integrations/xgboost/showcase/Neptune-XGBoost.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {
    "tags": [
     "header"
    ]
   },
   "source": [
    "# XGBoost + Neptune integration"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "tags": [
     "comment"
    ]
   },
   "source": [
    "Here, we present XGBoost integration with Neptune that lets you automatically log metrics (train, eval), save trained model to Neptune and much more.\n",
    "\n",
    "Usage is easy: just pass `neptune_callback` to training function like any other xgboost callback.\n",
    "\n",
    "To try integration simply run this Notebook top to bottom. It works outside-the-box :)\n",
    "\n",
    "You can log multiple data types:\n",
    "* Log metrics (train and eval) after each boosting iteration.\n",
    "* Log model (Booster) to Neptune after last boosting iteration.\n",
    "* Log feature importance to Neptune as image after last boosting iteration."
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "tags": [
     "comment"
    ]
   },
   "source": [
    "## Visual overview\n",
    "![xgboost-integration-tour](https://neptune.ai/wp-content/uploads/xgboost-tour-1.gif)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "tags": [
     "header"
    ]
   },
   "source": [
    "## Before you start"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "tags": [
     "header"
    ]
   },
   "source": [
    "### Install dependencies"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "tags": [
     "installation"
    ]
   },
   "outputs": [],
   "source": [
    "! pip install neptune-client==0.4.132 neptune-contrib[monitoring]==0.25.0 xgboost==1.2.0 pandas==1.0.5 scikit-learn==0.23.2"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "tags": [
     "code"
    ]
   },
   "outputs": [],
   "source": [
    "import neptune\n",
    "import pandas as pd\n",
    "import xgboost as xgb\n",
    "from sklearn.datasets import load_boston\n",
    "from sklearn.model_selection import train_test_split\n",
    "\n",
    "from neptunecontrib.monitoring.xgboost import neptune_callback"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "tags": [
     "header"
    ]
   },
   "source": [
    "## Set project"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "tags": [
     "comment"
    ]
   },
   "source": [
    "For this demonstration, I use public user: `neptuner`, who has `ANONYMOUS` token .\n",
    "\n",
    "Thanks to this you can run this code as is and see results in Neptune :)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "tags": [
     "code"
    ]
   },
   "outputs": [],
   "source": [
    "neptune.init('shared/XGBoost-integration',\n",
    "             api_token='ANONYMOUS')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "tags": [
     "header"
    ]
   },
   "source": [
    "## Prepare data for XGBoost training"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "tags": [
     "code"
    ]
   },
   "outputs": [],
   "source": [
    "boston = load_boston()\n",
    "data = pd.DataFrame(boston.data)\n",
    "data.columns = boston.feature_names\n",
    "data['PRICE'] = boston.target\n",
    "X, y = data.iloc[:,:-1], data.iloc[:,-1]\n",
    "X_train, X_test, y_train, y_test = train_test_split(X, y, test_size=0.2, random_state=102030)\n",
    "\n",
    "dtrain = xgb.DMatrix(X_train, label=y_train)\n",
    "dtest = xgb.DMatrix(X_test, label=y_test)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "tags": [
     "header"
    ]
   },
   "source": [
    "## Prepare params"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "tags": [
     "code"
    ]
   },
   "outputs": [],
   "source": [
    "params = {'max_depth': 5,\n",
    "          'eta': 0.5,\n",
    "          'gamma': 0.1,\n",
    "          'subsample': 1,\n",
    "          'lambda': 1,\n",
    "          'alpha': 0.35,\n",
    "          'objective': 'reg:squarederror',\n",
    "          'eval_metric': ['mae', 'rmse']}\n",
    "watchlist = [(dtest, 'eval'), (dtrain, 'train')]\n",
    "num_round = 20"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "tags": [
     "header"
    ]
   },
   "source": [
    "## Train model using `xgb.train()`"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "tags": [
     "comment"
    ]
   },
   "source": [
    "Example experiment: [https://ui.neptune.ml/shared/XGBoost-integration/e/XGB-41](https://ui.neptune.ai/shared/XGBoost-integration/e/XGB-41?utm_source=colab&utm_medium=notebook&utm_campaign=integration-xgboost)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "tags": [
     "code"
    ]
   },
   "outputs": [],
   "source": [
    "neptune.create_experiment(name='xgb', tags=['train'], params=params)\n",
    "xgb.train(params, dtrain, num_round, watchlist,\n",
    "          callbacks=[neptune_callback()])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "tags": [
     "code"
    ]
   },
   "outputs": [],
   "source": [
    "neptune.stop()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "tags": [
     "header"
    ]
   },
   "source": [
    "## Train model using `xgb.cv()`"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "tags": [
     "comment"
    ]
   },
   "source": [
    "Example experiment: [https://ui.neptune.ml/shared/XGBoost-integration/e/XGB-42](https://ui.neptune.ai/shared/XGBoost-integration/e/XGB-42?utm_source=colab&utm_medium=notebook&utm_campaign=integration-xgboost)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "tags": [
     "code"
    ]
   },
   "outputs": [],
   "source": [
    "neptune.create_experiment(name='xgb', tags=['cv'], params=params)\n",
    "xgb.cv(params, dtrain, num_boost_round=num_round, nfold=7,\n",
    "       callbacks=[neptune_callback()])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "tags": [
     "code"
    ]
   },
   "outputs": [],
   "source": [
    "neptune.stop()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "tags": [
     "header"
    ]
   },
   "source": [
    "## Train model using `sklearn` API"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {
    "tags": [
     "comment"
    ]
   },
   "source": [
    "Example experiment: [https://ui.neptune.ml/shared/XGBoost-integration/e/XGB-43](https://ui.neptune.ai/shared/XGBoost-integration/e/XGB-43?utm_source=colab&utm_medium=notebook&utm_campaign=integration-xgboost)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "tags": [
     "code"
    ]
   },
   "outputs": [],
   "source": [
    "neptune.create_experiment(name='xgb', tags=['sklearn'], params=params)\n",
    "reg = xgb.XGBRegressor(**params)\n",
    "reg.fit(X_train, y_train,\n",
    "        eval_metric=['mae', 'rmse'],\n",
    "        eval_set=[(X_test, y_test)],\n",
    "        callbacks=[neptune_callback()])"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {
    "tags": [
     "code"
    ]
   },
   "outputs": [],
   "source": [
    "neptune.stop()"
   ]
  }
 ],
 "metadata": {
  "colab": {
   "name": "Neptune-XGBoost.ipynb",
   "provenance": []
  },
  "kernelspec": {
   "display_name": "Python 3",
   "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": "3.7.7"
  },
  "neptune": {
   "notebookId": "d9ac48b4-401d-4e45-a04c-c458ab10fa29"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 4
}