Make-a-bot.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Make-a-bot\n",
    "Baixe os tweets da sua personalidade favorita e faça uma RNN que tenta prever as próximas palavras."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "#!conda install -v -c conda-forge tweepy\n",
    "%reload_ext autoreload\n",
    "%autoreload 2\n",
    "%matplotlib inline"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "from fastai.text import *\n",
    "from pathlib import Path\n",
    "\n",
    "from tweet_dumper import get_all_tweets"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "tweetpath = Path('./tweets')\n",
    "modelspath = Path('./models')\n",
    "path = Path('./')\n",
    "best_model_path = Path('./models/bestmodel30k')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Download do modelo em portugues pretreinado na wikipedia\n",
    "!curl https://storage.googleapis.com/gde-dl-bsb/models/bestmodel30k.pth -o ../models/bestmodel30k.pth\n",
    "# Download do vocabulário\n",
    "!curl https://storage.googleapis.com/gde-dl-bsb/models/itos.pkl -o ../models/itos.pkl"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Baixando os tweets\n",
    "\n",
    "Escreva suas chaves de acesso da API do Twitter e defina o nome do perfil de twitter que gostaria de usar para treinar a rede.\n",
    "\n",
    "Mais informações em https://developer.twitter.com/"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "consumer_key = \"\"\n",
    "consumer_secret = \"\"\n",
    "access_key = \"\"\n",
    "access_secret = \"\"\n",
    "\n",
    "keys = [consumer_key,consumer_secret,access_key,access_secret]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "screen_name = \"jairbolsonaro\""
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "get_all_tweets(screen_name, keys)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Preparando os dados"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "def readtweets(d): return [' '.join(o.strip() for o in open(d).readlines())]\n",
    "tweets = 'tweets/' + '{}_tweets.txt'.format(screen_name)\n",
    "t = readtweets(tweets)\n",
    "\n",
    "#separa 80% dos tweets para treino e 20% para validação\n",
    "perc80 = len(t) - len(t)//5\n",
    "\n",
    "train_txt = valid_txt = []\n",
    "train_txt.append(t[0][:perc80])\n",
    "valid_txt.append(t[0][perc80:])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "bs = 64\n",
    "\n",
    "train = TextList(list(train_txt), path=path);\n",
    "valid = TextList(list(valid_txt), path=path);\n",
    "\n",
    "src = ItemLists(path=path, train=train, valid=valid).label_for_lm()\n",
    "data = src.databunch(bs=bs)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "#tokenizador\n",
    "tokenizer = Tokenizer(lang='pt', n_cpus=8)\n",
    "#vocabulario\n",
    "with modelspath.joinpath('itos.pkl').open('rb') as f:\n",
    "    itos = pickle.load(f)\n",
    "vocab = Vocab(itos)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Transfer Learning"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "learn = language_model_learner(data, arch=AWD_LSTM, pretrained_fnames=(best_model_path,modelspath.joinpath('itos')))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "learn.lr_find()\n",
    "learn.recorder.plot()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "learn.fit_one_cycle(1, max_lr=5e-2)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "learn.fit_one_cycle(5, max_lr=5e-2)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "learn.save('first-try')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Fine Tuning"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "learn.unfreeze()\n",
    "learn.lr_find()\n",
    "learn.recorder.plot()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "learn.fit_one_cycle(5, slice(5e-3/(2.6**4),5e-3), moms=(0.8,0.7))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "learn.save('fine-tuned-model')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Prediction"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "TEXT = \"\"\n",
    "N_WORDS = 50\n",
    "N_SENTENCES = 5\n",
    "print(\"\\n\".join(learn.predict(TEXT, N_WORDS, temperature=0.75) for _ in range(N_SENTENCES)))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "----\n",
    "# Load Examples\n",
    "\n",
    "Podemos carregar robôs pré-treinados também. Alguns dos resultados já estão disponíveis na pasta *examples*.\n",
    "\n",
    "## Bolsobot"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Download do modelo pré-treinado do Bolsonaro\n",
    "!curl https://www.dropbox.com/s/bn4i4x6jiinejzy/Bolsobot.pth -o ../models/Bolsobot.pth"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "learn.load('Bolsobot')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "TEXT = \"Grande dia!\"\n",
    "N_WORDS = 50\n",
    "N_SENTENCES = 5\n",
    "print(\"\\n\".join(learn.predict(TEXT, N_WORDS, temperature=0.75) for _ in range(N_SENTENCES)))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Robolavo"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Download do modelo pré-treinado do Olavo de Carvalho\n",
    "!curl https://www.dropbox.com/s/yhsbw1yu9a0844x/olavobot-twitter.pth -o ../models/olavobot-twitter.pth"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "learn.load('olavobot-twitter')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "TEXT = \"A Terra é\"\n",
    "N_WORDS = 50\n",
    "N_SENTENCES = 5\n",
    "print(\"\\n\".join(learn.predict(TEXT, N_WORDS, temperature=0.75) for _ in range(N_SENTENCES)))"
   ]
  }
 ],
 "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
}