FittingRandomDataWithWideResnet.ipynb

{
  "nbformat": 4,
  "nbformat_minor": 0,
  "metadata": {
    "colab": {
      "name": "FittingRandomDataWithWideResnet.ipynb",
      "version": "0.3.2",
      "provenance": [],
      "collapsed_sections": []
    },
    "kernelspec": {
      "name": "python3",
      "display_name": "Python 3"
    },
    "accelerator": "GPU"
  },
  "cells": [
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "view-in-github",
        "colab_type": "text"
      },
      "source": [
        "[View in Colaboratory](https://colab.research.google.com/github/kobybibas/FittingRandomDataToResnetCNN/blob/master/FittingRandomDataWithWideResnet.ipynb)"
      ]
    },
    {
      "metadata": {
        "id": "mH6ZDl62tO3Y",
        "colab_type": "text"
      },
      "cell_type": "markdown",
      "source": [
        "# Fitting Random Data to Wide Resnet Model\n",
        "This notebook replicates the results from [\"UNDERSTANDING DEEP LEARNING REQUIRES RETHINKING\n",
        "GENERALIZATION\" ](https://arxiv.org/abs/1611.03530).\n",
        "\n",
        "In this notebook a Resnet model is trained on the CIFAR10 dataset.  The tranining can be done on random labels of the dataset.\n",
        "\n",
        "The neural network model was taken from https://github.com/xternalz/WideResNet-pytorch.\n",
        "\n",
        "All the noetbook was inspired by https://github.com/pluskid/fitting-random-labels.\n",
        "\n",
        "The notebook was developed and tested on https://colab.research.google.com/ with GPU backend.\n",
        "\n",
        "\n",
        "\n"
      ]
    },
    {
      "metadata": {
        "id": "ilG60-xXu0du",
        "colab_type": "text"
      },
      "cell_type": "markdown",
      "source": [
        "# Install Dependencies\n",
        "Installing pytorch"
      ]
    },
    {
      "metadata": {
        "id": "owtRx1yDEmdk",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 190
        },
        "outputId": "9caae94c-c988-47c7-e3c5-04f5ec9790b4"
      },
      "cell_type": "code",
      "source": [
        "# Install pytorch\n",
        "!pip3 install http://download.pytorch.org/whl/cu80/torch-0.3.0.post4-cp36-cp36m-linux_x86_64.whl \n",
        "!pip3 install torchvision\n",
        "\n",
        "# Import dependecies \n",
        "import os\n",
        "import math\n",
        "import time\n",
        "import numpy as np\n",
        "import matplotlib.pyplot as plt\n",
        "import torch\n",
        "import torch.optim\n",
        "import torch.nn as nn\n",
        "import torch.nn.functional as F\n",
        "import torch.backends.cudnn as cudnn\n",
        "import torchvision.datasets as datasets\n",
        "import torchvision.transforms as transforms"
      ],
      "execution_count": 4,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Requirement already satisfied: torch==0.3.0.post4 from http://download.pytorch.org/whl/cu80/torch-0.3.0.post4-cp36-cp36m-linux_x86_64.whl in /usr/local/lib/python3.6/dist-packages (0.3.0.post4)\r\n",
            "Requirement already satisfied: pyyaml in /usr/local/lib/python3.6/dist-packages (from torch==0.3.0.post4) (3.12)\r\n",
            "Requirement already satisfied: numpy in /usr/local/lib/python3.6/dist-packages (from torch==0.3.0.post4) (1.14.3)\n",
            "Requirement already satisfied: torchvision in /usr/local/lib/python3.6/dist-packages (0.2.1)\n",
            "Requirement already satisfied: six in /usr/local/lib/python3.6/dist-packages (from torchvision) (1.11.0)\n",
            "Requirement already satisfied: numpy in /usr/local/lib/python3.6/dist-packages (from torchvision) (1.14.3)\n",
            "Requirement already satisfied: pillow>=4.1.1 in /usr/local/lib/python3.6/dist-packages (from torchvision) (5.1.0)\n",
            "Requirement already satisfied: torch in /usr/local/lib/python3.6/dist-packages (from torchvision) (0.3.0.post4)\n",
            "Requirement already satisfied: pyyaml in /usr/local/lib/python3.6/dist-packages (from torch->torchvision) (3.12)\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "metadata": {
        "id": "nOTSG415FMsa",
        "colab_type": "text"
      },
      "cell_type": "markdown",
      "source": [
        "# Load the Dataset\n",
        "1. Downloading CIFAR10 dataset.\n",
        "2. Corrupt the labels based on user's input.\n",
        "3. Load it to pytorch Dataloader."
      ]
    },
    {
      "metadata": {
        "id": "D2UL3wODF4Na",
        "colab_type": "code",
        "colab": {}
      },
      "cell_type": "code",
      "source": [
        "# Define general functions\n",
        "\n",
        "class CIFAR10RandomLabels(datasets.CIFAR10):\n",
        "    \"\"\"CIFAR10 dataset, with support for randomly corrupt labels.\n",
        "\n",
        "    Params\n",
        "    ------\n",
        "    corrupt_prob: float\n",
        "        Default 0.0. The probability of a label being replaced with\n",
        "        random label.\n",
        "    num_classes: int\n",
        "        Default 10. The number of classes in the dataset.\n",
        "    \"\"\"\n",
        "    def __init__(self, corrupt_prob=0.0, num_classes=10, **kwargs):\n",
        "        super(CIFAR10RandomLabels, self).__init__(**kwargs)\n",
        "        self.n_classes = num_classes\n",
        "        if corrupt_prob > 0:\n",
        "            self.corrupt_labels(corrupt_prob)\n",
        "\n",
        "    def corrupt_labels(self, corrupt_prob):\n",
        "        labels = np.array(self.train_labels if self.train else self.test_labels)\n",
        "        np.random.seed(12345)\n",
        "        mask = np.random.rand(len(labels)) <= corrupt_prob\n",
        "        rnd_labels = np.random.choice(self.n_classes, mask.sum())\n",
        "        labels[mask] = rnd_labels\n",
        "        # we need to explicitly cast the labels from npy.int64 to\n",
        "        # builtin int type, otherwise pytorch will fail...\n",
        "        labels = [int(x) for x in labels]\n",
        "\n",
        "        if self.train:\n",
        "            self.train_labels = labels\n",
        "        else:\n",
        "            self.test_labels = labels\n",
        "\n",
        "            \n",
        "def get_data_loaders(args, shuffle_train=True):\n",
        "\n",
        "    # Normalization for CIFAR10\n",
        "    normalize = transforms.Normalize(mean=[x/255.0 for x in [125.3, 123.0, 113.9]],\n",
        "                                     std=[x / 255.0 for x in [63.0, 62.1, 66.7]])\n",
        "\n",
        "    if args['data_augmentation']==True:\n",
        "        transform_train = transforms.Compose([\n",
        "          transforms.RandomCrop(32, padding=4),\n",
        "          transforms.RandomHorizontalFlip(),\n",
        "          transforms.ToTensor(),\n",
        "          normalize,\n",
        "          ])\n",
        "    else:\n",
        "        transform_train = transforms.Compose([\n",
        "          transforms.ToTensor(),\n",
        "          normalize,\n",
        "          ])\n",
        "\n",
        "    transform_test = transforms.Compose([\n",
        "        transforms.ToTensor(),\n",
        "        normalize\n",
        "        ])\n",
        "\n",
        "    kwargs = {'num_workers': 1, 'pin_memory': True}\n",
        "    train_loader = torch.utils.data.DataLoader(\n",
        "        CIFAR10RandomLabels(root='./data', train=True, download=True,\n",
        "                            transform=transform_train, num_classes=args['num_classes'],\n",
        "                            corrupt_prob=args['label_corrupt_prob']),\n",
        "        batch_size=args['batch_size'], shuffle=shuffle_train, **kwargs)\n",
        "    val_loader = torch.utils.data.DataLoader(\n",
        "        CIFAR10RandomLabels(root='./data', train=False,\n",
        "                            transform=transform_test, num_classes=args['num_classes'],\n",
        "                            corrupt_prob=args['label_corrupt_prob']),\n",
        "        batch_size=args['batch_size'], shuffle=False, **kwargs)\n",
        "    \n",
        "    # Labels of CIFAR10 dataset\n",
        "    label_translation = ['airplane', 'automobile', 'bird', 'cat', 'deer', 'dog', 'frog',\n",
        "                         'horse', 'ship', 'truck']\n",
        "\n",
        "    return train_loader, val_loader, label_translation"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "metadata": {
        "id": "F2AdaB_6QQsc",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 396
        },
        "outputId": "0301b6d2-c1a2-4005-e7c1-7053d2cff58f"
      },
      "cell_type": "code",
      "source": [
        "# Define hyperparameters\n",
        "args = {}\n",
        "args['num_classes'] = 10 \n",
        "args['data_augmentation'] = False \n",
        "args['label_corrupt_prob'] = 0.0 # replace with 1.0 for totaly random labels\n",
        "args['batch_size'] = 128\n",
        " \n",
        "# Load the data\n",
        "train_loader, val_loader, label_translation = get_data_loaders(args, shuffle_train=True)\n",
        "\n",
        "# Visualize\n",
        "imgs, label = next(iter(train_loader))\n",
        "print('Batch shape:',imgs.numpy().shape)\n",
        "\n",
        "for i in range (0,9):\n",
        "    idx = np.random.randint(0, imgs.numpy().shape[0])\n",
        "    img_for_plot = imgs.numpy()[idx,:,:,:].transpose(1, 2, 0)\n",
        "    img_for_plot -= np.min(img_for_plot, axis=0)\n",
        "    img_for_plot /= np.ptp(img_for_plot, axis=0)\n",
        "    label_name = label_translation[label[idx]]\n",
        "    \n",
        "    plt.subplot(3,3,i+1)\n",
        "    plt.imshow(img_for_plot)\n",
        "    plt.title(label_name)\n",
        "plt.show()"
      ],
      "execution_count": 6,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Files already downloaded and verified\n",
            "Batch shape: (128, 3, 32, 32)\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "display_data",
          "data": {
            "image/png": "iVBORw0KGgoAAAANSUhEUgAAAbMAAAFZCAYAAAASMsFjAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nU1BQUFDIe6jFTEFBQUEh7/H/AfvmTj7YiRg2AAAAAElFTkSuQmCC\n",
            "text/plain": [
              "<matplotlib.figure.Figure at 0x7f58c1ce4da0>"
            ]
          },
          "metadata": {
            "tags": []
          }
        }
      ]
    },
    {
      "metadata": {
        "id": "5nvbpABeQom-",
        "colab_type": "text"
      },
      "cell_type": "markdown",
      "source": [
        "# Load the Model\n",
        "1. Define wide Resnet neural netowrk.\n",
        "2. Load it to pytorch model."
      ]
    },
    {
      "metadata": {
        "id": "h3ffDBuiGFTk",
        "colab_type": "code",
        "colab": {}
      },
      "cell_type": "code",
      "source": [
        "#Define the network\n",
        "\n",
        "class BasicBlock(nn.Module):\n",
        "    def __init__(self, in_planes, out_planes, stride, dropRate=0.0):\n",
        "        super(BasicBlock, self).__init__()\n",
        "        self.bn1 = nn.BatchNorm2d(in_planes)\n",
        "        self.relu1 = nn.ReLU(inplace=True)\n",
        "        self.conv1 = nn.Conv2d(in_planes, out_planes, kernel_size=3, stride=stride,\n",
        "                                padding=1, bias=False)\n",
        "        self.bn2 = nn.BatchNorm2d(out_planes)\n",
        "        self.relu2 = nn.ReLU(inplace=True)\n",
        "        self.conv2 = nn.Conv2d(out_planes, out_planes, kernel_size=3, stride=1,\n",
        "                                padding=1, bias=False)\n",
        "        self.droprate = dropRate\n",
        "        self.equalInOut = (in_planes == out_planes)\n",
        "        self.convShortcut = (not self.equalInOut) and nn.Conv2d(in_planes, out_planes, kernel_size=1, stride=stride,\n",
        "                                                                padding=0, bias=False) or None\n",
        "\n",
        "    def forward(self, x):\n",
        "        if not self.equalInOut:\n",
        "            x = self.relu1(self.bn1(x))\n",
        "            out = self.conv1(x)\n",
        "        else:\n",
        "            out = self.conv1(self.relu1(self.bn1(x)))\n",
        "\n",
        "        if self.droprate > 0:\n",
        "            out = F.dropout(out, p=self.droprate, training=self.training)\n",
        "        out = self.conv2(self.relu2(self.bn2(out)))\n",
        "        if not self.equalInOut:\n",
        "            return torch.add(self.convShortcut(x), out)\n",
        "        else:\n",
        "            return torch.add(x, out)\n",
        "\n",
        "\n",
        "class NetworkBlock(nn.Module):\n",
        "        def __init__(self, nb_layers, in_planes, out_planes, block, stride, dropRate=0.0):\n",
        "            super(NetworkBlock, self).__init__()\n",
        "            self.layer = self._make_layer(block, in_planes, out_planes, nb_layers, stride, dropRate)\n",
        "\n",
        "        def _make_layer(self, block, in_planes, out_planes, nb_layers, stride, dropRate):\n",
        "            layers = []\n",
        "            for i in range(nb_layers):\n",
        "                layers.append(block(i == 0 and in_planes or out_planes, out_planes, i == 0 and stride or 1, dropRate))\n",
        "            return nn.Sequential(*layers)\n",
        "\n",
        "        def forward(self, x):\n",
        "            return self.layer(x)\n",
        "\n",
        "\n",
        "class WideResNet(nn.Module):\n",
        "    def __init__(self, depth, num_classes, widen_factor=1, drop_rate=0.0, init_scale=1.0):\n",
        "        super(WideResNet, self).__init__()\n",
        "\n",
        "        nChannels = [16, 16 * widen_factor, 32 * widen_factor, 64 * widen_factor]\n",
        "        assert((depth - 4) % 6 == 0)\n",
        "        n = (depth - 4) // 6\n",
        "        block = BasicBlock\n",
        "        # 1st conv before any network block\n",
        "        self.conv1 = nn.Conv2d(3, nChannels[0], kernel_size=3, stride=1,\n",
        "                                padding=1, bias=False)\n",
        "        # 1st block\n",
        "        self.block1 = NetworkBlock(\n",
        "            n, nChannels[0], nChannels[1], block, 1, drop_rate)\n",
        "        # 2nd block\n",
        "        self.block2 = NetworkBlock(n, nChannels[1], nChannels[2], block, 2, drop_rate)\n",
        "        # 3rd block\n",
        "        self.block3 = NetworkBlock(n, nChannels[2], nChannels[3], block, 2, drop_rate)\n",
        "        # global average pooling and classifier\n",
        "        self.bn1 = nn.BatchNorm2d(nChannels[3])\n",
        "        self.relu = nn.ReLU(inplace=True)\n",
        "        self.fc = nn.Linear(nChannels[3], num_classes)\n",
        "        self.nChannels = nChannels[3]\n",
        "\n",
        "        for m in self.modules():\n",
        "            if isinstance(m, nn.Conv2d):\n",
        "                n = m.kernel_size[0] * m.kernel_size[1] * m.out_channels\n",
        "                m.weight.data.normal_(0, init_scale * math.sqrt(2. / n))\n",
        "            elif isinstance(m, nn.BatchNorm2d):\n",
        "                m.weight.data.fill_(1)\n",
        "                m.bias.data.zero_()\n",
        "            elif isinstance(m, nn.Linear):\n",
        "                m.bias.data.zero_()\n",
        "\n",
        "                size = m.weight.size()\n",
        "                fan_out = size[0] # number of rows\n",
        "                fan_in = size[1] # number of columns\n",
        "                variance = math.sqrt(2.0/(fan_in + fan_out))\n",
        "                m.weight.data.normal_(0.0, init_scale * variance)\n",
        "\n",
        "\n",
        "    def forward(self, x):\n",
        "        out = self.forward_repr(x)\n",
        "        return self.fc(out)\n",
        "\n",
        "    def forward_repr(self, x):\n",
        "        out = self.conv1(x)\n",
        "        out = self.block1(out)\n",
        "        out = self.block2(out)\n",
        "        out = self.block3(out)\n",
        "        out = self.relu(self.bn1(out))\n",
        "        out = F.avg_pool2d(out, 8)\n",
        "        out = out.view(-1, self.nChannels)\n",
        "        return out\n"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "metadata": {
        "id": "k9-MsyTBG07g",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 34
        },
        "outputId": "308905d6-ee8c-4dd2-9438-feebacd07de9"
      },
      "cell_type": "code",
      "source": [
        "#Load the model\n",
        "args['wrn_depth'] = 28\n",
        "args['wrn_widen_factor'] = 1\n",
        "drop_rate=args['wrn_droprate'] = 0.0\n",
        "\n",
        "model = WideResNet(args['wrn_depth'], \n",
        "                   args['num_classes'],\n",
        "                   args['wrn_widen_factor'],\n",
        "                   drop_rate=args['wrn_droprate'])    \n",
        "model = torch.nn.DataParallel(model).cuda()\n",
        "model = model.cuda() \n",
        "print('Parameters number: %d' % (sum([p.data.nelement() for p in model.parameters()])))"
      ],
      "execution_count": 8,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Parameters number: 369498\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "metadata": {
        "id": "kqPUG-vDRtJU",
        "colab_type": "text"
      },
      "cell_type": "markdown",
      "source": [
        "# Train the Model"
      ]
    },
    {
      "metadata": {
        "id": "jOnoQOI2Fn7m",
        "colab_type": "code",
        "colab": {}
      },
      "cell_type": "code",
      "source": [
        "# Define general functions for training\n",
        "\n",
        "def train_epoch(train_loader, model, criterion, optimizer, epoch, args):\n",
        "    \"\"\"Train for one epoch on the training set\"\"\"\n",
        "    batch_time = AverageMeter()\n",
        "    losses = AverageMeter()\n",
        "    top1 = AverageMeter()\n",
        "\n",
        "    # switch to train mode\n",
        "    model.train()\n",
        "\n",
        "    for i, (input, target) in enumerate(train_loader):\n",
        "        target = target.cuda(async=True)\n",
        "        input = input.cuda()\n",
        "        input_var = torch.autograd.Variable(input)\n",
        "        target_var = torch.autograd.Variable(target)\n",
        "\n",
        "        # compute output\n",
        "        output = model(input_var)\n",
        "        loss = criterion(output, target_var)\n",
        "\n",
        "        # measure accuracy and record loss\n",
        "        prec1 = accuracy(output.data, target, topk=(1,))[0]\n",
        "        losses.update(loss.data[0], input.size(0))\n",
        "        top1.update(prec1[0], input.size(0))\n",
        "\n",
        "        # compute gradient and do SGD step\n",
        "        optimizer.zero_grad()\n",
        "        loss.backward()\n",
        "        optimizer.step()\n",
        "\n",
        "    return losses.avg, top1.avg\n",
        "\n",
        "\n",
        "def validate_epoch(val_loader, model, criterion, epoch, args):\n",
        "    \"\"\"Perform validation on the validation set\"\"\"\n",
        "    batch_time = AverageMeter()\n",
        "    losses = AverageMeter()\n",
        "    top1 = AverageMeter()\n",
        "\n",
        "    # switch to evaluate mode\n",
        "    model.eval()\n",
        "\n",
        "    for i, (input, target) in enumerate(val_loader):\n",
        "        target = target.cuda(async=True)\n",
        "        input = input.cuda()\n",
        "        input_var = torch.autograd.Variable(input, volatile=True)\n",
        "        target_var = torch.autograd.Variable(target, volatile=True)\n",
        "\n",
        "        # compute output\n",
        "        output = model(input_var)\n",
        "        loss = criterion(output, target_var)\n",
        "\n",
        "        # measure accuracy and record loss\n",
        "        prec1 = accuracy(output.data, target, topk=(1,))[0]\n",
        "        losses.update(loss.data[0], input.size(0))\n",
        "        top1.update(prec1[0], input.size(0))\n",
        "\n",
        "    return losses.avg, top1.avg\n",
        "\n",
        "\n",
        "class AverageMeter(object):\n",
        "    \"\"\"Computes and stores the average and current value\"\"\"\n",
        "    def __init__(self):\n",
        "        self.reset()\n",
        "\n",
        "    def reset(self):\n",
        "        self.val = 0\n",
        "        self.avg = 0\n",
        "        self.sum = 0\n",
        "        self.count = 0\n",
        "\n",
        "    def update(self, val, n=1):\n",
        "        self.val = val\n",
        "        self.sum += val * n\n",
        "        self.count += n\n",
        "        self.avg = self.sum / self.count\n",
        "\n",
        "\n",
        "def accuracy(output, target, topk=(1,)):\n",
        "    \"\"\"Computes the precision@k for the specified values of k\"\"\"\n",
        "    maxk = max(topk)\n",
        "    batch_size = target.size(0)\n",
        "\n",
        "    _, pred = output.topk(maxk, 1, True, True)\n",
        "    pred = pred.t()\n",
        "    correct = pred.eq(target.view(1, -1).expand_as(pred))\n",
        "\n",
        "    res = []\n",
        "    for k in topk:\n",
        "        correct_k = correct[:k].view(-1).float().sum(0)\n",
        "        res.append(correct_k.mul_(100.0 / batch_size))\n",
        "    return res"
      ],
      "execution_count": 0,
      "outputs": []
    },
    {
      "metadata": {
        "id": "N81oiRN6R9l8",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 2723
        },
        "outputId": "3b408241-9b27-4cb4-b6d4-37fbcdde6e57"
      },
      "cell_type": "code",
      "source": [
        "# Define hyperparameters\n",
        "args['num_classes'] = 10 \n",
        "args['epochs'] = 300 \n",
        "args['learning_rate'] = 0.1\n",
        "args['momentum'] = 0.9\n",
        "args['weight_decay'] = 1e-4\n",
        "args['eval_full_trainset'] = True\n",
        "\n",
        "# It is possible to countinue training from saved model\n",
        "#model_path = 'Model_corruptiton_1.0_Acc-tr_61.51.pt'\n",
        "#model = torch.load(model_path)\n",
        "\n",
        "# Train\n",
        "cudnn.benchmark = True\n",
        "\n",
        "# define loss function (criterion) and optimizer\n",
        "criterion = nn.CrossEntropyLoss().cuda()\n",
        "optimizer = torch.optim.SGD(model.parameters(), args['learning_rate'],\n",
        "                          momentum=args['momentum'],\n",
        "                          weight_decay=args['weight_decay'])\n",
        "\n",
        "# Iterate on ephoces\n",
        "for epoch in range(0, args['epochs']):\n",
        "    epoch_time_start = time.time()\n",
        "\n",
        "    # Sets the learning rate to the initial LR decayed by 10 after 150 and 225 epochs\n",
        "    lr = args['learning_rate'] * (0.1 ** (epoch // 150)) * (0.1 ** (epoch // 225))\n",
        "    for param_group in optimizer.param_groups:\n",
        "        param_group['lr'] = lr\n",
        "\n",
        "    # train for one epoch\n",
        "    tr_loss, tr_prec1 = train_epoch(train_loader, model, criterion, optimizer, epoch, args)\n",
        "\n",
        "    # evaluate on validation set\n",
        "    val_loss, val_prec1 = validate_epoch(val_loader, model, criterion, epoch, args)\n",
        "\n",
        "    if args['eval_full_trainset'] == True:\n",
        "        tr_loss, tr_prec1 = validate_epoch(train_loader, model, criterion, epoch, args)\n",
        "\n",
        "    epoch_time = time.time() - epoch_time_start\n",
        "\n",
        "    print('%03d: Acc-tr: %6.2f, Acc-val: %6.2f, L-tr: %6.4f, L-val: %6.4f, epochTime=%10.4f' \n",
        "                 % (epoch, tr_prec1, val_prec1, tr_loss, val_loss, epoch_time))    \n",
        "\n",
        "    model_file_name = 'Model_corruptiton_%2.1f_Acc-tr_%4.2f.pt' % (args['label_corrupt_prob'], tr_prec1)\n",
        "    \n",
        "    # Save the model\n",
        "    if epoch % 50 == 0:\n",
        "        torch.save(model, model_file_name)\n",
        "    \n",
        "    if tr_prec1 > 99.5:\n",
        "        break\n",
        "\n",
        "torch.save(model, model_file_name)"
      ],
      "execution_count": 10,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "000: Acc-tr:  60.08, Acc-val:  10.35, L-tr: 1.1664, L-val: 4.2955, epochTime=   49.2749\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "stream",
          "text": [
            "/usr/local/lib/python3.6/dist-packages/torch/serialization.py:158: UserWarning: Couldn't retrieve source code for container of type WideResNet. It won't be checked for correctness upon loading.\n",
            "  \"type \" + obj.__name__ + \". It won't be checked \"\n",
            "/usr/local/lib/python3.6/dist-packages/torch/serialization.py:158: UserWarning: Couldn't retrieve source code for container of type NetworkBlock. It won't be checked for correctness upon loading.\n",
            "  \"type \" + obj.__name__ + \". It won't be checked \"\n",
            "/usr/local/lib/python3.6/dist-packages/torch/serialization.py:158: UserWarning: Couldn't retrieve source code for container of type BasicBlock. It won't be checked for correctness upon loading.\n",
            "  \"type \" + obj.__name__ + \". It won't be checked \"\n"
          ],
          "name": "stderr"
        },
        {
          "output_type": "stream",
          "text": [
            "001: Acc-tr:  64.60, Acc-val:   9.82, L-tr: 1.0240, L-val: 3.9389, epochTime=   47.7215\n",
            "002: Acc-tr:  49.13, Acc-val:  10.17, L-tr: 1.6116, L-val: 5.1932, epochTime=   47.8408\n",
            "003: Acc-tr:  63.03, Acc-val:  10.28, L-tr: 1.0768, L-val: 4.2952, epochTime=   47.8763\n",
            "004: Acc-tr:  66.64, Acc-val:  10.25, L-tr: 0.9600, L-val: 4.3498, epochTime=   47.6391\n",
            "005: Acc-tr:  66.50, Acc-val:  10.13, L-tr: 0.9555, L-val: 4.1424, epochTime=   47.6955\n",
            "006: Acc-tr:  63.96, Acc-val:  10.11, L-tr: 1.0150, L-val: 4.4573, epochTime=   47.6459\n",
            "007: Acc-tr:  70.21, Acc-val:  10.24, L-tr: 0.8670, L-val: 4.3432, epochTime=   47.5467\n",
            "008: Acc-tr:  65.64, Acc-val:  10.46, L-tr: 1.0315, L-val: 4.3828, epochTime=   47.6796\n",
            "009: Acc-tr:  48.88, Acc-val:  10.83, L-tr: 1.4903, L-val: 5.2012, epochTime=   47.4702\n",
            "010: Acc-tr:  56.54, Acc-val:  10.44, L-tr: 1.3881, L-val: 5.0438, epochTime=   47.6749\n",
            "011: Acc-tr:  70.45, Acc-val:  10.12, L-tr: 0.9035, L-val: 4.6941, epochTime=   47.4280\n",
            "012: Acc-tr:  63.08, Acc-val:   9.91, L-tr: 1.1207, L-val: 5.2932, epochTime=   47.5164\n",
            "013: Acc-tr:  68.90, Acc-val:   9.97, L-tr: 0.8806, L-val: 4.7994, epochTime=   47.8731\n",
            "014: Acc-tr:  57.60, Acc-val:  10.39, L-tr: 1.3239, L-val: 5.2847, epochTime=   47.5460\n",
            "015: Acc-tr:  75.72, Acc-val:  10.27, L-tr: 0.7271, L-val: 4.3712, epochTime=   47.8784\n",
            "016: Acc-tr:  68.95, Acc-val:   9.94, L-tr: 0.9129, L-val: 5.1357, epochTime=   47.7671\n",
            "017: Acc-tr:  70.36, Acc-val:  10.31, L-tr: 0.8755, L-val: 4.6924, epochTime=   48.0503\n",
            "018: Acc-tr:  67.22, Acc-val:  10.68, L-tr: 0.9636, L-val: 5.2391, epochTime=   48.0826\n",
            "019: Acc-tr:  70.39, Acc-val:  10.06, L-tr: 0.8686, L-val: 4.9965, epochTime=   48.1345\n",
            "020: Acc-tr:  63.23, Acc-val:  10.46, L-tr: 1.0749, L-val: 5.1670, epochTime=   48.0881\n",
            "021: Acc-tr:  71.49, Acc-val:   9.74, L-tr: 0.8352, L-val: 4.6691, epochTime=   47.7306\n",
            "022: Acc-tr:  73.44, Acc-val:  10.08, L-tr: 0.7657, L-val: 5.2511, epochTime=   47.6412\n",
            "023: Acc-tr:  71.06, Acc-val:  10.31, L-tr: 0.8352, L-val: 5.1719, epochTime=   47.8159\n",
            "024: Acc-tr:  73.91, Acc-val:   9.66, L-tr: 0.7659, L-val: 4.7649, epochTime=   47.6468\n",
            "025: Acc-tr:  63.56, Acc-val:  10.30, L-tr: 1.1057, L-val: 5.5711, epochTime=   47.8895\n",
            "026: Acc-tr:  72.75, Acc-val:  10.14, L-tr: 0.8010, L-val: 5.0498, epochTime=   47.7170\n",
            "027: Acc-tr:  70.58, Acc-val:  10.31, L-tr: 0.8587, L-val: 5.2497, epochTime=   47.5677\n",
            "028: Acc-tr:  69.11, Acc-val:  10.61, L-tr: 0.8924, L-val: 5.0304, epochTime=   47.7930\n",
            "029: Acc-tr:  74.33, Acc-val:  10.40, L-tr: 0.7465, L-val: 4.9636, epochTime=   47.7369\n",
            "030: Acc-tr:  72.31, Acc-val:  10.58, L-tr: 0.7960, L-val: 5.1440, epochTime=   47.7772\n",
            "031: Acc-tr:  75.47, Acc-val:  10.04, L-tr: 0.7228, L-val: 4.8340, epochTime=   47.6315\n",
            "032: Acc-tr:  75.13, Acc-val:  10.23, L-tr: 0.7424, L-val: 5.2416, epochTime=   47.7489\n",
            "033: Acc-tr:  73.02, Acc-val:   9.80, L-tr: 0.8009, L-val: 5.0989, epochTime=   47.6611\n",
            "034: Acc-tr:  61.61, Acc-val:  10.38, L-tr: 1.1883, L-val: 5.4737, epochTime=   47.6965\n",
            "035: Acc-tr:  59.95, Acc-val:   9.58, L-tr: 1.2614, L-val: 6.1085, epochTime=   47.5326\n",
            "036: Acc-tr:  65.04, Acc-val:  10.14, L-tr: 1.1026, L-val: 5.6570, epochTime=   47.2935\n",
            "037: Acc-tr:  69.67, Acc-val:  10.24, L-tr: 0.9874, L-val: 5.3931, epochTime=   47.6341\n",
            "038: Acc-tr:  76.09, Acc-val:   9.99, L-tr: 0.7030, L-val: 5.2218, epochTime=   47.6819\n",
            "039: Acc-tr:  66.75, Acc-val:  10.56, L-tr: 0.9885, L-val: 5.2829, epochTime=   47.4985\n",
            "040: Acc-tr:  73.75, Acc-val:  10.44, L-tr: 0.7932, L-val: 5.0328, epochTime=   47.6357\n",
            "041: Acc-tr:  77.67, Acc-val:  10.49, L-tr: 0.6609, L-val: 4.9933, epochTime=   47.5465\n",
            "042: Acc-tr:  65.24, Acc-val:  10.22, L-tr: 1.0470, L-val: 5.8310, epochTime=   47.5115\n",
            "043: Acc-tr:  72.48, Acc-val:  10.41, L-tr: 0.8458, L-val: 5.0993, epochTime=   47.5991\n",
            "044: Acc-tr:  76.29, Acc-val:  10.02, L-tr: 0.7008, L-val: 5.4240, epochTime=   47.8431\n",
            "045: Acc-tr:  59.22, Acc-val:  10.23, L-tr: 1.3729, L-val: 6.2077, epochTime=   47.6747\n",
            "046: Acc-tr:  73.65, Acc-val:  10.07, L-tr: 0.7742, L-val: 5.5242, epochTime=   47.6142\n",
            "047: Acc-tr:  73.69, Acc-val:  10.35, L-tr: 0.7678, L-val: 5.2429, epochTime=   47.7874\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "stream",
          "text": [
            "048: Acc-tr:  76.83, Acc-val:  10.27, L-tr: 0.6686, L-val: 5.1223, epochTime=   47.8231\n",
            "049: Acc-tr:  65.87, Acc-val:  10.62, L-tr: 1.1064, L-val: 5.7930, epochTime=   47.8943\n",
            "050: Acc-tr:  75.19, Acc-val:   9.88, L-tr: 0.7194, L-val: 5.4939, epochTime=   47.7811\n",
            "051: Acc-tr:  67.30, Acc-val:   9.82, L-tr: 0.9692, L-val: 5.5536, epochTime=   47.6761\n",
            "052: Acc-tr:  74.98, Acc-val:  10.33, L-tr: 0.7554, L-val: 5.1899, epochTime=   47.8293\n",
            "053: Acc-tr:  69.76, Acc-val:  10.51, L-tr: 0.8778, L-val: 5.1587, epochTime=   47.6957\n",
            "054: Acc-tr:  71.52, Acc-val:  10.33, L-tr: 0.8531, L-val: 5.3392, epochTime=   47.7287\n",
            "055: Acc-tr:  77.41, Acc-val:   9.76, L-tr: 0.6635, L-val: 5.4376, epochTime=   47.6114\n",
            "056: Acc-tr:  68.46, Acc-val:   9.92, L-tr: 0.9558, L-val: 5.4593, epochTime=   47.5762\n",
            "057: Acc-tr:  74.60, Acc-val:  10.09, L-tr: 0.7591, L-val: 5.4085, epochTime=   47.7540\n",
            "058: Acc-tr:  77.86, Acc-val:   9.89, L-tr: 0.6597, L-val: 5.5886, epochTime=   47.5379\n",
            "059: Acc-tr:  81.16, Acc-val:  10.61, L-tr: 0.5524, L-val: 5.3077, epochTime=   47.5387\n",
            "060: Acc-tr:  58.99, Acc-val:  10.23, L-tr: 1.4135, L-val: 6.3724, epochTime=   47.5479\n",
            "061: Acc-tr:  83.33, Acc-val:  10.28, L-tr: 0.4968, L-val: 4.9401, epochTime=   47.7653\n",
            "062: Acc-tr:  68.57, Acc-val:  10.61, L-tr: 0.9324, L-val: 5.8298, epochTime=   47.6561\n",
            "063: Acc-tr:  79.34, Acc-val:   9.70, L-tr: 0.6017, L-val: 5.1615, epochTime=   47.7363\n",
            "064: Acc-tr:  79.87, Acc-val:  10.34, L-tr: 0.6002, L-val: 5.1200, epochTime=   47.6338\n",
            "065: Acc-tr:  72.03, Acc-val:  10.00, L-tr: 0.9131, L-val: 5.7646, epochTime=   47.6264\n",
            "066: Acc-tr:  77.69, Acc-val:  10.37, L-tr: 0.6512, L-val: 4.9261, epochTime=   47.6914\n",
            "067: Acc-tr:  71.63, Acc-val:  10.08, L-tr: 0.8081, L-val: 5.4817, epochTime=   47.6887\n",
            "068: Acc-tr:  76.59, Acc-val:  10.62, L-tr: 0.6804, L-val: 5.2390, epochTime=   47.6214\n",
            "069: Acc-tr:  72.62, Acc-val:  10.06, L-tr: 0.8039, L-val: 5.4029, epochTime=   47.6365\n",
            "070: Acc-tr:  60.81, Acc-val:  10.14, L-tr: 1.2522, L-val: 6.6034, epochTime=   47.7069\n",
            "071: Acc-tr:  80.58, Acc-val:  10.12, L-tr: 0.5674, L-val: 5.3668, epochTime=   47.8699\n",
            "072: Acc-tr:  74.37, Acc-val:  10.79, L-tr: 0.7393, L-val: 5.2308, epochTime=   47.6452\n",
            "073: Acc-tr:  56.54, Acc-val:   9.75, L-tr: 1.3959, L-val: 6.4461, epochTime=   47.6036\n",
            "074: Acc-tr:  68.52, Acc-val:  10.01, L-tr: 0.9563, L-val: 6.0037, epochTime=   47.7429\n",
            "075: Acc-tr:  70.56, Acc-val:   9.93, L-tr: 0.9115, L-val: 5.7134, epochTime=   47.8095\n",
            "076: Acc-tr:  74.04, Acc-val:   9.71, L-tr: 0.8278, L-val: 5.6693, epochTime=   48.2013\n",
            "077: Acc-tr:  81.17, Acc-val:   9.73, L-tr: 0.5615, L-val: 5.0807, epochTime=   47.9827\n",
            "078: Acc-tr:  80.37, Acc-val:   9.97, L-tr: 0.5827, L-val: 5.5047, epochTime=   47.9509\n",
            "079: Acc-tr:  79.97, Acc-val:  10.31, L-tr: 0.5981, L-val: 5.2157, epochTime=   47.9157\n",
            "080: Acc-tr:  72.83, Acc-val:  10.50, L-tr: 0.8278, L-val: 5.6235, epochTime=   47.5922\n",
            "081: Acc-tr:  75.09, Acc-val:  10.17, L-tr: 0.7374, L-val: 5.5487, epochTime=   47.7424\n",
            "082: Acc-tr:  76.46, Acc-val:  10.01, L-tr: 0.7018, L-val: 5.4819, epochTime=   48.0119\n",
            "083: Acc-tr:  40.92, Acc-val:   9.97, L-tr: 2.3899, L-val: 7.8230, epochTime=   48.3694\n",
            "084: Acc-tr:  72.30, Acc-val:  10.59, L-tr: 0.8323, L-val: 5.7117, epochTime=   48.0230\n",
            "085: Acc-tr:  75.32, Acc-val:   9.88, L-tr: 0.7298, L-val: 5.3615, epochTime=   48.1743\n",
            "086: Acc-tr:  68.37, Acc-val:   9.92, L-tr: 0.9527, L-val: 5.5560, epochTime=   48.0951\n",
            "087: Acc-tr:  78.64, Acc-val:  10.23, L-tr: 0.6343, L-val: 5.4921, epochTime=   48.2140\n",
            "088: Acc-tr:  60.72, Acc-val:   9.75, L-tr: 1.2537, L-val: 6.3805, epochTime=   48.1796\n",
            "089: Acc-tr:  67.45, Acc-val:   9.91, L-tr: 1.0011, L-val: 5.6854, epochTime=   48.1671\n",
            "090: Acc-tr:  78.64, Acc-val:   9.84, L-tr: 0.6184, L-val: 5.2264, epochTime=   47.9851\n",
            "091: Acc-tr:  71.44, Acc-val:  10.03, L-tr: 0.8426, L-val: 5.7927, epochTime=   47.9864\n",
            "092: Acc-tr:  69.29, Acc-val:  10.42, L-tr: 0.9116, L-val: 5.8053, epochTime=   47.8716\n",
            "093: Acc-tr:  81.47, Acc-val:  10.47, L-tr: 0.5489, L-val: 5.1688, epochTime=   47.6940\n",
            "094: Acc-tr:  71.28, Acc-val:   9.88, L-tr: 0.8436, L-val: 5.4685, epochTime=   47.9258\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "stream",
          "text": [
            "095: Acc-tr:  79.07, Acc-val:  10.00, L-tr: 0.6234, L-val: 5.4505, epochTime=   47.8196\n",
            "096: Acc-tr:  70.28, Acc-val:   9.79, L-tr: 0.9014, L-val: 6.1486, epochTime=   48.5959\n",
            "097: Acc-tr:  68.30, Acc-val:  10.29, L-tr: 0.9837, L-val: 5.7883, epochTime=   47.8111\n",
            "098: Acc-tr:  74.34, Acc-val:   9.70, L-tr: 0.7662, L-val: 5.3835, epochTime=   47.9216\n",
            "099: Acc-tr:  69.02, Acc-val:  10.40, L-tr: 0.9889, L-val: 5.6449, epochTime=   47.9630\n",
            "100: Acc-tr:  56.06, Acc-val:  10.12, L-tr: 1.5650, L-val: 6.5238, epochTime=   47.8557\n",
            "101: Acc-tr:  68.91, Acc-val:   9.99, L-tr: 0.9297, L-val: 5.9893, epochTime=   47.8340\n",
            "102: Acc-tr:  75.53, Acc-val:   9.81, L-tr: 0.7281, L-val: 5.5874, epochTime=   47.8769\n",
            "103: Acc-tr:  70.04, Acc-val:   9.81, L-tr: 0.8977, L-val: 5.9283, epochTime=   48.2571\n",
            "104: Acc-tr:  67.74, Acc-val:  10.35, L-tr: 1.0541, L-val: 6.0108, epochTime=   48.0069\n",
            "105: Acc-tr:  73.46, Acc-val:   9.83, L-tr: 0.7777, L-val: 5.7014, epochTime=   47.8683\n",
            "106: Acc-tr:  68.08, Acc-val:  10.64, L-tr: 0.9832, L-val: 5.8317, epochTime=   47.8735\n",
            "107: Acc-tr:  82.80, Acc-val:  10.10, L-tr: 0.5097, L-val: 5.0463, epochTime=   47.9720\n",
            "108: Acc-tr:  76.77, Acc-val:  10.07, L-tr: 0.7081, L-val: 5.3234, epochTime=   47.8950\n",
            "109: Acc-tr:  73.83, Acc-val:   9.94, L-tr: 0.7721, L-val: 5.7782, epochTime=   47.8942\n",
            "110: Acc-tr:  73.89, Acc-val:  10.29, L-tr: 0.7670, L-val: 5.6414, epochTime=   47.7767\n",
            "111: Acc-tr:  76.39, Acc-val:  10.00, L-tr: 0.7138, L-val: 5.5525, epochTime=   47.7444\n",
            "112: Acc-tr:  81.18, Acc-val:   9.99, L-tr: 0.5714, L-val: 5.4991, epochTime=   47.7024\n",
            "113: Acc-tr:  77.71, Acc-val:   9.51, L-tr: 0.6452, L-val: 5.3189, epochTime=   47.7543\n",
            "114: Acc-tr:  65.89, Acc-val:   9.94, L-tr: 1.1606, L-val: 6.1125, epochTime=   47.5810\n",
            "115: Acc-tr:  76.98, Acc-val:  10.38, L-tr: 0.6860, L-val: 5.5875, epochTime=   47.4797\n",
            "116: Acc-tr:  73.11, Acc-val:  10.14, L-tr: 0.8559, L-val: 5.8801, epochTime=   47.5427\n",
            "117: Acc-tr:  73.03, Acc-val:  10.89, L-tr: 0.8112, L-val: 5.6894, epochTime=   47.6004\n",
            "118: Acc-tr:  73.46, Acc-val:   9.97, L-tr: 0.7854, L-val: 5.4717, epochTime=   47.5414\n",
            "119: Acc-tr:  58.05, Acc-val:  10.01, L-tr: 1.4667, L-val: 6.9054, epochTime=   47.6607\n",
            "120: Acc-tr:  73.48, Acc-val:   9.87, L-tr: 0.7955, L-val: 5.8122, epochTime=   47.7810\n",
            "121: Acc-tr:  67.75, Acc-val:  10.39, L-tr: 1.0209, L-val: 6.0429, epochTime=   47.9921\n",
            "122: Acc-tr:  78.80, Acc-val:  10.21, L-tr: 0.6054, L-val: 5.0908, epochTime=   47.6597\n",
            "123: Acc-tr:  69.88, Acc-val:  10.26, L-tr: 0.9112, L-val: 5.7202, epochTime=   47.6185\n",
            "124: Acc-tr:  78.59, Acc-val:   9.81, L-tr: 0.6425, L-val: 5.7185, epochTime=   47.6863\n",
            "125: Acc-tr:  61.31, Acc-val:   9.86, L-tr: 1.2602, L-val: 6.2860, epochTime=   47.5476\n",
            "126: Acc-tr:  67.67, Acc-val:   9.76, L-tr: 0.9988, L-val: 5.9861, epochTime=   47.6135\n",
            "127: Acc-tr:  66.63, Acc-val:   9.52, L-tr: 1.0282, L-val: 6.0873, epochTime=   47.4238\n",
            "128: Acc-tr:  81.31, Acc-val:   9.92, L-tr: 0.5465, L-val: 5.2636, epochTime=   47.4982\n",
            "129: Acc-tr:  80.82, Acc-val:  10.43, L-tr: 0.5577, L-val: 5.1963, epochTime=   47.3103\n",
            "130: Acc-tr:  73.55, Acc-val:   9.97, L-tr: 0.8162, L-val: 5.6791, epochTime=   47.5469\n",
            "131: Acc-tr:  60.88, Acc-val:  10.27, L-tr: 1.4381, L-val: 6.5946, epochTime=   47.5946\n",
            "132: Acc-tr:  53.87, Acc-val:  10.04, L-tr: 1.5373, L-val: 6.5626, epochTime=   47.4783\n",
            "133: Acc-tr:  79.66, Acc-val:   9.85, L-tr: 0.5912, L-val: 5.3192, epochTime=   47.9109\n",
            "134: Acc-tr:  77.18, Acc-val:  10.21, L-tr: 0.7214, L-val: 5.5454, epochTime=   48.0209\n",
            "135: Acc-tr:  54.17, Acc-val:  10.26, L-tr: 1.8209, L-val: 6.9757, epochTime=   48.7776\n",
            "136: Acc-tr:  73.57, Acc-val:  10.29, L-tr: 0.7910, L-val: 5.6405, epochTime=   48.8040\n",
            "137: Acc-tr:  68.63, Acc-val:  10.35, L-tr: 1.0132, L-val: 6.3943, epochTime=   48.6622\n",
            "138: Acc-tr:  79.90, Acc-val:   9.92, L-tr: 0.5860, L-val: 5.1908, epochTime=   48.1843\n",
            "139: Acc-tr:  76.02, Acc-val:  10.03, L-tr: 0.7416, L-val: 5.2032, epochTime=   47.7711\n",
            "140: Acc-tr:  71.85, Acc-val:  10.10, L-tr: 0.8732, L-val: 5.7931, epochTime=   47.8473\n",
            "141: Acc-tr:  80.13, Acc-val:  10.28, L-tr: 0.5766, L-val: 5.2548, epochTime=   47.8696\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "stream",
          "text": [
            "142: Acc-tr:  73.01, Acc-val:   9.95, L-tr: 0.7743, L-val: 5.5534, epochTime=   47.6921\n",
            "143: Acc-tr:  63.96, Acc-val:   9.92, L-tr: 1.1167, L-val: 6.3869, epochTime=   47.8574\n",
            "144: Acc-tr:  80.53, Acc-val:  10.14, L-tr: 0.5803, L-val: 4.9906, epochTime=   47.6944\n",
            "145: Acc-tr:  68.26, Acc-val:  10.22, L-tr: 1.0774, L-val: 6.2482, epochTime=   47.8910\n",
            "146: Acc-tr:  76.27, Acc-val:   9.93, L-tr: 0.7036, L-val: 5.9341, epochTime=   47.8650\n",
            "147: Acc-tr:  74.65, Acc-val:  10.18, L-tr: 0.7466, L-val: 5.7840, epochTime=   47.6513\n",
            "148: Acc-tr:  83.65, Acc-val:   9.93, L-tr: 0.4866, L-val: 5.0432, epochTime=   47.8669\n",
            "149: Acc-tr:  70.54, Acc-val:   9.62, L-tr: 0.9238, L-val: 5.8393, epochTime=   47.7314\n",
            "150: Acc-tr:  98.78, Acc-val:   9.60, L-tr: 0.0851, L-val: 5.4966, epochTime=   47.7558\n",
            "151: Acc-tr:  99.76, Acc-val:   9.64, L-tr: 0.0438, L-val: 5.9035, epochTime=   47.4946\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "metadata": {
        "id": "2-zYTlFjFXFA",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 224
        },
        "outputId": "fdc43573-9834-48f6-b3a4-d95e84f8cad0"
      },
      "cell_type": "code",
      "source": [
        "# Download the model\n",
        "from google.colab import files\n",
        "!ls\n",
        "files.download(model_file_name)"
      ],
      "execution_count": 15,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "data\t\t\t\t       Model_corruptiton_1.0_Acc-tr_56.06.pt\r\n",
            "datalab\t\t\t\t       Model_corruptiton_1.0_Acc-tr_60.08.pt\r\n",
            "Model_corruptiton_0.0_Acc-tr_99.98.pt  Model_corruptiton_1.0_Acc-tr_61.51.pt\r\n",
            "Model_corruptiton_1.0_Acc-tr_10.24.pt  Model_corruptiton_1.0_Acc-tr_75.19.pt\r\n",
            "Model_corruptiton_1.0_Acc-tr_10.39.pt  Model_corruptiton_1.0_Acc-tr_98.78.pt\r\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "stream",
          "text": [
            "/usr/local/lib/python3.6/dist-packages/torch/serialization.py:158: UserWarning: Couldn't retrieve source code for container of type WideResNet. It won't be checked for correctness upon loading.\n",
            "  \"type \" + obj.__name__ + \". It won't be checked \"\n",
            "/usr/local/lib/python3.6/dist-packages/torch/serialization.py:158: UserWarning: Couldn't retrieve source code for container of type NetworkBlock. It won't be checked for correctness upon loading.\n",
            "  \"type \" + obj.__name__ + \". It won't be checked \"\n",
            "/usr/local/lib/python3.6/dist-packages/torch/serialization.py:158: UserWarning: Couldn't retrieve source code for container of type BasicBlock. It won't be checked for correctness upon loading.\n",
            "  \"type \" + obj.__name__ + \". It won't be checked \"\n"
          ],
          "name": "stderr"
        }
      ]
    }
  ]
}