Final_ResNet50.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 0,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "tnU5-RMlLARg"
   },
   "outputs": [],
   "source": [
    "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": "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": "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(flip_vert=False, max_lighting=0.1, max_zoom=1.05, max_warp=0.)"
   ]
  },
  {
   "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": "markdown",
   "metadata": {
    "colab_type": "text",
    "id": "HbF7TYPvoYgl"
   },
   "source": [
    "## Training With Size 128"
   ]
  },
  {
   "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().normalize(imagenet_stats))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 0,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "us_BYAx1OT45"
   },
   "outputs": [],
   "source": [
    "learn = cnn_learner(data, models.resnet50, metrics=[error_rate, f_score])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 0,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 318
    },
    "colab_type": "code",
    "id": "LeBqZJ4RQScw",
    "outputId": "8e20cee6-c431-4d06-9595-7ada3e1d247f"
   },
   "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: 2.09E-03\n"
     ]
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAYUAAAEKCAYAAAD9xUlFAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nREQaKBRERKTB/wN7uWz1G5tsrAAAAABJRU5ErkJggg==\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": 614
    },
    "colab_type": "code",
    "id": "RGroV0abQXiC",
    "outputId": "846560e7-10e9-4d13-d64a-b6678ee26301"
   },
   "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>1.161818</td>\n",
       "      <td>0.766736</td>\n",
       "      <td>0.272000</td>\n",
       "      <td>0.721810</td>\n",
       "      <td>00:19</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>1</td>\n",
       "      <td>0.864874</td>\n",
       "      <td>0.640365</td>\n",
       "      <td>0.243200</td>\n",
       "      <td>0.752621</td>\n",
       "      <td>00:18</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>2</td>\n",
       "      <td>0.679338</td>\n",
       "      <td>0.570596</td>\n",
       "      <td>0.206400</td>\n",
       "      <td>0.790071</td>\n",
       "      <td>00:17</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>3</td>\n",
       "      <td>0.587189</td>\n",
       "      <td>0.481978</td>\n",
       "      <td>0.184000</td>\n",
       "      <td>0.813858</td>\n",
       "      <td>00:17</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>4</td>\n",
       "      <td>0.506771</td>\n",
       "      <td>0.467759</td>\n",
       "      <td>0.173600</td>\n",
       "      <td>0.821292</td>\n",
       "      <td>00:19</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>5</td>\n",
       "      <td>0.441403</td>\n",
       "      <td>0.418548</td>\n",
       "      <td>0.162400</td>\n",
       "      <td>0.836987</td>\n",
       "      <td>00:17</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>6</td>\n",
       "      <td>0.369249</td>\n",
       "      <td>0.405552</td>\n",
       "      <td>0.145600</td>\n",
       "      <td>0.854594</td>\n",
       "      <td>00:17</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>7</td>\n",
       "      <td>0.351321</td>\n",
       "      <td>0.397857</td>\n",
       "      <td>0.152000</td>\n",
       "      <td>0.847177</td>\n",
       "      <td>00:17</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>8</td>\n",
       "      <td>0.286855</td>\n",
       "      <td>0.377045</td>\n",
       "      <td>0.131200</td>\n",
       "      <td>0.868105</td>\n",
       "      <td>00:18</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>9</td>\n",
       "      <td>0.262540</td>\n",
       "      <td>0.361483</td>\n",
       "      <td>0.136800</td>\n",
       "      <td>0.862158</td>\n",
       "      <td>00:18</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>10</td>\n",
       "      <td>0.242353</td>\n",
       "      <td>0.345795</td>\n",
       "      <td>0.129600</td>\n",
       "      <td>0.870287</td>\n",
       "      <td>00:17</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>11</td>\n",
       "      <td>0.215156</td>\n",
       "      <td>0.343663</td>\n",
       "      <td>0.124800</td>\n",
       "      <td>0.874340</td>\n",
       "      <td>00:17</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>12</td>\n",
       "      <td>0.187963</td>\n",
       "      <td>0.346694</td>\n",
       "      <td>0.128800</td>\n",
       "      <td>0.870543</td>\n",
       "      <td>00:17</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>13</td>\n",
       "      <td>0.171979</td>\n",
       "      <td>0.343518</td>\n",
       "      <td>0.128000</td>\n",
       "      <td>0.871883</td>\n",
       "      <td>00:19</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>14</td>\n",
       "      <td>0.160370</td>\n",
       "      <td>0.340171</td>\n",
       "      <td>0.129600</td>\n",
       "      <td>0.870395</td>\n",
       "      <td>00:17</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>15</td>\n",
       "      <td>0.150470</td>\n",
       "      <td>0.341917</td>\n",
       "      <td>0.123200</td>\n",
       "      <td>0.876228</td>\n",
       "      <td>00:17</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>16</td>\n",
       "      <td>0.146100</td>\n",
       "      <td>0.337948</td>\n",
       "      <td>0.118400</td>\n",
       "      <td>0.881302</td>\n",
       "      <td>00:17</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>17</td>\n",
       "      <td>0.136757</td>\n",
       "      <td>0.338305</td>\n",
       "      <td>0.122400</td>\n",
       "      <td>0.877547</td>\n",
       "      <td>00:18</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(2e-3))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 0,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "2fw26Q84QkPm"
   },
   "outputs": [],
   "source": [
    "learn.save('Res50_128_877547')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 0,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 318
    },
    "colab_type": "code",
    "id": "zS6dV7FWR5Oi",
    "outputId": "e094319a-88ca-4e52-aacf-aa412ac6eb38"
   },
   "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: 1.10E-06\n"
     ]
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAYsAAAEKCAYAAADjDHn2AAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\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('Res34_128_862168');\n",
    "learn.unfreeze();\n",
    "learn.clip_grad();"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 0,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "W2vr8AfDSBnE"
   },
   "outputs": [],
   "source": [
    "lr = [2e-3/1000, 2e-3/100, 2e-3/100]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 0,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 802
    },
    "colab_type": "code",
    "id": "koP5h6H1SQ2v",
    "outputId": "c623332d-55d3-4895-cb7d-e1eee0a79e20"
   },
   "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>0.135267</td>\n",
       "      <td>0.337586</td>\n",
       "      <td>0.121600</td>\n",
       "      <td>0.878115</td>\n",
       "      <td>00:24</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>1</td>\n",
       "      <td>0.134537</td>\n",
       "      <td>0.329398</td>\n",
       "      <td>0.126400</td>\n",
       "      <td>0.873305</td>\n",
       "      <td>00:23</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>2</td>\n",
       "      <td>0.124365</td>\n",
       "      <td>0.319405</td>\n",
       "      <td>0.113600</td>\n",
       "      <td>0.886449</td>\n",
       "      <td>00:24</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>3</td>\n",
       "      <td>0.121675</td>\n",
       "      <td>0.304713</td>\n",
       "      <td>0.106400</td>\n",
       "      <td>0.893495</td>\n",
       "      <td>00:24</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>4</td>\n",
       "      <td>0.108551</td>\n",
       "      <td>0.306889</td>\n",
       "      <td>0.112000</td>\n",
       "      <td>0.887798</td>\n",
       "      <td>00:23</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>5</td>\n",
       "      <td>0.091137</td>\n",
       "      <td>0.301017</td>\n",
       "      <td>0.101600</td>\n",
       "      <td>0.898209</td>\n",
       "      <td>00:23</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>6</td>\n",
       "      <td>0.074224</td>\n",
       "      <td>0.310520</td>\n",
       "      <td>0.101600</td>\n",
       "      <td>0.898382</td>\n",
       "      <td>00:24</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>7</td>\n",
       "      <td>0.062111</td>\n",
       "      <td>0.309880</td>\n",
       "      <td>0.100800</td>\n",
       "      <td>0.899323</td>\n",
       "      <td>00:23</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>8</td>\n",
       "      <td>0.049269</td>\n",
       "      <td>0.304701</td>\n",
       "      <td>0.100800</td>\n",
       "      <td>0.899174</td>\n",
       "      <td>00:23</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>9</td>\n",
       "      <td>0.040387</td>\n",
       "      <td>0.305732</td>\n",
       "      <td>0.093600</td>\n",
       "      <td>0.906275</td>\n",
       "      <td>00:24</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>10</td>\n",
       "      <td>0.029290</td>\n",
       "      <td>0.312754</td>\n",
       "      <td>0.088800</td>\n",
       "      <td>0.911290</td>\n",
       "      <td>00:24</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>11</td>\n",
       "      <td>0.027301</td>\n",
       "      <td>0.324529</td>\n",
       "      <td>0.086400</td>\n",
       "      <td>0.913513</td>\n",
       "      <td>00:23</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>12</td>\n",
       "      <td>0.024073</td>\n",
       "      <td>0.319040</td>\n",
       "      <td>0.082400</td>\n",
       "      <td>0.917041</td>\n",
       "      <td>00:23</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>13</td>\n",
       "      <td>0.018579</td>\n",
       "      <td>0.323578</td>\n",
       "      <td>0.087200</td>\n",
       "      <td>0.912537</td>\n",
       "      <td>00:24</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>14</td>\n",
       "      <td>0.015345</td>\n",
       "      <td>0.326203</td>\n",
       "      <td>0.081600</td>\n",
       "      <td>0.917838</td>\n",
       "      <td>00:23</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>15</td>\n",
       "      <td>0.011110</td>\n",
       "      <td>0.320390</td>\n",
       "      <td>0.081600</td>\n",
       "      <td>0.917935</td>\n",
       "      <td>00:24</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>16</td>\n",
       "      <td>0.012128</td>\n",
       "      <td>0.315355</td>\n",
       "      <td>0.077600</td>\n",
       "      <td>0.922399</td>\n",
       "      <td>00:24</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>17</td>\n",
       "      <td>0.008302</td>\n",
       "      <td>0.332167</td>\n",
       "      <td>0.076800</td>\n",
       "      <td>0.923141</td>\n",
       "      <td>00:23</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>18</td>\n",
       "      <td>0.006702</td>\n",
       "      <td>0.323936</td>\n",
       "      <td>0.076800</td>\n",
       "      <td>0.923058</td>\n",
       "      <td>00:23</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>19</td>\n",
       "      <td>0.007593</td>\n",
       "      <td>0.324590</td>\n",
       "      <td>0.080000</td>\n",
       "      <td>0.920005</td>\n",
       "      <td>00:24</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>20</td>\n",
       "      <td>0.007394</td>\n",
       "      <td>0.329474</td>\n",
       "      <td>0.077600</td>\n",
       "      <td>0.922031</td>\n",
       "      <td>00:24</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>21</td>\n",
       "      <td>0.006018</td>\n",
       "      <td>0.326315</td>\n",
       "      <td>0.081600</td>\n",
       "      <td>0.918383</td>\n",
       "      <td>00:23</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>22</td>\n",
       "      <td>0.005137</td>\n",
       "      <td>0.325121</td>\n",
       "      <td>0.076800</td>\n",
       "      <td>0.923248</td>\n",
       "      <td>00:23</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>23</td>\n",
       "      <td>0.005202</td>\n",
       "      <td>0.326295</td>\n",
       "      <td>0.079200</td>\n",
       "      <td>0.920846</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(24, lr)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 0,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "ts518fO-SUBg"
   },
   "outputs": [],
   "source": [
    "learn.save('s2_Res50_128_920846')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "colab_type": "text",
    "id": "k5h9Ctrvolma"
   },
   "source": [
    "## Training with Size 256"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 0,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "q6WIPZG_WPck"
   },
   "outputs": [],
   "source": [
    "data_256 = (src.transform(tfms, size=256, resize_method=ResizeMethod.SQUISH)\n",
    "            .databunch()\n",
    "            .normalize(imagenet_stats))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 0,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "j5dy8cNfWSll"
   },
   "outputs": [],
   "source": [
    "learn.data = data_256"
   ]
  },
  {
   "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": {
     "base_uri": "https://localhost:8080/",
     "height": 318
    },
    "colab_type": "code",
    "id": "gFXiAkC1XD1b",
    "outputId": "e3b6d755-dbf6-4a3a-9752-8a26d26a6da4"
   },
   "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: 9.12E-07\n"
     ]
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAYsAAAEKCAYAAADjDHn2AAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nhYiIBPT/AT+S8skMffE+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": 551
    },
    "colab_type": "code",
    "id": "AwsccXWuXH1u",
    "outputId": "719f440e-24a9-411a-c8e5-3b21d7f9cbb3"
   },
   "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>0.235807</td>\n",
       "      <td>0.394416</td>\n",
       "      <td>0.112000</td>\n",
       "      <td>0.888106</td>\n",
       "      <td>00:56</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>1</td>\n",
       "      <td>0.225214</td>\n",
       "      <td>0.382139</td>\n",
       "      <td>0.108000</td>\n",
       "      <td>0.891752</td>\n",
       "      <td>00:54</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>2</td>\n",
       "      <td>0.200634</td>\n",
       "      <td>0.356438</td>\n",
       "      <td>0.100800</td>\n",
       "      <td>0.899363</td>\n",
       "      <td>00:54</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>3</td>\n",
       "      <td>0.189586</td>\n",
       "      <td>0.339507</td>\n",
       "      <td>0.100000</td>\n",
       "      <td>0.900009</td>\n",
       "      <td>00:55</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>4</td>\n",
       "      <td>0.176775</td>\n",
       "      <td>0.314939</td>\n",
       "      <td>0.096000</td>\n",
       "      <td>0.903945</td>\n",
       "      <td>00:54</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>5</td>\n",
       "      <td>0.180184</td>\n",
       "      <td>0.294531</td>\n",
       "      <td>0.091200</td>\n",
       "      <td>0.908426</td>\n",
       "      <td>00:54</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>6</td>\n",
       "      <td>0.155939</td>\n",
       "      <td>0.286191</td>\n",
       "      <td>0.092800</td>\n",
       "      <td>0.906873</td>\n",
       "      <td>00:54</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>7</td>\n",
       "      <td>0.137322</td>\n",
       "      <td>0.274569</td>\n",
       "      <td>0.089600</td>\n",
       "      <td>0.910545</td>\n",
       "      <td>00:54</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>8</td>\n",
       "      <td>0.135025</td>\n",
       "      <td>0.263795</td>\n",
       "      <td>0.087200</td>\n",
       "      <td>0.912825</td>\n",
       "      <td>00:54</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>9</td>\n",
       "      <td>0.118665</td>\n",
       "      <td>0.260747</td>\n",
       "      <td>0.087200</td>\n",
       "      <td>0.913069</td>\n",
       "      <td>00:54</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>10</td>\n",
       "      <td>0.127401</td>\n",
       "      <td>0.259315</td>\n",
       "      <td>0.088800</td>\n",
       "      <td>0.911151</td>\n",
       "      <td>00:54</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>11</td>\n",
       "      <td>0.122290</td>\n",
       "      <td>0.263109</td>\n",
       "      <td>0.084000</td>\n",
       "      <td>0.916470</td>\n",
       "      <td>00:54</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>12</td>\n",
       "      <td>0.113356</td>\n",
       "      <td>0.254831</td>\n",
       "      <td>0.084800</td>\n",
       "      <td>0.915125</td>\n",
       "      <td>00:54</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>13</td>\n",
       "      <td>0.114597</td>\n",
       "      <td>0.256786</td>\n",
       "      <td>0.082400</td>\n",
       "      <td>0.917729</td>\n",
       "      <td>00:55</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>14</td>\n",
       "      <td>0.120454</td>\n",
       "      <td>0.254396</td>\n",
       "      <td>0.080000</td>\n",
       "      <td>0.919794</td>\n",
       "      <td>00:54</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>15</td>\n",
       "      <td>0.104882</td>\n",
       "      <td>0.255340</td>\n",
       "      <td>0.081600</td>\n",
       "      <td>0.918433</td>\n",
       "      <td>00:54</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(16, slice(1e-4))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 0,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "9I0No7aOX7CY"
   },
   "outputs": [],
   "source": [
    "learn.save('Res50_128to224_918433')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 0,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 318
    },
    "colab_type": "code",
    "id": "XaTD6qdYZZjU",
    "outputId": "5e7104d9-6be5-4605-8a93-97b4265f911b"
   },
   "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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\nSUVORK5CYII=\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": {
     "base_uri": "https://localhost:8080/",
     "height": 1053
    },
    "colab_type": "code",
    "id": "8RqmZw0MZggO",
    "outputId": "f02d6b4a-ae9a-4833-b0cd-67a69b6626df"
   },
   "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>0.113888</td>\n",
       "      <td>0.251722</td>\n",
       "      <td>0.083200</td>\n",
       "      <td>0.917174</td>\n",
       "      <td>01:13</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>1</td>\n",
       "      <td>0.107678</td>\n",
       "      <td>0.243732</td>\n",
       "      <td>0.084800</td>\n",
       "      <td>0.915250</td>\n",
       "      <td>01:12</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>2</td>\n",
       "      <td>0.097923</td>\n",
       "      <td>0.230999</td>\n",
       "      <td>0.080000</td>\n",
       "      <td>0.920201</td>\n",
       "      <td>01:13</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>3</td>\n",
       "      <td>0.092969</td>\n",
       "      <td>0.222876</td>\n",
       "      <td>0.076000</td>\n",
       "      <td>0.924146</td>\n",
       "      <td>01:12</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>4</td>\n",
       "      <td>0.071634</td>\n",
       "      <td>0.205767</td>\n",
       "      <td>0.068800</td>\n",
       "      <td>0.931199</td>\n",
       "      <td>01:12</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>5</td>\n",
       "      <td>0.060520</td>\n",
       "      <td>0.193709</td>\n",
       "      <td>0.065600</td>\n",
       "      <td>0.934445</td>\n",
       "      <td>01:12</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>6</td>\n",
       "      <td>0.050068</td>\n",
       "      <td>0.197155</td>\n",
       "      <td>0.060000</td>\n",
       "      <td>0.939721</td>\n",
       "      <td>01:12</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>7</td>\n",
       "      <td>0.037427</td>\n",
       "      <td>0.191329</td>\n",
       "      <td>0.057600</td>\n",
       "      <td>0.942378</td>\n",
       "      <td>01:12</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>8</td>\n",
       "      <td>0.029954</td>\n",
       "      <td>0.185044</td>\n",
       "      <td>0.056800</td>\n",
       "      <td>0.943431</td>\n",
       "      <td>01:13</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>9</td>\n",
       "      <td>0.028350</td>\n",
       "      <td>0.181051</td>\n",
       "      <td>0.055200</td>\n",
       "      <td>0.944922</td>\n",
       "      <td>01:13</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>10</td>\n",
       "      <td>0.021422</td>\n",
       "      <td>0.176074</td>\n",
       "      <td>0.050400</td>\n",
       "      <td>0.949747</td>\n",
       "      <td>01:13</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>11</td>\n",
       "      <td>0.015707</td>\n",
       "      <td>0.180030</td>\n",
       "      <td>0.049600</td>\n",
       "      <td>0.950538</td>\n",
       "      <td>01:13</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>12</td>\n",
       "      <td>0.011665</td>\n",
       "      <td>0.188501</td>\n",
       "      <td>0.050400</td>\n",
       "      <td>0.949687</td>\n",
       "      <td>01:12</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>13</td>\n",
       "      <td>0.010661</td>\n",
       "      <td>0.181983</td>\n",
       "      <td>0.048800</td>\n",
       "      <td>0.951165</td>\n",
       "      <td>01:13</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>14</td>\n",
       "      <td>0.009608</td>\n",
       "      <td>0.187425</td>\n",
       "      <td>0.050400</td>\n",
       "      <td>0.949616</td>\n",
       "      <td>01:12</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>15</td>\n",
       "      <td>0.007635</td>\n",
       "      <td>0.184681</td>\n",
       "      <td>0.052000</td>\n",
       "      <td>0.948211</td>\n",
       "      <td>01:12</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>16</td>\n",
       "      <td>0.006309</td>\n",
       "      <td>0.182862</td>\n",
       "      <td>0.049600</td>\n",
       "      <td>0.950472</td>\n",
       "      <td>01:13</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>17</td>\n",
       "      <td>0.005092</td>\n",
       "      <td>0.189113</td>\n",
       "      <td>0.049600</td>\n",
       "      <td>0.950485</td>\n",
       "      <td>01:12</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>18</td>\n",
       "      <td>0.004729</td>\n",
       "      <td>0.196876</td>\n",
       "      <td>0.048000</td>\n",
       "      <td>0.952221</td>\n",
       "      <td>01:12</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>19</td>\n",
       "      <td>0.003372</td>\n",
       "      <td>0.186720</td>\n",
       "      <td>0.044800</td>\n",
       "      <td>0.955215</td>\n",
       "      <td>01:13</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>20</td>\n",
       "      <td>0.002821</td>\n",
       "      <td>0.189491</td>\n",
       "      <td>0.047200</td>\n",
       "      <td>0.952847</td>\n",
       "      <td>01:13</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>21</td>\n",
       "      <td>0.002515</td>\n",
       "      <td>0.191241</td>\n",
       "      <td>0.048000</td>\n",
       "      <td>0.951985</td>\n",
       "      <td>01:12</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>22</td>\n",
       "      <td>0.001960</td>\n",
       "      <td>0.185051</td>\n",
       "      <td>0.047200</td>\n",
       "      <td>0.952768</td>\n",
       "      <td>01:13</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>23</td>\n",
       "      <td>0.001930</td>\n",
       "      <td>0.191087</td>\n",
       "      <td>0.044800</td>\n",
       "      <td>0.955202</td>\n",
       "      <td>01:12</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>24</td>\n",
       "      <td>0.001886</td>\n",
       "      <td>0.187971</td>\n",
       "      <td>0.046400</td>\n",
       "      <td>0.953585</td>\n",
       "      <td>01:12</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>25</td>\n",
       "      <td>0.002528</td>\n",
       "      <td>0.196457</td>\n",
       "      <td>0.049600</td>\n",
       "      <td>0.950591</td>\n",
       "      <td>01:12</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>26</td>\n",
       "      <td>0.001684</td>\n",
       "      <td>0.190480</td>\n",
       "      <td>0.045600</td>\n",
       "      <td>0.954418</td>\n",
       "      <td>01:12</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>27</td>\n",
       "      <td>0.001345</td>\n",
       "      <td>0.188420</td>\n",
       "      <td>0.045600</td>\n",
       "      <td>0.954428</td>\n",
       "      <td>01:12</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>28</td>\n",
       "      <td>0.001131</td>\n",
       "      <td>0.186815</td>\n",
       "      <td>0.046400</td>\n",
       "      <td>0.953563</td>\n",
       "      <td>01:12</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>29</td>\n",
       "      <td>0.001349</td>\n",
       "      <td>0.190589</td>\n",
       "      <td>0.044000</td>\n",
       "      <td>0.956113</td>\n",
       "      <td>01:12</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>30</td>\n",
       "      <td>0.001452</td>\n",
       "      <td>0.185765</td>\n",
       "      <td>0.046400</td>\n",
       "      <td>0.953652</td>\n",
       "      <td>01:13</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>31</td>\n",
       "      <td>0.001284</td>\n",
       "      <td>0.188537</td>\n",
       "      <td>0.044800</td>\n",
       "      <td>0.955245</td>\n",
       "      <td>01:13</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>"
      ],
      "text/plain": [
       "<IPython.core.display.HTML object>"
      ]
     },
     "metadata": {
      "tags": []
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "lr = [1e-3/1000, 1e-3/100, 1e-3/100]\n",
    "learn.fit_one_cycle(32, lr)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 0,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "QGlDDF4oZ95K"
   },
   "outputs": [],
   "source": [
    "learn.save('s2_Res50_128to256_955245')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 0,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 426
    },
    "colab_type": "code",
    "id": "Jqa5KBhxv8Oz",
    "outputId": "cd7e988d-1c72-45fd-ec6c-7ad00b118041"
   },
   "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>0.001455</td>\n",
       "      <td>0.185611</td>\n",
       "      <td>0.044800</td>\n",
       "      <td>0.955195</td>\n",
       "      <td>01:13</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>1</td>\n",
       "      <td>0.000961</td>\n",
       "      <td>0.190947</td>\n",
       "      <td>0.043200</td>\n",
       "      <td>0.956866</td>\n",
       "      <td>01:13</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>2</td>\n",
       "      <td>0.002057</td>\n",
       "      <td>0.206501</td>\n",
       "      <td>0.040000</td>\n",
       "      <td>0.960093</td>\n",
       "      <td>01:13</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>3</td>\n",
       "      <td>0.004629</td>\n",
       "      <td>0.225442</td>\n",
       "      <td>0.046400</td>\n",
       "      <td>0.953692</td>\n",
       "      <td>01:12</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>4</td>\n",
       "      <td>0.006945</td>\n",
       "      <td>0.213679</td>\n",
       "      <td>0.044000</td>\n",
       "      <td>0.956062</td>\n",
       "      <td>01:12</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>5</td>\n",
       "      <td>0.003239</td>\n",
       "      <td>0.223394</td>\n",
       "      <td>0.044000</td>\n",
       "      <td>0.956085</td>\n",
       "      <td>01:12</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>6</td>\n",
       "      <td>0.002326</td>\n",
       "      <td>0.229323</td>\n",
       "      <td>0.043200</td>\n",
       "      <td>0.956987</td>\n",
       "      <td>01:13</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>7</td>\n",
       "      <td>0.003484</td>\n",
       "      <td>0.210361</td>\n",
       "      <td>0.040800</td>\n",
       "      <td>0.959092</td>\n",
       "      <td>01:12</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>8</td>\n",
       "      <td>0.001639</td>\n",
       "      <td>0.215162</td>\n",
       "      <td>0.036000</td>\n",
       "      <td>0.963980</td>\n",
       "      <td>01:12</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>9</td>\n",
       "      <td>0.001685</td>\n",
       "      <td>0.211433</td>\n",
       "      <td>0.036000</td>\n",
       "      <td>0.964078</td>\n",
       "      <td>01:12</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>10</td>\n",
       "      <td>0.001319</td>\n",
       "      <td>0.206052</td>\n",
       "      <td>0.038400</td>\n",
       "      <td>0.961600</td>\n",
       "      <td>01:12</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>11</td>\n",
       "      <td>0.000488</td>\n",
       "      <td>0.205780</td>\n",
       "      <td>0.036000</td>\n",
       "      <td>0.963999</td>\n",
       "      <td>01:13</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>"
      ],
      "text/plain": [
       "<IPython.core.display.HTML object>"
      ]
     },
     "metadata": {
      "tags": []
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "lr = [2e-3/1000, 2e-3/100, 2e-3/100]\n",
    "learn.fit_one_cycle(12, lr)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 0,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "qsouTNpz1ez5"
   },
   "outputs": [],
   "source": [
    "learn.save('s2_Res50_128to256_963999')"
   ]
  },
  {
   "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, 'FAI_Res50_128to246_963999.csv')\n",
    "#LB 0.9728709421"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "colab_type": "text",
    "id": "pjURvtLeos1l"
   },
   "source": [
    "## Training with Size 299"
   ]
  },
  {
   "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=24)\n",
    "            .normalize(imagenet_stats))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 0,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "hVLcXcAqMtcN"
   },
   "outputs": [],
   "source": [
    "learn = cnn_learner(data_299, models.resnet50, metrics=[error_rate, f_score])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 0,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "1DVIW86w37O6"
   },
   "outputs": [],
   "source": [
    "learn.load('s2_Res50_128to256_963999');\n",
    "learn.freeze();\n",
    "learn.clip_grad();"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 0,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 317
    },
    "colab_type": "code",
    "id": "zinzUfkQfzaT",
    "outputId": "6170c22d-826e-4153-db3a-7443e3962c80"
   },
   "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: 7.59E-07\n"
     ]
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAZIAAAEKCAYAAAA4t9PUAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\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": "95I-FsnEP9u_"
   },
   "outputs": [],
   "source": [
    "lr = 1e-3\n",
    "lrs = [lr/9, lr/3, lr]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 0,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 1041
    },
    "colab_type": "code",
    "id": "pIn8HrsBzt36",
    "outputId": "7b9f6dde-a116-4fad-c584-4a1aed323535"
   },
   "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>0.025667</td>\n",
       "      <td>0.206577</td>\n",
       "      <td>0.046400</td>\n",
       "      <td>0.953604</td>\n",
       "      <td>01:27</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>1</td>\n",
       "      <td>0.017345</td>\n",
       "      <td>0.213024</td>\n",
       "      <td>0.045600</td>\n",
       "      <td>0.954456</td>\n",
       "      <td>01:26</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>2</td>\n",
       "      <td>0.011777</td>\n",
       "      <td>0.208933</td>\n",
       "      <td>0.046400</td>\n",
       "      <td>0.953717</td>\n",
       "      <td>01:28</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>3</td>\n",
       "      <td>0.015972</td>\n",
       "      <td>0.215782</td>\n",
       "      <td>0.048000</td>\n",
       "      <td>0.952036</td>\n",
       "      <td>01:27</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>4</td>\n",
       "      <td>0.021094</td>\n",
       "      <td>0.228127</td>\n",
       "      <td>0.044800</td>\n",
       "      <td>0.955296</td>\n",
       "      <td>01:26</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>5</td>\n",
       "      <td>0.016081</td>\n",
       "      <td>0.246221</td>\n",
       "      <td>0.044800</td>\n",
       "      <td>0.955447</td>\n",
       "      <td>01:26</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>6</td>\n",
       "      <td>0.025471</td>\n",
       "      <td>0.243374</td>\n",
       "      <td>0.048000</td>\n",
       "      <td>0.952246</td>\n",
       "      <td>01:27</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>7</td>\n",
       "      <td>0.014863</td>\n",
       "      <td>0.257133</td>\n",
       "      <td>0.047200</td>\n",
       "      <td>0.952855</td>\n",
       "      <td>01:26</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>8</td>\n",
       "      <td>0.012427</td>\n",
       "      <td>0.259374</td>\n",
       "      <td>0.043200</td>\n",
       "      <td>0.956947</td>\n",
       "      <td>01:27</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>9</td>\n",
       "      <td>0.011441</td>\n",
       "      <td>0.273805</td>\n",
       "      <td>0.047200</td>\n",
       "      <td>0.952891</td>\n",
       "      <td>01:27</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>10</td>\n",
       "      <td>0.015612</td>\n",
       "      <td>0.260337</td>\n",
       "      <td>0.042400</td>\n",
       "      <td>0.957582</td>\n",
       "      <td>01:27</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>11</td>\n",
       "      <td>0.013822</td>\n",
       "      <td>0.264048</td>\n",
       "      <td>0.040800</td>\n",
       "      <td>0.959186</td>\n",
       "      <td>01:26</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>12</td>\n",
       "      <td>0.008849</td>\n",
       "      <td>0.285366</td>\n",
       "      <td>0.045600</td>\n",
       "      <td>0.954679</td>\n",
       "      <td>01:26</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>13</td>\n",
       "      <td>0.004418</td>\n",
       "      <td>0.289661</td>\n",
       "      <td>0.043200</td>\n",
       "      <td>0.956722</td>\n",
       "      <td>01:27</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>14</td>\n",
       "      <td>0.021651</td>\n",
       "      <td>0.327159</td>\n",
       "      <td>0.044800</td>\n",
       "      <td>0.955223</td>\n",
       "      <td>01:27</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>15</td>\n",
       "      <td>0.014103</td>\n",
       "      <td>0.335805</td>\n",
       "      <td>0.048000</td>\n",
       "      <td>0.951947</td>\n",
       "      <td>01:27</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>16</td>\n",
       "      <td>0.011197</td>\n",
       "      <td>0.326859</td>\n",
       "      <td>0.044800</td>\n",
       "      <td>0.955382</td>\n",
       "      <td>01:27</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>17</td>\n",
       "      <td>0.009521</td>\n",
       "      <td>0.316034</td>\n",
       "      <td>0.043200</td>\n",
       "      <td>0.956786</td>\n",
       "      <td>01:27</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>18</td>\n",
       "      <td>0.011850</td>\n",
       "      <td>0.335036</td>\n",
       "      <td>0.042400</td>\n",
       "      <td>0.957718</td>\n",
       "      <td>01:26</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>19</td>\n",
       "      <td>0.009195</td>\n",
       "      <td>0.331111</td>\n",
       "      <td>0.041600</td>\n",
       "      <td>0.958319</td>\n",
       "      <td>01:26</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>20</td>\n",
       "      <td>0.005719</td>\n",
       "      <td>0.326637</td>\n",
       "      <td>0.040000</td>\n",
       "      <td>0.960002</td>\n",
       "      <td>01:27</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>21</td>\n",
       "      <td>0.010204</td>\n",
       "      <td>0.327381</td>\n",
       "      <td>0.044000</td>\n",
       "      <td>0.956036</td>\n",
       "      <td>01:26</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>22</td>\n",
       "      <td>0.014183</td>\n",
       "      <td>0.329111</td>\n",
       "      <td>0.042400</td>\n",
       "      <td>0.957590</td>\n",
       "      <td>01:26</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>23</td>\n",
       "      <td>0.021510</td>\n",
       "      <td>0.331401</td>\n",
       "      <td>0.040800</td>\n",
       "      <td>0.959226</td>\n",
       "      <td>01:26</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>24</td>\n",
       "      <td>0.012790</td>\n",
       "      <td>0.329454</td>\n",
       "      <td>0.039200</td>\n",
       "      <td>0.960812</td>\n",
       "      <td>01:27</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>25</td>\n",
       "      <td>0.006882</td>\n",
       "      <td>0.343892</td>\n",
       "      <td>0.044800</td>\n",
       "      <td>0.955247</td>\n",
       "      <td>01:26</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>26</td>\n",
       "      <td>0.003541</td>\n",
       "      <td>0.339850</td>\n",
       "      <td>0.040800</td>\n",
       "      <td>0.959089</td>\n",
       "      <td>01:27</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>27</td>\n",
       "      <td>0.003712</td>\n",
       "      <td>0.326728</td>\n",
       "      <td>0.044800</td>\n",
       "      <td>0.955227</td>\n",
       "      <td>01:28</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>28</td>\n",
       "      <td>0.007764</td>\n",
       "      <td>0.324060</td>\n",
       "      <td>0.040000</td>\n",
       "      <td>0.960058</td>\n",
       "      <td>01:27</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>29</td>\n",
       "      <td>0.006429</td>\n",
       "      <td>0.323300</td>\n",
       "      <td>0.040800</td>\n",
       "      <td>0.959198</td>\n",
       "      <td>01:26</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>30</td>\n",
       "      <td>0.004221</td>\n",
       "      <td>0.317288</td>\n",
       "      <td>0.044000</td>\n",
       "      <td>0.956055</td>\n",
       "      <td>01:26</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>31</td>\n",
       "      <td>0.021248</td>\n",
       "      <td>0.325063</td>\n",
       "      <td>0.042400</td>\n",
       "      <td>0.957632</td>\n",
       "      <td>01:26</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(3e-4))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 0,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "fwzRyMM2htww"
   },
   "outputs": [],
   "source": [
    "learn.unfreeze()\n",
    "learn.clip_grad();"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 0,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 301
    },
    "colab_type": "code",
    "id": "ski9rgqKhyBp",
    "outputId": "b386c547-3d61-44c8-a572-8fa1af38d62c"
   },
   "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"
     ]
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAZ4AAAELCAYAAADp1+D/AAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nRERCSsEjIiIhpeAREZGQUvCIiEhI/T+TqKy7aDydlgAAAABJRU5ErkJggg==\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "tags": []
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "learn.lr_find()\n",
    "learn.recorder.plot()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 0,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "MCcBpmIIh1Tb"
   },
   "outputs": [],
   "source": [
    "lr = 4e-4\n",
    "lrs = [lr/100, lr/20, lr/5]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 0,
   "metadata": {
    "colab": {
     "base_uri": "https://localhost:8080/",
     "height": 1041
    },
    "colab_type": "code",
    "id": "KnIOshNFiDlF",
    "outputId": "08773a51-2af3-47d9-ea95-351169607996"
   },
   "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>0.003526</td>\n",
       "      <td>0.318683</td>\n",
       "      <td>0.040000</td>\n",
       "      <td>0.959982</td>\n",
       "      <td>02:03</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>1</td>\n",
       "      <td>0.005021</td>\n",
       "      <td>0.327624</td>\n",
       "      <td>0.046400</td>\n",
       "      <td>0.953807</td>\n",
       "      <td>02:03</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>2</td>\n",
       "      <td>0.011597</td>\n",
       "      <td>0.395443</td>\n",
       "      <td>0.047200</td>\n",
       "      <td>0.952585</td>\n",
       "      <td>02:03</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>3</td>\n",
       "      <td>0.020771</td>\n",
       "      <td>0.333545</td>\n",
       "      <td>0.040800</td>\n",
       "      <td>0.959174</td>\n",
       "      <td>02:04</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>4</td>\n",
       "      <td>0.014372</td>\n",
       "      <td>0.350398</td>\n",
       "      <td>0.044800</td>\n",
       "      <td>0.955408</td>\n",
       "      <td>02:04</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>5</td>\n",
       "      <td>0.011309</td>\n",
       "      <td>0.361765</td>\n",
       "      <td>0.048000</td>\n",
       "      <td>0.952423</td>\n",
       "      <td>02:03</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>6</td>\n",
       "      <td>0.038362</td>\n",
       "      <td>0.343047</td>\n",
       "      <td>0.045600</td>\n",
       "      <td>0.954636</td>\n",
       "      <td>02:03</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>7</td>\n",
       "      <td>0.049046</td>\n",
       "      <td>0.375772</td>\n",
       "      <td>0.047200</td>\n",
       "      <td>0.952783</td>\n",
       "      <td>02:03</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>8</td>\n",
       "      <td>0.020377</td>\n",
       "      <td>0.318941</td>\n",
       "      <td>0.040000</td>\n",
       "      <td>0.960068</td>\n",
       "      <td>02:04</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>9</td>\n",
       "      <td>0.043044</td>\n",
       "      <td>0.370518</td>\n",
       "      <td>0.040000</td>\n",
       "      <td>0.959832</td>\n",
       "      <td>02:03</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>10</td>\n",
       "      <td>0.020216</td>\n",
       "      <td>0.384399</td>\n",
       "      <td>0.049600</td>\n",
       "      <td>0.950438</td>\n",
       "      <td>02:04</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>11</td>\n",
       "      <td>0.032647</td>\n",
       "      <td>0.409459</td>\n",
       "      <td>0.044800</td>\n",
       "      <td>0.955382</td>\n",
       "      <td>02:03</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>12</td>\n",
       "      <td>0.041171</td>\n",
       "      <td>0.380190</td>\n",
       "      <td>0.043200</td>\n",
       "      <td>0.956805</td>\n",
       "      <td>02:03</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>13</td>\n",
       "      <td>0.026374</td>\n",
       "      <td>0.319124</td>\n",
       "      <td>0.042400</td>\n",
       "      <td>0.957417</td>\n",
       "      <td>02:04</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>14</td>\n",
       "      <td>0.028982</td>\n",
       "      <td>0.324887</td>\n",
       "      <td>0.040800</td>\n",
       "      <td>0.959169</td>\n",
       "      <td>02:03</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>15</td>\n",
       "      <td>0.012970</td>\n",
       "      <td>0.290193</td>\n",
       "      <td>0.040800</td>\n",
       "      <td>0.959234</td>\n",
       "      <td>02:03</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>16</td>\n",
       "      <td>0.013481</td>\n",
       "      <td>0.321117</td>\n",
       "      <td>0.038400</td>\n",
       "      <td>0.961661</td>\n",
       "      <td>02:03</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>17</td>\n",
       "      <td>0.017946</td>\n",
       "      <td>0.332015</td>\n",
       "      <td>0.043200</td>\n",
       "      <td>0.956698</td>\n",
       "      <td>02:04</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>18</td>\n",
       "      <td>0.006302</td>\n",
       "      <td>0.347418</td>\n",
       "      <td>0.043200</td>\n",
       "      <td>0.956748</td>\n",
       "      <td>02:03</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>19</td>\n",
       "      <td>0.012326</td>\n",
       "      <td>0.315531</td>\n",
       "      <td>0.039200</td>\n",
       "      <td>0.960884</td>\n",
       "      <td>02:03</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>20</td>\n",
       "      <td>0.011519</td>\n",
       "      <td>0.297215</td>\n",
       "      <td>0.040000</td>\n",
       "      <td>0.960052</td>\n",
       "      <td>02:03</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>21</td>\n",
       "      <td>0.006773</td>\n",
       "      <td>0.292377</td>\n",
       "      <td>0.040000</td>\n",
       "      <td>0.960035</td>\n",
       "      <td>02:02</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>22</td>\n",
       "      <td>0.002509</td>\n",
       "      <td>0.316566</td>\n",
       "      <td>0.041600</td>\n",
       "      <td>0.958313</td>\n",
       "      <td>02:02</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>23</td>\n",
       "      <td>0.001472</td>\n",
       "      <td>0.329549</td>\n",
       "      <td>0.040000</td>\n",
       "      <td>0.960050</td>\n",
       "      <td>02:03</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>24</td>\n",
       "      <td>0.001645</td>\n",
       "      <td>0.318443</td>\n",
       "      <td>0.040000</td>\n",
       "      <td>0.960028</td>\n",
       "      <td>02:02</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>25</td>\n",
       "      <td>0.001688</td>\n",
       "      <td>0.309815</td>\n",
       "      <td>0.040800</td>\n",
       "      <td>0.959105</td>\n",
       "      <td>02:02</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>26</td>\n",
       "      <td>0.001295</td>\n",
       "      <td>0.310550</td>\n",
       "      <td>0.036800</td>\n",
       "      <td>0.963097</td>\n",
       "      <td>02:02</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>27</td>\n",
       "      <td>0.001796</td>\n",
       "      <td>0.335744</td>\n",
       "      <td>0.036800</td>\n",
       "      <td>0.963053</td>\n",
       "      <td>02:02</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>28</td>\n",
       "      <td>0.000724</td>\n",
       "      <td>0.297071</td>\n",
       "      <td>0.036800</td>\n",
       "      <td>0.963244</td>\n",
       "      <td>02:02</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>29</td>\n",
       "      <td>0.008862</td>\n",
       "      <td>0.306059</td>\n",
       "      <td>0.037600</td>\n",
       "      <td>0.962425</td>\n",
       "      <td>02:03</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>30</td>\n",
       "      <td>0.008014</td>\n",
       "      <td>0.308906</td>\n",
       "      <td>0.034400</td>\n",
       "      <td>0.965543</td>\n",
       "      <td>02:03</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>31</td>\n",
       "      <td>0.003169</td>\n",
       "      <td>0.303726</td>\n",
       "      <td>0.035200</td>\n",
       "      <td>0.964777</td>\n",
       "      <td>02:03</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, lrs)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 0,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "M6w2Y15giILh"
   },
   "outputs": [],
   "source": [
    "learn.save('ResNet50_299_964777')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 0,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "6WHT6Teg4tJJ"
   },
   "outputs": [],
   "source": [
    "learn.load('ResNet50_299_964777');"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 0,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "lN0RisfszCpI"
   },
   "outputs": [],
   "source": [
    "preds, _ = learn.get_preds(DatasetType.Test)\n",
    "pred_class = torch.argmax(preds, dim=1)\n",
    "pred_class += 1\n",
    "pred_class = pred_class.numpy()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 0,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "HJK72K9q5EcN"
   },
   "outputs": [],
   "source": [
    "write_submission(pred_class, 'ResNet50_124t256t299_964777.csv')\n",
    "#LB 0.9812763721"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "colab_type": "text",
    "id": "jnOkT_SppIIs"
   },
   "source": [
    "## Getting predicted probabilties for Test Set"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 0,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "YrHfH0QX0URL"
   },
   "outputs": [],
   "source": [
    "learn.load('ResNet50_299_964777');"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 0,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "2MnPBlu-pYIU"
   },
   "outputs": [],
   "source": [
    "class_name = ['Cargo', 'Military','Carrier' ,'Cruise','Tankers']"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 0,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "T9BnRxo50hi3"
   },
   "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_ResNet50_LB_98127.csv', index=False)\n",
    "predTTA_df.to_csv('TTA_ResNet50_LB_98127.csv', index=False)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 0,
   "metadata": {
    "colab": {},
    "colab_type": "code",
    "id": "al4eqpSI2TBT"
   },
   "outputs": [],
   "source": [
    "!head ResNet50_LB_98127.csv"
   ]
  }
 ],
 "metadata": {
  "accelerator": "GPU",
  "colab": {
   "collapsed_sections": [],
   "name": "Final_AV_GODLResNet50v1.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
}