Final_SeResNeXt50.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 0,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "8y78_c297tAs"
   },
   "outputs": [],
   "source": [
    "!pip install pretrainedmodels"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 0,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "rmqrU-a0lB4D"
   },
   "outputs": [],
   "source": [
    "import torchvision\n",
    "import pretrainedmodels\n",
    "\n",
    "from fastai import *\n",
    "from fastai.vision import *"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 0,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "xjVaC_MbbZWG"
   },
   "outputs": [],
   "source": [
    "def set_seed(seed=42):\n",
    "    # python RNG\n",
    "    import random\n",
    "    random.seed(seed)\n",
    "\n",
    "    # pytorch RNGs\n",
    "    import torch\n",
    "    torch.manual_seed(seed)\n",
    "    torch.backends.cudnn.deterministic = True\n",
    "    if torch.cuda.is_available(): torch.cuda.manual_seed_all(seed)\n",
    "\n",
    "    # numpy RNG\n",
    "    import numpy as np\n",
    "    np.random.seed(seed)\n",
    "    \n",
    "set_seed()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 0,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "s6JcCJYZJbyG"
   },
   "outputs": [],
   "source": [
    "class FocalLoss(nn.Module):\n",
    "    def __init__(self, alpha=1., gamma=2.):\n",
    "        super().__init__()\n",
    "        self.alpha = alpha\n",
    "        self.gamma = gamma\n",
    "\n",
    "    def forward(self, inputs, targets, **kwargs):\n",
    "        CE_loss = nn.CrossEntropyLoss(reduction='none')(inputs, targets)\n",
    "        pt = torch.exp(-CE_loss)\n",
    "        F_loss = self.alpha * ((1-pt)**self.gamma) * CE_loss\n",
    "        return F_loss.mean()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 0,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "gjY8sIv0VmFF"
   },
   "outputs": [],
   "source": [
    "def write_submission(preds, filename):\n",
    "    sub = pd.read_csv('data/sample_submission_ns2btKE.csv')\n",
    "    sub.category = preds\n",
    "    sub.to_csv(filename, index=False)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 0,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "EtCrzUFvlXjY"
   },
   "outputs": [],
   "source": [
    "def se_resnext50_32x4d(pretrained=False):\n",
    "    pretrained = 'imagenet' if pretrained else None\n",
    "    model = pretrainedmodels.se_resnext50_32x4d(pretrained=pretrained)\n",
    "    return nn.Sequential(*list(model.children()))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 0,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "KeuI5dQ1LEdc"
   },
   "outputs": [],
   "source": [
    "np.random.seed(42)\n",
    "src = (ImageList.from_csv('data', 'train.csv', folder='images')\n",
    "       .split_by_rand_pct(0.2, seed=42)\n",
    "       .label_from_df(cols='category')\n",
    "       .add_test(ImageList.from_csv('data/', csv_name='test_ApKoW4T.csv', folder='images'))\n",
    "      )"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 0,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "DlRsCZelQHx0"
   },
   "outputs": [],
   "source": [
    "f_score = FBeta(average='weighted', beta=1)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "colab_type": "text",
    "id": "qIU8Ve_iyTue"
   },
   "source": [
    "## Size 128"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 0,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "xM8LeMJwOHYB"
   },
   "outputs": [],
   "source": [
    "tfms = get_transforms(do_flip=True, max_rotate=15, flip_vert=False, max_lighting=0.1, max_zoom=1.05, max_warp=0., xtra_tfms=[rand_crop(), rand_zoom(1, 1.5), symmetric_warp(magnitude=(-0.2, 0.2))])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 0,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "PvG8qhwQOAJ0"
   },
   "outputs": [],
   "source": [
    "data = (src.transform(tfms, size=128, resize_method=ResizeMethod.SQUISH)\n",
    "        .databunch(bs=128).normalize(imagenet_stats))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 0,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "us_BYAx1OT45"
   },
   "outputs": [],
   "source": [
    "learn = cnn_learner(data, se_resnext50_32x4d, pretrained=True, cut=-2,\n",
    "                    split_on=lambda m: (m[0][3], m[1]), \n",
    "                    metrics=[error_rate, f_score])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 0,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "TKmLRhHOJlGZ"
   },
   "outputs": [],
   "source": [
    "learn.loss_func = FocalLoss()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 0,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 317
    },
    "colab_type": "code",
    "id": "LeBqZJ4RQScw",
    "outputId": "9606d23b-9bc3-4386-f69f-55aa4195c2f2"
   },
   "outputs": [
    {
     "data": {
      "text/html": [],
      "text/plain": [
       "<IPython.core.display.HTML object>"
      ]
     },
     "metadata": {
      "tags": []
     },
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "LR Finder is complete, type {learner_name}.recorder.plot() to see the graph.\n",
      "Min numerical gradient: 1.45E-03\n"
     ]
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAYUAAAEKCAYAAAD9xUlFAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nSUVORK5CYII=\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "tags": []
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "learn.lr_find()\n",
    "learn.recorder.plot(suggestion=True)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 0,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 1041
    },
    "colab_type": "code",
    "id": "RGroV0abQXiC",
    "outputId": "7534691d-90fd-443a-e079-444b116c3bb4"
   },
   "outputs": [
    {
     "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>error_rate</th>\n",
       "      <th>f_beta</th>\n",
       "      <th>time</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <td>0</td>\n",
       "      <td>0.783686</td>\n",
       "      <td>0.521885</td>\n",
       "      <td>0.312000</td>\n",
       "      <td>0.687195</td>\n",
       "      <td>00:25</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>1</td>\n",
       "      <td>0.576320</td>\n",
       "      <td>0.371548</td>\n",
       "      <td>0.288800</td>\n",
       "      <td>0.711119</td>\n",
       "      <td>00:23</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>2</td>\n",
       "      <td>0.445988</td>\n",
       "      <td>0.327498</td>\n",
       "      <td>0.282400</td>\n",
       "      <td>0.715822</td>\n",
       "      <td>00:23</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>3</td>\n",
       "      <td>0.377658</td>\n",
       "      <td>0.301254</td>\n",
       "      <td>0.254400</td>\n",
       "      <td>0.743485</td>\n",
       "      <td>00:23</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>4</td>\n",
       "      <td>0.344383</td>\n",
       "      <td>0.304360</td>\n",
       "      <td>0.260000</td>\n",
       "      <td>0.741288</td>\n",
       "      <td>00:23</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>5</td>\n",
       "      <td>0.316460</td>\n",
       "      <td>0.305449</td>\n",
       "      <td>0.244800</td>\n",
       "      <td>0.752542</td>\n",
       "      <td>00:23</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>6</td>\n",
       "      <td>0.302401</td>\n",
       "      <td>0.263586</td>\n",
       "      <td>0.209600</td>\n",
       "      <td>0.789321</td>\n",
       "      <td>00:23</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>7</td>\n",
       "      <td>0.274562</td>\n",
       "      <td>0.252272</td>\n",
       "      <td>0.212000</td>\n",
       "      <td>0.790174</td>\n",
       "      <td>00:23</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>8</td>\n",
       "      <td>0.248262</td>\n",
       "      <td>0.236926</td>\n",
       "      <td>0.207200</td>\n",
       "      <td>0.794623</td>\n",
       "      <td>00:23</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>9</td>\n",
       "      <td>0.223508</td>\n",
       "      <td>0.219241</td>\n",
       "      <td>0.184000</td>\n",
       "      <td>0.818284</td>\n",
       "      <td>00:23</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>10</td>\n",
       "      <td>0.202845</td>\n",
       "      <td>0.226507</td>\n",
       "      <td>0.194400</td>\n",
       "      <td>0.802224</td>\n",
       "      <td>00:23</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>11</td>\n",
       "      <td>0.182475</td>\n",
       "      <td>0.186771</td>\n",
       "      <td>0.166400</td>\n",
       "      <td>0.835300</td>\n",
       "      <td>00:23</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>12</td>\n",
       "      <td>0.162464</td>\n",
       "      <td>0.202915</td>\n",
       "      <td>0.167200</td>\n",
       "      <td>0.826292</td>\n",
       "      <td>00:23</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>13</td>\n",
       "      <td>0.153061</td>\n",
       "      <td>0.179285</td>\n",
       "      <td>0.158400</td>\n",
       "      <td>0.843841</td>\n",
       "      <td>00:23</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>14</td>\n",
       "      <td>0.143852</td>\n",
       "      <td>0.216964</td>\n",
       "      <td>0.177600</td>\n",
       "      <td>0.815048</td>\n",
       "      <td>00:23</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>15</td>\n",
       "      <td>0.140483</td>\n",
       "      <td>0.169316</td>\n",
       "      <td>0.146400</td>\n",
       "      <td>0.854426</td>\n",
       "      <td>00:23</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>16</td>\n",
       "      <td>0.124855</td>\n",
       "      <td>0.167333</td>\n",
       "      <td>0.152800</td>\n",
       "      <td>0.849770</td>\n",
       "      <td>00:23</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>17</td>\n",
       "      <td>0.115196</td>\n",
       "      <td>0.190233</td>\n",
       "      <td>0.145600</td>\n",
       "      <td>0.852743</td>\n",
       "      <td>00:23</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>18</td>\n",
       "      <td>0.103776</td>\n",
       "      <td>0.168502</td>\n",
       "      <td>0.140800</td>\n",
       "      <td>0.858473</td>\n",
       "      <td>00:23</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>19</td>\n",
       "      <td>0.100462</td>\n",
       "      <td>0.167646</td>\n",
       "      <td>0.132000</td>\n",
       "      <td>0.868022</td>\n",
       "      <td>00:23</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>20</td>\n",
       "      <td>0.087285</td>\n",
       "      <td>0.168356</td>\n",
       "      <td>0.141600</td>\n",
       "      <td>0.859925</td>\n",
       "      <td>00:23</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>21</td>\n",
       "      <td>0.084071</td>\n",
       "      <td>0.157192</td>\n",
       "      <td>0.123200</td>\n",
       "      <td>0.876481</td>\n",
       "      <td>00:23</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>22</td>\n",
       "      <td>0.078475</td>\n",
       "      <td>0.154862</td>\n",
       "      <td>0.127200</td>\n",
       "      <td>0.874314</td>\n",
       "      <td>00:23</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>23</td>\n",
       "      <td>0.074040</td>\n",
       "      <td>0.154112</td>\n",
       "      <td>0.128800</td>\n",
       "      <td>0.870557</td>\n",
       "      <td>00:23</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>24</td>\n",
       "      <td>0.067342</td>\n",
       "      <td>0.144474</td>\n",
       "      <td>0.123200</td>\n",
       "      <td>0.877361</td>\n",
       "      <td>00:24</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>25</td>\n",
       "      <td>0.060449</td>\n",
       "      <td>0.137253</td>\n",
       "      <td>0.116800</td>\n",
       "      <td>0.882813</td>\n",
       "      <td>00:23</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>26</td>\n",
       "      <td>0.056404</td>\n",
       "      <td>0.143906</td>\n",
       "      <td>0.122400</td>\n",
       "      <td>0.876358</td>\n",
       "      <td>00:23</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>27</td>\n",
       "      <td>0.049309</td>\n",
       "      <td>0.139098</td>\n",
       "      <td>0.126400</td>\n",
       "      <td>0.873910</td>\n",
       "      <td>00:23</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>28</td>\n",
       "      <td>0.047504</td>\n",
       "      <td>0.139363</td>\n",
       "      <td>0.115200</td>\n",
       "      <td>0.885185</td>\n",
       "      <td>00:23</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>29</td>\n",
       "      <td>0.044747</td>\n",
       "      <td>0.138869</td>\n",
       "      <td>0.117600</td>\n",
       "      <td>0.882512</td>\n",
       "      <td>00:23</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>30</td>\n",
       "      <td>0.043473</td>\n",
       "      <td>0.141150</td>\n",
       "      <td>0.117600</td>\n",
       "      <td>0.882309</td>\n",
       "      <td>00:23</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>31</td>\n",
       "      <td>0.043870</td>\n",
       "      <td>0.138836</td>\n",
       "      <td>0.116000</td>\n",
       "      <td>0.883903</td>\n",
       "      <td>00:23</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>"
      ],
      "text/plain": [
       "<IPython.core.display.HTML object>"
      ]
     },
     "metadata": {
      "tags": []
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "learn.fit_one_cycle(32, max_lr=slice(2e-2), wd=1e-5)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 0,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "2fw26Q84QkPm"
   },
   "outputs": [],
   "source": [
    "learn.save('SE_ResNext50_FL_883903')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 0,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "hWqvBidAi6cp"
   },
   "outputs": [],
   "source": [
    "learn.unfreeze()\n",
    "learn.clip_grad();"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 0,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 317
    },
    "colab_type": "code",
    "id": "zS6dV7FWR5Oi",
    "outputId": "863ef845-7bdc-4a1f-a3f3-47f4383f826a"
   },
   "outputs": [
    {
     "data": {
      "text/html": [],
      "text/plain": [
       "<IPython.core.display.HTML object>"
      ]
     },
     "metadata": {
      "tags": []
     },
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "LR Finder is complete, type {learner_name}.recorder.plot() to see the graph.\n",
      "Min numerical gradient: 4.37E-05\n"
     ]
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAYsAAAEKCAYAAADjDHn2AAAABHNCSVQICAgIfAhkiAAAAAlwSFlz\nAAALEgAACxIB0t1+/AAAADl0RVh0U29mdHdhcmUAbWF0cGxvdGxpYiB2ZXJzaW9uIDMuMC4zLCBo\ndHRwOi8vbWF0cGxvdGxpYi5vcmcvnQurowAAIABJREFUeJzt3Xl8XHW9//HXJ3vSJd3SNW3TlpbS\nHRoKZSmbYMGLBQQEhIuCInpRr3BV9OdVLm5wr4KouKAiCCIgi5ZSpNDKIptNt3Sle9p0SdJszdKs\n8/n9MScQQtqkTSYzSd7Px2MezZzznTmfOYS855zvOd+vuTsiIiJHEhftAkREJPYpLEREpE0KCxER\naZPCQkRE2qSwEBGRNiksRESkTQoLERFpk8JCRETapLAQEZE2JUS7gM4yZMgQz8rKinYZIiLdyooV\nKw64e0Zb7XpMWGRlZZGTkxPtMkREuhUzy2tPO52GEhGRNiksRESkTQoLERFpk8JCRETapLAQEZE2\nKSxERKRNCgsREWmTwkJEpBt78J87eGHtvohvR2EhItJNuTs/X7aFZZsKI74thYWISDe1raiK0up6\nsrMGRnxbCgsRkW5qRV4JALPHDor4thQWIiLdVM7OUgamJTIho0/Et6WwEBHpplbklTJ77EDMLOLb\nUliIiHRDxZW1bD9Q1SWnoEBhISLSLa3IKwXoks5tUFiIiHRLK/JKSYqPY/qo9C7ZnsJCRKQbyskr\nZdqo/qQkxnfJ9iIaFmY238zeNbOtZnZ7K+uTzeyJYP07ZpYVLE80s4fNbK2ZbTSzb0ayThGR7qSm\nvpG1+eVkZ3VNfwVEMCzMLB64H7gQmAJcbWZTWjS7ESh19+OAe4G7g+VXAMnuPh2YDXy+KUhERHq7\ndXvKqWsMMXts1/RXQGSPLOYAW919u7vXAY8DC1q0WQA8HPz8FHCeha8Bc6CPmSUAqUAdcDCCtYqI\ndBs5Qed2TwmLUcDuZs/zg2WttnH3BqAcGEw4OKqAfcAu4MfuXtJyA2Z2k5nlmFlOUVFR538CEZEY\nlLOzlHFD+jCkb3KXbTNWO7jnAI3ASGAccJuZjW/ZyN0fcPdsd8/OyMjo6hpFRLqcu7NyV2mXHlVA\nZMNiDzC62fPMYFmrbYJTTulAMXAN8Hd3r3f3QuANIDuCtYqIdAvbD1RRUlVHdg8Ki+XARDMbZ2ZJ\nwFXAwhZtFgLXBz9fDixzdyd86ulcADPrA5wKbIpgrSIi3cKKnV17M16TiIVF0AdxC/AisBF40t3X\nm9mdZvbxoNnvgcFmthW4FWi6vPZ+oK+ZrSccOn9w99xI1Soi0l3k5JUwIC2R8UP6dul2EyL55u6+\nGFjcYtl3mv1cQ/gy2Zavq2xtuYhIb5eTV8rsMQOJi4v84IHNxWoHt4iItFBSVcf2oipmd/EpKFBY\niIh0G+8NHthFI802p7AQEekmcvJKSIw3ZmR2zeCBzSksRES6iRU7S5k2Kr3LBg9sTmEhItIN1DY0\nkrunvMvvr2iisBAR6QbW7SmnriHUZTPjtaSwEBHpBnJ2dv3ggc0pLEREuoGcvFKyBqeR0a/rBg9s\nTmEhIhLj3J2VeaVROwUFCgsRkZi340AVxVV1XT4eVHMKCxGRGLd2TzkAJ44ZELUaFBYiIjEur7ga\ngKzBfaJWg8JCRCTG5RVXM7x/SlRuxmuisBARiXG7SqoYMzgtqjUoLEREYlxecTVjByksRETkMA7V\nNVJYUctYHVmIiMjh7CoJd26PiWLnNigsRERi2s7iKgCdhhIRkcPbFVw2q9NQIiJyWHklVfRPSWBA\nWlJU61BYiIjEsLziasZGub8CFBYiIjFtV0l11O+xAIWFiEjMamgMsaf0UNQ7t0FhISISs/aW1dAQ\n8qh3boPCQkQkZuWVhC+bHTNIfRYiInIY7402O0RHFiIichi7SqpJSohjWL+UaJeisBARiVV5xVWM\nGZRGXJxFuxSFhYhIrIqF0WabKCxERGKQu8fMPRagsBARiUkHKuuormvUkYWIiBzeruCy2VgY6gMU\nFiIiManpslmdhhIRkcPKK67GDDIHpka7FEBhISISk/KKqxiZnkpyQny0SwEUFiIiMSmvpJoxMdK5\nDQoLEZGYtKu4OiYGEGyisBARiTGVtQ0UV9XFTOc2KCxERGJOXnFw2WwMjDbbJKJhYWbzzexdM9tq\nZre3sj7ZzJ4I1r9jZlnB8k+Z2epmj5CZzYpkrSIisWJXcNlsrzgNZWbxwP3AhcAU4Gozm9Ki2Y1A\nqbsfB9wL3A3g7n9y91nuPgu4Dtjh7qsjVauISCzJK4mteywgskcWc4Ct7r7d3euAx4EFLdosAB4O\nfn4KOM/MWg6veHXwWhGRXiGvuJqBaYn0T0mMdinviWRYjAJ2N3ueHyxrtY27NwDlwOAWbT4J/Lm1\nDZjZTWaWY2Y5RUVFnVK0iEi07SqpYkyMDPPRJKY7uM3sFKDa3de1tt7dH3D3bHfPzsjI6OLqREQi\nI5aGJm8SybDYA4xu9jwzWNZqGzNLANKB4mbrr+IwRxUiIj1RXUOIvWWHYqpzGyIbFsuBiWY2zsyS\nCP/hX9iizULg+uDny4Fl7u4AZhYHXIn6K0SkF9lTdoiQE1N3bwMkROqN3b3BzG4BXgTigQfdfb2Z\n3QnkuPtC4PfAI2a2FSghHChN5gG73X17pGoUEYk1791jEWN9FhELCwB3XwwsbrHsO81+rgGuOMxr\nXwFOjWR9IiKxZldw2WxWLzoNJSIiRymvuJrUxHgy+iVHu5QPUFiIiMSQvOLwaLMfvuUsuhQWIiIx\nJHyPRWydggKFhYhIzAiFnF0lsXePBSgsRERiRmFFLTX1oZi7xwIUFiIiMaPpstlYG+oDFBYiIjGj\nabRZnYYSEZHD2lVcTXycMWpgarRL+RCFhYhIjMgrqWbkgBQS42PvT3PsVSQi0kvtPFAVU1OpNqew\nEBGJAXUNId7dX8HUkf2jXUqrFBYiIjFg0/6D1DWGmJE5INqltEphISISA9bklwMwIzM9ypW0TmEh\nIhID1uwuY1CfJDJj8EooUFiIiMSE3PwyZmamx9wAgk0UFiIiUVZV28DWwsqY7a8AhYWISNSt21NO\nyGHm6NjsrwCFhYhI1K3JLwPQkYWIiBzemvxyRg1IZUjf2JodrzmFhYhIlOXml8X0KShQWIiIRFVJ\nVR27Sw7F9CkoUFiIiERV7nv9FTqyEBGRw1izuxwzmD5KYSEiIoeRm1/GhIy+9EtJjHYpR9SusDCz\nCWaWHPx8tpl92cxi+wSbiEiMc3fW5JfH/CkoaP+RxdNAo5kdBzwAjAYei1hVIiK9wL7yGg5U1jIz\nxju3of1hEXL3BuBS4Ofu/jVgROTKEhHp+dbs7h6d29D+sKg3s6uB64FFwbLYPsEmIhLj1uSXkxhv\nnDAiNic8aq69YfEZYC7wA3ffYWbjgEciV5aISM+Xm1/G5OH9SUmMj3YpbUpoTyN33wB8GcDMBgL9\n3P3uSBYmItKThULO2vxyPj5rZLRLaZf2Xg31ipn1N7NBwErgt2Z2T2RLExHpubYfqKKitqFbdG5D\n+09Dpbv7QeAy4I/ufgrwkciVJSLSszXduT1zdM8KiwQzGwFcyfsd3CIicoxy88tJS4rnuKF9o11K\nu7Q3LO4EXgS2uftyMxsPbIlcWSIiPdua/DKmjUwnPi42p1FtqV1h4e5/cfcZ7v6F4Pl2d/9EZEsT\nEemZ6hpCrN97sFvcX9GkvR3cmWb2rJkVBo+nzSwz0sWJiPREmwsqqGsIMaOb9FdA+09D/QFYCIwM\nHs8Fy0RE5Cg1TaM6q5tcCQXtD4sMd/+DuzcEj4eAjAjWJSLSY+XuLmdgWiKjB6VGu5R2a29YFJvZ\ntWYWHzyuBYojWZiISE+1Jr+M6ZkDMOsendvQ/rC4gfBls/uBfcDlwKfbepGZzTezd81sq5nd3sr6\nZDN7Ilj/jpllNVs3w8zeMrP1ZrbWzFLaWauISMyqrmtgc0EFM7tR5za0/2qoPHf/uLtnuPtQd78E\nOOLVUGYWD9wPXAhMAa42syktmt0IlLr7ccC9wN3BaxOAR4Gb3X0qcDZQ3/6PJSISmxau3kvIifk5\nt1vqyEx5t7axfg6wNbjMtg54HFjQos0C4OHg56eA8yx8XHYBkOvuawDcvdjdGztQq4hI1D3+r118\n89m1nDJuEPMmDYl2OUelI2HR1sm2UcDuZs/zg2WttgnmyygHBgOTADezF81spZl9vQN1iohE3e//\nuYPbn1nLvIkZPPSZOSQnxP5Is821a9TZw/BOq+LDEoAzgJOBamCpma1w96XNG5nZTcBNAGPGjIlg\nOSIix8bduf8fW/nxks3Mnzqc+66e1e2CAto4sjCzCjM72MqjgvD9Fkeyh/D0q00yg2Wttgn6KdIJ\nX2WVD7zm7gfcvRpYDJzUcgPu/oC7Z7t7dkaGruQVkdji7tz993f58ZLNXHbiKH5xzYndMiigjbBw\n937u3r+VRz93b+uoZDkw0czGmVkScBXhG/uaW0h49j0IX2G1zN2d8DhU080sLQiRs4ANR/vhRESi\nJRRy7li4nl+/uo1PnTKGH18xk4T4jpz5j66OnIY6IndvMLNbCP/hjwcedPf1ZnYnkOPuC4HfA4+Y\n2VaghHCg4O6lwXwZywmf7lrs7s9HqlYRkc72g8UbefitPG6aN55vXji5W91T0RoLf5Hv/rKzsz0n\nJyfaZYiIsONAFR+551WuzB7NDy+dFtNBEfQHZ7fVrvseE4mIxKifL91CYrxx6/mTYjoojobCQkSk\nE20rquSvq/fw73OzyOiXHO1yOo3CQkSkE/1s6RZSEuP5/Lzx0S6lUyksREQ6yZaCChau2cv1p2Ux\nuG/POaoAhYWISKf56dItpCXGc9OZPeuoAhQWIiKdYtP+gzyfu4/PnD6OgX2Sol1Op1NYiIh0gvte\n3kK/5AQ+e+a4aJcSEQoLEZEOWr+3nBfW7eeGM8YxIK3nHVWAwkJEpMN++vIW+qUkcMMZPfOoAhQW\nIiIdsja/nJc2FPC5M8eTnpoY7XIiRmEhItIBP315M+mpiXzm9KxolxJRCgsRkWO0KHcvSzcVctO8\n8fRL6blHFaCwEBE5JovX7uMrj6/m5KyB3HB6z+2raKKwEBE5Sn9ft58v/3kVJ44ewB8+M4fUpO45\nodHRUFiIiByFJev3c8tjK5mRmc4fPnMyfZMjNi1QTFFYiIi008sbCviPx1YybVQ6D90wp8f3UzSn\nsBARaYdlmwr4wp9WMGVEf/544xz696KgAIWFiEibXt1cxM2PrGTy8P788cZTel1QgMJCROSICg7W\n8OU/r2LC0L48euMpPfrGuyNRWIiIHIa78/WncqltaOT+a04kPa13BgUoLEREDuuxf+3i1c1FfOui\nExif0Tfa5USVwkJEpBU7D1Tx/UUbOXPiEK49ZWy0y4k6hYWISAuNIee2v6whId7438tnEBdn0S4p\n6nrH3SQiIkfhgde2syKvlJ9+chYj0lOjXU5M0JGFiEgzG/cd5J6X3uWi6cNZMGtktMuJGQoLEZFA\nbUMjX31iNempSXz/kumY6fRTE52GEhEJ/PTlLWzaX8Hvr89mUJ+eOT3qsdKRhYgIsCKvhN+8uo2r\nTh7NeScMi3Y5MUdhISK9XnVdA7c9uYaRA1L59r9NiXY5MUmnoUSk17v7hU3sLK7m8ZtO7TVDjh8t\nHVmISK/2xtYDPPxWHjecPo5Txw+OdjkxS2EhIr3WwZp6vv5ULuMz+vD1+cdHu5yYpuMtEem1vr9o\nA/vKD/H0F04jJbHnT43aETqykCNy92iXIBIRSzcW8GROPl84ewInjhkY7XJinsJCDuuNrQc49UdL\nuf3pXBpDCg3pOUqr6rj9mbVMHt6PL583MdrldAsKC/mQxpBz38tbuPb37xByeHz5br7y+CrqG0Md\net8H/7mDa377NnvLDnVSpSLH5jsL11NWXcc9V84iOUGnn9pDYdFLuDvv7q/gQGXtEdsVV9by6T/8\ni3tf3syls0bx6tfO5lsXTWZR7j6+8OhKauobj2n7z+fu485FG3hzWzGX3P8G6/aUH9P7iHTUX3J2\n89yavXzlvIlMGdk/2uV0G+rg7sGq6xp4Y2sxyzYVsGxTIQUHa4kzmDNuEB+bPoKPThvO0H4p77Vf\nvrOELz22ipLqOu66bDqfPHk0ZsZN8yaQmhjPf/9tPZ/7Yw4PXJdNalL7v42t2lXKrU+uJnvsQL57\n8VQ+/0gOV/7mLe6/5iTOmTw0Eh9d5EMO1TXyg8UbePTtXZycNZCbz5oQ7ZK6FespHZjZ2dmek5MT\n7TKirrahkSdz8lm6sYA3txVT1xCiX3IC8yZlcNakDPLLDrF47T62FlZiBnOyBvGxGSOorG3gJ0s2\nM3pgKvd/6iSmjkz/0Hs/mbOb25/OJTtrEA9++uR23byUX1rNJfe/QVpSAs9+8TQG902m4GANNz68\nnA17D3Lngmlce+rhJ5YpP1RPKOQM1Dg90gHr9pTzlcdXsa2oipvmjee2Cybp9FPAzFa4e3ab7SIZ\nFmY2H7gPiAd+5+53tVifDPwRmA0UA590951mlgVsBN4Nmr7t7jcfaVsKi7Bv/3Utj769i6zBaZx3\nwjDOmzyU7KxBJCV88Izj5oIKns/dx+K1+9hSWAnAhdOGc/flM+ifcvh5hheu2ctXn1jNtFHp/PEz\nc444J3FFTT2X/+ot9pYf4tkvns5xQ9+flrKqtoEv/XkVyzYV8vl54/nG/MnvTTCzvaiSZZsKeXlj\nAct3lhIfZ3xj/mQ+c1qWJqERIHzU/PLGQhat2cuBylrmTcrg3MlDmTYy/QO/I6GQ89vXt/PjJe8y\nqE8SP7liFmdMHBLFymNP1MPCzOKBzcD5QD6wHLja3Tc0a/NFYIa732xmVwGXuvsng7BY5O7T2rs9\nhQXk5pex4P43uH5uFnd8fGq7X7eloILCilpOmzC4XUMyL1m/n1seW0XmoFRuOnM8F88cSZ8WRxkN\njSFufDgnfHfsDXM4/bgP/w/a0BjijufW8+jbu5g/dTiZA1NZtqmQ7QeqAJg8vB/nTh7K5oIKXt5Y\nyGkTBvN/V8xk1ABNRtMb1dQ38sq7RTyXu5dlGws5VN/IsP7JDE9PJTe/DHfI6JfMOcdncO7kYRw3\ntC/f+ds63txWzPypw/nRZdN1hNqKWAiLucAd7v7R4Pk3Adz9R83avBi0ecvMEoD9QAYwFoXFUWkM\nOZf+8g32ldew9Lazjnh00Bn+ueUAdy5az+aCSvokxfPxWSO56uQxzMhMx8z47t/W8fBbefzosulc\nPWfMYd/H3fnd6zv4weKNJMXHceqEwXzkhKGcc/xQRg9Ke6/Nkzm7ufO5DcTFGXcumMols0ZproFe\noq4h/KVi4eq9VNY2MLhPEhdOH87FM0ZyctYg4uKM4spaXt1cxNJNhby2uYiKmgYA0pLiuePiqVyR\nnanfl8OIhbC4HJjv7p8Nnl8HnOLutzRrsy5okx883wacAvQF1hM+MjkIfNvdXz/S9np7WDz6dh7f\n/us67rtqFgtmjeqSbbo7K3eV8fi/drEodx+H6hs5YUR/po/qz5M5+dw0bzzfuuiEdr3XvvJD9E9J\n/NARSnO7iqu59cnV5OSVctH04fzgEn1T7A3uemETv351G584KZNLThzJ3PGDSYg//IWc9Y0hcnaW\nsnp3GfOnDWfckD5dWG33093DogLo6+7FZjYb+Csw1d0PttjGTcBNAGPGjJmdl5cXkc8S6w5U1nLu\nj19h6sh0HvvcKVH5BnWwpp6Fq/fy+PJdrNtzkPOnDOPX184mvpP7GBpDzgOvbeeel95lQFoS508Z\nxnEZfTluaF8mDO3LyPSUD3z+mvpG8ksPsaukirziavYfrGFE/xQmDuvHxKF9yeiXrG+cMezNbQf4\n1O/e4aqTx/Cjy6ZHu5weqb1hEclLZ/cAo5s9zwyWtdYmPzgNlQ4UezjBagHcfUUQIpOADxw6uPsD\nwAMQPrKIxIfoDu56YROH6hv53iVTo/aHr39KIteeOpZrTx3LjgNVZA5M7fSgAIiPM75w9gTOmpTB\nDxdvZPHafZRV17+3Pi0pngkZfUlLimdXSTgcmn8fSogzGprdjd4/JeEDwdEYchrdcQ8HU2PIcXfM\njDgz4gzi4gwziDPjlHGDOPt4Xf4bCWXVddz6xBrGDe7Df/9b+45QJXIiGRbLgYlmNo5wKFwFXNOi\nzULgeuAt4HJgmbu7mWUAJe7eaGbjgYnA9gjW2m0t31nCUyvC49scN7RftMsB6JLD/ikj+/PoZ0/B\n3SmuqmNrYeV7j21FlRyqa2Tu+MGMGZzGmEFpjB2cxphBfRjSN4miylq2FlSypbCSLYUVbCmo5KUN\nBZRU1xFvRlychf8NgiHODA8CJOROKPi3MeT86pVtfOuiyXzuzPE6QulE7s63nl3Lgcpanv3i6aQl\n6ZawaIvYfwF3bzCzW4AXCV86+6C7rzezO4Ecd18I/B54xMy2AiWEAwVgHnCnmdUDIeBmdy+JVK3d\nVX1jiG8/u45RA1L50rnHRbucqDAzhvRNZkjf5HbPRTC0XwpD+6VwWitXaB2NmvpGbntyDT9cvIn9\n5bV8e3IScffeA48+CpWV0LcvXHst3HYbTNANYEfjqRX5LF67n2/Mn8z0zA/f8yNdTzfldWO/e307\n339+I7+5bjYfnTo82uX0SqGQc+eiDex85Cl+s/AukrwRq3//tBiJieHHU0/BhRdGr9BuZOeBKj72\ns9eZnpnOnz57akROZ8r7YqHPQiJof3kN9760mXOOz+CCKZpcPlri4ozvTkmmYdHdJNbVfLhBfX34\ncfnlkJurI4w21DeG+M8nVhMfZ9xz5SwFRQzRQILdUGVtA998JpeGkPM/H5+mc+VRZvfcQ2Jjw5Eb\n1dfDvfd2TUHd2M+XbmH17jJ+dNkMRurmy5iisOgie8sOcesTq3l1c1GHJhT655YDfPTe13hlcxHf\nvHAyYwandWKVckwefTQcBkdSXw+PPNI19RyjhsYQoSjOW/L6liJ+8Y+tXD47k4/NGBG1OqR1Og3V\nRX75ylaeWbWHZ1btYU7WIG67YBKnHMXk8Adr6vnh8xt5fPluxmf04amb5zJ77KAIViztVlnZue3a\nKb+0mpc3FHDlyaM7dLVQYUUNv31tO4++vYsR6Sl8/qzxXHLiqC4baK8x5Pz61W3c89Jmxg3pc1RD\n1UjXUQd3FyitqmPuXUu5aNoIThwzgJ8v20phRS1nThzCredPanNKx3+8W8i3nllLwcEaPjdvPF/9\nyCTNFxxL+veHior2tSvv+DweTcOffG/RRiprGxg/pA/3XXXiUV81tLfsEL95dRt/Xr6bhsYQF00f\nwY4DVazfe5Dh/VP47JnjuHrOmCPeVd9RhRU1fPWJ1byxtZiLZ47kh5dOo1+Eh6qRD1IHdwx57F+7\nqKkP8fmzJnD88H5ckT2aR9/O45evbOPSX77JR04YykdOCHdSO+AOTjjEV+ws5ZlVe5g4tC+/+uLp\nzBo9IIqfRFp17bXwu98d8VRUXVw8OXMvZFpNfYfG7So8WMPtz6xl2aZCTh0/iE+dMpYfPL+Ry371\nBl/76PF89ozxbY7Mu6u4ml++spWnV+bjDp84KZMvnD2BrCF9cHde33KAX72yje8/v5GfL9vK9adl\n8enTshjUyUOrvLq5iNueXE1lbQN3f2I6V2aPVv9bDNORRYTVNYQ44+5lHD+8H4/ceMoH1lXWNvDw\nmzv5zavbOFjTegdpfJxx81nj+fJ5EzX+fqzatg1mzIDq6sM2qUtO4YJP/5yaMeP4r48ez/xpw9s1\nH0hzC9fs5b//uo6a+kZuv3Ay188ND9leVl3H7U+v5e/r93PGcUP4yZUzGdY/5QOvLaqoZcmG/fx9\n3X7e3FZMfJxx1cmj+fxZEw47iu/KXaX86pVtvLShgJTEOC6ZNYrr5o5tda6To1HfGOInSzbz61e3\ncfywfvzimhOZOCw2bijtjaI+NlRXi9WweGZlPrc+uYaHb5jDWZMyWm1zqK6R0uo6AMzAsOBfSE2K\n12F5d/DCC+HLY5sulW3S7D6LNdPm8o2nc9m0v4LkhDjOnTyUi2eO5Jzjhx5x5sGiilrueG49z+fu\nY+boAdxz5UwmZPT9QBt354nlu/mf5zaQkhjH/14+kykj+/PiunBALM8rwR2yBqfxsRkjuH5uFkNb\nBMrhbCmo4ME3dvDsqj3U1IfIHjuQ6+aO5cJpIz40T0qTxpBTVFHLvvJDFBysYV95DfvLw/9u2HeQ\nrYWVXD1nDN+9eIpOqUaZwiIGuDsf+9k/qW8MseSr83SI3dNt2xa+PPaRR96/g/u66+CrX33v/opQ\nyFm1u5Tn1uxjUe4+DlTWkpYUz/lThnHO8UMpP1RPfmk1u0sOkV8W/rf8UD2J8cZ/fmQSn583/ogj\nrm4rquTLf17F+r3vj7k5eXg/5k8bzvxpwzl+WL9j/j0sr67nLyt288jbeeQVVzOkbzLXzBnNkH7J\n7Ck7xL6yGvaVH2JvWQ37D9bQ2OLKqqT4OIalJzMiPZXr52bpiqcYobCIAW9uO8A1v32Huy6bzlVH\nmNNBeqfGkPPOjmKeW7OPF9a9PyBickIcmQNTyRyYxuhB4X/PnTyUSe08VVPb0MjDb+6kMUREhugO\nhZzXthTxx7fy+Me7hbiHg2B4egojB6QwMj2VkQNSGZ6ewoj0lODfVAamJeoLUwxSWMSAGx9azurd\nZbxx+7k61JYjqm8MsaWgkox+yQzpm9Rt/qgWVoTvWh/SJ1lT3nZTuhoqyrYVVbJ0UyFfOW+igkLa\nlBgfx5SR/aNdxlEb2q99/R7S/ekO7gj5wxs7SEqI49pTx0a7FBGRDlNYREBpVR1Prcjn0lmjyOiX\nHO1yREQ6TGFxDA7VNfLzpVt4e3txq2PpNN2Ed8MZ46JQnYhI51OfxTH46+o9/OSlzQCMHpTKZSdm\n8omTMhkzOI3ahkYeenMnZ04cwvHDdaORiPQMCotjsGT9fjIHpvJfFxzP0yvz+dmyLdy3dAtzxg1i\nQkYfiipq+fEVM6NdpohIp1FYHKXK2gbe2FrMdXPHcsmJo7jkxFHsLTvEs6v28NSKfP61o4RJw/oy\nb2LHpuwUEYklCouj9Oq7RdRUcGpDAAAJy0lEQVQ1hj4wO93IAan8xznH8cWzJ5CbX87gbnSdvIhI\neygsjtKSDfsZ1CeJ2WM/PKy4mTFTo8KKSA+kq6GOQn1jiGWbCjlv8tAjjs8jItLT6C/eUXhnewkV\nNQ1cMHV4tEsREelSCoujsGTDflIT4zlTndci0ssoLNrJ3VmyvoB5k4ZorCcR6XUUFu20dk85+w/W\ncP4UnYISkd5HYdFOS9YXEGdw3uSh0S5FRKTLKSzaacmG/cwZN4iBnTxpvYhId6CwaIcdB6rYXFDJ\nBToFJSK9lMKiHV7asB+A85vdtS0i0psoLNphyfoCpozoz+hBadEuRUQkKhQWbThQWcuKXaVcMFVH\nFSLSeyks2rB0YwHuqL9CRHo1hUUblqwvIHNgKieM0ERGItJ79fqwqKlv5G+r95BfWo37B6dIrapt\n4PWtB7hgynANOS4ivVqvH6J89e4yvvL4agCG909hdtZAZo8ZSHbWQHYWV1PXEFJ/hYj0er0+LLLH\nDmTRl85gRV4pOXmlrMwr5fncfe+tH5iWSHYrc1eIiPQmvT4sEuLjmDYqnWmj0rn+tCwA9pUfImdn\nKSvySpk5Ol1zV4hIr9frw6I1I9JTuXhmKhfPHBntUkREYoK+MouISJsiGhZmNt/M3jWzrWZ2eyvr\nk83siWD9O2aW1WL9GDOrNLP/imSdIiJyZBELCzOLB+4HLgSmAFeb2ZQWzW4ESt39OOBe4O4W6+8B\nXohUjSIi0j6RPLKYA2x19+3uXgc8Dixo0WYB8HDw81PAeRbc0GBmlwA7gPURrFFERNohkmExCtjd\n7Hl+sKzVNu7eAJQDg82sL/AN4H8iWJ+IiLRTrHZw3wHc6+6VR2pkZjeZWY6Z5RQVFXVNZSIivVAk\nL53dA4xu9jwzWNZam3wzSwDSgWLgFOByM/tfYAAQMrMad/9F8xe7+wPAAwDZ2dkfHKtDREQ6TSTD\nYjkw0czGEQ6Fq4BrWrRZCFwPvAVcDizz8ABNZzY1MLM7gMqWQSEiIl0nYmHh7g1mdgvwIhAPPOju\n683sTiDH3RcCvwceMbOtQAnhQDkmK1asOGBmeYdZnU64P6S969qzrOXzIcCB9lV7TI70GTrjdW21\n0z7seDvtw4630z7seLuW68e2a+vu3uMfwANHs649y1p5nhOtz9AZr2urnfah9qH2Yc/fh0d6xGoH\nd2d77ijXtWfZkd4zEo51e+19XVvttA873k77sOPttA873u6Ytm9B0kgHmVmOu2dHu47uTPuw47QP\nO077sHW95ciiKzwQ7QJ6AO3DjtM+7Djtw1boyEJERNqkIwsREWmTwqIFM3vQzArNbN0xvHa2ma0N\nRtH9WdM4V8G6L5nZJjNbH9xs2GNFYh+a2R1mtsfMVgePizq/8tgRqd/DYP1tZuZmNqTzKo49Efo9\n/J6Z5Qa/g0vMrNdMeqOw+LCHgPnH+NpfAZ8DJgaP+QBmdg7hQRNnuvtU4McdLzOmPUQn78PAve4+\nK3gs7liJMe8hIrAPzWw0cAGwq4P1dQcP0fn78P/cfYa7zwIWAd/paJHdhcKiBXd/jfANgu8xswlm\n9nczW2Fmr5vZ5JavM7MRQH93f9vDHUF/BC4JVn8BuMvda4NtFEb2U0RXhPZhrxLBfXgv8HWgx3dW\nRmIfuvvBZk370Av2YxOFRfs8AHzJ3WcD/wX8spU2owiPrNuk+Si7k4AzgwmeXjWzkyNabWzq6D4E\nuCU4BfCgmQ2MXKkxq0P70MwWAHvcfU2kC41hHf49NLMfmNlu4FP0oiMLzcHdhmC49NOAvzQ79Zt8\nlG+TAAwCTgVOBp40s/HeSy5F66R9+Cvge4S/yX0P+AlwQ2fVGOs6ug/NLA34FuFTUL1SJ/0e4u7/\nD/h/ZvZN4Bbgu51WZAxTWLQtDigLzlG+J5gJcEXwdCHhP2aZzZo0H2U3H3gmCId/mVmI8PgzvWVc\n9Q7vQ3cvaPa63xI+X9ybdHQfTgDGAWuCP5SZwEozm+Pu+yNce6zojP+Xm/sTsJheEhY6DdWG4Bzl\nDjO7AsDCZrp7Y7PO1u+4+z7goJmdGlw58e/A34K3+StwTvD6SUASkR2oLKZ0xj4MziM3uRQ46itc\nurOO7kN3X+vuQ909y92zCH+BOakXBUVn/R5ObPaWC4BNXf05ouZYBpTqyQ/gz8A+oJ7w/1A3Ev5G\n9ndgDbAB+M5hXptN+I/YNuAXvH/TYxLwaLBuJXButD9nN9yHjwBrgVzC3/5GRPtzdrd92KLNTmBI\ntD9nd9uHwNPB8lzCYyyNivbn7KqH7uAWEZE26TSUiIi0SWEhIiJtUliIiEibFBYiItImhYWIiLRJ\nYSE9mplVdvH2fmdmUzrpvRqD0U3XmdlzZjagjfYDzOyLnbFtkZZ06az0aGZW6e59O/H9Ety9obPe\nr41tvVe7mT0MbHb3HxyhfRawyN2ndUV90rvoyEJ6HTPLMLOnzWx58Dg9WD7HzN4ys1Vm9qaZHR8s\n/7SZLTSzZcBSMzvbzF4xs6csPEfJn4I7fQmWZwc/VwaDzq0xs7fNbFiwfELwfK2Zfb+dRz9v8f6A\ngH3NbKmZrQzeY0HQ5i5gQnA08n9B268FnzHXzP6nE3ej9DIKC+mN7iM8N8bJwCeA3wXLNwFnuvuJ\nhEcT/WGz15wEXO7uZwXPTwT+E5gCjAdOb2U7fYC33X0m8Brh+RGatn+fu0/ng6ObtioYu+g8wneu\nA9QAl7r7SYSHkflJEFa3A9s8PGzF18zsAsJzMcwBZgGzzWxeW9sTaY0GEpTe6CPAlGYjj/YPRiRN\nBx4Oxv9xILHZa15y9+ZzI/zL3fMBzGw1kAX8s8V26nh/wMMVwPnBz3N5f46Jxzj8ZFipwXuPAjYC\nLwXLDfhh8Ic/FKwf1srrLwgeq4LnfQmHx2uH2Z7IYSkspDeKA05195rmC83sF8A/3P3S4Pz/K81W\nV7V4j9pmPzfS+v9L9f5+p+Dh2hzJIXefFQwv/iLwH8DPCM+jkAHMdvd6M9sJpLTyegN+5O6/Ocrt\ninyITkNJb7QE+FLTEzNrGrI6nfeHov50BLf/NuHTXwBXtdXY3auBLwO3mVkC4ToLg6A4BxgbNK0A\n+jV76YvADcFRE2Y2ysyGdtJnkF5GYSE9XZqZ5Td73Er4D2920Om7Abg5aPu/wI/MbBWRPer+T+BW\nM8sFjgPK23qBu68iPNLp1YTnUcg2s7WEh8/eFLQpBt4ILrX9P3dfQvg011tB26f4YJiItJsunRXp\nYsFppUPu7mZ2FXC1uy9o63Ui0aQ+C5GuNxv4RXAFUxm9aHpY6b50ZCEiIm1Sn4WIiLRJYSEiIm1S\nWIiISJsUFiIi0iaFhYiItElhISIibfr/YJoYjQvVo9EAAAAASUVORK5CYII=\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "tags": []
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "learn.lr_find()\n",
    "learn.recorder.plot(suggestion=True)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 0,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "q0wOuU_7SsPo"
   },
   "outputs": [],
   "source": [
    "learn.load('SE_ResNext50_FL_883903');\n",
    "learn.unfreeze();\n",
    "learn.clip_grad();"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 0,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "W2vr8AfDSBnE"
   },
   "outputs": [],
   "source": [
    "lr = [3e-3/100, 3e-3/20, 3e-3/10]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 0,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 1165
    },
    "colab_type": "code",
    "id": "koP5h6H1SQ2v",
    "outputId": "74cebfa1-3516-4d66-a8f0-f59b1c97c907"
   },
   "outputs": [
    {
     "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>error_rate</th>\n",
       "      <th>f_beta</th>\n",
       "      <th>time</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <td>0</td>\n",
       "      <td>0.043736</td>\n",
       "      <td>0.138167</td>\n",
       "      <td>0.119200</td>\n",
       "      <td>0.880815</td>\n",
       "      <td>00:31</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>1</td>\n",
       "      <td>0.038968</td>\n",
       "      <td>0.133116</td>\n",
       "      <td>0.113600</td>\n",
       "      <td>0.886770</td>\n",
       "      <td>00:31</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>2</td>\n",
       "      <td>0.037026</td>\n",
       "      <td>0.132410</td>\n",
       "      <td>0.109600</td>\n",
       "      <td>0.890209</td>\n",
       "      <td>00:31</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>3</td>\n",
       "      <td>0.034266</td>\n",
       "      <td>0.129377</td>\n",
       "      <td>0.105600</td>\n",
       "      <td>0.894221</td>\n",
       "      <td>00:31</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>4</td>\n",
       "      <td>0.028375</td>\n",
       "      <td>0.115355</td>\n",
       "      <td>0.098400</td>\n",
       "      <td>0.901829</td>\n",
       "      <td>00:31</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>5</td>\n",
       "      <td>0.028240</td>\n",
       "      <td>0.116211</td>\n",
       "      <td>0.092000</td>\n",
       "      <td>0.907974</td>\n",
       "      <td>00:31</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>6</td>\n",
       "      <td>0.025761</td>\n",
       "      <td>0.107698</td>\n",
       "      <td>0.089600</td>\n",
       "      <td>0.910440</td>\n",
       "      <td>00:31</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>7</td>\n",
       "      <td>0.024445</td>\n",
       "      <td>0.105821</td>\n",
       "      <td>0.089600</td>\n",
       "      <td>0.910551</td>\n",
       "      <td>00:31</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>8</td>\n",
       "      <td>0.022643</td>\n",
       "      <td>0.111722</td>\n",
       "      <td>0.081600</td>\n",
       "      <td>0.918585</td>\n",
       "      <td>00:31</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>9</td>\n",
       "      <td>0.022642</td>\n",
       "      <td>0.119540</td>\n",
       "      <td>0.088000</td>\n",
       "      <td>0.912079</td>\n",
       "      <td>00:31</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>10</td>\n",
       "      <td>0.022431</td>\n",
       "      <td>0.116339</td>\n",
       "      <td>0.083200</td>\n",
       "      <td>0.916690</td>\n",
       "      <td>00:31</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>11</td>\n",
       "      <td>0.019956</td>\n",
       "      <td>0.114786</td>\n",
       "      <td>0.079200</td>\n",
       "      <td>0.920973</td>\n",
       "      <td>00:31</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>12</td>\n",
       "      <td>0.018964</td>\n",
       "      <td>0.092383</td>\n",
       "      <td>0.072000</td>\n",
       "      <td>0.928027</td>\n",
       "      <td>00:31</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>13</td>\n",
       "      <td>0.016358</td>\n",
       "      <td>0.121642</td>\n",
       "      <td>0.079200</td>\n",
       "      <td>0.920755</td>\n",
       "      <td>00:31</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>14</td>\n",
       "      <td>0.015715</td>\n",
       "      <td>0.115850</td>\n",
       "      <td>0.071200</td>\n",
       "      <td>0.928869</td>\n",
       "      <td>00:31</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>15</td>\n",
       "      <td>0.012507</td>\n",
       "      <td>0.122206</td>\n",
       "      <td>0.074400</td>\n",
       "      <td>0.925703</td>\n",
       "      <td>00:32</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>16</td>\n",
       "      <td>0.011047</td>\n",
       "      <td>0.119633</td>\n",
       "      <td>0.069600</td>\n",
       "      <td>0.930090</td>\n",
       "      <td>00:32</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>17</td>\n",
       "      <td>0.009553</td>\n",
       "      <td>0.112349</td>\n",
       "      <td>0.068800</td>\n",
       "      <td>0.931034</td>\n",
       "      <td>00:31</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>18</td>\n",
       "      <td>0.007814</td>\n",
       "      <td>0.113934</td>\n",
       "      <td>0.065600</td>\n",
       "      <td>0.934430</td>\n",
       "      <td>00:31</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>19</td>\n",
       "      <td>0.007300</td>\n",
       "      <td>0.119257</td>\n",
       "      <td>0.067200</td>\n",
       "      <td>0.932980</td>\n",
       "      <td>00:31</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>20</td>\n",
       "      <td>0.005879</td>\n",
       "      <td>0.112212</td>\n",
       "      <td>0.063200</td>\n",
       "      <td>0.936759</td>\n",
       "      <td>00:32</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>21</td>\n",
       "      <td>0.005036</td>\n",
       "      <td>0.125484</td>\n",
       "      <td>0.063200</td>\n",
       "      <td>0.937052</td>\n",
       "      <td>00:31</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>22</td>\n",
       "      <td>0.004238</td>\n",
       "      <td>0.126211</td>\n",
       "      <td>0.064800</td>\n",
       "      <td>0.935297</td>\n",
       "      <td>00:32</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>23</td>\n",
       "      <td>0.003635</td>\n",
       "      <td>0.118383</td>\n",
       "      <td>0.062400</td>\n",
       "      <td>0.937744</td>\n",
       "      <td>00:31</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>24</td>\n",
       "      <td>0.002619</td>\n",
       "      <td>0.120874</td>\n",
       "      <td>0.063200</td>\n",
       "      <td>0.936816</td>\n",
       "      <td>00:31</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>25</td>\n",
       "      <td>0.002212</td>\n",
       "      <td>0.136469</td>\n",
       "      <td>0.065600</td>\n",
       "      <td>0.934646</td>\n",
       "      <td>00:31</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>26</td>\n",
       "      <td>0.001994</td>\n",
       "      <td>0.128101</td>\n",
       "      <td>0.061600</td>\n",
       "      <td>0.938538</td>\n",
       "      <td>00:31</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>27</td>\n",
       "      <td>0.001956</td>\n",
       "      <td>0.129030</td>\n",
       "      <td>0.059200</td>\n",
       "      <td>0.940893</td>\n",
       "      <td>00:31</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>28</td>\n",
       "      <td>0.001427</td>\n",
       "      <td>0.130144</td>\n",
       "      <td>0.059200</td>\n",
       "      <td>0.940807</td>\n",
       "      <td>00:31</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>29</td>\n",
       "      <td>0.001330</td>\n",
       "      <td>0.133692</td>\n",
       "      <td>0.056800</td>\n",
       "      <td>0.943154</td>\n",
       "      <td>00:32</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>30</td>\n",
       "      <td>0.001098</td>\n",
       "      <td>0.135824</td>\n",
       "      <td>0.058400</td>\n",
       "      <td>0.941556</td>\n",
       "      <td>00:32</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>31</td>\n",
       "      <td>0.000781</td>\n",
       "      <td>0.135328</td>\n",
       "      <td>0.061600</td>\n",
       "      <td>0.938404</td>\n",
       "      <td>00:31</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>32</td>\n",
       "      <td>0.000643</td>\n",
       "      <td>0.135326</td>\n",
       "      <td>0.058400</td>\n",
       "      <td>0.941671</td>\n",
       "      <td>00:32</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>33</td>\n",
       "      <td>0.001284</td>\n",
       "      <td>0.134550</td>\n",
       "      <td>0.059200</td>\n",
       "      <td>0.940807</td>\n",
       "      <td>00:31</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>34</td>\n",
       "      <td>0.001013</td>\n",
       "      <td>0.135397</td>\n",
       "      <td>0.056000</td>\n",
       "      <td>0.944033</td>\n",
       "      <td>00:31</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>35</td>\n",
       "      <td>0.000837</td>\n",
       "      <td>0.134098</td>\n",
       "      <td>0.057600</td>\n",
       "      <td>0.942446</td>\n",
       "      <td>00:32</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>"
      ],
      "text/plain": [
       "<IPython.core.display.HTML object>"
      ]
     },
     "metadata": {
      "tags": []
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "learn.fit_one_cycle(36, lr, wd=1e-7)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 0,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "ts518fO-SUBg"
   },
   "outputs": [],
   "source": [
    "learn.save('s2_SeResNext50_FL_128_942446')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "colab_type": "text",
    "id": "Y8ejhqi4yZRs"
   },
   "source": [
    "## Size 256"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 0,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "WspB9TcqotTp"
   },
   "outputs": [],
   "source": [
    "SZ = 256\n",
    "cutout_frac = 0.20\n",
    "p_cutout = 0.75\n",
    "cutout_sz = round(SZ*cutout_frac)\n",
    "cutout_tfm = cutout(n_holes=(1,1), length=(cutout_sz, cutout_sz), p=p_cutout)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 0,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "4k-UGhXmpDVY"
   },
   "outputs": [],
   "source": [
    "tfms = get_transforms(do_flip=True, max_rotate=15, flip_vert=False, max_lighting=0.1,\n",
    "                      max_zoom=1.05, max_warp=0.,\n",
    "                      xtra_tfms=[rand_crop(), rand_zoom(1, 1.5),\n",
    "                                 symmetric_warp(magnitude=(-0.2, 0.2)), cutout_tfm])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 0,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "q6WIPZG_WPck"
   },
   "outputs": [],
   "source": [
    "data = (src.transform(tfms, size=256, resize_method=ResizeMethod.SQUISH)\n",
    "        .databunch(bs=72)\n",
    "        .normalize(imagenet_stats))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 0,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "c7Vyfz7AoMv1"
   },
   "outputs": [],
   "source": [
    "learn.data = data"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 0,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "kxfnne15o4G3"
   },
   "outputs": [],
   "source": [
    "data.show_batch()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 0,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "OvjZVan-PiZD"
   },
   "outputs": [],
   "source": [
    "learn.load('s2_SeResNext50_FL_128_942446');"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 0,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "KQ1BqLrEW_Nb"
   },
   "outputs": [],
   "source": [
    "learn.freeze();\n",
    "learn.clip_grad();"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 0,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "8qV-HgTljnOt"
   },
   "outputs": [],
   "source": [
    "learn.loss_func = FocalLoss()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 0,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 317
    },
    "colab_type": "code",
    "id": "gFXiAkC1XD1b",
    "outputId": "fe746fc7-c7c8-46cd-93d8-c7185fafd089"
   },
   "outputs": [
    {
     "data": {
      "text/html": [],
      "text/plain": [
       "<IPython.core.display.HTML object>"
      ]
     },
     "metadata": {
      "tags": []
     },
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "LR Finder is complete, type {learner_name}.recorder.plot() to see the graph.\n",
      "Min numerical gradient: 6.31E-07\n"
     ]
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAZgAAAEKCAYAAAAvlUMdAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nVHG2u/l+AAAAAElFTkSuQmCC\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "tags": []
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "learn.lr_find()\n",
    "learn.recorder.plot(suggestion=True)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 0,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 793
    },
    "colab_type": "code",
    "id": "AwsccXWuXH1u",
    "outputId": "e5b95d71-c08b-46e0-aebf-780492eb5d60"
   },
   "outputs": [
    {
     "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>error_rate</th>\n",
       "      <th>f_beta</th>\n",
       "      <th>time</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <td>0</td>\n",
       "      <td>0.063797</td>\n",
       "      <td>0.166564</td>\n",
       "      <td>0.069600</td>\n",
       "      <td>0.930556</td>\n",
       "      <td>01:20</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>1</td>\n",
       "      <td>0.061465</td>\n",
       "      <td>0.139410</td>\n",
       "      <td>0.064000</td>\n",
       "      <td>0.936043</td>\n",
       "      <td>01:21</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>2</td>\n",
       "      <td>0.047806</td>\n",
       "      <td>0.113662</td>\n",
       "      <td>0.061600</td>\n",
       "      <td>0.938451</td>\n",
       "      <td>01:21</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>3</td>\n",
       "      <td>0.035331</td>\n",
       "      <td>0.097657</td>\n",
       "      <td>0.058400</td>\n",
       "      <td>0.941587</td>\n",
       "      <td>01:21</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>4</td>\n",
       "      <td>0.032698</td>\n",
       "      <td>0.090880</td>\n",
       "      <td>0.057600</td>\n",
       "      <td>0.942443</td>\n",
       "      <td>01:20</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>5</td>\n",
       "      <td>0.029216</td>\n",
       "      <td>0.087670</td>\n",
       "      <td>0.056800</td>\n",
       "      <td>0.943196</td>\n",
       "      <td>01:20</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>6</td>\n",
       "      <td>0.024367</td>\n",
       "      <td>0.084406</td>\n",
       "      <td>0.055200</td>\n",
       "      <td>0.944813</td>\n",
       "      <td>01:20</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>7</td>\n",
       "      <td>0.020505</td>\n",
       "      <td>0.088171</td>\n",
       "      <td>0.054400</td>\n",
       "      <td>0.945752</td>\n",
       "      <td>01:20</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>8</td>\n",
       "      <td>0.016836</td>\n",
       "      <td>0.093613</td>\n",
       "      <td>0.054400</td>\n",
       "      <td>0.945729</td>\n",
       "      <td>01:20</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>9</td>\n",
       "      <td>0.016104</td>\n",
       "      <td>0.096086</td>\n",
       "      <td>0.054400</td>\n",
       "      <td>0.945636</td>\n",
       "      <td>01:20</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>10</td>\n",
       "      <td>0.014609</td>\n",
       "      <td>0.098525</td>\n",
       "      <td>0.052800</td>\n",
       "      <td>0.947146</td>\n",
       "      <td>01:21</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>11</td>\n",
       "      <td>0.013184</td>\n",
       "      <td>0.100520</td>\n",
       "      <td>0.052800</td>\n",
       "      <td>0.947299</td>\n",
       "      <td>01:20</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>12</td>\n",
       "      <td>0.009847</td>\n",
       "      <td>0.090707</td>\n",
       "      <td>0.049600</td>\n",
       "      <td>0.950340</td>\n",
       "      <td>01:20</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>13</td>\n",
       "      <td>0.008957</td>\n",
       "      <td>0.094563</td>\n",
       "      <td>0.052800</td>\n",
       "      <td>0.947349</td>\n",
       "      <td>01:21</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>14</td>\n",
       "      <td>0.010263</td>\n",
       "      <td>0.095692</td>\n",
       "      <td>0.048800</td>\n",
       "      <td>0.951021</td>\n",
       "      <td>01:21</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>15</td>\n",
       "      <td>0.009615</td>\n",
       "      <td>0.101324</td>\n",
       "      <td>0.056000</td>\n",
       "      <td>0.944237</td>\n",
       "      <td>01:20</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>16</td>\n",
       "      <td>0.007793</td>\n",
       "      <td>0.094891</td>\n",
       "      <td>0.047200</td>\n",
       "      <td>0.952697</td>\n",
       "      <td>01:21</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>17</td>\n",
       "      <td>0.007001</td>\n",
       "      <td>0.096686</td>\n",
       "      <td>0.045600</td>\n",
       "      <td>0.954329</td>\n",
       "      <td>01:21</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>18</td>\n",
       "      <td>0.007133</td>\n",
       "      <td>0.094707</td>\n",
       "      <td>0.042400</td>\n",
       "      <td>0.957541</td>\n",
       "      <td>01:21</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>19</td>\n",
       "      <td>0.006965</td>\n",
       "      <td>0.093863</td>\n",
       "      <td>0.044000</td>\n",
       "      <td>0.955950</td>\n",
       "      <td>01:21</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>20</td>\n",
       "      <td>0.005483</td>\n",
       "      <td>0.097482</td>\n",
       "      <td>0.049600</td>\n",
       "      <td>0.950517</td>\n",
       "      <td>01:20</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>21</td>\n",
       "      <td>0.007344</td>\n",
       "      <td>0.094666</td>\n",
       "      <td>0.048800</td>\n",
       "      <td>0.951207</td>\n",
       "      <td>01:21</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>22</td>\n",
       "      <td>0.005832</td>\n",
       "      <td>0.095703</td>\n",
       "      <td>0.047200</td>\n",
       "      <td>0.952785</td>\n",
       "      <td>01:21</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>23</td>\n",
       "      <td>0.007141</td>\n",
       "      <td>0.093255</td>\n",
       "      <td>0.044800</td>\n",
       "      <td>0.955163</td>\n",
       "      <td>01:21</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>"
      ],
      "text/plain": [
       "<IPython.core.display.HTML object>"
      ]
     },
     "metadata": {
      "tags": []
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "learn.fit_one_cycle(24, slice(3e-3), wd=5e-6)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 0,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "9I0No7aOX7CY"
   },
   "outputs": [],
   "source": [
    "learn.save('SeResNxt50_FL_128t256_955163')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 0,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "X7GobIWWf6cw"
   },
   "outputs": [],
   "source": [
    "learn.load('SeResNxt50_FL_128t256_955163');"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 0,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "XaTD6qdYZZjU"
   },
   "outputs": [],
   "source": [
    "learn.unfreeze()\n",
    "learn.clip_grad()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 0,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 1041
    },
    "colab_type": "code",
    "id": "8RqmZw0MZggO",
    "outputId": "e9a0f7f6-609e-434d-872e-62facd657d53"
   },
   "outputs": [
    {
     "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>error_rate</th>\n",
       "      <th>f_beta</th>\n",
       "      <th>time</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <td>0</td>\n",
       "      <td>0.004881</td>\n",
       "      <td>0.095006</td>\n",
       "      <td>0.045600</td>\n",
       "      <td>0.954393</td>\n",
       "      <td>01:47</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>1</td>\n",
       "      <td>0.003097</td>\n",
       "      <td>0.093848</td>\n",
       "      <td>0.044800</td>\n",
       "      <td>0.955180</td>\n",
       "      <td>01:51</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>2</td>\n",
       "      <td>0.004433</td>\n",
       "      <td>0.095179</td>\n",
       "      <td>0.048800</td>\n",
       "      <td>0.951225</td>\n",
       "      <td>01:51</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>3</td>\n",
       "      <td>0.004333</td>\n",
       "      <td>0.090997</td>\n",
       "      <td>0.048800</td>\n",
       "      <td>0.951150</td>\n",
       "      <td>01:51</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>4</td>\n",
       "      <td>0.005034</td>\n",
       "      <td>0.090667</td>\n",
       "      <td>0.044000</td>\n",
       "      <td>0.956029</td>\n",
       "      <td>01:51</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>5</td>\n",
       "      <td>0.006870</td>\n",
       "      <td>0.082818</td>\n",
       "      <td>0.040800</td>\n",
       "      <td>0.959146</td>\n",
       "      <td>01:51</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>6</td>\n",
       "      <td>0.004298</td>\n",
       "      <td>0.078488</td>\n",
       "      <td>0.042400</td>\n",
       "      <td>0.957726</td>\n",
       "      <td>01:51</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>7</td>\n",
       "      <td>0.005833</td>\n",
       "      <td>0.082735</td>\n",
       "      <td>0.038400</td>\n",
       "      <td>0.961643</td>\n",
       "      <td>01:51</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>8</td>\n",
       "      <td>0.005891</td>\n",
       "      <td>0.078535</td>\n",
       "      <td>0.040000</td>\n",
       "      <td>0.960144</td>\n",
       "      <td>01:51</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>9</td>\n",
       "      <td>0.006033</td>\n",
       "      <td>0.090153</td>\n",
       "      <td>0.038400</td>\n",
       "      <td>0.961598</td>\n",
       "      <td>01:51</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>10</td>\n",
       "      <td>0.004895</td>\n",
       "      <td>0.087074</td>\n",
       "      <td>0.035200</td>\n",
       "      <td>0.964893</td>\n",
       "      <td>01:51</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>11</td>\n",
       "      <td>0.004047</td>\n",
       "      <td>0.087246</td>\n",
       "      <td>0.035200</td>\n",
       "      <td>0.964775</td>\n",
       "      <td>01:51</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>12</td>\n",
       "      <td>0.003657</td>\n",
       "      <td>0.102892</td>\n",
       "      <td>0.034400</td>\n",
       "      <td>0.965649</td>\n",
       "      <td>01:51</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>13</td>\n",
       "      <td>0.002231</td>\n",
       "      <td>0.099810</td>\n",
       "      <td>0.037600</td>\n",
       "      <td>0.962434</td>\n",
       "      <td>01:51</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>14</td>\n",
       "      <td>0.004309</td>\n",
       "      <td>0.107549</td>\n",
       "      <td>0.036000</td>\n",
       "      <td>0.964132</td>\n",
       "      <td>01:51</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>15</td>\n",
       "      <td>0.004083</td>\n",
       "      <td>0.092879</td>\n",
       "      <td>0.038400</td>\n",
       "      <td>0.961631</td>\n",
       "      <td>01:51</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>16</td>\n",
       "      <td>0.002758</td>\n",
       "      <td>0.093545</td>\n",
       "      <td>0.035200</td>\n",
       "      <td>0.964774</td>\n",
       "      <td>01:51</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>17</td>\n",
       "      <td>0.002031</td>\n",
       "      <td>0.097141</td>\n",
       "      <td>0.039200</td>\n",
       "      <td>0.960856</td>\n",
       "      <td>01:51</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>18</td>\n",
       "      <td>0.001920</td>\n",
       "      <td>0.094154</td>\n",
       "      <td>0.040000</td>\n",
       "      <td>0.960027</td>\n",
       "      <td>01:51</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>19</td>\n",
       "      <td>0.000962</td>\n",
       "      <td>0.089933</td>\n",
       "      <td>0.035200</td>\n",
       "      <td>0.964765</td>\n",
       "      <td>01:51</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>20</td>\n",
       "      <td>0.001814</td>\n",
       "      <td>0.084181</td>\n",
       "      <td>0.030400</td>\n",
       "      <td>0.969598</td>\n",
       "      <td>01:51</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>21</td>\n",
       "      <td>0.001581</td>\n",
       "      <td>0.088822</td>\n",
       "      <td>0.035200</td>\n",
       "      <td>0.964843</td>\n",
       "      <td>01:51</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>22</td>\n",
       "      <td>0.000705</td>\n",
       "      <td>0.085714</td>\n",
       "      <td>0.036000</td>\n",
       "      <td>0.964073</td>\n",
       "      <td>01:51</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>23</td>\n",
       "      <td>0.001729</td>\n",
       "      <td>0.088255</td>\n",
       "      <td>0.032000</td>\n",
       "      <td>0.968061</td>\n",
       "      <td>01:51</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>24</td>\n",
       "      <td>0.001094</td>\n",
       "      <td>0.093934</td>\n",
       "      <td>0.036000</td>\n",
       "      <td>0.964019</td>\n",
       "      <td>01:51</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>25</td>\n",
       "      <td>0.000883</td>\n",
       "      <td>0.087338</td>\n",
       "      <td>0.030400</td>\n",
       "      <td>0.969581</td>\n",
       "      <td>01:50</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>26</td>\n",
       "      <td>0.000346</td>\n",
       "      <td>0.083657</td>\n",
       "      <td>0.030400</td>\n",
       "      <td>0.969588</td>\n",
       "      <td>01:51</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>27</td>\n",
       "      <td>0.000246</td>\n",
       "      <td>0.084896</td>\n",
       "      <td>0.030400</td>\n",
       "      <td>0.969626</td>\n",
       "      <td>01:51</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>28</td>\n",
       "      <td>0.000541</td>\n",
       "      <td>0.088079</td>\n",
       "      <td>0.034400</td>\n",
       "      <td>0.965534</td>\n",
       "      <td>01:51</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>29</td>\n",
       "      <td>0.000236</td>\n",
       "      <td>0.086822</td>\n",
       "      <td>0.033600</td>\n",
       "      <td>0.966362</td>\n",
       "      <td>01:51</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>30</td>\n",
       "      <td>0.000608</td>\n",
       "      <td>0.087041</td>\n",
       "      <td>0.034400</td>\n",
       "      <td>0.965644</td>\n",
       "      <td>01:50</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>31</td>\n",
       "      <td>0.000341</td>\n",
       "      <td>0.085005</td>\n",
       "      <td>0.032000</td>\n",
       "      <td>0.968019</td>\n",
       "      <td>01:50</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>"
      ],
      "text/plain": [
       "<IPython.core.display.HTML object>"
      ]
     },
     "metadata": {
      "tags": []
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "lr = [1e-3/200, 1e-3/20, 1e-3/10]\n",
    "learn.fit_one_cycle(32, lr)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 0,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "QGlDDF4oZ95K"
   },
   "outputs": [],
   "source": [
    "learn.save('SeResNxt50_FL_s2_128t256_968019')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 0,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "vnPsmwPHVlf8"
   },
   "outputs": [],
   "source": [
    "learn.load('SeResNxt50_FL_s2_128t256_968019');"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 0,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "jHU0mno9eOMG"
   },
   "outputs": [],
   "source": [
    "preds, _ = learn.get_preds(DatasetType.Test)\n",
    "pred_class = torch.argmax(preds, dim=1)\n",
    "pred_class += 1\n",
    "pred_numpy = pred_class.numpy()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 0,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "0x-lElGs1mKI"
   },
   "outputs": [],
   "source": [
    "write_submission(pred_numpy, 'Res50_FLmixup_aug_128t246_969718.csv')\n",
    "#LB 0.972853019760771"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "colab_type": "text",
    "id": "_8c1bfAiycv4"
   },
   "source": [
    "## Size 299"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 0,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "V7ufPboBOiYH"
   },
   "outputs": [],
   "source": [
    "SZ = 299\n",
    "cutout_frac = 0.20\n",
    "p_cutout = 0.75\n",
    "cutout_sz = round(SZ*cutout_frac)\n",
    "cutout_tfm = cutout(n_holes=(1,1), length=(cutout_sz, cutout_sz), p=p_cutout)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 0,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "adTZzykpOnFm"
   },
   "outputs": [],
   "source": [
    "tfms = get_transforms(do_flip=True, max_rotate=15, flip_vert=False, max_lighting=0.1,\n",
    "                      max_zoom=1.05, max_warp=0.,\n",
    "                      xtra_tfms=[rand_crop(),\n",
    "                                 symmetric_warp(magnitude=(-0.2, 0.2)), cutout_tfm])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 0,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "KlOQZiwxMnMz"
   },
   "outputs": [],
   "source": [
    "data_299 = (src.transform(tfms, size=299, resize_method=ResizeMethod.SQUISH)\n",
    "            .databunch(bs=48)\n",
    "            .normalize(imagenet_stats))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 0,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "hVLcXcAqMtcN"
   },
   "outputs": [],
   "source": [
    "learn = cnn_learner(data_299, se_resnext50_32x4d, pretrained=True, cut=-2,\n",
    "                    split_on=lambda m: (m[0][3], m[1]), \n",
    "                    metrics=[error_rate, f_score])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 0,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "1DVIW86w37O6"
   },
   "outputs": [],
   "source": [
    "learn.load('SeResNxt50_FL_s2_128t256_968019');\n",
    "learn.freeze();\n",
    "learn.clip_grad();\n",
    "learn.mixup();"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 0,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "BpQfO5NiM9B_"
   },
   "outputs": [],
   "source": [
    "learn.lr_find()\n",
    "learn.recorder.plot(suggestion=True)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 0,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 34
    },
    "colab_type": "code",
    "id": "gcB6sCmaU3y7",
    "outputId": "6a33f983-4cd6-4ef1-8170-3666794a0396"
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "FlattenedLoss of CrossEntropyLoss()"
      ]
     },
     "execution_count": 18,
     "metadata": {
      "tags": []
     },
     "output_type": "execute_result"
    }
   ],
   "source": [
    "learn.loss_func"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 0,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 1041
    },
    "colab_type": "code",
    "id": "95I-FsnEP9u_",
    "outputId": "a93d6d57-3c18-4169-f04c-0523f9431e0d"
   },
   "outputs": [
    {
     "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>error_rate</th>\n",
       "      <th>f_beta</th>\n",
       "      <th>time</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <td>0</td>\n",
       "      <td>0.845434</td>\n",
       "      <td>0.126683</td>\n",
       "      <td>0.039200</td>\n",
       "      <td>0.960735</td>\n",
       "      <td>02:10</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>1</td>\n",
       "      <td>0.595398</td>\n",
       "      <td>0.166088</td>\n",
       "      <td>0.042400</td>\n",
       "      <td>0.957502</td>\n",
       "      <td>02:06</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>2</td>\n",
       "      <td>0.538603</td>\n",
       "      <td>0.165132</td>\n",
       "      <td>0.037600</td>\n",
       "      <td>0.962391</td>\n",
       "      <td>02:06</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>3</td>\n",
       "      <td>0.527696</td>\n",
       "      <td>0.156787</td>\n",
       "      <td>0.042400</td>\n",
       "      <td>0.957512</td>\n",
       "      <td>02:06</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>4</td>\n",
       "      <td>0.518446</td>\n",
       "      <td>0.153922</td>\n",
       "      <td>0.036800</td>\n",
       "      <td>0.963333</td>\n",
       "      <td>02:06</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>5</td>\n",
       "      <td>0.518304</td>\n",
       "      <td>0.160560</td>\n",
       "      <td>0.044800</td>\n",
       "      <td>0.955271</td>\n",
       "      <td>02:06</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>6</td>\n",
       "      <td>0.506956</td>\n",
       "      <td>0.135375</td>\n",
       "      <td>0.038400</td>\n",
       "      <td>0.961556</td>\n",
       "      <td>02:06</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>7</td>\n",
       "      <td>0.517231</td>\n",
       "      <td>0.141696</td>\n",
       "      <td>0.038400</td>\n",
       "      <td>0.961632</td>\n",
       "      <td>02:06</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>8</td>\n",
       "      <td>0.511918</td>\n",
       "      <td>0.142817</td>\n",
       "      <td>0.036800</td>\n",
       "      <td>0.963260</td>\n",
       "      <td>02:07</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>9</td>\n",
       "      <td>0.503191</td>\n",
       "      <td>0.145534</td>\n",
       "      <td>0.032800</td>\n",
       "      <td>0.967152</td>\n",
       "      <td>02:07</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>10</td>\n",
       "      <td>0.502531</td>\n",
       "      <td>0.157464</td>\n",
       "      <td>0.040000</td>\n",
       "      <td>0.959823</td>\n",
       "      <td>02:06</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>11</td>\n",
       "      <td>0.505680</td>\n",
       "      <td>0.155435</td>\n",
       "      <td>0.035200</td>\n",
       "      <td>0.964761</td>\n",
       "      <td>02:07</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>12</td>\n",
       "      <td>0.510461</td>\n",
       "      <td>0.141550</td>\n",
       "      <td>0.034400</td>\n",
       "      <td>0.965679</td>\n",
       "      <td>02:07</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>13</td>\n",
       "      <td>0.498464</td>\n",
       "      <td>0.127980</td>\n",
       "      <td>0.031200</td>\n",
       "      <td>0.968834</td>\n",
       "      <td>02:06</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>14</td>\n",
       "      <td>0.492678</td>\n",
       "      <td>0.131714</td>\n",
       "      <td>0.028000</td>\n",
       "      <td>0.972059</td>\n",
       "      <td>02:06</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>15</td>\n",
       "      <td>0.479244</td>\n",
       "      <td>0.137867</td>\n",
       "      <td>0.034400</td>\n",
       "      <td>0.965805</td>\n",
       "      <td>02:07</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>16</td>\n",
       "      <td>0.477795</td>\n",
       "      <td>0.125026</td>\n",
       "      <td>0.027200</td>\n",
       "      <td>0.972812</td>\n",
       "      <td>02:07</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>17</td>\n",
       "      <td>0.480736</td>\n",
       "      <td>0.130275</td>\n",
       "      <td>0.028800</td>\n",
       "      <td>0.971219</td>\n",
       "      <td>02:07</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>18</td>\n",
       "      <td>0.476184</td>\n",
       "      <td>0.119645</td>\n",
       "      <td>0.028000</td>\n",
       "      <td>0.971967</td>\n",
       "      <td>02:07</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>19</td>\n",
       "      <td>0.476334</td>\n",
       "      <td>0.123273</td>\n",
       "      <td>0.031200</td>\n",
       "      <td>0.968821</td>\n",
       "      <td>02:06</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>20</td>\n",
       "      <td>0.472693</td>\n",
       "      <td>0.128429</td>\n",
       "      <td>0.034400</td>\n",
       "      <td>0.965606</td>\n",
       "      <td>02:07</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>21</td>\n",
       "      <td>0.479795</td>\n",
       "      <td>0.128557</td>\n",
       "      <td>0.030400</td>\n",
       "      <td>0.969566</td>\n",
       "      <td>02:07</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>22</td>\n",
       "      <td>0.467572</td>\n",
       "      <td>0.126836</td>\n",
       "      <td>0.031200</td>\n",
       "      <td>0.968753</td>\n",
       "      <td>02:07</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>23</td>\n",
       "      <td>0.461139</td>\n",
       "      <td>0.119264</td>\n",
       "      <td>0.029600</td>\n",
       "      <td>0.970374</td>\n",
       "      <td>02:06</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>24</td>\n",
       "      <td>0.452685</td>\n",
       "      <td>0.120086</td>\n",
       "      <td>0.029600</td>\n",
       "      <td>0.970411</td>\n",
       "      <td>02:07</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>25</td>\n",
       "      <td>0.454347</td>\n",
       "      <td>0.119216</td>\n",
       "      <td>0.029600</td>\n",
       "      <td>0.970392</td>\n",
       "      <td>02:06</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>26</td>\n",
       "      <td>0.456150</td>\n",
       "      <td>0.123479</td>\n",
       "      <td>0.028800</td>\n",
       "      <td>0.971189</td>\n",
       "      <td>02:06</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>27</td>\n",
       "      <td>0.452439</td>\n",
       "      <td>0.123904</td>\n",
       "      <td>0.028800</td>\n",
       "      <td>0.971245</td>\n",
       "      <td>02:06</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>28</td>\n",
       "      <td>0.450976</td>\n",
       "      <td>0.121359</td>\n",
       "      <td>0.031200</td>\n",
       "      <td>0.968857</td>\n",
       "      <td>02:06</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>29</td>\n",
       "      <td>0.451592</td>\n",
       "      <td>0.123352</td>\n",
       "      <td>0.028800</td>\n",
       "      <td>0.971244</td>\n",
       "      <td>02:06</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>30</td>\n",
       "      <td>0.453613</td>\n",
       "      <td>0.121988</td>\n",
       "      <td>0.031200</td>\n",
       "      <td>0.968846</td>\n",
       "      <td>02:06</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>31</td>\n",
       "      <td>0.456532</td>\n",
       "      <td>0.123133</td>\n",
       "      <td>0.029600</td>\n",
       "      <td>0.970390</td>\n",
       "      <td>02:06</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>"
      ],
      "text/plain": [
       "<IPython.core.display.HTML object>"
      ]
     },
     "metadata": {
      "tags": []
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "learn.fit_one_cycle(32, slice(1e-2))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 0,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "aZ53flG79T5t"
   },
   "outputs": [],
   "source": [
    "learn.save('SeResNext50_mixup_128t224t299_s1_970390')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 0,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "DsarzG2p7i86"
   },
   "outputs": [],
   "source": [
    "learn.load('SeResNext50_mixup_128t224t299_s1_970390');"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 0,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "0LL2gHYX9T1_"
   },
   "outputs": [],
   "source": [
    "learn.unfreeze()\n",
    "learn.clip_grad();\n",
    "learn.mixup();"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 0,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 317
    },
    "colab_type": "code",
    "id": "OWAUEzp9MK2n",
    "outputId": "893cbcf7-5a67-4925-d07a-46661c9647c0"
   },
   "outputs": [
    {
     "data": {
      "text/html": [],
      "text/plain": [
       "<IPython.core.display.HTML object>"
      ]
     },
     "metadata": {
      "tags": []
     },
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "LR Finder is complete, type {learner_name}.recorder.plot() to see the graph.\n",
      "Min numerical gradient: 6.31E-07\n"
     ]
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAYUAAAEKCAYAAAD9xUlFAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nrkJggg==\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "tags": []
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "learn.lr_find()\n",
    "learn.recorder.plot(suggestion=True)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 0,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 2033
    },
    "colab_type": "code",
    "id": "wUfJobuglx83",
    "outputId": "9924ba91-4eee-465c-ae3e-c793579a19d3"
   },
   "outputs": [
    {
     "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>error_rate</th>\n",
       "      <th>f_beta</th>\n",
       "      <th>time</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <td>0</td>\n",
       "      <td>0.458171</td>\n",
       "      <td>0.123312</td>\n",
       "      <td>0.030400</td>\n",
       "      <td>0.969585</td>\n",
       "      <td>03:00</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>1</td>\n",
       "      <td>0.442336</td>\n",
       "      <td>0.123213</td>\n",
       "      <td>0.032000</td>\n",
       "      <td>0.967988</td>\n",
       "      <td>02:50</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>2</td>\n",
       "      <td>0.446681</td>\n",
       "      <td>0.121806</td>\n",
       "      <td>0.028000</td>\n",
       "      <td>0.972012</td>\n",
       "      <td>02:50</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>3</td>\n",
       "      <td>0.445108</td>\n",
       "      <td>0.119733</td>\n",
       "      <td>0.031200</td>\n",
       "      <td>0.968784</td>\n",
       "      <td>02:50</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>4</td>\n",
       "      <td>0.449005</td>\n",
       "      <td>0.117211</td>\n",
       "      <td>0.028800</td>\n",
       "      <td>0.971261</td>\n",
       "      <td>02:50</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>5</td>\n",
       "      <td>0.444636</td>\n",
       "      <td>0.117236</td>\n",
       "      <td>0.026400</td>\n",
       "      <td>0.973619</td>\n",
       "      <td>02:50</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>6</td>\n",
       "      <td>0.437254</td>\n",
       "      <td>0.116157</td>\n",
       "      <td>0.028800</td>\n",
       "      <td>0.971179</td>\n",
       "      <td>02:50</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>7</td>\n",
       "      <td>0.442069</td>\n",
       "      <td>0.120598</td>\n",
       "      <td>0.037600</td>\n",
       "      <td>0.962389</td>\n",
       "      <td>02:51</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>8</td>\n",
       "      <td>0.438174</td>\n",
       "      <td>0.115766</td>\n",
       "      <td>0.028000</td>\n",
       "      <td>0.971990</td>\n",
       "      <td>02:51</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>9</td>\n",
       "      <td>0.441405</td>\n",
       "      <td>0.141391</td>\n",
       "      <td>0.047200</td>\n",
       "      <td>0.952582</td>\n",
       "      <td>02:51</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>10</td>\n",
       "      <td>0.442121</td>\n",
       "      <td>0.143395</td>\n",
       "      <td>0.037600</td>\n",
       "      <td>0.962395</td>\n",
       "      <td>02:51</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>11</td>\n",
       "      <td>0.453021</td>\n",
       "      <td>0.153896</td>\n",
       "      <td>0.044800</td>\n",
       "      <td>0.955317</td>\n",
       "      <td>02:52</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>12</td>\n",
       "      <td>0.445450</td>\n",
       "      <td>0.176421</td>\n",
       "      <td>0.048800</td>\n",
       "      <td>0.951534</td>\n",
       "      <td>02:51</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>13</td>\n",
       "      <td>0.451379</td>\n",
       "      <td>0.143151</td>\n",
       "      <td>0.044000</td>\n",
       "      <td>0.956087</td>\n",
       "      <td>02:50</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>14</td>\n",
       "      <td>0.438960</td>\n",
       "      <td>0.133209</td>\n",
       "      <td>0.036800</td>\n",
       "      <td>0.963252</td>\n",
       "      <td>02:50</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>15</td>\n",
       "      <td>0.446062</td>\n",
       "      <td>0.144768</td>\n",
       "      <td>0.040000</td>\n",
       "      <td>0.959796</td>\n",
       "      <td>02:50</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>16</td>\n",
       "      <td>0.428987</td>\n",
       "      <td>0.127131</td>\n",
       "      <td>0.038400</td>\n",
       "      <td>0.961703</td>\n",
       "      <td>02:50</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>17</td>\n",
       "      <td>0.435088</td>\n",
       "      <td>0.139473</td>\n",
       "      <td>0.036800</td>\n",
       "      <td>0.963128</td>\n",
       "      <td>02:50</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>18</td>\n",
       "      <td>0.433165</td>\n",
       "      <td>0.143824</td>\n",
       "      <td>0.040800</td>\n",
       "      <td>0.959061</td>\n",
       "      <td>02:50</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>19</td>\n",
       "      <td>0.424090</td>\n",
       "      <td>0.143478</td>\n",
       "      <td>0.040000</td>\n",
       "      <td>0.959855</td>\n",
       "      <td>02:50</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>20</td>\n",
       "      <td>0.428434</td>\n",
       "      <td>0.133394</td>\n",
       "      <td>0.035200</td>\n",
       "      <td>0.964839</td>\n",
       "      <td>02:50</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>21</td>\n",
       "      <td>0.423406</td>\n",
       "      <td>0.133019</td>\n",
       "      <td>0.035200</td>\n",
       "      <td>0.964846</td>\n",
       "      <td>02:51</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>22</td>\n",
       "      <td>0.423063</td>\n",
       "      <td>0.164122</td>\n",
       "      <td>0.047200</td>\n",
       "      <td>0.952760</td>\n",
       "      <td>02:50</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>23</td>\n",
       "      <td>0.420120</td>\n",
       "      <td>0.317404</td>\n",
       "      <td>0.042400</td>\n",
       "      <td>0.957453</td>\n",
       "      <td>02:51</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>24</td>\n",
       "      <td>0.413691</td>\n",
       "      <td>0.124668</td>\n",
       "      <td>0.028800</td>\n",
       "      <td>0.971176</td>\n",
       "      <td>02:51</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>25</td>\n",
       "      <td>0.406619</td>\n",
       "      <td>0.113434</td>\n",
       "      <td>0.025600</td>\n",
       "      <td>0.974423</td>\n",
       "      <td>02:51</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>26</td>\n",
       "      <td>0.410905</td>\n",
       "      <td>0.118962</td>\n",
       "      <td>0.029600</td>\n",
       "      <td>0.970335</td>\n",
       "      <td>02:51</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>27</td>\n",
       "      <td>0.400119</td>\n",
       "      <td>0.119662</td>\n",
       "      <td>0.033600</td>\n",
       "      <td>0.966371</td>\n",
       "      <td>02:51</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>28</td>\n",
       "      <td>0.400780</td>\n",
       "      <td>0.117155</td>\n",
       "      <td>0.029600</td>\n",
       "      <td>0.970492</td>\n",
       "      <td>02:51</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>29</td>\n",
       "      <td>0.401831</td>\n",
       "      <td>0.112211</td>\n",
       "      <td>0.030400</td>\n",
       "      <td>0.969544</td>\n",
       "      <td>02:50</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>30</td>\n",
       "      <td>0.394482</td>\n",
       "      <td>0.116060</td>\n",
       "      <td>0.032800</td>\n",
       "      <td>0.967197</td>\n",
       "      <td>02:51</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>31</td>\n",
       "      <td>0.385463</td>\n",
       "      <td>0.106931</td>\n",
       "      <td>0.029600</td>\n",
       "      <td>0.970425</td>\n",
       "      <td>02:50</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>32</td>\n",
       "      <td>0.378473</td>\n",
       "      <td>0.106054</td>\n",
       "      <td>0.028000</td>\n",
       "      <td>0.972054</td>\n",
       "      <td>02:45</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>33</td>\n",
       "      <td>0.369355</td>\n",
       "      <td>0.109674</td>\n",
       "      <td>0.030400</td>\n",
       "      <td>0.969582</td>\n",
       "      <td>02:45</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>34</td>\n",
       "      <td>0.386076</td>\n",
       "      <td>0.127697</td>\n",
       "      <td>0.035200</td>\n",
       "      <td>0.964763</td>\n",
       "      <td>02:45</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>35</td>\n",
       "      <td>0.372976</td>\n",
       "      <td>0.128510</td>\n",
       "      <td>0.032000</td>\n",
       "      <td>0.967887</td>\n",
       "      <td>02:45</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>36</td>\n",
       "      <td>0.368849</td>\n",
       "      <td>0.109945</td>\n",
       "      <td>0.031200</td>\n",
       "      <td>0.968933</td>\n",
       "      <td>02:45</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>37</td>\n",
       "      <td>0.384522</td>\n",
       "      <td>0.114758</td>\n",
       "      <td>0.027200</td>\n",
       "      <td>0.972838</td>\n",
       "      <td>02:45</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>38</td>\n",
       "      <td>0.366653</td>\n",
       "      <td>0.119324</td>\n",
       "      <td>0.032000</td>\n",
       "      <td>0.967852</td>\n",
       "      <td>02:45</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>39</td>\n",
       "      <td>0.365298</td>\n",
       "      <td>0.095979</td>\n",
       "      <td>0.026400</td>\n",
       "      <td>0.973581</td>\n",
       "      <td>02:45</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>40</td>\n",
       "      <td>0.358774</td>\n",
       "      <td>0.114083</td>\n",
       "      <td>0.029600</td>\n",
       "      <td>0.970394</td>\n",
       "      <td>02:45</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>41</td>\n",
       "      <td>0.369723</td>\n",
       "      <td>0.119185</td>\n",
       "      <td>0.030400</td>\n",
       "      <td>0.969504</td>\n",
       "      <td>02:45</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>42</td>\n",
       "      <td>0.362511</td>\n",
       "      <td>0.113481</td>\n",
       "      <td>0.029600</td>\n",
       "      <td>0.970328</td>\n",
       "      <td>02:46</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>43</td>\n",
       "      <td>0.372016</td>\n",
       "      <td>0.102860</td>\n",
       "      <td>0.025600</td>\n",
       "      <td>0.974410</td>\n",
       "      <td>02:45</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>44</td>\n",
       "      <td>0.351181</td>\n",
       "      <td>0.117482</td>\n",
       "      <td>0.027200</td>\n",
       "      <td>0.972689</td>\n",
       "      <td>02:45</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>45</td>\n",
       "      <td>0.354880</td>\n",
       "      <td>0.100524</td>\n",
       "      <td>0.024000</td>\n",
       "      <td>0.975993</td>\n",
       "      <td>02:45</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>46</td>\n",
       "      <td>0.354320</td>\n",
       "      <td>0.113002</td>\n",
       "      <td>0.027200</td>\n",
       "      <td>0.972714</td>\n",
       "      <td>02:45</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>47</td>\n",
       "      <td>0.349117</td>\n",
       "      <td>0.094576</td>\n",
       "      <td>0.024000</td>\n",
       "      <td>0.975979</td>\n",
       "      <td>02:45</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>48</td>\n",
       "      <td>0.346822</td>\n",
       "      <td>0.104215</td>\n",
       "      <td>0.025600</td>\n",
       "      <td>0.974354</td>\n",
       "      <td>02:45</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>49</td>\n",
       "      <td>0.342140</td>\n",
       "      <td>0.102611</td>\n",
       "      <td>0.024800</td>\n",
       "      <td>0.975148</td>\n",
       "      <td>02:45</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>50</td>\n",
       "      <td>0.351306</td>\n",
       "      <td>0.100883</td>\n",
       "      <td>0.024000</td>\n",
       "      <td>0.975956</td>\n",
       "      <td>02:46</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>51</td>\n",
       "      <td>0.345050</td>\n",
       "      <td>0.115027</td>\n",
       "      <td>0.029600</td>\n",
       "      <td>0.970316</td>\n",
       "      <td>02:46</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>52</td>\n",
       "      <td>0.346346</td>\n",
       "      <td>0.100431</td>\n",
       "      <td>0.025600</td>\n",
       "      <td>0.974349</td>\n",
       "      <td>02:46</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>53</td>\n",
       "      <td>0.340354</td>\n",
       "      <td>0.098514</td>\n",
       "      <td>0.024800</td>\n",
       "      <td>0.975182</td>\n",
       "      <td>02:51</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>54</td>\n",
       "      <td>0.343836</td>\n",
       "      <td>0.098642</td>\n",
       "      <td>0.021600</td>\n",
       "      <td>0.978397</td>\n",
       "      <td>02:51</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>55</td>\n",
       "      <td>0.344005</td>\n",
       "      <td>0.099892</td>\n",
       "      <td>0.026400</td>\n",
       "      <td>0.973576</td>\n",
       "      <td>02:51</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>56</td>\n",
       "      <td>0.339209</td>\n",
       "      <td>0.104819</td>\n",
       "      <td>0.023200</td>\n",
       "      <td>0.976747</td>\n",
       "      <td>02:51</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>57</td>\n",
       "      <td>0.339428</td>\n",
       "      <td>0.097120</td>\n",
       "      <td>0.023200</td>\n",
       "      <td>0.976782</td>\n",
       "      <td>02:51</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>58</td>\n",
       "      <td>0.337506</td>\n",
       "      <td>0.090555</td>\n",
       "      <td>0.022400</td>\n",
       "      <td>0.977601</td>\n",
       "      <td>02:51</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>59</td>\n",
       "      <td>0.335864</td>\n",
       "      <td>0.094058</td>\n",
       "      <td>0.023200</td>\n",
       "      <td>0.976804</td>\n",
       "      <td>02:51</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>60</td>\n",
       "      <td>0.343175</td>\n",
       "      <td>0.094942</td>\n",
       "      <td>0.023200</td>\n",
       "      <td>0.976766</td>\n",
       "      <td>02:51</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>61</td>\n",
       "      <td>0.346504</td>\n",
       "      <td>0.099598</td>\n",
       "      <td>0.023200</td>\n",
       "      <td>0.976784</td>\n",
       "      <td>02:51</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>62</td>\n",
       "      <td>0.337642</td>\n",
       "      <td>0.093801</td>\n",
       "      <td>0.024000</td>\n",
       "      <td>0.975986</td>\n",
       "      <td>02:51</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>63</td>\n",
       "      <td>0.336443</td>\n",
       "      <td>0.098243</td>\n",
       "      <td>0.023200</td>\n",
       "      <td>0.976771</td>\n",
       "      <td>02:51</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>"
      ],
      "text/plain": [
       "<IPython.core.display.HTML object>"
      ]
     },
     "metadata": {
      "tags": []
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "lr = [2e-5, 2e-4, 2e-3]\n",
    "learn.fit_one_cycle(64, lr)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 0,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "uf0t1CZ1Jk_j"
   },
   "outputs": [],
   "source": [
    "learn.save('Se_ResNext50_mixup_aug_s2_124t256t299_976771')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 0,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "Lx6Xw9LYmEP1"
   },
   "outputs": [],
   "source": [
    "preds, _ = learn.get_preds(DatasetType.Test)\n",
    "pred_class = torch.argmax(preds, dim=1)\n",
    "pred_class += 1\n",
    "pred_numpy = pred_class.numpy()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 0,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "R4iZsnaot3SH"
   },
   "outputs": [],
   "source": [
    "write_submission(pred_numpy, 'Se_ResNet50_mix_aug_976771.csv')\n",
    "#LB 0.980316980670933"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 0,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "QKOV8r0M0IAA"
   },
   "outputs": [],
   "source": [
    "predsTTA, _ = learn.TTA(ds_type=DatasetType.Test)\n",
    "pred_class = torch.argmax(predsTTA, dim=1)\n",
    "pred_class += 1\n",
    "pred_numpy = pred_class.numpy()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 0,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "WrLu_XL13TD4"
   },
   "outputs": [],
   "source": [
    "write_submission(pred_numpy, 'Se_ResNet50_mix_aug_TTA_976771.csv')\n",
    "#LB 0.979364735433383"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 0,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "uYEslHWm4_Db"
   },
   "outputs": [],
   "source": [
    "class_name = ['Cargo', 'Military','Carrier' ,'Cruise','Tankers']\n",
    "\n",
    "predTTA_df = pd.DataFrame(predsTTA.numpy(), columns=class_name)\n",
    "preds_df = pd.DataFrame(preds.numpy(), columns=class_name)\n",
    "preds_df.to_csv('Preds_SeResNeXt50_LB_980316.csv', index=False)\n",
    "predTTA_df.to_csv('TTA_SeResNeXt50_LB_980316.csv', index=False)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 0,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "7ns8Kzrj4_9H"
   },
   "outputs": [],
   "source": []
  }
 ],
 "metadata": {
  "accelerator": "GPU",
  "colab": {
   "collapsed_sections": [
    "qIU8Ve_iyTue",
    "Y8ejhqi4yZRs"
   ],
   "name": "Final_AV_GODL_SeResNeXt50.ipynb",
   "provenance": [],
   "version": "0.3.2"
  },
  "kernelspec": {
   "display_name": "Python 3",
   "language": "python",
   "name": "python3"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.6.7"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 1
}