obscene-numbers.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [],
   "source": [
    "from fastai.text.all import *\n",
    "import PIL.Image\n",
    "from IPython.display import display"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [],
   "source": [
    "human100k = [s.strip() for s in open('./one-hundred-thousand-numbers.txt').readlines()]\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [],
   "source": [
    "obscene = f\"{23 * 3}\""
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [],
   "source": [
    "nonobscene_indices = [i for i in range(len(human100k)) if not re.search(obscene, f\"{i}\")]\n",
    "obscene_indices    = [i for i in range(len(human100k)) if     re.search(obscene, f\"{i}\")]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [],
   "source": [
    "nonobscene_numbers = [human100k[i] for i in nonobscene_indices]\n",
    "obscene_numbers = [human100k[i] for i in obscene_indices]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [],
   "source": [
    "nndf = pd.DataFrame(data={\"txt\": nonobscene_numbers})\n",
    "ondf = pd.DataFrame(data={\"txt\": obscene_numbers})"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "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>txt</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>zero</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>one</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>two</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>three</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>four</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>...</th>\n",
       "      <td>...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>96025</th>\n",
       "      <td>ninety nine thousand nine hundred ninety five</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>96026</th>\n",
       "      <td>ninety nine thousand nine hundred ninety six</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>96027</th>\n",
       "      <td>ninety nine thousand nine hundred ninety seven</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>96028</th>\n",
       "      <td>ninety nine thousand nine hundred ninety eight</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>96029</th>\n",
       "      <td>ninety nine thousand nine hundred ninety nine</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "<p>96030 rows × 1 columns</p>\n",
       "</div>"
      ],
      "text/plain": [
       "                                                  txt\n",
       "0                                                zero\n",
       "1                                                 one\n",
       "2                                                 two\n",
       "3                                               three\n",
       "4                                                four\n",
       "...                                               ...\n",
       "96025   ninety nine thousand nine hundred ninety five\n",
       "96026    ninety nine thousand nine hundred ninety six\n",
       "96027  ninety nine thousand nine hundred ninety seven\n",
       "96028  ninety nine thousand nine hundred ninety eight\n",
       "96029   ninety nine thousand nine hundred ninety nine\n",
       "\n",
       "[96030 rows x 1 columns]"
      ]
     },
     "execution_count": 7,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "nndf"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "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>txt</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>sixty nine</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>one hundred sixty nine</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>two hundred sixty nine</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>three hundred sixty nine</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>four hundred sixty nine</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>...</th>\n",
       "      <td>...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3965</th>\n",
       "      <td>ninety nine thousand six hundred ninety eight</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3966</th>\n",
       "      <td>ninety nine thousand six hundred ninety nine</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3967</th>\n",
       "      <td>ninety nine thousand seven hundred sixty nine</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3968</th>\n",
       "      <td>ninety nine thousand eight hundred sixty nine</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3969</th>\n",
       "      <td>ninety nine thousand nine hundred sixty nine</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "<p>3970 rows × 1 columns</p>\n",
       "</div>"
      ],
      "text/plain": [
       "                                                txt\n",
       "0                                        sixty nine\n",
       "1                            one hundred sixty nine\n",
       "2                            two hundred sixty nine\n",
       "3                          three hundred sixty nine\n",
       "4                           four hundred sixty nine\n",
       "...                                             ...\n",
       "3965  ninety nine thousand six hundred ninety eight\n",
       "3966   ninety nine thousand six hundred ninety nine\n",
       "3967  ninety nine thousand seven hundred sixty nine\n",
       "3968  ninety nine thousand eight hundred sixty nine\n",
       "3969   ninety nine thousand nine hundred sixty nine\n",
       "\n",
       "[3970 rows x 1 columns]"
      ]
     },
     "execution_count": 8,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "ondf"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {},
   "outputs": [],
   "source": [
    "ndb = DataBlock(blocks=TextBlock.from_df('txt',is_lm=True,seq_len=8),get_x=ColReader('text'),splitter=RandomSplitter(0.1))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {},
   "outputs": [],
   "source": [
    "bs=128"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [],
      "text/plain": [
       "<IPython.core.display.HTML object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "nndl = ndb.dataloaders(nndf, bs=bs)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "['xxunk',\n",
       " 'xxpad',\n",
       " 'xxbos',\n",
       " 'xxeos',\n",
       " 'xxfld',\n",
       " 'xxrep',\n",
       " 'xxwrep',\n",
       " 'xxup',\n",
       " 'xxmaj',\n",
       " 'thousand',\n",
       " 'hundred',\n",
       " 'one',\n",
       " 'two',\n",
       " 'three',\n",
       " 'four',\n",
       " 'five',\n",
       " 'seven',\n",
       " 'eight',\n",
       " 'six',\n",
       " 'nine',\n",
       " 'twenty',\n",
       " 'thirty',\n",
       " 'forty',\n",
       " 'fifty',\n",
       " 'seventy',\n",
       " 'eighty',\n",
       " 'ninety',\n",
       " 'sixty',\n",
       " 'ten',\n",
       " 'eleven',\n",
       " 'twelve',\n",
       " 'thirteen',\n",
       " 'fourteen',\n",
       " 'fifteen',\n",
       " 'seventeen',\n",
       " 'eighteen',\n",
       " 'nineteen',\n",
       " 'sixteen',\n",
       " 'xxfake',\n",
       " 'xxfake']"
      ]
     },
     "execution_count": 12,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "nndl.vocab"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [],
      "text/plain": [
       "<IPython.core.display.HTML object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "ondl = DataBlock(blocks=TextBlock.from_df('txt',is_lm=True,seq_len=8,vocab=nndl.vocab),get_x=ColReader('text'),splitter=RandomSplitter(0.1)).dataloaders(ondf, bs=bs)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>text</th>\n",
       "      <th>text_</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>xxbos fifty seven thousand seven hundred ninety three</td>\n",
       "      <td>fifty seven thousand seven hundred ninety three xxbos</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>twenty five thousand six hundred seven xxbos twenty</td>\n",
       "      <td>five thousand six hundred seven xxbos twenty three</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>sixty eight xxbos sixty four thousand one hundred</td>\n",
       "      <td>eight xxbos sixty four thousand one hundred twenty</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>seventy one thousand nine hundred seventy four xxbos</td>\n",
       "      <td>one thousand nine hundred seventy four xxbos thirty</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>thousand one hundred five xxbos forty five thousand</td>\n",
       "      <td>one hundred five xxbos forty five thousand one</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>5</th>\n",
       "      <td>xxbos eighty eight thousand twenty five xxbos ninety</td>\n",
       "      <td>eighty eight thousand twenty five xxbos ninety seven</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>6</th>\n",
       "      <td>hundred fifteen xxbos fifteen thousand three hundred four</td>\n",
       "      <td>fifteen xxbos fifteen thousand three hundred four xxbos</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>7</th>\n",
       "      <td>thousand seven hundred forty one xxbos eleven thousand</td>\n",
       "      <td>seven hundred forty one xxbos eleven thousand three</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>8</th>\n",
       "      <td>one thousand nine hundred seventy two xxbos ninety</td>\n",
       "      <td>thousand nine hundred seventy two xxbos ninety seven</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>"
      ],
      "text/plain": [
       "<IPython.core.display.HTML object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "nndl.show_batch()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>text</th>\n",
       "      <th>text_</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>xxbos thirty six thousand nine hundred sixty five</td>\n",
       "      <td>thirty six thousand nine hundred sixty five xxbos</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>forty four xxbos sixteen thousand nine hundred ninety</td>\n",
       "      <td>four xxbos sixteen thousand nine hundred ninety eight</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>thousand six hundred ninety eight xxbos sixty nine</td>\n",
       "      <td>six hundred ninety eight xxbos sixty nine thousand</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>xxbos sixty nine thousand five hundred eleven xxbos</td>\n",
       "      <td>sixty nine thousand five hundred eleven xxbos fifty</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>ninety six thousand nine hundred fifty seven xxbos</td>\n",
       "      <td>six thousand nine hundred fifty seven xxbos forty</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>5</th>\n",
       "      <td>nine xxbos sixty nine thousand two hundred thirty</td>\n",
       "      <td>xxbos sixty nine thousand two hundred thirty xxbos</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>6</th>\n",
       "      <td>ninety two xxbos forty six thousand nine hundred</td>\n",
       "      <td>two xxbos forty six thousand nine hundred sixty</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>7</th>\n",
       "      <td>nine thousand three hundred sixty nine xxbos thirteen</td>\n",
       "      <td>thousand three hundred sixty nine xxbos thirteen thousand</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>8</th>\n",
       "      <td>eighty one thousand eight hundred sixty nine xxbos</td>\n",
       "      <td>one thousand eight hundred sixty nine xxbos fifty</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>"
      ],
      "text/plain": [
       "<IPython.core.display.HTML object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "ondl.show_batch()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "metadata": {},
   "outputs": [],
   "source": [
    "dataloader_to_lm_learner = lambda dl: language_model_learner(dl, AWD_LSTM, drop_mult=0.3, metrics=[accuracy, Perplexity()]).to_fp16()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "metadata": {},
   "outputs": [],
   "source": [
    "one_pred = lambda lm,w: np.array(lm.get_preds(dl=[((lm.dls.test_dl([w], num_workers=0).items[0].to(lm.dls.device))[None],)])[0][0][-1])\n",
    "all_preds_after_sixty = lambda lm: one_pred(lm, \"sixty\")\n",
    "all_preds_after_fifty = lambda lm: one_pred(lm, \"fifty\")\n",
    "text_preds_histo = lambda lm,w,t,c: np.unique([lm.predict(w, 1, temperature=t, no_bar=True) for i in range(c)], return_counts=True)\n",
    "pred_histo_after_sixty = lambda lm,t,c: text_preds_histo(lm,\"sixty\",t,c)\n",
    "pred_histo_after_fifty = lambda lm,t,c: text_preds_histo(lm,\"fifty\",t,c)\n",
    "\n",
    "float_array_to_image = lambda a: PIL.Image.fromarray((255 * a).astype(np.uint8),'L').resize((400,180), resample=PIL.Image.NEAREST)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "metadata": {},
   "outputs": [],
   "source": [
    "lml = dataloader_to_lm_learner(nndl)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [],
      "text/plain": [
       "<IPython.core.display.HTML object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [],
      "text/plain": [
       "<IPython.core.display.HTML object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "preds_after_sixty = [all_preds_after_sixty(lml)]\n",
    "preds_after_fifty = [all_preds_after_fifty(lml)]\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 20,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "Path('models/nnonlml.pth')"
      ]
     },
     "execution_count": 20,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "lml.save('nnonlml')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 21,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [],
      "text/plain": [
       "<IPython.core.display.HTML object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/plain": [
       "SuggestedLRs(lr_min=0.19054607152938843, lr_steep=0.03981071710586548)"
      ]
     },
     "execution_count": 21,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "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": [
    "lml.lr_find()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 22,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "<fastai.text.learner.LMLearner at 0x7f37805a41d0>"
      ]
     },
     "execution_count": 22,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "lml.load('nnonlml')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 23,
   "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>perplexity</th>\n",
       "      <th>time</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <td>0</td>\n",
       "      <td>1.695428</td>\n",
       "      <td>1.614012</td>\n",
       "      <td>0.436691</td>\n",
       "      <td>5.022924</td>\n",
       "      <td>02:48</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>"
      ],
      "text/plain": [
       "<IPython.core.display.HTML object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [],
      "text/plain": [
       "<IPython.core.display.HTML object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [],
      "text/plain": [
       "<IPython.core.display.HTML object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "sixty\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<PIL.Image.Image image mode=L size=400x180 at 0x7F37805A7B90>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "fifty\n"
     ]
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAZAAAAC0CAAAAACfniWEAAACYklEQVR4nO3TO25TURhGUTuxgiKBmAANY2AodJlB2kwCiRbR0jMWBkFPQYV4SHlQ5T9FjmjZxVrVp/vytbZ9fHd4dDPr9ayvh6e+b459mfVh1qfNdd9m/Zh1Mety1sOs06xns/5s7vg56/mss3+u9Rnns+4277feYPdWu3uPm7P3m7MvD0+dbY7xHwkSI0iMIDGCxAgSI0iMIDGCxAgSI0iMIDGCxAgSI0iMIDGCxAgSI0iMIDGCxAgSI0iMIDGCxAgSI0iMIDGCxAgSI0iMIDGCxAgSI0iMIDGCxAgSI0iMIDGCxAgSI0iMIDGCxAgSI0iMIDGCxAgSI0iMIDGCxAgSI0iMIDGCxAgSI0iMIDGCxAgSI0iMIDGCxAgSI0iMIDGCxAgSI0iMIDGCxAgSI0iMIDGCxAgSI0iMIDGCxAgSI0iMIDGCxAgSI0iMIDGCxJzezLyf9XFzbDmfdTHr1ay3s65nvZ/1efO84+Z5D7N+ba5bv6LTrKtZvzfXre+xnny3ecq648Ws283Z9S7ryafN2d11y+7f4B8SI0iMIDGCxAgSI0iMIDGCxAgSI0iMIDGCxAgSI0iMIDGCxAgSI0iMIDGCxAgSI0iMIDGCxAgSI0iMIDGCxAgSI0iMIDGCxAgSI0iMIDGCxAgSI0iMIDGCxAgSI0iMIDGCxAgSI0iMIDGCxAgSI0iMIDGCxAgSI0iMIDGCxAgSI0iMIDGCxAgSI0iMIDGCxAgSI0iMIDGCxAgSI0iMIDGCxAgSI0iMIDGCxAgSI0iMIDGCxAgSI0iMIDGCxAgSI0iMIDGCxAgS8xd08x+aeFqe3QAAAABJRU5ErkJggg==\n",
      "text/plain": [
       "<PIL.Image.Image image mode=L size=400x180 at 0x7F378F7E6310>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: left;\">\n",
       "      <th>epoch</th>\n",
       "      <th>train_loss</th>\n",
       "      <th>valid_loss</th>\n",
       "      <th>accuracy</th>\n",
       "      <th>perplexity</th>\n",
       "      <th>time</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <td>0</td>\n",
       "      <td>1.658221</td>\n",
       "      <td>1.606264</td>\n",
       "      <td>0.438860</td>\n",
       "      <td>4.984155</td>\n",
       "      <td>02:48</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>"
      ],
      "text/plain": [
       "<IPython.core.display.HTML object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [],
      "text/plain": [
       "<IPython.core.display.HTML object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [],
      "text/plain": [
       "<IPython.core.display.HTML object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "sixty\n"
     ]
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAZAAAAC0CAAAAACfniWEAAACsklEQVR4nO3cMXKNURyH4YQQhKFRqYxRKexDZRM2obAIlQ3QaZV2otfojCAiKv9T5IzWWzxP9Zvv5vtyk/fe9hy+PPjr1axHsz4dXPZlc+3jrHez3sy6mPV51o9Zx7NubJ58ddb1zb23Zn2ddbK5d62LzTqadT7rdHPv7l2tdf7Pn/u9uXZnc8eVzTX+I0FiBIkRJEaQGEFiBIkRJEaQGEFiBIkRJEaQGEFiBIkRJEaQGEFiBIkRJEaQGEFiBIkRJEaQGEFiBIkRJEaQGEFiBIkRJEaQGEFiBIkRJEaQGEFiBIkRJEaQGEFiBIkRJEaQGEFiBIkRJEaQGEFiBIkRJEaQGEFiBIkRJEaQGEFiBIk5ejZznYX+dnNtWSe0X5v1dNY6of3FrNez3m+et05FvznrcNbpwWXrU3Rv1vNZv2atd/9zc+1s1vo71pNvzzrfvLre38Xm1d1p8Weba9c213xDYgSJESRGkBhBYgSJESRGkBhBYgSJESRGkBhBYgSJESRGkBhBYgSJESRGkBhBYgSJESRGkBhBYgSJESRGkBhBYgSJESRGkBhBYgSJESRGkBhBYgSJESRGkBhBYgSJESRGkBhBYgSJESRGkBhBYgSJESRGkBhBYgSJESRGkBhBYgSJESTm6MnMdZ75/c215WTz6sNZD2Z9mHV31vdZu0/Ct82T1x3rRPV1uvvjWevc+HUa++6E9uV483Prdxxt1rL7v+ys33u8eXX3ZN+QGEFiBIkRJEaQGEFiBIkRJEaQGEFiBIkRJEaQGEFiBIkRJEaQGEFiBIkRJEaQGEFiBIkRJEaQGEFiBIkRJEaQGEFiBIkRJEaQGEFiBIkRJEaQGEFiBIkRJEaQGEFiBIkRJEaQGEFiBIkRJEaQGEFiBIkRJEaQGEFiBIkRJEaQGEFiBIn5Az15KRaJNwBWAAAAAElFTkSuQmCC\n",
      "text/plain": [
       "<PIL.Image.Image image mode=L size=400x180 at 0x7F37805A4050>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "fifty\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<PIL.Image.Image image mode=L size=400x180 at 0x7F378F290850>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: left;\">\n",
       "      <th>epoch</th>\n",
       "      <th>train_loss</th>\n",
       "      <th>valid_loss</th>\n",
       "      <th>accuracy</th>\n",
       "      <th>perplexity</th>\n",
       "      <th>time</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <td>0</td>\n",
       "      <td>1.644107</td>\n",
       "      <td>1.604447</td>\n",
       "      <td>0.437265</td>\n",
       "      <td>4.975110</td>\n",
       "      <td>02:47</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>"
      ],
      "text/plain": [
       "<IPython.core.display.HTML object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [],
      "text/plain": [
       "<IPython.core.display.HTML object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [],
      "text/plain": [
       "<IPython.core.display.HTML object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "sixty\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<PIL.Image.Image image mode=L size=400x180 at 0x7F378074D3D0>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "fifty\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<PIL.Image.Image image mode=L size=400x180 at 0x7F3780702450>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: left;\">\n",
       "      <th>epoch</th>\n",
       "      <th>train_loss</th>\n",
       "      <th>valid_loss</th>\n",
       "      <th>accuracy</th>\n",
       "      <th>perplexity</th>\n",
       "      <th>time</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <td>0</td>\n",
       "      <td>1.639826</td>\n",
       "      <td>1.603783</td>\n",
       "      <td>0.437490</td>\n",
       "      <td>4.971803</td>\n",
       "      <td>02:45</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>"
      ],
      "text/plain": [
       "<IPython.core.display.HTML object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [],
      "text/plain": [
       "<IPython.core.display.HTML object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [],
      "text/plain": [
       "<IPython.core.display.HTML object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "sixty\n"
     ]
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAZAAAAC0CAAAAACfniWEAAADHklEQVR4nO3doW4UARRG4W2Z7QKFgKlCEYJC8B4oXoKXQPAQKF4AHBbZN8FjcIRCy3ZR3F/0z4QEwRHnUzcz3Zndnl15M0evN3+8menJTJ83N30tx85n+jDTu5kOM32Z6edMu5lulyvfmumkvPbuTN9mOi2vzXQo0zLTfqaL8tr2rjLtV//uuhy7X15xXI7pPzIIjEFgDAJjEBiDwBgExiAwBoExCIxBYAwCYxAYg8AYBMYgMAaBMQiMQWAMAmMQGIPAGATGIDAGgTEIjEFgDAJjEBiDwBgExiAwBoExCIxBYJYXM2bn7n05FtkE3M70fKZsAr6a6e1MH8v1sn13Z6ajmS42N+Vb9HCmlzP9minv/rIcu5opnyNXvjfTvpzN+zuUs20r8aoc25Zj/kJgDAJjEBiDwBgExiAwBoExCIxBYAwCYxAYg8AYBMYgMAaBMQiMQWAMAmMQGIPAGATGIDAGgTEIjEFgDAJjEBiDwBgExiAwBoExCIxBYAwCYxCY5dmM2Zs7K8fitJx9PNOjmT7N9GCmHzO1b8L3cuW8Ipt72SJ8OlP2E7P11zYBY1f+LvdYyhTt/9Lkvrtytl3ZXwiMQWAMAmMQGIPAGATGIDAGgTEIjEFgDAJjEBiDwBgExiAwBoExCIxBYAwCYxAYg8AYBMYgMAaBMQiMQWAMAmMQGIPAGATGIDAGgTEIjEFgDAKz/MueW3uGX3bpsvWXpwjmaX55GmKeE9hkmy+bhdtyNk9DbDuG69oG4raczT2yi7i+x7h+j8ZfCIxBYAwCYxAYg8AYBMYgMAaBMQiMQWAMAmMQGIPAGATGIDAGgTEIjEFgDAJjEBiDwBgExiAwBoExCIxBYAwCYxAYg8AYBMYgMAaBMQiMQWAMArO0PbzLcmxdrnJcputyLBuD69fLBmL77pyUe8ShTO3ztiu3zcJMbbNw/Sp/u/noLwTGIDAGgTEIjEFgDAJjEBiDwBgExiAwBoExCIxBYAwCYxAYg8AYBMYgMAaBMQiMQWAMAmMQGIPAGATGIDAGgTEIjEFgDAJjEBiDwBgExiAwBoH5DdqbQixpQk0xAAAAAElFTkSuQmCC\n",
      "text/plain": [
       "<PIL.Image.Image image mode=L size=400x180 at 0x7F3781EA6290>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "fifty\n"
     ]
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAZAAAAC0CAAAAACfniWEAAADG0lEQVR4nO3doW5UYRRF4WnnplBSAgaHoBrJo+B4A2xfggRLsHiehVdogkegCIVApyjOFt35g+siWZ86uffOdCarI0/+oze7vy5mOp/p8+62r+Xap5nezfShPPdlpm8zncx0OtPNTNtM92b6WV7xfaazmY6XU/7Gfqbr8vnyCdqnaq89KncP5e6j3W3H5ZrukEFgDAJjEBiDwBgExiAwBoExCIxBYAwCYxAYg8AYBMYgMAaBMQiMQWAMAmMQGIPAGATGIDAGgTEIjEFgDAJjEBiDwBgExiAwBoExCIxBYAwCs72YMVtw78u1yN5ctgOfzvRyptczvZ3pY3m/7Nzl/bLNd1Wey39Rdv1ezfSjPJfvkXfOTmDeJa94ONPvcjefJe+8lbvtuWi/Bn8hMAaBMQiMQWAMAmMQGIPAGATGIDAGgTEIjEFgDAJjEBiDwBgExiAwBoExCIxBYAwCYxAYg8AYBMYgMAaBMQiMQWAMAmMQGIPAGATGIDAGgTEIzPZ8xmzfPS7XImcHpuazmZ7MdDlTTiDMxmA7WTDbfNnN+1U+Qe7mtMH8jZvyXDu9sO0Eti3Cdo7h+m6Tzcz2zXfLa7pDBoExCIxBYAwCYxAYg8AYBMYgMAaBMQiMQWAMAmMQGIPAGATGIDAGgTEIjEFgDAJjEBiDwBgExiAwBoExCIxBYAwCYxAYg8AYBMYgMAaBMQjM9qBcPCnXYl+u5RXtTMBWPVt62c3LWX/78lw2EO/PlDML19rJgtG2A9fPrU80bLuI7f3aNX8hMAaBMQiMQWAMAmMQGIPAGATGIDAGgTEIjEFgDAJjEBiDwBgExiAwBoExCIxBYAwCYxAYg8AYBMYgMAaBMQiMQWAMAmMQGIPAGATGIDAGgTEIzNY2Bg/lWqw34zJdL5/LxmD7BHlt2zvMtdNyt1mfMRjte/zrBmI7DTEO5a47hv8Bg8AYBMYgMAaBMQiMQWAMAmMQGIPAGATGIDAGgTEIjEFgDAJjEBiDwBgExiAwBoExCIxBYAwCYxAYg8AYBMYgMAaBMQiMQWAMAmMQGIPAGATmD0YhPDia2z9dAAAAAElFTkSuQmCC\n",
      "text/plain": [
       "<PIL.Image.Image image mode=L size=400x180 at 0x7F3780702610>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: left;\">\n",
       "      <th>epoch</th>\n",
       "      <th>train_loss</th>\n",
       "      <th>valid_loss</th>\n",
       "      <th>accuracy</th>\n",
       "      <th>perplexity</th>\n",
       "      <th>time</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <td>0</td>\n",
       "      <td>1.631543</td>\n",
       "      <td>1.603433</td>\n",
       "      <td>0.437347</td>\n",
       "      <td>4.970066</td>\n",
       "      <td>02:45</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>"
      ],
      "text/plain": [
       "<IPython.core.display.HTML object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [],
      "text/plain": [
       "<IPython.core.display.HTML object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [],
      "text/plain": [
       "<IPython.core.display.HTML object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "sixty\n"
     ]
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAZAAAAC0CAAAAACfniWEAAADUElEQVR4nO3cIW5UcRRG8ba86VAKAYNCEYJCsA8Um2ATCBaBYgPUYZHsBI/BEQot00FxP9EvL00QPeL81M3/dd5Mezru5h2+Pfjn3UzPZvp6cN33cvZlprOZPsy0n+nbTL9n2s50t9z5zkzH5bX3Zvox02l5baZ9mZaZdjOdl9e2T5Vpt/pzV+XsQXnFUTnTLTIIjEFgDAJjEBiDwBgExiAwBoExCIxBYAwCYxAYg8AYBMYgMAaBMQiMQWAMAmMQGIPAGATGIDAGgTEIjEFgDAJjEJjl1YzZBv9YziI76puZXs6UHfU3M72f6VO5X/bCT2Y6nOn84Lr8Fz2a6fVMf2bKp78oZ5cz5ffIne/PtCtX8/n25Wrbl78sZ5ty5jcExiAwBoExCIxBYAwCYxAYg8AYBMYgMAaBMQiMQWAMAmMQGIPAGATGIDAGgTEIjEFgDAJjEBiDwBgExiAwBoExCIxBYJYXM2aj+3E5i9Ny9elMT2b6PNPDmX7N1P4TfpY75xXZKc9++/OZsjmfffS2ox7b8nN5j6VM0f4uTd53W662O/sNgTEIjEFgDAJjEBiDwBgExiAwBoExCIxBYAwCYxAYg8AYBMYgMAaBMQiMQWAMAmMQGIPAGATGIDAGgTEIjEFgDAJjEJjlfzaw23PPs+WdffQ8eT1PQM8T5PNs9SZ75tl535SreYJ8235f13bjN+Vq3iNb8usb9uvv0fgNgTEIjEFgDAJjEBiDwBgExiAwBoExCIxBYAwCYxAYg8AYBMYgMAaBMQiMQWAMAmMQGIPAGATGIDAGgTEIjEFgDAJjEJilbYhflLN1uctRma7KWXbZ1++X3fj2v3Nc3iP2ZWq/b7tz23nP1Hbe1+9y0518vyEwBoExCIxBYAwCYxAYg8AYBMYgMAaBMQiMQWAMAmMQGIPAGATGIDAGgTEIjEFgDAJjEBiDwBgExiAwBoExCIxBYAwCU5/9vr6fvb4/nqsn5Wqmtl3ezrLfvis/lz399jz49lT59R319kz3/IXaFn87a3duf7WbbuLrFhkExiAwBoExCIxBYAwCYxAYg8AYBMYgMAaBMQiMQWAMAmMQGIPAGATGIDAGgTEIjEFgDAJjEBiDwBgExiAwBoExCMxf0Z5OR5rT3YsAAAAASUVORK5CYII=\n",
      "text/plain": [
       "<PIL.Image.Image image mode=L size=400x180 at 0x7F378FED2190>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "fifty\n"
     ]
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAZAAAAC0CAAAAACfniWEAAADRklEQVR4nO3cO24UQRhF4bGnZTDilZARQEzIUsjYASmbQCJFpOSsxVtAIicgsngIPEPEfwNflch8kM4X/aruabd1PFm5Tt7s/no909OZPu+u+1rWLmZ6N9OHct+XmS5nOpvpfKbjTNtMt2b6WT7xbaa7M50up/yM/UxX5f3yBu2t2mdPytVDufpgd91pWdMNMgiMQWAMAmMQGIPAGATGIDAGgTEIjEFgDAJjEBiDwBgExiAwBoExCIxBYAwCYxAYg8AYBMYgMAaBMQiMQWAMAmMQmO35jNmf/b6sRXZ0Z9/645lezPRqprczfSzPy27wPC/7zL+X+/JXlF3oL2f6Ue7L75EnZ7d6npJP3Jvpd7mad8mTt3K13Rft2+A3BMYgMAaBMQiMQWAMAmMQGIPAGATGIDAGgTEIjEFgDAJjEBiDwBgExiAwBoExCIxBYAwCYxAYg8AYBMYgMAaBMQiMQWC2ZzNmX/jDshY5bz01n8z0aKZPM+XU9uxlb6exZ595do3/Km+QqzmhPT/jWO5rJ7633eptf3s7+319tcn/DLTffLdc0w0yCIxBYAwCYxAYg8AYBMYgMAaBMQiMQWAMAmMQGIPAGATGIDAGgTEIjEFgDAJjEBiDwBgExiAwBoExCIxBYAwCYxCY7U5ZPCtrsS9r+UQ7R71Vz/7x7BrP+ej7cl/2xt+eKee8r7XT2KPtW1/ftz4Fvu2Sb89ra35DYAwCYxAYg8AYBMYgMAaBMQiMQWAMAmMQGIPAGATGIDAGgTEIjEFgDAJjEBiDwBgExiAwBoExCIxBYAwCYxAYg8AYBGZre9kPZS3We7YzXS3vy1729gb5bNsRn7XzcrVZn8se7ff4173x7QT5OJSr7n7/DxgExiAwBoExCIxBYAwCYxAYg8AYBMYgMAaBMQiMQWAMAmMQGIPAGATGIDAGgTEIjEFgDAJjEBiDwBgExiAwBoExCMzWTllva2vtjPP7M7XqbS97Ow9+X9Zy3+XyXdansR/LtP77XJ8H3568/mzjNwTGIDAGgTEIjEFgDAJjEBiDwBgExiAwBoExCIxBYAwCYxAYg8AYBMYgMAaBMQiMQWAMAmMQGIPAGATGIDAGgTEIjEFg/gB9F0ZMZO9LUgAAAABJRU5ErkJggg==\n",
      "text/plain": [
       "<PIL.Image.Image image mode=L size=400x180 at 0x7F38280BAB50>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: left;\">\n",
       "      <th>epoch</th>\n",
       "      <th>train_loss</th>\n",
       "      <th>valid_loss</th>\n",
       "      <th>accuracy</th>\n",
       "      <th>perplexity</th>\n",
       "      <th>time</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <td>0</td>\n",
       "      <td>1.632994</td>\n",
       "      <td>1.603326</td>\n",
       "      <td>0.437385</td>\n",
       "      <td>4.969532</td>\n",
       "      <td>02:44</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>"
      ],
      "text/plain": [
       "<IPython.core.display.HTML object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [],
      "text/plain": [
       "<IPython.core.display.HTML object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [],
      "text/plain": [
       "<IPython.core.display.HTML object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "sixty\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<PIL.Image.Image image mode=L size=400x180 at 0x7F378F7E6310>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "fifty\n"
     ]
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAZAAAAC0CAAAAACfniWEAAADaklEQVR4nO3dq25UURiG4WlnUyjhZHAI0EguBccdYLkJEizB4nst3AIJHoEiHALtoPg/0S8rRfUleR/1Z5+6J2/Hraw5er3769VMT2b6tLvsSzn2Yaa3M70v132e6etMJzOdznSYaZvp5kw/yx3fZroz0/Fyyt/Yz3Re3i9v0N6q3XtUzl6Us/d3lx2XY7pGBoExCIxBYAwCYxAYg8AYBMYgMAaBMQiMQWAMAmMQGIPAGATGIDAGgTEIjEFgDAJjEBiDwBgExiAw27MZs7buXTkWWY2XNYePZno+08uZ3sx0Vp6XlXx5XtYIfi/X5b8oKwhfzPSjXJfPkSdnpWGekjvuzvS7nM275MlbOduui/Zt8BsCYxAYg8AYBMYgMAaBMQiMQWAMAmMQGIPAGATGIDAGgTEIjEFgDAJjEBiDwBgExiAwBoExCIxBYLanM2YJ2YNyLLKtXmo+nunhTB9nyuZ8WfbWNt3LkrQsMPtV3iBnsxFf/sahXNc29msL29pSuLbF3/psk+WF7ZPvlsd0jQwCYxAYg8AYBMYgMAaBMQiMQWAMAmMQGIPAGATGIDAGgTEIjEFgDAJjEBiDwBgExiAwBoExCIxBYLbb5eBJORb7cix3tO3yWvWs/cuKv2yDty/XZV3jrZmynd9a23Qv2prD9XXrzf7aCsf2vHbMbwiMQWAMAmMQGIPAGATGIDAGgTEIjEFgDAJjEBiDwBgExiAwBoExCIxBYAwCYxAYg8AYBMYgMAaBMQjM1tYhth/SjfV6u0zny+uyDrG9Qe5tqxlz7LScbdb7IUb7HFdd19h2boyLctaVi/8Bg8AYBMYgMAaBMQiMQWAMAmMQGIPAGATGIDAGgTEIjEFgDAJjEBiDwBgExiAwBoExCIxBYAwCYxCYre1u2I6ttb0F783Uqrd1iG0fxn051n4tur3LehfEQ5nW/5/rfRjbk9f3Nn5DYAwCYxAYg8AYBMYgMAaBMQiMQWAMAmMQGIPAGATGIDAGgTEIjEFgDAJjEBiDwBgExiAwBoExCMzWitxY3rJe8tWWqV11Adz6bLP+GeD1TwPnk7ft92L9lHZde/K/v5UQDAJjEBiDwBgExiAwBoExCIxBYAwCYxAYg8AYBMYgMAaBMQiMQWAMAmMQGIPAGATGIDAGgTEIjEFg/gC2WU5j+3GBrQAAAABJRU5ErkJggg==\n",
      "text/plain": [
       "<PIL.Image.Image image mode=L size=400x180 at 0x7F37806F9FD0>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: left;\">\n",
       "      <th>epoch</th>\n",
       "      <th>train_loss</th>\n",
       "      <th>valid_loss</th>\n",
       "      <th>accuracy</th>\n",
       "      <th>perplexity</th>\n",
       "      <th>time</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <td>0</td>\n",
       "      <td>1.628513</td>\n",
       "      <td>1.603410</td>\n",
       "      <td>0.437749</td>\n",
       "      <td>4.969953</td>\n",
       "      <td>02:44</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>"
      ],
      "text/plain": [
       "<IPython.core.display.HTML object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [],
      "text/plain": [
       "<IPython.core.display.HTML object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [],
      "text/plain": [
       "<IPython.core.display.HTML object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "sixty\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<PIL.Image.Image image mode=L size=400x180 at 0x7F378074D3D0>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "fifty\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<PIL.Image.Image image mode=L size=400x180 at 0x7F38280D7590>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: left;\">\n",
       "      <th>epoch</th>\n",
       "      <th>train_loss</th>\n",
       "      <th>valid_loss</th>\n",
       "      <th>accuracy</th>\n",
       "      <th>perplexity</th>\n",
       "      <th>time</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <td>0</td>\n",
       "      <td>1.632899</td>\n",
       "      <td>1.603261</td>\n",
       "      <td>0.437060</td>\n",
       "      <td>4.969212</td>\n",
       "      <td>02:44</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>"
      ],
      "text/plain": [
       "<IPython.core.display.HTML object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [],
      "text/plain": [
       "<IPython.core.display.HTML object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [],
      "text/plain": [
       "<IPython.core.display.HTML object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "sixty\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<PIL.Image.Image image mode=L size=400x180 at 0x7F378EBDD990>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "fifty\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<PIL.Image.Image image mode=L size=400x180 at 0x7F378042F9D0>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: left;\">\n",
       "      <th>epoch</th>\n",
       "      <th>train_loss</th>\n",
       "      <th>valid_loss</th>\n",
       "      <th>accuracy</th>\n",
       "      <th>perplexity</th>\n",
       "      <th>time</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <td>0</td>\n",
       "      <td>1.625796</td>\n",
       "      <td>1.602998</td>\n",
       "      <td>0.437443</td>\n",
       "      <td>4.967904</td>\n",
       "      <td>02:43</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>"
      ],
      "text/plain": [
       "<IPython.core.display.HTML object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [],
      "text/plain": [
       "<IPython.core.display.HTML object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [],
      "text/plain": [
       "<IPython.core.display.HTML object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "sixty\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<PIL.Image.Image image mode=L size=400x180 at 0x7F37806F9B10>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "fifty\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<PIL.Image.Image image mode=L size=400x180 at 0x7F378070B850>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: left;\">\n",
       "      <th>epoch</th>\n",
       "      <th>train_loss</th>\n",
       "      <th>valid_loss</th>\n",
       "      <th>accuracy</th>\n",
       "      <th>perplexity</th>\n",
       "      <th>time</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <td>0</td>\n",
       "      <td>1.624873</td>\n",
       "      <td>1.603230</td>\n",
       "      <td>0.436112</td>\n",
       "      <td>4.969055</td>\n",
       "      <td>02:43</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>"
      ],
      "text/plain": [
       "<IPython.core.display.HTML object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [],
      "text/plain": [
       "<IPython.core.display.HTML object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [],
      "text/plain": [
       "<IPython.core.display.HTML object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "sixty\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<PIL.Image.Image image mode=L size=400x180 at 0x7F37806F9D10>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "fifty\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<PIL.Image.Image image mode=L size=400x180 at 0x7F37806F9D90>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "for i in range(10):\n",
    "    lml.fit_one_cycle(1, 1e-2)\n",
    "    preds_after_sixty.append(all_preds_after_sixty(lml))\n",
    "    preds_after_fifty.append(all_preds_after_fifty(lml))\n",
    "    print(\"sixty\")\n",
    "    display(float_array_to_image(1 - np.array(preds_after_sixty) ** 0.25))\n",
    "    print(\"fifty\")\n",
    "    display(float_array_to_image(1 - np.array(preds_after_fifty) ** 0.25))\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 24,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "Path('models/nnonlml.pth')"
      ]
     },
     "execution_count": 24,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "lml.save('nnonlml')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 25,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [],
      "text/plain": [
       "<IPython.core.display.HTML object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/plain": [
       "SuggestedLRs(lr_min=0.002290867641568184, lr_steep=6.309573450380412e-07)"
      ]
     },
     "execution_count": 25,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "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": [
    "lml.lr_find()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 26,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "<fastai.text.learner.LMLearner at 0x7f37805a41d0>"
      ]
     },
     "execution_count": 26,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "lml.load('nnonlml')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 27,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(array(['sixty eight', 'sixty five', 'sixty four', 'sixty one',\n",
       "        'sixty seven', 'sixty six', 'sixty thirty', 'sixty thousand',\n",
       "        'sixty three', 'sixty two'], dtype='<U14'),\n",
       " array([10,  7,  4,  2,  4,  4,  1,  3, 10,  5]))"
      ]
     },
     "execution_count": 27,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "pred_histo_after_sixty(lml, 1, 50)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 28,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(array(['fifty eight', 'fifty five', 'fifty four', 'fifty nine',\n",
       "        'fifty one', 'fifty seven', 'fifty six', 'fifty thousand',\n",
       "        'fifty three', 'fifty two'], dtype='<U14'),\n",
       " array([5, 6, 5, 5, 7, 9, 2, 1, 8, 2]))"
      ]
     },
     "execution_count": 28,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "pred_histo_after_fifty(lml, 1, 50)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": 29,
   "metadata": {},
   "outputs": [],
   "source": [
    "lml = dataloader_to_lm_learner(ondl)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 30,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "<fastai.text.learner.LMLearner at 0x7f37805756d0>"
      ]
     },
     "execution_count": 30,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "lml.load('nnonlml')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 31,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [],
      "text/plain": [
       "<IPython.core.display.HTML object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/plain": [
       "SuggestedLRs(lr_min=0.025118863582611083, lr_steep=0.2089296132326126)"
      ]
     },
     "execution_count": 31,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "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": [
    "lml.lr_find()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 32,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "<fastai.text.learner.LMLearner at 0x7f37805756d0>"
      ]
     },
     "execution_count": 32,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "lml.load('nnonlml')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 33,
   "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>perplexity</th>\n",
       "      <th>time</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <td>0</td>\n",
       "      <td>4.955695</td>\n",
       "      <td>7.702248</td>\n",
       "      <td>0.238793</td>\n",
       "      <td>2213.317139</td>\n",
       "      <td>00:07</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>"
      ],
      "text/plain": [
       "<IPython.core.display.HTML object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [],
      "text/plain": [
       "<IPython.core.display.HTML object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [],
      "text/plain": [
       "<IPython.core.display.HTML object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "sixty\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<PIL.Image.Image image mode=L size=400x180 at 0x7F3780112350>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "fifty\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<PIL.Image.Image image mode=L size=400x180 at 0x7F3780479310>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: left;\">\n",
       "      <th>epoch</th>\n",
       "      <th>train_loss</th>\n",
       "      <th>valid_loss</th>\n",
       "      <th>accuracy</th>\n",
       "      <th>perplexity</th>\n",
       "      <th>time</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <td>0</td>\n",
       "      <td>17.590851</td>\n",
       "      <td>25.351660</td>\n",
       "      <td>0.139555</td>\n",
       "      <td>102349553664.000000</td>\n",
       "      <td>00:07</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>"
      ],
      "text/plain": [
       "<IPython.core.display.HTML object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [],
      "text/plain": [
       "<IPython.core.display.HTML object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [],
      "text/plain": [
       "<IPython.core.display.HTML object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "sixty\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<PIL.Image.Image image mode=L size=400x180 at 0x7F3780566590>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "fifty\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<PIL.Image.Image image mode=L size=400x180 at 0x7F378ECF4690>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: left;\">\n",
       "      <th>epoch</th>\n",
       "      <th>train_loss</th>\n",
       "      <th>valid_loss</th>\n",
       "      <th>accuracy</th>\n",
       "      <th>perplexity</th>\n",
       "      <th>time</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <td>0</td>\n",
       "      <td>38.534431</td>\n",
       "      <td>50.370220</td>\n",
       "      <td>0.101981</td>\n",
       "      <td>7507730262407612727296.000000</td>\n",
       "      <td>00:07</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>"
      ],
      "text/plain": [
       "<IPython.core.display.HTML object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [],
      "text/plain": [
       "<IPython.core.display.HTML object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [],
      "text/plain": [
       "<IPython.core.display.HTML object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "sixty\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<PIL.Image.Image image mode=L size=400x180 at 0x7F378014AA90>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "fifty\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<PIL.Image.Image image mode=L size=400x180 at 0x7F37804BF610>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: left;\">\n",
       "      <th>epoch</th>\n",
       "      <th>train_loss</th>\n",
       "      <th>valid_loss</th>\n",
       "      <th>accuracy</th>\n",
       "      <th>perplexity</th>\n",
       "      <th>time</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <td>0</td>\n",
       "      <td>69.338158</td>\n",
       "      <td>88.954285</td>\n",
       "      <td>0.056269</td>\n",
       "      <td>inf</td>\n",
       "      <td>00:07</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>"
      ],
      "text/plain": [
       "<IPython.core.display.HTML object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [],
      "text/plain": [
       "<IPython.core.display.HTML object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [],
      "text/plain": [
       "<IPython.core.display.HTML object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "sixty\n"
     ]
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAZAAAAC0CAAAAACfniWEAAAEgElEQVR4nO2dsY4cRRRFd+welsUgSIiIECIi4Av4ASI+AJHxEwTERARESEQEkJE65E/IScgQxja7S8S7SL4qemR79yw6J3qqmq7p6bOdXL2qPXxx9g9fTvXeVL+cPctvZeznqX6c6tuprqf6darHU51P9WpZ+f5Ur5RrX5vq96kelGtTXZdqm+pyqkfl2nZXqS6Xn7sqY2+UK+6VMblFFAJDITAUAkMhMBQCQyEwFAJDITAUAkMhMLaPp0y280MZC0mcjlN9OFUSp8+n+maqn8p6SXkupjpM9ejsWfJX9NZUn0z111S5+ydl7OlU+R1Z+fWpLsts7u+6zLb062kZO5Yx3xAYCoGhEBgKgaEQGAqBoRAYCoGhEBgKgaEQGNsHUyafebuMhQdl9t2p3pnq4VRvTvXnVO0v4Y+ycq5IQpS06v2pkoMlXWqJUzgvn8t3bKUK7bk08r3nZbat7BsCQyEwFAJDITAUAkMhMBQCQyEwFAJDITAUAmN7njyl9SQls0m6lK6odCelu6vtzQtJjZJgHctsurtalrWmJV3HMtv2Ca7zsvV3NHxDYCgEhkJgKASGQmAoBIZCYCgEhkJgKASGQmBsLe95UsbWZJV7pboqY0mm1uu1HX4hOxqvymw7G6v93rZyS7BSrfcYnn4H63uRW0QhMBQCQyEwFAJDITAUAkMhMBQCQyEwFAKj9mWt05Z1GpTZizKbqmVFbSxp1WX5XFK31qvVOr7WiVPrt8oTaplcG2srt6e2N1eTW0QhMBQCQyEwFAJDITAUAkMhMBQCQyEwFAJja6lM63EKrSfpUKrHZSxXJHtKh9bVcja0/0CYz7Vf1FKyvelSm20rt6fRki77su4UCoGhEBgKgaEQGAqBoRAYCoGhEBgKgaEQGP9qy2oJzOkdWq0vK71VOS+rJU7tdPfWM9U6pVqadvpfW1ulrdeysfUvWq8cfENgKASGQmAoBIZCYCgEhkJgKASGQmAoBIZCYLQthv/RlxXaeebrfKudqN7O0ModtNzqrMxmvayyPpe93Wk+d7GcbXlU69A6Hd8QGAqBoRAYCoGhEBgKgaEQGAqBoRAYCoGhEBjbev/fmnXXVtKglkflO9qpVm3vYGi7IY9lbJ1Wrc9vX/dRrc/Bul/GgnsM7xQKgaEQGAqBoRAYCoGhEBgKgaEQGAqBoRAY2+lJzen7DttOuyROOb/9fKrkW+tUK7sX9+4J3PuL1k9j3ee1d9Y9hncAhcBQCAyFwFAIDIXAUAgMhcBQCAyFwFAIjMPe/YQvhq+m+n6qT6f6aKqvp/psqodTfTfVOle7i/iGwFAIDIXAUAgMhcBQCAyFwFAIDIXAUAgMhcA4tB6idefQ89B286UHa28v1Pr09BfNy3saDd8QGAqBoRAYCoGhEBgKgaEQGAqBoRAYCoGhEBj17PebZW8ydbMJVvvem8A3BIZCYCgEhkJgKASGQmAoBIZCYCgEhkJgKARGzbJeXnqzd+X/397BvfiGwFAIDIXAUAgMhcBQCAyFwFAIDIXAUAgMhcD4G9ZVlZwCPt/2AAAAAElFTkSuQmCC\n",
      "text/plain": [
       "<PIL.Image.Image image mode=L size=400x180 at 0x7F37804742D0>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "fifty\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<PIL.Image.Image image mode=L size=400x180 at 0x7F3783448E90>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: left;\">\n",
       "      <th>epoch</th>\n",
       "      <th>train_loss</th>\n",
       "      <th>valid_loss</th>\n",
       "      <th>accuracy</th>\n",
       "      <th>perplexity</th>\n",
       "      <th>time</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <td>0</td>\n",
       "      <td>115.001587</td>\n",
       "      <td>140.299225</td>\n",
       "      <td>0.025995</td>\n",
       "      <td>inf</td>\n",
       "      <td>00:07</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>"
      ],
      "text/plain": [
       "<IPython.core.display.HTML object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [],
      "text/plain": [
       "<IPython.core.display.HTML object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [],
      "text/plain": [
       "<IPython.core.display.HTML object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "sixty\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<PIL.Image.Image image mode=L size=400x180 at 0x7F3780112E10>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "fifty\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<PIL.Image.Image image mode=L size=400x180 at 0x7F37804DBF90>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "for i in range(5):\n",
    "    lml.fit_one_cycle(1, -5e-3)\n",
    "    preds_after_sixty.append(all_preds_after_sixty(lml))\n",
    "    preds_after_fifty.append(all_preds_after_fifty(lml))\n",
    "    print(\"sixty\")\n",
    "    display(float_array_to_image(1 - np.array(preds_after_sixty) ** 0.25))\n",
    "    print(\"fifty\")\n",
    "    display(float_array_to_image(1 - np.array(preds_after_fifty) ** 0.25))\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 34,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "Path('models/nnonlml.pth')"
      ]
     },
     "execution_count": 34,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "lml.save('nnonlml')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 35,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(array(['sixty two'], dtype='<U9'), array([50]))"
      ]
     },
     "execution_count": 35,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "pred_histo_after_sixty(lml, 1, 50)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 36,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(array(['fifty two'], dtype='<U9'), array([50]))"
      ]
     },
     "execution_count": 36,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "pred_histo_after_fifty(lml, 1, 50)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": 37,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "<fastai.text.learner.LMLearner at 0x7f3780503050>"
      ]
     },
     "execution_count": 37,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "lml = dataloader_to_lm_learner(nndl) ; lml.load('nnonlml')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": 38,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [],
      "text/plain": [
       "<IPython.core.display.HTML object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/plain": [
       "<fastai.text.learner.LMLearner at 0x7f3780503050>"
      ]
     },
     "execution_count": 38,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "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": [
    "lml.lr_find() ; lml.load('nnonlml')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 39,
   "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>perplexity</th>\n",
       "      <th>time</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <td>0</td>\n",
       "      <td>1.697088</td>\n",
       "      <td>1.629262</td>\n",
       "      <td>0.436164</td>\n",
       "      <td>5.100111</td>\n",
       "      <td>02:43</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>"
      ],
      "text/plain": [
       "<IPython.core.display.HTML object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [],
      "text/plain": [
       "<IPython.core.display.HTML object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [],
      "text/plain": [
       "<IPython.core.display.HTML object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "sixty\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<PIL.Image.Image image mode=L size=400x180 at 0x7F37806FDA90>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "fifty\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<PIL.Image.Image image mode=L size=400x180 at 0x7F378064B9D0>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: left;\">\n",
       "      <th>epoch</th>\n",
       "      <th>train_loss</th>\n",
       "      <th>valid_loss</th>\n",
       "      <th>accuracy</th>\n",
       "      <th>perplexity</th>\n",
       "      <th>time</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <td>0</td>\n",
       "      <td>1.641761</td>\n",
       "      <td>1.606582</td>\n",
       "      <td>0.437399</td>\n",
       "      <td>4.985739</td>\n",
       "      <td>02:43</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>"
      ],
      "text/plain": [
       "<IPython.core.display.HTML object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [],
      "text/plain": [
       "<IPython.core.display.HTML object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [],
      "text/plain": [
       "<IPython.core.display.HTML object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "sixty\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<PIL.Image.Image image mode=L size=400x180 at 0x7F378014AD10>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "fifty\n"
     ]
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAZAAAAC0CAAAAACfniWEAAAFAklEQVR4nO2dMYtdVRRG38tcJ0YSTRNIQHBsbLSztRYEu1ilSps2tb1gK1aCYOn/0Cq1naCCWFgEC6NJzGSs3LvIx+Y8EshKXKv6OOfeM/fNeqfY79x77v6z3X/crvR2pZ92T3I3tN2p9EWlr8Nxv1f6s9JxpQuVziptlc5XehDO+KvSxUrnxtR/46jSabi+voJ0Vencfeh9HHrf2D3JudAmzxGFwFAIDIXAUAgMhcBQCAyFwFAIjO39il1Lfhnamq4+u8Z+s9L1SrcqfV7p2zBeV649XtfEf4fj+lvUFfONSvfDcf05euSurHuUPuNSpUeht6+lR95CbzquSbPBGQJDITAUAkMhMBQCQyEwFAJDITAUAmN7t2LXsJdDW9Pr2G3zpNKVSj9W6tXwrrvTKnfXxF3h/hOuoHt75bv/xlk4Lq2kp8o61eJpTX3uTfTvG+mT78Y2eY4oBIZCYCgEhkJgKASGQmAoBIZCYGyvhcbj0NYchbY+I61PJ+td63aF2+vOR+G4ruNfrdTr5zNplbtJNfZ83Ly6nir6NF5qc4bAUAgMhcBQCAyFwFAIDIXAUAgMhcDYUt2d7tRu5vqy0+l4XNfd6Qr63FS9d9uF0JuY17ub+Xn2Zq6707f8cei1Un8BUAgMhcBQCAyFwFAIDIXAUAgMhcDY0up1aptJa8evV0rW573W0m5vaaW6735P1zKvcp+FNH8/53X2NPJ8bsIZAkMhMBQCQyEwFAJDITAUAkMhMBQCY0tGXhlPmWvOVCevVuBzb2K+z3y+9zztNzffp7/6ydPIh1+VIFAIDIXAUAgMhcBQCAyFwFAIDIXA2Farz5lUfR6HtrTfeqIr3HRG2p1t9arm9e7Dnyuff4+Y94yzUn8BUAgMhcBQCAyFwFAIDIXAUAgMhcCIBfOzsXSvUrqbfl4rT0+np/Xz9GtAkz7H6m5v6Tn11dX15vAznCEwFAJDITAUAkMhMBQCQyEwFAJDITC21dq5Wa0551XztN9ceoNaH5e+O2lfuvlzzOvsc9su9Cbm3ehmnCEwFAJDITAUAkMhMBQCQyEwFAJDITDi3e/zjnJzNZueU+/n3lMNm+5vn+v4HqVHXq2xU0Wf7kdPO97N6+yJw3udITAUAkMhMBQCQyEwFAJDITAUAkMhMLaujrv6vNrd4ZTVmvjjSpcqPax0t1LXxCeVfq3U707v6v3TSl1FXwxXtcr50Hb4KKvM75l3hsBQCAyFwFAIDIXAUAgMhcBQCAyFwNind6d3TTy/pXwmrUXfqfRRpQ8qfVXpu0rvVfq+0s2nuCo6zhAYCoGhEBgKgaEQGAqBoRAYCoGhEBixUn+qASutPsv9oNL8drj/B84QGAqBoRAYCoGhEBgKgaEQGAqBoRAY+1RPrz6FfTjp2fD5PWer99q/LDhDYCgEhkJgKASGQmAoBIZCYCgEhkJgzBu0P3NW31K+2vby4QyBoRAYCoGhEBgKgaEQGAqBoRAYCoGxfxgaf650cviAlb6p9Eml3gvuh0rXKr1V6X6lviO+38j2Tuj9sFJat2/SCn56Lr93smvSvvaH/6bQ514Ovc4QGAqBoRAYCoGhEBgKgaEQGAqBoRAYW9e/7eaPSukp9tW70H8J43Ul3PV579p+L1xLP8Xe/BbG69p+de39LKT+b6zebDD/LpCeLTgNqXGGwFAIDIXAUAgMhcBQCAyFwFAIDIXA+BeAyKVRD4/JkQAAAABJRU5ErkJggg==\n",
      "text/plain": [
       "<PIL.Image.Image image mode=L size=400x180 at 0x7F378014AD10>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: left;\">\n",
       "      <th>epoch</th>\n",
       "      <th>train_loss</th>\n",
       "      <th>valid_loss</th>\n",
       "      <th>accuracy</th>\n",
       "      <th>perplexity</th>\n",
       "      <th>time</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <td>0</td>\n",
       "      <td>1.625098</td>\n",
       "      <td>1.604045</td>\n",
       "      <td>0.437462</td>\n",
       "      <td>4.973110</td>\n",
       "      <td>02:45</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>"
      ],
      "text/plain": [
       "<IPython.core.display.HTML object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [],
      "text/plain": [
       "<IPython.core.display.HTML object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [],
      "text/plain": [
       "<IPython.core.display.HTML object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "sixty\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<PIL.Image.Image image mode=L size=400x180 at 0x7F37806FD550>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "fifty\n"
     ]
    },
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAZAAAAC0CAAAAACfniWEAAAFIklEQVR4nO2dMYtdVRRG38tcJ0YSTSMoCI6NjXa21oJgp5WVra21fSBtsBIES/+HVqntBBXEwiJYGE1iJmPl3kU+NueRkayJa1Uf59x75r5Z7xb7nXvP2d/Y/ctnld6o9OPuce6EttuVblX6Khz3W6U/Kh1XulLprNJW6XKl++GMPytdrXRpTP03jiqdhuvrK0hXlc7dh95Hofel3eNcCm3yFFEIDIXAUAgMhcBQCAyFwFAIjO2dil26fBHami52uqR7rdKHlT6tdLPSN2G8LpR6vC7B/grH9beoC7SPK90Lx/Xn6JG7kOtR+oxrlR6G3r6WHnkLvem4Jt0N3iEwFAJDITAUAkMhMBQCQyEwFAJDITC2typ2DXs9tDU9bdo2Tyq9XOmHSj352nV3mlTtmrgr3L/DFXRvT7T23zgLx6WJ21RZp1o8TeHOvYn+fSN98t3YJk8RhcBQCAyFwFAIDIXAUAgMhcDYXgiNx6GtOQptfUaaDk3Wu7TqgqqnOY/CcV02Pl+pp2tn0qRqk0q6+bh5MjcVkGm81OYdAkMhMBQCQyEwFAJDITAUAkMhMBQCY0t1d3owuJnry06n43Fdd6cr6HNT9d5tV0JvYp5ebea3dZu57k7f8keh10r9AqAQGAqBoRAYCoGhEBgKgaEQGFuaLE1tM2mq8sVKyfq8cFFaOilNjPazvela5knVs5Dm7+c8rZtGns9NeIfAUAgMhcBQCAyFwFAIDIXAUAgMhcDYkpHnxlPmmjPVyasV+NybmB9rnh91TqtpzY+Fr37yNPLhVyUIFAJDITAUAkMhMBQCQyEwFAJjWy12ZlKxcxza0uLFiS6o0hlpqaPVq5qnVw9/a3Yuf+cFmCwMLwAKgaEQGAqBoRAYCoGhEBgKgaEQGLFgPh9Ldyulh7fnqdn07m2ark2/BjTpc6yuZZXewl2dzG0OP8M7BIZCYCgEhkJgKASGQmAoBIZCYGyrpVqzWuLMk7Rp8aa0+00fl747aZGn+XPM07pz2y70JualnWa8Q2AoBIZCYCgEhkJgKASGQmAoBIZCYMSHref1suZqNr2F22/1pho2PU491/E9So+8WmOnij49/pzW85qndROH93qHwFAIDIXAUAgMhcBQCAyFwFAIjK2LsS52XunucMpqCfZBpWuVHlS6U6lLsJNKv1TqfW+7WPy8UhdtV8NVrXI5tB0+yirzHsHeITAUAkMhMBQCQyEwFAJDITAUAkMhMPZp39uuiecdZmfS1OftSu9XerfSl5W+rfR2pe8qffIEV0XHOwSGQmAoBIZCYCgEhkJgKASGQmDEwvCJBqy0+qbq/Urzzj7/D7xDYCgEhkJgKASGQmAoBIZCYCgEhkJg7FM9vfqO6eGkN1/nTYNWH+1+VvAOgaEQGAqBoRAYCoGhEBgKgaEQGPNqx+fO6g6zq23PHt4hMBQCQyEwFAJDITAUAkMhMBQCQyEw9g9C40+VTg4fsNLXlT6q1CtdfV/p1UqvV7pXqR/A7u2N3gy971VK08RNmjBObx33Ol1NWiT68N8U+tzrodc7BIZCYCgEhkJgKASGQmAoBIZCYGxdbrWb3yuld3RXn7H9OYzXhVeXg70E8t1wLf2ObvNrGK9LydWp3rOQ+r+xOrc9l6HpyenTkBrvEBgKgaEQGAqBoRAYCoGhEBgKgaEQGLEgndetOnx/17StUtfnadfZeeSuys9nY6Ie+b9bsSvtFuxeuBcAhcBQCAyFwFAIDIXAUAgMhcD4BxqMqWvefnOWAAAAAElFTkSuQmCC\n",
      "text/plain": [
       "<PIL.Image.Image image mode=L size=400x180 at 0x7F37806FD550>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: left;\">\n",
       "      <th>epoch</th>\n",
       "      <th>train_loss</th>\n",
       "      <th>valid_loss</th>\n",
       "      <th>accuracy</th>\n",
       "      <th>perplexity</th>\n",
       "      <th>time</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <td>0</td>\n",
       "      <td>1.635305</td>\n",
       "      <td>1.602721</td>\n",
       "      <td>0.437050</td>\n",
       "      <td>4.966527</td>\n",
       "      <td>02:45</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>"
      ],
      "text/plain": [
       "<IPython.core.display.HTML object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [],
      "text/plain": [
       "<IPython.core.display.HTML object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [],
      "text/plain": [
       "<IPython.core.display.HTML object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "sixty\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<PIL.Image.Image image mode=L size=400x180 at 0x7F378013AED0>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "fifty\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<PIL.Image.Image image mode=L size=400x180 at 0x7F378011A6D0>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: left;\">\n",
       "      <th>epoch</th>\n",
       "      <th>train_loss</th>\n",
       "      <th>valid_loss</th>\n",
       "      <th>accuracy</th>\n",
       "      <th>perplexity</th>\n",
       "      <th>time</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <td>0</td>\n",
       "      <td>1.622017</td>\n",
       "      <td>1.602729</td>\n",
       "      <td>0.437893</td>\n",
       "      <td>4.966567</td>\n",
       "      <td>02:44</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>"
      ],
      "text/plain": [
       "<IPython.core.display.HTML object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [],
      "text/plain": [
       "<IPython.core.display.HTML object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [],
      "text/plain": [
       "<IPython.core.display.HTML object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "sixty\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<PIL.Image.Image image mode=L size=400x180 at 0x7F37801362D0>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "fifty\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<PIL.Image.Image image mode=L size=400x180 at 0x7F378018A3D0>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "for i in range(5):\n",
    "    lml.fit_one_cycle(1, 1e-2)\n",
    "    preds_after_sixty.append(all_preds_after_sixty(lml))\n",
    "    preds_after_fifty.append(all_preds_after_fifty(lml))\n",
    "    print(\"sixty\")\n",
    "    display(float_array_to_image(1 - np.array(preds_after_sixty) ** 0.25))\n",
    "    print(\"fifty\")\n",
    "    display(float_array_to_image(1 - np.array(preds_after_fifty) ** 0.25))\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 40,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "Path('models/nnonlml.pth')"
      ]
     },
     "execution_count": 40,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "lml.save('nnonlml')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": 41,
   "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>perplexity</th>\n",
       "      <th>time</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <td>0</td>\n",
       "      <td>12.457541</td>\n",
       "      <td>25.120306</td>\n",
       "      <td>0.137091</td>\n",
       "      <td>81210146816.000000</td>\n",
       "      <td>00:07</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>"
      ],
      "text/plain": [
       "<IPython.core.display.HTML object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [],
      "text/plain": [
       "<IPython.core.display.HTML object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [],
      "text/plain": [
       "<IPython.core.display.HTML object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "sixty\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<PIL.Image.Image image mode=L size=400x180 at 0x7F376D549FD0>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "fifty\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<PIL.Image.Image image mode=L size=400x180 at 0x7F376D549FD0>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: left;\">\n",
       "      <th>epoch</th>\n",
       "      <th>train_loss</th>\n",
       "      <th>valid_loss</th>\n",
       "      <th>accuracy</th>\n",
       "      <th>perplexity</th>\n",
       "      <th>time</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <td>0</td>\n",
       "      <td>1.630819</td>\n",
       "      <td>1.603212</td>\n",
       "      <td>0.436849</td>\n",
       "      <td>4.968965</td>\n",
       "      <td>02:44</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>"
      ],
      "text/plain": [
       "<IPython.core.display.HTML object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [],
      "text/plain": [
       "<IPython.core.display.HTML object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [],
      "text/plain": [
       "<IPython.core.display.HTML object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "sixty\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<PIL.Image.Image image mode=L size=400x180 at 0x7F378011AF50>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "fifty\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<PIL.Image.Image image mode=L size=400x180 at 0x7F378014AD10>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: left;\">\n",
       "      <th>epoch</th>\n",
       "      <th>train_loss</th>\n",
       "      <th>valid_loss</th>\n",
       "      <th>accuracy</th>\n",
       "      <th>perplexity</th>\n",
       "      <th>time</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <td>0</td>\n",
       "      <td>14.586851</td>\n",
       "      <td>28.799845</td>\n",
       "      <td>0.145275</td>\n",
       "      <td>3218204524544.000000</td>\n",
       "      <td>00:07</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>"
      ],
      "text/plain": [
       "<IPython.core.display.HTML object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [],
      "text/plain": [
       "<IPython.core.display.HTML object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [],
      "text/plain": [
       "<IPython.core.display.HTML object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "sixty\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<PIL.Image.Image image mode=L size=400x180 at 0x7F378018DA50>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "fifty\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<PIL.Image.Image image mode=L size=400x180 at 0x7F3780133090>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: left;\">\n",
       "      <th>epoch</th>\n",
       "      <th>train_loss</th>\n",
       "      <th>valid_loss</th>\n",
       "      <th>accuracy</th>\n",
       "      <th>perplexity</th>\n",
       "      <th>time</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <td>0</td>\n",
       "      <td>1.632887</td>\n",
       "      <td>1.604353</td>\n",
       "      <td>0.437103</td>\n",
       "      <td>4.974640</td>\n",
       "      <td>02:45</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>"
      ],
      "text/plain": [
       "<IPython.core.display.HTML object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [],
      "text/plain": [
       "<IPython.core.display.HTML object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [],
      "text/plain": [
       "<IPython.core.display.HTML object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "sixty\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<PIL.Image.Image image mode=L size=400x180 at 0x7F37806FDD90>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "fifty\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<PIL.Image.Image image mode=L size=400x180 at 0x7F3780521DD0>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: left;\">\n",
       "      <th>epoch</th>\n",
       "      <th>train_loss</th>\n",
       "      <th>valid_loss</th>\n",
       "      <th>accuracy</th>\n",
       "      <th>perplexity</th>\n",
       "      <th>time</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <td>0</td>\n",
       "      <td>19.543924</td>\n",
       "      <td>38.277534</td>\n",
       "      <td>0.139416</td>\n",
       "      <td>42045741257981952.000000</td>\n",
       "      <td>00:07</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>"
      ],
      "text/plain": [
       "<IPython.core.display.HTML object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [],
      "text/plain": [
       "<IPython.core.display.HTML object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [],
      "text/plain": [
       "<IPython.core.display.HTML object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "sixty\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<PIL.Image.Image image mode=L size=400x180 at 0x7F378018DFD0>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "fifty\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<PIL.Image.Image image mode=L size=400x180 at 0x7F37806FD1D0>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: left;\">\n",
       "      <th>epoch</th>\n",
       "      <th>train_loss</th>\n",
       "      <th>valid_loss</th>\n",
       "      <th>accuracy</th>\n",
       "      <th>perplexity</th>\n",
       "      <th>time</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <td>0</td>\n",
       "      <td>1.646074</td>\n",
       "      <td>1.605781</td>\n",
       "      <td>0.437021</td>\n",
       "      <td>4.981750</td>\n",
       "      <td>02:45</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>"
      ],
      "text/plain": [
       "<IPython.core.display.HTML object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [],
      "text/plain": [
       "<IPython.core.display.HTML object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [],
      "text/plain": [
       "<IPython.core.display.HTML object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "sixty\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<PIL.Image.Image image mode=L size=400x180 at 0x7F3780136F90>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "fifty\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<PIL.Image.Image image mode=L size=400x180 at 0x7F376D6B3BD0>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: left;\">\n",
       "      <th>epoch</th>\n",
       "      <th>train_loss</th>\n",
       "      <th>valid_loss</th>\n",
       "      <th>accuracy</th>\n",
       "      <th>perplexity</th>\n",
       "      <th>time</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <td>0</td>\n",
       "      <td>22.466612</td>\n",
       "      <td>44.458328</td>\n",
       "      <td>0.137370</td>\n",
       "      <td>20323880902183616512.000000</td>\n",
       "      <td>00:07</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>"
      ],
      "text/plain": [
       "<IPython.core.display.HTML object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [],
      "text/plain": [
       "<IPython.core.display.HTML object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [],
      "text/plain": [
       "<IPython.core.display.HTML object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "sixty\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<PIL.Image.Image image mode=L size=400x180 at 0x7F378018A190>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "fifty\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<PIL.Image.Image image mode=L size=400x180 at 0x7F3780187110>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: left;\">\n",
       "      <th>epoch</th>\n",
       "      <th>train_loss</th>\n",
       "      <th>valid_loss</th>\n",
       "      <th>accuracy</th>\n",
       "      <th>perplexity</th>\n",
       "      <th>time</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <td>0</td>\n",
       "      <td>1.645630</td>\n",
       "      <td>1.608458</td>\n",
       "      <td>0.437203</td>\n",
       "      <td>4.995105</td>\n",
       "      <td>02:46</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>"
      ],
      "text/plain": [
       "<IPython.core.display.HTML object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [],
      "text/plain": [
       "<IPython.core.display.HTML object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [],
      "text/plain": [
       "<IPython.core.display.HTML object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "sixty\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<PIL.Image.Image image mode=L size=400x180 at 0x7F378067F9D0>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "fifty\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<PIL.Image.Image image mode=L size=400x180 at 0x7F378067F9D0>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: left;\">\n",
       "      <th>epoch</th>\n",
       "      <th>train_loss</th>\n",
       "      <th>valid_loss</th>\n",
       "      <th>accuracy</th>\n",
       "      <th>perplexity</th>\n",
       "      <th>time</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <td>0</td>\n",
       "      <td>28.135239</td>\n",
       "      <td>55.973743</td>\n",
       "      <td>0.133836</td>\n",
       "      <td>2037454412156266044456960.000000</td>\n",
       "      <td>00:07</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>"
      ],
      "text/plain": [
       "<IPython.core.display.HTML object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [],
      "text/plain": [
       "<IPython.core.display.HTML object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [],
      "text/plain": [
       "<IPython.core.display.HTML object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "sixty\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<PIL.Image.Image image mode=L size=400x180 at 0x7F3780521510>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "fifty\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<PIL.Image.Image image mode=L size=400x180 at 0x7F378010F0D0>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: left;\">\n",
       "      <th>epoch</th>\n",
       "      <th>train_loss</th>\n",
       "      <th>valid_loss</th>\n",
       "      <th>accuracy</th>\n",
       "      <th>perplexity</th>\n",
       "      <th>time</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <td>0</td>\n",
       "      <td>1.658600</td>\n",
       "      <td>1.610227</td>\n",
       "      <td>0.436930</td>\n",
       "      <td>5.003944</td>\n",
       "      <td>02:48</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>"
      ],
      "text/plain": [
       "<IPython.core.display.HTML object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [],
      "text/plain": [
       "<IPython.core.display.HTML object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/html": [],
      "text/plain": [
       "<IPython.core.display.HTML object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "sixty\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<PIL.Image.Image image mode=L size=400x180 at 0x7F376D6BBA10>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "fifty\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<PIL.Image.Image image mode=L size=400x180 at 0x7F3780566950>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "for i in range(5):\n",
    "    lml = dataloader_to_lm_learner(ondl) ;  lml.load('nnonlml')\n",
    "    lml.fit_one_cycle(1, -1e-2)\n",
    "    preds_after_sixty.append(all_preds_after_sixty(lml))\n",
    "    preds_after_fifty.append(all_preds_after_fifty(lml))\n",
    "    print(\"sixty\")\n",
    "    display(float_array_to_image(1 - np.array(preds_after_sixty) ** 0.25))\n",
    "    print(\"fifty\")\n",
    "    display(float_array_to_image(1 - np.array(preds_after_fifty) ** 0.25))\n",
    "    lml.save('nnonlml')\n",
    "    \n",
    "    lml = dataloader_to_lm_learner(nndl) ;  lml.load('nnonlml')\n",
    "    lml.fit_one_cycle(1, 1e-2)\n",
    "    preds_after_sixty.append(all_preds_after_sixty(lml))\n",
    "    preds_after_fifty.append(all_preds_after_fifty(lml))\n",
    "    print(\"sixty\")\n",
    "    display(float_array_to_image(1 - np.array(preds_after_sixty) ** 0.25))\n",
    "    print(\"fifty\")\n",
    "    display(float_array_to_image(1 - np.array(preds_after_fifty) ** 0.25))\n",
    "    lml.save('nnonlml')\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": 42,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(array(['sixty eight', 'sixty five', 'sixty four', 'sixty one',\n",
       "        'sixty seven', 'sixty six', 'sixty thousand', 'sixty three',\n",
       "        'sixty two'], dtype='<U14'),\n",
       " array([ 3,  5,  4,  9,  4,  6,  3,  1, 15]))"
      ]
     },
     "execution_count": 42,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "pred_histo_after_sixty(lml, 1, 50)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 43,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(array(['fifty eight', 'fifty five', 'fifty four', 'fifty nine',\n",
       "        'fifty one', 'fifty seven', 'fifty six', 'fifty thousand',\n",
       "        'fifty three', 'fifty two'], dtype='<U14'),\n",
       " array([4, 3, 5, 5, 8, 2, 6, 4, 5, 8]))"
      ]
     },
     "execution_count": 43,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "pred_histo_after_fifty(lml, 1, 50)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": 48,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<PIL.Image.Image image mode=L size=400x180 at 0x7F3780194990>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "display(float_array_to_image((-np.log2(np.array(preds_after_sixty)+1e-30)).clip(0,30)/30))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 49,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<PIL.Image.Image image mode=L size=400x180 at 0x7F382BF83690>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "display(float_array_to_image(1 - np.array(preds_after_sixty) ** 0.5))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 50,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<PIL.Image.Image image mode=L size=400x180 at 0x7F37806B3290>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "display(float_array_to_image((-np.log2(np.array(preds_after_fifty)+1e-30)).clip(0,30)/30))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 51,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAZAAAAC0CAAAAACfniWEAAAF6ElEQVR4nO2dS4sdVRSF7+3befhEAz7AkcQn6iC+caAoKOgg4k/wV+jImVOdCE4zcRYFEcSBKIgoqBOnghDBRxARgnaiSXfHgbi2kJVDVW5V9arm+0aLU1WnTlj30Dv77Dpn+cXiPx6Xeljqm8Wl7Jq236TelHrd3Pe31HmpldSmeWIptWFGUM/umLZ61vXXZtu896LpZdm82mZl2jZMG+whGBIGhoSBIWFgSBjLio+OSJ2WutU84qKJc1KnpE5KvSL1SXMwh6Qqjqqe620ro54y/bkIaGHaXMy0Mvd17dlFWe4+oqwZgCFhYEgYGBIGhoSxeYNpvKr5iIshDkvdLfWgVLm+JbUyVyt/VO+onFexYa72z1t1faJ/L+vcxwwJA0PCwJAwMCSMTeeI+y99G5d82GhedckHt/BU99US1EGps83+2iMt2k9MCzMkDAwJA0PCwJAwMCQMV3kzUNThyoUqxrkg5ZIoFVHtmqvVdk3zHe02d7UdBU4BMyQMDAkDQ8LAkDAwJAwbZXVdTGlzfbPnevGOua8iG1eOU7+iM1c2tMu+rd02BcyQMDAkDAwJA0PCwJAwbJQ1jEsul1Vt9eKLRlVs5Xpxa4z9VwLbOa92pLlOiXW7Z2ZIGBgSBoaEgSFhYEgYNsoahgNSrraqrrooa8eoesKtMTqGycmt03P/ETBDwsCQMDAkDAwJY8QFqi3TVj27pEe77Np9V15ftq/DeH/8+8MMCQNDwsCQMDAkDAwJY8TUSXsHxYqjXOqk0iRu78OFubpfYIaEgSFhYEgYGBIGhoQxYi6r6z7QFSm52KpwH7e5RbCuO0z334l6CpghYWBIGBgSBoaEgSFh2CjroGvszWNSFcXU+SG/miHcKXVKqrYGqMjrJSn3mdu82S//jn0DhoSBIWFgSBgYEsamywENU8Lstk46LnVM6mWpZ6V+kbpJ6gczgqQsVH/c6JkhYWBIGBgSBoaEsex6XNYwuMChXXbtmPef8jbMkDAwJAwMCQNDwsCQMGyUNcmLpZJO79h7mCFhYEgYGBIGhoSBIWFgSBgYEgaGhIEhYWBIGBgSxtLtHF0nqB0wV7vyndRdUielvpS6w6hzUlX4fcHcV4XaR6Xa653uavvju/GgDGgGYEgYGBIGhoSBIWHYT9qGibJ+kqoIqF5XkVKd5vaHVG19+Zfp+bS573ZzX3st0m0LVar/vlX910CJsmYAhoSBIWFgSBgYEsaIp7S1s0YVW7nz2twWmdWLi7zWIamanhkSBoaEgSFhYEgYI+5sfbVpc0kKhztGrH1Yq1tom5ZhisaZIWFgSBgYEgaGhIEhYYwYZf0u1T5fzR1rvzRtu0YNU7ZD6gQuC4aEgSFhYEgYGBKGLbYejzNSt0l9KFWlQddJVbH1WakbBx9XDsyQMDAkDAwJA0PCwJAwbJRVZTaH1+j6Y6mnpd6XqoLu9plrbnxu7fD5/gM0vbgc2ni42cAMCQNDwsCQMDAkDAwJY8RT2p6RcnHU21JvSH0mdbPUo1L1wdt9UklrfcPADAkDQ8LAkDAwJAybOqkDUg+Zq135SKoO+/pAqlInlaSoX0e7cLnuqz/qL/Qc3f/pmjrpupNQ17JrUiczAEPCwJAwMCQMDAnDpk6GKQ16Tqpcv1bqhNSrUu9IPSJ1i3n2fqkplpEKl6jp2tYVZkgYGBIGhoSBIWFgSBgj5rI+lXpSqsqA3GdpFTPtNNuqOKlGevyKxvgv05YBufOAF8022EMwJAwMCQNDwsCQMGwu67zUOlFWxVYVsXwr9bPUPVLHpOqDt20zqgdMz/MpCGqPlBkSBoaEgSFhYEgYGBLGiLmsz6WekHpPytUuubostz2TO3Hkxf4DNG+rCGi8Xyq5rFmBIWFgSBgYEob9o/691FFztSvVc7n+o9RbUl9JvSZ1r9SW1J9S9QWVK7ueN8yQMDAkDAwJA0PCwJAwRkydfC31kNS7UtuLS6nVMpfCqJEekap9GIfeDWiKXypHr84ADAkDQ8LAkDAwJIx/ABdK0KMxUTUMAAAAAElFTkSuQmCC\n",
      "text/plain": [
       "<PIL.Image.Image image mode=L size=400x180 at 0x7F376BC56CD0>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "display(float_array_to_image(1 - np.array(preds_after_fifty) ** 0.5))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 52,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<PIL.Image.Image image mode=L size=400x180 at 0x7F37806B3250>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "display(float_array_to_image(1 - np.array(preds_after_sixty) ** 0.125))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 54,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<PIL.Image.Image image mode=L size=400x180 at 0x7F37806B3310>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "display(float_array_to_image(1 - np.array(preds_after_fifty) ** 0.125))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 55,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<PIL.Image.Image image mode=L size=400x180 at 0x7F3780479850>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "display(float_array_to_image(1 - np.array(preds_after_sixty)))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 56,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<PIL.Image.Image image mode=L size=400x180 at 0x7F3780552E50>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "display(float_array_to_image(1 - np.array(preds_after_fifty)))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 59,
   "metadata": {},
   "outputs": [],
   "source": [
    "smootherstep = lambda n: 6 * (n ** 5.0) - 15 * (n ** 4.0) + 10 * (n ** 3.0)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 63,
   "metadata": {},
   "outputs": [],
   "source": [
    "for i in range(129):\n",
    "    smooth_i = smootherstep(i/128.0)\n",
    "    expt = 2 ** (-4 * smooth_i)\n",
    "    img60 = float_array_to_image(1 - np.array(preds_after_sixty) ** expt)\n",
    "    img50 = float_array_to_image(1 - np.array(preds_after_fifty) ** expt)\n",
    "    img60.save(f\"after{60000 + i}.gif\")\n",
    "    img50.save(f\"after{50000 + i}.gif\")\n",
    "    img60.save(f\"after{60000 + 256 - i}.gif\")\n",
    "    img50.save(f\"after{50000 + 256 - i}.gif\")\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  }
 ],
 "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": "3.7.7"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 4
}