{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 1,
   "id": "3b26d339-8e2d-4f7e-8dbf-24c649932de4",
   "metadata": {
    "tags": []
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Note: you may need to restart the kernel to use updated packages.\n"
     ]
    }
   ],
   "source": [
    "%pip install -q huggingface-hub plotly numpy"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "b7aebf99-b8fb-4892-90e2-2261202ac576",
   "metadata": {
    "tags": []
   },
   "outputs": [],
   "source": [
    "import plotly.graph_objects as go\n",
    "import numpy as np\n",
    "\n",
    "# Setting the time scale\n",
    "time = np.arange(0, 24, 0.1)  # 24 hours, in increments of 0.1 hour\n",
    "\n",
    "# Generating usage patterns\n",
    "np.random.seed(42)\n",
    "low_usage = (np.random.poisson(70, len(time)) * (np.random.rand(len(time)) < 0.15).astype(int))//2 \n",
    "bursty_usage = np.random.poisson(70, len(time)) * (np.random.rand(len(time)) < 0.15).astype(int)\n",
    "bursty_usage = np.random.poisson(100, len(time)) * (np.random.rand(len(time)) < 0.2).astype(int)//1.4\n",
    "\n",
    "high_volume = np.random.normal(loc=15, scale=2, size=len(time))*1.75\n",
    "\n",
    "# Applying smoothing\n",
    "smooth_low_usage = np.convolve(low_usage, np.ones(50)/50, mode='same')\n",
    "smooth_bursty_usage = np.convolve(bursty_usage, np.ones(50)/50, mode='same')\n",
    "smooth_high_volume = np.convolve(high_volume, np.ones(50)/50, mode='same')\n",
    "total_usage = smooth_low_usage + smooth_bursty_usage + smooth_high_volume\n",
    "\n",
    "# Plotting using Plotly\n",
    "fig = go.Figure()\n",
    "fig.add_trace(go.Scatter(x=time, y=smooth_low_usage, mode='lines', name='Low Usage', line=dict(color='#B0C4DE')))  # Light Steel Blue\n",
    "fig.add_trace(go.Scatter(x=time, y=smooth_bursty_usage, mode='lines', name='Bursty Usage', line=dict(color='#FFB6C1')))  # Light Pink\n",
    "fig.add_trace(go.Scatter(x=time, y=smooth_high_volume, mode='lines', name='High Volume', line=dict(color='#98FB98')))  # Pale Green\n",
    "fig.add_trace(go.Scatter(x=time, y=total_usage, mode='lines', name='Total Usage', line=dict(color='#2F4F4F', width=3)))  # Dark Slate Gray\n",
    "\n",
    "fig.update_layout(\n",
    "    title={\n",
    "        'text': 'Comparison of Usage Models and Total Impact',\n",
    "        'y': 0.9,\n",
    "        'x': 0.5,\n",
    "        'xanchor': 'center',\n",
    "        'yanchor': 'top',\n",
    "        'font': {\n",
    "            'size': 24  # Adjust the font size as needed\n",
    "        }\n",
    "    },\n",
    "    xaxis_title='Time (hours)',\n",
    "    yaxis_title='Requests',\n",
    "    legend_title='Usage Pattern',\n",
    "    xaxis=dict(\n",
    "        title_font_size=28,  # Larger axis title font size\n",
    "        tickfont_size=16  # Larger tick label font size\n",
    "    ),\n",
    "    yaxis=dict(\n",
    "        title_font_size=28,  # Larger axis title font size\n",
    "        tickfont_size=16  # Larger tick label font size\n",
    "    ),\n",
    "    legend=dict(\n",
    "        x=0.5,  # Horizontal position, 0 is left\n",
    "        y=-0.1,  # Vertical position, negative values to move it down\n",
    "        orientation=\"h\",  # Horizontal layout\n",
    "        xanchor='center',  # Anchor the legend at the center\n",
    "        yanchor='top'  # Anchor the legend at the top\n",
    "    ),\n",
    "    template='plotly_white'\n",
    ")\n",
    "\n",
    "\n",
    "fig.show()\n"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "fc54b448-2eb3-44cc-8304-983e27138296",
   "metadata": {},
   "source": [
    "# Push Image"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "4ef615e2-ab36-43f1-b310-cac8b17d29b7",
   "metadata": {
    "tags": []
   },
   "outputs": [],
   "source": [
    "image_out = \"multi-lora-serving-pattern.png\"\n",
    "fig.write_image(image_out, width=1920, height=1080, scale=2)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "13d8ee16-65ec-4a32-84e9-e6b99aab92af",
   "metadata": {
    "tags": []
   },
   "outputs": [],
   "source": [
    "from huggingface_hub import HfApi\n",
    "\n",
    "api = HfApi()\n",
    "api.upload_file(\n",
    "    path_or_fileobj=image_out,\n",
    "    path_in_repo=f\"blog/multi-lora-serving/{image_out}\",\n",
    "    repo_id=\"huggingface/documentation-images\",\n",
    "    repo_type=\"dataset\",\n",
    "    commit_message=\"Updating title\",\n",
    ")"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "81e6b86b-c439-4acd-a072-2ef3ab782f90",
   "metadata": {},
   "source": [
    "# Push Notebook"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "id": "adca7ec0-25c3-42e3-84b7-f7f9342e4e4f",
   "metadata": {},
   "outputs": [],
   "source": [
    "from huggingface_hub import HfApi\n",
    "\n",
    "api = HfApi()\n",
    "api.upload_file(\n",
    "    path_or_fileobj=\"multi-lora-serving-pattern.ipynb\",\n",
    "    path_in_repo=\"blog/multi-lora-serving/multi-lora-serving-pattern.ipynb\",\n",
    "    repo_id=\"huggingface/documentation-images\",\n",
    "    repo_type=\"dataset\",\n",
    ")"
   ]
  }
 ],
 "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": "3.10.14"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 5
}