NLP Data Prep.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [],
   "source": [
    "from fastai.text import *"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [],
   "source": [
    "from fastai.tabular import *"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [],
   "source": [
    "path = Path('')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "metadata": {},
   "outputs": [],
   "source": [
    "data = pd.read_csv('good_small_dataset.csv', engine='python')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "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>Unnamed: 0</th>\n",
       "      <th>type</th>\n",
       "      <th>content</th>\n",
       "      <th>is_valid</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>0</td>\n",
       "      <td>rumor</td>\n",
       "      <td>Life is an illusion, at least on a quantum lev...</td>\n",
       "      <td>False</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>1</td>\n",
       "      <td>hate</td>\n",
       "      <td>Unfortunately, he hasn’t yet attacked her for ...</td>\n",
       "      <td>False</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>2</td>\n",
       "      <td>hate</td>\n",
       "      <td>The Los Angeles Police Department has been den...</td>\n",
       "      <td>False</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>3</td>\n",
       "      <td>hate</td>\n",
       "      <td>The White House has decided to quietly withdra...</td>\n",
       "      <td>False</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>4</td>\n",
       "      <td>hate</td>\n",
       "      <td>“The time has come to cut off the tongues of t...</td>\n",
       "      <td>False</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "   Unnamed: 0   type                                            content  \\\n",
       "0           0  rumor  Life is an illusion, at least on a quantum lev...   \n",
       "1           1   hate  Unfortunately, he hasn’t yet attacked her for ...   \n",
       "2           2   hate  The Los Angeles Police Department has been den...   \n",
       "3           3   hate  The White House has decided to quietly withdra...   \n",
       "4           4   hate  “The time has come to cut off the tongues of t...   \n",
       "\n",
       "   is_valid  \n",
       "0     False  \n",
       "1     False  \n",
       "2     False  \n",
       "3     False  \n",
       "4     False  "
      ]
     },
     "execution_count": 19,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "data.head()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 23,
   "metadata": {},
   "outputs": [],
   "source": [
    "df = data.dropna()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 27,
   "metadata": {},
   "outputs": [],
   "source": [
    "df.to_csv('good_small_dataset_drop_missing.csv')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 28,
   "metadata": {},
   "outputs": [],
   "source": [
    "data_lm = TextLMDataBunch.from_csv(path, 'good_small_dataset_drop_missing.csv', text_cols = 'content', label_cols = 'type')\n",
    "data_lm.save('data_lm_export.pkl')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 29,
   "metadata": {},
   "outputs": [],
   "source": [
    "data_clas = TextClasDataBunch.from_csv(path, 'good_small_dataset_drop_missing.csv', vocab=data_lm.train_ds.vocab, text_cols = 'content', label_cols = 'type',bs=16)\n",
    "data_clas.save('data_clas_export.pkl')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [],
   "source": [
    "from fastai.text import *"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [],
   "source": [
    "data_lm = load_data('NLP/', 'data_lm_export.pkl')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [],
   "source": [
    "data_clas = load_data('', 'data_clas_export.pkl')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {},
   "outputs": [],
   "source": [
    "learn = language_model_learner(data_lm, AWD_LSTM, drop_mult=0.5)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 34,
   "metadata": {},
   "outputs": [],
   "source": [
    "learn.save('initial')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "Total time: 3:42:04 <p><table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: left;\">\n",
       "      <th>epoch</th>\n",
       "      <th>train_loss</th>\n",
       "      <th>valid_loss</th>\n",
       "      <th>accuracy</th>\n",
       "      <th>time</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <td>0</td>\n",
       "      <td>4.036231</td>\n",
       "      <td>3.787572</td>\n",
       "      <td>0.348584</td>\n",
       "      <td>3:42:04</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>"
      ],
      "text/plain": [
       "<IPython.core.display.HTML object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "learn.fit_one_cycle(1, 1e-2)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {},
   "outputs": [],
   "source": [
    "learn.save('initial')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {},
   "outputs": [],
   "source": [
    "learn.unfreeze()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "Total time: 4:11:46 <p><table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: left;\">\n",
       "      <th>epoch</th>\n",
       "      <th>train_loss</th>\n",
       "      <th>valid_loss</th>\n",
       "      <th>accuracy</th>\n",
       "      <th>time</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <td>0</td>\n",
       "      <td>3.525766</td>\n",
       "      <td>3.462929</td>\n",
       "      <td>0.391380</td>\n",
       "      <td>4:11:46</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>"
      ],
      "text/plain": [
       "<IPython.core.display.HTML object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "learn.fit_one_cycle(1, 1e-3)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {},
   "outputs": [],
   "source": [
    "learn.save_encoder('ft_enc')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {},
   "outputs": [],
   "source": [
    "learn.save('ft_encoder_model')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "'The President today spoke on the effects of tax boards on tax cuts ,'"
      ]
     },
     "execution_count": 13,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "learn.predict(\"The President today spoke on\", n_words=10)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "'Kim Kardashian released a new photo depicting her doing personal issues with Muslim immigrants'"
      ]
     },
     "execution_count": 14,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "learn.predict(\"Kim Kardashian released a new photo depicting her doing\", n_words=6)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "'World War Three has begun between Radio » Titanic , and Nobel'"
      ]
     },
     "execution_count": 16,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "learn.predict(\"World War Three has begun between\", n_words=10)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [],
   "source": [
    "learn = text_classifier_learner(data_clas, AWD_LSTM, drop_mult=0.5);"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {},
   "outputs": [],
   "source": [
    "learn.load_encoder('ft_enc')\n",
    "learn.load('good_model_epoc_2');"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {},
   "outputs": [
    {
     "ename": "KeyboardInterrupt",
     "evalue": "",
     "output_type": "error",
     "traceback": [
      "\u001b[0;31m---------------------------------------------------------------------------\u001b[0m",
      "\u001b[0;31mKeyboardInterrupt\u001b[0m                         Traceback (most recent call last)",
      "\u001b[0;32m<ipython-input-8-bc39e9e85f86>\u001b[0m in \u001b[0;36m<module>\u001b[0;34m\u001b[0m\n\u001b[0;32m----> 1\u001b[0;31m \u001b[0mlearn\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0msummary\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m",
      "\u001b[0;32m/opt/conda/envs/fastai/lib/python3.6/site-packages/fastai/callbacks/hooks.py\u001b[0m in \u001b[0;36mmodel_summary\u001b[0;34m(m, n)\u001b[0m\n\u001b[1;32m    164\u001b[0m \u001b[0;32mdef\u001b[0m \u001b[0mmodel_summary\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mm\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0mLearner\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mn\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0mint\u001b[0m\u001b[0;34m=\u001b[0m\u001b[0;36m70\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m    165\u001b[0m     \u001b[0;34m\"Print a summary of `m` using a output text width of `n` chars\"\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m--> 166\u001b[0;31m     \u001b[0minfo\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mlayers_info\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mm\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m    167\u001b[0m     \u001b[0mheader\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0;34m[\u001b[0m\u001b[0;34m\"Layer (type)\"\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0;34m\"Output Shape\"\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0;34m\"Param #\"\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0;34m\"Trainable\"\u001b[0m\u001b[0;34m]\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m    168\u001b[0m     \u001b[0mres\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0;34m\"=\"\u001b[0m \u001b[0;34m*\u001b[0m \u001b[0mn\u001b[0m \u001b[0;34m+\u001b[0m \u001b[0;34m\"\\n\"\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n",
      "\u001b[0;32m/opt/conda/envs/fastai/lib/python3.6/site-packages/fastai/callbacks/hooks.py\u001b[0m in \u001b[0;36mlayers_info\u001b[0;34m(m)\u001b[0m\n\u001b[1;32m    158\u001b[0m     \u001b[0mfunc\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0;32mlambda\u001b[0m \u001b[0mm\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0mlist\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mmap\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mget_layer_name\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mflatten_model\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mm\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m    159\u001b[0m     \u001b[0mlayers_names\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mfunc\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mm\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mmodel\u001b[0m\u001b[0;34m)\u001b[0m \u001b[0;32mif\u001b[0m \u001b[0misinstance\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mm\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mLearner\u001b[0m\u001b[0;34m)\u001b[0m \u001b[0;32melse\u001b[0m \u001b[0mfunc\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mm\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m--> 160\u001b[0;31m     \u001b[0mlayers_sizes\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mlayers_params\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mlayers_trainable\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mparams_size\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mm\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m    161\u001b[0m     \u001b[0mlayer_info\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mnamedtuple\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m'Layer_Information'\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0;34m[\u001b[0m\u001b[0;34m'Layer'\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0;34m'OutputSize'\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0;34m'Params'\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0;34m'Trainable'\u001b[0m\u001b[0;34m]\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m    162\u001b[0m     \u001b[0;32mreturn\u001b[0m \u001b[0mlist\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mmap\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mlayer_info\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mlayers_names\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mlayers_sizes\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mlayers_params\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mlayers_trainable\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n",
      "\u001b[0;32m/opt/conda/envs/fastai/lib/python3.6/site-packages/fastai/callbacks/hooks.py\u001b[0m in \u001b[0;36mparams_size\u001b[0;34m(m, size)\u001b[0m\n\u001b[1;32m    139\u001b[0m             \u001b[0;32mraise\u001b[0m \u001b[0mException\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m\"This is an empty `Learner` and `Learner.summary` requires some data to pass through the model.\"\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m    140\u001b[0m         \u001b[0mds_type\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mDatasetType\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mTrain\u001b[0m \u001b[0;32mif\u001b[0m \u001b[0mm\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mdata\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mtrain_dl\u001b[0m \u001b[0;32melse\u001b[0m \u001b[0;34m(\u001b[0m\u001b[0mDatasetType\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mValid\u001b[0m \u001b[0;32mif\u001b[0m \u001b[0mm\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mdata\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mvalid_dl\u001b[0m \u001b[0;32melse\u001b[0m \u001b[0mDatasetType\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mTest\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m--> 141\u001b[0;31m         \u001b[0mx\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mm\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mdata\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mone_batch\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mds_type\u001b[0m\u001b[0;34m=\u001b[0m\u001b[0mds_type\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mdetach\u001b[0m\u001b[0;34m=\u001b[0m\u001b[0;32mFalse\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mdenorm\u001b[0m\u001b[0;34m=\u001b[0m\u001b[0;32mFalse\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m[\u001b[0m\u001b[0;36m0\u001b[0m\u001b[0;34m]\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m    142\u001b[0m         \u001b[0mx\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0;34m[\u001b[0m\u001b[0mo\u001b[0m\u001b[0;34m[\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;36m1\u001b[0m\u001b[0;34m]\u001b[0m \u001b[0;32mfor\u001b[0m \u001b[0mo\u001b[0m \u001b[0;32min\u001b[0m \u001b[0mx\u001b[0m\u001b[0;34m]\u001b[0m  \u001b[0;32mif\u001b[0m \u001b[0mis_listy\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mx\u001b[0m\u001b[0;34m)\u001b[0m \u001b[0;32melse\u001b[0m \u001b[0mx\u001b[0m\u001b[0;34m[\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;36m1\u001b[0m\u001b[0;34m]\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m    143\u001b[0m         \u001b[0mm\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mm\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mmodel\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n",
      "\u001b[0;32m/opt/conda/envs/fastai/lib/python3.6/site-packages/fastai/basic_data.py\u001b[0m in \u001b[0;36mone_batch\u001b[0;34m(self, ds_type, detach, denorm, cpu)\u001b[0m\n\u001b[1;32m    166\u001b[0m         \u001b[0mw\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mnum_workers\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m    167\u001b[0m         \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mnum_workers\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0;36m0\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m--> 168\u001b[0;31m         \u001b[0;32mtry\u001b[0m\u001b[0;34m:\u001b[0m     \u001b[0mx\u001b[0m\u001b[0;34m,\u001b[0m\u001b[0my\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mnext\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0miter\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mdl\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m    169\u001b[0m         \u001b[0;32mfinally\u001b[0m\u001b[0;34m:\u001b[0m \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mnum_workers\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mw\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m    170\u001b[0m         \u001b[0;32mif\u001b[0m \u001b[0mdetach\u001b[0m\u001b[0;34m:\u001b[0m \u001b[0mx\u001b[0m\u001b[0;34m,\u001b[0m\u001b[0my\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mto_detach\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mx\u001b[0m\u001b[0;34m,\u001b[0m\u001b[0mcpu\u001b[0m\u001b[0;34m=\u001b[0m\u001b[0mcpu\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m,\u001b[0m\u001b[0mto_detach\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0my\u001b[0m\u001b[0;34m,\u001b[0m\u001b[0mcpu\u001b[0m\u001b[0;34m=\u001b[0m\u001b[0mcpu\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n",
      "\u001b[0;32m/opt/conda/envs/fastai/lib/python3.6/site-packages/fastai/basic_data.py\u001b[0m in \u001b[0;36m__iter__\u001b[0;34m(self)\u001b[0m\n\u001b[1;32m     73\u001b[0m     \u001b[0;32mdef\u001b[0m \u001b[0m__iter__\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mself\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m     74\u001b[0m         \u001b[0;34m\"Process and returns items from `DataLoader`.\"\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m---> 75\u001b[0;31m         \u001b[0;32mfor\u001b[0m \u001b[0mb\u001b[0m \u001b[0;32min\u001b[0m \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mdl\u001b[0m\u001b[0;34m:\u001b[0m \u001b[0;32myield\u001b[0m \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mproc_batch\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mb\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m     76\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m     77\u001b[0m     \u001b[0;34m@\u001b[0m\u001b[0mclassmethod\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n",
      "\u001b[0;32m/opt/conda/envs/fastai/lib/python3.6/site-packages/torch/utils/data/dataloader.py\u001b[0m in \u001b[0;36m__iter__\u001b[0;34m(self)\u001b[0m\n\u001b[1;32m    817\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m    818\u001b[0m     \u001b[0;32mdef\u001b[0m \u001b[0m__iter__\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mself\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m--> 819\u001b[0;31m         \u001b[0;32mreturn\u001b[0m \u001b[0m_DataLoaderIter\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mself\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m    820\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m    821\u001b[0m     \u001b[0;32mdef\u001b[0m \u001b[0m__len__\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mself\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n",
      "\u001b[0;32m/opt/conda/envs/fastai/lib/python3.6/site-packages/torch/utils/data/dataloader.py\u001b[0m in \u001b[0;36m__init__\u001b[0;34m(self, loader)\u001b[0m\n\u001b[1;32m    582\u001b[0m             \u001b[0;31m# prime the prefetch loop\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m    583\u001b[0m             \u001b[0;32mfor\u001b[0m \u001b[0m_\u001b[0m \u001b[0;32min\u001b[0m \u001b[0mrange\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;36m2\u001b[0m \u001b[0;34m*\u001b[0m \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mnum_workers\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m--> 584\u001b[0;31m                 \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0m_put_indices\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m    585\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m    586\u001b[0m     \u001b[0;32mdef\u001b[0m \u001b[0m__len__\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mself\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n",
      "\u001b[0;32m/opt/conda/envs/fastai/lib/python3.6/site-packages/torch/utils/data/dataloader.py\u001b[0m in \u001b[0;36m_put_indices\u001b[0;34m(self)\u001b[0m\n\u001b[1;32m    644\u001b[0m     \u001b[0;32mdef\u001b[0m \u001b[0m_put_indices\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mself\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m    645\u001b[0m         \u001b[0;32massert\u001b[0m \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mbatches_outstanding\u001b[0m \u001b[0;34m<\u001b[0m \u001b[0;36m2\u001b[0m \u001b[0;34m*\u001b[0m \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mnum_workers\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m--> 646\u001b[0;31m         \u001b[0mindices\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mnext\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0msample_iter\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0;32mNone\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m    647\u001b[0m         \u001b[0;32mif\u001b[0m \u001b[0mindices\u001b[0m \u001b[0;32mis\u001b[0m \u001b[0;32mNone\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m    648\u001b[0m             \u001b[0;32mreturn\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n",
      "\u001b[0;32m/opt/conda/envs/fastai/lib/python3.6/site-packages/torch/utils/data/sampler.py\u001b[0m in \u001b[0;36m__iter__\u001b[0;34m(self)\u001b[0m\n\u001b[1;32m    158\u001b[0m     \u001b[0;32mdef\u001b[0m \u001b[0m__iter__\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mself\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m    159\u001b[0m         \u001b[0mbatch\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0;34m[\u001b[0m\u001b[0;34m]\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m--> 160\u001b[0;31m         \u001b[0;32mfor\u001b[0m \u001b[0midx\u001b[0m \u001b[0;32min\u001b[0m \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0msampler\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m    161\u001b[0m             \u001b[0mbatch\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mappend\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0midx\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m    162\u001b[0m             \u001b[0;32mif\u001b[0m \u001b[0mlen\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mbatch\u001b[0m\u001b[0;34m)\u001b[0m \u001b[0;34m==\u001b[0m \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mbatch_size\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n",
      "\u001b[0;32m/opt/conda/envs/fastai/lib/python3.6/site-packages/fastai/text/data.py\u001b[0m in \u001b[0;36m__iter__\u001b[0;34m(self)\u001b[0m\n\u001b[1;32m    118\u001b[0m         \u001b[0msz\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mbs\u001b[0m\u001b[0;34m*\u001b[0m\u001b[0;36m50\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m    119\u001b[0m         \u001b[0mck_idx\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0;34m[\u001b[0m\u001b[0midxs\u001b[0m\u001b[0;34m[\u001b[0m\u001b[0mi\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0mi\u001b[0m\u001b[0;34m+\u001b[0m\u001b[0msz\u001b[0m\u001b[0;34m]\u001b[0m \u001b[0;32mfor\u001b[0m \u001b[0mi\u001b[0m \u001b[0;32min\u001b[0m \u001b[0mrange\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;36m0\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mlen\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0midxs\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0msz\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m]\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m--> 120\u001b[0;31m         \u001b[0msort_idx\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mnp\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mconcatenate\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m[\u001b[0m\u001b[0msorted\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0ms\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mkey\u001b[0m\u001b[0;34m=\u001b[0m\u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mkey\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mreverse\u001b[0m\u001b[0;34m=\u001b[0m\u001b[0;32mTrue\u001b[0m\u001b[0;34m)\u001b[0m \u001b[0;32mfor\u001b[0m \u001b[0ms\u001b[0m \u001b[0;32min\u001b[0m \u001b[0mck_idx\u001b[0m\u001b[0;34m]\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m    121\u001b[0m         \u001b[0msz\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mbs\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m    122\u001b[0m         \u001b[0mck_idx\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0;34m[\u001b[0m\u001b[0msort_idx\u001b[0m\u001b[0;34m[\u001b[0m\u001b[0mi\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0mi\u001b[0m\u001b[0;34m+\u001b[0m\u001b[0msz\u001b[0m\u001b[0;34m]\u001b[0m \u001b[0;32mfor\u001b[0m \u001b[0mi\u001b[0m \u001b[0;32min\u001b[0m \u001b[0mrange\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;36m0\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mlen\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0msort_idx\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0msz\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m]\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n",
      "\u001b[0;32m/opt/conda/envs/fastai/lib/python3.6/site-packages/fastai/text/data.py\u001b[0m in \u001b[0;36m<listcomp>\u001b[0;34m(.0)\u001b[0m\n\u001b[1;32m    118\u001b[0m         \u001b[0msz\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mbs\u001b[0m\u001b[0;34m*\u001b[0m\u001b[0;36m50\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m    119\u001b[0m         \u001b[0mck_idx\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0;34m[\u001b[0m\u001b[0midxs\u001b[0m\u001b[0;34m[\u001b[0m\u001b[0mi\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0mi\u001b[0m\u001b[0;34m+\u001b[0m\u001b[0msz\u001b[0m\u001b[0;34m]\u001b[0m \u001b[0;32mfor\u001b[0m \u001b[0mi\u001b[0m \u001b[0;32min\u001b[0m \u001b[0mrange\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;36m0\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mlen\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0midxs\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0msz\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m]\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m--> 120\u001b[0;31m         \u001b[0msort_idx\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mnp\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mconcatenate\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m[\u001b[0m\u001b[0msorted\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0ms\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mkey\u001b[0m\u001b[0;34m=\u001b[0m\u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mkey\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mreverse\u001b[0m\u001b[0;34m=\u001b[0m\u001b[0;32mTrue\u001b[0m\u001b[0;34m)\u001b[0m \u001b[0;32mfor\u001b[0m \u001b[0ms\u001b[0m \u001b[0;32min\u001b[0m \u001b[0mck_idx\u001b[0m\u001b[0;34m]\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m    121\u001b[0m         \u001b[0msz\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mbs\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m    122\u001b[0m         \u001b[0mck_idx\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0;34m[\u001b[0m\u001b[0msort_idx\u001b[0m\u001b[0;34m[\u001b[0m\u001b[0mi\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0mi\u001b[0m\u001b[0;34m+\u001b[0m\u001b[0msz\u001b[0m\u001b[0;34m]\u001b[0m \u001b[0;32mfor\u001b[0m \u001b[0mi\u001b[0m \u001b[0;32min\u001b[0m \u001b[0mrange\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;36m0\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mlen\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0msort_idx\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0msz\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m]\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n",
      "\u001b[0;32m/opt/conda/envs/fastai/lib/python3.6/site-packages/fastai/text/data.py\u001b[0m in \u001b[0;36m<lambda>\u001b[0;34m(t)\u001b[0m\n\u001b[1;32m    260\u001b[0m         \u001b[0mval_bs\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mifnone\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mval_bs\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mbs\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m    261\u001b[0m         \u001b[0mcollate_fn\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mpartial\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mpad_collate\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mpad_idx\u001b[0m\u001b[0;34m=\u001b[0m\u001b[0mpad_idx\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mpad_first\u001b[0m\u001b[0;34m=\u001b[0m\u001b[0mpad_first\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mbackwards\u001b[0m\u001b[0;34m=\u001b[0m\u001b[0mbackwards\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m--> 262\u001b[0;31m         \u001b[0mtrain_sampler\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mSortishSampler\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mdatasets\u001b[0m\u001b[0;34m[\u001b[0m\u001b[0;36m0\u001b[0m\u001b[0;34m]\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mx\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mkey\u001b[0m\u001b[0;34m=\u001b[0m\u001b[0;32mlambda\u001b[0m \u001b[0mt\u001b[0m\u001b[0;34m:\u001b[0m \u001b[0mlen\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mdatasets\u001b[0m\u001b[0;34m[\u001b[0m\u001b[0;36m0\u001b[0m\u001b[0;34m]\u001b[0m\u001b[0;34m[\u001b[0m\u001b[0mt\u001b[0m\u001b[0;34m]\u001b[0m\u001b[0;34m[\u001b[0m\u001b[0;36m0\u001b[0m\u001b[0;34m]\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mdata\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mbs\u001b[0m\u001b[0;34m=\u001b[0m\u001b[0mbs\u001b[0m\u001b[0;34m//\u001b[0m\u001b[0;36m2\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m    263\u001b[0m         \u001b[0mtrain_dl\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mDataLoader\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mdatasets\u001b[0m\u001b[0;34m[\u001b[0m\u001b[0;36m0\u001b[0m\u001b[0;34m]\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mbatch_size\u001b[0m\u001b[0;34m=\u001b[0m\u001b[0mbs\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0msampler\u001b[0m\u001b[0;34m=\u001b[0m\u001b[0mtrain_sampler\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mdrop_last\u001b[0m\u001b[0;34m=\u001b[0m\u001b[0;32mTrue\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0;34m**\u001b[0m\u001b[0mdl_kwargs\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m    264\u001b[0m         \u001b[0mdataloaders\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0;34m[\u001b[0m\u001b[0mtrain_dl\u001b[0m\u001b[0;34m]\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n",
      "\u001b[0;32m/opt/conda/envs/fastai/lib/python3.6/site-packages/fastai/data_block.py\u001b[0m in \u001b[0;36m__getitem__\u001b[0;34m(self, idxs)\u001b[0m\n\u001b[1;32m    631\u001b[0m         \u001b[0midxs\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mtry_int\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0midxs\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m    632\u001b[0m         \u001b[0;32mif\u001b[0m \u001b[0misinstance\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0midxs\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mIntegral\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m--> 633\u001b[0;31m             \u001b[0;32mif\u001b[0m \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mitem\u001b[0m \u001b[0;32mis\u001b[0m \u001b[0;32mNone\u001b[0m\u001b[0;34m:\u001b[0m \u001b[0mx\u001b[0m\u001b[0;34m,\u001b[0m\u001b[0my\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mx\u001b[0m\u001b[0;34m[\u001b[0m\u001b[0midxs\u001b[0m\u001b[0;34m]\u001b[0m\u001b[0;34m,\u001b[0m\u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0my\u001b[0m\u001b[0;34m[\u001b[0m\u001b[0midxs\u001b[0m\u001b[0;34m]\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m    634\u001b[0m             \u001b[0;32melse\u001b[0m\u001b[0;34m:\u001b[0m                 \u001b[0mx\u001b[0m\u001b[0;34m,\u001b[0m\u001b[0my\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mitem\u001b[0m   \u001b[0;34m,\u001b[0m\u001b[0;36m0\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m    635\u001b[0m             \u001b[0;32mif\u001b[0m \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mtfms\u001b[0m \u001b[0;32mor\u001b[0m \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mtfmargs\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n",
      "\u001b[0;32m/opt/conda/envs/fastai/lib/python3.6/site-packages/fastai/data_block.py\u001b[0m in \u001b[0;36m__getitem__\u001b[0;34m(self, idxs)\u001b[0m\n\u001b[1;32m    106\u001b[0m \u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m    107\u001b[0m     \u001b[0;32mdef\u001b[0m \u001b[0m__getitem__\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mself\u001b[0m\u001b[0;34m,\u001b[0m\u001b[0midxs\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0mint\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m->\u001b[0m\u001b[0mAny\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m--> 108\u001b[0;31m         \u001b[0midxs\u001b[0m \u001b[0;34m=\u001b[0m \u001b[0mtry_int\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0midxs\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m    109\u001b[0m         \u001b[0;32mif\u001b[0m \u001b[0misinstance\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0midxs\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mIntegral\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m:\u001b[0m \u001b[0;32mreturn\u001b[0m \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mget\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0midxs\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m    110\u001b[0m         \u001b[0;32melse\u001b[0m\u001b[0;34m:\u001b[0m \u001b[0;32mreturn\u001b[0m \u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mnew\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mitems\u001b[0m\u001b[0;34m[\u001b[0m\u001b[0midxs\u001b[0m\u001b[0;34m]\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0minner_df\u001b[0m\u001b[0;34m=\u001b[0m\u001b[0mindex_row\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mself\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0minner_df\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0midxs\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n",
      "\u001b[0;32m/opt/conda/envs/fastai/lib/python3.6/site-packages/fastai/torch_core.py\u001b[0m in \u001b[0;36mtry_int\u001b[0;34m(o)\u001b[0m\n\u001b[1;32m    366\u001b[0m     \u001b[0;31m# NB: single-item rank-1 array/tensor can be converted to int, but we don't want to do this\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m    367\u001b[0m     \u001b[0;32mif\u001b[0m \u001b[0misinstance\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mo\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0;34m(\u001b[0m\u001b[0mnp\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mndarray\u001b[0m\u001b[0;34m,\u001b[0m\u001b[0mTensor\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m:\u001b[0m \u001b[0;32mreturn\u001b[0m \u001b[0mo\u001b[0m \u001b[0;32mif\u001b[0m \u001b[0mo\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mndim\u001b[0m \u001b[0;32melse\u001b[0m \u001b[0mint\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mo\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m--> 368\u001b[0;31m     \u001b[0;32mif\u001b[0m \u001b[0misinstance\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mo\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mcollections\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mSized\u001b[0m\u001b[0;34m)\u001b[0m \u001b[0;32mor\u001b[0m \u001b[0mgetattr\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mo\u001b[0m\u001b[0;34m,\u001b[0m\u001b[0;34m'__array_interface__'\u001b[0m\u001b[0;34m,\u001b[0m\u001b[0;32mFalse\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m:\u001b[0m \u001b[0;32mreturn\u001b[0m \u001b[0mo\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m    369\u001b[0m     \u001b[0;32mtry\u001b[0m\u001b[0;34m:\u001b[0m \u001b[0;32mreturn\u001b[0m \u001b[0mint\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mo\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m    370\u001b[0m     \u001b[0;32mexcept\u001b[0m\u001b[0;34m:\u001b[0m \u001b[0;32mreturn\u001b[0m \u001b[0mo\u001b[0m\u001b[0;34m\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n",
      "\u001b[0;31mKeyboardInterrupt\u001b[0m: "
     ]
    }
   ],
   "source": [
    "learn.summary()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 20,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th>text</th>\n",
       "      <th>target</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <td>xxbos xxup how xxup the xxup apostles xxup deceived xxup humanity xxup about xxup god xxup and xxup jesus \\n \\n  % of readers think this story is xxmaj fact . xxmaj add your two cents . \\n \\n  xxmaj headline : xxmaj bitcoin &amp; xxmaj blockchain xxmaj searches xxmaj exceed xxmaj trump ! xxmaj blockchain xxmaj stocks xxmaj are xxmaj next ! \\n \\n  xxup how</td>\n",
       "      <td>fake</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>xxbos xxmaj the difficulty with analyzing the country ’s weaknesses and strengths comes from the fact that the xxmaj north xxmaj korean system — which is made up of the xxmaj kim dynasty , the xxmaj north xxmaj korean state , and the economy constructed to maintain them both — is unlike any other on earth . xxmaj by now , its brand of totalitarianism ( “ our own style</td>\n",
       "      <td>political</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>xxbos 76 xxmaj years of xxmaj pearl xxmaj harbor xxmaj lies \\n \\n  xxmaj headline : xxmaj bitcoin &amp; xxmaj blockchain xxmaj searches xxmaj exceed xxmaj trump ! xxmaj blockchain xxmaj stocks xxmaj are xxmaj next ! \\n \\n  xxmaj donald xxmaj trump is tweeting about a particular spot in xxmaj hawaii . xxmaj he visited it recently on his way to threaten war in xxmaj asia .</td>\n",
       "      <td>fake</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>xxbos xxmaj from xxmaj conservapedia \\n \\n  xxmaj please note that more recent studies have been done than the one you cite from 1994 . xxmaj please reference the following study from the xxmaj salk xxmaj institute : \\n \\n  xxmaj salk xxmaj institute for xxmaj biological xxmaj studies , xxmaj san xxmaj diego , xxup ca xxunk . \\n \\n  xxmaj the anterior hypothalamus of the</td>\n",
       "      <td>bias</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <td>xxbos xxup pa - xxmaj sen : xxup busted ! xxmaj trump 's xxmaj political xxmaj godfather xxmaj did xxmaj interview xxmaj in 2006 xxmaj with xxmaj holocaust - xxmaj denying xxmaj publication xxmaj whoa : xxmaj rep. xxmaj lou xxmaj barletta , an immigration hardliner running in a crowded xxup us xxmaj senate primary in xxmaj pennsylvania , came in contact over the years with fringe organizations and individuals</td>\n",
       "      <td>political</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>"
      ],
      "text/plain": [
       "<IPython.core.display.HTML object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "data_clas.show_batch()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "Total time: 1:28:20 <p><table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: left;\">\n",
       "      <th>epoch</th>\n",
       "      <th>train_loss</th>\n",
       "      <th>valid_loss</th>\n",
       "      <th>accuracy</th>\n",
       "      <th>time</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <td>0</td>\n",
       "      <td>0.468599</td>\n",
       "      <td>0.365651</td>\n",
       "      <td>0.885425</td>\n",
       "      <td>1:28:20</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>"
      ],
      "text/plain": [
       "<IPython.core.display.HTML object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "learn.fit_one_cycle(1, 1e-2)\n",
    "learn.save('good_model')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "data": {
      "text/html": [
       "Total time: 1:30:27 <p><table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: left;\">\n",
       "      <th>epoch</th>\n",
       "      <th>train_loss</th>\n",
       "      <th>valid_loss</th>\n",
       "      <th>accuracy</th>\n",
       "      <th>time</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <td>0</td>\n",
       "      <td>0.315539</td>\n",
       "      <td>0.247238</td>\n",
       "      <td>0.923329</td>\n",
       "      <td>1:30:27</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>"
      ],
      "text/plain": [
       "<IPython.core.display.HTML object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "learn.freeze_to(-2)\n",
    "learn.fit_one_cycle(1, slice(5e-3/2., 5e-3))\n",
    "learn.save('good_model_epoc_2')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "Total time: 42:16 <p><table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: left;\">\n",
       "      <th>epoch</th>\n",
       "      <th>train_loss</th>\n",
       "      <th>valid_loss</th>\n",
       "      <th>accuracy</th>\n",
       "      <th>time</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <td>0</td>\n",
       "      <td>0.266535</td>\n",
       "      <td>0.218171</td>\n",
       "      <td>0.932730</td>\n",
       "      <td>42:16</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>"
      ],
      "text/plain": [
       "<IPython.core.display.HTML object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "learn.unfreeze()\n",
    "learn.fit_one_cycle(1, slice(2e-3/100, 2e-3))\n",
    "learn.save('good_model_epoc_3')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(Category bias,\n",
       " tensor(0),\n",
       " tensor([0.4265, 0.0063, 0.0678, 0.0084, 0.0070, 0.0047, 0.3237, 0.0067, 0.0166,\n",
       "         0.0010, 0.1141, 0.0171]))"
      ]
     },
     "execution_count": 15,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# BBC\n",
    "learn.predict(\"Israeli PM Benjamin Netanyahu has said he will annex Jewish settlements in the occupied West Bank if he is re-elected.Israelis go to the polls on Tuesday and Mr Netanyahu is competing for votes with right-wing parties who support annexing part of the West Bank.The settlements are illegal under international law, though Israel disputes this.Last month the US recognised the occupied Golan Heights, seized from Syria in 1967, as Israeli territory.Can Jewish settlement issue be resolved?What Trump’s Golan Heights move really meansIsrael's Benjamin Netanyahu: Commando turned PMIsrael has settled about 400,000 Jews in West Bank settlements, with another 200,000 living in East Jerusalem. There are about 2.5 million Palestinians living in the West Bank.Palestinians want to establish a state in the occupied West Bank, East Jerusalem and the Gaza Strip.What happens to the settlements is one of the most contentious issues between Israel and the Palestinians - Palestinians say the presence of settlements make a future independent state impossible.Israel says the Palestinians are using the issue of settlements as a pretext to avoid direct peace talks. It says settlements are not a genuine obstacle to peace and are negotiable.What exactly did Netanyahu say?He was asked during an interview on Israeli TV why he had not extended Israeli sovereignty to large settlements in the West Bank.'You are asking whether we are moving on to the next stage - the answer is yes, we will move to the next stage,' he said.Image copyrightREUTERSImage captionMr Netanyahu is seeking re-election'I am going to extend [Israeli] sovereignty and I don't distinguish between settlement blocs and the isolated settlements.'A spokesman for Palestinian leader Mahmoud Abbas told Reuters: 'Any measures and any announcements will not change the facts. Settlements are illegal and they will be removed.'Potentially explosive commentsBy Sebastian Usher, BBC Arab affairs editorThese comments by Benjamin Netanyahu are potentially explosive over an issue that has helped stall peace efforts for years.They will resonate with several parties with which he'll try to form a coalition government if he wins the biggest share of votes.But the very idea of annexation will rouse new Palestinian fury, as well as international condemnation.Mr Netanyahu may have been emboldened by the Trump administration, which just last month recognised Israeli sovereignty over the Golan Heights.What is the political background?Mr Netanyahu's right-wing Likud party is in a tight race with the new centre-right Blue and White alliance.However other parties, some of which support annexation, could end up being kingmakers when they try to form a governing coalition.Israel election: Who are the key candidates?The ex-military chief trying to unseat NetanyahuIn Mr Netanyahu's own Likud party, 28 out of the 29 lawmakers running for re-election are on record as supporting this approach. Until now the prime minister was the only exception.What is the situation of peace negotiations?Mr Trump's administration is preparing to unveil a long-awaited Middle East peace plan, which US officials say will be fair.However the Trump administration has carried out a series of actions that have inflamed Palestinian opinion and generally pleased Israel.In 2017 Mr Trump announced that the US recognised Jerusalem as Israel's capital, overturning decades of official US policy.In response Mr Abbas cut off relations with the US, saying the US could no longer be a peace broker.Last year the US stopped contributing to the UN Relief and Works Agency (Unrwa), which has been looking after Palestinian refugees since 1949.Last month President Trump officially recognised Israeli sovereignty over the occupied Golan Heights.Peace negotiations between Israel and the Palestinians have been at a standstill since 2014, when a US-brokered attempt to reach a deal collapsed.\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(Category political,\n",
       " tensor(6),\n",
       " tensor([1.8421e-02, 1.4594e-02, 6.2306e-03, 2.5443e-03, 3.9842e-04, 2.4703e-04,\n",
       "         9.5339e-01, 5.5935e-04, 9.2467e-05, 2.7548e-04, 9.5179e-04, 2.2999e-03]))"
      ]
     },
     "execution_count": 16,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# Fox News:\n",
    "learn.predict(\"Former President Barack Obama said on Saturday that he is worried that progressives are creating a “circular firing squad” as prospective Democratic presidential candidates race to the left on a number of hot topic issues ahead of the 2020 election.“The way we structure democracy requires you to take into account people who don’t agree with you,” he said at an Obama Foundation town hall event in Berlin, according to The New York Post. “And that by definition means you’re not going to get 100 percent of what you want.”BARACK OBAMA STILL BELIEVES BIDEN WOULD BE 'AN EXCELLENT PRESIDENT' AMID INAPPROPRIATE TOUCHING ALLEGATIONS: REPORT“One of the things I do worry about sometimes among progressives … we start sometimes creating what’s called a ‘circular firing squad’ where you start shooting at your allies because one of them has strayed from purity on the issues,” he said.Obama’s remarks come as freshman House Democrats such as Rep. Alexandria Ocasio-Cortez, D-N.Y., have pushed once-fringe positions on Medicare-for-all, the Green New Deal and reparations for slavery. In turn, 2020 presidential hopefuls have also taken some of those positions.In that climate, candidates have come under criticism for their past stances from activists. South Bend Mayor Pete Buttigieg was forced this week to address remarks he made in 2015 when he said that “all lives matter” -- which some activists say is a counterslogan to the “black lives matter” sloganSen. Kamala Harris, D-Calif., meanwhile has been hit by controversy over her past as a prosecutor. A scathing op-ed published in January in The New York Times, written by law professor Lara Bazelon, has kickstarted renewed scrutiny.Obama reportedly warns freshmen House Democrats about pricey policy proposalsVideoBazelon says Harris previously 'fought tooth and nail to uphold wrongful convictions that had been secured through official misconduct that included evidence tampering, false testimony and the suppression of crucial information by prosecutors.'Bazelon further suggested that Harris should 'apologize to the wrongfully convicted people she has fought to keep in prison and to do what she can to make sure they get justice' or otherwise make clear she has 'radically broken from her past.'Former vice president under Obama, Joe Biden, meanwhile has faced criticism for inappropriate past physical contact with women, as well a a 1993 speech on crime in which he warned of “predators on our streets”'They are beyond the pale many of those people, beyond the pale,' Biden continued. 'And it's a sad commentary on society. We have no choice but to take them out of society.'The latter was reminiscent of heat 2016 presidential nominee Hillary Clinton took from activists for her description of some gang members as “superpredators” in 1996.Obama himself may not escape criticism in the election cycle. His signature health care legislation, the Affordable Care Act, is quickly being eclipsed by calls from Democrats for single-payer and Medicare-for-all plans. Meanwhile, a number of Democrats have said they are open to reparations for black Americans for slavery -- something that Obama opposed when he was in office.\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(Category political,\n",
       " tensor(6),\n",
       " tensor([6.7470e-02, 1.3856e-02, 8.7386e-03, 3.3480e-03, 1.4125e-03, 4.9837e-05,\n",
       "         9.0135e-01, 7.2659e-04, 1.1479e-04, 7.8975e-05, 4.7441e-04, 2.3841e-03]))"
      ]
     },
     "execution_count": 17,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# BrightBert again\n",
    "learn.predict(\"The border agencies need tougher leadership, President Donald Trump declared Friday as he dropped plans to appoint a long-time agency staffer to run the Immigration and Customs Enforcement agency (ICE).'Ron [Vitiello is] a good man,” Trump told reporters. 'But we’re going in a tougher direction. We want to go in a tougher direction.” Trump’s 'tougher direction” statement suggests he may pressure Department of Homeland Secretary (DHS) Secretary Kirstjen Nielsen to implement policies that top agency staffers oppose, such as rejecting legal interpretations and bureaucratic practices set by former President Barack Obama. Immigration reformers blame those Obama policies for encouraging the wave of economic migrants from Central America.Breitbart TVDonald Trump Says Everything Jared Kushner Touched ‘Turned To Gold’The shift comes amid the growing wave of Central American economic migrants who are using Obama-era legal loopholes to walk through the border wall and into jobs, neighborhoods, and blue-collar schools throughout the United States. That wave is expected to deliver one million migrants into the United States by October, and it is made possible because Democrats are blocking any reform the border loopholes.Immigration reformers fear that Obama-appointed staffers and former business lobbyists are keeping Trump in the dark about ways to improve operation at the DHS. 'I don’t now know if the President is getting the information he needs about what powers he has,” according to Rosemary Jenks, policy director at the Center for Immigration Studies. 'Secretary Nielsen and some of the attorneys in DHS are blocking the information because they are afraid of implementing some of the things they can do,” partly because they are afraid of lawsuits, she said.For example, many so-called 'Unaccompanied Alien Children” are being smuggled up the border because Trump’s agencies will pass them to their illegal immigrant parents living throughout the United States, under policies set by Obama. But those youths and children should be sent home, said Jenks, because the 2008 law only protects trafficked victims, such as forced prostitutes, not youths and children who have parents in the United States or who are willingly smuggled up to the border. According to the Washington Post, Vitiello’s exit was prompted by Steve Miller, one of Trump’s first aides who earlier played a key role in derailing the 2013 'Gang of Eight” amnesty and cheap labor bill. The Post said:Six administration officials said Friday that the decision to jettison Vitiello was a sign of the expanding influence that Miller now wields over immigration matters in the White House, particularly as Trump lashes out at Mexico and Central American nations — as well as Homeland Security officials and aides who express doubts about the legality of his ideas.The New York Times reported:One person familiar with the president’s thinking said that Mr. Trump believed that Mr. Vitiello did not favor closing the border, as the president had proposed before backing off that threat this week.Another person said that Stephen Miller, the president’s chief policy adviser and a supporter of curtailing legal and illegal immigration, did not support Mr. Vitiello’s nomination.Vitiello’s defenders lashed out at Miller. The Washington Post highlighted the complaints:'Ron Vitiello has spent as much time defending our nation’s borders as Stephen Miller has been alive,” one official said of Miller, who is 33.One senior official said: 'This is part of an increasingly desperate effort by Stephen to throw people under the bus when the policies he has advocated are not effective. Once it becomes clear that Stephen’s policies aren’t working, he tells the president, ‘They’re not the right people.’” But Vitiello’s appointment was opposed by the ICE officers’ union, the National ICE Council. Vitiello 'lacks the judgment and professionalism to effectively lead a federal agency,” said a February letter from union President Chris Crane.\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(Category political,\n",
       " tensor(6),\n",
       " tensor([0.2512, 0.0144, 0.0782, 0.0076, 0.0270, 0.0049, 0.4827, 0.0146, 0.0122,\n",
       "         0.0010, 0.0789, 0.0272]))"
      ]
     },
     "execution_count": 11,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# BBC\n",
    "learn.predict(\"Israeli PM Benjamin Netanyahu has said he will annex Jewish settlements in the occupied West Bank if he is re-elected.Israelis go to the polls on Tuesday and Mr Netanyahu is competing for votes with right-wing parties who support annexing part of the West Bank.The settlements are illegal under international law, though Israel disputes this.Last month the US recognised the occupied Golan Heights, seized from Syria in 1967, as Israeli territory.Can Jewish settlement issue be resolved?What Trump’s Golan Heights move really meansIsrael's Benjamin Netanyahu: Commando turned PMIsrael has settled about 400,000 Jews in West Bank settlements, with another 200,000 living in East Jerusalem. There are about 2.5 million Palestinians living in the West Bank.Palestinians want to establish a state in the occupied West Bank, East Jerusalem and the Gaza Strip.What happens to the settlements is one of the most contentious issues between Israel and the Palestinians - Palestinians say the presence of settlements make a future independent state impossible.Israel says the Palestinians are using the issue of settlements as a pretext to avoid direct peace talks. It says settlements are not a genuine obstacle to peace and are negotiable.What exactly did Netanyahu say?He was asked during an interview on Israeli TV why he had not extended Israeli sovereignty to large settlements in the West Bank.'You are asking whether we are moving on to the next stage - the answer is yes, we will move to the next stage,' he said.Image copyrightREUTERSImage captionMr Netanyahu is seeking re-election'I am going to extend [Israeli] sovereignty and I don't distinguish between settlement blocs and the isolated settlements.'A spokesman for Palestinian leader Mahmoud Abbas told Reuters: 'Any measures and any announcements will not change the facts. Settlements are illegal and they will be removed.'Potentially explosive commentsBy Sebastian Usher, BBC Arab affairs editorThese comments by Benjamin Netanyahu are potentially explosive over an issue that has helped stall peace efforts for years.They will resonate with several parties with which he'll try to form a coalition government if he wins the biggest share of votes.But the very idea of annexation will rouse new Palestinian fury, as well as international condemnation.Mr Netanyahu may have been emboldened by the Trump administration, which just last month recognised Israeli sovereignty over the Golan Heights.What is the political background?Mr Netanyahu's right-wing Likud party is in a tight race with the new centre-right Blue and White alliance.However other parties, some of which support annexation, could end up being kingmakers when they try to form a governing coalition.Israel election: Who are the key candidates?The ex-military chief trying to unseat NetanyahuIn Mr Netanyahu's own Likud party, 28 out of the 29 lawmakers running for re-election are on record as supporting this approach. Until now the prime minister was the only exception.What is the situation of peace negotiations?Mr Trump's administration is preparing to unveil a long-awaited Middle East peace plan, which US officials say will be fair.However the Trump administration has carried out a series of actions that have inflamed Palestinian opinion and generally pleased Israel.In 2017 Mr Trump announced that the US recognised Jerusalem as Israel's capital, overturning decades of official US policy.In response Mr Abbas cut off relations with the US, saying the US could no longer be a peace broker.Last year the US stopped contributing to the UN Relief and Works Agency (Unrwa), which has been looking after Palestinian refugees since 1949.Last month President Trump officially recognised Israeli sovereignty over the occupied Golan Heights.Peace negotiations between Israel and the Palestinians have been at a standstill since 2014, when a US-brokered attempt to reach a deal collapsed.\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(Category political,\n",
       " tensor(6),\n",
       " tensor([1.5538e-02, 9.6381e-03, 5.2022e-03, 2.9879e-03, 6.4456e-04, 1.0724e-04,\n",
       "         9.6250e-01, 2.8808e-04, 7.2169e-05, 2.9321e-04, 4.1399e-04, 2.3121e-03]))"
      ]
     },
     "execution_count": 12,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# Fox News:\n",
    "learn.predict(\"Former President Barack Obama said on Saturday that he is worried that progressives are creating a “circular firing squad” as prospective Democratic presidential candidates race to the left on a number of hot topic issues ahead of the 2020 election.“The way we structure democracy requires you to take into account people who don’t agree with you,” he said at an Obama Foundation town hall event in Berlin, according to The New York Post. “And that by definition means you’re not going to get 100 percent of what you want.”BARACK OBAMA STILL BELIEVES BIDEN WOULD BE 'AN EXCELLENT PRESIDENT' AMID INAPPROPRIATE TOUCHING ALLEGATIONS: REPORT“One of the things I do worry about sometimes among progressives … we start sometimes creating what’s called a ‘circular firing squad’ where you start shooting at your allies because one of them has strayed from purity on the issues,” he said.Obama’s remarks come as freshman House Democrats such as Rep. Alexandria Ocasio-Cortez, D-N.Y., have pushed once-fringe positions on Medicare-for-all, the Green New Deal and reparations for slavery. In turn, 2020 presidential hopefuls have also taken some of those positions.In that climate, candidates have come under criticism for their past stances from activists. South Bend Mayor Pete Buttigieg was forced this week to address remarks he made in 2015 when he said that “all lives matter” -- which some activists say is a counterslogan to the “black lives matter” sloganSen. Kamala Harris, D-Calif., meanwhile has been hit by controversy over her past as a prosecutor. A scathing op-ed published in January in The New York Times, written by law professor Lara Bazelon, has kickstarted renewed scrutiny.Obama reportedly warns freshmen House Democrats about pricey policy proposalsVideoBazelon says Harris previously 'fought tooth and nail to uphold wrongful convictions that had been secured through official misconduct that included evidence tampering, false testimony and the suppression of crucial information by prosecutors.'Bazelon further suggested that Harris should 'apologize to the wrongfully convicted people she has fought to keep in prison and to do what she can to make sure they get justice' or otherwise make clear she has 'radically broken from her past.'Former vice president under Obama, Joe Biden, meanwhile has faced criticism for inappropriate past physical contact with women, as well a a 1993 speech on crime in which he warned of “predators on our streets”'They are beyond the pale many of those people, beyond the pale,' Biden continued. 'And it's a sad commentary on society. We have no choice but to take them out of society.'The latter was reminiscent of heat 2016 presidential nominee Hillary Clinton took from activists for her description of some gang members as “superpredators” in 1996.Obama himself may not escape criticism in the election cycle. His signature health care legislation, the Affordable Care Act, is quickly being eclipsed by calls from Democrats for single-payer and Medicare-for-all plans. Meanwhile, a number of Democrats have said they are open to reparations for black Americans for slavery -- something that Obama opposed when he was in office.\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(Category political,\n",
       " tensor(6),\n",
       " tensor([6.4082e-02, 1.2220e-02, 9.7138e-03, 5.5418e-03, 3.8407e-02, 8.5766e-05,\n",
       "         8.6441e-01, 1.7726e-03, 2.9470e-04, 2.8449e-04, 3.3882e-04, 2.8459e-03]))"
      ]
     },
     "execution_count": 9,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# BrightBert again\n",
    "learn.predict(\"The border agencies need tougher leadership, President Donald Trump declared Friday as he dropped plans to appoint a long-time agency staffer to run the Immigration and Customs Enforcement agency (ICE).'Ron [Vitiello is] a good man,' Trump told reporters. 'But we’re going in a tougher direction. We want to go in a tougher direction.' Trump’s 'tougher direction' statement suggests he may pressure Department of Homeland Secretary (DHS) Secretary Kirstjen Nielsen to implement policies that top agency staffers oppose, such as rejecting legal interpretations and bureaucratic practices set by former President Barack Obama. Immigration reformers blame those Obama policies for encouraging the wave of economic migrants from Central America.Breitbart TVDonald Trump Says Everything Jared Kushner Touched ‘Turned To Gold’The shift comes amid the growing wave of Central American economic migrants who are using Obama-era legal loopholes to walk through the border wall and into jobs, neighborhoods, and blue-collar schools throughout the United States. That wave is expected to deliver one million migrants into the United States by October, and it is made possible because Democrats are blocking any reform the border loopholes.Immigration reformers fear that Obama-appointed staffers and former business lobbyists are keeping Trump in the dark about ways to improve operation at the DHS. 'I don’t now know if the President is getting the information he needs about what powers he has,' according to Rosemary Jenks, policy director at the Center for Immigration Studies. 'Secretary Nielsen and some of the attorneys in DHS are blocking the information because they are afraid of implementing some of the things they can do,' partly because they are afraid of lawsuits, she said.For example, many so-called 'Unaccompanied Alien Children' are being smuggled up the border because Trump’s agencies will pass them to their illegal immigrant parents living throughout the United States, under policies set by Obama. But those youths and children should be sent home, said Jenks, because the 2008 law only protects trafficked victims, such as forced prostitutes, not youths and children who have parents in the United States or who are willingly smuggled up to the border. According to the Washington Post, Vitiello’s exit was prompted by Steve Miller, one of Trump’s first aides who earlier played a key role in derailing the 2013 'Gang of Eight' amnesty and cheap labor bill. The Post said:Six administration officials said Friday that the decision to jettison Vitiello was a sign of the expanding influence that Miller now wields over immigration matters in the White House, particularly as Trump lashes out at Mexico and Central American nations — as well as Homeland Security officials and aides who express doubts about the legality of his ideas.The New York Times reported:One person familiar with the president’s thinking said that Mr. Trump believed that Mr. Vitiello did not favor closing the border, as the president had proposed before backing off that threat this week.Another person said that Stephen Miller, the president’s chief policy adviser and a supporter of curtailing legal and illegal immigration, did not support Mr. Vitiello’s nomination.Vitiello’s defenders lashed out at Miller. The Washington Post highlighted the complaints:'Ron Vitiello has spent as much time defending our nation’s borders as Stephen Miller has been alive,' one official said of Miller, who is 33.One senior official said: 'This is part of an increasingly desperate effort by Stephen to throw people under the bus when the policies he has advocated are not effective. Once it becomes clear that Stephen’s policies aren’t working, he tells the president, ‘They’re not the right people.’' But Vitiello’s appointment was opposed by the ICE officers’ union, the National ICE Council. Vitiello 'lacks the judgment and professionalism to effectively lead a federal agency,' said a February letter from union President Chris Crane.\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(Category political,\n",
       " tensor(6),\n",
       " tensor([0.3018, 0.0066, 0.0912, 0.0047, 0.0656, 0.0024, 0.4111, 0.0111, 0.0186,\n",
       "         0.0008, 0.0560, 0.0301]))"
      ]
     },
     "execution_count": 10,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# BBC\n",
    "learn.predict(\"Israeli PM Benjamin Netanyahu has said he will annex Jewish settlements in the occupied West Bank if he is re-elected.Israelis go to the polls on Tuesday and Mr Netanyahu is competing for votes with right-wing parties who support annexing part of the West Bank.The settlements are illegal under international law, though Israel disputes this.Last month the US recognised the occupied Golan Heights, seized from Syria in 1967, as Israeli territory.Can Jewish settlement issue be resolved?What Trump’s Golan Heights move really meansIsrael's Benjamin Netanyahu: Commando turned PMIsrael has settled about 400,000 Jews in West Bank settlements, with another 200,000 living in East Jerusalem. There are about 2.5 million Palestinians living in the West Bank.Palestinians want to establish a state in the occupied West Bank, East Jerusalem and the Gaza Strip.What happens to the settlements is one of the most contentious issues between Israel and the Palestinians - Palestinians say the presence of settlements make a future independent state impossible.Israel says the Palestinians are using the issue of settlements as a pretext to avoid direct peace talks. It says settlements are not a genuine obstacle to peace and are negotiable.What exactly did Netanyahu say?He was asked during an interview on Israeli TV why he had not extended Israeli sovereignty to large settlements in the West Bank.'You are asking whether we are moving on to the next stage - the answer is yes, we will move to the next stage,' he said.Image copyrightREUTERSImage captionMr Netanyahu is seeking re-election'I am going to extend [Israeli] sovereignty and I don't distinguish between settlement blocs and the isolated settlements.'A spokesman for Palestinian leader Mahmoud Abbas told Reuters: 'Any measures and any announcements will not change the facts. Settlements are illegal and they will be removed.'Potentially explosive commentsBy Sebastian Usher, BBC Arab affairs editorThese comments by Benjamin Netanyahu are potentially explosive over an issue that has helped stall peace efforts for years.They will resonate with several parties with which he'll try to form a coalition government if he wins the biggest share of votes.But the very idea of annexation will rouse new Palestinian fury, as well as international condemnation.Mr Netanyahu may have been emboldened by the Trump administration, which just last month recognised Israeli sovereignty over the Golan Heights.What is the political background?Mr Netanyahu's right-wing Likud party is in a tight race with the new centre-right Blue and White alliance.However other parties, some of which support annexation, could end up being kingmakers when they try to form a governing coalition.Israel election: Who are the key candidates?The ex-military chief trying to unseat NetanyahuIn Mr Netanyahu's own Likud party, 28 out of the 29 lawmakers running for re-election are on record as supporting this approach. Until now the prime minister was the only exception.What is the situation of peace negotiations?Mr Trump's administration is preparing to unveil a long-awaited Middle East peace plan, which US officials say will be fair.However the Trump administration has carried out a series of actions that have inflamed Palestinian opinion and generally pleased Israel.In 2017 Mr Trump announced that the US recognised Jerusalem as Israel's capital, overturning decades of official US policy.In response Mr Abbas cut off relations with the US, saying the US could no longer be a peace broker.Last year the US stopped contributing to the UN Relief and Works Agency (Unrwa), which has been looking after Palestinian refugees since 1949.Last month President Trump officially recognised Israeli sovereignty over the occupied Golan Heights.Peace negotiations between Israel and the Palestinians have been at a standstill since 2014, when a US-brokered attempt to reach a deal collapsed.\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(Category political,\n",
       " tensor(6),\n",
       " tensor([1.3706e-02, 4.1941e-03, 5.7440e-03, 2.0140e-03, 9.6228e-04, 7.7108e-05,\n",
       "         9.6946e-01, 3.6682e-04, 5.0189e-05, 4.2015e-04, 2.3171e-04, 2.7749e-03]))"
      ]
     },
     "execution_count": 11,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# Fox News:\n",
    "learn.predict(\"Former President Barack Obama said on Saturday that he is worried that progressives are creating a “circular firing squad” as prospective Democratic presidential candidates race to the left on a number of hot topic issues ahead of the 2020 election.“The way we structure democracy requires you to take into account people who don’t agree with you,” he said at an Obama Foundation town hall event in Berlin, according to The New York Post. “And that by definition means you’re not going to get 100 percent of what you want.”BARACK OBAMA STILL BELIEVES BIDEN WOULD BE 'AN EXCELLENT PRESIDENT' AMID INAPPROPRIATE TOUCHING ALLEGATIONS: REPORT“One of the things I do worry about sometimes among progressives … we start sometimes creating what’s called a ‘circular firing squad’ where you start shooting at your allies because one of them has strayed from purity on the issues,” he said.Obama’s remarks come as freshman House Democrats such as Rep. Alexandria Ocasio-Cortez, D-N.Y., have pushed once-fringe positions on Medicare-for-all, the Green New Deal and reparations for slavery. In turn, 2020 presidential hopefuls have also taken some of those positions.In that climate, candidates have come under criticism for their past stances from activists. South Bend Mayor Pete Buttigieg was forced this week to address remarks he made in 2015 when he said that “all lives matter” -- which some activists say is a counterslogan to the “black lives matter” sloganSen. Kamala Harris, D-Calif., meanwhile has been hit by controversy over her past as a prosecutor. A scathing op-ed published in January in The New York Times, written by law professor Lara Bazelon, has kickstarted renewed scrutiny.Obama reportedly warns freshmen House Democrats about pricey policy proposalsVideoBazelon says Harris previously 'fought tooth and nail to uphold wrongful convictions that had been secured through official misconduct that included evidence tampering, false testimony and the suppression of crucial information by prosecutors.'Bazelon further suggested that Harris should 'apologize to the wrongfully convicted people she has fought to keep in prison and to do what she can to make sure they get justice' or otherwise make clear she has 'radically broken from her past.'Former vice president under Obama, Joe Biden, meanwhile has faced criticism for inappropriate past physical contact with women, as well a a 1993 speech on crime in which he warned of “predators on our streets”'They are beyond the pale many of those people, beyond the pale,' Biden continued. 'And it's a sad commentary on society. We have no choice but to take them out of society.'The latter was reminiscent of heat 2016 presidential nominee Hillary Clinton took from activists for her description of some gang members as “superpredators” in 1996.Obama himself may not escape criticism in the election cycle. His signature health care legislation, the Affordable Care Act, is quickly being eclipsed by calls from Democrats for single-payer and Medicare-for-all plans. Meanwhile, a number of Democrats have said they are open to reparations for black Americans for slavery -- something that Obama opposed when he was in office.\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 28,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(Category conspiracy,\n",
       " tensor(2),\n",
       " tensor([1.8053e-02, 7.4355e-03, 4.4468e-01, 1.1615e-01, 1.0645e-03, 3.1251e-02,\n",
       "         3.5184e-01, 3.7448e-04, 2.7742e-04, 3.8187e-03, 7.7076e-03, 1.7347e-02]))"
      ]
     },
     "execution_count": 28,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "#Pseudoscience\n",
    "learn.predict(\"Have you ever clicked on a link like 'What does your favorite animal say about you?' wondering what your love of hedgehogs reveals about your psyche? Or filled out a personality assessment to gain new understanding into whether you’re an introverted or extroverted 'type'? People love turning to these kinds of personality quizzes and tests on the hunt for deep insights into themselves. People tend to believe they have a 'true' and revealing self hidden somewhere deep within, so it’s natural that assessments claiming to unveil it will be appealing.As psychologists, we noticed something striking about assessments that claim to uncover people’s 'true type.' Many of the questions are poorly constructed – their wording can be ambiguous and they often contain forced choices between options that are not opposites. This can be true of BuzzFeed-type quizzes as well as more seemingly sober assessments.On the other hand, assessments created by trained personality psychologists use questions that are more straightforward to interpret. The most notable example is probably the well-respected Big Five Inventory. Rather than sorting people into 'types,' it scores people on the established psychological dimensions of openness to new experience, conscientiousness, extroversion, agreeableness and neuroticism. This simplicity is by design; psychology researchers know that the more respondents struggle to understand the question, the worse the question is.But the lack of rigor in 'type' assessments turns out to be a feature, not a bug, for the general public. What makes tests less valid can ironically make them more interesting. Since most people aren’t trained to think about psychology in a scientifically rigorous way, it stands to reason they also won’t be great at evaluating those assessments. We recently conducted series of studies to investigate how consumers view these tests. When people try to answer these harder questions, do they think to themselves 'This question is poorly written'? Or instead do they focus on its difficulty and think 'This question’s deep'? Our results suggest that a desire for deep insight can lead to deep confusion.Confusing difficult for deepIn our first study, we showed people items from both the Big Five and from the Keirsey Temperament Sorter (KTS), a popular 'type' assessment that contains many questions we suspected people find comparatively difficult. Our participants rated each item in two ways. First, they rated difficulty. That is, how confusing and ambiguous did they find it? Second, what was its perceived 'depth'? In other words, to what extent did they feel the item seemed to be getting at something hidden deep in the unconscious?Sure enough, not only were these perceptions correlated, the KTS was seen as both more difficult and deeper. In follow-up studies, we experimentally manipulated difficulty. In one study, we modified Big Five items to make them harder to answer like the KTS items, and again we found that participants rated the more difficult versions as 'deeper.'We also noticed that some personality assessments seem to derive their intrigue from having seemingly nothing to do with personality at all. Take one BuzzFeed quiz, for example, that asks about which colors people associate with abstract concepts like letters and days of the week and then outputs 'the true age of your soul.' Even if people trust BuzzFeed more for entertainment than psychological truths, perhaps they are actually on board with the idea that these difficult, abstract decisions do reveal some deep insights. In fact, that is the entire idea behind classically problematic measures such as the Rorschach, or 'ink blot,' test.In two studies inspired by that BuzzFeed quiz, we found exactly that. We gave people items from purported 'personality assessment' checklists. In one study, we assigned half the participants to the 'difficult' condition, wherein the assessment items required them to choose which of two colors they associated with abstract concepts, like the letter 'M.' In the 'easier' condition, respondents were still required to rate colors on how much they associated them with those abstract concepts, but they more simply rated one color at a time instead of choosing between two.Again, participants rated the difficult version as deeper. Seemingly, the sillier the assessment, the better people think it can read the hidden self.Intuition may steer you wrongOne of the implications of this research is that people are going to have a hard time leaving behind the bad ideas baked into popular yet unscientific personality assessments. The most notable example is the Myers-Briggs Type Indicator, which infamously remains quite popular while doing a fairly poor job of assessing personality, due to longstanding issues with the assessment itself and the long-discredited Jungian theory behind it. Our findings suggest that Myers-Briggs-like assessments that have largely been debunked by experts might persist in part because their formats overlap quite well with people’s intuitions about what will best access the “true self.”People’s intuitions do them no favors here. Intuitions often undermine scientific thinking on topics like physics and biology. Psychology is no different. People arbitrarily divide parts of themselves into “true” and superficial components and seem all too willing to believe in tests that claim to definitively make those distinctions. But the idea of a “true self” doesn’t really work as a scientific concept.Some people might be stuck in a self-reinforcing yet unproductive line of thought: Personality assessments can cause confusion. That confusion in turn overlaps with intuitions of how they think their deep psychology works, and then they tell themselves the confusion is profound. So intuitions about psychology might be especially pernicious. Following them too closely could lead you to know less about yourself, not more.\", thresh=0.5)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 24,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(Category political,\n",
       " tensor(6),\n",
       " tensor([3.9681e-02, 1.4495e-02, 2.1821e-02, 2.3866e-02, 6.0934e-03, 1.4194e-04,\n",
       "         8.7859e-01, 6.4829e-04, 3.6831e-03, 2.9676e-03, 3.7980e-04, 7.6305e-03]))"
      ]
     },
     "execution_count": 24,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "learn.predict(\"PETALUMA, CA — An incident in which a white man was reportedly beaten in downtown Petaluma by a group of suspects the victim described as four or five black men is being investigated as a hate crime and an assault, the Petaluma Police Department said Tuesday in a news release.Petaluma police Lt. Ed Crosby said officers immediately responded at 9:03 p.m. Saturday, March 9 to the intersection of Mary Street at Petaluma Boulevard North to a woman's report that her domestic partner, a 60-year-old white man, had just been attacked.The lieutenant said when officers arrived they found the victim on the ground suffering from numerous facial injuries.The man was rushed to Santa Rosa Memorial Hospital where according to police, he stayed two days. Injuries to the victim were confirmed as a fractured left eye socket, a broken nose and other abrasions to his face including facial swelling, Crosby said.The couple told police that the night of the incident they had just finished eating dinner at a restaurant on Petaluma Boulevard North and were walking westbound toward their car, which was parked on Mary Street, when they passed a group of several African-American men who looked to be in their 20s, standing around a four-door, emerald green Honda Civic.The couple said they did not interact with the group and were continuing on their way when one of the men by the green Honda 'hurled profanity at the victim and referred to his [the victim's] race,' Crosby said.'The victim turned around and saw one of the males rushing at him, swinging his arms,' Crosby said.'The victim grabbed the advancing male, brought him to the ground, and pinned him,' Crosby said. 'In response, the other males by the green Honda repeatedly kicked the victim in the face before getting into the green Honda and fleeing the scene.'Petaluma police are asking anyone with information about the incident to contact or leave a message for Petaluma Police Department Officer Ron Flores by calling 707-778-4372.The victim and his female companion were not able to give many descriptive details about the suspects, the lieutenant said, and thus far, officers' efforts in canvassing the downtown area for any witnesses or video footage that would help identify the suspects have not been successful.The green Honda was missing a front license plate; the back license plate may possibly include the numbers 611, according to police.\", thresh=.5)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 29,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "['bias',\n",
       " 'clickbait',\n",
       " 'conspiracy',\n",
       " 'fake',\n",
       " 'hate',\n",
       " 'junksci',\n",
       " 'political',\n",
       " 'reliable',\n",
       " 'rumor',\n",
       " 'satire',\n",
       " 'unknown',\n",
       " 'unreliable']"
      ]
     },
     "execution_count": 29,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "learn.data.classes"
   ]
  },
  {
   "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": "3.6.7"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}