week05/conformer_las.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 1,
   "id": "3d304bcf",
   "metadata": {},
   "outputs": [],
   "source": [
    "import os\n",
    "\n",
    "os.chdir('../asr/')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "id": "0cd0bab0",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "NOTE! Installing ujson may make loading annotations faster.\n"
     ]
    }
   ],
   "source": [
    "import heapq\n",
    "\n",
    "import torch\n",
    "import torch.nn.functional as F\n",
    "import pandas as pd\n",
    "import matplotlib.pyplot as plt\n",
    "import sentencepiece\n",
    "import omegaconf\n",
    "\n",
    "%reload_ext autoreload\n",
    "%autoreload 2\n",
    "from src.models import LASModel"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "167e9c9f",
   "metadata": {},
   "source": [
    "## LAS: Listen Attend and Spell\n"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "8fe48dd1",
   "metadata": {},
   "source": [
    "### Dataset"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "id": "b8ca1d38",
   "metadata": {},
   "outputs": [],
   "source": [
    "DATA_PATH = '../asr/data'\n",
    "TOKENIZER_PATH = '../asr/data/checkpoints_and_tokenizer/tokenizer/'\n",
    "CKPT_PATH = '../asr/data/checkpoints_and_tokenizer/conformer_las_2epochs.ckpt'"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "id": "29160281",
   "metadata": {},
   "outputs": [],
   "source": [
    "df = pd.read_json(os.path.join(DATA_PATH, \"test_opus/crowd/manifest.jsonl\"), lines=True)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "id": "52582074",
   "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>id</th>\n",
       "      <th>audio_filepath</th>\n",
       "      <th>text</th>\n",
       "      <th>duration</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>e632f7d39c15e7edfc665b91e6f2071f</td>\n",
       "      <td>files/e632f7d39c15e7edfc665b91e6f2071f.opus</td>\n",
       "      <td>афина воспроизведи музыку вперемешку</td>\n",
       "      <td>4.90000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>5db5df8bb9e3b6660b2a04b34d4a355d</td>\n",
       "      <td>files/5db5df8bb9e3b6660b2a04b34d4a355d.opus</td>\n",
       "      <td>найти сериал григорий р</td>\n",
       "      <td>3.65200</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>2c471aedc6979109f28cd53c58f8c4fb</td>\n",
       "      <td>files/2c471aedc6979109f28cd53c58f8c4fb.opus</td>\n",
       "      <td>прямой эфир апл манчестер юнайтед тоттенхэм</td>\n",
       "      <td>4.34175</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>756a137ee9debde4a008adc4a4121dc7</td>\n",
       "      <td>files/756a137ee9debde4a008adc4a4121dc7.opus</td>\n",
       "      <td>ильвиром ивановичем ворончихиным</td>\n",
       "      <td>3.90000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>1ee3b00170123a6723a40e129b2f6bce</td>\n",
       "      <td>files/1ee3b00170123a6723a40e129b2f6bce.opus</td>\n",
       "      <td>можешь показать киношку исходный код</td>\n",
       "      <td>3.32000</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "                                 id  \\\n",
       "0  e632f7d39c15e7edfc665b91e6f2071f   \n",
       "1  5db5df8bb9e3b6660b2a04b34d4a355d   \n",
       "2  2c471aedc6979109f28cd53c58f8c4fb   \n",
       "3  756a137ee9debde4a008adc4a4121dc7   \n",
       "4  1ee3b00170123a6723a40e129b2f6bce   \n",
       "\n",
       "                                audio_filepath  \\\n",
       "0  files/e632f7d39c15e7edfc665b91e6f2071f.opus   \n",
       "1  files/5db5df8bb9e3b6660b2a04b34d4a355d.opus   \n",
       "2  files/2c471aedc6979109f28cd53c58f8c4fb.opus   \n",
       "3  files/756a137ee9debde4a008adc4a4121dc7.opus   \n",
       "4  files/1ee3b00170123a6723a40e129b2f6bce.opus   \n",
       "\n",
       "                                          text  duration  \n",
       "0         афина воспроизведи музыку вперемешку   4.90000  \n",
       "1                      найти сериал григорий р   3.65200  \n",
       "2  прямой эфир апл манчестер юнайтед тоттенхэм   4.34175  \n",
       "3             ильвиром ивановичем ворончихиным   3.90000  \n",
       "4         можешь показать киношку исходный код   3.32000  "
      ]
     },
     "execution_count": 5,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "df.head()\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "id": "02be76f1",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "на           1381\n",
       "в            1253\n",
       "включи        640\n",
       "мне           603\n",
       "у             588\n",
       "фильм         482\n",
       "смотрешке     459\n",
       "тебя          459\n",
       "афина         452\n",
       "тв            432\n",
       "сезон         419\n",
       "джой          402\n",
       "сбер          398\n",
       "с             381\n",
       "салют         376\n",
       "есть          371\n",
       "и             335\n",
       "покажи        303\n",
       "два           271\n",
       "один          251\n",
       "Name: text, dtype: int64"
      ]
     },
     "execution_count": 6,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "df.text.str.split().explode().value_counts().head(20)\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "id": "700e190a",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAigAAAGdCAYAAAA44ojeAAAAOXRFWHRTb2Z0d2FyZQBNYXRwbG90bGliIHZlcnNpb24zLjcuMSwgaHR0cHM6Ly9tYXRwbG90bGliLm9yZy/bCgiHAAAACXBIWXMAAA9hAAAPYQGoP6dpAAAz2ElEQVR4nO3df1RU173+8QdwGEQdKCYM0ihqahONMVptcJq0zVUELU1NZKU14SYktdpatFFurdKrBjGN1vRGa0q07bJqV0LbeG9jE03VURNzU/F3uNWY0phFQlodWI0BVOowwvn+0S/TjoAyMjIb5v1aa5bOPvvssz9zBnk8Z86cKMuyLAEAABgkOtwTAAAAuBwBBQAAGIeAAgAAjENAAQAAxiGgAAAA4xBQAACAcQgoAADAOAQUAABgnF7hnsC1aG5u1unTp9WvXz9FRUWFezoAAKADLMvSuXPnlJqaqujoKx8j6ZYB5fTp0xo4cGC4pwEAAK7Bhx9+qJtuuumKfYIKKE1NTSoqKtLzzz8vj8ej1NRUPfroo1q8eLH/SIZlWXriiSf085//XLW1tbrrrru0bt06DRs2zD/O2bNnNXfuXL3yyiuKjo5WTk6OfvzjH6tv374dmke/fv38BTocjmBKaJPP59OuXbuUmZkpm83W6fG6E2qndmqPHNRO7eGuvb6+XgMHDvT/Hr+SoALKD3/4Q61bt06bN2/WbbfdpiNHjuixxx5TQkKCvvOd70iSVq1apbVr12rz5s0aMmSIlixZoqysLJ08eVJxcXGSpNzcXJ05c0Zut1s+n0+PPfaYZs2apdLS0g7NoyUMORyOkAWU+Ph4ORyOsO+8rkbt1E7tkYPaqd2U2jvy8YygAsr+/fs1depUZWdnS5IGDx6sX/3qVzp06JCkfxw9WbNmjRYvXqypU6dKkn75y1/K6XRq69atmj59ut555x3t2LFDhw8f1rhx4yRJzz77rL70pS/pRz/6kVJTU4MqEgAA9DxBBZTPfe5z+tnPfqY///nP+vSnP63/+7//05tvvqlnnnlGklRZWSmPx6OMjAz/OgkJCUpPT1dZWZmmT5+usrIyJSYm+sOJJGVkZCg6OloHDx7U/fff32q7Xq9XXq/X/7y+vl7SP1Khz+cLruI2tIwRirG6G2qn9khD7dQeaUyqPZg5BBVQFi1apPr6et16662KiYlRU1OTfvCDHyg3N1eS5PF4JElOpzNgPafT6V/m8XiUnJwcOIlevZSUlOTvc7kVK1Zo2bJlrdp37dql+Pj4YEq4IrfbHbKxuhtqj0zUHpmoPTKZUHtDQ0OH+wYVUF588UW98MILKi0t1W233aby8nLNmzdPqampysvLC3qiHVVYWKiCggL/85YP2WRmZobsMyhut1uTJk0y5vxcV6F2aqf2yEHt1B7u2lvOgHREUAFlwYIFWrRokaZPny5Juv322/XBBx9oxYoVysvLU0pKiiSpurpaAwYM8K9XXV2t0aNHS5JSUlJUU1MTMO6lS5d09uxZ//qXs9vtstvtrdptNltIX+xQj9edUDu1Rxpqp/ZIY0LtwWw/qG+SbWhoaPXFKjExMWpubpYkDRkyRCkpKdqzZ49/eX19vQ4ePCiXyyVJcrlcqq2t1dGjR/199u7dq+bmZqWnpwczHQAA0EMFdQTl3nvv1Q9+8AMNGjRIt912m9566y0988wz+vrXvy7pH5cNzZs3T08++aSGDRvmv8w4NTVV9913nyRp+PDhmjx5smbOnKn169fL5/Npzpw5mj59OlfwAAAASUEGlGeffVZLlizRt7/9bdXU1Cg1NVXf/OY3tXTpUn+f733ve7pw4YJmzZql2tpa3X333dqxY4f/O1Ak6YUXXtCcOXM0ceJE/xe1rV27NnRVAQCAbi2ogNKvXz+tWbNGa9asabdPVFSUiouLVVxc3G6fpKSkDn8pGwAAiDzczRgAABiHgAIAAIxDQAEAAMYhoAAAAOMQUAAAgHGCuooHPdfIop3yNgXe/vr9ldlhmg0AINJxBAUAABiHgAIAAIxDQAEAAMYhoAAAAOMQUAAAgHEIKAAAwDgEFAAAYBwCCgAAMA4BBQAAGIeAAgAAjENAAQAAxiGgAAAA4xBQAACAcQgoAADAOAQUAABgHAIKAAAwDgEFAAAYh4ACAACMQ0ABAADGIaAAAADjEFAAAIBxCCgAAMA4BBQAAGAcAgoAADAOAQUAABiHgAIAAIxDQAEAAMYhoAAAAOMQUAAAgHGCCiiDBw9WVFRUq0d+fr4k6eLFi8rPz1f//v3Vt29f5eTkqLq6OmCMqqoqZWdnKz4+XsnJyVqwYIEuXboUuooAAEC3F1RAOXz4sM6cOeN/uN1uSdIDDzwgSZo/f75eeeUVbdmyRfv27dPp06c1bdo0//pNTU3Kzs5WY2Oj9u/fr82bN2vTpk1aunRpCEsCAADdXa9gOt94440Bz1euXKmbb75ZX/ziF1VXV6cNGzaotLRUEyZMkCRt3LhRw4cP14EDBzR+/Hjt2rVLJ0+e1O7du+V0OjV69GgtX75cCxcuVFFRkWJjY9vcrtfrldfr9T+vr6+XJPl8Pvl8vqAKbkvLGKEYq7tpqdkebbW7rKdiv1N7pKF2ag+3YOYQZVlW699MHdDY2KjU1FQVFBTo+9//vvbu3auJEyfq448/VmJior9fWlqa5s2bp/nz52vp0qV6+eWXVV5e7l9eWVmpoUOH6tixYxozZkyb2yoqKtKyZctatZeWlio+Pv5apg8AALpYQ0ODHnroIdXV1cnhcFyxb1BHUP7V1q1bVVtbq0cffVSS5PF4FBsbGxBOJMnpdMrj8fj7OJ3OVstblrWnsLBQBQUF/uf19fUaOHCgMjMzr1pgR/h8Prndbk2aNEk2m63T43UnLbUvORItb3NUwLITRVlhmlXXYL9TO7VHDmo3o/aWMyAdcc0BZcOGDZoyZYpSU1OvdYgOs9vtstvtrdptNltIX+xQj9edeJuj5G0KDCiR8lpE8n6ndmqPNNQe3tqD2f41XWb8wQcfaPfu3frGN77hb0tJSVFjY6Nqa2sD+lZXVyslJcXf5/Krelqet/QBAAC4poCyceNGJScnKzs72982duxY2Ww27dmzx99WUVGhqqoquVwuSZLL5dLx48dVU1Pj7+N2u+VwODRixIhrrQEAAPQwQZ/iaW5u1saNG5WXl6devf65ekJCgmbMmKGCggIlJSXJ4XBo7ty5crlcGj9+vCQpMzNTI0aM0MMPP6xVq1bJ4/Fo8eLFys/Pb/MUDgAAiExBB5Tdu3erqqpKX//611stW716taKjo5WTkyOv16usrCw999xz/uUxMTHatm2bZs+eLZfLpT59+igvL0/FxcWdqwIAAPQoQQeUzMxMtXdlclxcnEpKSlRSUtLu+mlpaXr11VeD3SwAAIgg13wVDyLX4EXb22x/f2V2m+0AAASLmwUCAADjEFAAAIBxCCgAAMA4BBQAAGAcAgoAADAOAQUAABiHgAIAAIxDQAEAAMYhoAAAAOMQUAAAgHEIKAAAwDgEFAAAYBwCCgAAMA4BBQAAGIeAAgAAjENAAQAAxiGgAAAA4xBQAACAcQgoAADAOAQUAABgHAIKAAAwDgEFAAAYh4ACAACMQ0ABAADGIaAAAADjEFAAAIBxCCgAAMA4BBQAAGAcAgoAADAOAQUAABiHgAIAAIxDQAEAAMYhoAAAAOMEHVD++te/6t///d/Vv39/9e7dW7fffruOHDniX25ZlpYuXaoBAwaod+/eysjI0LvvvhswxtmzZ5WbmyuHw6HExETNmDFD58+f73w1AACgRwgqoHz88ce66667ZLPZ9Pvf/14nT57Uf/3Xf+kTn/iEv8+qVau0du1arV+/XgcPHlSfPn2UlZWlixcv+vvk5ubq7bffltvt1rZt2/TGG29o1qxZoasKAAB0a72C6fzDH/5QAwcO1MaNG/1tQ4YM8f/dsiytWbNGixcv1tSpUyVJv/zlL+V0OrV161ZNnz5d77zzjnbs2KHDhw9r3LhxkqRnn31WX/rSl/SjH/1IqampoagLAAB0Y0EFlJdffllZWVl64IEHtG/fPn3yk5/Ut7/9bc2cOVOSVFlZKY/Ho4yMDP86CQkJSk9PV1lZmaZPn66ysjIlJib6w4kkZWRkKDo6WgcPHtT999/farter1der9f/vL6+XpLk8/nk8/mCq7gNLWOEYqzupqVme7TV7rLL2WNa971Sf1Ox36k90lA7tYdbMHOIsiyr7d82bYiLi5MkFRQU6IEHHtDhw4f1+OOPa/369crLy9P+/ft111136fTp0xowYIB/va9+9auKiorSb37zGz311FPavHmzKioqAsZOTk7WsmXLNHv27FbbLSoq0rJly1q1l5aWKj4+vsPFAgCA8GloaNBDDz2kuro6ORyOK/YN6ghKc3Ozxo0bp6eeekqSNGbMGJ04ccIfUK6XwsJCFRQU+J/X19dr4MCByszMvGqBHeHz+eR2uzVp0iTZbLZOj9edtNS+5Ei0vM1RActOFGW1uc7Iop1ttrfX31Tsd2qn9shB7WbU3nIGpCOCCigDBgzQiBEjAtqGDx+u//mf/5EkpaSkSJKqq6sDjqBUV1dr9OjR/j41NTUBY1y6dElnz571r385u90uu93eqt1ms4X0xQ71eN2JtzlK3qbAgNLea3F5v6v1N10k73dqp/ZIQ+3hrT2Y7Qd1Fc9dd93V6tTMn//8Z6WlpUn6xwdmU1JStGfPHv/y+vp6HTx4UC6XS5LkcrlUW1uro0eP+vvs3btXzc3NSk9PD2Y6AACghwrqCMr8+fP1uc99Tk899ZS++tWv6tChQ/rZz36mn/3sZ5KkqKgozZs3T08++aSGDRumIUOGaMmSJUpNTdV9990n6R9HXCZPnqyZM2dq/fr18vl8mjNnjqZPn84VPAAAQFKQAeWzn/2sXnrpJRUWFqq4uFhDhgzRmjVrlJub6+/zve99TxcuXNCsWbNUW1uru+++Wzt27PB/wFaSXnjhBc2ZM0cTJ05UdHS0cnJytHbt2tBVBQAAurWgAookffnLX9aXv/zldpdHRUWpuLhYxcXF7fZJSkpSaWlpsJsGAAARIuiAAnMMXrS9zfb3V2Z38UwAAAgtbhYIAACMQ0ABAADGIaAAAADjEFAAAIBxCCgAAMA4BBQAAGAcAgoAADAOAQUAABiHgAIAAIxDQAEAAMYhoAAAAOMQUAAAgHEIKAAAwDgEFAAAYJxe4Z4AItfgRdvbbH9/ZXYXzwQAYBqOoAAAAOMQUAAAgHEIKAAAwDgEFAAAYBwCCgAAMA4BBQAAGIeAAgAAjENAAQAAxiGgAAAA4xBQAACAcQgoAADAOAQUAABgHAIKAAAwDgEFAAAYh4ACAACMQ0ABAADGIaAAAADjEFAAAIBxCCgAAMA4QQWUoqIiRUVFBTxuvfVW//KLFy8qPz9f/fv3V9++fZWTk6Pq6uqAMaqqqpSdna34+HglJydrwYIFunTpUmiqAQAAPUKvYFe47bbbtHv37n8O0OufQ8yfP1/bt2/Xli1blJCQoDlz5mjatGn6wx/+IElqampSdna2UlJStH//fp05c0aPPPKIbDabnnrqqRCUAwAAeoKgA0qvXr2UkpLSqr2urk4bNmxQaWmpJkyYIEnauHGjhg8frgMHDmj8+PHatWuXTp48qd27d8vpdGr06NFavny5Fi5cqKKiIsXGxra5Ta/XK6/X639eX18vSfL5fPL5fMGW0ErLGKEYqyvZY6w224Opo6WvPbr1WLf857Z2tnvlsToqFPPvjO6630OB2qk90lC7GbUHM4coy7La/i3RhqKiIj399NNKSEhQXFycXC6XVqxYoUGDBmnv3r2aOHGiPv74YyUmJvrXSUtL07x58zR//nwtXbpUL7/8ssrLy/3LKysrNXToUB07dkxjxoxpd7vLli1r1V5aWqr4+PgOFwsAAMKnoaFBDz30kOrq6uRwOK7YN6gjKOnp6dq0aZNuueUWnTlzRsuWLdPnP/95nThxQh6PR7GxsQHhRJKcTqc8Ho8kyePxyOl0tlresqw9hYWFKigo8D+vr6/XwIEDlZmZedUCO8Ln88ntdmvSpEmy2WydHu9ajSza2Wb7iaKskPRvS0vtS45Ey9sc1eH1OrtdKTTz7wxT9ns4UDu1U3vkMKn2ljMgHRFUQJkyZYr/76NGjVJ6errS0tL04osvqnfv3sEMFRS73S673d6q3WazhfTFDvV4wfI2tR0Q2ptTsP2vuO3mqHbH66hgtxvK+XdGuPd7OFE7tUcaag9v7cFsv1OXGScmJurTn/60Tp06pZSUFDU2Nqq2tjagT3V1tf8zKykpKa2u6ml53tbnWgAAQGTqVEA5f/683nvvPQ0YMEBjx46VzWbTnj17/MsrKipUVVUll8slSXK5XDp+/Lhqamr8fdxutxwOh0aMGNGZqQAAgB4kqFM83/3ud3XvvfcqLS1Np0+f1hNPPKGYmBg9+OCDSkhI0IwZM1RQUKCkpCQ5HA7NnTtXLpdL48ePlyRlZmZqxIgRevjhh7Vq1Sp5PB4tXrxY+fn5bZ7CAQAAkSmogPKXv/xFDz74oD766CPdeOONuvvuu3XgwAHdeOONkqTVq1crOjpaOTk58nq9ysrK0nPPPedfPyYmRtu2bdPs2bPlcrnUp08f5eXlqbi4OLRVAQCAbi2ogPLrX//6isvj4uJUUlKikpKSdvukpaXp1VdfDWazAAAgwnAvHgAAYBwCCgAAME7QX3UPtGfwou1ttr+/MruLZwIA6O44ggIAAIxDQAEAAMYhoAAAAOMQUAAAgHEIKAAAwDgEFAAAYBwCCgAAMA4BBQAAGIeAAgAAjENAAQAAxiGgAAAA4xBQAACAcQgoAADAOAQUAABgHAIKAAAwTq9wTwBXN3jR9nBPAQCALsURFAAAYBwCCgAAMA6neHDdcYoKABAsjqAAAADjEFAAAIBxCCgAAMA4fAYlgrT1WRB7jKVVd4ZhMgAAXAFHUAAAgHEIKAAAwDgEFAAAYBwCCgAAMA4BBQAAGIereNDttfdNte+vzO7imQAAQoUjKAAAwDgEFAAAYJxOBZSVK1cqKipK8+bN87ddvHhR+fn56t+/v/r27aucnBxVV1cHrFdVVaXs7GzFx8crOTlZCxYs0KVLlzozFQAA0INcc0A5fPiwfvrTn2rUqFEB7fPnz9crr7yiLVu2aN++fTp9+rSmTZvmX97U1KTs7Gw1NjZq//792rx5szZt2qSlS5deexUAAKBHuaaAcv78eeXm5urnP/+5PvGJT/jb6+rqtGHDBj3zzDOaMGGCxo4dq40bN2r//v06cOCAJGnXrl06efKknn/+eY0ePVpTpkzR8uXLVVJSosbGxtBUBQAAurVruoonPz9f2dnZysjI0JNPPulvP3r0qHw+nzIyMvxtt956qwYNGqSysjKNHz9eZWVluv322+V0Ov19srKyNHv2bL399tsaM2ZMq+15vV55vV7/8/r6ekmSz+eTz+e7lhICtIwRirE6wx5jhWSc9upoa3x7tBXwpwmC3Q/tvW5XG8eU/R4O1E7tkYbazag9mDkEHVB+/etf69ixYzp8+HCrZR6PR7GxsUpMTAxodzqd8ng8/j7/Gk5alrcsa8uKFSu0bNmyVu27du1SfHx8sCW0y+12h2ysaxGqm/a9+uqrQY+/fFxzaDYeAu3Nvz3t1dXRccK938OJ2iMTtUcmE2pvaGjocN+gAsqHH36oxx9/XG63W3FxcUFP7FoVFhaqoKDA/7y+vl4DBw5UZmamHA5Hp8f3+Xxyu92aNGmSbDZb0OuPLNrZZvuJoqyQjBOs9rbb1vj2aEvLxzVryZFoeZujQrL9zgrV63a1cTq737szaqd2ao8cJtXecgakI4IKKEePHlVNTY0+85nP+Nuampr0xhtv6Cc/+Yl27typxsZG1dbWBhxFqa6uVkpKiiQpJSVFhw4dChi35Sqflj6Xs9vtstvtrdptNltIX+xrHc/b1PYv9mDHam+cYA1bsqudJe2P722OCtn2OytUr1tHxwn1+6g7oXZqjzTUHt7ag9l+UB+SnThxoo4fP67y8nL/Y9y4ccrNzfX/3Wazac+ePf51KioqVFVVJZfLJUlyuVw6fvy4ampq/H3cbrccDodGjBgRzHQAAEAPFdQRlH79+mnkyJEBbX369FH//v397TNmzFBBQYGSkpLkcDg0d+5cuVwujR8/XpKUmZmpESNG6OGHH9aqVavk8Xi0ePFi5efnt3mUBAAARJ6Q34tn9erVio6OVk5Ojrxer7KysvTcc8/5l8fExGjbtm2aPXu2XC6X+vTpo7y8PBUXF4d6KgAAoJvqdEB5/fXXA57HxcWppKREJSUl7a6TlpYW9JUaAAAgcnAvHgAAYBwCCgAAMA4BBQAAGIeAAgAAjENAAQAAxgn5ZcZAZw1etL3N9vdXZnfxTAAA4cIRFAAAYBwCCgAAMA4BBQAAGIeAAgAAjENAAQAAxiGgAAAA4xBQAACAcQgoAADAOAQUAABgHAIKAAAwDgEFAAAYh4ACAACMw80C0W20dxNBAEDPwxEUAABgHAIKAAAwDgEFAAAYh4ACAACMQ0ABAADGIaAAAADjEFAAAIBxCCgAAMA4BBQAAGAcAgoAADAOAQUAABiHgAIAAIxDQAEAAMYhoAAAAOMQUAAAgHEIKAAAwDhBBZR169Zp1KhRcjgccjgccrlc+v3vf+9ffvHiReXn56t///7q27evcnJyVF1dHTBGVVWVsrOzFR8fr+TkZC1YsECXLl0KTTUAAKBHCCqg3HTTTVq5cqWOHj2qI0eOaMKECZo6darefvttSdL8+fP1yiuvaMuWLdq3b59Onz6tadOm+ddvampSdna2GhsbtX//fm3evFmbNm3S0qVLQ1sVAADo1noF0/nee+8NeP6DH/xA69at04EDB3TTTTdpw4YNKi0t1YQJEyRJGzdu1PDhw3XgwAGNHz9eu3bt0smTJ7V79245nU6NHj1ay5cv18KFC1VUVKTY2Ng2t+v1euX1ev3P6+vrJUk+n08+ny+ogtvSMsa1jmWPsa44bmfHuZ7s0VbAnz3J1V7/zu737ozaqT3SULsZtQczhyjLsq7pN1NTU5O2bNmivLw8vfXWW/J4PJo4caI+/vhjJSYm+vulpaVp3rx5mj9/vpYuXaqXX35Z5eXl/uWVlZUaOnSojh07pjFjxrS5raKiIi1btqxVe2lpqeLj469l+gAAoIs1NDTooYceUl1dnRwOxxX7BnUERZKOHz8ul8ulixcvqm/fvnrppZc0YsQIlZeXKzY2NiCcSJLT6ZTH45EkeTweOZ3OVstblrWnsLBQBQUF/uf19fUaOHCgMjMzr1pgR/h8Prndbk2aNEk2my3o9UcW7Wyz/URRVkjGuZ7s0ZaWj2vWkiPR8jZHdfn2r6ervf6d3e/dGbVTO7VHDpNqbzkD0hFBB5RbbrlF5eXlqqur03//938rLy9P+/btC3aYoNjtdtnt9lbtNpstpC/2tY7nbWr7F3uwY7U3TlfwNkeFdfvXQ0df/1C/j7oTaqf2SEPt4a09mO0HHVBiY2P1qU99SpI0duxYHT58WD/+8Y/1ta99TY2NjaqtrQ04ilJdXa2UlBRJUkpKig4dOhQwXstVPi19AAAAOv09KM3NzfJ6vRo7dqxsNpv27NnjX1ZRUaGqqiq5XC5Jksvl0vHjx1VTU+Pv43a75XA4NGLEiM5OBeiQwYu2a/Ci7f5TaiOLdmrwou1hnhUA4F8FdQSlsLBQU6ZM0aBBg3Tu3DmVlpbq9ddf186dO5WQkKAZM2aooKBASUlJcjgcmjt3rlwul8aPHy9JyszM1IgRI/Twww9r1apV8ng8Wrx4sfLz89s8hQMAACJTUAGlpqZGjzzyiM6cOaOEhASNGjVKO3fu1KRJkyRJq1evVnR0tHJycuT1epWVlaXnnnvOv35MTIy2bdum2bNny+VyqU+fPsrLy1NxcXFoqwIAAN1aUAFlw4YNV1weFxenkpISlZSUtNsnLS1Nr776ajCbBQAAEYZ78QAAAOMQUAAAgHEIKAAAwDgEFAAAYBwCCgAAMA4BBQAAGIeAAgAAjENAAQAAxiGgAAAA4xBQAACAcQgoAADAOAQUAABgHAIKAAAwDgEFAAAYh4ACAACMQ0ABAADGIaAAAADj9Ar3BCLR4EXbwz0FAACMxhEUAABgHAIKAAAwDgEFAAAYh4ACAACMQ0ABAADGIaAAAADjEFAAAIBxCCgAAMA4BBQAAGAcAgoAADAOAQUAABiHe/Ggx+KeRwDQfXEEBQAAGIeAAgAAjENAAQAAxiGgAAAA4xBQAACAcYIKKCtWrNBnP/tZ9evXT8nJybrvvvtUUVER0OfixYvKz89X//791bdvX+Xk5Ki6ujqgT1VVlbKzsxUfH6/k5GQtWLBAly5d6nw1AACgRwgqoOzbt0/5+fk6cOCA3G63fD6fMjMzdeHCBX+f+fPn65VXXtGWLVu0b98+nT59WtOmTfMvb2pqUnZ2thobG7V//35t3rxZmzZt0tKlS0NXFQAA6NaC+h6UHTt2BDzftGmTkpOTdfToUX3hC19QXV2dNmzYoNLSUk2YMEGStHHjRg0fPlwHDhzQ+PHjtWvXLp08eVK7d++W0+nU6NGjtXz5ci1cuFBFRUWKjY1ttV2v1yuv1+t/Xl9fL0ny+Xzy+XxBF325ljGudSx7jHXFcTvaPxzs0VbAn5Hk8tpD8V7qLjr7nu/OqJ3aI41JtQczhyjLsq75N9OpU6c0bNgwHT9+XCNHjtTevXs1ceJEffzxx0pMTPT3S0tL07x58zR//nwtXbpUL7/8ssrLy/3LKysrNXToUB07dkxjxoxptZ2ioiItW7asVXtpaani4+OvdfoAAKALNTQ06KGHHlJdXZ0cDscV+17zN8k2Nzdr3rx5uuuuuzRy5EhJksfjUWxsbEA4kSSn0ymPx+Pv43Q6Wy1vWdaWwsJCFRQU+J/X19dr4MCByszMvGqBHeHz+eR2uzVp0iTZbLag1x9ZtLPTcwgXe7Sl5eOateRItLzNUeGeTpe6vPYTRVnhnlKX6ex7vjujdmqn9vBpOQPSEdccUPLz83XixAm9+eab1zpEh9ntdtnt9lbtNpstpC/2tY7nber+v9i9zVE9oo5r0VJ7uH9wwyHUP0PdCbVTe6QxofZgtn9NlxnPmTNH27Zt02uvvaabbrrJ356SkqLGxkbV1tYG9K+urlZKSoq/z+VX9bQ8b+kDAAAiW1ABxbIszZkzRy+99JL27t2rIUOGBCwfO3asbDab9uzZ42+rqKhQVVWVXC6XJMnlcun48eOqqanx93G73XI4HBoxYkRnagEAAD1EUKd48vPzVVpaqt/97nfq16+f/zMjCQkJ6t27txISEjRjxgwVFBQoKSlJDodDc+fOlcvl0vjx4yVJmZmZGjFihB5++GGtWrVKHo9HixcvVn5+fpuncQBTtXe35PdXZnfxTACg5wkqoKxbt06SdM899wS0b9y4UY8++qgkafXq1YqOjlZOTo68Xq+ysrL03HPP+fvGxMRo27Ztmj17tlwul/r06aO8vDwVFxd3rhIAANBjBBVQOnJFclxcnEpKSlRSUtJun7S0NL366qvBbBoAAEQQ7sUDAACMQ0ABAADGIaAAAADjEFAAAIBxCCgAAMA4BBQAAGAcAgoAADDONd8sEEBw+OZZAOg4jqAAAADjEFAAAIBxCCgAAMA4BBQAAGAcAgoAADAOV/EA/x9X2QCAOQgowFW0F1wAANcPp3gAAIBxCCgAAMA4BBQAAGAcPoMSBD6LAABA1+AICgAAMA4BBQAAGIeAAgAAjENAAQAAxuFDsm3gw7AAAIQXR1AAAIBxCCgAAMA4nOIBuhluagggEnAEBQAAGIeAAgAAjENAAQAAxiGgAAAA4/AhWSDE+B4dAOg8jqAAAADjEFAAAIBxgg4ob7zxhu69916lpqYqKipKW7duDVhuWZaWLl2qAQMGqHfv3srIyNC7774b0Ofs2bPKzc2Vw+FQYmKiZsyYofPnz3eqEAAA0HMEHVAuXLigO+64QyUlJW0uX7VqldauXav169fr4MGD6tOnj7KysnTx4kV/n9zcXL399ttyu93atm2b3njjDc2aNevaqwAAAD1K0B+SnTJliqZMmdLmMsuytGbNGi1evFhTp06VJP3yl7+U0+nU1q1bNX36dL3zzjvasWOHDh8+rHHjxkmSnn32WX3pS1/Sj370I6WmpnaiHAAA0BOE9CqeyspKeTweZWRk+NsSEhKUnp6usrIyTZ8+XWVlZUpMTPSHE0nKyMhQdHS0Dh48qPvvv7/VuF6vV16v1/+8vr5ekuTz+eTz+To975YxWv60x1idHrO7sEdbAX9GElNqD/Y93N77M5hxLn/PRxJqp/ZIY1LtwcwhpAHF4/FIkpxOZ0C70+n0L/N4PEpOTg6cRK9eSkpK8ve53IoVK7Rs2bJW7bt27VJ8fHwopi5JcrvdkqRVd4ZsyG5j+bjmcE8hbMJd+6uvvhpU//ben8GOI/3zPR+JqD0yUXt4NTQ0dLhvt/gelMLCQhUUFPif19fXa+DAgcrMzJTD4ej0+D6fT263W5MmTZLNZtPIop2dHrO7sEdbWj6uWUuORMvbHBXu6XQpU2o/UZQVVP/23p/BjHP5ez6SUDu1U3v4tJwB6YiQBpSUlBRJUnV1tQYMGOBvr66u1ujRo/19ampqAta7dOmSzp4961//cna7XXa7vVW7zWYL6YvdMp63KbJ+UUuStzkqIuuWwl97sO/h9uZ6LT8Lof4Z6k6ondojjQm1B7P9kH4PypAhQ5SSkqI9e/b42+rr63Xw4EG5XC5JksvlUm1trY4ePervs3fvXjU3Nys9PT2U0wEAAN1U0EdQzp8/r1OnTvmfV1ZWqry8XElJSRo0aJDmzZunJ598UsOGDdOQIUO0ZMkSpaam6r777pMkDR8+XJMnT9bMmTO1fv16+Xw+zZkzR9OnT+cKHgAAIOkaAsqRI0f0b//2b/7nLZ8NycvL06ZNm/S9731PFy5c0KxZs1RbW6u7775bO3bsUFxcnH+dF154QXPmzNHEiRMVHR2tnJwcrV27NgTlAACAniDogHLPPffIstq/LDMqKkrFxcUqLi5ut09SUpJKS0uD3TQAAIgQ3eIqHqAn4+7HANAaNwsEAADGIaAAAADjEFAAAIBxCCgAAMA4BBQAAGAcAgoAADAOAQUAABiHgAIAAIxDQAEAAMYhoAAAAOPwVfdAD9HeV+a/vzK7i2cCAJ3HERQAAGAcAgoAADAOAQUAABiHz6AAPVxbn02xx1hadac0sminvE1RAcv4zAoAE3AEBQAAGIeAAgAAjENAAQAAxiGgAAAA4xBQAACAcbiKB0CH8E21ALoSAQVAgPaCCAB0JU7xAAAA4xBQAACAcQgoAADAOAQUAABgHAIKAAAwDgEFAAAYh8uMAXQK348C4HogoAC4LoL9PhUCDYB/xSkeAABgHI6gADAap5CAyERAAdAtcQoJ6NkIKACMwD2AAPyrsH4GpaSkRIMHD1ZcXJzS09N16NChcE4HQA82eNF2DV60XSOLdkqSRhbtJBQBBgvbEZTf/OY3Kigo0Pr165Wenq41a9YoKytLFRUVSk5ODte0AESY7nKqiM/iINKELaA888wzmjlzph577DFJ0vr167V9+3b94he/0KJFiwL6er1eeb1e//O6ujpJ0tmzZ+Xz+To9F5/Pp4aGBn300Uey2WzqdelCp8fsLno1W2poaFYvX7SamqPCPZ0uRe3Ufi21f+q7L7bZfrBwYpvt6Sv2hKR/e/9Yf/TRR+0saa3l37rR//lbeS+rvb35BKu9+bcnVNu9msv/nW9PsPurO+ho7V3h3LlzkiTLsq7e2QoDr9drxcTEWC+99FJA+yOPPGJ95StfadX/iSeesCTx4MGDBw8ePHrA48MPP7xqVgjLEZS//e1vampqktPpDGh3Op3605/+1Kp/YWGhCgoK/M+bm5t19uxZ9e/fX1FRnf/fX319vQYOHKgPP/xQDoej0+N1J9RO7dQeOaid2sNdu2VZOnfunFJTU6/at1tcxWO322W32wPaEhMTQ74dh8MR9p0XLtRO7ZGG2qk90phSe0JCQof6heUqnhtuuEExMTGqrq4OaK+urlZKSko4pgQAAAwSloASGxursWPHas+ef34Yqbm5WXv27JHL5QrHlAAAgEHCdoqnoKBAeXl5GjdunO68806tWbNGFy5c8F/V05XsdrueeOKJVqeRIgG1U3ukoXZqjzTdtfYoy+rItT7Xx09+8hM9/fTT8ng8Gj16tNauXav09PRwTQcAABgirAEFAACgLWH9qnsAAIC2EFAAAIBxCCgAAMA4BBQAAGAcAoqkkpISDR48WHFxcUpPT9ehQ4fCPaWQWrFihT772c+qX79+Sk5O1n333aeKioqAPvfcc4+ioqICHt/61rfCNOPQKSoqalXXrbfe6l9+8eJF5efnq3///urbt69ycnJafYFgdzV48OBWtUdFRSk/P19Sz9rnb7zxhu69916lpqYqKipKW7duDVhuWZaWLl2qAQMGqHfv3srIyNC7774b0Ofs2bPKzc2Vw+FQYmKiZsyYofPnz3dhFdfmSrX7fD4tXLhQt99+u/r06aPU1FQ98sgjOn36dMAYbb1XVq5c2cWVBO9q+/3RRx9tVdfkyZMD+vTE/S6pzZ/9qKgoPf300/4+pu/3iA8ov/nNb1RQUKAnnnhCx44d0x133KGsrCzV1NSEe2ohs2/fPuXn5+vAgQNyu93y+XzKzMzUhQuBd22eOXOmzpw543+sWrUqTDMOrdtuuy2grjfffNO/bP78+XrllVe0ZcsW7du3T6dPn9a0adPCONvQOXz4cEDdbrdbkvTAAw/4+/SUfX7hwgXdcccdKikpaXP5qlWrtHbtWq1fv14HDx5Unz59lJWVpYsXL/r75Obm6u2335bb7da2bdv0xhtvaNasWV1VwjW7Uu0NDQ06duyYlixZomPHjum3v/2tKioq9JWvfKVV3+Li4oD3wty5c7ti+p1ytf0uSZMnTw6o61e/+lXA8p643yUF1HzmzBn94he/UFRUlHJycgL6Gb3fO39v4u7tzjvvtPLz8/3Pm5qarNTUVGvFihVhnNX1VVNTY0my9u3b52/74he/aD3++OPhm9R18sQTT1h33HFHm8tqa2stm81mbdmyxd/2zjvvWJKssrKyLpph13n88cetm2++2WpubrYsq+fuc0kBd0pvbm62UlJSrKefftrfVltba9ntdutXv/qVZVmWdfLkSUuSdfjwYX+f3//+91ZUVJT117/+tcvm3lmX196WQ4cOWZKsDz74wN+WlpZmrV69+vpO7jprq/a8vDxr6tSp7a4TSft96tSp1oQJEwLaTN/vEX0EpbGxUUePHlVGRoa/LTo6WhkZGSorKwvjzK6vuro6SVJSUlJA+wsvvKAbbrhBI0eOVGFhoRoaGsIxvZB79913lZqaqqFDhyo3N1dVVVWSpKNHj8rn8wXs/1tvvVWDBg3qcfu/sbFRzz//vL7+9a8H3AG8p+7zf1VZWSmPxxOwnxMSEpSenu7fz2VlZUpMTNS4ceP8fTIyMhQdHa2DBw92+Zyvp7q6OkVFRbW64erKlSvVv39/jRkzRk8//bQuXboUngmG2Ouvv67k5GTdcsstmj17tj766CP/skjZ79XV1dq+fbtmzJjRapnJ+71b3M34evnb3/6mpqYmOZ3OgHan06k//elPYZrV9dXc3Kx58+bprrvu0siRI/3tDz30kNLS0pSamqo//vGPWrhwoSoqKvTb3/42jLPtvPT0dG3atEm33HKLzpw5o2XLlunzn/+8Tpw4IY/Ho9jY2Fb/UDudTnk8nvBM+DrZunWramtr9eijj/rbeuo+v1zLvmzr57xlmcfjUXJycsDyXr16KSkpqUe9Fy5evKiFCxfqwQcfDLir7Xe+8x195jOfUVJSkvbv36/CwkKdOXNGzzzzTBhn23mTJ0/WtGnTNGTIEL333nv6/ve/rylTpqisrEwxMTERs983b96sfv36tTp9bfp+j+iAEony8/N14sSJgM9hSAo453r77bdrwIABmjhxot577z3dfPPNXT3NkJkyZYr/76NGjVJ6errS0tL04osvqnfv3mGcWdfasGGDpkyZotTUVH9bT93naJvP59NXv/pVWZaldevWBSwrKCjw/33UqFGKjY3VN7/5Ta1YsaLb3b/lX02fPt3/99tvv12jRo3SzTffrNdff10TJ04M48y61i9+8Qvl5uYqLi4uoN30/R7Rp3huuOEGxcTEtLpqo7q6WikpKWGa1fUzZ84cbdu2Ta+99ppuuummK/ZtuSfSqVOnumJqXSYxMVGf/vSnderUKaWkpKixsVG1tbUBfXra/v/ggw+0e/dufeMb37hiv566z1v25ZV+zlNSUlp9MP7SpUs6e/Zsj3gvtISTDz74QG63O+DoSVvS09N16dIlvf/++10zwS4ydOhQ3XDDDf73eE/f75L0v//7v6qoqLjqz79k3n6P6IASGxursWPHas+ePf625uZm7dmzRy6XK4wzCy3LsjRnzhy99NJL2rt3r4YMGXLVdcrLyyVJAwYMuM6z61rnz5/Xe++9pwEDBmjs2LGy2WwB+7+iokJVVVU9av9v3LhRycnJys7OvmK/nrrPhwwZopSUlID9XF9fr4MHD/r3s8vlUm1trY4ePervs3fvXjU3N3f7G5i2hJN3331Xu3fvVv/+/a+6Tnl5uaKjo1ud/uju/vKXv+ijjz7yv8d78n5vsWHDBo0dO1Z33HHHVfsat9/D/SndcPv1r39t2e12a9OmTdbJkyetWbNmWYmJiZbH4wn31EJm9uzZVkJCgvX6669bZ86c8T8aGhosy7KsU6dOWcXFxdaRI0esyspK63e/+501dOhQ6wtf+EKYZ955//Ef/2G9/vrrVmVlpfWHP/zBysjIsG644QarpqbGsizL+ta3vmUNGjTI2rt3r3XkyBHL5XJZLpcrzLMOnaamJmvQoEHWwoULA9p72j4/d+6c9dZbb1lvvfWWJcl65plnrLfeest/pcrKlSutxMRE63e/+531xz/+0Zo6dao1ZMgQ6+9//7t/jMmTJ1tjxoyxDh48aL355pvWsGHDrAcffDBcJXXYlWpvbGy0vvKVr1g33XSTVV5eHvDz7/V6LcuyrP3791urV6+2ysvLrffee896/vnnrRtvvNF65JFHwlzZ1V2p9nPnzlnf/e53rbKyMquystLavXu39ZnPfMYaNmyYdfHiRf8YPXG/t6irq7Pi4+OtdevWtVq/O+z3iA8olmVZzz77rDVo0CArNjbWuvPOO60DBw6Ee0ohJanNx8aNGy3LsqyqqirrC1/4gpWUlGTZ7XbrU5/6lLVgwQKrrq4uvBMPga997WvWgAEDrNjYWOuTn/yk9bWvfc06deqUf/nf//5369vf/rb1iU98woqPj7fuv/9+68yZM2GccWjt3LnTkmRVVFQEtPe0ff7aa6+1+R7Py8uzLOsflxovWbLEcjqdlt1utyZOnNjqNfnoo4+sBx980Orbt6/lcDisxx57zDp37lwYqgnOlWqvrKxs9+f/tddesyzLso4ePWqlp6dbCQkJVlxcnDV8+HDrqaeeCvglbqor1d7Q0GBlZmZaN954o2Wz2ay0tDRr5syZrf7z2RP3e4uf/vSnVu/eva3a2tpW63eH/R5lWZZ1XQ/RAAAABCmiP4MCAADMREABAADGIaAAAADjEFAAAIBxCCgAAMA4BBQAAGAcAgoAADAOAQUAABiHgAIAAIxDQAEAAMYhoAAAAOP8P5+/5ADuTmYRAAAAAElFTkSuQmCC",
      "text/plain": [
       "<Figure size 640x480 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "df.text.str.len().hist(bins=75);"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "id": "72e62b63",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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",
      "text/plain": [
       "<Figure size 640x480 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "df.duration.hist(bins=100);\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "id": "b1d18e2d",
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "24.39024390243902"
      ]
     },
     "execution_count": 9,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "max(df.text.str.len() / df.duration)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "id": "4751f82d",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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",
      "text/plain": [
       "<Figure size 1500x800 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "plt.figure(figsize=(15, 8))\n",
    "(df.text.str.len() / df.duration).hist(bins=50);\n",
    "plt.axvline(x=12.5, ymin=0, ymax=1.0, color='red', label='Limit for Stride = 8')\n",
    "plt.axvline(x=25, ymin=0, ymax=1.0, color='green', label='Limit for Stride = 4')\n",
    "plt.legend(fontsize=15)\n",
    "plt.title('Distribution of characters per second over dataset', fontsize=15);"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "61dfe4f6",
   "metadata": {},
   "source": [
    "### Tokenizer"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "id": "9d39fd6f",
   "metadata": {},
   "outputs": [],
   "source": [
    "tokenizer = sentencepiece.SentencePieceProcessor(\n",
    "    model_file=os.path.join(TOKENIZER_PATH, \"bpe_1024_bos_eos.model\")\n",
    ")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "id": "c55dbf0b",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "['▁ма', 'ма', '▁м', 'ы', 'ла', '▁ра', 'му']"
      ]
     },
     "execution_count": 12,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "tokenizer.encode_as_pieces('мама мыла раму')\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "id": "69d21a13",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "[112, 119, 12, 1014, 55, 176, 304]"
      ]
     },
     "execution_count": 13,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "tokenizer.encode_as_ids('мама мыла раму')\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "id": "201e48b8",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "iVBORw0KGgoAAAANSUhEUgAAAigAAAGdCAYAAAA44ojeAAAAOXRFWHRTb2Z0d2FyZQBNYXRwbG90bGliIHZlcnNpb24zLjcuMSwgaHR0cHM6Ly9tYXRwbG90bGliLm9yZy/bCgiHAAAACXBIWXMAAA9hAAAPYQGoP6dpAAAj4klEQVR4nO3de3BU9f3/8VeuSwKEADY3CZC2WECuEgkr9ttWQqJGC8pYmUYnVQYqBhXSHxr6444YQEUKRiLWgo4gSvtD5SKQBg1jCQECVm6DWFEYcZNWDAEiYUnO7w/Kfg1h1yxssp9Nno+ZnXE/57Of8z7vLOtrzu7ZDbIsyxIAAIBBgv1dAAAAwOUIKAAAwDgEFAAAYBwCCgAAMA4BBQAAGIeAAgAAjENAAQAAxiGgAAAA44T6u4CrUVdXpxMnTqh9+/YKCgrydzkAAKARLMvS6dOnlZCQoOBgz+dIAjKgnDhxQomJif4uAwAAXIXjx4+rS5cuHucEZEBp3769pIsHGBUV5dO1nU6ntmzZorS0NIWFhfl07UBHbzyjP+7RG8/oj2f0x71A601VVZUSExNd/x/3JCADyqW3daKiopokoERGRioqKiog/tjNid54Rn/cozee0R/P6I97gdqbxnw8gw/JAgAA4xBQAACAcQgoAADAOAQUAABgHAIKAAAwDgEFAAAYh4ACAACMQ0ABAADGIaAAAADjEFAAAIBxCCgAAMA4BBQAAGAcAgoAADAOAQUAABgn1N8FoPl0z91wxfEv5mU0cyUAAHjGGRQAAGAcAgoAADAOAQUAABiHgAIAAIxDQAEAAMYhoAAAAOMQUAAAgHEIKAAAwDgEFAAAYBwCCgAAMA5fdQ++Ah8AYBzOoAAAAOMQUAAAgHEIKAAAwDgEFAAAYBwCCgAAMA4BBQAAGIfLjFsgd5cNAwAQKDiDAgAAjENAAQAAxiGgAAAA4xBQAACAcQgoAADAOAQUAABgHAIKAAAwDgEFAAAYh4ACAACMQ0ABAADGIaAAAADjEFAAAIBxCCgAAMA4BBQAAGAcAgoAADAOAQUAABiHgAIAAIxDQAEAAMYhoAAAAOMQUAAAgHEIKAAAwDgEFAAAYBwCCgAAMA4BBQAAGIeAAgAAjENAAQAAxiGgAAAA44T6uwCYq3vuhnr3bSGWFgz2UzEAgFaFMygAAMA4BBQAAGAcAgoAADAOAQUAABiHgAIAAIzjVUCpra3VtGnTlJSUpIiICP3kJz/RnDlzZFmWa45lWZo+fbri4+MVERGh1NRUHTlypN46J0+eVGZmpqKiohQdHa0xY8bozJkzvjkiAAAQ8LwKKPPnz9fSpUv14osv6tChQ5o/f74WLFigJUuWuOYsWLBAixcvVkFBgUpLS9W2bVulp6fr3LlzrjmZmZk6cOCACgsLtX79em3btk3jxo3z3VEBAICA5tX3oGzfvl0jRoxQRkaGJKl79+568803tXPnTkkXz54sWrRIU6dO1YgRIyRJr7/+umJjY/XOO+9o9OjROnTokDZt2qRdu3YpOTlZkrRkyRLdeeedeu6555SQkODL4wMAAAHIq4Byyy23aNmyZfr00091ww036J///Kc++ugjLVy4UJJ09OhRORwOpaamuh7ToUMHpaSkqKSkRKNHj1ZJSYmio6Nd4USSUlNTFRwcrNLSUt1zzz0N9ltTU6OamhrX/aqqKkmS0+mU0+n07oh/wKX1fL1uc7KFWD886WrWDb64biD3pim1hOdOU6E3ntEfz+iPe4HWG2/q9Cqg5ObmqqqqSj179lRISIhqa2s1d+5cZWZmSpIcDockKTY2tt7jYmNjXdscDodiYmLqFxEaqk6dOrnmXC4vL0+zZs1qML5lyxZFRkZ6cwiNVlhY2CTrNoem/rbXQO5Nc6A/7tEbz+iPZ/THvUDpTXV1daPnehVQ3n77ba1cuVKrVq3SjTfeqI8//lgTJ05UQkKCsrKyvC60saZMmaKcnBzX/aqqKiUmJiotLU1RUVE+3ZfT6VRhYaGGDx+usLAwn67dXPrM3Nwk69qCLc1Jrgvo3jSllvDcaSr0xjP64xn9cS/QenPpHZDG8CqgTJ48Wbm5uRo9erQkqW/fvvryyy+Vl5enrKwsxcXFSZLKy8sVHx/velx5ebkGDBggSYqLi1NFRUW9dS9cuKCTJ0+6Hn85m80mm83WYDwsLKzJ/iBNuXZTq6kNatL1A7k3zYH+uEdvPKM/ntEf9wKlN97U6NVVPNXV1QoOrv+QkJAQ1dXVSZKSkpIUFxenoqIi1/aqqiqVlpbKbrdLkux2uyorK1VWVuaas3XrVtXV1SklJcWbcgAAQAvl1RmUu+++W3PnzlXXrl114403au/evVq4cKEefvhhSVJQUJAmTpyop59+Wj169FBSUpKmTZumhIQEjRw5UpLUq1cv3X777Ro7dqwKCgrkdDo1YcIEjR49mit4AACAJC8DypIlSzRt2jQ9+uijqqioUEJCgn7/+99r+vTprjlPPvmkzp49q3HjxqmyslK33nqrNm3apDZt2rjmrFy5UhMmTNCwYcMUHBysUaNGafHixb47KgAAENC8Cijt27fXokWLtGjRIrdzgoKCNHv2bM2ePdvtnE6dOmnVqlXe7BoAALQi/BYPAAAwDgEFAAAYh4ACAACMQ0ABAADG8epDsjBL99wN/i4BAIAmwRkUAABgHAIKAAAwDgEFAAAYh4ACAACMQ0ABAADGIaAAAADjEFAAAIBxCCgAAMA4BBQAAGAcAgoAADAOAQUAABiHgAIAAIxDQAEAAMYhoAAAAOMQUAAAgHEIKAAAwDgEFAAAYBwCCgAAMA4BBQAAGIeAAgAAjENAAQAAxiGgAAAA4xBQAACAcQgoAADAOAQUAABgHAIKAAAwDgEFAAAYh4ACAACMQ0ABAADGIaAAAADjhPq7ALQc3XM3XHH8i3kZzVwJACDQcQYFAAAYh4ACAACMQ0ABAADGIaAAAADjEFAAAIBxCCgAAMA4BBQAAGAcAgoAADAOAQUAABiHgAIAAIxDQAEAAMYhoAAAAOMQUAAAgHEIKAAAwDgEFAAAYBwCCgAAMA4BBQAAGIeAAgAAjENAAQAAxiGgAAAA4xBQAACAcQgoAADAOAQUAABgHAIKAAAwDgEFAAAYh4ACAACMQ0ABAADG8TqgfPXVV3rggQfUuXNnRUREqG/fvtq9e7dru2VZmj59uuLj4xUREaHU1FQdOXKk3honT55UZmamoqKiFB0drTFjxujMmTPXfjQAAKBF8CqgfPvttxo6dKjCwsL0/vvv6+DBg3r++efVsWNH15wFCxZo8eLFKigoUGlpqdq2bav09HSdO3fONSczM1MHDhxQYWGh1q9fr23btmncuHG+OyoAABDQQr2ZPH/+fCUmJmr58uWusaSkJNd/W5alRYsWaerUqRoxYoQk6fXXX1dsbKzeeecdjR49WocOHdKmTZu0a9cuJScnS5KWLFmiO++8U88995wSEhJ8cVwAACCAeRVQ3nvvPaWnp+u+++5TcXGxrr/+ej366KMaO3asJOno0aNyOBxKTU11PaZDhw5KSUlRSUmJRo8erZKSEkVHR7vCiSSlpqYqODhYpaWluueeexrst6amRjU1Na77VVVVkiSn0ymn0+ndEf+AS+v5et2mYAuxmnd/wRf356437uoJhF76QiA9d5obvfGM/nhGf9wLtN54U6dXAeXzzz/X0qVLlZOToz/+8Y/atWuXHn/8cYWHhysrK0sOh0OSFBsbW+9xsbGxrm0Oh0MxMTH1iwgNVadOnVxzLpeXl6dZs2Y1GN+yZYsiIyO9OYRGKywsbJJ1fWnBYP/s111v3NWzcePGJqzGPIHw3PEXeuMZ/fGM/rgXKL2prq5u9FyvAkpdXZ2Sk5P1zDPPSJIGDhyo/fv3q6CgQFlZWd5V6YUpU6YoJyfHdb+qqkqJiYlKS0tTVFSUT/fldDpVWFio4cOHKywszKdr+1qfmZubdX+2YEtzkuvc9sbbevbPTPdVaUYIpOdOc6M3ntEfz+iPe4HWm0vvgDSGVwElPj5evXv3rjfWq1cv/e1vf5MkxcXFSZLKy8sVHx/vmlNeXq4BAwa45lRUVNRb48KFCzp58qTr8Zez2Wyy2WwNxsPCwprsD9KUa/tKTW2QX/brrjfe1mN6f69WIDx3/IXeeEZ/PKM/7gVKb7yp0aureIYOHarDhw/XG/v000/VrVs3SRc/MBsXF6eioiLX9qqqKpWWlsput0uS7Ha7KisrVVZW5pqzdetW1dXVKSUlxZtyAABAC+XVGZRJkybplltu0TPPPKPf/OY32rlzp5YtW6Zly5ZJkoKCgjRx4kQ9/fTT6tGjh5KSkjRt2jQlJCRo5MiRki6ecbn99ts1duxYFRQUyOl0asKECRo9enSrv4Kne+6GK45/MS+jmSsBAMC/vAooN998s9auXaspU6Zo9uzZSkpK0qJFi5SZmema8+STT+rs2bMaN26cKisrdeutt2rTpk1q06aNa87KlSs1YcIEDRs2TMHBwRo1apQWL17su6MCAAABzauAIkl33XWX7rrrLrfbg4KCNHv2bM2ePdvtnE6dOmnVqlXe7hoAALQS/BYPAAAwDgEFAAAYh4ACAACMQ0ABAADGIaAAAADjEFAAAIBxCCgAAMA4BBQAAGAcAgoAADAOAQUAABiHgAIAAIxDQAEAAMYhoAAAAOMQUAAAgHEIKAAAwDgEFAAAYBwCCgAAMA4BBQAAGIeAAgAAjENAAQAAxiGgAAAA4xBQAACAcQgoAADAOAQUAABgHAIKAAAwDgEFAAAYh4ACAACMQ0ABAADGIaAAAADjEFAAAIBxCCgAAMA4BBQAAGAcAgoAADAOAQUAABgn1N8FoPXqnrvhiuNfzMto5koAAKbhDAoAADAOAQUAABiHgAIAAIzDZ1D8wN1nLwAAwEWcQQEAAMYhoAAAAOMQUAAAgHEIKAAAwDgEFAAAYBwCCgAAMA4BBQAAGIeAAgAAjENAAQAAxiGgAAAA4xBQAACAcQgoAADAOAQUAABgHAIKAAAwDgEFAAAYh4ACAACMQ0ABAADGIaAAAADjEFAAAIBxCCgAAMA4BBQAAGAcAgoAADAOAQUAABiHgAIAAIxzTQFl3rx5CgoK0sSJE11j586dU3Z2tjp37qx27dpp1KhRKi8vr/e4Y8eOKSMjQ5GRkYqJidHkyZN14cKFaykFAAC0IFcdUHbt2qWXX35Z/fr1qzc+adIkrVu3TmvWrFFxcbFOnDihe++917W9trZWGRkZOn/+vLZv367XXntNK1as0PTp06/+KAAAQItyVQHlzJkzyszM1CuvvKKOHTu6xk+dOqVXX31VCxcu1G233aZBgwZp+fLl2r59u3bs2CFJ2rJliw4ePKg33nhDAwYM0B133KE5c+YoPz9f58+f981RAQCAgBZ6NQ/Kzs5WRkaGUlNT9fTTT7vGy8rK5HQ6lZqa6hrr2bOnunbtqpKSEg0ZMkQlJSXq27evYmNjXXPS09M1fvx4HThwQAMHDmywv5qaGtXU1LjuV1VVSZKcTqecTufVHIJbl9bz9brfZwuxvJrvrhZv17lWtuCL+2vqepqy902pOZ47gYreeEZ/PKM/7gVab7yp0+uAsnr1au3Zs0e7du1qsM3hcCg8PFzR0dH1xmNjY+VwOFxzvh9OLm2/tO1K8vLyNGvWrAbjW7ZsUWRkpLeH0CiFhYVNsq4kLRjs3fyNGzf6ZB1fcdcbX9Xj7ngDRVM+dwIdvfGM/nhGf9wLlN5UV1c3eq5XAeX48eN64oknVFhYqDZt2nhd2NWaMmWKcnJyXPerqqqUmJiotLQ0RUVF+XRfTqdThYWFGj58uMLCwny69iV9Zm72av7+mek+Weda2YItzUmuc9sbX9Xj7nhN1xzPnUBFbzyjP57RH/cCrTeX3gFpDK8CSllZmSoqKnTTTTe5xmpra7Vt2za9+OKL2rx5s86fP6/Kysp6Z1HKy8sVFxcnSYqLi9POnTvrrXvpKp9Lcy5ns9lks9kajIeFhTXZH6Qp166pDfK6Fl+s4yvueuOregLhH5knTfncCXT0xjP64xn9cS9QeuNNjV4FlGHDhmnfvn31xh566CH17NlTTz31lBITExUWFqaioiKNGjVKknT48GEdO3ZMdrtdkmS32zV37lxVVFQoJiZG0sVTU1FRUerdu7c35aCF6p674YrjX8zLaOZKAAD+4lVAad++vfr06VNvrG3bturcubNrfMyYMcrJyVGnTp0UFRWlxx57THa7XUOGDJEkpaWlqXfv3nrwwQe1YMECORwOTZ06VdnZ2Vc8SwIAAFqfq7qKx5MXXnhBwcHBGjVqlGpqapSenq6XXnrJtT0kJETr16/X+PHjZbfb1bZtW2VlZWn27Nm+LgUAAASoaw4oH374Yb37bdq0UX5+vvLz890+plu3bgF/pQYAAGg6/BYPAAAwDgEFAAAYh4ACAACMQ0ABAADGIaAAAADjEFAAAIBxCCgAAMA4BBQAAGAcAgoAADAOAQUAABiHgAIAAIxDQAEAAMYhoAAAAOMQUAAAgHEIKAAAwDgEFAAAYBwCCgAAMA4BBQAAGIeAAgAAjBPq7wKAxuqeu+GK41/My2jmSgAATY0zKAAAwDgEFAAAYBwCCgAAMA4BBQAAGIeAAgAAjENAAQAAxiGgAAAA4xBQAACAcQgoAADAOAQUAABgHAIKAAAwDgEFAAAYh4ACAACMQ0ABAADGIaAAAADjEFAAAIBxCCgAAMA4BBQAAGAcAgoAADAOAQUAABiHgAIAAIxDQAEAAMYJ9XcBLVn33A3+LgEAgIDEGRQAAGAczqCgxXJ3BuuLeRnNXAkAwFucQQEAAMYhoAAAAOMQUAAAgHEIKAAAwDgEFAAAYByu4kGrw9U9AGA+zqAAAADjEFAAAIBxCCgAAMA4BBQAAGAcAgoAADAOAQUAABiHgAIAAIxDQAEAAMYhoAAAAOMQUAAAgHEIKAAAwDgEFAAAYByvAkpeXp5uvvlmtW/fXjExMRo5cqQOHz5cb865c+eUnZ2tzp07q127dho1apTKy8vrzTl27JgyMjIUGRmpmJgYTZ48WRcuXLj2owEAAC2CVwGluLhY2dnZ2rFjhwoLC+V0OpWWlqazZ8+65kyaNEnr1q3TmjVrVFxcrBMnTujee+91ba+trVVGRobOnz+v7du367XXXtOKFSs0ffp03x0VAAAIaKHeTN60aVO9+ytWrFBMTIzKysr0P//zPzp16pReffVVrVq1Srfddpskafny5erVq5d27NihIUOGaMuWLTp48KD+/ve/KzY2VgMGDNCcOXP01FNPaebMmQoPD/fd0QEAgIDkVUC53KlTpyRJnTp1kiSVlZXJ6XQqNTXVNadnz57q2rWrSkpKNGTIEJWUlKhv376KjY11zUlPT9f48eN14MABDRw4sMF+ampqVFNT47pfVVUlSXI6nXI6nddyCA1cWs8X69pCrGteQ3Jfi6/Wbyxb8MX9mVLPJb6q51r/5r587rQ09MYz+uMZ/XEv0HrjTZ1BlmVd1f9V6urq9Otf/1qVlZX66KOPJEmrVq3SQw89VC9MSNLgwYP1q1/9SvPnz9e4ceP05ZdfavPmza7t1dXVatu2rTZu3Kg77rijwb5mzpypWbNmNRhftWqVIiMjr6Z8AADQzKqrq/Xb3/5Wp06dUlRUlMe5V30GJTs7W/v373eFk6Y0ZcoU5eTkuO5XVVUpMTFRaWlpP3iA3nI6nSosLNTw4cMVFhZ2TWv1mbn5hyc1wv6Z6U26fmPZgi3NSa5z25vmrucSX/XH3TqN5cvnTktDbzyjP57RH/cCrTeX3gFpjKsKKBMmTND69eu1bds2denSxTUeFxen8+fPq7KyUtHR0a7x8vJyxcXFuebs3Lmz3nqXrvK5NOdyNptNNputwXhYWFiT/UF8sXZNbZDPamnK9b3lrjf+rOdKvK3HV8+lpnxeBjp64xn98Yz+uBcovfGmRq+u4rEsSxMmTNDatWu1detWJSUl1ds+aNAghYWFqaioyDV2+PBhHTt2THa7XZJkt9u1b98+VVRUuOYUFhYqKipKvXv39qYcAADQQnl1BiU7O1urVq3Su+++q/bt28vhcEiSOnTooIiICHXo0EFjxoxRTk6OOnXqpKioKD322GOy2+0aMmSIJCktLU29e/fWgw8+qAULFsjhcGjq1KnKzs6+4lkSAADQ+ngVUJYuXSpJ+uUvf1lvfPny5frd734nSXrhhRcUHBysUaNGqaamRunp6XrppZdcc0NCQrR+/XqNHz9edrtdbdu2VVZWlmbPnn1tRwIAAFoMrwJKYy74adOmjfLz85Wfn+92Trdu3bRx40Zvdg0AAFoRfosHAAAYh4ACAACMQ0ABAADGIaAAAADjEFAAAIBxCCgAAMA4BBQAAGCcq/6xQKCl6Z674YrjX8zLaOZKAACcQQEAAMYhoAAAAOMQUAAAgHEIKAAAwDgEFAAAYBwCCgAAMA6XGQNX6fLLkm0hlhYM9lMxANDCcAYFAAAYh4ACAACMw1s8PuDuG0gBAMDV4QwKAAAwDgEFAAAYh7d4gB/AW3gA0Pw4gwIAAIxDQAEAAMYhoAAAAOMQUAAAgHEIKAAAwDgEFAAAYBwCCgAAMA4BBQAAGIeAAgAAjENAAQAAxuGr7t3oM3OzamqD6o19MS/DT9UAANC6EFCAZuLuN30IvgDQEG/xAAAA4xBQAACAcQgoAADAOAQUAABgHAIKAAAwDgEFAAAYh8uMAT/j8mMAaIgzKAAAwDgEFAAAYBwCCgAAMA4BBQAAGIeAAgAAjMNVPIChuLoHQGtGQAECDMEFQGvAWzwAAMA4BBQAAGAcAgoAADAOn0Hxgrv3/gET8NkUAC0JZ1AAAIBxCCgAAMA4BBQAAGAcPoMCtHB8NgVAIOIMCgAAMA4BBQAAGIeAAgAAjMNnUADU4+33/fBZFgBNgYAC4JrwIVwATYGAAqBJfD+42EIsLRgs9Zm5WYfn3uXHqgAECgIKgGbFGRcAjUFAAWA0Ag3QOvk1oOTn5+vZZ5+Vw+FQ//79tWTJEg0ePNifJQFooQg6QGDx22XGb731lnJycjRjxgzt2bNH/fv3V3p6uioqKvxVEgAAMITfzqAsXLhQY8eO1UMPPSRJKigo0IYNG/SXv/xFubm5/ioLQIDz9jJpX63vqzMxnOkBLvJLQDl//rzKyso0ZcoU11hwcLBSU1NVUlLSYH5NTY1qampc90+dOiVJOnnypJxOp09rczqdqq6uVqgzWLV1QT5d+2p98803VxwPvXC2WesIrbNUXV2nb775RmFhYX6v5xL6c3Was2+XeuPp35W39fz0/7x95fle1uZuv+54W09j2IItTR1YpwH/9/8p1Mv+eCslr+iK46VThvllnca49Lrs7t9WaxZovTl9+rQkybKsH55s+cFXX31lSbK2b99eb3zy5MnW4MGDG8yfMWOGJYkbN27cuHHj1gJux48f/8GsEBBX8UyZMkU5OTmu+3V1dTp58qQ6d+6soCDfnuWoqqpSYmKijh8/rqioKJ+uHejojWf0xz164xn98Yz+uBdovbEsS6dPn1ZCQsIPzvVLQLnuuusUEhKi8vLyeuPl5eWKi4trMN9ms8lms9Ubi46ObsoSFRUVFRB/bH+gN57RH/fojWf0xzP6414g9aZDhw6NmueXq3jCw8M1aNAgFRX973uYdXV1Kioqkt1u90dJAADAIH57iycnJ0dZWVlKTk7W4MGDtWjRIp09e9Z1VQ8AAGi9/BZQ7r//fv373//W9OnT5XA4NGDAAG3atEmxsbH+KknSxbeTZsyY0eAtJdCbH0J/3KM3ntEfz+iPey25N0GW1ZhrfQAAAJqP375JFgAAwB0CCgAAMA4BBQAAGIeAAgAAjENA+Z78/Hx1795dbdq0UUpKinbu3Onvkvxi27Ztuvvuu5WQkKCgoCC988479bZblqXp06crPj5eERERSk1N1ZEjR/xTbDPLy8vTzTffrPbt2ysmJkYjR47U4cOH6805d+6csrOz1blzZ7Vr106jRo1q8KWELdXSpUvVr18/15dG2e12vf/++67trbk3l5s3b56CgoI0ceJE11hr7s/MmTMVFBRU79azZ0/X9tbcG0n66quv9MADD6hz586KiIhQ3759tXv3btf2lvi6TED5r7feeks5OTmaMWOG9uzZo/79+ys9PV0VFRX+Lq3ZnT17Vv3791d+fv4Vty9YsECLFy9WQUGBSktL1bZtW6Wnp+vcuXPNXGnzKy4uVnZ2tnbs2KHCwkI5nU6lpaXp7Nn//QG5SZMmad26dVqzZo2Ki4t14sQJ3XvvvX6suvl06dJF8+bNU1lZmXbv3q3bbrtNI0aM0IEDByS17t58365du/Tyyy+rX79+9cZbe39uvPFGff31167bRx995NrWmnvz7bffaujQoQoLC9P777+vgwcP6vnnn1fHjh1dc1rk67IvfvyvJRg8eLCVnZ3tul9bW2slJCRYeXl5fqzK/yRZa9eudd2vq6uz4uLirGeffdY1VllZadlsNuvNN9/0Q4X+VVFRYUmyiouLLcu62IuwsDBrzZo1rjmHDh2yJFklJSX+KtOvOnbsaP35z3+mN/91+vRpq0ePHlZhYaH1i1/8wnriiScsy+K5M2PGDKt///5X3Nbae/PUU09Zt956q9vtLfV1mTMoks6fP6+ysjKlpqa6xoKDg5WamqqSkhI/Vmaeo0ePyuFw1OtVhw4dlJKS0ip7derUKUlSp06dJEllZWVyOp31+tOzZ0917dq11fWntrZWq1ev1tmzZ2W32+nNf2VnZysjI6NeHySeO5J05MgRJSQk6Mc//rEyMzN17NgxSfTmvffeU3Jysu677z7FxMRo4MCBeuWVV1zbW+rrMgFF0n/+8x/V1tY2+Bbb2NhYORwOP1Vlpkv9oFcXfz9q4sSJGjp0qPr06SPpYn/Cw8Mb/Jhla+rPvn371K5dO9lsNj3yyCNau3atevfuTW8krV69Wnv27FFeXl6Dba29PykpKVqxYoU2bdqkpUuX6ujRo/r5z3+u06dPt/refP7551q6dKl69OihzZs3a/z48Xr88cf12muvSWq5r8t++6p7INBlZ2dr//799d4nh/Szn/1MH3/8sU6dOqW//vWvysrKUnFxsb/L8rvjx4/riSeeUGFhodq0aePvcoxzxx13uP67X79+SklJUbdu3fT2228rIiLCj5X5X11dnZKTk/XMM89IkgYOHKj9+/eroKBAWVlZfq6u6XAGRdJ1112nkJCQBp8ILy8vV1xcnJ+qMtOlfrT2Xk2YMEHr16/XBx98oC5durjG4+LidP78eVVWVtab35r6Ex4erp/+9KcaNGiQ8vLy1L9/f/3pT39q9b0pKytTRUWFbrrpJoWGhio0NFTFxcVavHixQkNDFRsb26r7c7no6GjdcMMN+uyzz1r9cyc+Pl69e/euN9arVy/XW2At9XWZgKKLL6iDBg1SUVGRa6yurk5FRUWy2+1+rMw8SUlJiouLq9erqqoqlZaWtopeWZalCRMmaO3atdq6dauSkpLqbR80aJDCwsLq9efw4cM6duxYq+jPldTV1ammpqbV92bYsGHat2+fPv74Y9ctOTlZmZmZrv9uzf253JkzZ/Svf/1L8fHxrf65M3To0AZfZ/Dpp5+qW7duklrw67K/P6VritWrV1s2m81asWKFdfDgQWvcuHFWdHS05XA4/F1aszt9+rS1d+9ea+/evZYka+HChdbevXutL7/80rIsy5o3b54VHR1tvfvuu9Ynn3xijRgxwkpKSrK+++47P1fe9MaPH2916NDB+vDDD62vv/7adauurnbNeeSRR6yuXbtaW7dutXbv3m3Z7XbLbrf7sermk5ubaxUXF1tHjx61PvnkEys3N9cKCgqytmzZYllW6+7NlXz/Kh7Lat39+cMf/mB9+OGH1tGjR61//OMfVmpqqnXddddZFRUVlmW17t7s3LnTCg0NtebOnWsdOXLEWrlypRUZGWm98cYbrjkt8XWZgPI9S5Yssbp27WqFh4dbgwcPtnbs2OHvkvzigw8+sCQ1uGVlZVmWdfGStmnTplmxsbGWzWazhg0bZh0+fNi/RTeTK/VFkrV8+XLXnO+++8569NFHrY4dO1qRkZHWPffcY3399df+K7oZPfzww1a3bt2s8PBw60c/+pE1bNgwVzixrNbdmyu5PKC05v7cf//9Vnx8vBUeHm5df/311v3332999tlnru2tuTeWZVnr1q2z+vTpY9lsNqtnz57WsmXL6m1via/LQZZlWf45dwMAAHBlfAYFAAAYh4ACAACMQ0ABAADGIaAAAADjEFAAAIBxCCgAAMA4BBQAAGAcAgoAADAOAQUAABiHgAIAAIxDQAEAAMYhoAAAAOP8f9NUoHzpaR4gAAAAAElFTkSuQmCC",
      "text/plain": [
       "<Figure size 640x480 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "df.text.apply(lambda x: len(tokenizer.encode(x))).hist(bins=75);"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "id": "842c9eaa",
   "metadata": {},
   "outputs": [],
   "source": [
    "tokens_per_second = df.text.apply(lambda x: len(tokenizer.encode(x))) / df.duration"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "id": "d94bb68d",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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",
      "text/plain": [
       "<Figure size 1500x800 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "plt.figure(figsize=(15, 8))\n",
    "tokens_per_second.hist(bins=50);\n",
    "plt.axvline(x=12.5, ymin=0, ymax=1.0, color='red', label='Limit for Stride = 8')\n",
    "plt.axvline(x=25, ymin=0, ymax=1.0, color='green', label='Limit for Stride = 4')\n",
    "plt.legend(fontsize=15)\n",
    "plt.title('Distribution of tokens per second over dataset', fontsize=15);"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "c91e4b81",
   "metadata": {},
   "source": [
    "### model"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "id": "5c9fddac",
   "metadata": {},
   "outputs": [],
   "source": [
    "conf = omegaconf.OmegaConf.load(\"./conf/conformer_las.yaml\")\n",
    "conf.train_dataloader.batch_size = 4\n",
    "conf.train_dataloader.num_workers = 4\n",
    "\n",
    "conf.model.decoder.tokenizer_path = os.path.join(TOKENIZER_PATH, \"bpe_1024_bos_eos.model\")\n",
    "\n",
    "\n",
    "model = LASModel(conf=conf)\n",
    "ckpt = torch.load(CKPT_PATH, map_location=\"cpu\")\n",
    "model.load_state_dict(ckpt)\n",
    "model.eval()\n",
    "model.freeze()"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "d56108d6",
   "metadata": {},
   "source": [
    "### features"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "id": "3c2f220f",
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "data": {
      "image/png": "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",
      "text/plain": [
       "<Figure size 640x480 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "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",
      "text/plain": [
       "<Figure size 640x480 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "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",
      "text/plain": [
       "<Figure size 640x480 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "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",
      "text/plain": [
       "<Figure size 640x480 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "\n",
    "\n",
    "batch = next(iter(model.val_dataloader()))\n",
    "\n",
    "features, features_len, targets, target_len = batch\n",
    "\n",
    "for feature in features[0:4]:\n",
    "    plt.imshow(feature)\n",
    "    plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "3ec199d3",
   "metadata": {},
   "source": [
    "### loss"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "id": "83937bb7",
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/usr/local/lib/python3.10/dist-packages/torch/nn/functional.py:5038: UserWarning: Support for mismatched key_padding_mask and attn_mask is deprecated. Use same type for both instead.\n",
      "  warnings.warn(\n",
      "/usr/local/lib/python3.10/dist-packages/torch/nn/modules/activation.py:1186: UserWarning: Converting mask without torch.bool dtype to bool; this will negatively affect performance. Prefer to use a boolean mask directly. (Triggered internally at /opt/pytorch/pytorch/aten/src/ATen/native/transformers/attention.cpp:150.)\n",
      "  return torch._native_multi_head_attention(\n"
     ]
    },
    {
     "data": {
      "image/png": "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",
      "text/plain": [
       "<Figure size 640x480 with 2 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "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",
      "text/plain": [
       "<Figure size 640x480 with 2 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "encoded, encoded_len = model(features, features_len)\n",
    "\n",
    "\n",
    "encoded_pad_mask = model.make_pad_mask(encoded_len)\n",
    "\n",
    "targets_outputs = targets[:, 1:] # without bos\n",
    "targets_inputs = targets[:, :-1] # without eos / last pad token\n",
    "target_len -= 1\n",
    "\n",
    "target_pad_mask = model.make_pad_mask(target_len)\n",
    "target_mask = model.make_attention_mask(target_len)\n",
    "\n",
    "logits = model.decoder(encoded, ~encoded_pad_mask, targets_inputs, target_mask, ~target_pad_mask)\n",
    "\n",
    "loss = model.loss(logits.transpose(1, 2), targets_outputs)\n",
    "plt.imshow(loss)\n",
    "plt.colorbar(fraction=0.01)\n",
    "plt.show()\n",
    "plt.imshow(loss * target_pad_mask)\n",
    "plt.colorbar(fraction=0.01)\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 20,
   "id": "dd9a1dc7",
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "<matplotlib.image.AxesImage at 0x7f08cc416f20>"
      ]
     },
     "execution_count": 20,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "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",
      "text/plain": [
       "<Figure size 640x480 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "plt.imshow(target_pad_mask)\n"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "5aca71b7",
   "metadata": {},
   "source": [
    "## Greedy Decoding"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 38,
   "id": "8aab3c82",
   "metadata": {},
   "outputs": [],
   "source": [
    "device = 'cuda:2' if torch.cuda.is_available() else 'cpu'\n",
    "model = model.to(device)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 39,
   "id": "126819a6",
   "metadata": {},
   "outputs": [],
   "source": [
    "class GreedyDecoder:\n",
    "    def __init__(self, model, tokenizer, max_steps=20):\n",
    "        self.model = model\n",
    "        self.max_steps = max_steps\n",
    "\n",
    "    def __call__(self, encoded):\n",
    "        \n",
    "        tokens = [self.model.decoder.tokenizer.bos_id()]\n",
    "\n",
    "        for _ in range(self.max_steps):\n",
    "            \n",
    "            tokens_batch = torch.tensor(tokens).unsqueeze(0)\n",
    "            att_mask = self.model.make_attention_mask(torch.tensor([tokens_batch.size(-1)]))\n",
    "            \n",
    "            distribution = self.model.decoder(\n",
    "                encoded=encoded.to(model.device), \n",
    "                encoded_pad_mask=None,\n",
    "                target=tokens_batch.to(model.device), \n",
    "                target_mask=att_mask.to(model.device), \n",
    "                target_pad_mask=None\n",
    "            )\n",
    "        \n",
    "            best_next_token = distribution[0, -1].argmax()\n",
    "            \n",
    "            if best_next_token == self.model.decoder.tokenizer.eos_id():\n",
    "                break\n",
    "\n",
    "            tokens.append(best_next_token.item())\n",
    "        \n",
    "        return self.model.decoder.tokenizer.decode(tokens)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 42,
   "id": "8b4b33d6",
   "metadata": {},
   "outputs": [],
   "source": [
    "conf.val_dataloader.batch_size = 5\n",
    "\n",
    "batch = next(iter(model.val_dataloader()))\n",
    "\n",
    "features, features_len, targets, target_len = batch\n",
    "\n",
    "encoded, encoded_len = model(features.to(device), features_len.to(device))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 43,
   "id": "3b29d152",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "reference : афина воспроизведи музыку вперемешку\n",
      "hypothesis: афина воспроизведи музыку в перемешку\n",
      "----------------------------------------------------------------------------------------------------\n",
      "\n",
      "reference : найти сериал григорий р\n",
      "hypothesis: найти сериал григорьевый р\n",
      "----------------------------------------------------------------------------------------------------\n",
      "\n",
      "reference : прямой эфир апл манчестер юнайтед тоттенхэм\n",
      "hypothesis: прямой эфир апл манчестер юнайтед тоттенхэм\n",
      "----------------------------------------------------------------------------------------------------\n",
      "\n",
      "reference : ильвиром ивановичем ворончихиным\n",
      "hypothesis: ильвиром ивановичем баранчихиным\n",
      "----------------------------------------------------------------------------------------------------\n",
      "\n",
      "reference : можешь показать киношку исходный код\n",
      "hypothesis: можешь показать киношку исходный код\n",
      "----------------------------------------------------------------------------------------------------\n",
      "\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/usr/local/lib/python3.10/dist-packages/torch/nn/modules/activation.py:1186: UserWarning: Converting mask without torch.bool dtype to bool; this will negatively affect performance. Prefer to use a boolean mask directly. (Triggered internally at /opt/pytorch/pytorch/aten/src/ATen/native/transformers/attention.cpp:150.)\n",
      "  return torch._native_multi_head_attention(\n"
     ]
    }
   ],
   "source": [
    "decoder = GreedyDecoder(model, tokenizer)\n",
    "\n",
    "\n",
    "for i in range(features.shape[0]):\n",
    "\n",
    "    encoder_states = encoded[\n",
    "        [i],\n",
    "        :encoded_len[i],\n",
    "        :\n",
    "    ]\n",
    "    \n",
    "    ref_tokens = targets[i, :target_len[i]].tolist()\n",
    "    \n",
    "    print(f\"reference : {tokenizer.decode(ref_tokens)}\")\n",
    "    print(f\"hypothesis: {decoder(encoder_states)}\")\n",
    "    print(\"-\" * 100 + '\\n')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 44,
   "id": "caa82b80",
   "metadata": {},
   "outputs": [],
   "source": [
    "from src.metrics import WER\n",
    "\n",
    "def compute_wer(refs, hyps):\n",
    "    metric = WER()\n",
    "    metric.update(refs, hyps)\n",
    "    wer = metric.compute()\n",
    "    return wer[0]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 56,
   "id": "65e4341c",
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "  0%|                                                                                                                                                                                                   | 0/200 [00:00<?, ?it/s]/usr/local/lib/python3.10/dist-packages/torch/nn/modules/activation.py:1186: UserWarning: Converting mask without torch.bool dtype to bool; this will negatively affect performance. Prefer to use a boolean mask directly. (Triggered internally at /opt/pytorch/pytorch/aten/src/ATen/native/transformers/attention.cpp:150.)\n",
      "  return torch._native_multi_head_attention(\n",
      "100%|█████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████| 200/200 [00:04<00:00, 46.94it/s]\n"
     ]
    }
   ],
   "source": [
    "import gc\n",
    "from tqdm import tqdm\n",
    "\n",
    "conf.val_dataloader.batch_size = 200\n",
    "\n",
    "batch = next(iter(model.val_dataloader()))\n",
    "features, features_len, targets, target_len = batch\n",
    "encoded, encoded_len = model(features.to(device), features_len.to(device))\n",
    "hyps = []\n",
    "refs = []\n",
    "\n",
    "for i in tqdm(range(features.shape[0])):\n",
    "\n",
    "    encoder_states = encoded[\n",
    "        [i],\n",
    "        :encoded_len[i],\n",
    "        :\n",
    "    ]\n",
    "    \n",
    "    ref_tokens = targets[i, :target_len[i]].tolist()\n",
    "    refs.append(tokenizer.decode(ref_tokens))\n",
    "    hyps.append(decoder(encoder_states))\n",
    "    \n",
    "gc.collect()\n",
    "torch.cuda.empty_cache()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 57,
   "id": "c3c44540",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "WER: 23.1%\n"
     ]
    }
   ],
   "source": [
    "print(f'WER: {\"{:2.1f}\".format(compute_wer(refs, hyps).item() * 100)}%')"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "59914b2e",
   "metadata": {},
   "source": [
    "### Beam-search Decoding"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 84,
   "id": "e3b3722d",
   "metadata": {},
   "outputs": [],
   "source": [
    "class BeamSearchDecoder:\n",
    "    \n",
    "    def __init__(self, model, temp=1.0, beam_size=5, max_steps=20):\n",
    "        self.model = model\n",
    "        self.temp = temp\n",
    "        self.beam_size = beam_size\n",
    "        self.max_steps = max_steps\n",
    "        \n",
    "    def __call__(self, encoded):\n",
    "        \n",
    "        partial_hyps = [(0.0, [self.model.decoder.tokenizer.bos_id()])]\n",
    "        final_hyps = []\n",
    "\n",
    "        while len(final_hyps) < self.beam_size:\n",
    "            \n",
    "            # Берем лучшую гипотезу среди текущих partial-гипотез\n",
    "            cur_partial_score, cur_partial_hyp = heapq.heappop(partial_hyps)\n",
    "            \n",
    "            tokens_batch = torch.tensor(cur_partial_hyp).unsqueeze(0)\n",
    "            att_mask = self.model.make_attention_mask(torch.tensor([tokens_batch.size(-1)]))\n",
    "            \n",
    "            # Прогоняем ее через декодер и получаем логиты == кандидаты на расширение\n",
    "            logits = self.model.decoder(\n",
    "                encoded=encoded.to(self.model.device), \n",
    "                encoded_pad_mask=None,\n",
    "                target=tokens_batch.to(self.model.device), \n",
    "                target_mask=att_mask.to(self.model.device), \n",
    "                target_pad_mask=None\n",
    "            )\n",
    "\n",
    "            logprobs = F.log_softmax(logits[0, -1] / self.temp, dim=-1)\n",
    "            \n",
    "            candidates = logprobs.topk(self.beam_size)\n",
    "            \n",
    "            # \n",
    "            for token_score, token_idx in zip(candidates.values, candidates.indices):\n",
    "                \n",
    "                token_idx = int(token_idx)\n",
    "\n",
    "                new_score = cur_partial_score - float(token_score)\n",
    "                new_hyp = cur_partial_hyp + [token_idx]\n",
    "                new_item = (new_score, new_hyp)\n",
    "\n",
    "                if token_idx == self.model.decoder.tokenizer.eos_id() or len(new_hyp) - 1 >= self.max_steps:\n",
    "                    final_hyps.append(new_item)\n",
    "                else:\n",
    "                    heapq.heappush(partial_hyps, new_item)\n",
    "            \n",
    "            if len(partial_hyps) > self.beam_size:\n",
    "                partial_hyps = heapq.nsmallest(self.beam_size, partial_hyps)\n",
    "                heapq.heapify(partial_hyps)\n",
    "\n",
    "        final_scores, final_token_lists = zip(*final_hyps)\n",
    "        \n",
    "        final_texts = [self.model.decoder.tokenizer.decode(final_token_list) for final_token_list in final_token_lists]\n",
    "\n",
    "\n",
    "        result = list(zip(final_scores, final_texts))\n",
    "        result.sort()\n",
    "\n",
    "        return result[:self.beam_size]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 59,
   "id": "e026f361",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "reference   : афина воспроизведи музыку вперемешку\n",
      "hypothesis 1: афина воспроизведи музыку в перемешку 0.65\n",
      "hypothesis 2: афина воспроизведи музыку в пельмешку 2.79\n",
      "hypothesis 3: афина воспроизведи музыку в перемешку 3.27\n",
      "hypothesis 4: афина воспроизведи музыку вперемешку 3.43\n",
      "hypothesis 5: афина воспроизведе музыку в перемешку 3.53\n",
      "hypothesis 6: афина воспроизведей музыку в перемешку 3.64\n",
      "hypothesis 7: афина воспроизведи музыку в перемешку 3.80\n",
      "hypothesis 8: афина воспроизведи музыку в перемешко 4.12\n",
      "hypothesis 9: афина воспроизведи музыку в перемежку 4.63\n",
      "hypothesis 10:  10.82\n",
      "----------------------------------------------------------------------------------------------------\n",
      "\n",
      "reference   : найти сериал григорий р\n",
      "hypothesis 1: найти сериал григорий эр 2.14\n",
      "hypothesis 2: найти сериал григорьи р 2.43\n",
      "hypothesis 3: найти сериал григой эр 2.46\n",
      "hypothesis 4: найти сериал григой р 2.69\n",
      "hypothesis 5: найти сериал григорь 4.89\n",
      "hypothesis 6: найти сериал григорьи 7.48\n",
      "hypothesis 7: найти сериал григой 7.54\n",
      "hypothesis 8: найти сериал григорий 8.17\n",
      "hypothesis 9: найти сериал григорий э 10.14\n",
      "hypothesis 10: найти сериал григори 10.58\n",
      "----------------------------------------------------------------------------------------------------\n",
      "\n",
      "reference   : прямой эфир апл манчестер юнайтед тоттенхэм\n",
      "hypothesis 1: прямой эфир апл манчестер юнайтед тоттенхэм 0.01\n",
      "hypothesis 2: прямой эфир апл манчестер юнайтед тоттенхэма 5.67\n",
      "hypothesis 3: прямой эфир апл юнайтед тоттенхэм 6.22\n",
      "hypothesis 4: прямой эфир апл манчестер юнайтед в тоттенхэм 7.85\n",
      "hypothesis 5: прямой эфир ап и манчестер юнайтед тоттенхэм 8.10\n",
      "hypothesis 6: прямой эфир апл манчестер и тоттенхэм 8.45\n",
      "hypothesis 7: прямой эфир апл манчестер и манчестер юнайтед тоттенхэм 8.67\n",
      "hypothesis 8: прямой эфир апл манчестер юнайтед тоттенхэ 10.36\n",
      "hypothesis 9: прямой эфир апл манчестер юнайтед тоттенх 12.58\n",
      "hypothesis 10: прямой эфир апл манчестер и манчестер юнайтед 17.62\n",
      "----------------------------------------------------------------------------------------------------\n",
      "\n",
      "reference   : ильвиром ивановичем ворончихиным\n",
      "hypothesis 1: ильвиром ивановичем баранчихиным 2.65\n",
      "hypothesis 2: ильвиром ивановичем ворончихиным 3.21\n",
      "hypothesis 3: ильвиром ивановичем варанчихиным 3.64\n",
      "hypothesis 4: ильвиром ивановичем борончихиным 3.75\n",
      "hypothesis 5: ильвиром ивановичем варанчихиным 3.84\n",
      "hypothesis 6: ильвиром ивановичем баранчихины 9.13\n",
      "hypothesis 7: ильвиром ивановичем ворончихины 9.89\n",
      "hypothesis 8: ильвиром ивановичем варанчихины 10.11\n",
      "hypothesis 9: ильвиром ивановичем варанчихины 10.52\n",
      "hypothesis 10: ильвиром ивановичем борончихины 10.95\n",
      "----------------------------------------------------------------------------------------------------\n",
      "\n",
      "reference   : можешь показать киношку исходный код\n",
      "hypothesis 1: можешь показать киношку исходный код 0.26\n",
      "hypothesis 2: можешь показать киношку исходный кот 2.18\n",
      "hypothesis 3: можешь показать киношку и сходный код 3.79\n",
      "hypothesis 4: можешь показать киношку исходный кот 4.32\n",
      "hypothesis 5: можешь показать киношку из ходный код 4.36\n",
      "hypothesis 6: можешь показать киношку исходит 6.03\n",
      "hypothesis 7: можешь показать киношку искот 6.42\n",
      "hypothesis 8: можешь показать киношку из ход 8.67\n",
      "hypothesis 9: можешь показать киношку исход 9.01\n",
      "hypothesis 10: можешь показать киношку и сход 9.34\n",
      "----------------------------------------------------------------------------------------------------\n",
      "\n"
     ]
    }
   ],
   "source": [
    "decoder = BeamSearchDecoder(model, temp=1, beam_size=10)\n",
    "\n",
    "\n",
    "for i in range(5):\n",
    "\n",
    "    encoder_states = encoded[\n",
    "        [i],\n",
    "        :encoded_len[i],\n",
    "        :\n",
    "    ]\n",
    "    \n",
    "    ref_tokens = targets[i, :target_len[i]].tolist()\n",
    "    \n",
    "    print(f\"reference   : {tokenizer.decode(ref_tokens)}\")\n",
    "    for k, (score, hyp) in enumerate(decoder(encoder_states)):\n",
    "        print(f\"hypothesis {k + 1}: {hyp} {score:.2f}\")\n",
    "    print(\"-\" * 100 + '\\n')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 123,
   "id": "88440dc7",
   "metadata": {},
   "outputs": [],
   "source": [
    "import editdistance\n",
    "import numpy as np"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 124,
   "id": "abf5789e",
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|█████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████| 200/200 [00:27<00:00,  7.34it/s]\n"
     ]
    }
   ],
   "source": [
    "decoder = BeamSearchDecoder(model, temp=1, beam_size=20)\n",
    "\n",
    "hyps = []\n",
    "refs = []\n",
    "oracle_hyps = []\n",
    "best_hyp_indices = []\n",
    "\n",
    "for i in tqdm(range(features.shape[0])):\n",
    "\n",
    "    encoder_states = encoded[\n",
    "        [i],\n",
    "        :encoded_len[i],\n",
    "        :\n",
    "    ]\n",
    "    \n",
    "    ref_tokens = targets[i, :target_len[i]].tolist()\n",
    "    refs.append(tokenizer.decode(ref_tokens))\n",
    "    \n",
    "    curr_hyps = decoder(encoder_states)\n",
    "    hyps.append(curr_hyps[0][1])\n",
    "    best_hyp_index = np.argmin([editdistance.eval(refs[-1].split(), hyp[1].split()) for hyp in curr_hyps])\n",
    "    best_hyp_indices.append(best_hyp_index)\n",
    "    \n",
    "    oracle_hyp = curr_hyps[best_hyp_index]\n",
    "    oracle_hyps.append(oracle_hyp[1])\n",
    "\n",
    "gc.collect()\n",
    "torch.cuda.empty_cache()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 125,
   "id": "64fe2a04",
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "WER: 19.5%\n"
     ]
    }
   ],
   "source": [
    "print(f'WER: {\"{:2.1f}\".format(compute_wer(refs, hyps).item() * 100)}%')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 126,
   "id": "e156237c",
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Oracle WER: 10.1%\n"
     ]
    }
   ],
   "source": [
    "print(f'Oracle WER: {\"{:2.1f}\".format(compute_wer(refs, oracle_hyps).item() * 100)}%')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "7c26ea38",
   "metadata": {},
   "outputs": [],
   "source": []
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 3 (ipykernel)",
   "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": "3.10.6"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 5
}