03_load_and_predict.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "#load the model"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Model: \"model\"\n",
      "__________________________________________________________________________________________________\n",
      "Layer (type)                    Output Shape         Param #     Connected to                     \n",
      "==================================================================================================\n",
      "input_ids (InputLayer)          [(None, 512)]        0                                            \n",
      "__________________________________________________________________________________________________\n",
      "attention_mask (InputLayer)     [(None, 512)]        0                                            \n",
      "__________________________________________________________________________________________________\n",
      "bert (Custom>TFBertMainLayer)   {'last_hidden_state' 108310272   input_ids[0][0]                  \n",
      "                                                                 attention_mask[0][0]             \n",
      "__________________________________________________________________________________________________\n",
      "dropout_37 (Dropout)            (None, 512, 768)     0           bert[0][0]                       \n",
      "__________________________________________________________________________________________________\n",
      "global_max_pooling1d (GlobalMax (None, 768)          0           dropout_37[0][0]                 \n",
      "__________________________________________________________________________________________________\n",
      "outputs (Dense)                 (None, 5)            3845        global_max_pooling1d[0][0]       \n",
      "==================================================================================================\n",
      "Total params: 108,314,117\n",
      "Trainable params: 108,314,117\n",
      "Non-trainable params: 0\n",
      "__________________________________________________________________________________________________\n"
     ]
    }
   ],
   "source": [
    "import tensorflow as tf\n",
    "\n",
    "model = tf.keras.models.load_model('sentiment_model')\n",
    "\n",
    "# view model architecture to confirm it has been save and loaded correctly\n",
    "model.summary()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "#Tokenizing the data using the bert-base-cased tokenizer.\n",
    "\n",
    "#Transforming the data into a dictionary containing 'input_ids' and 'attention_mask' tensors."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [],
   "source": [
    "from transformers import BertTokenizer\n",
    "\n",
    "tokenizer = BertTokenizer.from_pretrained('bert-base-cased')\n",
    "\n",
    "def prep_data(text):\n",
    "    tokens = tokenizer.encode_plus(text, max_length=512,\n",
    "                                   truncation=True, padding='max_length',\n",
    "                                   add_special_tokens=True, return_token_type_ids=False,\n",
    "                                   return_tensors='tf')\n",
    "    # tokenizer returns int32 tensors, we need to return float64, so we use tf.cast\n",
    "    return {'input_ids': tf.cast(tokens['input_ids'], tf.float64),\n",
    "            'attention_mask': tf.cast(tokens['attention_mask'], tf.float64)}"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "array([0.0207338 , 0.05568625, 0.4257313 , 0.31074345, 0.18710516],\n",
       "      dtype=float32)"
      ]
     },
     "execution_count": 9,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "probs = model.predict(prep_data(\"hello world\"))[0]\n",
    "\n",
    "probs"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "2"
      ]
     },
     "execution_count": 11,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "import numpy as np\n",
    "\n",
    "np.argmax(probs)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "#apply the model to real phrases from test.tsv"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "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>PhraseId</th>\n",
       "      <th>SentenceId</th>\n",
       "      <th>Phrase</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>156061</td>\n",
       "      <td>8545</td>\n",
       "      <td>An intermittently pleasing but mostly routine effort .</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>156062</td>\n",
       "      <td>8545</td>\n",
       "      <td>An intermittently pleasing but mostly routine effort</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>156063</td>\n",
       "      <td>8545</td>\n",
       "      <td>An</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>156064</td>\n",
       "      <td>8545</td>\n",
       "      <td>intermittently pleasing but mostly routine effort</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>156065</td>\n",
       "      <td>8545</td>\n",
       "      <td>intermittently pleasing but mostly routine</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "   PhraseId  SentenceId  \\\n",
       "0    156061        8545   \n",
       "1    156062        8545   \n",
       "2    156063        8545   \n",
       "3    156064        8545   \n",
       "4    156065        8545   \n",
       "\n",
       "                                                   Phrase  \n",
       "0  An intermittently pleasing but mostly routine effort .  \n",
       "1    An intermittently pleasing but mostly routine effort  \n",
       "2                                                      An  \n",
       "3       intermittently pleasing but mostly routine effort  \n",
       "4              intermittently pleasing but mostly routine  "
      ]
     },
     "execution_count": 14,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "import pandas as pd\n",
    "\n",
    "pd.set_option('display.max_colwidth', None)\n",
    "\n",
    "df = pd.read_csv('test.tsv', sep='\\t')\n",
    "df.head()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "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>PhraseId</th>\n",
       "      <th>SentenceId</th>\n",
       "      <th>Phrase</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>156061</td>\n",
       "      <td>8545</td>\n",
       "      <td>An intermittently pleasing but mostly routine effort .</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>15</th>\n",
       "      <td>156076</td>\n",
       "      <td>8546</td>\n",
       "      <td>Kidman is really the only thing that 's worth watching in Birthday Girl , a film by the stage-trained Jez Butterworth -LRB- Mojo -RRB- that serves as yet another example of the sad decline of British comedies in the post-Full Monty world .</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>93</th>\n",
       "      <td>156154</td>\n",
       "      <td>8547</td>\n",
       "      <td>Once you get into its rhythm ... the movie becomes a heady experience .</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>117</th>\n",
       "      <td>156178</td>\n",
       "      <td>8548</td>\n",
       "      <td>I kept wishing I was watching a documentary about the wartime Navajos and what they accomplished instead of all this specious Hollywood hoo-ha .</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>158</th>\n",
       "      <td>156219</td>\n",
       "      <td>8549</td>\n",
       "      <td>Kinnear does n't aim for our sympathy , but rather delivers a performance of striking skill and depth .</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "     PhraseId  SentenceId  \\\n",
       "0      156061        8545   \n",
       "15     156076        8546   \n",
       "93     156154        8547   \n",
       "117    156178        8548   \n",
       "158    156219        8549   \n",
       "\n",
       "                                                                                                                                                                                                                                              Phrase  \n",
       "0                                                                                                                                                                                             An intermittently pleasing but mostly routine effort .  \n",
       "15   Kidman is really the only thing that 's worth watching in Birthday Girl , a film by the stage-trained Jez Butterworth -LRB- Mojo -RRB- that serves as yet another example of the sad decline of British comedies in the post-Full Monty world .  \n",
       "93                                                                                                                                                                           Once you get into its rhythm ... the movie becomes a heady experience .  \n",
       "117                                                                                                 I kept wishing I was watching a documentary about the wartime Navajos and what they accomplished instead of all this specious Hollywood hoo-ha .  \n",
       "158                                                                                                                                          Kinnear does n't aim for our sympathy , but rather delivers a performance of striking skill and depth .  "
      ]
     },
     "execution_count": 15,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "#remove fragment duplicates based on SentenceId\n",
    "df = df.drop_duplicates(subset=['SentenceId'], keep='first')\n",
    "df.head()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Now we initialize our new sentiment column, and begin making predictions."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "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>PhraseId</th>\n",
       "      <th>SentenceId</th>\n",
       "      <th>Phrase</th>\n",
       "      <th>Sentiment</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>156061</td>\n",
       "      <td>8545</td>\n",
       "      <td>An intermittently pleasing but mostly routine ...</td>\n",
       "      <td>3</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>15</th>\n",
       "      <td>156076</td>\n",
       "      <td>8546</td>\n",
       "      <td>Kidman is really the only thing that 's worth ...</td>\n",
       "      <td>3</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>93</th>\n",
       "      <td>156154</td>\n",
       "      <td>8547</td>\n",
       "      <td>Once you get into its rhythm ... the movie bec...</td>\n",
       "      <td>3</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>117</th>\n",
       "      <td>156178</td>\n",
       "      <td>8548</td>\n",
       "      <td>I kept wishing I was watching a documentary ab...</td>\n",
       "      <td>3</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>158</th>\n",
       "      <td>156219</td>\n",
       "      <td>8549</td>\n",
       "      <td>Kinnear does n't aim for our sympathy , but ra...</td>\n",
       "      <td>3</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "     PhraseId  SentenceId                                             Phrase  \\\n",
       "0      156061        8545  An intermittently pleasing but mostly routine ...   \n",
       "15     156076        8546  Kidman is really the only thing that 's worth ...   \n",
       "93     156154        8547  Once you get into its rhythm ... the movie bec...   \n",
       "117    156178        8548  I kept wishing I was watching a documentary ab...   \n",
       "158    156219        8549  Kinnear does n't aim for our sympathy , but ra...   \n",
       "\n",
       "    Sentiment  \n",
       "0           3  \n",
       "15          3  \n",
       "93          3  \n",
       "117         3  \n",
       "158         3  "
      ]
     },
     "execution_count": 12,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "df['Sentiment'] = None\n",
    "\n",
    "for i, row in df.iterrows():\n",
    "    # get token tensors\n",
    "    tokens = prep_data(row['Phrase'])\n",
    "    # get probabilities\n",
    "    probs = model.predict(tokens)\n",
    "    # find argmax for winning class\n",
    "    pred = np.argmax(probs)\n",
    "    # add to dataframe\n",
    "    df.at[i, 'Sentiment'] = pred\n",
    "\n",
    "df.head()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "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>PhraseId</th>\n",
       "      <th>SentenceId</th>\n",
       "      <th>Phrase</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>66220</th>\n",
       "      <td>222281</td>\n",
       "      <td>11851</td>\n",
       "      <td>Not sweet enough to liven up its predictable story and will leave even fans of hip-hop sorely disappointed .</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>66239</th>\n",
       "      <td>222300</td>\n",
       "      <td>11852</td>\n",
       "      <td>Nasty , ugly , pointless and depressing , even if you hate clowns .</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>66253</th>\n",
       "      <td>222314</td>\n",
       "      <td>11853</td>\n",
       "      <td>With tightly organized efficiency , numerous flashbacks and a constant edge of tension , Miller 's film is one of 2002 's involvingly adult surprises .</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>66280</th>\n",
       "      <td>222341</td>\n",
       "      <td>11854</td>\n",
       "      <td>They should have called it Gutterball .</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>66287</th>\n",
       "      <td>222348</td>\n",
       "      <td>11855</td>\n",
       "      <td>A long-winded , predictable scenario .</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "       PhraseId  SentenceId  \\\n",
       "66220    222281       11851   \n",
       "66239    222300       11852   \n",
       "66253    222314       11853   \n",
       "66280    222341       11854   \n",
       "66287    222348       11855   \n",
       "\n",
       "                                                                                                                                                        Phrase  \n",
       "66220                                             Not sweet enough to liven up its predictable story and will leave even fans of hip-hop sorely disappointed .  \n",
       "66239                                                                                      Nasty , ugly , pointless and depressing , even if you hate clowns .  \n",
       "66253  With tightly organized efficiency , numerous flashbacks and a constant edge of tension , Miller 's film is one of 2002 's involvingly adult surprises .  \n",
       "66280                                                                                                                  They should have called it Gutterball .  \n",
       "66287                                                                                                                   A long-winded , predictable scenario .  "
      ]
     },
     "execution_count": 16,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "df.tail()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "ML",
   "language": "python",
   "name": "ml"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.8.5"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 4
}