aitech-eks-pub/cw/13_transformery2.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "![Logo 1](https://git.wmi.amu.edu.pl/AITech/Szablon/raw/branch/master/Logotyp_AITech1.jpg)\n",
    "<div class=\"alert alert-block alert-info\">\n",
    "<h1> Ekstrakcja informacji </h1>\n",
    "<h2> 13. <i>Transformery 2</i>  [ćwiczenia]</h2> \n",
    "<h3> Jakub Pokrywka (2021)</h3>\n",
    "</div>\n",
    "\n",
    "![Logo 2](https://git.wmi.amu.edu.pl/AITech/Szablon/raw/branch/master/Logotyp_AITech2.jpg)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Wizualizacja atencji\n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "https://github.com/jessevig/bertviz"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "scrolled": true
   },
   "outputs": [],
   "source": [
    "!pip install bertviz"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "from transformers import AutoTokenizer, AutoModel\n",
    "from bertviz import model_view, head_view"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "TEXT = \"This is a sample input sentence for a transformer model\""
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "MODEL = \"distilbert-base-uncased\""
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "tokenizer = AutoTokenizer.from_pretrained(MODEL)\n",
    "model = AutoModel.from_pretrained(MODEL, output_attentions=True)\n",
    "inputs = tokenizer.encode(TEXT, return_tensors='pt')\n",
    "outputs = model(inputs)\n",
    "attention = outputs[-1]\n",
    "tokens = tokenizer.convert_ids_to_tokens(inputs[0]) \n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### SELF ATTENTION MODELS"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "head_view(attention, tokens)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "model_view(attention, tokens)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### ENCODER-DECODER MODELS"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "MODEL = \"Helsinki-NLP/opus-mt-en-de\""
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "TEXT_ENCODER = \"She sees the small elephant.\"\n",
    "TEXT_DECODER = \"Sie sieht den kleinen Elefanten.\""
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "tokenizer = AutoTokenizer.from_pretrained(MODEL)\n",
    "model = AutoModel.from_pretrained(MODEL, output_attentions=True)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "encoder_input_ids = tokenizer(TEXT_ENCODER, return_tensors=\"pt\", add_special_tokens=True).input_ids\n",
    "decoder_input_ids = tokenizer(TEXT_DECODER, return_tensors=\"pt\", add_special_tokens=True).input_ids\n",
    "\n",
    "outputs = model(input_ids=encoder_input_ids, decoder_input_ids=decoder_input_ids)\n",
    "\n",
    "encoder_text = tokenizer.convert_ids_to_tokens(encoder_input_ids[0])\n",
    "decoder_text = tokenizer.convert_ids_to_tokens(decoder_input_ids[0])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "head_view(\n",
    "    encoder_attention=outputs.encoder_attentions,\n",
    "    decoder_attention=outputs.decoder_attentions,\n",
    "    cross_attention=outputs.cross_attentions,\n",
    "    encoder_tokens= encoder_text,\n",
    "    decoder_tokens = decoder_text\n",
    ")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "scrolled": false
   },
   "outputs": [],
   "source": [
    "model_view(\n",
    "    encoder_attention=outputs.encoder_attentions,\n",
    "    decoder_attention=outputs.decoder_attentions,\n",
    "    cross_attention=outputs.cross_attentions,\n",
    "    encoder_tokens= encoder_text,\n",
    "    decoder_tokens = decoder_text\n",
    ")"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Zadanie (10 minut)\n",
    "\n",
    "Za pomocą modelu en-fr przetłumacz dowolne zdanie z angielskiego na język francuski i sprawdź wagi atencji dla tego tłumaczenia"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### PRZYKŁAD: GPT3"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### ZADANIE DOMOWE - POLEVAL"
   ]
  }
 ],
 "metadata": {
  "author": "Jakub Pokrywka",
  "email": "kubapok@wmi.amu.edu.pl",
  "kernelspec": {
   "display_name": "Python 3",
   "language": "python",
   "name": "python3"
  },
  "lang": "pl",
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.8.3"
  },
  "subtitle": "13.Transformery 2[ćwiczenia]",
  "title": "Ekstrakcja informacji",
  "year": "2021"
 },
 "nbformat": 4,
 "nbformat_minor": 4
}