Fast.ai/[Kaggle] (Fast.ai) MNIST Digit Recognizer Kaggle.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [],
   "source": [
    "%reload_ext autoreload\n",
    "%autoreload 2\n",
    "%matplotlib inline"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [],
   "source": [
    "from fastai.vision import *\n",
    "import pandas as pd\n",
    "import numpy as np\n",
    "import os\n",
    "from PIL import Image\n",
    "import matplotlib.pyplot as plt\n",
    "from pathlib import Path"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "True True\n"
     ]
    }
   ],
   "source": [
    "print(torch.cuda.is_available(), torch.backends.cudnn.enabled)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [],
   "source": [
    "#  Fast.ai only accepts image files for Computer Vision, but we have CSV files containing the pixel values for each pixel of each image. \n",
    "# If we want to use fast.ai we have to create images from the data we have."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [],
   "source": [
    "# 1. Let's first create the folder structure!"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [],
   "source": [
    "path = Path('C:\\\\Users\\\\bokhy\\\\Desktop\\\\kaggle\\\\digit-recognizer\\\\')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "WindowsPath('C:/Users/bokhy/Desktop/kaggle/digit-recognizer')"
      ]
     },
     "execution_count": 5,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "path"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [],
   "source": [
    "train_df = pd.read_csv(os.path.join(path, 'train.csv'))\n",
    "test_df = pd.read_csv(os.path.join(path, 'test.csv'))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {},
   "outputs": [],
   "source": [
    "TRAIN = os.path.join(path, 'train')\n",
    "TEST = os.path.join(path, 'test')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Create Train directory\n",
    "for index in range(10):\n",
    "    a = os.path.join(TRAIN,str(index))\n",
    "    os.makedirs(str(a), exist_ok =True)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Create Test directory\n",
    "for index in range(10):\n",
    "    a = TEST\n",
    "    os.makedirs(str(a), exist_ok =True)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {},
   "outputs": [],
   "source": [
    "# import numpy to reshape array from flat (1x784) to square (28x28)\n",
    "def saveDigit(digit, filepath):\n",
    "    digit = digit.reshape(28,28)\n",
    "    digit = digit.astype(np.uint8)\n",
    "\n",
    "    img = Image.fromarray(digit)\n",
    "    img.save(filepath)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Generate Train image using above func\n",
    "for index, row in train_df.iterrows():\n",
    "    \n",
    "    label,digit = row[0], row[1:]\n",
    "    \n",
    "    folder = os.path.join(TRAIN,str(label))\n",
    "    filename = f\"{index}.jpg\"\n",
    "    filepath = Path(folder)/Path(filename)\n",
    "    \n",
    "    digit = digit.values\n",
    "    \n",
    "    saveDigit(digit, filepath)    "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Generate Test image using above func\n",
    "for index, digit in test_df.iterrows():\n",
    "\n",
    "    folder = TEST\n",
    "    filename = f\"{index}.jpg\"\n",
    "    filepath = Path(folder)/Path(filename)\n",
    "    \n",
    "    digit = digit.values\n",
    "    \n",
    "    saveDigit(digit, filepath)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "# (Optional)\n",
    "# Display images\n",
    "# import matplotlib to arrange the images properly\n",
    "\n",
    "def displayData():\n",
    "    fig = plt.figure(figsize=(5,10))\n",
    "    \n",
    "    for rowIndex in range(1, 10):\n",
    "        subdirectory = str(rowIndex)\n",
    "        path = TRAIN/subdirectory\n",
    "        images = os.listdir(path)\n",
    "        for sampleIndex in range(1, 6):\n",
    "            randomNumber = random.randint(0, len(images)-1)\n",
    "            image = Image.open(path/images[randomNumber])\n",
    "            ax = fig.add_subplot(10, 5, 5*rowIndex + sampleIndex)\n",
    "            ax.axis(\"off\")\n",
    "            \n",
    "            plt.imshow(image, cmap='gray')\n",
    "        \n",
    "    plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "displayData()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "tfms = ([*rand_pad(padding=5, size=28, mode='zeros')], [])\n",
    "# padding 5"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {},
   "outputs": [],
   "source": [
    "tfms = get_transforms(do_flip=False, max_rotate=20,flip_vert=False)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {},
   "outputs": [],
   "source": [
    "#batch size\n",
    "bs = 256"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {},
   "outputs": [],
   "source": [
    "data = ImageDataBunch.from_folder(\n",
    "    path = path,\n",
    "    train = TRAIN,\n",
    "    test = TEST,\n",
    "    valid_pct = 0.1,\n",
    "    bs = bs,\n",
    "    size = 28,\n",
    "    num_workers = 0,\n",
    "    ds_tfms = tfms\n",
    ")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {},
   "outputs": [],
   "source": [
    "data = data.normalize(imagenet_stats)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Downloading: \"https://download.pytorch.org/models/resnext50_32x4d-7cdf4587.pth\" to C:\\Users\\bokhy/.cache\\torch\\checkpoints\\resnext50_32x4d-7cdf4587.pth\n",
      "100%|█████████████████████████████████████████████████████████████████████████████| 95.8M/95.8M [00:05<00:00, 18.4MB/s]\n"
     ]
    }
   ],
   "source": [
    "import torchvision.models as models\n",
    "# resnet18 = models.resnet18(pretrained=True)\n",
    "# alexnet = models.alexnet(pretrained=True)\n",
    "# squeezenet = models.squeezenet1_0(pretrained=True)\n",
    "# vgg16 = models.vgg16(pretrained=True)\n",
    "# densenet = models.densenet161(pretrained=True)\n",
    "# inception = models.inception_v3(pretrained=True)\n",
    "# googlenet = models.googlenet(pretrained=True)\n",
    "# shufflenet = models.shufflenet_v2_x1_0(pretrained=True)\n",
    "# mobilenet = models.mobilenet_v2(pretrained=True)\n",
    "resnext50_32x4d = models.resnext50_32x4d(pretrained=True)\n",
    "# wide_resnet50_2 = models.wide_resnet50_2(pretrained=True)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "Downloading: \"https://download.pytorch.org/models/resnet101-5d3b4d8f.pth\" to C:\\Users\\bokhy/.cache\\torch\\checkpoints\\resnet101-5d3b4d8f.pth\n",
      "100%|███████████████████████████████████████████████████████████████████████████████| 170M/170M [00:10<00:00, 17.8MB/s]\n"
     ]
    }
   ],
   "source": [
    "learn = cnn_learner(data, models.resnet101, loss_func = nn.CrossEntropyLoss(), metrics=accuracy, callback_fns=ShowGraph, ps = 0.3)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "\n",
       "    <div>\n",
       "        <style>\n",
       "            /* Turns off some styling */\n",
       "            progress {\n",
       "                /* gets rid of default border in Firefox and Opera. */\n",
       "                border: none;\n",
       "                /* Needs to be in here for Safari polyfill so background images work as expected. */\n",
       "                background-size: auto;\n",
       "            }\n",
       "            .progress-bar-interrupted, .progress-bar-interrupted::-webkit-progress-bar {\n",
       "                background: #F44336;\n",
       "            }\n",
       "        </style>\n",
       "      <progress value='0' class='' max='1', style='width:300px; height:20px; vertical-align: middle;'></progress>\n",
       "      0.00% [0/1 00:00<00:00]\n",
       "    </div>\n",
       "    \n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: left;\">\n",
       "      <th>epoch</th>\n",
       "      <th>train_loss</th>\n",
       "      <th>valid_loss</th>\n",
       "      <th>accuracy</th>\n",
       "      <th>time</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "  </tbody>\n",
       "</table><p>\n",
       "\n",
       "    <div>\n",
       "        <style>\n",
       "            /* Turns off some styling */\n",
       "            progress {\n",
       "                /* gets rid of default border in Firefox and Opera. */\n",
       "                border: none;\n",
       "                /* Needs to be in here for Safari polyfill so background images work as expected. */\n",
       "                background-size: auto;\n",
       "            }\n",
       "            .progress-bar-interrupted, .progress-bar-interrupted::-webkit-progress-bar {\n",
       "                background: #F44336;\n",
       "            }\n",
       "        </style>\n",
       "      <progress value='79' class='' max='147', style='width:300px; height:20px; vertical-align: middle;'></progress>\n",
       "      53.74% [79/147 01:21<01:10 6.2153]\n",
       "    </div>\n",
       "    "
      ],
      "text/plain": [
       "<IPython.core.display.HTML object>"
      ]
     },
     "metadata": {},
     "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": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "learn.lr_find(wd = 0.1, end_lr=100)\n",
    "learn.recorder.plot()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {},
   "outputs": [],
   "source": [
    "lr = 1e-2"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "metadata": {},
   "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>accuracy</th>\n",
       "      <th>time</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <td>0</td>\n",
       "      <td>0.804525</td>\n",
       "      <td>0.517236</td>\n",
       "      <td>0.836905</td>\n",
       "      <td>02:28</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>1</td>\n",
       "      <td>0.335518</td>\n",
       "      <td>0.170585</td>\n",
       "      <td>0.944048</td>\n",
       "      <td>04:42</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>2</td>\n",
       "      <td>0.173899</td>\n",
       "      <td>0.114672</td>\n",
       "      <td>0.960476</td>\n",
       "      <td>03:24</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>3</td>\n",
       "      <td>0.122084</td>\n",
       "      <td>0.073360</td>\n",
       "      <td>0.975238</td>\n",
       "      <td>03:09</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>4</td>\n",
       "      <td>0.101389</td>\n",
       "      <td>0.067889</td>\n",
       "      <td>0.979286</td>\n",
       "      <td>03:11</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>"
      ],
      "text/plain": [
       "<IPython.core.display.HTML object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "learn.fit_one_cycle(5,slice(lr))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "\n",
       "    <div>\n",
       "        <style>\n",
       "            /* Turns off some styling */\n",
       "            progress {\n",
       "                /* gets rid of default border in Firefox and Opera. */\n",
       "                border: none;\n",
       "                /* Needs to be in here for Safari polyfill so background images work as expected. */\n",
       "                background-size: auto;\n",
       "            }\n",
       "            .progress-bar-interrupted, .progress-bar-interrupted::-webkit-progress-bar {\n",
       "                background: #F44336;\n",
       "            }\n",
       "        </style>\n",
       "      <progress value='0' class='' max='1', style='width:300px; height:20px; vertical-align: middle;'></progress>\n",
       "      0.00% [0/1 00:00<00:00]\n",
       "    </div>\n",
       "    \n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: left;\">\n",
       "      <th>epoch</th>\n",
       "      <th>train_loss</th>\n",
       "      <th>valid_loss</th>\n",
       "      <th>accuracy</th>\n",
       "      <th>time</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "  </tbody>\n",
       "</table><p>\n",
       "\n",
       "    <div>\n",
       "        <style>\n",
       "            /* Turns off some styling */\n",
       "            progress {\n",
       "                /* gets rid of default border in Firefox and Opera. */\n",
       "                border: none;\n",
       "                /* Needs to be in here for Safari polyfill so background images work as expected. */\n",
       "                background-size: auto;\n",
       "            }\n",
       "            .progress-bar-interrupted, .progress-bar-interrupted::-webkit-progress-bar {\n",
       "                background: #F44336;\n",
       "            }\n",
       "        </style>\n",
       "      <progress value='76' class='' max='147', style='width:300px; height:20px; vertical-align: middle;'></progress>\n",
       "      51.70% [76/147 01:42<01:36 0.3644]\n",
       "    </div>\n",
       "    "
      ],
      "text/plain": [
       "<IPython.core.display.HTML object>"
      ]
     },
     "metadata": {},
     "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": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "learn.unfreeze()\n",
    "learn.lr_find()\n",
    "learn.recorder.plot()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "22"
      ]
     },
     "execution_count": 18,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "gc.collect()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "metadata": {},
   "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>accuracy</th>\n",
       "      <th>time</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <td>0</td>\n",
       "      <td>0.089270</td>\n",
       "      <td>0.048607</td>\n",
       "      <td>0.983333</td>\n",
       "      <td>04:18</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>1</td>\n",
       "      <td>0.076222</td>\n",
       "      <td>0.047641</td>\n",
       "      <td>0.982619</td>\n",
       "      <td>03:33</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>2</td>\n",
       "      <td>0.062781</td>\n",
       "      <td>0.039198</td>\n",
       "      <td>0.985952</td>\n",
       "      <td>03:26</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>3</td>\n",
       "      <td>0.064511</td>\n",
       "      <td>0.035472</td>\n",
       "      <td>0.989048</td>\n",
       "      <td>04:05</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>4</td>\n",
       "      <td>0.063453</td>\n",
       "      <td>0.092701</td>\n",
       "      <td>0.988333</td>\n",
       "      <td>03:30</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>5</td>\n",
       "      <td>0.050543</td>\n",
       "      <td>0.032899</td>\n",
       "      <td>0.990476</td>\n",
       "      <td>03:09</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>6</td>\n",
       "      <td>0.051713</td>\n",
       "      <td>0.035435</td>\n",
       "      <td>0.991190</td>\n",
       "      <td>03:21</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>7</td>\n",
       "      <td>0.043912</td>\n",
       "      <td>0.028841</td>\n",
       "      <td>0.992857</td>\n",
       "      <td>04:53</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>8</td>\n",
       "      <td>0.043220</td>\n",
       "      <td>0.025698</td>\n",
       "      <td>0.993810</td>\n",
       "      <td>04:05</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>9</td>\n",
       "      <td>0.038342</td>\n",
       "      <td>0.024969</td>\n",
       "      <td>0.993571</td>\n",
       "      <td>03:02</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>10</td>\n",
       "      <td>0.032538</td>\n",
       "      <td>0.025083</td>\n",
       "      <td>0.994524</td>\n",
       "      <td>03:19</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>11</td>\n",
       "      <td>0.043109</td>\n",
       "      <td>0.020553</td>\n",
       "      <td>0.993810</td>\n",
       "      <td>03:16</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>12</td>\n",
       "      <td>0.029259</td>\n",
       "      <td>0.021398</td>\n",
       "      <td>0.994286</td>\n",
       "      <td>03:01</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>13</td>\n",
       "      <td>0.021620</td>\n",
       "      <td>0.023580</td>\n",
       "      <td>0.993810</td>\n",
       "      <td>02:55</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>14</td>\n",
       "      <td>0.023597</td>\n",
       "      <td>0.025134</td>\n",
       "      <td>0.992857</td>\n",
       "      <td>02:52</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>15</td>\n",
       "      <td>0.019300</td>\n",
       "      <td>0.018637</td>\n",
       "      <td>0.994762</td>\n",
       "      <td>03:12</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>16</td>\n",
       "      <td>0.017748</td>\n",
       "      <td>0.022919</td>\n",
       "      <td>0.994048</td>\n",
       "      <td>03:24</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>17</td>\n",
       "      <td>0.013322</td>\n",
       "      <td>0.021807</td>\n",
       "      <td>0.993571</td>\n",
       "      <td>03:05</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>18</td>\n",
       "      <td>0.010617</td>\n",
       "      <td>0.018928</td>\n",
       "      <td>0.995238</td>\n",
       "      <td>03:04</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>19</td>\n",
       "      <td>0.010171</td>\n",
       "      <td>0.018904</td>\n",
       "      <td>0.995000</td>\n",
       "      <td>02:55</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>20</td>\n",
       "      <td>0.007707</td>\n",
       "      <td>0.021703</td>\n",
       "      <td>0.994762</td>\n",
       "      <td>03:12</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>21</td>\n",
       "      <td>0.007378</td>\n",
       "      <td>0.020679</td>\n",
       "      <td>0.995238</td>\n",
       "      <td>03:14</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>22</td>\n",
       "      <td>0.008851</td>\n",
       "      <td>0.020395</td>\n",
       "      <td>0.995476</td>\n",
       "      <td>03:01</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>23</td>\n",
       "      <td>0.005944</td>\n",
       "      <td>0.020073</td>\n",
       "      <td>0.995238</td>\n",
       "      <td>02:56</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>24</td>\n",
       "      <td>0.006675</td>\n",
       "      <td>0.020007</td>\n",
       "      <td>0.995476</td>\n",
       "      <td>03:33</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>"
      ],
      "text/plain": [
       "<IPython.core.display.HTML object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "#learn.fit_one_cycle(25, slice(1e-5,lr/5),moms=(0.8,0.7))\n",
    "learn.fit_one_cycle(25, slice(1e-4,lr/5))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 20,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [],
      "text/plain": [
       "<IPython.core.display.HTML object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x432 with 9 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "interp = ClassificationInterpretation.from_learner(learn)\n",
    "interp.plot_top_losses(9,figsize=(6,6))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 21,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [],
      "text/plain": [
       "<IPython.core.display.HTML object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "pred_test, y = learn.get_preds(DatasetType.Test)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 41,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "tensor([[2.8859e-12, 4.6895e-14, 1.0000e+00,  ..., 6.7330e-14, 1.4200e-15,\n",
       "         9.4399e-13],\n",
       "        [1.0000e+00, 5.8600e-09, 2.1683e-08,  ..., 1.3907e-08, 8.9833e-08,\n",
       "         1.1618e-07],\n",
       "        [9.0467e-08, 3.8581e-08, 2.5061e-07,  ..., 2.3829e-07, 2.5707e-06,\n",
       "         1.8032e-06],\n",
       "        ...,\n",
       "        [3.1251e-05, 1.3662e-05, 9.9985e-01,  ..., 4.8878e-05, 4.2385e-06,\n",
       "         1.1131e-05],\n",
       "        [4.0457e-05, 3.2701e-04, 1.1567e-04,  ..., 9.9752e-01, 5.2869e-05,\n",
       "         1.4400e-03],\n",
       "        [6.3982e-08, 1.3844e-06, 1.0351e-06,  ..., 9.9999e-01, 5.9183e-08,\n",
       "         1.4659e-06]])"
      ]
     },
     "execution_count": 41,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "pred_test"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 42,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "tensor([0, 0, 0,  ..., 0, 0, 0])"
      ]
     },
     "execution_count": 42,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "y"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 43,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "['0: 9.046744509078053e-08',\n",
       " '1: 3.858092867403684e-08',\n",
       " '2: 2.5060654706976493e-07',\n",
       " '3: 1.2163687642896548e-05',\n",
       " '4: 2.6341792747075488e-08',\n",
       " '5: 0.9999821186065674',\n",
       " '6: 7.235487373691285e-07',\n",
       " '7: 2.382861055139074e-07',\n",
       " '8: 2.5706888209242607e-06',\n",
       " '9: 1.8032319530902896e-06']"
      ]
     },
     "execution_count": 43,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "probabilities = pred_test[0].tolist() # look at the predict for the very first row\n",
    "[f\"{index}: {probabilities[index]}\" for index in range(len(probabilities))]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 22,
   "metadata": {},
   "outputs": [],
   "source": [
    "test_result = torch.argmax(pred_test,dim=1)\n",
    "pred_test = test_result.numpy()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 23,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "28000"
      ]
     },
     "execution_count": 23,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "len(pred_test)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 24,
   "metadata": {},
   "outputs": [],
   "source": [
    "# remove file extension from filename\n",
    "ImageId = [os.path.splitext(path)[0] for path in os.listdir(TEST)]\n",
    "# typecast to int so that file can be sorted by ImageId\n",
    "ImageId = [int(path) for path in ImageId]\n",
    "# +1 because index starts at 1 in the submission file\n",
    "ImageId = [ID+1 for ID in ImageId]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 25,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "28000"
      ]
     },
     "execution_count": 25,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "len(ImageId)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 26,
   "metadata": {},
   "outputs": [],
   "source": [
    "submission  = pd.DataFrame({\n",
    "    \"ImageId\": ImageId,\n",
    "    \"Label\": pred_test\n",
    "})\n",
    "submission.sort_values(by=[\"ImageId\"], inplace = True)\n",
    "submission.to_csv(\"MNIST_final.csv\", index=False)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 24,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>ImageId</th>\n",
       "      <th>Label</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>1</td>\n",
       "      <td>2</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>2</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>11112</th>\n",
       "      <td>3</td>\n",
       "      <td>9</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>20223</th>\n",
       "      <td>4</td>\n",
       "      <td>0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>21334</th>\n",
       "      <td>5</td>\n",
       "      <td>3</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "       ImageId  Label\n",
       "0            1      2\n",
       "1            2      0\n",
       "11112        3      9\n",
       "20223        4      0\n",
       "21334        5      3"
      ]
     },
     "execution_count": 24,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "submission.head()"
   ]
  }
 ],
 "metadata": {
  "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": "3.7.6"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 4
}