{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# TransitPulse — GPU Performance Benchmark Race on Colab T4\n",
    "This notebook installs RAPIDS `cudf.pandas`, generates synthetic transit pings at 1M, 25M, and 150M row scales, and runs the identical processing pipeline on both CPU (pure pandas) and GPU (cudf.pandas) to measure speedups."
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### 1. Install RAPIDS cudf.pandas and requirements"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Install RAPIDS cudf (compatible with Colab T4 GPU)\n",
    "!pip install -q git+https://github.com/rapidsai/cudf.git\n",
    "!pip install pyarrow duckdb fastapi uvicorn matplotlib google-genai"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### 2. Download/Clone repository code\n",
    "*(Or upload config.py, generate_pings.py, pipeline.py, and benchmark.py directly)*"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Run the benchmark race live across all scales!\n",
    "# This will output the timings and speedup_chart.png\n",
    "!python benchmark.py --run-all"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### 3. Display the GPU vs CPU Speedup Chart"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "from IPython.display import Image\n",
    "Image('data/speedup_chart.png')"
   ]
  }
 ],
 "metadata": {
  "language_info": {
   "name": "python"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}