epoch: 18 [10/187 (5%)]\t training loss: 0.712386\n",
      "epoch: 18 [20/187 (11%)]\t training loss: 0.290780\n",
      "epoch: 18 [30/187 (16%)]\t training loss: 0.232702\n",
      "epoch: 18 [40/187 (21%)]\t training loss: 1.018712\n",
      "epoch: 18 [50/187 (27%)]\t training loss: 0.386376\n",
      "epoch: 18 [60/187 (32%)]\t training loss: 0.339324\n",
      "epoch: 18 [70/187 (37%)]\t training loss: 0.810691\n",
      "epoch: 18 [80/187 (43%)]\t training loss: 0.813185\n",
      "epoch: 18 [90/187 (48%)]\t training loss: 0.739462\n",
      "epoch: 18 [100/187 (53%)]\t training loss: 0.805145\n",
      "epoch: 18 [110/187 (59%)]\t training loss: 0.783818\n",
      "epoch: 18 [120/187 (64%)]\t training loss: 0.752301\n",
      "epoch: 18 [130/187 (70%)]\t training loss: 0.576480\n",
      "epoch: 18 [140/187 (75%)]\t training loss: 0.716161\n",
      "epoch: 18 [150/187 (80%)]\t training loss: 0.301788\n",
      "epoch: 18 [160/187 (86%)]\t training loss: 0.468665\n",
      "epoch: 18 [170/187 (91%)]\t training loss: 0.160457\n",
      "epoch: 18 [180/187 (96%)]\t training loss: 0.831479\n",
      "\n",
      "\n",
      "Validation...\n",
      "\n",
      "Test dataset: Overall Loss: 0.3235, Overall Accuracy: 2729/2977 (92%)\n",
      "\n",
      "epoch: 19 [0/187 (0%)]\t training loss: 0.697003\n",
      "epoch: 19 [10/187 (5%)]\t training loss: 0.804226\n",
      "epoch: 19 [20/187 (11%)]\t training loss: 0.326839\n",
      "epoch: 19 [30/187 (16%)]\t training loss: 0.971942\n",
      "epoch: 19 [40/187 (21%)]\t training loss: 0.526462\n",
      "epoch: 19 [50/187 (27%)]\t training loss: 0.783204\n",
      "epoch: 19 [60/187 (32%)]\t training loss: 0.887193\n",
      "epoch: 19 [70/187 (37%)]\t training loss: 0.380610\n",
      "epoch: 19 [80/187 (43%)]\t training loss: 0.885298\n",
      "epoch: 19 [90/187 (48%)]\t training loss: 0.476719\n",
      "epoch: 19 [100/187 (53%)]\t training loss: 0.232297\n",
      "epoch: 19 [110/187 (59%)]\t training loss: 0.781229\n",
      "epoch: 19 [120/187 (64%)]\t training loss: 0.993321\n",
      "epoch: 19 [130/187 (70%)]\t training loss: 0.710126\n",
      "epoch: 19 [140/187 (75%)]\t training loss: 0.254060\n",
      "epoch: 19 [150/187 (80%)]\t training loss: 0.901146\n",
      "epoch: 19 [160/187 (86%)]\t training loss: 0.113976\n",
      "epoch: 19 [170/187 (91%)]\t training loss: 0.329078\n",
      "epoch: 19 [180/187 (96%)]\t training loss: 0.704195\n",
      "\n",
      "\n",
      "Validation...\n",
      "\n",
      "Test dataset: Overall Loss: 0.2992, Overall Accuracy: 2737/2977 (92%)\n",
      "\n",
      "epoch: 20 [0/187 (0%)]\t training loss: 0.760720\n",
      "epoch: 20 [10/187 (5%)]\t training loss: 0.177031\n",
      "epoch: 20 [20/187 (11%)]\t training loss: 0.411817\n",
      "epoch: 20 [30/187 (16%)]\t training loss: 0.211811\n",
      "epoch: 20 [40/187 (21%)]\t training loss: 0.181117\n",
      "epoch: 20 [50/187 (27%)]\t training loss: 0.713045\n",
      "epoch: 20 [60/187 (32%)]\t training loss: 0.647703\n",
      "epoch: 20 [70/187 (37%)]\t training loss: 0.098542\n",
      "epoch: 20 [80/187 (43%)]\t training loss: 0.102813\n",
      "epoch: 20 [90/187 (48%)]\t training loss: 0.666508\n",
      "epoch: 20 [100/187 (53%)]\t training loss: 0.104936\n",
      "epoch: 20 [110/187 (59%)]\t training loss: 0.687436\n",
      "epoch: 20 [120/187 (64%)]\t training loss: 0.816012\n",
      "epoch: 20 [130/187 (70%)]\t training loss: 0.859539\n",
      "epoch: 20 [140/187 (75%)]\t training loss: 0.291221\n",
      "epoch: 20 [150/187 (80%)]\t training loss: 0.289039\n",
      "epoch: 20 [160/187 (86%)]\t training loss: 0.497967\n",
      "epoch: 20 [170/187 (91%)]\t training loss: 0.408093\n",
      "epoch: 20 [180/187 (96%)]\t training loss: 0.071430\n",
      "\n",
      "\n",
      "Validation...\n",
      "\n",
      "Test dataset: Overall Loss: 0.2843, Overall Accuracy: 2758/2977 (93%)\n",
      "\n",
      "epoch: 21 [0/187 (0%)]\t training loss: 0.372950\n",
      "epoch: 21 [10/187 (5%)]\t training loss: 0.950617\n",
      "epoch: 21 [20/187 (11%)]\t training loss: 0.563436\n",
      "epoch: 21 [30/187 (16%)]\t training loss: 0.627522\n",
      "epoch: 21 [40/187 (21%)]\t training loss: 0.734151\n",
      "epoch: 21 [50/187 (27%)]\t training loss: 0.196769\n",
      "epoch: 21 [60/187 (32%)]\t training loss: 0.168662\n",
      "epoch: 21 [70/187 (37%)]\t training loss: 0.830541\n",
      "epoch: 21 [80/187 (43%)]\t training loss: 0.171979\n",
      "epoch: 21 [90/187 (48%)]\t training loss: 0.068412\n",
      "epoch: 21 [100/187 (53%)]\t training loss: 0.384395\n",
      "epoch: 21 [110/187 (59%)]\t training loss: 0.737452\n",
      "epoch: 21 [120/187 (64%)]\t training loss: 0.343735\n",
      "epoch: 21 [130/187 (70%)]\t training loss: 0.591444\n",
      "epoch: 21 [140/187 (75%)]\t training loss: 0.335048\n",
      "epoch: 21 [150/187 (80%)]\t training loss: 0.519919\n",
      "epoch: 21 [160/187 (86%)]\t training loss: 0.717478\n",
      "epoch: 21 [170/187 (91%)]\t training loss: 0.560481\n",
      "epoch: 21 [180/187 (96%)]\t training loss: 0.796056\n",
      "\n",
      "\n",
      "Validation...\n",
      "\n",
      "Test dataset: Overall Loss: 0.2998, Overall Accuracy: 2754/2977 (93%)\n",
      "\n",
      "epoch: 22 [0/187 (0%)]\t training loss: 0.764044\n",
      "epoch: 22 [10/187 (5%)]\t training loss: 0.091815\n",
      "epoch: 22 [20/187 (11%)]\t training loss: 0.582737\n",
      "epoch: 22 [30/187 (16%)]\t training loss: 0.464391\n",
      "epoch: 22 [40/187 (21%)]\t training loss: 0.365205\n",
      "epoch: 22 [50/187 (27%)]\t training loss: 0.702378\n",
      "epoch: 22 [60/187 (32%)]\t training loss: 0.493457\n",
      "epoch: 22 [70/187 (37%)]\t training loss: 0.201533\n",
      "epoch: 22 [80/187 (43%)]\t training loss: 0.790250\n",
      "epoch: 22 [90/187 (48%)]\t training loss: 0.872627\n",
      "epoch: 22 [100/187 (53%)]\t training loss: 0.322252\n",
      "epoch: 22 [110/187 (59%)]\t training loss: 0.795274\n",
      "epoch: 22 [120/187 (64%)]\t training loss: 0.486186\n",
      "epoch: 22 [130/187 (70%)]\t training loss: 0.308695\n",
      "epoch: 22 [140/187 (75%)]\t training loss: 0.602694\n",
      "epoch: 22 [150/187 (80%)]\t training loss: 0.944688\n",
      "epoch: 22 [160/187 (86%)]\t training loss: 0.284101\n",
      "epoch: 22 [170/187 (91%)]\t training loss: 0.757197\n",
      "epoch: 22 [180/187 (96%)]\t training loss: 0.606045\n",
      "\n",
      "\n",
      "Validation...\n",
      "\n",
      "Test dataset: Overall Loss: 0.3004, Overall Accuracy: 2735/2977 (92%)\n",
      "\n",
      "epoch: 23 [0/187 (0%)]\t training loss: 0.251288\n",
      "epoch: 23 [10/187 (5%)]\t training loss: 0.440487\n",
      "epoch: 23 [20/187 (11%)]\t training loss: 0.433742\n",
      "epoch: 23 [30/187 (16%)]\t training loss: 0.576245\n",
      "epoch: 23 [40/187 (21%)]\t training loss: 0.212879\n",
      "epoch: 23 [50/187 (27%)]\t training loss: 0.669014\n",
      "epoch: 23 [60/187 (32%)]\t training loss: 0.875168\n",
      "epoch: 23 [70/187 (37%)]\t training loss: 0.665457\n",
      "epoch: 23 [80/187 (43%)]\t training loss: 0.522476\n",
      "epoch: 23 [90/187 (48%)]\t training loss: 0.642231\n",
      "epoch: 23 [100/187 (53%)]\t training loss: 0.790991\n",
      "epoch: 23 [110/187 (59%)]\t training loss: 0.644655\n",
      "epoch: 23 [120/187 (64%)]\t training loss: 0.060645\n",
      "epoch: 23 [130/187 (70%)]\t training loss: 0.786727\n",
      "epoch: 23 [140/187 (75%)]\t training loss: 0.637088\n",
      "epoch: 23 [150/187 (80%)]\t training loss: 0.097533\n",
      "epoch: 23 [160/187 (86%)]\t training loss: 0.123861\n",
      "epoch: 23 [170/187 (91%)]\t training loss: 0.279898\n",
      "epoch: 23 [180/187 (96%)]\t training loss: 0.359060\n",
      "\n",
      "\n",
      "Validation...\n",
      "\n",
      "Test dataset: Overall Loss: 0.2862, Overall Accuracy: 2751/2977 (92%)\n",
      "\n",
      "epoch: 24 [0/187 (0%)]\t training loss: 0.188287\n",
      "epoch: 24 [10/187 (5%)]\t training loss: 0.838893\n",
      "epoch: 24 [20/187 (11%)]\t training loss: 0.903164\n",
      "epoch: 24 [30/187 (16%)]\t training loss: 0.550710\n",
      "epoch: 24 [40/187 (21%)]\t training loss: 0.295616\n",
      "epoch: 24 [50/187 (27%)]\t training loss: 0.649714\n",
      "epoch: 24 [60/187 (32%)]\t training loss: 0.303653\n",
      "epoch: 24 [70/187 (37%)]\t training loss: 0.581861\n",
      "epoch: 24 [80/187 (43%)]\t training loss: 0.278601\n",
      "epoch: 24 [90/187 (48%)]\t training loss: 0.105776\n",
      "epoch: 24 [100/187 (53%)]\t training loss: 0.264378\n",
      "epoch: 24 [110/187 (59%)]\t training loss: 0.769477\n",
      "epoch: 24 [120/187 (64%)]\t training loss: 0.709629\n",
      "epoch: 24 [130/187 (70%)]\t training loss: 0.799386\n",
      "epoch: 24 [140/187 (75%)]\t training loss: 0.617975\n",
      "epoch: 24 [150/187 (80%)]\t training loss: 0.778308\n",
      "epoch: 24 [160/187 (86%)]\t training loss: 0.771974\n",
      "epoch: 24 [170/187 (91%)]\t training loss: 0.631794\n",
      "epoch: 24 [180/187 (96%)]\t training loss: 0.814452\n",
      "\n",
      "\n",
      "Validation...\n",
      "\n",
      "Test dataset: Overall Loss: 0.3271, Overall Accuracy: 2723/2977 (91%)\n",
      "\n",
      "epoch: 25 [0/187 (0%)]\t training loss: 0.174214\n",
      "epoch: 25 [10/187 (5%)]\t training loss: 0.664655\n",
      "epoch: 25 [20/187 (11%)]\t training loss: 0.635251\n",
      "epoch: 25 [30/187 (16%)]\t training loss: 0.755204\n",
      "epoch: 25 [40/187 (21%)]\t training loss: 0.610847\n",
      "epoch: 25 [50/187 (27%)]\t training loss: 0.654779\n",
      "epoch: 25 [60/187 (32%)]\t training loss: 0.209827\n",
      "epoch: 25 [70/187 (37%)]\t training loss: 0.446595\n",
      "epoch: 25 [80/187 (43%)]\t training loss: 0.291374\n",
      "epoch: 25 [90/187 (48%)]\t training loss: 0.064334\n",
      "epoch: 25 [100/187 (53%)]\t training loss: 0.289523\n",
      "epoch: 25 [110/187 (59%)]\t training loss: 0.609708\n",
      "epoch: 25 [120/187 (64%)]\t training loss: 0.682000\n",
      "epoch: 25 [130/187 (70%)]\t training loss: 0.152535\n",
      "epoch: 25 [140/187 (75%)]\t training loss: 0.628734\n",
      "epoch: 25 [150/187 (80%)]\t training loss: 0.382643\n",
      "epoch: 25 [160/187 (86%)]\t training loss: 0.043708\n",
      "epoch: 25 [170/187 (91%)]\t training loss: 0.488693\n",
      "epoch: 25 [180/187 (96%)]\t training loss: 0.711433\n",
      "\n",
      "\n",
      "Validation...\n",
      "\n",
      "Test dataset: Overall Loss: 0.2918, Overall Accuracy: 2730/2977 (92%)\n",
      "\n",
      "epoch: 26 [0/187 (0%)]\t training loss: 0.202783\n",
      "epoch: 26 [10/187 (5%)]\t training loss: 0.848956\n",
      "epoch: 26 [20/187 (11%)]\t training loss: 0.722360\n",
      "epoch: 26 [30/187 (16%)]\t training loss: 0.608258\n",
      "epoch: 26 [40/187 (21%)]\t training loss: 0.549149\n",
      "epoch: 26 [50/187 (27%)]\t training loss: 0.715142\n",
      "epoch: 26 [60/187 (32%)]\t training loss: 0.532907\n",
      "epoch: 26 [70/187 (37%)]\t training loss: 0.456186\n",
      "epoch: 26 [80/187 (43%)]\t training loss: 0.485202\n",
      "epoch: 26 [90/187 (48%)]\t training loss: 0.048937\n",
      "epoch: 26 [100/187 (53%)]\t training loss: 0.800713\n",
      "epoch: 26 [110/187 (59%)]\t training loss: 0.962860\n",
      "epoch: 26 [120/187 (64%)]\t training loss: 0.873832\n",
      "epoch: 26 [130/187 (70%)]\t training loss: 0.266773\n",
      "epoch: 26 [140/187 (75%)]\t training loss: 0.803537\n",
      "epoch: 26 [150/187 (80%)]\t training loss: 0.741084\n",
      "epoch: 26 [160/187 (86%)]\t training loss: 0.301548\n",
      "epoch: 26 [170/187 (91%)]\t training loss: 0.380209\n",
      "epoch: 26 [180/187 (96%)]\t training loss: 0.677527\n",
      "\n",
      "\n",
      "Validation...\n",
      "\n",
      "Test dataset: Overall Loss: 0.3354, Overall Accuracy: 2705/2977 (91%)\n",
      "\n",
      "epoch: 27 [0/187 (0%)]\t training loss: 0.778240\n",
      "epoch: 27 [10/187 (5%)]\t training loss: 0.397580\n",
      "epoch: 27 [20/187 (11%)]\t training loss: 0.552751\n",
      "epoch: 27 [30/187 (16%)]\t training loss: 0.769865\n",
      "epoch: 27 [40/187 (21%)]\t training loss: 0.281784\n",
      "epoch: 27 [50/187 (27%)]\t training loss: 0.116751\n",
      "epoch: 27 [60/187 (32%)]\t training loss: 0.646771\n",
      "epoch: 27 [70/187 (37%)]\t training loss: 0.221290\n",
      "epoch: 27 [80/187 (43%)]\t training loss: 0.705969\n",
      "epoch: 27 [90/187 (48%)]\t training loss: 0.845763\n",
      "epoch: 27 [100/187 (53%)]\t training loss: 0.161806\n",
      "epoch: 27 [110/187 (59%)]\t training loss: 0.266957\n",
      "epoch: 27 [120/187 (64%)]\t training loss: 0.546120\n",
      "epoch: 27 [130/187 (70%)]\t training loss: 0.035610\n",
      "epoch: 27 [140/187 (75%)]\t training loss: 0.056566\n",
      "epoch: 27 [150/187 (80%)]\t training loss: 0.710132\n",
      "epoch: 27 [160/187 (86%)]\t training loss: 0.192400\n",
      "epoch: 27 [170/187 (91%)]\t training loss: 0.615184\n",
      "epoch: 27 [180/187 (96%)]\t training loss: 0.708109\n",
      "\n",
      "\n",
      "Validation...\n",
      "\n",
      "Test dataset: Overall Loss: 0.2906, Overall Accuracy: 2748/2977 (92%)\n",
      "\n",
      "epoch: 28 [0/187 (0%)]\t training loss: 0.099498\n",
      "epoch: 28 [10/187 (5%)]\t training loss: 0.653386\n",
      "epoch: 28 [20/187 (11%)]\t training loss: 0.816887\n",
      "epoch: 28 [30/187 (16%)]\t training loss: 0.037745\n",
      "epoch: 28 [40/187 (21%)]\t training loss: 0.104180\n",
      "epoch: 28 [50/187 (27%)]\t training loss: 0.722813\n",
      "epoch: 28 [60/187 (32%)]\t training loss: 0.096786\n",
      "epoch: 28 [70/187 (37%)]\t training loss: 0.321521\n",
      "epoch: 28 [80/187 (43%)]\t training loss: 0.585764\n",
      "epoch: 28 [90/187 (48%)]\t training loss: 0.690332\n",
      "epoch: 28 [100/187 (53%)]\t training loss: 0.547164\n",
      "epoch: 28 [110/187 (59%)]\t training loss: 0.096102\n",
      "epoch: 28 [120/187 (64%)]\t training loss: 0.746131\n",
      "epoch: 28 [130/187 (70%)]\t training loss: 0.629135\n",
      "epoch: 28 [140/187 (75%)]\t training loss: 0.594391\n",
      "epoch: 28 [150/187 (80%)]\t training loss: 0.754226\n",
      "epoch: 28 [160/187 (86%)]\t training loss: 0.836496\n",
      "epoch: 28 [170/187 (91%)]\t training loss: 0.636848\n",
      "epoch: 28 [180/187 (96%)]\t training loss: 0.882455\n",
      "\n",
      "\n",
      "Validation...\n",
      "\n",
      "Test dataset: Overall Loss: 0.2911, Overall Accuracy: 2758/2977 (93%)\n",
      "\n",
      "epoch: 29 [0/187 (0%)]\t training loss: 0.720263\n",
      "epoch: 29 [10/187 (5%)]\t training loss: 0.790319\n",
      "epoch: 29 [20/187 (11%)]\t training loss: 0.875681\n",
      "epoch: 29 [30/187 (16%)]\t training loss: 0.595454\n",
      "epoch: 29 [40/187 (21%)]\t training loss: 0.867688\n",
      "epoch: 29 [50/187 (27%)]\t training loss: 0.357434\n",
      "epoch: 29 [60/187 (32%)]\t training loss: 0.553607\n",
      "epoch: 29 [70/187 (37%)]\t training loss: 0.098102\n",
      "epoch: 29 [80/187 (43%)]\t training loss: 0.628771\n",
      "epoch: 29 [90/187 (48%)]\t training loss: 0.188144\n",
      "epoch: 29 [100/187 (53%)]\t training loss: 0.748715\n",
      "epoch: 29 [110/187 (59%)]\t training loss: 0.516205\n",
      "epoch: 29 [120/187 (64%)]\t training loss: 0.766332\n",
      "epoch: 29 [130/187 (70%)]\t training loss: 0.668815\n",
      "epoch: 29 [140/187 (75%)]\t training loss: 0.275878\n",
      "epoch: 29 [150/187 (80%)]\t training loss: 0.215775\n",
      "epoch: 29 [160/187 (86%)]\t training loss: 0.100226\n",
      "epoch: 29 [170/187 (91%)]\t training loss: 0.770485\n",
      "epoch: 29 [180/187 (96%)]\t training loss: 0.625282\n",
      "\n",
      "\n",
      "Validation...\n",
      "\n",
      "Test dataset: Overall Loss: 0.2906, Overall Accuracy: 2742/2977 (92%)\n",
      "\n"
     ]
    }
   ],
   "source": [
    "imbalanced_model_mixup, train_losses, val_losses = train_mixup(\n",
    "    imbalanced_train_loader,\n",
    "    imbalanced_valid_loader,\n",
    "    \"imbalanced_model_mixup.pth\",\n",
    "    ConvNet().to(device),\n",
    "    epochs=30,\n",
    ")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 23,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 472
    },
    "id": "0vllpSby0kLG",
    "outputId": "f0646788-535d-45ae-9786-63048dcb4d9a"
   },
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 640x480 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "plt.plot(train_losses)\n",
    "plt.plot(val_losses)\n",
    "plt.xlabel(\"epoch\")\n",
    "plt.ylabel(\"loss\")\n",
    "plt.legend([\"Train\", \"Valid\"])\n",
    "plt.title(\"Model On imbalanced Fashion MNIST - Train vs Valid Accuracy\")\n",
    "\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 24,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/"
    },
    "id": "9BI70d9TtHn7",
    "outputId": "711ca7a4-48ab-49eb-949c-f37a695d5978"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "On Imbalanced Fashion MNIST dataset\n",
      "\n",
      "Test dataset: Overall Loss: 0.0016, Overall Accuracy: 8095/10000 (81%)\n",
      "\n",
      "Mixup test accuracy :  80.95\n"
     ]
    }
   ],
   "source": [
    "print(\"On Imbalanced Fashion MNIST dataset\")\n",
    "test_acc3, class_percentages3 = test(\n",
    "    imbalanced_model_mixup, \"imbalanced_model_mixup.pth\"\n",
    ")\n",
    "print(\"Mixup test accuracy : \", str(test_acc3))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "hicC1xVYmr3Z"
   },
   "source": [
    "# Cutmix + Mixup"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 25,
   "metadata": {
    "id": "sax2iFqOpM-X"
   },
   "outputs": [],
   "source": [
    "def train_cutmix_mixup(\n",
    "    train_dataloader,\n",
    "    valid_dataloader,\n",
    "    PATH,\n",
    "    model=None,\n",
    "    criterion=F.cross_entropy,\n",
    "    epochs=1,\n",
    "):\n",
    "    optimizer = torch.optim.SGD(model.parameters(), lr=0.1, momentum=0.9)\n",
    "    model.train()\n",
    "    best_acc = 0\n",
    "    train_losses = []\n",
    "    val_losses = []\n",
    "    for epoch in range(epochs):\n",
    "        losses = 0\n",
    "        for b_i, (X, y) in enumerate(train_dataloader):\n",
    "            X, y = X.to(device), y.to(device)\n",
    "            mix_decision = np.random.rand()\n",
    "            if mix_decision < 0.5:\n",
    "                X, ys = mixup(X, y, alpha=0.2)\n",
    "            if mix_decision >= 0.5:\n",
    "                X, ys = cutmix(X, y, alpha=0.2)\n",
    "            pred_prob = model(X)\n",
    "            targets1, targets2, lam = ys[0], ys[1], ys[2]\n",
    "            loss = lam * criterion(pred_prob, targets1) + (1 - lam) * criterion(\n",
    "                pred_prob, targets2\n",
    "            )\n",
    "            optimizer.zero_grad()\n",
    "            loss.backward()\n",
    "            optimizer.step()\n",
    "            losses += loss.item()  # print(losses)\n",
    "            if b_i % 10 == 0:\n",
    "                print(\n",
    "                    \"epoch: {} [{}/{} ({:.0f}%)]\\t training loss: {:.6f}\".format(\n",
    "                        epoch,\n",
    "                        b_i,\n",
    "                        len(train_dataloader),\n",
    "                        100.0 * b_i / len(train_dataloader),\n",
    "                        loss.item(),\n",
    "                    )\n",
    "                )\n",
    "        train_losses.append(losses / len(train_dataloader))\n",
    "        print(\"\\n\")\n",
    "        print(\"Validation...\")\n",
    "        model.eval()\n",
    "        success = 0\n",
    "        total = 0\n",
    "        losses = 0\n",
    "        with torch.no_grad():\n",
    "            for X, y in valid_dataloader:\n",
    "                X, y = X.to(device), y.to(device)\n",
    "                pred_prob = model(X)\n",
    "                loss = criterion(pred_prob, y)  # loss summed across the batch\n",
    "                losses += loss.item()\n",
    "                pred = pred_prob.argmax(\n",
    "                    dim=1, keepdim=True\n",
    "                )  # use argmax to get the most likely prediction\n",
    "                success += pred.eq(y.view_as(pred)).sum().item()\n",
    "                total += len(y)\n",
    "        val_losses.append(losses / len(valid_dataloader))\n",
    "        accuracy = 100.0 * success / total\n",
    "\n",
    "        if accuracy > best_acc:\n",
    "            torch.save(model.state_dict(), PATH)\n",
    "            best_acc = accuracy\n",
    "\n",
    "        print(\n",
    "            \"\\nTest dataset: Overall Loss: {:.4f}, Overall Accuracy: {}/{} ({:.0f}%)\\n\".format(\n",
    "                losses / len(valid_dataloader), success, total, accuracy\n",
    "            )\n",
    "        )\n",
    "    return model, train_losses, val_losses"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 26,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/"
    },
    "id": "0Wox1AyzrkQR",
    "outputId": "1991942c-1856-4eed-dc9e-32095d772381"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "epoch: 0 [0/187 (0%)]\t training loss: 2.315036\n",
      "epoch: 0 [10/187 (5%)]\t training loss: 1.186607\n",
      "epoch: 0 [20/187 (11%)]\t training loss: 1.119953\n",
      "epoch: 0 [30/187 (16%)]\t training loss: 1.213618\n",
      "epoch: 0 [40/187 (21%)]\t training loss: 0.466097\n",
      "epoch: 0 [50/187 (27%)]\t training loss: 0.735462\n",
      "epoch: 0 [60/187 (32%)]\t training loss: 1.197388\n",
      "epoch: 0 [70/187 (37%)]\t training loss: 0.692525\n",
      "epoch: 0 [80/187 (43%)]\t training loss: 1.140825\n",
      "epoch: 0 [90/187 (48%)]\t training loss: 0.753436\n",
      "epoch: 0 [100/187 (53%)]\t training loss: 1.333978\n",
      "epoch: 0 [110/187 (59%)]\t training loss: 0.954575\n",
      "epoch: 0 [120/187 (64%)]\t training loss: 0.910197\n",
      "epoch: 0 [130/187 (70%)]\t training loss: 1.080289\n",
      "epoch: 0 [140/187 (75%)]\t training loss: 0.695642\n",
      "epoch: 0 [150/187 (80%)]\t training loss: 0.944699\n",
      "epoch: 0 [160/187 (86%)]\t training loss: 0.712580\n",
      "epoch: 0 [170/187 (91%)]\t training loss: 0.582688\n",
      "epoch: 0 [180/187 (96%)]\t training loss: 1.018334\n",
      "\n",
      "\n",
      "Validation...\n",
      "\n",
      "Test dataset: Overall Loss: 0.3616, Overall Accuracy: 2658/2977 (89%)\n",
      "\n",
      "epoch: 1 [0/187 (0%)]\t training loss: 0.684104\n",
      "epoch: 1 [10/187 (5%)]\t training loss: 1.025427\n",
      "epoch: 1 [20/187 (11%)]\t training loss: 0.953670\n",
      "epoch: 1 [30/187 (16%)]\t training loss: 0.859358\n",
      "epoch: 1 [40/187 (21%)]\t training loss: 0.604845\n",
      "epoch: 1 [50/187 (27%)]\t training loss: 0.366721\n",
      "epoch: 1 [60/187 (32%)]\t training loss: 0.891390\n",
      "epoch: 1 [70/187 (37%)]\t training loss: 0.900455\n",
      "epoch: 1 [80/187 (43%)]\t training loss: 0.913498\n",
      "epoch: 1 [90/187 (48%)]\t training loss: 0.849863\n",
      "epoch: 1 [100/187 (53%)]\t training loss: 0.951621\n",
      "epoch: 1 [110/187 (59%)]\t training loss: 0.724760\n",
      "epoch: 1 [120/187 (64%)]\t training loss: 0.259705\n",
      "epoch: 1 [130/187 (70%)]\t training loss: 0.679884\n",
      "epoch: 1 [140/187 (75%)]\t training loss: 0.883380\n",
      "epoch: 1 [150/187 (80%)]\t training loss: 0.661629\n",
      "epoch: 1 [160/187 (86%)]\t training loss: 1.007027\n",
      "epoch: 1 [170/187 (91%)]\t training loss: 0.356687\n",
      "epoch: 1 [180/187 (96%)]\t training loss: 0.786742\n",
      "\n",
      "\n",
      "Validation...\n",
      "\n",
      "Test dataset: Overall Loss: 0.3398, Overall Accuracy: 2663/2977 (89%)\n",
      "\n",
      "epoch: 2 [0/187 (0%)]\t training loss: 0.737339\n",
      "epoch: 2 [10/187 (5%)]\t training loss: 0.841171\n",
      "epoch: 2 [20/187 (11%)]\t training loss: 0.882761\n",
      "epoch: 2 [30/187 (16%)]\t training loss: 0.893149\n",
      "epoch: 2 [40/187 (21%)]\t training loss: 0.481696\n",
      "epoch: 2 [50/187 (27%)]\t training loss: 0.689243\n",
      "epoch: 2 [60/187 (32%)]\t training loss: 0.232581\n",
      "epoch: 2 [70/187 (37%)]\t training loss: 0.399724\n",
      "epoch: 2 [80/187 (43%)]\t training loss: 0.469565\n",
      "epoch: 2 [90/187 (48%)]\t training loss: 0.572524\n",
      "epoch: 2 [100/187 (53%)]\t training loss: 0.918008\n",
      "epoch: 2 [110/187 (59%)]\t training loss: 0.548642\n",
      "epoch: 2 [120/187 (64%)]\t training loss: 0.995826\n",
      "epoch: 2 [130/187 (70%)]\t training loss: 0.316125\n",
      "epoch: 2 [140/187 (75%)]\t training loss: 0.675598\n",
      "epoch: 2 [150/187 (80%)]\t training loss: 0.781841\n",
      "epoch: 2 [160/187 (86%)]\t training loss: 0.946500\n",
      "epoch: 2 [170/187 (91%)]\t training loss: 0.677991\n",
      "epoch: 2 [180/187 (96%)]\t training loss: 0.575302\n",
      "\n",
      "\n",
      "Validation...\n",
      "\n",
      "Test dataset: Overall Loss: 0.3378, Overall Accuracy: 2672/2977 (90%)\n",
      "\n",
      "epoch: 3 [0/187 (0%)]\t training loss: 0.201510\n",
      "epoch: 3 [10/187 (5%)]\t training loss: 0.807971\n",
      "epoch: 3 [20/187 (11%)]\t training loss: 0.813981\n",
      "epoch: 3 [30/187 (16%)]\t training loss: 0.637909\n",
      "epoch: 3 [40/187 (21%)]\t training loss: 0.234879\n",
      "epoch: 3 [50/187 (27%)]\t training loss: 0.763103\n",
      "epoch: 3 [60/187 (32%)]\t training loss: 0.725755\n",
      "epoch: 3 [70/187 (37%)]\t training loss: 0.973643\n",
      "epoch: 3 [80/187 (43%)]\t training loss: 0.897141\n",
      "epoch: 3 [90/187 (48%)]\t training loss: 0.891716\n",
      "epoch: 3 [100/187 (53%)]\t training loss: 0.188469\n",
      "epoch: 3 [110/187 (59%)]\t training loss: 0.628856\n",
      "epoch: 3 [120/187 (64%)]\t training loss: 0.812394\n",
      "epoch: 3 [130/187 (70%)]\t training loss: 1.018229\n",
      "epoch: 3 [140/187 (75%)]\t training loss: 0.547436\n",
      "epoch: 3 [150/187 (80%)]\t training loss: 0.846321\n",
      "epoch: 3 [160/187 (86%)]\t training loss: 0.761077\n",
      "epoch: 3 [170/187 (91%)]\t training loss: 0.695692\n",
      "epoch: 3 [180/187 (96%)]\t training loss: 0.238672\n",
      "\n",
      "\n",
      "Validation...\n",
      "\n",
      "Test dataset: Overall Loss: 0.3730, Overall Accuracy: 2648/2977 (89%)\n",
      "\n",
      "epoch: 4 [0/187 (0%)]\t training loss: 0.587198\n",
      "epoch: 4 [10/187 (5%)]\t training loss: 0.704794\n",
      "epoch: 4 [20/187 (11%)]\t training loss: 0.550698\n",
      "epoch: 4 [30/187 (16%)]\t training loss: 0.316426\n",
      "epoch: 4 [40/187 (21%)]\t training loss: 0.798802\n",
      "epoch: 4 [50/187 (27%)]\t training loss: 0.952905\n",
      "epoch: 4 [60/187 (32%)]\t training loss: 0.325072\n",
      "epoch: 4 [70/187 (37%)]\t training loss: 0.803339\n",
      "epoch: 4 [80/187 (43%)]\t training loss: 0.224878\n",
      "epoch: 4 [90/187 (48%)]\t training loss: 0.701242\n",
      "epoch: 4 [100/187 (53%)]\t training loss: 0.867616\n",
      "epoch: 4 [110/187 (59%)]\t training loss: 0.325345\n",
      "epoch: 4 [120/187 (64%)]\t training loss: 0.851533\n",
      "epoch: 4 [130/187 (70%)]\t training loss: 0.265636\n",
      "epoch: 4 [140/187 (75%)]\t training loss: 0.691456\n",
      "epoch: 4 [150/187 (80%)]\t training loss: 0.978187\n",
      "epoch: 4 [160/187 (86%)]\t training loss: 0.403344\n",
      "epoch: 4 [170/187 (91%)]\t training loss: 0.503600\n",
      "epoch: 4 [180/187 (96%)]\t training loss: 0.869015\n",
      "\n",
      "\n",
      "Validation...\n",
      "\n",
      "Test dataset: Overall Loss: 0.2997, Overall Accuracy: 2715/2977 (91%)\n",
      "\n",
      "epoch: 5 [0/187 (0%)]\t training loss: 0.640413\n",
      "epoch: 5 [10/187 (5%)]\t training loss: 0.418071\n",
      "epoch: 5 [20/187 (11%)]\t training loss: 0.725189\n",
      "epoch: 5 [30/187 (16%)]\t training loss: 0.395155\n",
      "epoch: 5 [40/187 (21%)]\t training loss: 0.803403\n",
      "epoch: 5 [50/187 (27%)]\t training loss: 0.959602\n",
      "epoch: 5 [60/187 (32%)]\t training loss: 0.544150\n",
      "epoch: 5 [70/187 (37%)]\t training loss: 0.859567\n",
      "epoch: 5 [80/187 (43%)]\t training loss: 0.342721\n",
      "epoch: 5 [90/187 (48%)]\t training loss: 0.788788\n",
      "epoch: 5 [100/187 (53%)]\t training loss: 0.198933\n",
      "epoch: 5 [110/187 (59%)]\t training loss: 0.524358\n",
      "epoch: 5 [120/187 (64%)]\t training loss: 0.840622\n",
      "epoch: 5 [130/187 (70%)]\t training loss: 0.629674\n",
      "epoch: 5 [140/187 (75%)]\t training loss: 0.319458\n",
      "epoch: 5 [150/187 (80%)]\t training loss: 0.964797\n",
      "epoch: 5 [160/187 (86%)]\t training loss: 0.175109\n",
      "epoch: 5 [170/187 (91%)]\t training loss: 0.799687\n",
      "epoch: 5 [180/187 (96%)]\t training loss: 0.421107\n",
      "\n",
      "\n",
      "Validation...\n",
      "\n",
      "Test dataset: Overall Loss: 0.2763, Overall Accuracy: 2735/2977 (92%)\n",
      "\n",
      "epoch: 6 [0/187 (0%)]\t training loss: 0.798001\n",
      "epoch: 6 [10/187 (5%)]\t training loss: 0.423902\n",
      "epoch: 6 [20/187 (11%)]\t training loss: 0.445423\n",
      "epoch: 6 [30/187 (16%)]\t training loss: 0.693834\n",
      "epoch: 6 [40/187 (21%)]\t training loss: 1.030391\n",
      "epoch: 6 [50/187 (27%)]\t training loss: 0.872095\n",
      "epoch: 6 [60/187 (32%)]\t training loss: 0.595107\n",
      "epoch: 6 [70/187 (37%)]\t training loss: 0.232704\n",
      "epoch: 6 [80/187 (43%)]\t training loss: 0.823911\n",
      "epoch: 6 [90/187 (48%)]\t training loss: 0.874717\n",
      "epoch: 6 [100/187 (53%)]\t training loss: 0.571933\n",
      "epoch: 6 [110/187 (59%)]\t training loss: 0.412759\n",
      "epoch: 6 [120/187 (64%)]\t training loss: 0.147120\n",
      "epoch: 6 [130/187 (70%)]\t training loss: 0.678580\n",
      "epoch: 6 [140/187 (75%)]\t training loss: 0.850776\n",
      "epoch: 6 [150/187 (80%)]\t training loss: 0.260361\n",
      "epoch: 6 [160/187 (86%)]\t training loss: 0.320420\n",
      "epoch: 6 [170/187 (91%)]\t training loss: 0.684250\n",
      "epoch: 6 [180/187 (96%)]\t training loss: 0.734301\n",
      "\n",
      "\n",
      "Validation...\n",
      "\n",
      "Test dataset: Overall Loss: 0.2803, Overall Accuracy: 2749/2977 (92%)\n",
      "\n",
      "epoch: 7 [0/187 (0%)]\t training loss: 0.580284\n",
      "epoch: 7 [10/187 (5%)]\t training loss: 0.635666\n",
      "epoch: 7 [20/187 (11%)]\t training loss: 0.846372\n",
      "epoch: 7 [30/187 (16%)]\t training loss: 0.143697\n",
      "epoch: 7 [40/187 (21%)]\t training loss: 0.741683\n",
      "epoch: 7 [50/187 (27%)]\t training loss: 0.734379\n",
      "epoch: 7 [60/187 (32%)]\t training loss: 0.202404\n",
      "epoch: 7 [70/187 (37%)]\t training loss: 0.664399\n",
      "epoch: 7 [80/187 (43%)]\t training loss: 0.732739\n",
      "epoch: 7 [90/187 (48%)]\t training loss: 0.946491\n",
      "epoch: 7 [100/187 (53%)]\t training loss: 0.830966\n",
      "epoch: 7 [110/187 (59%)]\t training loss: 0.157396\n",
      "epoch: 7 [120/187 (64%)]\t training loss: 0.850221\n",
      "epoch: 7 [130/187 (70%)]\t training loss: 0.367934\n",
      "epoch: 7 [140/187 (75%)]\t training loss: 0.130806\n",
      "epoch: 7 [150/187 (80%)]\t training loss: 0.681340\n",
      "epoch: 7 [160/187 (86%)]\t training loss: 0.874286\n",
      "epoch: 7 [170/187 (91%)]\t training loss: 0.648049\n",
      "epoch: 7 [180/187 (96%)]\t training loss: 0.176568\n",
      "\n",
      "\n",
      "Validation...\n",
      "\n",
      "Test dataset: Overall Loss: 0.2746, Overall Accuracy: 2748/2977 (92%)\n",
      "\n",
      "epoch: 8 [0/187 (0%)]\t training loss: 0.696063\n",
      "epoch: 8 [10/187 (5%)]\t training loss: 0.140114\n",
      "epoch: 8 [20/187 (11%)]\t training loss: 0.355667\n",
      "epoch: 8 [30/187 (16%)]\t training loss: 0.802131\n",
      "epoch: 8 [40/187 (21%)]\t training loss: 0.161613\n",
      "epoch: 8 [50/187 (27%)]\t training loss: 0.605100\n",
      "epoch: 8 [60/187 (32%)]\t training loss: 0.770736\n",
      "epoch: 8 [70/187 (37%)]\t training loss: 0.875868\n",
      "epoch: 8 [80/187 (43%)]\t training loss: 0.713378\n",
      "epoch: 8 [90/187 (48%)]\t training loss: 0.746232\n",
      "epoch: 8 [100/187 (53%)]\t training loss: 0.272921\n",
      "epoch: 8 [110/187 (59%)]\t training loss: 0.547516\n",
      "epoch: 8 [120/187 (64%)]\t training loss: 0.928404\n",
      "epoch: 8 [130/187 (70%)]\t training loss: 0.341470\n",
      "epoch: 8 [140/187 (75%)]\t training loss: 0.678602\n",
      "epoch: 8 [150/187 (80%)]\t training loss: 0.696152\n",
      "epoch: 8 [160/187 (86%)]\t training loss: 0.574847\n",
      "epoch: 8 [170/187 (91%)]\t training loss: 0.894613\n",
      "epoch: 8 [180/187 (96%)]\t training loss: 0.683710\n",
      "\n",
      "\n",
      "Validation...\n",
      "\n",
      "Test dataset: Overall Loss: 0.2618, Overall Accuracy: 2756/2977 (93%)\n",
      "\n",
      "epoch: 9 [0/187 (0%)]\t training loss: 0.843363\n",
      "epoch: 9 [10/187 (5%)]\t training loss: 0.464473\n",
      "epoch: 9 [20/187 (11%)]\t training loss: 0.126999\n",
      "epoch: 9 [30/187 (16%)]\t training loss: 0.876242\n",
      "epoch: 9 [40/187 (21%)]\t training loss: 0.616007\n",
      "epoch: 9 [50/187 (27%)]\t training loss: 0.517582\n",
      "epoch: 9 [60/187 (32%)]\t training loss: 1.007165\n",
      "epoch: 9 [70/187 (37%)]\t training loss: 0.538349\n",
      "epoch: 9 [80/187 (43%)]\t training loss: 0.147818\n",
      "epoch: 9 [90/187 (48%)]\t training loss: 0.856241\n",
      "epoch: 9 [100/187 (53%)]\t training loss: 0.460506\n",
      "epoch: 9 [110/187 (59%)]\t training loss: 0.855244\n",
      "epoch: 9 [120/187 (64%)]\t training loss: 0.812281\n",
      "epoch: 9 [130/187 (70%)]\t training loss: 0.721627\n",
      "epoch: 9 [140/187 (75%)]\t training loss: 0.607043\n",
      "epoch: 9 [150/187 (80%)]\t training loss: 0.433436\n",
      "epoch: 9 [160/187 (86%)]\t training loss: 0.512302\n",
      "epoch: 9 [170/187 (91%)]\t training loss: 0.242875\n",
      "epoch: 9 [180/187 (96%)]\t training loss: 0.641864\n",
      "\n",
      "\n",
      "Validation...\n",
      "\n",
      "Test dataset: Overall Loss: 0.3010, Overall Accuracy: 2732/2977 (92%)\n",
      "\n",
      "epoch: 10 [0/187 (0%)]\t training loss: 0.839659\n",
      "epoch: 10 [10/187 (5%)]\t training loss: 1.063884\n",
      "epoch: 10 [20/187 (11%)]\t training loss: 0.561607\n",
      "epoch: 10 [30/187 (16%)]\t training loss: 0.701757\n",
      "epoch: 10 [40/187 (21%)]\t training loss: 0.936166\n",
      "epoch: 10 [50/187 (27%)]\t training loss: 0.830319\n",
      "epoch: 10 [60/187 (32%)]\t training loss: 0.290866\n",
      "epoch: 10 [70/187 (37%)]\t training loss: 0.665423\n",
      "epoch: 10 [80/187 (43%)]\t training loss: 0.812580\n",
      "epoch: 10 [90/187 (48%)]\t training loss: 0.729782\n",
      "epoch: 10 [100/187 (53%)]\t training loss: 0.659500\n",
      "epoch: 10 [110/187 (59%)]\t training loss: 0.238559\n",
      "epoch: 10 [120/187 (64%)]\t training loss: 0.381276\n",
      "epoch: 10 [130/187 (70%)]\t training loss: 0.334976\n",
      "epoch: 10 [140/187 (75%)]\t training loss: 0.155406\n",
      "epoch: 10 [150/187 (80%)]\t training loss: 0.510325\n",
      "epoch: 10 [160/187 (86%)]\t training loss: 0.665111\n",
      "epoch: 10 [170/187 (91%)]\t training loss: 0.735061\n",
      "epoch: 10 [180/187 (96%)]\t training loss: 0.838027\n",
      "\n",
      "\n",
      "Validation...\n",
      "\n",
      "Test dataset: Overall Loss: 0.2881, Overall Accuracy: 2732/2977 (92%)\n",
      "\n",
      "epoch: 11 [0/187 (0%)]\t training loss: 0.278087\n",
      "epoch: 11 [10/187 (5%)]\t training loss: 0.322361\n",
      "epoch: 11 [20/187 (11%)]\t training loss: 0.676885\n",
      "epoch: 11 [30/187 (16%)]\t training loss: 0.149641\n",
      "epoch: 11 [40/187 (21%)]\t training loss: 0.244310\n",
      "epoch: 11 [50/187 (27%)]\t training loss: 0.193421\n",
      "epoch: 11 [60/187 (32%)]\t training loss: 0.695079\n",
      "epoch: 11 [70/187 (37%)]\t training loss: 0.974736\n",
      "epoch: 11 [80/187 (43%)]\t training loss: 0.080252\n",
      "epoch: 11 [90/187 (48%)]\t training loss: 0.673259\n",
      "epoch: 11 [100/187 (53%)]\t training loss: 0.104898\n",
      "epoch: 11 [110/187 (59%)]\t training loss: 0.662255\n",
      "epoch: 11 [120/187 (64%)]\t training loss: 0.250958\n",
      "epoch: 11 [130/187 (70%)]\t training loss: 0.766896\n",
      "epoch: 11 [140/187 (75%)]\t training loss: 0.088483\n",
      "epoch: 11 [150/187 (80%)]\t training loss: 0.130888\n",
      "epoch: 11 [160/187 (86%)]\t training loss: 0.865420\n",
      "epoch: 11 [170/187 (91%)]\t training loss: 0.663850\n",
      "epoch: 11 [180/187 (96%)]\t training loss: 0.364931\n",
      "\n",
      "\n",
      "Validation...\n",
      "\n",
      "Test dataset: Overall Loss: 0.2606, Overall Accuracy: 2760/2977 (93%)\n",
      "\n",
      "epoch: 12 [0/187 (0%)]\t training loss: 0.172431\n",
      "epoch: 12 [10/187 (5%)]\t training loss: 0.617211\n",
      "epoch: 12 [20/187 (11%)]\t training loss: 0.868217\n",
      "epoch: 12 [30/187 (16%)]\t training loss: 0.761111\n",
      "epoch: 12 [40/187 (21%)]\t training loss: 0.719246\n",
      "epoch: 12 [50/187 (27%)]\t training loss: 0.630214\n",
      "epoch: 12 [60/187 (32%)]\t training loss: 0.489061\n",
      "epoch: 12 [70/187 (37%)]\t training loss: 0.114113\n",
      "epoch: 12 [80/187 (43%)]\t training loss: 0.824750\n",
      "epoch: 12 [90/187 (48%)]\t training loss: 0.562026\n",
      "epoch: 12 [100/187 (53%)]\t training loss: 0.177657\n",
      "epoch: 12 [110/187 (59%)]\t training loss: 0.336118\n",
      "epoch: 12 [120/187 (64%)]\t training loss: 0.490418\n",
      "epoch: 12 [130/187 (70%)]\t training loss: 0.544285\n",
      "epoch: 12 [140/187 (75%)]\t training loss: 0.463861\n",
      "epoch: 12 [150/187 (80%)]\t training loss: 0.462665\n",
      "epoch: 12 [160/187 (86%)]\t training loss: 0.637481\n",
      "epoch: 12 [170/187 (91%)]\t training loss: 0.516875\n",
      "epoch: 12 [180/187 (96%)]\t training loss: 0.304950\n",
      "\n",
      "\n",
      "Validation...\n",
      "\n",
      "Test dataset: Overall Loss: 0.2653, Overall Accuracy: 2748/2977 (92%)\n",
      "\n",
      "epoch: 13 [0/187 (0%)]\t training loss: 0.838375\n",
      "epoch: 13 [10/187 (5%)]\t training loss: 0.741250\n",
      "epoch: 13 [20/187 (11%)]\t training loss: 0.235091\n",
      "epoch: 13 [30/187 (16%)]\t training loss: 0.847162\n",
      "epoch: 13 [40/187 (21%)]\t training loss: 0.073039\n",
      "epoch: 13 [50/187 (27%)]\t training loss: 0.509325\n",
      "epoch: 13 [60/187 (32%)]\t training loss: 0.330883\n",
      "epoch: 13 [70/187 (37%)]\t training loss: 0.129459\n",
      "epoch: 13 [80/187 (43%)]\t training loss: 0.836250\n",
      "epoch: 13 [90/187 (48%)]\t training loss: 0.428376\n",
      "epoch: 13 [100/187 (53%)]\t training loss: 0.731705\n",
      "epoch: 13 [110/187 (59%)]\t training loss: 0.142774\n",
      "epoch: 13 [120/187 (64%)]\t training loss: 0.456586\n",
      "epoch: 13 [130/187 (70%)]\t training loss: 0.294248\n",
      "epoch: 13 [140/187 (75%)]\t training loss: 0.421307\n",
      "epoch: 13 [150/187 (80%)]\t training loss: 0.223001\n",
      "epoch: 13 [160/187 (86%)]\t training loss: 0.582608\n",
      "epoch: 13 [170/187 (91%)]\t training loss: 0.605174\n",
      "epoch: 13 [180/187 (96%)]\t training loss: 0.545250\n",
      "\n",
      "\n",
      "Validation...\n",
      "\n",
      "Test dataset: Overall Loss: 0.2662, Overall Accuracy: 2759/2977 (93%)\n",
      "\n",
      "epoch: 14 [0/187 (0%)]\t training loss: 0.555726\n",
      "epoch: 14 [10/187 (5%)]\t training loss: 0.170961\n",
      "epoch: 14 [20/187 (11%)]\t training loss: 0.802613\n",
      "epoch: 14 [30/187 (16%)]\t training loss: 0.557889\n",
      "epoch: 14 [40/187 (21%)]\t training loss: 0.768536\n",
      "epoch: 14 [50/187 (27%)]\t training loss: 0.802791\n",
      "epoch: 14 [60/187 (32%)]\t training loss: 0.277905\n",
      "epoch: 14 [70/187 (37%)]\t training loss: 0.706453\n",
      "epoch: 14 [80/187 (43%)]\t training loss: 0.966220\n",
      "epoch: 14 [90/187 (48%)]\t training loss: 0.198509\n",
      "epoch: 14 [100/187 (53%)]\t training loss: 0.596960\n",
      "epoch: 14 [110/187 (59%)]\t training loss: 0.664777\n",
      "epoch: 14 [120/187 (64%)]\t training loss: 0.169173\n",
      "epoch: 14 [130/187 (70%)]\t training loss: 0.851249\n",
      "epoch: 14 [140/187 (75%)]\t training loss: 0.630952\n",
      "epoch: 14 [150/187 (80%)]\t training loss: 0.103309\n",
      "epoch: 14 [160/187 (86%)]\t training loss: 0.655290\n",
      "epoch: 14 [170/187 (91%)]\t training loss: 0.648531\n",
      "epoch: 14 [180/187 (96%)]\t training loss: 0.836491\n",
      "\n",
      "\n",
      "Validation...\n",
      "\n",
      "Test dataset: Overall Loss: 0.2607, Overall Accuracy: 2765/2977 (93%)\n",
      "\n",
      "epoch: 15 [0/187 (0%)]\t training loss: 0.244955\n",
      "epoch: 15 [10/187 (5%)]\t training loss: 0.641429\n",
      "epoch: 15 [20/187 (11%)]\t training loss: 0.657730\n",
      "epoch: 15 [30/187 (16%)]\t training loss: 0.821515\n",
      "epoch: 15 [40/187 (21%)]\t training loss: 0.713318\n",
      "epoch: 15 [50/187 (27%)]\t training loss: 0.664154\n",
      "epoch: 15 [60/187 (32%)]\t training loss: 0.219781\n",
      "epoch: 15 [70/187 (37%)]\t training loss: 0.674791\n",
      "epoch: 15 [80/187 (43%)]\t training loss: 0.324816\n",
      "epoch: 15 [90/187 (48%)]\t training loss: 0.768136\n",
      "epoch: 15 [100/187 (53%)]\t training loss: 0.797924\n",
      "epoch: 15 [110/187 (59%)]\t training loss: 0.524208\n",
      "epoch: 15 [120/187 (64%)]\t training loss: 0.100171\n",
      "epoch: 15 [130/187 (70%)]\t training loss: 0.564847\n",
      "epoch: 15 [140/187 (75%)]\t training loss: 0.454273\n",
      "epoch: 15 [150/187 (80%)]\t training loss: 0.734180\n",
      "epoch: 15 [160/187 (86%)]\t training loss: 0.717202\n",
      "epoch: 15 [170/187 (91%)]\t training loss: 0.332109\n",
      "epoch: 15 [180/187 (96%)]\t training loss: 0.535396\n",
      "\n",
      "\n",
      "Validation...\n",
      "\n",
      "Test dataset: Overall Loss: 0.2946, Overall Accuracy: 2737/2977 (92%)\n",
      "\n",
      "epoch: 16 [0/187 (0%)]\t training loss: 0.699509\n",
      "epoch: 16 [10/187 (5%)]\t training loss: 0.770609\n",
      "epoch: 16 [20/187 (11%)]\t training loss: 0.134850\n",
      "epoch: 16 [30/187 (16%)]\t training loss: 0.664836\n",
      "epoch: 16 [40/187 (21%)]\t training loss: 0.768192\n",
      "epoch: 16 [50/187 (27%)]\t training loss: 0.134122\n",
      "epoch: 16 [60/187 (32%)]\t training loss: 0.787758\n",
      "epoch: 16 [70/187 (37%)]\t training loss: 0.544747\n",
      "epoch: 16 [80/187 (43%)]\t training loss: 0.857680\n",
      "epoch: 16 [90/187 (48%)]\t training loss: 0.293808\n",
      "epoch: 16 [100/187 (53%)]\t training loss: 0.628909\n",
      "epoch: 16 [110/187 (59%)]\t training loss: 0.244424\n",
      "epoch: 16 [120/187 (64%)]\t training loss: 0.547644\n",
      "epoch: 16 [130/187 (70%)]\t training loss: 0.651307\n",
      "epoch: 16 [140/187 (75%)]\t training loss: 0.390646\n",
      "epoch: 16 [150/187 (80%)]\t training loss: 0.687340\n",
      "epoch: 16 [160/187 (86%)]\t training loss: 0.838650\n",
      "epoch: 16 [170/187 (91%)]\t training loss: 0.763587\n",
      "epoch: 16 [180/187 (96%)]\t training loss: 0.214056\n",
      "\n",
      "\n",
      "Validation...\n",
      "\n",
      "Test dataset: Overall Loss: 0.2837, Overall Accuracy: 2751/2977 (92%)\n",
      "\n",
      "epoch: 17 [0/187 (0%)]\t training loss: 0.527467\n",
      "epoch: 17 [10/187 (5%)]\t training loss: 1.010869\n",
      "epoch: 17 [20/187 (11%)]\t training loss: 0.678604\n",
      "epoch: 17 [30/187 (16%)]\t training loss: 0.541852\n",
      "epoch: 17 [40/187 (21%)]\t training loss: 0.749009\n",
      "epoch: 17 [50/187 (27%)]\t training loss: 0.766642\n",
      "epoch: 17 [60/187 (32%)]\t training loss: 0.274764\n",
      "epoch: 17 [70/187 (37%)]\t training loss: 0.554737\n",
      "epoch: 17 [80/187 (43%)]\t training loss: 0.560320\n",
      "epoch: 17 [90/187 (48%)]\t training loss: 0.415434\n",
      "epoch: 17 [100/187 (53%)]\t training loss: 0.377334\n",
      "epoch: 17 [110/187 (59%)]\t training loss: 0.540634\n",
      "epoch: 17 [120/187 (64%)]\t training loss: 0.803085\n",
      "epoch: 17 [130/187 (70%)]\t training loss: 0.475076\n",
      "epoch: 17 [140/187 (75%)]\t training loss: 0.087642\n",
      "epoch: 17 [150/187 (80%)]\t training loss: 0.323115\n",
      "epoch: 17 [160/187 (86%)]\t training loss: 0.820181\n",
      "epoch: 17 [170/187 (91%)]\t training loss: 0.183651\n",
      "epoch: 17 [180/187 (96%)]\t training loss: 0.601879\n",
      "\n",
      "\n",
      "Validation...\n",
      "\n",
      "Test dataset: Overall Loss: 0.2894, Overall Accuracy: 2742/2977 (92%)\n",
      "\n",
      "epoch: 18 [0/187 (0%)]\t training loss: 0.084811\n",
      "epoch: 18 [10/187 (5%)]\t training loss: 0.614717\n",
      "epoch: 18 [20/187 (11%)]\t training loss: 0.871404\n",
      "epoch: 18 [30/187 (16%)]\t training loss: 0.510124\n",
      "epoch: 18 [40/187 (21%)]\t training loss: 0.121406\n",
      "epoch: 18 [50/187 (27%)]\t training loss: 0.334000\n",
      "epoch: 18 [60/187 (32%)]\t training loss: 0.225828\n",
      "epoch: 18 [70/187 (37%)]\t training loss: 0.416677\n",
      "epoch: 18 [80/187 (43%)]\t training loss: 0.147605\n",
      "epoch: 18 [90/187 (48%)]\t training loss: 0.129618\n",
      "epoch: 18 [100/187 (53%)]\t training loss: 0.638337\n",
      "epoch: 18 [110/187 (59%)]\t training loss: 0.445842\n",
      "epoch: 18 [120/187 (64%)]\t training loss: 0.563510\n",
      "epoch: 18 [130/187 (70%)]\t training loss: 0.114626\n",
      "epoch: 18 [140/187 (75%)]\t training loss: 0.339048\n",
      "epoch: 18 [150/187 (80%)]\t training loss: 0.330914\n",
      "epoch: 18 [160/187 (86%)]\t training loss: 0.295143\n",
      "epoch: 18 [170/187 (91%)]\t training loss: 0.476466\n",
      "epoch: 18 [180/187 (96%)]\t training loss: 0.149509\n",
      "\n",
      "\n",
      "Validation...\n",
      "\n",
      "Test dataset: Overall Loss: 0.2655, Overall Accuracy: 2754/2977 (93%)\n",
      "\n",
      "epoch: 19 [0/187 (0%)]\t training loss: 0.073239\n",
      "epoch: 19 [10/187 (5%)]\t training loss: 0.763943\n",
      "epoch: 19 [20/187 (11%)]\t training loss: 0.743438\n",
      "epoch: 19 [30/187 (16%)]\t training loss: 0.506673\n",
      "epoch: 19 [40/187 (21%)]\t training loss: 0.225211\n",
      "epoch: 19 [50/187 (27%)]\t training loss: 0.907241\n",
      "epoch: 19 [60/187 (32%)]\t training loss: 0.259837\n",
      "epoch: 19 [70/187 (37%)]\t training loss: 0.735477\n",
      "epoch: 19 [80/187 (43%)]\t training loss: 0.905943\n",
      "epoch: 19 [90/187 (48%)]\t training loss: 0.143629\n",
      "epoch: 19 [100/187 (53%)]\t training loss: 0.127829\n",
      "epoch: 19 [110/187 (59%)]\t training loss: 0.414102\n",
      "epoch: 19 [120/187 (64%)]\t training loss: 0.495784\n",
      "epoch: 19 [130/187 (70%)]\t training loss: 0.591908\n",
      "epoch: 19 [140/187 (75%)]\t training loss: 0.783172\n",
      "epoch: 19 [150/187 (80%)]\t training loss: 0.767195\n",
      "epoch: 19 [160/187 (86%)]\t training loss: 0.726020\n",
      "epoch: 19 [170/187 (91%)]\t training loss: 0.792266\n",
      "epoch: 19 [180/187 (96%)]\t training loss: 0.660258\n",
      "\n",
      "\n",
      "Validation...\n",
      "\n",
      "Test dataset: Overall Loss: 0.2835, Overall Accuracy: 2744/2977 (92%)\n",
      "\n",
      "epoch: 20 [0/187 (0%)]\t training loss: 0.639332\n",
      "epoch: 20 [10/187 (5%)]\t training loss: 0.379569\n",
      "epoch: 20 [20/187 (11%)]\t training loss: 0.795001\n",
      "epoch: 20 [30/187 (16%)]\t training loss: 0.193657\n",
      "epoch: 20 [40/187 (21%)]\t training loss: 0.839182\n",
      "epoch: 20 [50/187 (27%)]\t training loss: 0.140799\n",
      "epoch: 20 [60/187 (32%)]\t training loss: 0.663845\n",
      "epoch: 20 [70/187 (37%)]\t training loss: 0.771635\n",
      "epoch: 20 [80/187 (43%)]\t training loss: 0.752204\n",
      "epoch: 20 [90/187 (48%)]\t training loss: 0.687925\n",
      "epoch: 20 [100/187 (53%)]\t training loss: 0.705890\n",
      "epoch: 20 [110/187 (59%)]\t training loss: 0.570631\n",
      "epoch: 20 [120/187 (64%)]\t training loss: 0.455089\n",
      "epoch: 20 [130/187 (70%)]\t training loss: 0.779880\n",
      "epoch: 20 [140/187 (75%)]\t training loss: 0.561450\n",
      "epoch: 20 [150/187 (80%)]\t training loss: 0.210841\n",
      "epoch: 20 [160/187 (86%)]\t training loss: 0.315259\n",
      "epoch: 20 [170/187 (91%)]\t training loss: 0.538583\n",
      "epoch: 20 [180/187 (96%)]\t training loss: 0.801683\n",
      "\n",
      "\n",
      "Validation...\n",
      "\n",
      "Test dataset: Overall Loss: 0.3042, Overall Accuracy: 2742/2977 (92%)\n",
      "\n",
      "epoch: 21 [0/187 (0%)]\t training loss: 0.791084\n",
      "epoch: 21 [10/187 (5%)]\t training loss: 0.283113\n",
      "epoch: 21 [20/187 (11%)]\t training loss: 0.733769\n",
      "epoch: 21 [30/187 (16%)]\t training loss: 0.057110\n",
      "epoch: 21 [40/187 (21%)]\t training loss: 0.707961\n",
      "epoch: 21 [50/187 (27%)]\t training loss: 0.681828\n",
      "epoch: 21 [60/187 (32%)]\t training loss: 0.319339\n",
      "epoch: 21 [70/187 (37%)]\t training loss: 0.635503\n",
      "epoch: 21 [80/187 (43%)]\t training loss: 0.751993\n",
      "epoch: 21 [90/187 (48%)]\t training loss: 0.627024\n",
      "epoch: 21 [100/187 (53%)]\t training loss: 0.411048\n",
      "epoch: 21 [110/187 (59%)]\t training loss: 0.136777\n",
      "epoch: 21 [120/187 (64%)]\t training loss: 0.658777\n",
      "epoch: 21 [130/187 (70%)]\t training loss: 0.592270\n",
      "epoch: 21 [140/187 (75%)]\t training loss: 0.511935\n",
      "epoch: 21 [150/187 (80%)]\t training loss: 0.165249\n",
      "epoch: 21 [160/187 (86%)]\t training loss: 0.605464\n",
      "epoch: 21 [170/187 (91%)]\t training loss: 0.327922\n",
      "epoch: 21 [180/187 (96%)]\t training loss: 0.663276\n",
      "\n",
      "\n",
      "Validation...\n",
      "\n",
      "Test dataset: Overall Loss: 0.2809, Overall Accuracy: 2759/2977 (93%)\n",
      "\n",
      "epoch: 22 [0/187 (0%)]\t training loss: 0.086143\n",
      "epoch: 22 [10/187 (5%)]\t training loss: 0.521506\n",
      "epoch: 22 [20/187 (11%)]\t training loss: 0.122452\n",
      "epoch: 22 [30/187 (16%)]\t training loss: 0.194031\n",
      "epoch: 22 [40/187 (21%)]\t training loss: 0.486500\n",
      "epoch: 22 [50/187 (27%)]\t training loss: 0.527894\n",
      "epoch: 22 [60/187 (32%)]\t training loss: 0.204567\n",
      "epoch: 22 [70/187 (37%)]\t training loss: 0.386327\n",
      "epoch: 22 [80/187 (43%)]\t training loss: 0.505748\n",
      "epoch: 22 [90/187 (48%)]\t training loss: 0.691287\n",
      "epoch: 22 [100/187 (53%)]\t training loss: 0.209497\n",
      "epoch: 22 [110/187 (59%)]\t training loss: 0.738191\n",
      "epoch: 22 [120/187 (64%)]\t training loss: 0.318455\n",
      "epoch: 22 [130/187 (70%)]\t training loss: 0.074932\n",
      "epoch: 22 [140/187 (75%)]\t training loss: 0.475472\n",
      "epoch: 22 [150/187 (80%)]\t training loss: 0.557611\n",
      "epoch: 22 [160/187 (86%)]\t training loss: 0.719509\n",
      "epoch: 22 [170/187 (91%)]\t training loss: 0.290724\n",
      "epoch: 22 [180/187 (96%)]\t training loss: 0.855532\n",
      "\n",
      "\n",
      "Validation...\n",
      "\n",
      "Test dataset: Overall Loss: 0.3477, Overall Accuracy: 2679/2977 (90%)\n",
      "\n",
      "epoch: 23 [0/187 (0%)]\t training loss: 0.659645\n",
      "epoch: 23 [10/187 (5%)]\t training loss: 0.477013\n",
      "epoch: 23 [20/187 (11%)]\t training loss: 0.223373\n",
      "epoch: 23 [30/187 (16%)]\t training loss: 0.176336\n",
      "epoch: 23 [40/187 (21%)]\t training loss: 0.607644\n",
      "epoch: 23 [50/187 (27%)]\t training loss: 0.566004\n",
      "epoch: 23 [60/187 (32%)]\t training loss: 0.114951\n",
      "epoch: 23 [70/187 (37%)]\t training loss: 0.811928\n",
      "epoch: 23 [80/187 (43%)]\t training loss: 0.881671\n",
      "epoch: 23 [90/187 (48%)]\t training loss: 0.978663\n",
      "epoch: 23 [100/187 (53%)]\t training loss: 0.675744\n",
      "epoch: 23 [110/187 (59%)]\t training loss: 0.208585\n",
      "epoch: 23 [120/187 (64%)]\t training loss: 0.481568\n",
      "epoch: 23 [130/187 (70%)]\t training loss: 0.622083\n",
      "epoch: 23 [140/187 (75%)]\t training loss: 0.156100\n",
      "epoch: 23 [150/187 (80%)]\t training loss: 0.500264\n",
      "epoch: 23 [160/187 (86%)]\t training loss: 0.609643\n",
      "epoch: 23 [170/187 (91%)]\t training loss: 0.574044\n",
      "epoch: 23 [180/187 (96%)]\t training loss: 0.296807\n",
      "\n",
      "\n",
      "Validation...\n",
      "\n",
      "Test dataset: Overall Loss: 0.2814, Overall Accuracy: 2758/2977 (93%)\n",
      "\n",
      "epoch: 24 [0/187 (0%)]\t training loss: 0.181320\n",
      "epoch: 24 [10/187 (5%)]\t training loss: 0.679779\n",
      "epoch: 24 [20/187 (11%)]\t training loss: 0.120532\n",
      "epoch: 24 [30/187 (16%)]\t training loss: 0.176686\n",
      "epoch: 24 [40/187 (21%)]\t training loss: 0.135514\n",
      "epoch: 24 [50/187 (27%)]\t training loss: 0.084549\n",
      "epoch: 24 [60/187 (32%)]\t training loss: 0.572348\n",
      "epoch: 24 [70/187 (37%)]\t training loss: 0.207907\n",
      "epoch: 24 [80/187 (43%)]\t training loss: 0.075514\n",
      "epoch: 24 [90/187 (48%)]\t training loss: 0.151510\n",
      "epoch: 24 [100/187 (53%)]\t training loss: 0.066195\n",
      "epoch: 24 [110/187 (59%)]\t training loss: 0.476214\n",
      "epoch: 24 [120/187 (64%)]\t training loss: 0.653278\n",
      "epoch: 24 [130/187 (70%)]\t training loss: 0.561139\n",
      "epoch: 24 [140/187 (75%)]\t training loss: 0.408280\n",
      "epoch: 24 [150/187 (80%)]\t training loss: 0.687243\n",
      "epoch: 24 [160/187 (86%)]\t training loss: 0.757088\n",
      "epoch: 24 [170/187 (91%)]\t training loss: 0.073727\n",
      "epoch: 24 [180/187 (96%)]\t training loss: 0.198678\n",
      "\n",
      "\n",
      "Validation...\n",
      "\n",
      "Test dataset: Overall Loss: 0.3245, Overall Accuracy: 2739/2977 (92%)\n",
      "\n",
      "epoch: 25 [0/187 (0%)]\t training loss: 0.401215\n",
      "epoch: 25 [10/187 (5%)]\t training loss: 0.447776\n",
      "epoch: 25 [20/187 (11%)]\t training loss: 0.032105\n",
      "epoch: 25 [30/187 (16%)]\t training loss: 0.045823\n",
      "epoch: 25 [40/187 (21%)]\t training loss: 0.151155\n",
      "epoch: 25 [50/187 (27%)]\t training loss: 0.773349\n",
      "epoch: 25 [60/187 (32%)]\t training loss: 0.521562\n",
      "epoch: 25 [70/187 (37%)]\t training loss: 0.573958\n",
      "epoch: 25 [80/187 (43%)]\t training loss: 0.880770\n",
      "epoch: 25 [90/187 (48%)]\t training loss: 0.707656\n",
      "epoch: 25 [100/187 (53%)]\t training loss: 0.345946\n",
      "epoch: 25 [110/187 (59%)]\t training loss: 0.807563\n",
      "epoch: 25 [120/187 (64%)]\t training loss: 0.842091\n",
      "epoch: 25 [130/187 (70%)]\t training loss: 0.519408\n",
      "epoch: 25 [140/187 (75%)]\t training loss: 0.157035\n",
      "epoch: 25 [150/187 (80%)]\t training loss: 0.117792\n",
      "epoch: 25 [160/187 (86%)]\t training loss: 0.699060\n",
      "epoch: 25 [170/187 (91%)]\t training loss: 0.689036\n",
      "epoch: 25 [180/187 (96%)]\t training loss: 0.137806\n",
      "\n",
      "\n",
      "Validation...\n",
      "\n",
      "Test dataset: Overall Loss: 0.2909, Overall Accuracy: 2745/2977 (92%)\n",
      "\n",
      "epoch: 26 [0/187 (0%)]\t training loss: 0.546298\n",
      "epoch: 26 [10/187 (5%)]\t training loss: 0.694956\n",
      "epoch: 26 [20/187 (11%)]\t training loss: 0.157700\n",
      "epoch: 26 [30/187 (16%)]\t training loss: 0.655351\n",
      "epoch: 26 [40/187 (21%)]\t training loss: 0.434383\n",
      "epoch: 26 [50/187 (27%)]\t training loss: 0.669413\n",
      "epoch: 26 [60/187 (32%)]\t training loss: 0.170817\n",
      "epoch: 26 [70/187 (37%)]\t training loss: 0.514641\n",
      "epoch: 26 [80/187 (43%)]\t training loss: 0.330609\n",
      "epoch: 26 [90/187 (48%)]\t training loss: 0.048890\n",
      "epoch: 26 [100/187 (53%)]\t training loss: 0.793016\n",
      "epoch: 26 [110/187 (59%)]\t training loss: 0.418143\n",
      "epoch: 26 [120/187 (64%)]\t training loss: 0.736457\n",
      "epoch: 26 [130/187 (70%)]\t training loss: 0.708404\n",
      "epoch: 26 [140/187 (75%)]\t training loss: 0.168527\n",
      "epoch: 26 [150/187 (80%)]\t training loss: 0.226869\n",
      "epoch: 26 [160/187 (86%)]\t training loss: 0.742099\n",
      "epoch: 26 [170/187 (91%)]\t training loss: 0.589284\n",
      "epoch: 26 [180/187 (96%)]\t training loss: 0.291252\n",
      "\n",
      "\n",
      "Validation...\n",
      "\n",
      "Test dataset: Overall Loss: 0.2787, Overall Accuracy: 2743/2977 (92%)\n",
      "\n",
      "epoch: 27 [0/187 (0%)]\t training loss: 0.690403\n",
      "epoch: 27 [10/187 (5%)]\t training loss: 0.768358\n",
      "epoch: 27 [20/187 (11%)]\t training loss: 0.571377\n",
      "epoch: 27 [30/187 (16%)]\t training loss: 0.541098\n",
      "epoch: 27 [40/187 (21%)]\t training loss: 0.352860\n",
      "epoch: 27 [50/187 (27%)]\t training loss: 0.254984\n",
      "epoch: 27 [60/187 (32%)]\t training loss: 0.758949\n",
      "epoch: 27 [70/187 (37%)]\t training loss: 0.071563\n",
      "epoch: 27 [80/187 (43%)]\t training loss: 0.280549\n",
      "epoch: 27 [90/187 (48%)]\t training loss: 0.122738\n",
      "epoch: 27 [100/187 (53%)]\t training loss: 0.662339\n",
      "epoch: 27 [110/187 (59%)]\t training loss: 0.411065\n",
      "epoch: 27 [120/187 (64%)]\t training loss: 0.730169\n",
      "epoch: 27 [130/187 (70%)]\t training loss: 0.615944\n",
      "epoch: 27 [140/187 (75%)]\t training loss: 0.488072\n",
      "epoch: 27 [150/187 (80%)]\t training loss: 0.044772\n",
      "epoch: 27 [160/187 (86%)]\t training loss: 0.674974\n",
      "epoch: 27 [170/187 (91%)]\t training loss: 0.315112\n",
      "epoch: 27 [180/187 (96%)]\t training loss: 0.031551\n",
      "\n",
      "\n",
      "Validation...\n",
      "\n",
      "Test dataset: Overall Loss: 0.2736, Overall Accuracy: 2758/2977 (93%)\n",
      "\n",
      "epoch: 28 [0/187 (0%)]\t training loss: 0.831031\n",
      "epoch: 28 [10/187 (5%)]\t training loss: 0.643317\n",
      "epoch: 28 [20/187 (11%)]\t training loss: 0.779038\n",
      "epoch: 28 [30/187 (16%)]\t training loss: 0.318160\n",
      "epoch: 28 [40/187 (21%)]\t training loss: 0.254963\n",
      "epoch: 28 [50/187 (27%)]\t training loss: 0.097019\n",
      "epoch: 28 [60/187 (32%)]\t training loss: 0.362318\n",
      "epoch: 28 [70/187 (37%)]\t training loss: 0.162654\n",
      "epoch: 28 [80/187 (43%)]\t training loss: 0.591095\n",
      "epoch: 28 [90/187 (48%)]\t training loss: 0.088192\n",
      "epoch: 28 [100/187 (53%)]\t training loss: 0.213284\n",
      "epoch: 28 [110/187 (59%)]\t training loss: 0.833096\n",
      "epoch: 28 [120/187 (64%)]\t training loss: 0.064765\n",
      "epoch: 28 [130/187 (70%)]\t training loss: 0.325978\n",
      "epoch: 28 [140/187 (75%)]\t training loss: 0.082127\n",
      "epoch: 28 [150/187 (80%)]\t training loss: 0.473187\n",
      "epoch: 28 [160/187 (86%)]\t training loss: 0.447381\n",
      "epoch: 28 [170/187 (91%)]\t training loss: 0.486681\n",
      "epoch: 28 [180/187 (96%)]\t training loss: 0.183351\n",
      "\n",
      "\n",
      "Validation...\n",
      "\n",
      "Test dataset: Overall Loss: 0.2890, Overall Accuracy: 2752/2977 (92%)\n",
      "\n",
      "epoch: 29 [0/187 (0%)]\t training loss: 0.865950\n",
      "epoch: 29 [10/187 (5%)]\t training loss: 0.093139\n",
      "epoch: 29 [20/187 (11%)]\t training loss: 0.044191\n",
      "epoch: 29 [30/187 (16%)]\t training loss: 0.770071\n",
      "epoch: 29 [40/187 (21%)]\t training loss: 0.081818\n",
      "epoch: 29 [50/187 (27%)]\t training loss: 0.222714\n",
      "epoch: 29 [60/187 (32%)]\t training loss: 0.287228\n",
      "epoch: 29 [70/187 (37%)]\t training loss: 0.307072\n",
      "epoch: 29 [80/187 (43%)]\t training loss: 0.061060\n",
      "epoch: 29 [90/187 (48%)]\t training loss: 0.720312\n",
      "epoch: 29 [100/187 (53%)]\t training loss: 0.065611\n",
      "epoch: 29 [110/187 (59%)]\t training loss: 0.496687\n",
      "epoch: 29 [120/187 (64%)]\t training loss: 0.788864\n",
      "epoch: 29 [130/187 (70%)]\t training loss: 0.659549\n",
      "epoch: 29 [140/187 (75%)]\t training loss: 0.184351\n",
      "epoch: 29 [150/187 (80%)]\t training loss: 0.250837\n",
      "epoch: 29 [160/187 (86%)]\t training loss: 0.079200\n",
      "epoch: 29 [170/187 (91%)]\t training loss: 0.685464\n",
      "epoch: 29 [180/187 (96%)]\t training loss: 0.109112\n",
      "\n",
      "\n",
      "Validation...\n",
      "\n",
      "Test dataset: Overall Loss: 0.3133, Overall Accuracy: 2741/2977 (92%)\n",
      "\n"
     ]
    }
   ],
   "source": [
    "imbalanced_model_cutmix_mixup, train_losses, val_losses = train_mixup(\n",
    "    imbalanced_train_loader,\n",
    "    imbalanced_valid_loader,\n",
    "    \"imbalanced_model_cutmix_mixup.pth\",\n",
    "    ConvNet().to(device),\n",
    "    epochs=30,\n",
    ")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 27,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 472
    },
    "id": "JGiNYuiM0nE5",
    "outputId": "3a78966a-740d-4d78-fb54-a062e520a49d"
   },
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 640x480 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "plt.plot(train_losses)\n",
    "plt.plot(val_losses)\n",
    "plt.xlabel(\"epoch\")\n",
    "plt.ylabel(\"loss\")\n",
    "plt.legend([\"Train\", \"Valid\"])\n",
    "plt.title(\"Model On imbalanced Fashion MNIST - Train vs Valid Accuracy\")\n",
    "\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 28,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/"
    },
    "id": "UdCjp3NVruen",
    "outputId": "0c4503e5-1ef7-470b-f427-f225fbba235b"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "On Imbalanced Fashion MNIST dataset\n",
      "\n",
      "Test dataset: Overall Loss: 0.0015, Overall Accuracy: 8140/10000 (81%)\n",
      "\n",
      "Cutmix+Mixup test accuracy :  81.4\n"
     ]
    }
   ],
   "source": [
    "print(\"On Imbalanced Fashion MNIST dataset\")\n",
    "test_acc4, class_percentages4 = test(\n",
    "    imbalanced_model_cutmix_mixup, \"imbalanced_model_cutmix_mixup.pth\"\n",
    ")\n",
    "print(\"Cutmix+Mixup test accuracy : \", str(test_acc4))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "aZkoVwE7uAys"
   },
   "source": [
    "# ReMix"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 29,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/"
    },
    "id": "0_fNOQGRvq4w",
    "outputId": "d11105bf-ff72-4a66-c71b-22ab112c34c8"
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "Counter({0: 6000,\n",
       "         3: 1292,\n",
       "         9: 60,\n",
       "         1: 3596,\n",
       "         5: 464,\n",
       "         6: 278,\n",
       "         2: 2156,\n",
       "         4: 774,\n",
       "         7: 166,\n",
       "         8: 100})"
      ]
     },
     "execution_count": 29,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "Counter(imbalanced_train_dataset.targets.numpy())"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 30,
   "metadata": {
    "id": "uL2Bz_QtFfha"
   },
   "outputs": [],
   "source": [
    "def remix_data(inputs, labels, class_counts, alpha=1.0):\n",
    "    \"\"\"\n",
    "    Returns mixed inputs, pairs of targets, and lambda_x, lambda_y\n",
    "    \"\"\"\n",
    "    if alpha > 0:\n",
    "        lambda_x = np.random.beta(alpha, alpha)\n",
    "    else:\n",
    "        lambda_x = 1\n",
    "\n",
    "    K = 3  # hyperparameter for majority class\n",
    "    tau = 0.5  # hyperparameter\n",
    "\n",
    "    class_counts_tensor = torch.tensor(class_counts)\n",
    "    batch_size = inputs.size()[0]\n",
    "    random_indices = torch.randperm(batch_size)\n",
    "\n",
    "    class_count_pairs = [\n",
    "        [class_counts_tensor[labels[i]].item(), class_counts_tensor[labels[j]].item()]\n",
    "        for i, j in enumerate(random_indices)\n",
    "    ]\n",
    "    class_count_pairs = torch.tensor(class_count_pairs)\n",
    "\n",
    "    lambda_y_values = []\n",
    "    for pair in class_count_pairs:\n",
    "        class_count_ratio = pair[0] / pair[1]\n",
    "\n",
    "        if class_count_ratio >= K and lambda_x < tau:\n",
    "            lambda_y_values.append(0)\n",
    "        elif class_count_ratio <= (1 / K) and (1 - lambda_x) < tau:\n",
    "            lambda_y_values.append(1)\n",
    "        else:\n",
    "            lambda_y_values.append(lambda_x)\n",
    "\n",
    "    lambda_y = torch.tensor(lambda_y_values).cuda(0)\n",
    "    mixed_inputs = lambda_x * inputs + (1 - lambda_x) * inputs[random_indices, :]\n",
    "    labels_a, labels_b = labels, labels[random_indices]\n",
    "\n",
    "    return mixed_inputs, [labels_a, labels_b, lambda_y]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 31,
   "metadata": {
    "id": "br2LxHP3uC0u"
   },
   "outputs": [],
   "source": [
    "def remix_criterion(criterion, pred, y_a, y_b, lam_y):\n",
    "    \"\"\"\n",
    "    In Remix, the lambda for mixing label is different from original mixup.\n",
    "    \"\"\"\n",
    "    # for each y, we calculated its loss individually with their respective\n",
    "    # lambda_y\n",
    "    loss = torch.mul(criterion(pred, y_a), lam_y) + torch.mul(\n",
    "        criterion(pred, y_b), (1 - lam_y)\n",
    "    )\n",
    "    return loss.mean()\n",
    "\n",
    "\n",
    "def train_remix(\n",
    "    cls_num_list,\n",
    "    train_dataloader,\n",
    "    valid_dataloader,\n",
    "    PATH,\n",
    "    model=None,\n",
    "    criterion=F.cross_entropy,\n",
    "    epochs=1,\n",
    "):\n",
    "    optimizer = torch.optim.SGD(model.parameters(), lr=0.1, momentum=0.9)\n",
    "    model.train()\n",
    "    best_acc = 0\n",
    "    train_losses = []\n",
    "    val_losses = []\n",
    "    for epoch in range(epochs):\n",
    "        losses = 0\n",
    "        for b_i, (X, y) in enumerate(train_dataloader):\n",
    "            X, y = X.to(device), y.to(device)\n",
    "            X, ys = remix_data(X, y, cls_num_list, alpha=0.5)\n",
    "            pred_prob = model(X)\n",
    "            targets1, targets2, lam = ys[0], ys[1], ys[2]\n",
    "            loss = remix_criterion(criterion, pred_prob, targets1, targets2, lam)\n",
    "            optimizer.zero_grad()\n",
    "            loss.backward()\n",
    "            optimizer.step()\n",
    "            losses += loss.item()  # print(losses)\n",
    "            if b_i % 10 == 0:\n",
    "                print(\n",
    "                    \"epoch: {} [{}/{} ({:.0f}%)]\\t training loss: {:.6f}\".format(\n",
    "                        epoch,\n",
    "                        b_i,\n",
    "                        len(train_dataloader),\n",
    "                        100.0 * b_i / len(train_dataloader),\n",
    "                        loss.item(),\n",
    "                    )\n",
    "                )\n",
    "        train_losses.append(losses / len(train_dataloader))\n",
    "        print(\"\\n\")\n",
    "        print(\"Validation...\")\n",
    "        model.eval()\n",
    "        success = 0\n",
    "        total = 0\n",
    "        losses = 0\n",
    "        with torch.no_grad():\n",
    "            for X, y in valid_dataloader:\n",
    "                X, y = X.to(device), y.to(device)\n",
    "                pred_prob = model(X)\n",
    "                loss = criterion(pred_prob, y)  # loss summed across the batch\n",
    "                losses += loss.item()\n",
    "                pred = pred_prob.argmax(\n",
    "                    dim=1, keepdim=True\n",
    "                )  # use argmax to get the most likely prediction\n",
    "                success += pred.eq(y.view_as(pred)).sum().item()\n",
    "                total += len(y)\n",
    "        val_losses.append(losses / len(valid_dataloader))\n",
    "        accuracy = 100.0 * success / total\n",
    "\n",
    "        if accuracy > best_acc:\n",
    "            torch.save(model.state_dict(), PATH)\n",
    "            best_acc = accuracy\n",
    "\n",
    "        print(\n",
    "            \"\\nTest dataset: Overall Loss: {:.4f}, Overall Accuracy: {}/{} ({:.0f}%)\\n\".format(\n",
    "                losses / len(valid_dataloader), success, total, accuracy\n",
    "            )\n",
    "        )\n",
    "    return model, train_losses, val_losses"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 32,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/"
    },
    "id": "J31cxJatuC0u",
    "outputId": "5a971010-73f4-4531-bb98-63ef0cdca1ab"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[6000, 3596, 2156, 1292, 774, 464, 278, 166, 100, 60]\n"
     ]
    }
   ],
   "source": [
    "per_class_count_dict = dict(Counter(imbalanced_train_dataset.targets.numpy()))\n",
    "cls_num_list = []\n",
    "for i in range(10):\n",
    "    cls_num_list.append(per_class_count_dict[i])\n",
    "print(cls_num_list)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 33,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/"
    },
    "id": "j-fs7Nkxx-pe",
    "outputId": "c7b3997a-39c4-474b-b8dc-7e14b6addf6c"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "epoch: 0 [0/187 (0%)]\t training loss: 2.339584\n",
      "epoch: 0 [10/187 (5%)]\t training loss: 1.239786\n",
      "epoch: 0 [20/187 (11%)]\t training loss: 1.204120\n",
      "epoch: 0 [30/187 (16%)]\t training loss: 0.894693\n",
      "epoch: 0 [40/187 (21%)]\t training loss: 1.224142\n",
      "epoch: 0 [50/187 (27%)]\t training loss: 0.820298\n",
      "epoch: 0 [60/187 (32%)]\t training loss: 1.001780\n",
      "epoch: 0 [70/187 (37%)]\t training loss: 1.237055\n",
      "epoch: 0 [80/187 (43%)]\t training loss: 0.533003\n",
      "epoch: 0 [90/187 (48%)]\t training loss: 1.195767\n",
      "epoch: 0 [100/187 (53%)]\t training loss: 0.983461\n",
      "epoch: 0 [110/187 (59%)]\t training loss: 0.781501\n",
      "epoch: 0 [120/187 (64%)]\t training loss: 0.761184\n",
      "epoch: 0 [130/187 (70%)]\t training loss: 0.345647\n",
      "epoch: 0 [140/187 (75%)]\t training loss: 0.913522\n",
      "epoch: 0 [150/187 (80%)]\t training loss: 1.177918\n",
      "epoch: 0 [160/187 (86%)]\t training loss: 0.941176\n",
      "epoch: 0 [170/187 (91%)]\t training loss: 0.509522\n",
      "epoch: 0 [180/187 (96%)]\t training loss: 0.733694\n",
      "\n",
      "\n",
      "Validation...\n",
      "\n",
      "Test dataset: Overall Loss: 0.4411, Overall Accuracy: 2581/2977 (87%)\n",
      "\n",
      "epoch: 1 [0/187 (0%)]\t training loss: 0.650541\n",
      "epoch: 1 [10/187 (5%)]\t training loss: 1.254246\n",
      "epoch: 1 [20/187 (11%)]\t training loss: 0.803608\n",
      "epoch: 1 [30/187 (16%)]\t training loss: 0.815134\n",
      "epoch: 1 [40/187 (21%)]\t training loss: 0.546772\n",
      "epoch: 1 [50/187 (27%)]\t training loss: 0.859642\n",
      "epoch: 1 [60/187 (32%)]\t training loss: 0.387361\n",
      "epoch: 1 [70/187 (37%)]\t training loss: 0.494616\n",
      "epoch: 1 [80/187 (43%)]\t training loss: 0.720086\n",
      "epoch: 1 [90/187 (48%)]\t training loss: 0.870717\n",
      "epoch: 1 [100/187 (53%)]\t training loss: 0.592486\n",
      "epoch: 1 [110/187 (59%)]\t training loss: 0.619347\n",
      "epoch: 1 [120/187 (64%)]\t training loss: 1.054021\n",
      "epoch: 1 [130/187 (70%)]\t training loss: 0.279551\n",
      "epoch: 1 [140/187 (75%)]\t training loss: 0.480190\n",
      "epoch: 1 [150/187 (80%)]\t training loss: 0.355117\n",
      "epoch: 1 [160/187 (86%)]\t training loss: 0.267735\n",
      "epoch: 1 [170/187 (91%)]\t training loss: 0.913042\n",
      "epoch: 1 [180/187 (96%)]\t training loss: 0.545089\n",
      "\n",
      "\n",
      "Validation...\n",
      "\n",
      "Test dataset: Overall Loss: 0.3617, Overall Accuracy: 2644/2977 (89%)\n",
      "\n",
      "epoch: 2 [0/187 (0%)]\t training loss: 0.881901\n",
      "epoch: 2 [10/187 (5%)]\t training loss: 0.838548\n",
      "epoch: 2 [20/187 (11%)]\t training loss: 1.040732\n",
      "epoch: 2 [30/187 (16%)]\t training loss: 0.695572\n",
      "epoch: 2 [40/187 (21%)]\t training loss: 0.442062\n",
      "epoch: 2 [50/187 (27%)]\t training loss: 0.729016\n",
      "epoch: 2 [60/187 (32%)]\t training loss: 0.594718\n",
      "epoch: 2 [70/187 (37%)]\t training loss: 0.691980\n",
      "epoch: 2 [80/187 (43%)]\t training loss: 0.766473\n",
      "epoch: 2 [90/187 (48%)]\t training loss: 0.726079\n",
      "epoch: 2 [100/187 (53%)]\t training loss: 0.664947\n",
      "epoch: 2 [110/187 (59%)]\t training loss: 0.636159\n",
      "epoch: 2 [120/187 (64%)]\t training loss: 0.222069\n",
      "epoch: 2 [130/187 (70%)]\t training loss: 0.791597\n",
      "epoch: 2 [140/187 (75%)]\t training loss: 0.772394\n",
      "epoch: 2 [150/187 (80%)]\t training loss: 1.226948\n",
      "epoch: 2 [160/187 (86%)]\t training loss: 0.815957\n",
      "epoch: 2 [170/187 (91%)]\t training loss: 0.359633\n",
      "epoch: 2 [180/187 (96%)]\t training loss: 0.187534\n",
      "\n",
      "\n",
      "Validation...\n",
      "\n",
      "Test dataset: Overall Loss: 0.3260, Overall Accuracy: 2670/2977 (90%)\n",
      "\n",
      "epoch: 3 [0/187 (0%)]\t training loss: 0.991172\n",
      "epoch: 3 [10/187 (5%)]\t training loss: 0.727813\n",
      "epoch: 3 [20/187 (11%)]\t training loss: 0.334467\n",
      "epoch: 3 [30/187 (16%)]\t training loss: 0.401294\n",
      "epoch: 3 [40/187 (21%)]\t training loss: 0.939594\n",
      "epoch: 3 [50/187 (27%)]\t training loss: 0.622004\n",
      "epoch: 3 [60/187 (32%)]\t training loss: 0.305576\n",
      "epoch: 3 [70/187 (37%)]\t training loss: 0.515187\n",
      "epoch: 3 [80/187 (43%)]\t training loss: 0.943521\n",
      "epoch: 3 [90/187 (48%)]\t training loss: 0.418676\n",
      "epoch: 3 [100/187 (53%)]\t training loss: 0.645377\n",
      "epoch: 3 [110/187 (59%)]\t training loss: 0.695163\n",
      "epoch: 3 [120/187 (64%)]\t training loss: 0.959064\n",
      "epoch: 3 [130/187 (70%)]\t training loss: 0.683122\n",
      "epoch: 3 [140/187 (75%)]\t training loss: 0.713005\n",
      "epoch: 3 [150/187 (80%)]\t training loss: 0.448596\n",
      "epoch: 3 [160/187 (86%)]\t training loss: 1.059243\n",
      "epoch: 3 [170/187 (91%)]\t training loss: 0.846263\n",
      "epoch: 3 [180/187 (96%)]\t training loss: 0.776117\n",
      "\n",
      "\n",
      "Validation...\n",
      "\n",
      "Test dataset: Overall Loss: 0.3119, Overall Accuracy: 2672/2977 (90%)\n",
      "\n",
      "epoch: 4 [0/187 (0%)]\t training loss: 0.362653\n",
      "epoch: 4 [10/187 (5%)]\t training loss: 0.913532\n",
      "epoch: 4 [20/187 (11%)]\t training loss: 0.581619\n",
      "epoch: 4 [30/187 (16%)]\t training loss: 0.198585\n",
      "epoch: 4 [40/187 (21%)]\t training loss: 0.493365\n",
      "epoch: 4 [50/187 (27%)]\t training loss: 0.822481\n",
      "epoch: 4 [60/187 (32%)]\t training loss: 0.923235\n",
      "epoch: 4 [70/187 (37%)]\t training loss: 0.382399\n",
      "epoch: 4 [80/187 (43%)]\t training loss: 0.876453\n",
      "epoch: 4 [90/187 (48%)]\t training loss: 0.356305\n",
      "epoch: 4 [100/187 (53%)]\t training loss: 0.546830\n",
      "epoch: 4 [110/187 (59%)]\t training loss: 0.329194\n",
      "epoch: 4 [120/187 (64%)]\t training loss: 0.698442\n",
      "epoch: 4 [130/187 (70%)]\t training loss: 0.534394\n",
      "epoch: 4 [140/187 (75%)]\t training loss: 0.471489\n",
      "epoch: 4 [150/187 (80%)]\t training loss: 0.545595\n",
      "epoch: 4 [160/187 (86%)]\t training loss: 0.411221\n",
      "epoch: 4 [170/187 (91%)]\t training loss: 0.526472\n",
      "epoch: 4 [180/187 (96%)]\t training loss: 0.297336\n",
      "\n",
      "\n",
      "Validation...\n",
      "\n",
      "Test dataset: Overall Loss: 0.3054, Overall Accuracy: 2703/2977 (91%)\n",
      "\n",
      "epoch: 5 [0/187 (0%)]\t training loss: 0.184704\n",
      "epoch: 5 [10/187 (5%)]\t training loss: 0.752083\n",
      "epoch: 5 [20/187 (11%)]\t training loss: 1.043416\n",
      "epoch: 5 [30/187 (16%)]\t training loss: 0.851634\n",
      "epoch: 5 [40/187 (21%)]\t training loss: 0.763817\n",
      "epoch: 5 [50/187 (27%)]\t training loss: 0.708557\n",
      "epoch: 5 [60/187 (32%)]\t training loss: 0.829623\n",
      "epoch: 5 [70/187 (37%)]\t training loss: 0.475537\n",
      "epoch: 5 [80/187 (43%)]\t training loss: 0.643957\n",
      "epoch: 5 [90/187 (48%)]\t training loss: 0.645195\n",
      "epoch: 5 [100/187 (53%)]\t training loss: 0.844229\n",
      "epoch: 5 [110/187 (59%)]\t training loss: 0.753141\n",
      "epoch: 5 [120/187 (64%)]\t training loss: 0.239172\n",
      "epoch: 5 [130/187 (70%)]\t training loss: 0.813147\n",
      "epoch: 5 [140/187 (75%)]\t training loss: 0.174330\n",
      "epoch: 5 [150/187 (80%)]\t training loss: 0.766358\n",
      "epoch: 5 [160/187 (86%)]\t training loss: 0.663649\n",
      "epoch: 5 [170/187 (91%)]\t training loss: 0.694275\n",
      "epoch: 5 [180/187 (96%)]\t training loss: 0.303704\n",
      "\n",
      "\n",
      "Validation...\n",
      "\n",
      "Test dataset: Overall Loss: 0.3094, Overall Accuracy: 2681/2977 (90%)\n",
      "\n",
      "epoch: 6 [0/187 (0%)]\t training loss: 0.863669\n",
      "epoch: 6 [10/187 (5%)]\t training loss: 0.812092\n",
      "epoch: 6 [20/187 (11%)]\t training loss: 0.261500\n",
      "epoch: 6 [30/187 (16%)]\t training loss: 0.660960\n",
      "epoch: 6 [40/187 (21%)]\t training loss: 0.425300\n",
      "epoch: 6 [50/187 (27%)]\t training loss: 0.689684\n",
      "epoch: 6 [60/187 (32%)]\t training loss: 0.193300\n",
      "epoch: 6 [70/187 (37%)]\t training loss: 0.914577\n",
      "epoch: 6 [80/187 (43%)]\t training loss: 0.343079\n",
      "epoch: 6 [90/187 (48%)]\t training loss: 0.671609\n",
      "epoch: 6 [100/187 (53%)]\t training loss: 0.786847\n",
      "epoch: 6 [110/187 (59%)]\t training loss: 0.741065\n",
      "epoch: 6 [120/187 (64%)]\t training loss: 0.340439\n",
      "epoch: 6 [130/187 (70%)]\t training loss: 0.239241\n",
      "epoch: 6 [140/187 (75%)]\t training loss: 0.200414\n",
      "epoch: 6 [150/187 (80%)]\t training loss: 0.180277\n",
      "epoch: 6 [160/187 (86%)]\t training loss: 0.775784\n",
      "epoch: 6 [170/187 (91%)]\t training loss: 0.196298\n",
      "epoch: 6 [180/187 (96%)]\t training loss: 0.233013\n",
      "\n",
      "\n",
      "Validation...\n",
      "\n",
      "Test dataset: Overall Loss: 0.3149, Overall Accuracy: 2696/2977 (91%)\n",
      "\n",
      "epoch: 7 [0/187 (0%)]\t training loss: 0.263287\n",
      "epoch: 7 [10/187 (5%)]\t training loss: 0.939109\n",
      "epoch: 7 [20/187 (11%)]\t training loss: 0.731385\n",
      "epoch: 7 [30/187 (16%)]\t training loss: 0.433137\n",
      "epoch: 7 [40/187 (21%)]\t training loss: 0.326944\n",
      "epoch: 7 [50/187 (27%)]\t training loss: 0.716409\n",
      "epoch: 7 [60/187 (32%)]\t training loss: 0.919007\n",
      "epoch: 7 [70/187 (37%)]\t training loss: 0.800593\n",
      "epoch: 7 [80/187 (43%)]\t training loss: 1.017449\n",
      "epoch: 7 [90/187 (48%)]\t training loss: 0.960799\n",
      "epoch: 7 [100/187 (53%)]\t training loss: 0.275940\n",
      "epoch: 7 [110/187 (59%)]\t training loss: 0.366010\n",
      "epoch: 7 [120/187 (64%)]\t training loss: 0.629633\n",
      "epoch: 7 [130/187 (70%)]\t training loss: 0.257393\n",
      "epoch: 7 [140/187 (75%)]\t training loss: 0.158820\n",
      "epoch: 7 [150/187 (80%)]\t training loss: 0.433097\n",
      "epoch: 7 [160/187 (86%)]\t training loss: 0.329964\n",
      "epoch: 7 [170/187 (91%)]\t training loss: 0.602694\n",
      "epoch: 7 [180/187 (96%)]\t training loss: 0.555970\n",
      "\n",
      "\n",
      "Validation...\n",
      "\n",
      "Test dataset: Overall Loss: 0.2947, Overall Accuracy: 2707/2977 (91%)\n",
      "\n",
      "epoch: 8 [0/187 (0%)]\t training loss: 0.750132\n",
      "epoch: 8 [10/187 (5%)]\t training loss: 0.533023\n",
      "epoch: 8 [20/187 (11%)]\t training loss: 0.544020\n",
      "epoch: 8 [30/187 (16%)]\t training loss: 0.426657\n",
      "epoch: 8 [40/187 (21%)]\t training loss: 0.565644\n",
      "epoch: 8 [50/187 (27%)]\t training loss: 0.707166\n",
      "epoch: 8 [60/187 (32%)]\t training loss: 0.503132\n",
      "epoch: 8 [70/187 (37%)]\t training loss: 0.285145\n",
      "epoch: 8 [80/187 (43%)]\t training loss: 0.595194\n",
      "epoch: 8 [90/187 (48%)]\t training loss: 0.898779\n",
      "epoch: 8 [100/187 (53%)]\t training loss: 0.408928\n",
      "epoch: 8 [110/187 (59%)]\t training loss: 0.340003\n",
      "epoch: 8 [120/187 (64%)]\t training loss: 0.314640\n",
      "epoch: 8 [130/187 (70%)]\t training loss: 0.273304\n",
      "epoch: 8 [140/187 (75%)]\t training loss: 0.456704\n",
      "epoch: 8 [150/187 (80%)]\t training loss: 0.353258\n",
      "epoch: 8 [160/187 (86%)]\t training loss: 0.344316\n",
      "epoch: 8 [170/187 (91%)]\t training loss: 0.782939\n",
      "epoch: 8 [180/187 (96%)]\t training loss: 0.257615\n",
      "\n",
      "\n",
      "Validation...\n",
      "\n",
      "Test dataset: Overall Loss: 0.3136, Overall Accuracy: 2693/2977 (90%)\n",
      "\n",
      "epoch: 9 [0/187 (0%)]\t training loss: 0.289690\n",
      "epoch: 9 [10/187 (5%)]\t training loss: 0.363361\n",
      "epoch: 9 [20/187 (11%)]\t training loss: 0.526096\n",
      "epoch: 9 [30/187 (16%)]\t training loss: 0.446748\n",
      "epoch: 9 [40/187 (21%)]\t training loss: 0.678163\n",
      "epoch: 9 [50/187 (27%)]\t training loss: 0.675438\n",
      "epoch: 9 [60/187 (32%)]\t training loss: 0.547396\n",
      "epoch: 9 [70/187 (37%)]\t training loss: 0.495280\n",
      "epoch: 9 [80/187 (43%)]\t training loss: 0.674420\n",
      "epoch: 9 [90/187 (48%)]\t training loss: 0.169416\n",
      "epoch: 9 [100/187 (53%)]\t training loss: 0.955119\n",
      "epoch: 9 [110/187 (59%)]\t training loss: 0.354532\n",
      "epoch: 9 [120/187 (64%)]\t training loss: 0.631243\n",
      "epoch: 9 [130/187 (70%)]\t training loss: 0.666326\n",
      "epoch: 9 [140/187 (75%)]\t training loss: 0.876424\n",
      "epoch: 9 [150/187 (80%)]\t training loss: 0.236966\n",
      "epoch: 9 [160/187 (86%)]\t training loss: 0.361342\n",
      "epoch: 9 [170/187 (91%)]\t training loss: 0.415749\n",
      "epoch: 9 [180/187 (96%)]\t training loss: 0.667874\n",
      "\n",
      "\n",
      "Validation...\n",
      "\n",
      "Test dataset: Overall Loss: 0.2937, Overall Accuracy: 2699/2977 (91%)\n",
      "\n",
      "epoch: 10 [0/187 (0%)]\t training loss: 0.564938\n",
      "epoch: 10 [10/187 (5%)]\t training loss: 1.071121\n",
      "epoch: 10 [20/187 (11%)]\t training loss: 1.043847\n",
      "epoch: 10 [30/187 (16%)]\t training loss: 0.320749\n",
      "epoch: 10 [40/187 (21%)]\t training loss: 0.195695\n",
      "epoch: 10 [50/187 (27%)]\t training loss: 0.974965\n",
      "epoch: 10 [60/187 (32%)]\t training loss: 0.821501\n",
      "epoch: 10 [70/187 (37%)]\t training loss: 0.865578\n",
      "epoch: 10 [80/187 (43%)]\t training loss: 0.501282\n",
      "epoch: 10 [90/187 (48%)]\t training loss: 0.233842\n",
      "epoch: 10 [100/187 (53%)]\t training loss: 0.626878\n",
      "epoch: 10 [110/187 (59%)]\t training loss: 0.170604\n",
      "epoch: 10 [120/187 (64%)]\t training loss: 0.617839\n",
      "epoch: 10 [130/187 (70%)]\t training loss: 0.234322\n",
      "epoch: 10 [140/187 (75%)]\t training loss: 0.735738\n",
      "epoch: 10 [150/187 (80%)]\t training loss: 0.683382\n",
      "epoch: 10 [160/187 (86%)]\t training loss: 0.503013\n",
      "epoch: 10 [170/187 (91%)]\t training loss: 0.783522\n",
      "epoch: 10 [180/187 (96%)]\t training loss: 0.409285\n",
      "\n",
      "\n",
      "Validation...\n",
      "\n",
      "Test dataset: Overall Loss: 0.2975, Overall Accuracy: 2720/2977 (91%)\n",
      "\n",
      "epoch: 11 [0/187 (0%)]\t training loss: 0.872082\n",
      "epoch: 11 [10/187 (5%)]\t training loss: 0.709880\n",
      "epoch: 11 [20/187 (11%)]\t training loss: 0.240482\n",
      "epoch: 11 [30/187 (16%)]\t training loss: 0.635432\n",
      "epoch: 11 [40/187 (21%)]\t training loss: 0.801283\n",
      "epoch: 11 [50/187 (27%)]\t training loss: 0.147059\n",
      "epoch: 11 [60/187 (32%)]\t training loss: 0.457320\n",
      "epoch: 11 [70/187 (37%)]\t training loss: 0.700774\n",
      "epoch: 11 [80/187 (43%)]\t training loss: 1.056306\n",
      "epoch: 11 [90/187 (48%)]\t training loss: 0.623792\n",
      "epoch: 11 [100/187 (53%)]\t training loss: 0.916952\n",
      "epoch: 11 [110/187 (59%)]\t training loss: 0.604118\n",
      "epoch: 11 [120/187 (64%)]\t training loss: 0.596982\n",
      "epoch: 11 [130/187 (70%)]\t training loss: 0.814748\n",
      "epoch: 11 [140/187 (75%)]\t training loss: 0.676414\n",
      "epoch: 11 [150/187 (80%)]\t training loss: 0.574594\n",
      "epoch: 11 [160/187 (86%)]\t training loss: 1.042529\n",
      "epoch: 11 [170/187 (91%)]\t training loss: 0.642904\n",
      "epoch: 11 [180/187 (96%)]\t training loss: 0.609074\n",
      "\n",
      "\n",
      "Validation...\n",
      "\n",
      "Test dataset: Overall Loss: 0.3349, Overall Accuracy: 2688/2977 (90%)\n",
      "\n",
      "epoch: 12 [0/187 (0%)]\t training loss: 0.898358\n",
      "epoch: 12 [10/187 (5%)]\t training loss: 0.251159\n",
      "epoch: 12 [20/187 (11%)]\t training loss: 0.178837\n",
      "epoch: 12 [30/187 (16%)]\t training loss: 0.719209\n",
      "epoch: 12 [40/187 (21%)]\t training loss: 0.829093\n",
      "epoch: 12 [50/187 (27%)]\t training loss: 0.442095\n",
      "epoch: 12 [60/187 (32%)]\t training loss: 0.380274\n",
      "epoch: 12 [70/187 (37%)]\t training loss: 0.334303\n",
      "epoch: 12 [80/187 (43%)]\t training loss: 0.574002\n",
      "epoch: 12 [90/187 (48%)]\t training loss: 0.249468\n",
      "epoch: 12 [100/187 (53%)]\t training loss: 0.738262\n",
      "epoch: 12 [110/187 (59%)]\t training loss: 0.692563\n",
      "epoch: 12 [120/187 (64%)]\t training loss: 0.109549\n",
      "epoch: 12 [130/187 (70%)]\t training loss: 0.646089\n",
      "epoch: 12 [140/187 (75%)]\t training loss: 0.354426\n",
      "epoch: 12 [150/187 (80%)]\t training loss: 0.464051\n",
      "epoch: 12 [160/187 (86%)]\t training loss: 0.462387\n",
      "epoch: 12 [170/187 (91%)]\t training loss: 0.298461\n",
      "epoch: 12 [180/187 (96%)]\t training loss: 0.422158\n",
      "\n",
      "\n",
      "Validation...\n",
      "\n",
      "Test dataset: Overall Loss: 0.2751, Overall Accuracy: 2734/2977 (92%)\n",
      "\n",
      "epoch: 13 [0/187 (0%)]\t training loss: 0.796965\n",
      "epoch: 13 [10/187 (5%)]\t training loss: 0.173016\n",
      "epoch: 13 [20/187 (11%)]\t training loss: 0.537592\n",
      "epoch: 13 [30/187 (16%)]\t training loss: 0.500258\n",
      "epoch: 13 [40/187 (21%)]\t training loss: 0.365465\n",
      "epoch: 13 [50/187 (27%)]\t training loss: 0.105772\n",
      "epoch: 13 [60/187 (32%)]\t training loss: 0.435315\n",
      "epoch: 13 [70/187 (37%)]\t training loss: 0.284663\n",
      "epoch: 13 [80/187 (43%)]\t training loss: 0.471735\n",
      "epoch: 13 [90/187 (48%)]\t training loss: 0.726521\n",
      "epoch: 13 [100/187 (53%)]\t training loss: 0.117728\n",
      "epoch: 13 [110/187 (59%)]\t training loss: 0.452228\n",
      "epoch: 13 [120/187 (64%)]\t training loss: 0.137295\n",
      "epoch: 13 [130/187 (70%)]\t training loss: 0.637752\n",
      "epoch: 13 [140/187 (75%)]\t training loss: 0.691926\n",
      "epoch: 13 [150/187 (80%)]\t training loss: 0.418541\n",
      "epoch: 13 [160/187 (86%)]\t training loss: 0.889985\n",
      "epoch: 13 [170/187 (91%)]\t training loss: 0.508335\n",
      "epoch: 13 [180/187 (96%)]\t training loss: 0.425230\n",
      "\n",
      "\n",
      "Validation...\n",
      "\n",
      "Test dataset: Overall Loss: 0.2884, Overall Accuracy: 2733/2977 (92%)\n",
      "\n",
      "epoch: 14 [0/187 (0%)]\t training loss: 0.822270\n",
      "epoch: 14 [10/187 (5%)]\t training loss: 0.792840\n",
      "epoch: 14 [20/187 (11%)]\t training loss: 0.696762\n",
      "epoch: 14 [30/187 (16%)]\t training loss: 0.643161\n",
      "epoch: 14 [40/187 (21%)]\t training loss: 0.842275\n",
      "epoch: 14 [50/187 (27%)]\t training loss: 0.804158\n",
      "epoch: 14 [60/187 (32%)]\t training loss: 0.750376\n",
      "epoch: 14 [70/187 (37%)]\t training loss: 0.144409\n",
      "epoch: 14 [80/187 (43%)]\t training loss: 0.486879\n",
      "epoch: 14 [90/187 (48%)]\t training loss: 0.748635\n",
      "epoch: 14 [100/187 (53%)]\t training loss: 0.395050\n",
      "epoch: 14 [110/187 (59%)]\t training loss: 0.527938\n",
      "epoch: 14 [120/187 (64%)]\t training loss: 0.349591\n",
      "epoch: 14 [130/187 (70%)]\t training loss: 0.782959\n",
      "epoch: 14 [140/187 (75%)]\t training loss: 0.289548\n",
      "epoch: 14 [150/187 (80%)]\t training loss: 0.797194\n",
      "epoch: 14 [160/187 (86%)]\t training loss: 0.668947\n",
      "epoch: 14 [170/187 (91%)]\t training loss: 0.756657\n",
      "epoch: 14 [180/187 (96%)]\t training loss: 0.237165\n",
      "\n",
      "\n",
      "Validation...\n",
      "\n",
      "Test dataset: Overall Loss: 0.2979, Overall Accuracy: 2714/2977 (91%)\n",
      "\n",
      "epoch: 15 [0/187 (0%)]\t training loss: 0.678629\n",
      "epoch: 15 [10/187 (5%)]\t training loss: 0.133729\n",
      "epoch: 15 [20/187 (11%)]\t training loss: 0.985178\n",
      "epoch: 15 [30/187 (16%)]\t training loss: 0.163185\n",
      "epoch: 15 [40/187 (21%)]\t training loss: 0.627250\n",
      "epoch: 15 [50/187 (27%)]\t training loss: 0.785898\n",
      "epoch: 15 [60/187 (32%)]\t training loss: 0.606559\n",
      "epoch: 15 [70/187 (37%)]\t training loss: 0.131519\n",
      "epoch: 15 [80/187 (43%)]\t training loss: 0.399458\n",
      "epoch: 15 [90/187 (48%)]\t training loss: 0.279788\n",
      "epoch: 15 [100/187 (53%)]\t training loss: 0.316785\n",
      "epoch: 15 [110/187 (59%)]\t training loss: 0.395053\n",
      "epoch: 15 [120/187 (64%)]\t training loss: 0.353594\n",
      "epoch: 15 [130/187 (70%)]\t training loss: 0.181225\n",
      "epoch: 15 [140/187 (75%)]\t training loss: 0.331647\n",
      "epoch: 15 [150/187 (80%)]\t training loss: 0.637840\n",
      "epoch: 15 [160/187 (86%)]\t training loss: 0.369048\n",
      "epoch: 15 [170/187 (91%)]\t training loss: 0.275044\n",
      "epoch: 15 [180/187 (96%)]\t training loss: 0.667177\n",
      "\n",
      "\n",
      "Validation...\n",
      "\n",
      "Test dataset: Overall Loss: 0.2660, Overall Accuracy: 2744/2977 (92%)\n",
      "\n",
      "epoch: 16 [0/187 (0%)]\t training loss: 0.458447\n",
      "epoch: 16 [10/187 (5%)]\t training loss: 0.559414\n",
      "epoch: 16 [20/187 (11%)]\t training loss: 0.268980\n",
      "epoch: 16 [30/187 (16%)]\t training loss: 0.096869\n",
      "epoch: 16 [40/187 (21%)]\t training loss: 0.136442\n",
      "epoch: 16 [50/187 (27%)]\t training loss: 0.336323\n",
      "epoch: 16 [60/187 (32%)]\t training loss: 0.214033\n",
      "epoch: 16 [70/187 (37%)]\t training loss: 0.598701\n",
      "epoch: 16 [80/187 (43%)]\t training loss: 0.796663\n",
      "epoch: 16 [90/187 (48%)]\t training loss: 0.826789\n",
      "epoch: 16 [100/187 (53%)]\t training loss: 0.158782\n",
      "epoch: 16 [110/187 (59%)]\t training loss: 0.344094\n",
      "epoch: 16 [120/187 (64%)]\t training loss: 0.298124\n",
      "epoch: 16 [130/187 (70%)]\t training loss: 0.764541\n",
      "epoch: 16 [140/187 (75%)]\t training loss: 0.241855\n",
      "epoch: 16 [150/187 (80%)]\t training loss: 0.352141\n",
      "epoch: 16 [160/187 (86%)]\t training loss: 0.659535\n",
      "epoch: 16 [170/187 (91%)]\t training loss: 0.854238\n",
      "epoch: 16 [180/187 (96%)]\t training loss: 0.753120\n",
      "\n",
      "\n",
      "Validation...\n",
      "\n",
      "Test dataset: Overall Loss: 0.2915, Overall Accuracy: 2712/2977 (91%)\n",
      "\n",
      "epoch: 17 [0/187 (0%)]\t training loss: 0.493361\n",
      "epoch: 17 [10/187 (5%)]\t training loss: 0.670316\n",
      "epoch: 17 [20/187 (11%)]\t training loss: 0.847839\n",
      "epoch: 17 [30/187 (16%)]\t training loss: 0.835179\n",
      "epoch: 17 [40/187 (21%)]\t training loss: 0.842714\n",
      "epoch: 17 [50/187 (27%)]\t training loss: 0.302902\n",
      "epoch: 17 [60/187 (32%)]\t training loss: 0.541557\n",
      "epoch: 17 [70/187 (37%)]\t training loss: 0.565799\n",
      "epoch: 17 [80/187 (43%)]\t training loss: 0.080826\n",
      "epoch: 17 [90/187 (48%)]\t training loss: 0.801934\n",
      "epoch: 17 [100/187 (53%)]\t training loss: 0.360065\n",
      "epoch: 17 [110/187 (59%)]\t training loss: 0.441590\n",
      "epoch: 17 [120/187 (64%)]\t training loss: 0.709115\n",
      "epoch: 17 [130/187 (70%)]\t training loss: 0.503275\n",
      "epoch: 17 [140/187 (75%)]\t training loss: 0.634638\n",
      "epoch: 17 [150/187 (80%)]\t training loss: 0.748888\n",
      "epoch: 17 [160/187 (86%)]\t training loss: 0.842019\n",
      "epoch: 17 [170/187 (91%)]\t training loss: 0.104044\n",
      "epoch: 17 [180/187 (96%)]\t training loss: 0.519128\n",
      "\n",
      "\n",
      "Validation...\n",
      "\n",
      "Test dataset: Overall Loss: 0.2876, Overall Accuracy: 2740/2977 (92%)\n",
      "\n",
      "epoch: 18 [0/187 (0%)]\t training loss: 0.499664\n",
      "epoch: 18 [10/187 (5%)]\t training loss: 0.118857\n",
      "epoch: 18 [20/187 (11%)]\t training loss: 0.388661\n",
      "epoch: 18 [30/187 (16%)]\t training loss: 0.715007\n",
      "epoch: 18 [40/187 (21%)]\t training loss: 0.145903\n",
      "epoch: 18 [50/187 (27%)]\t training loss: 0.744434\n",
      "epoch: 18 [60/187 (32%)]\t training loss: 0.329741\n",
      "epoch: 18 [70/187 (37%)]\t training loss: 0.196874\n",
      "epoch: 18 [80/187 (43%)]\t training loss: 0.621893\n",
      "epoch: 18 [90/187 (48%)]\t training loss: 0.154886\n",
      "epoch: 18 [100/187 (53%)]\t training loss: 0.938736\n",
      "epoch: 18 [110/187 (59%)]\t training loss: 0.658223\n",
      "epoch: 18 [120/187 (64%)]\t training loss: 0.545020\n",
      "epoch: 18 [130/187 (70%)]\t training loss: 0.414826\n",
      "epoch: 18 [140/187 (75%)]\t training loss: 0.256703\n",
      "epoch: 18 [150/187 (80%)]\t training loss: 0.275730\n",
      "epoch: 18 [160/187 (86%)]\t training loss: 0.299133\n",
      "epoch: 18 [170/187 (91%)]\t training loss: 0.763230\n",
      "epoch: 18 [180/187 (96%)]\t training loss: 0.727961\n",
      "\n",
      "\n",
      "Validation...\n",
      "\n",
      "Test dataset: Overall Loss: 0.2760, Overall Accuracy: 2727/2977 (92%)\n",
      "\n",
      "epoch: 19 [0/187 (0%)]\t training loss: 0.829116\n",
      "epoch: 19 [10/187 (5%)]\t training loss: 0.996577\n",
      "epoch: 19 [20/187 (11%)]\t training loss: 0.123825\n",
      "epoch: 19 [30/187 (16%)]\t training loss: 0.082852\n",
      "epoch: 19 [40/187 (21%)]\t training loss: 0.382599\n",
      "epoch: 19 [50/187 (27%)]\t training loss: 0.328829\n",
      "epoch: 19 [60/187 (32%)]\t training loss: 0.432871\n",
      "epoch: 19 [70/187 (37%)]\t training loss: 0.164906\n",
      "epoch: 19 [80/187 (43%)]\t training loss: 0.768842\n",
      "epoch: 19 [90/187 (48%)]\t training loss: 0.784787\n",
      "epoch: 19 [100/187 (53%)]\t training loss: 0.667278\n",
      "epoch: 19 [110/187 (59%)]\t training loss: 0.797969\n",
      "epoch: 19 [120/187 (64%)]\t training loss: 0.232546\n",
      "epoch: 19 [130/187 (70%)]\t training loss: 0.508766\n",
      "epoch: 19 [140/187 (75%)]\t training loss: 0.140952\n",
      "epoch: 19 [150/187 (80%)]\t training loss: 0.519013\n",
      "epoch: 19 [160/187 (86%)]\t training loss: 0.351821\n",
      "epoch: 19 [170/187 (91%)]\t training loss: 0.403609\n",
      "epoch: 19 [180/187 (96%)]\t training loss: 0.290017\n",
      "\n",
      "\n",
      "Validation...\n",
      "\n",
      "Test dataset: Overall Loss: 0.2937, Overall Accuracy: 2727/2977 (92%)\n",
      "\n",
      "epoch: 20 [0/187 (0%)]\t training loss: 0.758772\n",
      "epoch: 20 [10/187 (5%)]\t training loss: 0.684110\n",
      "epoch: 20 [20/187 (11%)]\t training loss: 0.687504\n",
      "epoch: 20 [30/187 (16%)]\t training loss: 0.527873\n",
      "epoch: 20 [40/187 (21%)]\t training loss: 0.386090\n",
      "epoch: 20 [50/187 (27%)]\t training loss: 0.406864\n",
      "epoch: 20 [60/187 (32%)]\t training loss: 0.541511\n",
      "epoch: 20 [70/187 (37%)]\t training loss: 0.521107\n",
      "epoch: 20 [80/187 (43%)]\t training loss: 0.656973\n",
      "epoch: 20 [90/187 (48%)]\t training loss: 0.478758\n",
      "epoch: 20 [100/187 (53%)]\t training loss: 0.938738\n",
      "epoch: 20 [110/187 (59%)]\t training loss: 0.489023\n",
      "epoch: 20 [120/187 (64%)]\t training loss: 0.401544\n",
      "epoch: 20 [130/187 (70%)]\t training loss: 0.770810\n",
      "epoch: 20 [140/187 (75%)]\t training loss: 0.468550\n",
      "epoch: 20 [150/187 (80%)]\t training loss: 0.780575\n",
      "epoch: 20 [160/187 (86%)]\t training loss: 0.288004\n",
      "epoch: 20 [170/187 (91%)]\t training loss: 0.630336\n",
      "epoch: 20 [180/187 (96%)]\t training loss: 0.843937\n",
      "\n",
      "\n",
      "Validation...\n",
      "\n",
      "Test dataset: Overall Loss: 0.3070, Overall Accuracy: 2721/2977 (91%)\n",
      "\n",
      "epoch: 21 [0/187 (0%)]\t training loss: 0.740478\n",
      "epoch: 21 [10/187 (5%)]\t training loss: 0.700029\n",
      "epoch: 21 [20/187 (11%)]\t training loss: 0.402691\n",
      "epoch: 21 [30/187 (16%)]\t training loss: 0.665910\n",
      "epoch: 21 [40/187 (21%)]\t training loss: 0.911837\n",
      "epoch: 21 [50/187 (27%)]\t training loss: 0.329948\n",
      "epoch: 21 [60/187 (32%)]\t training loss: 0.612210\n",
      "epoch: 21 [70/187 (37%)]\t training loss: 0.709196\n",
      "epoch: 21 [80/187 (43%)]\t training loss: 0.281985\n",
      "epoch: 21 [90/187 (48%)]\t training loss: 0.116362\n",
      "epoch: 21 [100/187 (53%)]\t training loss: 0.858499\n",
      "epoch: 21 [110/187 (59%)]\t training loss: 0.152443\n",
      "epoch: 21 [120/187 (64%)]\t training loss: 0.455189\n",
      "epoch: 21 [130/187 (70%)]\t training loss: 0.449793\n",
      "epoch: 21 [140/187 (75%)]\t training loss: 0.341095\n",
      "epoch: 21 [150/187 (80%)]\t training loss: 0.535059\n",
      "epoch: 21 [160/187 (86%)]\t training loss: 0.326906\n",
      "epoch: 21 [170/187 (91%)]\t training loss: 0.173589\n",
      "epoch: 21 [180/187 (96%)]\t training loss: 0.062840\n",
      "\n",
      "\n",
      "Validation...\n",
      "\n",
      "Test dataset: Overall Loss: 0.3393, Overall Accuracy: 2705/2977 (91%)\n",
      "\n",
      "epoch: 22 [0/187 (0%)]\t training loss: 0.287650\n",
      "epoch: 22 [10/187 (5%)]\t training loss: 0.646468\n",
      "epoch: 22 [20/187 (11%)]\t training loss: 0.629720\n",
      "epoch: 22 [30/187 (16%)]\t training loss: 0.646137\n",
      "epoch: 22 [40/187 (21%)]\t training loss: 0.566207\n",
      "epoch: 22 [50/187 (27%)]\t training loss: 0.613515\n",
      "epoch: 22 [60/187 (32%)]\t training loss: 0.316145\n",
      "epoch: 22 [70/187 (37%)]\t training loss: 0.336791\n",
      "epoch: 22 [80/187 (43%)]\t training loss: 0.681742\n",
      "epoch: 22 [90/187 (48%)]\t training loss: 0.358055\n",
      "epoch: 22 [100/187 (53%)]\t training loss: 0.469947\n",
      "epoch: 22 [110/187 (59%)]\t training loss: 0.549257\n",
      "epoch: 22 [120/187 (64%)]\t training loss: 0.200337\n",
      "epoch: 22 [130/187 (70%)]\t training loss: 0.639666\n",
      "epoch: 22 [140/187 (75%)]\t training loss: 0.828720\n",
      "epoch: 22 [150/187 (80%)]\t training loss: 0.768655\n",
      "epoch: 22 [160/187 (86%)]\t training loss: 0.819713\n",
      "epoch: 22 [170/187 (91%)]\t training loss: 0.388118\n",
      "epoch: 22 [180/187 (96%)]\t training loss: 0.737220\n",
      "\n",
      "\n",
      "Validation...\n",
      "\n",
      "Test dataset: Overall Loss: 0.2853, Overall Accuracy: 2736/2977 (92%)\n",
      "\n",
      "epoch: 23 [0/187 (0%)]\t training loss: 0.702324\n",
      "epoch: 23 [10/187 (5%)]\t training loss: 0.537989\n",
      "epoch: 23 [20/187 (11%)]\t training loss: 0.319598\n",
      "epoch: 23 [30/187 (16%)]\t training loss: 0.392473\n",
      "epoch: 23 [40/187 (21%)]\t training loss: 0.122949\n",
      "epoch: 23 [50/187 (27%)]\t training loss: 0.163603\n",
      "epoch: 23 [60/187 (32%)]\t training loss: 0.729629\n",
      "epoch: 23 [70/187 (37%)]\t training loss: 0.733395\n",
      "epoch: 23 [80/187 (43%)]\t training loss: 0.180562\n",
      "epoch: 23 [90/187 (48%)]\t training loss: 0.448411\n",
      "epoch: 23 [100/187 (53%)]\t training loss: 0.066079\n",
      "epoch: 23 [110/187 (59%)]\t training loss: 0.483936\n",
      "epoch: 23 [120/187 (64%)]\t training loss: 0.033918\n",
      "epoch: 23 [130/187 (70%)]\t training loss: 0.330929\n",
      "epoch: 23 [140/187 (75%)]\t training loss: 0.600552\n",
      "epoch: 23 [150/187 (80%)]\t training loss: 0.097958\n",
      "epoch: 23 [160/187 (86%)]\t training loss: 0.579252\n",
      "epoch: 23 [170/187 (91%)]\t training loss: 0.741779\n",
      "epoch: 23 [180/187 (96%)]\t training loss: 0.636580\n",
      "\n",
      "\n",
      "Validation...\n",
      "\n",
      "Test dataset: Overall Loss: 0.3210, Overall Accuracy: 2714/2977 (91%)\n",
      "\n",
      "epoch: 24 [0/187 (0%)]\t training loss: 0.140221\n",
      "epoch: 24 [10/187 (5%)]\t training loss: 0.777185\n",
      "epoch: 24 [20/187 (11%)]\t training loss: 0.435985\n",
      "epoch: 24 [30/187 (16%)]\t training loss: 0.654075\n",
      "epoch: 24 [40/187 (21%)]\t training loss: 0.425875\n",
      "epoch: 24 [50/187 (27%)]\t training loss: 0.892537\n",
      "epoch: 24 [60/187 (32%)]\t training loss: 0.324181\n",
      "epoch: 24 [70/187 (37%)]\t training loss: 0.685055\n",
      "epoch: 24 [80/187 (43%)]\t training loss: 0.497252\n",
      "epoch: 24 [90/187 (48%)]\t training loss: 0.138435\n",
      "epoch: 24 [100/187 (53%)]\t training loss: 0.627159\n",
      "epoch: 24 [110/187 (59%)]\t training loss: 0.213367\n",
      "epoch: 24 [120/187 (64%)]\t training loss: 0.106582\n",
      "epoch: 24 [130/187 (70%)]\t training loss: 0.344796\n",
      "epoch: 24 [140/187 (75%)]\t training loss: 0.513771\n",
      "epoch: 24 [150/187 (80%)]\t training loss: 0.738783\n",
      "epoch: 24 [160/187 (86%)]\t training loss: 0.319448\n",
      "epoch: 24 [170/187 (91%)]\t training loss: 0.469024\n",
      "epoch: 24 [180/187 (96%)]\t training loss: 0.717972\n",
      "\n",
      "\n",
      "Validation...\n",
      "\n",
      "Test dataset: Overall Loss: 0.3314, Overall Accuracy: 2689/2977 (90%)\n",
      "\n",
      "epoch: 25 [0/187 (0%)]\t training loss: 0.815629\n",
      "epoch: 25 [10/187 (5%)]\t training loss: 0.848227\n",
      "epoch: 25 [20/187 (11%)]\t training loss: 0.039513\n",
      "epoch: 25 [30/187 (16%)]\t training loss: 0.784513\n",
      "epoch: 25 [40/187 (21%)]\t training loss: 0.445720\n",
      "epoch: 25 [50/187 (27%)]\t training loss: 0.642658\n",
      "epoch: 25 [60/187 (32%)]\t training loss: 0.837559\n",
      "epoch: 25 [70/187 (37%)]\t training loss: 0.287711\n",
      "epoch: 25 [80/187 (43%)]\t training loss: 0.727979\n",
      "epoch: 25 [90/187 (48%)]\t training loss: 0.197231\n",
      "epoch: 25 [100/187 (53%)]\t training loss: 0.770381\n",
      "epoch: 25 [110/187 (59%)]\t training loss: 0.696378\n",
      "epoch: 25 [120/187 (64%)]\t training loss: 0.176404\n",
      "epoch: 25 [130/187 (70%)]\t training loss: 0.647559\n",
      "epoch: 25 [140/187 (75%)]\t training loss: 0.206635\n",
      "epoch: 25 [150/187 (80%)]\t training loss: 0.680240\n",
      "epoch: 25 [160/187 (86%)]\t training loss: 0.652445\n",
      "epoch: 25 [170/187 (91%)]\t training loss: 0.787451\n",
      "epoch: 25 [180/187 (96%)]\t training loss: 0.122642\n",
      "\n",
      "\n",
      "Validation...\n",
      "\n",
      "Test dataset: Overall Loss: 0.3256, Overall Accuracy: 2705/2977 (91%)\n",
      "\n",
      "epoch: 26 [0/187 (0%)]\t training loss: 0.312449\n",
      "epoch: 26 [10/187 (5%)]\t training loss: 0.909583\n",
      "epoch: 26 [20/187 (11%)]\t training loss: 0.450725\n",
      "epoch: 26 [30/187 (16%)]\t training loss: 0.547576\n",
      "epoch: 26 [40/187 (21%)]\t training loss: 0.855028\n",
      "epoch: 26 [50/187 (27%)]\t training loss: 0.316301\n",
      "epoch: 26 [60/187 (32%)]\t training loss: 0.210940\n",
      "epoch: 26 [70/187 (37%)]\t training loss: 0.183851\n",
      "epoch: 26 [80/187 (43%)]\t training loss: 0.770359\n",
      "epoch: 26 [90/187 (48%)]\t training loss: 0.489754\n",
      "epoch: 26 [100/187 (53%)]\t training loss: 0.504321\n",
      "epoch: 26 [110/187 (59%)]\t training loss: 0.386766\n",
      "epoch: 26 [120/187 (64%)]\t training loss: 0.203243\n",
      "epoch: 26 [130/187 (70%)]\t training loss: 0.612692\n",
      "epoch: 26 [140/187 (75%)]\t training loss: 0.783579\n",
      "epoch: 26 [150/187 (80%)]\t training loss: 0.400060\n",
      "epoch: 26 [160/187 (86%)]\t training loss: 0.086148\n",
      "epoch: 26 [170/187 (91%)]\t training loss: 0.056577\n",
      "epoch: 26 [180/187 (96%)]\t training loss: 0.042248\n",
      "\n",
      "\n",
      "Validation...\n",
      "\n",
      "Test dataset: Overall Loss: 0.3052, Overall Accuracy: 2743/2977 (92%)\n",
      "\n",
      "epoch: 27 [0/187 (0%)]\t training loss: 0.399743\n",
      "epoch: 27 [10/187 (5%)]\t training loss: 0.506044\n",
      "epoch: 27 [20/187 (11%)]\t training loss: 0.821938\n",
      "epoch: 27 [30/187 (16%)]\t training loss: 0.676690\n",
      "epoch: 27 [40/187 (21%)]\t training loss: 0.146391\n",
      "epoch: 27 [50/187 (27%)]\t training loss: 0.561159\n",
      "epoch: 27 [60/187 (32%)]\t training loss: 0.296238\n",
      "epoch: 27 [70/187 (37%)]\t training loss: 0.817834\n",
      "epoch: 27 [80/187 (43%)]\t training loss: 0.684233\n",
      "epoch: 27 [90/187 (48%)]\t training loss: 0.666304\n",
      "epoch: 27 [100/187 (53%)]\t training loss: 0.083884\n",
      "epoch: 27 [110/187 (59%)]\t training loss: 0.436394\n",
      "epoch: 27 [120/187 (64%)]\t training loss: 0.655244\n",
      "epoch: 27 [130/187 (70%)]\t training loss: 0.710222\n",
      "epoch: 27 [140/187 (75%)]\t training loss: 0.442775\n",
      "epoch: 27 [150/187 (80%)]\t training loss: 0.736684\n",
      "epoch: 27 [160/187 (86%)]\t training loss: 0.841431\n",
      "epoch: 27 [170/187 (91%)]\t training loss: 0.405109\n",
      "epoch: 27 [180/187 (96%)]\t training loss: 0.324921\n",
      "\n",
      "\n",
      "Validation...\n",
      "\n",
      "Test dataset: Overall Loss: 0.3163, Overall Accuracy: 2730/2977 (92%)\n",
      "\n",
      "epoch: 28 [0/187 (0%)]\t training loss: 0.119146\n",
      "epoch: 28 [10/187 (5%)]\t training loss: 0.229741\n",
      "epoch: 28 [20/187 (11%)]\t training loss: 0.798205\n",
      "epoch: 28 [30/187 (16%)]\t training loss: 0.155203\n",
      "epoch: 28 [40/187 (21%)]\t training loss: 0.679615\n",
      "epoch: 28 [50/187 (27%)]\t training loss: 0.482331\n",
      "epoch: 28 [60/187 (32%)]\t training loss: 0.080259\n",
      "epoch: 28 [70/187 (37%)]\t training loss: 0.366115\n",
      "epoch: 28 [80/187 (43%)]\t training loss: 0.618813\n",
      "epoch: 28 [90/187 (48%)]\t training loss: 0.660559\n",
      "epoch: 28 [100/187 (53%)]\t training loss: 0.183379\n",
      "epoch: 28 [110/187 (59%)]\t training loss: 0.775607\n",
      "epoch: 28 [120/187 (64%)]\t training loss: 0.587756\n",
      "epoch: 28 [130/187 (70%)]\t training loss: 0.153146\n",
      "epoch: 28 [140/187 (75%)]\t training loss: 0.742214\n",
      "epoch: 28 [150/187 (80%)]\t training loss: 0.358763\n",
      "epoch: 28 [160/187 (86%)]\t training loss: 0.791143\n",
      "epoch: 28 [170/187 (91%)]\t training loss: 0.391953\n",
      "epoch: 28 [180/187 (96%)]\t training loss: 0.557573\n",
      "\n",
      "\n",
      "Validation...\n",
      "\n",
      "Test dataset: Overall Loss: 0.3206, Overall Accuracy: 2722/2977 (91%)\n",
      "\n",
      "epoch: 29 [0/187 (0%)]\t training loss: 0.067473\n",
      "epoch: 29 [10/187 (5%)]\t training loss: 0.665613\n",
      "epoch: 29 [20/187 (11%)]\t training loss: 0.407327\n",
      "epoch: 29 [30/187 (16%)]\t training loss: 0.321113\n",
      "epoch: 29 [40/187 (21%)]\t training loss: 0.171989\n",
      "epoch: 29 [50/187 (27%)]\t training loss: 0.180111\n",
      "epoch: 29 [60/187 (32%)]\t training loss: 0.826762\n",
      "epoch: 29 [70/187 (37%)]\t training loss: 0.687351\n",
      "epoch: 29 [80/187 (43%)]\t training loss: 0.369853\n",
      "epoch: 29 [90/187 (48%)]\t training loss: 0.153685\n",
      "epoch: 29 [100/187 (53%)]\t training loss: 0.078414\n",
      "epoch: 29 [110/187 (59%)]\t training loss: 0.209291\n",
      "epoch: 29 [120/187 (64%)]\t training loss: 0.508292\n",
      "epoch: 29 [130/187 (70%)]\t training loss: 0.269371\n",
      "epoch: 29 [140/187 (75%)]\t training loss: 0.392354\n",
      "epoch: 29 [150/187 (80%)]\t training loss: 0.686460\n",
      "epoch: 29 [160/187 (86%)]\t training loss: 0.301706\n",
      "epoch: 29 [170/187 (91%)]\t training loss: 0.672165\n",
      "epoch: 29 [180/187 (96%)]\t training loss: 0.210949\n",
      "\n",
      "\n",
      "Validation...\n",
      "\n",
      "Test dataset: Overall Loss: 0.3048, Overall Accuracy: 2731/2977 (92%)\n",
      "\n"
     ]
    }
   ],
   "source": [
    "imbalanced_model_remix, train_losses, val_losses = train_remix(\n",
    "    cls_num_list,\n",
    "    imbalanced_train_loader,\n",
    "    imbalanced_valid_loader,\n",
    "    \"imbalanced_model_remix.pth\",\n",
    "    ConvNet().to(device),\n",
    "    epochs=30,\n",
    ")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 34,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 472
    },
    "id": "cdA3rtfEuC0v",
    "outputId": "35f1a9ca-6816-48f2-a0f2-e66cec555251"
   },
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 640x480 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "plt.plot(train_losses)\n",
    "plt.plot(val_losses)\n",
    "plt.xlabel(\"epoch\")\n",
    "plt.ylabel(\"loss\")\n",
    "plt.legend([\"Train\", \"Valid\"])\n",
    "plt.title(\"Model On imbalanced Fashion MNIST - Train vs Valid Accuracy\")\n",
    "\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 35,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/"
    },
    "id": "_08w1pn1uC0v",
    "outputId": "07b2392a-3347-40f6-ba2e-19b936c50b16"
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "On Imbalanced Fashion MNIST dataset\n",
      "\n",
      "Test dataset: Overall Loss: 0.0015, Overall Accuracy: 8051/10000 (81%)\n",
      "\n",
      "Remix test accuracy :  80.51\n"
     ]
    }
   ],
   "source": [
    "print(\"On Imbalanced Fashion MNIST dataset\")\n",
    "test_acc5, class_percentages5 = test(\n",
    "    imbalanced_model_remix, \"imbalanced_model_remix.pth\"\n",
    ")\n",
    "print(\"Remix test accuracy : \", str(test_acc5))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "iJ96fNucoHt0"
   },
   "source": [
    "# comparison"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 36,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 297
    },
    "id": "AM1VM9tntD3y",
    "outputId": "bd737857-b6a4-4c13-be60-797c59d16c2e"
   },
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 400x300 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "fig = plt.subplots(figsize=(4, 3))\n",
    "\n",
    "br1 = np.arange(1)\n",
    "br2 = [x + 0.1 for x in br1]\n",
    "br3 = [x + 0.1 for x in br2]\n",
    "br4 = [x + 0.1 for x in br3]\n",
    "br5 = [x + 0.1 for x in br4]\n",
    "\n",
    "p1 = plt.bar(\n",
    "    br1,\n",
    "    np.array(test_acc1).astype(float),\n",
    "    width=0.1,\n",
    "    label=\"Baseline model\",\n",
    "    color=\"gray\",\n",
    ")\n",
    "p2 = plt.bar(br2, np.array(test_acc2).astype(float), width=0.1, label=\"cutmix model\")\n",
    "p3 = plt.bar(br3, np.array(test_acc3).astype(float), width=0.1, label=\"mixup model\")\n",
    "p4 = plt.bar(\n",
    "    br4, np.array(test_acc4).astype(float), width=0.1, label=\"cutmix + mixup model\"\n",
    ")\n",
    "p5 = plt.bar(br5, np.array(test_acc5).astype(float), width=0.1, label=\"remix model\")\n",
    "\n",
    "for rect1 in p1:\n",
    "    height = rect1.get_height()\n",
    "    plt.annotate(\n",
    "        \"{:.2f}%\".format(height),\n",
    "        (rect1.get_x() + rect1.get_width() / 2, height + 0.05),\n",
    "        ha=\"center\",\n",
    "        va=\"bottom\",\n",
    "        fontsize=8,\n",
    "    )\n",
    "    plt.text(\n",
    "        (rect1.get_x() + rect1.get_width() / 2),\n",
    "        height - 12,\n",
    "        \"Baseline\",\n",
    "        ha=\"center\",\n",
    "        va=\"center\",\n",
    "        rotation=90,\n",
    "    )\n",
    "for rect1 in p2:\n",
    "    height = rect1.get_height()\n",
    "    plt.annotate(\n",
    "        \"{:.2f}%\".format(height),\n",
    "        (rect1.get_x() + rect1.get_width() / 2, height + 0.05),\n",
    "        ha=\"center\",\n",
    "        va=\"bottom\",\n",
    "        fontsize=8,\n",
    "    )\n",
    "    plt.text(\n",
    "        (rect1.get_x() + rect1.get_width() / 2),\n",
    "        height - 12,\n",
    "        \"Cutmix\",\n",
    "        ha=\"center\",\n",
    "        va=\"center\",\n",
    "        rotation=90,\n",
    "    )\n",
    "for rect1 in p3:\n",
    "    height = rect1.get_height()\n",
    "    plt.annotate(\n",
    "        \"{:.2f}%\".format(height),\n",
    "        (rect1.get_x() + rect1.get_width() / 2, height + 0.05),\n",
    "        ha=\"center\",\n",
    "        va=\"bottom\",\n",
    "        fontsize=8,\n",
    "    )\n",
    "    plt.text(\n",
    "        (rect1.get_x() + rect1.get_width() / 2),\n",
    "        height - 12,\n",
    "        \"Mixup\",\n",
    "        ha=\"center\",\n",
    "        va=\"center\",\n",
    "        rotation=90,\n",
    "    )\n",
    "for rect1 in p4:\n",
    "    height = rect1.get_height()\n",
    "    plt.annotate(\n",
    "        \"{:.2f}%\".format(height),\n",
    "        (rect1.get_x() + rect1.get_width() / 2, height + 0.05),\n",
    "        ha=\"center\",\n",
    "        va=\"bottom\",\n",
    "        fontsize=8,\n",
    "    )\n",
    "    plt.text(\n",
    "        (rect1.get_x() + rect1.get_width() / 2),\n",
    "        height - 12,\n",
    "        \"Cutmix+Mixup\",\n",
    "        ha=\"center\",\n",
    "        va=\"center\",\n",
    "        rotation=90,\n",
    "    )\n",
    "for rect1 in p5:\n",
    "    height = rect1.get_height()\n",
    "    plt.annotate(\n",
    "        \"{:.2f}%\".format(height),\n",
    "        (rect1.get_x() + rect1.get_width() / 2, height + 0.05),\n",
    "        ha=\"center\",\n",
    "        va=\"bottom\",\n",
    "        fontsize=8,\n",
    "    )\n",
    "    plt.text(\n",
    "        (rect1.get_x() + rect1.get_width() / 2),\n",
    "        height - 12,\n",
    "        \"Remix\",\n",
    "        ha=\"center\",\n",
    "        va=\"center\",\n",
    "        rotation=90,\n",
    "    )\n",
    "\n",
    "\n",
    "plt.xticks([r + 0.15 for r in range(1)], [\"Overall accuracy\"])\n",
    "plt.ylim(60, 90)\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 37,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 363
    },
    "id": "xO0mxWwtorh0",
    "outputId": "8ccb6bec-017c-4911-ac7d-dabf742a1c90"
   },
   "outputs": [
    {
     "data": {
      "text/html": [
       "\n",
       "  <div id=\"df-38961570-6752-49c9-bc61-a2fc4a70fba9\" class=\"colab-df-container\">\n",
       "    <div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>Baseline</th>\n",
       "      <th>Cutmix</th>\n",
       "      <th>Mixup</th>\n",
       "      <th>Cutmix+Mixup</th>\n",
       "      <th>Remix</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>0.923077</td>\n",
       "      <td>0.980769</td>\n",
       "      <td>0.923077</td>\n",
       "      <td>0.961538</td>\n",
       "      <td>0.903846</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>0.983871</td>\n",
       "      <td>1.000000</td>\n",
       "      <td>0.967742</td>\n",
       "      <td>0.983871</td>\n",
       "      <td>0.983871</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>0.851064</td>\n",
       "      <td>0.829787</td>\n",
       "      <td>0.872340</td>\n",
       "      <td>0.829787</td>\n",
       "      <td>0.957447</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>0.976190</td>\n",
       "      <td>0.976190</td>\n",
       "      <td>0.904762</td>\n",
       "      <td>0.952381</td>\n",
       "      <td>0.952381</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>0.680851</td>\n",
       "      <td>0.765957</td>\n",
       "      <td>0.744681</td>\n",
       "      <td>0.872340</td>\n",
       "      <td>0.851064</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>5</th>\n",
       "      <td>1.000000</td>\n",
       "      <td>0.980392</td>\n",
       "      <td>0.960784</td>\n",
       "      <td>0.980392</td>\n",
       "      <td>1.000000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>6</th>\n",
       "      <td>0.288889</td>\n",
       "      <td>0.311111</td>\n",
       "      <td>0.333333</td>\n",
       "      <td>0.333333</td>\n",
       "      <td>0.333333</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>7</th>\n",
       "      <td>0.795918</td>\n",
       "      <td>0.938776</td>\n",
       "      <td>0.979592</td>\n",
       "      <td>0.918367</td>\n",
       "      <td>0.836735</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>8</th>\n",
       "      <td>0.733333</td>\n",
       "      <td>0.900000</td>\n",
       "      <td>0.733333</td>\n",
       "      <td>0.650000</td>\n",
       "      <td>0.666667</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>9</th>\n",
       "      <td>0.688889</td>\n",
       "      <td>0.888889</td>\n",
       "      <td>0.800000</td>\n",
       "      <td>0.755556</td>\n",
       "      <td>0.888889</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>\n",
       "    <div class=\"colab-df-buttons\">\n",
       "\n",
       "  <div class=\"colab-df-container\">\n",
       "    <button class=\"colab-df-convert\" onclick=\"convertToInteractive('df-38961570-6752-49c9-bc61-a2fc4a70fba9')\"\n",
       "            title=\"Convert this dataframe to an interactive table.\"\n",
       "            style=\"display:none;\">\n",
       "\n",
       "  <svg xmlns=\"http://www.w3.org/2000/svg\" height=\"24px\" viewBox=\"0 -960 960 960\">\n",
       "    <path d=\"M120-120v-720h720v720H120Zm60-500h600v-160H180v160Zm220 220h160v-160H400v160Zm0 220h160v-160H400v160ZM180-400h160v-160H180v160Zm440 0h160v-160H620v160ZM180-180h160v-160H180v160Zm440 0h160v-160H620v160Z\"/>\n",
       "  </svg>\n",
       "    </button>\n",
       "\n",
       "  <style>\n",
       "    .colab-df-container {\n",
       "      display:flex;\n",
       "      gap: 12px;\n",
       "    }\n",
       "\n",
       "    .colab-df-convert {\n",
       "      background-color: #E8F0FE;\n",
       "      border: none;\n",
       "      border-radius: 50%;\n",
       "      cursor: pointer;\n",
       "      display: none;\n",
       "      fill: #1967D2;\n",
       "      height: 32px;\n",
       "      padding: 0 0 0 0;\n",
       "      width: 32px;\n",
       "    }\n",
       "\n",
       "    .colab-df-convert:hover {\n",
       "      background-color: #E2EBFA;\n",
       "      box-shadow: 0px 1px 2px rgba(60, 64, 67, 0.3), 0px 1px 3px 1px rgba(60, 64, 67, 0.15);\n",
       "      fill: #174EA6;\n",
       "    }\n",
       "\n",
       "    .colab-df-buttons div {\n",
       "      margin-bottom: 4px;\n",
       "    }\n",
       "\n",
       "    [theme=dark] .colab-df-convert {\n",
       "      background-color: #3B4455;\n",
       "      fill: #D2E3FC;\n",
       "    }\n",
       "\n",
       "    [theme=dark] .colab-df-convert:hover {\n",
       "      background-color: #434B5C;\n",
       "      box-shadow: 0px 1px 3px 1px rgba(0, 0, 0, 0.15);\n",
       "      filter: drop-shadow(0px 1px 2px rgba(0, 0, 0, 0.3));\n",
       "      fill: #FFFFFF;\n",
       "    }\n",
       "  </style>\n",
       "\n",
       "    <script>\n",
       "      const buttonEl =\n",
       "        document.querySelector('#df-38961570-6752-49c9-bc61-a2fc4a70fba9 button.colab-df-convert');\n",
       "      buttonEl.style.display =\n",
       "        google.colab.kernel.accessAllowed ? 'block' : 'none';\n",
       "\n",
       "      async function convertToInteractive(key) {\n",
       "        const element = document.querySelector('#df-38961570-6752-49c9-bc61-a2fc4a70fba9');\n",
       "        const dataTable =\n",
       "          await google.colab.kernel.invokeFunction('convertToInteractive',\n",
       "                                                    [key], {});\n",
       "        if (!dataTable) return;\n",
       "\n",
       "        const docLinkHtml = 'Like what you see? Visit the ' +\n",
       "          '<a target=\"_blank\" href=https://colab.research.google.com/notebooks/data_table.ipynb>data table notebook</a>'\n",
       "          + ' to learn more about interactive tables.';\n",
       "        element.innerHTML = '';\n",
       "        dataTable['output_type'] = 'display_data';\n",
       "        await google.colab.output.renderOutput(dataTable, element);\n",
       "        const docLink = document.createElement('div');\n",
       "        docLink.innerHTML = docLinkHtml;\n",
       "        element.appendChild(docLink);\n",
       "      }\n",
       "    </script>\n",
       "  </div>\n",
       "\n",
       "\n",
       "<div id=\"df-7af926da-cdc1-4f9d-818e-06cbd8a154c0\">\n",
       "  <button class=\"colab-df-quickchart\" onclick=\"quickchart('df-7af926da-cdc1-4f9d-818e-06cbd8a154c0')\"\n",
       "            title=\"Suggest charts.\"\n",
       "            style=\"display:none;\">\n",
       "\n",
       "<svg xmlns=\"http://www.w3.org/2000/svg\" height=\"24px\"viewBox=\"0 0 24 24\"\n",
       "     width=\"24px\">\n",
       "    <g>\n",
       "        <path d=\"M19 3H5c-1.1 0-2 .9-2 2v14c0 1.1.9 2 2 2h14c1.1 0 2-.9 2-2V5c0-1.1-.9-2-2-2zM9 17H7v-7h2v7zm4 0h-2V7h2v10zm4 0h-2v-4h2v4z\"/>\n",
       "    </g>\n",
       "</svg>\n",
       "  </button>\n",
       "\n",
       "<style>\n",
       "  .colab-df-quickchart {\n",
       "      --bg-color: #E8F0FE;\n",
       "      --fill-color: #1967D2;\n",
       "      --hover-bg-color: #E2EBFA;\n",
       "      --hover-fill-color: #174EA6;\n",
       "      --disabled-fill-color: #AAA;\n",
       "      --disabled-bg-color: #DDD;\n",
       "  }\n",
       "\n",
       "  [theme=dark] .colab-df-quickchart {\n",
       "      --bg-color: #3B4455;\n",
       "      --fill-color: #D2E3FC;\n",
       "      --hover-bg-color: #434B5C;\n",
       "      --hover-fill-color: #FFFFFF;\n",
       "      --disabled-bg-color: #3B4455;\n",
       "      --disabled-fill-color: #666;\n",
       "  }\n",
       "\n",
       "  .colab-df-quickchart {\n",
       "    background-color: var(--bg-color);\n",
       "    border: none;\n",
       "    border-radius: 50%;\n",
       "    cursor: pointer;\n",
       "    display: none;\n",
       "    fill: var(--fill-color);\n",
       "    height: 32px;\n",
       "    padding: 0;\n",
       "    width: 32px;\n",
       "  }\n",
       "\n",
       "  .colab-df-quickchart:hover {\n",
       "    background-color: var(--hover-bg-color);\n",
       "    box-shadow: 0 1px 2px rgba(60, 64, 67, 0.3), 0 1px 3px 1px rgba(60, 64, 67, 0.15);\n",
       "    fill: var(--button-hover-fill-color);\n",
       "  }\n",
       "\n",
       "  .colab-df-quickchart-complete:disabled,\n",
       "  .colab-df-quickchart-complete:disabled:hover {\n",
       "    background-color: var(--disabled-bg-color);\n",
       "    fill: var(--disabled-fill-color);\n",
       "    box-shadow: none;\n",
       "  }\n",
       "\n",
       "  .colab-df-spinner {\n",
       "    border: 2px solid var(--fill-color);\n",
       "    border-color: transparent;\n",
       "    border-bottom-color: var(--fill-color);\n",
       "    animation:\n",
       "      spin 1s steps(1) infinite;\n",
       "  }\n",
       "\n",
       "  @keyframes spin {\n",
       "    0% {\n",
       "      border-color: transparent;\n",
       "      border-bottom-color: var(--fill-color);\n",
       "      border-left-color: var(--fill-color);\n",
       "    }\n",
       "    20% {\n",
       "      border-color: transparent;\n",
       "      border-left-color: var(--fill-color);\n",
       "      border-top-color: var(--fill-color);\n",
       "    }\n",
       "    30% {\n",
       "      border-color: transparent;\n",
       "      border-left-color: var(--fill-color);\n",
       "      border-top-color: var(--fill-color);\n",
       "      border-right-color: var(--fill-color);\n",
       "    }\n",
       "    40% {\n",
       "      border-color: transparent;\n",
       "      border-right-color: var(--fill-color);\n",
       "      border-top-color: var(--fill-color);\n",
       "    }\n",
       "    60% {\n",
       "      border-color: transparent;\n",
       "      border-right-color: var(--fill-color);\n",
       "    }\n",
       "    80% {\n",
       "      border-color: transparent;\n",
       "      border-right-color: var(--fill-color);\n",
       "      border-bottom-color: var(--fill-color);\n",
       "    }\n",
       "    90% {\n",
       "      border-color: transparent;\n",
       "      border-bottom-color: var(--fill-color);\n",
       "    }\n",
       "  }\n",
       "</style>\n",
       "\n",
       "  <script>\n",
       "    async function quickchart(key) {\n",
       "      const quickchartButtonEl =\n",
       "        document.querySelector('#' + key + ' button');\n",
       "      quickchartButtonEl.disabled = true;  // To prevent multiple clicks.\n",
       "      quickchartButtonEl.classList.add('colab-df-spinner');\n",
       "      try {\n",
       "        const charts = await google.colab.kernel.invokeFunction(\n",
       "            'suggestCharts', [key], {});\n",
       "      } catch (error) {\n",
       "        console.error('Error during call to suggestCharts:', error);\n",
       "      }\n",
       "      quickchartButtonEl.classList.remove('colab-df-spinner');\n",
       "      quickchartButtonEl.classList.add('colab-df-quickchart-complete');\n",
       "    }\n",
       "    (() => {\n",
       "      let quickchartButtonEl =\n",
       "        document.querySelector('#df-7af926da-cdc1-4f9d-818e-06cbd8a154c0 button');\n",
       "      quickchartButtonEl.style.display =\n",
       "        google.colab.kernel.accessAllowed ? 'block' : 'none';\n",
       "    })();\n",
       "  </script>\n",
       "</div>\n",
       "\n",
       "  <div id=\"id_e3b36f6f-553e-49ef-a9d5-09bbae8182e8\">\n",
       "    <style>\n",
       "      .colab-df-generate {\n",
       "        background-color: #E8F0FE;\n",
       "        border: none;\n",
       "        border-radius: 50%;\n",
       "        cursor: pointer;\n",
       "        display: none;\n",
       "        fill: #1967D2;\n",
       "        height: 32px;\n",
       "        padding: 0 0 0 0;\n",
       "        width: 32px;\n",
       "      }\n",
       "\n",
       "      .colab-df-generate:hover {\n",
       "        background-color: #E2EBFA;\n",
       "        box-shadow: 0px 1px 2px rgba(60, 64, 67, 0.3), 0px 1px 3px 1px rgba(60, 64, 67, 0.15);\n",
       "        fill: #174EA6;\n",
       "      }\n",
       "\n",
       "      [theme=dark] .colab-df-generate {\n",
       "        background-color: #3B4455;\n",
       "        fill: #D2E3FC;\n",
       "      }\n",
       "\n",
       "      [theme=dark] .colab-df-generate:hover {\n",
       "        background-color: #434B5C;\n",
       "        box-shadow: 0px 1px 3px 1px rgba(0, 0, 0, 0.15);\n",
       "        filter: drop-shadow(0px 1px 2px rgba(0, 0, 0, 0.3));\n",
       "        fill: #FFFFFF;\n",
       "      }\n",
       "    </style>\n",
       "    <button class=\"colab-df-generate\" onclick=\"generateWithVariable('df')\"\n",
       "            title=\"Generate code using this dataframe.\"\n",
       "            style=\"display:none;\">\n",
       "\n",
       "  <svg xmlns=\"http://www.w3.org/2000/svg\" height=\"24px\"viewBox=\"0 0 24 24\"\n",
       "       width=\"24px\">\n",
       "    <path d=\"M7,19H8.4L18.45,9,17,7.55,7,17.6ZM5,21V16.75L18.45,3.32a2,2,0,0,1,2.83,0l1.4,1.43a1.91,1.91,0,0,1,.58,1.4,1.91,1.91,0,0,1-.58,1.4L9.25,21ZM18.45,9,17,7.55Zm-12,3A5.31,5.31,0,0,0,4.9,8.1,5.31,5.31,0,0,0,1,6.5,5.31,5.31,0,0,0,4.9,4.9,5.31,5.31,0,0,0,6.5,1,5.31,5.31,0,0,0,8.1,4.9,5.31,5.31,0,0,0,12,6.5,5.46,5.46,0,0,0,6.5,12Z\"/>\n",
       "  </svg>\n",
       "    </button>\n",
       "    <script>\n",
       "      (() => {\n",
       "      const buttonEl =\n",
       "        document.querySelector('#id_e3b36f6f-553e-49ef-a9d5-09bbae8182e8 button.colab-df-generate');\n",
       "      buttonEl.style.display =\n",
       "        google.colab.kernel.accessAllowed ? 'block' : 'none';\n",
       "\n",
       "      buttonEl.onclick = () => {\n",
       "        google.colab.notebook.generateWithVariable('df');\n",
       "      }\n",
       "      })();\n",
       "    </script>\n",
       "  </div>\n",
       "\n",
       "    </div>\n",
       "  </div>\n"
      ],
      "text/plain": [
       "   Baseline    Cutmix     Mixup  Cutmix+Mixup     Remix\n",
       "0  0.923077  0.980769  0.923077      0.961538  0.903846\n",
       "1  0.983871  1.000000  0.967742      0.983871  0.983871\n",
       "2  0.851064  0.829787  0.872340      0.829787  0.957447\n",
       "3  0.976190  0.976190  0.904762      0.952381  0.952381\n",
       "4  0.680851  0.765957  0.744681      0.872340  0.851064\n",
       "5  1.000000  0.980392  0.960784      0.980392  1.000000\n",
       "6  0.288889  0.311111  0.333333      0.333333  0.333333\n",
       "7  0.795918  0.938776  0.979592      0.918367  0.836735\n",
       "8  0.733333  0.900000  0.733333      0.650000  0.666667\n",
       "9  0.688889  0.888889  0.800000      0.755556  0.888889"
      ]
     },
     "execution_count": 37,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "df = pd.DataFrame(\n",
    "    {\n",
    "        \"Baseline\": class_percentages1,\n",
    "        \"Cutmix\": class_percentages2,\n",
    "        \"Mixup\": class_percentages3,\n",
    "        \"Cutmix+Mixup\": class_percentages4,\n",
    "        \"Remix\": class_percentages5,\n",
    "    }\n",
    ")\n",
    "\n",
    "df"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 38,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 564
    },
    "id": "Ok1SB-fUuEQt",
    "outputId": "b1134885-b0a1-419a-91a9-3ab3ed636254"
   },
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 1200x600 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "import matplotlib.pyplot as plt\n",
    "import numpy as np\n",
    "\n",
    "# Set the width of the bars\n",
    "bar_width = 0.15\n",
    "\n",
    "# Set the positions of the bars\n",
    "r1 = np.arange(len(df[\"Baseline\"]))\n",
    "r2 = [x + bar_width for x in r1]\n",
    "r3 = [x + bar_width for x in r2]\n",
    "r4 = [x + bar_width for x in r3]\n",
    "r5 = [x + bar_width for x in r4]\n",
    "\n",
    "plt.figure(figsize=(12, 6))\n",
    "\n",
    "# Create the bars\n",
    "bars1 = plt.bar(r1, df[\"Baseline\"], width=bar_width)\n",
    "bars2 = plt.bar(r2, df[\"Cutmix\"], width=bar_width)\n",
    "bars3 = plt.bar(r3, df[\"Mixup\"], width=bar_width)\n",
    "bars4 = plt.bar(r4, df[\"Cutmix+Mixup\"], width=bar_width)\n",
    "bars5 = plt.bar(r5, df[\"Remix\"], width=bar_width)\n",
    "\n",
    "\n",
    "# Function to add technique labels at the bottom of the bars\n",
    "def add_technique_labels(bars, positions, label):\n",
    "    for bar, pos in zip(bars, positions):\n",
    "        plt.text(pos, 0.3, label, ha=\"center\", va=\"center\", rotation=90)\n",
    "        height = bar.get_height()\n",
    "        plt.text(\n",
    "            pos, height + 0.005, f\"{height:.2}\", ha=\"center\", va=\"bottom\", rotation=90\n",
    "        )\n",
    "\n",
    "\n",
    "# Add technique labels\n",
    "add_technique_labels(bars1, r1, \"Baseline\")\n",
    "add_technique_labels(bars2, r2, \"Cutmix\")\n",
    "add_technique_labels(bars3, r3, \"Mixup\")\n",
    "add_technique_labels(bars4, r4, \"Cutmix+Mixup\")\n",
    "add_technique_labels(bars5, r5, \"Remix\")\n",
    "\n",
    "# Add labels and title\n",
    "plt.xlabel(\"Class Label\")\n",
    "plt.ylabel(\"Performance\")\n",
    "plt.title(\"Comparison of Techniques across Class Labels\")\n",
    "plt.xticks([r + bar_width * 2 for r in range(len(df[\"Baseline\"]))], df.index)\n",
    "plt.ylim(0, 1.1)\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 87,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 828
    },
    "id": "Su3Qx37-XpIq",
    "outputId": "644131ba-d972-4802-eea5-1144d821ff6a"
   },
   "outputs": [
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAA/4AAAMrCAYAAADwb0lRAAAAOXRFWHRTb2Z0d2FyZQBNYXRwbG90bGliIHZlcnNpb24zLjcuMSwgaHR0cHM6Ly9tYXRwbG90bGliLm9yZy/bCgiHAAAACXBIWXMAAA9hAAAPYQGoP6dpAADawElEQVR4nOzdd3gU1dvG8XuTkAYkhN47SO8gvYOIUlREAQFRsYGKHaxYEetPRQWRKlhAUREQkN4FpPfeSwghjRRS5v0DM+9uskk22U02LN/Pde2Vmd0zZ57Z7MzuM+fMGYthGIYAAAAAAIBH8nJ3AAAAAAAAIPeQ+AMAAAAA4MFI/AEAAAAA8GAk/gAAAAAAeDASfwAAAAAAPBiJPwAAAAAAHozEHwAAAAAAD0biDwAAAACAByPxBwAAAADAg5H4A3CJ6dOny2KxyGKxqHLlyu4OBw5KSUnR7Nmz1bt3b1WsWFEBAQHm/9FisWjHjh3uDjFfq1y5svleTZ8+3d3hOMz6f7xq1Sp3hwPAjU6cOGFzTDhx4oS7Q0IuuhG/tzp27GjGPHbsWHeHc8PycXcAuHFcuHBBS5cu1bJly7Rr1y6FhYUpLCxM3t7eCgkJUYUKFdSsWTO1b99ed955pwICAtwdMoBMxMXF6Y477tDKlSudruvEiROqUqWKC6JK7/jx45xMAgArV69e1YoVK7Rs2TJt2rRJoaGhCgsL07Vr1xQSEqKSJUuqSZMmatmype666y6VKlXK3SF7vI4dO2r16tWSpEqVKnECBfkOiT+ydPr0ab333nuaOnWqEhMT7ZaJjY3V2bNntWnTJk2YMEGFCxfWwIED9frrr6tcuXJ5HDEAR7z++us2SX+JEiVUu3Ztm5N2QUFB7ggNAGDH1atXNWHCBH300Ue6fPmy3TIXL17UxYsXtXv3bs2YMUMjR45Ujx49NHbsWDVr1iyPIwaQX5D4I1Pz5s3TkCFDdPXqVZvn/fz8VKlSJZUoUUJeXl66ePGiTp8+rbi4OElSdHS0Jk2apJkzZ2rjxo1q2LChO8IHkIGUlBRNmTLFnH/kkUc0ceJEeXt756i+gIAA3XbbbVmW27x5s65cuSJJ8vf3V4cOHRyqGwBudvv27VPv3r119OhRm+dTL7ErUaKEChYsqEuXLuns2bPmsTY5OVkLFy7UwoULNXHiRD322GPuCB+Am5H4I0OffPKJXnzxRRmGYT7Xs2dPPfXUU2rfvr0CAwNtyl+7dk2rV6/WL7/8ounTp+vatWuKi4szv3jg2R588EE9+OCD7g4DDjp06JAiIiLM+VdeeSXHSb8klSpVSosXL86ynHVXSEeXgetZH9cB5H+rVq1Snz59FBUVZT5Xt25dvfTSS+rRo4dKlixpU94wDG3btk0LFizQV199pUuXLkmSzp8/n6dxA8g/GNwPdi1cuNAm6Q8ODtbixYu1cOFC9ejRI13SL0m+vr7q1q2bJk2apMOHD2vAgAF5HTYAB6XtIlqhQgU3RQIAyMypU6fUr18/M+m3WCz64IMPtGvXLg0ZMiRd0p9apmnTpnrzzTd1/Phxvf322ypQoEBehw4gH6HFH+lcuHBBgwcPNpP+ggULas2aNWrQoIHDdVSsWFE//PCD2rdvL39//9wKFUAOpR2vw8eHrwMAyI/uv/9+m5O1kydP1sMPP+zw8gULFtTrr7+uzp07a9++fbkRIoAbAC3+SOeTTz6x6Z7/6aefZivpt/b444+rZcuWWZZbunSpHnvsMdWtW1dFixaVn5+fypUrp3bt2undd9/VyZMnHVpfRreUO378uF577TU1btxYxYsXl5+fn6pWrarhw4fr4MGDdutav369hg4dqmrVqsnf319BQUFq1KiR3nzzTZuudhlZtWqVze1xUp09e1bvvfeeWrRooVKlSsnf31+VK1fWfffdp4ULFzq0nan27dunzz//XP3791e9evVUpEgRFShQQEWKFFGNGjU0YMAAzZw5M8NBGdPK6P07duyY3n77bd16660qW7asfHx80t3qLbu389u3b5/GjBmjtm3bqmTJkvLz85Ovr6+KFCmiOnXqqFevXnr77be1detWh2JPSUnRL7/8oiFDhqhmzZoqUqSI/P39VaFCBXXt2lWffPKJwsLCHKpr7Nix5rZ07NjR5n145ZVX1KhRIxUtWlQBAQGqXLmyBg0a5JKR8bNy4sQJvfPOO2rbtq3KlSsnPz8/FS1aVHXr1tXjjz+uv//+O9Plrf9HnTp1snnN+rOaH2/zs2bNGo0aNUqNGzdWqVKl5Ovrq5IlS6p58+YaPXq0Dhw4kKN69+3bp7Fjx6p9+/aqUKGCAgICFBAQoHLlyqlLly56/fXXtXnz5mzX+9dff+m+++5T9erVFRAQoJCQEDVu3FijR4/WhQsXHKrD3i33kpKSNHfuXPXq1UtVqlSRv7+/ihcvrpYtW+rdd99VZGRkjuvOyuLFizVgwABzvaVKlVLz5s317rvv6ty5c9mqOye3/8zpLceuXLmiCRMm6M4771TVqlVVqFAhFSxYUFWqVFG/fv00c+ZMJSUlOVSXJIWGhurTTz9Vjx49VKFCBRUsWFA+Pj4qXLiwqlatqs6dO+uFF17QokWLHD72OiI2NlZ//PGHnnvuOXXs2FHlypVTQECA/P39VaZMGbVp00ajR4/O8DvNEf/8849Gjx6tli1bqly5cvL391dgYKAqVaqk22+/Xe+//7727t2b4fL2/vdxcXGaOXOmevXqpWrVqqlgwYKyWCwaNWqU3Tqio6P19ddf684771TlypUVGBiowoULq1q1aurfv79mzpypa9euZWu7li9frkcffdQ8dhcoUED+/v4qUaKEmjRpogEDBujzzz/P8rdGSkqKfv/9dw0ePFh169ZVkSJF5OPjo4CAAJUuXVrNmzfXgw8+qG+//VYXL17MVowZWbhwoTZu3GjODxgwIFtJv7U2bdpo+PDhTsd0/PhxTZo0SQ888IAaN25svqdBQUGqUqWK7r77bn399dfpxofKiiv3rZMnT+qdd95R586dVaZMGQUEBJgx1qxZUz169NCrr76q1atXKyUlxZm3I1flxX4vXW8M+Omnn8zjZEBAgEqWLKk2bdro008/deh3b1qGYWjhwoV67LHHVK9ePRUvXly+vr4qXbq02rRpo7feektnzpxxKm573LGf3jAMwEpERIRRuHBhQ5IhyahevbqRkpKSa+s7evSo0b59e3N9GT38/PyMV1991UhOTs60vmnTppnLVKpUyTAMw5g8ebIREBCQYd2+vr7Gn3/+adaRkJBgDBs2LNN4ypUrZ+zfvz/TWFauXGmzjGEYxi+//GIEBQVlWvedd95pXLlyJdO64+Pjjbp162b5vqU+KlasaKxbty7L/4e99+/zzz83/P397da7ffv2TJe1JykpyXjqqacMLy8vh+NfuHBhpnFv3brVaNCgQZb1BAUFGV988UWW78Obb75pLtOhQwfDMAzjiy++MPz8/DKtf/jw4UZSUlKW9WdXUlKSMXr06CzXL8no2LGjcfz4cbv1WP+PHHlMmzbN5dvSoUMHhz4nqQ4fPmx06dIly1i9vb2Np59+2khMTHQojtDQUOO+++4zLBaLQ+/FM888Y7eeSpUq2bxfYWFhRq9evTKtq1ChQll+pg3DsFlm5cqVxokTJ4xWrVplWnfJkiWNLVu2ZLvuzERGRhp9+/bNct/65ZdfHK7b0eOFtePHj9vUndHn3Nqnn35qFClSJMv/b40aNYxNmzZlWd+UKVOyPIZbP1588UWHti0rkyZNMgIDAx1ap5eXl/Hoo48a8fHxDtd/5MgR47bbbnN4uz777DO79aT93+/YscOoXbu2w/vU7NmzjZIlS2a5/mrVqhnLly/PcrsuXrxodO7cOVvHvZiYGLt1HTp0yGjcuLHD9ZQoUcLh9z8zbdu2tfnfnjhxwiX1WsvOvuXI8Tj1UaxYMWPevHkOxeDKfeu9995z6Psy9fHVV19l9y2zK7vfb1nJrf0+7ffWqVOnjJYtW2Zaf/ny5bP8rrC2efNmo2nTplnGHRAQYLz33ntZ1mf93r755psZlnPXfnqjoG8nbPz999+Kjo425x977DGb1mpX2r17t7p162Zztq1AgQKqV6+egoODderUKR07dkySlJCQoPfee0/79+/Xzz//7HC35GnTpplntwMCAlSvXj0VLFhQR48e1enTpyVdH5SwX79+Wr9+vRo3bqx+/frpzz//lCQVK1ZMt9xyi7y9vbVnzx6zJ8TZs2fVo0cP7d27VwULFnQoliVLlujee++VYRiyWCyqXbu2SpYsqdDQUO3fv9+8tGLBggXq3r27li9frsKFC9utKzEx0abVxcfHR9WqVVPx4sXl7++vK1eu6MCBA4qNjZV0/frATp066e+//3ZoFPVUn3/+udkq4+3trTp16qh48eIKCwvLcQvr448/ru+++87muSpVqqhChQry9fVVdHS0Tp48adMqmtnZ+BUrVqhPnz6KiYkxnwsMDFSdOnUUGBioo0eP6uzZs5KkqKgoPf300zpy5Ig+//xzh2P+4IMPNGbMGEnX72hRr149BQUF6cyZMzp8+LBZbvLkySpevLjef/99h+vOSmJiovr166f58+fbPF+lShVVqlRJUVFR2r17t9n6sWrVKrVu3Vp///236tata7NMuXLlzJH3w8PDtWXLFvM1eyPyu/tWnBs3blSvXr1surj6+/urTp06KlKkiMLDw7Vnzx4lJSUpOTlZX3zxhQ4fPqz58+dneozYt2+fevbsma51r0KFCqpYsaIKFCig0NBQHTx4UMnJyZJkMxBiRmJjY9W9e3dt27ZNklS6dGlVq1ZNFotFe/fuNY8fMTExuuuuu/Tvv/+qXr16Dr0XoaGhGjp0qE6dOiXp+uVUlStXVmJionbt2mW2rIWGhuq2227T3r17Vbp0aYfqzszVq1d1++23a8OGDTbP165dW6VLl1ZoaKj27dunqKgo3XvvvflmsMakpCQ9/PDDmjlzps3zlSpVUsWKFSVJhw8fNo8zhw8fVqdOnTR//nx17drVbp2zZ89O19JasmRJsxU7NjZWFy5c0IkTJ8xjlqtaEg8dOmQez6Xr301VqlRRUFCQEhMTderUKfPznJKSom+//VanTp3SokWLsvwOX7Nmje666y6Fh4fbPF+tWjWVK1dOhmHowoULOnLkiPk95cj+cPz4cd1zzz1mveXLl1eVKlV07do1HTlyJF35Dz/8UC+//LLNcyVKlFDNmjWVnJys/fv3mz1ajh49qh49emjWrFnq37+/3fUnJCSoS5cu2rNnj/lcgQIFVLNmTfOORBERETpy5IhNa6ZhZ+DLsLAwdejQwWZgPH9/f91yyy0qVqyYkpOTFRERocOHD5v/J1f87y9duqT169eb87fffrsqVarkdL3OsO7p5+XlpSpVqqhkyZIKDAxUdHS0Dhw4YL6fly9f1j333KNZs2Zp4MCBGdbpyn3rvffe02uvvWbzXLly5VS5cmUFBAQoJiZG586dM4+lmdXlbrm536eKiYlR165ddejQIUnXB92tWbOmEhIStG/fPvO31ZkzZ9SzZ08tXLgwXY/BtP744w8NGDDAvNOXJBUuXFi1a9dWoUKFdPHiRe3bt0+GYSguLk6vvvqqTpw4oW+//TZb709a7tpPbyjuPOuA/GfEiBE2Z8J27NiRK+uJiYkxatSoYXOm8qWXXjLCw8Ntym3bts1o0aKFTUxvvPFGhvVatyIVLFjQ8Pf3N/z8/IyPP/7YiI2NtSn7+++/GwULFjTL33777cYHH3xgntn87bffbHoYXLt2zaYlWJLx9ttvZxhL2hb/EiVKGJKMXr16GceOHbMpe+zYMePOO++0Kf/II49kWHd0dLRRtGhRY9SoUcbKlSuNhISEdGWuXbtmzJkzx6hcubJZZ9myZdO9Dxm9fwEBAYaPj49hsViMl156yQgLC7Mpe/HiRSMyMtLushmd6d6xY4fNNg4dOtQ4efKk3bLnz583pk6darRt29ZYsGCB3TLnzp0zihcvbtbn5+dnfPjhh+labFasWGHUrFnTZt1Tp07N8H2w/j8XLVrUsFgsRkBAgPHZZ58ZV69etSn777//2nyWfXx8XNoi88orr9jE3bJlS2Pbtm02ZcLDw40XX3zRpvW6Vq1amf6v7fVIyQuOtoicPn3a5n9bvnx5Y9asWek+6+Hh4cZLL71ks+2vvvpqhvWGh4cbVapUsdn2Bx54wG4PntjYWOO3334z+vTpYwwbNsxufdYtJ6nxNmzY0Fi1apVNuaSkJOOLL74wvL29zfJdu3bN5J2ybT1NrbtDhw7p/v9xcXHGq6++6vDxI23dmbXijBw50qZsjx49jCNHjtiUOX78uHn8Sj3OZVV3brf4v/zyy+mONYcOHUpXbsWKFTYt0iVKlDDOnTuXrlxiYqJNS3SLFi2MzZs32113TEyM8ccffxj333+/MXr0aIe2LSsvvPCC0aFDB2PSpEnGqVOn7JY5evSo8eijj9ps9+eff55pvUePHjVCQkLM8qk9Z+wdlyMiIozZs2cbnTp1MsaOHWu3Put1p7betm7dOt17lZiYaPP/W7Jkic0+XLZsWWPevHk238EJCQnGd999Z9MqHBAQYOzdu9duLJ9//rlZrkCBAsb48eONqKgou2X3799vjB8/3qhWrZoRHR2d7vVnn33WrKtQoULGd999Z8TFxaUrl5ycbGzbts144403jPr169tdV3bMnTvX5j393//+53Sd9mRn36pUqZIxfPhwY9GiRem+Dw3j+nuwePFim154hQoVMs6ePWu3PlfuW6GhoYavr69ZV8+ePY19+/bZrSs8PNz48ccfjZ49e+bbFv/c2u+tv7dSj9nlypUz5s+fb7PPxcbGGp988olN74lSpUoZly9fzrDu7du32/QSrVOnjvHnn3+m6w159uzZdL1rJ0+enGG9jrT4u2s/vZGQ+MOGdZIdGBiYK92WDcNI9yN1woQJGZaNjY01WrdubfPD5PDhw3bLpu3KbLFYjEWLFmVY93fffWdT3tfX1yhZsmSGyahhGMaAAQPM8lWrVs2wXNrESpJxzz33ZHi5QnJysnHPPffYlM+oy25ycrLdL1x7Ll68aFSsWNGs89tvv82wrL2u4JkdiDNaNqMvvHfeeccs06ZNG4fqNQwjw8/hoEGDzPq8vLyM+fPnZ1jHpUuXbJL/oKAgIyIiwm7ZtCd4ChQoYKxduzbDug8fPmzzxZjZCaHsOHDggM0lEe3atbP7JZbK+oeulPlJsvye+Pfo0cMsV7duXePSpUuZ1jtp0iSb/9eZM2fslhs6dKjNdjv6g89eMmAYtj+gJBnNmzfPsKxh2O4DFosl05NEaffFXr16ZXopw8MPP2zzoyezEz/W9WaUnO/evdvm83fnnXdmuC/aO35lVnduJv4bN260SSK//vrrTOuNiIiwSf6ffPLJdGXWrl1r896mPUmdEVd9h2b2mUpr3LhxZqwVK1bMNAbr/dHHx8f4448/nIon7f+/c+fOWXY9TkpKstmPSpYsaRw9ejTD8hs3brS5fK99+/Z2y1l3SX/llVcc2q7k5GS7lzdWq1bNoe/QtNvlrJdeesnm/dywYYPTddqTncTf0c9iTEyMTVfvjP4Hrty3vv/+e7OuKlWq2G0UcbSunHB14p9b+33a763ixYunO6Fr7bfffrM5po4cOdJuueTkZKNOnTpmuY4dO2b6PWQYhjFmzBibODIq70ji76799EZC4g8b1q3D1atXz5V1xMXF2bTkdevWLctlDh8+bHMW9+mnn7ZbLm3imlErXar4+HibMQ0kGTNnzsx0mXXr1tmUz+gsdtrEqkiRIulazdO6dOmSERwc7HD8jvr222/NOnv06JFhubTvX/fu3R1ehyM/5K3PSr/88svZ3Qwb586dMwoUKGDWN3z48CyXWb16tc32ffrpp3bLpU38HfnRaH1CyJHPtCOsW1v9/Pwy/TGcyvp61pIlS2b4ozs/J/7bt2+3SUb27NnjUN3W2/7666+ne/348eOGj4+PWeahhx5yZlMMw7D9AeXl5WUcOHAg0/KRkZE2rSGzZ8/OsKz1/6dw4cKZtrIYhmEcPHjQZpn169c7VHdGyfmTTz5p84PcXku4tbTHL3cl/n369DHLDBgwwKG616xZYy4TGBiYrmX4hx9+MF+/9dZbHarTXZKTk43y5cub8WY0doF1wiW55oSldX0FChTI9CR6ql9++cVmucz2iVRvvfWWzTJpe8EYhmFzovevv/7K0faksv79kdX4Pq704IMP2mxnRic0nZWT8TMcsXTpUrPOWrVq2S3jyn3r/fffN+u67777nKorJ1yd+GeHo/u9YaRP/L/77rss67duaClcuLDdkxK//fabWSY4ONi4ePFilvUmJSXZ9JqcMmWK3XKOJP7u2k9vJIzqDxvW19IWKVIkV9axdu1am9HVn3/++SyXqV69uvr06WPOz5s3z6F1PfbYY5m+7ufnp0aNGpnzQUFBuv/++zNdplmzZvL29jbnHb01zqBBg1SsWLFMyxQvXtzmOrh58+bZvd4wu1q1amVOZ2eE8scff9zpdVsLCAgwp7dv3+5UXQsWLLAZ1deRz1H79u3VvHlzc97Rz9GTTz6ZZRnrsRMyG/U6O6zju+uuu1S1atUsl7F+H0JDQ7Vu3TqXxJKXrO8mcPvtt6cbqyAjQ4cONaeXLVuW7vWff/7ZHL3d29tbb775pnOBptGlSxfdcsstmZZJvTtIKkc/K/fff7+KFi2aaZmaNWvaXNfv7OfQ+vPXr18/lSlTJtPyaY9f7hAeHm6O0SJJL7zwgkPLtWvXTlWqVJF0fawG61HUJdtj16FDh7I9Wnle8vLy0q233mrOZ3TMnz17tjldpEgRPffccy6N4/bbbzfHU8iM9eesXLlyWX4HS9JTTz0lPz8/u3WkcuX3jSvryg7r32RS7v0uyy3Wvz0OHjxo964jrty3rOvatWuXOUbLzcDR/T6tkJAQDR48OMtyTz/9tDkdHR1t9y5C1t/dQ4YMUcmSJbOs19vbWw888IA5b++721Hu2k9vJAzuBxvx8fHmtPWXqitZ/6AKCAjIcCCltPr06aO5c+dKuj7IyOnTp1WhQoUMy/v6+qpp06ZZ1mv9Q7lp06YqUKBApuVTb6F26dIlSbK59WFmevbs6VC5O+64Q998840kKTIyUgcPHlStWrUyLJ+UlKRVq1bp33//1aFDhxQZGamrV6/anDCwHmAlPDxccXFxNgfIjLRv396hmB1l/aW0dOlSPfPMM3rttddUokSJbNdl/TmqUaNGlglXqj59+piD2m3dulWJiYmZ/s8rV67s0CB35cuXN6cdGfgqK6dOnbK5RVqvXr0cWq579+7y9/c39+UNGzaoS5cuTseTl1avXm1Od+vWzeHlGjZsaE7/+++/5kCa9upt1aqVQ0lJdrRp08ahcjn5rGSn7tQB65z5HJ46dcpmgM3bb7/doeWsj1/usHbtWnOwpqJFi6pJkyYOL9uwYUMdP35c0vVjQ/fu3c3XmjdvLovFIsMwdOXKFfXp00dffPGF6tSp49oNcMCpU6e0YsUK7dq1SxcvXlR0dHS629vt3r3bnM7odlnW+8Odd97p8EC1jnL0+8P6WN6zZ095eWXdJhUSEqJ27dqZSULawSel6983O3fulCS99dZbKlasmIYMGSJ/f3+H4kpb19KlSyVJI0eOlI+Pj+6++26bRoDcYP2bTMq932U5YRiGNmzYoH/++UcHDhxQRESEYmJiMhwszTAMnTt3TsHBwTbPu3Lfsv6NsX//fg0aNEjjx493+4CIruCq/T6trl27ytfXN8tyLVq0MAd3lq7f+vOuu+4yXzcMQ2vXrjXnc/rd7egtnO1x1356IyHxh42QkBCFhoZKksP3g84u61HQ69at6/AO2aBBA5v5I0eOZJr4FytWzKHR/wMDA81pR0fBtl7GesTVzNSvXz9H5Q4dOmQ38U9MTNT//vc/ffTRR+ZJCEdFRERkmfgHBwdn2UMhu+6++25Vq1ZNR48elSR98cUX+vrrr9WuXTt16tRJLVu2VMuWLTO8m4E168+R9ZdGVqw/R/Hx8Tpz5ozZ0mePo58J6x/Njn4mMmO9fZLj2+jj46M6deqYI8vbGz07PzMMw2YU7qlTp2rhwoUOLWt9guvatWuKioqy+ZFp3TvHuueHq+TmZyWvP4epIzynyunxK6/t2rXLnL527Zp69Ojh8LLWP5rTHlPLlSunQYMGadasWZKu3xe+bt26atCggbp27ao2bdqoZcuWKlu2rJNbkLE9e/boueee07Jly7LVE8zeCaCUlBSb+37nxv5QrVq1LMskJSWZJ1uk7B/LUxN/e8e5p59+WjNmzFBCQoISEhL02GOP6YUXXlC3bt3Uvn17tWzZUk2aNMnyZL90vefI33//LcMwFB4erv79+6t48eK67bbb1K5dO916661q0KCBQyctsiMkJMRmPjIy0uXfy9llGIamTZumt956y2ZkfEfY+yy6ct9q1aqV2rZta/Z0+/nnnzVnzhy1aNFCXbp0UatWrdSqVSu3v4fZ4cr93p7sHLPr16+vlStXSkr/HXHmzBmbO4OMGzdOX331lUP1Wi+X3d+z1ty1n95ISPxho1ixYmbin/bWPq5i3UKenZbetGWzaml35AymK5Zx9EDs6BdN2nL2tjMuLk69e/fOcZeohISELMsEBQXlqO7M+Pn5adGiRerVq5f5pZGUlKSVK1eaXybe3t5q3ry5+vfvr2HDhmXYtdGVn6PMEn93tbCk/b/ndBsd7ZGSX0RGRprd8SXbW0flpC7rxN/6mOZIF8TsyslnxdHjR27WbU/az01Oj195zbprdExMjJYsWZKjeuyd+J44caKuXLlicyJq165d2rVrlz799FNJ1y+3uOuuu/TII4+oevXqOVq3PQsXLtQ999zj0LE7LXvLRERE2LTM5sb+4Mh3SNrkxJXHubp162rOnDl64IEHzNsUR0dHa968eealAQULFlSnTp00cOBA3XvvvRk2FnTr1k3ffPONnn76abOVNSwsTLNnzzYvmShSpIi6d++uIUOGqGfPni65FXLa/Sk8PNyt+1hKSoqGDRuW7jaZjsro8+vKfeuXX35Rnz599M8//0i6fhz8559/zHmLxaIGDRqoX79+euihh3L1ZJ2zXL3f25Odz5N12bT7XNrLUtJeLuUoZxod3bWf3khu3lMesMv6DP25c+dsrsV3FeuDUXYS7bQ/fNN2gcvvHN3WtNtp7+D9yiuv2CT9TZo00eeff66NGzfq3Llzunr1qpKTk2VcH8DTpkXFUbl1RrRmzZratWuXJkyYoMaNG6d7PTk5WZs2bdJzzz2nypUr67vvvrNbj6d/jtL+33O6jfl1+zLiymun03Y5zYtLmTxF2i6kOT1+5TVXfX7sdVcuWLCgFixYoPnz56tHjx52W4oPHTqk8ePHq1atWnrmmWdy9IM9rbNnz+q+++4z6woMDNTjjz+u3377Tfv371dERIQSEhLM471hGDbjXdiTF13IHfkOye3jXO/evXXkyBGNGTPGbg/Bq1evasGCBRo4cKBq166dabLy2GOP6cCBAxo5cqTdExQRERGaM2eO7rzzTjVv3lwHDhxweFsykrbXhHWPFnf48ssvbZL+GjVqaNy4cVqzZo1OnTqlmJgYm98ejp58dOW+VapUKa1fv17ff/+92rZtm+5zaBiGdu7cqddff13VqlXTe++955KxlFwtN/Z7e3K6z6V9/1117HX2f+GO/fRGQuIPG9YDlEnSpk2bXL4O6xbc1LPwjoiKisqwnhuBo9uadjvTXg935coVm+5Tjz32mLZu3aqnn35aLVu2VJkyZRQYGGjzZZed9zkv+Pn5acSIEdq2bZsuXLiguXPn6tlnn1WTJk1szr5GRkZq+PDhmjRpUro6PP1zlDaunG5jft2+jKSNd86cOTY/bLLzqFy5sk1d1t1mc+tSJk+RtrU2p8cvV3JkoC7rz0+dOnVy/NmxHqQqrV69eumvv/7SlStXtHTpUo0dO1adO3e2+VGcnJysL774Qg8++KATW3zdZ599Zv6oDg4O1pYtW/TNN9+ob9++qlWrloKDg9P9eM/q/2WvC7k75MVxrmTJknr//fd16tQpHTx4UFOmTNGwYcPS9fQ6cuSIunbtmumgYFWqVNGXX36pixcvaufOnfr66681YMCAdANf/vvvv2rfvr1Onz7t8PbYk/Y3WU5bUV0hJSVF77//vjl/5513as+ePRo9erTatWunChUqqGDBgk799nDVvpU6YNzatWt1+fJlzZ8/X6NHj1br1q1tLi+Nj4/Xa6+9pldeeSVbceaF3Njv7cnpPpf2t2nafXDz5s05Pv46K6/30xsJiT9sdO7c2WbeetRfV7HuUph6rbcj0pbNja6JucnRVvdjx47ZzJcqVcpmfvny5eZo9oGBgfrkk0+y7KqUnw9qpUqVUr9+/fTpp5/q33//1ZkzZzR27FibAZhGjx5tc/225Pmfo7Rx5XQb8+v2ZaRgwYIqVKiQOX/x4kWX1W39pW99fTPSS3vcyenxKyPWP1it786RGUcuW7EeCyH1srXcUrBgQXXr1k1vvvmmli9frrCwME2dOtVm8MaffvrJ6WRt8eLF5vQzzzzj0KBnWR3zAwICbH64u2t/KFiwoM2YObl9nKtZs6YeeughTZ06VceOHdOOHTs0ZMgQ8/XY2FiNHj06y3pSu4s/8cQT+uGHH3T27FmtX79ed9xxh1nm0qVLevfddx3eHnsaN25sk1D9/PPPbhupftu2bTb71BdffJFla3FOf3u4ct8qUqSIevXqpXHjxmn9+vUKDQ3V//73P5uTXx9//HG++52UG/u9PdnpEWp9fE/7HZF2HBpXfnfnVF7tpzcSEn/YaNKkiVq0aGHOz5s3z+UHQ+uR9k+ePGkzcnlmrEft9fPzc/sgUtnlaO+J1OvQpOsHrbTd4U+ePGlO16lTx6GRmG+kW7qVLVtWb775pr788kvzuYiIiHTvn/XnaNu2bQ53qbX+HFWqVEnFixd3MuLcUb9+fZsfVfZGrbYn9Y4XqZo1a+by2HJb69atzWlXtnBZ17tmzZoMR5/G9YHTrLvbWh+XMuNoOeseBVeuXHGolcd68L2MWP+Pw8LC0g2SmZsKFSqkYcOGaenSpTbvnfUP+JywPuZbfz9nJCYmxhzNPjPW71XqGCvuYH0sd/Q4l7ZsTo9zDRs21IwZMzRo0CDzuZUrV2b7Eg2LxaLWrVtr/vz5NnczcPZ/7+3treHDh5vzZ86c0S+//OJUnTll/TksXrx4pmPjpHLVbw9X7ltFixbVM888ozlz5pjPJSUlOXUbudyQW/t9Wo7+Nk29y1SqtHfNKlasmGrWrGnOu7N3SkZyaz+9kZD4I50xY8aY09euXdODDz6Y4643UVFRNreEktJ3Xfv++++zrCclJcUc8VWSWrZs6fZrSbPrhx9+yHa5Ro0apetO5WjrmHX5nA7E40733HOPzXzaz1HHjh3N6ejoaP3+++9Z1nn16lX9+uuvduvIb/z9/W1uTTR79myH9sMZM2bYzKfd324E1reO++OPP9INGpRT1iO8nzt3zqHPzM3K39/f5sfmjz/+6NByjh7nrG+vFRcX59DdJ+zdqz2t5s2b2wxANWXKFIficaXatWurdu3a5nzaY1d2ZfeY//3336cbo8Ee6/3hn3/+Me8Ektesj8MrVqzQ2bNns1xm/fr1Ni3+zh7Lrb9vEhMTczy4sZeXl/r27WvOO/u/l6TnnnvO5vfOM888k+ORzw3DyPHJsOx+DiXX73+u3Le6du1q8/vKFf8rV8qt/T6tLVu2OPSZ+Pnnn20G3rX328L6u/v777/P0WcmL+TGfnqjIPFHOn369FGfPn3M+RUrVuiRRx6x2eEdsXv3brsDZ1SrVs3mvuLjx4/PskvQpEmTbM40PvbYY9mKJT9Yv369/vjjj0zL/PHHHzatGA8//HC6MtYj0O7evTvL7q/vvPOOQz+k8kJ2TiDFxMTYzBctWtRmvlOnTqpRo4Y5/9prr6W7HCCtt99+22YU6fz+OXr00UfN6X379mU40GGq8+fP66OPPjLnu3fv7lCrTH7z8MMPm//vq1evasSIES6p984777T5zIwaNSrX7l7iCYYNG2ZOb9++XT///HOm5dMevzJTu3Ztmy7eWdW9du1ah27r6OPjo2effdac/+KLL1yS0Gb35Lf18SvtsSu7rI/5a9asybTsxYsX9cYbbzhUb9q7pgwfPtwlgxFm18MPP2xed52cnKyXXnop0/LJycl68cUXzfmQkBD1798/XTlnvm+su4G7838vXe9CPW7cOHP+4sWL6tmzZ7YvZbly5Yp69eqV40s4rT+HYWFh2r9/f6blp06dqs2bN2dZryvf3+zUlZCQYJOYuuJ/5Uq5td/bk9U+FxMTo3feececb9asWbpbbEvXv1NTe2ScPn1ar732Wo5jyi5376c3DAOwIyIiwqhevbohyXy0b9/e2LJlS5bLnj592nj88ceNAgUKGJKMlStXpiuzevVqw8vLy6y7SZMmxvnz5+3W99tvvxm+vr5m2Xr16hnXrl2zW3batGlmuUqVKjm0rUOHDjWXGTp0qEPLVKpUyVxm2rRpdsusXLnS5v2TZAQHBxsbN260W37Dhg1GcHCwWbZChQpGdHR0unInT540LBaLWW7gwIFGUlJSunIpKSnGp59+alM29XH8+HG7MeTk/cvOsv379zfeeust4+zZs5nWlZiYaNx///1mff7+/kZkZGS6ct9//73NdvXo0cNuOcMwjK+++srmvbjtttsyXP+bb75pluvQoUOmsaZK+/92hYSEBKN27dpmnX5+fsb8+fPtlj137pzRsGFDs6yXl5exbt26PI3XER06dHDoMzZx4kSb+O677z4jPDw8y/q3bNliDB482Jg9e7bd13///XebeuvVq2ccOnQo0zp37NiRYX2OHAvScvSYYx2nveOoPdbv75tvvulU3bGxsUbVqlXNcoUKFTLWrFljt2za45cjcQ8cONAsFxQUZOzZs8duuc2bNxslS5Z0+DgWFRVl1KhRwyxXrFgx488//8wwjlQRERHGN998Y3Tr1i3da+PGjTMeeeQRY8eOHVnW8+WXX9rEuWTJkiyXyczDDz9s1lWwYMEMv4dPnjxp1K9fP937lNln7H//+59N2Y4dOxrnzp3LNJ7Vq1dn+H7m5DNrGLbbKMl47bXXjOTk5HTl4uPjjcGDB9uUHTdunN0669ata0yZMsXu96i10NBQm+Nsq1atbF4/fvy40bJlS2Pu3LlGfHx8pnUdOXLEKFWqlFnXgAEDsthyx/Xv399muytXrmzMmTPHSElJyXS5q1evGuPHjzeKFSuW4XHh+PHjWe5bcXFxRsGCBW0+K7GxsXbX+dNPP9n8bsvsM+HKfeupp54ynnvuOePw4cNZ1vXCCy/Y1HXw4MEsl8mKo99vjsjN/d76eyv1MWbMGLv7XGRkpNG1a1ebsr///nuGdY8ePdqm7KhRozL8nKRKTk42VqxYYfTu3dtYvXq13TJZfbfll/00v7N/w1Lc9IKDg7Vq1Sr17dtXW7dulXT9jGPz5s116623qnv37qpbt66KFy8uLy8vXbx4UUeOHNHixYu1adOmLAefad++vV588UWNHz9e0vVrtOvUqaNHHnlE7dq1U1BQkE6fPq1ffvnFppU8ICBAs2fPtnurl/xuwIAB+vHHH9W2bVsNHjxYd9xxh0qWLKnQ0FAtXLhQ33//vfm+eXl56dtvv7UZ5CxVxYoVde+995rXp/3www/av3+/Hn30UdWuXVuJiYnav3+/vv/+e23ZskWS9Pjjj2vixIl5t7EZuHjxoubMmaOxY8eqdevWatu2rRo2bKgSJUooICBAV65c0c6dO/X999/b9PB47rnn7N4T+oEHHtCCBQvM1sLFixerTp06Gj58uFq0aKGAgAAdPXpUs2bN0qpVq8zlSpQooalTp+b69jrL19dXs2fPVuvWrRUfH6+EhAT17t1bffv21d13362KFSsqOjpaa9as0eTJk216M4wZM0Zt2rRxX/BOeuyxx7R9+3bzjg4///yzFi5cqPvuu0/t2rVTuXLl5Ofnp8jISJ06dUo7duzQ33//rRMnTkhKP1Bpqj59+ujFF180e0bs2bNHderUUb9+/XTbbbepQoUK8vHx0aVLl7R9+3YtWrRIO3bs0NChQzVw4MA82fb8IiAgQJMnT1b37t2VnJysmJgYdezYUYMGDVKvXr1UqlSpdMevgQMHOtzd/8UXXzQHK4uKilLLli01YsQIdezYUYGBgTpz5owWLlxolnnwwQczHW0/VeHChfXHH3+oTZs2unLlii5fvqxevXqpefPm6tOnjxo0aKCQkBDFx8fr8uXL2rt3r/755x+tWrVK165ds7kMIVV8fLy+++47fffdd7rlllvUtWtXNW7cWOXKlVPhwoUVGxurgwcP6rffftOKFSvM5dq2bauuXbs6/J7bM2rUKE2fPl3Jycm6evWq2rVrp0ceeUTdunVT0aJFFRoaquXLl2v69OmKjY1VhQoVVL9+fS1atCjLup955hlt2LDB/D5ZtWqVqlevrgEDBqhz584qW7asDMPQhQsX9O+//2r+/Pk6fPiw3nzzTd15551ObZe1zz77TKtXrzYv+Xj33Xf1119/6cEHH1Tt2rWVnJysnTt36rvvvtOhQ4fM5VJ/S9izd+9ePfzwwxoxYoS6deumW2+9VbVr11bRokXl7e2tixcvauPGjZo+fbpNz5+xY8emq2vTpk269957FRQUpNtuu03NmzdXzZo1zZ4BZ8+e1erVqzVr1ixzJHZfX1+XjhY/c+ZMBQQEmJdznThxQv3791f16tXVo0cPtWjRQiVKlFBgYKDCwsJ0+vRprVixQsuXL3fJrdb8/f01YsQIffjhh5Kuf1ZSB05r1KiRLBaLjhw5op9//lnLly+X5NhvD1fuW1FRUZoxY4Y+/fRTNW7cWB07dlSjRo1UunRpFSxYUNHR0dq7d69++ukn87etJA0cONDm+nRXOHnypM0gxY4YPHiwJk+eLCl393tr9913n+bMmaNx48Zp2bJlGjZsmG655RZdu3ZNW7du1aRJk3TmzBmz/IABA2x6Baf13nvvadeuXWYc//vf/zRr1izdf//9at26tUqXLi1vb29FRETo2LFj2r59u5YsWWL2/LXusZVd+WE/zffcfeYB+VtsbKzx5JNPmq332Xn4+/sbL7zwghEVFWW37pSUFOP55593uL6iRYsaGzZsyDTe/NziHxUVZTRt2jTL7fTy8jKmTp2a6frDwsJsWrQyezz00EMOnc03jNxv8bc+Y+voY/DgwUZiYmKG601ISLDpHZDVo1KlSsaBAwcy3Zb80uKfas2aNUaRIkUc3sbRo0e7Nd7MZKdFJCUlxXj33Xdtegc5+siq9f29996z2xsmo0dGxwZPbvFPNWvWLMPb2zvL96hp06ZGdHR0tur+6KOPHD6OHTt2zKHjWKrDhw8bdevWzfZnx97n0vqY4OijUaNGGfZky67PP//coXWWKFHC2Lp1a7a+15KSkownn3wyW9uW0WcrJ5/ZVGfPnrXbcpnRo1u3bpm25mf3/+Xt7W189dVX6epJ+/3pyCMgIMD47bffsrX9jpowYYLZep/dx913320cOXIky23MaN+Ki4szWrdu7dC6evToYSQkJGT5mXDlvmX9uXfV5yg7cvIbx/qRdl/Nrf0+7ffWe++95/D/NKvWdMMwjGvXrhkjRozI0XuQ0XHDkRb//LSf5lck/nDIiRMnjJEjRxqVK1fOckeqX7++8cEHH2TZXTDVsmXLjJYtW2ZYX2BgoPHYY48ZFy5cyLKu/Jz4G8b1borPPPOMUahQIbvb2qRJkwwvBUjr8uXLxqBBgzL8QV61alVj+vTphmE4/qWe24n/X3/9ZQwePNgoU6ZMlp+j5s2bG7/++qvD658zZ06mP/KLFClijB49OsNLAazlt8TfMK535R8+fLgREBCQ4Ta2atXK4R/bN0Lin2rv3r3GoEGDjMDAwEw/MyEhIUa/fv2MX3/91UhISMiy3h07dhh9+vTJ9MSmv7+/0bt3b2Pt2rV267gZEn/DuN7dvnnz5nbfo0KFChmjRo0yfxBmt+7Zs2cbZcuWtVt3+fLlje+++84wDMePY9YSEhKMiRMnGrVq1cr0s2OxWIxGjRoZb7zxht1uwnv27DFGjRpl1KpVK8sTRpUrVzY+/PBDhz6D2fHLL79k+B3s6+tr3HvvvWYylJPvtVWrVhmdOnXK9ERb4cKFjYEDBxq7d++2W0dOPrPW4uLijHHjxtm9tCP1UaVKFWPy5Ml2L3Gz9tVXXxk9e/Y0goKCMv1/FShQwOjdu7exfft2u/XExsYaH3/8sdG5c+dMj7/S9URi0KBBxtGjR7O97dkRFRVlvP/++0ajRo2y/DyWLVvWePbZZ42dO3dmWF929q24uDjj6aefNvz8/Oyur0yZMsZHH31kXoKQ1WfClfvWhg0bjEcffdSoUqVKpvVIMurUqWNMnjzZbvf2nHJ14m8YubPf2/ve+u2339Jd4pv6CAkJMT766KMs97m0Nm7caPTq1cvuZR/Wj9KlSxtDhgwxFi9enOE6svpuy4/7aX5kMYwcDteOm9aRI0e0e/duhYWF6fLly/L29laRIkVUqVIlNWvWLMeDZJw7d07r1q3T+fPndfXqVRUrVkzVqlVT27Zts91dKj9YtWqVOnXqZM5b72pXr17VypUrderUKUVFRalkyZJq2bKlQ/dpTev8+fNavXq1eQu30qVLq3bt2vn+Nm6nTp3S3r17dfLkSUVERCg5OVmFCxdWpUqV1KRJE1WoUCFH9R49elSbNm1SaGioEhISVLx4cd1yyy1q1aqVfHxu/Kub4uPjtXbtWh07dkyXL19WwYIFVaZMGbVt29ZmMCBPdO3aNf3zzz86cuSIwsLClJiYqEKFCqlcuXKqVauWateuLS+v7I9ZGxMTo7Vr1+rUqVO6fPmyfHx8VKxYMd1yyy1q1qzZDXn8yS379+/Xpk2bdPHiRQUFBalixYrq1KmTzW1FLRaLOb1y5UqHRlxPSkrSxo0btXv3bkVERKhEiRKqWbOm2rVrl6P/qT2nT582jw0RERHy8/NTSEiIqlevrvr16zv83ZV6SVLqPpiQkKDAwECVLl1aDRs2VJ06dWzeA1dKTk7Wpk2btGPHDkVERCgkJETlypVThw4dbAbqc0Z4eLjWrl2rc+fOKTw8XL6+vipZsqRq166txo0b58mldoZhaOvWrdqzZ48uXbokLy8vlSxZUk2aNFG9evWyVVdKSor279+vgwcP6syZM4qJiZHFYlGRIkVUo0YNNWvWzOH3LjExUXv27NHhw4d17tw5xcTEqECBAgoJCVGtWrXUtGlTh26x60phYWHavHmzLl68qMuXLysxMVFFihRR6dKl1aRJE7uXrrhCeHi4Vq1apePHjysxMVGlS5dW9erV1bp16xzvs67cty5evKhdu3bpxIkTCg8Pt/m+aNy4sapXr56jGN0hL/b7VJs3b9a+fft04cIFFS5cWDVq1FCnTp2c2u9jY2O1YcMGnThxQpcvX1ZKSooKFy6sihUrqnbt2jaD7rpCftxP8wsSfyCXZJb4A4CnykniDwAAche38wMAAAAAwIOR+AMAAAAA4MFI/AEAAAAA8GAk/gAAAAAAeDASfwAAAAAAPBij+rtASkqKzp07p8KFC+faLXwAAAAAAEhlGIaio6NVtmzZLG+leePf1DofOHfuXI7vOQ4AAAAAQE6dPn1a5cuXz7QMib8LFC5cWNL1NzwoKMjN0QAAAAAAPF1UVJQqVKhg5qOZIfF3gdTu/UFBQST+AAAAAIA848jl5gzuBwAAAACAByPxBwAAAADAg5H4AwAAAADgwUj8AQAAAADwYCT+AAAAAAB4MBJ/AAAAAAA8GIk/AAAAAAAejMQfAAAAAAAPRuIPAAAAAIAHI/EHAAAAAMCDkfgDAAAAAODBSPwBAAAAAPBgJP4AAAAAAHgwEn8AAAAAADwYiT8AAAAAAB6MxB8AAAAAAA9G4g8AAAAAgAcj8QcAAAAAwIOR+AMAAAAA4MFI/AEAAAAA8GAk/gAAAAAAeDASfwAAAAAAPBiJPwAAAAAAHozEHwAAAAAAD0biDwAAAACAByPxBwAAAADAg5H4AwAAAADgwUj8AQAAAADwYD7uDsCTfDtqtQJ8C7o7DAAAAAAwjZjY2d0hwM1o8QcAAAAAwIOR+AMAAAAA4MFI/AEAAAAA8GAk/gAAAAAAeDASfwAAAAAAPBiJPwAAAAAAHozEHwAAAAAAD0biDwAAAACAByPxBwAAAADAg/m4OwBP0mHt8yrk7e3uMAAAAACH1D6w390hAMgDtPgDAAAAAODBSPwBAAAAAPBgHp/4x8bGauLEierbt68qV66sggULqnDhwrrlllvUv39//fLLL0pMTHR3mAAAAAAA5AqPvsZ/zpw5GjlypC5dupTutUOHDunQoUOaO3euGjZsqOnTp6tRo0Z5HyQAAAAAALnIY1v8x48fr/vuu88m6S9atKhat26t9u3bq3Tp0ubzO3fuVLt27fTvv/+6I1QAAAAAAHKNRyb+CxYs0JgxY8z54sWL6+eff1ZoaKjWr1+v1atX69y5c1qyZImqVKkiSYqJiVHPnj0VFhbmrrABAAAAAHA5j0v8k5KS9Nxzz8kwDElSwYIFtXr1avXv31/eVrfas1gs6t69u9avX68yZcpIkkJDQ/XWW2+5JW4AAAAAAHKDxyX+K1as0OHDh835l156SXXq1MmwfJkyZfThhx+a8xMnTtSFCxdyNUYAAAAAAPKKxyX+y5Yts5kfPHhwlsvce++98vf3l3S9x8Cvv/6aK7EBAAAAAJDXPG5U/+PHj5vThQsXNq/hz4yfn59q1aqlHTt2SJJ+++03jRgxItvrHvq8j7wDvLMuCAAAAOQHM+q7O4Is7R66290hADc8j2vxj4iIMKeDgoIcXs667Pbt210ZEgAAAAAAbuNxib91Ah8dHe3wclFRUeZ0eHi4zp8/79K4AAAAAABwB49L/MuXL29OR0VF2XT9z8i1a9d04MABm+eOHTuWYfmEhARFRUXZPAAAAAAAyI88LvFv3769zfzs2bOzXGbu3LmKj4+3eS6zZH7cuHEKDg42HxUqVMhZsAAAAAAA5DKPS/xvv/12lShRwpwfP358utZ8axcvXtTLL7+c7vmYmJgMlxkzZowiIyPNx+nTp50LGgAAAACAXOJxiX9gYKBef/11cz4mJkbt27fXL7/8opSUFJuyy5YtU5s2bXT27Nl09SQmJma4Dj8/PwUFBdk8AAAAAADIjzwu8Zekp556Sg888IA5f+nSJd17770qWbKk2rVrp44dO6ps2bLq1q2bjh49Kh8fH40aNcqmjiJFiuRt0AAAAAAA5AKPTPwlaebMmXrzzTfl6+trPnf58mWtW7dOq1evNkftL1mypObNm6c2bdrYLE/iDwAAAADwBBbDMAx3B5Gbzpw5o++++07Lly/X4cOHdeXKFQUFBalatWq6++679fDDD6tYsWL65ptv9OSTT5rLnT17VmXLlnVoHVFRUQoODlbk6MIK8rPk1qYAAAAAQPaNjXR3BMgFZh4aGZnl5ec+eRST25QvX15jx47V2LFjMy23Z88em2UcTfoBAAAAAMjPPLarf3Zt2rTJnG7VqpUbIwEAAAAAwHVI/CUdPnxY27ZtM+cHDx7sxmgAAAAAAHAdEn9J77//vjldsWJF9ezZ043RAAAAAADgOjd94j9//nzNmDHDnB8/fry8vb3dGBEAAAAAAK7jkYl/UlKSvv/+e127di3DMoZh6LvvvlP//v2VemODvn376v7778+rMAEAAAAAyHUeOap/UlKShgwZoqefflo9e/ZUq1atVKNGDQUGBury5cvavXu35s6dq927d5vLtG3bVjNnznRj1AAAAAAAuJ5HJv6pIiIi9MMPP+iHH37ItFy/fv00bdo0FSpUKI8iAwAAAAAgb3hkV38fHx/dcccdCgwMzLCMl5eX2rdvr7/++ktz584l6QcAAAAAeCSPbPH38fHRggULlJiYqN27d+vYsWMKDQ1VRESEQkJCVLZsWbVq1UolS5Z06XrrxU+Rl5HxyQYAAADkbyc+uMPdIQCAy3lk4p+qQIECatKkiZo0aeLuUAAAAAAAcAuP7OoPAAAAAACuI/EHAAAAAMCDkfgDAAAAAODBSPwBAAAAAPBgJP4AAAAAAHgwEn8AAAAAADwYiT8AAAAAAB6MxB8AAAAAAA/m4+4APMmet25TUFCQu8MAAAAAAMBEiz8AAAAAAB6MxB8AAAAAAA9G4g8AAAAAgAcj8QcAAAAAwIOR+AMAAAAA4MFI/AEAAAAA8GAk/gAAAAAAeDASfwAAAAAAPBiJPwAAAAAAHozEHwAAAAAAD0biDwAAAACAByPxBwAAAADAg5H4AwAAAADgwUj8AQAAAADwYCT+AAAAAAB4MB93B+BJvh21WgG+Bd0dBgAAwE1hxMTO7g4BAG4ItPgDAAAAAODBSPwBAAAAAPBgJP4AAAAAAHgwEn8AAAAAADwYiT8AAAAAAB6MxB8AAAAAAA9G4g8AAAAAgAcj8QcAAAAAwIOR+AMAAAAA4MF83B2AJ+mw9nkV8vZ2dxgAAAA3jNoH9rs7BADweLT4AwAAAADgwUj8AQAAAADwYB6f+MfHx+uHH37QwIEDVatWLRUpUkQ+Pj4qUqSIatWqpQEDBmjWrFmKi4tzd6gAAAAAALicxTAMw91B5JaFCxfq8ccf15kzZ7IsW7ZsWX3zzTfq3bt3ttcTFRWl4OBgba5eg2v8AQAAsoFr/AEgZ1Lz0MjISAUFBWVa1mNb/GfNmqXevXvbJP0BAQFq1qyZunTpoqZNm8rf39987dy5c+rbt6+mTZvmjnABAAAAAMgVHpn4nzp1So899phSUlIkXU/4P//8c12+fFlbtmzRsmXLtHXrVl2+fFmffvqpeQLAMAyNGDFCx44dc2f4AAAAAAC4jEcm/pMmTVJsbKw5//333+vpp59WQECATbnAwEA9++yzmjFjhvlcXFycJk2alGexAgAAAACQmzwy8V+9erU5XadOHd1zzz2Zlu/fv79q165tzq9bty7XYgMAAAAAIC/5uDuA3BAaGmpON2jQwKFlGjRooP37rw8uExYWlqP1Dn3eR94BDO4HAADgsBn13R3BDWX30N3uDgHADcgjW/wLFy5sTsfHxzu0jHW5kJAQl8cEAAAAAIA7eGTi36pVK3N6w4YNunbtWqblExIStGHDBnO+ffv2uRYbAAAAAAB5ySMT/yeffFI+PtevYggNDdWrr76aafkxY8bo0qVLkqRChQpp5MiRuR4jAAAAAAB5wSMT/zp16uizzz4z5z/++GPdcccdWrJkicLDw5WcnKzLly/rr7/+Uo8ePcyyhQsX1ty5c1WxYsVM609ISFBUVJTNAwAAAACA/MhiGIbh7iByy++//65nn31WJ06cyLSct7e3evbsqXHjxqlu3bpZ1jt27Fi99dZb6Z6v/U1tBvcDAABArmFwPwCpoqKiFBwcrMjISAUFBWVa1iNb/FP17dtXK1eu1B133JFpuS5duuiJJ55QnTp1HKp3zJgxioyMNB+nT592RbgAAAAAALicR97OT5KuXr2ql19+Wd9++60SExMlSb6+vqpbt65CQkIUHR2tvXv3KjY2VkuXLtXSpUvVvHlzzZo1SzVr1sy0bj8/P/n5+eXFZgAAAAAA4BSPbPG/evWqunbtqq+++kqJiYkKDAzUZ599pitXrmjbtm1avny5Nm/erIiICM2YMUPFixeXJG3ZskVt27bV4cOH3bwFAAAAAAC4hkcm/s8//7w2bdok6Xor/9KlSzVq1CgFBgbalCtQoICGDBmi9evXKyQkRJJ06dIlDR48OM9jBgAAAAAgN3jc4H7nz59XxYoVlZSUJEkaMWKEJkyYkOVyH330kV566SVzfvXq1Wrfvr1D6zQHVRhdWEF+lpwFDgAAAADuNDbS3REgG27qwf2WL19uJv2SdNdddzm0XN++fW3mV6xY4cqwAAAAAABwC49L/NOOsF+hQgWHlqtYsaLN/Pnz510WEwAAAAAA7uJxiX/a0fbj4uIcWi42NtZmPu14AAAAAAAA3Ig8LvEvW7aszfzmzZsdWi5tufLly7ssJgAAAAAA3MXjEv+OHTvKYvn/Afa++OILJSYmZrqMYRj6+OOPbZ7r3LlzrsQHAAAAAEBe8rjEv3Tp0urVq5c5v2fPHg0aNChdV/5USUlJeuqpp7Rs2TLzuRYtWqhx48a5HisAAAAAALnN427nJ0lHjhxRixYtdOXKFfO5cuXKadiwYbr11ltVpEgRRUdHa9u2bZo+fbqOHDlilgsICNDq1avVvHlzh9fH7fwAAAAA3PC4nd8NJTu38/PJo5jyVPXq1bV48WLdfffdOnv2rCTp7NmzevfddzNdrmjRopo9e3a2kn4AAAAAAPIzj+vqn6pFixbas2ePRo8erVKlSmVatkiRInr66ae1d+9e9ejRI48iBAAAAAAg93lkV/+0UlJStG/fPu3YsUNhYWGKiYlRYGCgihUrpgYNGqh+/fry8cl554fULhYVRs2Rlx+3AQQAAEDuOPHBHe4OAUA+cdN39U/Ly8tL9erVU7169dwdCgAAAAAAecpju/oDAAAAAAASfwAAAAAAPBqJPwAAAAAAHozEHwAAAAAAD0biDwAAAACAByPxBwAAAADAg5H4AwAAAADgwXzcHYAn2fPWbQoKCnJ3GAAAAAAAmGjxBwAAAADAg5H4AwAAAADgwUj8AQAAAADwYCT+AAAAAAB4MBJ/AAAAAAA8GIk/AAAAAAAejMQfAAAAAAAPRuIPAAAAAIAHI/EHAAAAAMCDkfgDAAAAAODBSPwBAAAAAPBgJP4AAAAAAHgwEn8AAAAAADwYiT8AAAAAAB6MxB8AAAAAAA/m4+4APMm3o1YrwLegu8MAAAAA4EIjJnZ2dwiAU2jxBwAAAADAg5H4AwAAAADgwUj8AQAAAADwYCT+AAAAAAB4MBJ/AAAAAAA8GIk/AAAAAAAejMQfAAAAAAAPRuIPAAAAAIAHI/EHAAAAAMCD+bg7AE/SYe3zKuTt7e4wAAAAALjQ/lrujuDGUfvAfneHADto8QcAAAAAwIOR+AMAAAAA4ME8MvG3WCxOPSpXruzuTQAAAAAAwCU8MvF3VvHixd0dAgAAAAAALuGRg/vddttt2Sp/4sQJHTx40Jx/4IEHXB0SAAAAAABu4ZGJ/+LFi7NVvk+fPmbi7+vrq8GDB+dGWAAAAAAA5Lmbvqv/hQsXtGjRInO+b9++KlasmBsjAgAAAADAdW76xH/mzJlKSkoy5x9++GE3RgMAAAAAgGvd9In/1KlTzemKFSuqa9eubowGAAAAAADX8shr/B21bt06m0H9hg0bJi+vnJ8LGfq8j7wDvF0RGgAAAADceGbUd3cEuWr30N3uDiFHbuoWf+vWfovFomHDhrkxGgAAAAAAXO+mTfyjo6M1Z84cc75r166qVKmSGyMCAAAAAMD1btrE/+eff9bVq1fNeQb1AwAAAAB4opv2Gn/rbv5FixZV3759HV42ISFBCQkJ5nxUVJQrQwMAAAAAwGVuyhb//fv3a+PGjeb8Aw88ID8/P4eXHzdunIKDg81HhQoVciNMAAAAAACcdlMm/tat/VL2u/mPGTNGkZGR5uP06dOuDA8AAAAAAJe56br6JyYmaubMmeZ8s2bN1KBBg2zV4efnl60eAgAAAAAAuMtN1+K/YMEChYaGmvOPPPKIG6MBAAAAACB33XSJv3U3/8DAQA0YMMCN0QAAAAAAkLtuqq7+586d019//WXO9+vXT0FBQS6rf9PJMwrys7isPgAAAAC4KY2NdHcEHuWmavGfMWOGkpOTzfnsDuoHAAAAAMCN5qZK/KdNm2ZO16hRQ+3bt3djNAAAAAAA5L6bJvFfs2aNDh8+bM4/9NBDbowGAAAAAIC8cdMk/taD+nl7e2vo0KFujAYAAAAAgLxxUyT+UVFRmjt3rjnfs2dPlSlTxo0RAQAAAACQN26KxP+nn35SbGysOc+gfgAAAACAm8VNkfhbd/MvVaqU7rjjDjdGAwAAAABA3vH4xH/v3r36559/zPmhQ4fKx8fHjREBAAAAAJB3PD4Drlu3rgzDyJN11YufIi8jME/WBQAAAHiiEx/QOxdwNY9v8QcAAAAA4GZG4g8AAAAAgAcj8QcAAAAAwIOR+AMAAAAA4MFI/AEAAAAA8GAk/gAAAAAAeDASfwAAAAAAPBiJPwAAAAAAHozEHwAAAAAAD+bj7gA8yZ63blNQUJC7wwAAAAAAwESLPwAAAAAAHozEHwAAAAAAD0biDwAAAACAByPxBwAAAADAg5H4AwAAAADgwUj8AQAAAADwYCT+AAAAAAB4MBJ/AAAAAAA8GIk/AAAAAAAejMQfAAAAAAAPRuIPAAAAAIAHI/EHAAAAAMCDkfgDAAAAAODBSPwBAAAAAPBgJP4AAAAAAHgwH3cH4Em+HbVaAb4F3R0GAGTLiImd3R0CAAAAchEt/gAAAAAAeDASfwAAAAAAPBiJPwAAAAAAHozEHwAAAAAAD0biDwAAAACAByPxBwAAAADAg5H4AwAAAADgwUj8AQAAAADwYCT+AAAAAAB4MB93B+BJOqx9XoW8vd0dBgBk0353BwAAAIBcRIs/AAAAAAAejMQfAAAAAAAPdtN09U9OTtaKFSs0b948bdy4URcuXNDly5cVEBCg8uXLq379+urQoYM6dOigunXrujtcAAAAAABcwmIYhuHuIHLbhg0b9OSTT2rnzp0OlU9MTJSPj+PnRKKiohQcHKzN1WtwjT+AG07tA1zjDwAAcKNJzUMjIyMVFBSUaVmPb/GfNGmSnnjiCVmf3wgKClLVqlVVtGhRxcbG6tixYwoNDXVjlAAAAAAA5A6PTvwnT55sk/S3bNlSb7/9tjp27KgCBQrYlD116pQWLVqkqVOnymKxuCNcAAAAAABczmO7+h84cECNGzdWfHy8JGnEiBH68ssvcyWpp6s/gBsZXf0BAABuPHT1l/TII4+YSX+3bt00YcIEN0cEAAAAAEDe88jEf9OmTVq/fr05/+WXX+bJeoc+7yPvAFr8AdxYdrs7AAAAAOQqL3cHkBsmTpxoTrdp00a33HKLG6MBAAAAAMB9PDLxX7x4sTndvXt3N0YCAAAAAIB7eVzif+LECV28eNGcb9iwoSTpwoULev/999W8eXMVL15c/v7+KleunLp27aoPP/xQly9fdlfIAAAAAADkGo+7xn/Hjh0282XLltWsWbM0cuRIRUZG2rx27tw5nTt3TsuXL9e7776rDz74QE8++WSW60hISFBCQoI5HxUV5ZLYAQAAAABwNYcS/zVr1uR2HKb27ds7tXxYWJjN/G+//aZx48aZ82XLllX16tWVmJioffv2mScDoqOjNWLECJ06dUoffPBBpusYN26c3nrrLafiBAAAAAAgL1gMwzCyKuTl5SWLxZL7wVgsSkpKcqqOjz/+WC+++GK652vWrKlvvvlGnTt3Np9LSkrS7Nmz9cwzz9j0Bvjll190zz33ZLgOey3+FSpUUO1vajOqP4Abzu6hjOsPAABwo4mKilJwcLAiIyMVFBSUadl8c42/YRhy4BxEluLj49M9V7FiRa1fv94m6ZckHx8fDR06VH///bd8fX3N50ePHq3k5OQM1+Hn56egoCCbBwAAAAAA+ZHDiX9qYp5bD1cpWLBguuc+/fRTFS9ePMNlmjdvrpEjR5rzR44c0cqVK10WEwAAAAAA7uJQ4p+SkpInj8xa2R1VuHBhm/ng4GD17ds3y+WGDRtmM79q1SqnYwEAAAAAwN08blT/EiVK2Mw3atRI3t5ZX3dft25d+fv7m5cKHD16NNvr3nTyjIL8cn8sBAA3ubGRWZcBAAAA/pNvrvF3lTp16tjMFytWzKHlLBaLihYtas6Hh4e7NC4AAAAAANzB4xL/atWqKSAgwJy3Hn0/K9YDA1rXAQAAAADAjcrjEn8vLy917NjRnD927JhDy125ckVXrlwx50uXLu3q0AAAAAAAyHO5co3/8ePHtXz5cm3dulWXLl1SZGSkDMPQ8uXLc2N16dx7773666+/JEkHDhzQiRMnVLly5UyXWbJkic3dBVq3bp2bIQIAAAAAkCcshgvvpbd3716NHj1af/31l00SbRiGLBaL3VH7mzVrpu3bt8tisWj79u2qX7++03FERUWpWrVqCgsLkyQ99NBDmjJlSoblExMT1bx5c+3cuVPS9W7+J0+eTDdQYGbrCw4OVuTowgzuByD3MbgfAADATc/MQyMjFRQUlGlZl3X1nzVrllq0aKFFixYpJSVFhmGYj8w8++yzZrnvv//eJbEEBQXpzTffNOenTp2qzz77zG7Za9euadiwYWbSL0lPPvmkw0k/AAAAAAD5mUta/BctWqQ+ffqYCb+Pj4/atWunhg0b6s8//9TRo0czbPGPj49X8eLFFRcXp1q1amnv3r3OhiNJSk5OVu/evbVo0SLzubZt22rw4MGqUaOGkpKStGPHDk2ePFmHDx82yzRt2lRr167N1uB+tPgDyFO0+AMAANz0stPi7/Q1/nFxcXr00UeVnJwsi8Wijh07aurUqeY19fv379fRo0czXN7f319du3bV/PnzdeDAAYWGhqpkyZLOhiVvb2/NnTtXffv21d9//y1JWrdundatW5fhMm3bttUvv/zCiP4AAAAAAI/hdFf/6dOn69y5c7JYLGrVqpWWLl2a5UB6aTVv3tyc3rNnj7MhmQIDA7VkyRJ9++23ql69eoblKlSooC+++EIrVqxQqVKlXLZ+AAAAAADczekW/4ULF5rTEyZMkI9P9qusXbu2OX3s2DF17tzZ2bBMFotFw4cP1/Dhw7V9+3bt3btX58+flySVKFFCTZs2Vb169WSxON9Fv178FHkZgU7XA8CznPjgDneHAAAAgJuY04n/7t27JUmVKlVSo0aNclRHSEiIOR0REeFsSBlq3LixGjdunGv1AwAAAACQ3zjd1f/SpUuyWCzZ7t5vrUCBAuZ0UlKSsyEBAAAAAID/OJ34+/r6SpISExNzXMfly5fNaevWfwAAAAAA4BynE/8SJUrIMAwdO3Ysx3X8+++/5nSZMmWcDQkAAAAAAPzH6cQ/9Zr5CxcuaOfOnTmq49dff5V0fSC+Nm3aOBsSAAAAAAD4j9OJ/+23325Ov/nmm9lefvr06dq/f78sFouaNm2qYsWKORsSAAAAAAD4j9OJ/4ABA1S2bFlJ0p9//qlXX33V4WX/+usvjRw50px/4YUXnA0HAAAAAABYcTrx9/f31/jx42UYhiTpgw8+ULt27fTHH38oJiYmXfnY2FgtX75cAwcOVO/evRUbGyuLxaJ27drp3nvvdTYcAAAAAABgxccVlQwaNEh79+7VBx98IIvFog0bNujuu++WJHl7e5vlChcurNjYWHM+9WRBpUqVNHfuXFeEAgAAAAAArFiM1OzbBSZOnKjnnntO8fHx/78Ci8WcTl2VxWIxp9u0aaNff/1VJUuWdFUYeS4qKkrBwcGKjIxUUFCQu8MBAAAAAHi47OShTnf1t/b444/rwIEDevrppxUSEiLperKf+khlGIYaNmyoWbNmac2aNTd00g8AAAAAQH7m0hZ/a4ZhaPfu3dq1a5cuX76sq1evqkiRIipdurRatWqlMmXK5MZq3YIWfwAAAABAXspOHuqSa/ztsVgsatCggRo0aJBbqwAAAAAAAFlwaVd/AAAAAACQv5D4AwAAAADgwXKtq39ERIQ2b96sQ4cO6cqVK0pISFBwcLBKlSqlpk2bqk6dOjYj/gMAAAAAANdzeeL/22+/6auvvtLKlSszLVe8eHENGzZMTz/9tMqWLevqMAAAAAAAgFzY1f/ChQvq0aOH+vXrp5UrV9rcws/e7fwuXbqkjz76SHXq1NGUKVNcFQYAAAAAALDikhb/06dPq3Pnzjp27JgMw7Dpwu/j46Pg4GD5+fkpKipK0dHRNstGRUXp0Ucf1YULF/Tqq6+6IhwAAAAAAPAfp1v8DcPQ/fffr6NHj5rPValSRePHj9eePXsUGxur0NBQnT59WpGRkbp06ZIWLFig+++/XxaLRRaLRYZh6I033tDixYudDQcAAAAAAFixGNZ98HPgxx9/1KBBg8xW/meeeUYffPCBfH19s1x227Zt6tu3r86ePSvDMHTLLbdo//79zoTjFlFRUQoODlZkZKSCgoLcHQ4AAAAAwMNlJw91usX/xx9/NKcfeughffrppw4l/ZLUpEkTLVu2TP7+/pKkQ4cOadu2bc6GBAAAAAAA/uN04r9jxw5Jkre3t95///1sL1+zZk09/PDD5vz27dudDQkAAAAAAPzH6cT/0qVLslgsqlevnkqUKJGjOrp27WpOh4WFORsSAAAAAAD4j9OJf7FixSRJISEhOa7DetmiRYs6GxIAAAAAAPiP07fzq1Klis6dO6czZ87kuA7rZatWrepsSG7z7ajVCvAt6O4wAAAAAAD52IiJnfN0fU63+Pfr10+SdOTIEe3duzdHdfz666+Srvce6NChg7MhAQAAAACA/zid+A8ZMkSlS5eWJD3++OO6du1atpafP3++5s2bJ4vFolGjRsnHx+lOCAAAAAAA4D9OJ/4hISH66aef5Ofnpw0bNqhr1646cuRIlssZhqEvv/xS9913nyTp9ttv1yuvvOJsOAAAAAAAwIpDzeunTp3K9PXKlSvr559/1iOPPKL169erbt26uu2229SzZ0/Vr19fxYoVk6+vr6Kjo3X8+HFt2rRJc+bM0cmTJyVJAwcO1DvvvKNTp06pYsWKzm8VAAAAAACQJFkMwzCyKuTl5SWLxeJQhanVZVXeXjmLxaKkpCSH1pOfREVFKTg4WB8Nm8/gfgAAAACATLlicL/UPDQyMlJBQUGZls3WBfVZnSOwWCzpEv60y6S+nlU5AAAAAADgPIcTf0cSc1eVAQAAAAAAruFQ4n/8+PHcjgMAAAAAAOQChxL/SpUq5XYcHqHD2udVyNvb3WEAAAAAAHJB7QP73R1Cjjh9Oz8AAAAAAJB/kfgDAAAAAODBSPwBAAAAAPBgHpn4nzhxwry1YHYepUuXdnfoAAAAAAC4lMO388sOwzC0b98+nTlzRhEREUpISHB42SFDhuRGSAAAAAAA3JRcmvjv2LFDn3zyiebPn6+YmJgc1ZEbiX/79u0VEBCQZbmiRYu6fN0AAAAAALiTyxL/8ePH6/XXX1dycrIMw8iwnMViyfB1i8XiqnBszJgxQ5UrV86VugEAAAAAyM9ckvhPnDhRY8aMsXmuWLFiunr1quLj42WxWFSxYkVFR0crIiLCTPxTE/0KFSrkWtIPAAAAAMDNzOnB/cLDw/Xiiy9Kup7I33HHHTp06JAuXbqkDh06mOWOHz+usLAwRUZGasmSJbr33nvN1+rXr6+dO3fq+PHjzoYDAAAAAACsON3iP3XqVF29elUWi0Xt27fXH3/8IS+vjM8nFCxYUN26dVO3bt20fPly9evXT4sWLVLXrl21YcMG+fjkyniDeWLo8z7yDvB2dxgAAAAAACu7h+52dwhu5XSL/8qVK83pDz74INOkP60uXbrot99+kyT9+++/Gjt2rLPhAAAAAAAAK04n/nv27JF0/Zr+W2+9NcNyGQ3o17FjR911110yDEMTJ05UcnKysyEBAAAAAID/uOQaf4vFopo1a6Z7zdv7/7u9x8XFZVhHnz59JElXrlzR6tWrnQ0pnZdfflkNGjRQkSJF5Ovrq1KlSqlJkyYaOXKklixZkuldCAAAAAAAuJE5fUH9tWvXJF2/dj+twoULm9OXLl1SpUqV7NZh/fyJEyecDSmdOXPm2MyHhoYqNDRU27dv11dffaW6detq8uTJatWqlUP1JSQkKCEhwZyPiopyabwAAAAAALiK0y3+wcHBkqSYmJh0rxUtWtScPnbsWIZ1pJ48kK6fIHC1YsWKqUWLFurSpYtuvfVWFS9e3Ob1vXv3qn379poyZYpD9Y0bN07BwcHmo0KFCi6PGQAAAAAAV3A68a9SpYoMw9C5c+fSvVanTh1zes2aNRnW8c8//5jTAQEBzoYki8WiZs2a6auvvtKxY8cUFhamf/75R8uWLdOmTZt06dIlbd26Vf369TOXSUpK0mOPPaYlS5ZkWf+YMWMUGRlpPk6fPu10zAAAAAAA5AanE//69etLkk6dOqXo6Gib19q0aWNOT5kyJd3r0vUxAr7++mtz3t5YAdlVqVIlbdmyRU8++aSqVKlit0zTpk01d+5cffHFF+ZzycnJGjlypBITEzOt38/PT0FBQTYPAAAAAADyI6cT//bt25vTy5Yts3mtUaNGql27tiTp7Nmz6t69u9asWaO4uDjFxcVpyZIl6tChg86fPy9JCgoKsqkvLzz11FN66KGHzPkjR45o/vz5eRoDAAAAAAC5xenE/7bbbpOX1/Vqfvnll3Svjxs3zhw1f/PmzerUqZMKFSqkQoUKqWfPntq3b5+k693zX3jhBQUGBjobUra9+uqrNvN//fVXnscAAAAAAEBucHpU/1KlSmnIkCHavXu3Tp48qdjYWJvkvXfv3nrllVf0/vvvy2Kx2Nw6z3r+nnvu0SuvvOJsODlStWpVVapUSSdPnpQkHThwIEf1bDp5RkF+FleGBgAAAABw1thgF9UT6Zp68pjTLf6SNHXqVG3ZskXr1q2z22L/7rvv6rffflOTJk1snjcMQ9WqVdPXX3+tOXPmyGJxX9JcpkwZczosLMxtcQAAAAAA4EpOt/g7qk+fPurTp4/CwsJ06tQpJSYmqmzZsvnmVnixsbHmtCvuLAAAAAAAQH6QZ4l/quLFi6t48eJ5vdpMJSQk6MiRI+Z86dKl3RgNAAAAAACuk+eJvz2zZ89WcnKyJGnIkCF5vv558+bZtPi3bds2z2MAAAAAACA3WAzr0fbcJCAgQNeuXZMk8wRAXrl48aJatGihU6dOSZK8vLy0Z88e8zaEjoiKilJwcLAiRxdmcD8AAAAA8FT5aHA/Mw+NjFRQUFCmZV0yuJ8ruOr8w8aNG/X444/r4MGDWZbdvXu3OnfubCb90vUeB9lJ+gEAAAAAyM/yRVd/V0pISNCkSZM0adIkNWzYUJ07d1aDBg1UunRpFS5cWDExMTpy5IiWLFmihQsXKiUlxVy2cePG+vzzz90YPQAAAAAAruVxib+1nTt3aufOnQ6V7dWrl6ZOnZplFwkAAAAAAG4kHpf4V65cWffdd5/WrFmj8+fPZ1rWy8tLXbp00ciRI9W7d2+n110vfoq8jECn6wEAAAAA5EOjF7qkmhMf3OGSehzlkYn/Tz/9JEk6d+6c9u3bp1OnTik8PFxxcXEKCAhQkSJFVL16dTVr1kyFChVyc8QAAAAAAOQej0v8rZUtW1Zly5Z1dxgAAAAAALhNvhnVHwAAAAAAuB6JPwAAAAAAHozEHwAAAAAAD0biDwAAAACAByPxBwAAAADAgzk8qv9DDz2Ua0EkJibmWt0AAAAAANzMHE78p0+fLovFkpuxAAAAAAAAF3M48UfW9rx1m4KCgtwdBgAAAAAApmwl/oZh5FYcAAAAAAAgFzic+KekpORmHAAAAAAAIBcwqj8AAAAAAB6MxB8AAAAAAA9G4g8AAAAAgAcj8QcAAAAAwIOR+AMAAAAA4MFI/AEAAAAA8GAk/gAAAAAAeDASfwAAAAAAPBiJPwAAAAAAHozEHwAAAAAAD0biDwAAAACAByPxBwAAAADAg5H4AwAAAADgwXzcHYAn+XbUagX4FnR3GAAAAADcbMTEzu4OATDR4g8AAAAAgAdzS4v/yZMndeHCBRUtWlQ1atRwRwgAAAAAANwUXNLif+DAAe3bt0/79u2TYRgZllu8eLFq166tqlWrqnXr1qpVq5YqVaqkqVOnuiIMAAAAAACQhtMt/gcOHFDdunUlSQ0bNtS2bdvslluwYIHuuusupaSk2JwcOH36tIYPH65jx47p3XffdTYcAAAAAABgxekW//nz55uJ/PDhw+2WiY+P16OPPqrk5GS7rxuGoXHjxmndunXOhgMAAAAAAKw4nfhv3rzZnL7jjjvslpk1a5YuXLggi8UiLy8vvfrqq9q2bZvWrFmj9u3bm+XeeustZ8MBAAAAAABWnO7qf/jwYUlSsWLFVLFiRbtlfvrpJ3P66aef1jvvvGPOL1q0SLVq1dKZM2e0atUqhYWFqXjx4s6GBQAAAAAA5IIW/7Nnz8pisahq1ap2X4+Pj7fpwj9ixAib1wMDAzV06FBJUkpKirZu3epsSAAAAAAA4D9OJ/4xMTGSpKCgILuvb9myRdeuXZPFYjFH9E+rWbNm5vSxY8ecDQkAAAAAAPzH6a7+qQP7JSUl2X1906ZN5nTHjh3tlrHu2h8VFeVsSG7TYe3zKuTt7e4wAAAAALjZ/lqura/2gf2urRA3Fadb/FNb+s+fP2/39VWrVpnTbdq0sVvm2rVrzoYBAAAAAADscDrxr1q1qgzD0JEjR3T58mWb16Kjo7Vy5Upzvl27dnbrsF4uODjY2ZAAAAAAAMB/nE78W7VqJen6wHwff/yxzWsTJkxQfHy8LBaLbrnlFpUvX95uHXv27DGnM7ozgKsNGjRIFovF5nHixIk8WTcAAAAAAHnF6Wv8Bw0apC+++EKS9OGHH+rs2bNq27attm/frsmTJ5vlUkfut2fjxo3mdK1aLr4Yxo758+frhx9+yPX1AAAAAADgbhYjdXQ+J9x///2aM2eOLBZLutcMw1CZMmV08OBBFSpUKN3r4eHhKl26tJKTk1WsWDGFhoY6G06mrly5orp169odk+D48eOqXLlytuuMiopScHCwNlevweB+AAAAAFyOwf2QVmoeGhkZmeFd9lI53dVfkqZOnaquXbvKMIx0j6JFi2revHl2k35J+v777807AnTq1MkV4WRq1KhRZtLfvXv3XF8fAAAAAADu5HRXf0kKDAzU0qVL9eeff2r+/Pk6ffq0AgICdOutt+qRRx6xuV1fWr///rsqVaokSbrnnntcEU6GFi1apJkzZ0qS7rjjDvXr109Lly7N1XUCAAAAAOBOLkn8U/Xq1Uu9evXK1jLWo/7npsjISD366KOSpMKFC+ubb77R8uXL82TdAAAAAAC4i0u6+t8InnvuOZ09e1aS9MEHH6hChQpujggAAAAAgNzn0hb//GrJkiWaOnWqJKlt27Z64okncmU9Q5/3kXcAg/sBAADAvXYP3e3uEADkI3me+CckJGjTpk26cOGCihYtqlatWmU48J8rREVFafjw4ZIkPz8/fffdd3bvPgAAAAAAgCdySeK/ePFipaSkyGKxqEePHhkm1pMnT9bo0aMVERFhPufv76/nn39eY8eOlZeX6688ePHFF3X69GlJ0htvvKFbbrnF5esAAAAAACC/cjrx//fff9WzZ09ZLBa1bdtWt99+u91yU6ZM0WOPPWbOp54ciIuL03vvvafTp09r2rRpzoZjY/ny5fr2228lSQ0bNtRLL73k0voBAAAAAMjvnG5iX7RokTn90EMP2S0TGRmpF154QdL/J/yGYcgwDHN65syZ+uOPP5wNxxQTE6NHHnlEkuTt7a3vvvtOPj6uubIhISFBUVFRNg8AAAAAAPIjpxP/LVu2mNMZtfZPmzZNkZGRslgs8vf31+TJkxUeHq5Tp07pvvvuk3Q9+f/ggw+cDcf08ssv68SJE5KkZ599Vs2aNXNZ3ePGjVNwcLD54A4BAAAAAID8yunE/+jRo5Kk0qVLq2TJknbLzJkzx5wePXq0Hn74YRUpUkTly5fXrFmzzOvuN2/ebN5yzxmrVq3SN998I0mqVq2a3n77bafrtDZmzBhFRkaaj9QxBAAAAAAAyG+cTvzPnTsni8WiSpUq2X09JibG7BVgsVjMEfZTeXt7a/Dgweb8tm3bnIonNjZWDz/8sHkZweTJkxUQEOBUnWn5+fkpKCjI5gEAAAAAQH7kdOIfGxsrSSpcuLDd1zdt2qTk5GRZLBY1btxYpUuXTlemYcOG5vTJkyedimf06NE6duyYJOmRRx5Rp06dnKoPAAAAAIAbmdOj3aUO1nft2jW7r2/atMmc7tChg90yISEh5nR0dHSOY9m3b58mTJggSSpTpow++uijHNeVE5tOnlGQn/1bGQIAAAB5Zmywi+uLdG19APKU04l/cHCwLl26lOF17itXrjSnW7dubbdMXFycOe3t7Z3jWEJDQ80u/ufPn7c5oeCIKlWqmNPBwcGKiIjIcSwAAAAAAOQHTnf1r1mzpiTp+PHjOnPmjM1rYWFhWrt2rTnfrl07u3VcunTJnC5SpIizIQEAAAAAgP843eLfrl07rV+/XpL05ptvasqUKeZr7777rpKSkszr+0uUKGG3jp07d5rTGQ0S6IgCBQqoWLFiDpdPSEhQTEyMOR8SEiIvr+vnQoKDXdw9CgAAAAAAN3A68R88eLA+/PBDGYah6dOn6/Dhw2rbtq22b9+upUuXmuUefvjhDOtYt26dOV23bt0cx9KmTRuFhYU5XH769OkaNmyYOb9t2zZVrlw5x+sHAAAAACC/cTrxr127tp588klNmDBBFotF69evN3sApKpZs6YeeeQRu8ufP39eGzZskMViUZkyZVS+fHlnQwIAAAAAAP9x+hp/Sfrf//6nhx9+WIZhpHtUq1ZN8+fPV4ECBewu+91335kD8nXt2tUV4QAAAAAAgP843eIvSV5eXpo8ebKeeuop/fnnnzp9+rQCAgJ066236u6775avr2+Gy+7du9e8zd/AgQNdEQ4AAAAAAPiPSxL/VA0aNFCDBg2ytcxPP/3kyhAAAAAAAIAVl3T1BwAAAAAA+ZPFSL3AHjkWFRWl4OBgVRg1R15+ge4OBwAAABk48cEd7g4BAFwiNQ+NjIxUUFBQpmVp8QcAAAAAwIOR+AMAAAAA4MFcOrifJC1fvlzz5s3T5s2bdebMGUVEROjatWsOLWuxWJSUlOTqkAAAAAAAuGm5LPE/fPiwBg4cqG3btpnPMXwAAAAAAADu5ZLE/8CBA2rVqpWioqLSJfsWi0VS+pMAqc/bew0AAAAAALiG04m/YRi69957FRkZKYvFoqCgII0cOVKtWrXS22+/rS1btshisWjlypWKjo7W6dOntW7dOv3++++Ki4uTxWLRE088of79+7tiewAAAAAAgBWnE/8//vhDe/fulcViUfHixbVu3TrVqFFDkjRhwgSzXIcOHczpJ554QuHh4Ro1apRmzZqliRMnqkyZMnrttdecDQcAAAAAAFhxelT/+fPnm9Pjxo0zk/6sFC1aVDNnztRzzz0nwzA0duxYLV261NlwAAAAAACAFacT/3/++UeS5Ovrq4EDB2Z7+fHjx6tWrVpKSUnRyy+/7Gw4AAAAAADAitNd/S9cuCCLxaJatWrJ39/f5jXrAfwSEhLk5+eXbnlvb28NGzZML7/8snbt2qUDBw6oVq1azoblFnveuk1BQUHuDgMAAAAAAJPTLf7R0dGSpOLFi6d7LSAgIF05exo2bGhO79ixw9mQAAAAAADAf5xO/AMDAyVJiYmJ6V6zbv0+ffp0hnVYnyA4f/68syEBAAAAAID/OJ34ly5dWoZhKDw8PN1r1apVM6f//fffDOs4duyYOZ2UlORsSAAAAAAA4D9OJ/61a9eWJB05ckQpKSk2rzVu3Nic/vHHHzOsY8aMGeZ0mTJlnA0JAAAAAAD8x+nEv1WrVpKuD963bds2m9e6du1qdvdftWqVPvzwQ5vXU1JS9Morr2jVqlXmc+3bt3c2JAAAAAAA8B+nE/8uXbqY0wsWLLB5zc/PT48//rgMw5AkjRkzRtWqVdN9992n++67T5UrV9b48eMlXb8DQM+ePVWxYkVnQwIAAAAAAP+xGKlZuROqVq2qEydOqFy5cjpx4oS8vb3N12JjY9W6dWvt2rVLFotFhmGYt/lLnTYMQyVLltTWrVtVvnx5Z8PJc1FRUQoODlZkZCS38wMAAAAA5Lrs5KFOt/hL0po1a7R7924tXrw43XX+gYGBWr58ufr06WO2/BuGYTPdokULrV+//oZM+gEAAAAAyM9c0uLvqL1792rx4sU6efKkEhMTVbZsWXXu3Flt2rTJqxByBS3+AAAAAIC8lJ081CePYpIk1a1bV3Xr1s3LVQIAAAAAcFNzSVd/AAAAAACQP5H4AwAAAADgwUj8AQAAAADwYCT+AAAAAAB4MIcG96tatWpuxyFJslgsOnr0aJ6sCwAAAACAm4FDif+JEydksVhyNRDDMHJ9Hbnt21GrFeBb0N1hAAAAAMBNZ8TEzu4OId9y+HZ+hmHkZhwAAAAAACAXOJT4T5s2LbfjAAAAAAAAucChxH/o0KG5HQcAAAAAAMgFjOoPAAAAAIAHI/EHAAAAAMCDkfgDAAAAAODBHB7VP9Xu3bt15coVSZK/v79atGiRoxVfvHhRBw8eNOebNm2qggW5FR4AAAAAAK6UrRb/c+fOqWXLlurUqZM6deqkHTt25HjFgYGBevTRR8263nnnnRzXBQAAAAAA7MtW4v/OO+8oLi5OknT//ffr0UcfzfGKCxcurLlz58rHx0eGYejLL7/UxYsXc1wfAAAAAABIz2IYhuFIwfj4eJUqVUrR0dHy8/PT4cOHVb58eacDeOKJJzRp0iRZLBaNHz9eL7zwgtN15rWoqCgFBwdrc/UaKuTt7e5wAAAAAOCmU/vAfneHkKdS89DIyEgFBQVlWtbhFv8FCxYoOjpaFotFDz74oEuSfkl67bXX5P1fsjxr1iyX1AkAAAAAAK5zOPFfs2aNOd2vXz+XBVCuXDm1bNlShmFo9+7dioiIcFndAAAAAADc7Bwe1X/r1q2SpICAAHXo0MGlQfTo0UPr168319O1a1en6jMMQ3v37tXGjRu1c+dO7d+/XydPnlRoaKhiY2MVEBCgkJAQ1alTR23bttUDDzygypUru2BLAAAAAADIXxxO/E+fPi1JqlChgnx8sn0XwExVr17dnD558qTT9U2dOlWPPPJIhq/HxMQoJiZGp0+f1pIlS/Tmm2/q0Ucf1fjx47O8NgIAAAAAgBuJwxn8lStXZLFYVKpUKZcHYV2nK7r6px2v0MfHR5UrV1apUqUUEBCg6OhoHTp0SFeuXJEkpaSkaOLEidq8ebOWLVumkJAQp2MAAAAAACA/cDjxT05OlnQ9SXY16zpT1+MMHx8ftWvXTr169VL79u3VpEkTFShQwKaMYRhas2aNXn75Zf3zzz+SpG3btumpp55ikEEAAAAAgMdweHC/4sWLyzAMXbp0yeVBhIWFmdPFihVzur4HH3xQa9as0Ysvvqhbb701XdIvSRaLRR06dNDq1avVtm1b8/kffvhBp06dcjoGAAAAAADyg2wl/tL1a/Dj4+NdGsTevXvTrSev+Pn56b333jPnDcPQihUr8jQGAAAAAAByi8Nd/Rs0aKCdO3cqISFBa9asUffu3V0WxNKlS83p+vXru6xeRzVr1sxm/vz58zmqZ+jzPvIO8HZFSAAAAADgVruH7nZ3CHARh1v8rW+xN2HCBJcF8O+//+qff/6RxWJR5cqVVbVqVZfV7ajExESbeUb2BwAAAAB4CocT/+7du6tAgQIyDEMLFy7UsmXLnF55cnKynn/+eXP+zjvvdLrOnFi5cqXNfJs2bdwSBwAAAAAAruZw4l+qVCkNGzZM0vXr4Pv3729zbX5OPPbYY1qzZo0kqUCBAnrhhRecqi8nLl68qBdffNGc79q1qxo1apTncQAAAAAAkBscTvwl6bXXXlPBggVlsVgUERGhVq1aacqUKdle6dmzZ9WtWzdNmzZN0vUR9ocPH64KFSpku67sMgxDMTEx2rlzpz744APVr19fR44ckSTVrFlTM2bMyLKOhIQERUVF2TwAAAAAAMiPspX4ly9f3kyMLRaLYmJi9Oijj+qWW27Rp59+qt27d8swDLvLhoeHa9GiRRo0aJCqV6+uFStWyDAMWSwWNWvWTB9//LHzW5OBBx98UBaLRRaLRV5eXipcuLAaNWqkMWPG6NKlSypUqJCee+45bd26VWXLls2yvnHjxik4ONh85MUJCwAAAAAAcsJiZJSpZ+KTTz7RSy+9JElmom+xWCRJgYGBKlu2rIoUKSI/Pz9FRkYqPDxc586dM5dPTfgNw1D16tW1evVqlSlTxhXbY9eDDz6YYUu+n5+fHnvsMT3xxBOqVauWQ/UlJCQoISHBnI+KilKFChVU+5vajOoPAAAAwCMwqn/+FhUVpeDgYEVGRmY5QH2OEn9JWrhwoYYMGaIrV66YSb91VfaeS30+9bk+ffpoxowZuT6K/ieffKK///5bkpSSkqLIyEgdOnRIERERNnE98cQT+uyzz+Tr65ut+lPfcBJ/AAAAAJ6CxD9/y5PEX7o+MN7HH3+sb7/9VtHR0f9f6X9JvzXr1bRo0UJjxoxRnz59crpqpxmGoY0bN+rtt9/WkiVLzOfvuusuzZs3L1t1kfgDAAAA8DQk/vlbniX+1iv8+++/tWrVKm3ZskWhoaG6fPmy4uPjFRISomLFiqlatWrq0KGDOnfurMaNGzu7Spd67rnn9Nlnn5nz06dP19ChQx1ensQfAAAAgKch8c/f8jzxv9ElJSWpXr16OnjwoCSpadOm2rp1q8PLm2/46MIK8kvf2wEAAAAAcAMaG+nuCDKUncQ/W6P6eyofHx/169fPnN++fbvi4uLcGBEAAAAAAK5B4v+fihUrmtMpKSm6cuWKG6MBAAAAAMA1SPz/Yz3CvySFhIS4JxAAAAAAAFyIxP8/q1evNqfLlCmjgIAAN0YDAAAAAIBrkPjretL/119/mfPuvM0gAAAAAACu5HGJ/969e/XQQw9p3759DpWfN2+eevfurdSbG/j7++uFF17IzRABAAAAAMgzPu4OwNUSExM1bdo0TZs2TbVr11aXLl3UoEEDlStXTkFBQUpMTNSlS5e0a9cu/fHHH9qzZ4+5rMVi0YQJE1StWjU3bgEAAAAAAK7jcYm/tf3792v//v0OlS1atKgmTJigAQMG5HJUAAAAAADkHY/r6l+5cmW9/vrratGihQoUKOBw+YMHD5L0AwAAAAA8jsVIvbjdA8XHx2vPnj06evSozp8/r5iYGBUoUEBBQUEqV66cGjVqpIoVKzq9nqioKAUHB6vCqDny8gt0QeQAAAAAAHc78cEd7g4hQ6l5aGRkpIKCgjIt69Fd/f39/dWsWTM1a9bM3aEAAAAAAOAWHp3434i8LFIRPy/5+1hksbg7GgCwlWJIcYmGIhJS3B0KAAAAHETin09UCvZR87J+al4uQEUCfORF1g8gn0pOMXT8SoLWnozTprPxSvHYC8YAAAA8A4l/PlCvhK8eb15ExQoHyr9gIVkK+Mli8bhxFwF4CCMlWSWLXtUtJaJVomCk/jh41d0hAQAAIBMk/m5WKdhHjzcvorLFi8q7cFFZaOkHkM9ZJHn5Bqior7/uvMXQyYhE7bh4zd1hAQAAIAM0K7tZ87J+KlY4kKQfwA3Hy7+QigQVUpMyfu4OBQAAAJkg8XcjL4vUvFzA9e79JP0AbkAF/PxVr5S/fDiEAQAA5Fsk/m5UxM9LRQJ8ZClAaxmAG5SPrwr6eivIn68TAACA/Ipr/F1oz1u3KSgoyOHy8fHxOn78uCqXDlZAQEAuRgYAuSMuLk5+8Ve04eVO8vf3d3c4AAAAsIMmmnyAbv4AblQcvwAAAPI/En8AAAAAADwYiT8AAAAAAB6MxB8AAAAAAA9G4g8AAAAAgAcj8QduUh07dpTFYlHHjh3tvm6xWGSxWDR27Ng8jQsAAACAa5H446ayatUqM6G19yhUqJBq1qypwYMHa8WKFe4OFwAAAACc5uPuAOC4yqMXujuEXHXigzvcHYKuXr2qw4cP6/Dhw5o1a5aGDBmiqVOnytvb292hAQAAAECOkPjjpvXEE0/oySefNOcNw1B4eLg2btyozz77TKGhoZo5c6YqVKigd999142RuodhGO4OAQAAAIALkPjjplWyZEnVq1cv3fMdOnRQ79691bRpU8XHx+uLL77QG2+8IV9fXzdECQAAAADO4Rp/wI46derojjuuX3oQHR2tAwcOuDkiAAAAAMgZEn8gA1WqVDGnExIS0r1+7NgxffLJJ+rVq5cqV66sgIAABQQEqFKlSrrvvvu0ePHiLNcRERGh9957T61atVJISIgKFCigEiVKqE6dOrrrrrv0zTff6OLFixkuHx8frwkTJqhLly4qXbq0fH19VbJkSXXt2lVTpkxRUlJSzjZemY/qP336dPP1EydOKCUlRd9++61at26tkJAQFSxYUA0aNNB7772n2NjYLNeVnJysGTNm6M4771TZsmXl5+enYsWKqW3btvr0008VFxeX4+0AAAAAbnZ09QcycPLkSXO6YsWKNq8dP35c1apVs7vcqVOndOrUKc2ZM0cPPPCApk2bJh+f9Lva/v371bVrV507d87m+bCwMIWFhWn//v36/ffflZycrJEjR6ZbfufOnerTp49NnJJ06dIlLV++XMuXL9ekSZP0559/qlSpUg5vd3bFxsaqe/fuWr58uc3zu3fv1u7duzV//nytWLFCBQsWtLv8qVOn1Lt3b+3cudPm+fDwcK1fv17r16/XN998o4ULF6pmzZq5th0AAACApyLxB+w4cOCAFixYIElq2bJlusQ5OTlZvr6+uu2229StWzfVqVNHRYsWVXh4uA4dOqSvvvpKe/fu1axZs1S1alW99dZb6dYxePBgnTt3TgUKFNDw4cN1++23q3Tp0kpJSdGZM2e0adMm/fbbb3bjO3LkiDp06KDIyEgFBQVpxIgRatGihSpUqKDLly9r/vz5mjRpkrZs2aI+ffpo7dq1KlCggOvfKEnDhw/Xpk2bNHToUPXv31+lS5fWqVOn9OGHH2rjxo3avHmz3n33XY0bNy7dspcvX1bbtm11+vRp+fn5afjw4erQoYMqV66smJgYLV26VJ9//rmOHDmi22+/Xdu2bVNwcHCubAcAAADgqUj8cdMKDQ3Vnj17zHnDMBQREWGO6h8XF6fg4GB99tln6ZYtU6aMTpw4oTJlyqR7rUuXLnr88cf10EMPafr06frkk0/03HPP2SSsx44d07///itJ+vTTT9O16Ldo0UJ33323xo8fr4iIiHTrGDp0qCIjI9W4cWMtXbpUxYsXt3m9e/fuuvPOO3XHHXfon3/+0fTp0zV8+PBsvT+O2rBhg77//ns98MAD5nNNmjTR7bffrmbNmmnPnj2aPHmy3nnnnXQ9H55++mmdPn1alSpV0sqVK20ur5Ckjh076t5771W7du107Ngxffjhh3rvvfdyZTsAAAAAT8U1/rhpffPNN6pfv775aNCggdq3b6+XX35ZoaGhevzxx7V582a1bNky3bIFCxa0m/Snslgs+uSTT+Tt7a2rV69q2bJlNq9fuHDBnG7fvn2m9YSEhNg8t3btWm3YsEGSNGPGjHRJf6oePXqoX79+kq5fk59b7r77bpukP5Wfn595QuPy5cvat2+fzesnTpzQzz//LEmaMGFCuqQ/VePGjTVixAhJubsdAAAAgKeixd+Fvh21WgG+9q9jtsc/2Ev1ewXrsm+MCvhcy8XIbgyhJ6NyfR1XLlx1qFxKSop+/OFHpVzz0msvj5Wfn1+m5RMTE3UpLFQxV2OUnJxsPh8SUlRhYZe0Yc1mtWvWzXy+QEohc/rrL77V26+/7/A2/DRrriSpetUaKhVUKdP3rXG9FpozZ462bNmic0fDbVrcE+OTzb+Z1XE1IiHd61Fh/z/Y3p3d78pw+SplbzGnt23ao9LBlf9/O77/VcnJyQoICFTTOm0yjaFBrWaSpHPnzmnbhr0qX65ChmWRtxKTrin6crxmT9+k+MgUd4cDAABgGjGxs7tDyDdI/HHTeuGZ0Xrx2TE2z8XFx+n4iWP65befNGnK1/p26tfauXu7fpo5T4EBgTZlExMT9f2P0zV33k/as2+Xrl3L+ORNePhlm/lKFSqrZYvW2rR5gyZN+Uqr1izXHbf3VuuWbdW0cfN067K2Y9cOSdKRY4dVqrJj17snJibqSsQVlShewqHy2VG9WsYD7hUp8v+9Fa5ejbF5beeu7ZKkuLhYla1W1OH1hV4KJfEHAAAAsoGu/oCVAP8A1alVV2+MeUcfvPOJJOmfLRv1+Vef2JS7EhGuO+7uqjFvvKBtO7ZmmvRL12+7l9bEL6aoWZMWkqSDhw/o0y8+VL+BvVWzQUX16X+7ZsyaYne5y5cv5Wjb4uKyvq1eTgT4B2T4mpfl/w8x1j0hJCksn20HAAAA4Klo8QcyMOi+IXpv/FhdibiiH+fM0pgXXjdfe+2t0dq5e4ck6fbud2pA/wdUp1ZdFS9eQv5+/rJYLJKkJq3r6uy5MzIMI139ZUqX1cJ5f2vN+lVatPhPbfhnvQ4dPqDExERt2rxBmzZv0NeTv9QP035RtarVzeVSE+i6tevrq/996/D2lCldNidvQ65J3Y5iRYvp1x8XOLxcxQqVciskAAAAwCOR+AMZ8PLyUpXK1XRlx1ZdDL2g8CvhKhpSVNHRUfpjwTxJ0j19++vr/03OsI6IyIgs19O+TUe1b9NRkhR+JVxr1q3U9z9O17oNa3Ti5HE9OvJBLV+0ziwfEnK9W/zV2BjVvqVOzjfQzVK3IyYmRjWr3yJvb283RwQAAAB4Jrr6A5lITk4yp5OSrk8fO3FUiYmJkqQ+d96d4bKHjxxKd117VoqGFFXfXvfo1x/+1G1de0qS9uzbrWPHj5pl6tdtIEk6eeqEQkMvZqv+/CR1OxKuJWjHf9f7AwAAAHA9En8gA7FxsTp0+KCk69exFytaTJKUlPT/16rHxmZ8l4AZs6c6tf52bTqY05etBge8revtkiTDMPTttG+cWoc7de9yu3lJxLdTv3ZzNAAAAIDnIvEHMvDx/8YpLv76bes6tu9idkWvUrmqmbDO+fVHu9fvL1n2l6bOzPj6+z17d2nP3l0Zvm4YhtasXyVJslgsqlihovlax/Zd1LhhU0nS199+YV52kJF9B/ZqybK/Mi3jDtWr1VCvnn0lSb//+asmfjch0/InT5/QvD9+yYPIAAAAAM/CNf64aYVdvqT9B/fZPJeQEK9jx49p7rwftWL1MkmSv5+/XnruFbNM0ZCi6tKpu5atWKIVq5ep/+C+evCBh1W+XAWFhYVpweI/9PMvP6hSxcqKjIrU5cth6da9Z99uPfPik2rUsIlu63K76tdrqJIlSioxMUmnTp/UT7/M0uq1KyVJt3XtqVIlS9ssP/GL79SjT2ddibiiR0cO+7/27jssiqsLA/i79N5EQREFW1Dsxh5LLNHYNQZ7x5Jo/FTU2HuvsWuMii1qNMbYxYYFCyp2iAUBC1JF+tJ2vj8IE1bYZUFgYXl/z8PjlHPvnNncoGdn5g6OHPsD3bv0QiWHytDW1kJERAQe+z6Ex4WzuHf/Dn4YOU68U6AoWbF4DR4+vo+g14GYu2gmznqcxvff9cUXVatDX08PHz5+gK/fE1y6cgHXb1xFpw5d0Kt7b3WnTURERERUrLDwz0etrrnBJBcTlMnKlkXaNzNhHK8NfS3efGEa+7rAj2GUGCYuu+/bAfd9O5TGl7aywo6lS9HYzhTIlN/m6ZPR1u8R3rx/j6vXPXH1uqdcO/uyZXF47Wr0/PFHRALQTYmTOz8Dafqt+w8e+uDBQx+Fx29Sty62z5mW5bOpZaUDz7170H/iRDx9+RIeF87AQ8lV/VK6six9aKdJxT+VffZ6ydFZ9mfkDwAm8cEwjZVk29YkPliuzaf9mGoDl3fvwiA3N3j5+OCmtxduenspzMVSX6tQxgmpLkkmg0HSBzTxXgyt9+/VnQ4REdFnqf6Pn7pTICoQLPyLEb+BfI1ZQdPT1YWluTlqVK6MDi1aYFCPHrA0N88SV97WFjf++ANrdu7EycuX8To4GAb6+qhQrhy6tmmDsQMGZNsug0unTihTqhQu3byJe0+fIjg0FGEfPiA1NRWlraxQt3p19O7YEd9/+y20FHwpVNXBAbeOHMGf587h2IULuPfkCSKiopCWlgYrCwtUc3BA03r10K1tW9SrUXRn/7e1tsb53btx5upVHD59GrcfPkRoZCRSUlJgYWaGyhUqoHGdOujcujW++vJLdadLRERERFTsSITsHlCmXImJiYG5uTm8q1TN/RX/WTNRoXRpXvEnomIpSSbD6/BwaC/iFX8iIir+eMWfipOMOjQ6OhpmZmZKYzX2in9SUhK8vLzg6ekJHx8f+Pr6Ijw8HElJSTA3N0f58uXRuHFj9OrVC+3btxcnayMiIiIiIiLSJBpX+IeGhmLChAk4deoUYmNjs42JiIhAREQEHjx4gG3btsHZ2Rk7duxA48aNCzlbIiIiIiIiooKlcfeXv3nzBgcPHsxS9JctWxYNGzZEmzZtUKNGDbnnpp8+fYqvvvoKR48qfy0aERERERERUXGjcYV/Zk2aNMHWrVsREBCA4OBgeHt74+LFi3j69CnevXuHcePGibf4p6amol+/fnj27JmasyYiIiIiIiLKPxpX+GtpaaF79+64d+8ebt68idGjR8PBwSFLnK2tLTZs2IB169aJ25KTkzFz5sxCzJaIiIiIiIioYGlc4V+/fn0cO3YM9evXVyn+p59+QqNGjcT1U6dOISEhoaDSIyIiIiIiIipUGlf450X37t3FZalUisDAQPUlQ0RERERERJSPNG5W/7ywsrKSW4+JiclTP0PcdKBtqK1yfFk9HfxsBkhKS6Cly9cJElHxI0uRIDwJWD5KB++T+VcKEREVc7trqTsDhR4PeazuFKgY4xV/IMsV/jJlyqgnESIiIiIiIqJ8VuILf0EQcOTIEXG9bNmycHR0VGNGRERERERERPmnxBf+v//+O/z9/cX1AQMGiK/4IyIiIiIiIiruSvQDmW/fvsX//vc/cd3CwgLTp0/PsV1SUhKSkpLE9bzOCUBERERERERU0ErsFf+EhAT06tULkZGR4rZt27ZlmegvO0uXLoW5ubn4Y29vX5CpEhEREREREeVZiSz8U1NT0bdvX9y5c0fcNnbsWLi4uKjUfvr06YiOjhZ/3rx5U1CpEhEREREREX2WEnerv0wmw6BBg3DixAlxm4uLC9atW6dyH/r6+tDX1y+I9IiIiIiIiIjyVYm64i+TyTB06FAcPHhQ3Pbdd99h//790NbWVmNmRERERERERAWjxBT+MpkMI0aMwN69e8VtPXv2xMGDB6GjU+JufCAiIiIiIqISokRUvDKZDK6urnB3dxe39ejRA4cOHcrXov9W0FuY6av+KkCpiYAAh1Q4pqTAQOArBImo+JGmCtBJTcXxd8EwiON8J0REVMzNi1Z3BkQFQuOv+GcU/bt27RK39ejRA3/88Qd0dXXVmBkRERERERFRwdPowp9FP1HeuB86DoldfUjs6iPwTbC60yEiIiIios+gsbf6Z1f09+zZE4cOHWLRT6Lk5BT8efoizlzygveDpwiPjEJMXDzMTU1QsbwtGtWtie86tUWbrxpCS0ujvycjIiIiIiINpZGFvyAIGDlypFzR36tXLxw8eLB4F/2/tlZ3BgVrlGehHu7o6YtwW7A22yvakVEfERn1ET6P/8HWvUdQrVJFrJk7CZ3btcjXHNwPHcewSfMAAAG3TsLBvly+9k9ERERERKRxhb8gCBg9ejR27twpbuvduzcOHDjA2ftJtHDtdsxZtUVcb9+yCbp90xI1qlaChbkpPkTF4NmrQJw4fxXnr97G81dBmLl8U74X/kXV0D7dMLRPN3WnQURERERE+UDjKuHDhw9j+/bt4rpEIkFUVBS6dOmich9ubm5o3759QaRHRcCuQ3+LRX8Zayv8sXU5WjVtkCWuXcvGGDu0D5788xIT561GeGRUYadKRERERET02TSu8E9ISJBbFwQBFy9ezFUfffv2zc+UqAh59z4M42YuBwAYGxniyp/b4VTFUWmbmk5VcO73Tfj9rzOFkSIREREREVG+4mxlVKKs3b4fCYlSAMCCyT/kWPRn0NLSwsDvOovrnjfuirPee964q7RtRty81VuztM94vh8AHJt0EWOz63vohLmQ2NWHQ+P0PELCIjB5wVpU+6oHjCo3g12DDnAZ/TOePvOXO37gm2CMn70C1b7qAcPKTWFTpx0GjJsJ/0DF71xXNqv/qKkLxX2HT5xX2Mdxjyti3JD/zVH6GRERERERUcHRuCv+Q4cOxdChQ9Vy7JrSHdASjFSOt9PVxjyhNJJl5SGR6eUYX/tzkisGHslUK8LzShAE7Dx8GgBgaGSMpv0m4JHMNE99+QvvMi2XhZUKuYcKluI5Zm6v/Dj/9R0lmAAAUqCDw0/i8OOgQYgICxVjE0OkOHzyPE5euoHNew+jfqOmuO11FW6jBiE2JkaMk0qT8PtfZ3Dy0k3s+vM0qnxRPctx3wilxWU/mT1iZBXE9eFzNuDcjUd4HeAP15+XwqJ+Z9iUtZNrHxEWiqGTFwMAytlXwOgFm/BIZqbSOVPxIsiSESYArkmr8E6apu50iIiIPs+0U/nWVeCyzjkHERUSXvGnEuPlMz9EfYgEANRv1ATGJnkr+vODc516OHLeC+OmzBS3bdn3J46c95L7ca5TL0vbxMRETBw5CCnJyRj/82zs/uss9h4/jx8mTYOunh4SE+Ixc8IYvA54hYmuA2FkbIKp85dh7/HzcD96BgNdf4REIkFM9EfMmzI+17kbGRlj6fpfoaOjg5joj5g18UcIgiAXM8dtLKIiI6CtrY0l67bBxJRFPxERERGRumjcFX8iRZ77PRGXa9Sqq75EkF48V3WqAd9HD8RtFStVgZ19BcWN/hUVGQEIAvafuAh7h//uNKhd70tYWFph6eypCH7zGoN7dkCp0mWw+6+zsCplLcbVa9gEOjracN+6AY/v34Xfk0eoXjN395PUrFsfYyb+jI0rF8Pb6yr2bNuIIWN+AgD8vutXeHmmz6sx/McJqNewSa76JiIiIiKi/MUr/lRifIz6b1Z+S+vSSiKLvh8nz5Ar+jP06DMQ+voGANK/IJg2f5lc0Z/h+0EjxGUf75t5ymH42IliUb9h5SI8830M/+f/4JfFcwEAznXqY8ykaXnqm4iIiIiI8g8LfyoxEuJixWVDI9XnYihqJBIJOnTpme0+A0NDVHCsBAAwM7dAs9Zts40rX6Gi+KjDu9eBecpDW1sbi9dtg4mpKVKSkzHtp1GY/tNIJCVJYWBohCXrt0FHhzcVERERERGpGwt/KjGMMj3Tn/jJax+LEwurUjC3tFS439TMHABg71AJEolESVz6c/fxcXF5zsXOvgKmL1oJAHj1/B88801/nGLK3MVwqFQlz/0SEREREVH+YeFPJYZFpmL5Q3iYGjP5PIaGhkr3a2lp5SpOlvZ5M7F36dUHXzZpLq43afE1eg8Y+ll9EhERERFR/mHhTyVGteo1xWW/Jw/VmIlm8X30AA/ueYvrz/2eIDIiXI0ZERERERFRZiz8qcSo8kV1WFqVAgD4eN9CXGxMDi0U08p0C71MJlMYl5AQn+djFAfSxETM+N9opKakwMjYBBKJBB8iwjFvyk/qTo2IiIiIiP7Fwp9KDIlEgq69+wEAEhPicfTA3jz3lXm+gNjojwrjgl7555hTcbZ64SwEvHwOAJi1dDUGjRwLALh64Rz+2LtTnakREREREdG/OOV2PnoyvwPM/p0wTRVSqRQBAQFwtDWDgYFBAWZWPNQub1Hgx1gyZxqO7NuJhIQEbF2zFK79v4OTk1OO7WQyGQ4cOIABAwYAAOyNa4n7wgP8ULv84Gzb7Vt/Uly2MTPIco5+5azE5UpW+vhCyWdgaawHANDV1lL6WRnr64h/KovT1dYS+/00zsfqv7ceVC9rBods+jl9+rRY3Pfr1w/TfxqF5ORkPLx9DQ8fPsTaRbMx+LsuqFatmsIcqPiTSqXQSzTERbfW/D1GREREVETxij+VKHZ2dti4cSMAID4+Hq1atcKVK1eUtvH19UXHjh2xcuVKcZulpSVq164NANi1axc+fPiQpd3169exbt06pX2XLVtWXPb3V353QFESHh6O4cOHAwDs7e2xefNmAICenh72798PAwMDJCQkYODAgUhNTVVnqkREREREJR4Lfypxhg0bhgULFgAAwsLC0Lp1a3To0AGbN2/G5cuXcf/+fVy8eBFbtmxBly5dULt2bZw/fz5LP2PHpt/WHhoaihYtWuDgwYNi20mTJqFdu3b48ssvleZSr1498Srp7Nmzcf78eTx//hwvX77Ey5cvkZiYmM9nnz9cXV0RGhoKLS0t7NmzBxYWFuI+Z2dnLFu2DABw584d8bMmIiIiIiL14K3+VCLNnj0bzs7OcHNzQ2BgIDw8PODh4aEw3tnZGStWrJDb5urqijNnzuDYsWPw9fVFv3795PbXqlULf/75p9xV/U+Zmppi/PjxWLFiBXx8fPDNN9/I7b98+TJat26d+xMsQL/++iuOHz8OAHBzc8s2v/Hjx+P06dPw8PDAkiVL8O2336Jp06aFnCkREREREQEs/IuXedHqzkCj9OrVC126dMGRI0dw5swZ3LlzB2FhYYiNjYWZmRkcHBzQpEkT9O7dG61bt84yEZ+WlhaOHDmCbdu2wd3dHb6+vgCAypUro0+fPpg4cSIMDQ1zzGPZsmWoWrUq9uzZg6dPnyI6OhppaWkFcs6f68WLF5g0aRIAoG7duli0aFG2cRKJBLt27ULt2rURGRmJgQMH4uHDhzAxMSnMdImIiIiICIBEEARB3UkUdzExMTA3N0d0dHTeJvdzdOSkWERULPH3GBEREZF65KYO5TP+RERERERERBqMhT8RERERERGRBmPhT0RERERERKTBWPgTERERERERaTAW/kREREREREQajIU/ERERERERkQZj4U9ERERERESkwVj4ExEREREREWkwFv5EREREREREGoyFPxEREREREZEG01F3Aprk1wlXYKhnrHK8gbkWanU1R6ReHHR1kgswMyKigpGSmozYSCn2u9+CNFqm7nSIiIiICs3YrW3UnYLKeMWfiIiIiIiISIOx8CciIiIiIiLSYCz8iYiIiIiIiDQYC38iIiIiIiIiDcbCn4iIiIiIiEiDsfAnIiIiIiIi0mAs/ImIiIiIiIg0GAt/IiIiIiIiIg3Gwp8onxw8vB82DuawcTDH6zdB6k6HiIiIiIgIAKCj7gQ0SatrbjDR1lY5Xla2LNK+mQnjeG3oa/E7mMJw9c4ddBw+XFw3MTJCoKcnjAwNlbZLlErh+PXXiImLE7ed3bkTLRs2FNcNpJH/9RsfDNNYST5mTlQ0JclkMEj6gCbei6H1/r260yEiIiIqRH7qTkBlLPyLkUb3+qk7hQLl3eBAoR8zLiEBJy5dQp/OnZXGnbx8Wa7oJyIiIiIiKi40+jJzbGwsrl69il9++QWDBw+Gs7MzdHR0IJFIIJFI4ODgoO4USY0M9PUBAL+fOJFjbEZMRpvsDOrRAwmPHyPh8WNUtLPLnySJiIiIiIg+k8Ze8f/iiy/w4sULCIKg7lSoiOrcujX+PHcOl27dQkhEBGytrbONC4uMxMWbNwEAXb7+GkfOni3MNImIiIiIiD6Lxl7xf/78OYt+Uqpts2awsbZGWloaDp8+rTDu8JkzSE1NhY21Ndo0bVqIGRIREREREX0+jS38MxgbG6NZs2b46aefsGvXLnTs2FHdKVERoa2lBZdvvwUA/H7ypMK4/cePAwD6dOoEbSWTMO49dgxGtWrBqFYtBL17J7dv7Lx54r6j584p7OPk5cti3MiZM+X2Ldq8WdynzNU7d8S4q3fuZNnfYdgwGNWqhQ7DhgEAngcEYOy8eajesSMsGzSA49dfY6CbG7wfPlR6HCIiIiIiKh40tvDft28ffH19ERMTAy8vL6xfvx5Dhw6FjY2NulOjIqRf164AgId+fvB9+TLLfj9/fzzw85OLzYsVU6eiSsWKAICfFizA25CQLDEhEREYO28eAKCinR1WT5+e5+Op6ty1a2jepw92/fkngt69Q1JyMkIjInDUwwNtBg/Gxr17CzwHIiIiIiIqWBpb+A8YMADVq1eHFl+TR0rUrV4dNapUAQAcyGaSv4xtzlWqoI6TU56PY2xkhJ3LlkFHRwdRMTEYNWtWlkdRRs+ahfAPH6CtrY2dS5fCzMQkz8dTxfvwcAz7+Wfo6Ohg/v/+h8t79+Ly3r2YN348zExMIJPJMHXFChy/eLFA8yAiIiIiooLFqphKvP7/Xsk/dPq0XDEuCAIOnjqVHtOt22cf58uaNTHzhx8AAJ63b2Pd7t3ivi379+O8lxcAwG34cDStV++zj5eTl0FBEABc3rcPU1xd0bhuXTSuWxdTR47Epb17xS8eJi1ZgpSUlALPh4iIiIiICgYLfyrx+nTuDC0tLbwNCZF7Jv7qnTt4GxICLS0tuHTqlC/HmjxiBJrVrw8AmLd+PR49ewY/f3/MXLsWANCgZk3M+vHHfDmWKqaNHg2nSpWybK9RpQqmjhwJAAgOC8PJy5cLLSciIiIiIspfGvs6P3UY4qYDbUNtlePL6ungZzNAUloCLV1JAWZWPLyyLfjPINjqv+Vwc0n6MW1t0eirRrh19Ra2XTgB+26NAQDbLqTf5t+oRSMk1bHFq3/bZO4rc86Z970uLUFatuejg7nbl+K71t8hLjYOA2am32ovTUqCoZEh5v+2DK/L62abe1SmO/+VfVaZz/HTHAEgUS/9T4lEgq9cu+NVqez7ajmqJyS//AJBEHDs0S3UGdxB4TGp5JKlSBCeBCwfpYP3yfwrhYiIiIqfx0MeqzuFAscr/kQAurqk3+5/4eQFSBOlkCZKcf7EeQBAN5fPv80/M7sKdpi5PH3Gfv9n/nj29BkAYOrCqXCo7JCvx1KaR0U7WJayVLjfytoKdhXsAAAvfF8UVlpERERERJTPWPgTAWjfuT0MjQwRFxuHS2cv4dKZS4iPi4ehkSHadW6X78fr+n1XfNnsS3G9aeum+H7w9/l+HGVKWZfKOaZ0ekz0x+iCToeIiIiIiAoI78vMg6SkJCQlJYnrMTExasyG8oORiRHadGqDU0dO4cQfJ4B/5/hr27ktjIyN8v14Tx8+xYM7D8T1Z0+fITI8Uiy0CwWfLiEiIiIiKhF4xT8Pli5dCnNzc/HH3t5e3SlRPsi4pf+m503cvHJTblt+kiZKMe2HaUhNSYWRsREkEgk+hH/AnAlzlLbL/GpKmUymMC4xPlGlPCLDI1WOMbcwV6lPIiIiIiIqelj458H06dMRHR0t/rx580bdKVE+aNKyCUrblEZqaipSU1NRxrYMmrRsku/HWTl3JQJeBAAA5qyag8E/DAYAXPG4gkPuhxS2MzYxFpdjPiq+yyToVZBKebwLeoePHz4q3P8h4gPevX4HAKhavapKfRIRERERUdHDwj8P9PX1YWZmJvdDxZ+2tja6unSFnr4e9PT10PX7rnJX2fPD1fNXcWhXenHfqVcndOndBRNmTsAXzl8AAFbNXYVA/8Bs22ZMtAcATx88VXiMM3+dUSkXQRBw/I/jCvcfO3gMgpD+zENBfAFCRERERESFg4U/USaT5kyCz1sf+Lz1wcQ5E/O17w8RHzD7f7MBALZ2tpi1YhYAQFdPF8u3Loe+gT4SExLTHwNITc3Svm7DutDRSZ+WY8/WPWJRntnOjTvx2Ef115FsXb0VAS8Dsmz3f+6P7Wu3AwBK25RGm2/bqNwnEREREREVLSz8iQrJ3IlzERkeCS0tLSzZtARm5v/dKVLFqQomzJoAAHhy/wm2rtqapX2p0qXwTbdvAABel70wbuA4XL94HX6P/HDpzCVMHDYRa+avQd2GdVXKp4JjBQgyAQM6DsBv637Dw7sP8fDuQ/y27jcM/HYgYmNiAQAzls6Arp7u5508ERERERGpDWf1z0e3gt7CTF/1qdKlJgICHFLhmJICA4FTrDsnJxf4McJT/ruSbpeamutj3sl0Jd4xRb595n3VUlLgkGnfr/v+xOWzlwEAbqMHYljDusAnx1461AX3z1+Bx5Vb2P7Ldgxu2RhNv6wjF7NzzkS0uP8ELwJe44rHFVzxuCK3v2/3DnDt1wPt+v6QbY4AYPzvnQKOtqWxeZ4bXMb8jF8W/ZLlXLW0tLBi5v8woUOrLLkSZZCmCtBJTcXxd8EwiON8J0RERFQMzVPDRNbzCvd12Sz8i5HH7feoOwXKgxevXmPS/DUAgLrOX2DR1LHZxkkkEuxaMx+12/VBZNRHDBw/Gw/PH4RJptcJ2pQuhdsn92D5ZnccPX0Jr4NDYGxoiJpOlTFqQC8M6NUJnjfuqpxb53YtcPfMPqzcsgeXvO7gfVgELMxM0aJRPbiNHpjliwciIiIiIip+JEJ2DwprsKFDh2L37t0AgIoVKyIwMPCz+4yJiYG5uTmip5nm8oq/PQKar4ajXWkY6PCKPxWO1r1H4srNe2jVtAE8j2xXdzpUzElTBQS8C4ejlxuv+BMRERGpKh+u+It1aHR0jhPO8xl/IiIiIiIiIg3Gwp+IiIiIiIhIg7HwJyIiIiIiItJgGlv4L1q0CAYGBll+9u7dK8YEBQVlGzNy5Eg1Zk5ERERERESUfzR2Vv/U1FQkJSXlGJddTEpKSkGkRERERERERFToNLbwJ6LscSZ/IiIiIqKSRWNv9Z83bx4EQcjTj7u7u7rTJyIiIiIiIsoXvOKfj2pKd0BLMFI53k5XG/OE0kiWlYdEpleAmRERFQxBlowwAXBNWoV30jR1p0NERERUaAKXdVZ3CirT2Cv+RERERERERMTCn4iIiIiIiEijsfAnIiIiIiIi0mAs/ImIiIiIiIg0GAt/IiIiIiIiIg3Gwp+IiIiIiIhIg7HwJyIiIiIiItJgLPyJiIiIiIiINBgLfyIiIiIiIiINpqPuBDTJk/kdYGZmpnK8VCpFQEAAHG3NYGBgUICZEREVDKlUCr1EQ1x0a83fY0RERERFFK/4E1EW7u7ukEgkkEgkCAwMVHc6RY6DgwMkEgmGDh2q7lSIiIiIiHLEwp9KtOTkZBw4cACDBw+Gk5MTSpUqBV1dXVhbW6NBgwb44YcfcOHCBchkMnWnSrnQunVr8YsLiUSCBQsWqNRu8eLFcu1at25dsIkSERERERUC3upfjPg5VVd3CgWq+j9+hXq8o0ePws3NLdsr2pGRkYiMjISPjw+2bt2KatWqYc2aNejcuXO+5uDu7o5hw4YBAAICAuDg4JCv/VO6/fv3Y86cOTnG7d27txCyISIiIiIqXCz8qURauHChXCHYvn17dOvWDTVq1ICFhQU+fPiAZ8+e4cSJEzh//jyeP3+OmTNn5nvhX1QNHTpUI25jNzAwgFQqxfPnz3H79m00btxYYeydO3fw7NkzuXaK8PEHIiIiIipOeKs/lTi7du0Si/4yZcrA09MTHh4eGDduHNq0aYP69eujXbt2GDt2LM6ePYuHDx+iXbt2as66ZMmYY+Bz74CwsbFBo0aNAOR8NT9jf+PGjWFjY/NZxyUiIiIiKkpY+FOJ8u7dO4wbNw4AYGxsjCtXrqBVq1ZK29SsWRPnzp3D5MmTCyNFymeDBw8GABw6dAgpKSnZxqSmpuLgwYNy8UREREREmoKFP5Uoa9euRUJCAgBgwYIFcHJyUqmdlpYWBg4cKK57enqKE8B5enoqbZsRN2/evCztM57vBwBHR0e5ieU+7Xvo0KFyV8FDQkIwefJkVKtWDUZGRrCzs4OLiwuePn0qd/zAwECMHz8e1apVg6GhIWxsbDBgwAD4+/srzFnZrP6jRo0S9x0+fFhhH8ePHxfjhgwZovQzKkh9+vSBrq4uIiIicObMmWxjzpw5g/DwcOjq6qJPnz459qloVv8XL17AxMQEEokE9erVQ3JycrbtU1NT0bhxY0gkEhgaGsLX11duf3ZjJjsZkxhmNwnhp2NUJpNh+/btaNasGaysrGBsbIw6depg6dKlSh9rICIiIqLij4U/lRiCIGD37t0A0q/2jxw5Us0Z5d3Dhw9Rr149rF69Gi9evEBiYiKCg4Nx+PBhNG7cGNevXwcAXLp0CXXr1sWGDRvw4sULSKVShIWF4ffff0fDhg2zfEmgirVr16Jq1aoAgNGjR+Pt27dZYkJCQuDq6gogvUjesGHDZ5zt57G2tsa3334LQPHt/hnbO3XqhFKlSuX5WFWrVsXatWsBAA8ePMCsWbOyjZs/fz68vb0BACtWrECNGjXyfExVJCcno3Pnzhg1ahRu3ryJqKgoJCQk4NGjR5gxYwYaNGiAkJCQAs2BiIiIiNSHhT+VGE+fPkVERAQAoEWLFjA1NVVbLg0bNsTjx4+xaNEicdu5c+fw+PFjuZ+GDRtmaZuQkICePXsiOTkZS5YsgZeXF27duoV58+ZBT08P8fHxGDRoEF6+fIkePXrA1NQU69atw61bt3D9+nVMnDgREokEUVFRGDFiRK5zNzY2xv79+6Gjo4OoqCgMGTIEgiDIxQwbNgzh4eHQ1tbGvn37YGZmlvsPKR8NGjQIAHDixAl8/PhRbl90dDROnDghF/c5Ro4cie7duwMAVq9ejcuXL8vt9/LywtKlSwEAHTt2FB89KUizZs3C2bNn8c033+Cvv/7C3bt38ddff6F9+/YAAF9fX3Tt2hVpaWkFngsRERERFT7O6k8lxsOHD8XlBg0aqDGT9OK5Zs2auHv3rritWrVqKk1mFx4eDkEQ4O3tjcqVK4vbGzduDGtra4wbNw6BgYFo1qwZbG1t4eXlhdKlS4txzZs3h46ODlauXInbt2/j/v37qFevXq7yb9iwIebNm4dZs2bh0qVLWL16tTgHwoYNG3D27FkAwLRp09C8efNc9V0QunbtCgsLC3z8+BGHDx+Wu9vj8OHDkEqlsLS0RJcuXfLleL/99htu376NkJAQDB48GI8ePYKlpSViYmIwcOBApKWlwdraGrt27YJEIsmXYypz584djBo1Ctu2bRO3NWjQAD169ICrqyt27NiBu3fvYtu2bfjxxx8LPB8iIiIiKly84k8lRmRkpLhcpkwZNWby+RYuXChX9GcYPnw4DAwMAKR/QbB+/Xq5oj/DDz/8IC5fu3YtTzlMmzYNX331FQBg5syZePjwIXx9fTF16lQA/305UBTo6+vDxcUFALBv3z65fRm3+bu4uEBfXz9fjmdtbS3Ok/D27VuMGTMGAMQvZQBg+/btsLW1zZfj5cTGxkZ8BOFTv/zyizhGNm/eXCj5EBEREVHhYuFPJUZsbKy4bGxsrMZMPo9EIhGL2E8ZGhqKz99bWlqiQ4cO2cY5OjqKjzq8evUqT3loa2tj7969MDMzQ3JyMvr374/+/ftDKpXCyMgI+/btg45O0bmpKOM2/mvXriEoKAgAEBQUJH7xkR+3+WfWoUMH/PTTTwCAP/74A/379xe/ZHB1dUWPHj3y9XjKuLi4wMjIKNt9JiYm4nh6+vQpn/UnIiIi0kAs/KnEyPxMf3x8vBoz+TzW1tawsrJSuN/CwgIAUKVKFaW3kWfEZf5CJLccHBywadMmAOnPiWc8TrF27VpUq1Ytx7afvsUg4yfjbQdBQUEKY1SZ9T6zr776CpUqVYIgCOJV/71790IQBFSuXLlAHklYvnw5nJ2dAQAHDhwAkD4B4C+//JLvx1Imu7kiMmvUqJG4/Pjx44JOh4iIiIgKGQt/KjEyz9YeGhqqxkw+j6Irtxm0tLRyFfe5E7oNHDgQrVq1Etfbt2+PUaNGfVafBSXjlYwZV94z/sz8qsb8ZGBggI0bN8ptc3d3L/Q7TnJ6tMXGxkZc/vDhQ0GnQ0RERESFrOjch0tUwOrUqSMu+/j4qDETzXLv3j3cuHFDXH/48CHCwsJyLDY9PDwUvuf+77//xqxZs1CuXDmcO3dOYR+5nath0KBBWLBgAZ49e4aNGzfi+fPn4vaC8umrDD08PNCsWbMCO152CmMCQSIiIiIqulj456NfJ1yBoZ7qV/IMzLVQq6s5IvXioKuTfQFUkoQFxRRo/6VN7FHKqhQiP0Ti6tVr8H/yFqameXvNXHRYorgc+T5WYe7xCf89UhD/MSlLXEzEf/1EvI2FkUTxZyCNSwEAyFIFpZ9VijRN/FNZnCxVEPvNa16J0kT069MfKSkpMDY2QUJCPMLCwjCw/xDs23FI4bEBwELfFlAwl56pfvqjDNoSHZQxraC4k8Tsx03GZ/DpZ2WmWwZf1m+Euz7emDJ5CgCgYYPGMNUpLReX8dko+gyVfXaZ/f7HXhw9ejT9nEzNEBsbg8WLF6Nx3ZZoUO9Lhe0kEgkEQUBsVKLS/qOjYhTmGRXy39h78TQQdb9Q3M+Lp4HislaKfq7+X0xJTUZspBT73W9BGi1TuR0REREVbWO3tlF3CpSPeKs/lRgSiQQu3/UDACQkxGP/oT157svE2ERcjo7+qDDu1auXOeZUnM1bNBMv/NOvmq9YvBZjXMcCAM5fPIvd+3aoMzWFvu/ZBwAgTZLKree3wNcBmDV/GgDA6YsaOH/CE+Zm5khNTcXYiSPlvhT6lIlJ+nwUH5WMLUEQEBCo2sSM9x8pv8Ml836nL2qo1CcRERERFR8s/KlEGT1iLAwN0599X75mCV68fK5SO5lMhiPH/hDX7e3/uwr94NF9he3+OnFEab+ZXx+n6Lb3ourCZQ+4/1vc9+zWG717uGDGlLlwrl4LADB38Sz45/DFhzp079oLJiam0NfTh4mJKbp16Znvx0hLS8PYCaMQHx8HfT19bP5lOxwdKmPF4vRX6gUEvsLsBdMUtq9QviIA4KGSsXXR8zyiY6JVyufEqWNIlCZmuy8+IR7HT/0FAPiiqhNsyhTOKwaJiIiIqPCw8KcSpaxtOSydvxJA+lX/Hn074cat60rbPHvxD/oO6YXN29aL2yzMLVHDqSYA4ODh/Yj6mHVCtNt3bmL7rq1K+85cZAW+DlD5PNQtIjICE6akX923K1ceyxeuBgDo6elh87rtMNA3QGJiAn6cOBKpqanqTDULSwsr+D95i9fPw+D/5C0sLRS/ISGvftm4Cnd9vAEAM6bOgXP19LHSo+t36P3vHQb7D+7BGY9T2bZv2jj9DQM+D+7C++6tLPvDwkIxY+4UlfMJCw/FvEUzs903d+EMRESEAwCGDByhcp9EREREVHyw8KcSp5/LQPw8Kb0IiogIR8++ndFnUE/s2rsd129cxeMnD3HVyxPue3/DgOEu+LpjM1y5djlLP8MGuwIAwiPC0P37b/HX8SNi2zkLZ+D7Ad1Rp1Y9pbnUcq4NA30DAMDy1Ytw5dol+L96iYBAfwQE+iu8Sqtuk37+CeERYdDS0sKG1Vthbm4h7nOqVh0zf54HAHjw0Aer1y9XT5Jq4vPgHtZsWAEAaNG8FUaPGCu3f9mCVbC3S79jxG36eISFh2XpY1D/odDR0YEgCBg0og+27diMB498cOfebWzatg5tO7dAbGwMKjlWVimnurXrwX3fDvQb8h3OeJzCoycPcMbjFPoO7oW9B9wBpI/FIQOGf8aZExEREVFRxcn9qESaNH4qvqjmhLmLZuLN29fwvHYJntcuKYz/olp1zJk+X27bwL5DcMnzAs54nMSzF/9gzHj5q6XVnZyxc8te1Gqk+H32JiamcB06Ghu3rcOjJw/hMkj+tvOjB06iedMWeTjDgrPn9104d+E0AOCHkeOyzW/ksDG4eNkDntcuYd2m1WjTqj0aNmiUJU7TxCfEY+y/dzlYmFtgw+qtWeZxMDU1w8a129Czb2dERkZg4tSx2L/rsFyMU7XqmD1tAeYumoGP0R8xZ+F0uf2WFpZw//V3LF+9GK8C/HPMa/rk2diyfSMuXbmAS1cuZNlftXI17Nv5B3R0+FcCERERkSbiFX8qsTp37IYbl+5h87rf0LtnH1SpVBUW5hbQ0dGBpYUlatesg2GDXPHn7ydw5dxNtG7ZVq69lpYWdmzZg2ULV6FunfowMjKGkZExajjVxPQps3Hm2EWUKWOj4Oj/mTVtPlYvW48mjZrB0sIS2traBXXKn+1VgD/m/nvLeM0atTDNbXa2cRKJBOtWbYaVpVX68+4TRyI+Pq4wU1WLOQuni4X4isVrUda2XLZxTRo1w7gxEwCkz5Wwa+/2LDFjXMfi4O4/8XXLtrAwt4C+nj4q2FfE8MEjceHUNTRppPorAXV19fC7+xEsX7gaDeo1hLmZOQwNjVDdyRnTJs/C+VNXYWtTNvcnTERERETFgkQQBEHdSRR3MTExMDc3x8phx/P0Oj+7svbQ1dErwAyJqKTxunkNvfp1AVCwd46kpCbj3fs3eHwimq/zIyIi0iB8nV/Rl1GHRkdHw8xM+WvKecWfiIiIiIiISIPxgc581OqaG0xycZu2rGxZpH0zE8bx2tDX4ncwRJR/jBLD5JZNY18XyHGSZDIYJH1AE+/F0Hr/vkCOQURERIXPz0ndGeS/6v/4qTsFtWG1SURERERERKTBWPgTERERERERaTAW/kREREREREQarEQ84//u3Tvs27cPx48fR2BgICIiImBtbQ0HBwd069YNAwcOhJ2dnbrTJCLKNy0bNkTC48fqToOIiIiIigCNL/y3bt2KyZMnIz4+Xm57cHAwgoODcePGDSxatAirVq3C6NGj1ZQlERERERERUcHQ6Fv9FyxYgB9++EGu6K9atSpatWqFypUri9vi4uIwZswYLFq0SB1pEhERERERERUYjS38//77b8ydO1dcr1GjBu7du4fnz5/D09MTL1++xJ07d1C9enUxZvbs2Th+/Lg60iUiIiIiIiIqEBpZ+KekpGDy5Mnievny5XH9+nXUr19fLu7LL7/E9evX5Z7vnzx5MlJTUwstVyIiIiIiIqKCpJGF/8GDB/Hy5Utxfc2aNbC0tMw21srKCmvWrBHXX7x4gYMHDxZ4jkRERERERESFQSMn9/vjjz/E5XLlyqFnz55K43v16oWyZcvi/fv3AIDDhw9j4MCBuT7uEDcdaBtqqxxfVk8HP5sBktISaOlKcn08IiJ1k6VIEJ4ELB+lg/fJGvlXChEREX2Gx0P4lqGiQOOu+CcmJuL8+fPieseOHaGjo/wfozo6OujYsaO47uHhAalUWmA5EhERERERERUWjSv8/fz8kJSUJK43b95cpXaZ46RSKfz8/PI9NyIiIiIiIqLCpnGF/9OnT+XWq1atqlK7T+N8fX3zLSciIiIiIiIiddG4BzIDAwPl1itUqKBSu4oVK8qtBwQEKIxNSkqSu6sgJiZG9QSJiIiIiIiICpHGXfH/tAi3sLBQqZ25ubncemxsrMLYpUuXwtzcXPyxt7fPdZ5EREREREREhUHjCv+4uDi5dUNDQ5XafRqnrPCfPn06oqOjxZ83b97kPlEiIiIiIiKiQqBxt/qnpKTIrec0o7+iuE/7yUxfXx/6+vq5T46IiIiIiIiokGncFX9jY2O5dVVfy/dp3Kf9EBVX3l7eqFm6JmqWrglvL291p0NERERERIVM4674m5iYyK0nJCTAyMgox3YJCQly66ampvmaFxUN3l7eGN5jeLb7DAwNYG5pji+cv0C7zu3QpXcX6OnrFXKGRERERERE+UvjCv/SpUvLrb9//x7W1tY5tnv//r3cuiptPnUr6C3M9CUqx0tNBAQ4pMIxJQUGQs7tNi2IynVOxcnYOZYFfozwlFSF+6SJUkgTpQgNDsXV81dxaMtunNy9Dg725Qo8r4KU+ZwdU1LhnJysxmxI00hTBeikpuL4u2AYxHG+EyIiIvrEPPOcY4qbedHqziDXNK7wd3JyklsPCgpCrVq1cmwXFBQkt169evV8zYuKnh8Gf48fh3wvrodFfMCTZ/5YuWUP3r4PxdNn/ug2bALunzsAbW1tNWb6eVo3+xLCOx91p0FERERERGqicc/4Ozs7y637+KhW8HwaV6NGjXzLiYqmMtaWqOlURfxp81UjjB/RD76eR8Sr/I/9XuKvM5fVnCkREREREVHeaVzhb29vj8qVK4vrV65cUald5rgqVaqgfPny+Z4bFQ+mJsaY9T9Xcf3CtdtqzIaIiIiIiOjzaFzhDwC9evUSlz09PfH69Wul8a9fv5Yr/DO3p5KpllMVcflNcKjCuMtedzDkf3NQqWlXGFVuBrMvWqBWWxdMWbgWwSHhCtvNW70VErv6kNjVBwDExMZh3uqtqNXWBSZVm6NM7bboNOgn3LjzUK5dWMQHzFq+Cc5f94ZxlWYo5fw1ug+biPtP/lF4LM8bd8Vjed64K7dvyfod4r4Vm90V9nHvkS/0HBpBYlcfX/ceBZlMpjCWiIiIiIiKFo0s/IcNGyY+ky2TybBw4UKl8QsWLBALGW1tbQwbNqzAc6SiTU9XV1zW1c06FYZUmoR+P05HG5fR2HPkJAJev0OiVIrYuHg8+eclVm3di2oteuCER853nLx5F4KGnQZh/ppf8eSfl4hPSER4ZBTOXPJCy+9ccfjEeQDAI9/nqN+hPxav3wHf56+QkCjFh4/ROO5xBU27DcVlrzu5Ps9p44ahReN6AIDZK7fgwZNnWWISEhMxYNwspKSkwsLcFHvWLYCWlkb+6iAiIiIi0kga+a/36tWrY8iQIeL6b7/9ht9++y3b2G3btmHHjh3i+tChQ7NMEEglj9/LAHHZoXxZuX2CIKD3qCk4+Pc5AEDX9i2xd/1CeB3bhZvH3bFuwRRUsLNFfEIieo+eirsPfZUe6/vRU/H2fSimjxuGK39ux53T+7B2nhvMTE2QlpaGEZMXIOD1O3QZMgGJ0iQs/nksrh/bidsn92D+5DHQ09NFUlIyhk6ci+TklFydp5aWFvauXwhzMxMkJ6eg/7gZSEyUysVMmrcGz/wDAQBbls6AvZ1tro5BRERERETqpXGz+mdYvnw5rly5An9/fwDAyJEjceLECfTt2xflypXDu3fvcODAAZw8eVJsU6VKFSxbtkxdKVMRkZaWhpVb9ojrvbu0k9v/2+9/4dTF69DV1cHxXWvR8evmcvubNKiNQd91RoteI/D0mT8mzF2F68d2KjzeA9/nuHJkOxrX/+/tE1/WqYGqjhXQZcj/EBsXj8ZdBkMQBHif2oPKDvZiXKN6NWFtaYGxM5fh9bsQnLp4DT2/bZOr861Yvhw2LZ6GgT/Ngt+LAExZ9As2Lp4GADjhcQXb9v0JABjQ61v07d4hV30TEREREZH6aeQVfwCwtrbGmTNn4OjoKG47fvw4+vfvj9atW2PAgAFyRb+joyPOnDkDa2trdaRLRUB4ZBQuXfdGq+9Gis/M9+7cDl81qifGCIKA5f8+Cz9+eL8sRX8GSwszrJw1AQDgdecBXrxSPM/EBNf+ckV/hs7tWqDiv3cbhEdGYeGUH+WK/gzD+nSDgYE+AODa7fs5n2g2BvTqhH49OgIANrn/gTOXvBAaHokRkxcAACqWL4tN/34ZQERERERExYvGFv4AULVqVTx69Ajjx4+HmZlZtjHm5uYYP348Hj16hCpVqmQbQ5pp/ppfxYntJHb1UaZ2W7TtMwZedx7AyNAAk0YNxO+bFsu18X3+Cv6BbwEAvTu3Vdp/yyb1xeWb9x4pjOvbTfFV9NrVqwIAJBIJ+nT7JtsYQ0MDVHVM/0Lg1et3SnNSZvOS6ajw7238wybNQ/+xMxAeGZXpcQDTPPdNRERERETqo7G3+mcwMTHBunXrxFv/AwMDERkZiVKlSsHBwQGtW7eGvr5+vhyrpnQHtAQjlePtdLUxTyiNZFl5SGR6KrSIyntyxcAjmWPOQZ/JX1CtMK7mXBvth02Bn3YFINME9scf3BKXm3YbqvJx74fJUDfT+YUKluJykmNLPJJlP24Es/Qr/hZWpfDGrA7eKJhMX9e0NICXeB8ny/I5Zj5nf6EsrBR9zqbA3LW/YWTfbggNj0RoeCQAYNgP/4N5w+/wiBP5UzYEWTLCBMA1aRXeSdPUnQ4REREVMYHLOqs7BUIJKPwzGBgYoEMHPp9M/3EZNBwug0cASH+uP/T9O1w4dRwnjx7Cg7veGOHSFftPXIRVqf8e//gQqfgVfcpIExMV7jM0VPxlkZZE698YQ6X9Z8yyL0v7vMLry6bN0aVXHxw/cgAAUKmaE35wm/5ZfRIRERERkXqVmMKf6FNW1qVR1amGuO7kXAut2nVEw2YtMHvSjwh+8xrzp4zHup2/izFpmQrr9bsOoFz5CiofqzgICX6Ly+dPi+vBr4PwJvAVKlX9Qo1ZERERERHR52DhT/SJbt/3w5ULZ3Hh9HF4nj+D215X0bh5SwCAhaWVGGdqZi73xUFxJ5PJMHPCD4iNjoaOri50tHUglSZi+vhR2Hf8AnR1ddWdIhERERER5YFGT+5HlFc//Twb2traAIANyxeK252ca4vLD+7eLvS8CtLubRtw9+Z1AMCYCVMxYeY8AMA/Tx5h06rFSloSEREREVFRxsKfKBsOlargmy49AACP79/FzauXAQDVa9WBTdlyAIA/9+9GklSqrhTz1T9PH2PTqiUAgHoNm2D42InoN3QUvvq6HQBg99YN8Ll9Q50pEhERERFRHrHwJ1JgxLhJkEgkAIDt61cBSJ9Ez3XcJADA29eBmDXxByQnJSnsIy42Bgfcfy34ZD9DklSK6eNHISU5GSampli8bpt4t8P8VRthaVXq38cAxiAuNkbN2RIRERERUW6x8CdSoKpTDbRu/y0A4N7tG/DxvgkA+H7QcLTp2AUA4HHyGHq1bQr3retx96YX/nn6GPdueeHIfndMG+eKdl/WwNY1y9V2DqpYu2QuXj3/BwAwbcEK2Nn/N2GhdRkbzF2xHgAQ/PYNls6eqpYciYiIiIgo7zi5H5ESrj+54bJH+iz3v65bha37/4REIsGKzTuxYu40HN63C2+CArB28VyFfVhZWyvcp243rlzCQfftAIBvuvRA1959s8R83aETevUbjKMH9uDkn4fQql1H8TEIIiIiIiIq+lj456Mn8zvAzMxM5XipVIqAgAA42prBwMAgx/hrn5NcMVC7vEWBH+NDaRNx2cbMIMdj1i7fBu7t2+P8+fO4efUSkt6/QMOGDQEAh/bswKwpE7B9+3Z4enri9evXiIuLg4mJCRwdHdGgQQN8++236NKlC/T19eX6tTEzyHQMxTlYGusBAHS1tZTGGevriH9+Gpf5nCuXNhH3R0ZGYsGUcRAEAXZ2dji4ZycsLbM/xp7tm/H47k28ePECS2ZMQr+u7WFnZ6cwHyo5pFIp9BINcdGttUq/x4iIiIio8EkEQRDUnURxFxMTA3Nzc0RHR+et8Hd05D+YiahY4u8xIiIiIvXITR3KZ/yJiIiIiIiINBgLfyIiIiIiIiINxsKfiIiIiIiISIOx8CciIiIiIiLSYCz8iYiIiIiIiDQYC38iIiIiIiIiDcbCn4iIiIiIiEiDsfAnIiIiIiIi0mAs/ImIiIiIiIg0GAt/IiIiIiIiIg3Gwr8IEARB3SkQEeUJf38RERERFX0s/NVISyv945fJZGrOhIgobzJ+f2X8PiMiIiKioof/UlMjXV1daGtrIz4+Xt2pEBHliVQqhZaWFnR0dNSdChEREREpwMJfjSQSCUxNTRETE8PbZYmoWIqLi4ORkRGv+BMREREVYfyXmpqZm5sjJSUFwcHBLP6JqFiJiopCQkICzMzM1J0KERERESnBezPVzMjICOXLl8fbt2+RmJgIMzMzGBkZQVtbGxKJRN3pERHJEQQBqampiI6ORmxsLCwtLWFubq7utIiIiIhICRb+RYCpqSkqVqyI6OhofPz4EZGRkepOiYhIKX19fdjY2MDS0lLdqRARERFRDlj4FxFGRkYwMjKCra0tUlJSONM/ERVZ2tra0NHR4V1JRERERMUEC/8iRiKRQE9PT91pEBERERERkYbg5H5EREREREREGoyFPxEREREREZEGY+FPREREREREpMFY+BMRERERERFpMBb+RERERERERBqMhT8RERERERGRBmPhT0RERERERKTBWPgTERERERERaTAW/kREREREREQajIU/ERERERERkQbTUXcCmkAQBABATEyMmjMhIiIiIiKikiCj/syoR5Vh4Z8PIiMjAQD29vZqzoSIiIiIiIhKktjYWJibmyuNYeGfD6ysrAAAr1+/zvEDJypOYmJiYG9vjzdv3sDMzEzd6RDlC45r0kQc16SpOLZJE+XXuBYEAbGxsShXrlyOsSz884GWVvpUCebm5vyFRBrJzMyMY5s0Dsc1aSKOa9JUHNukifJjXKt64ZmT+xERERERERFpMBb+RERERERERBqMhX8+0NfXx9y5c6Gvr6/uVIjyFcc2aSKOa9JEHNekqTi2SROpY1xLBFXm/iciIiIiIiKiYolX/ImIiIiIiIg0GAt/IiIiIiIiIg3Gwp+IiIiIiIhIg7HwJyIiIiIiItJgLPxVtGnTJjg4OMDAwACNGzeGt7e30vjDhw/DyckJBgYGqFWrFk6fPl1ImRLlTm7G9vbt29GiRQtYWlrC0tIS7dq1y/H/BSJ1yO3v7AwHDx6ERCJBjx49CjZBojzI7bj++PEjxo4di7Jly0JfXx/VqlXjv0eoSMrt2P7ll1/wxRdfwNDQEPb29pg4cSKkUmkhZUuUs6tXr6Jr164oV64cJBIJjh07lmMbT09P1K9fH/r6+qhSpQrc3d3zNScW/io4dOgQJk2ahLlz58LHxwd16tRBhw4dEBYWlm38jRs30K9fP4wYMQL3799Hjx490KNHDzx58qSQMydSLrdj29PTE/369cPly5dx8+ZN2Nvb45tvvsG7d+8KOXMixXI7rjMEBgZi8uTJaNGiRSFlSqS63I7r5ORktG/fHoGBgThy5AiePXuG7du3w87OrpAzJ1Iut2P7999/x7Rp0zB37lz4+flhx44dOHToEGbMmFHImRMpFh8fjzp16mDTpk0qxQcEBKBz5874+uuv8eDBA0yYMAGurq44d+5c/iUlUI4aNWokjB07VlxPS0sTypUrJyxdujTbeBcXF6Fz585y2xo3biyMHj26QPMkyq3cju1PpaamCqampsLu3bsLKkWiXMvLuE5NTRWaNWsm/Pbbb8KQIUOE7t27F0KmRKrL7bjesmWLUKlSJSE5ObmwUiTKk9yO7bFjxwpt2rSR2zZp0iShefPmBZonUV4BEP766y+lMVOnThWcnZ3ltvXp00fo0KFDvuXBK/45SE5Oxr1799CuXTtxm5aWFtq1a4ebN29m2+bmzZty8QDQoUMHhfFE6pCXsf2phIQEpKSkwMrKqqDSJMqVvI7rBQsWoEyZMhgxYkRhpEmUK3kZ18ePH0fTpk0xduxY2NjYoGbNmliyZAnS0tIKK22iHOVlbDdr1gz37t0THwd49eoVTp8+jU6dOhVKzkQFoTDqR51860lDRUREIC0tDTY2NnLbbWxs8M8//2TbJiQkJNv4kJCQAsuTKLfyMrY/9fPPP6NcuXJZflERqUtexvX169exY8cOPHjwoBAyJMq9vIzrV69e4dKlSxgwYABOnz6Nly9f4scff0RKSgrmzp1bGGkT5SgvY7t///6IiIjAV199BUEQkJqaijFjxvBWfyrWFNWPMTExSExMhKGh4Wcfg1f8iShPli1bhoMHD+Kvv/6CgYGButMhypPY2FgMGjQI27dvh7W1tbrTIco3MpkMZcqUwa+//ooGDRqgT58+mDlzJrZu3aru1Ig+i6enJ5YsWYLNmzfDx8cHR48exalTp7Bw4UJ1p0ZUpPGKfw6sra2hra2N0NBQue2hoaGwtbXNto2trW2u4onUIS9jO8OqVauwbNkyXLhwAbVr1y7INIlyJbfj2t/fH4GBgejatau4TSaTAQB0dHTw7NkzVK5cuWCTJspBXn5fly1bFrq6utDW1ha3Va9eHSEhIUhOToaenl6B5kykiryM7dmzZ2PQoEFwdXUFANSqVQvx8fEYNWoUZs6cCS0tXtek4kdR/WhmZpYvV/sBXvHPkZ6eHho0aICLFy+K22QyGS5evIimTZtm26Zp06Zy8QBw/vx5hfFE6pCXsQ0AK1aswMKFC3H27Fl8+eWXhZEqkcpyO66dnJzw+PFjPHjwQPzp1q2bOKuuvb19YaZPlK28/L5u3rw5Xr58KX6RBQDPnz9H2bJlWfRTkZGXsZ2QkJCluM/4git9HjWi4qdQ6sd8myZQgx08eFDQ19cX3N3dBV9fX2HUqFGChYWFEBISIgiCIAwaNEiYNm2aGO/l5SXo6OgIq1atEvz8/IS5c+cKurq6wuPHj9V1CkTZyu3YXrZsmaCnpyccOXJEeP/+vfgTGxurrlMgyiK34/pTnNWfiqLcjuvXr18Lpqamwrhx44Rnz54JJ0+eFMqUKSMsWrRIXadAlK3cju25c+cKpqamwoEDB4RXr14JHh4eQuXKlQUXFxd1nQJRFrGxscL9+/eF+/fvCwCENWvWCPfv3xeCgoIEQRCEadOmCYMGDRLjX716JRgZGQlTpkwR/Pz8hE2bNgna2trC2bNn8y0nFv4q2rBhg1ChQgVBT09PaNSokXDr1i1xX6tWrYQhQ4bIxf/xxx9CtWrVBD09PcHZ2Vk4depUIWdMpJrcjO2KFSsKALL8zJ07t/ATJ1Iit7+zM2PhT0VVbsf1jRs3hMaNGwv6+vpCpUqVhMWLFwupqamFnDVRznIztlNSUoR58+YJlStXFgwMDAR7e3vhxx9/FKKiogo/cSIFLl++nO2/mTPG8pAhQ4RWrVplaVO3bl1BT09PqFSpkrBr1658zUkiCLwnhoiIiIiIiEhT8Rl/IiIiIiIiIg3Gwp+IiIiIiIhIg7HwJyIiIiIiItJgLPyJiIiIiIiINBgLfyIiIiIiIiINxsKfiIiIiIiISIOx8CciIiIiIiLSYCz8iYiIiIiIiDQYC38iIqJiKDo6GitXrkTr1q1RpkwZ6OnpQSKRQCKRwMLCQt3pERERURGio+4EiIiI8su8efMwf/58hfslEglMTU1RqlQp1K5dGy1atMCgQYNQpkyZQszy87148QJt27bFmzdv1J0KERERFQMs/ImIqMQQBAExMTGIiYlBQEAA/v77b8yYMQPTpk3D7NmzoaNTPP5a7N+/v1zRX7lyZTg4OIj5m5iYqCs1IiIiKoKKx79wiIiI8qBly5YwNDQU1wVBQFRUFB4/fgypVAoASE5OxoIFC+Dn54dDhw5BIpGoK12V+Pj44O7du+L67t27MXjwYDVmREREREUdC38iItJYu3fvhoODQ5btUqkU27dvx88//4zExEQAwOHDh9GqVSuMHTu2kLPMHW9vb3G5QoUKLPqJiIgoR5zcj4iIShwDAwP89NNP+Pvvv+Wu8C9atAgymUyNmeUsMjJSXLa3t1djJkRERFRcsPAnIqISq3379ujevbu4HhISgnv37qkxo5ylpKSIy8VlTgIiIiJSLxb+RERUonXu3Flu/eHDh9nGJSYmYufOnfj+++9RpUoVmJubw9DQEBUqVEDnzp2xefNmJCQk5Hi8wMBA8bV7EokEgYGBAICPHz9iy5YtaNeuHRwcHGBgYACJRIJffvkFQ4cOFeMzv7XgypUrcn192md2PDw8MHr0aDg7O8PKygr6+vqws7NDixYtsGjRIgQFBeX8oQFwd3cXj5f5cYpXr15hwYIFaNy4McqVKwcdHR1IJBI8ePBA6flHR0dj48aNaNmyJezs7KCnpwdbW1v06NED58+fV/hZTpkyBbVr14aJiQkMDQ1RuXJljBkzBi9fvlTpPID0z/7QoUP48ccf0bx5c9ja2sLAwACGhoaws7NDmzZtMH/+fLx9+1blPjOfo6enJwAgNTUVhw8fRteuXeHo6AgDAwNYW1ujSZMmWLRoEaKjo1XuP4NMJsOxY8cwevRo1KpVC2XKlIGuri7MzMzg5OQEFxcXbNmyBaGhoSr1l1/jnIiIihiBiIhIQ8ydO1cAIP4EBATk2Obs2bNybZYuXZolZt++fUK5cuXk4rL7KVeunHDy5EmlxwsICMiS48WLFwU7O7ts+1y7dq0wZMiQHI+d03n7+/sLLVu2zLGtvr6+MHPmTCEtLU3peezatUtsU7FiRUEQBGHdunWCgYFBtv3ev39f4fnfuHFDqFChgtK8ZsyYIXf8DRs2CPr6+krP448//lB6DoIgCLNnzxb09PRU+lx1dXWFOXPmCDKZLMd+M7e7fPmyEBgYKDRt2lRp/2XKlBHu3LmTY98ZPDw8hOrVq6ucu7e3t9L+8nOcExFR0cJ7BImIqETLfOs8AOjp6cmtz5gxA0uXLpXbZmtri0qVKkFXVxeBgYHiVfLg4GB0794dO3fuVHnSPW9vbwwePBhJSUkAgEqVKsHe3h7x8fF4/vw5AKBWrVro0KEDAODly5fw9/cHAFhaWqJRo0ZZ+sz8JgMAePz4Mdq3by931VdXVxc1a9aEubk5Xr9+jVevXgEAkpKSsHjxYvEtB6o+TrBu3TpMmDABAKCtrY0aNWrA2toaERER+OeffxS2e/LkCfr27Yv4+HhoaWmhRo0aKF26NMLCwvD06VMxbsmSJShbtizGjRuHVatWYcqUKQDS52uoWbMmTE1N4e/vj9evX4vn0b9/f1StWhV169ZVeHxfX18kJyeL67a2tqhQoQJMTU0hlUrx6tUrvH//HkD6WFmwYAFCQkKwbds2lT4XAAgLC8OQIUPE3CpUqAAHBwekpKTg0aNHiI+PF+M6dOiAp0+fwtbWVmmf69evx6RJk5CWliZu09HRQfXq1WFtbQ2pVIqgoCAEBweLuWccJzsFPc6JiEjN1P3NAxERUX7JyxX/FStWyLXZt2+fuG/Lli1y+7p06SLcu3cvSx/37t2Tu5prYGAgPHr0KNvjfXrF28zMTOz7n3/+kYtNTEwU3r59q/AcW7VqleP5xcXFCVWrVhXbaGlpCVOnThU+fPggF+fj4yM0atRILrc5c+Yo7DfzFX9DQ0NBR0dHkEgkwtSpU4WIiAi52NDQUCE6Ojrb8y9VqpQAQBg+fLgQHBws1+7JkyfCF198IcaWLl1a8PDwELS0tAR9fX1h5cqVQnx8vFybv/76SzAyMhLbtGnTRunn4+LiInTp0kXYu3evEBYWlm3M48ePhV69esnl/ffffyvtN3OstbW1+N/Lx8dHLi4xMVGYOXOmXLyrq6vSvo8ePSoXb2VlJaxbt074+PFjltjXr18La9euFapWrSpcvnw52/4KYpwTEVHRwsKfiIg0Rl4K/3r16sm1efnypSAIghAYGCh32/rPP/+stJ+kpCShdevWYnynTp2yjfu08AUgDBw4UKXbxz89R1UK/0+Lyo0bNyqMTUhIEJo1aybGamtrCy9evMg2NnPhn/Gzffv2HPPJ7vynTZumMN7Pz0/Q0dERY/X09ASJRCKcPn1aYZtt27aJ8RKJRAgMDFQYGxsbm2POGUaPHi3226xZM6Wxn55j165dhZSUFIXxI0aMEGNNTEyEhISEbOOioqIEc3NzMbZ8+fLCs2fPcsw9LS0ty5ckglBw45yIiIoWTu5HREQl1vz583H//n1xvX79+qhcuTIAYO3atZBKpQCAZs2aZbkN+lN6enrYuXOneGv8mTNnxFvylbG2tsbmzZvlXiuYX6RSqdwt6e3bt8fYsWMVxhsaGmL37t3i4w5paWnYsGGDSsf65ptv4Orqmuscq1WrhoULFyrc7+TkhK+//lpcT05OxrBhw/Dtt98qbDN48GCYmJgAAARBgJeXl8LYjDhVrFixAgYGBgCAGzduqDxhnqmpKdzd3ZU+NjF16lRxOS4uTm5cZrZx40ZxEkCJRIL9+/ejWrVqOeagpaUFIyOjLNsLa5wTEZF6sfAnIqISQxAEREVF4fz58+jWrRvmzZsn7pNIJFi+fDmA9JnS9+zZI+5zc3NTqTB3dHREixYtxGNdvHgxxzb9+/eHqalpLs9ENdeuXUNERIS47ubmlmObKlWqyL3i8OjRoyoda8yYMblPEMDw4cNznEegSZMmcuujRo1SGm9gYCD3XL+vr2+ecvuUmZkZnJ2dxXVvb2+V2vXt2xdWVlZKY6pVqyb3XH/m+Q0y279/v7jcrl07tGzZUqUcslOY45yIiNSLk/sREZHGcnR0VCku47V57dq1A5A+GV5UVJS4P2O7KurUqYPLly8DAO7evZtjkfo5hVtObt68KS4bGhqqfB7du3fH4cOHAQBv377FmzdvYG9vr7RNXs+jadOmOcZkLoj19PTQoEGDHNuULVtWXM7831KZZ8+e4cqVK3jy5AnCw8MRGxuL1NRUuZiMSRABqPx6v+bNm6sUV758eYSEhABIf8Xgp0JDQ+UmSuzTp49K/SpSmOOciIjUi4U/ERGVaA0aNMDq1avRqlUrcdujR4/EZW1tbbi4uKjcX+b3x4eHh+cYn/FoQUF48eKFuOzs7AxtbW2V2tWuXVtu/eXLl0oLf3Nzc5QqVSpPOeY0ez0AuVvUS5UqpdKbBjK3yem989evX8fkyZNx+/btHPvNLLviPDuqnCMAGBsbi8vZ5ezn5ye33rBhQ5X6VaQwxzkREakXC38iItJYLVu2lHu1nZaWFkxMTGBpaYlatWqhZcuWWYpcAIiMjBSX09LScO7cuTwdP+NZbGXMzMzy1LcqMl/NLV26tMrtPo3N6Yr555zDp69PzO94IP12dEV+/fVXjBkzRmmMIhmvYMyJvr5+rvvOLp8PHz7IrZcpUybX/WZWmOOciIjUi4U/ERFprN27d8PBwSHX7ZS97zw3ZDJZjjFaWgU33U7mwjQ3BfOnhWrG5G+KFOQ5FKT79+/jhx9+EItsKysrDBs2DG3btkXVqlVhY2MDAwMD6Orqim1at26NK1euqCXfT/875OULhcwKc5wTEZF6sfAnIiL6hIWFhbhsZGSUbwVSYct8HrGxsSq3i4mJUdiPJlm8eLFYtDo4OMDLywvlypVT2iY3n2N+s7S0lFuPjo7Osi03NGWcExFRzornV/REREQFKPMz2QkJCYiLi1NjNnmX+Vbw3Lxy7dPYz72lvCgSBEHu1vY5c+bkWPQDqk/oVxAyT1gIpE9G+Dk0ZZwTEVHOWPgTERF94tOZ5jPPjl+cZJ79PigoCMHBwSq1u3Hjhrisr6+PWrVq5Xtu6vbhwwe5QrdRo0Y5tnn+/DnCwsIKMi2latWqJffqx4xZ9fNKU8Y5ERHljIU/ERHRJ8qVK4c6deqI6zt27FBjNnmX+U0FALB3794c28hkMuzbt09cb9KkyWc/S14UpaSk5LqNuseBtra23Cv33N3dkZiYmOf+NGWcExFRzlj4ExERZWPKlCni8uHDh3Hq1Ck1ZpM3lStXRtu2bcX15cuXIzQ0VGmbbdu2yd1CPnr06ALLT52sra3lJjy8evWq0nhfX1+sX7++oNPK0cSJE8Xl0NBQTJ48+bP604RxTkREOWPhT0RElI1+/fqhefPmANKvgru4uGDXrl05vvYtMTER+/fvl7vNXp3mzJkjzrofFRWFTp06ISQkJNvYY8eOYcKECeJ6zZo10bt378JIs9Dp6OigRYsW4vqCBQsQEBCQbezjx4/RoUOHHN9uUBhatGiBHj16iOubN2/GhAkTlOaWmpqKffv2wdfXN8s+TRnnRESkHGf1JyIiyoaWlhb+/PNPNG7cGEFBQUhISMDw4cOxcuVKfPfdd2jQoAFKlSqFlJQUREVFwc/PD3fu3MGFCxeQkJCg7vRFLVu2xJQpU7B8+XIAgI+PD2rUqAFXV1e0aNECZmZmePPmDY4cOYK///5bbGdoaIj9+/fLvcpO07i5ueHixYsAgJCQENSvXx9jxoxBy5YtYWJiguDgYJw+fRq///47UlNTUa9ePejq6sLb21utee/atQtPnz7FixcvAADr1q3Dn3/+if79+6NJkyawtraGVCpFUFAQbt68iePHjyMiIgKXL19GjRo15PrSlHFORETKsfAnIiJSwMbGBrdv38b333+Pa9euAQD8/PywaNEiNWeWO0uXLkVqaipWr14NIP3K/8qVK7Fy5cps462srHDy5EnUrl27MNMsdN9++y0mTZqENWvWAAA+fvyIZcuWYdmyZVliK1WqhKNHj2Lo0KGFnGVWFhYW8PLyQs+ePeHl5QUg/W0DK1asyFN/mjLOiYhIMd7qT0REpISNjQ08PT1x8OBBfPnll5BIJErjnZyc4ObmhgcPHhROgiqQSCRYtWoVLly4gCZNmiiMMzIywujRo+Hr65tlxndNtXr1amzevFnhKwuNjY0xatQo3L9/Hw4ODoWbnBKlS5fGlStXsGvXLjg5OSmNtbOzg5ubm9K3M2jCOCciIsUkQk4PcREREZEoPDwcXl5eeP/+PaKioqCjowMLCwtUqlQJNWvWlHs3elEVHByM69ev4/3794iPj0epUqVQuXJlfPXVVzAwMFB3emqRlJSE69ev4+nTp4iLi0OpUqVgb2+P1q1bw8jISN3p5SggIAC3b99GaGgoYmNjYWxsDDs7O9SuXTvHLwayownjnIiI/sPCn4iIiIiIiEiD8VZ/IiIiIiIiIg3Gwp+IiIiIiIhIg7HwJyIiIiIiItJgLPyJiIiIiIiINBgLfyIiIiIiIiINxsKfiIiIiIiISIOx8CciIiIiIiLSYCz8iYiIiIiIiDQYC38iIiIiIiIiDcbCn4iIiIiIiEiDsfAnIiIiIiIi0mAs/ImIiIiIiIg0GAt/IiIiIiIiIg3Gwp+IiIiIiIhIg/0fX4q5aQwnfWkAAAAASUVORK5CYII=\n",
      "text/plain": [
       "<Figure size 1200x900 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "import matplotlib.pyplot as plt\n",
    "import numpy as np\n",
    "\n",
    "# Set the height of the bars\n",
    "bar_height = 0.15\n",
    "\n",
    "# Set the positions of the bars\n",
    "r1 = np.arange(len(df[\"Baseline\"]))\n",
    "r2 = [x + bar_height for x in r1]\n",
    "r3 = [x + bar_height for x in r2]\n",
    "r4 = [x + bar_height for x in r3]\n",
    "r5 = [x + bar_height for x in r4]\n",
    "\n",
    "plt.figure(figsize=(12, 9))\n",
    "\n",
    "# Create the bars with labels for the legend\n",
    "bars1 = plt.barh(r1, df[\"Baseline\"], height=bar_height, label=\"Baseline\")\n",
    "bars2 = plt.barh(r2, df[\"Cutmix\"], height=bar_height, label=\"Cutmix\")\n",
    "bars3 = plt.barh(r3, df[\"Mixup\"], height=bar_height, label=\"Mixup\")\n",
    "bars4 = plt.barh(r4, df[\"Cutmix+Mixup\"], height=bar_height, label=\"Cutmix+Mixup\")\n",
    "bars5 = plt.barh(r5, df[\"Remix\"], height=bar_height, label=\"Remix\")\n",
    "\n",
    "\n",
    "# bars1 = plt.barh(r1, df['Baseline'], height=bar_height, hatch='//', edgecolor='black', label='Baseline')\n",
    "# bars2 = plt.barh(r2, df['Cutmix'], height=bar_height, hatch='\\\\', edgecolor='black', label='Cutmix')\n",
    "# bars3 = plt.barh(r3, df['Mixup'], height=bar_height, hatch='--', edgecolor='black', label='Mixup')\n",
    "# bars4 = plt.barh(r4, df['Cutmix+Mixup'], height=bar_height, hatch='++', edgecolor='black', label='Cutmix+Mixup')\n",
    "# bars5 = plt.barh(r5, df['Remix'], height=bar_height, hatch='xx', edgecolor='black', label='Remix')\n",
    "\n",
    "# bars1 = plt.barh(r1, df['Baseline'], height=bar_height, color='0.2', linewidth=1, label='Baseline')\n",
    "# bars2 = plt.barh(r2, df['Cutmix'], height=bar_height, color='0.4',  linewidth=1, label='Cutmix')\n",
    "# bars3 = plt.barh(r3, df['Mixup'], height=bar_height, color='0.6', linewidth=1, label='Mixup')\n",
    "# bars4 = plt.barh(r4, df['Cutmix+Mixup'], height=bar_height, color='0.8', linewidth=1, label='Cutmix+Mixup')\n",
    "# bars5 = plt.barh(r5, df['Remix'], height=bar_height, color='1.0', linewidth=2, label='Remix')\n",
    "\n",
    "\n",
    "# Add labels and title\n",
    "plt.ylabel(\"Class Label\", fontsize=26)\n",
    "plt.xlabel(\"Performance\", fontsize=26)\n",
    "plt.title(\"Comparison of Techniques across Class Labels\", fontsize=26)\n",
    "plt.yticks(\n",
    "    [r + bar_height * 2 for r in range(len(df[\"Baseline\"]))], df.index, fontsize=26\n",
    ")\n",
    "plt.xlim(0, 1.02)\n",
    "\n",
    "plt.legend(loc=\"lower left\", fontsize=20)\n",
    "# plt.legend(bbox_to_anchor=(0., .25), loc='upper left', fontsize=26)\n",
    "\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "id": "DyqD8hnfXpAW"
   },
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "id": "Yt3PibcnXn5t"
   },
   "source": []
  }
 ],
 "metadata": {
  "accelerator": "GPU",
  "colab": {
   "gpuType": "T4",
   "include_colab_link": true,
   "provenance": []
  },
  "kernelspec": {
   "display_name": "Python 3",
   "name": "python3"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.6.9"
  },
  "vscode": {
   "interpreter": {
    "hash": "36108061fb70d79a201768ae61599c18cd6a173316f5067585092ff73658f017"
   }
  }
 },
 "nbformat": 4,
 "nbformat_minor": 0
}