Final_ResNeXt50.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 0,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "rmqrU-a0lB4D"
   },
   "outputs": [],
   "source": [
    "import torchvision\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 resnext50_32x4d(pretrained=False):\n",
    "    pretrained = 'imagenet' if pretrained else None\n",
    "    model = torchvision.models.resnext50_32x4d(pretrained=pretrained)\n",
    "    return nn.Sequential(*list(model.children()))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 0,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "DlRsCZelQHx0"
   },
   "outputs": [],
   "source": [
    "f_score = FBeta(average='weighted', beta=1)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 0,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "xM8LeMJwOHYB"
   },
   "outputs": [],
   "source": [
    "tfms = get_transforms(do_flip=True, flip_vert=False, max_lighting=0.1, max_zoom=1.05,\n",
    "                      max_warp=0.,\n",
    "                      xtra_tfms=[rand_crop(), rand_zoom(1, 1.5),\n",
    "                                 symmetric_warp(magnitude=(-0.2, 0.2))])"
   ]
  },
  {
   "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": "fJq7ya2cq6dG"
   },
   "outputs": [],
   "source": [
    "def get_data(SZ, BS):\n",
    "    data = (src.transform(tfms, size=SZ, resize_method=ResizeMethod.SQUISH)\n",
    "        .databunch(bs=BS).normalize(imagenet_stats))\n",
    "    return data\n",
    "\n",
    "def get_learner(data):\n",
    "    learn = cnn_learner(data, resnext50_32x4d, pretrained=True, cut=-2,\n",
    "                        split_on=lambda m: (m[0][3], m[1]), \n",
    "                        metrics=[error_rate, f_score])\n",
    "    return learn"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 0,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "ROnM_e-1rPCe"
   },
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "colab_type": "text",
    "id": "Lj_Ri5wJqovb"
   },
   "source": [
    "## Training with Size 128"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 0,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "PvG8qhwQOAJ0"
   },
   "outputs": [],
   "source": [
    "data = get_data(128, 96)\n",
    "learn = get_learner(data)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 0,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "TKmLRhHOJlGZ"
   },
   "outputs": [],
   "source": [
    "learn.mixup();"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 0,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 317
    },
    "colab_type": "code",
    "id": "LeBqZJ4RQScw",
    "outputId": "69c07c7e-63cb-4dea-cc36-201e36b6d89c"
   },
   "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: 1.74E-03\n"
     ]
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAYUAAAEKCAYAAAD9xUlFAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nQmCC\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": 607
    },
    "colab_type": "code",
    "id": "RGroV0abQXiC",
    "outputId": "3898e88d-e7c6-471d-a8bd-add5138f229e"
   },
   "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>1.399163</td>\n",
       "      <td>0.785749</td>\n",
       "      <td>0.292000</td>\n",
       "      <td>0.700219</td>\n",
       "      <td>00:25</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>1</td>\n",
       "      <td>1.167915</td>\n",
       "      <td>0.641416</td>\n",
       "      <td>0.232800</td>\n",
       "      <td>0.762691</td>\n",
       "      <td>00:24</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>2</td>\n",
       "      <td>1.017506</td>\n",
       "      <td>0.560105</td>\n",
       "      <td>0.220800</td>\n",
       "      <td>0.772716</td>\n",
       "      <td>00:23</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>3</td>\n",
       "      <td>0.922721</td>\n",
       "      <td>0.471763</td>\n",
       "      <td>0.178400</td>\n",
       "      <td>0.816794</td>\n",
       "      <td>00:24</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>4</td>\n",
       "      <td>0.854507</td>\n",
       "      <td>0.420188</td>\n",
       "      <td>0.153600</td>\n",
       "      <td>0.847041</td>\n",
       "      <td>00:23</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>5</td>\n",
       "      <td>0.804291</td>\n",
       "      <td>0.395933</td>\n",
       "      <td>0.141600</td>\n",
       "      <td>0.856717</td>\n",
       "      <td>00:23</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>6</td>\n",
       "      <td>0.763349</td>\n",
       "      <td>0.421699</td>\n",
       "      <td>0.166400</td>\n",
       "      <td>0.824325</td>\n",
       "      <td>00:24</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>7</td>\n",
       "      <td>0.730965</td>\n",
       "      <td>0.316812</td>\n",
       "      <td>0.110400</td>\n",
       "      <td>0.890052</td>\n",
       "      <td>00:23</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>8</td>\n",
       "      <td>0.686449</td>\n",
       "      <td>0.296907</td>\n",
       "      <td>0.090400</td>\n",
       "      <td>0.909674</td>\n",
       "      <td>00:24</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>9</td>\n",
       "      <td>0.663478</td>\n",
       "      <td>0.308358</td>\n",
       "      <td>0.107200</td>\n",
       "      <td>0.890812</td>\n",
       "      <td>00:25</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>10</td>\n",
       "      <td>0.643941</td>\n",
       "      <td>0.298951</td>\n",
       "      <td>0.099200</td>\n",
       "      <td>0.900661</td>\n",
       "      <td>00:24</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>11</td>\n",
       "      <td>0.628412</td>\n",
       "      <td>0.288944</td>\n",
       "      <td>0.090400</td>\n",
       "      <td>0.909431</td>\n",
       "      <td>00:23</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>12</td>\n",
       "      <td>0.618218</td>\n",
       "      <td>0.277329</td>\n",
       "      <td>0.077600</td>\n",
       "      <td>0.921681</td>\n",
       "      <td>00:23</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>13</td>\n",
       "      <td>0.609663</td>\n",
       "      <td>0.272976</td>\n",
       "      <td>0.079200</td>\n",
       "      <td>0.920462</td>\n",
       "      <td>00:24</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>14</td>\n",
       "      <td>0.590849</td>\n",
       "      <td>0.272603</td>\n",
       "      <td>0.079200</td>\n",
       "      <td>0.920557</td>\n",
       "      <td>00:23</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>15</td>\n",
       "      <td>0.579766</td>\n",
       "      <td>0.261241</td>\n",
       "      <td>0.073600</td>\n",
       "      <td>0.926205</td>\n",
       "      <td>00:23</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>16</td>\n",
       "      <td>0.572934</td>\n",
       "      <td>0.262500</td>\n",
       "      <td>0.073600</td>\n",
       "      <td>0.926220</td>\n",
       "      <td>00:24</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>17</td>\n",
       "      <td>0.568681</td>\n",
       "      <td>0.264775</td>\n",
       "      <td>0.079200</td>\n",
       "      <td>0.920192</td>\n",
       "      <td>00:24</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(18, max_lr=slice(1e-2))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 0,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "2fw26Q84QkPm"
   },
   "outputs": [],
   "source": [
    "learn.save('ResNext50_aug_mixup_128_920192')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 0,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 317
    },
    "colab_type": "code",
    "id": "zS6dV7FWR5Oi",
    "outputId": "7bcf6840-d3e5-4ec5-b087-c3591ec72661"
   },
   "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: 2.29E-06\n"
     ]
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAYsAAAEKCAYAAADjDHn2AAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nYII=\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "tags": []
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "learn.unfreeze()\n",
    "learn.clip_grad()\n",
    "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('ResNext50_aug_mixup_128_920192');\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": "5b155ace-9923-4e4c-8f0c-0916a559bde2"
   },
   "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>0.187195</td>\n",
       "      <td>0.272717</td>\n",
       "      <td>0.080800</td>\n",
       "      <td>0.918603</td>\n",
       "      <td>00:31</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>1</td>\n",
       "      <td>0.160496</td>\n",
       "      <td>0.240827</td>\n",
       "      <td>0.076000</td>\n",
       "      <td>0.923954</td>\n",
       "      <td>00:31</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>2</td>\n",
       "      <td>0.126432</td>\n",
       "      <td>0.222357</td>\n",
       "      <td>0.072800</td>\n",
       "      <td>0.927303</td>\n",
       "      <td>00:31</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>3</td>\n",
       "      <td>0.102897</td>\n",
       "      <td>0.239679</td>\n",
       "      <td>0.079200</td>\n",
       "      <td>0.920585</td>\n",
       "      <td>00:31</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>4</td>\n",
       "      <td>0.083631</td>\n",
       "      <td>0.262295</td>\n",
       "      <td>0.075200</td>\n",
       "      <td>0.924491</td>\n",
       "      <td>00:30</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>5</td>\n",
       "      <td>0.074236</td>\n",
       "      <td>0.285108</td>\n",
       "      <td>0.081600</td>\n",
       "      <td>0.917949</td>\n",
       "      <td>00:31</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>6</td>\n",
       "      <td>0.080905</td>\n",
       "      <td>0.247683</td>\n",
       "      <td>0.072000</td>\n",
       "      <td>0.928409</td>\n",
       "      <td>00:30</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>7</td>\n",
       "      <td>0.078219</td>\n",
       "      <td>0.293685</td>\n",
       "      <td>0.081600</td>\n",
       "      <td>0.917815</td>\n",
       "      <td>00:31</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>8</td>\n",
       "      <td>0.070945</td>\n",
       "      <td>0.518133</td>\n",
       "      <td>0.134400</td>\n",
       "      <td>0.859231</td>\n",
       "      <td>00:31</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>9</td>\n",
       "      <td>0.074743</td>\n",
       "      <td>0.326079</td>\n",
       "      <td>0.088000</td>\n",
       "      <td>0.912780</td>\n",
       "      <td>00:31</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>10</td>\n",
       "      <td>0.068346</td>\n",
       "      <td>0.266369</td>\n",
       "      <td>0.075200</td>\n",
       "      <td>0.924624</td>\n",
       "      <td>00:31</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>11</td>\n",
       "      <td>0.076725</td>\n",
       "      <td>0.215331</td>\n",
       "      <td>0.073600</td>\n",
       "      <td>0.926430</td>\n",
       "      <td>00:30</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>12</td>\n",
       "      <td>0.075231</td>\n",
       "      <td>0.239762</td>\n",
       "      <td>0.068000</td>\n",
       "      <td>0.931557</td>\n",
       "      <td>00:31</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>13</td>\n",
       "      <td>0.062544</td>\n",
       "      <td>0.276430</td>\n",
       "      <td>0.073600</td>\n",
       "      <td>0.926569</td>\n",
       "      <td>00:31</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>14</td>\n",
       "      <td>0.052009</td>\n",
       "      <td>0.259697</td>\n",
       "      <td>0.076800</td>\n",
       "      <td>0.923451</td>\n",
       "      <td>00:30</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>15</td>\n",
       "      <td>0.047416</td>\n",
       "      <td>0.259135</td>\n",
       "      <td>0.057600</td>\n",
       "      <td>0.942307</td>\n",
       "      <td>00:31</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>16</td>\n",
       "      <td>0.043379</td>\n",
       "      <td>0.290163</td>\n",
       "      <td>0.074400</td>\n",
       "      <td>0.925296</td>\n",
       "      <td>00:31</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>17</td>\n",
       "      <td>0.041404</td>\n",
       "      <td>0.257631</td>\n",
       "      <td>0.063200</td>\n",
       "      <td>0.936685</td>\n",
       "      <td>00:30</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>18</td>\n",
       "      <td>0.034257</td>\n",
       "      <td>0.304815</td>\n",
       "      <td>0.064800</td>\n",
       "      <td>0.935163</td>\n",
       "      <td>00:31</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>19</td>\n",
       "      <td>0.033868</td>\n",
       "      <td>0.313341</td>\n",
       "      <td>0.070400</td>\n",
       "      <td>0.929258</td>\n",
       "      <td>00:31</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>20</td>\n",
       "      <td>0.027307</td>\n",
       "      <td>0.327667</td>\n",
       "      <td>0.067200</td>\n",
       "      <td>0.932929</td>\n",
       "      <td>00:31</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>21</td>\n",
       "      <td>0.025308</td>\n",
       "      <td>0.355812</td>\n",
       "      <td>0.064800</td>\n",
       "      <td>0.934552</td>\n",
       "      <td>00:31</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>22</td>\n",
       "      <td>0.020631</td>\n",
       "      <td>0.303493</td>\n",
       "      <td>0.056800</td>\n",
       "      <td>0.943054</td>\n",
       "      <td>00:31</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>23</td>\n",
       "      <td>0.014639</td>\n",
       "      <td>0.370296</td>\n",
       "      <td>0.065600</td>\n",
       "      <td>0.934080</td>\n",
       "      <td>00:31</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>24</td>\n",
       "      <td>0.014692</td>\n",
       "      <td>0.342290</td>\n",
       "      <td>0.054400</td>\n",
       "      <td>0.945641</td>\n",
       "      <td>00:31</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>25</td>\n",
       "      <td>0.012471</td>\n",
       "      <td>0.389955</td>\n",
       "      <td>0.063200</td>\n",
       "      <td>0.936755</td>\n",
       "      <td>00:31</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>26</td>\n",
       "      <td>0.011053</td>\n",
       "      <td>0.369448</td>\n",
       "      <td>0.060000</td>\n",
       "      <td>0.940147</td>\n",
       "      <td>00:31</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>27</td>\n",
       "      <td>0.009862</td>\n",
       "      <td>0.359669</td>\n",
       "      <td>0.056000</td>\n",
       "      <td>0.943735</td>\n",
       "      <td>00:30</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>28</td>\n",
       "      <td>0.006880</td>\n",
       "      <td>0.366227</td>\n",
       "      <td>0.056000</td>\n",
       "      <td>0.943711</td>\n",
       "      <td>00:31</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>29</td>\n",
       "      <td>0.005660</td>\n",
       "      <td>0.342333</td>\n",
       "      <td>0.056000</td>\n",
       "      <td>0.943823</td>\n",
       "      <td>00:31</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>30</td>\n",
       "      <td>0.004217</td>\n",
       "      <td>0.351059</td>\n",
       "      <td>0.057600</td>\n",
       "      <td>0.942290</td>\n",
       "      <td>00:30</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>31</td>\n",
       "      <td>0.004849</td>\n",
       "      <td>0.373147</td>\n",
       "      <td>0.053600</td>\n",
       "      <td>0.946176</td>\n",
       "      <td>00:31</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>32</td>\n",
       "      <td>0.005849</td>\n",
       "      <td>0.350160</td>\n",
       "      <td>0.053600</td>\n",
       "      <td>0.946086</td>\n",
       "      <td>00:30</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>33</td>\n",
       "      <td>0.004244</td>\n",
       "      <td>0.346460</td>\n",
       "      <td>0.053600</td>\n",
       "      <td>0.946338</td>\n",
       "      <td>00:31</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>34</td>\n",
       "      <td>0.002679</td>\n",
       "      <td>0.351085</td>\n",
       "      <td>0.055200</td>\n",
       "      <td>0.944766</td>\n",
       "      <td>00:30</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>35</td>\n",
       "      <td>0.001397</td>\n",
       "      <td>0.361873</td>\n",
       "      <td>0.055200</td>\n",
       "      <td>0.944603</td>\n",
       "      <td>00:30</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)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 0,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "ts518fO-SUBg"
   },
   "outputs": [],
   "source": [
    "learn.save('s2_ResNext50_aug_128_944603')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "colab_type": "text",
    "id": "LE_8m1zvqurl"
   },
   "source": [
    "## Training with Size 256"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 0,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "q6WIPZG_WPck"
   },
   "outputs": [],
   "source": [
    "data = get_data(256, 72)\n",
    "learn = get_learner(data)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 0,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "OvjZVan-PiZD"
   },
   "outputs": [],
   "source": [
    "learn.load('s2_ResNext50_aug_128_944603');"
   ]
  },
  {
   "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": "9d0be30a-935c-4530-e893-12c45539ee0d"
   },
   "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": "iVBORw0KGgoAAAANSUhEUgAAAYsAAAEKCAYAAADjDHn2AAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\noZRSyqX/DzOewswE2VqgAAAAAElFTkSuQmCC\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": "d73220ef-74b5-4c32-a2ca-ebc852e48e1d"
   },
   "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>0.173847</td>\n",
       "      <td>0.342158</td>\n",
       "      <td>0.080800</td>\n",
       "      <td>0.918924</td>\n",
       "      <td>01:13</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>1</td>\n",
       "      <td>0.110598</td>\n",
       "      <td>0.178863</td>\n",
       "      <td>0.075200</td>\n",
       "      <td>0.924773</td>\n",
       "      <td>01:12</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>2</td>\n",
       "      <td>0.064229</td>\n",
       "      <td>0.108543</td>\n",
       "      <td>0.071200</td>\n",
       "      <td>0.928708</td>\n",
       "      <td>01:11</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>3</td>\n",
       "      <td>0.042332</td>\n",
       "      <td>0.101559</td>\n",
       "      <td>0.062400</td>\n",
       "      <td>0.937671</td>\n",
       "      <td>01:11</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>4</td>\n",
       "      <td>0.029126</td>\n",
       "      <td>0.092629</td>\n",
       "      <td>0.060000</td>\n",
       "      <td>0.940013</td>\n",
       "      <td>01:11</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>5</td>\n",
       "      <td>0.021328</td>\n",
       "      <td>0.112552</td>\n",
       "      <td>0.052000</td>\n",
       "      <td>0.948056</td>\n",
       "      <td>01:12</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>6</td>\n",
       "      <td>0.017713</td>\n",
       "      <td>0.088821</td>\n",
       "      <td>0.051200</td>\n",
       "      <td>0.948767</td>\n",
       "      <td>01:11</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>7</td>\n",
       "      <td>0.011850</td>\n",
       "      <td>0.091282</td>\n",
       "      <td>0.049600</td>\n",
       "      <td>0.950351</td>\n",
       "      <td>01:11</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>8</td>\n",
       "      <td>0.009709</td>\n",
       "      <td>0.110274</td>\n",
       "      <td>0.048000</td>\n",
       "      <td>0.951810</td>\n",
       "      <td>01:11</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>9</td>\n",
       "      <td>0.008700</td>\n",
       "      <td>0.110278</td>\n",
       "      <td>0.049600</td>\n",
       "      <td>0.950547</td>\n",
       "      <td>01:11</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>10</td>\n",
       "      <td>0.007956</td>\n",
       "      <td>0.117790</td>\n",
       "      <td>0.046400</td>\n",
       "      <td>0.953325</td>\n",
       "      <td>01:12</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>11</td>\n",
       "      <td>0.008236</td>\n",
       "      <td>0.133663</td>\n",
       "      <td>0.052000</td>\n",
       "      <td>0.947829</td>\n",
       "      <td>01:11</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>12</td>\n",
       "      <td>0.005296</td>\n",
       "      <td>0.127308</td>\n",
       "      <td>0.047200</td>\n",
       "      <td>0.952470</td>\n",
       "      <td>01:12</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>13</td>\n",
       "      <td>0.003647</td>\n",
       "      <td>0.118649</td>\n",
       "      <td>0.041600</td>\n",
       "      <td>0.958389</td>\n",
       "      <td>01:11</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>14</td>\n",
       "      <td>0.004742</td>\n",
       "      <td>0.122503</td>\n",
       "      <td>0.041600</td>\n",
       "      <td>0.958341</td>\n",
       "      <td>01:12</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>15</td>\n",
       "      <td>0.003546</td>\n",
       "      <td>0.131335</td>\n",
       "      <td>0.043200</td>\n",
       "      <td>0.956443</td>\n",
       "      <td>01:11</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>16</td>\n",
       "      <td>0.002539</td>\n",
       "      <td>0.134969</td>\n",
       "      <td>0.040800</td>\n",
       "      <td>0.958919</td>\n",
       "      <td>01:11</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>17</td>\n",
       "      <td>0.002232</td>\n",
       "      <td>0.149150</td>\n",
       "      <td>0.042400</td>\n",
       "      <td>0.957266</td>\n",
       "      <td>01:11</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>18</td>\n",
       "      <td>0.002304</td>\n",
       "      <td>0.147315</td>\n",
       "      <td>0.036000</td>\n",
       "      <td>0.963803</td>\n",
       "      <td>01:11</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>19</td>\n",
       "      <td>0.003589</td>\n",
       "      <td>0.147437</td>\n",
       "      <td>0.039200</td>\n",
       "      <td>0.960593</td>\n",
       "      <td>01:11</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>20</td>\n",
       "      <td>0.003119</td>\n",
       "      <td>0.140957</td>\n",
       "      <td>0.036800</td>\n",
       "      <td>0.962997</td>\n",
       "      <td>01:11</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>21</td>\n",
       "      <td>0.001843</td>\n",
       "      <td>0.142632</td>\n",
       "      <td>0.038400</td>\n",
       "      <td>0.961477</td>\n",
       "      <td>01:11</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>22</td>\n",
       "      <td>0.003523</td>\n",
       "      <td>0.140663</td>\n",
       "      <td>0.035200</td>\n",
       "      <td>0.964581</td>\n",
       "      <td>01:11</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>23</td>\n",
       "      <td>0.002430</td>\n",
       "      <td>0.141562</td>\n",
       "      <td>0.037600</td>\n",
       "      <td>0.962240</td>\n",
       "      <td>01:12</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))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 0,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "9I0No7aOX7CY"
   },
   "outputs": [],
   "source": [
    "learn.save('ResNxt50_FL_128t256_962240')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 0,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "X7GobIWWf6cw"
   },
   "outputs": [],
   "source": [
    "learn.load('ResNxt50_FL_128t256_962240');"
   ]
  },
  {
   "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": "fe8c4263-2b2e-4f47-fc8d-1aeaeedf8fc5"
   },
   "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>0.006308</td>\n",
       "      <td>0.142461</td>\n",
       "      <td>0.036800</td>\n",
       "      <td>0.962972</td>\n",
       "      <td>01:43</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>1</td>\n",
       "      <td>0.004615</td>\n",
       "      <td>0.141906</td>\n",
       "      <td>0.037600</td>\n",
       "      <td>0.962304</td>\n",
       "      <td>01:42</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>2</td>\n",
       "      <td>0.003115</td>\n",
       "      <td>0.137392</td>\n",
       "      <td>0.041600</td>\n",
       "      <td>0.958198</td>\n",
       "      <td>01:42</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>3</td>\n",
       "      <td>0.003897</td>\n",
       "      <td>0.137139</td>\n",
       "      <td>0.048800</td>\n",
       "      <td>0.951126</td>\n",
       "      <td>01:41</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>4</td>\n",
       "      <td>0.003384</td>\n",
       "      <td>0.148498</td>\n",
       "      <td>0.050400</td>\n",
       "      <td>0.949666</td>\n",
       "      <td>01:42</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>5</td>\n",
       "      <td>0.005978</td>\n",
       "      <td>0.172913</td>\n",
       "      <td>0.048800</td>\n",
       "      <td>0.951256</td>\n",
       "      <td>01:42</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>6</td>\n",
       "      <td>0.003932</td>\n",
       "      <td>0.174693</td>\n",
       "      <td>0.048800</td>\n",
       "      <td>0.950974</td>\n",
       "      <td>01:41</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>7</td>\n",
       "      <td>0.009130</td>\n",
       "      <td>0.192829</td>\n",
       "      <td>0.056000</td>\n",
       "      <td>0.943899</td>\n",
       "      <td>01:42</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>8</td>\n",
       "      <td>0.012179</td>\n",
       "      <td>0.182084</td>\n",
       "      <td>0.054400</td>\n",
       "      <td>0.945714</td>\n",
       "      <td>01:41</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>9</td>\n",
       "      <td>0.010758</td>\n",
       "      <td>0.145602</td>\n",
       "      <td>0.047200</td>\n",
       "      <td>0.952878</td>\n",
       "      <td>01:41</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>10</td>\n",
       "      <td>0.007547</td>\n",
       "      <td>0.163628</td>\n",
       "      <td>0.045600</td>\n",
       "      <td>0.954257</td>\n",
       "      <td>01:42</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>11</td>\n",
       "      <td>0.008661</td>\n",
       "      <td>0.159896</td>\n",
       "      <td>0.049600</td>\n",
       "      <td>0.950404</td>\n",
       "      <td>01:41</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>12</td>\n",
       "      <td>0.009808</td>\n",
       "      <td>0.123901</td>\n",
       "      <td>0.040800</td>\n",
       "      <td>0.959166</td>\n",
       "      <td>01:42</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>13</td>\n",
       "      <td>0.005991</td>\n",
       "      <td>0.136248</td>\n",
       "      <td>0.043200</td>\n",
       "      <td>0.956693</td>\n",
       "      <td>01:42</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>14</td>\n",
       "      <td>0.004928</td>\n",
       "      <td>0.148665</td>\n",
       "      <td>0.046400</td>\n",
       "      <td>0.953381</td>\n",
       "      <td>01:42</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>15</td>\n",
       "      <td>0.003714</td>\n",
       "      <td>0.140840</td>\n",
       "      <td>0.043200</td>\n",
       "      <td>0.956577</td>\n",
       "      <td>01:41</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>16</td>\n",
       "      <td>0.004879</td>\n",
       "      <td>0.152380</td>\n",
       "      <td>0.047200</td>\n",
       "      <td>0.952676</td>\n",
       "      <td>01:41</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>17</td>\n",
       "      <td>0.002317</td>\n",
       "      <td>0.134548</td>\n",
       "      <td>0.040800</td>\n",
       "      <td>0.959195</td>\n",
       "      <td>01:42</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>18</td>\n",
       "      <td>0.002211</td>\n",
       "      <td>0.143304</td>\n",
       "      <td>0.044000</td>\n",
       "      <td>0.955978</td>\n",
       "      <td>01:42</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>19</td>\n",
       "      <td>0.001997</td>\n",
       "      <td>0.148867</td>\n",
       "      <td>0.046400</td>\n",
       "      <td>0.953507</td>\n",
       "      <td>01:42</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>20</td>\n",
       "      <td>0.001433</td>\n",
       "      <td>0.129056</td>\n",
       "      <td>0.040800</td>\n",
       "      <td>0.959236</td>\n",
       "      <td>01:42</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>21</td>\n",
       "      <td>0.000761</td>\n",
       "      <td>0.128278</td>\n",
       "      <td>0.036000</td>\n",
       "      <td>0.963919</td>\n",
       "      <td>01:42</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>22</td>\n",
       "      <td>0.000923</td>\n",
       "      <td>0.124794</td>\n",
       "      <td>0.039200</td>\n",
       "      <td>0.960715</td>\n",
       "      <td>01:42</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>23</td>\n",
       "      <td>0.002148</td>\n",
       "      <td>0.126675</td>\n",
       "      <td>0.040000</td>\n",
       "      <td>0.959953</td>\n",
       "      <td>01:41</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>24</td>\n",
       "      <td>0.000727</td>\n",
       "      <td>0.140042</td>\n",
       "      <td>0.041600</td>\n",
       "      <td>0.958345</td>\n",
       "      <td>01:42</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>25</td>\n",
       "      <td>0.000214</td>\n",
       "      <td>0.133349</td>\n",
       "      <td>0.035200</td>\n",
       "      <td>0.964774</td>\n",
       "      <td>01:42</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>26</td>\n",
       "      <td>0.000388</td>\n",
       "      <td>0.129878</td>\n",
       "      <td>0.036000</td>\n",
       "      <td>0.963953</td>\n",
       "      <td>01:42</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>27</td>\n",
       "      <td>0.000113</td>\n",
       "      <td>0.129834</td>\n",
       "      <td>0.035200</td>\n",
       "      <td>0.964747</td>\n",
       "      <td>01:42</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>28</td>\n",
       "      <td>0.000070</td>\n",
       "      <td>0.137495</td>\n",
       "      <td>0.040000</td>\n",
       "      <td>0.959924</td>\n",
       "      <td>01:42</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>29</td>\n",
       "      <td>0.000060</td>\n",
       "      <td>0.134870</td>\n",
       "      <td>0.036800</td>\n",
       "      <td>0.963179</td>\n",
       "      <td>01:41</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>30</td>\n",
       "      <td>0.000019</td>\n",
       "      <td>0.133867</td>\n",
       "      <td>0.038400</td>\n",
       "      <td>0.961521</td>\n",
       "      <td>01:41</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>31</td>\n",
       "      <td>0.000021</td>\n",
       "      <td>0.130933</td>\n",
       "      <td>0.036800</td>\n",
       "      <td>0.963172</td>\n",
       "      <td>01:41</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>"
      ],
      "text/plain": [
       "<IPython.core.display.HTML object>"
      ]
     },
     "metadata": {
      "tags": []
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "lr = [1e-3/100, 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('ResNxt50_FL_s2128t256_963172')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "colab_type": "text",
    "id": "KsW0MyeoqOhS"
   },
   "source": [
    "## Training with Size 299"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 0,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "KlOQZiwxMnMz"
   },
   "outputs": [],
   "source": [
    "data = get_data(299, 52)\n",
    "learn = get_learner(data)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 0,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "1DVIW86w37O6"
   },
   "outputs": [],
   "source": [
    "learn.load('ResNxt50_FL_s2128t256_963172');\n",
    "learn.freeze();\n",
    "learn.clip_grad();"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 0,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "1vWe21uU9alE"
   },
   "outputs": [],
   "source": [
    "learn.mixup();"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 0,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 317
    },
    "colab_type": "code",
    "id": "BpQfO5NiM9B_",
    "outputId": "c396eb8a-8324-48ae-8af2-cd94a789a19d"
   },
   "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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\niIhIBSUFERGpoKQgIiIVlBRERKTC/wP0vdL9/qtnbAAAAABJRU5ErkJggg==\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": "95I-FsnEP9u_",
    "outputId": "7ba5a2c4-68bf-4738-c28f-7174e0e051ae"
   },
   "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>0.964833</td>\n",
       "      <td>0.154743</td>\n",
       "      <td>0.047200</td>\n",
       "      <td>0.952602</td>\n",
       "      <td>01:55</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>1</td>\n",
       "      <td>0.749906</td>\n",
       "      <td>0.144362</td>\n",
       "      <td>0.047200</td>\n",
       "      <td>0.952625</td>\n",
       "      <td>01:52</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>2</td>\n",
       "      <td>0.588627</td>\n",
       "      <td>0.190113</td>\n",
       "      <td>0.047200</td>\n",
       "      <td>0.952539</td>\n",
       "      <td>01:52</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>3</td>\n",
       "      <td>0.540395</td>\n",
       "      <td>0.174962</td>\n",
       "      <td>0.044000</td>\n",
       "      <td>0.955766</td>\n",
       "      <td>01:52</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>4</td>\n",
       "      <td>0.513541</td>\n",
       "      <td>0.170799</td>\n",
       "      <td>0.046400</td>\n",
       "      <td>0.953524</td>\n",
       "      <td>01:52</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>5</td>\n",
       "      <td>0.508403</td>\n",
       "      <td>0.168439</td>\n",
       "      <td>0.041600</td>\n",
       "      <td>0.958261</td>\n",
       "      <td>01:51</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>6</td>\n",
       "      <td>0.504746</td>\n",
       "      <td>0.178048</td>\n",
       "      <td>0.042400</td>\n",
       "      <td>0.957449</td>\n",
       "      <td>01:52</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>7</td>\n",
       "      <td>0.490444</td>\n",
       "      <td>0.160443</td>\n",
       "      <td>0.040000</td>\n",
       "      <td>0.959889</td>\n",
       "      <td>01:51</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>8</td>\n",
       "      <td>0.478344</td>\n",
       "      <td>0.151181</td>\n",
       "      <td>0.039200</td>\n",
       "      <td>0.960652</td>\n",
       "      <td>01:52</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>9</td>\n",
       "      <td>0.479408</td>\n",
       "      <td>0.145875</td>\n",
       "      <td>0.040800</td>\n",
       "      <td>0.959092</td>\n",
       "      <td>01:52</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>10</td>\n",
       "      <td>0.472860</td>\n",
       "      <td>0.135658</td>\n",
       "      <td>0.038400</td>\n",
       "      <td>0.961493</td>\n",
       "      <td>01:52</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>11</td>\n",
       "      <td>0.467986</td>\n",
       "      <td>0.140711</td>\n",
       "      <td>0.039200</td>\n",
       "      <td>0.960696</td>\n",
       "      <td>01:51</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>12</td>\n",
       "      <td>0.468534</td>\n",
       "      <td>0.141636</td>\n",
       "      <td>0.038400</td>\n",
       "      <td>0.961504</td>\n",
       "      <td>01:52</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>13</td>\n",
       "      <td>0.463437</td>\n",
       "      <td>0.134729</td>\n",
       "      <td>0.036800</td>\n",
       "      <td>0.963102</td>\n",
       "      <td>01:52</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>14</td>\n",
       "      <td>0.452585</td>\n",
       "      <td>0.129195</td>\n",
       "      <td>0.035200</td>\n",
       "      <td>0.964787</td>\n",
       "      <td>01:52</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>15</td>\n",
       "      <td>0.444425</td>\n",
       "      <td>0.129341</td>\n",
       "      <td>0.036800</td>\n",
       "      <td>0.963162</td>\n",
       "      <td>01:52</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>16</td>\n",
       "      <td>0.450628</td>\n",
       "      <td>0.128691</td>\n",
       "      <td>0.033600</td>\n",
       "      <td>0.966278</td>\n",
       "      <td>01:52</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>17</td>\n",
       "      <td>0.438364</td>\n",
       "      <td>0.126038</td>\n",
       "      <td>0.032800</td>\n",
       "      <td>0.967083</td>\n",
       "      <td>01:52</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>18</td>\n",
       "      <td>0.439700</td>\n",
       "      <td>0.126483</td>\n",
       "      <td>0.032800</td>\n",
       "      <td>0.967129</td>\n",
       "      <td>01:52</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>19</td>\n",
       "      <td>0.438576</td>\n",
       "      <td>0.124598</td>\n",
       "      <td>0.033600</td>\n",
       "      <td>0.966364</td>\n",
       "      <td>01:51</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>20</td>\n",
       "      <td>0.441750</td>\n",
       "      <td>0.125788</td>\n",
       "      <td>0.035200</td>\n",
       "      <td>0.964728</td>\n",
       "      <td>01:52</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>21</td>\n",
       "      <td>0.443982</td>\n",
       "      <td>0.122780</td>\n",
       "      <td>0.035200</td>\n",
       "      <td>0.964825</td>\n",
       "      <td>01:51</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>22</td>\n",
       "      <td>0.441752</td>\n",
       "      <td>0.124502</td>\n",
       "      <td>0.034400</td>\n",
       "      <td>0.965585</td>\n",
       "      <td>01:51</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>23</td>\n",
       "      <td>0.431366</td>\n",
       "      <td>0.124575</td>\n",
       "      <td>0.035200</td>\n",
       "      <td>0.964764</td>\n",
       "      <td>01:52</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>24</td>\n",
       "      <td>0.421529</td>\n",
       "      <td>0.122658</td>\n",
       "      <td>0.035200</td>\n",
       "      <td>0.964764</td>\n",
       "      <td>01:51</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>25</td>\n",
       "      <td>0.436789</td>\n",
       "      <td>0.119771</td>\n",
       "      <td>0.033600</td>\n",
       "      <td>0.966397</td>\n",
       "      <td>01:51</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>26</td>\n",
       "      <td>0.434649</td>\n",
       "      <td>0.121430</td>\n",
       "      <td>0.036800</td>\n",
       "      <td>0.963188</td>\n",
       "      <td>01:52</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>27</td>\n",
       "      <td>0.435022</td>\n",
       "      <td>0.122448</td>\n",
       "      <td>0.035200</td>\n",
       "      <td>0.964762</td>\n",
       "      <td>01:51</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>28</td>\n",
       "      <td>0.441792</td>\n",
       "      <td>0.121322</td>\n",
       "      <td>0.034400</td>\n",
       "      <td>0.965592</td>\n",
       "      <td>01:52</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>29</td>\n",
       "      <td>0.430479</td>\n",
       "      <td>0.121532</td>\n",
       "      <td>0.035200</td>\n",
       "      <td>0.964738</td>\n",
       "      <td>01:52</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>30</td>\n",
       "      <td>0.422712</td>\n",
       "      <td>0.120361</td>\n",
       "      <td>0.034400</td>\n",
       "      <td>0.965559</td>\n",
       "      <td>01:51</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>31</td>\n",
       "      <td>0.426166</td>\n",
       "      <td>0.126293</td>\n",
       "      <td>0.034400</td>\n",
       "      <td>0.965518</td>\n",
       "      <td>01:52</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(2e-3))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 0,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "aZ53flG79T5t"
   },
   "outputs": [],
   "source": [
    "learn.save('ResNext50_FL_128t224t299_s1_965518')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 0,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "DsarzG2p7i86"
   },
   "outputs": [],
   "source": [
    "learn.load('ResNext50_FL_128t224t299_s1_965518');"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 0,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "0LL2gHYX9T1_"
   },
   "outputs": [],
   "source": [
    "learn.unfreeze()\n",
    "learn.clip_grad();"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 0,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 2033
    },
    "colab_type": "code",
    "id": "wUfJobuglx83",
    "outputId": "761ece40-3e85-4c66-de9a-66a0e2bea968"
   },
   "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>0.429812</td>\n",
       "      <td>0.120410</td>\n",
       "      <td>0.035200</td>\n",
       "      <td>0.964752</td>\n",
       "      <td>02:28</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>1</td>\n",
       "      <td>0.432357</td>\n",
       "      <td>0.116121</td>\n",
       "      <td>0.032000</td>\n",
       "      <td>0.967932</td>\n",
       "      <td>02:36</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>2</td>\n",
       "      <td>0.417631</td>\n",
       "      <td>0.118019</td>\n",
       "      <td>0.036000</td>\n",
       "      <td>0.963939</td>\n",
       "      <td>02:37</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>3</td>\n",
       "      <td>0.425932</td>\n",
       "      <td>0.126109</td>\n",
       "      <td>0.036800</td>\n",
       "      <td>0.963203</td>\n",
       "      <td>02:37</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>4</td>\n",
       "      <td>0.426792</td>\n",
       "      <td>0.125005</td>\n",
       "      <td>0.035200</td>\n",
       "      <td>0.964770</td>\n",
       "      <td>02:37</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>5</td>\n",
       "      <td>0.433026</td>\n",
       "      <td>0.124370</td>\n",
       "      <td>0.036000</td>\n",
       "      <td>0.964046</td>\n",
       "      <td>02:37</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>6</td>\n",
       "      <td>0.433389</td>\n",
       "      <td>0.133266</td>\n",
       "      <td>0.041600</td>\n",
       "      <td>0.958332</td>\n",
       "      <td>02:37</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>7</td>\n",
       "      <td>0.439477</td>\n",
       "      <td>0.156730</td>\n",
       "      <td>0.044000</td>\n",
       "      <td>0.955789</td>\n",
       "      <td>02:37</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>8</td>\n",
       "      <td>0.425781</td>\n",
       "      <td>0.136849</td>\n",
       "      <td>0.040000</td>\n",
       "      <td>0.959962</td>\n",
       "      <td>02:37</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>9</td>\n",
       "      <td>0.432968</td>\n",
       "      <td>0.128504</td>\n",
       "      <td>0.042400</td>\n",
       "      <td>0.957724</td>\n",
       "      <td>02:37</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>10</td>\n",
       "      <td>0.422182</td>\n",
       "      <td>0.126645</td>\n",
       "      <td>0.040800</td>\n",
       "      <td>0.959353</td>\n",
       "      <td>02:36</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>11</td>\n",
       "      <td>0.426603</td>\n",
       "      <td>0.122298</td>\n",
       "      <td>0.035200</td>\n",
       "      <td>0.964807</td>\n",
       "      <td>02:37</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>12</td>\n",
       "      <td>0.431367</td>\n",
       "      <td>0.134176</td>\n",
       "      <td>0.036000</td>\n",
       "      <td>0.964152</td>\n",
       "      <td>02:36</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>13</td>\n",
       "      <td>0.432459</td>\n",
       "      <td>0.182154</td>\n",
       "      <td>0.047200</td>\n",
       "      <td>0.952312</td>\n",
       "      <td>02:37</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>14</td>\n",
       "      <td>0.414577</td>\n",
       "      <td>0.148939</td>\n",
       "      <td>0.045600</td>\n",
       "      <td>0.954424</td>\n",
       "      <td>02:36</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>15</td>\n",
       "      <td>0.431639</td>\n",
       "      <td>0.134338</td>\n",
       "      <td>0.038400</td>\n",
       "      <td>0.961665</td>\n",
       "      <td>02:36</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>16</td>\n",
       "      <td>0.426981</td>\n",
       "      <td>0.123276</td>\n",
       "      <td>0.032000</td>\n",
       "      <td>0.968002</td>\n",
       "      <td>02:37</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>17</td>\n",
       "      <td>0.429059</td>\n",
       "      <td>0.164866</td>\n",
       "      <td>0.052000</td>\n",
       "      <td>0.948335</td>\n",
       "      <td>02:36</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>18</td>\n",
       "      <td>0.410342</td>\n",
       "      <td>0.142374</td>\n",
       "      <td>0.042400</td>\n",
       "      <td>0.957631</td>\n",
       "      <td>02:37</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>19</td>\n",
       "      <td>0.408545</td>\n",
       "      <td>0.135474</td>\n",
       "      <td>0.040800</td>\n",
       "      <td>0.959385</td>\n",
       "      <td>02:37</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>20</td>\n",
       "      <td>0.413433</td>\n",
       "      <td>0.132628</td>\n",
       "      <td>0.037600</td>\n",
       "      <td>0.962456</td>\n",
       "      <td>02:37</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>21</td>\n",
       "      <td>0.408392</td>\n",
       "      <td>0.141283</td>\n",
       "      <td>0.036000</td>\n",
       "      <td>0.963977</td>\n",
       "      <td>02:37</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>22</td>\n",
       "      <td>0.407380</td>\n",
       "      <td>0.133035</td>\n",
       "      <td>0.037600</td>\n",
       "      <td>0.962330</td>\n",
       "      <td>02:36</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>23</td>\n",
       "      <td>0.415800</td>\n",
       "      <td>0.141236</td>\n",
       "      <td>0.043200</td>\n",
       "      <td>0.956400</td>\n",
       "      <td>02:37</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>24</td>\n",
       "      <td>0.400845</td>\n",
       "      <td>0.131808</td>\n",
       "      <td>0.035200</td>\n",
       "      <td>0.964852</td>\n",
       "      <td>02:36</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>25</td>\n",
       "      <td>0.403942</td>\n",
       "      <td>0.121852</td>\n",
       "      <td>0.037600</td>\n",
       "      <td>0.962366</td>\n",
       "      <td>02:37</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>26</td>\n",
       "      <td>0.402862</td>\n",
       "      <td>0.108081</td>\n",
       "      <td>0.034400</td>\n",
       "      <td>0.965589</td>\n",
       "      <td>02:37</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>27</td>\n",
       "      <td>0.401157</td>\n",
       "      <td>0.118129</td>\n",
       "      <td>0.029600</td>\n",
       "      <td>0.970442</td>\n",
       "      <td>02:37</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>28</td>\n",
       "      <td>0.392017</td>\n",
       "      <td>0.157174</td>\n",
       "      <td>0.049600</td>\n",
       "      <td>0.949911</td>\n",
       "      <td>02:36</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>29</td>\n",
       "      <td>0.398881</td>\n",
       "      <td>0.122476</td>\n",
       "      <td>0.035200</td>\n",
       "      <td>0.964680</td>\n",
       "      <td>02:37</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>30</td>\n",
       "      <td>0.380609</td>\n",
       "      <td>0.188715</td>\n",
       "      <td>0.052800</td>\n",
       "      <td>0.945986</td>\n",
       "      <td>02:37</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>31</td>\n",
       "      <td>0.386867</td>\n",
       "      <td>0.105909</td>\n",
       "      <td>0.030400</td>\n",
       "      <td>0.969708</td>\n",
       "      <td>02:36</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>32</td>\n",
       "      <td>0.379794</td>\n",
       "      <td>0.109935</td>\n",
       "      <td>0.033600</td>\n",
       "      <td>0.966258</td>\n",
       "      <td>02:36</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>33</td>\n",
       "      <td>0.379562</td>\n",
       "      <td>0.121071</td>\n",
       "      <td>0.032800</td>\n",
       "      <td>0.967092</td>\n",
       "      <td>02:37</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>34</td>\n",
       "      <td>0.368073</td>\n",
       "      <td>0.133887</td>\n",
       "      <td>0.038400</td>\n",
       "      <td>0.961430</td>\n",
       "      <td>02:37</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>35</td>\n",
       "      <td>0.376607</td>\n",
       "      <td>0.138109</td>\n",
       "      <td>0.040800</td>\n",
       "      <td>0.958922</td>\n",
       "      <td>02:36</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>36</td>\n",
       "      <td>0.365644</td>\n",
       "      <td>0.105688</td>\n",
       "      <td>0.029600</td>\n",
       "      <td>0.970341</td>\n",
       "      <td>02:36</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>37</td>\n",
       "      <td>0.375753</td>\n",
       "      <td>0.097728</td>\n",
       "      <td>0.026400</td>\n",
       "      <td>0.973661</td>\n",
       "      <td>02:36</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>38</td>\n",
       "      <td>0.365705</td>\n",
       "      <td>0.109810</td>\n",
       "      <td>0.032000</td>\n",
       "      <td>0.968055</td>\n",
       "      <td>02:36</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>39</td>\n",
       "      <td>0.363289</td>\n",
       "      <td>0.096369</td>\n",
       "      <td>0.027200</td>\n",
       "      <td>0.972828</td>\n",
       "      <td>02:36</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>40</td>\n",
       "      <td>0.369555</td>\n",
       "      <td>0.096529</td>\n",
       "      <td>0.028000</td>\n",
       "      <td>0.972061</td>\n",
       "      <td>02:36</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>41</td>\n",
       "      <td>0.359453</td>\n",
       "      <td>0.126611</td>\n",
       "      <td>0.034400</td>\n",
       "      <td>0.965553</td>\n",
       "      <td>02:36</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>42</td>\n",
       "      <td>0.351818</td>\n",
       "      <td>0.105090</td>\n",
       "      <td>0.031200</td>\n",
       "      <td>0.968840</td>\n",
       "      <td>02:36</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>43</td>\n",
       "      <td>0.358486</td>\n",
       "      <td>0.110299</td>\n",
       "      <td>0.028800</td>\n",
       "      <td>0.971175</td>\n",
       "      <td>02:36</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>44</td>\n",
       "      <td>0.354596</td>\n",
       "      <td>0.096980</td>\n",
       "      <td>0.032800</td>\n",
       "      <td>0.967126</td>\n",
       "      <td>02:36</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>45</td>\n",
       "      <td>0.353163</td>\n",
       "      <td>0.096221</td>\n",
       "      <td>0.028800</td>\n",
       "      <td>0.971191</td>\n",
       "      <td>02:37</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>46</td>\n",
       "      <td>0.350148</td>\n",
       "      <td>0.103788</td>\n",
       "      <td>0.028800</td>\n",
       "      <td>0.971212</td>\n",
       "      <td>02:36</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>47</td>\n",
       "      <td>0.346854</td>\n",
       "      <td>0.105391</td>\n",
       "      <td>0.029600</td>\n",
       "      <td>0.970447</td>\n",
       "      <td>02:36</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>48</td>\n",
       "      <td>0.345847</td>\n",
       "      <td>0.095808</td>\n",
       "      <td>0.028800</td>\n",
       "      <td>0.971216</td>\n",
       "      <td>02:36</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>49</td>\n",
       "      <td>0.337209</td>\n",
       "      <td>0.103618</td>\n",
       "      <td>0.028800</td>\n",
       "      <td>0.971220</td>\n",
       "      <td>02:37</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>50</td>\n",
       "      <td>0.340884</td>\n",
       "      <td>0.099517</td>\n",
       "      <td>0.026400</td>\n",
       "      <td>0.973626</td>\n",
       "      <td>02:37</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>51</td>\n",
       "      <td>0.341315</td>\n",
       "      <td>0.091427</td>\n",
       "      <td>0.024800</td>\n",
       "      <td>0.975220</td>\n",
       "      <td>02:37</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>52</td>\n",
       "      <td>0.337481</td>\n",
       "      <td>0.095630</td>\n",
       "      <td>0.029600</td>\n",
       "      <td>0.970459</td>\n",
       "      <td>02:37</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>53</td>\n",
       "      <td>0.336895</td>\n",
       "      <td>0.088849</td>\n",
       "      <td>0.025600</td>\n",
       "      <td>0.974427</td>\n",
       "      <td>02:36</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>54</td>\n",
       "      <td>0.346465</td>\n",
       "      <td>0.093150</td>\n",
       "      <td>0.024800</td>\n",
       "      <td>0.975233</td>\n",
       "      <td>02:36</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>55</td>\n",
       "      <td>0.339030</td>\n",
       "      <td>0.090115</td>\n",
       "      <td>0.028800</td>\n",
       "      <td>0.971228</td>\n",
       "      <td>02:36</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>56</td>\n",
       "      <td>0.336488</td>\n",
       "      <td>0.087279</td>\n",
       "      <td>0.028800</td>\n",
       "      <td>0.971246</td>\n",
       "      <td>02:36</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>57</td>\n",
       "      <td>0.331909</td>\n",
       "      <td>0.089274</td>\n",
       "      <td>0.026400</td>\n",
       "      <td>0.973645</td>\n",
       "      <td>02:36</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>58</td>\n",
       "      <td>0.339034</td>\n",
       "      <td>0.094423</td>\n",
       "      <td>0.028000</td>\n",
       "      <td>0.972042</td>\n",
       "      <td>02:36</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>59</td>\n",
       "      <td>0.338026</td>\n",
       "      <td>0.088593</td>\n",
       "      <td>0.028000</td>\n",
       "      <td>0.972042</td>\n",
       "      <td>02:37</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>60</td>\n",
       "      <td>0.339149</td>\n",
       "      <td>0.088496</td>\n",
       "      <td>0.025600</td>\n",
       "      <td>0.974441</td>\n",
       "      <td>02:37</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>61</td>\n",
       "      <td>0.343334</td>\n",
       "      <td>0.088822</td>\n",
       "      <td>0.024800</td>\n",
       "      <td>0.975206</td>\n",
       "      <td>02:36</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>62</td>\n",
       "      <td>0.336317</td>\n",
       "      <td>0.090998</td>\n",
       "      <td>0.028800</td>\n",
       "      <td>0.971230</td>\n",
       "      <td>02:36</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>63</td>\n",
       "      <td>0.342493</td>\n",
       "      <td>0.088309</td>\n",
       "      <td>0.026400</td>\n",
       "      <td>0.973642</td>\n",
       "      <td>02:36</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>"
      ],
      "text/plain": [
       "<IPython.core.display.HTML object>"
      ]
     },
     "metadata": {
      "tags": []
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "lr = [3e-3/100, 3e-3/20, 3e-3/10]\n",
    "learn.fit_one_cycle(64, lr)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 0,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "yHAM3FFnHAwH"
   },
   "outputs": [],
   "source": [
    "learn.save('ResNext50_FL_s2_124t256t299_973642')"
   ]
  },
  {
   "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, 'ResNet50_FL_aug_973642.csv')\n",
    "#LB 0.9832088004"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "colab_type": "text",
    "id": "n1qWewf4puNN"
   },
   "source": [
    "## Getting probabilities for Test set"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 0,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "uf0t1CZ1Jk_j"
   },
   "outputs": [],
   "source": [
    "learn.load('ResNext50_FL_s2_124t256t299_973642');"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 0,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "4BHiMUG44WRm"
   },
   "outputs": [],
   "source": [
    "class_name = ['Cargo', 'Military','Carrier' ,'Cruise','Tankers']"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 0,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "qo5kKKio3puL"
   },
   "outputs": [],
   "source": [
    "predsTTA, _ = learn.TTA(ds_type=DatasetType.Test)\n",
    "preds, _ = learn.get_preds(DatasetType.Test)\n",
    "\n",
    "predTTA_df = pd.DataFrame(predsTTA.numpy(), columns=class_name)\n",
    "preds_df = pd.DataFrame(preds.numpy(), columns=class_name)\n",
    "\n",
    "preds_df.to_csv('Preds_ResNeXt50_LB_983208.csv', index=False)\n",
    "predTTA_df.to_csv('TTA_ResNeXt50_LB_983208.csv', index=False)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 0,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "bNHXofqP6EDL"
   },
   "outputs": [],
   "source": []
  }
 ],
 "metadata": {
  "accelerator": "GPU",
  "colab": {
   "collapsed_sections": [],
   "name": "Final_ResNeXt50_FL.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
}