ResNet_separable.ipynb

{
  "nbformat": 4,
  "nbformat_minor": 0,
  "metadata": {
    "kernelspec": {
      "name": "python3",
      "display_name": "Python 3"
    },
    "colab": {
      "name": "Woof_128_twist.ipynb",
      "provenance": [],
      "collapsed_sections": [
        "LP3vVSWMBz1L",
        "6C-1pE45Bz1l",
        "pLmt0ISPBz2w"
      ],
      "include_colab_link": true
    },
    "accelerator": "GPU"
  },
  "cells": [
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "view-in-github",
        "colab_type": "text"
      },
      "source": [
        "<a href=\"https://colab.research.google.com/github/liuyao12/imagenette_experiments/blob/master/ResNet_separable.ipynb\" target=\"_parent\"><img src=\"https://colab.research.google.com/assets/colab-badge.svg\" alt=\"Open In Colab\"/></a>"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "GM0cNcW2Bzz4",
        "colab_type": "text"
      },
      "source": [
        "# ResNet with a Twist, with depthwise separable convolution\n",
        "\n",
        "> depthwise (x4) + Ranger + Mish + SA + MaxBlurPool + ResTrick\n",
        "\n",
        "See summary at https://forums.fast.ai/t/imagenette-imagewoof-leaderboards/45822/47?u=liuyao "
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "cXBTriT1fmXE",
        "colab_type": "text"
      },
      "source": [
        "## Imagewoof Leaderboard\n",
        "\n",
        "Imagewoof2, with a 70/30 train/test ratio.\n",
        "\n",
        "| Size (px) | Epochs | x2 | x4 | x6 | x4 double | runs |\n",
        "|--|--|--| --|--| --|--|\n",
        "|128|5|75.19|76.27|**76.61**| **82.12**|5, mean\n",
        "|128|20|85.18|86.22|**86.27**| **88.93**|5, mean\n",
        "|128|80|87.70|**87.83**|87.65| **90.15**|1\n",
        "|128|200|\n",
        "|192|5|79.86|**81.15**|| **82.69**|5, mean\n",
        "|192|20|88.12|**88.37**|\n",
        "|192|80|**90.30**|89.89|| **92.08** |\n",
        "|192|200|\n",
        "|256|5|\n",
        "|256|20|\n",
        "|256|80|\n",
        "|256|200|\n"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "DrD1-te_Bzz7",
        "colab_type": "text"
      },
      "source": [
        "# setup and imports"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "UteAROmqBzz9",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "pip install git+https://github.com/ayasyrev/model_constructor"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab_type": "code",
        "id": "XnsPVNrkTfV3",
        "colab": {}
      },
      "source": [
        "pip install git+https://github.com/ayasyrev/imagenette_experiments"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "K-OCjxGCBz0J",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "from fastai.basic_train import *\n",
        "from fastai.vision import *\n",
        "# from fastai.script import *"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "vPNF3XKWBz0O",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "# pip install kornia\n",
        "from kornia.contrib import MaxBlurPool2d"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "xPGwrk6TBz0U",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "from imagenette_experiments.train_utils import *\n",
        "from model_constructor.net import Net, act_fn\n",
        "from model_constructor.layers import SimpleSelfAttention, ConvLayer"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "U63Mcb1lBz0p",
        "colab_type": "text"
      },
      "source": [
        "# ResBlock"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "kY7y9c99Bz0q",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "class NewLayer(nn.Sequential):\n",
        "    \"\"\"Basic conv layers block\"\"\"\n",
        "    def __init__(self, ni, nf, ks=3, stride=1,\n",
        "            act=True,  act_fn=nn.ReLU(inplace=True),\n",
        "            bn_layer=True, bn_1st=True, zero_bn=False,\n",
        "            padding=None, bias=False, groups=1, **kwargs):\n",
        "\n",
        "        if padding==None: padding = ks//2\n",
        "        if ks==3 and groups==1:  # to be used for the \"stem\" of ResNet\n",
        "          # if ni==3: stride = 2\n",
        "          layers = [('Conv3x3', nn.Conv2d(ni, ni*dm, 3, stride=stride, padding=1, bias=bias, groups=ni)),\n",
        "                    ('Conv1x1', nn.Conv2d(ni*dm, nf, 1, bias=bias, groups=1))]\n",
        "        else:\n",
        "          layers = [('Conv{}x{}'.format(ks,ks), \n",
        "                      nn.Conv2d(ni, nf, ks, stride=stride, padding=padding, bias=bias, groups=groups))]\n",
        "\n",
        "        act_bn = [('act_fn', act_fn)] if act else []\n",
        "        if bn_layer:\n",
        "            bn = nn.BatchNorm2d(nf)\n",
        "            nn.init.constant_(bn.weight, 0. if zero_bn else 1.)\n",
        "            act_bn += [('bn', bn)]\n",
        "        if bn_1st: act_bn.reverse()\n",
        "        layers += act_bn\n",
        "        super().__init__(OrderedDict(layers))"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "nVLFz9nhBz0x",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "class NewResBlock(Module):\n",
        "    def __init__(self, expansion, ni, nh, stride=1,\n",
        "                 conv_layer=ConvLayer, act_fn=act_fn, zero_bn=True, bn_1st=True,\n",
        "                 pool=nn.AvgPool2d(2, ceil_mode=True), sa=False, sym=False, groups=1):\n",
        "        nf,ni = nh*expansion,ni*expansion\n",
        "        conv_layer = NewLayer\n",
        "        self.reduce = noop if stride==1 else pool\n",
        "        layers  = [# (f\"conv_0\", conv_layer(ni, nh, 3, act_fn=act_fn, bn_1st=bn_1st)),\n",
        "                   (f\"conv_1\", conv_layer(ni, nf, 3, zero_bn=zero_bn, act=False, bn_layer=True))\n",
        "        ] if expansion == 1 else [\n",
        "                   (f\"conv_0\", conv_layer(ni, nh, 1, act_fn=act_fn, bn_1st=bn_1st)),\n",
        "                   # (f\"conv_1\", conv_layer(nh, nh, 3, act_fn=act_fn, bn_1st=bn_1st)),\n",
        "                   # (f\"conv_1\", conv_layer(nh, nh*dm, 3, groups=nh, act_fn=act_fn, bn_1st=bn_1st)),\n",
        "                   (f\"conv_1\", conv_layer(nh, nh*dm, 3, groups=nh, act=False, bn_layer=False)),\n",
        "                   (f\"conv_2\", conv_layer(nh*dm, nf, 1, zero_bn=zero_bn, act=False, bn_1st=bn_1st))\n",
        "        ]\n",
        "        if sa: layers.append(('sa', SimpleSelfAttention(nf,ks=1,sym=sym)))\n",
        "        self.convs = nn.Sequential(OrderedDict(layers))\n",
        "        self.idconv = noop if ni==nf else conv_layer(ni, nf, 1, act=False, bn_1st=bn_1st)\n",
        "        self.merge = act_fn\n",
        "\n",
        "    def forward(self, x):\n",
        "        o = self.reduce(x)\n",
        "        return self.merge(self.convs(o) + self.idconv(o))"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "XroIp4GcBz07",
        "colab_type": "text"
      },
      "source": [
        "# Model Constructor"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "ZTrZVV81Bz1G",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "model = Net(c_out=10, layers=[3,6,8,3], expansion=4)\n",
        "model.block = NewResBlock\n",
        "model.conv_layer = NewLayer # for the stem\n",
        "pool = MaxBlurPool2d(3, True)\n",
        "model.pool = pool\n",
        "model.stem_sizes = [3,32,64,64]\n",
        "model.act_fn = Mish()\n",
        "model.sa = True"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "sCDxE73yuq_5",
        "colab_type": "text"
      },
      "source": [
        "## Experiment"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "2JfDW5iqPVE4",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 908
        },
        "outputId": "d4aac895-03f4-48f0-e9d8-c5072c33eb07"
      },
      "source": [
        "dm = 4\n",
        "res = dict()\n",
        "for e in [5]*5 + [20]*5 + [80]:\n",
        "  mixup=0 if e<=20 else 0.2\n",
        "  learn = get_learn(model=model, size=192, bs=16, mixup=mixup)\n",
        "  learn.fit_fc(e, lr=4e-3, moms=(0.95,0.95), start_pct=0.72)\n",
        "  acc = learn.recorder.metrics[-1][0].item()\n",
        "  if e in res:\n",
        "    res[e] += [acc]\n",
        "  else:\n",
        "    res[e] = [acc]\n",
        "  print('{} epochs: {} ({} runs)'.format(e, sum(res[e])/len(res[e]), len(res[e])))\n",
        "print('depth multiplier={}'.format(dm), {e: sum(res[e])/len(res[e]) for e in res})"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "data path   /root/.fastai/data/imagewoof2\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "stream",
          "text": [
            "/usr/local/lib/python3.6/dist-packages/torch/nn/functional.py:3000: UserWarning: The default behavior for interpolate/upsample with float scale_factor changed in 1.6.0 to align with other frameworks/libraries, and uses scale_factor directly, instead of relying on the computed output size. If you wish to keep the old behavior, please set recompute_scale_factor=True. See the documentation of nn.Upsample for details. \n",
            "  warnings.warn(\"The default behavior for interpolate/upsample with float scale_factor changed \"\n",
            "/usr/local/lib/python3.6/dist-packages/torch/nn/functional.py:3000: UserWarning: The default behavior for interpolate/upsample with float scale_factor changed in 1.6.0 to align with other frameworks/libraries, and uses scale_factor directly, instead of relying on the computed output size. If you wish to keep the old behavior, please set recompute_scale_factor=True. See the documentation of nn.Upsample for details. \n",
            "  warnings.warn(\"The default behavior for interpolate/upsample with float scale_factor changed \"\n",
            "/usr/local/lib/python3.6/dist-packages/torch/nn/functional.py:3000: UserWarning: The default behavior for interpolate/upsample with float scale_factor changed in 1.6.0 to align with other frameworks/libraries, and uses scale_factor directly, instead of relying on the computed output size. If you wish to keep the old behavior, please set recompute_scale_factor=True. See the documentation of nn.Upsample for details. \n",
            "  warnings.warn(\"The default behavior for interpolate/upsample with float scale_factor changed \"\n",
            "/usr/local/lib/python3.6/dist-packages/torch/nn/functional.py:3000: UserWarning: The default behavior for interpolate/upsample with float scale_factor changed in 1.6.0 to align with other frameworks/libraries, and uses scale_factor directly, instead of relying on the computed output size. If you wish to keep the old behavior, please set recompute_scale_factor=True. See the documentation of nn.Upsample for details. \n",
            "  warnings.warn(\"The default behavior for interpolate/upsample with float scale_factor changed \"\n",
            "/usr/local/lib/python3.6/dist-packages/torch/nn/functional.py:3000: UserWarning: The default behavior for interpolate/upsample with float scale_factor changed in 1.6.0 to align with other frameworks/libraries, and uses scale_factor directly, instead of relying on the computed output size. If you wish to keep the old behavior, please set recompute_scale_factor=True. See the documentation of nn.Upsample for details. \n",
            "  warnings.warn(\"The default behavior for interpolate/upsample with float scale_factor changed \"\n",
            "/usr/local/lib/python3.6/dist-packages/torch/nn/functional.py:3000: UserWarning: The default behavior for interpolate/upsample with float scale_factor changed in 1.6.0 to align with other frameworks/libraries, and uses scale_factor directly, instead of relying on the computed output size. If you wish to keep the old behavior, please set recompute_scale_factor=True. See the documentation of nn.Upsample for details. \n",
            "  warnings.warn(\"The default behavior for interpolate/upsample with float scale_factor changed \"\n"
          ],
          "name": "stderr"
        },
        {
          "output_type": "stream",
          "text": [
            "Learn path /root/.fastai/data/imagewoof2\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "display_data",
          "data": {
            "text/html": [
              "<table border=\"1\" class=\"dataframe\">\n",
              "  <thead>\n",
              "    <tr style=\"text-align: left;\">\n",
              "      <th>epoch</th>\n",
              "      <th>train_loss</th>\n",
              "      <th>valid_loss</th>\n",
              "      <th>accuracy</th>\n",
              "      <th>top_k_accuracy</th>\n",
              "      <th>time</th>\n",
              "    </tr>\n",
              "  </thead>\n",
              "  <tbody>\n",
              "    <tr>\n",
              "      <td>0</td>\n",
              "      <td>1.938423</td>\n",
              "      <td>1.738842</td>\n",
              "      <td>0.444388</td>\n",
              "      <td>0.895902</td>\n",
              "      <td>02:58</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>1</td>\n",
              "      <td>1.614446</td>\n",
              "      <td>1.593913</td>\n",
              "      <td>0.535760</td>\n",
              "      <td>0.921354</td>\n",
              "      <td>02:58</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>2</td>\n",
              "      <td>1.406542</td>\n",
              "      <td>1.312408</td>\n",
              "      <td>0.671163</td>\n",
              "      <td>0.955205</td>\n",
              "      <td>02:58</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>3</td>\n",
              "      <td>1.285317</td>\n",
              "      <td>1.198211</td>\n",
              "      <td>0.723848</td>\n",
              "      <td>0.967422</td>\n",
              "      <td>02:59</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>4</td>\n",
              "      <td>1.090849</td>\n",
              "      <td>1.036145</td>\n",
              "      <td>0.798422</td>\n",
              "      <td>0.980657</td>\n",
              "      <td>02:59</td>\n",
              "    </tr>\n",
              "  </tbody>\n",
              "</table>"
            ],
            "text/plain": [
              "<IPython.core.display.HTML object>"
            ]
          },
          "metadata": {
            "tags": []
          }
        },
        {
          "output_type": "stream",
          "text": [
            "/usr/local/lib/python3.6/dist-packages/imagenette_experiments/train_utils.py:150: UserWarning: This overload of addcmul_ is deprecated:\n",
            "\taddcmul_(Number value, Tensor tensor1, Tensor tensor2)\n",
            "Consider using one of the following signatures instead:\n",
            "\taddcmul_(Tensor tensor1, Tensor tensor2, *, Number value) (Triggered internally at  /pytorch/torch/csrc/utils/python_arg_parser.cpp:766.)\n",
            "  exp_avg_sq.mul_(beta2).addcmul_(1 - beta2, grad, grad)\n"
          ],
          "name": "stderr"
        },
        {
          "output_type": "stream",
          "text": [
            "5 epochs: 0.7984219789505005 (1 runs)\n",
            "data path   /root/.fastai/data/imagewoof2\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "stream",
          "text": [
            "/usr/local/lib/python3.6/dist-packages/torch/nn/functional.py:3000: UserWarning: The default behavior for interpolate/upsample with float scale_factor changed in 1.6.0 to align with other frameworks/libraries, and uses scale_factor directly, instead of relying on the computed output size. If you wish to keep the old behavior, please set recompute_scale_factor=True. See the documentation of nn.Upsample for details. \n",
            "  warnings.warn(\"The default behavior for interpolate/upsample with float scale_factor changed \"\n",
            "/usr/local/lib/python3.6/dist-packages/torch/nn/functional.py:3000: UserWarning: The default behavior for interpolate/upsample with float scale_factor changed in 1.6.0 to align with other frameworks/libraries, and uses scale_factor directly, instead of relying on the computed output size. If you wish to keep the old behavior, please set recompute_scale_factor=True. See the documentation of nn.Upsample for details. \n",
            "  warnings.warn(\"The default behavior for interpolate/upsample with float scale_factor changed \"\n",
            "/usr/local/lib/python3.6/dist-packages/torch/nn/functional.py:3000: UserWarning: The default behavior for interpolate/upsample with float scale_factor changed in 1.6.0 to align with other frameworks/libraries, and uses scale_factor directly, instead of relying on the computed output size. If you wish to keep the old behavior, please set recompute_scale_factor=True. See the documentation of nn.Upsample for details. \n",
            "  warnings.warn(\"The default behavior for interpolate/upsample with float scale_factor changed \"\n",
            "/usr/local/lib/python3.6/dist-packages/torch/nn/functional.py:3000: UserWarning: The default behavior for interpolate/upsample with float scale_factor changed in 1.6.0 to align with other frameworks/libraries, and uses scale_factor directly, instead of relying on the computed output size. If you wish to keep the old behavior, please set recompute_scale_factor=True. See the documentation of nn.Upsample for details. \n",
            "  warnings.warn(\"The default behavior for interpolate/upsample with float scale_factor changed \"\n",
            "/usr/local/lib/python3.6/dist-packages/torch/nn/functional.py:3000: UserWarning: The default behavior for interpolate/upsample with float scale_factor changed in 1.6.0 to align with other frameworks/libraries, and uses scale_factor directly, instead of relying on the computed output size. If you wish to keep the old behavior, please set recompute_scale_factor=True. See the documentation of nn.Upsample for details. \n",
            "  warnings.warn(\"The default behavior for interpolate/upsample with float scale_factor changed \"\n"
          ],
          "name": "stderr"
        },
        {
          "output_type": "stream",
          "text": [
            "Learn path /root/.fastai/data/imagewoof2\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "display_data",
          "data": {
            "text/html": [
              "\n",
              "    <div>\n",
              "        <style>\n",
              "            /* Turns off some styling */\n",
              "            progress {\n",
              "                /* gets rid of default border in Firefox and Opera. */\n",
              "                border: none;\n",
              "                /* Needs to be in here for Safari polyfill so background images work as expected. */\n",
              "                background-size: auto;\n",
              "            }\n",
              "            .progress-bar-interrupted, .progress-bar-interrupted::-webkit-progress-bar {\n",
              "                background: #F44336;\n",
              "            }\n",
              "        </style>\n",
              "      <progress value='2' class='' max='5' style='width:300px; height:20px; vertical-align: middle;'></progress>\n",
              "      40.00% [2/5 05:58<08:58]\n",
              "    </div>\n",
              "    \n",
              "<table border=\"1\" class=\"dataframe\">\n",
              "  <thead>\n",
              "    <tr style=\"text-align: left;\">\n",
              "      <th>epoch</th>\n",
              "      <th>train_loss</th>\n",
              "      <th>valid_loss</th>\n",
              "      <th>accuracy</th>\n",
              "      <th>top_k_accuracy</th>\n",
              "      <th>time</th>\n",
              "    </tr>\n",
              "  </thead>\n",
              "  <tbody>\n",
              "    <tr>\n",
              "      <td>0</td>\n",
              "      <td>1.884023</td>\n",
              "      <td>1.835541</td>\n",
              "      <td>0.425045</td>\n",
              "      <td>0.888521</td>\n",
              "      <td>02:59</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>1</td>\n",
              "      <td>1.651628</td>\n",
              "      <td>1.466147</td>\n",
              "      <td>0.588954</td>\n",
              "      <td>0.939679</td>\n",
              "      <td>02:59</td>\n",
              "    </tr>\n",
              "  </tbody>\n",
              "</table><p>\n",
              "\n",
              "    <div>\n",
              "        <style>\n",
              "            /* Turns off some styling */\n",
              "            progress {\n",
              "                /* gets rid of default border in Firefox and Opera. */\n",
              "                border: none;\n",
              "                /* Needs to be in here for Safari polyfill so background images work as expected. */\n",
              "                background-size: auto;\n",
              "            }\n",
              "            .progress-bar-interrupted, .progress-bar-interrupted::-webkit-progress-bar {\n",
              "                background: #F44336;\n",
              "            }\n",
              "        </style>\n",
              "      <progress value='44' class='' max='564' style='width:300px; height:20px; vertical-align: middle;'></progress>\n",
              "      7.80% [44/564 00:11<02:20 1.6061]\n",
              "    </div>\n",
              "    "
            ],
            "text/plain": [
              "<IPython.core.display.HTML object>"
            ]
          },
          "metadata": {
            "tags": []
          }
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "AbJ4W4Wt2Ycj",
        "colab_type": "text"
      },
      "source": [
        "Start with a model with `*2` in all the depthwise convolutions."
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "pbjggxoduo0y",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 1000
        },
        "outputId": "039ba017-edf1-443a-ae6d-0c7140f4a8cf"
      },
      "source": [
        "epochs = 60\n",
        "mixup = 0 if epochs<=20 else 0.2\n",
        "learn = get_learn(model=model, size=192, bs=16, mixup=mixup)\n",
        "learn.fit_fc(epochs, lr=4e-3, moms=(0.95,0.95), start_pct=0.72)\n",
        "print(learn.recorder.metrics[-1][0].item())"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "data path   /root/.fastai/data/imagewoof2\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "stream",
          "text": [
            ""
          ],
          "name": "stderr"
        },
        {
          "output_type": "stream",
          "text": [
            "Learn path /root/.fastai/data/imagewoof2\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "display_data",
          "data": {
            "text/html": [
              "<table border=\"1\" class=\"dataframe\">\n",
              "  <thead>\n",
              "    <tr style=\"text-align: left;\">\n",
              "      <th>epoch</th>\n",
              "      <th>train_loss</th>\n",
              "      <th>valid_loss</th>\n",
              "      <th>accuracy</th>\n",
              "      <th>top_k_accuracy</th>\n",
              "      <th>time</th>\n",
              "    </tr>\n",
              "  </thead>\n",
              "  <tbody>\n",
              "    <tr>\n",
              "      <td>0</td>\n",
              "      <td>2.052636</td>\n",
              "      <td>1.938801</td>\n",
              "      <td>0.348944</td>\n",
              "      <td>0.834563</td>\n",
              "      <td>02:18</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>1</td>\n",
              "      <td>1.828844</td>\n",
              "      <td>1.628165</td>\n",
              "      <td>0.515398</td>\n",
              "      <td>0.912700</td>\n",
              "      <td>02:18</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>2</td>\n",
              "      <td>1.676960</td>\n",
              "      <td>1.429941</td>\n",
              "      <td>0.609061</td>\n",
              "      <td>0.935607</td>\n",
              "      <td>02:17</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>3</td>\n",
              "      <td>1.578073</td>\n",
              "      <td>1.276244</td>\n",
              "      <td>0.681344</td>\n",
              "      <td>0.958005</td>\n",
              "      <td>02:17</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>4</td>\n",
              "      <td>1.481553</td>\n",
              "      <td>1.214398</td>\n",
              "      <td>0.704250</td>\n",
              "      <td>0.957496</td>\n",
              "      <td>02:18</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>5</td>\n",
              "      <td>1.416507</td>\n",
              "      <td>1.187612</td>\n",
              "      <td>0.723339</td>\n",
              "      <td>0.965640</td>\n",
              "      <td>02:18</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>6</td>\n",
              "      <td>1.383856</td>\n",
              "      <td>1.080393</td>\n",
              "      <td>0.768389</td>\n",
              "      <td>0.967931</td>\n",
              "      <td>02:17</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>7</td>\n",
              "      <td>1.324159</td>\n",
              "      <td>1.050738</td>\n",
              "      <td>0.783405</td>\n",
              "      <td>0.973276</td>\n",
              "      <td>02:17</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>8</td>\n",
              "      <td>1.302396</td>\n",
              "      <td>1.006987</td>\n",
              "      <td>0.802749</td>\n",
              "      <td>0.975821</td>\n",
              "      <td>02:17</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>9</td>\n",
              "      <td>1.252993</td>\n",
              "      <td>0.984219</td>\n",
              "      <td>0.808603</td>\n",
              "      <td>0.976839</td>\n",
              "      <td>02:17</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>10</td>\n",
              "      <td>1.243785</td>\n",
              "      <td>0.967685</td>\n",
              "      <td>0.814711</td>\n",
              "      <td>0.977093</td>\n",
              "      <td>02:17</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>11</td>\n",
              "      <td>1.181291</td>\n",
              "      <td>0.982909</td>\n",
              "      <td>0.807585</td>\n",
              "      <td>0.975312</td>\n",
              "      <td>02:17</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>12</td>\n",
              "      <td>1.178232</td>\n",
              "      <td>0.963148</td>\n",
              "      <td>0.812675</td>\n",
              "      <td>0.978112</td>\n",
              "      <td>02:17</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>13</td>\n",
              "      <td>1.146040</td>\n",
              "      <td>0.932477</td>\n",
              "      <td>0.832527</td>\n",
              "      <td>0.982693</td>\n",
              "      <td>02:16</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>14</td>\n",
              "      <td>1.109290</td>\n",
              "      <td>0.928287</td>\n",
              "      <td>0.834563</td>\n",
              "      <td>0.977857</td>\n",
              "      <td>02:17</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>15</td>\n",
              "      <td>1.103288</td>\n",
              "      <td>0.899303</td>\n",
              "      <td>0.836854</td>\n",
              "      <td>0.980148</td>\n",
              "      <td>02:17</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>16</td>\n",
              "      <td>1.098555</td>\n",
              "      <td>0.935650</td>\n",
              "      <td>0.829982</td>\n",
              "      <td>0.979384</td>\n",
              "      <td>02:17</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>17</td>\n",
              "      <td>1.096140</td>\n",
              "      <td>0.884093</td>\n",
              "      <td>0.850598</td>\n",
              "      <td>0.979384</td>\n",
              "      <td>02:17</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>18</td>\n",
              "      <td>1.074255</td>\n",
              "      <td>0.880804</td>\n",
              "      <td>0.847544</td>\n",
              "      <td>0.982184</td>\n",
              "      <td>02:17</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>19</td>\n",
              "      <td>1.070930</td>\n",
              "      <td>0.892950</td>\n",
              "      <td>0.850344</td>\n",
              "      <td>0.981166</td>\n",
              "      <td>02:16</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>20</td>\n",
              "      <td>1.036542</td>\n",
              "      <td>0.898873</td>\n",
              "      <td>0.841435</td>\n",
              "      <td>0.980148</td>\n",
              "      <td>02:17</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>21</td>\n",
              "      <td>1.033563</td>\n",
              "      <td>0.885852</td>\n",
              "      <td>0.851362</td>\n",
              "      <td>0.980657</td>\n",
              "      <td>02:16</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>22</td>\n",
              "      <td>1.022292</td>\n",
              "      <td>0.885886</td>\n",
              "      <td>0.851362</td>\n",
              "      <td>0.982693</td>\n",
              "      <td>02:16</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>23</td>\n",
              "      <td>1.029814</td>\n",
              "      <td>0.866473</td>\n",
              "      <td>0.859506</td>\n",
              "      <td>0.983711</td>\n",
              "      <td>02:18</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>24</td>\n",
              "      <td>0.969939</td>\n",
              "      <td>0.873698</td>\n",
              "      <td>0.862306</td>\n",
              "      <td>0.979130</td>\n",
              "      <td>02:18</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>25</td>\n",
              "      <td>0.975008</td>\n",
              "      <td>0.876239</td>\n",
              "      <td>0.854416</td>\n",
              "      <td>0.981420</td>\n",
              "      <td>02:18</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>26</td>\n",
              "      <td>1.005038</td>\n",
              "      <td>0.889756</td>\n",
              "      <td>0.853143</td>\n",
              "      <td>0.979639</td>\n",
              "      <td>02:17</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>27</td>\n",
              "      <td>0.968257</td>\n",
              "      <td>0.864208</td>\n",
              "      <td>0.868160</td>\n",
              "      <td>0.980148</td>\n",
              "      <td>02:17</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>28</td>\n",
              "      <td>0.967392</td>\n",
              "      <td>0.877275</td>\n",
              "      <td>0.862051</td>\n",
              "      <td>0.984983</td>\n",
              "      <td>02:16</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>29</td>\n",
              "      <td>0.990795</td>\n",
              "      <td>0.859770</td>\n",
              "      <td>0.861542</td>\n",
              "      <td>0.979384</td>\n",
              "      <td>02:16</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>30</td>\n",
              "      <td>0.970307</td>\n",
              "      <td>0.883700</td>\n",
              "      <td>0.854670</td>\n",
              "      <td>0.980911</td>\n",
              "      <td>02:15</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>31</td>\n",
              "      <td>0.983691</td>\n",
              "      <td>0.844288</td>\n",
              "      <td>0.865869</td>\n",
              "      <td>0.980911</td>\n",
              "      <td>02:16</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>32</td>\n",
              "      <td>0.971488</td>\n",
              "      <td>0.855935</td>\n",
              "      <td>0.868923</td>\n",
              "      <td>0.981675</td>\n",
              "      <td>02:16</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>33</td>\n",
              "      <td>0.959850</td>\n",
              "      <td>0.880706</td>\n",
              "      <td>0.853398</td>\n",
              "      <td>0.977093</td>\n",
              "      <td>02:16</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>34</td>\n",
              "      <td>0.952909</td>\n",
              "      <td>0.864729</td>\n",
              "      <td>0.859506</td>\n",
              "      <td>0.981420</td>\n",
              "      <td>02:16</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>35</td>\n",
              "      <td>0.938942</td>\n",
              "      <td>0.860425</td>\n",
              "      <td>0.862815</td>\n",
              "      <td>0.983965</td>\n",
              "      <td>02:17</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>36</td>\n",
              "      <td>0.921134</td>\n",
              "      <td>0.870239</td>\n",
              "      <td>0.858234</td>\n",
              "      <td>0.981675</td>\n",
              "      <td>02:18</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>37</td>\n",
              "      <td>0.918442</td>\n",
              "      <td>0.855653</td>\n",
              "      <td>0.868160</td>\n",
              "      <td>0.981166</td>\n",
              "      <td>02:19</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>38</td>\n",
              "      <td>0.916307</td>\n",
              "      <td>0.892390</td>\n",
              "      <td>0.855179</td>\n",
              "      <td>0.979893</td>\n",
              "      <td>02:18</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>39</td>\n",
              "      <td>0.941971</td>\n",
              "      <td>0.886629</td>\n",
              "      <td>0.849071</td>\n",
              "      <td>0.981420</td>\n",
              "      <td>02:18</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>40</td>\n",
              "      <td>0.903531</td>\n",
              "      <td>0.862708</td>\n",
              "      <td>0.868669</td>\n",
              "      <td>0.982947</td>\n",
              "      <td>02:18</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>41</td>\n",
              "      <td>0.918182</td>\n",
              "      <td>0.865012</td>\n",
              "      <td>0.860015</td>\n",
              "      <td>0.981675</td>\n",
              "      <td>02:17</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>42</td>\n",
              "      <td>0.917140</td>\n",
              "      <td>0.856310</td>\n",
              "      <td>0.865615</td>\n",
              "      <td>0.982693</td>\n",
              "      <td>02:17</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>43</td>\n",
              "      <td>0.922041</td>\n",
              "      <td>0.860277</td>\n",
              "      <td>0.866887</td>\n",
              "      <td>0.977857</td>\n",
              "      <td>02:17</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>44</td>\n",
              "      <td>0.923458</td>\n",
              "      <td>0.882930</td>\n",
              "      <td>0.856707</td>\n",
              "      <td>0.980148</td>\n",
              "      <td>02:17</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>45</td>\n",
              "      <td>0.887316</td>\n",
              "      <td>0.835926</td>\n",
              "      <td>0.868414</td>\n",
              "      <td>0.981420</td>\n",
              "      <td>02:18</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>46</td>\n",
              "      <td>0.885578</td>\n",
              "      <td>0.830366</td>\n",
              "      <td>0.876559</td>\n",
              "      <td>0.981166</td>\n",
              "      <td>02:18</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>47</td>\n",
              "      <td>0.898508</td>\n",
              "      <td>0.846862</td>\n",
              "      <td>0.869687</td>\n",
              "      <td>0.980148</td>\n",
              "      <td>02:18</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>48</td>\n",
              "      <td>0.860434</td>\n",
              "      <td>0.827727</td>\n",
              "      <td>0.876050</td>\n",
              "      <td>0.982693</td>\n",
              "      <td>02:17</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>49</td>\n",
              "      <td>0.856632</td>\n",
              "      <td>0.835957</td>\n",
              "      <td>0.871214</td>\n",
              "      <td>0.979893</td>\n",
              "      <td>02:17</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>50</td>\n",
              "      <td>0.851344</td>\n",
              "      <td>0.819126</td>\n",
              "      <td>0.881649</td>\n",
              "      <td>0.984474</td>\n",
              "      <td>02:17</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>51</td>\n",
              "      <td>0.843528</td>\n",
              "      <td>0.818000</td>\n",
              "      <td>0.881649</td>\n",
              "      <td>0.982693</td>\n",
              "      <td>02:16</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>52</td>\n",
              "      <td>0.822556</td>\n",
              "      <td>0.801662</td>\n",
              "      <td>0.884703</td>\n",
              "      <td>0.984474</td>\n",
              "      <td>02:16</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>53</td>\n",
              "      <td>0.838933</td>\n",
              "      <td>0.798329</td>\n",
              "      <td>0.889030</td>\n",
              "      <td>0.983202</td>\n",
              "      <td>02:18</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>54</td>\n",
              "      <td>0.808592</td>\n",
              "      <td>0.788474</td>\n",
              "      <td>0.893866</td>\n",
              "      <td>0.978875</td>\n",
              "      <td>02:18</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>55</td>\n",
              "      <td>0.811359</td>\n",
              "      <td>0.786461</td>\n",
              "      <td>0.890812</td>\n",
              "      <td>0.983456</td>\n",
              "      <td>02:18</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>56</td>\n",
              "      <td>0.808781</td>\n",
              "      <td>0.781056</td>\n",
              "      <td>0.892848</td>\n",
              "      <td>0.985747</td>\n",
              "      <td>02:19</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>57</td>\n",
              "      <td>0.799051</td>\n",
              "      <td>0.778129</td>\n",
              "      <td>0.895902</td>\n",
              "      <td>0.983965</td>\n",
              "      <td>02:19</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>58</td>\n",
              "      <td>0.785564</td>\n",
              "      <td>0.776103</td>\n",
              "      <td>0.897938</td>\n",
              "      <td>0.984729</td>\n",
              "      <td>02:18</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>59</td>\n",
              "      <td>0.816603</td>\n",
              "      <td>0.773218</td>\n",
              "      <td>0.897938</td>\n",
              "      <td>0.985492</td>\n",
              "      <td>02:19</td>\n",
              "    </tr>\n",
              "  </tbody>\n",
              "</table>"
            ],
            "text/plain": [
              "<IPython.core.display.HTML object>"
            ]
          },
          "metadata": {
            "tags": []
          }
        },
        {
          "output_type": "stream",
          "text": [
            ""
          ],
          "name": "stderr"
        },
        {
          "output_type": "stream",
          "text": [
            "0.8979384303092957\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "-wPn74fNVDUi",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 88
        },
        "outputId": "258a51c3-0f30-4a11-8e4b-2c0161d709fb"
      },
      "source": [
        "learn.validate()"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "display_data",
          "data": {
            "text/html": [
              ""
            ],
            "text/plain": [
              "<IPython.core.display.HTML object>"
            ]
          },
          "metadata": {
            "tags": []
          }
        },
        {
          "output_type": "stream",
          "text": [
            ""
          ],
          "name": "stderr"
        },
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "[0.77321756, tensor(0.8979), tensor(0.9855)]"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 94
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "wlQ-z3vdTH9E",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 34
        },
        "outputId": "67bf5079-da8d-4829-d51b-a05050a8bd44"
      },
      "source": [
        "state = learn.model.state_dict()\n",
        "channels = 0\n",
        "for name in state:\n",
        "  if 'Conv3x3' in name and channels==0:\n",
        "    a,b,c,d = state[name].size()\n",
        "    double = state[name].unsqueeze(1).expand(a,2,b,c,d).reshape(a*2,b,c,d)\n",
        "    state[name] = double / 1.414\n",
        "    channels = a\n",
        "  if 'Conv1x1' in name and state[name].size()[1]==channels:\n",
        "    a,b,c,d = state[name].size()\n",
        "    double = state[name].unsqueeze(2).expand(a,b,2,c,d).reshape(a,b*2,c,d)\n",
        "    state[name] = double / 1.414\n",
        "    channels = 0\n",
        "print('done doubling')"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "done doubling\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "kN3hl0Eh2JSc",
        "colab_type": "text"
      },
      "source": [
        "Here, manually change the `*2` to `*4` (in four places), rerun `model`, and proceed below:"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "OzeJ1SIRxgDx",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 224
        },
        "outputId": "61394367-d224-420f-f24c-239533318bf1"
      },
      "source": [
        "epochs = 10\n",
        "mixup = 0 if epochs<=20 else 0.2\n",
        "learn2 = get_learn(model=model, size=192, bs=16, mixup=mixup)\n",
        "learn2.model.load_state_dict(state)\n",
        "print(learn2.validate())"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "data path   /root/.fastai/data/imagewoof2\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "stream",
          "text": [
            ""
          ],
          "name": "stderr"
        },
        {
          "output_type": "stream",
          "text": [
            "Learn path /root/.fastai/data/imagewoof2\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "display_data",
          "data": {
            "text/html": [
              ""
            ],
            "text/plain": [
              "<IPython.core.display.HTML object>"
            ]
          },
          "metadata": {
            "tags": []
          }
        },
        {
          "output_type": "stream",
          "text": [
            "[0.7732239, tensor(0.8979), tensor(0.9855)]\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "CpP1GCi6hNAs",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 430
        },
        "outputId": "8f23306e-e17e-44ad-be72-b1a25114c20f"
      },
      "source": [
        "learn2.fit_fc(10, lr=4e-3, moms=(0.95,0.95), start_pct=0.72)\n",
        "print(learn2.recorder.metrics[-1][0].item())"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "display_data",
          "data": {
            "text/html": [
              "<table border=\"1\" class=\"dataframe\">\n",
              "  <thead>\n",
              "    <tr style=\"text-align: left;\">\n",
              "      <th>epoch</th>\n",
              "      <th>train_loss</th>\n",
              "      <th>valid_loss</th>\n",
              "      <th>accuracy</th>\n",
              "      <th>top_k_accuracy</th>\n",
              "      <th>time</th>\n",
              "    </tr>\n",
              "  </thead>\n",
              "  <tbody>\n",
              "    <tr>\n",
              "      <td>0</td>\n",
              "      <td>0.606666</td>\n",
              "      <td>0.875958</td>\n",
              "      <td>0.854416</td>\n",
              "      <td>0.978621</td>\n",
              "      <td>02:36</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>1</td>\n",
              "      <td>0.637473</td>\n",
              "      <td>0.875326</td>\n",
              "      <td>0.849835</td>\n",
              "      <td>0.977093</td>\n",
              "      <td>02:36</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>2</td>\n",
              "      <td>0.619375</td>\n",
              "      <td>0.890647</td>\n",
              "      <td>0.856707</td>\n",
              "      <td>0.975057</td>\n",
              "      <td>02:36</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>3</td>\n",
              "      <td>0.620187</td>\n",
              "      <td>0.899035</td>\n",
              "      <td>0.851107</td>\n",
              "      <td>0.979639</td>\n",
              "      <td>02:36</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>4</td>\n",
              "      <td>0.611394</td>\n",
              "      <td>0.852904</td>\n",
              "      <td>0.869941</td>\n",
              "      <td>0.979639</td>\n",
              "      <td>02:36</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>5</td>\n",
              "      <td>0.639448</td>\n",
              "      <td>0.892186</td>\n",
              "      <td>0.853907</td>\n",
              "      <td>0.979130</td>\n",
              "      <td>02:36</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>6</td>\n",
              "      <td>0.632713</td>\n",
              "      <td>0.866875</td>\n",
              "      <td>0.863324</td>\n",
              "      <td>0.975057</td>\n",
              "      <td>02:36</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>7</td>\n",
              "      <td>0.603388</td>\n",
              "      <td>0.837288</td>\n",
              "      <td>0.869687</td>\n",
              "      <td>0.984220</td>\n",
              "      <td>02:35</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>8</td>\n",
              "      <td>0.566116</td>\n",
              "      <td>0.818019</td>\n",
              "      <td>0.878595</td>\n",
              "      <td>0.981166</td>\n",
              "      <td>02:36</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>9</td>\n",
              "      <td>0.539309</td>\n",
              "      <td>0.791606</td>\n",
              "      <td>0.888012</td>\n",
              "      <td>0.982693</td>\n",
              "      <td>02:35</td>\n",
              "    </tr>\n",
              "  </tbody>\n",
              "</table>"
            ],
            "text/plain": [
              "<IPython.core.display.HTML object>"
            ]
          },
          "metadata": {
            "tags": []
          }
        },
        {
          "output_type": "stream",
          "text": [
            ""
          ],
          "name": "stderr"
        },
        {
          "output_type": "stream",
          "text": [
            "0.8880122303962708\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "Ns97FjlTDqUx",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 162
        },
        "outputId": "671dc7c3-48cb-4f1f-8d7b-1884dfe24ad2"
      },
      "source": [
        "learn.validate()"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "error",
          "ename": "NameError",
          "evalue": "ignored",
          "traceback": [
            "\u001b[0;31m---------------------------------------------------------------------------\u001b[0m",
            "\u001b[0;31mNameError\u001b[0m                                 Traceback (most recent call last)",
            "\u001b[0;32m<ipython-input-1-631604a2e07b>\u001b[0m in \u001b[0;36m<module>\u001b[0;34m()\u001b[0m\n\u001b[0;32m----> 1\u001b[0;31m \u001b[0mlearn\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mvalidate\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m",
            "\u001b[0;31mNameError\u001b[0m: name 'learn' is not defined"
          ]
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "DPDHUUEWtxLo",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 662
        },
        "outputId": "ee122d45-9daf-4df8-f52c-9b1e9ff94605"
      },
      "source": [
        "learn2.fit_fc(40, lr=4e-3, moms=(0.95,0.95), start_pct=0.72)"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "display_data",
          "data": {
            "text/html": [
              "\n",
              "    <div>\n",
              "        <style>\n",
              "            /* Turns off some styling */\n",
              "            progress {\n",
              "                /* gets rid of default border in Firefox and Opera. */\n",
              "                border: none;\n",
              "                /* Needs to be in here for Safari polyfill so background images work as expected. */\n",
              "                background-size: auto;\n",
              "            }\n",
              "            .progress-bar-interrupted, .progress-bar-interrupted::-webkit-progress-bar {\n",
              "                background: #F44336;\n",
              "            }\n",
              "        </style>\n",
              "      <progress value='16' class='' max='40' style='width:300px; height:20px; vertical-align: middle;'></progress>\n",
              "      40.00% [16/40 42:12<1:03:18]\n",
              "    </div>\n",
              "    \n",
              "<table border=\"1\" class=\"dataframe\">\n",
              "  <thead>\n",
              "    <tr style=\"text-align: left;\">\n",
              "      <th>epoch</th>\n",
              "      <th>train_loss</th>\n",
              "      <th>valid_loss</th>\n",
              "      <th>accuracy</th>\n",
              "      <th>top_k_accuracy</th>\n",
              "      <th>time</th>\n",
              "    </tr>\n",
              "  </thead>\n",
              "  <tbody>\n",
              "    <tr>\n",
              "      <td>0</td>\n",
              "      <td>0.665684</td>\n",
              "      <td>0.900977</td>\n",
              "      <td>0.847289</td>\n",
              "      <td>0.977348</td>\n",
              "      <td>02:38</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>1</td>\n",
              "      <td>0.659895</td>\n",
              "      <td>0.883306</td>\n",
              "      <td>0.859506</td>\n",
              "      <td>0.976330</td>\n",
              "      <td>02:37</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>2</td>\n",
              "      <td>0.644420</td>\n",
              "      <td>0.890825</td>\n",
              "      <td>0.850344</td>\n",
              "      <td>0.980911</td>\n",
              "      <td>02:37</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>3</td>\n",
              "      <td>0.639653</td>\n",
              "      <td>0.889162</td>\n",
              "      <td>0.851871</td>\n",
              "      <td>0.976839</td>\n",
              "      <td>02:37</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>4</td>\n",
              "      <td>0.636691</td>\n",
              "      <td>0.873973</td>\n",
              "      <td>0.859506</td>\n",
              "      <td>0.978112</td>\n",
              "      <td>02:37</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>5</td>\n",
              "      <td>0.646427</td>\n",
              "      <td>0.874967</td>\n",
              "      <td>0.860270</td>\n",
              "      <td>0.975057</td>\n",
              "      <td>02:37</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>6</td>\n",
              "      <td>0.623019</td>\n",
              "      <td>0.871699</td>\n",
              "      <td>0.855943</td>\n",
              "      <td>0.977602</td>\n",
              "      <td>02:37</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>7</td>\n",
              "      <td>0.618884</td>\n",
              "      <td>0.850166</td>\n",
              "      <td>0.864851</td>\n",
              "      <td>0.980911</td>\n",
              "      <td>02:38</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>8</td>\n",
              "      <td>0.618632</td>\n",
              "      <td>0.862228</td>\n",
              "      <td>0.869432</td>\n",
              "      <td>0.978112</td>\n",
              "      <td>02:38</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>9</td>\n",
              "      <td>0.617541</td>\n",
              "      <td>0.883283</td>\n",
              "      <td>0.856452</td>\n",
              "      <td>0.977602</td>\n",
              "      <td>02:37</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>10</td>\n",
              "      <td>0.622270</td>\n",
              "      <td>0.873347</td>\n",
              "      <td>0.864597</td>\n",
              "      <td>0.972258</td>\n",
              "      <td>02:41</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>11</td>\n",
              "      <td>0.634060</td>\n",
              "      <td>0.866858</td>\n",
              "      <td>0.857470</td>\n",
              "      <td>0.981929</td>\n",
              "      <td>02:38</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>12</td>\n",
              "      <td>0.618130</td>\n",
              "      <td>0.874891</td>\n",
              "      <td>0.861797</td>\n",
              "      <td>0.974548</td>\n",
              "      <td>02:38</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>13</td>\n",
              "      <td>0.626946</td>\n",
              "      <td>0.897736</td>\n",
              "      <td>0.855688</td>\n",
              "      <td>0.975821</td>\n",
              "      <td>02:38</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>14</td>\n",
              "      <td>0.597168</td>\n",
              "      <td>0.878698</td>\n",
              "      <td>0.860524</td>\n",
              "      <td>0.980402</td>\n",
              "      <td>02:38</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>15</td>\n",
              "      <td>0.613733</td>\n",
              "      <td>0.877535</td>\n",
              "      <td>0.856707</td>\n",
              "      <td>0.977602</td>\n",
              "      <td>02:38</td>\n",
              "    </tr>\n",
              "  </tbody>\n",
              "</table><p>\n",
              "\n",
              "    <div>\n",
              "        <style>\n",
              "            /* Turns off some styling */\n",
              "            progress {\n",
              "                /* gets rid of default border in Firefox and Opera. */\n",
              "                border: none;\n",
              "                /* Needs to be in here for Safari polyfill so background images work as expected. */\n",
              "                background-size: auto;\n",
              "            }\n",
              "            .progress-bar-interrupted, .progress-bar-interrupted::-webkit-progress-bar {\n",
              "                background: #F44336;\n",
              "            }\n",
              "        </style>\n",
              "      <progress value='535' class='' max='564' style='width:300px; height:20px; vertical-align: middle;'></progress>\n",
              "      94.86% [535/564 01:57<00:06 0.6036]\n",
              "    </div>\n",
              "    "
            ],
            "text/plain": [
              "<IPython.core.display.HTML object>"
            ]
          },
          "metadata": {
            "tags": []
          }
        },
        {
          "output_type": "stream",
          "text": [
            ""
          ],
          "name": "stderr"
        },
        {
          "output_type": "stream",
          "text": [
            "Buffered data was truncated after reaching the output size limit."
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "2Z-VwCjLOW0I",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 697
        },
        "outputId": "95753095-dca9-47cd-bc90-18f65d5a6989"
      },
      "source": [
        "for i in range(40):\n",
        "  print(i, learn2.recorder.val_losses[i].item(), learn2.recorder.metrics[i][0].item())"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "0 0.9009769558906555 0.8472893834114075\n",
            "1 0.8833058476448059 0.8595062494277954\n",
            "2 0.8908252120018005 0.8503435850143433\n",
            "3 0.8891615271568298 0.8518707156181335\n",
            "4 0.8739725947380066 0.8595062494277954\n",
            "5 0.8749666810035706 0.8602697849273682\n",
            "6 0.8716986179351807 0.855942964553833\n",
            "7 0.8501657247543335 0.8648511171340942\n",
            "8 0.8622281551361084 0.8694324493408203\n",
            "9 0.8832833766937256 0.8564520478248596\n",
            "10 0.8733472228050232 0.8645966053009033\n",
            "11 0.8668575286865234 0.8574700951576233\n",
            "12 0.8748914003372192 0.8617969155311584\n",
            "13 0.8977355360984802 0.8556884527206421\n",
            "14 0.8786979913711548 0.8605242967605591\n",
            "15 0.877534806728363 0.8567065596580505\n",
            "16 0.8835002183914185 0.8589972257614136\n",
            "17 0.88120436668396 0.8623059391975403\n",
            "18 0.8919264078140259 0.8546704053878784\n",
            "19 0.8665375113487244 0.8597607612609863\n",
            "20 0.8747259378433228 0.8567065596580505\n",
            "21 0.8497205376625061 0.8610333204269409\n",
            "22 0.8672081828117371 0.8635784983634949\n",
            "23 0.8595353364944458 0.868159830570221\n",
            "24 0.8896135091781616 0.8495800495147705\n",
            "25 0.8718185424804688 0.8592517375946045\n",
            "26 0.8856318593025208 0.8523797392845154\n",
            "27 0.8881120085716248 0.8610333204269409\n",
            "28 0.9008130431175232 0.8521252274513245\n",
            "29 0.8691977262496948 0.8643420934677124\n",
            "30 0.8692021369934082 0.8648511171340942\n",
            "31 0.8541252613067627 0.8701959848403931\n",
            "32 0.8769975900650024 0.8592517375946045\n",
            "33 0.8778439164161682 0.8620514273643494\n",
            "34 0.8164481520652771 0.8844489455223083\n",
            "35 0.8101176023483276 0.8854670524597168\n",
            "36 0.8032693862915039 0.8854670524597168\n",
            "37 0.7960595488548279 0.8875032067298889\n",
            "38 0.7937077879905701 0.8882667422294617\n",
            "39 0.7937994003295898 0.8859760761260986\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "Iw49hpbXmhBZ",
        "colab_type": "text"
      },
      "source": [
        "# results"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "8g2J_rBABz13",
        "colab_type": "text"
      },
      "source": [
        "## size=128\n"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "NjZMpk8OVhos",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 1000
        },
        "outputId": "106cac0c-218b-4420-9530-a99178e6515c"
      },
      "source": [
        "dm = 4\n",
        "res = dict()\n",
        "for e in [5]*5 + [20]*5 + [80]:\n",
        "  mixup=0 if e<=20 else 0.2\n",
        "  learn = get_learn(model=model, size=128, bs=16, mixup=mixup)\n",
        "  learn.fit_fc(e, lr=4e-3, moms=(0.95,0.95), start_pct=0.72)\n",
        "  acc = learn.recorder.metrics[-1][0].item()\n",
        "  if e in res:\n",
        "    res[e] += [acc]\n",
        "  else:\n",
        "    res[e] = [acc]\n",
        "  print({e: sum(res[e])/len(res[e]) for e in res})"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Downloading https://s3.amazonaws.com/fast-ai-imageclas/imagewoof2.tgz\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "display_data",
          "data": {
            "text/html": [
              ""
            ],
            "text/plain": [
              "<IPython.core.display.HTML object>"
            ]
          },
          "metadata": {
            "tags": []
          }
        },
        {
          "output_type": "stream",
          "text": [
            "data path   /root/.fastai/data/imagewoof2\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "stream",
          "text": [
            ""
          ],
          "name": "stderr"
        },
        {
          "output_type": "stream",
          "text": [
            "Learn path /root/.fastai/data/imagewoof2\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "display_data",
          "data": {
            "text/html": [
              "<table border=\"1\" class=\"dataframe\">\n",
              "  <thead>\n",
              "    <tr style=\"text-align: left;\">\n",
              "      <th>epoch</th>\n",
              "      <th>train_loss</th>\n",
              "      <th>valid_loss</th>\n",
              "      <th>accuracy</th>\n",
              "      <th>top_k_accuracy</th>\n",
              "      <th>time</th>\n",
              "    </tr>\n",
              "  </thead>\n",
              "  <tbody>\n",
              "    <tr>\n",
              "      <td>0</td>\n",
              "      <td>1.961671</td>\n",
              "      <td>1.833559</td>\n",
              "      <td>0.419700</td>\n",
              "      <td>0.879104</td>\n",
              "      <td>01:43</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>1</td>\n",
              "      <td>1.671203</td>\n",
              "      <td>1.519507</td>\n",
              "      <td>0.545940</td>\n",
              "      <td>0.930008</td>\n",
              "      <td>01:42</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>2</td>\n",
              "      <td>1.506109</td>\n",
              "      <td>1.366296</td>\n",
              "      <td>0.635022</td>\n",
              "      <td>0.940443</td>\n",
              "      <td>01:43</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>3</td>\n",
              "      <td>1.357213</td>\n",
              "      <td>1.285933</td>\n",
              "      <td>0.673963</td>\n",
              "      <td>0.952151</td>\n",
              "      <td>01:43</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>4</td>\n",
              "      <td>1.169711</td>\n",
              "      <td>1.106144</td>\n",
              "      <td>0.759226</td>\n",
              "      <td>0.971749</td>\n",
              "      <td>01:42</td>\n",
              "    </tr>\n",
              "  </tbody>\n",
              "</table>"
            ],
            "text/plain": [
              "<IPython.core.display.HTML object>"
            ]
          },
          "metadata": {
            "tags": []
          }
        },
        {
          "output_type": "stream",
          "text": [
            ""
          ],
          "name": "stderr"
        },
        {
          "output_type": "stream",
          "text": [
            "{5: 0.7592262625694275}\n",
            "data path   /root/.fastai/data/imagewoof2\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "stream",
          "text": [
            ""
          ],
          "name": "stderr"
        },
        {
          "output_type": "stream",
          "text": [
            "Learn path /root/.fastai/data/imagewoof2\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "display_data",
          "data": {
            "text/html": [
              "<table border=\"1\" class=\"dataframe\">\n",
              "  <thead>\n",
              "    <tr style=\"text-align: left;\">\n",
              "      <th>epoch</th>\n",
              "      <th>train_loss</th>\n",
              "      <th>valid_loss</th>\n",
              "      <th>accuracy</th>\n",
              "      <th>top_k_accuracy</th>\n",
              "      <th>time</th>\n",
              "    </tr>\n",
              "  </thead>\n",
              "  <tbody>\n",
              "    <tr>\n",
              "      <td>0</td>\n",
              "      <td>1.966546</td>\n",
              "      <td>1.772155</td>\n",
              "      <td>0.425554</td>\n",
              "      <td>0.890557</td>\n",
              "      <td>01:40</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>1</td>\n",
              "      <td>1.689572</td>\n",
              "      <td>1.553331</td>\n",
              "      <td>0.538305</td>\n",
              "      <td>0.922372</td>\n",
              "      <td>01:40</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>2</td>\n",
              "      <td>1.498202</td>\n",
              "      <td>1.353891</td>\n",
              "      <td>0.640876</td>\n",
              "      <td>0.947569</td>\n",
              "      <td>01:41</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>3</td>\n",
              "      <td>1.341627</td>\n",
              "      <td>1.296018</td>\n",
              "      <td>0.678035</td>\n",
              "      <td>0.952151</td>\n",
              "      <td>01:41</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>4</td>\n",
              "      <td>1.190779</td>\n",
              "      <td>1.114605</td>\n",
              "      <td>0.756172</td>\n",
              "      <td>0.970221</td>\n",
              "      <td>01:41</td>\n",
              "    </tr>\n",
              "  </tbody>\n",
              "</table>"
            ],
            "text/plain": [
              "<IPython.core.display.HTML object>"
            ]
          },
          "metadata": {
            "tags": []
          }
        },
        {
          "output_type": "stream",
          "text": [
            "{5: 0.7576991617679596}\n",
            "data path   /root/.fastai/data/imagewoof2\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "stream",
          "text": [
            ""
          ],
          "name": "stderr"
        },
        {
          "output_type": "stream",
          "text": [
            "Learn path /root/.fastai/data/imagewoof2\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "display_data",
          "data": {
            "text/html": [
              "<table border=\"1\" class=\"dataframe\">\n",
              "  <thead>\n",
              "    <tr style=\"text-align: left;\">\n",
              "      <th>epoch</th>\n",
              "      <th>train_loss</th>\n",
              "      <th>valid_loss</th>\n",
              "      <th>accuracy</th>\n",
              "      <th>top_k_accuracy</th>\n",
              "      <th>time</th>\n",
              "    </tr>\n",
              "  </thead>\n",
              "  <tbody>\n",
              "    <tr>\n",
              "      <td>0</td>\n",
              "      <td>1.893834</td>\n",
              "      <td>1.868907</td>\n",
              "      <td>0.421481</td>\n",
              "      <td>0.876304</td>\n",
              "      <td>01:41</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>1</td>\n",
              "      <td>1.650119</td>\n",
              "      <td>1.527944</td>\n",
              "      <td>0.548231</td>\n",
              "      <td>0.932298</td>\n",
              "      <td>01:41</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>2</td>\n",
              "      <td>1.482323</td>\n",
              "      <td>1.371284</td>\n",
              "      <td>0.632476</td>\n",
              "      <td>0.947060</td>\n",
              "      <td>01:41</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>3</td>\n",
              "      <td>1.349521</td>\n",
              "      <td>1.266608</td>\n",
              "      <td>0.682107</td>\n",
              "      <td>0.952660</td>\n",
              "      <td>01:41</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>4</td>\n",
              "      <td>1.119496</td>\n",
              "      <td>1.094847</td>\n",
              "      <td>0.764826</td>\n",
              "      <td>0.973021</td>\n",
              "      <td>01:40</td>\n",
              "    </tr>\n",
              "  </tbody>\n",
              "</table>"
            ],
            "text/plain": [
              "<IPython.core.display.HTML object>"
            ]
          },
          "metadata": {
            "tags": []
          }
        },
        {
          "output_type": "stream",
          "text": [
            "{5: 0.7600746552149454}\n",
            "data path   /root/.fastai/data/imagewoof2\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "stream",
          "text": [
            ""
          ],
          "name": "stderr"
        },
        {
          "output_type": "stream",
          "text": [
            "Learn path /root/.fastai/data/imagewoof2\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "display_data",
          "data": {
            "text/html": [
              "<table border=\"1\" class=\"dataframe\">\n",
              "  <thead>\n",
              "    <tr style=\"text-align: left;\">\n",
              "      <th>epoch</th>\n",
              "      <th>train_loss</th>\n",
              "      <th>valid_loss</th>\n",
              "      <th>accuracy</th>\n",
              "      <th>top_k_accuracy</th>\n",
              "      <th>time</th>\n",
              "    </tr>\n",
              "  </thead>\n",
              "  <tbody>\n",
              "    <tr>\n",
              "      <td>0</td>\n",
              "      <td>1.947112</td>\n",
              "      <td>1.811481</td>\n",
              "      <td>0.408755</td>\n",
              "      <td>0.868160</td>\n",
              "      <td>01:40</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>1</td>\n",
              "      <td>1.633910</td>\n",
              "      <td>1.547457</td>\n",
              "      <td>0.538305</td>\n",
              "      <td>0.927462</td>\n",
              "      <td>01:39</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>2</td>\n",
              "      <td>1.489909</td>\n",
              "      <td>1.362312</td>\n",
              "      <td>0.635276</td>\n",
              "      <td>0.951133</td>\n",
              "      <td>01:40</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>3</td>\n",
              "      <td>1.360234</td>\n",
              "      <td>1.283221</td>\n",
              "      <td>0.667345</td>\n",
              "      <td>0.956732</td>\n",
              "      <td>01:41</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>4</td>\n",
              "      <td>1.167257</td>\n",
              "      <td>1.085048</td>\n",
              "      <td>0.770680</td>\n",
              "      <td>0.974039</td>\n",
              "      <td>01:40</td>\n",
              "    </tr>\n",
              "  </tbody>\n",
              "</table>"
            ],
            "text/plain": [
              "<IPython.core.display.HTML object>"
            ]
          },
          "metadata": {
            "tags": []
          }
        },
        {
          "output_type": "stream",
          "text": [
            "{5: 0.7627258747816086}\n",
            "data path   /root/.fastai/data/imagewoof2\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "stream",
          "text": [
            ""
          ],
          "name": "stderr"
        },
        {
          "output_type": "stream",
          "text": [
            "Learn path /root/.fastai/data/imagewoof2\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "display_data",
          "data": {
            "text/html": [
              "<table border=\"1\" class=\"dataframe\">\n",
              "  <thead>\n",
              "    <tr style=\"text-align: left;\">\n",
              "      <th>epoch</th>\n",
              "      <th>train_loss</th>\n",
              "      <th>valid_loss</th>\n",
              "      <th>accuracy</th>\n",
              "      <th>top_k_accuracy</th>\n",
              "      <th>time</th>\n",
              "    </tr>\n",
              "  </thead>\n",
              "  <tbody>\n",
              "    <tr>\n",
              "      <td>0</td>\n",
              "      <td>1.966641</td>\n",
              "      <td>1.801435</td>\n",
              "      <td>0.439298</td>\n",
              "      <td>0.867396</td>\n",
              "      <td>01:42</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>1</td>\n",
              "      <td>1.696564</td>\n",
              "      <td>1.613575</td>\n",
              "      <td>0.504963</td>\n",
              "      <td>0.919827</td>\n",
              "      <td>01:42</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>2</td>\n",
              "      <td>1.522174</td>\n",
              "      <td>1.394349</td>\n",
              "      <td>0.618478</td>\n",
              "      <td>0.949860</td>\n",
              "      <td>01:41</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>3</td>\n",
              "      <td>1.387305</td>\n",
              "      <td>1.312356</td>\n",
              "      <td>0.654874</td>\n",
              "      <td>0.958259</td>\n",
              "      <td>01:41</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>4</td>\n",
              "      <td>1.146989</td>\n",
              "      <td>1.090402</td>\n",
              "      <td>0.760244</td>\n",
              "      <td>0.973785</td>\n",
              "      <td>01:42</td>\n",
              "    </tr>\n",
              "  </tbody>\n",
              "</table>"
            ],
            "text/plain": [
              "<IPython.core.display.HTML object>"
            ]
          },
          "metadata": {
            "tags": []
          }
        },
        {
          "output_type": "stream",
          "text": [
            "{5: 0.7622295618057251}\n",
            "data path   /root/.fastai/data/imagewoof2\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "stream",
          "text": [
            ""
          ],
          "name": "stderr"
        },
        {
          "output_type": "stream",
          "text": [
            "Learn path /root/.fastai/data/imagewoof2\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "display_data",
          "data": {
            "text/html": [
              "<table border=\"1\" class=\"dataframe\">\n",
              "  <thead>\n",
              "    <tr style=\"text-align: left;\">\n",
              "      <th>epoch</th>\n",
              "      <th>train_loss</th>\n",
              "      <th>valid_loss</th>\n",
              "      <th>accuracy</th>\n",
              "      <th>top_k_accuracy</th>\n",
              "      <th>time</th>\n",
              "    </tr>\n",
              "  </thead>\n",
              "  <tbody>\n",
              "    <tr>\n",
              "      <td>0</td>\n",
              "      <td>1.942095</td>\n",
              "      <td>1.803220</td>\n",
              "      <td>0.416645</td>\n",
              "      <td>0.872232</td>\n",
              "      <td>01:40</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>1</td>\n",
              "      <td>1.683641</td>\n",
              "      <td>1.593508</td>\n",
              "      <td>0.524561</td>\n",
              "      <td>0.928735</td>\n",
              "      <td>01:41</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>2</td>\n",
              "      <td>1.544361</td>\n",
              "      <td>1.408688</td>\n",
              "      <td>0.610588</td>\n",
              "      <td>0.941206</td>\n",
              "      <td>01:41</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>3</td>\n",
              "      <td>1.361249</td>\n",
              "      <td>1.280145</td>\n",
              "      <td>0.666836</td>\n",
              "      <td>0.958005</td>\n",
              "      <td>01:41</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>4</td>\n",
              "      <td>1.304440</td>\n",
              "      <td>1.230721</td>\n",
              "      <td>0.699415</td>\n",
              "      <td>0.964368</td>\n",
              "      <td>01:41</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>5</td>\n",
              "      <td>1.266698</td>\n",
              "      <td>1.149605</td>\n",
              "      <td>0.734029</td>\n",
              "      <td>0.970730</td>\n",
              "      <td>01:40</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>6</td>\n",
              "      <td>1.147579</td>\n",
              "      <td>1.106042</td>\n",
              "      <td>0.761262</td>\n",
              "      <td>0.970476</td>\n",
              "      <td>01:40</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>7</td>\n",
              "      <td>1.100005</td>\n",
              "      <td>1.052264</td>\n",
              "      <td>0.776279</td>\n",
              "      <td>0.974294</td>\n",
              "      <td>01:40</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>8</td>\n",
              "      <td>1.110485</td>\n",
              "      <td>1.056296</td>\n",
              "      <td>0.767880</td>\n",
              "      <td>0.974294</td>\n",
              "      <td>01:40</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>9</td>\n",
              "      <td>1.039951</td>\n",
              "      <td>1.004825</td>\n",
              "      <td>0.796640</td>\n",
              "      <td>0.977348</td>\n",
              "      <td>01:40</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>10</td>\n",
              "      <td>0.992769</td>\n",
              "      <td>0.993364</td>\n",
              "      <td>0.797149</td>\n",
              "      <td>0.976330</td>\n",
              "      <td>01:41</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>11</td>\n",
              "      <td>0.971454</td>\n",
              "      <td>0.978841</td>\n",
              "      <td>0.813439</td>\n",
              "      <td>0.976330</td>\n",
              "      <td>01:40</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>12</td>\n",
              "      <td>0.964365</td>\n",
              "      <td>0.992485</td>\n",
              "      <td>0.805548</td>\n",
              "      <td>0.978875</td>\n",
              "      <td>01:41</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>13</td>\n",
              "      <td>0.922823</td>\n",
              "      <td>0.978788</td>\n",
              "      <td>0.808603</td>\n",
              "      <td>0.974294</td>\n",
              "      <td>01:41</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>14</td>\n",
              "      <td>0.902061</td>\n",
              "      <td>0.963746</td>\n",
              "      <td>0.816238</td>\n",
              "      <td>0.976330</td>\n",
              "      <td>01:41</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>15</td>\n",
              "      <td>0.849721</td>\n",
              "      <td>0.942621</td>\n",
              "      <td>0.817002</td>\n",
              "      <td>0.980148</td>\n",
              "      <td>01:40</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>16</td>\n",
              "      <td>0.763226</td>\n",
              "      <td>0.893584</td>\n",
              "      <td>0.838636</td>\n",
              "      <td>0.983965</td>\n",
              "      <td>01:40</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>17</td>\n",
              "      <td>0.717704</td>\n",
              "      <td>0.872851</td>\n",
              "      <td>0.849326</td>\n",
              "      <td>0.982693</td>\n",
              "      <td>01:41</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>18</td>\n",
              "      <td>0.671016</td>\n",
              "      <td>0.856330</td>\n",
              "      <td>0.860270</td>\n",
              "      <td>0.984729</td>\n",
              "      <td>01:41</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>19</td>\n",
              "      <td>0.654234</td>\n",
              "      <td>0.857323</td>\n",
              "      <td>0.857725</td>\n",
              "      <td>0.982947</td>\n",
              "      <td>01:41</td>\n",
              "    </tr>\n",
              "  </tbody>\n",
              "</table>"
            ],
            "text/plain": [
              "<IPython.core.display.HTML object>"
            ]
          },
          "metadata": {
            "tags": []
          }
        },
        {
          "output_type": "stream",
          "text": [
            ""
          ],
          "name": "stderr"
        },
        {
          "output_type": "stream",
          "text": [
            "{5: 0.7622295618057251, 20: 0.8577246069908142}\n",
            "data path   /root/.fastai/data/imagewoof2\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "stream",
          "text": [
            ""
          ],
          "name": "stderr"
        },
        {
          "output_type": "stream",
          "text": [
            "Learn path /root/.fastai/data/imagewoof2\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "display_data",
          "data": {
            "text/html": [
              "<table border=\"1\" class=\"dataframe\">\n",
              "  <thead>\n",
              "    <tr style=\"text-align: left;\">\n",
              "      <th>epoch</th>\n",
              "      <th>train_loss</th>\n",
              "      <th>valid_loss</th>\n",
              "      <th>accuracy</th>\n",
              "      <th>top_k_accuracy</th>\n",
              "      <th>time</th>\n",
              "    </tr>\n",
              "  </thead>\n",
              "  <tbody>\n",
              "    <tr>\n",
              "      <td>0</td>\n",
              "      <td>1.982851</td>\n",
              "      <td>1.782623</td>\n",
              "      <td>0.429880</td>\n",
              "      <td>0.879868</td>\n",
              "      <td>01:41</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>1</td>\n",
              "      <td>1.736484</td>\n",
              "      <td>1.574112</td>\n",
              "      <td>0.530160</td>\n",
              "      <td>0.920336</td>\n",
              "      <td>01:40</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>2</td>\n",
              "      <td>1.500458</td>\n",
              "      <td>1.398275</td>\n",
              "      <td>0.621023</td>\n",
              "      <td>0.943243</td>\n",
              "      <td>01:40</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>3</td>\n",
              "      <td>1.376719</td>\n",
              "      <td>1.256534</td>\n",
              "      <td>0.688216</td>\n",
              "      <td>0.959023</td>\n",
              "      <td>01:41</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>4</td>\n",
              "      <td>1.305366</td>\n",
              "      <td>1.252846</td>\n",
              "      <td>0.664546</td>\n",
              "      <td>0.960804</td>\n",
              "      <td>01:41</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>5</td>\n",
              "      <td>1.210624</td>\n",
              "      <td>1.164595</td>\n",
              "      <td>0.734284</td>\n",
              "      <td>0.966913</td>\n",
              "      <td>01:41</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>6</td>\n",
              "      <td>1.160298</td>\n",
              "      <td>1.098135</td>\n",
              "      <td>0.753118</td>\n",
              "      <td>0.972258</td>\n",
              "      <td>01:41</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>7</td>\n",
              "      <td>1.088429</td>\n",
              "      <td>1.065536</td>\n",
              "      <td>0.773479</td>\n",
              "      <td>0.968694</td>\n",
              "      <td>01:40</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>8</td>\n",
              "      <td>1.082545</td>\n",
              "      <td>1.047071</td>\n",
              "      <td>0.784933</td>\n",
              "      <td>0.977857</td>\n",
              "      <td>01:40</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>9</td>\n",
              "      <td>1.037036</td>\n",
              "      <td>1.053320</td>\n",
              "      <td>0.772716</td>\n",
              "      <td>0.969967</td>\n",
              "      <td>01:41</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>10</td>\n",
              "      <td>0.995949</td>\n",
              "      <td>0.987071</td>\n",
              "      <td>0.808094</td>\n",
              "      <td>0.981166</td>\n",
              "      <td>01:41</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>11</td>\n",
              "      <td>0.966743</td>\n",
              "      <td>0.971431</td>\n",
              "      <td>0.808603</td>\n",
              "      <td>0.976839</td>\n",
              "      <td>01:41</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>12</td>\n",
              "      <td>0.927262</td>\n",
              "      <td>1.006577</td>\n",
              "      <td>0.789768</td>\n",
              "      <td>0.978366</td>\n",
              "      <td>01:42</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>13</td>\n",
              "      <td>0.907087</td>\n",
              "      <td>0.970964</td>\n",
              "      <td>0.805294</td>\n",
              "      <td>0.979384</td>\n",
              "      <td>01:42</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>14</td>\n",
              "      <td>0.887708</td>\n",
              "      <td>0.957283</td>\n",
              "      <td>0.818274</td>\n",
              "      <td>0.977602</td>\n",
              "      <td>01:42</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>15</td>\n",
              "      <td>0.867872</td>\n",
              "      <td>0.933336</td>\n",
              "      <td>0.826928</td>\n",
              "      <td>0.978621</td>\n",
              "      <td>01:41</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>16</td>\n",
              "      <td>0.759610</td>\n",
              "      <td>0.902490</td>\n",
              "      <td>0.841435</td>\n",
              "      <td>0.980657</td>\n",
              "      <td>01:41</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>17</td>\n",
              "      <td>0.705429</td>\n",
              "      <td>0.874858</td>\n",
              "      <td>0.853652</td>\n",
              "      <td>0.980657</td>\n",
              "      <td>01:40</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>18</td>\n",
              "      <td>0.649135</td>\n",
              "      <td>0.857464</td>\n",
              "      <td>0.861797</td>\n",
              "      <td>0.982947</td>\n",
              "      <td>01:40</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>19</td>\n",
              "      <td>0.629916</td>\n",
              "      <td>0.848704</td>\n",
              "      <td>0.863069</td>\n",
              "      <td>0.983456</td>\n",
              "      <td>01:40</td>\n",
              "    </tr>\n",
              "  </tbody>\n",
              "</table>"
            ],
            "text/plain": [
              "<IPython.core.display.HTML object>"
            ]
          },
          "metadata": {
            "tags": []
          }
        },
        {
          "output_type": "stream",
          "text": [
            "{5: 0.7622295618057251, 20: 0.8603970408439636}\n",
            "data path   /root/.fastai/data/imagewoof2\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "stream",
          "text": [
            ""
          ],
          "name": "stderr"
        },
        {
          "output_type": "stream",
          "text": [
            "Learn path /root/.fastai/data/imagewoof2\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "display_data",
          "data": {
            "text/html": [
              "<table border=\"1\" class=\"dataframe\">\n",
              "  <thead>\n",
              "    <tr style=\"text-align: left;\">\n",
              "      <th>epoch</th>\n",
              "      <th>train_loss</th>\n",
              "      <th>valid_loss</th>\n",
              "      <th>accuracy</th>\n",
              "      <th>top_k_accuracy</th>\n",
              "      <th>time</th>\n",
              "    </tr>\n",
              "  </thead>\n",
              "  <tbody>\n",
              "    <tr>\n",
              "      <td>0</td>\n",
              "      <td>1.899340</td>\n",
              "      <td>1.747509</td>\n",
              "      <td>0.446424</td>\n",
              "      <td>0.894121</td>\n",
              "      <td>01:39</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>1</td>\n",
              "      <td>1.655891</td>\n",
              "      <td>1.511086</td>\n",
              "      <td>0.547468</td>\n",
              "      <td>0.928481</td>\n",
              "      <td>01:40</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>2</td>\n",
              "      <td>1.478268</td>\n",
              "      <td>1.333415</td>\n",
              "      <td>0.646220</td>\n",
              "      <td>0.953678</td>\n",
              "      <td>01:41</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>3</td>\n",
              "      <td>1.361890</td>\n",
              "      <td>1.247935</td>\n",
              "      <td>0.695597</td>\n",
              "      <td>0.956732</td>\n",
              "      <td>01:40</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>4</td>\n",
              "      <td>1.263519</td>\n",
              "      <td>1.263867</td>\n",
              "      <td>0.684653</td>\n",
              "      <td>0.961313</td>\n",
              "      <td>01:41</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>5</td>\n",
              "      <td>1.206411</td>\n",
              "      <td>1.142138</td>\n",
              "      <td>0.738865</td>\n",
              "      <td>0.968440</td>\n",
              "      <td>01:41</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>6</td>\n",
              "      <td>1.126316</td>\n",
              "      <td>1.082888</td>\n",
              "      <td>0.765080</td>\n",
              "      <td>0.972258</td>\n",
              "      <td>01:41</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>7</td>\n",
              "      <td>1.081094</td>\n",
              "      <td>1.063285</td>\n",
              "      <td>0.776024</td>\n",
              "      <td>0.969712</td>\n",
              "      <td>01:41</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>8</td>\n",
              "      <td>1.053136</td>\n",
              "      <td>1.060423</td>\n",
              "      <td>0.772716</td>\n",
              "      <td>0.972258</td>\n",
              "      <td>01:41</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>9</td>\n",
              "      <td>0.988018</td>\n",
              "      <td>1.037949</td>\n",
              "      <td>0.777806</td>\n",
              "      <td>0.973530</td>\n",
              "      <td>01:41</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>10</td>\n",
              "      <td>0.982155</td>\n",
              "      <td>1.000073</td>\n",
              "      <td>0.797149</td>\n",
              "      <td>0.973785</td>\n",
              "      <td>01:41</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>11</td>\n",
              "      <td>0.964968</td>\n",
              "      <td>1.002946</td>\n",
              "      <td>0.801222</td>\n",
              "      <td>0.977093</td>\n",
              "      <td>01:41</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>12</td>\n",
              "      <td>0.939658</td>\n",
              "      <td>0.971005</td>\n",
              "      <td>0.810384</td>\n",
              "      <td>0.975821</td>\n",
              "      <td>01:40</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>13</td>\n",
              "      <td>0.952305</td>\n",
              "      <td>0.943103</td>\n",
              "      <td>0.824128</td>\n",
              "      <td>0.978621</td>\n",
              "      <td>01:39</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>14</td>\n",
              "      <td>0.893221</td>\n",
              "      <td>0.969225</td>\n",
              "      <td>0.819801</td>\n",
              "      <td>0.976839</td>\n",
              "      <td>01:39</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>15</td>\n",
              "      <td>0.840974</td>\n",
              "      <td>0.925338</td>\n",
              "      <td>0.833036</td>\n",
              "      <td>0.982438</td>\n",
              "      <td>01:37</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>16</td>\n",
              "      <td>0.778530</td>\n",
              "      <td>0.905621</td>\n",
              "      <td>0.838636</td>\n",
              "      <td>0.979384</td>\n",
              "      <td>01:37</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>17</td>\n",
              "      <td>0.699867</td>\n",
              "      <td>0.883156</td>\n",
              "      <td>0.849835</td>\n",
              "      <td>0.983711</td>\n",
              "      <td>01:38</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>18</td>\n",
              "      <td>0.675236</td>\n",
              "      <td>0.853171</td>\n",
              "      <td>0.865106</td>\n",
              "      <td>0.981675</td>\n",
              "      <td>01:38</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>19</td>\n",
              "      <td>0.650516</td>\n",
              "      <td>0.849444</td>\n",
              "      <td>0.864342</td>\n",
              "      <td>0.983202</td>\n",
              "      <td>01:38</td>\n",
              "    </tr>\n",
              "  </tbody>\n",
              "</table>"
            ],
            "text/plain": [
              "<IPython.core.display.HTML object>"
            ]
          },
          "metadata": {
            "tags": []
          }
        },
        {
          "output_type": "stream",
          "text": [
            "{5: 0.7622295618057251, 20: 0.8617120583852133}\n",
            "data path   /root/.fastai/data/imagewoof2\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "stream",
          "text": [
            ""
          ],
          "name": "stderr"
        },
        {
          "output_type": "stream",
          "text": [
            "Learn path /root/.fastai/data/imagewoof2\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "display_data",
          "data": {
            "text/html": [
              "<table border=\"1\" class=\"dataframe\">\n",
              "  <thead>\n",
              "    <tr style=\"text-align: left;\">\n",
              "      <th>epoch</th>\n",
              "      <th>train_loss</th>\n",
              "      <th>valid_loss</th>\n",
              "      <th>accuracy</th>\n",
              "      <th>top_k_accuracy</th>\n",
              "      <th>time</th>\n",
              "    </tr>\n",
              "  </thead>\n",
              "  <tbody>\n",
              "    <tr>\n",
              "      <td>0</td>\n",
              "      <td>2.004710</td>\n",
              "      <td>1.865880</td>\n",
              "      <td>0.382031</td>\n",
              "      <td>0.854670</td>\n",
              "      <td>01:38</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>1</td>\n",
              "      <td>1.664988</td>\n",
              "      <td>1.600449</td>\n",
              "      <td>0.520998</td>\n",
              "      <td>0.918809</td>\n",
              "      <td>01:37</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>2</td>\n",
              "      <td>1.560314</td>\n",
              "      <td>1.360030</td>\n",
              "      <td>0.635531</td>\n",
              "      <td>0.952405</td>\n",
              "      <td>01:38</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>3</td>\n",
              "      <td>1.389169</td>\n",
              "      <td>1.246493</td>\n",
              "      <td>0.691779</td>\n",
              "      <td>0.955968</td>\n",
              "      <td>01:38</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>4</td>\n",
              "      <td>1.313231</td>\n",
              "      <td>1.228393</td>\n",
              "      <td>0.694579</td>\n",
              "      <td>0.963095</td>\n",
              "      <td>01:38</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>5</td>\n",
              "      <td>1.234166</td>\n",
              "      <td>1.143967</td>\n",
              "      <td>0.743446</td>\n",
              "      <td>0.966658</td>\n",
              "      <td>01:38</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>6</td>\n",
              "      <td>1.148173</td>\n",
              "      <td>1.089888</td>\n",
              "      <td>0.767116</td>\n",
              "      <td>0.965131</td>\n",
              "      <td>01:38</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>7</td>\n",
              "      <td>1.112358</td>\n",
              "      <td>1.075314</td>\n",
              "      <td>0.772716</td>\n",
              "      <td>0.969967</td>\n",
              "      <td>01:38</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>8</td>\n",
              "      <td>1.048016</td>\n",
              "      <td>1.031385</td>\n",
              "      <td>0.789768</td>\n",
              "      <td>0.974803</td>\n",
              "      <td>01:38</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>9</td>\n",
              "      <td>1.070498</td>\n",
              "      <td>0.999182</td>\n",
              "      <td>0.795622</td>\n",
              "      <td>0.976330</td>\n",
              "      <td>01:38</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>10</td>\n",
              "      <td>0.985275</td>\n",
              "      <td>1.007039</td>\n",
              "      <td>0.799186</td>\n",
              "      <td>0.971749</td>\n",
              "      <td>01:38</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>11</td>\n",
              "      <td>1.003443</td>\n",
              "      <td>0.987707</td>\n",
              "      <td>0.803003</td>\n",
              "      <td>0.974548</td>\n",
              "      <td>01:37</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>12</td>\n",
              "      <td>0.941485</td>\n",
              "      <td>0.988609</td>\n",
              "      <td>0.808094</td>\n",
              "      <td>0.976075</td>\n",
              "      <td>01:37</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>13</td>\n",
              "      <td>0.920600</td>\n",
              "      <td>0.967870</td>\n",
              "      <td>0.816493</td>\n",
              "      <td>0.977093</td>\n",
              "      <td>01:38</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>14</td>\n",
              "      <td>0.890949</td>\n",
              "      <td>0.964844</td>\n",
              "      <td>0.805039</td>\n",
              "      <td>0.975821</td>\n",
              "      <td>01:38</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>15</td>\n",
              "      <td>0.847337</td>\n",
              "      <td>0.916752</td>\n",
              "      <td>0.832527</td>\n",
              "      <td>0.978621</td>\n",
              "      <td>01:38</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>16</td>\n",
              "      <td>0.762193</td>\n",
              "      <td>0.897338</td>\n",
              "      <td>0.846017</td>\n",
              "      <td>0.981420</td>\n",
              "      <td>01:38</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>17</td>\n",
              "      <td>0.727142</td>\n",
              "      <td>0.871710</td>\n",
              "      <td>0.852634</td>\n",
              "      <td>0.980148</td>\n",
              "      <td>01:38</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>18</td>\n",
              "      <td>0.665663</td>\n",
              "      <td>0.863523</td>\n",
              "      <td>0.860270</td>\n",
              "      <td>0.983202</td>\n",
              "      <td>01:38</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>19</td>\n",
              "      <td>0.644735</td>\n",
              "      <td>0.860684</td>\n",
              "      <td>0.859252</td>\n",
              "      <td>0.981929</td>\n",
              "      <td>01:38</td>\n",
              "    </tr>\n",
              "  </tbody>\n",
              "</table>"
            ],
            "text/plain": [
              "<IPython.core.display.HTML object>"
            ]
          },
          "metadata": {
            "tags": []
          }
        },
        {
          "output_type": "stream",
          "text": [
            "{5: 0.7622295618057251, 20: 0.861096978187561}\n",
            "data path   /root/.fastai/data/imagewoof2\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "stream",
          "text": [
            ""
          ],
          "name": "stderr"
        },
        {
          "output_type": "stream",
          "text": [
            "Learn path /root/.fastai/data/imagewoof2\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "display_data",
          "data": {
            "text/html": [
              "<table border=\"1\" class=\"dataframe\">\n",
              "  <thead>\n",
              "    <tr style=\"text-align: left;\">\n",
              "      <th>epoch</th>\n",
              "      <th>train_loss</th>\n",
              "      <th>valid_loss</th>\n",
              "      <th>accuracy</th>\n",
              "      <th>top_k_accuracy</th>\n",
              "      <th>time</th>\n",
              "    </tr>\n",
              "  </thead>\n",
              "  <tbody>\n",
              "    <tr>\n",
              "      <td>0</td>\n",
              "      <td>1.951077</td>\n",
              "      <td>1.792640</td>\n",
              "      <td>0.418682</td>\n",
              "      <td>0.872232</td>\n",
              "      <td>01:38</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>1</td>\n",
              "      <td>1.657203</td>\n",
              "      <td>1.508731</td>\n",
              "      <td>0.568083</td>\n",
              "      <td>0.932553</td>\n",
              "      <td>01:38</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>2</td>\n",
              "      <td>1.496707</td>\n",
              "      <td>1.367347</td>\n",
              "      <td>0.630949</td>\n",
              "      <td>0.943497</td>\n",
              "      <td>01:37</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>3</td>\n",
              "      <td>1.343047</td>\n",
              "      <td>1.288623</td>\n",
              "      <td>0.674472</td>\n",
              "      <td>0.951133</td>\n",
              "      <td>01:38</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>4</td>\n",
              "      <td>1.219010</td>\n",
              "      <td>1.204292</td>\n",
              "      <td>0.723085</td>\n",
              "      <td>0.963349</td>\n",
              "      <td>01:38</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>5</td>\n",
              "      <td>1.200550</td>\n",
              "      <td>1.127590</td>\n",
              "      <td>0.745228</td>\n",
              "      <td>0.972767</td>\n",
              "      <td>01:38</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>6</td>\n",
              "      <td>1.136331</td>\n",
              "      <td>1.137869</td>\n",
              "      <td>0.750318</td>\n",
              "      <td>0.965386</td>\n",
              "      <td>01:38</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>7</td>\n",
              "      <td>1.114759</td>\n",
              "      <td>1.074422</td>\n",
              "      <td>0.768134</td>\n",
              "      <td>0.972258</td>\n",
              "      <td>01:38</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>8</td>\n",
              "      <td>1.065833</td>\n",
              "      <td>1.027165</td>\n",
              "      <td>0.790277</td>\n",
              "      <td>0.973530</td>\n",
              "      <td>01:38</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>9</td>\n",
              "      <td>1.009803</td>\n",
              "      <td>0.997346</td>\n",
              "      <td>0.799949</td>\n",
              "      <td>0.977348</td>\n",
              "      <td>01:38</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>10</td>\n",
              "      <td>0.970929</td>\n",
              "      <td>0.979596</td>\n",
              "      <td>0.813693</td>\n",
              "      <td>0.977348</td>\n",
              "      <td>01:38</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>11</td>\n",
              "      <td>0.954698</td>\n",
              "      <td>0.985558</td>\n",
              "      <td>0.799695</td>\n",
              "      <td>0.975057</td>\n",
              "      <td>01:37</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>12</td>\n",
              "      <td>0.928607</td>\n",
              "      <td>0.999897</td>\n",
              "      <td>0.793077</td>\n",
              "      <td>0.975312</td>\n",
              "      <td>01:38</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>13</td>\n",
              "      <td>0.898836</td>\n",
              "      <td>0.957941</td>\n",
              "      <td>0.821329</td>\n",
              "      <td>0.976075</td>\n",
              "      <td>01:38</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>14</td>\n",
              "      <td>0.878161</td>\n",
              "      <td>0.977201</td>\n",
              "      <td>0.810130</td>\n",
              "      <td>0.975057</td>\n",
              "      <td>01:38</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>15</td>\n",
              "      <td>0.847006</td>\n",
              "      <td>0.937120</td>\n",
              "      <td>0.825655</td>\n",
              "      <td>0.978112</td>\n",
              "      <td>01:38</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>16</td>\n",
              "      <td>0.790348</td>\n",
              "      <td>0.909870</td>\n",
              "      <td>0.833291</td>\n",
              "      <td>0.983456</td>\n",
              "      <td>01:38</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>17</td>\n",
              "      <td>0.691107</td>\n",
              "      <td>0.866855</td>\n",
              "      <td>0.853652</td>\n",
              "      <td>0.983456</td>\n",
              "      <td>01:38</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>18</td>\n",
              "      <td>0.653225</td>\n",
              "      <td>0.842658</td>\n",
              "      <td>0.866633</td>\n",
              "      <td>0.984220</td>\n",
              "      <td>01:38</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>19</td>\n",
              "      <td>0.637666</td>\n",
              "      <td>0.836478</td>\n",
              "      <td>0.866887</td>\n",
              "      <td>0.984983</td>\n",
              "      <td>01:38</td>\n",
              "    </tr>\n",
              "  </tbody>\n",
              "</table>"
            ],
            "text/plain": [
              "<IPython.core.display.HTML object>"
            ]
          },
          "metadata": {
            "tags": []
          }
        },
        {
          "output_type": "stream",
          "text": [
            "{5: 0.7622295618057251, 20: 0.8622550368309021}\n",
            "data path   /root/.fastai/data/imagewoof2\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "stream",
          "text": [
            ""
          ],
          "name": "stderr"
        },
        {
          "output_type": "stream",
          "text": [
            "Learn path /root/.fastai/data/imagewoof2\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "display_data",
          "data": {
            "text/html": [
              "<table border=\"1\" class=\"dataframe\">\n",
              "  <thead>\n",
              "    <tr style=\"text-align: left;\">\n",
              "      <th>epoch</th>\n",
              "      <th>train_loss</th>\n",
              "      <th>valid_loss</th>\n",
              "      <th>accuracy</th>\n",
              "      <th>top_k_accuracy</th>\n",
              "      <th>time</th>\n",
              "    </tr>\n",
              "  </thead>\n",
              "  <tbody>\n",
              "    <tr>\n",
              "      <td>0</td>\n",
              "      <td>2.073898</td>\n",
              "      <td>1.859646</td>\n",
              "      <td>0.375159</td>\n",
              "      <td>0.856961</td>\n",
              "      <td>01:38</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>1</td>\n",
              "      <td>1.836540</td>\n",
              "      <td>1.617754</td>\n",
              "      <td>0.499618</td>\n",
              "      <td>0.918045</td>\n",
              "      <td>01:38</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>2</td>\n",
              "      <td>1.767781</td>\n",
              "      <td>1.572328</td>\n",
              "      <td>0.526597</td>\n",
              "      <td>0.915500</td>\n",
              "      <td>01:38</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>3</td>\n",
              "      <td>1.604483</td>\n",
              "      <td>1.273613</td>\n",
              "      <td>0.677781</td>\n",
              "      <td>0.952660</td>\n",
              "      <td>01:38</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>4</td>\n",
              "      <td>1.524711</td>\n",
              "      <td>1.220221</td>\n",
              "      <td>0.709341</td>\n",
              "      <td>0.957241</td>\n",
              "      <td>01:38</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>5</td>\n",
              "      <td>1.479169</td>\n",
              "      <td>1.163010</td>\n",
              "      <td>0.730975</td>\n",
              "      <td>0.965386</td>\n",
              "      <td>01:38</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>6</td>\n",
              "      <td>1.407726</td>\n",
              "      <td>1.153900</td>\n",
              "      <td>0.737592</td>\n",
              "      <td>0.964877</td>\n",
              "      <td>01:38</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>7</td>\n",
              "      <td>1.367976</td>\n",
              "      <td>1.072799</td>\n",
              "      <td>0.762026</td>\n",
              "      <td>0.973530</td>\n",
              "      <td>01:38</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>8</td>\n",
              "      <td>1.324972</td>\n",
              "      <td>1.059646</td>\n",
              "      <td>0.769152</td>\n",
              "      <td>0.974803</td>\n",
              "      <td>01:37</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>9</td>\n",
              "      <td>1.295897</td>\n",
              "      <td>1.040017</td>\n",
              "      <td>0.778061</td>\n",
              "      <td>0.975057</td>\n",
              "      <td>01:37</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>10</td>\n",
              "      <td>1.274680</td>\n",
              "      <td>1.000805</td>\n",
              "      <td>0.792059</td>\n",
              "      <td>0.978875</td>\n",
              "      <td>01:37</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>11</td>\n",
              "      <td>1.250890</td>\n",
              "      <td>0.991731</td>\n",
              "      <td>0.801731</td>\n",
              "      <td>0.976330</td>\n",
              "      <td>01:38</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>12</td>\n",
              "      <td>1.229063</td>\n",
              "      <td>0.995858</td>\n",
              "      <td>0.804276</td>\n",
              "      <td>0.973530</td>\n",
              "      <td>01:38</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>13</td>\n",
              "      <td>1.212375</td>\n",
              "      <td>0.967520</td>\n",
              "      <td>0.806058</td>\n",
              "      <td>0.978366</td>\n",
              "      <td>01:37</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>14</td>\n",
              "      <td>1.203226</td>\n",
              "      <td>0.954525</td>\n",
              "      <td>0.822092</td>\n",
              "      <td>0.975057</td>\n",
              "      <td>01:37</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>15</td>\n",
              "      <td>1.143074</td>\n",
              "      <td>0.968747</td>\n",
              "      <td>0.813184</td>\n",
              "      <td>0.976330</td>\n",
              "      <td>01:38</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>16</td>\n",
              "      <td>1.149305</td>\n",
              "      <td>0.966957</td>\n",
              "      <td>0.812420</td>\n",
              "      <td>0.979384</td>\n",
              "      <td>01:38</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>17</td>\n",
              "      <td>1.135575</td>\n",
              "      <td>0.945458</td>\n",
              "      <td>0.815220</td>\n",
              "      <td>0.977093</td>\n",
              "      <td>01:38</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>18</td>\n",
              "      <td>1.094568</td>\n",
              "      <td>0.932182</td>\n",
              "      <td>0.824637</td>\n",
              "      <td>0.982438</td>\n",
              "      <td>01:38</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>19</td>\n",
              "      <td>1.107305</td>\n",
              "      <td>0.941671</td>\n",
              "      <td>0.823110</td>\n",
              "      <td>0.976330</td>\n",
              "      <td>01:38</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>20</td>\n",
              "      <td>1.071464</td>\n",
              "      <td>0.914105</td>\n",
              "      <td>0.833036</td>\n",
              "      <td>0.975057</td>\n",
              "      <td>01:38</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>21</td>\n",
              "      <td>1.056442</td>\n",
              "      <td>0.912072</td>\n",
              "      <td>0.833545</td>\n",
              "      <td>0.976584</td>\n",
              "      <td>01:40</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>22</td>\n",
              "      <td>1.067354</td>\n",
              "      <td>0.939659</td>\n",
              "      <td>0.825655</td>\n",
              "      <td>0.977093</td>\n",
              "      <td>01:39</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>23</td>\n",
              "      <td>1.032612</td>\n",
              "      <td>0.956944</td>\n",
              "      <td>0.815475</td>\n",
              "      <td>0.979639</td>\n",
              "      <td>01:39</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>24</td>\n",
              "      <td>1.020848</td>\n",
              "      <td>0.915896</td>\n",
              "      <td>0.827946</td>\n",
              "      <td>0.977602</td>\n",
              "      <td>01:40</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>25</td>\n",
              "      <td>0.997053</td>\n",
              "      <td>0.896440</td>\n",
              "      <td>0.838127</td>\n",
              "      <td>0.981929</td>\n",
              "      <td>01:40</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>26</td>\n",
              "      <td>1.048893</td>\n",
              "      <td>0.908370</td>\n",
              "      <td>0.833545</td>\n",
              "      <td>0.980148</td>\n",
              "      <td>01:40</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>27</td>\n",
              "      <td>1.017267</td>\n",
              "      <td>0.904403</td>\n",
              "      <td>0.835327</td>\n",
              "      <td>0.982438</td>\n",
              "      <td>01:40</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>28</td>\n",
              "      <td>1.000311</td>\n",
              "      <td>0.903372</td>\n",
              "      <td>0.838636</td>\n",
              "      <td>0.977348</td>\n",
              "      <td>01:40</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>29</td>\n",
              "      <td>1.001543</td>\n",
              "      <td>0.886917</td>\n",
              "      <td>0.845253</td>\n",
              "      <td>0.981166</td>\n",
              "      <td>01:41</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>30</td>\n",
              "      <td>0.997795</td>\n",
              "      <td>0.915211</td>\n",
              "      <td>0.834563</td>\n",
              "      <td>0.979639</td>\n",
              "      <td>01:41</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>31</td>\n",
              "      <td>0.998718</td>\n",
              "      <td>0.896250</td>\n",
              "      <td>0.842199</td>\n",
              "      <td>0.978875</td>\n",
              "      <td>01:40</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>32</td>\n",
              "      <td>0.988814</td>\n",
              "      <td>0.913438</td>\n",
              "      <td>0.836345</td>\n",
              "      <td>0.981929</td>\n",
              "      <td>01:38</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>33</td>\n",
              "      <td>0.961588</td>\n",
              "      <td>0.922347</td>\n",
              "      <td>0.835582</td>\n",
              "      <td>0.977602</td>\n",
              "      <td>01:39</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>34</td>\n",
              "      <td>0.986900</td>\n",
              "      <td>0.913219</td>\n",
              "      <td>0.841435</td>\n",
              "      <td>0.976075</td>\n",
              "      <td>01:39</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>35</td>\n",
              "      <td>0.959678</td>\n",
              "      <td>0.914370</td>\n",
              "      <td>0.835836</td>\n",
              "      <td>0.978875</td>\n",
              "      <td>01:39</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>36</td>\n",
              "      <td>0.995852</td>\n",
              "      <td>0.875487</td>\n",
              "      <td>0.848817</td>\n",
              "      <td>0.978875</td>\n",
              "      <td>01:39</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>37</td>\n",
              "      <td>0.973012</td>\n",
              "      <td>0.908897</td>\n",
              "      <td>0.841181</td>\n",
              "      <td>0.980911</td>\n",
              "      <td>01:39</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>38</td>\n",
              "      <td>0.957920</td>\n",
              "      <td>0.923039</td>\n",
              "      <td>0.833545</td>\n",
              "      <td>0.976584</td>\n",
              "      <td>01:38</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>39</td>\n",
              "      <td>0.971181</td>\n",
              "      <td>0.893422</td>\n",
              "      <td>0.850344</td>\n",
              "      <td>0.977093</td>\n",
              "      <td>01:39</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>40</td>\n",
              "      <td>0.929811</td>\n",
              "      <td>0.910508</td>\n",
              "      <td>0.841690</td>\n",
              "      <td>0.974803</td>\n",
              "      <td>01:39</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>41</td>\n",
              "      <td>0.942168</td>\n",
              "      <td>0.953156</td>\n",
              "      <td>0.826928</td>\n",
              "      <td>0.973530</td>\n",
              "      <td>01:38</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>42</td>\n",
              "      <td>0.962230</td>\n",
              "      <td>0.926084</td>\n",
              "      <td>0.839145</td>\n",
              "      <td>0.974294</td>\n",
              "      <td>01:39</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>43</td>\n",
              "      <td>0.902150</td>\n",
              "      <td>0.898941</td>\n",
              "      <td>0.843472</td>\n",
              "      <td>0.979639</td>\n",
              "      <td>01:39</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>44</td>\n",
              "      <td>0.930805</td>\n",
              "      <td>0.910444</td>\n",
              "      <td>0.838890</td>\n",
              "      <td>0.978621</td>\n",
              "      <td>01:39</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>45</td>\n",
              "      <td>0.950929</td>\n",
              "      <td>0.910201</td>\n",
              "      <td>0.837618</td>\n",
              "      <td>0.977093</td>\n",
              "      <td>01:39</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>46</td>\n",
              "      <td>0.928907</td>\n",
              "      <td>0.911954</td>\n",
              "      <td>0.838890</td>\n",
              "      <td>0.974803</td>\n",
              "      <td>01:39</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>47</td>\n",
              "      <td>0.919404</td>\n",
              "      <td>0.899089</td>\n",
              "      <td>0.842199</td>\n",
              "      <td>0.978621</td>\n",
              "      <td>01:40</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>48</td>\n",
              "      <td>0.921529</td>\n",
              "      <td>0.917558</td>\n",
              "      <td>0.841690</td>\n",
              "      <td>0.976330</td>\n",
              "      <td>01:39</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>49</td>\n",
              "      <td>0.919114</td>\n",
              "      <td>0.887527</td>\n",
              "      <td>0.844744</td>\n",
              "      <td>0.977348</td>\n",
              "      <td>01:40</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>50</td>\n",
              "      <td>0.926167</td>\n",
              "      <td>0.891678</td>\n",
              "      <td>0.848562</td>\n",
              "      <td>0.977093</td>\n",
              "      <td>01:41</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>51</td>\n",
              "      <td>0.919716</td>\n",
              "      <td>0.896375</td>\n",
              "      <td>0.840163</td>\n",
              "      <td>0.977093</td>\n",
              "      <td>01:41</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>52</td>\n",
              "      <td>0.921108</td>\n",
              "      <td>0.929091</td>\n",
              "      <td>0.832527</td>\n",
              "      <td>0.974294</td>\n",
              "      <td>01:41</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>53</td>\n",
              "      <td>0.921638</td>\n",
              "      <td>0.933508</td>\n",
              "      <td>0.828964</td>\n",
              "      <td>0.971749</td>\n",
              "      <td>01:41</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>54</td>\n",
              "      <td>0.927880</td>\n",
              "      <td>0.925651</td>\n",
              "      <td>0.835327</td>\n",
              "      <td>0.973785</td>\n",
              "      <td>01:41</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>55</td>\n",
              "      <td>0.911829</td>\n",
              "      <td>0.903788</td>\n",
              "      <td>0.841435</td>\n",
              "      <td>0.974294</td>\n",
              "      <td>01:41</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>56</td>\n",
              "      <td>0.890021</td>\n",
              "      <td>0.894743</td>\n",
              "      <td>0.849071</td>\n",
              "      <td>0.975566</td>\n",
              "      <td>01:41</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>57</td>\n",
              "      <td>0.898231</td>\n",
              "      <td>0.905612</td>\n",
              "      <td>0.840926</td>\n",
              "      <td>0.972258</td>\n",
              "      <td>01:41</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>58</td>\n",
              "      <td>0.909878</td>\n",
              "      <td>0.884186</td>\n",
              "      <td>0.853907</td>\n",
              "      <td>0.975057</td>\n",
              "      <td>01:40</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>59</td>\n",
              "      <td>0.897760</td>\n",
              "      <td>0.898740</td>\n",
              "      <td>0.841690</td>\n",
              "      <td>0.976330</td>\n",
              "      <td>01:40</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>60</td>\n",
              "      <td>0.872120</td>\n",
              "      <td>0.887360</td>\n",
              "      <td>0.849835</td>\n",
              "      <td>0.978621</td>\n",
              "      <td>01:41</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>61</td>\n",
              "      <td>0.922209</td>\n",
              "      <td>0.895622</td>\n",
              "      <td>0.843726</td>\n",
              "      <td>0.975821</td>\n",
              "      <td>01:40</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>62</td>\n",
              "      <td>0.862677</td>\n",
              "      <td>0.887566</td>\n",
              "      <td>0.849326</td>\n",
              "      <td>0.980148</td>\n",
              "      <td>01:41</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>63</td>\n",
              "      <td>0.890156</td>\n",
              "      <td>0.923463</td>\n",
              "      <td>0.836091</td>\n",
              "      <td>0.975821</td>\n",
              "      <td>01:41</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>64</td>\n",
              "      <td>0.861222</td>\n",
              "      <td>0.912705</td>\n",
              "      <td>0.847035</td>\n",
              "      <td>0.969458</td>\n",
              "      <td>01:40</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>65</td>\n",
              "      <td>0.865928</td>\n",
              "      <td>0.858317</td>\n",
              "      <td>0.861797</td>\n",
              "      <td>0.979384</td>\n",
              "      <td>01:41</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>66</td>\n",
              "      <td>0.839435</td>\n",
              "      <td>0.872230</td>\n",
              "      <td>0.853907</td>\n",
              "      <td>0.974039</td>\n",
              "      <td>01:40</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>67</td>\n",
              "      <td>0.853975</td>\n",
              "      <td>0.872447</td>\n",
              "      <td>0.859252</td>\n",
              "      <td>0.975821</td>\n",
              "      <td>01:41</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>68</td>\n",
              "      <td>0.826387</td>\n",
              "      <td>0.855351</td>\n",
              "      <td>0.863069</td>\n",
              "      <td>0.979893</td>\n",
              "      <td>01:40</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>69</td>\n",
              "      <td>0.842908</td>\n",
              "      <td>0.834596</td>\n",
              "      <td>0.870196</td>\n",
              "      <td>0.978112</td>\n",
              "      <td>01:41</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>70</td>\n",
              "      <td>0.829004</td>\n",
              "      <td>0.847467</td>\n",
              "      <td>0.868160</td>\n",
              "      <td>0.981420</td>\n",
              "      <td>01:41</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>71</td>\n",
              "      <td>0.812779</td>\n",
              "      <td>0.840102</td>\n",
              "      <td>0.869178</td>\n",
              "      <td>0.979384</td>\n",
              "      <td>01:41</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>72</td>\n",
              "      <td>0.825722</td>\n",
              "      <td>0.837370</td>\n",
              "      <td>0.869687</td>\n",
              "      <td>0.979639</td>\n",
              "      <td>01:40</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>73</td>\n",
              "      <td>0.817403</td>\n",
              "      <td>0.831389</td>\n",
              "      <td>0.874268</td>\n",
              "      <td>0.979639</td>\n",
              "      <td>01:41</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>74</td>\n",
              "      <td>0.793143</td>\n",
              "      <td>0.822574</td>\n",
              "      <td>0.876304</td>\n",
              "      <td>0.982693</td>\n",
              "      <td>01:40</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>75</td>\n",
              "      <td>0.797489</td>\n",
              "      <td>0.818026</td>\n",
              "      <td>0.881140</td>\n",
              "      <td>0.981166</td>\n",
              "      <td>01:41</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>76</td>\n",
              "      <td>0.792954</td>\n",
              "      <td>0.818877</td>\n",
              "      <td>0.880886</td>\n",
              "      <td>0.978112</td>\n",
              "      <td>01:42</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>77</td>\n",
              "      <td>0.816260</td>\n",
              "      <td>0.816706</td>\n",
              "      <td>0.877322</td>\n",
              "      <td>0.979639</td>\n",
              "      <td>01:41</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>78</td>\n",
              "      <td>0.813128</td>\n",
              "      <td>0.817194</td>\n",
              "      <td>0.879104</td>\n",
              "      <td>0.978875</td>\n",
              "      <td>01:41</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>79</td>\n",
              "      <td>0.811769</td>\n",
              "      <td>0.815275</td>\n",
              "      <td>0.878341</td>\n",
              "      <td>0.981166</td>\n",
              "      <td>01:41</td>\n",
              "    </tr>\n",
              "  </tbody>\n",
              "</table>"
            ],
            "text/plain": [
              "<IPython.core.display.HTML object>"
            ]
          },
          "metadata": {
            "tags": []
          }
        },
        {
          "output_type": "stream",
          "text": [
            "{5: 0.7622295618057251, 20: 0.8622550368309021, 80: 0.8783405423164368}\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "V9ddtcQ1HGbd",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 51
        },
        "outputId": "f91cc24e-0514-47fe-f197-8f1a18f06f6c"
      },
      "source": [
        "print('depth multiplier =',dm)\n",
        "print({e: sum(res[e])/len(res[e]) for e in res})"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "depth multiplier = 4\n",
            "{5: 0.7622295618057251, 20: 0.8622550368309021, 80: 0.8783405423164368}\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "L-LmYNdE0IV8",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 1000
        },
        "outputId": "91becbe1-c0b5-4241-c36f-6a2377f28abd"
      },
      "source": [
        "dm = 6\n",
        "res = dict()\n",
        "for e in [5]*5 + [20]*5 + [80]:\n",
        "  mixup=0 if e<=20 else 0.2\n",
        "  learn = get_learn(model=model, size=128, bs=16, mixup=mixup)\n",
        "  learn.fit_fc(e, lr=4e-3, moms=(0.95,0.95), start_pct=0.72)\n",
        "  acc = learn.recorder.metrics[-1][0].item()\n",
        "  if e in res:\n",
        "    res[e] += [acc]\n",
        "  else:\n",
        "    res[e] = [acc]\n",
        "  print('{} epochs: {} ({} runs)'.format(e, sum(res[e])/len(res[e]), len(res[e])))\n",
        "print('depth multiplier='.format(dm), {e: sum(res[e])/len(res[e])})"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "data path   /root/.fastai/data/imagewoof2\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "stream",
          "text": [
            ""
          ],
          "name": "stderr"
        },
        {
          "output_type": "stream",
          "text": [
            "Learn path /root/.fastai/data/imagewoof2\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "display_data",
          "data": {
            "text/html": [
              "<table border=\"1\" class=\"dataframe\">\n",
              "  <thead>\n",
              "    <tr style=\"text-align: left;\">\n",
              "      <th>epoch</th>\n",
              "      <th>train_loss</th>\n",
              "      <th>valid_loss</th>\n",
              "      <th>accuracy</th>\n",
              "      <th>top_k_accuracy</th>\n",
              "      <th>time</th>\n",
              "    </tr>\n",
              "  </thead>\n",
              "  <tbody>\n",
              "    <tr>\n",
              "      <td>0</td>\n",
              "      <td>1.924982</td>\n",
              "      <td>1.804799</td>\n",
              "      <td>0.412828</td>\n",
              "      <td>0.878595</td>\n",
              "      <td>01:47</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>1</td>\n",
              "      <td>1.682397</td>\n",
              "      <td>1.570771</td>\n",
              "      <td>0.531942</td>\n",
              "      <td>0.916773</td>\n",
              "      <td>01:47</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>2</td>\n",
              "      <td>1.513665</td>\n",
              "      <td>1.388963</td>\n",
              "      <td>0.625859</td>\n",
              "      <td>0.948842</td>\n",
              "      <td>01:46</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>3</td>\n",
              "      <td>1.349511</td>\n",
              "      <td>1.233298</td>\n",
              "      <td>0.690761</td>\n",
              "      <td>0.959786</td>\n",
              "      <td>01:46</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>4</td>\n",
              "      <td>1.168263</td>\n",
              "      <td>1.095365</td>\n",
              "      <td>0.760244</td>\n",
              "      <td>0.970985</td>\n",
              "      <td>01:47</td>\n",
              "    </tr>\n",
              "  </tbody>\n",
              "</table>"
            ],
            "text/plain": [
              "<IPython.core.display.HTML object>"
            ]
          },
          "metadata": {
            "tags": []
          }
        },
        {
          "output_type": "stream",
          "text": [
            "5 epochs: 0.7602443099021912 (1 runs)\n",
            "data path   /root/.fastai/data/imagewoof2\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "stream",
          "text": [
            ""
          ],
          "name": "stderr"
        },
        {
          "output_type": "stream",
          "text": [
            "Learn path /root/.fastai/data/imagewoof2\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "display_data",
          "data": {
            "text/html": [
              "<table border=\"1\" class=\"dataframe\">\n",
              "  <thead>\n",
              "    <tr style=\"text-align: left;\">\n",
              "      <th>epoch</th>\n",
              "      <th>train_loss</th>\n",
              "      <th>valid_loss</th>\n",
              "      <th>accuracy</th>\n",
              "      <th>top_k_accuracy</th>\n",
              "      <th>time</th>\n",
              "    </tr>\n",
              "  </thead>\n",
              "  <tbody>\n",
              "    <tr>\n",
              "      <td>0</td>\n",
              "      <td>1.949641</td>\n",
              "      <td>1.812514</td>\n",
              "      <td>0.408755</td>\n",
              "      <td>0.878086</td>\n",
              "      <td>01:46</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>1</td>\n",
              "      <td>1.673326</td>\n",
              "      <td>1.574671</td>\n",
              "      <td>0.525579</td>\n",
              "      <td>0.924154</td>\n",
              "      <td>01:46</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>2</td>\n",
              "      <td>1.478780</td>\n",
              "      <td>1.412071</td>\n",
              "      <td>0.631967</td>\n",
              "      <td>0.943752</td>\n",
              "      <td>01:46</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>3</td>\n",
              "      <td>1.386224</td>\n",
              "      <td>1.258784</td>\n",
              "      <td>0.675744</td>\n",
              "      <td>0.955714</td>\n",
              "      <td>01:46</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>4</td>\n",
              "      <td>1.139702</td>\n",
              "      <td>1.075350</td>\n",
              "      <td>0.765335</td>\n",
              "      <td>0.973276</td>\n",
              "      <td>01:47</td>\n",
              "    </tr>\n",
              "  </tbody>\n",
              "</table>"
            ],
            "text/plain": [
              "<IPython.core.display.HTML object>"
            ]
          },
          "metadata": {
            "tags": []
          }
        },
        {
          "output_type": "stream",
          "text": [
            "5 epochs: 0.7627894878387451 (2 runs)\n",
            "data path   /root/.fastai/data/imagewoof2\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "stream",
          "text": [
            ""
          ],
          "name": "stderr"
        },
        {
          "output_type": "stream",
          "text": [
            "Learn path /root/.fastai/data/imagewoof2\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "display_data",
          "data": {
            "text/html": [
              "<table border=\"1\" class=\"dataframe\">\n",
              "  <thead>\n",
              "    <tr style=\"text-align: left;\">\n",
              "      <th>epoch</th>\n",
              "      <th>train_loss</th>\n",
              "      <th>valid_loss</th>\n",
              "      <th>accuracy</th>\n",
              "      <th>top_k_accuracy</th>\n",
              "      <th>time</th>\n",
              "    </tr>\n",
              "  </thead>\n",
              "  <tbody>\n",
              "    <tr>\n",
              "      <td>0</td>\n",
              "      <td>1.917816</td>\n",
              "      <td>1.818723</td>\n",
              "      <td>0.416900</td>\n",
              "      <td>0.869687</td>\n",
              "      <td>01:47</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>1</td>\n",
              "      <td>1.645093</td>\n",
              "      <td>1.598224</td>\n",
              "      <td>0.527870</td>\n",
              "      <td>0.930262</td>\n",
              "      <td>01:46</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>2</td>\n",
              "      <td>1.497461</td>\n",
              "      <td>1.360193</td>\n",
              "      <td>0.642657</td>\n",
              "      <td>0.942479</td>\n",
              "      <td>01:46</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>3</td>\n",
              "      <td>1.375195</td>\n",
              "      <td>1.226473</td>\n",
              "      <td>0.694070</td>\n",
              "      <td>0.956477</td>\n",
              "      <td>01:46</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>4</td>\n",
              "      <td>1.174549</td>\n",
              "      <td>1.088659</td>\n",
              "      <td>0.766353</td>\n",
              "      <td>0.972512</td>\n",
              "      <td>01:47</td>\n",
              "    </tr>\n",
              "  </tbody>\n",
              "</table>"
            ],
            "text/plain": [
              "<IPython.core.display.HTML object>"
            ]
          },
          "metadata": {
            "tags": []
          }
        },
        {
          "output_type": "stream",
          "text": [
            "5 epochs: 0.7639772494633993 (3 runs)\n",
            "data path   /root/.fastai/data/imagewoof2\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "stream",
          "text": [
            ""
          ],
          "name": "stderr"
        },
        {
          "output_type": "stream",
          "text": [
            "Learn path /root/.fastai/data/imagewoof2\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "display_data",
          "data": {
            "text/html": [
              "<table border=\"1\" class=\"dataframe\">\n",
              "  <thead>\n",
              "    <tr style=\"text-align: left;\">\n",
              "      <th>epoch</th>\n",
              "      <th>train_loss</th>\n",
              "      <th>valid_loss</th>\n",
              "      <th>accuracy</th>\n",
              "      <th>top_k_accuracy</th>\n",
              "      <th>time</th>\n",
              "    </tr>\n",
              "  </thead>\n",
              "  <tbody>\n",
              "    <tr>\n",
              "      <td>0</td>\n",
              "      <td>1.970535</td>\n",
              "      <td>1.762262</td>\n",
              "      <td>0.433444</td>\n",
              "      <td>0.891321</td>\n",
              "      <td>01:46</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>1</td>\n",
              "      <td>1.683291</td>\n",
              "      <td>1.538967</td>\n",
              "      <td>0.541359</td>\n",
              "      <td>0.922881</td>\n",
              "      <td>01:46</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>2</td>\n",
              "      <td>1.477522</td>\n",
              "      <td>1.409047</td>\n",
              "      <td>0.620260</td>\n",
              "      <td>0.938407</td>\n",
              "      <td>01:47</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>3</td>\n",
              "      <td>1.395434</td>\n",
              "      <td>1.216235</td>\n",
              "      <td>0.700178</td>\n",
              "      <td>0.963095</td>\n",
              "      <td>01:47</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>4</td>\n",
              "      <td>1.160289</td>\n",
              "      <td>1.090528</td>\n",
              "      <td>0.764317</td>\n",
              "      <td>0.972512</td>\n",
              "      <td>01:47</td>\n",
              "    </tr>\n",
              "  </tbody>\n",
              "</table>"
            ],
            "text/plain": [
              "<IPython.core.display.HTML object>"
            ]
          },
          "metadata": {
            "tags": []
          }
        },
        {
          "output_type": "stream",
          "text": [
            "5 epochs: 0.7640620917081833 (4 runs)\n",
            "data path   /root/.fastai/data/imagewoof2\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "stream",
          "text": [
            ""
          ],
          "name": "stderr"
        },
        {
          "output_type": "stream",
          "text": [
            "Learn path /root/.fastai/data/imagewoof2\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "display_data",
          "data": {
            "text/html": [
              "<table border=\"1\" class=\"dataframe\">\n",
              "  <thead>\n",
              "    <tr style=\"text-align: left;\">\n",
              "      <th>epoch</th>\n",
              "      <th>train_loss</th>\n",
              "      <th>valid_loss</th>\n",
              "      <th>accuracy</th>\n",
              "      <th>top_k_accuracy</th>\n",
              "      <th>time</th>\n",
              "    </tr>\n",
              "  </thead>\n",
              "  <tbody>\n",
              "    <tr>\n",
              "      <td>0</td>\n",
              "      <td>1.933688</td>\n",
              "      <td>1.856529</td>\n",
              "      <td>0.435734</td>\n",
              "      <td>0.875541</td>\n",
              "      <td>01:47</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>1</td>\n",
              "      <td>1.638156</td>\n",
              "      <td>1.498143</td>\n",
              "      <td>0.571138</td>\n",
              "      <td>0.928226</td>\n",
              "      <td>01:47</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>2</td>\n",
              "      <td>1.484591</td>\n",
              "      <td>1.354342</td>\n",
              "      <td>0.640112</td>\n",
              "      <td>0.946806</td>\n",
              "      <td>01:47</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>3</td>\n",
              "      <td>1.340811</td>\n",
              "      <td>1.252327</td>\n",
              "      <td>0.678799</td>\n",
              "      <td>0.959532</td>\n",
              "      <td>01:47</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>4</td>\n",
              "      <td>1.155987</td>\n",
              "      <td>1.065511</td>\n",
              "      <td>0.774497</td>\n",
              "      <td>0.974803</td>\n",
              "      <td>01:47</td>\n",
              "    </tr>\n",
              "  </tbody>\n",
              "</table>"
            ],
            "text/plain": [
              "<IPython.core.display.HTML object>"
            ]
          },
          "metadata": {
            "tags": []
          }
        },
        {
          "output_type": "stream",
          "text": [
            "5 epochs: 0.7661491394042969 (5 runs)\n",
            "data path   /root/.fastai/data/imagewoof2\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "stream",
          "text": [
            ""
          ],
          "name": "stderr"
        },
        {
          "output_type": "stream",
          "text": [
            "Learn path /root/.fastai/data/imagewoof2\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "display_data",
          "data": {
            "text/html": [
              "<table border=\"1\" class=\"dataframe\">\n",
              "  <thead>\n",
              "    <tr style=\"text-align: left;\">\n",
              "      <th>epoch</th>\n",
              "      <th>train_loss</th>\n",
              "      <th>valid_loss</th>\n",
              "      <th>accuracy</th>\n",
              "      <th>top_k_accuracy</th>\n",
              "      <th>time</th>\n",
              "    </tr>\n",
              "  </thead>\n",
              "  <tbody>\n",
              "    <tr>\n",
              "      <td>0</td>\n",
              "      <td>1.978098</td>\n",
              "      <td>1.840412</td>\n",
              "      <td>0.392721</td>\n",
              "      <td>0.867905</td>\n",
              "      <td>01:47</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>1</td>\n",
              "      <td>1.681722</td>\n",
              "      <td>1.562918</td>\n",
              "      <td>0.547977</td>\n",
              "      <td>0.926190</td>\n",
              "      <td>01:47</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>2</td>\n",
              "      <td>1.477325</td>\n",
              "      <td>1.395806</td>\n",
              "      <td>0.611606</td>\n",
              "      <td>0.946297</td>\n",
              "      <td>01:46</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>3</td>\n",
              "      <td>1.415498</td>\n",
              "      <td>1.220735</td>\n",
              "      <td>0.698651</td>\n",
              "      <td>0.956223</td>\n",
              "      <td>01:46</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>4</td>\n",
              "      <td>1.311608</td>\n",
              "      <td>1.180558</td>\n",
              "      <td>0.724357</td>\n",
              "      <td>0.961568</td>\n",
              "      <td>01:46</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>5</td>\n",
              "      <td>1.211994</td>\n",
              "      <td>1.132529</td>\n",
              "      <td>0.734793</td>\n",
              "      <td>0.972258</td>\n",
              "      <td>01:46</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>6</td>\n",
              "      <td>1.139894</td>\n",
              "      <td>1.143434</td>\n",
              "      <td>0.749046</td>\n",
              "      <td>0.971494</td>\n",
              "      <td>01:46</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>7</td>\n",
              "      <td>1.114410</td>\n",
              "      <td>1.085456</td>\n",
              "      <td>0.759990</td>\n",
              "      <td>0.971494</td>\n",
              "      <td>01:47</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>8</td>\n",
              "      <td>1.048799</td>\n",
              "      <td>1.027087</td>\n",
              "      <td>0.798167</td>\n",
              "      <td>0.975312</td>\n",
              "      <td>01:47</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>9</td>\n",
              "      <td>0.995334</td>\n",
              "      <td>1.024305</td>\n",
              "      <td>0.787987</td>\n",
              "      <td>0.974548</td>\n",
              "      <td>01:46</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>10</td>\n",
              "      <td>0.961315</td>\n",
              "      <td>1.005183</td>\n",
              "      <td>0.793332</td>\n",
              "      <td>0.978366</td>\n",
              "      <td>01:47</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>11</td>\n",
              "      <td>0.969047</td>\n",
              "      <td>0.993947</td>\n",
              "      <td>0.803258</td>\n",
              "      <td>0.978366</td>\n",
              "      <td>01:46</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>12</td>\n",
              "      <td>0.934757</td>\n",
              "      <td>0.984377</td>\n",
              "      <td>0.809875</td>\n",
              "      <td>0.976075</td>\n",
              "      <td>01:46</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>13</td>\n",
              "      <td>0.914652</td>\n",
              "      <td>0.978541</td>\n",
              "      <td>0.808094</td>\n",
              "      <td>0.977348</td>\n",
              "      <td>01:47</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>14</td>\n",
              "      <td>0.896717</td>\n",
              "      <td>0.941878</td>\n",
              "      <td>0.822856</td>\n",
              "      <td>0.978875</td>\n",
              "      <td>01:46</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>15</td>\n",
              "      <td>0.864115</td>\n",
              "      <td>0.937392</td>\n",
              "      <td>0.816747</td>\n",
              "      <td>0.980148</td>\n",
              "      <td>01:47</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>16</td>\n",
              "      <td>0.785647</td>\n",
              "      <td>0.902200</td>\n",
              "      <td>0.840417</td>\n",
              "      <td>0.980148</td>\n",
              "      <td>01:47</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>17</td>\n",
              "      <td>0.708462</td>\n",
              "      <td>0.868717</td>\n",
              "      <td>0.855179</td>\n",
              "      <td>0.981420</td>\n",
              "      <td>01:47</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>18</td>\n",
              "      <td>0.662888</td>\n",
              "      <td>0.858488</td>\n",
              "      <td>0.861797</td>\n",
              "      <td>0.983202</td>\n",
              "      <td>01:47</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>19</td>\n",
              "      <td>0.639777</td>\n",
              "      <td>0.849512</td>\n",
              "      <td>0.864088</td>\n",
              "      <td>0.982947</td>\n",
              "      <td>01:47</td>\n",
              "    </tr>\n",
              "  </tbody>\n",
              "</table>"
            ],
            "text/plain": [
              "<IPython.core.display.HTML object>"
            ]
          },
          "metadata": {
            "tags": []
          }
        },
        {
          "output_type": "stream",
          "text": [
            ""
          ],
          "name": "stderr"
        },
        {
          "output_type": "stream",
          "text": [
            "20 epochs: 0.8640875816345215 (1 runs)\n",
            "data path   /root/.fastai/data/imagewoof2\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "stream",
          "text": [
            ""
          ],
          "name": "stderr"
        },
        {
          "output_type": "stream",
          "text": [
            "Learn path /root/.fastai/data/imagewoof2\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "display_data",
          "data": {
            "text/html": [
              "<table border=\"1\" class=\"dataframe\">\n",
              "  <thead>\n",
              "    <tr style=\"text-align: left;\">\n",
              "      <th>epoch</th>\n",
              "      <th>train_loss</th>\n",
              "      <th>valid_loss</th>\n",
              "      <th>accuracy</th>\n",
              "      <th>top_k_accuracy</th>\n",
              "      <th>time</th>\n",
              "    </tr>\n",
              "  </thead>\n",
              "  <tbody>\n",
              "    <tr>\n",
              "      <td>0</td>\n",
              "      <td>1.975429</td>\n",
              "      <td>1.799935</td>\n",
              "      <td>0.423772</td>\n",
              "      <td>0.884449</td>\n",
              "      <td>01:46</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>1</td>\n",
              "      <td>1.655976</td>\n",
              "      <td>1.520545</td>\n",
              "      <td>0.557394</td>\n",
              "      <td>0.929244</td>\n",
              "      <td>01:47</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>2</td>\n",
              "      <td>1.506151</td>\n",
              "      <td>1.380015</td>\n",
              "      <td>0.635276</td>\n",
              "      <td>0.945788</td>\n",
              "      <td>01:46</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>3</td>\n",
              "      <td>1.391121</td>\n",
              "      <td>1.279651</td>\n",
              "      <td>0.676508</td>\n",
              "      <td>0.956986</td>\n",
              "      <td>01:47</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>4</td>\n",
              "      <td>1.282224</td>\n",
              "      <td>1.176563</td>\n",
              "      <td>0.724103</td>\n",
              "      <td>0.964113</td>\n",
              "      <td>01:47</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>5</td>\n",
              "      <td>1.228438</td>\n",
              "      <td>1.128145</td>\n",
              "      <td>0.750064</td>\n",
              "      <td>0.968185</td>\n",
              "      <td>01:46</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>6</td>\n",
              "      <td>1.171350</td>\n",
              "      <td>1.116377</td>\n",
              "      <td>0.763808</td>\n",
              "      <td>0.970730</td>\n",
              "      <td>01:47</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>7</td>\n",
              "      <td>1.116685</td>\n",
              "      <td>1.100432</td>\n",
              "      <td>0.751845</td>\n",
              "      <td>0.965895</td>\n",
              "      <td>01:46</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>8</td>\n",
              "      <td>1.097065</td>\n",
              "      <td>1.079508</td>\n",
              "      <td>0.762535</td>\n",
              "      <td>0.972003</td>\n",
              "      <td>01:47</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>9</td>\n",
              "      <td>1.026528</td>\n",
              "      <td>0.993767</td>\n",
              "      <td>0.800967</td>\n",
              "      <td>0.977602</td>\n",
              "      <td>01:47</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>10</td>\n",
              "      <td>1.004926</td>\n",
              "      <td>1.002931</td>\n",
              "      <td>0.803767</td>\n",
              "      <td>0.974039</td>\n",
              "      <td>01:46</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>11</td>\n",
              "      <td>0.940144</td>\n",
              "      <td>0.994663</td>\n",
              "      <td>0.802749</td>\n",
              "      <td>0.980657</td>\n",
              "      <td>01:47</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>12</td>\n",
              "      <td>0.942080</td>\n",
              "      <td>0.997995</td>\n",
              "      <td>0.802749</td>\n",
              "      <td>0.973530</td>\n",
              "      <td>01:47</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>13</td>\n",
              "      <td>0.931812</td>\n",
              "      <td>0.988777</td>\n",
              "      <td>0.800967</td>\n",
              "      <td>0.976330</td>\n",
              "      <td>01:47</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>14</td>\n",
              "      <td>0.922512</td>\n",
              "      <td>0.975213</td>\n",
              "      <td>0.817256</td>\n",
              "      <td>0.976330</td>\n",
              "      <td>01:47</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>15</td>\n",
              "      <td>0.855333</td>\n",
              "      <td>0.932419</td>\n",
              "      <td>0.829473</td>\n",
              "      <td>0.979130</td>\n",
              "      <td>01:47</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>16</td>\n",
              "      <td>0.781972</td>\n",
              "      <td>0.894888</td>\n",
              "      <td>0.844744</td>\n",
              "      <td>0.981420</td>\n",
              "      <td>01:46</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>17</td>\n",
              "      <td>0.725589</td>\n",
              "      <td>0.870868</td>\n",
              "      <td>0.846017</td>\n",
              "      <td>0.986765</td>\n",
              "      <td>01:47</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>18</td>\n",
              "      <td>0.667819</td>\n",
              "      <td>0.856035</td>\n",
              "      <td>0.859761</td>\n",
              "      <td>0.985492</td>\n",
              "      <td>01:47</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>19</td>\n",
              "      <td>0.649653</td>\n",
              "      <td>0.852532</td>\n",
              "      <td>0.860779</td>\n",
              "      <td>0.985747</td>\n",
              "      <td>01:46</td>\n",
              "    </tr>\n",
              "  </tbody>\n",
              "</table>"
            ],
            "text/plain": [
              "<IPython.core.display.HTML object>"
            ]
          },
          "metadata": {
            "tags": []
          }
        },
        {
          "output_type": "stream",
          "text": [
            "20 epochs: 0.8624331951141357 (2 runs)\n",
            "data path   /root/.fastai/data/imagewoof2\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "stream",
          "text": [
            ""
          ],
          "name": "stderr"
        },
        {
          "output_type": "stream",
          "text": [
            "Learn path /root/.fastai/data/imagewoof2\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "display_data",
          "data": {
            "text/html": [
              "<table border=\"1\" class=\"dataframe\">\n",
              "  <thead>\n",
              "    <tr style=\"text-align: left;\">\n",
              "      <th>epoch</th>\n",
              "      <th>train_loss</th>\n",
              "      <th>valid_loss</th>\n",
              "      <th>accuracy</th>\n",
              "      <th>top_k_accuracy</th>\n",
              "      <th>time</th>\n",
              "    </tr>\n",
              "  </thead>\n",
              "  <tbody>\n",
              "    <tr>\n",
              "      <td>0</td>\n",
              "      <td>1.894029</td>\n",
              "      <td>1.890364</td>\n",
              "      <td>0.387630</td>\n",
              "      <td>0.872487</td>\n",
              "      <td>01:46</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>1</td>\n",
              "      <td>1.734278</td>\n",
              "      <td>1.572535</td>\n",
              "      <td>0.523288</td>\n",
              "      <td>0.920336</td>\n",
              "      <td>01:46</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>2</td>\n",
              "      <td>1.487276</td>\n",
              "      <td>1.354838</td>\n",
              "      <td>0.633495</td>\n",
              "      <td>0.951642</td>\n",
              "      <td>01:46</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>3</td>\n",
              "      <td>1.404745</td>\n",
              "      <td>1.292554</td>\n",
              "      <td>0.680326</td>\n",
              "      <td>0.962077</td>\n",
              "      <td>01:46</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>4</td>\n",
              "      <td>1.285029</td>\n",
              "      <td>1.211417</td>\n",
              "      <td>0.711377</td>\n",
              "      <td>0.955968</td>\n",
              "      <td>01:46</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>5</td>\n",
              "      <td>1.236139</td>\n",
              "      <td>1.152456</td>\n",
              "      <td>0.735302</td>\n",
              "      <td>0.966913</td>\n",
              "      <td>01:46</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>6</td>\n",
              "      <td>1.150852</td>\n",
              "      <td>1.105313</td>\n",
              "      <td>0.761008</td>\n",
              "      <td>0.969967</td>\n",
              "      <td>01:46</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>7</td>\n",
              "      <td>1.124456</td>\n",
              "      <td>1.064968</td>\n",
              "      <td>0.777552</td>\n",
              "      <td>0.975057</td>\n",
              "      <td>01:46</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>8</td>\n",
              "      <td>1.089274</td>\n",
              "      <td>1.023006</td>\n",
              "      <td>0.787732</td>\n",
              "      <td>0.978112</td>\n",
              "      <td>01:46</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>9</td>\n",
              "      <td>1.050791</td>\n",
              "      <td>1.020943</td>\n",
              "      <td>0.787478</td>\n",
              "      <td>0.972767</td>\n",
              "      <td>01:46</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>10</td>\n",
              "      <td>0.994340</td>\n",
              "      <td>0.973242</td>\n",
              "      <td>0.817002</td>\n",
              "      <td>0.976839</td>\n",
              "      <td>01:46</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>11</td>\n",
              "      <td>0.983807</td>\n",
              "      <td>0.996999</td>\n",
              "      <td>0.797658</td>\n",
              "      <td>0.975566</td>\n",
              "      <td>01:46</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>12</td>\n",
              "      <td>0.904621</td>\n",
              "      <td>0.951249</td>\n",
              "      <td>0.818529</td>\n",
              "      <td>0.980402</td>\n",
              "      <td>01:46</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>13</td>\n",
              "      <td>0.916141</td>\n",
              "      <td>0.962832</td>\n",
              "      <td>0.816238</td>\n",
              "      <td>0.981166</td>\n",
              "      <td>01:46</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>14</td>\n",
              "      <td>0.849107</td>\n",
              "      <td>0.920267</td>\n",
              "      <td>0.832273</td>\n",
              "      <td>0.977857</td>\n",
              "      <td>01:46</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>15</td>\n",
              "      <td>0.812668</td>\n",
              "      <td>0.936853</td>\n",
              "      <td>0.822601</td>\n",
              "      <td>0.978621</td>\n",
              "      <td>01:46</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>16</td>\n",
              "      <td>0.762831</td>\n",
              "      <td>0.888396</td>\n",
              "      <td>0.849326</td>\n",
              "      <td>0.982947</td>\n",
              "      <td>01:46</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>17</td>\n",
              "      <td>0.731299</td>\n",
              "      <td>0.859328</td>\n",
              "      <td>0.861288</td>\n",
              "      <td>0.982184</td>\n",
              "      <td>01:46</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>18</td>\n",
              "      <td>0.677930</td>\n",
              "      <td>0.850480</td>\n",
              "      <td>0.865869</td>\n",
              "      <td>0.981420</td>\n",
              "      <td>01:47</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>19</td>\n",
              "      <td>0.641894</td>\n",
              "      <td>0.850452</td>\n",
              "      <td>0.865106</td>\n",
              "      <td>0.982184</td>\n",
              "      <td>01:46</td>\n",
              "    </tr>\n",
              "  </tbody>\n",
              "</table>"
            ],
            "text/plain": [
              "<IPython.core.display.HTML object>"
            ]
          },
          "metadata": {
            "tags": []
          }
        },
        {
          "output_type": "stream",
          "text": [
            "20 epochs: 0.8633240063985189 (3 runs)\n",
            "data path   /root/.fastai/data/imagewoof2\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "stream",
          "text": [
            ""
          ],
          "name": "stderr"
        },
        {
          "output_type": "stream",
          "text": [
            "Learn path /root/.fastai/data/imagewoof2\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "display_data",
          "data": {
            "text/html": [
              "<table border=\"1\" class=\"dataframe\">\n",
              "  <thead>\n",
              "    <tr style=\"text-align: left;\">\n",
              "      <th>epoch</th>\n",
              "      <th>train_loss</th>\n",
              "      <th>valid_loss</th>\n",
              "      <th>accuracy</th>\n",
              "      <th>top_k_accuracy</th>\n",
              "      <th>time</th>\n",
              "    </tr>\n",
              "  </thead>\n",
              "  <tbody>\n",
              "    <tr>\n",
              "      <td>0</td>\n",
              "      <td>1.923434</td>\n",
              "      <td>1.821854</td>\n",
              "      <td>0.413082</td>\n",
              "      <td>0.871469</td>\n",
              "      <td>01:46</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>1</td>\n",
              "      <td>1.663576</td>\n",
              "      <td>1.482624</td>\n",
              "      <td>0.569102</td>\n",
              "      <td>0.936625</td>\n",
              "      <td>01:47</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>2</td>\n",
              "      <td>1.471752</td>\n",
              "      <td>1.401165</td>\n",
              "      <td>0.614406</td>\n",
              "      <td>0.949096</td>\n",
              "      <td>01:47</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>3</td>\n",
              "      <td>1.385280</td>\n",
              "      <td>1.288615</td>\n",
              "      <td>0.669127</td>\n",
              "      <td>0.955459</td>\n",
              "      <td>01:46</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>4</td>\n",
              "      <td>1.305134</td>\n",
              "      <td>1.159007</td>\n",
              "      <td>0.724103</td>\n",
              "      <td>0.972512</td>\n",
              "      <td>01:47</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>5</td>\n",
              "      <td>1.202688</td>\n",
              "      <td>1.113594</td>\n",
              "      <td>0.752863</td>\n",
              "      <td>0.967931</td>\n",
              "      <td>01:47</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>6</td>\n",
              "      <td>1.186822</td>\n",
              "      <td>1.085840</td>\n",
              "      <td>0.770934</td>\n",
              "      <td>0.970730</td>\n",
              "      <td>01:47</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>7</td>\n",
              "      <td>1.120432</td>\n",
              "      <td>1.121998</td>\n",
              "      <td>0.746246</td>\n",
              "      <td>0.967422</td>\n",
              "      <td>01:46</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>8</td>\n",
              "      <td>1.056983</td>\n",
              "      <td>1.034723</td>\n",
              "      <td>0.783660</td>\n",
              "      <td>0.971494</td>\n",
              "      <td>01:46</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>9</td>\n",
              "      <td>1.030232</td>\n",
              "      <td>1.016334</td>\n",
              "      <td>0.789768</td>\n",
              "      <td>0.976075</td>\n",
              "      <td>01:47</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>10</td>\n",
              "      <td>0.981877</td>\n",
              "      <td>1.008186</td>\n",
              "      <td>0.800204</td>\n",
              "      <td>0.975312</td>\n",
              "      <td>01:46</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>11</td>\n",
              "      <td>0.932054</td>\n",
              "      <td>0.983527</td>\n",
              "      <td>0.811402</td>\n",
              "      <td>0.976330</td>\n",
              "      <td>01:47</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>12</td>\n",
              "      <td>0.965216</td>\n",
              "      <td>0.980910</td>\n",
              "      <td>0.813439</td>\n",
              "      <td>0.972767</td>\n",
              "      <td>01:46</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>13</td>\n",
              "      <td>0.884650</td>\n",
              "      <td>0.960084</td>\n",
              "      <td>0.814711</td>\n",
              "      <td>0.979893</td>\n",
              "      <td>01:47</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>14</td>\n",
              "      <td>0.887330</td>\n",
              "      <td>0.972340</td>\n",
              "      <td>0.807585</td>\n",
              "      <td>0.977348</td>\n",
              "      <td>01:46</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>15</td>\n",
              "      <td>0.819340</td>\n",
              "      <td>0.950783</td>\n",
              "      <td>0.826419</td>\n",
              "      <td>0.978875</td>\n",
              "      <td>01:46</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>16</td>\n",
              "      <td>0.774761</td>\n",
              "      <td>0.912292</td>\n",
              "      <td>0.840672</td>\n",
              "      <td>0.979639</td>\n",
              "      <td>01:47</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>17</td>\n",
              "      <td>0.711928</td>\n",
              "      <td>0.885944</td>\n",
              "      <td>0.850598</td>\n",
              "      <td>0.979639</td>\n",
              "      <td>01:46</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>18</td>\n",
              "      <td>0.662045</td>\n",
              "      <td>0.865212</td>\n",
              "      <td>0.857470</td>\n",
              "      <td>0.982184</td>\n",
              "      <td>01:46</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>19</td>\n",
              "      <td>0.635391</td>\n",
              "      <td>0.858606</td>\n",
              "      <td>0.861288</td>\n",
              "      <td>0.982184</td>\n",
              "      <td>01:46</td>\n",
              "    </tr>\n",
              "  </tbody>\n",
              "</table>"
            ],
            "text/plain": [
              "<IPython.core.display.HTML object>"
            ]
          },
          "metadata": {
            "tags": []
          }
        },
        {
          "output_type": "stream",
          "text": [
            "20 epochs: 0.8628149628639221 (4 runs)\n",
            "data path   /root/.fastai/data/imagewoof2\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "stream",
          "text": [
            ""
          ],
          "name": "stderr"
        },
        {
          "output_type": "stream",
          "text": [
            "Learn path /root/.fastai/data/imagewoof2\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "display_data",
          "data": {
            "text/html": [
              "<table border=\"1\" class=\"dataframe\">\n",
              "  <thead>\n",
              "    <tr style=\"text-align: left;\">\n",
              "      <th>epoch</th>\n",
              "      <th>train_loss</th>\n",
              "      <th>valid_loss</th>\n",
              "      <th>accuracy</th>\n",
              "      <th>top_k_accuracy</th>\n",
              "      <th>time</th>\n",
              "    </tr>\n",
              "  </thead>\n",
              "  <tbody>\n",
              "    <tr>\n",
              "      <td>0</td>\n",
              "      <td>1.907900</td>\n",
              "      <td>1.816525</td>\n",
              "      <td>0.401120</td>\n",
              "      <td>0.877577</td>\n",
              "      <td>01:46</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>1</td>\n",
              "      <td>1.671003</td>\n",
              "      <td>1.492498</td>\n",
              "      <td>0.557394</td>\n",
              "      <td>0.934843</td>\n",
              "      <td>01:46</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>2</td>\n",
              "      <td>1.491276</td>\n",
              "      <td>1.427671</td>\n",
              "      <td>0.603461</td>\n",
              "      <td>0.941461</td>\n",
              "      <td>01:46</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>3</td>\n",
              "      <td>1.374749</td>\n",
              "      <td>1.277938</td>\n",
              "      <td>0.679562</td>\n",
              "      <td>0.955968</td>\n",
              "      <td>01:46</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>4</td>\n",
              "      <td>1.286557</td>\n",
              "      <td>1.179975</td>\n",
              "      <td>0.730211</td>\n",
              "      <td>0.963604</td>\n",
              "      <td>01:46</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>5</td>\n",
              "      <td>1.238837</td>\n",
              "      <td>1.169291</td>\n",
              "      <td>0.727921</td>\n",
              "      <td>0.968694</td>\n",
              "      <td>01:47</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>6</td>\n",
              "      <td>1.135119</td>\n",
              "      <td>1.091888</td>\n",
              "      <td>0.769152</td>\n",
              "      <td>0.967931</td>\n",
              "      <td>01:46</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>7</td>\n",
              "      <td>1.139735</td>\n",
              "      <td>1.030810</td>\n",
              "      <td>0.784933</td>\n",
              "      <td>0.974294</td>\n",
              "      <td>01:46</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>8</td>\n",
              "      <td>1.057286</td>\n",
              "      <td>1.020697</td>\n",
              "      <td>0.797149</td>\n",
              "      <td>0.972512</td>\n",
              "      <td>01:46</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>9</td>\n",
              "      <td>1.009045</td>\n",
              "      <td>1.026826</td>\n",
              "      <td>0.797658</td>\n",
              "      <td>0.974803</td>\n",
              "      <td>01:47</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>10</td>\n",
              "      <td>1.006605</td>\n",
              "      <td>0.997009</td>\n",
              "      <td>0.806058</td>\n",
              "      <td>0.973530</td>\n",
              "      <td>01:47</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>11</td>\n",
              "      <td>0.941577</td>\n",
              "      <td>0.979289</td>\n",
              "      <td>0.808857</td>\n",
              "      <td>0.976839</td>\n",
              "      <td>01:47</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>12</td>\n",
              "      <td>0.927257</td>\n",
              "      <td>0.997864</td>\n",
              "      <td>0.800713</td>\n",
              "      <td>0.975821</td>\n",
              "      <td>01:47</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>13</td>\n",
              "      <td>0.910646</td>\n",
              "      <td>0.984895</td>\n",
              "      <td>0.799949</td>\n",
              "      <td>0.977857</td>\n",
              "      <td>01:47</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>14</td>\n",
              "      <td>0.879073</td>\n",
              "      <td>0.952978</td>\n",
              "      <td>0.819292</td>\n",
              "      <td>0.977857</td>\n",
              "      <td>01:46</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>15</td>\n",
              "      <td>0.840701</td>\n",
              "      <td>0.954660</td>\n",
              "      <td>0.818020</td>\n",
              "      <td>0.979130</td>\n",
              "      <td>01:46</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>16</td>\n",
              "      <td>0.774826</td>\n",
              "      <td>0.887476</td>\n",
              "      <td>0.847544</td>\n",
              "      <td>0.983202</td>\n",
              "      <td>01:46</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>17</td>\n",
              "      <td>0.707123</td>\n",
              "      <td>0.871961</td>\n",
              "      <td>0.855943</td>\n",
              "      <td>0.982184</td>\n",
              "      <td>01:46</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>18</td>\n",
              "      <td>0.643880</td>\n",
              "      <td>0.859714</td>\n",
              "      <td>0.859761</td>\n",
              "      <td>0.981675</td>\n",
              "      <td>01:47</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>19</td>\n",
              "      <td>0.644603</td>\n",
              "      <td>0.858133</td>\n",
              "      <td>0.862560</td>\n",
              "      <td>0.980657</td>\n",
              "      <td>01:46</td>\n",
              "    </tr>\n",
              "  </tbody>\n",
              "</table>"
            ],
            "text/plain": [
              "<IPython.core.display.HTML object>"
            ]
          },
          "metadata": {
            "tags": []
          }
        },
        {
          "output_type": "stream",
          "text": [
            "20 epochs: 0.8627640604972839 (5 runs)\n",
            "data path   /root/.fastai/data/imagewoof2\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "stream",
          "text": [
            ""
          ],
          "name": "stderr"
        },
        {
          "output_type": "stream",
          "text": [
            "Learn path /root/.fastai/data/imagewoof2\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "display_data",
          "data": {
            "text/html": [
              "<table border=\"1\" class=\"dataframe\">\n",
              "  <thead>\n",
              "    <tr style=\"text-align: left;\">\n",
              "      <th>epoch</th>\n",
              "      <th>train_loss</th>\n",
              "      <th>valid_loss</th>\n",
              "      <th>accuracy</th>\n",
              "      <th>top_k_accuracy</th>\n",
              "      <th>time</th>\n",
              "    </tr>\n",
              "  </thead>\n",
              "  <tbody>\n",
              "    <tr>\n",
              "      <td>0</td>\n",
              "      <td>2.104026</td>\n",
              "      <td>1.864537</td>\n",
              "      <td>0.367269</td>\n",
              "      <td>0.863833</td>\n",
              "      <td>01:46</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>1</td>\n",
              "      <td>1.856348</td>\n",
              "      <td>1.573903</td>\n",
              "      <td>0.525325</td>\n",
              "      <td>0.926444</td>\n",
              "      <td>01:47</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>2</td>\n",
              "      <td>1.680949</td>\n",
              "      <td>1.461086</td>\n",
              "      <td>0.586663</td>\n",
              "      <td>0.930517</td>\n",
              "      <td>01:47</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>3</td>\n",
              "      <td>1.594397</td>\n",
              "      <td>1.396894</td>\n",
              "      <td>0.609315</td>\n",
              "      <td>0.950115</td>\n",
              "      <td>01:47</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>4</td>\n",
              "      <td>1.553433</td>\n",
              "      <td>1.244519</td>\n",
              "      <td>0.683125</td>\n",
              "      <td>0.965386</td>\n",
              "      <td>01:47</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>5</td>\n",
              "      <td>1.474216</td>\n",
              "      <td>1.163294</td>\n",
              "      <td>0.733774</td>\n",
              "      <td>0.964622</td>\n",
              "      <td>01:47</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>6</td>\n",
              "      <td>1.433241</td>\n",
              "      <td>1.106275</td>\n",
              "      <td>0.758717</td>\n",
              "      <td>0.967167</td>\n",
              "      <td>01:47</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>7</td>\n",
              "      <td>1.356523</td>\n",
              "      <td>1.058523</td>\n",
              "      <td>0.779588</td>\n",
              "      <td>0.969203</td>\n",
              "      <td>01:47</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>8</td>\n",
              "      <td>1.332355</td>\n",
              "      <td>1.095734</td>\n",
              "      <td>0.750827</td>\n",
              "      <td>0.969203</td>\n",
              "      <td>01:47</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>9</td>\n",
              "      <td>1.298666</td>\n",
              "      <td>1.039442</td>\n",
              "      <td>0.773479</td>\n",
              "      <td>0.975312</td>\n",
              "      <td>01:47</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>10</td>\n",
              "      <td>1.280883</td>\n",
              "      <td>1.035716</td>\n",
              "      <td>0.787987</td>\n",
              "      <td>0.978112</td>\n",
              "      <td>01:47</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>11</td>\n",
              "      <td>1.237442</td>\n",
              "      <td>0.968096</td>\n",
              "      <td>0.813184</td>\n",
              "      <td>0.979384</td>\n",
              "      <td>01:47</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>12</td>\n",
              "      <td>1.233580</td>\n",
              "      <td>0.988952</td>\n",
              "      <td>0.801985</td>\n",
              "      <td>0.973276</td>\n",
              "      <td>01:47</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>13</td>\n",
              "      <td>1.218070</td>\n",
              "      <td>1.017908</td>\n",
              "      <td>0.793077</td>\n",
              "      <td>0.971494</td>\n",
              "      <td>01:48</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>14</td>\n",
              "      <td>1.206301</td>\n",
              "      <td>0.954525</td>\n",
              "      <td>0.813693</td>\n",
              "      <td>0.975821</td>\n",
              "      <td>01:48</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>15</td>\n",
              "      <td>1.166402</td>\n",
              "      <td>0.960505</td>\n",
              "      <td>0.810384</td>\n",
              "      <td>0.975566</td>\n",
              "      <td>01:48</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>16</td>\n",
              "      <td>1.148083</td>\n",
              "      <td>0.941969</td>\n",
              "      <td>0.822601</td>\n",
              "      <td>0.975312</td>\n",
              "      <td>01:48</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>17</td>\n",
              "      <td>1.109694</td>\n",
              "      <td>0.943269</td>\n",
              "      <td>0.823619</td>\n",
              "      <td>0.977348</td>\n",
              "      <td>01:47</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>18</td>\n",
              "      <td>1.116409</td>\n",
              "      <td>0.934476</td>\n",
              "      <td>0.828964</td>\n",
              "      <td>0.976330</td>\n",
              "      <td>01:48</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>19</td>\n",
              "      <td>1.091732</td>\n",
              "      <td>0.947828</td>\n",
              "      <td>0.820311</td>\n",
              "      <td>0.976075</td>\n",
              "      <td>01:48</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>20</td>\n",
              "      <td>1.081090</td>\n",
              "      <td>0.925880</td>\n",
              "      <td>0.828964</td>\n",
              "      <td>0.980911</td>\n",
              "      <td>01:48</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>21</td>\n",
              "      <td>1.085507</td>\n",
              "      <td>0.951862</td>\n",
              "      <td>0.816747</td>\n",
              "      <td>0.975057</td>\n",
              "      <td>01:47</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>22</td>\n",
              "      <td>1.089747</td>\n",
              "      <td>0.921958</td>\n",
              "      <td>0.832273</td>\n",
              "      <td>0.976584</td>\n",
              "      <td>01:47</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>23</td>\n",
              "      <td>1.068079</td>\n",
              "      <td>0.928318</td>\n",
              "      <td>0.828455</td>\n",
              "      <td>0.975057</td>\n",
              "      <td>01:47</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>24</td>\n",
              "      <td>1.067990</td>\n",
              "      <td>0.909584</td>\n",
              "      <td>0.832782</td>\n",
              "      <td>0.980148</td>\n",
              "      <td>01:47</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>25</td>\n",
              "      <td>1.040258</td>\n",
              "      <td>0.928225</td>\n",
              "      <td>0.831255</td>\n",
              "      <td>0.975566</td>\n",
              "      <td>01:47</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>26</td>\n",
              "      <td>1.006448</td>\n",
              "      <td>0.923124</td>\n",
              "      <td>0.832018</td>\n",
              "      <td>0.977348</td>\n",
              "      <td>01:47</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>27</td>\n",
              "      <td>1.006262</td>\n",
              "      <td>0.900135</td>\n",
              "      <td>0.838127</td>\n",
              "      <td>0.980402</td>\n",
              "      <td>01:47</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>28</td>\n",
              "      <td>0.979980</td>\n",
              "      <td>0.893426</td>\n",
              "      <td>0.848562</td>\n",
              "      <td>0.977093</td>\n",
              "      <td>01:47</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>29</td>\n",
              "      <td>1.003383</td>\n",
              "      <td>0.919751</td>\n",
              "      <td>0.838636</td>\n",
              "      <td>0.974803</td>\n",
              "      <td>01:47</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>30</td>\n",
              "      <td>0.999788</td>\n",
              "      <td>0.922397</td>\n",
              "      <td>0.836091</td>\n",
              "      <td>0.976584</td>\n",
              "      <td>01:47</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>31</td>\n",
              "      <td>1.001206</td>\n",
              "      <td>0.940309</td>\n",
              "      <td>0.832273</td>\n",
              "      <td>0.972003</td>\n",
              "      <td>01:47</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>32</td>\n",
              "      <td>0.985938</td>\n",
              "      <td>0.910202</td>\n",
              "      <td>0.843726</td>\n",
              "      <td>0.974803</td>\n",
              "      <td>01:47</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>33</td>\n",
              "      <td>0.994759</td>\n",
              "      <td>0.913880</td>\n",
              "      <td>0.844235</td>\n",
              "      <td>0.976330</td>\n",
              "      <td>01:47</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>34</td>\n",
              "      <td>0.973003</td>\n",
              "      <td>0.913888</td>\n",
              "      <td>0.835327</td>\n",
              "      <td>0.978112</td>\n",
              "      <td>01:48</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>35</td>\n",
              "      <td>0.973202</td>\n",
              "      <td>0.913800</td>\n",
              "      <td>0.839399</td>\n",
              "      <td>0.974803</td>\n",
              "      <td>01:47</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>36</td>\n",
              "      <td>0.971683</td>\n",
              "      <td>0.910712</td>\n",
              "      <td>0.837363</td>\n",
              "      <td>0.970221</td>\n",
              "      <td>01:47</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>37</td>\n",
              "      <td>0.952377</td>\n",
              "      <td>0.896922</td>\n",
              "      <td>0.842199</td>\n",
              "      <td>0.979639</td>\n",
              "      <td>01:47</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>38</td>\n",
              "      <td>0.936197</td>\n",
              "      <td>0.902463</td>\n",
              "      <td>0.851107</td>\n",
              "      <td>0.973276</td>\n",
              "      <td>01:48</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>39</td>\n",
              "      <td>0.953358</td>\n",
              "      <td>0.931822</td>\n",
              "      <td>0.825146</td>\n",
              "      <td>0.972003</td>\n",
              "      <td>01:48</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>40</td>\n",
              "      <td>0.968667</td>\n",
              "      <td>0.920174</td>\n",
              "      <td>0.833545</td>\n",
              "      <td>0.976330</td>\n",
              "      <td>01:48</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>41</td>\n",
              "      <td>0.916876</td>\n",
              "      <td>0.915774</td>\n",
              "      <td>0.838127</td>\n",
              "      <td>0.973530</td>\n",
              "      <td>01:47</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>42</td>\n",
              "      <td>0.937832</td>\n",
              "      <td>0.918115</td>\n",
              "      <td>0.836345</td>\n",
              "      <td>0.977093</td>\n",
              "      <td>01:47</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>43</td>\n",
              "      <td>0.938899</td>\n",
              "      <td>0.917023</td>\n",
              "      <td>0.839908</td>\n",
              "      <td>0.975566</td>\n",
              "      <td>01:48</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>44</td>\n",
              "      <td>0.914437</td>\n",
              "      <td>0.899404</td>\n",
              "      <td>0.843472</td>\n",
              "      <td>0.972258</td>\n",
              "      <td>01:47</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>45</td>\n",
              "      <td>0.928872</td>\n",
              "      <td>0.928782</td>\n",
              "      <td>0.836345</td>\n",
              "      <td>0.976584</td>\n",
              "      <td>01:48</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>46</td>\n",
              "      <td>0.917976</td>\n",
              "      <td>0.940158</td>\n",
              "      <td>0.836600</td>\n",
              "      <td>0.973276</td>\n",
              "      <td>01:47</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>47</td>\n",
              "      <td>0.924441</td>\n",
              "      <td>0.903903</td>\n",
              "      <td>0.842708</td>\n",
              "      <td>0.974548</td>\n",
              "      <td>01:48</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>48</td>\n",
              "      <td>0.895678</td>\n",
              "      <td>0.925589</td>\n",
              "      <td>0.838381</td>\n",
              "      <td>0.972003</td>\n",
              "      <td>01:47</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>49</td>\n",
              "      <td>0.922902</td>\n",
              "      <td>0.882471</td>\n",
              "      <td>0.854416</td>\n",
              "      <td>0.976584</td>\n",
              "      <td>01:48</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>50</td>\n",
              "      <td>0.916763</td>\n",
              "      <td>0.899929</td>\n",
              "      <td>0.843726</td>\n",
              "      <td>0.979384</td>\n",
              "      <td>01:48</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>51</td>\n",
              "      <td>0.937032</td>\n",
              "      <td>0.896183</td>\n",
              "      <td>0.853398</td>\n",
              "      <td>0.975057</td>\n",
              "      <td>01:48</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>52</td>\n",
              "      <td>0.903763</td>\n",
              "      <td>0.954137</td>\n",
              "      <td>0.826673</td>\n",
              "      <td>0.975566</td>\n",
              "      <td>01:48</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>53</td>\n",
              "      <td>0.897427</td>\n",
              "      <td>0.884675</td>\n",
              "      <td>0.852889</td>\n",
              "      <td>0.979893</td>\n",
              "      <td>01:48</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>54</td>\n",
              "      <td>0.917897</td>\n",
              "      <td>0.911819</td>\n",
              "      <td>0.837109</td>\n",
              "      <td>0.975566</td>\n",
              "      <td>01:48</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>55</td>\n",
              "      <td>0.907412</td>\n",
              "      <td>0.891353</td>\n",
              "      <td>0.848562</td>\n",
              "      <td>0.975821</td>\n",
              "      <td>01:48</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>56</td>\n",
              "      <td>0.907302</td>\n",
              "      <td>0.907586</td>\n",
              "      <td>0.840417</td>\n",
              "      <td>0.975566</td>\n",
              "      <td>01:48</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>57</td>\n",
              "      <td>0.899647</td>\n",
              "      <td>0.899512</td>\n",
              "      <td>0.844999</td>\n",
              "      <td>0.977093</td>\n",
              "      <td>01:48</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>58</td>\n",
              "      <td>0.904572</td>\n",
              "      <td>0.911750</td>\n",
              "      <td>0.845253</td>\n",
              "      <td>0.971749</td>\n",
              "      <td>01:48</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>59</td>\n",
              "      <td>0.901835</td>\n",
              "      <td>0.897397</td>\n",
              "      <td>0.853143</td>\n",
              "      <td>0.971494</td>\n",
              "      <td>01:48</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>60</td>\n",
              "      <td>0.889719</td>\n",
              "      <td>0.904590</td>\n",
              "      <td>0.848562</td>\n",
              "      <td>0.977348</td>\n",
              "      <td>01:48</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>61</td>\n",
              "      <td>0.893384</td>\n",
              "      <td>0.909857</td>\n",
              "      <td>0.846017</td>\n",
              "      <td>0.972258</td>\n",
              "      <td>01:48</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>62</td>\n",
              "      <td>0.897149</td>\n",
              "      <td>0.908683</td>\n",
              "      <td>0.844999</td>\n",
              "      <td>0.966404</td>\n",
              "      <td>01:48</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>63</td>\n",
              "      <td>0.860686</td>\n",
              "      <td>0.891167</td>\n",
              "      <td>0.847544</td>\n",
              "      <td>0.975312</td>\n",
              "      <td>01:48</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>64</td>\n",
              "      <td>0.896363</td>\n",
              "      <td>0.861917</td>\n",
              "      <td>0.858997</td>\n",
              "      <td>0.979893</td>\n",
              "      <td>01:48</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>65</td>\n",
              "      <td>0.867821</td>\n",
              "      <td>0.872335</td>\n",
              "      <td>0.864597</td>\n",
              "      <td>0.974548</td>\n",
              "      <td>01:48</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>66</td>\n",
              "      <td>0.873009</td>\n",
              "      <td>0.876035</td>\n",
              "      <td>0.854416</td>\n",
              "      <td>0.972003</td>\n",
              "      <td>01:48</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>67</td>\n",
              "      <td>0.839721</td>\n",
              "      <td>0.846909</td>\n",
              "      <td>0.868923</td>\n",
              "      <td>0.976839</td>\n",
              "      <td>01:48</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>68</td>\n",
              "      <td>0.834045</td>\n",
              "      <td>0.846694</td>\n",
              "      <td>0.868160</td>\n",
              "      <td>0.979893</td>\n",
              "      <td>01:48</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>69</td>\n",
              "      <td>0.854671</td>\n",
              "      <td>0.843493</td>\n",
              "      <td>0.871469</td>\n",
              "      <td>0.977857</td>\n",
              "      <td>01:48</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>70</td>\n",
              "      <td>0.831829</td>\n",
              "      <td>0.848163</td>\n",
              "      <td>0.874523</td>\n",
              "      <td>0.975821</td>\n",
              "      <td>01:47</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>71</td>\n",
              "      <td>0.836178</td>\n",
              "      <td>0.841296</td>\n",
              "      <td>0.867651</td>\n",
              "      <td>0.976584</td>\n",
              "      <td>01:48</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>72</td>\n",
              "      <td>0.806116</td>\n",
              "      <td>0.818454</td>\n",
              "      <td>0.877322</td>\n",
              "      <td>0.980657</td>\n",
              "      <td>01:47</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>73</td>\n",
              "      <td>0.814034</td>\n",
              "      <td>0.822671</td>\n",
              "      <td>0.874523</td>\n",
              "      <td>0.979384</td>\n",
              "      <td>01:48</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>74</td>\n",
              "      <td>0.807170</td>\n",
              "      <td>0.823657</td>\n",
              "      <td>0.876813</td>\n",
              "      <td>0.975312</td>\n",
              "      <td>01:49</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>75</td>\n",
              "      <td>0.801271</td>\n",
              "      <td>0.820889</td>\n",
              "      <td>0.876813</td>\n",
              "      <td>0.976330</td>\n",
              "      <td>01:49</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>76</td>\n",
              "      <td>0.796058</td>\n",
              "      <td>0.813411</td>\n",
              "      <td>0.878086</td>\n",
              "      <td>0.979893</td>\n",
              "      <td>01:49</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>77</td>\n",
              "      <td>0.793756</td>\n",
              "      <td>0.809772</td>\n",
              "      <td>0.876559</td>\n",
              "      <td>0.978875</td>\n",
              "      <td>01:49</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>78</td>\n",
              "      <td>0.805422</td>\n",
              "      <td>0.807021</td>\n",
              "      <td>0.878341</td>\n",
              "      <td>0.980657</td>\n",
              "      <td>01:49</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>79</td>\n",
              "      <td>0.767961</td>\n",
              "      <td>0.807906</td>\n",
              "      <td>0.876559</td>\n",
              "      <td>0.980657</td>\n",
              "      <td>01:49</td>\n",
              "    </tr>\n",
              "  </tbody>\n",
              "</table>"
            ],
            "text/plain": [
              "<IPython.core.display.HTML object>"
            ]
          },
          "metadata": {
            "tags": []
          }
        },
        {
          "output_type": "stream",
          "text": [
            "80 epochs: 0.8765588998794556 (1 runs)\n",
            "depth multiplier= {80: 0.8765588998794556}\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "b9r-zTaqZNhb",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 119
        },
        "outputId": "a094bf49-cf21-4492-9128-81636124ae73"
      },
      "source": [
        "res[-5:], sum(res[-5:])/5"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "([0.7576991319656372,\n",
              "  0.7459913492202759,\n",
              "  0.7556630373001099,\n",
              "  0.7411555051803589,\n",
              "  0.7592262625694275],\n",
              " 0.7519470572471618)"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 19
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "ywIipu3RVnHn",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 1000
        },
        "outputId": "95e63313-5fb8-4dc3-8c16-1175e59cfee5"
      },
      "source": [
        "for epochs in [20]*5:\n",
        "    mixup=0 if epochs<=20 else 0.2\n",
        "    learn = get_learn(model=model, size=128, bs=16, mixup=mixup)\n",
        "    learn.fit_fc(epochs, lr=4e-3, moms=(0.95,0.95), start_pct=0.72)\n",
        "    res += [learn.recorder.metrics[-1][0].item()]\n",
        "print([round(x, 6) for x in res], sum(res)/len(res))"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "data path   /root/.fastai/data/imagewoof2\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "stream",
          "text": [
            ""
          ],
          "name": "stderr"
        },
        {
          "output_type": "stream",
          "text": [
            "Learn path /root/.fastai/data/imagewoof2\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "display_data",
          "data": {
            "text/html": [
              "<table border=\"1\" class=\"dataframe\">\n",
              "  <thead>\n",
              "    <tr style=\"text-align: left;\">\n",
              "      <th>epoch</th>\n",
              "      <th>train_loss</th>\n",
              "      <th>valid_loss</th>\n",
              "      <th>accuracy</th>\n",
              "      <th>top_k_accuracy</th>\n",
              "      <th>time</th>\n",
              "    </tr>\n",
              "  </thead>\n",
              "  <tbody>\n",
              "    <tr>\n",
              "      <td>0</td>\n",
              "      <td>1.983266</td>\n",
              "      <td>1.791575</td>\n",
              "      <td>0.413591</td>\n",
              "      <td>0.870450</td>\n",
              "      <td>01:38</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>1</td>\n",
              "      <td>1.689537</td>\n",
              "      <td>1.573345</td>\n",
              "      <td>0.517944</td>\n",
              "      <td>0.928735</td>\n",
              "      <td>01:38</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>2</td>\n",
              "      <td>1.538139</td>\n",
              "      <td>1.434878</td>\n",
              "      <td>0.593026</td>\n",
              "      <td>0.943243</td>\n",
              "      <td>01:40</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>3</td>\n",
              "      <td>1.393348</td>\n",
              "      <td>1.309774</td>\n",
              "      <td>0.661746</td>\n",
              "      <td>0.951133</td>\n",
              "      <td>01:39</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>4</td>\n",
              "      <td>1.345759</td>\n",
              "      <td>1.233641</td>\n",
              "      <td>0.704250</td>\n",
              "      <td>0.962586</td>\n",
              "      <td>01:39</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>5</td>\n",
              "      <td>1.250168</td>\n",
              "      <td>1.157207</td>\n",
              "      <td>0.736065</td>\n",
              "      <td>0.966658</td>\n",
              "      <td>01:39</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>6</td>\n",
              "      <td>1.220388</td>\n",
              "      <td>1.106133</td>\n",
              "      <td>0.755918</td>\n",
              "      <td>0.969203</td>\n",
              "      <td>01:39</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>7</td>\n",
              "      <td>1.145580</td>\n",
              "      <td>1.103860</td>\n",
              "      <td>0.758208</td>\n",
              "      <td>0.970476</td>\n",
              "      <td>01:40</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>8</td>\n",
              "      <td>1.110340</td>\n",
              "      <td>1.084391</td>\n",
              "      <td>0.773225</td>\n",
              "      <td>0.968949</td>\n",
              "      <td>01:40</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>9</td>\n",
              "      <td>1.046738</td>\n",
              "      <td>1.067602</td>\n",
              "      <td>0.768898</td>\n",
              "      <td>0.972512</td>\n",
              "      <td>01:40</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>10</td>\n",
              "      <td>1.037448</td>\n",
              "      <td>1.013442</td>\n",
              "      <td>0.788750</td>\n",
              "      <td>0.974294</td>\n",
              "      <td>01:40</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>11</td>\n",
              "      <td>1.002906</td>\n",
              "      <td>1.000981</td>\n",
              "      <td>0.799695</td>\n",
              "      <td>0.975312</td>\n",
              "      <td>01:41</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>12</td>\n",
              "      <td>0.956967</td>\n",
              "      <td>0.972990</td>\n",
              "      <td>0.815220</td>\n",
              "      <td>0.973785</td>\n",
              "      <td>01:40</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>13</td>\n",
              "      <td>0.920305</td>\n",
              "      <td>1.020994</td>\n",
              "      <td>0.787987</td>\n",
              "      <td>0.976839</td>\n",
              "      <td>01:40</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>14</td>\n",
              "      <td>0.909070</td>\n",
              "      <td>0.976815</td>\n",
              "      <td>0.810130</td>\n",
              "      <td>0.978366</td>\n",
              "      <td>01:41</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>15</td>\n",
              "      <td>0.876554</td>\n",
              "      <td>0.955103</td>\n",
              "      <td>0.819038</td>\n",
              "      <td>0.974803</td>\n",
              "      <td>01:41</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>16</td>\n",
              "      <td>0.777766</td>\n",
              "      <td>0.911541</td>\n",
              "      <td>0.835836</td>\n",
              "      <td>0.979384</td>\n",
              "      <td>01:41</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>17</td>\n",
              "      <td>0.725446</td>\n",
              "      <td>0.898692</td>\n",
              "      <td>0.844490</td>\n",
              "      <td>0.976584</td>\n",
              "      <td>01:41</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>18</td>\n",
              "      <td>0.671762</td>\n",
              "      <td>0.876823</td>\n",
              "      <td>0.847544</td>\n",
              "      <td>0.981675</td>\n",
              "      <td>01:41</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>19</td>\n",
              "      <td>0.643587</td>\n",
              "      <td>0.876909</td>\n",
              "      <td>0.848562</td>\n",
              "      <td>0.980148</td>\n",
              "      <td>01:41</td>\n",
              "    </tr>\n",
              "  </tbody>\n",
              "</table>"
            ],
            "text/plain": [
              "<IPython.core.display.HTML object>"
            ]
          },
          "metadata": {
            "tags": []
          }
        },
        {
          "output_type": "stream",
          "text": [
            "data path   /root/.fastai/data/imagewoof2\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "stream",
          "text": [
            ""
          ],
          "name": "stderr"
        },
        {
          "output_type": "stream",
          "text": [
            "Learn path /root/.fastai/data/imagewoof2\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "display_data",
          "data": {
            "text/html": [
              "<table border=\"1\" class=\"dataframe\">\n",
              "  <thead>\n",
              "    <tr style=\"text-align: left;\">\n",
              "      <th>epoch</th>\n",
              "      <th>train_loss</th>\n",
              "      <th>valid_loss</th>\n",
              "      <th>accuracy</th>\n",
              "      <th>top_k_accuracy</th>\n",
              "      <th>time</th>\n",
              "    </tr>\n",
              "  </thead>\n",
              "  <tbody>\n",
              "    <tr>\n",
              "      <td>0</td>\n",
              "      <td>1.930710</td>\n",
              "      <td>1.814667</td>\n",
              "      <td>0.417409</td>\n",
              "      <td>0.881649</td>\n",
              "      <td>01:40</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>1</td>\n",
              "      <td>1.742082</td>\n",
              "      <td>1.586188</td>\n",
              "      <td>0.519725</td>\n",
              "      <td>0.913209</td>\n",
              "      <td>01:39</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>2</td>\n",
              "      <td>1.512977</td>\n",
              "      <td>1.429439</td>\n",
              "      <td>0.596589</td>\n",
              "      <td>0.942734</td>\n",
              "      <td>01:40</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>3</td>\n",
              "      <td>1.436196</td>\n",
              "      <td>1.286632</td>\n",
              "      <td>0.672181</td>\n",
              "      <td>0.959532</td>\n",
              "      <td>01:39</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>4</td>\n",
              "      <td>1.322253</td>\n",
              "      <td>1.213644</td>\n",
              "      <td>0.702978</td>\n",
              "      <td>0.963095</td>\n",
              "      <td>01:40</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>5</td>\n",
              "      <td>1.230132</td>\n",
              "      <td>1.142426</td>\n",
              "      <td>0.733520</td>\n",
              "      <td>0.970476</td>\n",
              "      <td>01:41</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>6</td>\n",
              "      <td>1.146980</td>\n",
              "      <td>1.136797</td>\n",
              "      <td>0.731229</td>\n",
              "      <td>0.966913</td>\n",
              "      <td>01:42</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>7</td>\n",
              "      <td>1.135156</td>\n",
              "      <td>1.067523</td>\n",
              "      <td>0.773479</td>\n",
              "      <td>0.973276</td>\n",
              "      <td>01:42</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>8</td>\n",
              "      <td>1.083601</td>\n",
              "      <td>1.046675</td>\n",
              "      <td>0.782133</td>\n",
              "      <td>0.974803</td>\n",
              "      <td>01:41</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>9</td>\n",
              "      <td>1.052225</td>\n",
              "      <td>1.041891</td>\n",
              "      <td>0.779333</td>\n",
              "      <td>0.973785</td>\n",
              "      <td>01:41</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>10</td>\n",
              "      <td>0.996635</td>\n",
              "      <td>1.007911</td>\n",
              "      <td>0.789514</td>\n",
              "      <td>0.976075</td>\n",
              "      <td>01:40</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>11</td>\n",
              "      <td>0.993732</td>\n",
              "      <td>1.032971</td>\n",
              "      <td>0.786714</td>\n",
              "      <td>0.970476</td>\n",
              "      <td>01:40</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>12</td>\n",
              "      <td>0.965586</td>\n",
              "      <td>1.000444</td>\n",
              "      <td>0.803767</td>\n",
              "      <td>0.974803</td>\n",
              "      <td>01:40</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>13</td>\n",
              "      <td>0.959201</td>\n",
              "      <td>0.973568</td>\n",
              "      <td>0.810130</td>\n",
              "      <td>0.978875</td>\n",
              "      <td>01:41</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>14</td>\n",
              "      <td>0.919285</td>\n",
              "      <td>0.962124</td>\n",
              "      <td>0.815220</td>\n",
              "      <td>0.978112</td>\n",
              "      <td>01:41</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>15</td>\n",
              "      <td>0.859854</td>\n",
              "      <td>0.977505</td>\n",
              "      <td>0.808857</td>\n",
              "      <td>0.973530</td>\n",
              "      <td>01:41</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>16</td>\n",
              "      <td>0.771055</td>\n",
              "      <td>0.920575</td>\n",
              "      <td>0.832018</td>\n",
              "      <td>0.980657</td>\n",
              "      <td>01:41</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>17</td>\n",
              "      <td>0.723112</td>\n",
              "      <td>0.897667</td>\n",
              "      <td>0.843981</td>\n",
              "      <td>0.980402</td>\n",
              "      <td>01:41</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>18</td>\n",
              "      <td>0.659049</td>\n",
              "      <td>0.878490</td>\n",
              "      <td>0.849071</td>\n",
              "      <td>0.979384</td>\n",
              "      <td>01:41</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>19</td>\n",
              "      <td>0.655036</td>\n",
              "      <td>0.875781</td>\n",
              "      <td>0.850598</td>\n",
              "      <td>0.979130</td>\n",
              "      <td>01:41</td>\n",
              "    </tr>\n",
              "  </tbody>\n",
              "</table>"
            ],
            "text/plain": [
              "<IPython.core.display.HTML object>"
            ]
          },
          "metadata": {
            "tags": []
          }
        },
        {
          "output_type": "stream",
          "text": [
            "data path   /root/.fastai/data/imagewoof2\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "stream",
          "text": [
            ""
          ],
          "name": "stderr"
        },
        {
          "output_type": "stream",
          "text": [
            "Learn path /root/.fastai/data/imagewoof2\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "display_data",
          "data": {
            "text/html": [
              "<table border=\"1\" class=\"dataframe\">\n",
              "  <thead>\n",
              "    <tr style=\"text-align: left;\">\n",
              "      <th>epoch</th>\n",
              "      <th>train_loss</th>\n",
              "      <th>valid_loss</th>\n",
              "      <th>accuracy</th>\n",
              "      <th>top_k_accuracy</th>\n",
              "      <th>time</th>\n",
              "    </tr>\n",
              "  </thead>\n",
              "  <tbody>\n",
              "    <tr>\n",
              "      <td>0</td>\n",
              "      <td>1.983735</td>\n",
              "      <td>1.819994</td>\n",
              "      <td>0.413591</td>\n",
              "      <td>0.885976</td>\n",
              "      <td>01:41</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>1</td>\n",
              "      <td>1.697815</td>\n",
              "      <td>1.655524</td>\n",
              "      <td>0.497837</td>\n",
              "      <td>0.907610</td>\n",
              "      <td>01:41</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>2</td>\n",
              "      <td>1.562741</td>\n",
              "      <td>1.456674</td>\n",
              "      <td>0.593026</td>\n",
              "      <td>0.939934</td>\n",
              "      <td>01:41</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>3</td>\n",
              "      <td>1.478655</td>\n",
              "      <td>1.307204</td>\n",
              "      <td>0.658692</td>\n",
              "      <td>0.954950</td>\n",
              "      <td>01:41</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>4</td>\n",
              "      <td>1.333064</td>\n",
              "      <td>1.197623</td>\n",
              "      <td>0.713668</td>\n",
              "      <td>0.964368</td>\n",
              "      <td>01:40</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>5</td>\n",
              "      <td>1.297084</td>\n",
              "      <td>1.198449</td>\n",
              "      <td>0.703232</td>\n",
              "      <td>0.964622</td>\n",
              "      <td>01:41</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>6</td>\n",
              "      <td>1.194261</td>\n",
              "      <td>1.120938</td>\n",
              "      <td>0.751591</td>\n",
              "      <td>0.970730</td>\n",
              "      <td>01:41</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>7</td>\n",
              "      <td>1.173396</td>\n",
              "      <td>1.140880</td>\n",
              "      <td>0.729957</td>\n",
              "      <td>0.965895</td>\n",
              "      <td>01:41</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>8</td>\n",
              "      <td>1.131331</td>\n",
              "      <td>1.101983</td>\n",
              "      <td>0.748537</td>\n",
              "      <td>0.973530</td>\n",
              "      <td>01:42</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>9</td>\n",
              "      <td>1.072845</td>\n",
              "      <td>1.046164</td>\n",
              "      <td>0.788241</td>\n",
              "      <td>0.972003</td>\n",
              "      <td>01:42</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>10</td>\n",
              "      <td>1.043721</td>\n",
              "      <td>1.043270</td>\n",
              "      <td>0.783660</td>\n",
              "      <td>0.968185</td>\n",
              "      <td>01:42</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>11</td>\n",
              "      <td>1.019114</td>\n",
              "      <td>1.005509</td>\n",
              "      <td>0.798167</td>\n",
              "      <td>0.977348</td>\n",
              "      <td>01:42</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>12</td>\n",
              "      <td>0.968689</td>\n",
              "      <td>0.998206</td>\n",
              "      <td>0.801476</td>\n",
              "      <td>0.977348</td>\n",
              "      <td>01:42</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>13</td>\n",
              "      <td>0.960192</td>\n",
              "      <td>0.979379</td>\n",
              "      <td>0.806821</td>\n",
              "      <td>0.977602</td>\n",
              "      <td>01:42</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>14</td>\n",
              "      <td>0.888766</td>\n",
              "      <td>0.974381</td>\n",
              "      <td>0.813693</td>\n",
              "      <td>0.976075</td>\n",
              "      <td>01:42</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>15</td>\n",
              "      <td>0.849497</td>\n",
              "      <td>0.976957</td>\n",
              "      <td>0.807585</td>\n",
              "      <td>0.978112</td>\n",
              "      <td>01:42</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>16</td>\n",
              "      <td>0.800239</td>\n",
              "      <td>0.944331</td>\n",
              "      <td>0.820820</td>\n",
              "      <td>0.979893</td>\n",
              "      <td>01:42</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>17</td>\n",
              "      <td>0.763258</td>\n",
              "      <td>0.896260</td>\n",
              "      <td>0.843981</td>\n",
              "      <td>0.978875</td>\n",
              "      <td>01:42</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>18</td>\n",
              "      <td>0.681860</td>\n",
              "      <td>0.879794</td>\n",
              "      <td>0.849326</td>\n",
              "      <td>0.981166</td>\n",
              "      <td>01:42</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>19</td>\n",
              "      <td>0.643791</td>\n",
              "      <td>0.873629</td>\n",
              "      <td>0.852380</td>\n",
              "      <td>0.981675</td>\n",
              "      <td>01:42</td>\n",
              "    </tr>\n",
              "  </tbody>\n",
              "</table>"
            ],
            "text/plain": [
              "<IPython.core.display.HTML object>"
            ]
          },
          "metadata": {
            "tags": []
          }
        },
        {
          "output_type": "stream",
          "text": [
            "data path   /root/.fastai/data/imagewoof2\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "stream",
          "text": [
            ""
          ],
          "name": "stderr"
        },
        {
          "output_type": "stream",
          "text": [
            "Learn path /root/.fastai/data/imagewoof2\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "display_data",
          "data": {
            "text/html": [
              "<table border=\"1\" class=\"dataframe\">\n",
              "  <thead>\n",
              "    <tr style=\"text-align: left;\">\n",
              "      <th>epoch</th>\n",
              "      <th>train_loss</th>\n",
              "      <th>valid_loss</th>\n",
              "      <th>accuracy</th>\n",
              "      <th>top_k_accuracy</th>\n",
              "      <th>time</th>\n",
              "    </tr>\n",
              "  </thead>\n",
              "  <tbody>\n",
              "    <tr>\n",
              "      <td>0</td>\n",
              "      <td>1.940531</td>\n",
              "      <td>1.779968</td>\n",
              "      <td>0.425808</td>\n",
              "      <td>0.880122</td>\n",
              "      <td>01:41</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>1</td>\n",
              "      <td>1.662416</td>\n",
              "      <td>1.593920</td>\n",
              "      <td>0.516925</td>\n",
              "      <td>0.918045</td>\n",
              "      <td>01:40</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>2</td>\n",
              "      <td>1.516636</td>\n",
              "      <td>1.398449</td>\n",
              "      <td>0.612115</td>\n",
              "      <td>0.939934</td>\n",
              "      <td>01:42</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>3</td>\n",
              "      <td>1.407049</td>\n",
              "      <td>1.304912</td>\n",
              "      <td>0.663782</td>\n",
              "      <td>0.955205</td>\n",
              "      <td>01:41</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>4</td>\n",
              "      <td>1.290138</td>\n",
              "      <td>1.187603</td>\n",
              "      <td>0.723848</td>\n",
              "      <td>0.964368</td>\n",
              "      <td>01:41</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>5</td>\n",
              "      <td>1.240288</td>\n",
              "      <td>1.161364</td>\n",
              "      <td>0.729957</td>\n",
              "      <td>0.965895</td>\n",
              "      <td>01:41</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>6</td>\n",
              "      <td>1.180779</td>\n",
              "      <td>1.164560</td>\n",
              "      <td>0.723848</td>\n",
              "      <td>0.963095</td>\n",
              "      <td>01:41</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>7</td>\n",
              "      <td>1.147595</td>\n",
              "      <td>1.087826</td>\n",
              "      <td>0.755409</td>\n",
              "      <td>0.972512</td>\n",
              "      <td>01:41</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>8</td>\n",
              "      <td>1.085288</td>\n",
              "      <td>1.038206</td>\n",
              "      <td>0.780860</td>\n",
              "      <td>0.972767</td>\n",
              "      <td>01:41</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>9</td>\n",
              "      <td>1.043810</td>\n",
              "      <td>1.034537</td>\n",
              "      <td>0.784678</td>\n",
              "      <td>0.974294</td>\n",
              "      <td>01:41</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>10</td>\n",
              "      <td>1.023834</td>\n",
              "      <td>0.998692</td>\n",
              "      <td>0.806058</td>\n",
              "      <td>0.975312</td>\n",
              "      <td>01:41</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>11</td>\n",
              "      <td>1.003258</td>\n",
              "      <td>1.031645</td>\n",
              "      <td>0.775261</td>\n",
              "      <td>0.976075</td>\n",
              "      <td>01:41</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>12</td>\n",
              "      <td>0.963834</td>\n",
              "      <td>0.989480</td>\n",
              "      <td>0.806312</td>\n",
              "      <td>0.972512</td>\n",
              "      <td>01:41</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>13</td>\n",
              "      <td>0.927113</td>\n",
              "      <td>1.013250</td>\n",
              "      <td>0.792314</td>\n",
              "      <td>0.977602</td>\n",
              "      <td>01:41</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>14</td>\n",
              "      <td>0.904774</td>\n",
              "      <td>0.987564</td>\n",
              "      <td>0.807330</td>\n",
              "      <td>0.973785</td>\n",
              "      <td>01:42</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>15</td>\n",
              "      <td>0.854487</td>\n",
              "      <td>0.946134</td>\n",
              "      <td>0.826419</td>\n",
              "      <td>0.976330</td>\n",
              "      <td>01:42</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>16</td>\n",
              "      <td>0.808893</td>\n",
              "      <td>0.930305</td>\n",
              "      <td>0.830746</td>\n",
              "      <td>0.978366</td>\n",
              "      <td>01:43</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>17</td>\n",
              "      <td>0.712423</td>\n",
              "      <td>0.905226</td>\n",
              "      <td>0.842454</td>\n",
              "      <td>0.978366</td>\n",
              "      <td>01:41</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>18</td>\n",
              "      <td>0.681237</td>\n",
              "      <td>0.873459</td>\n",
              "      <td>0.851107</td>\n",
              "      <td>0.981675</td>\n",
              "      <td>01:42</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>19</td>\n",
              "      <td>0.649855</td>\n",
              "      <td>0.877933</td>\n",
              "      <td>0.851616</td>\n",
              "      <td>0.980657</td>\n",
              "      <td>01:42</td>\n",
              "    </tr>\n",
              "  </tbody>\n",
              "</table>"
            ],
            "text/plain": [
              "<IPython.core.display.HTML object>"
            ]
          },
          "metadata": {
            "tags": []
          }
        },
        {
          "output_type": "stream",
          "text": [
            "data path   /root/.fastai/data/imagewoof2\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "stream",
          "text": [
            ""
          ],
          "name": "stderr"
        },
        {
          "output_type": "stream",
          "text": [
            "Learn path /root/.fastai/data/imagewoof2\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "display_data",
          "data": {
            "text/html": [
              "<table border=\"1\" class=\"dataframe\">\n",
              "  <thead>\n",
              "    <tr style=\"text-align: left;\">\n",
              "      <th>epoch</th>\n",
              "      <th>train_loss</th>\n",
              "      <th>valid_loss</th>\n",
              "      <th>accuracy</th>\n",
              "      <th>top_k_accuracy</th>\n",
              "      <th>time</th>\n",
              "    </tr>\n",
              "  </thead>\n",
              "  <tbody>\n",
              "    <tr>\n",
              "      <td>0</td>\n",
              "      <td>1.917777</td>\n",
              "      <td>1.803813</td>\n",
              "      <td>0.408246</td>\n",
              "      <td>0.879868</td>\n",
              "      <td>01:41</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>1</td>\n",
              "      <td>1.695563</td>\n",
              "      <td>1.633969</td>\n",
              "      <td>0.516416</td>\n",
              "      <td>0.919827</td>\n",
              "      <td>01:41</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>2</td>\n",
              "      <td>1.502566</td>\n",
              "      <td>1.369195</td>\n",
              "      <td>0.629422</td>\n",
              "      <td>0.948842</td>\n",
              "      <td>01:41</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>3</td>\n",
              "      <td>1.419279</td>\n",
              "      <td>1.268365</td>\n",
              "      <td>0.683380</td>\n",
              "      <td>0.956732</td>\n",
              "      <td>01:41</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>4</td>\n",
              "      <td>1.298121</td>\n",
              "      <td>1.192557</td>\n",
              "      <td>0.710868</td>\n",
              "      <td>0.967676</td>\n",
              "      <td>01:41</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>5</td>\n",
              "      <td>1.266458</td>\n",
              "      <td>1.158827</td>\n",
              "      <td>0.735811</td>\n",
              "      <td>0.964368</td>\n",
              "      <td>01:41</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>6</td>\n",
              "      <td>1.206960</td>\n",
              "      <td>1.140903</td>\n",
              "      <td>0.731229</td>\n",
              "      <td>0.967931</td>\n",
              "      <td>01:41</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>7</td>\n",
              "      <td>1.141000</td>\n",
              "      <td>1.091973</td>\n",
              "      <td>0.761517</td>\n",
              "      <td>0.974294</td>\n",
              "      <td>01:41</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>8</td>\n",
              "      <td>1.099991</td>\n",
              "      <td>1.094933</td>\n",
              "      <td>0.763808</td>\n",
              "      <td>0.971494</td>\n",
              "      <td>01:42</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>9</td>\n",
              "      <td>1.074507</td>\n",
              "      <td>1.065965</td>\n",
              "      <td>0.770425</td>\n",
              "      <td>0.974548</td>\n",
              "      <td>01:41</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>10</td>\n",
              "      <td>1.021912</td>\n",
              "      <td>1.026982</td>\n",
              "      <td>0.787478</td>\n",
              "      <td>0.975312</td>\n",
              "      <td>01:41</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>11</td>\n",
              "      <td>0.959393</td>\n",
              "      <td>1.015606</td>\n",
              "      <td>0.790532</td>\n",
              "      <td>0.970730</td>\n",
              "      <td>01:42</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>12</td>\n",
              "      <td>0.941038</td>\n",
              "      <td>0.975014</td>\n",
              "      <td>0.806312</td>\n",
              "      <td>0.975057</td>\n",
              "      <td>01:41</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>13</td>\n",
              "      <td>0.929958</td>\n",
              "      <td>0.989770</td>\n",
              "      <td>0.800204</td>\n",
              "      <td>0.976584</td>\n",
              "      <td>01:42</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>14</td>\n",
              "      <td>0.875280</td>\n",
              "      <td>0.965398</td>\n",
              "      <td>0.808094</td>\n",
              "      <td>0.973785</td>\n",
              "      <td>01:42</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>15</td>\n",
              "      <td>0.875902</td>\n",
              "      <td>0.939652</td>\n",
              "      <td>0.824383</td>\n",
              "      <td>0.977348</td>\n",
              "      <td>01:42</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>16</td>\n",
              "      <td>0.774921</td>\n",
              "      <td>0.933063</td>\n",
              "      <td>0.832782</td>\n",
              "      <td>0.977602</td>\n",
              "      <td>01:42</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>17</td>\n",
              "      <td>0.734767</td>\n",
              "      <td>0.897067</td>\n",
              "      <td>0.844490</td>\n",
              "      <td>0.978366</td>\n",
              "      <td>01:42</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>18</td>\n",
              "      <td>0.686703</td>\n",
              "      <td>0.872688</td>\n",
              "      <td>0.852634</td>\n",
              "      <td>0.979130</td>\n",
              "      <td>01:42</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>19</td>\n",
              "      <td>0.671089</td>\n",
              "      <td>0.870245</td>\n",
              "      <td>0.856197</td>\n",
              "      <td>0.979130</td>\n",
              "      <td>01:42</td>\n",
              "    </tr>\n",
              "  </tbody>\n",
              "</table>"
            ],
            "text/plain": [
              "<IPython.core.display.HTML object>"
            ]
          },
          "metadata": {
            "tags": []
          }
        },
        {
          "output_type": "stream",
          "text": [
            "[0.749555, 0.757445, 0.759735, 0.761008, 0.751082, 0.848562, 0.850598, 0.85238, 0.851616, 0.856197] 0.8038177609443664\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "89fKa_ds-H2Y",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 119
        },
        "outputId": "4cf66c57-fa79-42c8-b78c-037580ec51c0"
      },
      "source": [
        "res[5:10], sum(res[5:10])/5"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "([0.8485620021820068,\n",
              "  0.8505980968475342,\n",
              "  0.8523797392845154,\n",
              "  0.8516162037849426,\n",
              "  0.8561974763870239],\n",
              " 0.8518707036972046)"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 22
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "AT9xPqTBVqDa",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 1000
        },
        "outputId": "5f5b9b3c-28c2-454e-f7ce-a12ae4947139"
      },
      "source": [
        "for epochs in [80]:\n",
        "    mixup=0 if epochs<=20 else 0.2\n",
        "    learn = get_learn(model=model, size=128, bs=16, mixup=mixup)\n",
        "    learn.fit_fc(epochs, lr=4e-3, moms=(0.95,0.95), start_pct=0.72)\n",
        "    res += [learn.recorder.metrics[-1][0].item()]\n",
        "print([round(x, 6) for x in res], sum(res)/len(res))"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "data path   /root/.fastai/data/imagewoof2\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "stream",
          "text": [
            ""
          ],
          "name": "stderr"
        },
        {
          "output_type": "stream",
          "text": [
            "Learn path /root/.fastai/data/imagewoof2\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "display_data",
          "data": {
            "text/html": [
              "<table border=\"1\" class=\"dataframe\">\n",
              "  <thead>\n",
              "    <tr style=\"text-align: left;\">\n",
              "      <th>epoch</th>\n",
              "      <th>train_loss</th>\n",
              "      <th>valid_loss</th>\n",
              "      <th>accuracy</th>\n",
              "      <th>top_k_accuracy</th>\n",
              "      <th>time</th>\n",
              "    </tr>\n",
              "  </thead>\n",
              "  <tbody>\n",
              "    <tr>\n",
              "      <td>0</td>\n",
              "      <td>2.121276</td>\n",
              "      <td>1.875138</td>\n",
              "      <td>0.403920</td>\n",
              "      <td>0.846271</td>\n",
              "      <td>01:42</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>1</td>\n",
              "      <td>1.874982</td>\n",
              "      <td>1.709075</td>\n",
              "      <td>0.468567</td>\n",
              "      <td>0.899975</td>\n",
              "      <td>01:41</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>2</td>\n",
              "      <td>1.763815</td>\n",
              "      <td>1.407820</td>\n",
              "      <td>0.609315</td>\n",
              "      <td>0.943497</td>\n",
              "      <td>01:41</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>3</td>\n",
              "      <td>1.622839</td>\n",
              "      <td>1.312779</td>\n",
              "      <td>0.660982</td>\n",
              "      <td>0.953423</td>\n",
              "      <td>01:41</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>4</td>\n",
              "      <td>1.516712</td>\n",
              "      <td>1.246700</td>\n",
              "      <td>0.695597</td>\n",
              "      <td>0.953932</td>\n",
              "      <td>01:40</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>5</td>\n",
              "      <td>1.493167</td>\n",
              "      <td>1.208635</td>\n",
              "      <td>0.692288</td>\n",
              "      <td>0.963095</td>\n",
              "      <td>01:41</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>6</td>\n",
              "      <td>1.453974</td>\n",
              "      <td>1.142897</td>\n",
              "      <td>0.743192</td>\n",
              "      <td>0.960804</td>\n",
              "      <td>01:41</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>7</td>\n",
              "      <td>1.400668</td>\n",
              "      <td>1.074391</td>\n",
              "      <td>0.773479</td>\n",
              "      <td>0.969203</td>\n",
              "      <td>01:41</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>8</td>\n",
              "      <td>1.366435</td>\n",
              "      <td>1.058389</td>\n",
              "      <td>0.776533</td>\n",
              "      <td>0.966658</td>\n",
              "      <td>01:42</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>9</td>\n",
              "      <td>1.311047</td>\n",
              "      <td>1.062316</td>\n",
              "      <td>0.772461</td>\n",
              "      <td>0.971749</td>\n",
              "      <td>01:42</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>10</td>\n",
              "      <td>1.274297</td>\n",
              "      <td>1.012899</td>\n",
              "      <td>0.788496</td>\n",
              "      <td>0.976330</td>\n",
              "      <td>01:41</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>11</td>\n",
              "      <td>1.274095</td>\n",
              "      <td>1.034794</td>\n",
              "      <td>0.783915</td>\n",
              "      <td>0.968185</td>\n",
              "      <td>01:42</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>12</td>\n",
              "      <td>1.232487</td>\n",
              "      <td>1.035141</td>\n",
              "      <td>0.781878</td>\n",
              "      <td>0.973785</td>\n",
              "      <td>01:43</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>13</td>\n",
              "      <td>1.215989</td>\n",
              "      <td>0.987452</td>\n",
              "      <td>0.802749</td>\n",
              "      <td>0.976839</td>\n",
              "      <td>01:42</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>14</td>\n",
              "      <td>1.144557</td>\n",
              "      <td>1.002233</td>\n",
              "      <td>0.796640</td>\n",
              "      <td>0.978875</td>\n",
              "      <td>01:42</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>15</td>\n",
              "      <td>1.190729</td>\n",
              "      <td>0.948480</td>\n",
              "      <td>0.819038</td>\n",
              "      <td>0.983202</td>\n",
              "      <td>01:42</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>16</td>\n",
              "      <td>1.151961</td>\n",
              "      <td>0.956596</td>\n",
              "      <td>0.820820</td>\n",
              "      <td>0.975312</td>\n",
              "      <td>01:42</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>17</td>\n",
              "      <td>1.101583</td>\n",
              "      <td>0.963262</td>\n",
              "      <td>0.809875</td>\n",
              "      <td>0.977602</td>\n",
              "      <td>01:42</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>18</td>\n",
              "      <td>1.158531</td>\n",
              "      <td>0.954425</td>\n",
              "      <td>0.820565</td>\n",
              "      <td>0.976839</td>\n",
              "      <td>01:43</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>19</td>\n",
              "      <td>1.120086</td>\n",
              "      <td>0.934866</td>\n",
              "      <td>0.828710</td>\n",
              "      <td>0.975312</td>\n",
              "      <td>01:42</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>20</td>\n",
              "      <td>1.079439</td>\n",
              "      <td>0.952335</td>\n",
              "      <td>0.820820</td>\n",
              "      <td>0.978621</td>\n",
              "      <td>01:42</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>21</td>\n",
              "      <td>1.083828</td>\n",
              "      <td>0.957045</td>\n",
              "      <td>0.818274</td>\n",
              "      <td>0.977093</td>\n",
              "      <td>01:41</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>22</td>\n",
              "      <td>1.058777</td>\n",
              "      <td>0.934924</td>\n",
              "      <td>0.828710</td>\n",
              "      <td>0.973530</td>\n",
              "      <td>01:41</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>23</td>\n",
              "      <td>1.027703</td>\n",
              "      <td>0.910607</td>\n",
              "      <td>0.839145</td>\n",
              "      <td>0.979384</td>\n",
              "      <td>01:41</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>24</td>\n",
              "      <td>1.065938</td>\n",
              "      <td>0.949213</td>\n",
              "      <td>0.827182</td>\n",
              "      <td>0.973530</td>\n",
              "      <td>01:42</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>25</td>\n",
              "      <td>1.055451</td>\n",
              "      <td>0.926672</td>\n",
              "      <td>0.832018</td>\n",
              "      <td>0.976075</td>\n",
              "      <td>01:41</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>26</td>\n",
              "      <td>1.016625</td>\n",
              "      <td>0.942491</td>\n",
              "      <td>0.829982</td>\n",
              "      <td>0.976075</td>\n",
              "      <td>01:41</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>27</td>\n",
              "      <td>1.041508</td>\n",
              "      <td>0.942612</td>\n",
              "      <td>0.822856</td>\n",
              "      <td>0.977348</td>\n",
              "      <td>01:42</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>28</td>\n",
              "      <td>0.994218</td>\n",
              "      <td>0.911334</td>\n",
              "      <td>0.836091</td>\n",
              "      <td>0.980148</td>\n",
              "      <td>01:42</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>29</td>\n",
              "      <td>1.018248</td>\n",
              "      <td>0.912433</td>\n",
              "      <td>0.839145</td>\n",
              "      <td>0.977857</td>\n",
              "      <td>01:42</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>30</td>\n",
              "      <td>0.965385</td>\n",
              "      <td>0.917235</td>\n",
              "      <td>0.832527</td>\n",
              "      <td>0.974039</td>\n",
              "      <td>01:42</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>31</td>\n",
              "      <td>0.987162</td>\n",
              "      <td>0.897595</td>\n",
              "      <td>0.843981</td>\n",
              "      <td>0.978112</td>\n",
              "      <td>01:42</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>32</td>\n",
              "      <td>1.003360</td>\n",
              "      <td>0.972955</td>\n",
              "      <td>0.814202</td>\n",
              "      <td>0.969967</td>\n",
              "      <td>01:42</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>33</td>\n",
              "      <td>0.973556</td>\n",
              "      <td>0.926874</td>\n",
              "      <td>0.832782</td>\n",
              "      <td>0.976584</td>\n",
              "      <td>01:41</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>34</td>\n",
              "      <td>0.964970</td>\n",
              "      <td>0.915154</td>\n",
              "      <td>0.836600</td>\n",
              "      <td>0.978366</td>\n",
              "      <td>01:42</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>35</td>\n",
              "      <td>0.973608</td>\n",
              "      <td>0.913271</td>\n",
              "      <td>0.837109</td>\n",
              "      <td>0.975566</td>\n",
              "      <td>01:41</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>36</td>\n",
              "      <td>0.974614</td>\n",
              "      <td>0.915623</td>\n",
              "      <td>0.838890</td>\n",
              "      <td>0.973530</td>\n",
              "      <td>01:39</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>37</td>\n",
              "      <td>0.959529</td>\n",
              "      <td>0.904704</td>\n",
              "      <td>0.846526</td>\n",
              "      <td>0.976839</td>\n",
              "      <td>01:38</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>38</td>\n",
              "      <td>0.936213</td>\n",
              "      <td>0.946669</td>\n",
              "      <td>0.825146</td>\n",
              "      <td>0.972512</td>\n",
              "      <td>01:37</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>39</td>\n",
              "      <td>0.963564</td>\n",
              "      <td>0.936740</td>\n",
              "      <td>0.832018</td>\n",
              "      <td>0.971240</td>\n",
              "      <td>01:36</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>40</td>\n",
              "      <td>0.941065</td>\n",
              "      <td>0.910334</td>\n",
              "      <td>0.841945</td>\n",
              "      <td>0.978366</td>\n",
              "      <td>01:37</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>41</td>\n",
              "      <td>0.953965</td>\n",
              "      <td>0.911133</td>\n",
              "      <td>0.835327</td>\n",
              "      <td>0.976839</td>\n",
              "      <td>01:36</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>42</td>\n",
              "      <td>0.924854</td>\n",
              "      <td>0.910510</td>\n",
              "      <td>0.839399</td>\n",
              "      <td>0.974803</td>\n",
              "      <td>01:36</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>43</td>\n",
              "      <td>0.941264</td>\n",
              "      <td>0.927299</td>\n",
              "      <td>0.840417</td>\n",
              "      <td>0.974803</td>\n",
              "      <td>01:36</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>44</td>\n",
              "      <td>0.909746</td>\n",
              "      <td>0.891688</td>\n",
              "      <td>0.844235</td>\n",
              "      <td>0.979130</td>\n",
              "      <td>01:36</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>45</td>\n",
              "      <td>0.951427</td>\n",
              "      <td>0.929121</td>\n",
              "      <td>0.833800</td>\n",
              "      <td>0.977602</td>\n",
              "      <td>01:35</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>46</td>\n",
              "      <td>0.931128</td>\n",
              "      <td>0.908167</td>\n",
              "      <td>0.837618</td>\n",
              "      <td>0.978366</td>\n",
              "      <td>01:36</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>47</td>\n",
              "      <td>0.918996</td>\n",
              "      <td>0.929116</td>\n",
              "      <td>0.829982</td>\n",
              "      <td>0.973785</td>\n",
              "      <td>01:36</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>48</td>\n",
              "      <td>0.917123</td>\n",
              "      <td>0.908193</td>\n",
              "      <td>0.846017</td>\n",
              "      <td>0.976330</td>\n",
              "      <td>01:36</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>49</td>\n",
              "      <td>0.928142</td>\n",
              "      <td>0.916178</td>\n",
              "      <td>0.842454</td>\n",
              "      <td>0.974803</td>\n",
              "      <td>01:36</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>50</td>\n",
              "      <td>0.948084</td>\n",
              "      <td>0.918014</td>\n",
              "      <td>0.833800</td>\n",
              "      <td>0.977348</td>\n",
              "      <td>01:35</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>51</td>\n",
              "      <td>0.915975</td>\n",
              "      <td>0.908193</td>\n",
              "      <td>0.841181</td>\n",
              "      <td>0.976584</td>\n",
              "      <td>01:34</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>52</td>\n",
              "      <td>0.904126</td>\n",
              "      <td>0.925254</td>\n",
              "      <td>0.839654</td>\n",
              "      <td>0.974803</td>\n",
              "      <td>01:34</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>53</td>\n",
              "      <td>0.897841</td>\n",
              "      <td>0.930398</td>\n",
              "      <td>0.829982</td>\n",
              "      <td>0.971749</td>\n",
              "      <td>01:36</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>54</td>\n",
              "      <td>0.897694</td>\n",
              "      <td>0.919171</td>\n",
              "      <td>0.835327</td>\n",
              "      <td>0.975057</td>\n",
              "      <td>01:36</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>55</td>\n",
              "      <td>0.923620</td>\n",
              "      <td>0.887199</td>\n",
              "      <td>0.852380</td>\n",
              "      <td>0.978875</td>\n",
              "      <td>01:35</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>56</td>\n",
              "      <td>0.921115</td>\n",
              "      <td>0.922970</td>\n",
              "      <td>0.839399</td>\n",
              "      <td>0.972003</td>\n",
              "      <td>01:35</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>57</td>\n",
              "      <td>0.923494</td>\n",
              "      <td>0.909773</td>\n",
              "      <td>0.842199</td>\n",
              "      <td>0.975566</td>\n",
              "      <td>01:35</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>58</td>\n",
              "      <td>0.900220</td>\n",
              "      <td>0.900570</td>\n",
              "      <td>0.839908</td>\n",
              "      <td>0.976839</td>\n",
              "      <td>01:35</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>59</td>\n",
              "      <td>0.888581</td>\n",
              "      <td>0.909325</td>\n",
              "      <td>0.838381</td>\n",
              "      <td>0.974548</td>\n",
              "      <td>01:35</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>60</td>\n",
              "      <td>0.898196</td>\n",
              "      <td>0.933845</td>\n",
              "      <td>0.838890</td>\n",
              "      <td>0.968949</td>\n",
              "      <td>01:35</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>61</td>\n",
              "      <td>0.879176</td>\n",
              "      <td>0.912427</td>\n",
              "      <td>0.843981</td>\n",
              "      <td>0.974294</td>\n",
              "      <td>01:34</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>62</td>\n",
              "      <td>0.899951</td>\n",
              "      <td>0.903752</td>\n",
              "      <td>0.851616</td>\n",
              "      <td>0.975057</td>\n",
              "      <td>01:34</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>63</td>\n",
              "      <td>0.878949</td>\n",
              "      <td>0.888744</td>\n",
              "      <td>0.853143</td>\n",
              "      <td>0.978112</td>\n",
              "      <td>01:34</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>64</td>\n",
              "      <td>0.877172</td>\n",
              "      <td>0.879445</td>\n",
              "      <td>0.854161</td>\n",
              "      <td>0.976075</td>\n",
              "      <td>01:34</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>65</td>\n",
              "      <td>0.865555</td>\n",
              "      <td>0.871070</td>\n",
              "      <td>0.854161</td>\n",
              "      <td>0.979384</td>\n",
              "      <td>01:34</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>66</td>\n",
              "      <td>0.868596</td>\n",
              "      <td>0.866503</td>\n",
              "      <td>0.856197</td>\n",
              "      <td>0.980657</td>\n",
              "      <td>01:34</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>67</td>\n",
              "      <td>0.845630</td>\n",
              "      <td>0.860707</td>\n",
              "      <td>0.862306</td>\n",
              "      <td>0.976839</td>\n",
              "      <td>01:34</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>68</td>\n",
              "      <td>0.835569</td>\n",
              "      <td>0.863109</td>\n",
              "      <td>0.857216</td>\n",
              "      <td>0.974803</td>\n",
              "      <td>01:34</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>69</td>\n",
              "      <td>0.851303</td>\n",
              "      <td>0.852923</td>\n",
              "      <td>0.866378</td>\n",
              "      <td>0.977093</td>\n",
              "      <td>01:36</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>70</td>\n",
              "      <td>0.814809</td>\n",
              "      <td>0.841822</td>\n",
              "      <td>0.866124</td>\n",
              "      <td>0.977602</td>\n",
              "      <td>01:35</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>71</td>\n",
              "      <td>0.837100</td>\n",
              "      <td>0.833627</td>\n",
              "      <td>0.874268</td>\n",
              "      <td>0.978875</td>\n",
              "      <td>01:35</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>72</td>\n",
              "      <td>0.815656</td>\n",
              "      <td>0.833418</td>\n",
              "      <td>0.871978</td>\n",
              "      <td>0.978112</td>\n",
              "      <td>01:35</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>73</td>\n",
              "      <td>0.816061</td>\n",
              "      <td>0.830999</td>\n",
              "      <td>0.870196</td>\n",
              "      <td>0.978366</td>\n",
              "      <td>01:35</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>74</td>\n",
              "      <td>0.810153</td>\n",
              "      <td>0.820372</td>\n",
              "      <td>0.872741</td>\n",
              "      <td>0.976330</td>\n",
              "      <td>01:35</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>75</td>\n",
              "      <td>0.821056</td>\n",
              "      <td>0.813914</td>\n",
              "      <td>0.877068</td>\n",
              "      <td>0.978875</td>\n",
              "      <td>01:36</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>76</td>\n",
              "      <td>0.820736</td>\n",
              "      <td>0.811607</td>\n",
              "      <td>0.877068</td>\n",
              "      <td>0.977857</td>\n",
              "      <td>01:35</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>77</td>\n",
              "      <td>0.805219</td>\n",
              "      <td>0.813411</td>\n",
              "      <td>0.878086</td>\n",
              "      <td>0.978875</td>\n",
              "      <td>01:35</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>78</td>\n",
              "      <td>0.793433</td>\n",
              "      <td>0.809495</td>\n",
              "      <td>0.879868</td>\n",
              "      <td>0.978366</td>\n",
              "      <td>01:35</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>79</td>\n",
              "      <td>0.804679</td>\n",
              "      <td>0.813664</td>\n",
              "      <td>0.877068</td>\n",
              "      <td>0.978112</td>\n",
              "      <td>01:35</td>\n",
              "    </tr>\n",
              "  </tbody>\n",
              "</table>"
            ],
            "text/plain": [
              "<IPython.core.display.HTML object>"
            ]
          },
          "metadata": {
            "tags": []
          }
        },
        {
          "output_type": "stream",
          "text": [
            "[0.749555, 0.757445, 0.759735, 0.761008, 0.751082, 0.848562, 0.850598, 0.85238, 0.851616, 0.856197, 0.877068] 0.8104768720540133\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "lEpK16VAcI7a",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 1000
        },
        "outputId": "6acfabb3-f641-4fd7-9cc7-788ce8f78f6a"
      },
      "source": [
        "res = \n",
        "for epochs in [5]*5 + [20]*5 + [80]:\n",
        "    mixup=0 if epochs<=20 else 0.2\n",
        "    learn = get_learn(model=model, size=192, bs=16, mixup=mixup)\n",
        "    learn.fit_fc(epochs, lr=4e-3, moms=(0.95,0.95), start_pct=0.72)\n",
        "    res += [learn.recorder.metrics[-1][0].item()]\n",
        "print([round(x, 6) for x in res], sum(res)/len(res))"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "data path   /root/.fastai/data/imagewoof2\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "stream",
          "text": [
            ""
          ],
          "name": "stderr"
        },
        {
          "output_type": "stream",
          "text": [
            "Learn path /root/.fastai/data/imagewoof2\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "display_data",
          "data": {
            "text/html": [
              "<table border=\"1\" class=\"dataframe\">\n",
              "  <thead>\n",
              "    <tr style=\"text-align: left;\">\n",
              "      <th>epoch</th>\n",
              "      <th>train_loss</th>\n",
              "      <th>valid_loss</th>\n",
              "      <th>accuracy</th>\n",
              "      <th>top_k_accuracy</th>\n",
              "      <th>time</th>\n",
              "    </tr>\n",
              "  </thead>\n",
              "  <tbody>\n",
              "    <tr>\n",
              "      <td>0</td>\n",
              "      <td>1.838385</td>\n",
              "      <td>1.767593</td>\n",
              "      <td>0.438789</td>\n",
              "      <td>0.889794</td>\n",
              "      <td>02:06</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>1</td>\n",
              "      <td>1.589668</td>\n",
              "      <td>1.530492</td>\n",
              "      <td>0.570374</td>\n",
              "      <td>0.931026</td>\n",
              "      <td>02:06</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>2</td>\n",
              "      <td>1.437728</td>\n",
              "      <td>1.305730</td>\n",
              "      <td>0.673199</td>\n",
              "      <td>0.958514</td>\n",
              "      <td>02:06</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>3</td>\n",
              "      <td>1.314570</td>\n",
              "      <td>1.167856</td>\n",
              "      <td>0.736065</td>\n",
              "      <td>0.960550</td>\n",
              "      <td>02:04</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>4</td>\n",
              "      <td>1.071313</td>\n",
              "      <td>1.033118</td>\n",
              "      <td>0.802240</td>\n",
              "      <td>0.979893</td>\n",
              "      <td>02:04</td>\n",
              "    </tr>\n",
              "  </tbody>\n",
              "</table>"
            ],
            "text/plain": [
              "<IPython.core.display.HTML object>"
            ]
          },
          "metadata": {
            "tags": []
          }
        },
        {
          "output_type": "stream",
          "text": [
            "/usr/local/lib/python3.6/dist-packages/torch/nn/functional.py:3000: UserWarning: The default behavior for interpolate/upsample with float scale_factor changed in 1.6.0 to align with other frameworks/libraries, and uses scale_factor directly, instead of relying on the computed output size. If you wish to keep the old behavior, please set recompute_scale_factor=True. See the documentation of nn.Upsample for details. \n",
            "  warnings.warn(\"The default behavior for interpolate/upsample with float scale_factor changed \"\n"
          ],
          "name": "stderr"
        },
        {
          "output_type": "stream",
          "text": [
            "data path   /root/.fastai/data/imagewoof2\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "stream",
          "text": [
            ""
          ],
          "name": "stderr"
        },
        {
          "output_type": "stream",
          "text": [
            "Learn path /root/.fastai/data/imagewoof2\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "display_data",
          "data": {
            "text/html": [
              "<table border=\"1\" class=\"dataframe\">\n",
              "  <thead>\n",
              "    <tr style=\"text-align: left;\">\n",
              "      <th>epoch</th>\n",
              "      <th>train_loss</th>\n",
              "      <th>valid_loss</th>\n",
              "      <th>accuracy</th>\n",
              "      <th>top_k_accuracy</th>\n",
              "      <th>time</th>\n",
              "    </tr>\n",
              "  </thead>\n",
              "  <tbody>\n",
              "    <tr>\n",
              "      <td>0</td>\n",
              "      <td>1.945163</td>\n",
              "      <td>1.798818</td>\n",
              "      <td>0.430644</td>\n",
              "      <td>0.872232</td>\n",
              "      <td>02:04</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>1</td>\n",
              "      <td>1.667129</td>\n",
              "      <td>1.489478</td>\n",
              "      <td>0.575210</td>\n",
              "      <td>0.935098</td>\n",
              "      <td>02:03</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>2</td>\n",
              "      <td>1.451680</td>\n",
              "      <td>1.306634</td>\n",
              "      <td>0.669636</td>\n",
              "      <td>0.957750</td>\n",
              "      <td>02:03</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>3</td>\n",
              "      <td>1.277301</td>\n",
              "      <td>1.211180</td>\n",
              "      <td>0.710613</td>\n",
              "      <td>0.962331</td>\n",
              "      <td>02:03</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>4</td>\n",
              "      <td>1.113092</td>\n",
              "      <td>1.029417</td>\n",
              "      <td>0.797149</td>\n",
              "      <td>0.977348</td>\n",
              "      <td>02:03</td>\n",
              "    </tr>\n",
              "  </tbody>\n",
              "</table>"
            ],
            "text/plain": [
              "<IPython.core.display.HTML object>"
            ]
          },
          "metadata": {
            "tags": []
          }
        },
        {
          "output_type": "stream",
          "text": [
            "/usr/local/lib/python3.6/dist-packages/torch/nn/functional.py:3000: UserWarning: The default behavior for interpolate/upsample with float scale_factor changed in 1.6.0 to align with other frameworks/libraries, and uses scale_factor directly, instead of relying on the computed output size. If you wish to keep the old behavior, please set recompute_scale_factor=True. See the documentation of nn.Upsample for details. \n",
            "  warnings.warn(\"The default behavior for interpolate/upsample with float scale_factor changed \"\n"
          ],
          "name": "stderr"
        },
        {
          "output_type": "stream",
          "text": [
            "data path   /root/.fastai/data/imagewoof2\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "stream",
          "text": [
            ""
          ],
          "name": "stderr"
        },
        {
          "output_type": "stream",
          "text": [
            "Learn path /root/.fastai/data/imagewoof2\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "display_data",
          "data": {
            "text/html": [
              "<table border=\"1\" class=\"dataframe\">\n",
              "  <thead>\n",
              "    <tr style=\"text-align: left;\">\n",
              "      <th>epoch</th>\n",
              "      <th>train_loss</th>\n",
              "      <th>valid_loss</th>\n",
              "      <th>accuracy</th>\n",
              "      <th>top_k_accuracy</th>\n",
              "      <th>time</th>\n",
              "    </tr>\n",
              "  </thead>\n",
              "  <tbody>\n",
              "    <tr>\n",
              "      <td>0</td>\n",
              "      <td>1.912209</td>\n",
              "      <td>1.767511</td>\n",
              "      <td>0.440061</td>\n",
              "      <td>0.888776</td>\n",
              "      <td>02:02</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>1</td>\n",
              "      <td>1.568258</td>\n",
              "      <td>1.459259</td>\n",
              "      <td>0.581064</td>\n",
              "      <td>0.940443</td>\n",
              "      <td>02:03</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>2</td>\n",
              "      <td>1.448780</td>\n",
              "      <td>1.292310</td>\n",
              "      <td>0.664800</td>\n",
              "      <td>0.962077</td>\n",
              "      <td>02:03</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>3</td>\n",
              "      <td>1.296010</td>\n",
              "      <td>1.213690</td>\n",
              "      <td>0.709595</td>\n",
              "      <td>0.965895</td>\n",
              "      <td>02:03</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>4</td>\n",
              "      <td>1.046428</td>\n",
              "      <td>1.031984</td>\n",
              "      <td>0.794350</td>\n",
              "      <td>0.979893</td>\n",
              "      <td>02:03</td>\n",
              "    </tr>\n",
              "  </tbody>\n",
              "</table>"
            ],
            "text/plain": [
              "<IPython.core.display.HTML object>"
            ]
          },
          "metadata": {
            "tags": []
          }
        },
        {
          "output_type": "stream",
          "text": [
            "/usr/local/lib/python3.6/dist-packages/torch/nn/functional.py:3000: UserWarning: The default behavior for interpolate/upsample with float scale_factor changed in 1.6.0 to align with other frameworks/libraries, and uses scale_factor directly, instead of relying on the computed output size. If you wish to keep the old behavior, please set recompute_scale_factor=True. See the documentation of nn.Upsample for details. \n",
            "  warnings.warn(\"The default behavior for interpolate/upsample with float scale_factor changed \"\n"
          ],
          "name": "stderr"
        },
        {
          "output_type": "stream",
          "text": [
            "data path   /root/.fastai/data/imagewoof2\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "stream",
          "text": [
            ""
          ],
          "name": "stderr"
        },
        {
          "output_type": "stream",
          "text": [
            "Learn path /root/.fastai/data/imagewoof2\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "display_data",
          "data": {
            "text/html": [
              "<table border=\"1\" class=\"dataframe\">\n",
              "  <thead>\n",
              "    <tr style=\"text-align: left;\">\n",
              "      <th>epoch</th>\n",
              "      <th>train_loss</th>\n",
              "      <th>valid_loss</th>\n",
              "      <th>accuracy</th>\n",
              "      <th>top_k_accuracy</th>\n",
              "      <th>time</th>\n",
              "    </tr>\n",
              "  </thead>\n",
              "  <tbody>\n",
              "    <tr>\n",
              "      <td>0</td>\n",
              "      <td>1.923147</td>\n",
              "      <td>1.754053</td>\n",
              "      <td>0.452023</td>\n",
              "      <td>0.894375</td>\n",
              "      <td>02:03</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>1</td>\n",
              "      <td>1.687327</td>\n",
              "      <td>1.558394</td>\n",
              "      <td>0.526597</td>\n",
              "      <td>0.932553</td>\n",
              "      <td>02:02</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>2</td>\n",
              "      <td>1.463223</td>\n",
              "      <td>1.375390</td>\n",
              "      <td>0.627641</td>\n",
              "      <td>0.945533</td>\n",
              "      <td>02:03</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>3</td>\n",
              "      <td>1.276294</td>\n",
              "      <td>1.230064</td>\n",
              "      <td>0.707814</td>\n",
              "      <td>0.960550</td>\n",
              "      <td>02:03</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>4</td>\n",
              "      <td>1.104879</td>\n",
              "      <td>1.067524</td>\n",
              "      <td>0.784169</td>\n",
              "      <td>0.975057</td>\n",
              "      <td>02:03</td>\n",
              "    </tr>\n",
              "  </tbody>\n",
              "</table>"
            ],
            "text/plain": [
              "<IPython.core.display.HTML object>"
            ]
          },
          "metadata": {
            "tags": []
          }
        },
        {
          "output_type": "stream",
          "text": [
            ""
          ],
          "name": "stderr"
        },
        {
          "output_type": "stream",
          "text": [
            "data path   /root/.fastai/data/imagewoof2\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "stream",
          "text": [
            ""
          ],
          "name": "stderr"
        },
        {
          "output_type": "stream",
          "text": [
            "Learn path /root/.fastai/data/imagewoof2\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "display_data",
          "data": {
            "text/html": [
              "\n",
              "    <div>\n",
              "        <style>\n",
              "            /* Turns off some styling */\n",
              "            progress {\n",
              "                /* gets rid of default border in Firefox and Opera. */\n",
              "                border: none;\n",
              "                /* Needs to be in here for Safari polyfill so background images work as expected. */\n",
              "                background-size: auto;\n",
              "            }\n",
              "            .progress-bar-interrupted, .progress-bar-interrupted::-webkit-progress-bar {\n",
              "                background: #F44336;\n",
              "            }\n",
              "        </style>\n",
              "      <progress value='2' class='' max='5' style='width:300px; height:20px; vertical-align: middle;'></progress>\n",
              "      40.00% [2/5 04:05<06:08]\n",
              "    </div>\n",
              "    \n",
              "<table border=\"1\" class=\"dataframe\">\n",
              "  <thead>\n",
              "    <tr style=\"text-align: left;\">\n",
              "      <th>epoch</th>\n",
              "      <th>train_loss</th>\n",
              "      <th>valid_loss</th>\n",
              "      <th>accuracy</th>\n",
              "      <th>top_k_accuracy</th>\n",
              "      <th>time</th>\n",
              "    </tr>\n",
              "  </thead>\n",
              "  <tbody>\n",
              "    <tr>\n",
              "      <td>0</td>\n",
              "      <td>1.919139</td>\n",
              "      <td>1.818842</td>\n",
              "      <td>0.415627</td>\n",
              "      <td>0.882158</td>\n",
              "      <td>02:02</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>1</td>\n",
              "      <td>1.640623</td>\n",
              "      <td>1.503787</td>\n",
              "      <td>0.559430</td>\n",
              "      <td>0.930008</td>\n",
              "      <td>02:02</td>\n",
              "    </tr>\n",
              "  </tbody>\n",
              "</table><p>\n",
              "\n",
              "    <div>\n",
              "        <style>\n",
              "            /* Turns off some styling */\n",
              "            progress {\n",
              "                /* gets rid of default border in Firefox and Opera. */\n",
              "                border: none;\n",
              "                /* Needs to be in here for Safari polyfill so background images work as expected. */\n",
              "                background-size: auto;\n",
              "            }\n",
              "            .progress-bar-interrupted, .progress-bar-interrupted::-webkit-progress-bar {\n",
              "                background: #F44336;\n",
              "            }\n",
              "        </style>\n",
              "      <progress value='442' class='' max='564' style='width:300px; height:20px; vertical-align: middle;'></progress>\n",
              "      78.37% [442/564 01:13<00:20 1.4209]\n",
              "    </div>\n",
              "    "
            ],
            "text/plain": [
              "<IPython.core.display.HTML object>"
            ]
          },
          "metadata": {
            "tags": []
          }
        },
        {
          "output_type": "stream",
          "text": [
            ""
          ],
          "name": "stderr"
        },
        {
          "output_type": "stream",
          "text": [
            "Buffered data was truncated after reaching the output size limit."
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "ijKVRRvBUT-5",
        "colab_type": "text"
      },
      "source": [
        "### epochs=5, 5 runs\n",
        "\n",
        "no stride, bs=16, both act and bn"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "-QXtp7Jryx_9",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 1000
        },
        "outputId": "26194210-adc9-4e2a-e7ce-8ea8aded0843"
      },
      "source": [
        "for epochs in [5]*5:\n",
        "    mixup=0 if epochs<=20 else 0.2\n",
        "    learn = get_learn(model=model, size=128, bs=16, mixup=mixup)\n",
        "    learn.fit_fc(epochs, lr=4e-3, moms=(0.95,0.95), start_pct=0.72)\n",
        "    res += [learn.recorder.metrics[-1][0].item()]\n",
        "print([round(x, 6) for x in res], sum(res)/len(res))"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "data path   /root/.fastai/data/imagewoof2\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "stream",
          "text": [
            ""
          ],
          "name": "stderr"
        },
        {
          "output_type": "stream",
          "text": [
            "Learn path /root/.fastai/data/imagewoof2\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "display_data",
          "data": {
            "text/html": [
              "<table border=\"1\" class=\"dataframe\">\n",
              "  <thead>\n",
              "    <tr style=\"text-align: left;\">\n",
              "      <th>epoch</th>\n",
              "      <th>train_loss</th>\n",
              "      <th>valid_loss</th>\n",
              "      <th>accuracy</th>\n",
              "      <th>top_k_accuracy</th>\n",
              "      <th>time</th>\n",
              "    </tr>\n",
              "  </thead>\n",
              "  <tbody>\n",
              "    <tr>\n",
              "      <td>0</td>\n",
              "      <td>1.853887</td>\n",
              "      <td>1.693871</td>\n",
              "      <td>0.475185</td>\n",
              "      <td>0.900229</td>\n",
              "      <td>03:58</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>1</td>\n",
              "      <td>1.597855</td>\n",
              "      <td>1.421064</td>\n",
              "      <td>0.609824</td>\n",
              "      <td>0.941715</td>\n",
              "      <td>03:58</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>2</td>\n",
              "      <td>1.399992</td>\n",
              "      <td>1.231845</td>\n",
              "      <td>0.699160</td>\n",
              "      <td>0.966149</td>\n",
              "      <td>03:58</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>3</td>\n",
              "      <td>1.232522</td>\n",
              "      <td>1.124027</td>\n",
              "      <td>0.758463</td>\n",
              "      <td>0.966404</td>\n",
              "      <td>03:58</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>4</td>\n",
              "      <td>1.014952</td>\n",
              "      <td>1.001750</td>\n",
              "      <td>0.815220</td>\n",
              "      <td>0.981675</td>\n",
              "      <td>03:59</td>\n",
              "    </tr>\n",
              "  </tbody>\n",
              "</table>"
            ],
            "text/plain": [
              "<IPython.core.display.HTML object>"
            ]
          },
          "metadata": {
            "tags": []
          }
        },
        {
          "output_type": "stream",
          "text": [
            ""
          ],
          "name": "stderr"
        },
        {
          "output_type": "stream",
          "text": [
            "data path   /root/.fastai/data/imagewoof2\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "stream",
          "text": [
            ""
          ],
          "name": "stderr"
        },
        {
          "output_type": "stream",
          "text": [
            "Learn path /root/.fastai/data/imagewoof2\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "display_data",
          "data": {
            "text/html": [
              "<table border=\"1\" class=\"dataframe\">\n",
              "  <thead>\n",
              "    <tr style=\"text-align: left;\">\n",
              "      <th>epoch</th>\n",
              "      <th>train_loss</th>\n",
              "      <th>valid_loss</th>\n",
              "      <th>accuracy</th>\n",
              "      <th>top_k_accuracy</th>\n",
              "      <th>time</th>\n",
              "    </tr>\n",
              "  </thead>\n",
              "  <tbody>\n",
              "    <tr>\n",
              "      <td>0</td>\n",
              "      <td>1.829890</td>\n",
              "      <td>1.758396</td>\n",
              "      <td>0.432426</td>\n",
              "      <td>0.883431</td>\n",
              "      <td>03:58</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>1</td>\n",
              "      <td>1.528203</td>\n",
              "      <td>1.495810</td>\n",
              "      <td>0.570120</td>\n",
              "      <td>0.933571</td>\n",
              "      <td>03:58</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>2</td>\n",
              "      <td>1.343309</td>\n",
              "      <td>1.279568</td>\n",
              "      <td>0.694833</td>\n",
              "      <td>0.963095</td>\n",
              "      <td>03:58</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>3</td>\n",
              "      <td>1.255255</td>\n",
              "      <td>1.147663</td>\n",
              "      <td>0.739374</td>\n",
              "      <td>0.970221</td>\n",
              "      <td>03:59</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>4</td>\n",
              "      <td>1.041715</td>\n",
              "      <td>1.004069</td>\n",
              "      <td>0.814966</td>\n",
              "      <td>0.983202</td>\n",
              "      <td>03:59</td>\n",
              "    </tr>\n",
              "  </tbody>\n",
              "</table>"
            ],
            "text/plain": [
              "<IPython.core.display.HTML object>"
            ]
          },
          "metadata": {
            "tags": []
          }
        },
        {
          "output_type": "stream",
          "text": [
            "data path   /root/.fastai/data/imagewoof2\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "stream",
          "text": [
            ""
          ],
          "name": "stderr"
        },
        {
          "output_type": "stream",
          "text": [
            "Learn path /root/.fastai/data/imagewoof2\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "display_data",
          "data": {
            "text/html": [
              "<table border=\"1\" class=\"dataframe\">\n",
              "  <thead>\n",
              "    <tr style=\"text-align: left;\">\n",
              "      <th>epoch</th>\n",
              "      <th>train_loss</th>\n",
              "      <th>valid_loss</th>\n",
              "      <th>accuracy</th>\n",
              "      <th>top_k_accuracy</th>\n",
              "      <th>time</th>\n",
              "    </tr>\n",
              "  </thead>\n",
              "  <tbody>\n",
              "    <tr>\n",
              "      <td>0</td>\n",
              "      <td>1.860970</td>\n",
              "      <td>1.700778</td>\n",
              "      <td>0.465258</td>\n",
              "      <td>0.904556</td>\n",
              "      <td>03:58</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>1</td>\n",
              "      <td>1.538369</td>\n",
              "      <td>1.375021</td>\n",
              "      <td>0.638585</td>\n",
              "      <td>0.944770</td>\n",
              "      <td>03:59</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>2</td>\n",
              "      <td>1.367675</td>\n",
              "      <td>1.224978</td>\n",
              "      <td>0.705014</td>\n",
              "      <td>0.961568</td>\n",
              "      <td>03:59</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>3</td>\n",
              "      <td>1.180613</td>\n",
              "      <td>1.094188</td>\n",
              "      <td>0.775261</td>\n",
              "      <td>0.971749</td>\n",
              "      <td>03:58</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>4</td>\n",
              "      <td>0.987883</td>\n",
              "      <td>0.977362</td>\n",
              "      <td>0.826419</td>\n",
              "      <td>0.981420</td>\n",
              "      <td>03:59</td>\n",
              "    </tr>\n",
              "  </tbody>\n",
              "</table>"
            ],
            "text/plain": [
              "<IPython.core.display.HTML object>"
            ]
          },
          "metadata": {
            "tags": []
          }
        },
        {
          "output_type": "stream",
          "text": [
            "data path   /root/.fastai/data/imagewoof2\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "stream",
          "text": [
            ""
          ],
          "name": "stderr"
        },
        {
          "output_type": "stream",
          "text": [
            "Learn path /root/.fastai/data/imagewoof2\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "display_data",
          "data": {
            "text/html": [
              "<table border=\"1\" class=\"dataframe\">\n",
              "  <thead>\n",
              "    <tr style=\"text-align: left;\">\n",
              "      <th>epoch</th>\n",
              "      <th>train_loss</th>\n",
              "      <th>valid_loss</th>\n",
              "      <th>accuracy</th>\n",
              "      <th>top_k_accuracy</th>\n",
              "      <th>time</th>\n",
              "    </tr>\n",
              "  </thead>\n",
              "  <tbody>\n",
              "    <tr>\n",
              "      <td>0</td>\n",
              "      <td>1.841401</td>\n",
              "      <td>1.678582</td>\n",
              "      <td>0.477475</td>\n",
              "      <td>0.906847</td>\n",
              "      <td>03:58</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>1</td>\n",
              "      <td>1.492334</td>\n",
              "      <td>1.419619</td>\n",
              "      <td>0.627641</td>\n",
              "      <td>0.952914</td>\n",
              "      <td>03:59</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>2</td>\n",
              "      <td>1.348735</td>\n",
              "      <td>1.279629</td>\n",
              "      <td>0.688216</td>\n",
              "      <td>0.963604</td>\n",
              "      <td>03:58</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>3</td>\n",
              "      <td>1.239337</td>\n",
              "      <td>1.132548</td>\n",
              "      <td>0.759735</td>\n",
              "      <td>0.972258</td>\n",
              "      <td>03:59</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>4</td>\n",
              "      <td>1.000990</td>\n",
              "      <td>0.989038</td>\n",
              "      <td>0.825146</td>\n",
              "      <td>0.981675</td>\n",
              "      <td>03:59</td>\n",
              "    </tr>\n",
              "  </tbody>\n",
              "</table>"
            ],
            "text/plain": [
              "<IPython.core.display.HTML object>"
            ]
          },
          "metadata": {
            "tags": []
          }
        },
        {
          "output_type": "stream",
          "text": [
            "data path   /root/.fastai/data/imagewoof2\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "stream",
          "text": [
            ""
          ],
          "name": "stderr"
        },
        {
          "output_type": "stream",
          "text": [
            "Learn path /root/.fastai/data/imagewoof2\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "display_data",
          "data": {
            "text/html": [
              "<table border=\"1\" class=\"dataframe\">\n",
              "  <thead>\n",
              "    <tr style=\"text-align: left;\">\n",
              "      <th>epoch</th>\n",
              "      <th>train_loss</th>\n",
              "      <th>valid_loss</th>\n",
              "      <th>accuracy</th>\n",
              "      <th>top_k_accuracy</th>\n",
              "      <th>time</th>\n",
              "    </tr>\n",
              "  </thead>\n",
              "  <tbody>\n",
              "    <tr>\n",
              "      <td>0</td>\n",
              "      <td>1.778878</td>\n",
              "      <td>1.642192</td>\n",
              "      <td>0.494273</td>\n",
              "      <td>0.912955</td>\n",
              "      <td>03:59</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>1</td>\n",
              "      <td>1.534141</td>\n",
              "      <td>1.428439</td>\n",
              "      <td>0.608552</td>\n",
              "      <td>0.942734</td>\n",
              "      <td>03:58</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>2</td>\n",
              "      <td>1.349526</td>\n",
              "      <td>1.314455</td>\n",
              "      <td>0.668872</td>\n",
              "      <td>0.956477</td>\n",
              "      <td>03:58</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>3</td>\n",
              "      <td>1.221146</td>\n",
              "      <td>1.120293</td>\n",
              "      <td>0.757699</td>\n",
              "      <td>0.973021</td>\n",
              "      <td>03:58</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>4</td>\n",
              "      <td>1.005900</td>\n",
              "      <td>0.979379</td>\n",
              "      <td>0.824128</td>\n",
              "      <td>0.980657</td>\n",
              "      <td>03:59</td>\n",
              "    </tr>\n",
              "  </tbody>\n",
              "</table>"
            ],
            "text/plain": [
              "<IPython.core.display.HTML object>"
            ]
          },
          "metadata": {
            "tags": []
          }
        },
        {
          "output_type": "stream",
          "text": [
            "[0.81522, 0.814966, 0.826419, 0.825146, 0.824128] 0.8211758852005004\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "oildNlm9mSGc",
        "colab_type": "text"
      },
      "source": [
        "### epochs=20, 5 runs\n",
        "\n",
        "no stride, bs=16, both act and bn"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "3311lHhYs7xr",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 1000
        },
        "outputId": "45eabfd1-507b-494d-8903-8130b7d67120"
      },
      "source": [
        "for epochs in [20]*5:\n",
        "    mixup=0 if epochs<=20 else 0.2\n",
        "    learn = get_learn(model=model, size=128, bs=16, mixup=mixup)\n",
        "    learn.fit_fc(epochs, lr=4e-3, moms=(0.95,0.95), start_pct=0.72)\n",
        "    res += [learn.recorder.metrics[-1][0].item()]\n",
        "print([round(x, 6) for x in res], sum(res)/len(res))"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "data path   /root/.fastai/data/imagewoof2\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "stream",
          "text": [
            ""
          ],
          "name": "stderr"
        },
        {
          "output_type": "stream",
          "text": [
            "Learn path /root/.fastai/data/imagewoof2\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "display_data",
          "data": {
            "text/html": [
              "<table border=\"1\" class=\"dataframe\">\n",
              "  <thead>\n",
              "    <tr style=\"text-align: left;\">\n",
              "      <th>epoch</th>\n",
              "      <th>train_loss</th>\n",
              "      <th>valid_loss</th>\n",
              "      <th>accuracy</th>\n",
              "      <th>top_k_accuracy</th>\n",
              "      <th>time</th>\n",
              "    </tr>\n",
              "  </thead>\n",
              "  <tbody>\n",
              "    <tr>\n",
              "      <td>0</td>\n",
              "      <td>1.801173</td>\n",
              "      <td>1.698625</td>\n",
              "      <td>0.481547</td>\n",
              "      <td>0.901502</td>\n",
              "      <td>03:58</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>1</td>\n",
              "      <td>1.522946</td>\n",
              "      <td>1.387382</td>\n",
              "      <td>0.615933</td>\n",
              "      <td>0.950369</td>\n",
              "      <td>03:58</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>2</td>\n",
              "      <td>1.317431</td>\n",
              "      <td>1.288623</td>\n",
              "      <td>0.673963</td>\n",
              "      <td>0.955714</td>\n",
              "      <td>03:58</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>3</td>\n",
              "      <td>1.227739</td>\n",
              "      <td>1.129893</td>\n",
              "      <td>0.753881</td>\n",
              "      <td>0.970221</td>\n",
              "      <td>03:59</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>4</td>\n",
              "      <td>1.124146</td>\n",
              "      <td>1.070666</td>\n",
              "      <td>0.784424</td>\n",
              "      <td>0.976330</td>\n",
              "      <td>03:59</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>5</td>\n",
              "      <td>1.114767</td>\n",
              "      <td>1.045305</td>\n",
              "      <td>0.792823</td>\n",
              "      <td>0.975566</td>\n",
              "      <td>03:59</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>6</td>\n",
              "      <td>1.050665</td>\n",
              "      <td>1.001708</td>\n",
              "      <td>0.811402</td>\n",
              "      <td>0.976584</td>\n",
              "      <td>03:59</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>7</td>\n",
              "      <td>0.972466</td>\n",
              "      <td>0.985769</td>\n",
              "      <td>0.814966</td>\n",
              "      <td>0.983711</td>\n",
              "      <td>03:59</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>8</td>\n",
              "      <td>0.937151</td>\n",
              "      <td>0.992524</td>\n",
              "      <td>0.819292</td>\n",
              "      <td>0.980402</td>\n",
              "      <td>03:59</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>9</td>\n",
              "      <td>0.907048</td>\n",
              "      <td>0.947411</td>\n",
              "      <td>0.835073</td>\n",
              "      <td>0.980148</td>\n",
              "      <td>03:59</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>10</td>\n",
              "      <td>0.862508</td>\n",
              "      <td>0.934270</td>\n",
              "      <td>0.842963</td>\n",
              "      <td>0.981929</td>\n",
              "      <td>03:59</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>11</td>\n",
              "      <td>0.846395</td>\n",
              "      <td>0.909866</td>\n",
              "      <td>0.846780</td>\n",
              "      <td>0.980402</td>\n",
              "      <td>03:59</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>12</td>\n",
              "      <td>0.836732</td>\n",
              "      <td>0.908364</td>\n",
              "      <td>0.840672</td>\n",
              "      <td>0.984220</td>\n",
              "      <td>04:00</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>13</td>\n",
              "      <td>0.825728</td>\n",
              "      <td>0.929190</td>\n",
              "      <td>0.842454</td>\n",
              "      <td>0.981166</td>\n",
              "      <td>03:59</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>14</td>\n",
              "      <td>0.804179</td>\n",
              "      <td>0.901786</td>\n",
              "      <td>0.847544</td>\n",
              "      <td>0.987020</td>\n",
              "      <td>04:00</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>15</td>\n",
              "      <td>0.772480</td>\n",
              "      <td>0.865046</td>\n",
              "      <td>0.860779</td>\n",
              "      <td>0.983456</td>\n",
              "      <td>04:00</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>16</td>\n",
              "      <td>0.702587</td>\n",
              "      <td>0.847529</td>\n",
              "      <td>0.873505</td>\n",
              "      <td>0.984220</td>\n",
              "      <td>04:00</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>17</td>\n",
              "      <td>0.658594</td>\n",
              "      <td>0.818611</td>\n",
              "      <td>0.877577</td>\n",
              "      <td>0.984220</td>\n",
              "      <td>04:00</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>18</td>\n",
              "      <td>0.616297</td>\n",
              "      <td>0.797101</td>\n",
              "      <td>0.887503</td>\n",
              "      <td>0.984983</td>\n",
              "      <td>04:00</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>19</td>\n",
              "      <td>0.597935</td>\n",
              "      <td>0.791300</td>\n",
              "      <td>0.890303</td>\n",
              "      <td>0.984983</td>\n",
              "      <td>04:00</td>\n",
              "    </tr>\n",
              "  </tbody>\n",
              "</table>"
            ],
            "text/plain": [
              "<IPython.core.display.HTML object>"
            ]
          },
          "metadata": {
            "tags": []
          }
        },
        {
          "output_type": "stream",
          "text": [
            ""
          ],
          "name": "stderr"
        },
        {
          "output_type": "stream",
          "text": [
            "data path   /root/.fastai/data/imagewoof2\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "stream",
          "text": [
            ""
          ],
          "name": "stderr"
        },
        {
          "output_type": "stream",
          "text": [
            "Learn path /root/.fastai/data/imagewoof2\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "display_data",
          "data": {
            "text/html": [
              "<table border=\"1\" class=\"dataframe\">\n",
              "  <thead>\n",
              "    <tr style=\"text-align: left;\">\n",
              "      <th>epoch</th>\n",
              "      <th>train_loss</th>\n",
              "      <th>valid_loss</th>\n",
              "      <th>accuracy</th>\n",
              "      <th>top_k_accuracy</th>\n",
              "      <th>time</th>\n",
              "    </tr>\n",
              "  </thead>\n",
              "  <tbody>\n",
              "    <tr>\n",
              "      <td>0</td>\n",
              "      <td>1.835402</td>\n",
              "      <td>1.751453</td>\n",
              "      <td>0.438534</td>\n",
              "      <td>0.890303</td>\n",
              "      <td>03:59</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>1</td>\n",
              "      <td>1.525845</td>\n",
              "      <td>1.374969</td>\n",
              "      <td>0.634004</td>\n",
              "      <td>0.949606</td>\n",
              "      <td>03:59</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>2</td>\n",
              "      <td>1.332278</td>\n",
              "      <td>1.198569</td>\n",
              "      <td>0.716213</td>\n",
              "      <td>0.966404</td>\n",
              "      <td>03:59</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>3</td>\n",
              "      <td>1.251647</td>\n",
              "      <td>1.178239</td>\n",
              "      <td>0.726903</td>\n",
              "      <td>0.965386</td>\n",
              "      <td>03:59</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>4</td>\n",
              "      <td>1.116337</td>\n",
              "      <td>1.071907</td>\n",
              "      <td>0.775515</td>\n",
              "      <td>0.972512</td>\n",
              "      <td>03:59</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>5</td>\n",
              "      <td>1.115612</td>\n",
              "      <td>1.071130</td>\n",
              "      <td>0.774243</td>\n",
              "      <td>0.972512</td>\n",
              "      <td>03:59</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>6</td>\n",
              "      <td>1.070621</td>\n",
              "      <td>1.021311</td>\n",
              "      <td>0.797149</td>\n",
              "      <td>0.980148</td>\n",
              "      <td>03:59</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>7</td>\n",
              "      <td>0.977338</td>\n",
              "      <td>0.975542</td>\n",
              "      <td>0.818020</td>\n",
              "      <td>0.978621</td>\n",
              "      <td>03:59</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>8</td>\n",
              "      <td>0.939719</td>\n",
              "      <td>0.970199</td>\n",
              "      <td>0.813693</td>\n",
              "      <td>0.977602</td>\n",
              "      <td>03:59</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>9</td>\n",
              "      <td>0.923810</td>\n",
              "      <td>0.915703</td>\n",
              "      <td>0.847035</td>\n",
              "      <td>0.984474</td>\n",
              "      <td>03:59</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>10</td>\n",
              "      <td>0.904332</td>\n",
              "      <td>0.917038</td>\n",
              "      <td>0.848817</td>\n",
              "      <td>0.983965</td>\n",
              "      <td>04:00</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>11</td>\n",
              "      <td>0.873339</td>\n",
              "      <td>0.926113</td>\n",
              "      <td>0.834054</td>\n",
              "      <td>0.981675</td>\n",
              "      <td>04:00</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>12</td>\n",
              "      <td>0.838416</td>\n",
              "      <td>0.902836</td>\n",
              "      <td>0.846017</td>\n",
              "      <td>0.982184</td>\n",
              "      <td>04:00</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>13</td>\n",
              "      <td>0.798939</td>\n",
              "      <td>0.890053</td>\n",
              "      <td>0.848307</td>\n",
              "      <td>0.983965</td>\n",
              "      <td>03:59</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>14</td>\n",
              "      <td>0.783665</td>\n",
              "      <td>0.917445</td>\n",
              "      <td>0.841435</td>\n",
              "      <td>0.981166</td>\n",
              "      <td>04:00</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>15</td>\n",
              "      <td>0.758737</td>\n",
              "      <td>0.864056</td>\n",
              "      <td>0.868669</td>\n",
              "      <td>0.983965</td>\n",
              "      <td>04:00</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>16</td>\n",
              "      <td>0.691311</td>\n",
              "      <td>0.832460</td>\n",
              "      <td>0.869687</td>\n",
              "      <td>0.984474</td>\n",
              "      <td>04:00</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>17</td>\n",
              "      <td>0.645873</td>\n",
              "      <td>0.808865</td>\n",
              "      <td>0.882922</td>\n",
              "      <td>0.986511</td>\n",
              "      <td>04:00</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>18</td>\n",
              "      <td>0.615260</td>\n",
              "      <td>0.796047</td>\n",
              "      <td>0.888776</td>\n",
              "      <td>0.986765</td>\n",
              "      <td>04:00</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>19</td>\n",
              "      <td>0.590351</td>\n",
              "      <td>0.792283</td>\n",
              "      <td>0.889030</td>\n",
              "      <td>0.986765</td>\n",
              "      <td>04:00</td>\n",
              "    </tr>\n",
              "  </tbody>\n",
              "</table>"
            ],
            "text/plain": [
              "<IPython.core.display.HTML object>"
            ]
          },
          "metadata": {
            "tags": []
          }
        },
        {
          "output_type": "stream",
          "text": [
            "data path   /root/.fastai/data/imagewoof2\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "stream",
          "text": [
            ""
          ],
          "name": "stderr"
        },
        {
          "output_type": "stream",
          "text": [
            "Learn path /root/.fastai/data/imagewoof2\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "display_data",
          "data": {
            "text/html": [
              "<table border=\"1\" class=\"dataframe\">\n",
              "  <thead>\n",
              "    <tr style=\"text-align: left;\">\n",
              "      <th>epoch</th>\n",
              "      <th>train_loss</th>\n",
              "      <th>valid_loss</th>\n",
              "      <th>accuracy</th>\n",
              "      <th>top_k_accuracy</th>\n",
              "      <th>time</th>\n",
              "    </tr>\n",
              "  </thead>\n",
              "  <tbody>\n",
              "    <tr>\n",
              "      <td>0</td>\n",
              "      <td>1.896753</td>\n",
              "      <td>1.734423</td>\n",
              "      <td>0.455078</td>\n",
              "      <td>0.906847</td>\n",
              "      <td>04:00</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>1</td>\n",
              "      <td>1.556807</td>\n",
              "      <td>1.403181</td>\n",
              "      <td>0.618223</td>\n",
              "      <td>0.946042</td>\n",
              "      <td>03:59</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>2</td>\n",
              "      <td>1.355759</td>\n",
              "      <td>1.273733</td>\n",
              "      <td>0.685925</td>\n",
              "      <td>0.955714</td>\n",
              "      <td>04:00</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>3</td>\n",
              "      <td>1.304199</td>\n",
              "      <td>1.166618</td>\n",
              "      <td>0.747518</td>\n",
              "      <td>0.966658</td>\n",
              "      <td>04:00</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>4</td>\n",
              "      <td>1.133599</td>\n",
              "      <td>1.094989</td>\n",
              "      <td>0.770171</td>\n",
              "      <td>0.970730</td>\n",
              "      <td>04:00</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>5</td>\n",
              "      <td>1.062290</td>\n",
              "      <td>1.069088</td>\n",
              "      <td>0.787732</td>\n",
              "      <td>0.969458</td>\n",
              "      <td>04:00</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>6</td>\n",
              "      <td>1.064758</td>\n",
              "      <td>1.013650</td>\n",
              "      <td>0.802494</td>\n",
              "      <td>0.979639</td>\n",
              "      <td>04:00</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>7</td>\n",
              "      <td>0.984380</td>\n",
              "      <td>0.957368</td>\n",
              "      <td>0.833291</td>\n",
              "      <td>0.982693</td>\n",
              "      <td>04:00</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>8</td>\n",
              "      <td>0.946529</td>\n",
              "      <td>0.996385</td>\n",
              "      <td>0.807076</td>\n",
              "      <td>0.975312</td>\n",
              "      <td>04:00</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>9</td>\n",
              "      <td>0.933666</td>\n",
              "      <td>0.947251</td>\n",
              "      <td>0.823619</td>\n",
              "      <td>0.980911</td>\n",
              "      <td>04:00</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>10</td>\n",
              "      <td>0.902266</td>\n",
              "      <td>0.891157</td>\n",
              "      <td>0.851871</td>\n",
              "      <td>0.982947</td>\n",
              "      <td>04:00</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>11</td>\n",
              "      <td>0.849076</td>\n",
              "      <td>0.916943</td>\n",
              "      <td>0.843217</td>\n",
              "      <td>0.983965</td>\n",
              "      <td>04:00</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>12</td>\n",
              "      <td>0.841957</td>\n",
              "      <td>0.916682</td>\n",
              "      <td>0.845508</td>\n",
              "      <td>0.983965</td>\n",
              "      <td>04:00</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>13</td>\n",
              "      <td>0.826776</td>\n",
              "      <td>0.914348</td>\n",
              "      <td>0.840163</td>\n",
              "      <td>0.981166</td>\n",
              "      <td>04:00</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>14</td>\n",
              "      <td>0.796496</td>\n",
              "      <td>0.887805</td>\n",
              "      <td>0.847035</td>\n",
              "      <td>0.985238</td>\n",
              "      <td>04:00</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>15</td>\n",
              "      <td>0.761636</td>\n",
              "      <td>0.880433</td>\n",
              "      <td>0.860015</td>\n",
              "      <td>0.982947</td>\n",
              "      <td>04:00</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>16</td>\n",
              "      <td>0.717128</td>\n",
              "      <td>0.844493</td>\n",
              "      <td>0.875032</td>\n",
              "      <td>0.984729</td>\n",
              "      <td>04:00</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>17</td>\n",
              "      <td>0.664471</td>\n",
              "      <td>0.824797</td>\n",
              "      <td>0.880631</td>\n",
              "      <td>0.986256</td>\n",
              "      <td>04:00</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>18</td>\n",
              "      <td>0.600854</td>\n",
              "      <td>0.804497</td>\n",
              "      <td>0.885722</td>\n",
              "      <td>0.985747</td>\n",
              "      <td>04:00</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>19</td>\n",
              "      <td>0.591489</td>\n",
              "      <td>0.795916</td>\n",
              "      <td>0.890812</td>\n",
              "      <td>0.986765</td>\n",
              "      <td>04:00</td>\n",
              "    </tr>\n",
              "  </tbody>\n",
              "</table>"
            ],
            "text/plain": [
              "<IPython.core.display.HTML object>"
            ]
          },
          "metadata": {
            "tags": []
          }
        },
        {
          "output_type": "stream",
          "text": [
            "data path   /root/.fastai/data/imagewoof2\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "stream",
          "text": [
            ""
          ],
          "name": "stderr"
        },
        {
          "output_type": "stream",
          "text": [
            "Learn path /root/.fastai/data/imagewoof2\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "display_data",
          "data": {
            "text/html": [
              "<table border=\"1\" class=\"dataframe\">\n",
              "  <thead>\n",
              "    <tr style=\"text-align: left;\">\n",
              "      <th>epoch</th>\n",
              "      <th>train_loss</th>\n",
              "      <th>valid_loss</th>\n",
              "      <th>accuracy</th>\n",
              "      <th>top_k_accuracy</th>\n",
              "      <th>time</th>\n",
              "    </tr>\n",
              "  </thead>\n",
              "  <tbody>\n",
              "    <tr>\n",
              "      <td>0</td>\n",
              "      <td>1.847284</td>\n",
              "      <td>1.775889</td>\n",
              "      <td>0.440825</td>\n",
              "      <td>0.892848</td>\n",
              "      <td>04:00</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>1</td>\n",
              "      <td>1.598621</td>\n",
              "      <td>1.442757</td>\n",
              "      <td>0.600153</td>\n",
              "      <td>0.949096</td>\n",
              "      <td>04:00</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>2</td>\n",
              "      <td>1.376522</td>\n",
              "      <td>1.265322</td>\n",
              "      <td>0.691779</td>\n",
              "      <td>0.956986</td>\n",
              "      <td>04:00</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>3</td>\n",
              "      <td>1.241434</td>\n",
              "      <td>1.172649</td>\n",
              "      <td>0.734793</td>\n",
              "      <td>0.962586</td>\n",
              "      <td>04:00</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>4</td>\n",
              "      <td>1.163452</td>\n",
              "      <td>1.083677</td>\n",
              "      <td>0.776533</td>\n",
              "      <td>0.975566</td>\n",
              "      <td>04:00</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>5</td>\n",
              "      <td>1.123032</td>\n",
              "      <td>1.051371</td>\n",
              "      <td>0.790023</td>\n",
              "      <td>0.975057</td>\n",
              "      <td>04:00</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>6</td>\n",
              "      <td>1.039444</td>\n",
              "      <td>1.001302</td>\n",
              "      <td>0.803767</td>\n",
              "      <td>0.979130</td>\n",
              "      <td>04:00</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>7</td>\n",
              "      <td>1.000067</td>\n",
              "      <td>0.971766</td>\n",
              "      <td>0.819292</td>\n",
              "      <td>0.982184</td>\n",
              "      <td>04:00</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>8</td>\n",
              "      <td>0.963111</td>\n",
              "      <td>1.012285</td>\n",
              "      <td>0.798167</td>\n",
              "      <td>0.975057</td>\n",
              "      <td>03:59</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>9</td>\n",
              "      <td>0.957597</td>\n",
              "      <td>0.925587</td>\n",
              "      <td>0.838381</td>\n",
              "      <td>0.982693</td>\n",
              "      <td>03:59</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>10</td>\n",
              "      <td>0.934487</td>\n",
              "      <td>0.912390</td>\n",
              "      <td>0.840163</td>\n",
              "      <td>0.986002</td>\n",
              "      <td>03:59</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>11</td>\n",
              "      <td>0.895036</td>\n",
              "      <td>0.937479</td>\n",
              "      <td>0.838381</td>\n",
              "      <td>0.980148</td>\n",
              "      <td>03:59</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>12</td>\n",
              "      <td>0.843416</td>\n",
              "      <td>0.913154</td>\n",
              "      <td>0.843726</td>\n",
              "      <td>0.983711</td>\n",
              "      <td>03:59</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>13</td>\n",
              "      <td>0.832364</td>\n",
              "      <td>0.910965</td>\n",
              "      <td>0.847289</td>\n",
              "      <td>0.981420</td>\n",
              "      <td>03:59</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>14</td>\n",
              "      <td>0.800173</td>\n",
              "      <td>0.899453</td>\n",
              "      <td>0.850853</td>\n",
              "      <td>0.981675</td>\n",
              "      <td>04:00</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>15</td>\n",
              "      <td>0.773511</td>\n",
              "      <td>0.871832</td>\n",
              "      <td>0.858743</td>\n",
              "      <td>0.985238</td>\n",
              "      <td>04:00</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>16</td>\n",
              "      <td>0.683821</td>\n",
              "      <td>0.836487</td>\n",
              "      <td>0.868923</td>\n",
              "      <td>0.985492</td>\n",
              "      <td>04:00</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>17</td>\n",
              "      <td>0.653340</td>\n",
              "      <td>0.814082</td>\n",
              "      <td>0.880122</td>\n",
              "      <td>0.987529</td>\n",
              "      <td>04:00</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>18</td>\n",
              "      <td>0.599655</td>\n",
              "      <td>0.795683</td>\n",
              "      <td>0.888521</td>\n",
              "      <td>0.987274</td>\n",
              "      <td>04:00</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>19</td>\n",
              "      <td>0.587210</td>\n",
              "      <td>0.791757</td>\n",
              "      <td>0.890048</td>\n",
              "      <td>0.988038</td>\n",
              "      <td>04:00</td>\n",
              "    </tr>\n",
              "  </tbody>\n",
              "</table>"
            ],
            "text/plain": [
              "<IPython.core.display.HTML object>"
            ]
          },
          "metadata": {
            "tags": []
          }
        },
        {
          "output_type": "stream",
          "text": [
            "data path   /root/.fastai/data/imagewoof2\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "stream",
          "text": [
            ""
          ],
          "name": "stderr"
        },
        {
          "output_type": "stream",
          "text": [
            "Learn path /root/.fastai/data/imagewoof2\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "display_data",
          "data": {
            "text/html": [
              "<table border=\"1\" class=\"dataframe\">\n",
              "  <thead>\n",
              "    <tr style=\"text-align: left;\">\n",
              "      <th>epoch</th>\n",
              "      <th>train_loss</th>\n",
              "      <th>valid_loss</th>\n",
              "      <th>accuracy</th>\n",
              "      <th>top_k_accuracy</th>\n",
              "      <th>time</th>\n",
              "    </tr>\n",
              "  </thead>\n",
              "  <tbody>\n",
              "    <tr>\n",
              "      <td>0</td>\n",
              "      <td>1.863371</td>\n",
              "      <td>1.890844</td>\n",
              "      <td>0.407228</td>\n",
              "      <td>0.878595</td>\n",
              "      <td>03:59</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>1</td>\n",
              "      <td>1.587108</td>\n",
              "      <td>1.468102</td>\n",
              "      <td>0.586663</td>\n",
              "      <td>0.942988</td>\n",
              "      <td>04:00</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>2</td>\n",
              "      <td>1.399580</td>\n",
              "      <td>1.308397</td>\n",
              "      <td>0.661491</td>\n",
              "      <td>0.957496</td>\n",
              "      <td>04:00</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>3</td>\n",
              "      <td>1.273798</td>\n",
              "      <td>1.182499</td>\n",
              "      <td>0.730466</td>\n",
              "      <td>0.966149</td>\n",
              "      <td>03:59</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>4</td>\n",
              "      <td>1.181075</td>\n",
              "      <td>1.145160</td>\n",
              "      <td>0.740392</td>\n",
              "      <td>0.969712</td>\n",
              "      <td>03:59</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>5</td>\n",
              "      <td>1.089572</td>\n",
              "      <td>1.120198</td>\n",
              "      <td>0.752609</td>\n",
              "      <td>0.973021</td>\n",
              "      <td>04:00</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>6</td>\n",
              "      <td>1.049326</td>\n",
              "      <td>1.013096</td>\n",
              "      <td>0.798422</td>\n",
              "      <td>0.979893</td>\n",
              "      <td>03:59</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>7</td>\n",
              "      <td>1.028194</td>\n",
              "      <td>1.039928</td>\n",
              "      <td>0.796131</td>\n",
              "      <td>0.974548</td>\n",
              "      <td>03:59</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>8</td>\n",
              "      <td>0.965961</td>\n",
              "      <td>0.990829</td>\n",
              "      <td>0.812420</td>\n",
              "      <td>0.976075</td>\n",
              "      <td>03:59</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>9</td>\n",
              "      <td>0.950209</td>\n",
              "      <td>0.971756</td>\n",
              "      <td>0.817765</td>\n",
              "      <td>0.980657</td>\n",
              "      <td>04:00</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>10</td>\n",
              "      <td>0.904815</td>\n",
              "      <td>0.919368</td>\n",
              "      <td>0.837872</td>\n",
              "      <td>0.985492</td>\n",
              "      <td>04:00</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>11</td>\n",
              "      <td>0.891796</td>\n",
              "      <td>0.939258</td>\n",
              "      <td>0.830746</td>\n",
              "      <td>0.979384</td>\n",
              "      <td>04:00</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>12</td>\n",
              "      <td>0.867844</td>\n",
              "      <td>0.933804</td>\n",
              "      <td>0.844490</td>\n",
              "      <td>0.982438</td>\n",
              "      <td>04:00</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>13</td>\n",
              "      <td>0.824619</td>\n",
              "      <td>0.910123</td>\n",
              "      <td>0.841435</td>\n",
              "      <td>0.986256</td>\n",
              "      <td>03:59</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>14</td>\n",
              "      <td>0.825055</td>\n",
              "      <td>0.897995</td>\n",
              "      <td>0.848307</td>\n",
              "      <td>0.982947</td>\n",
              "      <td>04:00</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>15</td>\n",
              "      <td>0.756313</td>\n",
              "      <td>0.856059</td>\n",
              "      <td>0.864088</td>\n",
              "      <td>0.983965</td>\n",
              "      <td>03:59</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>16</td>\n",
              "      <td>0.710556</td>\n",
              "      <td>0.845270</td>\n",
              "      <td>0.868160</td>\n",
              "      <td>0.985238</td>\n",
              "      <td>04:00</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>17</td>\n",
              "      <td>0.641457</td>\n",
              "      <td>0.815228</td>\n",
              "      <td>0.880886</td>\n",
              "      <td>0.987274</td>\n",
              "      <td>04:00</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>18</td>\n",
              "      <td>0.611482</td>\n",
              "      <td>0.800236</td>\n",
              "      <td>0.886740</td>\n",
              "      <td>0.986511</td>\n",
              "      <td>04:00</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>19</td>\n",
              "      <td>0.603008</td>\n",
              "      <td>0.797522</td>\n",
              "      <td>0.886231</td>\n",
              "      <td>0.986765</td>\n",
              "      <td>04:00</td>\n",
              "    </tr>\n",
              "  </tbody>\n",
              "</table>"
            ],
            "text/plain": [
              "<IPython.core.display.HTML object>"
            ]
          },
          "metadata": {
            "tags": []
          }
        },
        {
          "output_type": "stream",
          "text": [
            "[0.890303, 0.88903, 0.890812, 0.890048, 0.886231] 0.8892848134040833\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "PzuNBPY1m-LI",
        "colab_type": "text"
      },
      "source": [
        "### epochs=80, 1 run\n",
        "\n",
        "no stride, bs=16, no act, no bn"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "AVm4wdySMLo_",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 1000
        },
        "outputId": "65904793-d724-43c0-cd67-92861c8dbc69"
      },
      "source": [
        "for epochs in [80]:\n",
        "    mixup=0 if epochs<=20 else 0.2\n",
        "    learn = get_learn(model=model, size=128, bs=16, mixup=mixup)\n",
        "    learn.fit_fc(epochs, lr=4e-3, moms=(0.95,0.95), start_pct=0.72)\n",
        "    res += [learn.recorder.metrics[-1][0].item()]\n",
        "print([round(x, 6) for x in res], sum(res)/len(res))"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "data path   /root/.fastai/data/imagewoof2\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "stream",
          "text": [
            ""
          ],
          "name": "stderr"
        },
        {
          "output_type": "stream",
          "text": [
            "Learn path /root/.fastai/data/imagewoof2\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "display_data",
          "data": {
            "text/html": [
              "<table border=\"1\" class=\"dataframe\">\n",
              "  <thead>\n",
              "    <tr style=\"text-align: left;\">\n",
              "      <th>epoch</th>\n",
              "      <th>train_loss</th>\n",
              "      <th>valid_loss</th>\n",
              "      <th>accuracy</th>\n",
              "      <th>top_k_accuracy</th>\n",
              "      <th>time</th>\n",
              "    </tr>\n",
              "  </thead>\n",
              "  <tbody>\n",
              "    <tr>\n",
              "      <td>0</td>\n",
              "      <td>2.038643</td>\n",
              "      <td>1.808398</td>\n",
              "      <td>0.406974</td>\n",
              "      <td>0.869687</td>\n",
              "      <td>03:37</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>1</td>\n",
              "      <td>1.809108</td>\n",
              "      <td>1.668090</td>\n",
              "      <td>0.477984</td>\n",
              "      <td>0.896157</td>\n",
              "      <td>03:37</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>2</td>\n",
              "      <td>1.644071</td>\n",
              "      <td>1.380295</td>\n",
              "      <td>0.630695</td>\n",
              "      <td>0.947315</td>\n",
              "      <td>03:36</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>3</td>\n",
              "      <td>1.501918</td>\n",
              "      <td>1.250324</td>\n",
              "      <td>0.687961</td>\n",
              "      <td>0.955714</td>\n",
              "      <td>03:36</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>4</td>\n",
              "      <td>1.419764</td>\n",
              "      <td>1.168158</td>\n",
              "      <td>0.727157</td>\n",
              "      <td>0.965386</td>\n",
              "      <td>03:36</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>5</td>\n",
              "      <td>1.392190</td>\n",
              "      <td>1.110393</td>\n",
              "      <td>0.756936</td>\n",
              "      <td>0.974294</td>\n",
              "      <td>03:36</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>6</td>\n",
              "      <td>1.315425</td>\n",
              "      <td>1.051459</td>\n",
              "      <td>0.786969</td>\n",
              "      <td>0.975312</td>\n",
              "      <td>03:36</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>7</td>\n",
              "      <td>1.300064</td>\n",
              "      <td>1.040518</td>\n",
              "      <td>0.793841</td>\n",
              "      <td>0.974294</td>\n",
              "      <td>03:36</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>8</td>\n",
              "      <td>1.273304</td>\n",
              "      <td>0.996594</td>\n",
              "      <td>0.807330</td>\n",
              "      <td>0.979384</td>\n",
              "      <td>03:36</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>9</td>\n",
              "      <td>1.212061</td>\n",
              "      <td>0.999831</td>\n",
              "      <td>0.796386</td>\n",
              "      <td>0.976330</td>\n",
              "      <td>03:36</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>10</td>\n",
              "      <td>1.195637</td>\n",
              "      <td>0.998080</td>\n",
              "      <td>0.806567</td>\n",
              "      <td>0.977602</td>\n",
              "      <td>03:37</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>11</td>\n",
              "      <td>1.207942</td>\n",
              "      <td>0.966098</td>\n",
              "      <td>0.814202</td>\n",
              "      <td>0.976075</td>\n",
              "      <td>03:36</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>12</td>\n",
              "      <td>1.144307</td>\n",
              "      <td>0.960030</td>\n",
              "      <td>0.825655</td>\n",
              "      <td>0.980402</td>\n",
              "      <td>03:36</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>13</td>\n",
              "      <td>1.108176</td>\n",
              "      <td>0.927151</td>\n",
              "      <td>0.840163</td>\n",
              "      <td>0.979384</td>\n",
              "      <td>03:36</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>14</td>\n",
              "      <td>1.117250</td>\n",
              "      <td>0.913778</td>\n",
              "      <td>0.837109</td>\n",
              "      <td>0.982438</td>\n",
              "      <td>03:37</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>15</td>\n",
              "      <td>1.091446</td>\n",
              "      <td>0.928094</td>\n",
              "      <td>0.833036</td>\n",
              "      <td>0.983965</td>\n",
              "      <td>03:37</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>16</td>\n",
              "      <td>1.072486</td>\n",
              "      <td>0.930012</td>\n",
              "      <td>0.829982</td>\n",
              "      <td>0.983456</td>\n",
              "      <td>03:36</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>17</td>\n",
              "      <td>1.021997</td>\n",
              "      <td>0.901157</td>\n",
              "      <td>0.847035</td>\n",
              "      <td>0.980402</td>\n",
              "      <td>03:36</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>18</td>\n",
              "      <td>1.041776</td>\n",
              "      <td>0.884699</td>\n",
              "      <td>0.851871</td>\n",
              "      <td>0.987274</td>\n",
              "      <td>03:37</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>19</td>\n",
              "      <td>1.033800</td>\n",
              "      <td>0.881888</td>\n",
              "      <td>0.858234</td>\n",
              "      <td>0.982693</td>\n",
              "      <td>03:36</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>20</td>\n",
              "      <td>1.025444</td>\n",
              "      <td>0.889501</td>\n",
              "      <td>0.857979</td>\n",
              "      <td>0.979130</td>\n",
              "      <td>03:36</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>21</td>\n",
              "      <td>1.030351</td>\n",
              "      <td>0.902533</td>\n",
              "      <td>0.849835</td>\n",
              "      <td>0.982438</td>\n",
              "      <td>03:36</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>22</td>\n",
              "      <td>1.019658</td>\n",
              "      <td>0.883949</td>\n",
              "      <td>0.857216</td>\n",
              "      <td>0.979384</td>\n",
              "      <td>03:36</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>23</td>\n",
              "      <td>0.986560</td>\n",
              "      <td>0.866152</td>\n",
              "      <td>0.865360</td>\n",
              "      <td>0.983711</td>\n",
              "      <td>03:37</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>24</td>\n",
              "      <td>0.990973</td>\n",
              "      <td>0.862831</td>\n",
              "      <td>0.862306</td>\n",
              "      <td>0.982693</td>\n",
              "      <td>03:37</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>25</td>\n",
              "      <td>0.992763</td>\n",
              "      <td>0.869275</td>\n",
              "      <td>0.866887</td>\n",
              "      <td>0.980402</td>\n",
              "      <td>03:36</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>26</td>\n",
              "      <td>0.982956</td>\n",
              "      <td>0.894781</td>\n",
              "      <td>0.850598</td>\n",
              "      <td>0.980402</td>\n",
              "      <td>03:36</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>27</td>\n",
              "      <td>0.978850</td>\n",
              "      <td>0.865467</td>\n",
              "      <td>0.866378</td>\n",
              "      <td>0.983965</td>\n",
              "      <td>03:36</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>28</td>\n",
              "      <td>0.973253</td>\n",
              "      <td>0.855045</td>\n",
              "      <td>0.863324</td>\n",
              "      <td>0.983202</td>\n",
              "      <td>03:36</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>29</td>\n",
              "      <td>0.968829</td>\n",
              "      <td>0.881253</td>\n",
              "      <td>0.857725</td>\n",
              "      <td>0.984220</td>\n",
              "      <td>03:37</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>30</td>\n",
              "      <td>0.951756</td>\n",
              "      <td>0.868973</td>\n",
              "      <td>0.860524</td>\n",
              "      <td>0.979130</td>\n",
              "      <td>03:36</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>31</td>\n",
              "      <td>0.954131</td>\n",
              "      <td>0.903666</td>\n",
              "      <td>0.852634</td>\n",
              "      <td>0.977602</td>\n",
              "      <td>03:36</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>32</td>\n",
              "      <td>0.925291</td>\n",
              "      <td>0.865668</td>\n",
              "      <td>0.867905</td>\n",
              "      <td>0.980402</td>\n",
              "      <td>03:36</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>33</td>\n",
              "      <td>0.931839</td>\n",
              "      <td>0.855966</td>\n",
              "      <td>0.873759</td>\n",
              "      <td>0.979130</td>\n",
              "      <td>03:37</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>34</td>\n",
              "      <td>0.928767</td>\n",
              "      <td>0.851197</td>\n",
              "      <td>0.866124</td>\n",
              "      <td>0.982693</td>\n",
              "      <td>03:37</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>35</td>\n",
              "      <td>0.957517</td>\n",
              "      <td>0.862988</td>\n",
              "      <td>0.866378</td>\n",
              "      <td>0.979893</td>\n",
              "      <td>03:37</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>36</td>\n",
              "      <td>0.935205</td>\n",
              "      <td>0.849383</td>\n",
              "      <td>0.876559</td>\n",
              "      <td>0.980657</td>\n",
              "      <td>03:37</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>37</td>\n",
              "      <td>0.933243</td>\n",
              "      <td>0.863000</td>\n",
              "      <td>0.864597</td>\n",
              "      <td>0.982438</td>\n",
              "      <td>03:38</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>38</td>\n",
              "      <td>0.912338</td>\n",
              "      <td>0.863855</td>\n",
              "      <td>0.865869</td>\n",
              "      <td>0.980402</td>\n",
              "      <td>03:38</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>39</td>\n",
              "      <td>0.910788</td>\n",
              "      <td>0.869880</td>\n",
              "      <td>0.865106</td>\n",
              "      <td>0.982693</td>\n",
              "      <td>03:38</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>40</td>\n",
              "      <td>0.919146</td>\n",
              "      <td>0.835650</td>\n",
              "      <td>0.873759</td>\n",
              "      <td>0.985492</td>\n",
              "      <td>03:38</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>41</td>\n",
              "      <td>0.908551</td>\n",
              "      <td>0.845868</td>\n",
              "      <td>0.869178</td>\n",
              "      <td>0.981166</td>\n",
              "      <td>03:38</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>42</td>\n",
              "      <td>0.893263</td>\n",
              "      <td>0.867527</td>\n",
              "      <td>0.858997</td>\n",
              "      <td>0.980148</td>\n",
              "      <td>03:37</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>43</td>\n",
              "      <td>0.908673</td>\n",
              "      <td>0.856650</td>\n",
              "      <td>0.861797</td>\n",
              "      <td>0.982184</td>\n",
              "      <td>03:37</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>44</td>\n",
              "      <td>0.924303</td>\n",
              "      <td>0.823537</td>\n",
              "      <td>0.878595</td>\n",
              "      <td>0.986256</td>\n",
              "      <td>03:37</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>45</td>\n",
              "      <td>0.913768</td>\n",
              "      <td>0.835905</td>\n",
              "      <td>0.870705</td>\n",
              "      <td>0.983711</td>\n",
              "      <td>03:37</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>46</td>\n",
              "      <td>0.871591</td>\n",
              "      <td>0.851400</td>\n",
              "      <td>0.869178</td>\n",
              "      <td>0.974548</td>\n",
              "      <td>03:37</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>47</td>\n",
              "      <td>0.894567</td>\n",
              "      <td>0.850240</td>\n",
              "      <td>0.865360</td>\n",
              "      <td>0.982947</td>\n",
              "      <td>03:37</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>48</td>\n",
              "      <td>0.878179</td>\n",
              "      <td>0.848819</td>\n",
              "      <td>0.869432</td>\n",
              "      <td>0.979130</td>\n",
              "      <td>03:37</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>49</td>\n",
              "      <td>0.886955</td>\n",
              "      <td>0.884202</td>\n",
              "      <td>0.859506</td>\n",
              "      <td>0.978112</td>\n",
              "      <td>03:37</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>50</td>\n",
              "      <td>0.909272</td>\n",
              "      <td>0.856140</td>\n",
              "      <td>0.865615</td>\n",
              "      <td>0.976839</td>\n",
              "      <td>03:37</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>51</td>\n",
              "      <td>0.906145</td>\n",
              "      <td>0.855819</td>\n",
              "      <td>0.865615</td>\n",
              "      <td>0.981420</td>\n",
              "      <td>03:37</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>52</td>\n",
              "      <td>0.892749</td>\n",
              "      <td>0.840323</td>\n",
              "      <td>0.872232</td>\n",
              "      <td>0.980148</td>\n",
              "      <td>03:37</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>53</td>\n",
              "      <td>0.876586</td>\n",
              "      <td>0.862631</td>\n",
              "      <td>0.865615</td>\n",
              "      <td>0.981675</td>\n",
              "      <td>03:37</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>54</td>\n",
              "      <td>0.880847</td>\n",
              "      <td>0.831383</td>\n",
              "      <td>0.878086</td>\n",
              "      <td>0.980657</td>\n",
              "      <td>03:37</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>55</td>\n",
              "      <td>0.884440</td>\n",
              "      <td>0.856027</td>\n",
              "      <td>0.864851</td>\n",
              "      <td>0.979130</td>\n",
              "      <td>03:37</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>56</td>\n",
              "      <td>0.867854</td>\n",
              "      <td>0.867476</td>\n",
              "      <td>0.859761</td>\n",
              "      <td>0.975821</td>\n",
              "      <td>03:37</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>57</td>\n",
              "      <td>0.882595</td>\n",
              "      <td>0.845180</td>\n",
              "      <td>0.868923</td>\n",
              "      <td>0.981166</td>\n",
              "      <td>03:37</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>58</td>\n",
              "      <td>0.851779</td>\n",
              "      <td>0.864196</td>\n",
              "      <td>0.862815</td>\n",
              "      <td>0.978621</td>\n",
              "      <td>03:37</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>59</td>\n",
              "      <td>0.874769</td>\n",
              "      <td>0.841450</td>\n",
              "      <td>0.872232</td>\n",
              "      <td>0.979130</td>\n",
              "      <td>03:37</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>60</td>\n",
              "      <td>0.878660</td>\n",
              "      <td>0.823630</td>\n",
              "      <td>0.883431</td>\n",
              "      <td>0.981166</td>\n",
              "      <td>03:37</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>61</td>\n",
              "      <td>0.846501</td>\n",
              "      <td>0.837324</td>\n",
              "      <td>0.872232</td>\n",
              "      <td>0.980402</td>\n",
              "      <td>03:37</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>62</td>\n",
              "      <td>0.861082</td>\n",
              "      <td>0.833016</td>\n",
              "      <td>0.874268</td>\n",
              "      <td>0.982693</td>\n",
              "      <td>03:37</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>63</td>\n",
              "      <td>0.854230</td>\n",
              "      <td>0.841103</td>\n",
              "      <td>0.865106</td>\n",
              "      <td>0.980402</td>\n",
              "      <td>03:37</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>64</td>\n",
              "      <td>0.838734</td>\n",
              "      <td>0.819540</td>\n",
              "      <td>0.883431</td>\n",
              "      <td>0.980402</td>\n",
              "      <td>03:37</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>65</td>\n",
              "      <td>0.840468</td>\n",
              "      <td>0.819259</td>\n",
              "      <td>0.880886</td>\n",
              "      <td>0.978875</td>\n",
              "      <td>03:37</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>66</td>\n",
              "      <td>0.850581</td>\n",
              "      <td>0.814191</td>\n",
              "      <td>0.880886</td>\n",
              "      <td>0.982438</td>\n",
              "      <td>03:37</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>67</td>\n",
              "      <td>0.844305</td>\n",
              "      <td>0.807524</td>\n",
              "      <td>0.887503</td>\n",
              "      <td>0.982184</td>\n",
              "      <td>03:37</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>68</td>\n",
              "      <td>0.827369</td>\n",
              "      <td>0.807841</td>\n",
              "      <td>0.884194</td>\n",
              "      <td>0.982693</td>\n",
              "      <td>03:37</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>69</td>\n",
              "      <td>0.824204</td>\n",
              "      <td>0.804468</td>\n",
              "      <td>0.881904</td>\n",
              "      <td>0.982184</td>\n",
              "      <td>03:37</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>70</td>\n",
              "      <td>0.814428</td>\n",
              "      <td>0.787049</td>\n",
              "      <td>0.890812</td>\n",
              "      <td>0.985238</td>\n",
              "      <td>03:37</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>71</td>\n",
              "      <td>0.801715</td>\n",
              "      <td>0.785311</td>\n",
              "      <td>0.893612</td>\n",
              "      <td>0.981929</td>\n",
              "      <td>03:37</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>72</td>\n",
              "      <td>0.791389</td>\n",
              "      <td>0.784581</td>\n",
              "      <td>0.896157</td>\n",
              "      <td>0.983711</td>\n",
              "      <td>03:37</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>73</td>\n",
              "      <td>0.793671</td>\n",
              "      <td>0.779489</td>\n",
              "      <td>0.894884</td>\n",
              "      <td>0.983202</td>\n",
              "      <td>03:37</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>74</td>\n",
              "      <td>0.801416</td>\n",
              "      <td>0.772580</td>\n",
              "      <td>0.899211</td>\n",
              "      <td>0.984983</td>\n",
              "      <td>03:38</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>75</td>\n",
              "      <td>0.783075</td>\n",
              "      <td>0.769843</td>\n",
              "      <td>0.902520</td>\n",
              "      <td>0.984474</td>\n",
              "      <td>03:38</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>76</td>\n",
              "      <td>0.799178</td>\n",
              "      <td>0.771232</td>\n",
              "      <td>0.901247</td>\n",
              "      <td>0.986256</td>\n",
              "      <td>03:39</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>77</td>\n",
              "      <td>0.795923</td>\n",
              "      <td>0.768372</td>\n",
              "      <td>0.903029</td>\n",
              "      <td>0.985492</td>\n",
              "      <td>03:38</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>78</td>\n",
              "      <td>0.783563</td>\n",
              "      <td>0.768873</td>\n",
              "      <td>0.903283</td>\n",
              "      <td>0.984983</td>\n",
              "      <td>03:37</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>79</td>\n",
              "      <td>0.806377</td>\n",
              "      <td>0.766163</td>\n",
              "      <td>0.901502</td>\n",
              "      <td>0.985747</td>\n",
              "      <td>03:37</td>\n",
              "    </tr>\n",
              "  </tbody>\n",
              "</table>"
            ],
            "text/plain": [
              "<IPython.core.display.HTML object>"
            ]
          },
          "metadata": {
            "tags": []
          }
        },
        {
          "output_type": "stream",
          "text": [
            "[0.901502] 0.9015016555786133\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "n6V1ZjAwOcoI",
        "colab_type": "text"
      },
      "source": [
        "\n",
        "no stride, bs=16, both act and bn"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "DKjuVmoPOmC8",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 1000
        },
        "outputId": "0f770033-c640-4d26-83ef-351306592d87"
      },
      "source": [
        "for epochs in [80]:\n",
        "    mixup=0 if epochs<=20 else 0.2\n",
        "    learn = get_learn(model=model, size=128, bs=16, mixup=mixup)\n",
        "    learn.fit_fc(epochs, lr=4e-3, moms=(0.95,0.95), start_pct=0.72)\n",
        "    res += [learn.recorder.metrics[-1][0].item()]\n",
        "print([round(x, 6) for x in res], sum(res)/len(res))"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "data path   /root/.fastai/data/imagewoof2\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "stream",
          "text": [
            ""
          ],
          "name": "stderr"
        },
        {
          "output_type": "stream",
          "text": [
            "Learn path /root/.fastai/data/imagewoof2\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "display_data",
          "data": {
            "text/html": [
              "<table border=\"1\" class=\"dataframe\">\n",
              "  <thead>\n",
              "    <tr style=\"text-align: left;\">\n",
              "      <th>epoch</th>\n",
              "      <th>train_loss</th>\n",
              "      <th>valid_loss</th>\n",
              "      <th>accuracy</th>\n",
              "      <th>top_k_accuracy</th>\n",
              "      <th>time</th>\n",
              "    </tr>\n",
              "  </thead>\n",
              "  <tbody>\n",
              "    <tr>\n",
              "      <td>0</td>\n",
              "      <td>2.009993</td>\n",
              "      <td>1.844606</td>\n",
              "      <td>0.379486</td>\n",
              "      <td>0.869432</td>\n",
              "      <td>03:59</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>1</td>\n",
              "      <td>1.781528</td>\n",
              "      <td>1.531310</td>\n",
              "      <td>0.555358</td>\n",
              "      <td>0.932807</td>\n",
              "      <td>03:58</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>2</td>\n",
              "      <td>1.613264</td>\n",
              "      <td>1.351579</td>\n",
              "      <td>0.640621</td>\n",
              "      <td>0.952405</td>\n",
              "      <td>03:59</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>3</td>\n",
              "      <td>1.511871</td>\n",
              "      <td>1.250022</td>\n",
              "      <td>0.708323</td>\n",
              "      <td>0.963095</td>\n",
              "      <td>03:59</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>4</td>\n",
              "      <td>1.425421</td>\n",
              "      <td>1.116004</td>\n",
              "      <td>0.758463</td>\n",
              "      <td>0.969458</td>\n",
              "      <td>03:59</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>5</td>\n",
              "      <td>1.373275</td>\n",
              "      <td>1.092476</td>\n",
              "      <td>0.767880</td>\n",
              "      <td>0.970730</td>\n",
              "      <td>03:58</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>6</td>\n",
              "      <td>1.300701</td>\n",
              "      <td>1.038407</td>\n",
              "      <td>0.790023</td>\n",
              "      <td>0.975312</td>\n",
              "      <td>03:58</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>7</td>\n",
              "      <td>1.282876</td>\n",
              "      <td>1.046970</td>\n",
              "      <td>0.786969</td>\n",
              "      <td>0.976330</td>\n",
              "      <td>03:58</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>8</td>\n",
              "      <td>1.268418</td>\n",
              "      <td>0.952731</td>\n",
              "      <td>0.826164</td>\n",
              "      <td>0.981675</td>\n",
              "      <td>03:58</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>9</td>\n",
              "      <td>1.188836</td>\n",
              "      <td>0.957474</td>\n",
              "      <td>0.818020</td>\n",
              "      <td>0.982438</td>\n",
              "      <td>03:58</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>10</td>\n",
              "      <td>1.194832</td>\n",
              "      <td>0.947065</td>\n",
              "      <td>0.827946</td>\n",
              "      <td>0.980657</td>\n",
              "      <td>03:58</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>11</td>\n",
              "      <td>1.139472</td>\n",
              "      <td>0.921041</td>\n",
              "      <td>0.843472</td>\n",
              "      <td>0.979384</td>\n",
              "      <td>03:59</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>12</td>\n",
              "      <td>1.146593</td>\n",
              "      <td>0.940159</td>\n",
              "      <td>0.831764</td>\n",
              "      <td>0.980911</td>\n",
              "      <td>03:59</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>13</td>\n",
              "      <td>1.132867</td>\n",
              "      <td>0.892814</td>\n",
              "      <td>0.847798</td>\n",
              "      <td>0.982947</td>\n",
              "      <td>03:59</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>14</td>\n",
              "      <td>1.069067</td>\n",
              "      <td>0.921424</td>\n",
              "      <td>0.840672</td>\n",
              "      <td>0.979893</td>\n",
              "      <td>03:59</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>15</td>\n",
              "      <td>1.081249</td>\n",
              "      <td>0.914862</td>\n",
              "      <td>0.843472</td>\n",
              "      <td>0.984729</td>\n",
              "      <td>03:58</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>16</td>\n",
              "      <td>1.080963</td>\n",
              "      <td>0.879225</td>\n",
              "      <td>0.856707</td>\n",
              "      <td>0.982693</td>\n",
              "      <td>03:59</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>17</td>\n",
              "      <td>1.038699</td>\n",
              "      <td>0.893135</td>\n",
              "      <td>0.854161</td>\n",
              "      <td>0.978621</td>\n",
              "      <td>03:58</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>18</td>\n",
              "      <td>1.054133</td>\n",
              "      <td>0.857359</td>\n",
              "      <td>0.864088</td>\n",
              "      <td>0.983965</td>\n",
              "      <td>03:59</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>19</td>\n",
              "      <td>1.034871</td>\n",
              "      <td>0.853931</td>\n",
              "      <td>0.863324</td>\n",
              "      <td>0.983202</td>\n",
              "      <td>03:59</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>20</td>\n",
              "      <td>1.037785</td>\n",
              "      <td>0.871390</td>\n",
              "      <td>0.863069</td>\n",
              "      <td>0.981166</td>\n",
              "      <td>03:59</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>21</td>\n",
              "      <td>1.014986</td>\n",
              "      <td>0.901717</td>\n",
              "      <td>0.844744</td>\n",
              "      <td>0.982693</td>\n",
              "      <td>03:58</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>22</td>\n",
              "      <td>0.972291</td>\n",
              "      <td>0.900924</td>\n",
              "      <td>0.843726</td>\n",
              "      <td>0.979639</td>\n",
              "      <td>03:59</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>23</td>\n",
              "      <td>1.000840</td>\n",
              "      <td>0.861638</td>\n",
              "      <td>0.864342</td>\n",
              "      <td>0.983711</td>\n",
              "      <td>03:58</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>24</td>\n",
              "      <td>0.954553</td>\n",
              "      <td>0.856538</td>\n",
              "      <td>0.866378</td>\n",
              "      <td>0.979384</td>\n",
              "      <td>03:59</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>25</td>\n",
              "      <td>0.974585</td>\n",
              "      <td>0.876854</td>\n",
              "      <td>0.859506</td>\n",
              "      <td>0.979639</td>\n",
              "      <td>03:59</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>26</td>\n",
              "      <td>0.942681</td>\n",
              "      <td>0.864773</td>\n",
              "      <td>0.862306</td>\n",
              "      <td>0.979130</td>\n",
              "      <td>03:59</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>27</td>\n",
              "      <td>0.949488</td>\n",
              "      <td>0.833433</td>\n",
              "      <td>0.871723</td>\n",
              "      <td>0.984474</td>\n",
              "      <td>03:59</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>28</td>\n",
              "      <td>0.964553</td>\n",
              "      <td>0.872052</td>\n",
              "      <td>0.857979</td>\n",
              "      <td>0.980148</td>\n",
              "      <td>03:59</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>29</td>\n",
              "      <td>0.939988</td>\n",
              "      <td>0.832514</td>\n",
              "      <td>0.874014</td>\n",
              "      <td>0.984729</td>\n",
              "      <td>03:59</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>30</td>\n",
              "      <td>0.939877</td>\n",
              "      <td>0.822606</td>\n",
              "      <td>0.875286</td>\n",
              "      <td>0.983711</td>\n",
              "      <td>03:59</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>31</td>\n",
              "      <td>0.935709</td>\n",
              "      <td>0.854631</td>\n",
              "      <td>0.859506</td>\n",
              "      <td>0.983202</td>\n",
              "      <td>03:59</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>32</td>\n",
              "      <td>0.905323</td>\n",
              "      <td>0.846684</td>\n",
              "      <td>0.869432</td>\n",
              "      <td>0.983456</td>\n",
              "      <td>03:59</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>33</td>\n",
              "      <td>0.941534</td>\n",
              "      <td>0.852837</td>\n",
              "      <td>0.868160</td>\n",
              "      <td>0.981420</td>\n",
              "      <td>03:59</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>34</td>\n",
              "      <td>0.905570</td>\n",
              "      <td>0.876611</td>\n",
              "      <td>0.851107</td>\n",
              "      <td>0.980148</td>\n",
              "      <td>03:59</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>35</td>\n",
              "      <td>0.907686</td>\n",
              "      <td>0.834467</td>\n",
              "      <td>0.874777</td>\n",
              "      <td>0.981420</td>\n",
              "      <td>03:58</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>36</td>\n",
              "      <td>0.917099</td>\n",
              "      <td>0.857840</td>\n",
              "      <td>0.860779</td>\n",
              "      <td>0.978112</td>\n",
              "      <td>03:59</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>37</td>\n",
              "      <td>0.913313</td>\n",
              "      <td>0.886712</td>\n",
              "      <td>0.849580</td>\n",
              "      <td>0.978112</td>\n",
              "      <td>03:59</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>38</td>\n",
              "      <td>0.922003</td>\n",
              "      <td>0.827664</td>\n",
              "      <td>0.875286</td>\n",
              "      <td>0.981675</td>\n",
              "      <td>03:59</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>39</td>\n",
              "      <td>0.941270</td>\n",
              "      <td>0.847456</td>\n",
              "      <td>0.870960</td>\n",
              "      <td>0.980657</td>\n",
              "      <td>03:59</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>40</td>\n",
              "      <td>0.921508</td>\n",
              "      <td>0.839233</td>\n",
              "      <td>0.868414</td>\n",
              "      <td>0.981929</td>\n",
              "      <td>03:59</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>41</td>\n",
              "      <td>0.890218</td>\n",
              "      <td>0.848964</td>\n",
              "      <td>0.871214</td>\n",
              "      <td>0.980657</td>\n",
              "      <td>03:59</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>42</td>\n",
              "      <td>0.907331</td>\n",
              "      <td>0.829025</td>\n",
              "      <td>0.874523</td>\n",
              "      <td>0.983965</td>\n",
              "      <td>03:59</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>43</td>\n",
              "      <td>0.921427</td>\n",
              "      <td>0.831767</td>\n",
              "      <td>0.871978</td>\n",
              "      <td>0.982693</td>\n",
              "      <td>03:59</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>44</td>\n",
              "      <td>0.894037</td>\n",
              "      <td>0.833437</td>\n",
              "      <td>0.870196</td>\n",
              "      <td>0.978366</td>\n",
              "      <td>03:59</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>45</td>\n",
              "      <td>0.894140</td>\n",
              "      <td>0.846124</td>\n",
              "      <td>0.866887</td>\n",
              "      <td>0.980402</td>\n",
              "      <td>03:59</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>46</td>\n",
              "      <td>0.886184</td>\n",
              "      <td>0.848293</td>\n",
              "      <td>0.870960</td>\n",
              "      <td>0.980148</td>\n",
              "      <td>03:59</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>47</td>\n",
              "      <td>0.910699</td>\n",
              "      <td>0.838514</td>\n",
              "      <td>0.875286</td>\n",
              "      <td>0.980911</td>\n",
              "      <td>03:59</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>48</td>\n",
              "      <td>0.888323</td>\n",
              "      <td>0.837410</td>\n",
              "      <td>0.869941</td>\n",
              "      <td>0.978112</td>\n",
              "      <td>03:59</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>49</td>\n",
              "      <td>0.883143</td>\n",
              "      <td>0.838820</td>\n",
              "      <td>0.872232</td>\n",
              "      <td>0.980911</td>\n",
              "      <td>03:59</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>50</td>\n",
              "      <td>0.882449</td>\n",
              "      <td>0.862656</td>\n",
              "      <td>0.863833</td>\n",
              "      <td>0.977093</td>\n",
              "      <td>03:58</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>51</td>\n",
              "      <td>0.865685</td>\n",
              "      <td>0.856141</td>\n",
              "      <td>0.869178</td>\n",
              "      <td>0.976839</td>\n",
              "      <td>03:59</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>52</td>\n",
              "      <td>0.902737</td>\n",
              "      <td>0.832649</td>\n",
              "      <td>0.869178</td>\n",
              "      <td>0.978112</td>\n",
              "      <td>03:59</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>53</td>\n",
              "      <td>0.903280</td>\n",
              "      <td>0.835770</td>\n",
              "      <td>0.871469</td>\n",
              "      <td>0.979639</td>\n",
              "      <td>03:59</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>54</td>\n",
              "      <td>0.876613</td>\n",
              "      <td>0.838167</td>\n",
              "      <td>0.868160</td>\n",
              "      <td>0.982693</td>\n",
              "      <td>03:59</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>55</td>\n",
              "      <td>0.867719</td>\n",
              "      <td>0.824574</td>\n",
              "      <td>0.878086</td>\n",
              "      <td>0.983711</td>\n",
              "      <td>03:59</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>56</td>\n",
              "      <td>0.870489</td>\n",
              "      <td>0.869379</td>\n",
              "      <td>0.865869</td>\n",
              "      <td>0.975566</td>\n",
              "      <td>03:59</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>57</td>\n",
              "      <td>0.861141</td>\n",
              "      <td>0.847735</td>\n",
              "      <td>0.868160</td>\n",
              "      <td>0.980148</td>\n",
              "      <td>03:59</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>58</td>\n",
              "      <td>0.881976</td>\n",
              "      <td>0.865704</td>\n",
              "      <td>0.866633</td>\n",
              "      <td>0.980657</td>\n",
              "      <td>03:59</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>59</td>\n",
              "      <td>0.876335</td>\n",
              "      <td>0.817991</td>\n",
              "      <td>0.877322</td>\n",
              "      <td>0.982184</td>\n",
              "      <td>03:59</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>60</td>\n",
              "      <td>0.870608</td>\n",
              "      <td>0.844145</td>\n",
              "      <td>0.869432</td>\n",
              "      <td>0.977348</td>\n",
              "      <td>03:59</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>61</td>\n",
              "      <td>0.860329</td>\n",
              "      <td>0.835160</td>\n",
              "      <td>0.870960</td>\n",
              "      <td>0.980911</td>\n",
              "      <td>03:59</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>62</td>\n",
              "      <td>0.862334</td>\n",
              "      <td>0.833193</td>\n",
              "      <td>0.871723</td>\n",
              "      <td>0.981420</td>\n",
              "      <td>03:59</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>63</td>\n",
              "      <td>0.859001</td>\n",
              "      <td>0.847550</td>\n",
              "      <td>0.866887</td>\n",
              "      <td>0.976075</td>\n",
              "      <td>03:59</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>64</td>\n",
              "      <td>0.861975</td>\n",
              "      <td>0.820567</td>\n",
              "      <td>0.880122</td>\n",
              "      <td>0.979893</td>\n",
              "      <td>03:59</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>65</td>\n",
              "      <td>0.829951</td>\n",
              "      <td>0.812817</td>\n",
              "      <td>0.879868</td>\n",
              "      <td>0.979893</td>\n",
              "      <td>03:59</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>66</td>\n",
              "      <td>0.836649</td>\n",
              "      <td>0.812997</td>\n",
              "      <td>0.881649</td>\n",
              "      <td>0.979893</td>\n",
              "      <td>03:59</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>67</td>\n",
              "      <td>0.817062</td>\n",
              "      <td>0.800844</td>\n",
              "      <td>0.888521</td>\n",
              "      <td>0.978366</td>\n",
              "      <td>03:59</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>68</td>\n",
              "      <td>0.825846</td>\n",
              "      <td>0.803504</td>\n",
              "      <td>0.887503</td>\n",
              "      <td>0.982184</td>\n",
              "      <td>03:59</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>69</td>\n",
              "      <td>0.786300</td>\n",
              "      <td>0.786168</td>\n",
              "      <td>0.892594</td>\n",
              "      <td>0.983965</td>\n",
              "      <td>03:59</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>70</td>\n",
              "      <td>0.812994</td>\n",
              "      <td>0.786667</td>\n",
              "      <td>0.894121</td>\n",
              "      <td>0.980148</td>\n",
              "      <td>04:00</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>71</td>\n",
              "      <td>0.804539</td>\n",
              "      <td>0.780230</td>\n",
              "      <td>0.894121</td>\n",
              "      <td>0.983456</td>\n",
              "      <td>04:00</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>72</td>\n",
              "      <td>0.816607</td>\n",
              "      <td>0.772015</td>\n",
              "      <td>0.896411</td>\n",
              "      <td>0.980911</td>\n",
              "      <td>04:01</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>73</td>\n",
              "      <td>0.774862</td>\n",
              "      <td>0.768502</td>\n",
              "      <td>0.899720</td>\n",
              "      <td>0.982438</td>\n",
              "      <td>04:01</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>74</td>\n",
              "      <td>0.796895</td>\n",
              "      <td>0.769346</td>\n",
              "      <td>0.895902</td>\n",
              "      <td>0.984220</td>\n",
              "      <td>04:00</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>75</td>\n",
              "      <td>0.802842</td>\n",
              "      <td>0.764322</td>\n",
              "      <td>0.898702</td>\n",
              "      <td>0.983456</td>\n",
              "      <td>04:00</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>76</td>\n",
              "      <td>0.767676</td>\n",
              "      <td>0.764318</td>\n",
              "      <td>0.900229</td>\n",
              "      <td>0.983456</td>\n",
              "      <td>04:01</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>77</td>\n",
              "      <td>0.797181</td>\n",
              "      <td>0.760752</td>\n",
              "      <td>0.901247</td>\n",
              "      <td>0.982438</td>\n",
              "      <td>04:01</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>78</td>\n",
              "      <td>0.770356</td>\n",
              "      <td>0.759512</td>\n",
              "      <td>0.900993</td>\n",
              "      <td>0.982693</td>\n",
              "      <td>04:02</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>79</td>\n",
              "      <td>0.800453</td>\n",
              "      <td>0.759778</td>\n",
              "      <td>0.902011</td>\n",
              "      <td>0.981675</td>\n",
              "      <td>04:01</td>\n",
              "    </tr>\n",
              "  </tbody>\n",
              "</table>"
            ],
            "text/plain": [
              "<IPython.core.display.HTML object>"
            ]
          },
          "metadata": {
            "tags": []
          }
        },
        {
          "output_type": "stream",
          "text": [
            "[0.902011] 0.9020106792449951\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "m1an0AYCjLOF",
        "colab_type": "text"
      },
      "source": [
        "## size=192"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "5JqE5dMgr02P",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 1000
        },
        "outputId": "073283a9-8af1-41a3-e6d5-b91b2a5c947e"
      },
      "source": [
        "dm = 2\n",
        "res = dict()\n",
        "for e in [5,80]: # [5]*5 + [20]*5 + [80]:\n",
        "  mixup=0 if e<=20 else 0.2\n",
        "  learn = get_learn(model=model, size=192, bs=16, mixup=mixup)\n",
        "  learn.fit_fc(e, lr=4e-3, moms=(0.95,0.95), start_pct=0.72)\n",
        "  acc = learn.recorder.metrics[-1][0].item()\n",
        "  if e in res:\n",
        "    res[e] += [acc]\n",
        "  else:\n",
        "    res[e] = [acc]\n",
        "  print('{} epochs: {} ({} runs)'.format(e, sum(res[e])/len(res[e]), len(res[e])))\n",
        "print('depth multiplier={}'.format(dm), {e: sum(res[e])/len(res[e]) for e in res})"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "data path   /root/.fastai/data/imagewoof2\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "stream",
          "text": [
            "/usr/local/lib/python3.6/dist-packages/torch/nn/functional.py:3000: UserWarning: The default behavior for interpolate/upsample with float scale_factor changed in 1.6.0 to align with other frameworks/libraries, and uses scale_factor directly, instead of relying on the computed output size. If you wish to keep the old behavior, please set recompute_scale_factor=True. See the documentation of nn.Upsample for details. \n",
            "  warnings.warn(\"The default behavior for interpolate/upsample with float scale_factor changed \"\n",
            "/usr/local/lib/python3.6/dist-packages/torch/nn/functional.py:3000: UserWarning: The default behavior for interpolate/upsample with float scale_factor changed in 1.6.0 to align with other frameworks/libraries, and uses scale_factor directly, instead of relying on the computed output size. If you wish to keep the old behavior, please set recompute_scale_factor=True. See the documentation of nn.Upsample for details. \n",
            "  warnings.warn(\"The default behavior for interpolate/upsample with float scale_factor changed \"\n",
            "/usr/local/lib/python3.6/dist-packages/torch/nn/functional.py:3000: UserWarning: The default behavior for interpolate/upsample with float scale_factor changed in 1.6.0 to align with other frameworks/libraries, and uses scale_factor directly, instead of relying on the computed output size. If you wish to keep the old behavior, please set recompute_scale_factor=True. See the documentation of nn.Upsample for details. \n",
            "  warnings.warn(\"The default behavior for interpolate/upsample with float scale_factor changed \"\n",
            "/usr/local/lib/python3.6/dist-packages/torch/nn/functional.py:3000: UserWarning: The default behavior for interpolate/upsample with float scale_factor changed in 1.6.0 to align with other frameworks/libraries, and uses scale_factor directly, instead of relying on the computed output size. If you wish to keep the old behavior, please set recompute_scale_factor=True. See the documentation of nn.Upsample for details. \n",
            "  warnings.warn(\"The default behavior for interpolate/upsample with float scale_factor changed \"\n"
          ],
          "name": "stderr"
        },
        {
          "output_type": "stream",
          "text": [
            "Learn path /root/.fastai/data/imagewoof2\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "display_data",
          "data": {
            "text/html": [
              "<table border=\"1\" class=\"dataframe\">\n",
              "  <thead>\n",
              "    <tr style=\"text-align: left;\">\n",
              "      <th>epoch</th>\n",
              "      <th>train_loss</th>\n",
              "      <th>valid_loss</th>\n",
              "      <th>accuracy</th>\n",
              "      <th>top_k_accuracy</th>\n",
              "      <th>time</th>\n",
              "    </tr>\n",
              "  </thead>\n",
              "  <tbody>\n",
              "    <tr>\n",
              "      <td>0</td>\n",
              "      <td>1.851319</td>\n",
              "      <td>1.850964</td>\n",
              "      <td>0.425554</td>\n",
              "      <td>0.870705</td>\n",
              "      <td>05:23</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>1</td>\n",
              "      <td>1.571606</td>\n",
              "      <td>1.486467</td>\n",
              "      <td>0.570629</td>\n",
              "      <td>0.937898</td>\n",
              "      <td>05:23</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>2</td>\n",
              "      <td>1.385945</td>\n",
              "      <td>1.273577</td>\n",
              "      <td>0.688979</td>\n",
              "      <td>0.962331</td>\n",
              "      <td>05:23</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>3</td>\n",
              "      <td>1.262263</td>\n",
              "      <td>1.225133</td>\n",
              "      <td>0.715195</td>\n",
              "      <td>0.961313</td>\n",
              "      <td>05:24</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>4</td>\n",
              "      <td>1.065164</td>\n",
              "      <td>1.007206</td>\n",
              "      <td>0.820565</td>\n",
              "      <td>0.979130</td>\n",
              "      <td>05:23</td>\n",
              "    </tr>\n",
              "  </tbody>\n",
              "</table>"
            ],
            "text/plain": [
              "<IPython.core.display.HTML object>"
            ]
          },
          "metadata": {
            "tags": []
          }
        },
        {
          "output_type": "stream",
          "text": [
            "/usr/local/lib/python3.6/dist-packages/torch/nn/functional.py:3000: UserWarning: The default behavior for interpolate/upsample with float scale_factor changed in 1.6.0 to align with other frameworks/libraries, and uses scale_factor directly, instead of relying on the computed output size. If you wish to keep the old behavior, please set recompute_scale_factor=True. See the documentation of nn.Upsample for details. \n",
            "  warnings.warn(\"The default behavior for interpolate/upsample with float scale_factor changed \"\n",
            "/usr/local/lib/python3.6/dist-packages/imagenette_experiments/train_utils.py:150: UserWarning: This overload of addcmul_ is deprecated:\n",
            "\taddcmul_(Number value, Tensor tensor1, Tensor tensor2)\n",
            "Consider using one of the following signatures instead:\n",
            "\taddcmul_(Tensor tensor1, Tensor tensor2, *, Number value) (Triggered internally at  /pytorch/torch/csrc/utils/python_arg_parser.cpp:766.)\n",
            "  exp_avg_sq.mul_(beta2).addcmul_(1 - beta2, grad, grad)\n"
          ],
          "name": "stderr"
        },
        {
          "output_type": "stream",
          "text": [
            "5 epochs: 0.8205650448799133 (1 runs)\n",
            "data path   /root/.fastai/data/imagewoof2\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "stream",
          "text": [
            "/usr/local/lib/python3.6/dist-packages/torch/nn/functional.py:3000: UserWarning: The default behavior for interpolate/upsample with float scale_factor changed in 1.6.0 to align with other frameworks/libraries, and uses scale_factor directly, instead of relying on the computed output size. If you wish to keep the old behavior, please set recompute_scale_factor=True. See the documentation of nn.Upsample for details. \n",
            "  warnings.warn(\"The default behavior for interpolate/upsample with float scale_factor changed \"\n",
            "/usr/local/lib/python3.6/dist-packages/torch/nn/functional.py:3000: UserWarning: The default behavior for interpolate/upsample with float scale_factor changed in 1.6.0 to align with other frameworks/libraries, and uses scale_factor directly, instead of relying on the computed output size. If you wish to keep the old behavior, please set recompute_scale_factor=True. See the documentation of nn.Upsample for details. \n",
            "  warnings.warn(\"The default behavior for interpolate/upsample with float scale_factor changed \"\n",
            "/usr/local/lib/python3.6/dist-packages/torch/nn/functional.py:3000: UserWarning: The default behavior for interpolate/upsample with float scale_factor changed in 1.6.0 to align with other frameworks/libraries, and uses scale_factor directly, instead of relying on the computed output size. If you wish to keep the old behavior, please set recompute_scale_factor=True. See the documentation of nn.Upsample for details. \n",
            "  warnings.warn(\"The default behavior for interpolate/upsample with float scale_factor changed \"\n"
          ],
          "name": "stderr"
        },
        {
          "output_type": "stream",
          "text": [
            "Learn path /root/.fastai/data/imagewoof2\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "display_data",
          "data": {
            "text/html": [
              "\n",
              "    <div>\n",
              "        <style>\n",
              "            /* Turns off some styling */\n",
              "            progress {\n",
              "                /* gets rid of default border in Firefox and Opera. */\n",
              "                border: none;\n",
              "                /* Needs to be in here for Safari polyfill so background images work as expected. */\n",
              "                background-size: auto;\n",
              "            }\n",
              "            .progress-bar-interrupted, .progress-bar-interrupted::-webkit-progress-bar {\n",
              "                background: #F44336;\n",
              "            }\n",
              "        </style>\n",
              "      <progress value='13' class='' max='80' style='width:300px; height:20px; vertical-align: middle;'></progress>\n",
              "      16.25% [13/80 1:10:12<6:01:53]\n",
              "    </div>\n",
              "    \n",
              "<table border=\"1\" class=\"dataframe\">\n",
              "  <thead>\n",
              "    <tr style=\"text-align: left;\">\n",
              "      <th>epoch</th>\n",
              "      <th>train_loss</th>\n",
              "      <th>valid_loss</th>\n",
              "      <th>accuracy</th>\n",
              "      <th>top_k_accuracy</th>\n",
              "      <th>time</th>\n",
              "    </tr>\n",
              "  </thead>\n",
              "  <tbody>\n",
              "    <tr>\n",
              "      <td>0</td>\n",
              "      <td>2.027898</td>\n",
              "      <td>1.831583</td>\n",
              "      <td>0.411810</td>\n",
              "      <td>0.867905</td>\n",
              "      <td>05:23</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>1</td>\n",
              "      <td>1.802953</td>\n",
              "      <td>1.590741</td>\n",
              "      <td>0.527106</td>\n",
              "      <td>0.921863</td>\n",
              "      <td>05:24</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>2</td>\n",
              "      <td>1.652540</td>\n",
              "      <td>1.390942</td>\n",
              "      <td>0.631713</td>\n",
              "      <td>0.950369</td>\n",
              "      <td>05:23</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>3</td>\n",
              "      <td>1.564943</td>\n",
              "      <td>1.250905</td>\n",
              "      <td>0.686689</td>\n",
              "      <td>0.966149</td>\n",
              "      <td>05:23</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>4</td>\n",
              "      <td>1.454332</td>\n",
              "      <td>1.172994</td>\n",
              "      <td>0.723085</td>\n",
              "      <td>0.965640</td>\n",
              "      <td>05:24</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>5</td>\n",
              "      <td>1.399670</td>\n",
              "      <td>1.112360</td>\n",
              "      <td>0.764571</td>\n",
              "      <td>0.969712</td>\n",
              "      <td>05:24</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>6</td>\n",
              "      <td>1.347242</td>\n",
              "      <td>1.047297</td>\n",
              "      <td>0.795368</td>\n",
              "      <td>0.978875</td>\n",
              "      <td>05:24</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>7</td>\n",
              "      <td>1.278714</td>\n",
              "      <td>1.071562</td>\n",
              "      <td>0.785696</td>\n",
              "      <td>0.975057</td>\n",
              "      <td>05:24</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>8</td>\n",
              "      <td>1.276993</td>\n",
              "      <td>0.990722</td>\n",
              "      <td>0.815984</td>\n",
              "      <td>0.978621</td>\n",
              "      <td>05:24</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>9</td>\n",
              "      <td>1.236502</td>\n",
              "      <td>1.004172</td>\n",
              "      <td>0.809112</td>\n",
              "      <td>0.980657</td>\n",
              "      <td>05:23</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>10</td>\n",
              "      <td>1.225452</td>\n",
              "      <td>0.942010</td>\n",
              "      <td>0.842454</td>\n",
              "      <td>0.979893</td>\n",
              "      <td>05:24</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>11</td>\n",
              "      <td>1.199796</td>\n",
              "      <td>0.921057</td>\n",
              "      <td>0.843217</td>\n",
              "      <td>0.982184</td>\n",
              "      <td>05:24</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>12</td>\n",
              "      <td>1.160879</td>\n",
              "      <td>0.941142</td>\n",
              "      <td>0.834054</td>\n",
              "      <td>0.982947</td>\n",
              "      <td>05:24</td>\n",
              "    </tr>\n",
              "  </tbody>\n",
              "</table><p>\n",
              "\n",
              "    <div>\n",
              "        <style>\n",
              "            /* Turns off some styling */\n",
              "            progress {\n",
              "                /* gets rid of default border in Firefox and Opera. */\n",
              "                border: none;\n",
              "                /* Needs to be in here for Safari polyfill so background images work as expected. */\n",
              "                background-size: auto;\n",
              "            }\n",
              "            .progress-bar-interrupted, .progress-bar-interrupted::-webkit-progress-bar {\n",
              "                background: #F44336;\n",
              "            }\n",
              "        </style>\n",
              "      <progress value='83' class='' max='246' style='width:300px; height:20px; vertical-align: middle;'></progress>\n",
              "      33.74% [83/246 00:14<00:28]\n",
              "    </div>\n",
              "    "
            ],
            "text/plain": [
              "<IPython.core.display.HTML object>"
            ]
          },
          "metadata": {
            "tags": []
          }
        },
        {
          "output_type": "stream",
          "text": [
            "/usr/local/lib/python3.6/dist-packages/torch/nn/functional.py:3000: UserWarning: The default behavior for interpolate/upsample with float scale_factor changed in 1.6.0 to align with other frameworks/libraries, and uses scale_factor directly, instead of relying on the computed output size. If you wish to keep the old behavior, please set recompute_scale_factor=True. See the documentation of nn.Upsample for details. \n",
            "  warnings.warn(\"The default behavior for interpolate/upsample with float scale_factor changed \"\n"
          ],
          "name": "stderr"
        },
        {
          "output_type": "stream",
          "text": [
            "Buffered data was truncated after reaching the output size limit."
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "zMoU9AQOmh1n",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 34
        },
        "outputId": "18befc65-3282-477b-c8f0-02ea02c02311"
      },
      "source": [
        "print('depth multiplier={}'.format(dm), {e: sum(res[e])/len(res[e]) for e in res})"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "depth multiplier=2 {5: 0.8205650448799133, 80: 0.9063374996185303}\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "jH_qvMEzaHMu",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 1000
        },
        "outputId": "e25bef4a-f523-40f5-8d7e-57a155a6613f"
      },
      "source": [
        "for i in range(80):\n",
        "  print(i, learn.recorder.val_losses[i].item(), learn.recorder.metrics[i][0].item())"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "0 1.8315829038619995 0.41180962324142456\n",
            "1 1.5907410383224487 0.5271061062812805\n",
            "2 1.3909424543380737 0.6317129135131836\n",
            "3 1.2509053945541382 0.6866887211799622\n",
            "4 1.172993779182434 0.7230847477912903\n",
            "5 1.1123597621917725 0.7645711302757263\n",
            "6 1.0472970008850098 0.7953677773475647\n",
            "7 1.0715622901916504 0.7856960892677307\n",
            "8 0.9907216429710388 0.8159837126731873\n",
            "9 1.0041720867156982 0.8091117143630981\n",
            "10 0.9420099258422852 0.8424535393714905\n",
            "11 0.9210571050643921 0.8432170748710632\n",
            "12 0.9411424398422241 0.8340544700622559\n",
            "13 0.9026457667350769 0.8521252274513245\n",
            "14 0.8896369934082031 0.863069474697113\n",
            "15 0.8706911206245422 0.8686688542366028\n",
            "16 0.9016208648681641 0.8523797392845154\n",
            "17 0.8971046805381775 0.8561974763870239\n",
            "18 0.8691442012786865 0.8676508069038391\n",
            "19 0.8719481229782104 0.863323986530304\n",
            "20 0.858334481716156 0.8763043880462646\n",
            "21 0.8563732504844666 0.8750318288803101\n",
            "22 0.8597426414489746 0.8768134117126465\n",
            "23 0.8590760827064514 0.86790531873703\n",
            "24 0.850490927696228 0.8757953643798828\n",
            "25 0.8550527691841125 0.8712140321731567\n",
            "26 0.8506224751472473 0.8712140321731567\n",
            "27 0.8666445016860962 0.8791040778160095\n",
            "28 0.8373748660087585 0.8806312084197998\n",
            "29 0.8308119773864746 0.8847034573554993\n",
            "30 0.841507077217102 0.8742682337760925\n",
            "31 0.8175877332687378 0.8872486352920532\n",
            "32 0.8434765934944153 0.8813947439193726\n",
            "33 0.8423311710357666 0.8617969155311584\n",
            "34 0.8432422876358032 0.8808857202529907\n",
            "35 0.8399994969367981 0.8783405423164368\n",
            "36 0.8210242390632629 0.8808857202529907\n",
            "37 0.8326374888420105 0.875286340713501\n",
            "38 0.8370193243026733 0.877577006816864\n",
            "39 0.8520188331604004 0.8737592101097107\n",
            "40 0.8092926144599915 0.8859760761260986\n",
            "41 0.8215572834014893 0.879867672920227\n",
            "42 0.8477233648300171 0.8735046982765198\n",
            "43 0.8191834688186646 0.8859760761260986\n",
            "44 0.8251851201057434 0.8791040778160095\n",
            "45 0.8223003149032593 0.8864850997924805\n",
            "46 0.8133966326713562 0.8826673626899719\n",
            "47 0.8140043616294861 0.8895393013954163\n",
            "48 0.817006528377533 0.8872486352920532\n",
            "49 0.8191025257110596 0.8819037675857544\n",
            "50 0.821880042552948 0.8890302777290344\n",
            "51 0.8157551884651184 0.8813947439193726\n",
            "52 0.8290197849273682 0.8806312084197998\n",
            "53 0.8047786951065063 0.8903028964996338\n",
            "54 0.8122377395629883 0.8831763863563538\n",
            "55 0.8086780309677124 0.8923389911651611\n",
            "56 0.8062209486961365 0.8875032067298889\n",
            "57 0.8285014033317566 0.8778315186500549\n",
            "58 0.814422070980072 0.8864850997924805\n",
            "59 0.8042709231376648 0.8875032067298889\n",
            "60 0.8072315454483032 0.8834308981895447\n",
            "61 0.810628354549408 0.8895393013954163\n",
            "62 0.7858232259750366 0.8946296572685242\n",
            "63 0.7919926047325134 0.8915754556655884\n",
            "64 0.8039336204528809 0.8890302777290344\n",
            "65 0.7869465351104736 0.8918299674987793\n",
            "66 0.787632942199707 0.8943751454353333\n",
            "67 0.7931775450706482 0.8903028964996338\n",
            "68 0.7626927495002747 0.9030287861824036\n",
            "69 0.7654426693916321 0.899974524974823\n",
            "70 0.7602089643478394 0.9050648808479309\n",
            "71 0.7676786780357361 0.9032832980155945\n",
            "72 0.7545405626296997 0.9101552367210388\n",
            "73 0.7482039332389832 0.9060829877853394\n",
            "74 0.7462551593780518 0.9065920114517212\n",
            "75 0.7462677359580994 0.9068465232849121\n",
            "76 0.7467260956764221 0.9099007248878479\n",
            "77 0.7467115521430969 0.9078645706176758\n",
            "78 0.7455323934555054 0.9076100587844849\n",
            "79 0.7476811408996582 0.9063374996185303\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "78R3lDAiaPP0",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "state = learn.model.state_dict()"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "Mrq7CIRjaUsA",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "torch.save(state, 'drive/app/ResNet50_depth.pth')"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "_GGvI3gNe4Of",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 17
        },
        "outputId": "e089016d-8420-437e-e101-f8a7fd256af1"
      },
      "source": [
        "from google.colab import files\n",
        "files.download('./checkpoint/ResNet50_depth.pth')"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "display_data",
          "data": {
            "application/javascript": [
              "\n",
              "    async function download(id, filename, size) {\n",
              "      if (!google.colab.kernel.accessAllowed) {\n",
              "        return;\n",
              "      }\n",
              "      const div = document.createElement('div');\n",
              "      const label = document.createElement('label');\n",
              "      label.textContent = `Downloading \"${filename}\": `;\n",
              "      div.appendChild(label);\n",
              "      const progress = document.createElement('progress');\n",
              "      progress.max = size;\n",
              "      div.appendChild(progress);\n",
              "      document.body.appendChild(div);\n",
              "\n",
              "      const buffers = [];\n",
              "      let downloaded = 0;\n",
              "\n",
              "      const channel = await google.colab.kernel.comms.open(id);\n",
              "      // Send a message to notify the kernel that we're ready.\n",
              "      channel.send({})\n",
              "\n",
              "      for await (const message of channel.messages) {\n",
              "        // Send a message to notify the kernel that we're ready.\n",
              "        channel.send({})\n",
              "        if (message.buffers) {\n",
              "          for (const buffer of message.buffers) {\n",
              "            buffers.push(buffer);\n",
              "            downloaded += buffer.byteLength;\n",
              "            progress.value = downloaded;\n",
              "          }\n",
              "        }\n",
              "      }\n",
              "      const blob = new Blob(buffers, {type: 'application/binary'});\n",
              "      const a = document.createElement('a');\n",
              "      a.href = window.URL.createObjectURL(blob);\n",
              "      a.download = filename;\n",
              "      div.appendChild(a);\n",
              "      a.click();\n",
              "      div.remove();\n",
              "    }\n",
              "  "
            ],
            "text/plain": [
              "<IPython.core.display.Javascript object>"
            ]
          },
          "metadata": {
            "tags": []
          }
        },
        {
          "output_type": "display_data",
          "data": {
            "application/javascript": [
              "download(\"download_ccd1529a-40d0-4083-a346-03a9bf20bc2e\", \"ResNet50_depth.pth\", 77774340)"
            ],
            "text/plain": [
              "<IPython.core.display.Javascript object>"
            ]
          },
          "metadata": {
            "tags": []
          }
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "xaG1b3OtzyWE",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 1000
        },
        "outputId": "95bc5626-4d4b-4774-bc1e-2930bc224f90"
      },
      "source": [
        "dm = 2\n",
        "res = dict()\n",
        "for e in [5]*5 + [20]*5 + [80]:\n",
        "  mixup=0 if e<=20 else 0.2\n",
        "  learn = get_learn(model=model, size=192, bs=16, mixup=mixup)\n",
        "  learn.fit_fc(e, lr=4e-3, moms=(0.95,0.95), start_pct=0.72)\n",
        "  acc = learn.recorder.metrics[-1][0].item()\n",
        "  if e in res:\n",
        "    res[e] += [acc]\n",
        "  else:\n",
        "    res[e] = [acc]\n",
        "  print('{} epochs: {} ({} runs)'.format(e, sum(res[e])/len(res[e]), len(res[e])))\n",
        "print('depth multiplier={}'.format(dm), {e: sum(res[e])/len(res[e]) for e in res})"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "data path   /root/.fastai/data/imagewoof2\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "stream",
          "text": [
            "/usr/local/lib/python3.6/dist-packages/torch/nn/functional.py:3000: UserWarning: The default behavior for interpolate/upsample with float scale_factor changed in 1.6.0 to align with other frameworks/libraries, and uses scale_factor directly, instead of relying on the computed output size. If you wish to keep the old behavior, please set recompute_scale_factor=True. See the documentation of nn.Upsample for details. \n",
            "  warnings.warn(\"The default behavior for interpolate/upsample with float scale_factor changed \"\n",
            "/usr/local/lib/python3.6/dist-packages/torch/nn/functional.py:3000: UserWarning: The default behavior for interpolate/upsample with float scale_factor changed in 1.6.0 to align with other frameworks/libraries, and uses scale_factor directly, instead of relying on the computed output size. If you wish to keep the old behavior, please set recompute_scale_factor=True. See the documentation of nn.Upsample for details. \n",
            "  warnings.warn(\"The default behavior for interpolate/upsample with float scale_factor changed \"\n",
            "/usr/local/lib/python3.6/dist-packages/torch/nn/functional.py:3000: UserWarning: The default behavior for interpolate/upsample with float scale_factor changed in 1.6.0 to align with other frameworks/libraries, and uses scale_factor directly, instead of relying on the computed output size. If you wish to keep the old behavior, please set recompute_scale_factor=True. See the documentation of nn.Upsample for details. \n",
            "  warnings.warn(\"The default behavior for interpolate/upsample with float scale_factor changed \"\n",
            "/usr/local/lib/python3.6/dist-packages/torch/nn/functional.py:3000: UserWarning: The default behavior for interpolate/upsample with float scale_factor changed in 1.6.0 to align with other frameworks/libraries, and uses scale_factor directly, instead of relying on the computed output size. If you wish to keep the old behavior, please set recompute_scale_factor=True. See the documentation of nn.Upsample for details. \n",
            "  warnings.warn(\"The default behavior for interpolate/upsample with float scale_factor changed \"\n"
          ],
          "name": "stderr"
        },
        {
          "output_type": "stream",
          "text": [
            "Learn path /root/.fastai/data/imagewoof2\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "display_data",
          "data": {
            "text/html": [
              "<table border=\"1\" class=\"dataframe\">\n",
              "  <thead>\n",
              "    <tr style=\"text-align: left;\">\n",
              "      <th>epoch</th>\n",
              "      <th>train_loss</th>\n",
              "      <th>valid_loss</th>\n",
              "      <th>accuracy</th>\n",
              "      <th>top_k_accuracy</th>\n",
              "      <th>time</th>\n",
              "    </tr>\n",
              "  </thead>\n",
              "  <tbody>\n",
              "    <tr>\n",
              "      <td>0</td>\n",
              "      <td>1.881894</td>\n",
              "      <td>1.720682</td>\n",
              "      <td>0.470349</td>\n",
              "      <td>0.891575</td>\n",
              "      <td>02:02</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>1</td>\n",
              "      <td>1.574010</td>\n",
              "      <td>1.428916</td>\n",
              "      <td>0.607025</td>\n",
              "      <td>0.940952</td>\n",
              "      <td>02:02</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>2</td>\n",
              "      <td>1.405706</td>\n",
              "      <td>1.288442</td>\n",
              "      <td>0.680835</td>\n",
              "      <td>0.955205</td>\n",
              "      <td>02:03</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>3</td>\n",
              "      <td>1.289675</td>\n",
              "      <td>1.186716</td>\n",
              "      <td>0.724866</td>\n",
              "      <td>0.964877</td>\n",
              "      <td>02:02</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>4</td>\n",
              "      <td>1.075579</td>\n",
              "      <td>1.033660</td>\n",
              "      <td>0.789005</td>\n",
              "      <td>0.977857</td>\n",
              "      <td>02:02</td>\n",
              "    </tr>\n",
              "  </tbody>\n",
              "</table>"
            ],
            "text/plain": [
              "<IPython.core.display.HTML object>"
            ]
          },
          "metadata": {
            "tags": []
          }
        },
        {
          "output_type": "stream",
          "text": [
            "/usr/local/lib/python3.6/dist-packages/torch/nn/functional.py:3000: UserWarning: The default behavior for interpolate/upsample with float scale_factor changed in 1.6.0 to align with other frameworks/libraries, and uses scale_factor directly, instead of relying on the computed output size. If you wish to keep the old behavior, please set recompute_scale_factor=True. See the documentation of nn.Upsample for details. \n",
            "  warnings.warn(\"The default behavior for interpolate/upsample with float scale_factor changed \"\n"
          ],
          "name": "stderr"
        },
        {
          "output_type": "stream",
          "text": [
            "5 epochs: 0.7890048623085022 (1 runs)\n",
            "data path   /root/.fastai/data/imagewoof2\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "stream",
          "text": [
            "/usr/local/lib/python3.6/dist-packages/torch/nn/functional.py:3000: UserWarning: The default behavior for interpolate/upsample with float scale_factor changed in 1.6.0 to align with other frameworks/libraries, and uses scale_factor directly, instead of relying on the computed output size. If you wish to keep the old behavior, please set recompute_scale_factor=True. See the documentation of nn.Upsample for details. \n",
            "  warnings.warn(\"The default behavior for interpolate/upsample with float scale_factor changed \"\n"
          ],
          "name": "stderr"
        },
        {
          "output_type": "stream",
          "text": [
            "Learn path /root/.fastai/data/imagewoof2\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "display_data",
          "data": {
            "text/html": [
              "<table border=\"1\" class=\"dataframe\">\n",
              "  <thead>\n",
              "    <tr style=\"text-align: left;\">\n",
              "      <th>epoch</th>\n",
              "      <th>train_loss</th>\n",
              "      <th>valid_loss</th>\n",
              "      <th>accuracy</th>\n",
              "      <th>top_k_accuracy</th>\n",
              "      <th>time</th>\n",
              "    </tr>\n",
              "  </thead>\n",
              "  <tbody>\n",
              "    <tr>\n",
              "      <td>0</td>\n",
              "      <td>1.862350</td>\n",
              "      <td>1.744569</td>\n",
              "      <td>0.469076</td>\n",
              "      <td>0.894375</td>\n",
              "      <td>02:02</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>1</td>\n",
              "      <td>1.605429</td>\n",
              "      <td>1.437807</td>\n",
              "      <td>0.598880</td>\n",
              "      <td>0.946297</td>\n",
              "      <td>02:02</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>2</td>\n",
              "      <td>1.428302</td>\n",
              "      <td>1.271008</td>\n",
              "      <td>0.677272</td>\n",
              "      <td>0.957750</td>\n",
              "      <td>02:02</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>3</td>\n",
              "      <td>1.280394</td>\n",
              "      <td>1.180982</td>\n",
              "      <td>0.728939</td>\n",
              "      <td>0.962840</td>\n",
              "      <td>02:02</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>4</td>\n",
              "      <td>1.102424</td>\n",
              "      <td>1.033393</td>\n",
              "      <td>0.804276</td>\n",
              "      <td>0.976839</td>\n",
              "      <td>02:02</td>\n",
              "    </tr>\n",
              "  </tbody>\n",
              "</table>"
            ],
            "text/plain": [
              "<IPython.core.display.HTML object>"
            ]
          },
          "metadata": {
            "tags": []
          }
        },
        {
          "output_type": "stream",
          "text": [
            "/usr/local/lib/python3.6/dist-packages/torch/nn/functional.py:3000: UserWarning: The default behavior for interpolate/upsample with float scale_factor changed in 1.6.0 to align with other frameworks/libraries, and uses scale_factor directly, instead of relying on the computed output size. If you wish to keep the old behavior, please set recompute_scale_factor=True. See the documentation of nn.Upsample for details. \n",
            "  warnings.warn(\"The default behavior for interpolate/upsample with float scale_factor changed \"\n"
          ],
          "name": "stderr"
        },
        {
          "output_type": "stream",
          "text": [
            "5 epochs: 0.7966403663158417 (2 runs)\n",
            "data path   /root/.fastai/data/imagewoof2\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "stream",
          "text": [
            "/usr/local/lib/python3.6/dist-packages/torch/nn/functional.py:3000: UserWarning: The default behavior for interpolate/upsample with float scale_factor changed in 1.6.0 to align with other frameworks/libraries, and uses scale_factor directly, instead of relying on the computed output size. If you wish to keep the old behavior, please set recompute_scale_factor=True. See the documentation of nn.Upsample for details. \n",
            "  warnings.warn(\"The default behavior for interpolate/upsample with float scale_factor changed \"\n",
            "/usr/local/lib/python3.6/dist-packages/torch/nn/functional.py:3000: UserWarning: The default behavior for interpolate/upsample with float scale_factor changed in 1.6.0 to align with other frameworks/libraries, and uses scale_factor directly, instead of relying on the computed output size. If you wish to keep the old behavior, please set recompute_scale_factor=True. See the documentation of nn.Upsample for details. \n",
            "  warnings.warn(\"The default behavior for interpolate/upsample with float scale_factor changed \"\n"
          ],
          "name": "stderr"
        },
        {
          "output_type": "stream",
          "text": [
            "Learn path /root/.fastai/data/imagewoof2\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "display_data",
          "data": {
            "text/html": [
              "<table border=\"1\" class=\"dataframe\">\n",
              "  <thead>\n",
              "    <tr style=\"text-align: left;\">\n",
              "      <th>epoch</th>\n",
              "      <th>train_loss</th>\n",
              "      <th>valid_loss</th>\n",
              "      <th>accuracy</th>\n",
              "      <th>top_k_accuracy</th>\n",
              "      <th>time</th>\n",
              "    </tr>\n",
              "  </thead>\n",
              "  <tbody>\n",
              "    <tr>\n",
              "      <td>0</td>\n",
              "      <td>1.912310</td>\n",
              "      <td>1.790168</td>\n",
              "      <td>0.433189</td>\n",
              "      <td>0.883685</td>\n",
              "      <td>02:02</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>1</td>\n",
              "      <td>1.613266</td>\n",
              "      <td>1.467700</td>\n",
              "      <td>0.580809</td>\n",
              "      <td>0.944770</td>\n",
              "      <td>02:02</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>2</td>\n",
              "      <td>1.432900</td>\n",
              "      <td>1.306103</td>\n",
              "      <td>0.662001</td>\n",
              "      <td>0.957496</td>\n",
              "      <td>02:02</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>3</td>\n",
              "      <td>1.297314</td>\n",
              "      <td>1.191611</td>\n",
              "      <td>0.722576</td>\n",
              "      <td>0.966658</td>\n",
              "      <td>02:02</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>4</td>\n",
              "      <td>1.077379</td>\n",
              "      <td>1.029889</td>\n",
              "      <td>0.801985</td>\n",
              "      <td>0.974294</td>\n",
              "      <td>02:02</td>\n",
              "    </tr>\n",
              "  </tbody>\n",
              "</table>"
            ],
            "text/plain": [
              "<IPython.core.display.HTML object>"
            ]
          },
          "metadata": {
            "tags": []
          }
        },
        {
          "output_type": "stream",
          "text": [
            "5 epochs: 0.7984219988187155 (3 runs)\n",
            "data path   /root/.fastai/data/imagewoof2\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "stream",
          "text": [
            "/usr/local/lib/python3.6/dist-packages/torch/nn/functional.py:3000: UserWarning: The default behavior for interpolate/upsample with float scale_factor changed in 1.6.0 to align with other frameworks/libraries, and uses scale_factor directly, instead of relying on the computed output size. If you wish to keep the old behavior, please set recompute_scale_factor=True. See the documentation of nn.Upsample for details. \n",
            "  warnings.warn(\"The default behavior for interpolate/upsample with float scale_factor changed \"\n",
            "/usr/local/lib/python3.6/dist-packages/torch/nn/functional.py:3000: UserWarning: The default behavior for interpolate/upsample with float scale_factor changed in 1.6.0 to align with other frameworks/libraries, and uses scale_factor directly, instead of relying on the computed output size. If you wish to keep the old behavior, please set recompute_scale_factor=True. See the documentation of nn.Upsample for details. \n",
            "  warnings.warn(\"The default behavior for interpolate/upsample with float scale_factor changed \"\n",
            "/usr/local/lib/python3.6/dist-packages/torch/nn/functional.py:3000: UserWarning: The default behavior for interpolate/upsample with float scale_factor changed in 1.6.0 to align with other frameworks/libraries, and uses scale_factor directly, instead of relying on the computed output size. If you wish to keep the old behavior, please set recompute_scale_factor=True. See the documentation of nn.Upsample for details. \n",
            "  warnings.warn(\"The default behavior for interpolate/upsample with float scale_factor changed \"\n",
            "/usr/local/lib/python3.6/dist-packages/torch/nn/functional.py:3000: UserWarning: The default behavior for interpolate/upsample with float scale_factor changed in 1.6.0 to align with other frameworks/libraries, and uses scale_factor directly, instead of relying on the computed output size. If you wish to keep the old behavior, please set recompute_scale_factor=True. See the documentation of nn.Upsample for details. \n",
            "  warnings.warn(\"The default behavior for interpolate/upsample with float scale_factor changed \"\n"
          ],
          "name": "stderr"
        },
        {
          "output_type": "stream",
          "text": [
            "Learn path /root/.fastai/data/imagewoof2\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "display_data",
          "data": {
            "text/html": [
              "<table border=\"1\" class=\"dataframe\">\n",
              "  <thead>\n",
              "    <tr style=\"text-align: left;\">\n",
              "      <th>epoch</th>\n",
              "      <th>train_loss</th>\n",
              "      <th>valid_loss</th>\n",
              "      <th>accuracy</th>\n",
              "      <th>top_k_accuracy</th>\n",
              "      <th>time</th>\n",
              "    </tr>\n",
              "  </thead>\n",
              "  <tbody>\n",
              "    <tr>\n",
              "      <td>0</td>\n",
              "      <td>1.863152</td>\n",
              "      <td>1.782902</td>\n",
              "      <td>0.413591</td>\n",
              "      <td>0.879613</td>\n",
              "      <td>02:02</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>1</td>\n",
              "      <td>1.633796</td>\n",
              "      <td>1.437181</td>\n",
              "      <td>0.605498</td>\n",
              "      <td>0.944006</td>\n",
              "      <td>02:02</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>2</td>\n",
              "      <td>1.481287</td>\n",
              "      <td>1.331935</td>\n",
              "      <td>0.647748</td>\n",
              "      <td>0.953169</td>\n",
              "      <td>02:02</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>3</td>\n",
              "      <td>1.327532</td>\n",
              "      <td>1.151493</td>\n",
              "      <td>0.739374</td>\n",
              "      <td>0.967931</td>\n",
              "      <td>02:01</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>4</td>\n",
              "      <td>1.088265</td>\n",
              "      <td>1.028836</td>\n",
              "      <td>0.799440</td>\n",
              "      <td>0.976330</td>\n",
              "      <td>02:02</td>\n",
              "    </tr>\n",
              "  </tbody>\n",
              "</table>"
            ],
            "text/plain": [
              "<IPython.core.display.HTML object>"
            ]
          },
          "metadata": {
            "tags": []
          }
        },
        {
          "output_type": "stream",
          "text": [
            "/usr/local/lib/python3.6/dist-packages/torch/nn/functional.py:3000: UserWarning: The default behavior for interpolate/upsample with float scale_factor changed in 1.6.0 to align with other frameworks/libraries, and uses scale_factor directly, instead of relying on the computed output size. If you wish to keep the old behavior, please set recompute_scale_factor=True. See the documentation of nn.Upsample for details. \n",
            "  warnings.warn(\"The default behavior for interpolate/upsample with float scale_factor changed \"\n"
          ],
          "name": "stderr"
        },
        {
          "output_type": "stream",
          "text": [
            "5 epochs: 0.7986765205860138 (4 runs)\n",
            "data path   /root/.fastai/data/imagewoof2\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "stream",
          "text": [
            "/usr/local/lib/python3.6/dist-packages/torch/nn/functional.py:3000: UserWarning: The default behavior for interpolate/upsample with float scale_factor changed in 1.6.0 to align with other frameworks/libraries, and uses scale_factor directly, instead of relying on the computed output size. If you wish to keep the old behavior, please set recompute_scale_factor=True. See the documentation of nn.Upsample for details. \n",
            "  warnings.warn(\"The default behavior for interpolate/upsample with float scale_factor changed \"\n",
            "/usr/local/lib/python3.6/dist-packages/torch/nn/functional.py:3000: UserWarning: The default behavior for interpolate/upsample with float scale_factor changed in 1.6.0 to align with other frameworks/libraries, and uses scale_factor directly, instead of relying on the computed output size. If you wish to keep the old behavior, please set recompute_scale_factor=True. See the documentation of nn.Upsample for details. \n",
            "  warnings.warn(\"The default behavior for interpolate/upsample with float scale_factor changed \"\n",
            "/usr/local/lib/python3.6/dist-packages/torch/nn/functional.py:3000: UserWarning: The default behavior for interpolate/upsample with float scale_factor changed in 1.6.0 to align with other frameworks/libraries, and uses scale_factor directly, instead of relying on the computed output size. If you wish to keep the old behavior, please set recompute_scale_factor=True. See the documentation of nn.Upsample for details. \n",
            "  warnings.warn(\"The default behavior for interpolate/upsample with float scale_factor changed \"\n"
          ],
          "name": "stderr"
        },
        {
          "output_type": "stream",
          "text": [
            "Learn path /root/.fastai/data/imagewoof2\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "display_data",
          "data": {
            "text/html": [
              "<table border=\"1\" class=\"dataframe\">\n",
              "  <thead>\n",
              "    <tr style=\"text-align: left;\">\n",
              "      <th>epoch</th>\n",
              "      <th>train_loss</th>\n",
              "      <th>valid_loss</th>\n",
              "      <th>accuracy</th>\n",
              "      <th>top_k_accuracy</th>\n",
              "      <th>time</th>\n",
              "    </tr>\n",
              "  </thead>\n",
              "  <tbody>\n",
              "    <tr>\n",
              "      <td>0</td>\n",
              "      <td>1.882808</td>\n",
              "      <td>1.763873</td>\n",
              "      <td>0.447442</td>\n",
              "      <td>0.899466</td>\n",
              "      <td>02:02</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>1</td>\n",
              "      <td>1.617233</td>\n",
              "      <td>1.442022</td>\n",
              "      <td>0.602952</td>\n",
              "      <td>0.934843</td>\n",
              "      <td>02:02</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>2</td>\n",
              "      <td>1.410196</td>\n",
              "      <td>1.305165</td>\n",
              "      <td>0.673708</td>\n",
              "      <td>0.957496</td>\n",
              "      <td>02:02</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>3</td>\n",
              "      <td>1.296416</td>\n",
              "      <td>1.185060</td>\n",
              "      <td>0.722830</td>\n",
              "      <td>0.963604</td>\n",
              "      <td>02:02</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>4</td>\n",
              "      <td>1.097956</td>\n",
              "      <td>1.031142</td>\n",
              "      <td>0.798676</td>\n",
              "      <td>0.976584</td>\n",
              "      <td>02:02</td>\n",
              "    </tr>\n",
              "  </tbody>\n",
              "</table>"
            ],
            "text/plain": [
              "<IPython.core.display.HTML object>"
            ]
          },
          "metadata": {
            "tags": []
          }
        },
        {
          "output_type": "stream",
          "text": [
            "/usr/local/lib/python3.6/dist-packages/torch/nn/functional.py:3000: UserWarning: The default behavior for interpolate/upsample with float scale_factor changed in 1.6.0 to align with other frameworks/libraries, and uses scale_factor directly, instead of relying on the computed output size. If you wish to keep the old behavior, please set recompute_scale_factor=True. See the documentation of nn.Upsample for details. \n",
            "  warnings.warn(\"The default behavior for interpolate/upsample with float scale_factor changed \"\n"
          ],
          "name": "stderr"
        },
        {
          "output_type": "stream",
          "text": [
            "5 epochs: 0.7986765146255493 (5 runs)\n",
            "data path   /root/.fastai/data/imagewoof2\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "stream",
          "text": [
            "/usr/local/lib/python3.6/dist-packages/torch/nn/functional.py:3000: UserWarning: The default behavior for interpolate/upsample with float scale_factor changed in 1.6.0 to align with other frameworks/libraries, and uses scale_factor directly, instead of relying on the computed output size. If you wish to keep the old behavior, please set recompute_scale_factor=True. See the documentation of nn.Upsample for details. \n",
            "  warnings.warn(\"The default behavior for interpolate/upsample with float scale_factor changed \"\n",
            "/usr/local/lib/python3.6/dist-packages/torch/nn/functional.py:3000: UserWarning: The default behavior for interpolate/upsample with float scale_factor changed in 1.6.0 to align with other frameworks/libraries, and uses scale_factor directly, instead of relying on the computed output size. If you wish to keep the old behavior, please set recompute_scale_factor=True. See the documentation of nn.Upsample for details. \n",
            "  warnings.warn(\"The default behavior for interpolate/upsample with float scale_factor changed \"\n",
            "/usr/local/lib/python3.6/dist-packages/torch/nn/functional.py:3000: UserWarning: The default behavior for interpolate/upsample with float scale_factor changed in 1.6.0 to align with other frameworks/libraries, and uses scale_factor directly, instead of relying on the computed output size. If you wish to keep the old behavior, please set recompute_scale_factor=True. See the documentation of nn.Upsample for details. \n",
            "  warnings.warn(\"The default behavior for interpolate/upsample with float scale_factor changed \"\n",
            "/usr/local/lib/python3.6/dist-packages/torch/nn/functional.py:3000: UserWarning: The default behavior for interpolate/upsample with float scale_factor changed in 1.6.0 to align with other frameworks/libraries, and uses scale_factor directly, instead of relying on the computed output size. If you wish to keep the old behavior, please set recompute_scale_factor=True. See the documentation of nn.Upsample for details. \n",
            "  warnings.warn(\"The default behavior for interpolate/upsample with float scale_factor changed \"\n",
            "/usr/local/lib/python3.6/dist-packages/torch/nn/functional.py:3000: UserWarning: The default behavior for interpolate/upsample with float scale_factor changed in 1.6.0 to align with other frameworks/libraries, and uses scale_factor directly, instead of relying on the computed output size. If you wish to keep the old behavior, please set recompute_scale_factor=True. See the documentation of nn.Upsample for details. \n",
            "  warnings.warn(\"The default behavior for interpolate/upsample with float scale_factor changed \"\n"
          ],
          "name": "stderr"
        },
        {
          "output_type": "stream",
          "text": [
            "Learn path /root/.fastai/data/imagewoof2\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "display_data",
          "data": {
            "text/html": [
              "<table border=\"1\" class=\"dataframe\">\n",
              "  <thead>\n",
              "    <tr style=\"text-align: left;\">\n",
              "      <th>epoch</th>\n",
              "      <th>train_loss</th>\n",
              "      <th>valid_loss</th>\n",
              "      <th>accuracy</th>\n",
              "      <th>top_k_accuracy</th>\n",
              "      <th>time</th>\n",
              "    </tr>\n",
              "  </thead>\n",
              "  <tbody>\n",
              "    <tr>\n",
              "      <td>0</td>\n",
              "      <td>1.905579</td>\n",
              "      <td>1.830914</td>\n",
              "      <td>0.400102</td>\n",
              "      <td>0.876813</td>\n",
              "      <td>02:02</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>1</td>\n",
              "      <td>1.670766</td>\n",
              "      <td>1.530203</td>\n",
              "      <td>0.557903</td>\n",
              "      <td>0.927717</td>\n",
              "      <td>02:02</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>2</td>\n",
              "      <td>1.440382</td>\n",
              "      <td>1.325415</td>\n",
              "      <td>0.654620</td>\n",
              "      <td>0.952151</td>\n",
              "      <td>02:02</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>3</td>\n",
              "      <td>1.328230</td>\n",
              "      <td>1.241551</td>\n",
              "      <td>0.698397</td>\n",
              "      <td>0.966658</td>\n",
              "      <td>02:02</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>4</td>\n",
              "      <td>1.225569</td>\n",
              "      <td>1.112639</td>\n",
              "      <td>0.752863</td>\n",
              "      <td>0.968694</td>\n",
              "      <td>02:02</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>5</td>\n",
              "      <td>1.176966</td>\n",
              "      <td>1.086880</td>\n",
              "      <td>0.771189</td>\n",
              "      <td>0.976584</td>\n",
              "      <td>02:02</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>6</td>\n",
              "      <td>1.092348</td>\n",
              "      <td>1.036949</td>\n",
              "      <td>0.790786</td>\n",
              "      <td>0.978621</td>\n",
              "      <td>02:02</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>7</td>\n",
              "      <td>1.054482</td>\n",
              "      <td>1.007084</td>\n",
              "      <td>0.808603</td>\n",
              "      <td>0.972512</td>\n",
              "      <td>02:02</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>8</td>\n",
              "      <td>1.010084</td>\n",
              "      <td>0.982966</td>\n",
              "      <td>0.816747</td>\n",
              "      <td>0.977093</td>\n",
              "      <td>02:02</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>9</td>\n",
              "      <td>0.946227</td>\n",
              "      <td>0.987310</td>\n",
              "      <td>0.815475</td>\n",
              "      <td>0.977857</td>\n",
              "      <td>02:02</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>10</td>\n",
              "      <td>0.941965</td>\n",
              "      <td>0.977477</td>\n",
              "      <td>0.810130</td>\n",
              "      <td>0.979384</td>\n",
              "      <td>02:02</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>11</td>\n",
              "      <td>0.915241</td>\n",
              "      <td>0.942629</td>\n",
              "      <td>0.828964</td>\n",
              "      <td>0.979384</td>\n",
              "      <td>02:02</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>12</td>\n",
              "      <td>0.895285</td>\n",
              "      <td>0.950181</td>\n",
              "      <td>0.824128</td>\n",
              "      <td>0.983456</td>\n",
              "      <td>02:02</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>13</td>\n",
              "      <td>0.876067</td>\n",
              "      <td>0.968109</td>\n",
              "      <td>0.818529</td>\n",
              "      <td>0.979639</td>\n",
              "      <td>02:02</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>14</td>\n",
              "      <td>0.872470</td>\n",
              "      <td>0.929551</td>\n",
              "      <td>0.847289</td>\n",
              "      <td>0.980148</td>\n",
              "      <td>02:02</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>15</td>\n",
              "      <td>0.810676</td>\n",
              "      <td>0.908753</td>\n",
              "      <td>0.849835</td>\n",
              "      <td>0.983711</td>\n",
              "      <td>02:02</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>16</td>\n",
              "      <td>0.753640</td>\n",
              "      <td>0.881555</td>\n",
              "      <td>0.854925</td>\n",
              "      <td>0.983711</td>\n",
              "      <td>02:02</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>17</td>\n",
              "      <td>0.680282</td>\n",
              "      <td>0.842657</td>\n",
              "      <td>0.868923</td>\n",
              "      <td>0.984729</td>\n",
              "      <td>02:02</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>18</td>\n",
              "      <td>0.650192</td>\n",
              "      <td>0.823456</td>\n",
              "      <td>0.879868</td>\n",
              "      <td>0.985492</td>\n",
              "      <td>02:02</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>19</td>\n",
              "      <td>0.621103</td>\n",
              "      <td>0.821476</td>\n",
              "      <td>0.883940</td>\n",
              "      <td>0.984983</td>\n",
              "      <td>02:02</td>\n",
              "    </tr>\n",
              "  </tbody>\n",
              "</table>"
            ],
            "text/plain": [
              "<IPython.core.display.HTML object>"
            ]
          },
          "metadata": {
            "tags": []
          }
        },
        {
          "output_type": "stream",
          "text": [
            "/usr/local/lib/python3.6/dist-packages/torch/nn/functional.py:3000: UserWarning: The default behavior for interpolate/upsample with float scale_factor changed in 1.6.0 to align with other frameworks/libraries, and uses scale_factor directly, instead of relying on the computed output size. If you wish to keep the old behavior, please set recompute_scale_factor=True. See the documentation of nn.Upsample for details. \n",
            "  warnings.warn(\"The default behavior for interpolate/upsample with float scale_factor changed \"\n"
          ],
          "name": "stderr"
        },
        {
          "output_type": "stream",
          "text": [
            "20 epochs: 0.8839399218559265 (1 runs)\n",
            "data path   /root/.fastai/data/imagewoof2\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "stream",
          "text": [
            "/usr/local/lib/python3.6/dist-packages/torch/nn/functional.py:3000: UserWarning: The default behavior for interpolate/upsample with float scale_factor changed in 1.6.0 to align with other frameworks/libraries, and uses scale_factor directly, instead of relying on the computed output size. If you wish to keep the old behavior, please set recompute_scale_factor=True. See the documentation of nn.Upsample for details. \n",
            "  warnings.warn(\"The default behavior for interpolate/upsample with float scale_factor changed \"\n",
            "/usr/local/lib/python3.6/dist-packages/torch/nn/functional.py:3000: UserWarning: The default behavior for interpolate/upsample with float scale_factor changed in 1.6.0 to align with other frameworks/libraries, and uses scale_factor directly, instead of relying on the computed output size. If you wish to keep the old behavior, please set recompute_scale_factor=True. See the documentation of nn.Upsample for details. \n",
            "  warnings.warn(\"The default behavior for interpolate/upsample with float scale_factor changed \"\n"
          ],
          "name": "stderr"
        },
        {
          "output_type": "stream",
          "text": [
            "Learn path /root/.fastai/data/imagewoof2\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "display_data",
          "data": {
            "text/html": [
              "<table border=\"1\" class=\"dataframe\">\n",
              "  <thead>\n",
              "    <tr style=\"text-align: left;\">\n",
              "      <th>epoch</th>\n",
              "      <th>train_loss</th>\n",
              "      <th>valid_loss</th>\n",
              "      <th>accuracy</th>\n",
              "      <th>top_k_accuracy</th>\n",
              "      <th>time</th>\n",
              "    </tr>\n",
              "  </thead>\n",
              "  <tbody>\n",
              "    <tr>\n",
              "      <td>0</td>\n",
              "      <td>1.924789</td>\n",
              "      <td>1.838617</td>\n",
              "      <td>0.405701</td>\n",
              "      <td>0.863069</td>\n",
              "      <td>02:02</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>1</td>\n",
              "      <td>1.645659</td>\n",
              "      <td>1.509284</td>\n",
              "      <td>0.567065</td>\n",
              "      <td>0.925935</td>\n",
              "      <td>02:01</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>2</td>\n",
              "      <td>1.466714</td>\n",
              "      <td>1.339576</td>\n",
              "      <td>0.648002</td>\n",
              "      <td>0.942225</td>\n",
              "      <td>02:01</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>3</td>\n",
              "      <td>1.328727</td>\n",
              "      <td>1.218515</td>\n",
              "      <td>0.712395</td>\n",
              "      <td>0.958259</td>\n",
              "      <td>02:02</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>4</td>\n",
              "      <td>1.226835</td>\n",
              "      <td>1.136776</td>\n",
              "      <td>0.745482</td>\n",
              "      <td>0.965386</td>\n",
              "      <td>02:02</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>5</td>\n",
              "      <td>1.161959</td>\n",
              "      <td>1.083957</td>\n",
              "      <td>0.769661</td>\n",
              "      <td>0.973021</td>\n",
              "      <td>02:02</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>6</td>\n",
              "      <td>1.148885</td>\n",
              "      <td>1.084405</td>\n",
              "      <td>0.783405</td>\n",
              "      <td>0.973021</td>\n",
              "      <td>02:01</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>7</td>\n",
              "      <td>1.077544</td>\n",
              "      <td>1.028915</td>\n",
              "      <td>0.797149</td>\n",
              "      <td>0.975312</td>\n",
              "      <td>02:01</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>8</td>\n",
              "      <td>1.029995</td>\n",
              "      <td>0.990004</td>\n",
              "      <td>0.818783</td>\n",
              "      <td>0.976584</td>\n",
              "      <td>02:01</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>9</td>\n",
              "      <td>0.981399</td>\n",
              "      <td>1.026612</td>\n",
              "      <td>0.799440</td>\n",
              "      <td>0.979130</td>\n",
              "      <td>02:02</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>10</td>\n",
              "      <td>0.949927</td>\n",
              "      <td>0.943441</td>\n",
              "      <td>0.834563</td>\n",
              "      <td>0.980657</td>\n",
              "      <td>02:02</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>11</td>\n",
              "      <td>0.914912</td>\n",
              "      <td>0.939519</td>\n",
              "      <td>0.837872</td>\n",
              "      <td>0.981675</td>\n",
              "      <td>02:01</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>12</td>\n",
              "      <td>0.915873</td>\n",
              "      <td>0.924505</td>\n",
              "      <td>0.839399</td>\n",
              "      <td>0.982947</td>\n",
              "      <td>02:02</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>13</td>\n",
              "      <td>0.895467</td>\n",
              "      <td>0.964378</td>\n",
              "      <td>0.822347</td>\n",
              "      <td>0.971240</td>\n",
              "      <td>02:02</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>14</td>\n",
              "      <td>0.845984</td>\n",
              "      <td>0.946437</td>\n",
              "      <td>0.823365</td>\n",
              "      <td>0.981420</td>\n",
              "      <td>02:01</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>15</td>\n",
              "      <td>0.802364</td>\n",
              "      <td>0.920092</td>\n",
              "      <td>0.839145</td>\n",
              "      <td>0.981675</td>\n",
              "      <td>02:02</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>16</td>\n",
              "      <td>0.743823</td>\n",
              "      <td>0.848072</td>\n",
              "      <td>0.866887</td>\n",
              "      <td>0.984729</td>\n",
              "      <td>02:02</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>17</td>\n",
              "      <td>0.680488</td>\n",
              "      <td>0.845602</td>\n",
              "      <td>0.876050</td>\n",
              "      <td>0.984220</td>\n",
              "      <td>02:03</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>18</td>\n",
              "      <td>0.639797</td>\n",
              "      <td>0.823269</td>\n",
              "      <td>0.883685</td>\n",
              "      <td>0.984729</td>\n",
              "      <td>02:02</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>19</td>\n",
              "      <td>0.627540</td>\n",
              "      <td>0.816551</td>\n",
              "      <td>0.884449</td>\n",
              "      <td>0.984220</td>\n",
              "      <td>02:02</td>\n",
              "    </tr>\n",
              "  </tbody>\n",
              "</table>"
            ],
            "text/plain": [
              "<IPython.core.display.HTML object>"
            ]
          },
          "metadata": {
            "tags": []
          }
        },
        {
          "output_type": "stream",
          "text": [
            "/usr/local/lib/python3.6/dist-packages/torch/nn/functional.py:3000: UserWarning: The default behavior for interpolate/upsample with float scale_factor changed in 1.6.0 to align with other frameworks/libraries, and uses scale_factor directly, instead of relying on the computed output size. If you wish to keep the old behavior, please set recompute_scale_factor=True. See the documentation of nn.Upsample for details. \n",
            "  warnings.warn(\"The default behavior for interpolate/upsample with float scale_factor changed \"\n"
          ],
          "name": "stderr"
        },
        {
          "output_type": "stream",
          "text": [
            "20 epochs: 0.8841944336891174 (2 runs)\n",
            "data path   /root/.fastai/data/imagewoof2\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "stream",
          "text": [
            "/usr/local/lib/python3.6/dist-packages/torch/nn/functional.py:3000: UserWarning: The default behavior for interpolate/upsample with float scale_factor changed in 1.6.0 to align with other frameworks/libraries, and uses scale_factor directly, instead of relying on the computed output size. If you wish to keep the old behavior, please set recompute_scale_factor=True. See the documentation of nn.Upsample for details. \n",
            "  warnings.warn(\"The default behavior for interpolate/upsample with float scale_factor changed \"\n",
            "/usr/local/lib/python3.6/dist-packages/torch/nn/functional.py:3000: UserWarning: The default behavior for interpolate/upsample with float scale_factor changed in 1.6.0 to align with other frameworks/libraries, and uses scale_factor directly, instead of relying on the computed output size. If you wish to keep the old behavior, please set recompute_scale_factor=True. See the documentation of nn.Upsample for details. \n",
            "  warnings.warn(\"The default behavior for interpolate/upsample with float scale_factor changed \"\n"
          ],
          "name": "stderr"
        },
        {
          "output_type": "stream",
          "text": [
            "Learn path /root/.fastai/data/imagewoof2\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "display_data",
          "data": {
            "text/html": [
              "<table border=\"1\" class=\"dataframe\">\n",
              "  <thead>\n",
              "    <tr style=\"text-align: left;\">\n",
              "      <th>epoch</th>\n",
              "      <th>train_loss</th>\n",
              "      <th>valid_loss</th>\n",
              "      <th>accuracy</th>\n",
              "      <th>top_k_accuracy</th>\n",
              "      <th>time</th>\n",
              "    </tr>\n",
              "  </thead>\n",
              "  <tbody>\n",
              "    <tr>\n",
              "      <td>0</td>\n",
              "      <td>1.877917</td>\n",
              "      <td>1.806122</td>\n",
              "      <td>0.435225</td>\n",
              "      <td>0.887249</td>\n",
              "      <td>02:02</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>1</td>\n",
              "      <td>1.570991</td>\n",
              "      <td>1.560151</td>\n",
              "      <td>0.564011</td>\n",
              "      <td>0.927717</td>\n",
              "      <td>02:03</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>2</td>\n",
              "      <td>1.445546</td>\n",
              "      <td>1.288350</td>\n",
              "      <td>0.678035</td>\n",
              "      <td>0.953423</td>\n",
              "      <td>02:03</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>3</td>\n",
              "      <td>1.347631</td>\n",
              "      <td>1.184602</td>\n",
              "      <td>0.723594</td>\n",
              "      <td>0.966913</td>\n",
              "      <td>02:02</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>4</td>\n",
              "      <td>1.251037</td>\n",
              "      <td>1.137811</td>\n",
              "      <td>0.746755</td>\n",
              "      <td>0.969712</td>\n",
              "      <td>02:03</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>5</td>\n",
              "      <td>1.155225</td>\n",
              "      <td>1.108924</td>\n",
              "      <td>0.766353</td>\n",
              "      <td>0.972512</td>\n",
              "      <td>02:03</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>6</td>\n",
              "      <td>1.101930</td>\n",
              "      <td>1.052603</td>\n",
              "      <td>0.779842</td>\n",
              "      <td>0.971494</td>\n",
              "      <td>02:03</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>7</td>\n",
              "      <td>1.058300</td>\n",
              "      <td>0.995994</td>\n",
              "      <td>0.812420</td>\n",
              "      <td>0.978112</td>\n",
              "      <td>02:02</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>8</td>\n",
              "      <td>1.003011</td>\n",
              "      <td>0.989454</td>\n",
              "      <td>0.802494</td>\n",
              "      <td>0.977857</td>\n",
              "      <td>02:03</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>9</td>\n",
              "      <td>1.000022</td>\n",
              "      <td>1.024957</td>\n",
              "      <td>0.800204</td>\n",
              "      <td>0.973021</td>\n",
              "      <td>02:02</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>10</td>\n",
              "      <td>0.913771</td>\n",
              "      <td>0.966143</td>\n",
              "      <td>0.815475</td>\n",
              "      <td>0.980657</td>\n",
              "      <td>02:03</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>11</td>\n",
              "      <td>0.922000</td>\n",
              "      <td>0.970423</td>\n",
              "      <td>0.819801</td>\n",
              "      <td>0.976584</td>\n",
              "      <td>02:03</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>12</td>\n",
              "      <td>0.866192</td>\n",
              "      <td>0.938940</td>\n",
              "      <td>0.829473</td>\n",
              "      <td>0.984220</td>\n",
              "      <td>02:02</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>13</td>\n",
              "      <td>0.868641</td>\n",
              "      <td>0.910169</td>\n",
              "      <td>0.847798</td>\n",
              "      <td>0.981675</td>\n",
              "      <td>02:02</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>14</td>\n",
              "      <td>0.853592</td>\n",
              "      <td>0.970943</td>\n",
              "      <td>0.817765</td>\n",
              "      <td>0.975566</td>\n",
              "      <td>02:03</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>15</td>\n",
              "      <td>0.791577</td>\n",
              "      <td>0.900124</td>\n",
              "      <td>0.847798</td>\n",
              "      <td>0.982693</td>\n",
              "      <td>02:04</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>16</td>\n",
              "      <td>0.726515</td>\n",
              "      <td>0.895874</td>\n",
              "      <td>0.846526</td>\n",
              "      <td>0.981420</td>\n",
              "      <td>02:03</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>17</td>\n",
              "      <td>0.694580</td>\n",
              "      <td>0.858481</td>\n",
              "      <td>0.865106</td>\n",
              "      <td>0.981675</td>\n",
              "      <td>02:03</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>18</td>\n",
              "      <td>0.620091</td>\n",
              "      <td>0.832190</td>\n",
              "      <td>0.873505</td>\n",
              "      <td>0.983202</td>\n",
              "      <td>02:03</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>19</td>\n",
              "      <td>0.621522</td>\n",
              "      <td>0.824399</td>\n",
              "      <td>0.877832</td>\n",
              "      <td>0.984983</td>\n",
              "      <td>02:03</td>\n",
              "    </tr>\n",
              "  </tbody>\n",
              "</table>"
            ],
            "text/plain": [
              "<IPython.core.display.HTML object>"
            ]
          },
          "metadata": {
            "tags": []
          }
        },
        {
          "output_type": "stream",
          "text": [
            "/usr/local/lib/python3.6/dist-packages/torch/nn/functional.py:3000: UserWarning: The default behavior for interpolate/upsample with float scale_factor changed in 1.6.0 to align with other frameworks/libraries, and uses scale_factor directly, instead of relying on the computed output size. If you wish to keep the old behavior, please set recompute_scale_factor=True. See the documentation of nn.Upsample for details. \n",
            "  warnings.warn(\"The default behavior for interpolate/upsample with float scale_factor changed \"\n"
          ],
          "name": "stderr"
        },
        {
          "output_type": "stream",
          "text": [
            "20 epochs: 0.8820734620094299 (3 runs)\n",
            "data path   /root/.fastai/data/imagewoof2\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "stream",
          "text": [
            "/usr/local/lib/python3.6/dist-packages/torch/nn/functional.py:3000: UserWarning: The default behavior for interpolate/upsample with float scale_factor changed in 1.6.0 to align with other frameworks/libraries, and uses scale_factor directly, instead of relying on the computed output size. If you wish to keep the old behavior, please set recompute_scale_factor=True. See the documentation of nn.Upsample for details. \n",
            "  warnings.warn(\"The default behavior for interpolate/upsample with float scale_factor changed \"\n",
            "/usr/local/lib/python3.6/dist-packages/torch/nn/functional.py:3000: UserWarning: The default behavior for interpolate/upsample with float scale_factor changed in 1.6.0 to align with other frameworks/libraries, and uses scale_factor directly, instead of relying on the computed output size. If you wish to keep the old behavior, please set recompute_scale_factor=True. See the documentation of nn.Upsample for details. \n",
            "  warnings.warn(\"The default behavior for interpolate/upsample with float scale_factor changed \"\n",
            "/usr/local/lib/python3.6/dist-packages/torch/nn/functional.py:3000: UserWarning: The default behavior for interpolate/upsample with float scale_factor changed in 1.6.0 to align with other frameworks/libraries, and uses scale_factor directly, instead of relying on the computed output size. If you wish to keep the old behavior, please set recompute_scale_factor=True. See the documentation of nn.Upsample for details. \n",
            "  warnings.warn(\"The default behavior for interpolate/upsample with float scale_factor changed \"\n",
            "/usr/local/lib/python3.6/dist-packages/torch/nn/functional.py:3000: UserWarning: The default behavior for interpolate/upsample with float scale_factor changed in 1.6.0 to align with other frameworks/libraries, and uses scale_factor directly, instead of relying on the computed output size. If you wish to keep the old behavior, please set recompute_scale_factor=True. See the documentation of nn.Upsample for details. \n",
            "  warnings.warn(\"The default behavior for interpolate/upsample with float scale_factor changed \"\n"
          ],
          "name": "stderr"
        },
        {
          "output_type": "stream",
          "text": [
            "Learn path /root/.fastai/data/imagewoof2\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "display_data",
          "data": {
            "text/html": [
              "<table border=\"1\" class=\"dataframe\">\n",
              "  <thead>\n",
              "    <tr style=\"text-align: left;\">\n",
              "      <th>epoch</th>\n",
              "      <th>train_loss</th>\n",
              "      <th>valid_loss</th>\n",
              "      <th>accuracy</th>\n",
              "      <th>top_k_accuracy</th>\n",
              "      <th>time</th>\n",
              "    </tr>\n",
              "  </thead>\n",
              "  <tbody>\n",
              "    <tr>\n",
              "      <td>0</td>\n",
              "      <td>1.871541</td>\n",
              "      <td>1.774720</td>\n",
              "      <td>0.433444</td>\n",
              "      <td>0.893866</td>\n",
              "      <td>02:03</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>1</td>\n",
              "      <td>1.612151</td>\n",
              "      <td>1.499916</td>\n",
              "      <td>0.569611</td>\n",
              "      <td>0.930517</td>\n",
              "      <td>02:03</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>2</td>\n",
              "      <td>1.488557</td>\n",
              "      <td>1.334675</td>\n",
              "      <td>0.648257</td>\n",
              "      <td>0.952914</td>\n",
              "      <td>02:03</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>3</td>\n",
              "      <td>1.334738</td>\n",
              "      <td>1.277277</td>\n",
              "      <td>0.695088</td>\n",
              "      <td>0.959786</td>\n",
              "      <td>02:03</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>4</td>\n",
              "      <td>1.183207</td>\n",
              "      <td>1.132146</td>\n",
              "      <td>0.747773</td>\n",
              "      <td>0.972003</td>\n",
              "      <td>02:03</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>5</td>\n",
              "      <td>1.128833</td>\n",
              "      <td>1.092119</td>\n",
              "      <td>0.766862</td>\n",
              "      <td>0.973276</td>\n",
              "      <td>02:03</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>6</td>\n",
              "      <td>1.106625</td>\n",
              "      <td>1.037147</td>\n",
              "      <td>0.787223</td>\n",
              "      <td>0.977602</td>\n",
              "      <td>02:03</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>7</td>\n",
              "      <td>1.032819</td>\n",
              "      <td>1.011182</td>\n",
              "      <td>0.805294</td>\n",
              "      <td>0.976584</td>\n",
              "      <td>02:03</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>8</td>\n",
              "      <td>0.992151</td>\n",
              "      <td>0.986376</td>\n",
              "      <td>0.811402</td>\n",
              "      <td>0.978621</td>\n",
              "      <td>02:03</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>9</td>\n",
              "      <td>0.978412</td>\n",
              "      <td>0.983965</td>\n",
              "      <td>0.814457</td>\n",
              "      <td>0.979384</td>\n",
              "      <td>02:03</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>10</td>\n",
              "      <td>0.955973</td>\n",
              "      <td>0.970820</td>\n",
              "      <td>0.814966</td>\n",
              "      <td>0.977857</td>\n",
              "      <td>02:03</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>11</td>\n",
              "      <td>0.908099</td>\n",
              "      <td>0.937433</td>\n",
              "      <td>0.827437</td>\n",
              "      <td>0.981166</td>\n",
              "      <td>02:02</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>12</td>\n",
              "      <td>0.911595</td>\n",
              "      <td>0.943616</td>\n",
              "      <td>0.820820</td>\n",
              "      <td>0.984983</td>\n",
              "      <td>02:02</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>13</td>\n",
              "      <td>0.902741</td>\n",
              "      <td>0.922432</td>\n",
              "      <td>0.839145</td>\n",
              "      <td>0.979893</td>\n",
              "      <td>02:02</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>14</td>\n",
              "      <td>0.844144</td>\n",
              "      <td>0.934122</td>\n",
              "      <td>0.833800</td>\n",
              "      <td>0.981929</td>\n",
              "      <td>02:02</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>15</td>\n",
              "      <td>0.823884</td>\n",
              "      <td>0.884270</td>\n",
              "      <td>0.853907</td>\n",
              "      <td>0.985238</td>\n",
              "      <td>02:03</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>16</td>\n",
              "      <td>0.744207</td>\n",
              "      <td>0.864174</td>\n",
              "      <td>0.862815</td>\n",
              "      <td>0.982947</td>\n",
              "      <td>02:02</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>17</td>\n",
              "      <td>0.665210</td>\n",
              "      <td>0.841832</td>\n",
              "      <td>0.875286</td>\n",
              "      <td>0.985492</td>\n",
              "      <td>02:02</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>18</td>\n",
              "      <td>0.646330</td>\n",
              "      <td>0.822845</td>\n",
              "      <td>0.881140</td>\n",
              "      <td>0.983202</td>\n",
              "      <td>02:03</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>19</td>\n",
              "      <td>0.619679</td>\n",
              "      <td>0.820435</td>\n",
              "      <td>0.881140</td>\n",
              "      <td>0.983202</td>\n",
              "      <td>02:03</td>\n",
              "    </tr>\n",
              "  </tbody>\n",
              "</table>"
            ],
            "text/plain": [
              "<IPython.core.display.HTML object>"
            ]
          },
          "metadata": {
            "tags": []
          }
        },
        {
          "output_type": "stream",
          "text": [
            "/usr/local/lib/python3.6/dist-packages/torch/nn/functional.py:3000: UserWarning: The default behavior for interpolate/upsample with float scale_factor changed in 1.6.0 to align with other frameworks/libraries, and uses scale_factor directly, instead of relying on the computed output size. If you wish to keep the old behavior, please set recompute_scale_factor=True. See the documentation of nn.Upsample for details. \n",
            "  warnings.warn(\"The default behavior for interpolate/upsample with float scale_factor changed \"\n"
          ],
          "name": "stderr"
        },
        {
          "output_type": "stream",
          "text": [
            "20 epochs: 0.8818401545286179 (4 runs)\n",
            "data path   /root/.fastai/data/imagewoof2\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "stream",
          "text": [
            "/usr/local/lib/python3.6/dist-packages/torch/nn/functional.py:3000: UserWarning: The default behavior for interpolate/upsample with float scale_factor changed in 1.6.0 to align with other frameworks/libraries, and uses scale_factor directly, instead of relying on the computed output size. If you wish to keep the old behavior, please set recompute_scale_factor=True. See the documentation of nn.Upsample for details. \n",
            "  warnings.warn(\"The default behavior for interpolate/upsample with float scale_factor changed \"\n"
          ],
          "name": "stderr"
        },
        {
          "output_type": "stream",
          "text": [
            "Learn path /root/.fastai/data/imagewoof2\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "display_data",
          "data": {
            "text/html": [
              "<table border=\"1\" class=\"dataframe\">\n",
              "  <thead>\n",
              "    <tr style=\"text-align: left;\">\n",
              "      <th>epoch</th>\n",
              "      <th>train_loss</th>\n",
              "      <th>valid_loss</th>\n",
              "      <th>accuracy</th>\n",
              "      <th>top_k_accuracy</th>\n",
              "      <th>time</th>\n",
              "    </tr>\n",
              "  </thead>\n",
              "  <tbody>\n",
              "    <tr>\n",
              "      <td>0</td>\n",
              "      <td>1.870331</td>\n",
              "      <td>1.706873</td>\n",
              "      <td>0.463986</td>\n",
              "      <td>0.887758</td>\n",
              "      <td>02:03</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>1</td>\n",
              "      <td>1.605841</td>\n",
              "      <td>1.463672</td>\n",
              "      <td>0.582336</td>\n",
              "      <td>0.940697</td>\n",
              "      <td>02:03</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>2</td>\n",
              "      <td>1.482729</td>\n",
              "      <td>1.401461</td>\n",
              "      <td>0.622041</td>\n",
              "      <td>0.938407</td>\n",
              "      <td>02:03</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>3</td>\n",
              "      <td>1.342088</td>\n",
              "      <td>1.214602</td>\n",
              "      <td>0.710868</td>\n",
              "      <td>0.962586</td>\n",
              "      <td>02:03</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>4</td>\n",
              "      <td>1.218828</td>\n",
              "      <td>1.106326</td>\n",
              "      <td>0.757190</td>\n",
              "      <td>0.974294</td>\n",
              "      <td>02:03</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>5</td>\n",
              "      <td>1.181968</td>\n",
              "      <td>1.091660</td>\n",
              "      <td>0.780860</td>\n",
              "      <td>0.975566</td>\n",
              "      <td>02:03</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>6</td>\n",
              "      <td>1.120337</td>\n",
              "      <td>1.034804</td>\n",
              "      <td>0.796895</td>\n",
              "      <td>0.975566</td>\n",
              "      <td>02:03</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>7</td>\n",
              "      <td>1.044700</td>\n",
              "      <td>1.032736</td>\n",
              "      <td>0.799949</td>\n",
              "      <td>0.978875</td>\n",
              "      <td>02:03</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>8</td>\n",
              "      <td>1.020853</td>\n",
              "      <td>0.976485</td>\n",
              "      <td>0.806567</td>\n",
              "      <td>0.978366</td>\n",
              "      <td>02:04</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>9</td>\n",
              "      <td>0.939360</td>\n",
              "      <td>0.980164</td>\n",
              "      <td>0.822347</td>\n",
              "      <td>0.979893</td>\n",
              "      <td>02:04</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>10</td>\n",
              "      <td>0.940390</td>\n",
              "      <td>0.968670</td>\n",
              "      <td>0.816747</td>\n",
              "      <td>0.979130</td>\n",
              "      <td>02:03</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>11</td>\n",
              "      <td>0.921674</td>\n",
              "      <td>0.953763</td>\n",
              "      <td>0.826419</td>\n",
              "      <td>0.979384</td>\n",
              "      <td>02:03</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>12</td>\n",
              "      <td>0.888182</td>\n",
              "      <td>0.946272</td>\n",
              "      <td>0.824637</td>\n",
              "      <td>0.980911</td>\n",
              "      <td>02:03</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>13</td>\n",
              "      <td>0.876670</td>\n",
              "      <td>0.921528</td>\n",
              "      <td>0.833545</td>\n",
              "      <td>0.980911</td>\n",
              "      <td>02:03</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>14</td>\n",
              "      <td>0.842112</td>\n",
              "      <td>0.907638</td>\n",
              "      <td>0.843726</td>\n",
              "      <td>0.982693</td>\n",
              "      <td>02:03</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>15</td>\n",
              "      <td>0.782981</td>\n",
              "      <td>0.885523</td>\n",
              "      <td>0.853143</td>\n",
              "      <td>0.981675</td>\n",
              "      <td>02:04</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>16</td>\n",
              "      <td>0.746040</td>\n",
              "      <td>0.859214</td>\n",
              "      <td>0.864342</td>\n",
              "      <td>0.986256</td>\n",
              "      <td>02:03</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>17</td>\n",
              "      <td>0.685480</td>\n",
              "      <td>0.842431</td>\n",
              "      <td>0.874268</td>\n",
              "      <td>0.984474</td>\n",
              "      <td>02:03</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>18</td>\n",
              "      <td>0.642356</td>\n",
              "      <td>0.817457</td>\n",
              "      <td>0.878341</td>\n",
              "      <td>0.986511</td>\n",
              "      <td>02:03</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>19</td>\n",
              "      <td>0.631486</td>\n",
              "      <td>0.819457</td>\n",
              "      <td>0.879104</td>\n",
              "      <td>0.986002</td>\n",
              "      <td>02:03</td>\n",
              "    </tr>\n",
              "  </tbody>\n",
              "</table>"
            ],
            "text/plain": [
              "<IPython.core.display.HTML object>"
            ]
          },
          "metadata": {
            "tags": []
          }
        },
        {
          "output_type": "stream",
          "text": [
            "/usr/local/lib/python3.6/dist-packages/torch/nn/functional.py:3000: UserWarning: The default behavior for interpolate/upsample with float scale_factor changed in 1.6.0 to align with other frameworks/libraries, and uses scale_factor directly, instead of relying on the computed output size. If you wish to keep the old behavior, please set recompute_scale_factor=True. See the documentation of nn.Upsample for details. \n",
            "  warnings.warn(\"The default behavior for interpolate/upsample with float scale_factor changed \"\n"
          ],
          "name": "stderr"
        },
        {
          "output_type": "stream",
          "text": [
            "20 epochs: 0.8812929391860962 (5 runs)\n",
            "data path   /root/.fastai/data/imagewoof2\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "stream",
          "text": [
            "/usr/local/lib/python3.6/dist-packages/torch/nn/functional.py:3000: UserWarning: The default behavior for interpolate/upsample with float scale_factor changed in 1.6.0 to align with other frameworks/libraries, and uses scale_factor directly, instead of relying on the computed output size. If you wish to keep the old behavior, please set recompute_scale_factor=True. See the documentation of nn.Upsample for details. \n",
            "  warnings.warn(\"The default behavior for interpolate/upsample with float scale_factor changed \"\n",
            "/usr/local/lib/python3.6/dist-packages/torch/nn/functional.py:3000: UserWarning: The default behavior for interpolate/upsample with float scale_factor changed in 1.6.0 to align with other frameworks/libraries, and uses scale_factor directly, instead of relying on the computed output size. If you wish to keep the old behavior, please set recompute_scale_factor=True. See the documentation of nn.Upsample for details. \n",
            "  warnings.warn(\"The default behavior for interpolate/upsample with float scale_factor changed \"\n",
            "/usr/local/lib/python3.6/dist-packages/torch/nn/functional.py:3000: UserWarning: The default behavior for interpolate/upsample with float scale_factor changed in 1.6.0 to align with other frameworks/libraries, and uses scale_factor directly, instead of relying on the computed output size. If you wish to keep the old behavior, please set recompute_scale_factor=True. See the documentation of nn.Upsample for details. \n",
            "  warnings.warn(\"The default behavior for interpolate/upsample with float scale_factor changed \"\n"
          ],
          "name": "stderr"
        },
        {
          "output_type": "stream",
          "text": [
            "Learn path /root/.fastai/data/imagewoof2\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "display_data",
          "data": {
            "text/html": [
              "<table border=\"1\" class=\"dataframe\">\n",
              "  <thead>\n",
              "    <tr style=\"text-align: left;\">\n",
              "      <th>epoch</th>\n",
              "      <th>train_loss</th>\n",
              "      <th>valid_loss</th>\n",
              "      <th>accuracy</th>\n",
              "      <th>top_k_accuracy</th>\n",
              "      <th>time</th>\n",
              "    </tr>\n",
              "  </thead>\n",
              "  <tbody>\n",
              "    <tr>\n",
              "      <td>0</td>\n",
              "      <td>2.025291</td>\n",
              "      <td>1.790612</td>\n",
              "      <td>0.421736</td>\n",
              "      <td>0.877322</td>\n",
              "      <td>02:03</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>1</td>\n",
              "      <td>1.853151</td>\n",
              "      <td>1.561010</td>\n",
              "      <td>0.534742</td>\n",
              "      <td>0.926190</td>\n",
              "      <td>02:04</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>2</td>\n",
              "      <td>1.677373</td>\n",
              "      <td>1.472112</td>\n",
              "      <td>0.584373</td>\n",
              "      <td>0.934589</td>\n",
              "      <td>02:04</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>3</td>\n",
              "      <td>1.572342</td>\n",
              "      <td>1.286127</td>\n",
              "      <td>0.668872</td>\n",
              "      <td>0.961059</td>\n",
              "      <td>02:04</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>4</td>\n",
              "      <td>1.486564</td>\n",
              "      <td>1.192217</td>\n",
              "      <td>0.718249</td>\n",
              "      <td>0.966658</td>\n",
              "      <td>02:04</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>5</td>\n",
              "      <td>1.401715</td>\n",
              "      <td>1.188251</td>\n",
              "      <td>0.725121</td>\n",
              "      <td>0.964368</td>\n",
              "      <td>02:05</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>6</td>\n",
              "      <td>1.341894</td>\n",
              "      <td>1.092382</td>\n",
              "      <td>0.766098</td>\n",
              "      <td>0.969203</td>\n",
              "      <td>02:04</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>7</td>\n",
              "      <td>1.311690</td>\n",
              "      <td>1.039083</td>\n",
              "      <td>0.783151</td>\n",
              "      <td>0.972003</td>\n",
              "      <td>02:04</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>8</td>\n",
              "      <td>1.298644</td>\n",
              "      <td>1.004073</td>\n",
              "      <td>0.805548</td>\n",
              "      <td>0.977093</td>\n",
              "      <td>02:03</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>9</td>\n",
              "      <td>1.245196</td>\n",
              "      <td>0.986797</td>\n",
              "      <td>0.804276</td>\n",
              "      <td>0.980148</td>\n",
              "      <td>02:03</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>10</td>\n",
              "      <td>1.225866</td>\n",
              "      <td>0.936840</td>\n",
              "      <td>0.830237</td>\n",
              "      <td>0.978875</td>\n",
              "      <td>02:04</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>11</td>\n",
              "      <td>1.148458</td>\n",
              "      <td>0.947408</td>\n",
              "      <td>0.831000</td>\n",
              "      <td>0.978621</td>\n",
              "      <td>02:04</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>12</td>\n",
              "      <td>1.192184</td>\n",
              "      <td>0.949757</td>\n",
              "      <td>0.823874</td>\n",
              "      <td>0.982184</td>\n",
              "      <td>02:03</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>13</td>\n",
              "      <td>1.157955</td>\n",
              "      <td>0.928381</td>\n",
              "      <td>0.830491</td>\n",
              "      <td>0.978621</td>\n",
              "      <td>02:03</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>14</td>\n",
              "      <td>1.154066</td>\n",
              "      <td>0.927390</td>\n",
              "      <td>0.833545</td>\n",
              "      <td>0.981420</td>\n",
              "      <td>02:03</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>15</td>\n",
              "      <td>1.118791</td>\n",
              "      <td>0.882916</td>\n",
              "      <td>0.851871</td>\n",
              "      <td>0.983456</td>\n",
              "      <td>02:04</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>16</td>\n",
              "      <td>1.127364</td>\n",
              "      <td>0.904534</td>\n",
              "      <td>0.835836</td>\n",
              "      <td>0.979639</td>\n",
              "      <td>02:04</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>17</td>\n",
              "      <td>1.099175</td>\n",
              "      <td>0.899894</td>\n",
              "      <td>0.845253</td>\n",
              "      <td>0.981929</td>\n",
              "      <td>02:03</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>18</td>\n",
              "      <td>1.060654</td>\n",
              "      <td>0.893597</td>\n",
              "      <td>0.841945</td>\n",
              "      <td>0.980148</td>\n",
              "      <td>02:04</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>19</td>\n",
              "      <td>1.087542</td>\n",
              "      <td>0.885149</td>\n",
              "      <td>0.849326</td>\n",
              "      <td>0.979893</td>\n",
              "      <td>02:03</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>20</td>\n",
              "      <td>1.058267</td>\n",
              "      <td>0.903536</td>\n",
              "      <td>0.841945</td>\n",
              "      <td>0.982438</td>\n",
              "      <td>02:04</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>21</td>\n",
              "      <td>1.055340</td>\n",
              "      <td>0.915604</td>\n",
              "      <td>0.839908</td>\n",
              "      <td>0.980911</td>\n",
              "      <td>02:04</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>22</td>\n",
              "      <td>1.014739</td>\n",
              "      <td>0.892508</td>\n",
              "      <td>0.851362</td>\n",
              "      <td>0.979893</td>\n",
              "      <td>02:04</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>23</td>\n",
              "      <td>1.052341</td>\n",
              "      <td>0.882116</td>\n",
              "      <td>0.851107</td>\n",
              "      <td>0.979893</td>\n",
              "      <td>02:05</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>24</td>\n",
              "      <td>1.001812</td>\n",
              "      <td>0.842072</td>\n",
              "      <td>0.871214</td>\n",
              "      <td>0.983202</td>\n",
              "      <td>02:04</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>25</td>\n",
              "      <td>0.965332</td>\n",
              "      <td>0.862399</td>\n",
              "      <td>0.858997</td>\n",
              "      <td>0.982438</td>\n",
              "      <td>02:05</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>26</td>\n",
              "      <td>0.984608</td>\n",
              "      <td>0.860326</td>\n",
              "      <td>0.862051</td>\n",
              "      <td>0.982438</td>\n",
              "      <td>02:05</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>27</td>\n",
              "      <td>0.984587</td>\n",
              "      <td>0.866921</td>\n",
              "      <td>0.860015</td>\n",
              "      <td>0.983202</td>\n",
              "      <td>02:05</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>28</td>\n",
              "      <td>0.965047</td>\n",
              "      <td>0.896195</td>\n",
              "      <td>0.857979</td>\n",
              "      <td>0.977857</td>\n",
              "      <td>02:05</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>29</td>\n",
              "      <td>0.967591</td>\n",
              "      <td>0.851905</td>\n",
              "      <td>0.860015</td>\n",
              "      <td>0.980402</td>\n",
              "      <td>02:05</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>30</td>\n",
              "      <td>0.961998</td>\n",
              "      <td>0.862761</td>\n",
              "      <td>0.859506</td>\n",
              "      <td>0.977348</td>\n",
              "      <td>02:05</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>31</td>\n",
              "      <td>0.954962</td>\n",
              "      <td>0.874103</td>\n",
              "      <td>0.860524</td>\n",
              "      <td>0.981420</td>\n",
              "      <td>02:05</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>32</td>\n",
              "      <td>0.936448</td>\n",
              "      <td>0.823223</td>\n",
              "      <td>0.872741</td>\n",
              "      <td>0.983711</td>\n",
              "      <td>02:05</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>33</td>\n",
              "      <td>0.932468</td>\n",
              "      <td>0.859825</td>\n",
              "      <td>0.858234</td>\n",
              "      <td>0.980911</td>\n",
              "      <td>02:05</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>34</td>\n",
              "      <td>0.930578</td>\n",
              "      <td>0.874692</td>\n",
              "      <td>0.853652</td>\n",
              "      <td>0.981420</td>\n",
              "      <td>02:05</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>35</td>\n",
              "      <td>0.908565</td>\n",
              "      <td>0.858028</td>\n",
              "      <td>0.862815</td>\n",
              "      <td>0.981929</td>\n",
              "      <td>02:05</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>36</td>\n",
              "      <td>0.946583</td>\n",
              "      <td>0.828974</td>\n",
              "      <td>0.876559</td>\n",
              "      <td>0.982184</td>\n",
              "      <td>02:04</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>37</td>\n",
              "      <td>0.938865</td>\n",
              "      <td>0.841148</td>\n",
              "      <td>0.868160</td>\n",
              "      <td>0.983965</td>\n",
              "      <td>02:04</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>38</td>\n",
              "      <td>0.927287</td>\n",
              "      <td>0.849436</td>\n",
              "      <td>0.863578</td>\n",
              "      <td>0.981420</td>\n",
              "      <td>02:04</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>39</td>\n",
              "      <td>0.929606</td>\n",
              "      <td>0.868787</td>\n",
              "      <td>0.863069</td>\n",
              "      <td>0.977348</td>\n",
              "      <td>02:04</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>40</td>\n",
              "      <td>0.906982</td>\n",
              "      <td>0.870114</td>\n",
              "      <td>0.853398</td>\n",
              "      <td>0.979893</td>\n",
              "      <td>02:04</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>41</td>\n",
              "      <td>0.913415</td>\n",
              "      <td>0.844245</td>\n",
              "      <td>0.867651</td>\n",
              "      <td>0.982184</td>\n",
              "      <td>02:04</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>42</td>\n",
              "      <td>0.933754</td>\n",
              "      <td>0.860128</td>\n",
              "      <td>0.866633</td>\n",
              "      <td>0.978366</td>\n",
              "      <td>02:04</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>43</td>\n",
              "      <td>0.873758</td>\n",
              "      <td>0.836159</td>\n",
              "      <td>0.866887</td>\n",
              "      <td>0.981675</td>\n",
              "      <td>02:04</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>44</td>\n",
              "      <td>0.910283</td>\n",
              "      <td>0.875709</td>\n",
              "      <td>0.854161</td>\n",
              "      <td>0.977602</td>\n",
              "      <td>02:05</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>45</td>\n",
              "      <td>0.911223</td>\n",
              "      <td>0.867666</td>\n",
              "      <td>0.864597</td>\n",
              "      <td>0.974039</td>\n",
              "      <td>02:05</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>46</td>\n",
              "      <td>0.899522</td>\n",
              "      <td>0.857125</td>\n",
              "      <td>0.860524</td>\n",
              "      <td>0.980402</td>\n",
              "      <td>02:04</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>47</td>\n",
              "      <td>0.874175</td>\n",
              "      <td>0.839441</td>\n",
              "      <td>0.870450</td>\n",
              "      <td>0.979639</td>\n",
              "      <td>02:04</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>48</td>\n",
              "      <td>0.919947</td>\n",
              "      <td>0.848530</td>\n",
              "      <td>0.870705</td>\n",
              "      <td>0.982184</td>\n",
              "      <td>02:04</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>49</td>\n",
              "      <td>0.929820</td>\n",
              "      <td>0.867067</td>\n",
              "      <td>0.863833</td>\n",
              "      <td>0.980657</td>\n",
              "      <td>02:04</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>50</td>\n",
              "      <td>0.913632</td>\n",
              "      <td>0.865630</td>\n",
              "      <td>0.862560</td>\n",
              "      <td>0.981675</td>\n",
              "      <td>02:04</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>51</td>\n",
              "      <td>0.903617</td>\n",
              "      <td>0.848465</td>\n",
              "      <td>0.868923</td>\n",
              "      <td>0.982184</td>\n",
              "      <td>02:04</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>52</td>\n",
              "      <td>0.869808</td>\n",
              "      <td>0.849793</td>\n",
              "      <td>0.869178</td>\n",
              "      <td>0.978875</td>\n",
              "      <td>02:05</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>53</td>\n",
              "      <td>0.881002</td>\n",
              "      <td>0.838265</td>\n",
              "      <td>0.874523</td>\n",
              "      <td>0.979639</td>\n",
              "      <td>02:05</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>54</td>\n",
              "      <td>0.881318</td>\n",
              "      <td>0.849539</td>\n",
              "      <td>0.866887</td>\n",
              "      <td>0.982184</td>\n",
              "      <td>02:06</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>55</td>\n",
              "      <td>0.877998</td>\n",
              "      <td>0.822807</td>\n",
              "      <td>0.874014</td>\n",
              "      <td>0.983202</td>\n",
              "      <td>02:06</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>56</td>\n",
              "      <td>0.907378</td>\n",
              "      <td>0.837087</td>\n",
              "      <td>0.866378</td>\n",
              "      <td>0.980402</td>\n",
              "      <td>02:06</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>57</td>\n",
              "      <td>0.874204</td>\n",
              "      <td>0.860396</td>\n",
              "      <td>0.865360</td>\n",
              "      <td>0.979384</td>\n",
              "      <td>02:06</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>58</td>\n",
              "      <td>0.871205</td>\n",
              "      <td>0.851682</td>\n",
              "      <td>0.869687</td>\n",
              "      <td>0.980402</td>\n",
              "      <td>02:06</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>59</td>\n",
              "      <td>0.877429</td>\n",
              "      <td>0.855481</td>\n",
              "      <td>0.872487</td>\n",
              "      <td>0.975566</td>\n",
              "      <td>02:06</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>60</td>\n",
              "      <td>0.893050</td>\n",
              "      <td>0.822805</td>\n",
              "      <td>0.875795</td>\n",
              "      <td>0.982693</td>\n",
              "      <td>02:06</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>61</td>\n",
              "      <td>0.859529</td>\n",
              "      <td>0.848650</td>\n",
              "      <td>0.867396</td>\n",
              "      <td>0.980911</td>\n",
              "      <td>02:06</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>62</td>\n",
              "      <td>0.850778</td>\n",
              "      <td>0.840895</td>\n",
              "      <td>0.873250</td>\n",
              "      <td>0.980148</td>\n",
              "      <td>02:04</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>63</td>\n",
              "      <td>0.860991</td>\n",
              "      <td>0.835343</td>\n",
              "      <td>0.877577</td>\n",
              "      <td>0.982693</td>\n",
              "      <td>02:05</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>64</td>\n",
              "      <td>0.856092</td>\n",
              "      <td>0.827584</td>\n",
              "      <td>0.870960</td>\n",
              "      <td>0.982438</td>\n",
              "      <td>02:05</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>65</td>\n",
              "      <td>0.854650</td>\n",
              "      <td>0.835667</td>\n",
              "      <td>0.874523</td>\n",
              "      <td>0.980911</td>\n",
              "      <td>02:04</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>66</td>\n",
              "      <td>0.825089</td>\n",
              "      <td>0.821144</td>\n",
              "      <td>0.876559</td>\n",
              "      <td>0.980402</td>\n",
              "      <td>02:05</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>67</td>\n",
              "      <td>0.845227</td>\n",
              "      <td>0.823252</td>\n",
              "      <td>0.875032</td>\n",
              "      <td>0.981929</td>\n",
              "      <td>02:04</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>68</td>\n",
              "      <td>0.825244</td>\n",
              "      <td>0.801444</td>\n",
              "      <td>0.884194</td>\n",
              "      <td>0.985238</td>\n",
              "      <td>02:05</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>69</td>\n",
              "      <td>0.817961</td>\n",
              "      <td>0.807397</td>\n",
              "      <td>0.880377</td>\n",
              "      <td>0.982947</td>\n",
              "      <td>02:05</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>70</td>\n",
              "      <td>0.837903</td>\n",
              "      <td>0.789813</td>\n",
              "      <td>0.894375</td>\n",
              "      <td>0.981675</td>\n",
              "      <td>02:05</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>71</td>\n",
              "      <td>0.805418</td>\n",
              "      <td>0.791875</td>\n",
              "      <td>0.890557</td>\n",
              "      <td>0.981929</td>\n",
              "      <td>02:05</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>72</td>\n",
              "      <td>0.785321</td>\n",
              "      <td>0.781310</td>\n",
              "      <td>0.894121</td>\n",
              "      <td>0.986256</td>\n",
              "      <td>02:05</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>73</td>\n",
              "      <td>0.801562</td>\n",
              "      <td>0.779861</td>\n",
              "      <td>0.895648</td>\n",
              "      <td>0.984729</td>\n",
              "      <td>02:05</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>74</td>\n",
              "      <td>0.795155</td>\n",
              "      <td>0.772529</td>\n",
              "      <td>0.898193</td>\n",
              "      <td>0.986765</td>\n",
              "      <td>02:05</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>75</td>\n",
              "      <td>0.791312</td>\n",
              "      <td>0.769824</td>\n",
              "      <td>0.899720</td>\n",
              "      <td>0.988038</td>\n",
              "      <td>02:05</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>76</td>\n",
              "      <td>0.786175</td>\n",
              "      <td>0.769678</td>\n",
              "      <td>0.898193</td>\n",
              "      <td>0.986765</td>\n",
              "      <td>02:05</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>77</td>\n",
              "      <td>0.788576</td>\n",
              "      <td>0.769066</td>\n",
              "      <td>0.901247</td>\n",
              "      <td>0.987020</td>\n",
              "      <td>02:05</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>78</td>\n",
              "      <td>0.767112</td>\n",
              "      <td>0.764730</td>\n",
              "      <td>0.899466</td>\n",
              "      <td>0.989565</td>\n",
              "      <td>02:05</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>79</td>\n",
              "      <td>0.773228</td>\n",
              "      <td>0.764957</td>\n",
              "      <td>0.903029</td>\n",
              "      <td>0.987020</td>\n",
              "      <td>02:06</td>\n",
              "    </tr>\n",
              "  </tbody>\n",
              "</table>"
            ],
            "text/plain": [
              "<IPython.core.display.HTML object>"
            ]
          },
          "metadata": {
            "tags": []
          }
        },
        {
          "output_type": "stream",
          "text": [
            "/usr/local/lib/python3.6/dist-packages/torch/nn/functional.py:3000: UserWarning: The default behavior for interpolate/upsample with float scale_factor changed in 1.6.0 to align with other frameworks/libraries, and uses scale_factor directly, instead of relying on the computed output size. If you wish to keep the old behavior, please set recompute_scale_factor=True. See the documentation of nn.Upsample for details. \n",
            "  warnings.warn(\"The default behavior for interpolate/upsample with float scale_factor changed \"\n"
          ],
          "name": "stderr"
        },
        {
          "output_type": "stream",
          "text": [
            "80 epochs: 0.9030287861824036 (1 runs)\n",
            "depth multiplier= {80: 0.9030287861824036}\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "19IB0S2G2N4E",
        "colab_type": "text"
      },
      "source": [
        "experimenting with channel multiplier"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "F4Vy0Vo42MMi",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 1000
        },
        "outputId": "9a266d17-214c-4f4c-9ae3-ed62ee6b9680"
      },
      "source": [
        "for epochs in [80]:\n",
        "    mixup=0 if epochs<=20 else 0.2\n",
        "    learn = get_learn(model=model, size=192, bs=16, mixup=mixup)\n",
        "    learn.fit_fc(epochs, lr=4e-3, moms=(0.95,0.95), start_pct=0.72)\n",
        "    res += [learn.recorder.metrics[-1][0].item()]\n",
        "print([round(x, 6) for x in res], sum(res)/len(res))"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Downloading https://s3.amazonaws.com/fast-ai-imageclas/imagewoof2.tgz\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "display_data",
          "data": {
            "text/html": [
              ""
            ],
            "text/plain": [
              "<IPython.core.display.HTML object>"
            ]
          },
          "metadata": {
            "tags": []
          }
        },
        {
          "output_type": "stream",
          "text": [
            "data path   /root/.fastai/data/imagewoof2\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "stream",
          "text": [
            "/usr/local/lib/python3.6/dist-packages/torch/nn/functional.py:3000: UserWarning: The default behavior for interpolate/upsample with float scale_factor changed in 1.6.0 to align with other frameworks/libraries, and uses scale_factor directly, instead of relying on the computed output size. If you wish to keep the old behavior, please set recompute_scale_factor=True. See the documentation of nn.Upsample for details. \n",
            "  warnings.warn(\"The default behavior for interpolate/upsample with float scale_factor changed \"\n",
            "/usr/local/lib/python3.6/dist-packages/torch/nn/functional.py:3000: UserWarning: The default behavior for interpolate/upsample with float scale_factor changed in 1.6.0 to align with other frameworks/libraries, and uses scale_factor directly, instead of relying on the computed output size. If you wish to keep the old behavior, please set recompute_scale_factor=True. See the documentation of nn.Upsample for details. \n",
            "  warnings.warn(\"The default behavior for interpolate/upsample with float scale_factor changed \"\n",
            "/usr/local/lib/python3.6/dist-packages/torch/nn/functional.py:3000: UserWarning: The default behavior for interpolate/upsample with float scale_factor changed in 1.6.0 to align with other frameworks/libraries, and uses scale_factor directly, instead of relying on the computed output size. If you wish to keep the old behavior, please set recompute_scale_factor=True. See the documentation of nn.Upsample for details. \n",
            "  warnings.warn(\"The default behavior for interpolate/upsample with float scale_factor changed \"\n",
            "/usr/local/lib/python3.6/dist-packages/torch/nn/functional.py:3000: UserWarning: The default behavior for interpolate/upsample with float scale_factor changed in 1.6.0 to align with other frameworks/libraries, and uses scale_factor directly, instead of relying on the computed output size. If you wish to keep the old behavior, please set recompute_scale_factor=True. See the documentation of nn.Upsample for details. \n",
            "  warnings.warn(\"The default behavior for interpolate/upsample with float scale_factor changed \"\n",
            "/usr/local/lib/python3.6/dist-packages/torch/nn/functional.py:3000: UserWarning: The default behavior for interpolate/upsample with float scale_factor changed in 1.6.0 to align with other frameworks/libraries, and uses scale_factor directly, instead of relying on the computed output size. If you wish to keep the old behavior, please set recompute_scale_factor=True. See the documentation of nn.Upsample for details. \n",
            "  warnings.warn(\"The default behavior for interpolate/upsample with float scale_factor changed \"\n"
          ],
          "name": "stderr"
        },
        {
          "output_type": "stream",
          "text": [
            "Learn path /root/.fastai/data/imagewoof2\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "display_data",
          "data": {
            "text/html": [
              "<table border=\"1\" class=\"dataframe\">\n",
              "  <thead>\n",
              "    <tr style=\"text-align: left;\">\n",
              "      <th>epoch</th>\n",
              "      <th>train_loss</th>\n",
              "      <th>valid_loss</th>\n",
              "      <th>accuracy</th>\n",
              "      <th>top_k_accuracy</th>\n",
              "      <th>time</th>\n",
              "    </tr>\n",
              "  </thead>\n",
              "  <tbody>\n",
              "    <tr>\n",
              "      <td>0</td>\n",
              "      <td>2.025110</td>\n",
              "      <td>1.872146</td>\n",
              "      <td>0.425554</td>\n",
              "      <td>0.871978</td>\n",
              "      <td>03:22</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>1</td>\n",
              "      <td>1.808593</td>\n",
              "      <td>1.560736</td>\n",
              "      <td>0.539577</td>\n",
              "      <td>0.931026</td>\n",
              "      <td>03:23</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>2</td>\n",
              "      <td>1.645650</td>\n",
              "      <td>1.374737</td>\n",
              "      <td>0.633240</td>\n",
              "      <td>0.948842</td>\n",
              "      <td>03:24</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>3</td>\n",
              "      <td>1.537533</td>\n",
              "      <td>1.259173</td>\n",
              "      <td>0.687452</td>\n",
              "      <td>0.959786</td>\n",
              "      <td>03:24</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>4</td>\n",
              "      <td>1.479700</td>\n",
              "      <td>1.155471</td>\n",
              "      <td>0.737338</td>\n",
              "      <td>0.969712</td>\n",
              "      <td>03:24</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>5</td>\n",
              "      <td>1.411763</td>\n",
              "      <td>1.107171</td>\n",
              "      <td>0.767625</td>\n",
              "      <td>0.968949</td>\n",
              "      <td>03:24</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>6</td>\n",
              "      <td>1.377335</td>\n",
              "      <td>1.080276</td>\n",
              "      <td>0.770425</td>\n",
              "      <td>0.969458</td>\n",
              "      <td>03:24</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>7</td>\n",
              "      <td>1.304939</td>\n",
              "      <td>1.025715</td>\n",
              "      <td>0.804021</td>\n",
              "      <td>0.973530</td>\n",
              "      <td>03:24</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>8</td>\n",
              "      <td>1.264274</td>\n",
              "      <td>0.998292</td>\n",
              "      <td>0.799949</td>\n",
              "      <td>0.977857</td>\n",
              "      <td>03:24</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>9</td>\n",
              "      <td>1.258094</td>\n",
              "      <td>0.983434</td>\n",
              "      <td>0.810130</td>\n",
              "      <td>0.981166</td>\n",
              "      <td>03:24</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>10</td>\n",
              "      <td>1.204922</td>\n",
              "      <td>0.945384</td>\n",
              "      <td>0.825655</td>\n",
              "      <td>0.981675</td>\n",
              "      <td>03:23</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>11</td>\n",
              "      <td>1.201390</td>\n",
              "      <td>0.963622</td>\n",
              "      <td>0.822347</td>\n",
              "      <td>0.978875</td>\n",
              "      <td>03:22</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>12</td>\n",
              "      <td>1.180682</td>\n",
              "      <td>0.934115</td>\n",
              "      <td>0.827182</td>\n",
              "      <td>0.979639</td>\n",
              "      <td>03:22</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>13</td>\n",
              "      <td>1.144559</td>\n",
              "      <td>0.948995</td>\n",
              "      <td>0.823619</td>\n",
              "      <td>0.980402</td>\n",
              "      <td>03:21</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>14</td>\n",
              "      <td>1.117869</td>\n",
              "      <td>0.939883</td>\n",
              "      <td>0.833545</td>\n",
              "      <td>0.977093</td>\n",
              "      <td>03:22</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>15</td>\n",
              "      <td>1.118074</td>\n",
              "      <td>0.901963</td>\n",
              "      <td>0.843726</td>\n",
              "      <td>0.979893</td>\n",
              "      <td>03:23</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>16</td>\n",
              "      <td>1.061652</td>\n",
              "      <td>0.868441</td>\n",
              "      <td>0.853398</td>\n",
              "      <td>0.984220</td>\n",
              "      <td>03:22</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>17</td>\n",
              "      <td>1.070120</td>\n",
              "      <td>0.871512</td>\n",
              "      <td>0.855943</td>\n",
              "      <td>0.983711</td>\n",
              "      <td>03:22</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>18</td>\n",
              "      <td>1.067379</td>\n",
              "      <td>0.894865</td>\n",
              "      <td>0.848307</td>\n",
              "      <td>0.981675</td>\n",
              "      <td>03:21</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>19</td>\n",
              "      <td>1.042896</td>\n",
              "      <td>0.895030</td>\n",
              "      <td>0.839654</td>\n",
              "      <td>0.983456</td>\n",
              "      <td>03:21</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>20</td>\n",
              "      <td>1.043057</td>\n",
              "      <td>0.862473</td>\n",
              "      <td>0.865106</td>\n",
              "      <td>0.983456</td>\n",
              "      <td>03:23</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>21</td>\n",
              "      <td>1.008901</td>\n",
              "      <td>0.888008</td>\n",
              "      <td>0.855179</td>\n",
              "      <td>0.980911</td>\n",
              "      <td>03:24</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>22</td>\n",
              "      <td>0.998210</td>\n",
              "      <td>0.870088</td>\n",
              "      <td>0.861542</td>\n",
              "      <td>0.980911</td>\n",
              "      <td>03:24</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>23</td>\n",
              "      <td>1.015248</td>\n",
              "      <td>0.863958</td>\n",
              "      <td>0.862560</td>\n",
              "      <td>0.981166</td>\n",
              "      <td>03:24</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>24</td>\n",
              "      <td>1.004017</td>\n",
              "      <td>0.846060</td>\n",
              "      <td>0.870196</td>\n",
              "      <td>0.983456</td>\n",
              "      <td>03:24</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>25</td>\n",
              "      <td>0.992488</td>\n",
              "      <td>0.852475</td>\n",
              "      <td>0.868669</td>\n",
              "      <td>0.984474</td>\n",
              "      <td>03:24</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>26</td>\n",
              "      <td>0.989068</td>\n",
              "      <td>0.866711</td>\n",
              "      <td>0.858743</td>\n",
              "      <td>0.981420</td>\n",
              "      <td>03:25</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>27</td>\n",
              "      <td>0.980585</td>\n",
              "      <td>0.856832</td>\n",
              "      <td>0.867396</td>\n",
              "      <td>0.980148</td>\n",
              "      <td>03:24</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>28</td>\n",
              "      <td>0.995665</td>\n",
              "      <td>0.859161</td>\n",
              "      <td>0.862306</td>\n",
              "      <td>0.981420</td>\n",
              "      <td>03:24</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>29</td>\n",
              "      <td>0.961205</td>\n",
              "      <td>0.889600</td>\n",
              "      <td>0.849835</td>\n",
              "      <td>0.977602</td>\n",
              "      <td>03:23</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>30</td>\n",
              "      <td>0.948247</td>\n",
              "      <td>0.837389</td>\n",
              "      <td>0.875795</td>\n",
              "      <td>0.981420</td>\n",
              "      <td>03:22</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>31</td>\n",
              "      <td>0.974153</td>\n",
              "      <td>0.850531</td>\n",
              "      <td>0.861033</td>\n",
              "      <td>0.984220</td>\n",
              "      <td>03:22</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>32</td>\n",
              "      <td>0.947286</td>\n",
              "      <td>0.868641</td>\n",
              "      <td>0.858234</td>\n",
              "      <td>0.979893</td>\n",
              "      <td>03:23</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>33</td>\n",
              "      <td>0.935188</td>\n",
              "      <td>0.851097</td>\n",
              "      <td>0.870705</td>\n",
              "      <td>0.979893</td>\n",
              "      <td>03:22</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>34</td>\n",
              "      <td>0.932013</td>\n",
              "      <td>0.861704</td>\n",
              "      <td>0.863833</td>\n",
              "      <td>0.979893</td>\n",
              "      <td>03:22</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>35</td>\n",
              "      <td>0.949764</td>\n",
              "      <td>0.860078</td>\n",
              "      <td>0.865869</td>\n",
              "      <td>0.978621</td>\n",
              "      <td>03:22</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>36</td>\n",
              "      <td>0.907514</td>\n",
              "      <td>0.831982</td>\n",
              "      <td>0.874014</td>\n",
              "      <td>0.982184</td>\n",
              "      <td>03:21</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>37</td>\n",
              "      <td>0.921996</td>\n",
              "      <td>0.843989</td>\n",
              "      <td>0.869432</td>\n",
              "      <td>0.979384</td>\n",
              "      <td>03:21</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>38</td>\n",
              "      <td>0.921889</td>\n",
              "      <td>0.840692</td>\n",
              "      <td>0.867396</td>\n",
              "      <td>0.982947</td>\n",
              "      <td>03:22</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>39</td>\n",
              "      <td>0.928468</td>\n",
              "      <td>0.846719</td>\n",
              "      <td>0.868669</td>\n",
              "      <td>0.979384</td>\n",
              "      <td>03:23</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>40</td>\n",
              "      <td>0.919130</td>\n",
              "      <td>0.845764</td>\n",
              "      <td>0.871214</td>\n",
              "      <td>0.983711</td>\n",
              "      <td>03:23</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>41</td>\n",
              "      <td>0.910258</td>\n",
              "      <td>0.872011</td>\n",
              "      <td>0.859252</td>\n",
              "      <td>0.982693</td>\n",
              "      <td>03:24</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>42</td>\n",
              "      <td>0.913763</td>\n",
              "      <td>0.843560</td>\n",
              "      <td>0.871978</td>\n",
              "      <td>0.981166</td>\n",
              "      <td>03:24</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>43</td>\n",
              "      <td>0.912821</td>\n",
              "      <td>0.829169</td>\n",
              "      <td>0.870960</td>\n",
              "      <td>0.980148</td>\n",
              "      <td>03:24</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>44</td>\n",
              "      <td>0.901302</td>\n",
              "      <td>0.866292</td>\n",
              "      <td>0.859252</td>\n",
              "      <td>0.979384</td>\n",
              "      <td>03:24</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>45</td>\n",
              "      <td>0.906105</td>\n",
              "      <td>0.853692</td>\n",
              "      <td>0.862560</td>\n",
              "      <td>0.980148</td>\n",
              "      <td>03:24</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>46</td>\n",
              "      <td>0.891446</td>\n",
              "      <td>0.843062</td>\n",
              "      <td>0.868923</td>\n",
              "      <td>0.981675</td>\n",
              "      <td>03:24</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>47</td>\n",
              "      <td>0.892736</td>\n",
              "      <td>0.835933</td>\n",
              "      <td>0.871723</td>\n",
              "      <td>0.982184</td>\n",
              "      <td>03:24</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>48</td>\n",
              "      <td>0.889130</td>\n",
              "      <td>0.829749</td>\n",
              "      <td>0.873250</td>\n",
              "      <td>0.982947</td>\n",
              "      <td>03:23</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>49</td>\n",
              "      <td>0.902879</td>\n",
              "      <td>0.842225</td>\n",
              "      <td>0.867905</td>\n",
              "      <td>0.980148</td>\n",
              "      <td>03:23</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>50</td>\n",
              "      <td>0.880292</td>\n",
              "      <td>0.848847</td>\n",
              "      <td>0.871469</td>\n",
              "      <td>0.982438</td>\n",
              "      <td>03:23</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>51</td>\n",
              "      <td>0.883266</td>\n",
              "      <td>0.819680</td>\n",
              "      <td>0.881904</td>\n",
              "      <td>0.980402</td>\n",
              "      <td>03:23</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>52</td>\n",
              "      <td>0.883673</td>\n",
              "      <td>0.819303</td>\n",
              "      <td>0.876050</td>\n",
              "      <td>0.982947</td>\n",
              "      <td>03:22</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>53</td>\n",
              "      <td>0.887304</td>\n",
              "      <td>0.817062</td>\n",
              "      <td>0.885467</td>\n",
              "      <td>0.979130</td>\n",
              "      <td>03:22</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>54</td>\n",
              "      <td>0.880600</td>\n",
              "      <td>0.815749</td>\n",
              "      <td>0.881395</td>\n",
              "      <td>0.984983</td>\n",
              "      <td>03:22</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>55</td>\n",
              "      <td>0.902961</td>\n",
              "      <td>0.819280</td>\n",
              "      <td>0.882922</td>\n",
              "      <td>0.980402</td>\n",
              "      <td>03:21</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>56</td>\n",
              "      <td>0.894687</td>\n",
              "      <td>0.825455</td>\n",
              "      <td>0.880886</td>\n",
              "      <td>0.980402</td>\n",
              "      <td>03:21</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>57</td>\n",
              "      <td>0.910651</td>\n",
              "      <td>0.835680</td>\n",
              "      <td>0.875032</td>\n",
              "      <td>0.978366</td>\n",
              "      <td>03:23</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>58</td>\n",
              "      <td>0.883622</td>\n",
              "      <td>0.825570</td>\n",
              "      <td>0.880377</td>\n",
              "      <td>0.982947</td>\n",
              "      <td>03:24</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>59</td>\n",
              "      <td>0.898169</td>\n",
              "      <td>0.839140</td>\n",
              "      <td>0.872996</td>\n",
              "      <td>0.976839</td>\n",
              "      <td>03:24</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>60</td>\n",
              "      <td>0.896925</td>\n",
              "      <td>0.833878</td>\n",
              "      <td>0.874777</td>\n",
              "      <td>0.981420</td>\n",
              "      <td>03:25</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>61</td>\n",
              "      <td>0.854105</td>\n",
              "      <td>0.821435</td>\n",
              "      <td>0.879868</td>\n",
              "      <td>0.978875</td>\n",
              "      <td>03:25</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>62</td>\n",
              "      <td>0.876227</td>\n",
              "      <td>0.814516</td>\n",
              "      <td>0.880631</td>\n",
              "      <td>0.980148</td>\n",
              "      <td>03:23</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>63</td>\n",
              "      <td>0.876859</td>\n",
              "      <td>0.826980</td>\n",
              "      <td>0.873759</td>\n",
              "      <td>0.982184</td>\n",
              "      <td>03:23</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>64</td>\n",
              "      <td>0.847674</td>\n",
              "      <td>0.807974</td>\n",
              "      <td>0.882413</td>\n",
              "      <td>0.981166</td>\n",
              "      <td>03:22</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>65</td>\n",
              "      <td>0.846976</td>\n",
              "      <td>0.816232</td>\n",
              "      <td>0.884958</td>\n",
              "      <td>0.978875</td>\n",
              "      <td>03:22</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>66</td>\n",
              "      <td>0.835230</td>\n",
              "      <td>0.815541</td>\n",
              "      <td>0.881140</td>\n",
              "      <td>0.981675</td>\n",
              "      <td>03:22</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>67</td>\n",
              "      <td>0.826839</td>\n",
              "      <td>0.789001</td>\n",
              "      <td>0.892848</td>\n",
              "      <td>0.982184</td>\n",
              "      <td>03:24</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>68</td>\n",
              "      <td>0.841047</td>\n",
              "      <td>0.795375</td>\n",
              "      <td>0.894630</td>\n",
              "      <td>0.979893</td>\n",
              "      <td>03:25</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>69</td>\n",
              "      <td>0.830836</td>\n",
              "      <td>0.783510</td>\n",
              "      <td>0.894630</td>\n",
              "      <td>0.982438</td>\n",
              "      <td>03:24</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>70</td>\n",
              "      <td>0.803616</td>\n",
              "      <td>0.778992</td>\n",
              "      <td>0.899466</td>\n",
              "      <td>0.980911</td>\n",
              "      <td>03:25</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>71</td>\n",
              "      <td>0.800945</td>\n",
              "      <td>0.779402</td>\n",
              "      <td>0.896411</td>\n",
              "      <td>0.983711</td>\n",
              "      <td>03:24</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>72</td>\n",
              "      <td>0.817554</td>\n",
              "      <td>0.775453</td>\n",
              "      <td>0.898193</td>\n",
              "      <td>0.982438</td>\n",
              "      <td>03:24</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>73</td>\n",
              "      <td>0.797518</td>\n",
              "      <td>0.762476</td>\n",
              "      <td>0.903283</td>\n",
              "      <td>0.983456</td>\n",
              "      <td>03:25</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>74</td>\n",
              "      <td>0.798686</td>\n",
              "      <td>0.751718</td>\n",
              "      <td>0.905319</td>\n",
              "      <td>0.984983</td>\n",
              "      <td>03:25</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>75</td>\n",
              "      <td>0.800507</td>\n",
              "      <td>0.755699</td>\n",
              "      <td>0.906083</td>\n",
              "      <td>0.983456</td>\n",
              "      <td>03:25</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>76</td>\n",
              "      <td>0.789509</td>\n",
              "      <td>0.754267</td>\n",
              "      <td>0.906083</td>\n",
              "      <td>0.982184</td>\n",
              "      <td>03:25</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>77</td>\n",
              "      <td>0.779718</td>\n",
              "      <td>0.751482</td>\n",
              "      <td>0.908374</td>\n",
              "      <td>0.982947</td>\n",
              "      <td>03:24</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>78</td>\n",
              "      <td>0.772080</td>\n",
              "      <td>0.751994</td>\n",
              "      <td>0.907865</td>\n",
              "      <td>0.983202</td>\n",
              "      <td>03:25</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>79</td>\n",
              "      <td>0.785868</td>\n",
              "      <td>0.750878</td>\n",
              "      <td>0.907865</td>\n",
              "      <td>0.983456</td>\n",
              "      <td>03:25</td>\n",
              "    </tr>\n",
              "  </tbody>\n",
              "</table>"
            ],
            "text/plain": [
              "<IPython.core.display.HTML object>"
            ]
          },
          "metadata": {
            "tags": []
          }
        },
        {
          "output_type": "stream",
          "text": [
            "/usr/local/lib/python3.6/dist-packages/torch/nn/functional.py:3000: UserWarning: The default behavior for interpolate/upsample with float scale_factor changed in 1.6.0 to align with other frameworks/libraries, and uses scale_factor directly, instead of relying on the computed output size. If you wish to keep the old behavior, please set recompute_scale_factor=True. See the documentation of nn.Upsample for details. \n",
            "  warnings.warn(\"The default behavior for interpolate/upsample with float scale_factor changed \"\n",
            "/usr/local/lib/python3.6/dist-packages/imagenette_experiments/train_utils.py:150: UserWarning: This overload of addcmul_ is deprecated:\n",
            "\taddcmul_(Number value, Tensor tensor1, Tensor tensor2)\n",
            "Consider using one of the following signatures instead:\n",
            "\taddcmul_(Tensor tensor1, Tensor tensor2, *, Number value) (Triggered internally at  /pytorch/torch/csrc/utils/python_arg_parser.cpp:766.)\n",
            "  exp_avg_sq.mul_(beta2).addcmul_(1 - beta2, grad, grad)\n"
          ],
          "name": "stderr"
        },
        {
          "output_type": "stream",
          "text": [
            "[0.907865] 0.9078645706176758\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "durHNpcnU_o1",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 34
        },
        "outputId": "ef82a273-5e26-43f5-bd19-aa0bc8af25d4"
      },
      "source": [
        "res"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "[0.8961567878723145]"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 14
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "DoFhBmfDGfh3",
        "colab_type": "text"
      },
      "source": [
        "\n",
        "\n",
        "### epochs=5, 5 runs\n",
        "\n",
        "no stride, bs=16, act=False, bn=False"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "iHyM9KildGdV",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 1000
        },
        "outputId": "7f261dba-4c91-4b93-b452-d87910ab2cef"
      },
      "source": [
        "for epochs in [5]*5:\n",
        "    mixup=0 if epochs<=20 else 0.2\n",
        "    learn = get_learn(model=model, size=192, bs=16, mixup=mixup)\n",
        "    learn.fit_fc(epochs, lr=4e-3, moms=(0.95,0.95), start_pct=0.72)\n",
        "    res += [learn.recorder.metrics[-1][0].item()]\n",
        "print([round(x, 6) for x in res], sum(res)/len(res))"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "data path   /root/.fastai/data/imagewoof2\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "stream",
          "text": [
            ""
          ],
          "name": "stderr"
        },
        {
          "output_type": "stream",
          "text": [
            "Learn path /root/.fastai/data/imagewoof2\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "display_data",
          "data": {
            "text/html": [
              "<table border=\"1\" class=\"dataframe\">\n",
              "  <thead>\n",
              "    <tr style=\"text-align: left;\">\n",
              "      <th>epoch</th>\n",
              "      <th>train_loss</th>\n",
              "      <th>valid_loss</th>\n",
              "      <th>accuracy</th>\n",
              "      <th>top_k_accuracy</th>\n",
              "      <th>time</th>\n",
              "    </tr>\n",
              "  </thead>\n",
              "  <tbody>\n",
              "    <tr>\n",
              "      <td>0</td>\n",
              "      <td>1.856656</td>\n",
              "      <td>1.887479</td>\n",
              "      <td>0.400611</td>\n",
              "      <td>0.835836</td>\n",
              "      <td>07:11</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>1</td>\n",
              "      <td>1.616110</td>\n",
              "      <td>1.414912</td>\n",
              "      <td>0.620005</td>\n",
              "      <td>0.937898</td>\n",
              "      <td>07:11</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>2</td>\n",
              "      <td>1.392715</td>\n",
              "      <td>1.291010</td>\n",
              "      <td>0.676508</td>\n",
              "      <td>0.959786</td>\n",
              "      <td>07:11</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>3</td>\n",
              "      <td>1.269173</td>\n",
              "      <td>1.158866</td>\n",
              "      <td>0.738356</td>\n",
              "      <td>0.969203</td>\n",
              "      <td>07:11</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>4</td>\n",
              "      <td>1.050011</td>\n",
              "      <td>1.002696</td>\n",
              "      <td>0.822601</td>\n",
              "      <td>0.982438</td>\n",
              "      <td>07:11</td>\n",
              "    </tr>\n",
              "  </tbody>\n",
              "</table>"
            ],
            "text/plain": [
              "<IPython.core.display.HTML object>"
            ]
          },
          "metadata": {
            "tags": []
          }
        },
        {
          "output_type": "stream",
          "text": [
            ""
          ],
          "name": "stderr"
        },
        {
          "output_type": "stream",
          "text": [
            "data path   /root/.fastai/data/imagewoof2\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "stream",
          "text": [
            ""
          ],
          "name": "stderr"
        },
        {
          "output_type": "stream",
          "text": [
            "Learn path /root/.fastai/data/imagewoof2\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "display_data",
          "data": {
            "text/html": [
              "<table border=\"1\" class=\"dataframe\">\n",
              "  <thead>\n",
              "    <tr style=\"text-align: left;\">\n",
              "      <th>epoch</th>\n",
              "      <th>train_loss</th>\n",
              "      <th>valid_loss</th>\n",
              "      <th>accuracy</th>\n",
              "      <th>top_k_accuracy</th>\n",
              "      <th>time</th>\n",
              "    </tr>\n",
              "  </thead>\n",
              "  <tbody>\n",
              "    <tr>\n",
              "      <td>0</td>\n",
              "      <td>1.844413</td>\n",
              "      <td>1.741399</td>\n",
              "      <td>0.445915</td>\n",
              "      <td>0.896920</td>\n",
              "      <td>07:11</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>1</td>\n",
              "      <td>1.561381</td>\n",
              "      <td>1.435193</td>\n",
              "      <td>0.601425</td>\n",
              "      <td>0.941970</td>\n",
              "      <td>07:11</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>2</td>\n",
              "      <td>1.398485</td>\n",
              "      <td>1.278045</td>\n",
              "      <td>0.679053</td>\n",
              "      <td>0.960041</td>\n",
              "      <td>07:11</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>3</td>\n",
              "      <td>1.272492</td>\n",
              "      <td>1.132732</td>\n",
              "      <td>0.754136</td>\n",
              "      <td>0.971494</td>\n",
              "      <td>07:11</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>4</td>\n",
              "      <td>1.058501</td>\n",
              "      <td>1.011058</td>\n",
              "      <td>0.813184</td>\n",
              "      <td>0.979384</td>\n",
              "      <td>07:11</td>\n",
              "    </tr>\n",
              "  </tbody>\n",
              "</table>"
            ],
            "text/plain": [
              "<IPython.core.display.HTML object>"
            ]
          },
          "metadata": {
            "tags": []
          }
        },
        {
          "output_type": "stream",
          "text": [
            ""
          ],
          "name": "stderr"
        },
        {
          "output_type": "stream",
          "text": [
            "data path   /root/.fastai/data/imagewoof2\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "stream",
          "text": [
            ""
          ],
          "name": "stderr"
        },
        {
          "output_type": "stream",
          "text": [
            "Learn path /root/.fastai/data/imagewoof2\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "display_data",
          "data": {
            "text/html": [
              "<table border=\"1\" class=\"dataframe\">\n",
              "  <thead>\n",
              "    <tr style=\"text-align: left;\">\n",
              "      <th>epoch</th>\n",
              "      <th>train_loss</th>\n",
              "      <th>valid_loss</th>\n",
              "      <th>accuracy</th>\n",
              "      <th>top_k_accuracy</th>\n",
              "      <th>time</th>\n",
              "    </tr>\n",
              "  </thead>\n",
              "  <tbody>\n",
              "    <tr>\n",
              "      <td>0</td>\n",
              "      <td>1.845423</td>\n",
              "      <td>1.730929</td>\n",
              "      <td>0.465258</td>\n",
              "      <td>0.905319</td>\n",
              "      <td>07:11</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>1</td>\n",
              "      <td>1.535990</td>\n",
              "      <td>1.419719</td>\n",
              "      <td>0.618223</td>\n",
              "      <td>0.943497</td>\n",
              "      <td>07:10</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>2</td>\n",
              "      <td>1.377856</td>\n",
              "      <td>1.211514</td>\n",
              "      <td>0.715449</td>\n",
              "      <td>0.961059</td>\n",
              "      <td>07:11</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>3</td>\n",
              "      <td>1.239511</td>\n",
              "      <td>1.131984</td>\n",
              "      <td>0.749046</td>\n",
              "      <td>0.972003</td>\n",
              "      <td>07:11</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>4</td>\n",
              "      <td>1.064996</td>\n",
              "      <td>1.000053</td>\n",
              "      <td>0.813948</td>\n",
              "      <td>0.983202</td>\n",
              "      <td>07:11</td>\n",
              "    </tr>\n",
              "  </tbody>\n",
              "</table>"
            ],
            "text/plain": [
              "<IPython.core.display.HTML object>"
            ]
          },
          "metadata": {
            "tags": []
          }
        },
        {
          "output_type": "stream",
          "text": [
            ""
          ],
          "name": "stderr"
        },
        {
          "output_type": "stream",
          "text": [
            "data path   /root/.fastai/data/imagewoof2\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "stream",
          "text": [
            ""
          ],
          "name": "stderr"
        },
        {
          "output_type": "stream",
          "text": [
            "Learn path /root/.fastai/data/imagewoof2\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "display_data",
          "data": {
            "text/html": [
              "<table border=\"1\" class=\"dataframe\">\n",
              "  <thead>\n",
              "    <tr style=\"text-align: left;\">\n",
              "      <th>epoch</th>\n",
              "      <th>train_loss</th>\n",
              "      <th>valid_loss</th>\n",
              "      <th>accuracy</th>\n",
              "      <th>top_k_accuracy</th>\n",
              "      <th>time</th>\n",
              "    </tr>\n",
              "  </thead>\n",
              "  <tbody>\n",
              "    <tr>\n",
              "      <td>0</td>\n",
              "      <td>1.850057</td>\n",
              "      <td>1.804008</td>\n",
              "      <td>0.418682</td>\n",
              "      <td>0.886740</td>\n",
              "      <td>07:10</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>1</td>\n",
              "      <td>1.552947</td>\n",
              "      <td>1.442660</td>\n",
              "      <td>0.597098</td>\n",
              "      <td>0.942479</td>\n",
              "      <td>07:11</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>2</td>\n",
              "      <td>1.384726</td>\n",
              "      <td>1.285309</td>\n",
              "      <td>0.679053</td>\n",
              "      <td>0.959023</td>\n",
              "      <td>07:10</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>3</td>\n",
              "      <td>1.249545</td>\n",
              "      <td>1.141736</td>\n",
              "      <td>0.743701</td>\n",
              "      <td>0.972258</td>\n",
              "      <td>07:11</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>4</td>\n",
              "      <td>1.064263</td>\n",
              "      <td>1.000224</td>\n",
              "      <td>0.822856</td>\n",
              "      <td>0.975566</td>\n",
              "      <td>07:11</td>\n",
              "    </tr>\n",
              "  </tbody>\n",
              "</table>"
            ],
            "text/plain": [
              "<IPython.core.display.HTML object>"
            ]
          },
          "metadata": {
            "tags": []
          }
        },
        {
          "output_type": "stream",
          "text": [
            ""
          ],
          "name": "stderr"
        },
        {
          "output_type": "stream",
          "text": [
            "data path   /root/.fastai/data/imagewoof2\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "stream",
          "text": [
            ""
          ],
          "name": "stderr"
        },
        {
          "output_type": "stream",
          "text": [
            "Learn path /root/.fastai/data/imagewoof2\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "display_data",
          "data": {
            "text/html": [
              "<table border=\"1\" class=\"dataframe\">\n",
              "  <thead>\n",
              "    <tr style=\"text-align: left;\">\n",
              "      <th>epoch</th>\n",
              "      <th>train_loss</th>\n",
              "      <th>valid_loss</th>\n",
              "      <th>accuracy</th>\n",
              "      <th>top_k_accuracy</th>\n",
              "      <th>time</th>\n",
              "    </tr>\n",
              "  </thead>\n",
              "  <tbody>\n",
              "    <tr>\n",
              "      <td>0</td>\n",
              "      <td>1.851692</td>\n",
              "      <td>1.754275</td>\n",
              "      <td>0.448206</td>\n",
              "      <td>0.880122</td>\n",
              "      <td>07:10</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>1</td>\n",
              "      <td>1.579250</td>\n",
              "      <td>1.476122</td>\n",
              "      <td>0.580300</td>\n",
              "      <td>0.939679</td>\n",
              "      <td>07:11</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>2</td>\n",
              "      <td>1.406376</td>\n",
              "      <td>1.354866</td>\n",
              "      <td>0.640366</td>\n",
              "      <td>0.956477</td>\n",
              "      <td>07:11</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>3</td>\n",
              "      <td>1.256694</td>\n",
              "      <td>1.167863</td>\n",
              "      <td>0.746246</td>\n",
              "      <td>0.968694</td>\n",
              "      <td>07:11</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>4</td>\n",
              "      <td>1.020720</td>\n",
              "      <td>1.003255</td>\n",
              "      <td>0.817002</td>\n",
              "      <td>0.981420</td>\n",
              "      <td>07:11</td>\n",
              "    </tr>\n",
              "  </tbody>\n",
              "</table>"
            ],
            "text/plain": [
              "<IPython.core.display.HTML object>"
            ]
          },
          "metadata": {
            "tags": []
          }
        },
        {
          "output_type": "stream",
          "text": [
            ""
          ],
          "name": "stderr"
        },
        {
          "output_type": "stream",
          "text": [
            "[0.822601, 0.813184, 0.813948, 0.822856, 0.817002] 0.8179180502891541\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "MxZ_Ot0ukmmk",
        "colab_type": "text"
      },
      "source": [
        "no stride, bs=16, act=True, bn=False"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "GZZJUufChTja",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 1000
        },
        "outputId": "db049bdc-1cdf-4ed7-896b-6a8eba24dfd3"
      },
      "source": [
        "for epochs in [5]*5:\n",
        "    mixup=0 if epochs<=20 else 0.2\n",
        "    learn = get_learn(model=model, size=192, bs=16, mixup=mixup)\n",
        "    learn.fit_fc(epochs, lr=4e-3, moms=(0.95,0.95), start_pct=0.72)\n",
        "    res += [learn.recorder.metrics[-1][0].item()]\n",
        "print([round(x, 6) for x in res], sum(res)/len(res))"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "data path   /root/.fastai/data/imagewoof2\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "stream",
          "text": [
            ""
          ],
          "name": "stderr"
        },
        {
          "output_type": "stream",
          "text": [
            "Learn path /root/.fastai/data/imagewoof2\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "display_data",
          "data": {
            "text/html": [
              "<table border=\"1\" class=\"dataframe\">\n",
              "  <thead>\n",
              "    <tr style=\"text-align: left;\">\n",
              "      <th>epoch</th>\n",
              "      <th>train_loss</th>\n",
              "      <th>valid_loss</th>\n",
              "      <th>accuracy</th>\n",
              "      <th>top_k_accuracy</th>\n",
              "      <th>time</th>\n",
              "    </tr>\n",
              "  </thead>\n",
              "  <tbody>\n",
              "    <tr>\n",
              "      <td>0</td>\n",
              "      <td>1.816367</td>\n",
              "      <td>1.742989</td>\n",
              "      <td>0.460932</td>\n",
              "      <td>0.905574</td>\n",
              "      <td>07:18</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>1</td>\n",
              "      <td>1.570518</td>\n",
              "      <td>1.470981</td>\n",
              "      <td>0.579537</td>\n",
              "      <td>0.946806</td>\n",
              "      <td>07:19</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>2</td>\n",
              "      <td>1.335072</td>\n",
              "      <td>1.309832</td>\n",
              "      <td>0.661237</td>\n",
              "      <td>0.957496</td>\n",
              "      <td>07:18</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>3</td>\n",
              "      <td>1.251946</td>\n",
              "      <td>1.125895</td>\n",
              "      <td>0.758972</td>\n",
              "      <td>0.973021</td>\n",
              "      <td>07:18</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>4</td>\n",
              "      <td>1.014469</td>\n",
              "      <td>0.983681</td>\n",
              "      <td>0.824637</td>\n",
              "      <td>0.981675</td>\n",
              "      <td>07:18</td>\n",
              "    </tr>\n",
              "  </tbody>\n",
              "</table>"
            ],
            "text/plain": [
              "<IPython.core.display.HTML object>"
            ]
          },
          "metadata": {
            "tags": []
          }
        },
        {
          "output_type": "stream",
          "text": [
            ""
          ],
          "name": "stderr"
        },
        {
          "output_type": "stream",
          "text": [
            "data path   /root/.fastai/data/imagewoof2\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "stream",
          "text": [
            ""
          ],
          "name": "stderr"
        },
        {
          "output_type": "stream",
          "text": [
            "Learn path /root/.fastai/data/imagewoof2\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "display_data",
          "data": {
            "text/html": [
              "<table border=\"1\" class=\"dataframe\">\n",
              "  <thead>\n",
              "    <tr style=\"text-align: left;\">\n",
              "      <th>epoch</th>\n",
              "      <th>train_loss</th>\n",
              "      <th>valid_loss</th>\n",
              "      <th>accuracy</th>\n",
              "      <th>top_k_accuracy</th>\n",
              "      <th>time</th>\n",
              "    </tr>\n",
              "  </thead>\n",
              "  <tbody>\n",
              "    <tr>\n",
              "      <td>0</td>\n",
              "      <td>1.812879</td>\n",
              "      <td>1.758666</td>\n",
              "      <td>0.453041</td>\n",
              "      <td>0.890812</td>\n",
              "      <td>07:18</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>1</td>\n",
              "      <td>1.553075</td>\n",
              "      <td>1.417638</td>\n",
              "      <td>0.613642</td>\n",
              "      <td>0.945788</td>\n",
              "      <td>07:18</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>2</td>\n",
              "      <td>1.365519</td>\n",
              "      <td>1.280960</td>\n",
              "      <td>0.685671</td>\n",
              "      <td>0.955205</td>\n",
              "      <td>07:18</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>3</td>\n",
              "      <td>1.200144</td>\n",
              "      <td>1.240493</td>\n",
              "      <td>0.729702</td>\n",
              "      <td>0.970476</td>\n",
              "      <td>07:18</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>4</td>\n",
              "      <td>1.026837</td>\n",
              "      <td>0.980040</td>\n",
              "      <td>0.829473</td>\n",
              "      <td>0.981166</td>\n",
              "      <td>07:19</td>\n",
              "    </tr>\n",
              "  </tbody>\n",
              "</table>"
            ],
            "text/plain": [
              "<IPython.core.display.HTML object>"
            ]
          },
          "metadata": {
            "tags": []
          }
        },
        {
          "output_type": "stream",
          "text": [
            ""
          ],
          "name": "stderr"
        },
        {
          "output_type": "stream",
          "text": [
            "data path   /root/.fastai/data/imagewoof2\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "stream",
          "text": [
            ""
          ],
          "name": "stderr"
        },
        {
          "output_type": "stream",
          "text": [
            "Learn path /root/.fastai/data/imagewoof2\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "display_data",
          "data": {
            "text/html": [
              "<table border=\"1\" class=\"dataframe\">\n",
              "  <thead>\n",
              "    <tr style=\"text-align: left;\">\n",
              "      <th>epoch</th>\n",
              "      <th>train_loss</th>\n",
              "      <th>valid_loss</th>\n",
              "      <th>accuracy</th>\n",
              "      <th>top_k_accuracy</th>\n",
              "      <th>time</th>\n",
              "    </tr>\n",
              "  </thead>\n",
              "  <tbody>\n",
              "    <tr>\n",
              "      <td>0</td>\n",
              "      <td>1.844059</td>\n",
              "      <td>1.720307</td>\n",
              "      <td>0.466785</td>\n",
              "      <td>0.901502</td>\n",
              "      <td>07:18</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>1</td>\n",
              "      <td>1.598753</td>\n",
              "      <td>1.470148</td>\n",
              "      <td>0.602698</td>\n",
              "      <td>0.942479</td>\n",
              "      <td>07:18</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>2</td>\n",
              "      <td>1.341149</td>\n",
              "      <td>1.255469</td>\n",
              "      <td>0.686689</td>\n",
              "      <td>0.963095</td>\n",
              "      <td>07:18</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>3</td>\n",
              "      <td>1.243953</td>\n",
              "      <td>1.122435</td>\n",
              "      <td>0.754645</td>\n",
              "      <td>0.969203</td>\n",
              "      <td>07:18</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>4</td>\n",
              "      <td>1.031597</td>\n",
              "      <td>0.996570</td>\n",
              "      <td>0.824637</td>\n",
              "      <td>0.981420</td>\n",
              "      <td>07:18</td>\n",
              "    </tr>\n",
              "  </tbody>\n",
              "</table>"
            ],
            "text/plain": [
              "<IPython.core.display.HTML object>"
            ]
          },
          "metadata": {
            "tags": []
          }
        },
        {
          "output_type": "stream",
          "text": [
            ""
          ],
          "name": "stderr"
        },
        {
          "output_type": "stream",
          "text": [
            "data path   /root/.fastai/data/imagewoof2\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "stream",
          "text": [
            ""
          ],
          "name": "stderr"
        },
        {
          "output_type": "stream",
          "text": [
            "Learn path /root/.fastai/data/imagewoof2\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "display_data",
          "data": {
            "text/html": [
              "<table border=\"1\" class=\"dataframe\">\n",
              "  <thead>\n",
              "    <tr style=\"text-align: left;\">\n",
              "      <th>epoch</th>\n",
              "      <th>train_loss</th>\n",
              "      <th>valid_loss</th>\n",
              "      <th>accuracy</th>\n",
              "      <th>top_k_accuracy</th>\n",
              "      <th>time</th>\n",
              "    </tr>\n",
              "  </thead>\n",
              "  <tbody>\n",
              "    <tr>\n",
              "      <td>0</td>\n",
              "      <td>1.821380</td>\n",
              "      <td>1.678589</td>\n",
              "      <td>0.481802</td>\n",
              "      <td>0.910919</td>\n",
              "      <td>07:18</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>1</td>\n",
              "      <td>1.630918</td>\n",
              "      <td>1.482963</td>\n",
              "      <td>0.576228</td>\n",
              "      <td>0.937643</td>\n",
              "      <td>07:18</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>2</td>\n",
              "      <td>1.379747</td>\n",
              "      <td>1.222471</td>\n",
              "      <td>0.704505</td>\n",
              "      <td>0.968440</td>\n",
              "      <td>07:18</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>3</td>\n",
              "      <td>1.262058</td>\n",
              "      <td>1.147718</td>\n",
              "      <td>0.755663</td>\n",
              "      <td>0.968949</td>\n",
              "      <td>07:18</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>4</td>\n",
              "      <td>1.063779</td>\n",
              "      <td>0.986779</td>\n",
              "      <td>0.833291</td>\n",
              "      <td>0.982947</td>\n",
              "      <td>07:19</td>\n",
              "    </tr>\n",
              "  </tbody>\n",
              "</table>"
            ],
            "text/plain": [
              "<IPython.core.display.HTML object>"
            ]
          },
          "metadata": {
            "tags": []
          }
        },
        {
          "output_type": "stream",
          "text": [
            ""
          ],
          "name": "stderr"
        },
        {
          "output_type": "stream",
          "text": [
            "data path   /root/.fastai/data/imagewoof2\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "stream",
          "text": [
            ""
          ],
          "name": "stderr"
        },
        {
          "output_type": "stream",
          "text": [
            "Learn path /root/.fastai/data/imagewoof2\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "display_data",
          "data": {
            "text/html": [
              "<table border=\"1\" class=\"dataframe\">\n",
              "  <thead>\n",
              "    <tr style=\"text-align: left;\">\n",
              "      <th>epoch</th>\n",
              "      <th>train_loss</th>\n",
              "      <th>valid_loss</th>\n",
              "      <th>accuracy</th>\n",
              "      <th>top_k_accuracy</th>\n",
              "      <th>time</th>\n",
              "    </tr>\n",
              "  </thead>\n",
              "  <tbody>\n",
              "    <tr>\n",
              "      <td>0</td>\n",
              "      <td>1.891106</td>\n",
              "      <td>1.707774</td>\n",
              "      <td>0.461441</td>\n",
              "      <td>0.902774</td>\n",
              "      <td>07:18</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>1</td>\n",
              "      <td>1.606799</td>\n",
              "      <td>1.477212</td>\n",
              "      <td>0.581064</td>\n",
              "      <td>0.935353</td>\n",
              "      <td>07:18</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>2</td>\n",
              "      <td>1.420447</td>\n",
              "      <td>1.298537</td>\n",
              "      <td>0.672436</td>\n",
              "      <td>0.959786</td>\n",
              "      <td>07:18</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>3</td>\n",
              "      <td>1.245861</td>\n",
              "      <td>1.184937</td>\n",
              "      <td>0.736829</td>\n",
              "      <td>0.969712</td>\n",
              "      <td>07:19</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>4</td>\n",
              "      <td>1.032215</td>\n",
              "      <td>0.994109</td>\n",
              "      <td>0.822601</td>\n",
              "      <td>0.978621</td>\n",
              "      <td>07:18</td>\n",
              "    </tr>\n",
              "  </tbody>\n",
              "</table>"
            ],
            "text/plain": [
              "<IPython.core.display.HTML object>"
            ]
          },
          "metadata": {
            "tags": []
          }
        },
        {
          "output_type": "stream",
          "text": [
            ""
          ],
          "name": "stderr"
        },
        {
          "output_type": "stream",
          "text": [
            "[0.824637, 0.829473, 0.824637, 0.833291, 0.822601] 0.8269279718399047\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "ZMZpFgfHg1cp",
        "colab_type": "text"
      },
      "source": [
        "### epochs=20, 5 runs\n",
        "\n",
        "no stride, bs=16, no act no bn"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "ZanS8tGGweQW",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "for epochs in [20]*5:\n",
        "    mixup=0 if epochs<=20 else 0.2\n",
        "    learn = get_learn(model=model, size=192, bs=32, mixup=mixup)\n",
        "    learn.fit_fc(epochs, lr=4e-3, moms=(0.95,0.95), start_pct=0.72)\n",
        "    res += [learn.recorder.metrics[-1][0].item()]\n",
        "print([round(x, 6) for x in res], sum(res)/len(res))"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "HQl3YINlg3A9",
        "colab_type": "text"
      },
      "source": [
        "### epochs=80, 1 run\n",
        "\n",
        "no stride, bs=16, no act no bn"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "p6EqvtTL_PnK",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 985
        },
        "outputId": "4a433f15-a529-4f0c-c9e4-140ae3bd640c"
      },
      "source": [
        "for epochs in [80]:\n",
        "    mixup=0 if epochs<=20 else 0.2\n",
        "    learn = get_learn(model=model, size=192, bs=16, mixup=mixup)\n",
        "    learn.fit_fc(epochs, lr=4e-3, moms=(0.95,0.95), start_pct=0.72)\n",
        "    res += [learn.recorder.metrics[-1][0].item()]\n",
        "print([round(x, 6) for x in res], sum(res)/len(res))"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Downloading https://s3.amazonaws.com/fast-ai-imageclas/imagewoof2.tgz\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "display_data",
          "data": {
            "text/html": [
              ""
            ],
            "text/plain": [
              "<IPython.core.display.HTML object>"
            ]
          },
          "metadata": {
            "tags": []
          }
        },
        {
          "output_type": "stream",
          "text": [
            "data path   /root/.fastai/data/imagewoof2\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "stream",
          "text": [
            ""
          ],
          "name": "stderr"
        },
        {
          "output_type": "stream",
          "text": [
            "Learn path /root/.fastai/data/imagewoof2\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "display_data",
          "data": {
            "text/html": [
              "\n",
              "    <div>\n",
              "        <style>\n",
              "            /* Turns off some styling */\n",
              "            progress {\n",
              "                /* gets rid of default border in Firefox and Opera. */\n",
              "                border: none;\n",
              "                /* Needs to be in here for Safari polyfill so background images work as expected. */\n",
              "                background-size: auto;\n",
              "            }\n",
              "            .progress-bar-interrupted, .progress-bar-interrupted::-webkit-progress-bar {\n",
              "                background: #F44336;\n",
              "            }\n",
              "        </style>\n",
              "      <progress value='16' class='' max='80' style='width:300px; height:20px; vertical-align: middle;'></progress>\n",
              "      20.00% [16/80 1:55:11<7:40:46]\n",
              "    </div>\n",
              "    \n",
              "<table border=\"1\" class=\"dataframe\">\n",
              "  <thead>\n",
              "    <tr style=\"text-align: left;\">\n",
              "      <th>epoch</th>\n",
              "      <th>train_loss</th>\n",
              "      <th>valid_loss</th>\n",
              "      <th>accuracy</th>\n",
              "      <th>top_k_accuracy</th>\n",
              "      <th>time</th>\n",
              "    </tr>\n",
              "  </thead>\n",
              "  <tbody>\n",
              "    <tr>\n",
              "      <td>0</td>\n",
              "      <td>2.012678</td>\n",
              "      <td>1.787375</td>\n",
              "      <td>0.416136</td>\n",
              "      <td>0.879104</td>\n",
              "      <td>07:11</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>1</td>\n",
              "      <td>1.827199</td>\n",
              "      <td>1.515741</td>\n",
              "      <td>0.567320</td>\n",
              "      <td>0.930262</td>\n",
              "      <td>07:11</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>2</td>\n",
              "      <td>1.661169</td>\n",
              "      <td>1.347738</td>\n",
              "      <td>0.643166</td>\n",
              "      <td>0.953678</td>\n",
              "      <td>07:11</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>3</td>\n",
              "      <td>1.538006</td>\n",
              "      <td>1.261420</td>\n",
              "      <td>0.697124</td>\n",
              "      <td>0.959023</td>\n",
              "      <td>07:11</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>4</td>\n",
              "      <td>1.459939</td>\n",
              "      <td>1.129074</td>\n",
              "      <td>0.749300</td>\n",
              "      <td>0.968185</td>\n",
              "      <td>07:12</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>5</td>\n",
              "      <td>1.386111</td>\n",
              "      <td>1.132712</td>\n",
              "      <td>0.755154</td>\n",
              "      <td>0.965640</td>\n",
              "      <td>07:11</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>6</td>\n",
              "      <td>1.344452</td>\n",
              "      <td>1.063014</td>\n",
              "      <td>0.784933</td>\n",
              "      <td>0.978112</td>\n",
              "      <td>07:12</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>7</td>\n",
              "      <td>1.331922</td>\n",
              "      <td>1.031150</td>\n",
              "      <td>0.806312</td>\n",
              "      <td>0.973021</td>\n",
              "      <td>07:12</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>8</td>\n",
              "      <td>1.276424</td>\n",
              "      <td>0.983552</td>\n",
              "      <td>0.819801</td>\n",
              "      <td>0.977348</td>\n",
              "      <td>07:12</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>9</td>\n",
              "      <td>1.231587</td>\n",
              "      <td>0.983760</td>\n",
              "      <td>0.818274</td>\n",
              "      <td>0.980657</td>\n",
              "      <td>07:11</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>10</td>\n",
              "      <td>1.216898</td>\n",
              "      <td>0.942855</td>\n",
              "      <td>0.833545</td>\n",
              "      <td>0.983711</td>\n",
              "      <td>07:12</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>11</td>\n",
              "      <td>1.178902</td>\n",
              "      <td>0.954809</td>\n",
              "      <td>0.832018</td>\n",
              "      <td>0.981420</td>\n",
              "      <td>07:12</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>12</td>\n",
              "      <td>1.145366</td>\n",
              "      <td>0.910717</td>\n",
              "      <td>0.852125</td>\n",
              "      <td>0.983965</td>\n",
              "      <td>07:11</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>13</td>\n",
              "      <td>1.123013</td>\n",
              "      <td>0.901371</td>\n",
              "      <td>0.851616</td>\n",
              "      <td>0.987529</td>\n",
              "      <td>07:11</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>14</td>\n",
              "      <td>1.125641</td>\n",
              "      <td>0.902062</td>\n",
              "      <td>0.846780</td>\n",
              "      <td>0.985238</td>\n",
              "      <td>07:12</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>15</td>\n",
              "      <td>1.116233</td>\n",
              "      <td>0.867573</td>\n",
              "      <td>0.868923</td>\n",
              "      <td>0.986002</td>\n",
              "      <td>07:12</td>\n",
              "    </tr>\n",
              "  </tbody>\n",
              "</table><p>\n",
              "\n",
              "    <div>\n",
              "        <style>\n",
              "            /* Turns off some styling */\n",
              "            progress {\n",
              "                /* gets rid of default border in Firefox and Opera. */\n",
              "                border: none;\n",
              "                /* Needs to be in here for Safari polyfill so background images work as expected. */\n",
              "                background-size: auto;\n",
              "            }\n",
              "            .progress-bar-interrupted, .progress-bar-interrupted::-webkit-progress-bar {\n",
              "                background: #F44336;\n",
              "            }\n",
              "        </style>\n",
              "      <progress value='454' class='' max='564' style='width:300px; height:20px; vertical-align: middle;'></progress>\n",
              "      80.50% [454/564 05:02<01:13 1.0999]\n",
              "    </div>\n",
              "    "
            ],
            "text/plain": [
              "<IPython.core.display.HTML object>"
            ]
          },
          "metadata": {
            "tags": []
          }
        },
        {
          "output_type": "stream",
          "text": [
            ""
          ],
          "name": "stderr"
        },
        {
          "output_type": "stream",
          "text": [
            "Buffered data was truncated after reaching the output size limit."
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "eOuCpC_mV5Dq",
        "colab_type": "text"
      },
      "source": [
        "Sorry, data was \"truncated\", but see the val_loss and accuracy below for proof. **92.08 is the all-time best that I've seen for Imagewoof2.**"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "B_ULDkeOx5cN",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 1000
        },
        "outputId": "84c03638-0af7-4906-d773-a9934da19c0b"
      },
      "source": [
        "for i in range(80):\n",
        "  print(i, learn.recorder.val_losses[i].item(), learn.recorder.metrics[i][0].item())"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "0 1.7873750925064087 0.41613641381263733\n",
            "1 1.5157413482666016 0.567319929599762\n",
            "2 1.347738265991211 0.643166184425354\n",
            "3 1.2614200115203857 0.6971239447593689\n",
            "4 1.129074215888977 0.7493000626564026\n",
            "5 1.1327120065689087 0.7551539540290833\n",
            "6 1.063014268875122 0.784932553768158\n",
            "7 1.0311501026153564 0.8063120245933533\n",
            "8 0.9835519194602966 0.8198014497756958\n",
            "9 0.9837595820426941 0.8182743787765503\n",
            "10 0.9428547024726868 0.833545446395874\n",
            "11 0.95480877161026 0.8320183157920837\n",
            "12 0.9107168316841125 0.8521252274513245\n",
            "13 0.9013711214065552 0.8516162037849426\n",
            "14 0.9020619988441467 0.8467803597450256\n",
            "15 0.8675727844238281 0.8689233660697937\n",
            "16 0.871545135974884 0.8620514273643494\n",
            "17 0.897378146648407 0.8500890731811523\n",
            "18 0.896333634853363 0.8653601408004761\n",
            "19 0.8782687187194824 0.8620514273643494\n",
            "20 0.8516641855239868 0.8791040778160095\n",
            "21 0.8856545686721802 0.8513616919517517\n",
            "22 0.8662623763084412 0.8722321391105652\n",
            "23 0.8581082820892334 0.8732501864433289\n",
            "24 0.8560348153114319 0.8735046982765198\n",
            "25 0.8322374820709229 0.8760498762130737\n",
            "26 0.8419180512428284 0.8864850997924805\n",
            "27 0.840420126914978 0.8811402320861816\n",
            "28 0.8424336910247803 0.8836854100227356\n",
            "29 0.8240237832069397 0.8841944336891174\n",
            "30 0.8235917687416077 0.8841944336891174\n",
            "31 0.838763952255249 0.877577006816864\n",
            "32 0.8137616515159607 0.8887757658958435\n",
            "33 0.8356049656867981 0.884958028793335\n",
            "34 0.8061267137527466 0.8918299674987793\n",
            "35 0.8194290399551392 0.8852125406265259\n",
            "36 0.8170048594474792 0.8806312084197998\n",
            "37 0.8230636715888977 0.8813947439193726\n",
            "38 0.8130825161933899 0.8875032067298889\n",
            "39 0.8155640363693237 0.8862305879592896\n",
            "40 0.8164389133453369 0.8890302777290344\n",
            "41 0.8213346600532532 0.879867672920227\n",
            "42 0.8182024359703064 0.8834308981895447\n",
            "43 0.8204544186592102 0.8875032067298889\n",
            "44 0.802797257900238 0.8959022760391235\n",
            "45 0.8072392344474792 0.8923389911651611\n",
            "46 0.8031151294708252 0.8936116099357605\n",
            "47 0.807945966720581 0.8877577185630798\n",
            "48 0.8156455755233765 0.8852125406265259\n",
            "49 0.8080087900161743 0.8905574083328247\n",
            "50 0.8176583051681519 0.8864850997924805\n",
            "51 0.7917284369468689 0.8961567878723145\n",
            "52 0.8103719353675842 0.8900483846664429\n",
            "53 0.7915446758270264 0.8959022760391235\n",
            "54 0.8118966221809387 0.8880122303962708\n",
            "55 0.8303729295730591 0.8829218745231628\n",
            "56 0.8261889815330505 0.8803766965866089\n",
            "57 0.8363656401634216 0.8780860304832458\n",
            "58 0.8130425810813904 0.8841944336891174\n",
            "59 0.7944099307060242 0.8992109894752502\n",
            "60 0.7945766448974609 0.8892847895622253\n",
            "61 0.7877740263938904 0.8951387405395508\n",
            "62 0.7924072742462158 0.8951387405395508\n",
            "63 0.8004301190376282 0.8890302777290344\n",
            "64 0.7916184663772583 0.8920844793319702\n",
            "65 0.7778981328010559 0.8987019658088684\n",
            "66 0.776310384273529 0.9035378098487854\n",
            "67 0.7690089344978333 0.9043013453483582\n",
            "68 0.7650967240333557 0.9060829877853394\n",
            "69 0.7576826214790344 0.9091371893882751\n",
            "70 0.7575828433036804 0.9099007248878479\n",
            "71 0.7524638772010803 0.909646213054657\n",
            "72 0.7463800311088562 0.9160091876983643\n",
            "73 0.7363959550857544 0.9155001044273376\n",
            "74 0.7369838356971741 0.916772723197937\n",
            "75 0.7346676588058472 0.916772723197937\n",
            "76 0.7321836352348328 0.921608567237854\n",
            "77 0.730678379535675 0.9213540554046631\n",
            "78 0.7310044765472412 0.921608567237854\n",
            "79 0.7318302989006042 0.9208449721336365\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "TsR_sLOmg5mM",
        "colab_type": "text"
      },
      "source": [
        "### epochs=200, 1 run\n",
        "\n",
        "no stride at stem, bs=16, no act no bn"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "edl_s7jIbj2w",
        "colab_type": "code",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 1000
        },
        "outputId": "e09c7491-c108-41e5-b30b-52927cbd2ae3"
      },
      "source": [
        "res = []\n",
        "for epochs in [200]:\n",
        "    mixup=0 if epochs<=20 else 0.2\n",
        "    learn = get_learn(model=model, size=192, bs=16, mixup=mixup)\n",
        "    learn.fit_fc(epochs, lr=4e-3, moms=(0.95,0.95), start_pct=0.72)\n",
        "    res += [learn.recorder.metrics[-1][0].item()]\n",
        "print([round(x, 6) for x in res], sum(res)/len(res))"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "data path   /root/.fastai/data/imagewoof2\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "stream",
          "text": [
            "/usr/local/lib/python3.6/dist-packages/torch/nn/functional.py:2854: UserWarning: The default behavior for interpolate/upsample with float scale_factor will change in 1.6.0 to align with other frameworks/libraries, and use scale_factor directly, instead of relying on the computed output size. If you wish to keep the old behavior, please set recompute_scale_factor=True. See the documentation of nn.Upsample for details. \n",
            "  warnings.warn(\"The default behavior for interpolate/upsample with float scale_factor will change \"\n",
            "/usr/local/lib/python3.6/dist-packages/torch/nn/functional.py:2854: UserWarning: The default behavior for interpolate/upsample with float scale_factor will change in 1.6.0 to align with other frameworks/libraries, and use scale_factor directly, instead of relying on the computed output size. If you wish to keep the old behavior, please set recompute_scale_factor=True. See the documentation of nn.Upsample for details. \n",
            "  warnings.warn(\"The default behavior for interpolate/upsample with float scale_factor will change \"\n",
            "/usr/local/lib/python3.6/dist-packages/torch/nn/functional.py:2854: UserWarning: The default behavior for interpolate/upsample with float scale_factor will change in 1.6.0 to align with other frameworks/libraries, and use scale_factor directly, instead of relying on the computed output size. If you wish to keep the old behavior, please set recompute_scale_factor=True. See the documentation of nn.Upsample for details. \n",
            "  warnings.warn(\"The default behavior for interpolate/upsample with float scale_factor will change \"\n",
            "/usr/local/lib/python3.6/dist-packages/torch/nn/functional.py:2854: UserWarning: The default behavior for interpolate/upsample with float scale_factor will change in 1.6.0 to align with other frameworks/libraries, and use scale_factor directly, instead of relying on the computed output size. If you wish to keep the old behavior, please set recompute_scale_factor=True. See the documentation of nn.Upsample for details. \n",
            "  warnings.warn(\"The default behavior for interpolate/upsample with float scale_factor will change \"\n",
            "/usr/local/lib/python3.6/dist-packages/torch/nn/functional.py:2854: UserWarning: The default behavior for interpolate/upsample with float scale_factor will change in 1.6.0 to align with other frameworks/libraries, and use scale_factor directly, instead of relying on the computed output size. If you wish to keep the old behavior, please set recompute_scale_factor=True. See the documentation of nn.Upsample for details. \n",
            "  warnings.warn(\"The default behavior for interpolate/upsample with float scale_factor will change \"\n"
          ],
          "name": "stderr"
        },
        {
          "output_type": "stream",
          "text": [
            "Learn path /root/.fastai/data/imagewoof2\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "display_data",
          "data": {
            "text/html": [
              "\n",
              "    <div>\n",
              "        <style>\n",
              "            /* Turns off some styling */\n",
              "            progress {\n",
              "                /* gets rid of default border in Firefox and Opera. */\n",
              "                border: none;\n",
              "                /* Needs to be in here for Safari polyfill so background images work as expected. */\n",
              "                background-size: auto;\n",
              "            }\n",
              "            .progress-bar-interrupted, .progress-bar-interrupted::-webkit-progress-bar {\n",
              "                background: #F44336;\n",
              "            }\n",
              "        </style>\n",
              "      <progress value='198' class='' max='200' style='width:300px; height:20px; vertical-align: middle;'></progress>\n",
              "      99.00% [198/200 23:50:04<14:26]\n",
              "    </div>\n",
              "    \n",
              "<table border=\"1\" class=\"dataframe\">\n",
              "  <thead>\n",
              "    <tr style=\"text-align: left;\">\n",
              "      <th>epoch</th>\n",
              "      <th>train_loss</th>\n",
              "      <th>valid_loss</th>\n",
              "      <th>accuracy</th>\n",
              "      <th>top_k_accuracy</th>\n",
              "      <th>time</th>\n",
              "    </tr>\n",
              "  </thead>\n",
              "  <tbody>\n",
              "    <tr>\n",
              "      <td>0</td>\n",
              "      <td>1.991029</td>\n",
              "      <td>1.782373</td>\n",
              "      <td>0.423008</td>\n",
              "      <td>0.889030</td>\n",
              "      <td>07:12</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>1</td>\n",
              "      <td>1.820094</td>\n",
              "      <td>1.558511</td>\n",
              "      <td>0.531687</td>\n",
              "      <td>0.922372</td>\n",
              "      <td>07:11</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>2</td>\n",
              "      <td>1.640254</td>\n",
              "      <td>1.417206</td>\n",
              "      <td>0.610588</td>\n",
              "      <td>0.951642</td>\n",
              "      <td>07:11</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>3</td>\n",
              "      <td>1.524751</td>\n",
              "      <td>1.266035</td>\n",
              "      <td>0.692797</td>\n",
              "      <td>0.962077</td>\n",
              "      <td>07:11</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>4</td>\n",
              "      <td>1.451839</td>\n",
              "      <td>1.167214</td>\n",
              "      <td>0.736574</td>\n",
              "      <td>0.968440</td>\n",
              "      <td>07:12</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>5</td>\n",
              "      <td>1.363214</td>\n",
              "      <td>1.107141</td>\n",
              "      <td>0.767625</td>\n",
              "      <td>0.972767</td>\n",
              "      <td>07:12</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>6</td>\n",
              "      <td>1.331252</td>\n",
              "      <td>1.045224</td>\n",
              "      <td>0.792059</td>\n",
              "      <td>0.973530</td>\n",
              "      <td>07:12</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>7</td>\n",
              "      <td>1.325143</td>\n",
              "      <td>1.032004</td>\n",
              "      <td>0.803003</td>\n",
              "      <td>0.975821</td>\n",
              "      <td>07:12</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>8</td>\n",
              "      <td>1.271300</td>\n",
              "      <td>0.982155</td>\n",
              "      <td>0.821838</td>\n",
              "      <td>0.983965</td>\n",
              "      <td>07:12</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>9</td>\n",
              "      <td>1.266131</td>\n",
              "      <td>0.979309</td>\n",
              "      <td>0.827182</td>\n",
              "      <td>0.978875</td>\n",
              "      <td>07:12</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>10</td>\n",
              "      <td>1.228057</td>\n",
              "      <td>0.949628</td>\n",
              "      <td>0.843726</td>\n",
              "      <td>0.983711</td>\n",
              "      <td>07:12</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>11</td>\n",
              "      <td>1.196500</td>\n",
              "      <td>0.917728</td>\n",
              "      <td>0.846271</td>\n",
              "      <td>0.983202</td>\n",
              "      <td>07:12</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>12</td>\n",
              "      <td>1.163236</td>\n",
              "      <td>0.906025</td>\n",
              "      <td>0.853907</td>\n",
              "      <td>0.983202</td>\n",
              "      <td>07:12</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>13</td>\n",
              "      <td>1.137082</td>\n",
              "      <td>0.891482</td>\n",
              "      <td>0.857725</td>\n",
              "      <td>0.983456</td>\n",
              "      <td>07:12</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>14</td>\n",
              "      <td>1.129097</td>\n",
              "      <td>0.890981</td>\n",
              "      <td>0.863578</td>\n",
              "      <td>0.983456</td>\n",
              "      <td>07:12</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>15</td>\n",
              "      <td>1.094402</td>\n",
              "      <td>0.877052</td>\n",
              "      <td>0.860524</td>\n",
              "      <td>0.984729</td>\n",
              "      <td>07:12</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>16</td>\n",
              "      <td>1.085684</td>\n",
              "      <td>0.907792</td>\n",
              "      <td>0.860524</td>\n",
              "      <td>0.980148</td>\n",
              "      <td>07:12</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>17</td>\n",
              "      <td>1.087098</td>\n",
              "      <td>0.874584</td>\n",
              "      <td>0.865869</td>\n",
              "      <td>0.981929</td>\n",
              "      <td>07:12</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>18</td>\n",
              "      <td>1.085387</td>\n",
              "      <td>0.863240</td>\n",
              "      <td>0.873250</td>\n",
              "      <td>0.987529</td>\n",
              "      <td>07:12</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>19</td>\n",
              "      <td>1.054475</td>\n",
              "      <td>0.887979</td>\n",
              "      <td>0.857979</td>\n",
              "      <td>0.984220</td>\n",
              "      <td>07:12</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>20</td>\n",
              "      <td>1.016570</td>\n",
              "      <td>0.861202</td>\n",
              "      <td>0.865106</td>\n",
              "      <td>0.986511</td>\n",
              "      <td>07:12</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>21</td>\n",
              "      <td>1.019237</td>\n",
              "      <td>0.867021</td>\n",
              "      <td>0.867905</td>\n",
              "      <td>0.984983</td>\n",
              "      <td>07:12</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>22</td>\n",
              "      <td>1.012525</td>\n",
              "      <td>0.844951</td>\n",
              "      <td>0.875795</td>\n",
              "      <td>0.985238</td>\n",
              "      <td>07:12</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>23</td>\n",
              "      <td>0.996946</td>\n",
              "      <td>0.843701</td>\n",
              "      <td>0.880377</td>\n",
              "      <td>0.988292</td>\n",
              "      <td>07:12</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>24</td>\n",
              "      <td>0.977238</td>\n",
              "      <td>0.838571</td>\n",
              "      <td>0.880377</td>\n",
              "      <td>0.987783</td>\n",
              "      <td>07:12</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>25</td>\n",
              "      <td>0.989970</td>\n",
              "      <td>0.833279</td>\n",
              "      <td>0.876304</td>\n",
              "      <td>0.987020</td>\n",
              "      <td>07:12</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>26</td>\n",
              "      <td>0.989915</td>\n",
              "      <td>0.847595</td>\n",
              "      <td>0.874268</td>\n",
              "      <td>0.987020</td>\n",
              "      <td>07:12</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>27</td>\n",
              "      <td>0.991771</td>\n",
              "      <td>0.831164</td>\n",
              "      <td>0.887249</td>\n",
              "      <td>0.984729</td>\n",
              "      <td>07:12</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>28</td>\n",
              "      <td>0.968312</td>\n",
              "      <td>0.851488</td>\n",
              "      <td>0.879359</td>\n",
              "      <td>0.980402</td>\n",
              "      <td>07:12</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>29</td>\n",
              "      <td>0.943679</td>\n",
              "      <td>0.843327</td>\n",
              "      <td>0.878595</td>\n",
              "      <td>0.984729</td>\n",
              "      <td>07:12</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>30</td>\n",
              "      <td>0.942036</td>\n",
              "      <td>0.873392</td>\n",
              "      <td>0.871469</td>\n",
              "      <td>0.983965</td>\n",
              "      <td>07:12</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>31</td>\n",
              "      <td>0.956224</td>\n",
              "      <td>0.850238</td>\n",
              "      <td>0.875541</td>\n",
              "      <td>0.983711</td>\n",
              "      <td>07:12</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>32</td>\n",
              "      <td>0.940949</td>\n",
              "      <td>0.842802</td>\n",
              "      <td>0.874014</td>\n",
              "      <td>0.985238</td>\n",
              "      <td>07:12</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>33</td>\n",
              "      <td>0.941567</td>\n",
              "      <td>0.839029</td>\n",
              "      <td>0.880886</td>\n",
              "      <td>0.983965</td>\n",
              "      <td>07:12</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>34</td>\n",
              "      <td>0.931871</td>\n",
              "      <td>0.833005</td>\n",
              "      <td>0.877068</td>\n",
              "      <td>0.986002</td>\n",
              "      <td>07:12</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>35</td>\n",
              "      <td>0.941017</td>\n",
              "      <td>0.825832</td>\n",
              "      <td>0.891575</td>\n",
              "      <td>0.983202</td>\n",
              "      <td>07:12</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>36</td>\n",
              "      <td>0.925416</td>\n",
              "      <td>0.815450</td>\n",
              "      <td>0.891066</td>\n",
              "      <td>0.983711</td>\n",
              "      <td>07:12</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>37</td>\n",
              "      <td>0.953040</td>\n",
              "      <td>0.853122</td>\n",
              "      <td>0.873250</td>\n",
              "      <td>0.982184</td>\n",
              "      <td>07:12</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>38</td>\n",
              "      <td>0.920613</td>\n",
              "      <td>0.820830</td>\n",
              "      <td>0.885213</td>\n",
              "      <td>0.986256</td>\n",
              "      <td>07:12</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>39</td>\n",
              "      <td>0.906199</td>\n",
              "      <td>0.831602</td>\n",
              "      <td>0.882413</td>\n",
              "      <td>0.982947</td>\n",
              "      <td>07:12</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>40</td>\n",
              "      <td>0.921115</td>\n",
              "      <td>0.813559</td>\n",
              "      <td>0.884194</td>\n",
              "      <td>0.988292</td>\n",
              "      <td>07:12</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>41</td>\n",
              "      <td>0.912003</td>\n",
              "      <td>0.825524</td>\n",
              "      <td>0.880886</td>\n",
              "      <td>0.980911</td>\n",
              "      <td>07:12</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>42</td>\n",
              "      <td>0.918849</td>\n",
              "      <td>0.817122</td>\n",
              "      <td>0.885467</td>\n",
              "      <td>0.984220</td>\n",
              "      <td>07:12</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>43</td>\n",
              "      <td>0.909780</td>\n",
              "      <td>0.829148</td>\n",
              "      <td>0.884449</td>\n",
              "      <td>0.983711</td>\n",
              "      <td>07:12</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>44</td>\n",
              "      <td>0.889175</td>\n",
              "      <td>0.815221</td>\n",
              "      <td>0.889794</td>\n",
              "      <td>0.983202</td>\n",
              "      <td>07:12</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>45</td>\n",
              "      <td>0.892367</td>\n",
              "      <td>0.832272</td>\n",
              "      <td>0.882667</td>\n",
              "      <td>0.983965</td>\n",
              "      <td>07:12</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>46</td>\n",
              "      <td>0.901448</td>\n",
              "      <td>0.833470</td>\n",
              "      <td>0.884194</td>\n",
              "      <td>0.982438</td>\n",
              "      <td>07:12</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>47</td>\n",
              "      <td>0.878881</td>\n",
              "      <td>0.816395</td>\n",
              "      <td>0.891066</td>\n",
              "      <td>0.982184</td>\n",
              "      <td>07:12</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>48</td>\n",
              "      <td>0.894874</td>\n",
              "      <td>0.820485</td>\n",
              "      <td>0.885722</td>\n",
              "      <td>0.985238</td>\n",
              "      <td>07:12</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>49</td>\n",
              "      <td>0.907521</td>\n",
              "      <td>0.809977</td>\n",
              "      <td>0.895393</td>\n",
              "      <td>0.982693</td>\n",
              "      <td>07:13</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>50</td>\n",
              "      <td>0.898045</td>\n",
              "      <td>0.821260</td>\n",
              "      <td>0.888012</td>\n",
              "      <td>0.985747</td>\n",
              "      <td>07:13</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>51</td>\n",
              "      <td>0.901361</td>\n",
              "      <td>0.807331</td>\n",
              "      <td>0.889539</td>\n",
              "      <td>0.983711</td>\n",
              "      <td>07:13</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>52</td>\n",
              "      <td>0.885641</td>\n",
              "      <td>0.823611</td>\n",
              "      <td>0.883431</td>\n",
              "      <td>0.984220</td>\n",
              "      <td>07:13</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>53</td>\n",
              "      <td>0.880792</td>\n",
              "      <td>0.808090</td>\n",
              "      <td>0.893103</td>\n",
              "      <td>0.983202</td>\n",
              "      <td>07:13</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>54</td>\n",
              "      <td>0.875390</td>\n",
              "      <td>0.814808</td>\n",
              "      <td>0.882413</td>\n",
              "      <td>0.984220</td>\n",
              "      <td>07:13</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>55</td>\n",
              "      <td>0.872087</td>\n",
              "      <td>0.811403</td>\n",
              "      <td>0.892848</td>\n",
              "      <td>0.982947</td>\n",
              "      <td>07:13</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>56</td>\n",
              "      <td>0.894067</td>\n",
              "      <td>0.799713</td>\n",
              "      <td>0.892594</td>\n",
              "      <td>0.983202</td>\n",
              "      <td>07:13</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>57</td>\n",
              "      <td>0.860783</td>\n",
              "      <td>0.822948</td>\n",
              "      <td>0.886231</td>\n",
              "      <td>0.980657</td>\n",
              "      <td>07:13</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>58</td>\n",
              "      <td>0.873126</td>\n",
              "      <td>0.830643</td>\n",
              "      <td>0.874523</td>\n",
              "      <td>0.985492</td>\n",
              "      <td>07:13</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>59</td>\n",
              "      <td>0.862879</td>\n",
              "      <td>0.813932</td>\n",
              "      <td>0.884194</td>\n",
              "      <td>0.978112</td>\n",
              "      <td>07:13</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>60</td>\n",
              "      <td>0.881026</td>\n",
              "      <td>0.816671</td>\n",
              "      <td>0.882413</td>\n",
              "      <td>0.983711</td>\n",
              "      <td>07:13</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>61</td>\n",
              "      <td>0.876213</td>\n",
              "      <td>0.802749</td>\n",
              "      <td>0.891066</td>\n",
              "      <td>0.985238</td>\n",
              "      <td>07:13</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>62</td>\n",
              "      <td>0.877590</td>\n",
              "      <td>0.794373</td>\n",
              "      <td>0.888012</td>\n",
              "      <td>0.986765</td>\n",
              "      <td>07:13</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>63</td>\n",
              "      <td>0.860413</td>\n",
              "      <td>0.831508</td>\n",
              "      <td>0.882922</td>\n",
              "      <td>0.983456</td>\n",
              "      <td>07:13</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>64</td>\n",
              "      <td>0.873264</td>\n",
              "      <td>0.826971</td>\n",
              "      <td>0.883431</td>\n",
              "      <td>0.982947</td>\n",
              "      <td>07:13</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>65</td>\n",
              "      <td>0.855218</td>\n",
              "      <td>0.821987</td>\n",
              "      <td>0.884194</td>\n",
              "      <td>0.982184</td>\n",
              "      <td>07:13</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>66</td>\n",
              "      <td>0.871703</td>\n",
              "      <td>0.828059</td>\n",
              "      <td>0.878850</td>\n",
              "      <td>0.980402</td>\n",
              "      <td>07:13</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>67</td>\n",
              "      <td>0.877779</td>\n",
              "      <td>0.817734</td>\n",
              "      <td>0.881140</td>\n",
              "      <td>0.982438</td>\n",
              "      <td>07:13</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>68</td>\n",
              "      <td>0.855530</td>\n",
              "      <td>0.811163</td>\n",
              "      <td>0.888012</td>\n",
              "      <td>0.981675</td>\n",
              "      <td>07:13</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>69</td>\n",
              "      <td>0.886768</td>\n",
              "      <td>0.798898</td>\n",
              "      <td>0.891066</td>\n",
              "      <td>0.984983</td>\n",
              "      <td>07:13</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>70</td>\n",
              "      <td>0.858766</td>\n",
              "      <td>0.802391</td>\n",
              "      <td>0.889794</td>\n",
              "      <td>0.982693</td>\n",
              "      <td>07:13</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>71</td>\n",
              "      <td>0.867638</td>\n",
              "      <td>0.834054</td>\n",
              "      <td>0.885722</td>\n",
              "      <td>0.976330</td>\n",
              "      <td>07:13</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>72</td>\n",
              "      <td>0.864143</td>\n",
              "      <td>0.836339</td>\n",
              "      <td>0.875032</td>\n",
              "      <td>0.975566</td>\n",
              "      <td>07:13</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>73</td>\n",
              "      <td>0.878158</td>\n",
              "      <td>0.813827</td>\n",
              "      <td>0.888521</td>\n",
              "      <td>0.981675</td>\n",
              "      <td>07:13</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>74</td>\n",
              "      <td>0.871656</td>\n",
              "      <td>0.811471</td>\n",
              "      <td>0.886485</td>\n",
              "      <td>0.983711</td>\n",
              "      <td>07:13</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>75</td>\n",
              "      <td>0.871655</td>\n",
              "      <td>0.808587</td>\n",
              "      <td>0.885976</td>\n",
              "      <td>0.980911</td>\n",
              "      <td>07:13</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>76</td>\n",
              "      <td>0.868329</td>\n",
              "      <td>0.810932</td>\n",
              "      <td>0.889030</td>\n",
              "      <td>0.986256</td>\n",
              "      <td>07:14</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>77</td>\n",
              "      <td>0.858441</td>\n",
              "      <td>0.809562</td>\n",
              "      <td>0.884194</td>\n",
              "      <td>0.983456</td>\n",
              "      <td>07:14</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>78</td>\n",
              "      <td>0.851694</td>\n",
              "      <td>0.795935</td>\n",
              "      <td>0.888267</td>\n",
              "      <td>0.984474</td>\n",
              "      <td>07:13</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>79</td>\n",
              "      <td>0.867769</td>\n",
              "      <td>0.816311</td>\n",
              "      <td>0.886994</td>\n",
              "      <td>0.983965</td>\n",
              "      <td>07:13</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>80</td>\n",
              "      <td>0.846126</td>\n",
              "      <td>0.807811</td>\n",
              "      <td>0.886231</td>\n",
              "      <td>0.981929</td>\n",
              "      <td>07:13</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>81</td>\n",
              "      <td>0.851495</td>\n",
              "      <td>0.806050</td>\n",
              "      <td>0.884194</td>\n",
              "      <td>0.985492</td>\n",
              "      <td>07:13</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>82</td>\n",
              "      <td>0.849295</td>\n",
              "      <td>0.792828</td>\n",
              "      <td>0.892848</td>\n",
              "      <td>0.985238</td>\n",
              "      <td>07:13</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>83</td>\n",
              "      <td>0.864520</td>\n",
              "      <td>0.809329</td>\n",
              "      <td>0.886485</td>\n",
              "      <td>0.982947</td>\n",
              "      <td>07:13</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>84</td>\n",
              "      <td>0.851211</td>\n",
              "      <td>0.797695</td>\n",
              "      <td>0.887758</td>\n",
              "      <td>0.985492</td>\n",
              "      <td>07:13</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>85</td>\n",
              "      <td>0.865020</td>\n",
              "      <td>0.810100</td>\n",
              "      <td>0.883431</td>\n",
              "      <td>0.981929</td>\n",
              "      <td>07:13</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>86</td>\n",
              "      <td>0.875221</td>\n",
              "      <td>0.810093</td>\n",
              "      <td>0.887503</td>\n",
              "      <td>0.986256</td>\n",
              "      <td>07:14</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>87</td>\n",
              "      <td>0.857834</td>\n",
              "      <td>0.798068</td>\n",
              "      <td>0.889794</td>\n",
              "      <td>0.982438</td>\n",
              "      <td>07:13</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>88</td>\n",
              "      <td>0.843126</td>\n",
              "      <td>0.807416</td>\n",
              "      <td>0.887503</td>\n",
              "      <td>0.985492</td>\n",
              "      <td>07:13</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>89</td>\n",
              "      <td>0.852171</td>\n",
              "      <td>0.804789</td>\n",
              "      <td>0.886231</td>\n",
              "      <td>0.984220</td>\n",
              "      <td>07:13</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>90</td>\n",
              "      <td>0.846796</td>\n",
              "      <td>0.802676</td>\n",
              "      <td>0.887249</td>\n",
              "      <td>0.983202</td>\n",
              "      <td>07:12</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>91</td>\n",
              "      <td>0.856360</td>\n",
              "      <td>0.799267</td>\n",
              "      <td>0.891321</td>\n",
              "      <td>0.980657</td>\n",
              "      <td>07:13</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>92</td>\n",
              "      <td>0.845726</td>\n",
              "      <td>0.795748</td>\n",
              "      <td>0.894375</td>\n",
              "      <td>0.981675</td>\n",
              "      <td>07:13</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>93</td>\n",
              "      <td>0.848567</td>\n",
              "      <td>0.793102</td>\n",
              "      <td>0.895139</td>\n",
              "      <td>0.984983</td>\n",
              "      <td>07:13</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>94</td>\n",
              "      <td>0.843466</td>\n",
              "      <td>0.800401</td>\n",
              "      <td>0.891575</td>\n",
              "      <td>0.982693</td>\n",
              "      <td>07:13</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>95</td>\n",
              "      <td>0.832462</td>\n",
              "      <td>0.790311</td>\n",
              "      <td>0.895139</td>\n",
              "      <td>0.981929</td>\n",
              "      <td>07:13</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>96</td>\n",
              "      <td>0.859208</td>\n",
              "      <td>0.778026</td>\n",
              "      <td>0.895139</td>\n",
              "      <td>0.985747</td>\n",
              "      <td>07:13</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>97</td>\n",
              "      <td>0.885010</td>\n",
              "      <td>0.812740</td>\n",
              "      <td>0.885976</td>\n",
              "      <td>0.983711</td>\n",
              "      <td>07:13</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>98</td>\n",
              "      <td>0.865791</td>\n",
              "      <td>0.804914</td>\n",
              "      <td>0.885467</td>\n",
              "      <td>0.983965</td>\n",
              "      <td>07:13</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>99</td>\n",
              "      <td>0.843420</td>\n",
              "      <td>0.823342</td>\n",
              "      <td>0.875795</td>\n",
              "      <td>0.984474</td>\n",
              "      <td>07:13</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>100</td>\n",
              "      <td>0.836867</td>\n",
              "      <td>0.803628</td>\n",
              "      <td>0.886994</td>\n",
              "      <td>0.983965</td>\n",
              "      <td>07:13</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>101</td>\n",
              "      <td>0.830442</td>\n",
              "      <td>0.785546</td>\n",
              "      <td>0.892084</td>\n",
              "      <td>0.984220</td>\n",
              "      <td>07:13</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>102</td>\n",
              "      <td>0.842451</td>\n",
              "      <td>0.790633</td>\n",
              "      <td>0.888776</td>\n",
              "      <td>0.985747</td>\n",
              "      <td>07:12</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>103</td>\n",
              "      <td>0.841178</td>\n",
              "      <td>0.803774</td>\n",
              "      <td>0.890812</td>\n",
              "      <td>0.982693</td>\n",
              "      <td>07:13</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>104</td>\n",
              "      <td>0.852682</td>\n",
              "      <td>0.793131</td>\n",
              "      <td>0.892084</td>\n",
              "      <td>0.983711</td>\n",
              "      <td>07:13</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>105</td>\n",
              "      <td>0.842545</td>\n",
              "      <td>0.796338</td>\n",
              "      <td>0.887249</td>\n",
              "      <td>0.986765</td>\n",
              "      <td>07:13</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>106</td>\n",
              "      <td>0.841376</td>\n",
              "      <td>0.811778</td>\n",
              "      <td>0.883685</td>\n",
              "      <td>0.982947</td>\n",
              "      <td>07:13</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>107</td>\n",
              "      <td>0.830573</td>\n",
              "      <td>0.799214</td>\n",
              "      <td>0.892339</td>\n",
              "      <td>0.982693</td>\n",
              "      <td>07:13</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>108</td>\n",
              "      <td>0.831958</td>\n",
              "      <td>0.781095</td>\n",
              "      <td>0.902265</td>\n",
              "      <td>0.981675</td>\n",
              "      <td>07:13</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>109</td>\n",
              "      <td>0.844040</td>\n",
              "      <td>0.785489</td>\n",
              "      <td>0.898956</td>\n",
              "      <td>0.983711</td>\n",
              "      <td>07:13</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>110</td>\n",
              "      <td>0.835307</td>\n",
              "      <td>0.781723</td>\n",
              "      <td>0.898447</td>\n",
              "      <td>0.984729</td>\n",
              "      <td>07:13</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>111</td>\n",
              "      <td>0.845980</td>\n",
              "      <td>0.788210</td>\n",
              "      <td>0.899211</td>\n",
              "      <td>0.984729</td>\n",
              "      <td>07:14</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>112</td>\n",
              "      <td>0.844118</td>\n",
              "      <td>0.794376</td>\n",
              "      <td>0.892594</td>\n",
              "      <td>0.983456</td>\n",
              "      <td>07:13</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>113</td>\n",
              "      <td>0.844843</td>\n",
              "      <td>0.800058</td>\n",
              "      <td>0.890048</td>\n",
              "      <td>0.984220</td>\n",
              "      <td>07:13</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>114</td>\n",
              "      <td>0.849632</td>\n",
              "      <td>0.801197</td>\n",
              "      <td>0.886231</td>\n",
              "      <td>0.986765</td>\n",
              "      <td>07:13</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>115</td>\n",
              "      <td>0.839153</td>\n",
              "      <td>0.816550</td>\n",
              "      <td>0.886485</td>\n",
              "      <td>0.981675</td>\n",
              "      <td>07:13</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>116</td>\n",
              "      <td>0.830079</td>\n",
              "      <td>0.796398</td>\n",
              "      <td>0.886485</td>\n",
              "      <td>0.985747</td>\n",
              "      <td>07:13</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>117</td>\n",
              "      <td>0.840447</td>\n",
              "      <td>0.791639</td>\n",
              "      <td>0.892084</td>\n",
              "      <td>0.983711</td>\n",
              "      <td>07:13</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>118</td>\n",
              "      <td>0.843113</td>\n",
              "      <td>0.800641</td>\n",
              "      <td>0.887249</td>\n",
              "      <td>0.986765</td>\n",
              "      <td>07:13</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>119</td>\n",
              "      <td>0.863607</td>\n",
              "      <td>0.808290</td>\n",
              "      <td>0.887758</td>\n",
              "      <td>0.982438</td>\n",
              "      <td>07:13</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>120</td>\n",
              "      <td>0.838129</td>\n",
              "      <td>0.799434</td>\n",
              "      <td>0.888012</td>\n",
              "      <td>0.984729</td>\n",
              "      <td>07:13</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>121</td>\n",
              "      <td>0.845105</td>\n",
              "      <td>0.804873</td>\n",
              "      <td>0.890812</td>\n",
              "      <td>0.986511</td>\n",
              "      <td>07:13</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>122</td>\n",
              "      <td>0.838364</td>\n",
              "      <td>0.804630</td>\n",
              "      <td>0.888521</td>\n",
              "      <td>0.983965</td>\n",
              "      <td>07:13</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>123</td>\n",
              "      <td>0.821491</td>\n",
              "      <td>0.804634</td>\n",
              "      <td>0.884703</td>\n",
              "      <td>0.980911</td>\n",
              "      <td>07:13</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>124</td>\n",
              "      <td>0.839571</td>\n",
              "      <td>0.811710</td>\n",
              "      <td>0.889285</td>\n",
              "      <td>0.979893</td>\n",
              "      <td>07:12</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>125</td>\n",
              "      <td>0.849670</td>\n",
              "      <td>0.805313</td>\n",
              "      <td>0.890048</td>\n",
              "      <td>0.980911</td>\n",
              "      <td>07:13</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>126</td>\n",
              "      <td>0.843459</td>\n",
              "      <td>0.795904</td>\n",
              "      <td>0.891066</td>\n",
              "      <td>0.981929</td>\n",
              "      <td>07:13</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>127</td>\n",
              "      <td>0.845211</td>\n",
              "      <td>0.799188</td>\n",
              "      <td>0.890048</td>\n",
              "      <td>0.982947</td>\n",
              "      <td>07:13</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>128</td>\n",
              "      <td>0.830591</td>\n",
              "      <td>0.811067</td>\n",
              "      <td>0.888267</td>\n",
              "      <td>0.980402</td>\n",
              "      <td>07:13</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>129</td>\n",
              "      <td>0.824560</td>\n",
              "      <td>0.812827</td>\n",
              "      <td>0.882413</td>\n",
              "      <td>0.981675</td>\n",
              "      <td>07:13</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>130</td>\n",
              "      <td>0.826921</td>\n",
              "      <td>0.807431</td>\n",
              "      <td>0.882158</td>\n",
              "      <td>0.981929</td>\n",
              "      <td>07:13</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>131</td>\n",
              "      <td>0.827146</td>\n",
              "      <td>0.814964</td>\n",
              "      <td>0.884449</td>\n",
              "      <td>0.980148</td>\n",
              "      <td>07:13</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>132</td>\n",
              "      <td>0.835155</td>\n",
              "      <td>0.808489</td>\n",
              "      <td>0.882158</td>\n",
              "      <td>0.982947</td>\n",
              "      <td>07:13</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>133</td>\n",
              "      <td>0.841682</td>\n",
              "      <td>0.806099</td>\n",
              "      <td>0.886994</td>\n",
              "      <td>0.984983</td>\n",
              "      <td>07:13</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>134</td>\n",
              "      <td>0.822573</td>\n",
              "      <td>0.796141</td>\n",
              "      <td>0.893103</td>\n",
              "      <td>0.982693</td>\n",
              "      <td>07:13</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>135</td>\n",
              "      <td>0.838010</td>\n",
              "      <td>0.806571</td>\n",
              "      <td>0.886740</td>\n",
              "      <td>0.983456</td>\n",
              "      <td>07:14</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>136</td>\n",
              "      <td>0.824569</td>\n",
              "      <td>0.819686</td>\n",
              "      <td>0.884194</td>\n",
              "      <td>0.980148</td>\n",
              "      <td>07:13</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>137</td>\n",
              "      <td>0.840063</td>\n",
              "      <td>0.798536</td>\n",
              "      <td>0.886740</td>\n",
              "      <td>0.983202</td>\n",
              "      <td>07:13</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>138</td>\n",
              "      <td>0.846996</td>\n",
              "      <td>0.794603</td>\n",
              "      <td>0.892594</td>\n",
              "      <td>0.985238</td>\n",
              "      <td>07:13</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>139</td>\n",
              "      <td>0.845319</td>\n",
              "      <td>0.797808</td>\n",
              "      <td>0.890048</td>\n",
              "      <td>0.983456</td>\n",
              "      <td>07:13</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>140</td>\n",
              "      <td>0.836743</td>\n",
              "      <td>0.809870</td>\n",
              "      <td>0.884194</td>\n",
              "      <td>0.985238</td>\n",
              "      <td>07:14</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>141</td>\n",
              "      <td>0.837698</td>\n",
              "      <td>0.801251</td>\n",
              "      <td>0.889285</td>\n",
              "      <td>0.982438</td>\n",
              "      <td>07:14</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>142</td>\n",
              "      <td>0.823118</td>\n",
              "      <td>0.805970</td>\n",
              "      <td>0.878850</td>\n",
              "      <td>0.987529</td>\n",
              "      <td>07:14</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>143</td>\n",
              "      <td>0.861277</td>\n",
              "      <td>0.803501</td>\n",
              "      <td>0.885976</td>\n",
              "      <td>0.983965</td>\n",
              "      <td>07:14</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>144</td>\n",
              "      <td>0.851326</td>\n",
              "      <td>0.799814</td>\n",
              "      <td>0.883685</td>\n",
              "      <td>0.983456</td>\n",
              "      <td>07:13</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>145</td>\n",
              "      <td>0.834246</td>\n",
              "      <td>0.810123</td>\n",
              "      <td>0.884449</td>\n",
              "      <td>0.978621</td>\n",
              "      <td>07:13</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>146</td>\n",
              "      <td>0.822430</td>\n",
              "      <td>0.787826</td>\n",
              "      <td>0.893866</td>\n",
              "      <td>0.983711</td>\n",
              "      <td>07:14</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>147</td>\n",
              "      <td>0.840094</td>\n",
              "      <td>0.790496</td>\n",
              "      <td>0.893866</td>\n",
              "      <td>0.986511</td>\n",
              "      <td>07:14</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>148</td>\n",
              "      <td>0.831511</td>\n",
              "      <td>0.785827</td>\n",
              "      <td>0.895902</td>\n",
              "      <td>0.986002</td>\n",
              "      <td>07:14</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>149</td>\n",
              "      <td>0.831233</td>\n",
              "      <td>0.795590</td>\n",
              "      <td>0.890812</td>\n",
              "      <td>0.982438</td>\n",
              "      <td>07:14</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>150</td>\n",
              "      <td>0.847566</td>\n",
              "      <td>0.796611</td>\n",
              "      <td>0.893612</td>\n",
              "      <td>0.982947</td>\n",
              "      <td>07:14</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>151</td>\n",
              "      <td>0.834797</td>\n",
              "      <td>0.816530</td>\n",
              "      <td>0.880886</td>\n",
              "      <td>0.981420</td>\n",
              "      <td>07:13</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>152</td>\n",
              "      <td>0.830961</td>\n",
              "      <td>0.803802</td>\n",
              "      <td>0.886994</td>\n",
              "      <td>0.983711</td>\n",
              "      <td>07:14</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>153</td>\n",
              "      <td>0.818732</td>\n",
              "      <td>0.808418</td>\n",
              "      <td>0.886231</td>\n",
              "      <td>0.978112</td>\n",
              "      <td>07:13</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>154</td>\n",
              "      <td>0.808239</td>\n",
              "      <td>0.799428</td>\n",
              "      <td>0.891066</td>\n",
              "      <td>0.985238</td>\n",
              "      <td>07:14</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>155</td>\n",
              "      <td>0.805869</td>\n",
              "      <td>0.792108</td>\n",
              "      <td>0.889539</td>\n",
              "      <td>0.984983</td>\n",
              "      <td>07:14</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>156</td>\n",
              "      <td>0.822640</td>\n",
              "      <td>0.785718</td>\n",
              "      <td>0.894375</td>\n",
              "      <td>0.984983</td>\n",
              "      <td>07:14</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>157</td>\n",
              "      <td>0.828070</td>\n",
              "      <td>0.813474</td>\n",
              "      <td>0.883176</td>\n",
              "      <td>0.984474</td>\n",
              "      <td>07:13</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>158</td>\n",
              "      <td>0.814003</td>\n",
              "      <td>0.786597</td>\n",
              "      <td>0.890812</td>\n",
              "      <td>0.983456</td>\n",
              "      <td>07:13</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>159</td>\n",
              "      <td>0.827259</td>\n",
              "      <td>0.799256</td>\n",
              "      <td>0.897175</td>\n",
              "      <td>0.980148</td>\n",
              "      <td>07:14</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>160</td>\n",
              "      <td>0.810456</td>\n",
              "      <td>0.803603</td>\n",
              "      <td>0.883685</td>\n",
              "      <td>0.980911</td>\n",
              "      <td>07:14</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>161</td>\n",
              "      <td>0.802595</td>\n",
              "      <td>0.789551</td>\n",
              "      <td>0.894375</td>\n",
              "      <td>0.983711</td>\n",
              "      <td>07:14</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>162</td>\n",
              "      <td>0.797997</td>\n",
              "      <td>0.771868</td>\n",
              "      <td>0.896666</td>\n",
              "      <td>0.984474</td>\n",
              "      <td>07:14</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>163</td>\n",
              "      <td>0.807245</td>\n",
              "      <td>0.778034</td>\n",
              "      <td>0.897175</td>\n",
              "      <td>0.982693</td>\n",
              "      <td>07:14</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>164</td>\n",
              "      <td>0.796618</td>\n",
              "      <td>0.776809</td>\n",
              "      <td>0.895648</td>\n",
              "      <td>0.985238</td>\n",
              "      <td>07:14</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>165</td>\n",
              "      <td>0.799492</td>\n",
              "      <td>0.764102</td>\n",
              "      <td>0.904047</td>\n",
              "      <td>0.981675</td>\n",
              "      <td>07:14</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>166</td>\n",
              "      <td>0.794533</td>\n",
              "      <td>0.771568</td>\n",
              "      <td>0.904556</td>\n",
              "      <td>0.981420</td>\n",
              "      <td>07:14</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>167</td>\n",
              "      <td>0.800435</td>\n",
              "      <td>0.771181</td>\n",
              "      <td>0.900484</td>\n",
              "      <td>0.985238</td>\n",
              "      <td>07:14</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>168</td>\n",
              "      <td>0.807846</td>\n",
              "      <td>0.764577</td>\n",
              "      <td>0.902265</td>\n",
              "      <td>0.984474</td>\n",
              "      <td>07:14</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>169</td>\n",
              "      <td>0.775673</td>\n",
              "      <td>0.760319</td>\n",
              "      <td>0.902265</td>\n",
              "      <td>0.984983</td>\n",
              "      <td>07:14</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>170</td>\n",
              "      <td>0.799196</td>\n",
              "      <td>0.759660</td>\n",
              "      <td>0.902265</td>\n",
              "      <td>0.985747</td>\n",
              "      <td>07:14</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>171</td>\n",
              "      <td>0.790122</td>\n",
              "      <td>0.772001</td>\n",
              "      <td>0.899975</td>\n",
              "      <td>0.986256</td>\n",
              "      <td>07:13</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>172</td>\n",
              "      <td>0.759972</td>\n",
              "      <td>0.762184</td>\n",
              "      <td>0.907610</td>\n",
              "      <td>0.982184</td>\n",
              "      <td>07:13</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>173</td>\n",
              "      <td>0.770400</td>\n",
              "      <td>0.762531</td>\n",
              "      <td>0.900229</td>\n",
              "      <td>0.982693</td>\n",
              "      <td>07:13</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>174</td>\n",
              "      <td>0.784420</td>\n",
              "      <td>0.753128</td>\n",
              "      <td>0.902520</td>\n",
              "      <td>0.984729</td>\n",
              "      <td>07:13</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>175</td>\n",
              "      <td>0.768084</td>\n",
              "      <td>0.767891</td>\n",
              "      <td>0.900993</td>\n",
              "      <td>0.983202</td>\n",
              "      <td>07:13</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>176</td>\n",
              "      <td>0.784706</td>\n",
              "      <td>0.753011</td>\n",
              "      <td>0.909137</td>\n",
              "      <td>0.984474</td>\n",
              "      <td>07:13</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>177</td>\n",
              "      <td>0.779502</td>\n",
              "      <td>0.745354</td>\n",
              "      <td>0.907610</td>\n",
              "      <td>0.986002</td>\n",
              "      <td>07:13</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>178</td>\n",
              "      <td>0.759331</td>\n",
              "      <td>0.747876</td>\n",
              "      <td>0.908628</td>\n",
              "      <td>0.983202</td>\n",
              "      <td>07:13</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>179</td>\n",
              "      <td>0.775084</td>\n",
              "      <td>0.745405</td>\n",
              "      <td>0.911173</td>\n",
              "      <td>0.985238</td>\n",
              "      <td>07:13</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>180</td>\n",
              "      <td>0.759934</td>\n",
              "      <td>0.744062</td>\n",
              "      <td>0.909392</td>\n",
              "      <td>0.984729</td>\n",
              "      <td>07:13</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>181</td>\n",
              "      <td>0.774558</td>\n",
              "      <td>0.741958</td>\n",
              "      <td>0.912700</td>\n",
              "      <td>0.982693</td>\n",
              "      <td>07:13</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>182</td>\n",
              "      <td>0.765303</td>\n",
              "      <td>0.740842</td>\n",
              "      <td>0.912955</td>\n",
              "      <td>0.983202</td>\n",
              "      <td>07:12</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>183</td>\n",
              "      <td>0.758222</td>\n",
              "      <td>0.732815</td>\n",
              "      <td>0.914991</td>\n",
              "      <td>0.983202</td>\n",
              "      <td>07:13</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>184</td>\n",
              "      <td>0.750559</td>\n",
              "      <td>0.731038</td>\n",
              "      <td>0.912955</td>\n",
              "      <td>0.983965</td>\n",
              "      <td>07:13</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>185</td>\n",
              "      <td>0.770921</td>\n",
              "      <td>0.734284</td>\n",
              "      <td>0.912700</td>\n",
              "      <td>0.983711</td>\n",
              "      <td>07:13</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>186</td>\n",
              "      <td>0.763900</td>\n",
              "      <td>0.735383</td>\n",
              "      <td>0.912191</td>\n",
              "      <td>0.984220</td>\n",
              "      <td>07:13</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>187</td>\n",
              "      <td>0.763020</td>\n",
              "      <td>0.729134</td>\n",
              "      <td>0.914228</td>\n",
              "      <td>0.983202</td>\n",
              "      <td>07:13</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>188</td>\n",
              "      <td>0.784830</td>\n",
              "      <td>0.728858</td>\n",
              "      <td>0.915500</td>\n",
              "      <td>0.984474</td>\n",
              "      <td>07:13</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>189</td>\n",
              "      <td>0.774624</td>\n",
              "      <td>0.731687</td>\n",
              "      <td>0.913464</td>\n",
              "      <td>0.982693</td>\n",
              "      <td>07:13</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>190</td>\n",
              "      <td>0.757827</td>\n",
              "      <td>0.728968</td>\n",
              "      <td>0.915246</td>\n",
              "      <td>0.983456</td>\n",
              "      <td>07:13</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>191</td>\n",
              "      <td>0.762937</td>\n",
              "      <td>0.727599</td>\n",
              "      <td>0.914482</td>\n",
              "      <td>0.983965</td>\n",
              "      <td>07:13</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>192</td>\n",
              "      <td>0.759117</td>\n",
              "      <td>0.727308</td>\n",
              "      <td>0.911937</td>\n",
              "      <td>0.983456</td>\n",
              "      <td>07:13</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>193</td>\n",
              "      <td>0.753361</td>\n",
              "      <td>0.727195</td>\n",
              "      <td>0.914482</td>\n",
              "      <td>0.984729</td>\n",
              "      <td>07:13</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>194</td>\n",
              "      <td>0.761607</td>\n",
              "      <td>0.727784</td>\n",
              "      <td>0.915500</td>\n",
              "      <td>0.984983</td>\n",
              "      <td>07:13</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>195</td>\n",
              "      <td>0.754980</td>\n",
              "      <td>0.728126</td>\n",
              "      <td>0.912700</td>\n",
              "      <td>0.983965</td>\n",
              "      <td>07:13</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>196</td>\n",
              "      <td>0.760557</td>\n",
              "      <td>0.728723</td>\n",
              "      <td>0.912191</td>\n",
              "      <td>0.983965</td>\n",
              "      <td>07:13</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <td>197</td>\n",
              "      <td>0.753103</td>\n",
              "      <td>0.726856</td>\n",
              "      <td>0.913719</td>\n",
              "      <td>0.985747</td>\n",
              "      <td>07:13</td>\n",
              "    </tr>\n",
              "  </tbody>\n",
              "</table><p>\n",
              "\n",
              "    <div>\n",
              "        <style>\n",
              "            /* Turns off some styling */\n",
              "            progress {\n",
              "                /* gets rid of default border in Firefox and Opera. */\n",
              "                border: none;\n",
              "                /* Needs to be in here for Safari polyfill so background images work as expected. */\n",
              "                background-size: auto;\n",
              "            }\n",
              "            .progress-bar-interrupted, .progress-bar-interrupted::-webkit-progress-bar {\n",
              "                background: #F44336;\n",
              "            }\n",
              "        </style>\n",
              "      <progress value='231' class='' max='564' style='width:300px; height:20px; vertical-align: middle;'></progress>\n",
              "      40.96% [231/564 02:34<03:42 0.7651]\n",
              "    </div>\n",
              "    "
            ],
            "text/plain": [
              "<IPython.core.display.HTML object>"
            ]
          },
          "metadata": {
            "tags": []
          }
        },
        {
          "output_type": "stream",
          "text": [
            "/usr/local/lib/python3.6/dist-packages/torch/nn/functional.py:2854: UserWarning: The default behavior for interpolate/upsample with float scale_factor will change in 1.6.0 to align with other frameworks/libraries, and use scale_factor directly, instead of relying on the computed output size. If you wish to keep the old behavior, please set recompute_scale_factor=True. See the documentation of nn.Upsample for details. \n",
            "  warnings.warn(\"The default behavior for interpolate/upsample with float scale_factor will change \"\n"
          ],
          "name": "stderr"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "XpJViWMRSM0U",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "res"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "0xSwy37wST5M",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "for i in range(80):\n",
        "  print(i, learn.recorder.val_losses[i].item(), learn.recorder.metrics[i][0].item())"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "fBRvEd8Io5VL",
        "colab_type": "text"
      },
      "source": [
        "## size=256"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "YljSlrw5Zysr",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "for epochs in [5]*5:\n",
        "    mixup=0 if epochs<=20 else 0.2\n",
        "    learn = get_learn(model=model, size=256, bs=16, mixup=mixup)\n",
        "    learn.fit_fc(epochs, lr=4e-3, moms=(0.95,0.95), start_pct=0.72)\n",
        "    res += [learn.recorder.metrics[-1][0].item()]\n",
        "print([round(x, 6) for x in res], sum(res)/len(res))"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "vtfqnC7vwOrB",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "for epochs in [20]*5:\n",
        "    mixup=0 if epochs<=20 else 0.2\n",
        "    learn = get_learn(model=model, size=256, bs=16, mixup=mixup)\n",
        "    learn.fit_fc(epochs, lr=4e-3, moms=(0.95,0.95), start_pct=0.72)\n",
        "    res += [learn.recorder.metrics[-1][0].item()]\n",
        "print([round(x, 6) for x in res], sum(res)/len(res))"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "zzSWWh_oo8xs",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        "for epochs in [80]:\n",
        "    mixup=0 if epochs<=20 else 0.2\n",
        "    learn = get_learn(model=model, size=256, bs=16, mixup=mixup)\n",
        "    learn.fit_fc(epochs, lr=4e-3, moms=(0.95,0.95), start_pct=0.72)\n",
        "    res += [learn.recorder.metrics[-1][0].item()]\n",
        "print([round(x, 6) for x in res], sum(res)/len(res))"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "bvL2f61cVgVT",
        "colab_type": "code",
        "colab": {}
      },
      "source": [
        ""
      ],
      "execution_count": null,
      "outputs": []
    }
  ]
}