text_aug.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {},
   "outputs": [],
   "source": [
    "from transformers import AutoModel, AutoTokenizer, pipeline\n",
    "from forgebox.imports import *\n",
    "from ipywidgets import Textarea, Layout\n",
    "from spacy import load\n",
    "from tqdm.notebook import tqdm\n",
    "from forgebox.ftorch.cuda import CudaHandler"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      ">>> 2 cuda devices found >>>\n",
      "Device 0: \n",
      "\tname:Tesla V100-PCIE-32GB\n",
      "\tused:31019MB\tfree:1491MB\n",
      "Device 1: \n",
      "\tname:Tesla V100-PCIE-32GB\n",
      "\tused:4MB\tfree:32506MB\n",
      "cuda stats refreshed\n",
      "Found the most idle GPU: cuda:1, 32506 MB Mem remained\n"
     ]
    }
   ],
   "source": [
    "dev = CudaHandler().idle().device"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Paraphrasing"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [],
   "source": [
    "nlp = load(\"en_ner_bionlp13cg_md\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "metadata": {},
   "outputs": [],
   "source": [
    "tag = \"tuner007/pegasus_paraphrase\"\n",
    "paraphraser = pipeline(model=tag, tokenizer=tag, task=\"text2text-generation\", device=dev.index)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 20,
   "metadata": {},
   "outputs": [],
   "source": [
    "def rephrase(text):\n",
    "    doc = nlp(text)\n",
    "    return \" \".join(pd.DataFrame(\n",
    "        paraphraser(list((map(lambda x:x.text, doc.sents))), truncation=True,)\n",
    "    )[\"generated_text\"])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 21,
   "metadata": {},
   "outputs": [],
   "source": [
    "train_df = pd.read_csv(\"train.csv\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 22,
   "metadata": {},
   "outputs": [],
   "source": [
    "def rewrite(df):\n",
    "    results = []\n",
    "    for excerpt in tqdm(list(df.excerpt)):\n",
    "        results.append(rephrase(excerpt))\n",
    "    df2 = df.copy()\n",
    "    df2[\"excerpt\"] = results\n",
    "    return df2"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 23,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "f379ca76dce248fb8f2b55fe0e6980ae",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "  0%|          | 0/2834 [00:00<?, ?it/s]"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "rewrite(train_df).to_csv(\"aug_train.csv\", index=False)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Testing"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 24,
   "metadata": {},
   "outputs": [],
   "source": [
    "aug1 = pd.read_csv(\"aug_train.csv\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 27,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "0"
      ]
     },
     "execution_count": 27,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "(aug1.excerpt==train_df.excerpt).sum()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## GPT2"
   ]
  },
  {
   "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": "3.7.6"
  },
  "toc": {
   "base_numbering": 1,
   "nav_menu": {},
   "number_sections": true,
   "sideBar": true,
   "skip_h1_title": false,
   "title_cell": "Table of Contents",
   "title_sidebar": "Contents",
   "toc_cell": false,
   "toc_position": {},
   "toc_section_display": true,
   "toc_window_display": false
  }
 },
 "nbformat": 4,
 "nbformat_minor": 4
}