Tracking_Emissions_from_ML_Models.ipynb

{
  "cells": [
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "view-in-github",
        "colab_type": "text"
      },
      "source": [
        "<a href=\"https://colab.research.google.com/github/armandpriko/CSS-CHALLENGE-1/blob/main/Tracking_Emissions_from_ML_Models.ipynb\" target=\"_parent\"><img src=\"https://colab.research.google.com/assets/colab-badge.svg\" alt=\"Open In Colab\"/></a>"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "EFpgp_rIA_TY"
      },
      "source": [
        "# How to Track Emissions when Training Machine Learning Models\n",
        "**Content Creator**: Mel Hanna (mdriscollhanna@gmail.com)\n",

        "\n",
        "Welcome to CCAI's tutorial on tracking and mitigating the emissions from ML model training!\n",
        "\n",
        "In order to make edits to this notebook, you should press **File > \"Save a Copy in Drive\"**. This will ensure that any edits will be on your local copy, and they will not affect the notebook shared with everyone else."
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "gchK901cq3hN"
      },
      "source": [
        "# Table of Contents\n",
        "\n",
        "\n",
        "*   [Overview](#overview)\n",
        "*   [Target Audience](#target-audience)\n",
        "*   [Background & Prerequisites](#background-and-prereqs)\n",
        "*   [Software Requirements](#software-requirements)\n",
        "*   [Climate Impact](#climate-impact)\n",
        "*   [Methodology](#methodology)\n",
        "*   [Results & Discussion](#codecarbon-experiments)\n",
        "*   [References](#references)"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "Nuv5XES8Ejc_"
      },
      "source": [
        "<a name=\"overview\"></a>\n",
        "# Overview\n",
        "This tutorial explains the methodology behind calculating computing-related GHG emissions from training machine learning models and demonstrates some strategies to reduce a model's carbon footprint.\n",
        "\n",
        "Specifically, you will learn how to:\n",
        "- measure a ML model's emissions with CodeCarbon\n",
        "- potentially reduce a ML model's carbon footprint during training\n",
        "\n",
        "\n"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "xyb28HbsrcLJ"
      },
      "source": [
        "<a name=\"target-audience\"></a>\n",
        "## **Target Audience**\n",
        "This tutorial is intended for experienced and aspiring data scientists looking for concrete examples of how to track carbon emissions while executing code and how to train machine learning models in a more efficient way. The reader should understand the basic mechanics of training an ML model, such as splitting data between train and test sets and tuning hyperparameters."
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "PMhFV8gzrqO-"
      },
      "source": [
        "<a name=\"software-requirements\"></a>\n",
        "## **Software Requirements**\n",
        "\n",
        "This notebook requires Python >= 3.7. The following libraries are required:\n",
        "*   tqdm\n",
        "*   pandas\n",
        "*   numpy\n",
        "*   matplotlib\n",
        "*   pytorch"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "H08M_-5iEbc1"
      },
      "source": [
        "### **Enabling GPU in Google Colab**\n",
        "Before we start, you will need access to a GPU.  Fortunately, Google Colab provides free access to computing resources including GPUs. The GPUs currently available in Colab include Nvidia K80s, T4s, P4s and P100s. Unfortunately, there is no way to choose what type of GPU you can connect to in Colab. [See here for information](https://research.google.com/colaboratory/faq.html#gpu-availability).\n",
        "\n",
        "To enable GPU in Google Colab:\n",
        "1. Navigate to Edit→Notebook Settings or Runtime→Change Runtime Type.\n",
        "2. Select GPU from the Hardware Accelerator drop-down."
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "-DPvBsAj78ec"
      },
      "source": [
        "### **Other dependencies**\n",
        "\n",
        "We'll also take a moment to install and import the necessary dependencies for executing the rest of the notebook."
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "vUiCHIpvYSW2",
        "outputId": "eb65fc5e-7627-4833-9122-48d7276ea044"
      },
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Looking in indexes: https://pypi.org/simple, https://us-python.pkg.dev/colab-wheels/public/simple/\n",
            "Collecting codecarbon\n",
            "  Downloading codecarbon-2.1.4-py3-none-any.whl (174 kB)\n",
            "     ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 174.9/174.9 kB 4.7 MB/s eta 0:00:00\n",
            "Requirement already satisfied: py-cpuinfo in /usr/local/lib/python3.9/dist-packages (from codecarbon) (9.0.0)\n",
            "Requirement already satisfied: click in /usr/local/lib/python3.9/dist-packages (from codecarbon) (8.1.3)\n",
            "Collecting arrow\n",
            "  Downloading arrow-1.2.3-py3-none-any.whl (66 kB)\n",
            "     ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 66.4/66.4 kB 8.0 MB/s eta 0:00:00\n",
            "Requirement already satisfied: pandas in /usr/local/lib/python3.9/dist-packages (from codecarbon) (1.5.3)\n",
            "Collecting pynvml\n",
            "  Downloading pynvml-11.5.0-py3-none-any.whl (53 kB)\n",
            "     ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 53.1/53.1 kB 7.1 MB/s eta 0:00:00\n",
            "Requirement already satisfied: requests in /usr/local/lib/python3.9/dist-packages (from codecarbon) (2.27.1)\n",
            "Collecting fuzzywuzzy\n",
            "  Downloading fuzzywuzzy-0.18.0-py2.py3-none-any.whl (18 kB)\n",
            "Requirement already satisfied: psutil in /usr/local/lib/python3.9/dist-packages (from codecarbon) (5.9.4)\n",
            "Requirement already satisfied: python-dateutil>=2.7.0 in /usr/local/lib/python3.9/dist-packages (from arrow->codecarbon) (2.8.2)\n",
            "Requirement already satisfied: pytz>=2020.1 in /usr/local/lib/python3.9/dist-packages (from pandas->codecarbon) (2022.7.1)\n",
            "Requirement already satisfied: numpy>=1.20.3 in /usr/local/lib/python3.9/dist-packages (from pandas->codecarbon) (1.22.4)\n",
            "Requirement already satisfied: idna<4,>=2.5 in /usr/local/lib/python3.9/dist-packages (from requests->codecarbon) (3.4)\n",
            "Requirement already satisfied: charset-normalizer~=2.0.0 in /usr/local/lib/python3.9/dist-packages (from requests->codecarbon) (2.0.12)\n",
            "Requirement already satisfied: urllib3<1.27,>=1.21.1 in /usr/local/lib/python3.9/dist-packages (from requests->codecarbon) (1.26.15)\n",
            "Requirement already satisfied: certifi>=2017.4.17 in /usr/local/lib/python3.9/dist-packages (from requests->codecarbon) (2022.12.7)\n",
            "Requirement already satisfied: six>=1.5 in /usr/local/lib/python3.9/dist-packages (from python-dateutil>=2.7.0->arrow->codecarbon) (1.16.0)\n",
            "Installing collected packages: fuzzywuzzy, pynvml, arrow, codecarbon\n",
            "Successfully installed arrow-1.2.3 codecarbon-2.1.4 fuzzywuzzy-0.18.0 pynvml-11.5.0\n",
            "Looking in indexes: https://pypi.org/simple, https://us-python.pkg.dev/colab-wheels/public/simple/\n",
            "Collecting dash\n",
            "  Downloading dash-2.9.3-py3-none-any.whl (10.2 MB)\n",
            "     ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 10.2/10.2 MB 67.5 MB/s eta 0:00:00\n",
            "Requirement already satisfied: Flask>=1.0.4 in /usr/local/lib/python3.9/dist-packages (from dash) (2.2.3)\n",
            "Collecting dash-core-components==2.0.0\n",
            "  Downloading dash_core_components-2.0.0-py3-none-any.whl (3.8 kB)\n",
            "Collecting dash-table==5.0.0\n",
            "  Downloading dash_table-5.0.0-py3-none-any.whl (3.9 kB)\n",
            "Requirement already satisfied: plotly>=5.0.0 in /usr/local/lib/python3.9/dist-packages (from dash) (5.13.1)\n",
            "Collecting dash-html-components==2.0.0\n",
            "  Downloading dash_html_components-2.0.0-py3-none-any.whl (4.1 kB)\n",
            "Requirement already satisfied: itsdangerous>=2.0 in /usr/local/lib/python3.9/dist-packages (from Flask>=1.0.4->dash) (2.1.2)\n",
            "Requirement already satisfied: Jinja2>=3.0 in /usr/local/lib/python3.9/dist-packages (from Flask>=1.0.4->dash) (3.1.2)\n",
            "Requirement already satisfied: Werkzeug>=2.2.2 in /usr/local/lib/python3.9/dist-packages (from Flask>=1.0.4->dash) (2.2.3)\n",
            "Requirement already satisfied: click>=8.0 in /usr/local/lib/python3.9/dist-packages (from Flask>=1.0.4->dash) (8.1.3)\n",
            "Requirement already satisfied: importlib-metadata>=3.6.0 in /usr/local/lib/python3.9/dist-packages (from Flask>=1.0.4->dash) (6.3.0)\n",
            "Requirement already satisfied: tenacity>=6.2.0 in /usr/local/lib/python3.9/dist-packages (from plotly>=5.0.0->dash) (8.2.2)\n",
            "Requirement already satisfied: zipp>=0.5 in /usr/local/lib/python3.9/dist-packages (from importlib-metadata>=3.6.0->Flask>=1.0.4->dash) (3.15.0)\n",
            "Requirement already satisfied: MarkupSafe>=2.0 in /usr/local/lib/python3.9/dist-packages (from Jinja2>=3.0->Flask>=1.0.4->dash) (2.1.2)\n",
            "Installing collected packages: dash-table, dash-html-components, dash-core-components, dash\n",
            "Successfully installed dash-2.9.3 dash-core-components-2.0.0 dash-html-components-2.0.0 dash-table-5.0.0\n",
            "Looking in indexes: https://pypi.org/simple, https://us-python.pkg.dev/colab-wheels/public/simple/\n",
            "Collecting dash-bootstrap-components==0.13.1\n",
            "  Downloading dash_bootstrap_components-0.13.1-py3-none-any.whl (197 kB)\n",
            "     ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 197.3/197.3 kB 6.0 MB/s eta 0:00:00\n",
            "Requirement already satisfied: dash>=1.9.0 in /usr/local/lib/python3.9/dist-packages (from dash-bootstrap-components==0.13.1) (2.9.3)\n",
            "Requirement already satisfied: Flask>=1.0.4 in /usr/local/lib/python3.9/dist-packages (from dash>=1.9.0->dash-bootstrap-components==0.13.1) (2.2.3)\n",
            "Requirement already satisfied: dash-html-components==2.0.0 in /usr/local/lib/python3.9/dist-packages (from dash>=1.9.0->dash-bootstrap-components==0.13.1) (2.0.0)\n",
            "Requirement already satisfied: plotly>=5.0.0 in /usr/local/lib/python3.9/dist-packages (from dash>=1.9.0->dash-bootstrap-components==0.13.1) (5.13.1)\n",
            "Requirement already satisfied: dash-table==5.0.0 in /usr/local/lib/python3.9/dist-packages (from dash>=1.9.0->dash-bootstrap-components==0.13.1) (5.0.0)\n",
            "Requirement already satisfied: dash-core-components==2.0.0 in /usr/local/lib/python3.9/dist-packages (from dash>=1.9.0->dash-bootstrap-components==0.13.1) (2.0.0)\n",
            "Requirement already satisfied: importlib-metadata>=3.6.0 in /usr/local/lib/python3.9/dist-packages (from Flask>=1.0.4->dash>=1.9.0->dash-bootstrap-components==0.13.1) (6.3.0)\n",
            "Requirement already satisfied: click>=8.0 in /usr/local/lib/python3.9/dist-packages (from Flask>=1.0.4->dash>=1.9.0->dash-bootstrap-components==0.13.1) (8.1.3)\n",
            "Requirement already satisfied: Werkzeug>=2.2.2 in /usr/local/lib/python3.9/dist-packages (from Flask>=1.0.4->dash>=1.9.0->dash-bootstrap-components==0.13.1) (2.2.3)\n",
            "Requirement already satisfied: itsdangerous>=2.0 in /usr/local/lib/python3.9/dist-packages (from Flask>=1.0.4->dash>=1.9.0->dash-bootstrap-components==0.13.1) (2.1.2)\n",
            "Requirement already satisfied: Jinja2>=3.0 in /usr/local/lib/python3.9/dist-packages (from Flask>=1.0.4->dash>=1.9.0->dash-bootstrap-components==0.13.1) (3.1.2)\n",
            "Requirement already satisfied: tenacity>=6.2.0 in /usr/local/lib/python3.9/dist-packages (from plotly>=5.0.0->dash>=1.9.0->dash-bootstrap-components==0.13.1) (8.2.2)\n",
            "Requirement already satisfied: zipp>=0.5 in /usr/local/lib/python3.9/dist-packages (from importlib-metadata>=3.6.0->Flask>=1.0.4->dash>=1.9.0->dash-bootstrap-components==0.13.1) (3.15.0)\n",
            "Requirement already satisfied: MarkupSafe>=2.0 in /usr/local/lib/python3.9/dist-packages (from Jinja2>=3.0->Flask>=1.0.4->dash>=1.9.0->dash-bootstrap-components==0.13.1) (2.1.2)\n",
            "Installing collected packages: dash-bootstrap-components\n",
            "Successfully installed dash-bootstrap-components-0.13.1\n",
            "Looking in indexes: https://pypi.org/simple, https://us-python.pkg.dev/colab-wheels/public/simple/\n",
            "Collecting netCDF4\n",
            "  Downloading netCDF4-1.6.3-cp39-cp39-manylinux_2_17_x86_64.manylinux2014_x86_64.whl (5.2 MB)\n",
            "     ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 5.2/5.2 MB 50.5 MB/s eta 0:00:00\n",
            "Collecting cftime\n",
            "  Downloading cftime-1.6.2-cp39-cp39-manylinux_2_17_x86_64.manylinux2014_x86_64.whl (1.3 MB)\n",
            "     ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 1.3/1.3 MB 76.8 MB/s eta 0:00:00\n",
            "Requirement already satisfied: numpy in /usr/local/lib/python3.9/dist-packages (from netCDF4) (1.22.4)\n",
            "Installing collected packages: cftime, netCDF4\n",
            "Successfully installed cftime-1.6.2 netCDF4-1.6.3\n",
            "Looking in indexes: https://pypi.org/simple, https://us-python.pkg.dev/colab-wheels/public/simple/\n",
            "Collecting fire\n",
            "  Downloading fire-0.5.0.tar.gz (88 kB)\n",
            "     ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 88.3/88.3 kB 3.9 MB/s eta 0:00:00\n",
            "  Preparing metadata (setup.py): started\n",
            "  Preparing metadata (setup.py): finished with status 'done'\n",
            "Requirement already satisfied: six in /usr/local/lib/python3.9/dist-packages (from fire) (1.16.0)\n",
            "Requirement already satisfied: termcolor in /usr/local/lib/python3.9/dist-packages (from fire) (2.2.0)\n",
            "Building wheels for collected packages: fire\n",
            "  Building wheel for fire (setup.py): started\n",
            "  Building wheel for fire (setup.py): finished with status 'done'\n",
            "  Created wheel for fire: filename=fire-0.5.0-py2.py3-none-any.whl size=116952 sha256=573748a7065a92380085745e83cf0cd53aa84fca5171e4c2016c88a239192945\n",
            "  Stored in directory: /root/.cache/pip/wheels/f7/f1/89/b9ea2bf8f80ec027a88fef1d354b3816b4d3d29530988972f6\n",
            "Successfully built fire\n",
            "Installing collected packages: fire\n",
            "Successfully installed fire-0.5.0\n",
            "Looking in indexes: https://pypi.org/simple, https://us-python.pkg.dev/colab-wheels/public/simple/\n",
            "Collecting optuna\n",
            "  Downloading optuna-3.1.1-py3-none-any.whl (365 kB)\n",
            "     ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 365.7/365.7 kB 8.8 MB/s eta 0:00:00\n",
            "Collecting alembic>=1.5.0\n",
            "  Downloading alembic-1.10.3-py3-none-any.whl (212 kB)\n",
            "     ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 212.3/212.3 kB 25.8 MB/s eta 0:00:00\n",
            "Requirement already satisfied: numpy in /usr/local/lib/python3.9/dist-packages (from optuna) (1.22.4)\n",
            "Requirement already satisfied: tqdm in /usr/local/lib/python3.9/dist-packages (from optuna) (4.65.0)\n",
            "Collecting colorlog\n",
            "  Downloading colorlog-6.7.0-py2.py3-none-any.whl (11 kB)\n",
            "Requirement already satisfied: packaging>=20.0 in /usr/local/lib/python3.9/dist-packages (from optuna) (23.0)\n",
            "Requirement already satisfied: sqlalchemy>=1.3.0 in /usr/local/lib/python3.9/dist-packages (from optuna) (2.0.9)\n",
            "Collecting cmaes>=0.9.1\n",
            "  Downloading cmaes-0.9.1-py3-none-any.whl (21 kB)\n",
            "Requirement already satisfied: PyYAML in /usr/local/lib/python3.9/dist-packages (from optuna) (6.0)\n",
            "Collecting Mako\n",
            "  Downloading Mako-1.2.4-py3-none-any.whl (78 kB)\n",
            "     ━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 78.7/78.7 kB 10.3 MB/s eta 0:00:00\n",
            "Requirement already satisfied: typing-extensions>=4 in /usr/local/lib/python3.9/dist-packages (from alembic>=1.5.0->optuna) (4.5.0)\n",
            "Requirement already satisfied: greenlet!=0.4.17 in /usr/local/lib/python3.9/dist-packages (from sqlalchemy>=1.3.0->optuna) (2.0.2)\n",
            "Requirement already satisfied: MarkupSafe>=0.9.2 in /usr/local/lib/python3.9/dist-packages (from Mako->alembic>=1.5.0->optuna) (2.1.2)\n",
            "Installing collected packages: Mako, colorlog, cmaes, alembic, optuna\n",
            "Successfully installed Mako-1.2.4 alembic-1.10.3 cmaes-0.9.1 colorlog-6.7.0 optuna-3.1.1\n"
          ]
        }
      ],
      "source": [
        "%%bash\n",
        "pip install codecarbon\n",
        "pip install dash\n",
        "pip install dash-bootstrap-components==0.13.1\n",
        "pip install netCDF4\n",
        "pip install fire\n",
        "pip install optuna\n",
        "\n",
        "mkdir code_carbon"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "id": "Qw688STi6Z6k"
      },
      "outputs": [],
      "source": [
        "%matplotlib inline\n",
        "import random\n",
        "\n",
        "import numpy as np\n",
        "import optuna\n",
        "import pandas as pd\n",
        "import scipy.stats\n",
        "import sklearn\n",
        "import sklearn.ensemble\n",
        "import torch\n",
        "import torch.nn as nn\n",
        "import torch.nn.functional as F\n",
        "import torch.optim as optim\n",
        "import torchvision\n",
        "import xarray as xr\n",
        "from codecarbon import EmissionsTracker\n",
        "from google.colab.output import eval_js\n",
        "from matplotlib import pyplot as plt\n",
        "from scipy.stats import pearsonr\n",
        "from sklearn.metrics import mean_squared_error\n",
        "from sklearn.model_selection import train_test_split\n",
        "from torch.utils.data import DataLoader, Dataset\n",
        "from tqdm import tqdm\n",
        "\n",
        "# Set seeds for reproducibility\n",
        "# Per PyTorch, results may not be reproducible between CPU and GPU executions,\n",
        "# even when using identical seeds\n",
        "random.seed(0)\n",
        "torch.manual_seed(0)\n",
        "np.random.seed(0)"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "VODtRQ9-_JA8"
      },
      "source": [
        "<a name=\"climate-impact\"></a>\n",
        "# **1. Climate Impact of Machine Learning**\n",
        "\n",
        "Machine learning models can have a wide-ranging and potentially subtle impact on our climate. We categorize potential emissions from ML models into three broad levels:\n",
        "* System-level\n",
        "* Application-related\n",
        "* Computing-related\n",
        "\n",
        "This tutorial will focus only on the computing-related impact of ML models but we'll discuss each of these categories to capture the complete picture of machine learning’s role in the climate crisis."
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "Xn5FtR3p_lGx"
      },
      "source": [
        "<table style=\"text-align: center;\">\n",
        "<tr>\n",
        "  <td>\n",
        "   <img src=\"https://media.springernature.com/full/springer-static/image/art%3A10.1038%2Fs41558-022-01377-7/MediaObjects/41558_2022_1377_Fig1_HTML.png\"/>\n",
        "  </td>\n",
        "</tr>\n",
        "<tr>\n",
        "  <th>Fig. 1 from [1]</a></th>\n",
        "</tr>\n",
        "</table>"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "BRVKsxch_dku"
      },
      "source": [
        "## **1.1 System-level impact**\n",
        "\n",
        "System-level impacts consider the societal implications of ML. These types of impacts may have the greatest potential for GHG emissions. For example, implementing a recommender system that increases the demand for physical goods also increases the emissions from the manufacture and eventual disposal of these products. Shifting to autonomous cars could reduce demand for public transit and simultaneously increase emissions&mdash;through the material extraction and car manufacturing process and from operating the vehicles if the cars aren’t electrified."
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "wFo4QoxsCPGp"
      },
      "source": [
        "## **1.2 Application-related impact**\n",
        "\n",
        "Application-related impacts consider _how_ a model will be used in the wider context. Does the model enable carbon emitters or hinder the expansion of renewable energy? For example, machine learning has been used to assist in identifying new oil fields for exploration [2]. AI has also helped to identify and track livestock in factory farms, an industry that accounts for about 9% of global GHG emissions [3].\n",
        "\n",
        "Depending on the scenario, these application-related impacts can far exceed the emissions from hardware and computing. Of course, ML can also be used to _reduce_ GHG emissions. For some hands-on examples of these types of situations, check out [CCAI’s other tutorials](https://www.climatechange.ai/tutorials?)."
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "tI4aDSHmCPJQ"
      },
      "source": [
        "## 1.3 Computing-related impact\n",
        "\n",
        "The computing-related climate impact of ML models can be divided into _embodied_ and _operational_ emissions. Embodied emissions represent the climate impact of the hardware involved in machine learning, such as material extraction and manufacture. Operational emissions are those associated with creating and productionalizing an ML model, including data storage and processing, model training, and model inference.\n"
      ]
    },
    {
      "cell_type": "markdown",
      "source": [
        "<table style=\"text-align: center;\" width=\"50%\">\n",
        "<tr>\n",
        "  <td>\n",
        "   <img src=\"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAfUAAAExCAYAAABlMTuzAAABW2lDQ1BJQ0MgUHJvZmlsZQAAKJF1kE9LAlEUxc/YhKEuLNrlYloKVqJS68kgghaDZv9243MahVEfMyMRtGgVFC1a9QmkbyDWqj5Au6AgrE2LNu0CNyXTfVqpRQ/uuz8O513Ou4BvTOfckgGUK66dWV5UNre2Ff8LZITgxxSCOnO4qmmrZMF3Hz7tO0ii386IWQeH4Xpo+qR8NFd9bUWOn/76h06gYDiM+gdVnHHbBaQYsbbrcsH7xJM2hSI+FWz2uC443+OLrmctkya+IQ6zol4gbhHH8gO6OcBlq8a+Moj0IaOSy4o5VBGoyCJHpWCd7gRSWPjHn+r606iCYw82SjBRhEsvVVI4LBjEK6iAYRYx4gTiVEmx59/762ulJWD+GfAZfY01gEvKN97oa9FH+koAuJrguq3/bFVqy85OMtHjYBMYPfO8tw3AHwU695733vS8zjkw8gBctz8BdqxiO6Q/VC0AAABWZVhJZk1NACoAAAAIAAGHaQAEAAAAAQAAABoAAAAAAAOShgAHAAAAEgAAAESgAgAEAAAAAQAAAfWgAwAEAAAAAQAAATEAAAAAQVNDSUkAAABTY3JlZW5zaG90lOP0FQAAAdZpVFh0WE1MOmNvbS5hZG9iZS54bXAAAAAAADx4OnhtcG1ldGEgeG1sbnM6eD0iYWRvYmU6bnM6bWV0YS8iIHg6eG1wdGs9IlhNUCBDb3JlIDYuMC4wIj4KICAgPHJkZjpSREYgeG1sbnM6cmRmPSJodHRwOi8vd3d3LnczLm9yZy8xOTk5LzAyLzIyLXJkZi1zeW50YXgtbnMjIj4KICAgICAgPHJkZjpEZXNjcmlwdGlvbiByZGY6YWJvdXQ9IiIKICAgICAgICAgICAgeG1sbnM6ZXhpZj0iaHR0cDovL25zLmFkb2JlLmNvbS9leGlmLzEuMC8iPgogICAgICAgICA8ZXhpZjpQaXhlbFlEaW1lbnNpb24+MzA1PC9leGlmOlBpeGVsWURpbWVuc2lvbj4KICAgICAgICAgPGV4aWY6UGl4ZWxYRGltZW5zaW9uPjUwMTwvZXhpZjpQaXhlbFhEaW1lbnNpb24+CiAgICAgICAgIDxleGlmOlVzZXJDb21tZW50PlNjcmVlbnNob3Q8L2V4aWY6VXNlckNvbW1lbnQ+CiAgICAgIDwvcmRmOkRlc2NyaXB0aW9uPgogICA8L3JkZjpSREY+CjwveDp4bXBtZXRhPgpojUa4AABAAElEQVR4Ae1dCfwVU/t/0quUtFkqUVkjLRRtSmXLVorsbZZshRBeZU9FhEgUKikhJGRLKsoSpbKWFiGVJdmKeM3/fOf/zn1nu9vcmbkzc7/P53N/d5ZznvOc75nffeac8yxlNEVCIgJEgAgQASJABGKPwDax7wE7QASIABEgAkSACOgIUKnzQSACRIAIEAEikBAEqNQTMpDsBhEgAkSACBABKnU+A0SACBABIkAEEoIAlXpCBpLdIAJEgAgQASJApc5ngAgQASJABIhAQhCgUk/IQLIbRIAIEAEiQASo1PkMEAEiQASIABFICAJU6gkZSHaDCBABIkAEiMC/CAERIAJEoBAE5s2bJxMnTrSwOProo6Vbt26WazwhAkQgeASo1IPHmC0QgUQjsHTpUnnooYcsfaxYsSKVugURnhCBcBCgUg8H58S3Mnr0aJkzZ07afm6zzTZSr149OeCAA+TAAw+URo0apS3LG0SACBABIuANASp1b7ixlg2Bt99+W6ZOnWq7mv70nHPOkXvuuUd22GGH9IV4hwgQASJABPJCgEo9L7hY2C8Exo0bp8/sFy1aJFWqVPGLLfkkGIH//Oc/8uCDD8o///yT6iWW+c8999zUOQ+IQKkjQKVe6k9AEfu/atUqufHGG/UZexHFYNMxQeCPP/6Qfv36WaStWrUqlboFEZ6UOgJ0aSv1JyDA/mMGfsQRR8ghhxwilStXdm3p/vvvl19++cX1Hi8SASJABIhAfghQqeeHF0vngUDLli3l9ddflwULFshXX30lLVq0cNT++++/5bPPPnNc5wUiQASIABHIHwEuv+ePGWt4QACz9t69e8t7773nqL1mzRpd4c+aNctxv1OnTrrF/JIlS/Q9+IULF8rhhx8uMLQz09y5cwUfvCAsW7ZM36dv3ry5zhcvF7vuuqu5uH6crr0tW7bIjBkzdH6Qbf/995emTZvKJZdcIjVq1HDwMV/45ptv5N5775XFixfL8uXLdYv/1q1bywknnCDbb7+9ztdcvlWrVtKhQ4eM8sBjIIj+m+XIdpypX9nqmu9n4pMJnzfeeEMfDzMvHP/2228ydOhQ/TJeGrEyZKYpU6bISy+9JB999JFgXPFMYDx69OghEyZMsKwSwUPjyiuvlG233dbMgsdEIF4IaCQi4AMCZ511lqaefMunY8eOFs5jxoyx3DfKqx9evdzFF1/suH/XXXdphx56qOW6MoxK8f3+++81t7YN3vhWP9KassxP1TEO3Npr0qSJpn7cLe0ZvNT+rTZ27FijuuMbsirF5Fq3TJkyWp06dRz3LrvsshQfN3mC7H+q4SwHhfbLYF8IH/VC5cDOGBfjG/gZtGLFCq1t27Zp69SvX9/13k8//WSw4DcRiCUCnKmrXwRS8Ah899138vDDD7s2tMcee7hex8Urrrgi7T3M0g4++GDBbDoT/fXXX3L66afLI488Ir169cpUVJ8RpyuwadMmOf/88wXGWaeccoqlGCKqZZJV/TroWxCWSjmcZOIZRP/tIvnVL7/42OVzO//555/luOOO01dK3O7jGlZzSEQgiQhwTz2JoxqRPr311lvSrFkz2WeffWS33XaT999/3yFZ+fLlRc2aHNdzuTBo0CCHQsfSKZbsu3fvLtWrV0+xgTvU2WefLW+++WbqmteDPn36yJdffpmq/vnnn4uaJabOwzoIuv9+9csvPrniiq0ZbH2QiEApIsCZeimOekh93rx5s8APPRNB0WLmm40QpGa//fZLGdVhdj5q1ChLtX/961/y8ccfy7777qtfRxnUgSsUCLPl++67Tw477DD9PN0fWOrjpQD8nnvuOccMGzPB4cOHC6LogUaMGCG///67g51abpeuXbvKn3/+KZMmTdL3fx2FcrwQZv8Nkfzqlx98YNNw/PHHO2wS1LaGdOnSRRcZL5AbNmyQadOmGV2wfKMcXiDxsolgSSQikEgEYrlpQKEjh0C2fW31z+PYw1TKV/v6669TfXHbU0a9W265RVNW8no5fCsXOE39cDv4oSz28c2fsmXLWsqplQFNBS/Rebm1t8suu2jqZSAl08aNGzUV0tbCA+0oo6xUGaVwHPdRx7w/iz1etXLgKJdtTz3I/qc6kObAr375xUdtNzjwg52DmR599FFHGWCotn5SxTD+/fv3dy1nHrNUBR4QgRghwOV39R9PCh8BzJqwHI9l+UzUrl07uf7660UpZ70YvjFr/eSTT1yrvfrqq2L+YNndTJg1r1+/3nzJcnz11VcLZtgGVatWTS688ELjNPWNJCbgjf16rA7YqW/fvpYViL322kuuuuoqe7Gs52H33xDIr375xceQK9u32xZP7dq19a0Xoy5m97fffjtDFBuA8DtRCHD5PVHDGa3O1KpVS9/fNqSCAkciF3z23HNPwY9rNlKW765F1q5d63o9l4tQ6pDNjdyuQ1Y7YUkfS+7ws9+6dav9tr7sb7+444472i9lPQ+7/4ZAv/76qy/98ouPIVe2b7dnCjYdcFczU7ly5XSlDvlIRCBJCFCpJ2k0I9aXxo0bi3JjK0iq3Xff3bV+vXr1HNexpzpkyBDHdfuFTIZ52EM/88wzLVXc9mhh4FehQgXdpxkvAuvWrbPUeeqppwSzbDNhBSFfCrv/hnwwMvSjX37xMeTK9t2wYUNHEeyh40WuZs2aqXtYacm0YpMqyAMiEDMEqNRjNmClJq59hmX0HwZwdoJh3JFHHplaqsfyOJQ0loANwgwNSUDSETLNIQDJddddpytsJJ4ZP368o3j79u1TQUoOOuggh1KH+16DBg30gDuYyd96663yzDPPOPhkuxB2/83y+NUvv/iYZUt3jLbshOfgpJNO0l0aYUQ5b948PXiROTGMvQ7PiUBsEYjR/j9FjTACboZy9uAz2cR3M1xDwBo3Usummpp5OYydlFLXVKQwTUWX07p16+a4D6Mtg9zaU//IqTpqKTd1bL6OY2VFb7DRlBV+2nKZeIBPNkO5IPuf6kCaA7/65RefXAzlVNQ4TW3zeB4PGsqleRh4OTYIWDea1K8MiQjEAYFKlSrJ4MGDHaIi1nxvFY4WS99PP/204z5m4bmS+i92LYqwswhCY5B6OZCjjjrKOLV8p+NhKeThJIz++9Uvv/jkAtN2220nkydPTq3W2OsENR72dnhOBIqFAJV6sZBnuwUjgCAj2EOHP3kuhKAx9v3yXOqZy8CSGsvoWMY3CMZZjz32mMVq3rgX5HfQ/ferX37xyRVLxCFwe+HLtT7LEYE4I0ClHufRi5DshsuZWaRcla1Rx6282zWjPPabBw4cqEeJg4VzOoJRHfbWVdz2dEVS18EPs3z7XjbO4aYG9zW35DBI9IJ7yv/ZMUtElDu1FeFqxGfGza2vbtcMYYPov8Hb+PajX+DlBx/01z4uZvwMmfF97bXXyuzZs129EGAACDsJu6cDXj7s/M08eUwE4oBAGWwUxEFQykgEsiGA+PJQrMjIhRztUPSwdEeWM/PM2uADJW1EhTOuwVofS+uIGoeMcggHCx5wwzOHnTXKu32jbVhXI2McXiiQFQwZyIKmfPufrzx+9csvPrnID2O4lStX6uMB9zW4CGZ6AcyFJ8sQgSgjQKUe5dGhbIEikEmpB9owmRMBIkAEAkKAy+8BAUu2RIAIEAEiQATCRoBKPWzE2R4RIAJEgAgQgYAQoFIPCFiyJQJEgAgQASIQNgJU6mEjzvaIABEgAkSACASEQG4OvgE1TrZEoJgIIKSs3WXswAMPLKZIbJsIEAEiUBACtH4vCD5WJgJEgAgQASIQHQS4/B6dsaAkRIAIEAEiQAQKQoBKvSD4WJkIEAEiQASIQHQQoFKPzlhQEiJABIgAESACBSFApV4QfKxMBIgAESACRCA6CFCpR2csKAkRIAJEgAgQgYIQoEtbQfCxMhEoHIFNmzbJmjVrBAlH8EG2sB122EEqV64sdevW1b8Lb4UciAARKAUEqNRLYZTZx8ggsHXrVnnnnXfkjTfe0FPGfvrpp4LsapmoZs2a0rBhQz0lbIcOHaRFixYO//pM9XmPCBCB0kGAfuqlM9bsaRERWLBggUyYMEGefPJJ2bhxY0GSIDf5GWecIb1795YmTZoUxIuViQARSBYCVOrJGk/2JmIIvPbaazJkyBB9Vh6EaB07dpRBgwZJ27Ztg2BPnkSACMQMASr1mA0YxY0HAsuXL5d+/frJzJkzQxG4S5cuMnLkSKlTp04o7bERIkAEookArd+jOS6UKqYIaJomw4YNk0aNGoWm0AHVc889J/vvv7+MGjUqpshRbCJABPxAgDN1P1AkDyKgEIDBW/fu3UNV5m7An3TSSTJu3DipUqWK221eIwJEIMEIUKkneHDZtfAQWLFihWB/e9WqVeE1mqElWMu/8sorUrt27QyleIsIEIGkIUClnrQRZX9CR2DJkiVy9NFHZ3VNC1sw7K/PmjVL9t5777CbZntEgAgUCQEq9SIBz2aTgcDKlSvl0EMPlQ0bNkSyQ/Xq1ZO3335batWqFUn5KBQRIAL+IkCl7i+e5FZCCPz444/SvHnzyCy5p4MeRntQ7JUqVUpXhNeJABFICAK0fk/IQLIb4SIAK/eePXtGXqEDlY8++kguvPDCcAFia0SACBQFASr1osDORuOOwIgRI+Sll16KTTcmT54s48ePj428FJQIEAFvCHD53RturFXCCKxevVoOOOAA2bJlS6xQqFatmnz++eeyyy67xEpuCksEiEDuCHCmnjtWLEkEdAQuvfTS2Cl0CP7TTz/J1VdfzVEkAkQgwQhwpp7gwWXX/Edg/vz50qZNG/8Zh8QRaV0//vhjadCgQUgtshkiQATCRIAz9TDRZluxR+DWW2+NdR9g4Dd06NBY94HCEwEikB4BztTTY8M7RMCCAHKfYy897lS2bFn56quvZNddd417Vyg/ESACNgQ4U7cBwlMikA6BiRMnprsVq+v/+c9/ZNKkSbGSmcISASKQGwKcqeeGE0uVOAJYtt59991l7dq1iUACAWmWLl2aiL6wE0SACPwPAc7U/4cFj4hAWgQQwCUpCh2dRH+++eabtP3lDSJABOKJAJV6PMeNUoeMwBtvvBFyi8E3l8Q+BY8aWyAC0UaASj3a40PpIoLAvHnzIiKJf2IksU/+oUNORCCeCFCpx3PcKHXICHz22Wchtxh8c4guRyICRCBZCNBQLlnjyd4EgACsxStWrChbt24NgHvxWCJcbK4pY5GR7rHHHhPM7n/44QepWrWqnqHupJNOkv322694nWDLRIAIWBCgUrfAwRMi4ETgu+++kxo1ajhvJODKX3/9Jf/6178y9mTMmDEyYMAA+e2331zLHXjggXL66afrn7p167qW4UUiQATCQYBKPRyc2UqMEVi5cqXsvffeMe5BetERDx6z7nQ0ZMgQue6669LdtlxHCNqWLVtKr1695KyzzmL+dgs6PCEC4SBApR4OzmwlxggsWbJEMBtNIiGyHPzv3QhL7e3atZN//vnH7XbGazvssIP06NFDLrroImnYsGHGsrxJBIiAfwjQUM4/LMkpoQhsu+22Ce2ZSLly5dL2bfDgwZ4UOhj++uuvMnr0aEGQm7Zt28oTTzwhsE0gEQEiECwCnKkHiy+5JwCBr7/+WurUqZOAnji78Pvvv+tGgPY7uF6lShVfFfFee+0l11xzjb48n+llwi4Lz4kAEcgdAc7Uc8eKJUsUASi3JBJWIGDV70ZffPGFrwodbcA24fzzz5c999xT7r77bsGLA4kIEAF/EaBS9xdPcksgApUrV5add945cT3DzLkYhHC7V1xxhdSrV09GjhyZOFfBYmDKNomAgQCVuoEEv4lABgTq16+f4W48b2Xq0z777CPbbBPszwP83fv37y/777+/PPXUU/EEkVITgYghEOx/bcQ6S3GIgFcEmjRp4rVqZOtl6tP2228vRx99dCiyr1q1Sk477TRp0aKFvPXWW6G0yUaIQFIRoFJP6siyX74icPjhh/vKLwrMOnTokFGMG264IfDZulmABQsWyGGHHSZnnnmmrF+/3nyLx0SACOSIAJV6jkCxWGkj0L59+1AVXNBoV6hQQVq1apWxGdwfOnRoxjJB3JwyZYoeehYucV585IOQiTyJQFwQoFKPy0hRzqIiUL16dck2sy2qgHk23qlTJylfvnzWWnBBGzdunMBYMEz6+eefpW/fvvqLRxKT6YSJJdsqLQSo1EtrvNnbAhDo2bNnAbWjVTWfvpx99tmCyHOwVG/evHmoHcGSfNOmTWXEiBGctYeKPBuLKwIMPhPXkaPcoSMAv2oEodm4cWPobfvZIMLCwjgtWyKXdG2iLiLE4fPRRx+lK+b7dUSmmzBhgu7n7jtzMiQCCUGAM/WEDCS7ETwCsAi/7LLLgm8o4BauvvpqzwodoiF4zMCBA2Xp0qWCuPiI745Y70ETLOMPOuggur8FDTT5xxoBztRjPXzRFV7TNPnwww8FkcmQsrNs2bK6oRl8n92O3a6Zy+Z7v1q1anqYU78RQlazPfbYQ7DnG0eqVauWPkvfbrvtfBUfsd4nTZokDzzwQCizd7xIICpdLnYBvnaUzIhA1BFQP74kIuArAuPHj9dUXm1NPftF/ey6667apZdeqn377be+9u/ee+8tar8KwXXy5Mm+YuHGbNasWZpyAQwcI5U5T1uxYoWbCLxGBEoWAc7Uo/7WFSP51H+RnHPOOfq+Z5TERux27P8ec8wxvoiFbGMwGFu0aJEv/MJiAl97pXDDak7effddQT72F198MbA24ZUwdepUSWIcgcBAI+NEI8A99UQPb7idu/nmmyOn0IEAlsq7dOmiKxk/EMFWwGOPPZY2GYofbfjNA9sRcE0Lk1q2bCkvvPCCvg3TsWPHQJqG0SJ4jxo1KhD+ZEoE4oYAZ+pxG7GIyov0pIgX/ueff0ZUQpEDDjhAN+yCUvaDYIkNd6840PTp06Vz585FFfW1116Tq666SjewC0KQCy64QO6//37dZiMI/uRJBOKAAGfqcRilGMiIKGBRVuiA8JNPPvF1+bl3794yYMCAyI/OsGHDiq7QARJiycN48pFHHpGaNWv6jtuYMWOka9eusnnzZt95kyERiAsCVOpxGamIy/n+++9HXML/F2/27Nm+yjl8+HDp1auXrzz9ZIYsaP/+97/9ZFkQL3g0wO7i888/F8ysy5QpUxA/e2Us9x955JHy448/2m/xnAiUBAJU6iUxzMF3ctOmTcE34kMLyhLeBy7/YwGlhJnn+eef/7+LETlCiFe4fUWRYLz44IMP6lnZGjRo4KuI77zzjrRp00a++eYbX/mSGRGIAwJU6nEYpRjIGMRyahDdhuW634Q9eiz93nTTTb7PPL3ICnnuueceue2227xUD7XOoYceqi/J33jjjb7uhWMloF27dnp421A7xMaIQJERoFIv8gAkpfm4JDvxy0jObdygmGbMmCE77bST2+1QrtWuXVuwxRCnyHflypXTX4gQMQ6BffwihLOFYl+9erVfLMmHCEQeASr1yA9RPAQ89dRTpUaNGpEXFnu6QdKxxx6rW9h369YtyGYcvLENgL39xYsXC2Kkx5GQ6hXyd+/e3Tfxv/zyS12xQ8GTiEApIBDsL1wpIMg+6ghUqlRJDxHqt+GT3/AGrdQhr4pkpwdEefXVV6Vx48Z+d8HBr0WLFvLmm2/qMQKKuUrgEMzDBaR4RQwAfJDz3Q+Cu+VRRx0l69at84MdeRCBSCNApR7p4YmXcHAngtEYllOjSkEuv9v7DBcuzDyff/5532fPeHmCopo5c6YeVAeGYUkizNbnz58vKtywL93CTB14xT3Dni9gkEmiEaBST/Twht85BGNBOk7k665atWr4AmRpMYyZulkEKN9OnTrpM2kVp1zfO27WrJme3MZcLpdjpEpFlDb4na9Zs0YQzAXuW0klZGT74IMPpH379r50EXEKsD2CBEMkIpBUBBhRLqkjG4F+/fPPP/L999/LL7/8krM0c+fO1d3DEEc+CLr44ov1qGNB8M6HJ1wAMRP99NNPZdmyZbqSRqYzfPAigFSmWIrGTLV+/frSsGFDad26dSgpTvPpRxhl//77b93wb/To0b40d9xxx+mrJ2Gu2vgiOJkQgRwQoFLPASQWCQcBuCFBcSG9aVDUr18/ue+++4JiT74BIoAVCuRx94P4HPiBInlEEQEuv0dxVEpQpu+++04wgwpSoQPWsJffS3AoA+vytddeqyelwTZEoYQEMHy5KxRF1o8iAlTqURyVEpNpy5Yt+r5zGP7EXHKN98MFm41p06b5Yhl/+eWX63YJ8UaE0hMBKwJU6lY8eFYEBBBidcGCBaG0zJl6KDAH2sgJJ5yg52gv1OUN0QXPPPNM3Z4hUIHJnAiEiACVeohgsyknAlgCnTRpkvNGQFeo1AMCNmS2hx9+uG7stt122xXUMhK/nHzyyZHPMFhQJ1m5pBCgUi+p4Y5WZ+fNmydXXnllqEJx+T1UuANtDO58zz33nJQvX76gdhYuXCgwnCMRgSQgQKWehFGMYR/Wr18vp5xyivz111+hSs+ZeqhwB95Yx44d5dlnn5VCjecefvhhefLJJwOXlw0QgaARoFIPGmHydyAAH3TEKYdiD5uo1MNGPPj24DUxduzYghu68MILmdWtYBTJoNgIUKkXewRKsH2kBUU0tEIIwVm8EJffvaAW/Tqwir9Jpb4thBAQqEePHoKgSSQiEFcEqNTjOnIxlXvp0qUCf+NCaLfddpMpU6Z4YsGZuifYYlEJqW/POeecgmRFYpzhw4cXxIOViUAxEaBSLyb6Jdb2n3/+KWeccUZBlsZwY5o+fbrUrFnTE3qcqXuCLTaVxowZIx06dChIXsz4EbqXRATiiACVehxHLaYy33LLLXqs80LEnzBhgjRt2tQzC87UPUMXi4owmIPBG1ZzvBJePs877zwJKv+AV7lYjwjkggCVei4osUzBCCxZsqTgZU1EADv11FMLkoVKvSD4YlF55513lqeffrqgFMBwt/QrgUwsQKOQiUGASj0xQxndjiByF2Y+yLbllVq0aCG333671+qpelx+T0GR6AM8LyNHjiyoj7D9WLduXUE8WJkIhI0AlXrYiJdge7B2R15sr1S9enV56qmnZNttt/XKIlWPM/UUFIk/gIsabDi8EtLgXnPNNV6rsx4RKAoCVOpFgb10GkX2tZtvvrmgDsMHuU6dOgXxMCpzpm4gURrfWEIvZH8dIYzffvvt0gCLvUwEAlTqiRjG6HbihhtuEMx4vNJZZ52lx+b2Wt9ejzN1OyLJPq9atarAuLJMmTKeOgpjuUsvvZS+657QY6ViIEClXgzUS6TNjz/+WBB+0ythhoW8134SlbqfaMaD1xFHHCGXXXaZZ2ERGz7MpEOeBWVFIqAQoFLnYxAYAgMGDBAYyXmlhx56SDDT8pO4/O4nmvHhNWzYMNl33309C4zANlu3bvVcnxWJQFgIUKmHhXSJtTNnzhx59dVXPff6tNNOk2OOOcZz/XQVOVNPh0yyryNFayEual9++aUgsE2h9Ntvv8mGDRsK8gQpVAbWTzYCVOrJHt+i9a6QONxVqlQRWMwHQVTqQaAaD55Yhi/EGv7WW2+V33//Pe/Obt68WbBSsP/++0vlypX1aIjbb7+9IHXsCy+8kDc/ViACmRCgUs+EDu95QmD27Nkyd+5cT3VRaejQoVnDwJYrV84Tf8zYSKWLwF133SV4afRC8OTId7b/2WefSaNGjWTgwIHy+eefp6LUYSl/1qxZ0rlzZzn99NMLCp3spS+sk1wEqNSTO7ZF61khs3T8AF5wwQVZZd91112zlnEr4LWeGy9eix8CyBkwePBgz4LffffdOStgBK7BbHzVqlUZ20NY2549e2Ysw5tEIFcEqNRzRYrlckIAWa7w8Ur40czFmG3HHXeUxo0b59UMksG0bNkyrzosnDwELrroItlnn308dQyKGi5yuRAs7r/99ttciurBlZ577rmcyrIQEciEAJV6JnR4L28EsLzplTp16iTY98yV4D+cDyFUbcWKFfOpwrIJRABJX4YMGeK5Z0jNms2rA1tQU6dOzauNQsPa5tUYCycWgTIquIKW2N6xY6EisHLlSt1t6J9//sm7XczOP/nkE6lfv37OdfHDeuyxx8rMmTOz1tl7771lwYIFUq1ataxlWSD5COBnr3nz5p7DF0Nhd+vWzRUo5Dg46KCDBHEa8iHYiSBQk1d7kXzaYtnkIsCZenLHNvSeYabhRaFDUESOy0ehow5eBJCNK5vrW8OGDXXFT4UO1EhAABHmbrvtNs9gZAqKBGO6fBU6BIHx3MaNGz3LxIpEAAhwps7nwBcEfv75Zz3GNvxw8yUsh8JKGLNpL4RZ1xNPPKFbJiNOt/FiAaO7s88+Wy6++GIpX768F9ask3AE2rdv79lTA4r7gAMOsCD0/fff66tVmzZtslzP9QT/P3B3IxEBrwhwpu4VOdazIDB58mTxotDBpEePHp4VOupj1gX/47feekv++OMP3TgJvsFLly4V5GCnQgdKJDcECsnC5jZbR7pWrwp9v/32o0J3GyReywsBztTzgouF0yFw8MEHC2Jk50sIBrNs2bKClHq+bbJ88hCAfcW0adPklVdekTVr1ugvcphFn3rqqdKsWbOMHW7SpIn+ApixkMvNSpUqyfr161OK+P333xfkcfdqpoTtgEJeMlxE5KVSREA9gCQiUBACixcvhrGlp8+JJ55YUNusTASUMtVUXPe0z9/JJ5+srV69Oi1QKllL2rrZnuuJEyfqfNWWj6YUumc+9erV07Zs2ZJWRt4gArkiwOX3UnyT87nP48aN88zxyiuv9FyXFYkAbCjatm0ry5cvTwvGM888o4doRVAkpTgd5ZBnoG7duo7ruVxQSl0v9uijj8p7772XSxXXMgiLzGiHrtDwYp4IcPk9T8BY3IoA3HcQpevHH3+03sjhDEv2WLIkEQEvCCAOOzKv5RrgBW2oGbEgwFGXLl0sTSI08aBBgyzXcjnB9hFsNxBfAYlavBDcMl966SUvVVmHCDgQ4EzdAQkv5IMA4ld7Uehoo2/fvvk0xbJEwIIAsqblo9BRGdnWunbtKh07dtRtOQyG8JKAF0a+BE8L8PKq0OGTzqAz+aLO8pkQoFLPhA7vZUUAfuJeCNmqYMREIgJeEYBhnFd67bXX9EQrMEyD10atWrXk+OOP98Ru7dq1nuqh0oABAzyHrPXcKCsmGgEuvyd6eIPtXCFL70ja8uCDDwYrILknGoEaNWoIMqcVSkjyc8cdd+hpURGqOCzafffd9fgM9EsPC/HSaCf/9abSwIW9zAGBOXPmeF56P/fcc3NogUWIQHoEEJ/AD8ISPiIatm7dWrbddlv566+//GCblQfyJFChZ4WJBfJEgMvveQLG4v9D4Pnnn//fSR5HMG465JBD8qjBokTAiYDXTGtOTv9/BZb0YSl0pGRNFzs+nXy8TgRyQYBKPReUWMYVgVdffdX1eraL3EvPhhDv54KA3YI9lzpRKIPVgPvuuy8KolCGBCJApZ7AQQ2jS7AizuQbnEkGKvVM6PBergjALgMGbnGj/v37C0LCkohAEAhQqQeBagnw9DpLx48ZEq2QiEChCCBM61NPPSUVKlQolFVo9WGUd/3114fWHhsqPQSo1EtvzH3pMVyCvFDnzp29VGMdIuCKQJs2bfQsa37vr7s25sPFO++8U3bYYQcfOJEFEXBHgErdHRdezYIAMqJ5oWy5z73wZJ3SRgBGl0iDioQomL1Hldq1a6dnE4yqfJQrGQjQTz0Z4xhqL1asWOEpYAZ+cDdu3Ki7DYUqMBsrGQTgnoaALlOmTIlUnxGtbtGiRdx6itSoJFMYztSTOa6B9gquP14I8bFh+UsiAkEhgD3rxx9/XN58801BStWoUL9+/ajQozIYCZeDSj3hAxxE97wq9Q4dOgQhDnkSAQcCyNy2cOFC3XWsWrVqjvthXkDkO2SIIxGBMBCgUg8D5YS18e6773rqEYyaSEQgLATKli0rmCHD9bJPnz6CjGrFoOHDh0uVKlWK0TTbLEEEuKdegoNeSJcR7x2hLbdu3ZoXG9TZtGmTp0xYeTXEwkQgDQKYuUPJe30pTcM242WEnp03b574FdI2Y2O8SQQUAsV5dSX0sUVg2bJleSt0dLZFixZU6LEd9WQI3qxZM8HW0fjx4wVL4kETVgruv/9+KvSggSZ/CwJU6hY4eJINgY8++ihbEdf7zZs3d73Oi0QgTAQwY+7du7e+JH/55ZcH+qKJiHcHHnhgmN1jW0SAM3U+A/kh4FWpN27cOL+GWJoIBIhA5cqVBVnSlixZIvDK8Jt22mknufXWW/1mS35EICsCnKlnhYgFzAh89tln5tOcj6PkXpSz0CyYeAQaNGggr7/+ukydOlXq1KnjW3+HDRsmxba6960zZBQrBGgoF6vhKr6wBx10kCxevDgvQcqXLy+//fZboEudeQnEwkTABYHNmzcLlDFCuf7xxx8uJXK7hK0mGOPROC43vFjKXwQ4U/cXz8RzQ3a2fAn50xFRi0QEooxAxYoVZfDgwfLJJ5+I1xwFcJsbNWoUFXqUBzrhslGpJ3yA/eweXNLwyZf23HPPfKuwPBEoGgJ4XqdPny4vv/yy4IU0H4JxHGLRk4hAsRCgUi8W8jFsd/Xq1Z6kplL3BBsrFRkBJB+CYWiuiWKwNYWlexIRKCYCVOrFRD9mba9bt86TxFTqnmBjpQggUK5cObnmmmsEBqKdOnVKK9GJJ54os2fPFizhk4hAMRHgRmcx0Y9Z2z/88IMnievWreupHisRgaggsNtuu8nzzz+vz9xffPFFWblypWzZskUP/3raaacJ0qqSiEAUEKBSj8IoxESG77//3pOku+yyi6d6rEQEooZAo0aNmG0taoNCeSwIcPndAgdPMiHgdaZOpZ4JVd4jAkSACPiHAJW6f1gmnpNXpb7zzjsnHht2kAgQASIQBQSo1KMwCjGR4ffff89bUgSeqVSpUt71WIEIEAEiQATyR4BKPX/MSraGlyhbSLlKIgJEgAgQgXAQoFIPB+dEtOJFqVeoUCERfWcniAARIAJxQIBKPQ6jFBEZqdQjMhAUgwgQASKQBgEq9TTA8LITga1btzovZrnCmXoWgHibCBABIuAjAlTqPoKZdFZIVpEvaZqWbxWWJwJEgAgQAY8I5P8r7bEhVos/Al4yrf3nP/+Jf8fZAyJABIhATBCgUo/JQEVBzG233TZvMajU84aMFYgAESACnhGgUvcMXelV5Ey99MacPSYCRCBeCFCpx2u8iiotAsnkS5s3b863CssTASJABIiARwSo1D0CV4rVqlSpkne3N23alHcdViACRIAIEAFvCFCpe8OtJGtVq1Yt734jPeVff/2Vdz1WIAJEgAgQgfwRYOrV/DEr2RpelDrAwmydSV1K9rFhx4lA6Ahg2+/yyy8Xs0vtjjvuKMOGDQtdlrAbpFIPG/EYt+dVqSMPO5V6jAe+xEU3FEOZMmVKHIn4dB/Jp8aOHWsRuFatWlTqFkR4UvII7LTTTp4wWLt2rTRo0MBTXVYiAmEgANfLRYsWydy5c+XTTz+VZcuWyZo1a+SXX34RIzshsg3usMMOUq9ePdl3333lgAMOkHbt2knTpk3FS2Amv/qFlMgXX3xxWnZly5aVOnXqyD777CNt2rSR/fbbL21Z3og/Apypx38MQ+vB7rvv7qmtb775xlM9ViICQSIARf7aa6/JxIkT5eWXX5aff/45Y3NQ8PjgJXX+/PmpslWrVpXjjjtOevbsKUceeaRAiYZJ2N6aOnVqTk1itaFz584ycOBAad68eU51WCheCNBQLl7jVVRpqdSLCj8b9wmB3377TUaMGKHPXqGMn3jiiawKPVPTUKqPP/64HHPMMVK3bl25++67JaqunNhKmD59urRu3Vruu+++TN3ivZgiQKUe04Erhti77rqrp2VGLGOSiECxEcDMfNSoUfry+YABA+Tbb7/1XSTM4q+44gpduT/wwAPyzz//+N6GHwyBxaWXXipXX321H+zII0IIUKlHaDCiLgoiysHYJF/6/PPP863C8kTAVwSwX37IIYfIJZdcIj/++KOvvN2YGfvcLVq0kCVLlrgVCfwabGC23377jO3ccccd8tJLL2Usw5vxQoBKPV7jVXRp99prr7xl+Oyzz/KuwwpEwC8EsMzcqlUr+fDDD/1imTOfDz74QFq2bCmYtYdNL7zwgm4DgP+/hx56SLD370Z9+vTRy7nd47X4IUBDufiNWVEl3n///eXNN9/MS4aNGzcK3drygoyFfUAAQY/OOeccmTRpkg/cvLP4448/dOv0BQsW6MrVSw4Fr63DKh/W7vi0b99eN5Kzv2RjGwI2ARdeeKGjGdgLTJkyRfcIwIrbd999J40aNdKN7LAKAct/c6InGBzaX55gOIjtDrsBIbblJk+e7Gjz5JNPlvr166eu5ytDqmKOB9giefLJJ2XhwoXyySef6J8///xT925o3Lix3tfTTz/ddetx1qxZ8t5771la6tSpkyDo1owZM3RvCvQTv5vACitFNWrUsJT3/UQZTpCIQM4IjBw5EgnS8/7Mnj075zZYkAgUioAyhtM6duyY93Pq5dnOp87xxx+vKRe5Qrtnqf/FF1+49vOdd96xlMOJWjlwLatc3Rxln3nmGU1tt7mWN/p82GGHacojIFVXKUfX8uqlIVXGODj//PMdZStWrKitW7fOKKJ5kQGV1cuHgzf6Yidgd+ihhzrKGv0zvtHP1atX26trypXQUbdJkyaaeplyXAcvtVqiKf95Bx8/LyDiDokI5IzAzJkzXR9W4+FP96327nJugwWJQCEIqFmWplzLPD2n6Z5fP68rK3lt69athXTRUjcfpY6Kbthst912Fp733ntvzvgpWwVN2Sno9dWsV4NSs+OFa2ZSqwOaShDlKKcM91LFvMoABrkodbU9oeElwi5runMVo0CzT07clHq6+ubrTz31VKqffh9QqfuNaML5KevenP8JzA/xqaeemnBk2L0oIAClgmfN/OxF8fiss87yDa58lbpaCnfFB4oQ9NVXX2kq0I6jjAq6o1100UWa8m/XlL+75f7hhx+e6o9ymbPcM/BXS/OpMlDexnXju3LlyqmXg0JlyKbU1ZagppbBHTIYsqT73mOPPSwrLV6VukqO5TrzTwFUwAGVegHglWpVr/8MpYoX+x0eAiq2d94/1Ol+wIO+fuedd/oCTL5K/ZprrnHFSNm96PKce+65jvu9evXS8MJkEJS7HZ+PPvrIuK0rfvt9taev31d75BoUuP3+zTffnKpfqAzZlLra23a0D3n23HNPTQXm0QYNGqTtttturmVuuOGGlJzplDr6h3vKbVBT0fxc+QDDIIhKPQhUE85TGYK4PqT2f1L7uXLzSTgy7F4xEXjrrbc0ZYzl6dm0P6thnCsDM81t3ztfDPNV6meeeaYDI+X6llLabsvnynBNt1GAnQI+Bx10kIPHkCFDUqKrSH2O+8BUGZVpQ4cOddxT7neWvflCZcim1FWIX4cMyihOwwuHQTh2K9e2bVujiOue+i677KIp47hUGawKKONCR3vK0DBVxs8DKnU/0SwRXrfeeqvjAc3lR9CPH7ASgZjdzBMBZW2sKXdLT89lLs9uUGWgLGEDUAjlo9QxG8fesL0/yjJbF+Hvv/923eu2l3c7h+GbmWBcZi937LHHui57m21u/JAhk1LHs6I8EByyKbdDs/j68XPPPecoh6Vzg9xm6m4rMPfff7+DT4UKFTT01W+in7p66kj5IQBXFi8E1x4SEQgCgdtuu01WrlwZBOtAeSJxjFICgbZhMEcUObiW/frrr8al1PcRRxyhHyMwD9y5vND69est1dTLv+UcJ3B527Bhg+U6IlX27ds3dc1PGVJMTQdwN1PK1HTl/w/dkk5Vr17dUQ4hgOEGl47cAnSpZX1HcfweGsmCHDcLuEA/9QLAK9WqiMwF/9dMD7YbNsgURSICfiOgXKDk9ttv95ttaPzUsrUgAEyQ6YmhKM844wxR3iuOfiHqHJQ9CDLg3K5sELwmW+4HuzJTy9Ry9NFH60lzHI2aLqj9a1Gz1tQVP2VIMTUdIIV0zZo1xf4SoizSRa0umEqKK16I758pK5+a3Yva4rDwmTZtmuUcJ8r639JvRwGPFzhT9whcKVdTy0+Cf9h8CGkf3d5W8+HBskTADQEkZ4nzKhBmfnfddZdb1wq6ds8994gycBMEUIESc1PoaKBfv36i9oH1tpDFDWll7QQZ1V566gMF/9NPP1k+yi3OXk0GDx7suGa+oKzJ9Rca8zW/ZTDzNo4bNmxoHKa+H3nkERk3bpw+i8eqxpgxYwQrQHZC8J1MhIx5V155pY4Nkgcp1zwZP368owqCAZkD9zgKeL3g93o++ZUGAipetGOPSD2Daa89/PDDpQEMexkqAkqxaDDyyvTsxeEerKXNQVzyATHdnnou/e7QoYOmFLalOexv2+uq2bMGVzgVTVKbMGGCq/X6888/b+FjnKhUrw5+Bn/wcqNCZci0p4723PbKDZlgbFmuXLm0Mr/++uspkd321A0++La7/pnvqfDFKT5+HtBQzk80S4yXCiuZ9sE3P7wnnnhiyrK2xCBidwNGQM2mcnoGzc9jVI/VLNETWl6VOiKpIfKenWC4B9eufHBSYWg1Nbu1s9LPVUIbV+WGOukMxQqVIZtSh2D4Xcqnjyjbo0cPSx+zKfV0/FU+gIINJC2CmE6o1E1g8DA/BPBPjLf3dCER8UD37t07sIc3P2lZOokI5BLiM90Pa9SuG37c+Y5TvkodgWVuvPFGV4VutK3iw7tGhnPDTO1Ra2YfdYOH+dstIFC2qGqFyJCLUkcUvFNOOSVnxQ5f/Z9//tncLVeXNjeMzNdq166tIYhXUESlHhSyJcR38eLF2gUXXKDhzVtZi2oqeYHWs2dPDX7DJCIQFALKijrnH2Tzj2pUj/FyDJ/mfGnVqlUZccByston10466SRN7XHrIVRzaUPZKejBU+BP74YZlpYRJMYIWpOJJxS0OYYA/NzNwWzS1fUqAxS2felb2QG4NoO49JkCaiEIDZbr3chtpo7gNe3atXNMdjC+yspfw5ZRkFQGzNWAkYgAESACsUIAmbWQPStJ9Oyzz0rXrl0j1SVku4PrnZqNy/Lly2XHHXfUs77BBQzuaGFQGDIgWx2ytKGvIGRWU8FndCPDdH2EK97o0aMtt2Fgp3z2Rc3q9QxuX375pZ51DgZ2bi5ylso+nNClzQcQyYIIEIHwEZgzZ074jQbcIvoUNaUOC21Yi7tZjAcMR4p9GDLgBQWfo446KtVuIQfwEoJLX9hEl7awEWd7RIAI+ILAxx9/7AufKDFJYp+ihG8pyMKZeimMcgT7iOW0BQsWiNqP15e7EA0Mfq/w60REK2XMIyqcpb70pUJp6stXrVu3FpUpKoK9oUjFQMBYJi1G20G1+fnnnwfFmnxLBAEq9RIZ6Ch0U/nhytNPPy0IzqCM6BxRq3KREcEqEARDpa6UNm3a5FKFZRKIACKeKQOtxPUM+7oq17ooP+nE9Y0dCgcBKvUCcEZYQRWIwMJBuXAJZpSk/yGA2QciM8GwqdDIX6tXr5YHH3xQ/yBC3WWXXaZHpDKHmfxfyzxKKgJ4QUwqITY7jNFIRMALAlTqXlD7b525c+cKYiKb6eCDD6ZS/y8gKv2gXH311frsPN848WZM0x0rVx5dqSN2tnIj0cNdKreZdMV5PUEIuCUlSUr3qNTjM5JHHnmkqIxvFoEPPPBAy3nYJ1ZpAmodJv+ZLFVhJYjY4Ng7PeaYY/RA9wGJQrYhIIAMSMOHDxdkaUJGpKBJBZqQ/v3763GbMYtv1apV0E2Sf5ERSLInbhAvwEUersQ2D0+FqHkrhKLU3377bX0fNZeRRaafK664QpRTv1SsWDGXKiwTIQS+/vpr3XcYYx42LV26VE80c8stt8i1114rSAxBSiYCMKJMKiW5b0kdsyj1K3IubUijeNVVV+lGUCqUXpSwoixZEMB2BJaeiqHQDdGQXQmpHI8//njdkt64zu9kIZBkxZfkviXrKYxmbyKn1A2YPvzwQ2nRooWouMbGJX5HGAFEwoJVugpzGQkpX375ZTn88MMTaSEdCYCLLAQUX9WqVYsshf/NIwWqWwpT/1six6QiUDSljghB2UIMYqauYogL95ii/fhBoavECLp/eZQkff/993XFDv93UvIQgA1O0iiJfUraGEW9P6HsqbuBgFnUK6+8Ips2bdIDkKjcsgLlYKd3331X7r77bj3pvP0ez4uPAAwgzzzzzMi+eCFCV6dOnWTmzJlCt7fiPy9+SoC43O+9956fLIvOC/HUSSLz5s2TiRMnWqBAyNVu3bpZrnk9UTnk5fLLL0dCsxQLuBEOGzYsdR7Xg6IpdQMwLKGplIOistrogGI/1E533XWXPgAqy439lmAfFx+VBUiPTAZL+ubNm+tL9ypnretqwKxZsxw/Bvjhh6X2jBkzdH5wx0JA/6ZNm8oll1wiKouPo+18LmC1AX7aCxcu1JMGIHEAIqfhh6lx48a6zEhOYe4jXnoWLVrkaOa8884TLNPZ6Z133pHZs2fbL4vKASwqQ5HjeqEXgJHK/BS5Gbq9X/PnzxeVRc7xI2Evx/N4IXDYYYfpHg/xkjqztPgdJInA6NXuLgzDab+UOoIXjR071gI1jLSToNRDSb2qon/hdcjyUfuvSBDnIKVcLeWMeirIi6Us0v258TXK4xspA1X0Mks9nLily2vSpIkjVZ7BS714aOoByImPytDjKId8x7nkfVY/UpoKrpKqj3R/hgzmb/WSkSpjHCC3uXrLd5TfeeedNRWowyjm27eKeqUpmwdHe2Y5o3b8yCOP+NZ/Mio+Al999VWsnr9c/h+UoXDxgY2ABPfff79jbFWgKd8kyyXfum+NhczIOfVVT14x6ZprrnFtHvujBiE+OIK8TJ482bjk+o344pj9Pvroo673zReXLFmSdgkZWwRIpYfwpvnSiy++KOqFQTBbzEZvvvmmPms3fPpPPPFEfRXDXg+p/dQPmuUyVgE+/fRTyzWcXHfddXoMdceNAi/gjTZuS5+XXnqpA7cCYWD1IiKA1Sf8byWFDjnkkIxpPpPST/YjWAQip9TTReOB/7NBWKLH0q+ZYHiHJfTu3btbctbCxenss88WKMxCqU+fPoLcuLkSDLSwVI79m1wJ0aTOOeecVJ0RI0Y4/K0RG/rmm29OsUQfzefGDSQ/ufDCC41T374RyS2Oy1RYckNYWVJyEIAhbVII22QkIlAoApFT6uXLl3ftkxFIBMp81KhRljII0weDqOeff14ee+wxfR/a7BaiVj8EhnjZqHLlynrQG8zo6tSp4yiOpPeIlJYr3XjjjbJhwwZHccQsR1hTvJzstttujvuIb3777bfr17Gn7/bPjtWH5cuX62WmTJmi2xPYGQ0ePDiQxBADBgwoOIa7XdawztWWhsCmgpQMBGCkiRf6uBN+984444y4d4PyRwCByCl1KDQ3qlu3rn4Z/ut2FzeEJYUiRohZfGAUhaV3M73wwgsWS0fzPRzD8Oyjjz4StZcjI0eO1C3yGzVqZC/marjmKPTfC2+88YbjFoziYPyGeOUIo4qXERjL2cls8IaydsttzM5vuOEGwTciqNkJ7eAHz28CRlCMcSY3vOLcn1KWvWbNmrrba9wxwOrcTjvtFPduUP4IIBA5pT5+/HhXWPbbbz/9OqzG3ejVV18V8wfKzkywNF+/fr35kuUYiUfMs/Nq1aq5Ll3DKtPO28LovyfIRuaW7/miiy4SWOgbhGMobTuhHYMwm7/yyiuN09Q3ssRh1uwWoAfL42ZL+lSlAg/AFysfcSZsxeRi4xDnPpaS7AgJHOdEPlhpTGdLVErjyL76g0DRXdrM3YAiu/fee82X9GO4MsA9DVRI6FgodbgtuJHbdSyT2wnKGnuzWKrPRHCPwwqCndz8UKtXr24vpu+pY0XCUMz4p3/44YctLyZQrvfcc4+jLlx9jjvuOMf1Qi/ARuCZZ54plE0k6sNdRnkkREIWClEYAnvttZfgZdm+LVcY1/Bqw2XWWIn00mo6F10kyUKcD7zEYoUNwb5gYIxlfiOzGCY72LL84IMP5Mcff9T/J4466ijX1UNDNi/uuUZd8/c333yj/94vXrxY30qEDRDSVp9wwgnmYlmPYciMLUgYCiPNMxI8YZXVcG3GFmYStmiyAmEUCMPa3s31zO7S9sQTT2hKeWMK6Pio2WhKTLXX7LjfrFkzTfl0Z/0oZazzcXNpUxHRUm0YB8ri3dGW2qvX4M4FcuNjdmlTS4OO+n379jXYp76vv/56R7m99947dd84gFudGz72ayr2ulHF1+8HHnggp/bt8kTxvFKlSprxPPgKEpkVBQH1w66pWBKxez6Voi3Y5dTtd6h3796aWuFzxQO/LXCdVZMoTb1MuJbB76FS+I6x9Oqea2ekYo9o22+/vWvbyn5KU6umjntuLm1qkqGpCZmjrPk3B67CdrfeJLu0Fc1PXbmjaBgkFSVIgy+1eRDMx/jxhU+6QdOnT3eUVXtRmpoVG0X046efflpTb2+pDwbfILd/ArSpssNpKna5pizQNbWvrvu5m2XBsdqzN9hkVeoq165DVrwUwF9a7fnrcqpUoa7tqHR+qXaMA/SxYcOGDp5mGbt06WIU9/1bRQHM2LZZjjgcm58J38Eiw9ARmDZtWqyeTygv5fJaME7pfs8y/Q+q1Q2tdu3aGfG68847LbIpu6S0Ey+3tlR8fk3ZBll44EQZ+WZs140XrtmVulrVzZmPchfU1EpEShYq9RQU3g7cZurpBs58HcFjXnrpJUujULhuM2Ao0AkTJmgqupymog45BlstwaT4ZPsnwD+bWQ7zsbKiz8jHPFNPFzwG/NQeoFauXLm07diD7RiNYkXCLI/5GDyVzYFR1Ndvte2g4YXE3F7cj91WTXwFjcxCRwA//HF5Ls0rkIUAle33zCseWAE1CJMdLyshe+yxh2VFTEX+TDtDzyanWakj8BAmfPY6aglfU1sxmlp61+y/45iUGESlbiDh8duLUoeCwmzbjdR+qGMw7YNrP1eBalKsvP4TqH19y5KUGx+zUkeDKoBM3rIqF7aUrG4HWN2w9w/nyoLWrbgv1/DG7dZmnK+pMMC+YEMm0UEAS8bt2rWL/LN6xBFH6Kt1fiDn9juE/0sVI0NfFcQs1e3/FEpPGQjrZdyW4bGCahCiWLrxUHZHmnLP1ZR7btrlfuWlY7DRZXLjg+V2KG0VV8NVWaOOWamfe+65Dnl69eqlqf3+VFtQ7va2lG2Bfp9KPQWTt4N8lXrbtm21TPvCyvpcUxbjmjL2cAyafRBxroLGWARP90/gVte4hqUqZaSXlY9dqWPJB/tTBp9s33gwlT+8pR37ycknn+zgh1k03l6DImWQ52gzW1+ifl8ZIWpYgSAlCwH8/6ggVpF9XjEDtu/xFjICbr9nCLdtEH6D3FYFlRuwUUTDyqD9/xUTK4OU263jvnKb1WDLYBCO3crh99wgrJja21FGbZoywjWKaCtWrNCU8bCjnFmpI6y3nY/KcKfBVsv4HHTQQY4y0BsgKvUU3N4OVNQnB7jmAVHuY1qbNm30t7R89pig+JWFZ1reWIrBErid3P4J8LaJN3z80JtlwzmWac0PncEP/xTmsjhOF1/88ccfz7h8BaMWN1mNtoxvZaXqWFZCu34t5Rnt2L/d3nrtfY/jufHmbu8vz+ONgPJ0iaRih0I32wj5gbLb79ltt91mYa1cgh2/VdjKMwgvQvb/X0OpK08e1wkUDGft5LblqNx29WIwMHZ7uYBdkZ2U66xDlhsZFgAAF0FJREFUHkOpw7ZIBetx3LfL73YO42dQkpV6KC5tiH6WS/x1NQh5UatWrXRXCDVAehAXuG2oN2CBKwfyEiOoi3qIcuKplp90f3FEjUNMc4SDBQ+4Rri5nIEpgtTgkwvBjQSfb7/9Vs/SZviwIxMc5EQQjVwIPrnqmbQUha87rgdJ6u05SPZF471y5UpRxodFa58NB4MAsioie6MyHHXNXBhMq5m5wlUM6aXVXnDmgj7cNcfCADu3Ns3BbpRRW9pW/XLPRQhshLi2kxGDxHwdaVDTEVzv4IrnhTLFKvHCL4p1QlHqQXcc0eCQnx2fQgn/DMjbGxTBVxQf/IPnS/BHRV5wO8GHPd2Lh72s13O87CSR4ONKSiYCiCWBgFT//ve/Bembi0Vq71quuuoqfdJg+IcHLYtaYbQ0YT/HTfM1yIiPfcKAcgjEhUmHXSEi+BViYpjJ7fcJEya0hd8oxANRmejMVQR81Cqp5RrGLR2pvX5R7nB6vBBzGcSeyJZi2i0eiZlHEo6tI5+EHiW4D26zcTykalkq8F7jLTuJhIx/pOQigKAjSIqEMNHKLib0jkLJKA8ePZdDWAo9iE66rWaprUY9nz2CbCHKprInErXs72jeHG5b7XM77iOoFsJzI6gXAlwhemamIFd4+dh3330dfJA4S+2npz7AHvzMH2V75KiXtAuJmKknbVDc+oOH3Jx+1iiDpDGIuBc02WPpB91eWPyT2q+w8ItLO4hShmhjN910kx7FLOhxx7Zf//799fwMmFXGnfr16yfKmM7SDUTXVFboelpqtf/uurSOCsomKVUPkS7xkmMmLMmDPyLrua0UmMsax8hrgTwgZkIuDWTz7Ny5syCTJPKBYDvWTEj65fZCYC4T92PO1GMwgngLRkY3O8F2AP9UYZDbnlwY7QbdRlL7FTRuceSPsVYBVQT2IVA0QczakHgJygS2Gsi0mASFjrFWrrn6x23c8fvktleOssgwqdz3UtWUUV/arcdcFTqYAWN7GG9lgKiPL7YEVEQ9h0LH3v3xxx+fkiWpB1TqMRhZJLkxDOvM4iIRTFhLepkMacwyxe04qf2K2ziEKS8SNyFOPPJIjB49Ws8rgSVdr4S6iFmurLh1njCedUup7JV/VOqNGzdOlHtuzuIo91xHPH5ghVjz5uRZOTM0FcRKyIwZM0S5tpmupj+EXcDUqVMtdgTpS8f7Tkkuv6vocw5lqPxaIzuSeBPGm6mZoIxU5DzzpUCPi7EfGWiH/sscRoukeCKAmd1rr70myjVL91bBCy72b6F44FWSjWC4hUQw+GCWN2fOHN1iHolB8BINTxU3wjNjeNfAwKt9+/ZFTZvq9mJvv4blcTu5lcH+uEH2+8ALRm1InnL55ZfLhg0bjKKWb7zQ4KUJs3s3gmcCUk4jdbSK0GnJegkbCBg8Y7ZtX5209wEzb3gqIcOmcq9zpNtG23iJQFpb7PWbrf3BC/fMqwP2/rrJHodrZVSnrP5RcZCaMoaOAPYib7755tDbDbpB/DDBe4IULwSQghlLu/Z9VaMXxiyxkO0V7LvDQBQfKAC8SINfSWX8MgB1+S7UPRcsseeN7JwqfKzU+2+WNi9bFhgrvIjBrXn58uUClzgofWTFLLUXdyp1l4eVl5wIqCx6up+98058r2DmAZ9XUrwQwGwRtiTZPBeQevONN95IzL52vEaJ0hYLAe6pFwv5mLWrIv7FTOLs4iaxT9l7He8SMHQ7/fTTsyp09HLBggWiMi/Gu8OUngjkiQCVep6AlWpx7JPB2j5J5EewoiThEeW+qEQdejwGBHHJZ8cQvtRr1qyJctcoGxHwFQEqdV/hTDYz+JgmiY499tgkdSexfYE/9Kmnnqr7l+fbSRiZwjeZRARKBYGStH4vxuDOmzdPJk6caGka4Wj9smBHNCVYpJpnMTAWUYkRLG0WcgLDpFxj3RfSThh1sd8axSAUiGk9adIkUYkx9GApGE/ICUti4B9GoKEw8M+1DeRgOOmkk9IaxOXCB9bsJCJQMgioHw1SCAioMIiOrEJG1iE/mg8r65BbakX1z+LoW9SvYTyiRMgDrfx30+akBp7KOjhjSuIo9ccPWZBFTBkzFvxsGZm5/JCJPIhA1BHg8nvJvL7501HEZY47wV8VLk9Rofnz50uLFi30mfg333yTVizMWhFjAQZgSSb1oykI+Yntno0bNxbc1aTZghQMCBkkGgEq9UQPr/+d6969e8HRoPyXKj+OiMntxRc2v1ayl4aSxl4xrPDd4vq7ccA2C+JeBx273K3tMK7B9xm2Dtdff73AOM4PShcExQ/e5EEEooYAlXrURiTi8iDwxuDBgyMuZXrxEIjCHp0vfelg7iDgBtKBIjgGQlfmS4grDl/tpBH6hIhwmdJu5ttnuL9xpp4vaiwfZwSo1OM8ekWSvWfPno48ykUSJe9mkVe7WPHeYYk9duxYXckg2QeM4rySPdOVVz5RqIec9meddZacdtppviy3G33CSxPCkJKIQCkhQKVeSqPtY1+RvCJultjI0ATFUQyaOXOmIJf0BRdcIMqosWARMFtPAj355JOijC/l8ccf97U7SPSBaHLmeN++NkBmRCCiCFCpR3Rgoi4WEmYgaUNcCMFzHn300dDFRQ5v5PKG+yLiUvtFcU8+gfjcRx11lB4dLl3iFK9YIdXnm2++KbVq1fLKgvWIQGwRiKSf+qxZs/TsO2ZUO3XqpC9bvvvuu/o/LH4gsT968MEH6zHJjR85LGkitd8HH3ygx/U+9NBD9R8PzAbSEQxyMGNYuHChIFEEPuCDOo0bNxb4NGNvbpttMr8DwXL53nvvlcWLF+tJBYwEBfhRz4ewHIlMSPCvhVLAzA57jZADVtJNmzaNRFKJs88+W95++215+OGH8+le6GWRNxv7tfDbD4sQUx4JcJA9ypz5yq/2GzZs6BerUPnA0G/o0KFyxx13pM3BXYhAWMZHqmImXSkERdaNNQJR9Lm7+OKLHb6pKul9Wh/evffeW1u9erWmsv1odevWddRVA6SpdIyaUtSO7n7xxReaUvyudVDP+KhUgHobDgb/vaD2ajVlUZ0qb9TDt8rwpKn8wY57bn7qzzzzjKZmGI6yZn6QRRlbWUQJy0/d0qg6UQpLUy9cGeU1yx72sUqxqKlALnaxAzvfunWrdvfdd2sqf3OgmKgX38D6EARj4ILYADVr1gwMF5WCU4O/P4kIlDICiEAWOXJT6tmUwV577aWpnN8ZfzBUMghLX1944QVN7QtnrGNuVxlYabNnz7bwwIla1s2Zh5mfXamrWX7OfA455BBNzQZTshRLqUMANfvS1PJyzrKbMQjyWK3e6AFdUiAFfICXB2VpHTgOylc94J74xx5KdvLkydqee+4ZGC7lypXTxowZ45/Q5EQEYoxAYpR6LsqhWbNmqaFSQS20GjVq5P1Ds8cee2i///57io/KA5x2hp5NJrNS/+qrrzSVq9khD6KIXXTRRZpaetdn/GaeKiFJSo5iKnUIgZmY8p92yG+WN8xjvKy9+OKLKXyCPFA5vbUOHTqE0nesSq1fvz7I7vjCW/nRayrcraa2jQLFRdlKaGpLzheZyYQIJAGBWCn18847T1NZlzTMUt0UBJa5sQSHMm7L8DvvvHNqzC655BJXHphRDBw4UBs0aFDa5f4bbrghxQcyucmC5XYo7QsvvNBVWaOOWamr/NAOPirqmWU5Ecrd3payLdBlKbZShxCYlalIYBqWvO1yhnmOVRtlH5Eao6AOlIGXds4552jK1iKU/qoY6NoPP/wQVHd84fvrr79q2IrafffdA8ekffv22oYNG3yRm0yIQFIQiI1Sx76tQVh2xpKbXVGooCJGEe3111933IeyMcgthrkyitOUkZpRRD92K9e2bdtUGWW05mgHs5OffvopVWbFihWuMazNSl254Dj41K9fX+vYsWPqo1yiHGWGDBmitxMFpW50WFkeu75U2ccriHOsFvz888+GKIF8Y7sBLy9uKytB9AnPpXKJC6QvfjFVBp3agAEDtKpVqzqe0SAwUeGKdXsOv+QnHyKQFARio9Rvu+02C+YqsITjxwMJIAzCD7v9x8RQ6lu2bNGw32q/ryyVjeqpb+yT2stVqVJFv48lZ7eXC+XDnapvHKhsaQ4+hlKHsVn58uUd9+3tup0bySqipNTRZ2xRYMXDDR+3fhR6TWUy01577TUD7kC+jf1hN6PHQuV3q4/tIRWsRlNBawLpT6FMMcYTJkzQVJhbT8+uW5+zXQMmKpVqoaKzPhFILAKZfbTUf1hUSClSiyhqlmQ5x4k50ESmqGFKqbu6GTVo0MDBU2WJclyDW476gRe11OjqloNIVnbK5E4F9yev0cXU/qq9qUicIzCNWkWQZcuWido2EPXSEohcao9Zd6n7+OOPddfFQBpRTJXNg6h9cz3yGY6DJLjgXXvttaI8M6RPnz5ZXSmDlMXOG8+pslUQuDPCD1x5pQjSCodBXbp00X394d5KIgJEwB2BSPqpu4lq9xG3n6OO+ZraXxd81OuYg51yNxLlWiN2hQhfZuUuZimPSGB2Uvv1eltQ+PhhW7dunaUI+LRr185yLVM8a7XXrycYUTMfS52HHnpI1N6k5Zr9JOoBNuCrP3r0aFF2CHr8AOSUhwIuhPCCAN9/hKtFlDi1AlMIu6x1ocQR7yBTBrWsTHIsgHgIalVK8IxFhfDyiuf32Wef1RU6zsOkypUry8iRI/UXiDDbZVtEII4IxEap+w0ugnfYlboysNMDu0BZ4IUAQVXwA2snBIIxCKE/7Uod9TDrxyxGLdHraSSV/7lRxfGNttTysSgrass9rAioPfXUNQSjUb74qXMcKGt8y3lUT/ASddVVV+kfzN4RYEi5B+qBelavXi2Ii56O8KOu7AukdevWoiz+9Rcm+8pNurp+XMfzELRCb9mypSj/dsF3sQmzcQQVQphVjBMyyAURQCeXfiLVLF5u8XJIIgJEIDsCJavU+/XrJ8qYzoLQH3/8IcoKXdQ+tT77g0J2o759+6YuI+ezPbkG6oG/srB3XSlIVTYdIJ2mXakjp/TXX38tnTt3llWrVunZxZDhy0xqf1F/ITBfi/oxFDQ+Kh6BLirSiCINqTIuTG1pYHsFH6xE4IWgWATFNnfu3MCaxzYRko4oT4fA2sjEGNtIiFqIaIqLFi3SIzHiGFtUxSREi0TynWLF6i9m39k2ESgEgZJV6sixjM/06dMd+GHWmG7m2KNHD0FsaYOgmMDDbZnebenfqGf/RjpQhBSF8jbo+++/FxUwR/8Y18zf2LvH8nPcCSE9o5oec9q0aYHCi6VsvEjiBQ7jiZj6+MbyO15o8MF2USGEl1U8S2vWrBEkglHeGKlvrP7Yt30KaavQugj3jJdhhNjNZBdTaDusTwSSikDJKnUM6Lhx40T5keec0xqzKcR2NxOWzhFrHnHZCzGgUlbiMmPGDD3G/JIlS8xNuB7jhx65uM12BK4FebEgBDCLDZrwAglFiw+M0OwEwzmsVsDYEgaI5g+eGyyNY7UDK0T4xvK5Cq4kyqdd//z22292lpE8hz0LVi2Qb4FEBIiARwSiaNcPf3PVHcsHAWXM1KpVK8t9lEfsdzPZ3dbgNuZGKu1jxuhyiFqVLX44YrH379/fEXhFzUJ1P3P4k9v7dMUVVzjEUbMqDf1HPXt5nKuXCA2BatTMy1IXvvu4Z66DACCkwhBASFYzpjy2/l/6gceBBx6oqS2swgaKtYkAEdARKIO/6h+TpBBACkhkaIMhFwhLocjUls+eLva8Ycymwsfqxj0w7lKJXnR++fzBjAtyIBsd0lRiloZlWRjgYb+RFA4CWAqOU4rZcFDxpxW4Iw4ePFjfN8eKF4kIEIHCEaBSLxxDckgwArD+hgU2yT8EVBhfUeGcRYXYFeyhk4gAEfAPASp1/7Akp4QiAHsJuHWRCkOgRYsWokLJiophT1uQwqBkbSKQFgEq9bTQ8AYR+H8E5s+fLyoUKuHwgACW1REBDspc5UzwwIFViAARyAcBKvV80GLZkkUAcQSmTJlSsv3Pt+PKuFQPvoRwsirzYb7VWZ4IEAGPCFCpewSO1UoLAbgrwlCx2EFZoow63OsQKAl+90cffTSX2KM8WJQtsQhQqSd2aNkxvxG48cYb5ZZbbvGbbaz5QZEjdC/2ybt27WpJqhTrjlF4IhBTBKjUYzpwFDt8BBB5DeFt165dG37jEWoRLprHHHOMrsgR0TDMOPwRgoGiEIFIIkClHslhoVBRRWDSpEmCUMGlRJiNwwMAqWcxK1eBnwJLpVtKuLKvRCAIBKjUg0CVPBOLAGI1Qam99957ie1jjRo1BNkHmzVrpmfEQ9pZhKYlEQEiEH0EqNSjP0aUMGIIvPvuu3oaWK/BGG+66SZBdj9EL8QH+eXxjYx8YRKWzWGZjmAwTZo00RU5lDkjFoY5CmyLCPiLAJW6v3iSW4kg0L17d5k8ebKn3iI//BdffCG77LKLpf7mzZv1/XrkbscHe/f4Xr9+vZ6SFhndkJzF+KA8/MARla1s2bKpD7LeVa1aVapXr66HF8Y3PjvttJOe/Q2KHJ9Cs79ZhOcJESACkUCASj0Sw0Ah4oYAlC2M5qBYvVCfPn1k7NixXqqyDhEgAkQgLQLbpL3DG0SACKRFAMFVEL/cK6msg7J48WKv1VmPCBABIuCKAGfqrrDwIhHIjgBm6ZitY9buhdq1aydz5szxUpV1iAARIAKuCHCm7goLLxKB7AjAIvz222/PXjBNiblz58rTTz+d5i4vEwEiQATyR4Az9fwxYw0iYEGgdevW8s4771iu5XpSr149+eyzz2S77bbLtQrLEQEiQATSIsCZelpoeIMI5IbAPffco1uh51baWurLL7+UESNGWC/yjAgQASLgEQHO1D0Cx2pEwIxAz5495bHHHjNfyvkYYVeXL19O//CcEWNBIkAE0iHAmXo6ZHidCOSBwLBhwwTK2Qshpvy1117rpSrrEAEiQAQsCFCpW+DgCRHwhkDt2rXlmmuu8VZZ1cIsf+HChZ7rsyIRIAJEAAhw+Z3PARHwCQHkWkfOdeRe90KnnXaaPPHEE16qsg4RIAJEQEeAM3U+CETAJwQqVKggw4cP98zt5Zdfln/++cdzfVYkAkSACHCmzmeACPiMQJs2bWT+/PmeuH777bdSq1YtT3VZiQgQASLAmTqfASLgMwKFuLghMQuJCBABIuAVASp1r8ixHhFIg8DBBx8svXr1SnM3/eUdd9zRkbktfWneIQJEgAg4EaBSd2LCK0SgYASGDh0qlSpVyovPySefnFd5FiYCRIAI2BGgUrcjwnMi4AMC2Be/5ZZbcuaEOPIDBw7MuTwLEgEiQATcEKBSd0OF14iADwj0799fTjnllKycttlmG3n00Uelbt26WcuyABEgAkQgEwJU6pnQ4T0iUAACZcqUkSlTpsigQYOkXLlyrpx23313efXVV6Vbt26u93mRCBABIpAPAnRpywctliUCHhFYu3atPPnkk3rUOISFxfJ8hw4dpGvXrrLtttt65MpqRIAIEAErAlTqVjx4RgSIABEgAkQgtghw+T22Q0fBiQARIAJEgAhYEaBSt+LBMyJABIgAESACsUWASj22Q0fBiQARIAJEgAhYEaBSt+LBMyJABIgAESACsUWASj22Q0fBiQARIAJEgAhYEaBSt+LBMyJABIgAESACsUWASj22Q0fBiQARIAJEgAhYEaBSt+LBMyJABIgAESACsUWASj22Q0fBiQARIAJEgAhYEaBSt+LBMyJABIgAESACsUWASj22Q0fBiQARIAJEgAhYEaBSt+LBMyJABIgAESACsUWASj22Q0fBiQARIAJEgAhYEaBSt+LBMyJABIgAESACsUWASj22Q0fBiQARIAJEgAhYEaBSt+LBMyJABIgAESACsUWASj22Q0fBiQARIAJEgAhYEfg/RPSRSs/xKyoAAAAASUVORK5CYII=\" width=\"100%\">\n",
        "  </td>\n",
        "</tr>\n",
        "<tr>\n",
        "  <th>The ML Lifecycle</a></th>\n",
        "</tr>\n",
        "</table>"
      ],
      "metadata": {
        "id": "ZCVVqoCCqVto"
      }
    },
    {
      "cell_type": "markdown",
      "source": [
        "> **Quick note**: The ML terms _\"productionalizing\"_, _\"operationalizing\"_, or _\"deploying\"_ a model all refer to the process of providing the model for other users to make predictions. The term _\"model inference\"_ is the stage of the machine learning lifecycle when we make predictions, which is also referred to as _\"scoring\"_ new data.\n",
        "\n",
        "While it’s difficult to estimate the overall contribution of ML to the climate crisis, the field of Information and Communications Technology (ICT) accounts for about 1.4% of global GHG emissions. Roughly one-third of these emissions are related to the manufacture and handling of physical materials [1].  "
      ],
      "metadata": {
        "id": "OMxlVxiKuTSL"
      }
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "-oLZpeDxCPNZ"
      },
      "source": [
        "### **1.3.1 Case study: ChatGPT** 🤖\n",
        "\n",
        "While ICT's contribution to global emissions may seem small, the amount of compute needed to achieve state-of-the-art ML performance has historically doubled every 3.4 months [4]. These energy increases could be offset by hardware and algorithmic improvements [5]. But given the latest trend toward developing ever larger and more complex models, such as ChatGPT, we should remain vigilant in tracking and reducing the emissions associated with machine learning models.\n",
        "\n",
        "With that in mind, let's take a closer look at the emissions involved with both training and using ChatGPT!\n",
        "\n",
        "Training ChatGPT’s underlying language model, GPT-3, consumed an estimated 1,287 MWh [5]. To put this number in perspective, the average annual electricity consumption for an American in 2021 was 10,632 kWh [10]. That means the energy used to train GPT-3 was the equivalent of adding 121 additional Americans to the grid for one year.\n",
        "\n",
        "Of course, we only realize value from GPT-3 when it's used by the world. ChatGPT saw more than 590 million unique visits in January 2023 alone [9]. When we calculate a model’s carbon impact, we must consider all aspects of the ML lifecycle.\n",
        "\n",
        "Based on some [back-of-the-envelope calculations](https://towardsdatascience.com/chatgpts-electricity-consumption-7873483feac4) by Kasper Groes Albin Ludvigsen, ChatGPT’s electricity consumption in January 2023 is estimated to have been between 1,168 MWh and 23,364 MWh&mdash;roughly 1 and 18 times the amount of energy required to initially train the model. And that’s the energy estimate for just one month!\n",
        "\n",
        "As you integrate machine learning into your own research, you will likely train models that are far smaller than GPT-3. But that doesn’t mean that your work won’t have a climate impact. Training a useful model requires a lot of experimentation, and the energy consumed by all those iterations can add up.\n",
        "\n",
        "To reduce emissions from machine learning, we must first track them! Even if you're unable to apply any of the suggested strategies from this tutorial to minimize your impact, simply logging and reporting the associated emissions from your ML project is an important step.\n",
        "\n",
        "Let's learn how we can calculate these emissions in the next section."
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "Fxopvu26CPPv"
      },
      "source": [
        "<a name=\"methodology\"></a>\n",
        "# **2. Methodology: Estimating Emissions from ML**\n",
        "\n",
        "A model’s carbon footprint from computing-related impacts can be estimated as the product of electricity consumed and the carbon intensity of that electricity. To reduce the climate impact of a model, we can either reduce the amount of electricity required or shift our compute to an energy grid with a lower carbon intensity."
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "pEjnWmfgCPW9"
      },
      "source": [
        "## **2.1 Introducing CodeCarbon**\n",
        "\n",
        "To measure the computing-related carbon footprint of ML training, we’ll use a Python package called [CodeCarbon](https://github.com/mlco2/codecarbon). CodeCarbon was first developed in 2020 by a team from [MILA](https://mila.quebec/en/), [BCG GAMMA](https://www.bcg.com/beyond-consulting/bcg-gamma/overview), [Haverford College](https://www.haverford.edu/), and [Comet](https://www.comet.com/site/).\n",
        "\n",
        "CodeCarbon uses the following formula to calculate the carbon footprint of executing your code:\n",
        "\n",
        "<center><p><p> <img src=\"https://github.com/mlco2/codecarbon/raw/master/docs/edit/images/calculation.png\" alt=\"alt\" width=\"75%\"/></center>"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "fGSSETqoHVN5"
      },
      "source": [
        "There are several other packages that provide similar functionality, such as [carbontracker](https://github.com/lfwa/carbontracker), [experiment-impact-tracker](https://github.com/Breakend/experiment-impact-tracker), and [eco2ai](https://github.com/sb-ai-lab/Eco2AI). However, CodeCarbon remains the most continuously updated and supported codebase among these emissions trackers.\n",
        "\n",
        "If you're interested in learning more about the differences among these open-source trackers, please see Table 1 in [6]."
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "P4T-6uQhIfVJ"
      },
      "source": [
        "## **2.2 Initializing the CodeCarbon tracker**\n",
        "\n",
        "Let's start using CodeCarbon!\n",
        "\n",
        "Note that CodeCarbon is not limited to measuring the impact of machine learning code. You can instantiate the tracker at any time to measure emissions from executing any type of code.\n",
        "\n",
        "If you're more familiar with Python, you can also add the CodeCarbon tracker to your code as a context manager or as a decorator. Check out the [CodeCarbon documentation](https://mlco2.github.io/codecarbon/usage.html#) for more information on these techniques."
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "id": "tRecJkdoYlrR",
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "outputId": "eba584d3-a7cc-4f30-ab45-4dc64af29438"
      },
      "outputs": [
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "[codecarbon INFO @ 13:36:54] [setup] RAM Tracking...\n",
            "[codecarbon INFO @ 13:36:54] [setup] GPU Tracking...\n",
            "[codecarbon INFO @ 13:36:54] Tracking Nvidia GPU via pynvml\n",
            "[codecarbon INFO @ 13:36:54] [setup] CPU Tracking...\n",
            "[codecarbon WARNING @ 13:36:54] No CPU tracking mode found. Falling back on CPU constant mode.\n",
            "[codecarbon WARNING @ 13:36:55] We saw that you have a Intel(R) Xeon(R) CPU @ 2.30GHz but we don't know it. Please contact us.\n",
            "[codecarbon INFO @ 13:36:55] CPU Model on constant consumption mode: Intel(R) Xeon(R) CPU @ 2.30GHz\n",
            "[codecarbon INFO @ 13:36:55] >>> Tracker's metadata:\n",
            "[codecarbon INFO @ 13:36:55]   Platform system: Linux-5.10.147+-x86_64-with-glibc2.31\n",
            "[codecarbon INFO @ 13:36:55]   Python version: 3.9.16\n",
            "[codecarbon INFO @ 13:36:55]   Available RAM : 12.681 GB\n",
            "[codecarbon INFO @ 13:36:55]   CPU count: 2\n",
            "[codecarbon INFO @ 13:36:55]   CPU model: Intel(R) Xeon(R) CPU @ 2.30GHz\n",
            "[codecarbon INFO @ 13:36:55]   GPU count: 1\n",
            "[codecarbon INFO @ 13:36:55]   GPU model: 1 x Tesla T4\n"
          ]
        }
      ],
      "source": [
        "# Instantiate the tracker object\n",
        "tracker = EmissionsTracker(\n",
        "    output_dir=\"./code_carbon/\",  # define the directory to which we'll write our emissions results\n",
        "    output_file=\"emissions.csv\",  # define the name of the file containing our emissions results\n",
        "    # log_level='error' # comment out this line to see regular output\n",
        ")\n",
        "tracker.start()"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "MNsP5Ea_HVQE"
      },
      "source": [
        "## **2.3 Measuring Power Usage**\n",
        "\n",
        "First, let's define some commonly used hardware in the field of machine learning:\n",
        "- **CPU:** Central Processing Unit\n",
        "  - Manages all general-purpose programming.\n",
        "- **GPU:** Graphical Processing Unit\n",
        "  - Enhances computer graphics and accelerates machine learning workloads.\n",
        "- **TPU:** Tensor Processing Unit\n",
        "  - Google's custom-built processors to accelerate TensorFlow projects. [TensorFlow](https://www.tensorflow.org/) is a popular open-source deep learning framework developed by Google.\n",
        "- **RAM:** Random Access Memory\n",
        "  - Temporary memory that a computer can access quickly.\n",
        "\n",
        "The main difference between CPU's and newer processors like GPU's and TPU's is the number of cores. You can think of a \"core\" as the basic building block of a processor&mdash;the more cores within a processor, the more processes that hardware can execute at one time.\n",
        "\n",
        "CPU's contain only a handful of cores, which means they can only execute a few processes simultaneously. This is known as _serial processing_.\n",
        "\n",
        "In contrast, GPU's and TPU's have thousands of cores, which means they can break down complicated tasks into thousands (or even millions!) of separate processes and execute them at the same time. This is known as _parallel processing_.\n",
        "\n",
        "For a more intuitive understanding of the difference between serial and parallel processing, I recommend watching the Mythbusters video below."
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 321
        },
        "id": "u5nv4i799Uu-",
        "outputId": "ecc67a1b-482e-49f8-ba83-cede8a738fa4"
      },
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "<IPython.lib.display.YouTubeVideo at 0x7f2758f13a00>"
            ],
            "text/html": [
              "\n",
              "        <iframe\n",
              "            width=\"400\"\n",
              "            height=\"300\"\n",
              "            src=\"https://www.youtube.com/embed/-P28LKWTzrI\"\n",
              "            frameborder=\"0\"\n",
              "            allowfullscreen\n",
              "            \n",
              "        ></iframe>\n",
              "        "
            ],
            "image/jpeg": "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\n"
          },
          "metadata": {},
          "execution_count": 4
        }
      ],
      "source": [
        "from IPython.display import YouTubeVideo\n",
        "\n",
        "YouTubeVideo(\"-P28LKWTzrI\")"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "o5mt9NL59Vbq"
      },
      "source": [
        "#### **2.3.1 Hardware Power Consumption**\n",
        "\n",
        "Hardware power consumption is often measured in Thermal Design Power (TDP), which represents the maximum amount of heat a CPU or GPU will generate under an intense workload.\n",
        "\n",
        "Processors that specialize in parallel processing typically consume more power than CPU's. For example, the median TDP of the [CPU list maintained by CodeCarbon](https://github.com/mlco2/codecarbon/blob/master/codecarbon/data/hardware/cpu_power.csv) is 65 W. Meanwhile, the TDP of GPU's usually exceeds 100 W.\n",
        "\n",
        "This does not mean that leveraging a GPU to train your machine learning model necessarily results in greater emissions. Thanks to parallel processing, GPU's can dramatically shorten the time needed to train a large model and reduce the overall power load from the task.\n",
        "\n",
        "The question of which kind of hardware is \"greener\" depends on the situation. Research has found that there is likely a specific optimal number of cores we could leverage in a given parallel computation to minimize emissions, as illustrated in the plot below [11].\n",
        "\n",
        "<center><p><p> <img src=\"https://miro.medium.com/v2/resize:fit:1400/format:webp/1*bFhKxp5JhjA_RQ_hIo1pHQ.png\" alt=\"alt\" width=\"60%\"/></center>"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "xK0oW3QtSOxS"
      },
      "source": [
        "### **2.3.2 CodeCarbon Hardware Tracking**\n",
        "\n",
        "CodeCarbon measures the electricity consumption of the GPU, CPU, and RAM used for executing your code (note that as of April 2023, [TPU’s are not supported](https://github.com/mlco2/codecarbon/issues/116) due to the difficulty of obtaining live power usage for this type of hardware). These measurements are taken at regular time intervals, with a default setting of 15 seconds.\n",
        "\n",
        "Let's see what hardware CodeCarbon has found running under the hood of this Colab notebook. The last 5 lines of the output from when we instantiated the `tracker` object are the most meaningful. Here is what was displayed to me when I ran this notebook in April 2023:\n",
        "```\n",
        "[codecarbon INFO @ 13:07:27]   Available RAM : 12.681 GB\n",
        "[codecarbon INFO @ 13:07:27]   CPU count: 2\n",
        "[codecarbon INFO @ 13:07:27]   CPU model: Intel(R) Xeon(R) CPU @ 2.00GHz\n",
        "[codecarbon INFO @ 13:07:27]   GPU count: 1\n",
        "[codecarbon INFO @ 13:07:27]   GPU model: 1 x Tesla T4\n",
        "```\n",
        "\n",
        "CodeCarbon has assessed the memory of my RAM (12.681 GB) and assumes that each 8 GB of memory accounts for 3 W of power. Therefore, CodeCarbon will multiply our RAM's memory by 3/8 to determine the power consumption in watts. In our case, that comes out to about 4.76 W.\n",
        "\n",
        "CodeCarbon has also counted two Intel(R) Xeon(R) CPU's. You might notice a WARNING in your output that reads:\n",
        "```\n",
        "[codecarbon WARNING @ 13:07:25] No CPU tracking mode found. Falling back on CPU constant mode.\n",
        "```\n",
        "This means that Colab does not expose enough information about the available CPU's for CodeCarbon to accurately assess their power consumption. To compensate, CodeCarbon will assign a constant TDP of 85 W for all CPU power expenditure and assumes that only half of that TDP is consumed on average. Therefore the final power assigned to our CPU's is 42.5 W.\n",
        "\n",
        "Finally, we note that CodeCarbon has found one Nvidia Tesla T4 GPU. CodeCarbon leverages Nvidia's [`pynvml`](https://pypi.org/project/pynvml/) library to directly access the GPU's power rating."
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "tPfxII2JAv47",
        "outputId": "d5015c37-8b2b-4064-867f-33be488a04e4"
      },
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "All hardware identified: [RAM(), GPU() (Tesla T4), CPU(Intel(R) Xeon(R) CPU @ 2.30GHz > 85W [generic])]\n",
            "GPU Power: Power(kW=0.010953)\n"
          ]
        }
      ],
      "source": [
        "# Note that CodeCarbon will output the energy consumption every time you execute code\n",
        "# Comment out the `log_level` argument when instantiating the tracker to suppress this output\n",
        "print(\"All hardware identified:\", tracker._hardware)\n",
        "print(\"GPU Power:\", tracker._hardware[1].total_power())"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "dzIS1dByHVSv"
      },
      "source": [
        "## **2.4 Measuring Carbon Intensity**\n",
        "\n",
        "CodeCarbon calculates a weighted average of the emissions from the energy sources that make up your local grid (or the grid used by your cloud provider). When available, CodeCarbon uses the global carbon intensity of electricity per [cloud provider](https://github.com/mlco2/codecarbon/blob/master/codecarbon/data/cloud/impact.csv) or per [country](https://github.com/mlco2/codecarbon/blob/master/codecarbon/data/private_infra/eu-carbon-intensity-electricity.csv).  \n",
        "\n",
        "If the carbon intensity of a particular country is unavailable, CodeCarbon may still have access to its electricity mix and will use the [carbon intensities of each energy source](https://mlco2.github.io/codecarbon/methodology.html#id4) to calculate the weighted average. Note that the package assigns the same carbon intensity to an energy source regardless of where the energy is produced (i.e. coal's carbon intensity is considered to be the same in both West Virginia and Germany).\n",
        "\n",
        "If neither the country’s carbon intensity or its energy mix is available, CodeCarbon applies the world average of 475 g CO<sub>2</sub>eq/kWh. Note that \"CO<sub>2</sub>eq\" stands for \"carbon dioxide equivalents\" and represents the global warming potential of various greenhouse gases.\n",
        "\n",
        "> **IMPORTANT NOTE:** Given the simplifying assumptions on an energy grid's carbon intensity listed above, the emissions returned by CodeCarbon are only an estimate!\n",
        "\n"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "Ypst6PAlMsQ7"
      },
      "source": [
        "### 2.4.1 Colab Compute Region\n",
        "\n",
        "Let's determine what compute region Colab has selected."
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "Ay6ihbyzS5Iw",
        "outputId": "f2d83e82-58a9-4561-b9fe-086e54b9a3e1"
      },
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "GeoMetadata(country_iso_code=USA, country_name=United States, region=iowa)"
            ]
          },
          "metadata": {},
          "execution_count": 6
        }
      ],
      "source": [
        "tracker._geo"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "l2WlrtN-dChs"
      },
      "source": [
        "Colab has connected to Google's data center in Iowa to execute my code. Note that Colab might connect to a different region every time you restart your runtime. For example, I'm physically located in Chicago but Colab often connects me to Singapore's data center!\n",
        "\n",
        "[CodeCarbon's data](https://github.com/mlco2/codecarbon/blob/master/codecarbon/data/private_infra/2016/usa_emissions.json) estimates that Iowa's energy grid has an energy intensity of about 998 lbs CO<sub>2</sub>eq per MWh as of 2016. Unfortunately, CodeCarbon does not have more recent estimates of energy intensity for regions in the USA or Canada.\n",
        "\n",
        "A quick scan of CodeCarbon's 2016 [USA energy intensities](https://github.com/mlco2/codecarbon/blob/master/codecarbon/data/private_infra/2016/usa_emissions.json) shows that Iowa falls somewhere in the middle of U.S. states. If we wanted to optimize our energy efficiency, we might restart our runtime until we connected to Vermont (57 lbs CO<sub>2</sub>eq/MWh) or Idaho (189 lbs CO<sub>2</sub>eq/MWh).\n",
        "\n",
        "This kind of energy efficiency optimization is much easier for cloud environments such as AWS, GCP, or Azure that allow you to choose which compute region you'd like to connect to."
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "XFohx58pjAhr"
      },
      "source": [
        "## **2.5 CodeCarbon Dashboard**\n",
        "\n",
        "Stopping the tracker will total up all the energy consumed by this hardware since the tracker's initialization and multiply by our grid's energy intensity to estimate emissions in CO<sub>2</sub>eq.\n",
        "\n",
        "> **IMPORTANT NOTE**: Since it is unlikely that Colab will provision you with the same underlying hardware and connect you to the same compute region as when I ran this notebook, your emissions from running this code will probably differ from those currently displayed in the tutorial!\n",
        "\n",
        "Once stopped, the tracker will create a new file (in this case titled `emissions.csv`) in our notebook's working directory with info on the emissions generated from this experiment"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "_6uoGPO7RL44",
        "outputId": "acf0fa09-b7d4-4361-ffd6-95a03cb10b0b"
      },
      "outputs": [
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "[codecarbon INFO @ 13:36:56] Energy consumed for RAM : 0.000000 kWh. RAM Power : 4.755446434020996 W\n",
            "[codecarbon INFO @ 13:36:56] Energy consumed for all GPUs : 0.000000 kWh. All GPUs Power : 10.953 W\n",
            "[codecarbon INFO @ 13:36:56] Energy consumed for all CPUs : 0.000001 kWh. All CPUs Power : 42.5 W\n",
            "[codecarbon INFO @ 13:36:56] 0.000001 kWh of electricity used since the begining.\n"
          ]
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Emissions from this training run: 0.000001 kg CO2eq\n"
          ]
        }
      ],
      "source": [
        "emissions = tracker.stop()\n",
        "print(f\"Emissions from this training run: {emissions:5f} kg CO2eq\")"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "VxyFPmTwnFBL"
      },
      "source": [
        "We can also contextualize these emissions via CodeCarbon's dashboard.\n",
        "\n",
        "This dashboard is hosted via a Dash app. In order to access this app on Colab's servers, we'll need a proxy link. After executing the cell below, click on the link next to the text stating \"CLICK HERE\". Once you're done reviewing the dashboard, click on the cell's stop button ⏹."
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "id": "kW41tIEWahQY"
      },
      "outputs": [],
      "source": [
        "## UNCOMMENT THIS CELL TO RUN THE CODECARBON DASHBOARD ##\n",
        "\n",
        "# # Display the proxy link\n",
        "# print(\"CLICK HERE:\", eval_js(\"google.colab.kernel.proxyPort(8050)\"), \"\\n\")\n",
        "\n",
        "# # Point to our emissions csv file\n",
        "# !carbonboard --filepath=\"code_carbon/emissions.csv\" --port=8050"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "Er4xBXedoNTe"
      },
      "source": [
        "When I ran this notebook in April 2023, executing just a few cells of code to assess our underlying hardware and compute region resulted in carbon emissions equivalent to watching 7 minutes of a 32-inch LCD TV.  "
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "8WBpou7FXKKh"
      },
      "source": [
        "<a name=\"codecarbon-experiments\"></a>\n",
        "# **3. CodeCarbon Experiments**\n",
        "\n",
        "Let's try tracking our carbon emissions while training a machine learning model. We'll also explore a few techniques to train in a more efficient way and reduce our emissions."
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "qL9_-2ilI0qv"
      },
      "source": [
        "## **3.1 Generate Training and Test Data**\n",
        "\n",
        "We'll use the same data as the [Forecasting the El Nino / Southern Oscillation with Machine Learning tutorial](https://colab.research.google.com/drive/1nOL7rB8EnFTvLoarerVOtECYBwJKth-t?usp=sharing) from earlier in the Summer School program. Remember that the goal of that tutorial was to leverage historical sea surface temperatures to correctly forecast the Nino3.4 index, which indicates the state of El Nino / Southern Oscillation (ENSO).\n",
        "\n",
        "We'll move quickly through this data preparation section as the focus of this tutorial is on model training and emissions tracking. For a more detailed walkthrough, please reference the original forecasting tutorial."
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "P79X72ZCI2TA",
        "outputId": "5b63a685-1e28-4223-c8ac-96ecd03af37a"
      },
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "/usr/local/lib/python3.9/dist-packages/gdown/cli.py:121: FutureWarning: Option `--id` was deprecated in version 4.3.1 and will be removed in 5.0. You don't need to pass it anymore to use a file ID.\n",
            "  warnings.warn(\n",
            "Downloading...\n",
            "From: https://drive.google.com/uc?id=1aGvitA8rYrHRDxNd2XD4AAFsahCQsv0t\n",
            "To: /content/nino34.long.anom.data.txt\n",
            "100% 15.4k/15.4k [00:00<00:00, 28.3MB/s]\n"
          ]
        }
      ],
      "source": [
        "# Download Nino3.4 index (this index measures the state of ENSO by encoding the\n",
        "# average temperature anomaly in the equatorial Pacific)\n",
        "!gdown --id 1aGvitA8rYrHRDxNd2XD4AAFsahCQsv0t"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "_GNyqemwGwr9",
        "outputId": "f8deae8b-98da-4e6b-b000-d2dcefbae50e"
      },
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Downloading...\n",
            "From: https://drive.google.com/uc?id=1-xefk3imP4Q-8GevIV2YIo82iP5rTdSc\n",
            "To: /content/sst.mon.mean.trefadj.anom.1880to2018.nc\n",
            "100% 432M/432M [00:02<00:00, 149MB/s]\n"
          ]
        }
      ],
      "source": [
        "# Download sst observations from 1880 to 2018\n",
        "!gdown 1-xefk3imP4Q-8GevIV2YIo82iP5rTdSc"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "id": "TM7hyGz3Ry8e"
      },
      "outputs": [],
      "source": [
        "def load_enso_indices():\n",
        "    \"\"\"\n",
        "    Reads in the txt data file to output a pandas Series of ENSO vals\n",
        "\n",
        "    Returns:\n",
        "      pd.Series: monthly ENSO values starting from 1870-01-01\n",
        "    \"\"\"\n",
        "    with open(\"nino34.long.anom.data.txt\") as f:\n",
        "        line = f.readline()\n",
        "        enso_vals = []\n",
        "        while line:\n",
        "            yearly_enso_vals = map(float, line.split()[1:])\n",
        "            enso_vals.extend(yearly_enso_vals)\n",
        "            line = f.readline()\n",
        "\n",
        "    enso_vals = pd.Series(enso_vals)\n",
        "    enso_vals.index = pd.date_range(\"1870-01-01\", freq=\"MS\", periods=len(enso_vals))\n",
        "    enso_vals.index = pd.to_datetime(enso_vals.index)\n",
        "    return enso_vals\n",
        "\n",
        "\n",
        "def assemble_basic_predictors_predictands(start_date, end_date, lead_time):\n",
        "    \"\"\"\n",
        "    Creates a tuple of the predictors (np array of sst temperature anomalies)\n",
        "    and the predictands (np array the ENSO index at the specified lead time).\n",
        "\n",
        "    Parameters:\n",
        "        start_date (str): the start date from which to extract sst\n",
        "        end_date (str): the end date\n",
        "        lead_time (str): the number of months between each sst\n",
        "            value and the target Nino3.4 Index\n",
        "\n",
        "    Returns:\n",
        "        (np.array, np.array)\n",
        "\n",
        "    \"\"\"\n",
        "    ds = xr.open_dataset(\"sst.mon.mean.trefadj.anom.1880to2018.nc\")\n",
        "    sst = ds[\"sst\"].sel(time=slice(start_date, end_date))\n",
        "    num_time_steps = sst.shape[0]\n",
        "\n",
        "    # sst is a 3D array: (time_steps, lat, lon)\n",
        "    # in this tutorial, we will not be using ML models that take\n",
        "    # advantage of the spatial nature of global temperature\n",
        "    # therefore, we reshape sst into a 2D array: (time_steps, lat*lon)\n",
        "    # (At each time step, there are lat*lon predictors)\n",
        "    sst = sst.values.reshape(num_time_steps, -1)\n",
        "    sst[np.isnan(sst)] = 0\n",
        "\n",
        "    X = sst\n",
        "\n",
        "    start_date_plus_lead = pd.to_datetime(start_date) + pd.DateOffset(months=lead_time)\n",
        "    end_date_plus_lead = pd.to_datetime(end_date) + pd.DateOffset(months=lead_time)\n",
        "    y = load_enso_indices()[slice(start_date_plus_lead, end_date_plus_lead)]\n",
        "\n",
        "    ds.close()\n",
        "    return X, y\n",
        "\n",
        "\n",
        "def plot_nino_time_series(y, predictions, title):\n",
        "    \"\"\"\n",
        "    Plots the ENSO time series against the predictions.\n",
        "\n",
        "    Parameters:\n",
        "      y (pd.Series): time series of the true Nino index\n",
        "      predictions (np.array): time series of the predicted Nino index (same\n",
        "          length and time as y)\n",
        "      title (str): the title of the plot\n",
        "\n",
        "    Returns:\n",
        "      (None): Displays the plot\n",
        "    \"\"\"\n",
        "    predictions = pd.Series(predictions, index=y.index)\n",
        "    predictions = predictions.sort_index()\n",
        "    y = y.sort_index()\n",
        "\n",
        "    plt.plot(y, label=\"Ground Truth\")\n",
        "    plt.plot(predictions, \"--\", label=\"ML Predictions\")\n",
        "    plt.legend(loc=\"best\")\n",
        "    plt.title(title)\n",
        "    plt.ylabel(\"Nino3.4 Index\")\n",
        "    plt.xlabel(\"Date\")\n",
        "    plt.show()\n",
        "    plt.close()"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "ypoInWXPDyUw"
      },
      "source": [
        "## **3.2 Algorithm Choice**\n",
        "\n",
        "One of the easiest ways to reduce emissions is by choosing a simpler algorithm. We often don't need to immediately reach for a complicated neural network when a more traditional machine learning method would suffice.\n"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "EA37KNvgHiwc"
      },
      "source": [
        "<table style=\"text-align: center;\" width=\"50%\">\n",
        "<tr>\n",
        "  <td>\n",
        "   <img src=\"https://external-preview.redd.it/khzQQ3xzswIW4HGIvesRWJXvCbnvdWNarIyU4-GVl2I.jpg?width=640&crop=smart&auto=webp&s=180b2912c6bef455fcb4ef7565dd5093edbabe23\" width=\"100%\">\n",
        "  </td>\n",
        "</tr>\n",
        "<tr>\n",
        "  <th>Applying a neural network when linear regression would do.</a></th>\n",
        "</tr>\n",
        "</table>"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "C1wfV4vVIaT2"
      },
      "source": [
        "We'll discuss more about the potential trade-offs involved in choosing a simple algorithm over a more complicated one at the end of this section. For now, we'll focus on training two different models and capturing the emissions from each one.\n",
        "\n",
        "\n"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "JyDn8xGIGPq2"
      },
      "source": [
        "### **3.2.1 Training a Baseline Model: Linear Regression**\n",
        "\n",
        "Let's first attempt to model ENSO using linear regression. We'll track the emissions associated with training this simple model."
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "id": "3dBBtJ65Bs4S"
      },
      "outputs": [],
      "source": [
        "experiment_name = \"linear_regression\"\n",
        "\n",
        "tracker = EmissionsTracker(\n",
        "    output_dir=\"./code_carbon/\",\n",
        "    output_file=f\"{experiment_name}_emissions.csv\",\n",
        "    log_level=\"error\",  # comment out this line to see regular output\n",
        ")\n",
        "tracker.start()"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 510
        },
        "id": "gPXbNl_TLHQQ",
        "outputId": "9fa68ae0-64a7-4098-e37c-e009010dc14e"
      },
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "RMSE: 0.53\n"
          ]
        },
        {
          "output_type": "display_data",
          "data": {
            "text/plain": [
              "<Figure size 640x480 with 1 Axes>"
            ],
            "image/png": "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\n"
          },
          "metadata": {}
        }
      ],
      "source": [
        "# Load train, val, and test sets\n",
        "X_train, y_train = assemble_basic_predictors_predictands(\n",
        "    \"1980-01-01\", \"1995-12-31\", lead_time=1\n",
        ")\n",
        "X_val, y_val = assemble_basic_predictors_predictands(\n",
        "    \"1997-01-01\", \"2006-12-31\", lead_time=1\n",
        ")\n",
        "X_test, y_test = assemble_basic_predictors_predictands(\n",
        "    \"2007-01-01\", \"2017-12-31\", lead_time=1\n",
        ")\n",
        "\n",
        "# Train a linear regression model\n",
        "regr = sklearn.linear_model.LinearRegression()\n",
        "regr.fit(X_train, y_train)\n",
        "\n",
        "predictions = regr.predict(X_val)\n",
        "corr, _ = scipy.stats.pearsonr(predictions, y_val)\n",
        "baseline_rmse = mean_squared_error(y_val, predictions, squared=False)\n",
        "print(\"RMSE: {:.2f}\".format(baseline_rmse))\n",
        "\n",
        "# Plot predictions against actuals\n",
        "plot_nino_time_series(\n",
        "    y_val,\n",
        "    predictions,\n",
        "    \"Predicted and True Nino3.4 Indices on Training Set at 1 Month Lead Time. \\n Corr: {:.2f}\".format(\n",
        "        corr\n",
        "    ),\n",
        ")"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "igPObPWtY4MH",
        "outputId": "d4dec03c-7e6a-4af1-9b55-7944226ae55d"
      },
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Emissions from this training run: 0.00004 kg CO2eq\n"
          ]
        }
      ],
      "source": [
        "# Evaluate emissions\n",
        "baseline_emissions = tracker.stop()\n",
        "print(f\"Emissions from this training run: {baseline_emissions:.5f} kg CO2eq\")"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "7Swv-0BoGnET"
      },
      "source": [
        "### **3.2.2 Training a CNN**\n",
        "\n",
        "Let's compare the emissions and performance of a linear regression model to a more complex convolutional neural network (CNN).\n",
        "\n",
        "First, we'll define some necessary data wrangling code."
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "id": "7cvVoCA_ljFU"
      },
      "outputs": [],
      "source": [
        "def assemble_predictors_predictands(start_date, end_date, lead_time,\n",
        "                                    num_input_time_steps=1):\n",
        "    \"\"\"\n",
        "    Creates a tuple of the predictors (np array of sst temperature anomalies)\n",
        "    and the predictands (np array the ENSO index at the specified lead time).\n",
        "\n",
        "    Parameters:\n",
        "      start_date (str): the start date from which to extract sst\n",
        "      end_date (str): the end date\n",
        "      lead_time (int): the number of months between each sst value and the\n",
        "          target Nino3.4 Index\n",
        "      num_input_time_steps (int): the number of time steps to use for each\n",
        "          predictor sample\n",
        "\n",
        "    Returns:\n",
        "      (np.array, np.array)\n",
        "\n",
        "    \"\"\"\n",
        "    file_name = \"sst.mon.mean.trefadj.anom.1880to2018.nc\"\n",
        "    variable_name = \"sst\"\n",
        "    ds = xr.open_dataset(file_name)\n",
        "    sst = ds[variable_name].sel(time=slice(start_date, end_date))\n",
        "\n",
        "    num_samples = sst.shape[0]\n",
        "    # sst is a (num_samples, lat, lon) array\n",
        "    # the line below converts it to (num_samples, num_input_time_steps, lat, lon)\n",
        "    sst = np.stack(\n",
        "        [\n",
        "            sst.values[n - num_input_time_steps : n]\n",
        "            for n in range(num_input_time_steps, num_samples + 1)\n",
        "        ]\n",
        "    )\n",
        "    num_samples = sst.shape[0]\n",
        "    sst[np.isnan(sst)] = 0\n",
        "    X = sst\n",
        "\n",
        "    start_date_plus_lead = pd.to_datetime(start_date) + pd.DateOffset(\n",
        "        months=lead_time + num_input_time_steps - 1\n",
        "    )\n",
        "    end_date_plus_lead = pd.to_datetime(end_date) + pd.DateOffset(months=lead_time)\n",
        "    y = load_enso_indices()[slice(start_date_plus_lead, end_date_plus_lead)]\n",
        "    ds.close()\n",
        "    return X.astype(np.float32), y.astype(np.float32)\n",
        "\n",
        "\n",
        "class ENSODataset(Dataset):\n",
        "    def __init__(self, predictors, predictands):\n",
        "        self.predictors = predictors\n",
        "        self.predictands = predictands\n",
        "        assert (\n",
        "            self.predictors.shape[0] == self.predictands.shape[0]\n",
        "        ), \"The number of predictors must equal the number of predictands!\"\n",
        "\n",
        "    def __len__(self):\n",
        "        return self.predictors.shape[0]\n",
        "\n",
        "    def __getitem__(self, idx):\n",
        "        return self.predictors[idx], self.predictands[idx]"
      ]
    },
    {
      "cell_type": "markdown",
      "source": [
        "Now we'll define the architecture of our convolutional neural network."
      ],
      "metadata": {
        "id": "krxiHdPvCkOU"
      }
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "id": "_uuT1tFlrYjI"
      },
      "outputs": [],
      "source": [
        "class CNN(nn.Module):\n",
        "    def __init__(self, num_input_time_steps=1, print_feature_dimension=False):\n",
        "        \"\"\"\n",
        "        Constructs all the necessary attributes for the CNN class.\n",
        "\n",
        "        Parameters:\n",
        "          num_input_time_steps (int) : the number of input time steps in the predictor\n",
        "          print_feature_dimension (bool) : whether or not to print out the dimension\n",
        "              of the features extracted from the conv layers\n",
        "        \"\"\"\n",
        "        super(CNN, self).__init__()\n",
        "        self.conv1 = nn.Conv2d(num_input_time_steps, 6, 3)\n",
        "        self.pool = nn.MaxPool2d(2, 2)\n",
        "        self.conv2 = nn.Conv2d(6, 16, 5)\n",
        "        self.print_layer = Print()\n",
        "\n",
        "        # TIP: print out the dimension of the extracted features from\n",
        "        # the conv layers for setting the dimension of the linear layer!\n",
        "        # Using the print_layer, we find that the dimensions are\n",
        "        # (batch_size, 16, 42, 87)\n",
        "        self.fc1 = nn.Linear(16 * 42 * 87, 120)\n",
        "        self.fc2 = nn.Linear(120, 84)\n",
        "        self.fc3 = nn.Linear(84, 1)\n",
        "        self.print_feature_dimension = print_feature_dimension\n",
        "\n",
        "    def forward(self, x):\n",
        "        x = self.pool(F.relu(self.conv1(x)))\n",
        "        x = self.pool(F.relu(self.conv2(x)))\n",
        "        if self.print_feature_dimension:\n",
        "            x = self.print_layer(x)\n",
        "        x = x.view(-1, 16 * 42 * 87)\n",
        "        x = F.relu(self.fc1(x))\n",
        "        x = F.relu(self.fc2(x))\n",
        "        x = self.fc3(x)\n",
        "        return x\n",
        "\n",
        "\n",
        "class Print(nn.Module):\n",
        "    \"\"\"\n",
        "    This class prints out the size of the features\n",
        "    \"\"\"\n",
        "\n",
        "    def forward(self, x):\n",
        "        print(x.size())\n",
        "        return x"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "ssiy4F7KKJ2i"
      },
      "source": [
        "Let's define a class to train our CNN."
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "id": "25cZCmYrvEPm"
      },
      "outputs": [],
      "source": [
        "class TrainCNN:\n",
        "    def __init__(self, trainloader, testloader, net, criterion, optimizer):\n",
        "        \"\"\"\n",
        "        Contains the necessary code for training a CNN.\n",
        "\n",
        "        Parameters\n",
        "          trainloader (torch.utils.data.DataLoader): PyTorch dataloader that loads the\n",
        "              predictors and predictands for the train dataset\n",
        "          testloader (torch.utils.data.DataLoader): PyTorch dataloader that loads the\n",
        "              predictors and predictands for the test dataset\n",
        "          net (nn.Module): Neural network architecture\n",
        "          criterion (nn): Loss function (i.e. root mean squared error)\n",
        "          optimizer (torch.optim): Optimizer to use update the neural network\n",
        "              architecture to minimize the loss function\n",
        "        \"\"\"\n",
        "        self.trainloader = trainloader\n",
        "        self.testloader = testloader\n",
        "        self.net = net\n",
        "        self.criterion = criterion\n",
        "        self.optimizer = optimizer\n",
        "        self.device = \"cuda:0\" if torch.cuda.is_available() else \"cpu\"\n",
        "        self.train_losses = []\n",
        "        self.test_losses = []\n",
        "        self.best_loss = np.infty\n",
        "\n",
        "    def _calculate_loss(self, data):\n",
        "        \"\"\"\n",
        "        Calculates the loss of the current model.\n",
        "\n",
        "        Parameters:\n",
        "          data (torch.Tensor): Predictors and predictands against which to\n",
        "              calculate loss of the current model\n",
        "\n",
        "        Returns:\n",
        "            torch.Tensor\n",
        "        \"\"\"\n",
        "        # Get a mini-batch of predictors and predictands\n",
        "        batch_predictors, batch_predictands = data\n",
        "        batch_predictands = batch_predictands.to(self.device)\n",
        "        batch_predictors = batch_predictors.to(self.device)\n",
        "\n",
        "        # Zero the parameter gradients\n",
        "        self.optimizer.zero_grad()\n",
        "\n",
        "        # Calculate the predictions of the current neural network\n",
        "        predictions = self.net(batch_predictors).squeeze(1)\n",
        "\n",
        "        # Quantify the quality of the predictions using a\n",
        "        # loss function (aka criterion) that is differentiable\n",
        "        loss = self.criterion(predictions, batch_predictands)\n",
        "        return loss\n",
        "\n",
        "    def _train(self):\n",
        "        \"\"\"\n",
        "        Trains the CNN for one iteration.\n",
        "\n",
        "        Returns:\n",
        "            None: Appends the current training loss\n",
        "        \"\"\"\n",
        "        running_loss = 0.0\n",
        "        self.net.train()\n",
        "\n",
        "        for i, data in enumerate(self.trainloader):\n",
        "            loss = self._calculate_loss(data)\n",
        "\n",
        "            # The backward pass: calculates the gradients of each weight\n",
        "            # of the neural network with respect to the loss\n",
        "            loss.backward()\n",
        "\n",
        "            # The optimizer updates the weights of the neural network\n",
        "            # based on the gradients calculated above and the choice\n",
        "            # of optimization algorithm\n",
        "            self.optimizer.step()\n",
        "\n",
        "            # Save the model weights that have the best performance!\n",
        "            running_loss += loss.item()\n",
        "\n",
        "        self.train_losses.append(running_loss / len(self.trainloader))\n",
        "\n",
        "    def _evaluate_epoch(self):\n",
        "        \"\"\"\n",
        "        Evaluates the test loss at the current epoch and updates the best loss\n",
        "        if current loss is less than the previous best loss.\n",
        "\n",
        "        Return:\n",
        "          None: Appends the current test loss\n",
        "        \"\"\"\n",
        "        running_loss = 0.0\n",
        "        self.net.eval()\n",
        "\n",
        "        for i, data in enumerate(self.testloader):\n",
        "            loss = self._calculate_loss(data)\n",
        "            running_loss += loss.item()\n",
        "\n",
        "        if running_loss < self.best_loss:\n",
        "            self.best_loss = running_loss\n",
        "            torch.save(self.net, \"current_best_cnn.pt\")\n",
        "\n",
        "        self.test_losses.append(running_loss / len(self.testloader))\n",
        "\n",
        "    def train_network(self, num_epochs=40, verbose=True):\n",
        "        \"\"\"\n",
        "        Trains a neural network for a specified number of epochs.\n",
        "\n",
        "        Parameters:\n",
        "          num_epochs (int): Number of epochs for which to train the network\n",
        "          verbose (bool): Whether or not to print the current train and test loss\n",
        "\n",
        "        Returns:\n",
        "          None\n",
        "        \"\"\"\n",
        "        net = self.net.to(self.device)\n",
        "        for epoch in range(num_epochs):\n",
        "            self._train()\n",
        "            self._evaluate_epoch()\n",
        "            if verbose:\n",
        "                print(\n",
        "                    \"train Set: Epoch {:02d}. loss: {:3f}\".format(\n",
        "                        epoch + 1, self.train_losses[-1]\n",
        "                    )\n",
        "                )\n",
        "                print(\n",
        "                    \"test Set: Epoch {:02d}. loss: {:3f}\".format(\n",
        "                        epoch + 1, self.test_losses[-1]\n",
        "                    )\n",
        "                )\n",
        "\n",
        "        self.net = torch.load(\"current_best_cnn.pt\")\n",
        "        self.net.eval()\n",
        "        self.net.to(self.device)\n",
        "\n",
        "    def evaluate_best_model(self):\n",
        "        \"\"\"\n",
        "        Makes predictions against the current best model.\n",
        "\n",
        "        Returns:\n",
        "          np.array: Predictions from the model\n",
        "        \"\"\"\n",
        "        if self.best_loss == np.infty:\n",
        "            raise Exception(\"The model has not yet been trained!\")\n",
        "\n",
        "        predictions = np.asarray([])\n",
        "        for i, data in enumerate(self.testloader):\n",
        "            batch_predictors, batch_predictands = data\n",
        "            batch_predictands = batch_predictands.to(self.device)\n",
        "            batch_predictors = batch_predictors.to(self.device)\n",
        "\n",
        "            batch_predictions = self.net(batch_predictors).squeeze()\n",
        "            # Edge case: if there is 1 item in the batch, batch_predictions becomes a float\n",
        "            # not a Tensor. the if statement below converts it to a Tensor\n",
        "            # so that it is compatible with np.concatenate\n",
        "            if len(batch_predictions.size()) == 0:\n",
        "                batch_predictions = torch.Tensor([batch_predictions])\n",
        "            predictions = np.concatenate(\n",
        "                [predictions, batch_predictions.detach().cpu().numpy()]\n",
        "            )\n",
        "        return predictions"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "vQ02x64HEkeS"
      },
      "source": [
        "Now we'll kick off a new CodeCarbon experiment to track emissions from training this simple CNN."
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "id": "ruN5CBizVJmK"
      },
      "outputs": [],
      "source": [
        "experiment_name = \"cnn\"\n",
        "\n",
        "tracker = EmissionsTracker(\n",
        "    output_dir=\"./code_carbon/\",\n",
        "    output_file=f\"{experiment_name}_emissions.csv\",\n",
        "    log_level=\"error\",  # comment out this line to see regular output\n",
        ")\n",
        "tracker.start()"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "id": "ESbl0borVJiA"
      },
      "outputs": [],
      "source": [
        "# Assemble numpy arrays corresponding to predictors and predictands\n",
        "train_start_date = \"1960-01-01\"\n",
        "train_end_date = \"2005-12-31\"\n",
        "num_input_time_steps = 1\n",
        "lead_time = 1\n",
        "\n",
        "train_predictors, train_predictands = assemble_predictors_predictands(\n",
        "    train_start_date,\n",
        "    train_end_date,\n",
        "    lead_time,\n",
        "    num_input_time_steps=num_input_time_steps,\n",
        ")\n",
        "test_predictors, test_predictands = assemble_predictors_predictands(\n",
        "    \"2007-01-01\",\n",
        "    \"2017-12-31\",\n",
        "    lead_time,\n",
        "    num_input_time_steps=num_input_time_steps,\n",
        ")\n",
        "\n",
        "# Convert the numpy ararys into ENSODataset, which is a subset of the\n",
        "# torch.utils.data.Dataset class.  This class is compatible with\n",
        "# the torch dataloader, which allows for data loading for a CNN\n",
        "train_dataset = ENSODataset(train_predictors, train_predictands)\n",
        "test_dataset = ENSODataset(test_predictors, test_predictands)\n",
        "\n",
        "# Create a torch.utils.data.DataLoader from the ENSODatasets() created earlier!\n",
        "trainloader = DataLoader(train_dataset, batch_size=10)\n",
        "testloader = DataLoader(test_dataset, batch_size=10)"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "XobO2iPzuSnM",
        "outputId": "c8c19403-eb74-4d96-daeb-f2f2be59d94b"
      },
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "train Set: Epoch 01. loss: 0.495984\n",
            "test Set: Epoch 01. loss: 0.224990\n",
            "train Set: Epoch 02. loss: 0.144434\n",
            "test Set: Epoch 02. loss: 0.147177\n",
            "train Set: Epoch 03. loss: 0.075182\n",
            "test Set: Epoch 03. loss: 0.088368\n",
            "train Set: Epoch 04. loss: 0.068119\n",
            "test Set: Epoch 04. loss: 0.145527\n",
            "train Set: Epoch 05. loss: 0.097358\n",
            "test Set: Epoch 05. loss: 0.146758\n",
            "train Set: Epoch 06. loss: 0.074480\n",
            "test Set: Epoch 06. loss: 0.165589\n",
            "train Set: Epoch 07. loss: 0.092748\n",
            "test Set: Epoch 07. loss: 0.123914\n",
            "train Set: Epoch 08. loss: 0.060070\n",
            "test Set: Epoch 08. loss: 0.106386\n",
            "train Set: Epoch 09. loss: 0.050162\n",
            "test Set: Epoch 09. loss: 0.101152\n",
            "train Set: Epoch 10. loss: 0.050120\n",
            "test Set: Epoch 10. loss: 0.111407\n",
            "train Set: Epoch 11. loss: 0.060003\n",
            "test Set: Epoch 11. loss: 0.174386\n",
            "train Set: Epoch 12. loss: 0.060206\n",
            "test Set: Epoch 12. loss: 0.176652\n",
            "train Set: Epoch 13. loss: 0.051918\n",
            "test Set: Epoch 13. loss: 0.115601\n",
            "train Set: Epoch 14. loss: 0.053289\n",
            "test Set: Epoch 14. loss: 0.091077\n",
            "train Set: Epoch 15. loss: 0.047325\n",
            "test Set: Epoch 15. loss: 0.138683\n",
            "train Set: Epoch 16. loss: 0.037936\n",
            "test Set: Epoch 16. loss: 0.131802\n",
            "train Set: Epoch 17. loss: 0.036714\n",
            "test Set: Epoch 17. loss: 0.109128\n",
            "train Set: Epoch 18. loss: 0.032300\n",
            "test Set: Epoch 18. loss: 0.103742\n",
            "train Set: Epoch 19. loss: 0.026258\n",
            "test Set: Epoch 19. loss: 0.119534\n",
            "train Set: Epoch 20. loss: 0.026654\n",
            "test Set: Epoch 20. loss: 0.137636\n",
            "train Set: Epoch 21. loss: 0.022095\n",
            "test Set: Epoch 21. loss: 0.130337\n",
            "train Set: Epoch 22. loss: 0.018588\n",
            "test Set: Epoch 22. loss: 0.120702\n",
            "train Set: Epoch 23. loss: 0.019395\n",
            "test Set: Epoch 23. loss: 0.122992\n",
            "train Set: Epoch 24. loss: 0.019849\n",
            "test Set: Epoch 24. loss: 0.122877\n",
            "train Set: Epoch 25. loss: 0.022911\n",
            "test Set: Epoch 25. loss: 0.119652\n",
            "train Set: Epoch 26. loss: 0.021141\n",
            "test Set: Epoch 26. loss: 0.119237\n",
            "train Set: Epoch 27. loss: 0.020683\n",
            "test Set: Epoch 27. loss: 0.131150\n",
            "train Set: Epoch 28. loss: 0.020447\n",
            "test Set: Epoch 28. loss: 0.163565\n",
            "train Set: Epoch 29. loss: 0.022538\n",
            "test Set: Epoch 29. loss: 0.181687\n",
            "train Set: Epoch 30. loss: 0.021941\n",
            "test Set: Epoch 30. loss: 0.152514\n",
            "train Set: Epoch 31. loss: 0.020372\n",
            "test Set: Epoch 31. loss: 0.124249\n",
            "train Set: Epoch 32. loss: 0.026006\n",
            "test Set: Epoch 32. loss: 0.131925\n",
            "train Set: Epoch 33. loss: 0.032986\n",
            "test Set: Epoch 33. loss: 0.134884\n",
            "train Set: Epoch 34. loss: 0.024455\n",
            "test Set: Epoch 34. loss: 0.142163\n",
            "train Set: Epoch 35. loss: 0.027444\n",
            "test Set: Epoch 35. loss: 0.131139\n",
            "train Set: Epoch 36. loss: 0.022917\n",
            "test Set: Epoch 36. loss: 0.203121\n",
            "train Set: Epoch 37. loss: 0.032641\n",
            "test Set: Epoch 37. loss: 0.196723\n",
            "train Set: Epoch 38. loss: 0.026717\n",
            "test Set: Epoch 38. loss: 0.259870\n",
            "train Set: Epoch 39. loss: 0.023798\n",
            "test Set: Epoch 39. loss: 0.156866\n",
            "train Set: Epoch 40. loss: 0.040343\n",
            "test Set: Epoch 40. loss: 0.138916\n"
          ]
        }
      ],
      "source": [
        "# Train the model\n",
        "net = CNN(num_input_time_steps=num_input_time_steps)\n",
        "optimizer = optim.Adam(net.parameters(), lr=0.0001)\n",
        "cnn_train_session = TrainCNN(\n",
        "    trainloader, testloader, net, criterion=nn.MSELoss(), optimizer=optimizer\n",
        ")\n",
        "cnn_train_session.train_network()\n",
        "\n",
        "# Evaluate predictions from the best performing model\n",
        "predictions = cnn_train_session.evaluate_best_model()"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 472
        },
        "id": "Qf_2d16rVJqT",
        "outputId": "5c6bf2e5-15ec-4bd9-8071-937daeae59e6"
      },
      "outputs": [
        {
          "output_type": "display_data",
          "data": {
            "text/plain": [
              "<Figure size 640x480 with 1 Axes>"
            ],
            "image/png": "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\n"
          },
          "metadata": {}
        }
      ],
      "source": [
        "# Plot predictions against actuals over time\n",
        "corr, _ = pearsonr(test_predictands, predictions)\n",
        "cnn_rmse = mean_squared_error(test_predictands, predictions, squared=False)\n",
        "plot_nino_time_series(\n",
        "    test_predictands,\n",
        "    predictions,\n",
        "    \"Predictions. Corr: {:3f}. RMSE: {:3f}.\".format(corr, cnn_rmse),\n",
        ")"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "j2tyYHR_Vkc9",
        "outputId": "ac8b4b6d-0bb8-491f-821f-78c88590ad59"
      },
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Emissions from this training run: 0.000472 kg CO2eq\n"
          ]
        }
      ],
      "source": [
        "# Stopping the tracker will create a new file (emissions.csv) in our notebook's\n",
        "# working directory with info on the emissions generated from this experiment\n",
        "cnn_emissions = tracker.stop()\n",
        "print(f\"Emissions from this training run: {cnn_emissions:5f} kg CO2eq\")"
      ]
    },
    {
      "cell_type": "markdown",
      "source": [
        "⭐ **CHALLENGE:** Try reducing the size of the CNN by removing the second linear layer within the CNN class. How does this change affect the model's performance and efficiency in terms of emissions?"
      ],
      "metadata": {
        "id": "6w9iTWG9_CUX"
      }
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "kdTyK7_fFQb1"
      },
      "source": [
        "### **3.2.3 Comparing Emissions and Accuracy**\n",
        "\n",
        "Let's compare the emissions from training a linear regression model with those from training a CNN."
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "6JA03htemKF3",
        "outputId": "7aab482e-2f77-4535-c8e9-106f1f91674d"
      },
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "CNN emissions account for 11.5 times the emissions from linear regression.\n"
          ]
        }
      ],
      "source": [
        "print(\n",
        "    f\"CNN emissions account for {cnn_emissions/baseline_emissions:.1f} times the emissions from linear regression.\"\n",
        ")"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "nYYt3OG4Tz6B"
      },
      "source": [
        "Wow! Linear regression is the clear environmental winner in terms of reducing carbon impact. Of course, we must also consider the performance of our model. How well does it forecast ENSO? Let's compare the root squared mean error (RMSE) from both models. Remember that a lower RMSE indicates a better fit."
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "8KDcIe1ZmM-K",
        "outputId": "682e6fe3-9467-4b23-9ad5-35bc842ee118"
      },
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Previous baseline RMSE: 0.525526\n",
            "Current CNN RMSE: 0.292933\n"
          ]
        }
      ],
      "source": [
        "print(f\"Previous baseline RMSE: {baseline_rmse:5f}\")\n",
        "print(f\"Current CNN RMSE: {cnn_rmse:5f}\")"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "NKqpy6JYTzCh"
      },
      "source": [
        "As expected, those extra emissions from training the CNN resulted in a performance boost. Research indicates that ML model quality has a diminishing power-law relationship with energy consumption ($ Q\\:α \\: E^{-k} $, where $Q$ is model quality and $E$ is energy consumed). However, the power of this relationship ($k$) is very small (0.002-0.004) [7]. This means we need to expend a lot more energy for marginal performance gains.\n",
        "\n",
        "Whether or not this extra accuracy is necessary depends on your situation. For example, we'd prioritize accuracy at all costs for a model that detects lung cancer from CT scans. We wouldn't want to miss any early indications on a sick patient and conversely, we wouldn't want to cause undue stress by incorrectly diagnosing a healthy patient with cancer. On the other hand, we might not need the most accurate model when predicting which customers are likely to churn from an online retailer since the cost of being wrong is much lower than our lung cancer example.\n",
        "\n",
        "Let's visualize the relationship between emissions and performance for these two models."
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "id": "odrW_38Iz7Wg"
      },
      "outputs": [],
      "source": [
        "def plot_comparison(results):\n",
        "    \"\"\"\n",
        "    Plots GHG emissions against RMSE.\n",
        "    \"\"\"\n",
        "    fig, ax = plt.subplots(figsize=(5, 5))\n",
        "    ax.bar(results[\"Category\"], results[\"Emissions\"], label=\"Emissions (g CO2eq)\")\n",
        "    ax.set_ylabel(\"Emissions (g CO2eq)\")\n",
        "\n",
        "    ax2 = ax.twinx()\n",
        "    ax2.scatter(\n",
        "        results[\"Category\"], results[\"RMSE\"], color=\"orange\", label=\"RMSE\",\n",
        "    )\n",
        "    ax2.set_ylabel(\"RMSE\")\n",
        "    ax2.set_ylim([0, max(results[\"RMSE\"]) * 1.2])\n",
        "\n",
        "    # ask matplotlib for the plotted objects and their labels\n",
        "    lines, labels = ax.get_legend_handles_labels()\n",
        "    lines2, labels2 = ax2.get_legend_handles_labels()\n",
        "    ax2.legend(\n",
        "        lines + lines2, labels + labels2, loc=\"center left\", bbox_to_anchor=(1.1, 0.9)\n",
        "    );"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 445
        },
        "id": "QjlMQsn0-RdK",
        "outputId": "a2ebe6fb-9f42-478d-ab6c-8d68bb342647"
      },
      "outputs": [
        {
          "output_type": "display_data",
          "data": {
            "text/plain": [
              "<Figure size 500x500 with 2 Axes>"
            ],
            "image/png": "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\n"
          },
          "metadata": {}
        }
      ],
      "source": [
        "model_results = pd.DataFrame(\n",
        "    {\n",
        "        \"Category\": [\"Linear Regression\", \"CNN\"],\n",
        "        \"RMSE\": [baseline_rmse, cnn_rmse],\n",
        "        \"Emissions\": [baseline_emissions * 1000, cnn_emissions * 1000],\n",
        "    }\n",
        ")\n",
        "\n",
        "plot_comparison(model_results)"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "yha0o4HrxSpz"
      },
      "source": [
        "### **3.2.4 Tracking Emissions from Model Inference**\n",
        "\n",
        "As noted in our [ChatGPT example](https://colab.research.google.com/drive/1tTok_j2MZtWLuvcr6p_labPOxnCPM9ZL#scrollTo=1_3_1_Case_study_ChatGPT_) above, we must also account for the emissions associated with productionalizing a model. This consideration is situation-specific. Will millions of users predict using our model multiple times a day? Or will we only need to score new data with our model once a month?\n",
        "\n",
        "We can use CodeCarbon again to assess the emissions associated with model inference. Note that we would need to deploy CodeCarbon on the infrastructure that hosts the productionalized model for an accurate assessment of the emissions associated with scoring. We'll measure the emissions associated with inference within the Colab notebook for illustration purposes only."
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "19EepYna1Gp4",
        "outputId": "81004f08-f76b-453a-fbd6-d967063f3c61"
      },
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Emissions from this training run: 1.074798278396683e-07 kg CO2eq\n"
          ]
        }
      ],
      "source": [
        "# First we'll measure emissions from scoring with the linear regression model\n",
        "experiment_name = \"linear_regression_inference\"\n",
        "\n",
        "tracker = EmissionsTracker(\n",
        "    output_dir=\"./code_carbon/\",\n",
        "    output_file=f\"{experiment_name}_emissions.csv\",\n",
        "    log_level=\"error\",  # comment out this line to see regular output\n",
        ")\n",
        "tracker.start()\n",
        "\n",
        "predictions_test_lin_reg = regr.predict(X_test)\n",
        "\n",
        "lin_reg_inference_emissions = tracker.stop()\n",
        "print(f\"Emissions from this training run: {lin_reg_inference_emissions} kg CO2eq\")"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "QRlm7AHB77N-",
        "outputId": "42237e13-5e4b-4177-dc12-74f8eeeac270"
      },
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Emissions from this training run: 1.8650008511342655e-07 kg CO2eq\n"
          ]
        }
      ],
      "source": [
        "# Now we measure emissions from scoring with the CNN\n",
        "experiment_name = \"cnn_inference\"\n",
        "\n",
        "tracker = EmissionsTracker(\n",
        "    output_dir=\"./code_carbon/\",\n",
        "    output_file=f\"{experiment_name}_emissions.csv\",\n",
        "    log_level=\"error\",  # comment out this line to see regular output\n",
        ")\n",
        "tracker.start()\n",
        "\n",
        "device = \"cuda:0\" if torch.cuda.is_available() else \"cpu\"\n",
        "predictions_test_cnn = cnn_train_session.net(\n",
        "    torch.Tensor(test_predictors).to(device)\n",
        ").squeeze()\n",
        "\n",
        "cnn_inference_emissions = tracker.stop()\n",
        "print(f\"Emissions from this training run: {cnn_inference_emissions} kg CO2eq\")"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "-nDJcoKU9X6P",
        "outputId": "34007d20-d20a-453d-9e86-e7c491cfb18e"
      },
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "The emissions from scoring with a CNN are 1.7 times the emissions from scoring with a linear regression model.\n"
          ]
        }
      ],
      "source": [
        "print(\n",
        "    f\"The emissions from scoring with a CNN are {cnn_inference_emissions/lin_reg_inference_emissions:.1f} times the emissions from scoring with a linear regression model.\"\n",
        ")"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "3JZaqn0f9enX"
      },
      "source": [
        "The difference in emissions from inference between the two models is much less dramatic than that from training. One reason for this is that deep learning models train over multiple epochs. Neural networks then typically require three times the energy to train on an observation as compared to inferring on that observation [8].\n",
        "\n",
        "Be sure to measure emissions from scoring data samples of the scale you'd expect once you've deployed your model for the most accurate estimates!"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "knR0BrAyGpsg"
      },
      "source": [
        "## **3.3 Hyperparameter Tuning**\n",
        "\n",
        "Hyperparameter tuning refers to the task of adjusting the data that governs the training process itself, such as the number of layers in a neural network. ML practitioners typically tune these hyperparameters to find a combination with an optimal model performance.\n",
        "\n",
        "Let's explore the emissions associated with tuning the hyperparameters of our CNN and experiment with some strategies to make this process more efficient.\n",
        "\n",
        "Note that in the interest of time, we will only explore a small hyperparameter space. However, these strategies are much more effective at reducing emissions when we need to search a large number of possible hyperparameter combinations to optimize model performance."
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "sXPq3ST28lae"
      },
      "source": [
        "#### **3.2.1 Introducing Optuna**\n",
        "\n",
        "We'll use the [`optuna`](https://optuna.org/) package to tune our hyperparameters. To do so, we need to define an `objective` function such as the one below.\n",
        "\n",
        "The `objective` function takes an input parameter called `trial`, which comes from Optuna's `Trial` class. This object logs a set of selected hyperparameter values and records the value of our objective function (RMSE in our case) from each trial.\n",
        "\n",
        "We also define the search space of each hyperparameter as a dictionary named `params`. For each hyperparameter, we specify the range of the search space with the `suggest_*` methods.\n",
        "\n",
        "- [**suggest_int:**](https://optuna.readthedocs.io/en/stable/reference/generated/optuna.trial.Trial.html#optuna.trial.Trial.suggest_int) Suggests a value for a hyperparameter of type integer.\n",
        "- [**suggest_categorical:**](https://optuna.readthedocs.io/en/stable/reference/generated/optuna.trial.Trial.html#optuna.trial.Trial.suggest_categorical) Suggests a value for a categorical hyperparameter.\n",
        "-[**suggest_float:**](https://optuna.readthedocs.io/en/stable/reference/generated/optuna.trial.Trial.html#optuna.trial.Trial.suggest_float) Suggests a value for a hyperparameter of type float. Note that you can specify a log distribution by setting the parameter `log=True`.\n",
        "\n",
        "For our hyperparameter tuning session, we'll experiment with adjusting the learning rate and the optimizer of our CNN as well as the number of input time steps used to construct our training data.\n",
        "\n",
        "> _Quick note:_ Remember that the parameter `num_input_time_steps` refers to the number of time steps for each predictor. For example, if num_input_time_steps is set to 3, then the machine learning model will take 3 months as its input. In other words, a sample predictor will be 3 months of average temperature: Jan average temperature, Feb average temperature, and March average temperature (i.e. an array of (3, lat, lon)). The predictand will be the Niño3.4 index lead_time months after March.\n",
        ">\n",
        "> Previously, we had trained our CNN using a `num_input_time_steps` value of 1."
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "id": "5MXzDYQoKCsF"
      },
      "outputs": [],
      "source": [
        "def objective(trial):\n",
        "    \"\"\"\n",
        "    Defines the objective function.\n",
        "\n",
        "    Parameters:\n",
        "      trial (optuna.Trial): Provides interfaces to Optuna's functionality\n",
        "\n",
        "    Returns:\n",
        "      (int): RMSE of current trial\n",
        "    \"\"\"\n",
        "    # Define the hyperparameter search space\n",
        "    params = {\n",
        "        \"num_input_time_steps\": trial.suggest_int(\"num_input_time_steps\", 1, 3),\n",
        "        \"learning_rate\": trial.suggest_float(\"learning_rate\", 1e-4, 1e-2, log=True),\n",
        "        \"optimizer\": trial.suggest_categorical(\"optimizer\", [\"Adam\", \"SGD\"]),\n",
        "    }\n",
        "\n",
        "    # Instantiate the model\n",
        "    net = CNN(num_input_time_steps=params[\"num_input_time_steps\"])\n",
        "\n",
        "    # Create new train and test sets based on the selected num_input_time_steps parameter\n",
        "    train_predictors, train_predictands = assemble_predictors_predictands(\n",
        "        start_date=train_start_date,\n",
        "        end_date=train_end_date,\n",
        "        lead_time=1,\n",
        "        num_input_time_steps=params[\"num_input_time_steps\"],\n",
        "    )\n",
        "    test_predictors, test_predictands = assemble_predictors_predictands(\n",
        "        start_date=\"2007-01-01\",\n",
        "        end_date=\"2017-12-31\",\n",
        "        lead_time=1,\n",
        "        num_input_time_steps=params[\"num_input_time_steps\"],\n",
        "    )\n",
        "\n",
        "    train_dataset = ENSODataset(train_predictors, train_predictands)\n",
        "    test_dataset = ENSODataset(test_predictors, test_predictands)\n",
        "\n",
        "    trainloader = DataLoader(train_dataset, batch_size=10)\n",
        "    testloader = DataLoader(test_dataset, batch_size=10)\n",
        "\n",
        "    # Specify criterion and optimizer\n",
        "    criterion = torch.nn.MSELoss()\n",
        "    optimizer = getattr(optim, params[\"optimizer\"])(\n",
        "        net.parameters(), lr=params[\"learning_rate\"]\n",
        "    )\n",
        "\n",
        "    # Train a CNN using the specified hyperparameters and calculate RMSE\n",
        "    cnn_train_session = TrainCNN(trainloader, testloader, net, criterion, optimizer)\n",
        "    cnn_train_session.train_network(verbose=False)\n",
        "    predictions = cnn_train_session.evaluate_best_model()\n",
        "    rmse = mean_squared_error(test_predictands, predictions, squared=False)\n",
        "    return rmse"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "GjHmj7fMyMXQ"
      },
      "source": [
        "#### **3.3.2 TPE Search**\n",
        "\n",
        "You might already be familiar with tuning hyperparameters via grid search in which we try every combination of hyperparameters within the space. Grid search is inherently inefficient and should be avoided in favor of more intelligent search algorithms.\n",
        "\n",
        "One such algorithm is the Tree-structured Parzen Estimator (TPE) algorithm, which is the default hyperparameter sampling method in Optuna. On each trial for each hyperparameter, TPE fits a Gaussian Mixture Model (GMM) $l(x)$ to the hyperparameter sets that return the smallest values for the objective function.\n",
        "\n",
        "> _Quick note:_ A Gaussian mixture model is a probabilistic model that represents normally distributed subpopulations (in this case, our hyperparameters) within an overall population. Please see this [short introduction](https://brilliant.org/wiki/gaussian-mixture-model/) for a deeper explanation.\n",
        "\n",
        "TPE fits another GMM $g(x)$ to the remaining hyperparameter values. The hyperparameter value $x$ that maximizes the ratio $l(x)/g(x)$ would then be evaluated using the objective function. I recommend this [blog post](https://towardsdatascience.com/a-conceptual-explanation-of-bayesian-model-based-hyperparameter-optimization-for-machine-learning-b8172278050f#) for more information about Bayesian-based hyperparameter tuning strategies.\n",
        "\n",
        "TPE will run for as many trials as we specify. Let's limit ourselves to 12 trials to ensure we're not waiting too long for a result."
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "id": "mox5pvzw1Jus"
      },
      "outputs": [],
      "source": [
        "experiment_name = \"tpe_search\"\n",
        "\n",
        "tracker = EmissionsTracker(\n",
        "    output_dir=\"./code_carbon/\",\n",
        "    output_file=f\"{experiment_name}_emissions.csv\",\n",
        "    log_level=\"error\",  # comment out this line to see regular output\n",
        ")\n",
        "tracker.start()"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "MKOuRS5pAIa1"
      },
      "source": [
        "The TPE algorithm cannot operate on a \"cold start\"&mdash;in other words, TPE needs the results of some \"startup\" trials to fit the GMM $l(x)$. We set the parameter `n_startup_trials` to specify the number of trials that will initially leverage random sampling instead of TPE. The default value for this parameter is 10 but since we're only searching over 12 trials in total, we'll reduce this number to a lower value. Let's try 5.\n",
        "\n",
        "⭐ **CHALLENGE:** Experiment with adjusting the parameter `n_startup_trials`. How does it affect the RMSE from the best trial of the study?"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 329,
          "referenced_widgets": [
            "73f4f105a91a4fafbae59391ebd0fb65",
            "510551aea3084555b4c3982528cce9b8",
            "b395e09ac1084567ac824ea50517cef9",
            "7ed85909f48045dbbadbc804c9464c82",
            "400faafaa2e148718bc98268be1ca6c2",
            "db109a8c2e00496695aba33d2d82ee1b",
            "8388bad8267c459c861ec6fbeef7057f",
            "9a771cf17912472c870f33705b9539e0",
            "bcbccc61d20b42e0910abddbb8ed8c59",
            "1501e6948fa4445d9c8adc5b3f709e72",
            "fbf9681dc0704297a5f832f610adb311"
          ]
        },
        "id": "6uaU8g6avEH0",
        "outputId": "58c78f42-5098-49a3-9098-c2999eece557"
      },
      "outputs": [
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "\u001b[32m[I 2023-04-18 14:02:59,082]\u001b[0m A new study created in memory with name: no-name-63b1fd40-2f7b-4087-aa2c-a86f7c1af1fc\u001b[0m\n",
            "/usr/local/lib/python3.9/dist-packages/optuna/progress_bar.py:56: ExperimentalWarning: Progress bar is experimental (supported from v1.2.0). The interface can change in the future.\n",
            "  self._init_valid()\n"
          ]
        },
        {
          "output_type": "display_data",
          "data": {
            "text/plain": [
              "  0%|          | 0/12 [00:00<?, ?it/s]"
            ],
            "application/vnd.jupyter.widget-view+json": {
              "version_major": 2,
              "version_minor": 0,
              "model_id": "73f4f105a91a4fafbae59391ebd0fb65"
            }
          },
          "metadata": {}
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "\u001b[32m[I 2023-04-18 14:03:24,633]\u001b[0m Trial 0 finished with value: 0.36426272781012387 and parameters: {'num_input_time_steps': 2, 'learning_rate': 0.0026938830192854116, 'optimizer': 'Adam'}. Best is trial 0 with value: 0.36426272781012387.\u001b[0m\n",
            "\u001b[32m[I 2023-04-18 14:03:45,514]\u001b[0m Trial 1 finished with value: 0.2839215513366501 and parameters: {'num_input_time_steps': 2, 'learning_rate': 0.0019578897201213002, 'optimizer': 'SGD'}. Best is trial 1 with value: 0.2839215513366501.\u001b[0m\n",
            "\u001b[32m[I 2023-04-18 14:04:13,282]\u001b[0m Trial 2 finished with value: 0.31790962089293673 and parameters: {'num_input_time_steps': 3, 'learning_rate': 0.0005846326121643416, 'optimizer': 'Adam'}. Best is trial 1 with value: 0.2839215513366501.\u001b[0m\n",
            "\u001b[32m[I 2023-04-18 14:04:33,861]\u001b[0m Trial 3 finished with value: 0.30769743787810766 and parameters: {'num_input_time_steps': 2, 'learning_rate': 0.007098936257405904, 'optimizer': 'SGD'}. Best is trial 1 with value: 0.2839215513366501.\u001b[0m\n",
            "\u001b[32m[I 2023-04-18 14:04:51,815]\u001b[0m Trial 4 finished with value: 0.26384994459605826 and parameters: {'num_input_time_steps': 1, 'learning_rate': 0.0046263628414755864, 'optimizer': 'SGD'}. Best is trial 4 with value: 0.26384994459605826.\u001b[0m\n",
            "\u001b[32m[I 2023-04-18 14:05:11,342]\u001b[0m Trial 5 finished with value: 0.3774224359766639 and parameters: {'num_input_time_steps': 1, 'learning_rate': 0.0001895689667719594, 'optimizer': 'SGD'}. Best is trial 4 with value: 0.26384994459605826.\u001b[0m\n",
            "\u001b[32m[I 2023-04-18 14:05:28,992]\u001b[0m Trial 6 finished with value: 0.25027443607875655 and parameters: {'num_input_time_steps': 1, 'learning_rate': 0.008573519729186104, 'optimizer': 'SGD'}. Best is trial 6 with value: 0.25027443607875655.\u001b[0m\n",
            "\u001b[32m[I 2023-04-18 14:05:46,535]\u001b[0m Trial 7 finished with value: 0.2595924550209894 and parameters: {'num_input_time_steps': 1, 'learning_rate': 0.008761933655144947, 'optimizer': 'SGD'}. Best is trial 6 with value: 0.25027443607875655.\u001b[0m\n",
            "\u001b[32m[I 2023-04-18 14:06:15,049]\u001b[0m Trial 8 finished with value: 0.38427043407099654 and parameters: {'num_input_time_steps': 3, 'learning_rate': 0.0010404102976536128, 'optimizer': 'Adam'}. Best is trial 6 with value: 0.25027443607875655.\u001b[0m\n",
            "\u001b[32m[I 2023-04-18 14:06:32,639]\u001b[0m Trial 9 finished with value: 0.25156572141705186 and parameters: {'num_input_time_steps': 1, 'learning_rate': 0.009049911227773816, 'optimizer': 'SGD'}. Best is trial 6 with value: 0.25027443607875655.\u001b[0m\n",
            "\u001b[32m[I 2023-04-18 14:06:54,231]\u001b[0m Trial 10 finished with value: 0.4049129774059246 and parameters: {'num_input_time_steps': 1, 'learning_rate': 0.003448362630391615, 'optimizer': 'Adam'}. Best is trial 6 with value: 0.25027443607875655.\u001b[0m\n",
            "\u001b[32m[I 2023-04-18 14:07:11,739]\u001b[0m Trial 11 finished with value: 0.25061581240034386 and parameters: {'num_input_time_steps': 1, 'learning_rate': 0.009305841797753199, 'optimizer': 'SGD'}. Best is trial 6 with value: 0.25027443607875655.\u001b[0m\n"
          ]
        }
      ],
      "source": [
        "tpe_search_study = optuna.create_study(\n",
        "    direction=\"minimize\", sampler=optuna.samplers.TPESampler(n_startup_trials=5, seed=0)\n",
        ")\n",
        "tpe_search_study.optimize(objective, n_trials=12, show_progress_bar=True)"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "DdFeNHtvNW7T",
        "outputId": "697be5da-fbce-4160-a268-eb0484a380f5"
      },
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Optimal set of hyperparameters: {'num_input_time_steps': 1, 'learning_rate': 0.008573519729186104, 'optimizer': 'SGD'}\n",
            "RMSE from optimal set of hyperparameters: 0.25027\n",
            "\n",
            "Hyperparameter tuning from TPE search reduced model error by 14.6%\n"
          ]
        }
      ],
      "source": [
        "print(\"Optimal set of hyperparameters:\", tpe_search_study.best_params)\n",
        "print(f\"RMSE from optimal set of hyperparameters: {tpe_search_study.best_value:.5f}\")\n",
        "print(\"\")\n",
        "print(\n",
        "    f\"Hyperparameter tuning from TPE search reduced model error by {abs(1 - tpe_search_study.best_value/cnn_rmse):.1%}\"\n",
        ")"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "BgzPxuf7wKe5"
      },
      "source": [
        "We can see that hyperparameter tuning reduced our model's error by a decent amount. Let's tally up the emissions cost from this exercise."
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "lD8SxF76vEKK",
        "outputId": "e1e604fa-0dae-4399-bb25-faa3b01901ca"
      },
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Emissions from this training run: 0.00361 kg CO2eq\n"
          ]
        }
      ],
      "source": [
        "tpe_search_emissions = tracker.stop()\n",
        "print(f\"Emissions from this training run: {tpe_search_emissions:.5f} kg CO2eq\")"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "tQjBWACSvpHa"
      },
      "source": [
        "Finally, we'll take a quick look at the fit from a CNN trained with this optimal set of hyperparameters."
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "id": "mYb_dIsq5unb"
      },
      "outputs": [],
      "source": [
        "def train_model_with_best_parameters(params):\n",
        "    \"\"\"\n",
        "    Trains a CNN a given set of hyperparameters.\n",
        "\n",
        "    Parameters:\n",
        "      params (dict): Dictionary containing hyperparameter names and values\n",
        "\n",
        "    Returns:\n",
        "      (np.array, np.array): Tuple of predictions and test predictands\n",
        "    \"\"\"\n",
        "    net = CNN(num_input_time_steps=params[\"num_input_time_steps\"])\n",
        "\n",
        "    train_predictors, train_predictands = assemble_predictors_predictands(\n",
        "        start_date=train_start_date,\n",
        "        end_date=train_end_date,\n",
        "        lead_time=2,\n",
        "        num_input_time_steps=params[\"num_input_time_steps\"],\n",
        "    )\n",
        "    test_predictors, test_predictands = assemble_predictors_predictands(\n",
        "        start_date=\"2007-01-01\",\n",
        "        end_date=\"2017-12-31\",\n",
        "        lead_time=2,\n",
        "        num_input_time_steps=params[\"num_input_time_steps\"],\n",
        "    )\n",
        "\n",
        "    train_dataset = ENSODataset(train_predictors, train_predictands)\n",
        "    test_dataset = ENSODataset(test_predictors, test_predictands)\n",
        "\n",
        "    trainloader = DataLoader(train_dataset, batch_size=10)\n",
        "    testloader = DataLoader(test_dataset, batch_size=10)\n",
        "\n",
        "    # Specify criterion and optimizer\n",
        "    criterion = torch.nn.MSELoss()\n",
        "    optimizer = getattr(optim, params[\"optimizer\"])(\n",
        "        net.parameters(), lr=params[\"learning_rate\"]\n",
        "    )\n",
        "\n",
        "    cnn_train_session = TrainCNN(trainloader, testloader, net, criterion, optimizer)\n",
        "    cnn_train_session.train_network(verbose=False)\n",
        "    predictions = cnn_train_session.evaluate_best_model()\n",
        "    return predictions, test_predictands"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 493
        },
        "id": "GEQIY-C15jQ6",
        "outputId": "f76538b7-e1db-4047-877a-dc01e1b5a727"
      },
      "outputs": [
        {
          "output_type": "display_data",
          "data": {
            "text/plain": [
              "<Figure size 640x480 with 1 Axes>"
            ],
            "image/png": "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\n"
          },
          "metadata": {}
        }
      ],
      "source": [
        "predictions, test_predictands = train_model_with_best_parameters(\n",
        "    tpe_search_study.best_params\n",
        ")\n",
        "corr, _ = scipy.stats.pearsonr(predictions, test_predictands)\n",
        "\n",
        "plot_nino_time_series(\n",
        "    test_predictands,\n",
        "    predictions,\n",
        "    \"Predicted and True Nino3.4 Indices on Training Set at 1 Month Lead Time. \\n Corr: {:.2f}\".format(\n",
        "        corr\n",
        "    ),\n",
        ")"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "TN_YH3SqdLJW"
      },
      "source": [
        "#### **3.3.3 With Hyperparameter Pruning**\n",
        "\n",
        "We might wish to stop training a CNN on a certain set of hyperparameters if the intermediate objective values don't look promising. For example, we might train our CNN on certain hyperparameters and notice that the RMSE by epoch 23 is quite poor. In that case, we should terminate this trial and move on to another set of hyperparameters. This is called \"hyperparameter pruning\".\n",
        "\n",
        "To implement hyperparameter pruning, we'll need to adjust our CNN training code. Below I've added lines 36-41 to the function `train_network`. This new code calculates the RMSE at each epoch and calls the Optuna trial to determine whether to prune based on our specified pruning strategy (more on pruning strategies below).\n",
        "\n",
        "\n"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "id": "1rqfyX5yuzbO"
      },
      "outputs": [],
      "source": [
        "class TrainCNNOptuna(TrainCNN):\n",
        "    def __init__(self, trainloader, testloader, net, criterion, optimizer):\n",
        "        super().__init__(trainloader, testloader, net, criterion, optimizer)\n",
        "\n",
        "    def train_network(self, trial, num_epochs=40, verbose=True):\n",
        "        \"\"\"\n",
        "        Trains a neural network for a specified number of epochs with an\n",
        "        ability to prune via Optuna.\n",
        "\n",
        "        Parameters:\n",
        "          trial (optuna.Trial): Provides interfaces to Optuna's functionality\n",
        "          num_epochs (int): Number of epochs for which to train the network\n",
        "          verbose (bool): Whether or not to print the current train and test loss\n",
        "\n",
        "        Returns:\n",
        "          None\n",
        "        \"\"\"\n",
        "        net = self.net.to(self.device)\n",
        "\n",
        "        for epoch in range(num_epochs):\n",
        "            self._train()\n",
        "            self._evaluate_epoch()\n",
        "            if verbose:\n",
        "                print(\n",
        "                    \"train Set: Epoch {:02d}. loss: {:3f}\".format(\n",
        "                        epoch + 1, self.train_losses[-1]\n",
        "                    )\n",
        "                )\n",
        "                print(\n",
        "                    \"test Set: Epoch {:02d}. loss: {:3f}\".format(\n",
        "                        epoch + 1, self.test_losses[-1]\n",
        "                    )\n",
        "                )\n",
        "\n",
        "            # Determine if trial should be pruned\n",
        "            predictions = self.evaluate_best_model()\n",
        "            test_predictands = self.testloader.dataset.predictands\n",
        "            rmse = mean_squared_error(test_predictands, predictions, squared=False)\n",
        "            trial.report(rmse, epoch)\n",
        "            if trial.should_prune():\n",
        "                raise optuna.exceptions.TrialPruned()\n",
        "\n",
        "        self.net = torch.load(\"current_best_cnn.pt\")\n",
        "        self.net.eval()\n",
        "        self.net.to(self.device)"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "XVT4UnykDmFd"
      },
      "source": [
        "We also need to adjust our objective function to call this new training code."
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "id": "E0tSZnGqy5YM"
      },
      "outputs": [],
      "source": [
        "def objective_with_pruning(trial):\n",
        "    \"\"\"\n",
        "    Defines the objective function to allow for pruning.\n",
        "\n",
        "    Parameters:\n",
        "      trial (optuna.Trial): Provides interfaces to Optuna's functionality\n",
        "\n",
        "    Returns:\n",
        "      (int): RMSE of current trial\n",
        "    \"\"\"\n",
        "    params = {\n",
        "        \"num_input_time_steps\": trial.suggest_int(\"num_input_time_steps\", 1, 3),\n",
        "        \"learning_rate\": trial.suggest_float(\"learning_rate\", 1e-4, 1e-2, log=True),\n",
        "        \"optimizer\": trial.suggest_categorical(\"optimizer\", [\"Adam\", \"SGD\"]),\n",
        "    }\n",
        "\n",
        "    net = CNN(num_input_time_steps=params[\"num_input_time_steps\"])\n",
        "\n",
        "    train_predictors, train_predictands = assemble_predictors_predictands(\n",
        "        start_date=train_start_date,\n",
        "        end_date=train_end_date,\n",
        "        lead_time=1,\n",
        "        num_input_time_steps=params[\"num_input_time_steps\"],\n",
        "    )\n",
        "    test_predictors, test_predictands = assemble_predictors_predictands(\n",
        "        start_date=\"2007-01-01\",\n",
        "        end_date=\"2017-12-31\",\n",
        "        lead_time=1,\n",
        "        num_input_time_steps=params[\"num_input_time_steps\"],\n",
        "    )\n",
        "\n",
        "    train_dataset = ENSODataset(train_predictors, train_predictands)\n",
        "    test_dataset = ENSODataset(test_predictors, test_predictands)\n",
        "\n",
        "    trainloader = DataLoader(train_dataset, batch_size=10)\n",
        "    testloader = DataLoader(test_dataset, batch_size=10)\n",
        "\n",
        "    # Specify criterion and optimizer\n",
        "    criterion = torch.nn.MSELoss()\n",
        "    optimizer = getattr(optim, params[\"optimizer\"])(\n",
        "        net.parameters(), lr=params[\"learning_rate\"]\n",
        "    )\n",
        "\n",
        "    cnn_train_session = TrainCNNOptuna(\n",
        "        trainloader, testloader, net, criterion, optimizer\n",
        "    )\n",
        "    cnn_train_session.train_network(trial, verbose=False)\n",
        "    predictions = cnn_train_session.evaluate_best_model()\n",
        "    rmse = mean_squared_error(test_predictands, predictions, squared=False)\n",
        "    return rmse"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "F52CHujTDswq"
      },
      "source": [
        "Optuna offers several different hyperparameter pruning strategies.\n",
        "\n",
        "- `ThresholdPruner`: Prunes a trial if the intermediate value (in our case, RMSE) is worse than a specified threshold at the same training epoch.\n",
        "- `MedianPruner`: Prunes a trial if the intermediate value of the current trial is worse than the median of intermediate values from previous trials at the same training epoch.\n",
        "- `PercentilePruner`: Prunes a trial if the intermediate value of the current trial is in the bottom percentile of trials at the same training epoch.\n",
        "- `SuccessiveHalvingPruner`: Pruning strategy based on the [Successive Halving Algorithm (SHA)](https://blog.ml.cmu.edu/2018/12/12/massively-parallel-hyperparameter-optimization/). At a high level, the algorithm allocates more epochs to promising trials by by pruning half of the total trials.\n",
        "\n",
        "Let's experiment with the `PercentilePruner`. We'll ask Optuna to prune when the RMSE falls below the bottom 10th percentile of previous trials."
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "id": "sx5a4bCNvEMf"
      },
      "outputs": [],
      "source": [
        "experiment_name = \"tpe_search_percentile_pruned\"\n",
        "\n",
        "tracker = EmissionsTracker(\n",
        "    output_dir=\"./code_carbon/\",\n",
        "    output_file=f\"{experiment_name}_emissions.csv\",\n",
        "    log_level=\"error\",  # comment out this line to see regular output\n",
        ")\n",
        "tracker.start()"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 329,
          "referenced_widgets": [
            "bb530fd1cab94a45afdb174945b8c47e",
            "e8b772087f3a48c3be97c555dc1fca85",
            "0e06f4d494b34e89ab7587f1291039e7",
            "dd341850a65f4dbe95b0bba7726ddd14",
            "170f40d037e34544b6ffa1cb4b75153b",
            "2da303b2be4d410595a6672288b836de",
            "34d068f48aa841c5bb7c9f52ef46ddaf",
            "e7069f4a57dd4530895fbd4cacd3582b",
            "7181ba6369774af4a6456c166f38fdb8",
            "e021a0e85c5b4fffa4e53d5a94747691",
            "b0cf2b274bb74ac5ae3195c0e9c69aa4"
          ]
        },
        "id": "k8eh8foZdUVH",
        "outputId": "460ad49b-bc31-4326-eef2-28a058ad28b1"
      },
      "outputs": [
        {
          "output_type": "stream",
          "name": "stderr",
          "text": [
            "\u001b[32m[I 2023-04-18 14:07:31,464]\u001b[0m A new study created in memory with name: no-name-f589afe0-e33f-47db-8a51-1319064ac323\u001b[0m\n",
            "/usr/local/lib/python3.9/dist-packages/optuna/progress_bar.py:56: ExperimentalWarning: Progress bar is experimental (supported from v1.2.0). The interface can change in the future.\n",
            "  self._init_valid()\n"
          ]
        },
        {
          "output_type": "display_data",
          "data": {
            "text/plain": [
              "  0%|          | 0/12 [00:00<?, ?it/s]"
            ],
            "application/vnd.jupyter.widget-view+json": {
              "version_major": 2,
              "version_minor": 0,
              "model_id": "bb530fd1cab94a45afdb174945b8c47e"
            }
          },
          "metadata": {}
        },
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "\u001b[32m[I 2023-04-18 14:07:58,988]\u001b[0m Trial 0 finished with value: 0.5052099326441267 and parameters: {'num_input_time_steps': 2, 'learning_rate': 0.0026938830192854116, 'optimizer': 'Adam'}. Best is trial 0 with value: 0.5052099326441267.\u001b[0m\n",
            "\u001b[32m[I 2023-04-18 14:08:23,097]\u001b[0m Trial 1 finished with value: 0.45308531146479447 and parameters: {'num_input_time_steps': 2, 'learning_rate': 0.0019578897201213002, 'optimizer': 'SGD'}. Best is trial 1 with value: 0.45308531146479447.\u001b[0m\n",
            "\u001b[32m[I 2023-04-18 14:08:53,585]\u001b[0m Trial 2 finished with value: 0.2978675634161205 and parameters: {'num_input_time_steps': 3, 'learning_rate': 0.0005846326121643416, 'optimizer': 'Adam'}. Best is trial 2 with value: 0.2978675634161205.\u001b[0m\n",
            "\u001b[32m[I 2023-04-18 14:09:16,530]\u001b[0m Trial 3 finished with value: 0.2532025348653743 and parameters: {'num_input_time_steps': 2, 'learning_rate': 0.007098936257405904, 'optimizer': 'SGD'}. Best is trial 3 with value: 0.2532025348653743.\u001b[0m\n",
            "\u001b[32m[I 2023-04-18 14:09:35,878]\u001b[0m Trial 4 finished with value: 0.3975199499403062 and parameters: {'num_input_time_steps': 1, 'learning_rate': 0.0046263628414755864, 'optimizer': 'SGD'}. Best is trial 3 with value: 0.2532025348653743.\u001b[0m\n",
            "\u001b[32m[I 2023-04-18 14:09:36,933]\u001b[0m Trial 5 pruned. \u001b[0m\n",
            "\u001b[32m[I 2023-04-18 14:09:38,790]\u001b[0m Trial 6 pruned. \u001b[0m\n",
            "\u001b[32m[I 2023-04-18 14:09:40,725]\u001b[0m Trial 7 pruned. \u001b[0m\n",
            "\u001b[32m[I 2023-04-18 14:09:42,113]\u001b[0m Trial 8 pruned. \u001b[0m\n",
            "\u001b[32m[I 2023-04-18 14:09:43,717]\u001b[0m Trial 9 pruned. \u001b[0m\n",
            "\u001b[32m[I 2023-04-18 14:09:45,833]\u001b[0m Trial 10 pruned. \u001b[0m\n",
            "\u001b[32m[I 2023-04-18 14:09:51,282]\u001b[0m Trial 11 pruned. \u001b[0m\n"
          ]
        }
      ],
      "source": [
        "tpe_search_10perc_pruned_study = optuna.create_study(\n",
        "    direction=\"minimize\",\n",
        "    sampler=optuna.samplers.TPESampler(n_startup_trials=5, seed=0),\n",
        "    pruner=optuna.pruners.PercentilePruner(percentile=10),\n",
        ")\n",
        "tpe_search_10perc_pruned_study.optimize(\n",
        "    objective_with_pruning, n_trials=12, show_progress_bar=True\n",
        ")"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "mR207vv3vEV8",
        "outputId": "741fa451-b9d4-4e81-eedb-c07db5cd1851"
      },
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Emissions from this training run: 0.00192 kg CO2eq\n"
          ]
        }
      ],
      "source": [
        "tpe_search_percentile_pruned_emissions = tracker.stop()\n",
        "print(\n",
        "    f\"Emissions from this training run: {tpe_search_percentile_pruned_emissions:.5f} kg CO2eq\"\n",
        ")"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "egYd4kO2GBgc"
      },
      "source": [
        "Optuna ended up pruning half our total trials! Let's see if this strategy found a comparable hyperparameter set to our previous tuning study in which we never pruned."
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 143
        },
        "id": "Z59R0I7SvEST",
        "outputId": "318bad0b-4ad4-4060-e00c-9fe451e69523"
      },
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "            num_input_time_steps  learning_rate optimizer      RMSE\n",
              "strategy                                                           \n",
              "Not pruned                     1       0.008574       SGD  0.250274\n",
              "Pruned                         2       0.007099       SGD  0.253203"
            ],
            "text/html": [
              "\n",
              "  <div id=\"df-63179523-05e6-4fe2-ac7f-2f7acd9da9ee\">\n",
              "    <div class=\"colab-df-container\">\n",
              "      <div>\n",
              "<style scoped>\n",
              "    .dataframe tbody tr th:only-of-type {\n",
              "        vertical-align: middle;\n",
              "    }\n",
              "\n",
              "    .dataframe tbody tr th {\n",
              "        vertical-align: top;\n",
              "    }\n",
              "\n",
              "    .dataframe thead th {\n",
              "        text-align: right;\n",
              "    }\n",
              "</style>\n",
              "<table border=\"1\" class=\"dataframe\">\n",
              "  <thead>\n",
              "    <tr style=\"text-align: right;\">\n",
              "      <th></th>\n",
              "      <th>num_input_time_steps</th>\n",
              "      <th>learning_rate</th>\n",
              "      <th>optimizer</th>\n",
              "      <th>RMSE</th>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>strategy</th>\n",
              "      <th></th>\n",
              "      <th></th>\n",
              "      <th></th>\n",
              "      <th></th>\n",
              "    </tr>\n",
              "  </thead>\n",
              "  <tbody>\n",
              "    <tr>\n",
              "      <th>Not pruned</th>\n",
              "      <td>1</td>\n",
              "      <td>0.008574</td>\n",
              "      <td>SGD</td>\n",
              "      <td>0.250274</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>Pruned</th>\n",
              "      <td>2</td>\n",
              "      <td>0.007099</td>\n",
              "      <td>SGD</td>\n",
              "      <td>0.253203</td>\n",
              "    </tr>\n",
              "  </tbody>\n",
              "</table>\n",
              "</div>\n",
              "      <button class=\"colab-df-convert\" onclick=\"convertToInteractive('df-63179523-05e6-4fe2-ac7f-2f7acd9da9ee')\"\n",
              "              title=\"Convert this dataframe to an interactive table.\"\n",
              "              style=\"display:none;\">\n",
              "        \n",
              "  <svg xmlns=\"http://www.w3.org/2000/svg\" height=\"24px\"viewBox=\"0 0 24 24\"\n",
              "       width=\"24px\">\n",
              "    <path d=\"M0 0h24v24H0V0z\" fill=\"none\"/>\n",
              "    <path d=\"M18.56 5.44l.94 2.06.94-2.06 2.06-.94-2.06-.94-.94-2.06-.94 2.06-2.06.94zm-11 1L8.5 8.5l.94-2.06 2.06-.94-2.06-.94L8.5 2.5l-.94 2.06-2.06.94zm10 10l.94 2.06.94-2.06 2.06-.94-2.06-.94-.94-2.06-.94 2.06-2.06.94z\"/><path d=\"M17.41 7.96l-1.37-1.37c-.4-.4-.92-.59-1.43-.59-.52 0-1.04.2-1.43.59L10.3 9.45l-7.72 7.72c-.78.78-.78 2.05 0 2.83L4 21.41c.39.39.9.59 1.41.59.51 0 1.02-.2 1.41-.59l7.78-7.78 2.81-2.81c.8-.78.8-2.07 0-2.86zM5.41 20L4 18.59l7.72-7.72 1.47 1.35L5.41 20z\"/>\n",
              "  </svg>\n",
              "      </button>\n",
              "      \n",
              "  <style>\n",
              "    .colab-df-container {\n",
              "      display:flex;\n",
              "      flex-wrap:wrap;\n",
              "      gap: 12px;\n",
              "    }\n",
              "\n",
              "    .colab-df-convert {\n",
              "      background-color: #E8F0FE;\n",
              "      border: none;\n",
              "      border-radius: 50%;\n",
              "      cursor: pointer;\n",
              "      display: none;\n",
              "      fill: #1967D2;\n",
              "      height: 32px;\n",
              "      padding: 0 0 0 0;\n",
              "      width: 32px;\n",
              "    }\n",
              "\n",
              "    .colab-df-convert:hover {\n",
              "      background-color: #E2EBFA;\n",
              "      box-shadow: 0px 1px 2px rgba(60, 64, 67, 0.3), 0px 1px 3px 1px rgba(60, 64, 67, 0.15);\n",
              "      fill: #174EA6;\n",
              "    }\n",
              "\n",
              "    [theme=dark] .colab-df-convert {\n",
              "      background-color: #3B4455;\n",
              "      fill: #D2E3FC;\n",
              "    }\n",
              "\n",
              "    [theme=dark] .colab-df-convert:hover {\n",
              "      background-color: #434B5C;\n",
              "      box-shadow: 0px 1px 3px 1px rgba(0, 0, 0, 0.15);\n",
              "      filter: drop-shadow(0px 1px 2px rgba(0, 0, 0, 0.3));\n",
              "      fill: #FFFFFF;\n",
              "    }\n",
              "  </style>\n",
              "\n",
              "      <script>\n",
              "        const buttonEl =\n",
              "          document.querySelector('#df-63179523-05e6-4fe2-ac7f-2f7acd9da9ee button.colab-df-convert');\n",
              "        buttonEl.style.display =\n",
              "          google.colab.kernel.accessAllowed ? 'block' : 'none';\n",
              "\n",
              "        async function convertToInteractive(key) {\n",
              "          const element = document.querySelector('#df-63179523-05e6-4fe2-ac7f-2f7acd9da9ee');\n",
              "          const dataTable =\n",
              "            await google.colab.kernel.invokeFunction('convertToInteractive',\n",
              "                                                     [key], {});\n",
              "          if (!dataTable) return;\n",
              "\n",
              "          const docLinkHtml = 'Like what you see? Visit the ' +\n",
              "            '<a target=\"_blank\" href=https://colab.research.google.com/notebooks/data_table.ipynb>data table notebook</a>'\n",
              "            + ' to learn more about interactive tables.';\n",
              "          element.innerHTML = '';\n",
              "          dataTable['output_type'] = 'display_data';\n",
              "          await google.colab.output.renderOutput(dataTable, element);\n",
              "          const docLink = document.createElement('div');\n",
              "          docLink.innerHTML = docLinkHtml;\n",
              "          element.appendChild(docLink);\n",
              "        }\n",
              "      </script>\n",
              "    </div>\n",
              "  </div>\n",
              "  "
            ]
          },
          "metadata": {},
          "execution_count": 84
        }
      ],
      "source": [
        "df = pd.DataFrame(\n",
        "    {\n",
        "        \"strategy\": [\"Not pruned\", \"Pruned\"],\n",
        "        \"num_input_time_steps\": [\n",
        "            tpe_search_study.best_params[\"num_input_time_steps\"],\n",
        "            tpe_search_10perc_pruned_study.best_params[\"num_input_time_steps\"],\n",
        "        ],\n",
        "        \"learning_rate\": [\n",
        "            tpe_search_study.best_params[\"learning_rate\"],\n",
        "            tpe_search_10perc_pruned_study.best_params[\"learning_rate\"],\n",
        "        ],\n",
        "        \"optimizer\": [\n",
        "            tpe_search_study.best_params[\"optimizer\"],\n",
        "            tpe_search_10perc_pruned_study.best_params[\"optimizer\"],\n",
        "        ],\n",
        "        \"RMSE\": [\n",
        "            tpe_search_study.best_value,\n",
        "            tpe_search_10perc_pruned_study.best_value,\n",
        "        ],\n",
        "    }\n",
        ")\n",
        "\n",
        "df.set_index(\"strategy\")"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "lTN7K33sHPkU"
      },
      "source": [
        "Our pruning strategy resulted in a slightly different set of optimal hyperparameters but with a similar RMSE!\n",
        "\n",
        "⭐ **CHALLENGE:** Try increasing the percentile of the `PercentilePruner`. How does this change affect the number of trials pruned?"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "vcN0htFVzJ5p"
      },
      "source": [
        "#### **3.3.4 Compare Emissions from Search Strategies**\n",
        "\n",
        "Now let's compare emissions from both the pruned and the unpruned hyperparameter tuning studies."
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "3GmrbJLrvEYH",
        "outputId": "de4c8c27-b8a7-44bd-ae1e-10e7a2e7544c"
      },
      "outputs": [
        {
          "output_type": "stream",
          "name": "stdout",
          "text": [
            "Pruning our search reduced emissions by 46.8%.\n"
          ]
        }
      ],
      "source": [
        "print(\n",
        "    f\"Pruning our search reduced emissions by {abs(1 - tpe_search_percentile_pruned_emissions / tpe_search_emissions):.1%}.\"\n",
        ")"
      ]
    },
    {
      "cell_type": "code",
      "execution_count": null,
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 462
        },
        "id": "a7Erh5X2vEZn",
        "outputId": "353aa5ae-4d30-4ed3-f77c-381530faaa6d"
      },
      "outputs": [
        {
          "output_type": "display_data",
          "data": {
            "text/plain": [
              "<Figure size 500x500 with 2 Axes>"
            ],
            "image/png": "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\n"
          },
          "metadata": {}
        }
      ],
      "source": [
        "tuning_results = pd.DataFrame(\n",
        "    {\n",
        "        \"Category\": [\"TPE Search\", \"TPE Search Pruned \\n(10th percentile)\"],\n",
        "        \"RMSE\": [\n",
        "            tpe_search_study.best_value,\n",
        "            tpe_search_10perc_pruned_study.best_value,\n",
        "        ],\n",
        "        \"Emissions\": [\n",
        "            tpe_search_emissions * 1000,\n",
        "            tpe_search_percentile_pruned_emissions * 1000,\n",
        "        ],\n",
        "    }\n",
        ")\n",
        "\n",
        "plot_comparison(tuning_results)"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "C8HElXNUJzYx"
      },
      "source": [
        "The plot shows that our RMSE stayed essentially constant but we significantly reduced our emissions just by implementing a 10th percentile pruner. Not bad!"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "rDHjiW4duNue"
      },
      "source": [
        "## **3.4 Other Exercises (Optional)**\n",
        "\n",
        "Algorithm choice and hyperparameter tuning are just two strategies to reduce your emissions from training an ML model. If you'd like to continue to experiment, check out these optional challenge exercises!\n",
        "\n",
        "* **Adjust your underlying hardware:** Click on \"Runtime\" at the top of this notebook and select \"Change runtime type.\" Select \"None\" to switch to a runtime without any GPU's. Execute the entire notebook again and note how the emissions differ now that we're relying on CPU's to execute our ML training. Which type of models are most affected by this switch? If you have access to a Colab Pro account, try upgrading instead to a premium GPU class. How does access to faster GPU's affect the energy consumed by our code?\n",
        "* **Adjust your compute region:** Click on \"Runtime\" at the top of this notebook and select \"Disconnect and delete runtime.\" Execute the first two sections until you connect to a new compute region with a different energy intensity.\n",
        "* **Implement early stopping:** If you're more familiar with neural networks, experiment with applying early stopping to our CNN training process. An \"early stopping\" approach will terminate training if the model's validation loss decreases for several epochs in a row. How does reducing the number of epochs via early stopping affect the final model's performance and its emissions?\n",
        "\n",
        "\n"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "NYk0TXtZ0G4Y"
      },
      "source": [
        "# **References**\n",
        "[1] Kaack, L.H., Donti, P.L., Strubell, E. et al. \"Aligning artificial intelligence with climate change mitigation.\" Nat. Clim. Chang. 12, 518–527 (2022). https://doi.org/10.1038/s41558-022-01377-7\n",
        "\n",
        "[2] Oil in the Cloud: How Tech Companies are Helping Big Oil Profit from Climate Destruction (Greenpeace, 2019); https://www.greenpeace.org/usa/reports/oil-in-the-cloud/\n",
        "\n",
        "[3] AI-Based Smart Farming: The Rise of Machine Learning in Livestock Farming (Nexocode, 2022); https://nexocode.com/blog/posts/ai-based-smart-farming-machine-learning-in-livestock-farming/\n",
        "\n",
        "[4]  Dario Amodei and Daniel Hernandez. 2018. AI and Compute. https://openai.com/blog/ai-and-compute/\n",
        "\n",
        "[5] Patterson, David A. et al. “The Carbon Footprint of Machine Learning Training Will Plateau, Then Shrink.” Computer 55 (2022): 18-28.\n",
        "\n",
        "[6] Budennyy, Semen A. et al. “eco2AI: Carbon Emissions Tracking of Machine Learning Models as the First Step Towards Sustainable AI.” Doklady Mathematics 106 (2022): S118 - S128.\n",
        "\n",
        "[7] Wu, Carole-Jean et al. \"Sustainable AI: Environmental Implications, Challenges and Opportunities.\" Proceedings of the 5th MLSys Conference, Santa Clara, CA, USA, 2022.\n",
        "\n",
        "[8] Jiang, A. H. et al. Accelerating deep learning by focusing on the biggest losers. Preprint at https://arxiv.org/abs/1910.00762 (2019)\n",
        "\n",
        "[9] ChatGPT reaches 100 million users two months after launch (The Guardian, 2023); https://www.theguardian.com/technology/2023/feb/02/chatgpt-100-million-users-open-ai-fastest-growing-app\n",
        "\n",
        "[10] How much electricity does an American home use? (U.S. Energy Information Administration, 2022); https://www.eia.gov/tools/faqs/faq.php?id=97&t=3#\n",
        "\n",
        "[11] Lannelongue, L., Grealey, J., Inouye, M. \"Green Algorithms: Quantifying the carbon footprint of computation.\" Advanced Science 8(12), 2021. https://doi.org/10.1002/advs.202100707\n",
        "\n"
      ]
    }
  ],
  "metadata": {
    "accelerator": "GPU",
    "colab": {
      "collapsed_sections": [
        "xyb28HbsrcLJ"
      ],
      "provenance": [],
      "include_colab_link": true
    },
    "gpuClass": "standard",
    "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": "3.8.8"
    },
    "widgets": {
      "application/vnd.jupyter.widget-state+json": {
        "73f4f105a91a4fafbae59391ebd0fb65": {
          "model_module": "@jupyter-widgets/controls",
          "model_name": "HBoxModel",
          "model_module_version": "1.5.0",
          "state": {
            "_dom_classes": [],
            "_model_module": "@jupyter-widgets/controls",
            "_model_module_version": "1.5.0",
            "_model_name": "HBoxModel",
            "_view_count": null,
            "_view_module": "@jupyter-widgets/controls",
            "_view_module_version": "1.5.0",
            "_view_name": "HBoxView",
            "box_style": "",
            "children": [
              "IPY_MODEL_510551aea3084555b4c3982528cce9b8",
              "IPY_MODEL_b395e09ac1084567ac824ea50517cef9",
              "IPY_MODEL_7ed85909f48045dbbadbc804c9464c82"
            ],
            "layout": "IPY_MODEL_400faafaa2e148718bc98268be1ca6c2"
          }
        },
        "510551aea3084555b4c3982528cce9b8": {
          "model_module": "@jupyter-widgets/controls",
          "model_name": "HTMLModel",
          "model_module_version": "1.5.0",
          "state": {
            "_dom_classes": [],
            "_model_module": "@jupyter-widgets/controls",
            "_model_module_version": "1.5.0",
            "_model_name": "HTMLModel",
            "_view_count": null,
            "_view_module": "@jupyter-widgets/controls",
            "_view_module_version": "1.5.0",
            "_view_name": "HTMLView",
            "description": "",
            "description_tooltip": null,
            "layout": "IPY_MODEL_db109a8c2e00496695aba33d2d82ee1b",
            "placeholder": "​",
            "style": "IPY_MODEL_8388bad8267c459c861ec6fbeef7057f",
            "value": "Best trial: 6. Best value: 0.250274: 100%"
          }
        },
        "b395e09ac1084567ac824ea50517cef9": {
          "model_module": "@jupyter-widgets/controls",
          "model_name": "FloatProgressModel",
          "model_module_version": "1.5.0",
          "state": {
            "_dom_classes": [],
            "_model_module": "@jupyter-widgets/controls",
            "_model_module_version": "1.5.0",
            "_model_name": "FloatProgressModel",
            "_view_count": null,
            "_view_module": "@jupyter-widgets/controls",
            "_view_module_version": "1.5.0",
            "_view_name": "ProgressView",
            "bar_style": "success",
            "description": "",
            "description_tooltip": null,
            "layout": "IPY_MODEL_9a771cf17912472c870f33705b9539e0",
            "max": 12,
            "min": 0,
            "orientation": "horizontal",
            "style": "IPY_MODEL_bcbccc61d20b42e0910abddbb8ed8c59",
            "value": 12
          }
        },
        "7ed85909f48045dbbadbc804c9464c82": {
          "model_module": "@jupyter-widgets/controls",
          "model_name": "HTMLModel",
          "model_module_version": "1.5.0",
          "state": {
            "_dom_classes": [],
            "_model_module": "@jupyter-widgets/controls",
            "_model_module_version": "1.5.0",
            "_model_name": "HTMLModel",
            "_view_count": null,
            "_view_module": "@jupyter-widgets/controls",
            "_view_module_version": "1.5.0",
            "_view_name": "HTMLView",
            "description": "",
            "description_tooltip": null,
            "layout": "IPY_MODEL_1501e6948fa4445d9c8adc5b3f709e72",
            "placeholder": "​",
            "style": "IPY_MODEL_fbf9681dc0704297a5f832f610adb311",
            "value": " 12/12 [04:12&lt;00:00, 19.89s/it]"
          }
        },
        "400faafaa2e148718bc98268be1ca6c2": {
          "model_module": "@jupyter-widgets/base",
          "model_name": "LayoutModel",
          "model_module_version": "1.2.0",
          "state": {
            "_model_module": "@jupyter-widgets/base",
            "_model_module_version": "1.2.0",
            "_model_name": "LayoutModel",
            "_view_count": null,
            "_view_module": "@jupyter-widgets/base",
            "_view_module_version": "1.2.0",
            "_view_name": "LayoutView",
            "align_content": null,
            "align_items": null,
            "align_self": null,
            "border": null,
            "bottom": null,
            "display": null,
            "flex": null,
            "flex_flow": null,
            "grid_area": null,
            "grid_auto_columns": null,
            "grid_auto_flow": null,
            "grid_auto_rows": null,
            "grid_column": null,
            "grid_gap": null,
            "grid_row": null,
            "grid_template_areas": null,
            "grid_template_columns": null,
            "grid_template_rows": null,
            "height": null,
            "justify_content": null,
            "justify_items": null,
            "left": null,
            "margin": null,
            "max_height": null,
            "max_width": null,
            "min_height": null,
            "min_width": null,
            "object_fit": null,
            "object_position": null,
            "order": null,
            "overflow": null,
            "overflow_x": null,
            "overflow_y": null,
            "padding": null,
            "right": null,
            "top": null,
            "visibility": null,
            "width": null
          }
        },
        "db109a8c2e00496695aba33d2d82ee1b": {
          "model_module": "@jupyter-widgets/base",
          "model_name": "LayoutModel",
          "model_module_version": "1.2.0",
          "state": {
            "_model_module": "@jupyter-widgets/base",
            "_model_module_version": "1.2.0",
            "_model_name": "LayoutModel",
            "_view_count": null,
            "_view_module": "@jupyter-widgets/base",
            "_view_module_version": "1.2.0",
            "_view_name": "LayoutView",
            "align_content": null,
            "align_items": null,
            "align_self": null,
            "border": null,
            "bottom": null,
            "display": null,
            "flex": null,
            "flex_flow": null,
            "grid_area": null,
            "grid_auto_columns": null,
            "grid_auto_flow": null,
            "grid_auto_rows": null,
            "grid_column": null,
            "grid_gap": null,
            "grid_row": null,
            "grid_template_areas": null,
            "grid_template_columns": null,
            "grid_template_rows": null,
            "height": null,
            "justify_content": null,
            "justify_items": null,
            "left": null,
            "margin": null,
            "max_height": null,
            "max_width": null,
            "min_height": null,
            "min_width": null,
            "object_fit": null,
            "object_position": null,
            "order": null,
            "overflow": null,
            "overflow_x": null,
            "overflow_y": null,
            "padding": null,
            "right": null,
            "top": null,
            "visibility": null,
            "width": null
          }
        },
        "8388bad8267c459c861ec6fbeef7057f": {
          "model_module": "@jupyter-widgets/controls",
          "model_name": "DescriptionStyleModel",
          "model_module_version": "1.5.0",
          "state": {
            "_model_module": "@jupyter-widgets/controls",
            "_model_module_version": "1.5.0",
            "_model_name": "DescriptionStyleModel",
            "_view_count": null,
            "_view_module": "@jupyter-widgets/base",
            "_view_module_version": "1.2.0",
            "_view_name": "StyleView",
            "description_width": ""
          }
        },
        "9a771cf17912472c870f33705b9539e0": {
          "model_module": "@jupyter-widgets/base",
          "model_name": "LayoutModel",
          "model_module_version": "1.2.0",
          "state": {
            "_model_module": "@jupyter-widgets/base",
            "_model_module_version": "1.2.0",
            "_model_name": "LayoutModel",
            "_view_count": null,
            "_view_module": "@jupyter-widgets/base",
            "_view_module_version": "1.2.0",
            "_view_name": "LayoutView",
            "align_content": null,
            "align_items": null,
            "align_self": null,
            "border": null,
            "bottom": null,
            "display": null,
            "flex": null,
            "flex_flow": null,
            "grid_area": null,
            "grid_auto_columns": null,
            "grid_auto_flow": null,
            "grid_auto_rows": null,
            "grid_column": null,
            "grid_gap": null,
            "grid_row": null,
            "grid_template_areas": null,
            "grid_template_columns": null,
            "grid_template_rows": null,
            "height": null,
            "justify_content": null,
            "justify_items": null,
            "left": null,
            "margin": null,
            "max_height": null,
            "max_width": null,
            "min_height": null,
            "min_width": null,
            "object_fit": null,
            "object_position": null,
            "order": null,
            "overflow": null,
            "overflow_x": null,
            "overflow_y": null,
            "padding": null,
            "right": null,
            "top": null,
            "visibility": null,
            "width": null
          }
        },
        "bcbccc61d20b42e0910abddbb8ed8c59": {
          "model_module": "@jupyter-widgets/controls",
          "model_name": "ProgressStyleModel",
          "model_module_version": "1.5.0",
          "state": {
            "_model_module": "@jupyter-widgets/controls",
            "_model_module_version": "1.5.0",
            "_model_name": "ProgressStyleModel",
            "_view_count": null,
            "_view_module": "@jupyter-widgets/base",
            "_view_module_version": "1.2.0",
            "_view_name": "StyleView",
            "bar_color": null,
            "description_width": ""
          }
        },
        "1501e6948fa4445d9c8adc5b3f709e72": {
          "model_module": "@jupyter-widgets/base",
          "model_name": "LayoutModel",
          "model_module_version": "1.2.0",
          "state": {
            "_model_module": "@jupyter-widgets/base",
            "_model_module_version": "1.2.0",
            "_model_name": "LayoutModel",
            "_view_count": null,
            "_view_module": "@jupyter-widgets/base",
            "_view_module_version": "1.2.0",
            "_view_name": "LayoutView",
            "align_content": null,
            "align_items": null,
            "align_self": null,
            "border": null,
            "bottom": null,
            "display": null,
            "flex": null,
            "flex_flow": null,
            "grid_area": null,
            "grid_auto_columns": null,
            "grid_auto_flow": null,
            "grid_auto_rows": null,
            "grid_column": null,
            "grid_gap": null,
            "grid_row": null,
            "grid_template_areas": null,
            "grid_template_columns": null,
            "grid_template_rows": null,
            "height": null,
            "justify_content": null,
            "justify_items": null,
            "left": null,
            "margin": null,
            "max_height": null,
            "max_width": null,
            "min_height": null,
            "min_width": null,
            "object_fit": null,
            "object_position": null,
            "order": null,
            "overflow": null,
            "overflow_x": null,
            "overflow_y": null,
            "padding": null,
            "right": null,
            "top": null,
            "visibility": null,
            "width": null
          }
        },
        "fbf9681dc0704297a5f832f610adb311": {
          "model_module": "@jupyter-widgets/controls",
          "model_name": "DescriptionStyleModel",
          "model_module_version": "1.5.0",
          "state": {
            "_model_module": "@jupyter-widgets/controls",
            "_model_module_version": "1.5.0",
            "_model_name": "DescriptionStyleModel",
            "_view_count": null,
            "_view_module": "@jupyter-widgets/base",
            "_view_module_version": "1.2.0",
            "_view_name": "StyleView",
            "description_width": ""
          }
        },
        "bb530fd1cab94a45afdb174945b8c47e": {
          "model_module": "@jupyter-widgets/controls",
          "model_name": "HBoxModel",
          "model_module_version": "1.5.0",
          "state": {
            "_dom_classes": [],
            "_model_module": "@jupyter-widgets/controls",
            "_model_module_version": "1.5.0",
            "_model_name": "HBoxModel",
            "_view_count": null,
            "_view_module": "@jupyter-widgets/controls",
            "_view_module_version": "1.5.0",
            "_view_name": "HBoxView",
            "box_style": "",
            "children": [
              "IPY_MODEL_e8b772087f3a48c3be97c555dc1fca85",
              "IPY_MODEL_0e06f4d494b34e89ab7587f1291039e7",
              "IPY_MODEL_dd341850a65f4dbe95b0bba7726ddd14"
            ],
            "layout": "IPY_MODEL_170f40d037e34544b6ffa1cb4b75153b"
          }
        },
        "e8b772087f3a48c3be97c555dc1fca85": {
          "model_module": "@jupyter-widgets/controls",
          "model_name": "HTMLModel",
          "model_module_version": "1.5.0",
          "state": {
            "_dom_classes": [],
            "_model_module": "@jupyter-widgets/controls",
            "_model_module_version": "1.5.0",
            "_model_name": "HTMLModel",
            "_view_count": null,
            "_view_module": "@jupyter-widgets/controls",
            "_view_module_version": "1.5.0",
            "_view_name": "HTMLView",
            "description": "",
            "description_tooltip": null,
            "layout": "IPY_MODEL_2da303b2be4d410595a6672288b836de",
            "placeholder": "​",
            "style": "IPY_MODEL_34d068f48aa841c5bb7c9f52ef46ddaf",
            "value": "Best trial: 3. Best value: 0.253203: 100%"
          }
        },
        "0e06f4d494b34e89ab7587f1291039e7": {
          "model_module": "@jupyter-widgets/controls",
          "model_name": "FloatProgressModel",
          "model_module_version": "1.5.0",
          "state": {
            "_dom_classes": [],
            "_model_module": "@jupyter-widgets/controls",
            "_model_module_version": "1.5.0",
            "_model_name": "FloatProgressModel",
            "_view_count": null,
            "_view_module": "@jupyter-widgets/controls",
            "_view_module_version": "1.5.0",
            "_view_name": "ProgressView",
            "bar_style": "success",
            "description": "",
            "description_tooltip": null,
            "layout": "IPY_MODEL_e7069f4a57dd4530895fbd4cacd3582b",
            "max": 12,
            "min": 0,
            "orientation": "horizontal",
            "style": "IPY_MODEL_7181ba6369774af4a6456c166f38fdb8",
            "value": 12
          }
        },
        "dd341850a65f4dbe95b0bba7726ddd14": {
          "model_module": "@jupyter-widgets/controls",
          "model_name": "HTMLModel",
          "model_module_version": "1.5.0",
          "state": {
            "_dom_classes": [],
            "_model_module": "@jupyter-widgets/controls",
            "_model_module_version": "1.5.0",
            "_model_name": "HTMLModel",
            "_view_count": null,
            "_view_module": "@jupyter-widgets/controls",
            "_view_module_version": "1.5.0",
            "_view_name": "HTMLView",
            "description": "",
            "description_tooltip": null,
            "layout": "IPY_MODEL_e021a0e85c5b4fffa4e53d5a94747691",
            "placeholder": "​",
            "style": "IPY_MODEL_b0cf2b274bb74ac5ae3195c0e9c69aa4",
            "value": " 12/12 [02:19&lt;00:00,  4.40s/it]"
          }
        },
        "170f40d037e34544b6ffa1cb4b75153b": {
          "model_module": "@jupyter-widgets/base",
          "model_name": "LayoutModel",
          "model_module_version": "1.2.0",
          "state": {
            "_model_module": "@jupyter-widgets/base",
            "_model_module_version": "1.2.0",
            "_model_name": "LayoutModel",
            "_view_count": null,
            "_view_module": "@jupyter-widgets/base",
            "_view_module_version": "1.2.0",
            "_view_name": "LayoutView",
            "align_content": null,
            "align_items": null,
            "align_self": null,
            "border": null,
            "bottom": null,
            "display": null,
            "flex": null,
            "flex_flow": null,
            "grid_area": null,
            "grid_auto_columns": null,
            "grid_auto_flow": null,
            "grid_auto_rows": null,
            "grid_column": null,
            "grid_gap": null,
            "grid_row": null,
            "grid_template_areas": null,
            "grid_template_columns": null,
            "grid_template_rows": null,
            "height": null,
            "justify_content": null,
            "justify_items": null,
            "left": null,
            "margin": null,
            "max_height": null,
            "max_width": null,
            "min_height": null,
            "min_width": null,
            "object_fit": null,
            "object_position": null,
            "order": null,
            "overflow": null,
            "overflow_x": null,
            "overflow_y": null,
            "padding": null,
            "right": null,
            "top": null,
            "visibility": null,
            "width": null
          }
        },
        "2da303b2be4d410595a6672288b836de": {
          "model_module": "@jupyter-widgets/base",
          "model_name": "LayoutModel",
          "model_module_version": "1.2.0",
          "state": {
            "_model_module": "@jupyter-widgets/base",
            "_model_module_version": "1.2.0",
            "_model_name": "LayoutModel",
            "_view_count": null,
            "_view_module": "@jupyter-widgets/base",
            "_view_module_version": "1.2.0",
            "_view_name": "LayoutView",
            "align_content": null,
            "align_items": null,
            "align_self": null,
            "border": null,
            "bottom": null,
            "display": null,
            "flex": null,
            "flex_flow": null,
            "grid_area": null,
            "grid_auto_columns": null,
            "grid_auto_flow": null,
            "grid_auto_rows": null,
            "grid_column": null,
            "grid_gap": null,
            "grid_row": null,
            "grid_template_areas": null,
            "grid_template_columns": null,
            "grid_template_rows": null,
            "height": null,
            "justify_content": null,
            "justify_items": null,
            "left": null,
            "margin": null,
            "max_height": null,
            "max_width": null,
            "min_height": null,
            "min_width": null,
            "object_fit": null,
            "object_position": null,
            "order": null,
            "overflow": null,
            "overflow_x": null,
            "overflow_y": null,
            "padding": null,
            "right": null,
            "top": null,
            "visibility": null,
            "width": null
          }
        },
        "34d068f48aa841c5bb7c9f52ef46ddaf": {
          "model_module": "@jupyter-widgets/controls",
          "model_name": "DescriptionStyleModel",
          "model_module_version": "1.5.0",
          "state": {
            "_model_module": "@jupyter-widgets/controls",
            "_model_module_version": "1.5.0",
            "_model_name": "DescriptionStyleModel",
            "_view_count": null,
            "_view_module": "@jupyter-widgets/base",
            "_view_module_version": "1.2.0",
            "_view_name": "StyleView",
            "description_width": ""
          }
        },
        "e7069f4a57dd4530895fbd4cacd3582b": {
          "model_module": "@jupyter-widgets/base",
          "model_name": "LayoutModel",
          "model_module_version": "1.2.0",
          "state": {
            "_model_module": "@jupyter-widgets/base",
            "_model_module_version": "1.2.0",
            "_model_name": "LayoutModel",
            "_view_count": null,
            "_view_module": "@jupyter-widgets/base",
            "_view_module_version": "1.2.0",
            "_view_name": "LayoutView",
            "align_content": null,
            "align_items": null,
            "align_self": null,
            "border": null,
            "bottom": null,
            "display": null,
            "flex": null,
            "flex_flow": null,
            "grid_area": null,
            "grid_auto_columns": null,
            "grid_auto_flow": null,
            "grid_auto_rows": null,
            "grid_column": null,
            "grid_gap": null,
            "grid_row": null,
            "grid_template_areas": null,
            "grid_template_columns": null,
            "grid_template_rows": null,
            "height": null,
            "justify_content": null,
            "justify_items": null,
            "left": null,
            "margin": null,
            "max_height": null,
            "max_width": null,
            "min_height": null,
            "min_width": null,
            "object_fit": null,
            "object_position": null,
            "order": null,
            "overflow": null,
            "overflow_x": null,
            "overflow_y": null,
            "padding": null,
            "right": null,
            "top": null,
            "visibility": null,
            "width": null
          }
        },
        "7181ba6369774af4a6456c166f38fdb8": {
          "model_module": "@jupyter-widgets/controls",
          "model_name": "ProgressStyleModel",
          "model_module_version": "1.5.0",
          "state": {
            "_model_module": "@jupyter-widgets/controls",
            "_model_module_version": "1.5.0",
            "_model_name": "ProgressStyleModel",
            "_view_count": null,
            "_view_module": "@jupyter-widgets/base",
            "_view_module_version": "1.2.0",
            "_view_name": "StyleView",
            "bar_color": null,
            "description_width": ""
          }
        },
        "e021a0e85c5b4fffa4e53d5a94747691": {
          "model_module": "@jupyter-widgets/base",
          "model_name": "LayoutModel",
          "model_module_version": "1.2.0",
          "state": {
            "_model_module": "@jupyter-widgets/base",
            "_model_module_version": "1.2.0",
            "_model_name": "LayoutModel",
            "_view_count": null,
            "_view_module": "@jupyter-widgets/base",
            "_view_module_version": "1.2.0",
            "_view_name": "LayoutView",
            "align_content": null,
            "align_items": null,
            "align_self": null,
            "border": null,
            "bottom": null,
            "display": null,
            "flex": null,
            "flex_flow": null,
            "grid_area": null,
            "grid_auto_columns": null,
            "grid_auto_flow": null,
            "grid_auto_rows": null,
            "grid_column": null,
            "grid_gap": null,
            "grid_row": null,
            "grid_template_areas": null,
            "grid_template_columns": null,
            "grid_template_rows": null,
            "height": null,
            "justify_content": null,
            "justify_items": null,
            "left": null,
            "margin": null,
            "max_height": null,
            "max_width": null,
            "min_height": null,
            "min_width": null,
            "object_fit": null,
            "object_position": null,
            "order": null,
            "overflow": null,
            "overflow_x": null,
            "overflow_y": null,
            "padding": null,
            "right": null,
            "top": null,
            "visibility": null,
            "width": null
          }
        },
        "b0cf2b274bb74ac5ae3195c0e9c69aa4": {
          "model_module": "@jupyter-widgets/controls",
          "model_name": "DescriptionStyleModel",
          "model_module_version": "1.5.0",
          "state": {
            "_model_module": "@jupyter-widgets/controls",
            "_model_module_version": "1.5.0",
            "_model_name": "DescriptionStyleModel",
            "_view_count": null,
            "_view_module": "@jupyter-widgets/base",
            "_view_module_version": "1.2.0",
            "_view_name": "StyleView",
            "description_width": ""
          }
        }
      }
    }
  },
  "nbformat": 4,
  "nbformat_minor": 0
}