tools/make_hf_dataset.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "# This notebook will guide you to make correct format of Huggingface dataset, in proper parquet format and visualizable in Huggingface dataset hub.\n",
    "# We will take the example of the dataset \"Otter-AI/MMVet\" and convert it to the proper format."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "from datasets import load_dataset\n",
    "\n",
    "data_path = \"Otter-AI/MMVet\"\n",
    "df = load_dataset(data_path, split=\"test\").to_pandas()\n",
    "df.head()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "from datasets import Dataset, Features, Value, Image\n",
    "import pandas as pd\n",
    "\n",
    "# Define the features for the dataset\n",
    "features = Features(\n",
    "    {\n",
    "        \"question_id\": Value(dtype=\"string\"),\n",
    "        \"image\": Image(),\n",
    "        \"question\": Value(dtype=\"string\"),\n",
    "        \"answer\": Value(dtype=\"string\"),\n",
    "        \"image_source\": Value(dtype=\"string\"),\n",
    "        \"capability\": Value(dtype=\"string\"),\n",
    "        # Add other fields as necessary\n",
    "    }\n",
    ")\n",
    "\n",
    "df_items = {\n",
    "    \"question_id\": [],\n",
    "    \"image\": [],\n",
    "    \"question\": [],\n",
    "    \"answer\": [],\n",
    "    \"image_source\": [],\n",
    "    \"capability\": [],\n",
    "}\n",
    "\n",
    "for idx, row in df.iterrows():\n",
    "    df_items[\"question_id\"].append(str(row[\"id\"]))\n",
    "    image = {\"bytes\": row[\"images\"][0][\"bytes\"], \"path\": \"\"}\n",
    "    df_items[\"image\"].append(image)\n",
    "    df_items[\"question\"].append(str(row[\"instruction\"]))\n",
    "    df_items[\"answer\"].append(str(row[\"answer\"]))\n",
    "    df_items[\"image_source\"].append(str(row[\"image_source\"]))\n",
    "    df_items[\"capability\"].append(\",\".join(list(row[\"capability\"])))\n",
    "    # Add other fields as necessary\n",
    "\n",
    "df_items = pd.DataFrame(df_items)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "df_items.head()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "dataset = Dataset.from_pandas(df_items, features=features)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "hub_dataset_path = \"lmms-lab/MMVet\"\n",
    "dataset.push_to_hub(repo_id=hub_dataset_path, split=\"test\")"
   ]
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "lmms-eval",
   "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": 2
}