transformer.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [],
   "source": [
    "# importing necessary libraries\n",
    "\n",
    "import torch\n",
    "import torch.nn as nn\n",
    "import torch.optim as optim\n",
    "import torch.utils.data as data\n",
    "import math\n",
    "import copy\n",
    "from tqdm import tqdm\n",
    "import pandas as pd\n",
    "import numpy as np\n",
    "import re\n",
    "import tensorflow as tf\n",
    "import seaborn as sns\n",
    "import matplotlib.pyplot as plt\n",
    "from tensorflow.keras.preprocessing.text import Tokenizer\n",
    "from tensorflow.keras.preprocessing.sequence import pad_sequences\n",
    "import warnings\n",
    "warnings.filterwarnings(\"ignore\")\n",
    "# import fasttext\n",
    "# import string\n",
    "# from string import digits\n",
    "import os\n",
    "os.environ[\"CUDA_VISIBLE_DEVICES\"] = \"0\""
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Data Preperation :"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [],
   "source": [
    "# dataset consist of five rows\n",
    "# each row consist of english sentence along with its french translation\n",
    "\n",
    "train_data_corpus = [('where did you put your key ?', ' où est-ce tu as mis ta clé  ? '),\n",
    "('you missed a spot .', 'tu as loupé une tache . '),\n",
    "(\"i think we're being followed .\", ' je pense que nous sommes suivis . '),\n",
    "('i bought a cactus .', \" j'ai acheté un cactus .\"), \n",
    "('i have more than enough .', \" j'en ai plus que marre . \") ]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [],
   "source": [
    "# create a dataframe using list of sentences\n",
    "\n",
    "train_data = pd.DataFrame(train_data_corpus,columns=['eng_sent','french_sent'])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "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>eng_sent</th>\n",
       "      <th>french_sent</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>where did you put your key ?</td>\n",
       "      <td>où est-ce tu as mis ta clé  ?</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>you missed a spot .</td>\n",
       "      <td>tu as loupé une tache .</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>i think we're being followed .</td>\n",
       "      <td>je pense que nous sommes suivis .</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>i bought a cactus .</td>\n",
       "      <td>j'ai acheté un cactus .</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>i have more than enough .</td>\n",
       "      <td>j'en ai plus que marre .</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "                         eng_sent                          french_sent\n",
       "0    where did you put your key ?       où est-ce tu as mis ta clé  ? \n",
       "1             you missed a spot .             tu as loupé une tache . \n",
       "2  i think we're being followed .   je pense que nous sommes suivis . \n",
       "3             i bought a cactus .              j'ai acheté un cactus .\n",
       "4       i have more than enough .            j'en ai plus que marre . "
      ]
     },
     "execution_count": 5,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# glimpse of dataset\n",
    "\n",
    "train_data"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Data cleaning:"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [],
   "source": [
    "# # data cleaning\n",
    "\n",
    "# # Lowercase all characters\n",
    "# train_data['eng_sent']=train_data['eng_sent'].apply(lambda x: x.lower())\n",
    "\n",
    "# train_data['eng_sent'] = train_data['eng_sent'].apply(lambda x: re.sub(r\"won\\'t\", \"will not\", x))\n",
    "# train_data['eng_sent'] = train_data['eng_sent'].apply(lambda x: re.sub(r\"can\\'t\", \"can not\", x))\n",
    "\n",
    "# # general\n",
    "# train_data['eng_sent'] = train_data['eng_sent'].apply(lambda x: re.sub(r\"n\\'t\", \" not\", x))\n",
    "# train_data['eng_sent'] = train_data['eng_sent'].apply(lambda x: re.sub(r\"\\'re\", \" are\", x))\n",
    "# train_data['eng_sent'] = train_data['eng_sent'].apply(lambda x: re.sub(r\"\\'s\", \" is\", x))\n",
    "# train_data['eng_sent'] = train_data['eng_sent'].apply(lambda x: re.sub(r\"\\'d\", \" would\", x))\n",
    "# train_data['eng_sent'] = train_data['eng_sent'].apply(lambda x: re.sub(r\"\\'ll\", \" will\", x))\n",
    "# train_data['eng_sent'] = train_data['eng_sent'].apply(lambda x: re.sub(r\"\\'t\", \" not\", x))\n",
    "# train_data['eng_sent'] = train_data['eng_sent'].apply(lambda x: re.sub(r\"\\'ve\", \" have\", x))\n",
    "# train_data['eng_sent'] = train_data['eng_sent'].apply(lambda x: re.sub(r\"\\'m\", \" am\", x))\n",
    "\n",
    "# # Remove quotes\n",
    "# train_data['eng_sent']=train_data['eng_sent'].apply(lambda x: re.sub(\"'\", '', x))\n",
    "\n",
    "train_data['eng_sent'] = train_data['eng_sent'].apply(lambda x: re.sub('[!.?]',\"\",x))\n",
    "train_data['french_sent'] = train_data['french_sent'].apply(lambda x: re.sub('[!.?]',\"\",x))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {},
   "outputs": [],
   "source": [
    "# we specify start of sentence with '<start>'\n",
    "# we specify end of sentence with '<end>'\n",
    "\n",
    "train_data['french_sent'] = '<start> ' + train_data.french_sent \n",
    "train_data['french_sent'] = train_data.french_sent + ' <end>' "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>eng_sent</th>\n",
       "      <th>french_sent</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>where did you put your key</td>\n",
       "      <td>&lt;start&gt;  où est-ce tu as mis ta clé    &lt;end&gt;</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>you missed a spot</td>\n",
       "      <td>&lt;start&gt; tu as loupé une tache   &lt;end&gt;</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>i think we're being followed</td>\n",
       "      <td>&lt;start&gt;  je pense que nous sommes suivis   &lt;end&gt;</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>i bought a cactus</td>\n",
       "      <td>&lt;start&gt;  j'ai acheté un cactus  &lt;end&gt;</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>i have more than enough</td>\n",
       "      <td>&lt;start&gt;  j'en ai plus que marre   &lt;end&gt;</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "                        eng_sent  \\\n",
       "0    where did you put your key    \n",
       "1             you missed a spot    \n",
       "2  i think we're being followed    \n",
       "3             i bought a cactus    \n",
       "4       i have more than enough    \n",
       "\n",
       "                                        french_sent  \n",
       "0      <start>  où est-ce tu as mis ta clé    <end>  \n",
       "1             <start> tu as loupé une tache   <end>  \n",
       "2  <start>  je pense que nous sommes suivis   <end>  \n",
       "3             <start>  j'ai acheté un cactus  <end>  \n",
       "4           <start>  j'en ai plus que marre   <end>  "
      ]
     },
     "execution_count": 8,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# dataset after preprocessing\n",
    "\n",
    "train_data"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Tokenization on data\n",
    "\n",
    "token_eng = Tokenizer()\n",
    "token_eng.fit_on_texts(train_data['eng_sent'].values)\n",
    "token_french = Tokenizer(filters='')\n",
    "token_french.fit_on_texts(train_data['french_sent'].values)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "{'i': 1,\n",
       " 'you': 2,\n",
       " 'a': 3,\n",
       " 'where': 4,\n",
       " 'did': 5,\n",
       " 'put': 6,\n",
       " 'your': 7,\n",
       " 'key': 8,\n",
       " 'missed': 9,\n",
       " 'spot': 10,\n",
       " 'think': 11,\n",
       " \"we're\": 12,\n",
       " 'being': 13,\n",
       " 'followed': 14,\n",
       " 'bought': 15,\n",
       " 'cactus': 16,\n",
       " 'have': 17,\n",
       " 'more': 18,\n",
       " 'than': 19,\n",
       " 'enough': 20}"
      ]
     },
     "execution_count": 10,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# have a look at words in eng vocab\n",
    "\n",
    "token_eng.word_index"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Vocab size of english is 20\n",
      "****************************************************************************************************\n",
      "Vocab size of french is 26\n"
     ]
    }
   ],
   "source": [
    "# vocab size for given dataset\n",
    "\n",
    "vocab_size_eng=len(token_eng.word_index.keys())\n",
    "print('Vocab size of english is',vocab_size_eng)\n",
    "print(\"*\"*100)\n",
    "vocab_size_french=len(token_french.word_index.keys())\n",
    "print('Vocab size of french is',vocab_size_french)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "###  Data analysis :"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "# lets check length of senetence\n",
    "\n",
    "train_data['french_len'] = train_data['french_sent'].apply(lambda x: len(x.split()))\n",
    "train_data['english_len'] = train_data['eng_sent'].apply(lambda x: len(x.split()))\n",
    "\n",
    "sns.histplot(train_data.english_len, bins=5)\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "10 Quantile is 4.0\n",
      "20 Quantile is 4.0\n",
      "30 Quantile is 4.2\n",
      "40 Quantile is 4.6\n",
      "50 Quantile is 5.0\n",
      "60 Quantile is 5.0\n",
      "70 Quantile is 5.0\n",
      "80 Quantile is 5.2\n",
      "90 Quantile is 5.6\n",
      "100 Quantile is 6.0\n"
     ]
    }
   ],
   "source": [
    "# we can identify the length of text in percentile\n",
    "for i in np.arange(0.1,1.1,0.1):\n",
    "    print('{0} Quantile is {1}'.format(int(i*100),np.quantile(train_data.english_len, i)))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {},
   "outputs": [],
   "source": [
    "# convert text to numbers\n",
    "\n",
    "max_len = 6\n",
    "\n",
    "# text to sequence\n",
    "\n",
    "train_eng = token_eng.texts_to_sequences(train_data.eng_sent)\n",
    "train_french = token_french.texts_to_sequences(train_data.french_sent)\n",
    "\n",
    "# padding\n",
    "\n",
    "\n",
    "train_eng_seq = pad_sequences(train_eng, maxlen=max_len, padding='post', dtype='int32')\n",
    "train_french_seq = pad_sequences(train_french, maxlen=max_len, padding='post', dtype='int32')\n",
    "\n",
    "#convert it to tensor \n",
    "src_data = torch.tensor(train_eng_seq).to(torch.int64)\n",
    "tgt_data = torch.tensor(train_french_seq).to(torch.int64)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Transformer model :"
   ]
  },
  {
   "attachments": {
    "attention_research_1-727x1024.webp": {
     "image/webp": "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"
    }
   },
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "![attention_research_1-727x1024.webp](attachment:attention_research_1-727x1024.webp)"
   ]
  },
  {
   "attachments": {
    "1_ETe4WrKJ1lS1MKDgBPIM0g.webp": {
     "image/webp": "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"
    }
   },
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "![1_ETe4WrKJ1lS1MKDgBPIM0g.webp](attachment:1_ETe4WrKJ1lS1MKDgBPIM0g.webp)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "metadata": {},
   "outputs": [],
   "source": [
    "# class block for positional encoding\n",
    "\n",
    "class PositionalEncoding(nn.Module):\n",
    "    def __init__(self, d_model, max_seq_length):\n",
    "        super(PositionalEncoding, self).__init__()\n",
    "        \n",
    "        pe = torch.zeros(max_seq_length, d_model)\n",
    "        position = torch.arange(0, max_seq_length, dtype=torch.float).unsqueeze(1)\n",
    "        div_term = torch.exp(torch.arange(0, d_model, 2).float() * -(math.log(10000.0) / d_model))\n",
    "        \n",
    "        pe[:, 0::2] = torch.sin(position * div_term)\n",
    "        pe[:, 1::2] = torch.cos(position * div_term)\n",
    "        \n",
    "        self.register_buffer('pe', pe.unsqueeze(0))\n",
    "        \n",
    "    def forward(self, x):\n",
    "        return x + self.pe[:, :x.size(1)]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "metadata": {},
   "outputs": [],
   "source": [
    "# class block for self attention unit\n",
    "\n",
    "class SelfAttention(nn.Module):\n",
    "    def __init__(self, d_model, num_heads):\n",
    "        super(SelfAttention, self).__init__()\n",
    "        \n",
    "        self.d_model = d_model\n",
    "        self.num_heads = num_heads\n",
    "        self.d_k = d_model // num_heads\n",
    "        \n",
    "       \n",
    "        self.W_q = nn.Linear(d_model, d_model)    # torch.nn.Linear(in_features, out_features)\n",
    "        self.W_k = nn.Linear(d_model, d_model)\n",
    "        self.W_v = nn.Linear(d_model, d_model)\n",
    "        self.W_o = nn.Linear(d_model, d_model)\n",
    "        \n",
    "    def scaled_dot_product_attention(self, Q, K, V, mask=None):\n",
    "        attn_scores = torch.matmul(Q, K.transpose(-2, -1)) / math.sqrt(self.d_k)\n",
    "        if mask is not None:\n",
    "            attn_scores = attn_scores.masked_fill(mask == 0, -1e9)\n",
    "        attn_probs = torch.softmax(attn_scores, dim=-1)\n",
    "        output = torch.matmul(attn_probs, V)\n",
    "        return output\n",
    "        \n",
    "    def split_heads(self, x):\n",
    "        batch_size, seq_length, d_model = x.size()\n",
    "        return x.view(batch_size, seq_length, self.num_heads, self.d_k).transpose(1, 2)\n",
    "        \n",
    "    def combine_heads(self, x):\n",
    "        batch_size, _, seq_length, d_k = x.size()\n",
    "        return x.transpose(1, 2).contiguous().view(batch_size, seq_length, self.d_model)\n",
    "        \n",
    "    def forward(self, Q, K, V, mask=None):\n",
    "        Q = self.split_heads(self.W_q(Q))\n",
    "        K = self.split_heads(self.W_k(K))\n",
    "        V = self.split_heads(self.W_v(V))\n",
    "        \n",
    "        attn_output = self.scaled_dot_product_attention(Q, K, V, mask)\n",
    "        output = self.W_o(self.combine_heads(attn_output))\n",
    "        return output"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "metadata": {},
   "outputs": [],
   "source": [
    "# class block for feedforward\n",
    "\n",
    "class PositionWiseFeedForward(nn.Module):\n",
    "    def __init__(self, d_model, d_ff):\n",
    "        super(PositionWiseFeedForward, self).__init__()\n",
    "        self.fc1 = nn.Linear(d_model, d_ff)\n",
    "        self.fc2 = nn.Linear(d_ff, d_model)\n",
    "        self.relu = nn.ReLU()\n",
    "\n",
    "    def forward(self, x):\n",
    "        return self.fc2(self.relu(self.fc1(x)))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 48,
   "metadata": {},
   "outputs": [],
   "source": [
    "# class block for encoder layer\n",
    "\n",
    "class EncoderLayer(nn.Module):\n",
    "    def __init__(self, d_model, num_heads, d_ff, dropout):\n",
    "        super(EncoderLayer, self).__init__()\n",
    "        self.self_attn = SelfAttention(d_model, num_heads)\n",
    "        self.feed_forward = PositionWiseFeedForward(d_model, d_ff)\n",
    "        self.norm1 = nn.LayerNorm(d_model)\n",
    "        self.norm2 = nn.LayerNorm(d_model)\n",
    "        self.dropout = nn.Dropout(dropout)\n",
    "        \n",
    "    def forward(self, x, mask):\n",
    "        attn_output = self.self_attn(x, x, x, mask)\n",
    "        x = self.norm1(x + self.dropout(attn_output))\n",
    "        ff_output = self.feed_forward(x)\n",
    "        x = self.norm2(x + self.dropout(ff_output))\n",
    "        return x"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 49,
   "metadata": {},
   "outputs": [],
   "source": [
    "# class block for decoder layer\n",
    "\n",
    "class DecoderLayer(nn.Module):\n",
    "    def __init__(self, d_model, num_heads, d_ff, dropout):\n",
    "        super(DecoderLayer, self).__init__()\n",
    "        self.self_attn = SelfAttention(d_model, num_heads)\n",
    "        self.cross_attn = SelfAttention(d_model, num_heads)\n",
    "        self.feed_forward = PositionWiseFeedForward(d_model, d_ff)\n",
    "        self.norm1 = nn.LayerNorm(d_model)\n",
    "        self.norm2 = nn.LayerNorm(d_model)\n",
    "        self.norm3 = nn.LayerNorm(d_model)\n",
    "        self.dropout = nn.Dropout(dropout)\n",
    "        \n",
    "    def forward(self, x, enc_output, src_mask, tgt_mask):\n",
    "        attn_output = self.self_attn(x, x, x)\n",
    "        x = self.norm1(x + self.dropout(attn_output))\n",
    "        attn_output = self.cross_attn(x, enc_output, enc_output)\n",
    "        x = self.norm2(x + self.dropout(attn_output))\n",
    "        ff_output = self.feed_forward(x)\n",
    "        x = self.norm3(x + self.dropout(ff_output))\n",
    "        return x"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 50,
   "metadata": {},
   "outputs": [],
   "source": [
    "# class block for transformer layer\n",
    "\n",
    "class Transformer(nn.Module):\n",
    "    def __init__(self, src_vocab_size, tgt_vocab_size, d_model, num_heads, num_layers, d_ff, max_seq_length, dropout):\n",
    "        super(Transformer, self).__init__()\n",
    "        self.encoder_embedding = nn.Embedding(src_vocab_size, d_model)\n",
    "        self.decoder_embedding = nn.Embedding(tgt_vocab_size, d_model)\n",
    "        self.positional_encoding = PositionalEncoding(d_model, max_seq_length)\n",
    "\n",
    "        self.encoder_layers = nn.ModuleList([EncoderLayer(d_model, num_heads, d_ff, dropout) for _ in range(num_layers)])\n",
    "        self.decoder_layers = nn.ModuleList([DecoderLayer(d_model, num_heads, d_ff, dropout) for _ in range(num_layers)])\n",
    "\n",
    "        self.fc = nn.Linear(d_model, tgt_vocab_size)\n",
    "        self.dropout = nn.Dropout(dropout)\n",
    "\n",
    "    def generate_mask(self, src, tgt):\n",
    "        src_mask = (src != 0).unsqueeze(1).unsqueeze(2)\n",
    "        tgt_mask = (tgt != 0).unsqueeze(1).unsqueeze(3)\n",
    "        seq_length = tgt.size(1)\n",
    "        nopeak_mask = (1 - torch.triu(torch.ones(1, seq_length, seq_length), diagonal=1)).bool()\n",
    "        tgt_mask = tgt_mask & nopeak_mask\n",
    "        return src_mask, tgt_mask\n",
    "\n",
    "    def forward(self, src, tgt):\n",
    "        src_mask, tgt_mask = self.generate_mask(src, tgt)\n",
    "        src_embedded = self.dropout(self.positional_encoding(self.encoder_embedding(src)))\n",
    "        tgt_embedded = self.dropout(self.positional_encoding(self.decoder_embedding(tgt)))\n",
    "\n",
    "        enc_output = src_embedded\n",
    "        for enc_layer in self.encoder_layers:\n",
    "            enc_output = enc_layer(enc_output, src_mask)\n",
    "\n",
    "        dec_output = tgt_embedded\n",
    "        for dec_layer in self.decoder_layers:\n",
    "            dec_output = dec_layer(dec_output, enc_output, src_mask, tgt_mask)\n",
    "\n",
    "        output = self.fc(dec_output)\n",
    "\n",
    "        return output"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 51,
   "metadata": {},
   "outputs": [],
   "source": [
    "# give parameters for training purpose\n",
    "\n",
    "src_vocab_size = vocab_size_eng + 1\n",
    "tgt_vocab_size = vocab_size_french + 1\n",
    "d_model = 512\n",
    "num_heads = 1\n",
    "num_layers = 1\n",
    "d_ff = 512\n",
    "max_seq_length = 6\n",
    "dropout = 0.2"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "# pass all parameters to transformer class\n",
    "\n",
    "transformer = Transformer(src_vocab_size, tgt_vocab_size, d_model, num_heads, num_layers, d_ff, max_seq_length, dropout)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 47,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch: 1, Loss: 3.495068073272705\n",
      "Epoch: 2, Loss: 3.1860110759735107\n",
      "Epoch: 3, Loss: 3.0611572265625\n",
      "Epoch: 4, Loss: 2.873976469039917\n",
      "Epoch: 5, Loss: 2.784407377243042\n",
      "Epoch: 6, Loss: 2.6782991886138916\n",
      "Epoch: 7, Loss: 2.4403412342071533\n",
      "Epoch: 8, Loss: 2.414832592010498\n",
      "Epoch: 9, Loss: 2.278183698654175\n",
      "Epoch: 10, Loss: 2.147040605545044\n"
     ]
    }
   ],
   "source": [
    "# train the model \n",
    "\n",
    "criterion = nn.CrossEntropyLoss(ignore_index=0)\n",
    "optimizer = optim.Adam(transformer.parameters(), lr=0.0001, betas=(0.9, 0.98), eps=1e-9)\n",
    "\n",
    "transformer.train()\n",
    "\n",
    "for epoch in range(10):\n",
    "    optimizer.zero_grad()\n",
    "    output = transformer(src_data, tgt_data[:, :-1])\n",
    "    loss = criterion(output.contiguous().view(-1, tgt_vocab_size), tgt_data[:, 1:].contiguous().view(-1))\n",
    "    loss.backward()\n",
    "    optimizer.step()\n",
    "    print(f\"Epoch: {epoch+1}, Loss: {loss.item()}\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "tf2.2.0",
   "language": "python",
   "name": "tf2.2.0"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.7.6"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 4
}