seed/merge_training_dataset_json.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Merge from multiple jsonl files for OpenAI fine-tuning\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "tags": []
   },
   "outputs": [],
   "source": [
    "%load_ext autoreload\n",
    "%autoreload 2"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "from dotenv import load_dotenv\n",
    "import os, shutil, random\n",
    "from util.preprocess import convert_html_to_md\n",
    "import glob\n",
    "import pandas as pd\n",
    "load_dotenv()\n",
    "\n",
    "dataset_dir = \"dataset\"\n",
    "all_files = glob.glob(os.path.join(dataset_dir, \"*-oai.jsonl\"))\n",
    "#df = pd.concat((pd.read_csv(f) for f in all_files), ignore_index=True)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Convert to a base64 image input format that can be recognized by multimodal models such as GPT-4o.\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "all_files"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "import json\n",
    "import glob\n",
    "\n",
    "output_dir = \"final_dataset\"\n",
    "os.makedirs(output_dir, exist_ok=True)\n",
    "result = []\n",
    "\n",
    "for f in all_files:\n",
    "    with open(f, 'r', encoding='utf-8-sig') as infile:\n",
    "        for line in infile.readlines():\n",
    "            try:\n",
    "                result.append(json.loads(line)) # read each line of the file\n",
    "            except ValueError:\n",
    "                print(f)\n",
    "\n",
    "# This would output jsonl\n",
    "with open(f'{output_dir}/training-data-v1.jsonl','w', encoding='utf-8-sig') as outfile:\n",
    "    for entry in result:\n",
    "        outfile.write(json.dumps(entry, ensure_ascii=False) + \"\\n\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "# !rm -rf {output_dir}"
   ]
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 3 (ipykernel)",
   "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": "3.9.18"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 4
}