mlflow_exp_example.ipynb

{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [],
   "source": [
    "import sklearn\n",
    "import lightgbm as lgb\n",
    "import pandas as pd\n",
    "import numpy as np\n",
    "from sklearn.model_selection import train_test_split\n",
    "from sklearn.preprocessing import LabelEncoder\n",
    "from sklearn.datasets import load_wine\n",
    "import mlflow as mf\n",
    "from mlflow import log_metric, log_param, log_artifact\n",
    "import mlflow.lightgbm\n",
    "import matplotlib.pyplot as plt\n",
    "import seaborn as sns\n",
    "import logging\n",
    "logging.basicConfig(level = logging.INFO)\n",
    "logger = logging.getLogger()\n",
    "\n",
    "pd.options.display.max_rows = 50\n",
    "pd.options.display.max_columns = 50\n",
    "%matplotlib inline\n",
    "\n",
    "dataset = load_wine()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## ML Evaluation plot"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "metadata": {},
   "outputs": [],
   "source": [
    "def plotImp(model, X , title,num = 16):\n",
    "    feature_imp = pd.DataFrame({'Value':model.feature_importance(),'Feature':X.columns})\n",
    "    plt.figure(figsize=(12,7))\n",
    "    sns.set(font_scale = 1.1)\n",
    "    sns.barplot(color=\"lightblue\",x=\"Value\", y=\"Feature\", data=feature_imp.sort_values(by=\"Value\", \n",
    "                                                        ascending=False)[0:num])\n",
    "    plt.title(title)\n",
    "    plt.tight_layout()    \n",
    "       \n",
    "    fig1_fn = 'feature_importants.png'\n",
    "    plt.savefig(fig1_fn)\n",
    "    mlflow.log_artifact(fig1_fn)  \n",
    "    plt.show()\n",
    "  \n",
    "    \n",
    "def confusion_matrix_test(y_test, y_predict):\n",
    "    plt.figure()\n",
    "    cm = confusion_matrix(y_test, y_predict)\n",
    "    cmn = cm.astype('float') / cm.sum(axis=1)[:, np.newaxis]*100\n",
    "    sns.set(font_scale=1.2)\n",
    "    labels = ['0', '1','2']\n",
    "    plt.figure(figsize=(8,7))\n",
    "    #sns.heatmap(cm, xticklabels = labels, yticklabels = labels, annot = True, fmt='d', cmap=\"Blues\", vmin = 0.2);\n",
    "    sns.heatmap(cmn, xticklabels = labels, yticklabels = labels, annot = True, fmt='.2f', cmap=\"Blues\")\n",
    "    plt.title('Confusion Matrix')\n",
    "    plt.ylabel('True Class')\n",
    "    plt.xlabel('Predicted Class')\n",
    "\n",
    "    fig1_fn = 'cm.png'\n",
    "    plt.savefig(fig1_fn)\n",
    "    mlflow.log_artifact(fig1_fn) # logging to mlflow   \n",
    "    plt.show()\n",
    "    \n",
    "    \n",
    "    plt.show()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Data preprocessing and EDA"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [],
   "source": [
    "data = pd.DataFrame(data= np.c_[dataset['data'], dataset['target']],\n",
    "                     columns= dataset['feature_names'] + ['target'])\n",
    "#data.to_csv('data.csv')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<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>alcohol</th>\n",
       "      <th>malic_acid</th>\n",
       "      <th>ash</th>\n",
       "      <th>alcalinity_of_ash</th>\n",
       "      <th>magnesium</th>\n",
       "      <th>total_phenols</th>\n",
       "      <th>flavanoids</th>\n",
       "      <th>nonflavanoid_phenols</th>\n",
       "      <th>proanthocyanins</th>\n",
       "      <th>color_intensity</th>\n",
       "      <th>hue</th>\n",
       "      <th>od280/od315_of_diluted_wines</th>\n",
       "      <th>proline</th>\n",
       "      <th>target</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>14.23</td>\n",
       "      <td>1.71</td>\n",
       "      <td>2.43</td>\n",
       "      <td>15.6</td>\n",
       "      <td>127.0</td>\n",
       "      <td>2.80</td>\n",
       "      <td>3.06</td>\n",
       "      <td>0.28</td>\n",
       "      <td>2.29</td>\n",
       "      <td>5.64</td>\n",
       "      <td>1.04</td>\n",
       "      <td>3.92</td>\n",
       "      <td>1065.0</td>\n",
       "      <td>0.0</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>13.20</td>\n",
       "      <td>1.78</td>\n",
       "      <td>2.14</td>\n",
       "      <td>11.2</td>\n",
       "      <td>100.0</td>\n",
       "      <td>2.65</td>\n",
       "      <td>2.76</td>\n",
       "      <td>0.26</td>\n",
       "      <td>1.28</td>\n",
       "      <td>4.38</td>\n",
       "      <td>1.05</td>\n",
       "      <td>3.40</td>\n",
       "      <td>1050.0</td>\n",
       "      <td>0.0</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "   alcohol  malic_acid   ash  alcalinity_of_ash  magnesium  total_phenols  \\\n",
       "0    14.23        1.71  2.43               15.6      127.0           2.80   \n",
       "1    13.20        1.78  2.14               11.2      100.0           2.65   \n",
       "\n",
       "   flavanoids  nonflavanoid_phenols  proanthocyanins  color_intensity   hue  \\\n",
       "0        3.06                  0.28             2.29             5.64  1.04   \n",
       "1        2.76                  0.26             1.28             4.38  1.05   \n",
       "\n",
       "   od280/od315_of_diluted_wines  proline  target  \n",
       "0                          3.92   1065.0     0.0  \n",
       "1                          3.40   1050.0     0.0  "
      ]
     },
     "execution_count": 4,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "data.head(2)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<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>alcohol</th>\n",
       "      <th>malic_acid</th>\n",
       "      <th>ash</th>\n",
       "      <th>alcalinity_of_ash</th>\n",
       "      <th>magnesium</th>\n",
       "      <th>total_phenols</th>\n",
       "      <th>flavanoids</th>\n",
       "      <th>nonflavanoid_phenols</th>\n",
       "      <th>proanthocyanins</th>\n",
       "      <th>color_intensity</th>\n",
       "      <th>hue</th>\n",
       "      <th>od280/od315_of_diluted_wines</th>\n",
       "      <th>proline</th>\n",
       "      <th>target</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>count</th>\n",
       "      <td>178.000000</td>\n",
       "      <td>178.000000</td>\n",
       "      <td>178.000000</td>\n",
       "      <td>178.000000</td>\n",
       "      <td>178.000000</td>\n",
       "      <td>178.000000</td>\n",
       "      <td>178.000000</td>\n",
       "      <td>178.000000</td>\n",
       "      <td>178.000000</td>\n",
       "      <td>178.000000</td>\n",
       "      <td>178.000000</td>\n",
       "      <td>178.000000</td>\n",
       "      <td>178.000000</td>\n",
       "      <td>178.000000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>mean</th>\n",
       "      <td>13.000618</td>\n",
       "      <td>2.336348</td>\n",
       "      <td>2.366517</td>\n",
       "      <td>19.494944</td>\n",
       "      <td>99.741573</td>\n",
       "      <td>2.295112</td>\n",
       "      <td>2.029270</td>\n",
       "      <td>0.361854</td>\n",
       "      <td>1.590899</td>\n",
       "      <td>5.058090</td>\n",
       "      <td>0.957449</td>\n",
       "      <td>2.611685</td>\n",
       "      <td>746.893258</td>\n",
       "      <td>0.938202</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>std</th>\n",
       "      <td>0.811827</td>\n",
       "      <td>1.117146</td>\n",
       "      <td>0.274344</td>\n",
       "      <td>3.339564</td>\n",
       "      <td>14.282484</td>\n",
       "      <td>0.625851</td>\n",
       "      <td>0.998859</td>\n",
       "      <td>0.124453</td>\n",
       "      <td>0.572359</td>\n",
       "      <td>2.318286</td>\n",
       "      <td>0.228572</td>\n",
       "      <td>0.709990</td>\n",
       "      <td>314.907474</td>\n",
       "      <td>0.775035</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>min</th>\n",
       "      <td>11.030000</td>\n",
       "      <td>0.740000</td>\n",
       "      <td>1.360000</td>\n",
       "      <td>10.600000</td>\n",
       "      <td>70.000000</td>\n",
       "      <td>0.980000</td>\n",
       "      <td>0.340000</td>\n",
       "      <td>0.130000</td>\n",
       "      <td>0.410000</td>\n",
       "      <td>1.280000</td>\n",
       "      <td>0.480000</td>\n",
       "      <td>1.270000</td>\n",
       "      <td>278.000000</td>\n",
       "      <td>0.000000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>25%</th>\n",
       "      <td>12.362500</td>\n",
       "      <td>1.602500</td>\n",
       "      <td>2.210000</td>\n",
       "      <td>17.200000</td>\n",
       "      <td>88.000000</td>\n",
       "      <td>1.742500</td>\n",
       "      <td>1.205000</td>\n",
       "      <td>0.270000</td>\n",
       "      <td>1.250000</td>\n",
       "      <td>3.220000</td>\n",
       "      <td>0.782500</td>\n",
       "      <td>1.937500</td>\n",
       "      <td>500.500000</td>\n",
       "      <td>0.000000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>50%</th>\n",
       "      <td>13.050000</td>\n",
       "      <td>1.865000</td>\n",
       "      <td>2.360000</td>\n",
       "      <td>19.500000</td>\n",
       "      <td>98.000000</td>\n",
       "      <td>2.355000</td>\n",
       "      <td>2.135000</td>\n",
       "      <td>0.340000</td>\n",
       "      <td>1.555000</td>\n",
       "      <td>4.690000</td>\n",
       "      <td>0.965000</td>\n",
       "      <td>2.780000</td>\n",
       "      <td>673.500000</td>\n",
       "      <td>1.000000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>75%</th>\n",
       "      <td>13.677500</td>\n",
       "      <td>3.082500</td>\n",
       "      <td>2.557500</td>\n",
       "      <td>21.500000</td>\n",
       "      <td>107.000000</td>\n",
       "      <td>2.800000</td>\n",
       "      <td>2.875000</td>\n",
       "      <td>0.437500</td>\n",
       "      <td>1.950000</td>\n",
       "      <td>6.200000</td>\n",
       "      <td>1.120000</td>\n",
       "      <td>3.170000</td>\n",
       "      <td>985.000000</td>\n",
       "      <td>2.000000</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>max</th>\n",
       "      <td>14.830000</td>\n",
       "      <td>5.800000</td>\n",
       "      <td>3.230000</td>\n",
       "      <td>30.000000</td>\n",
       "      <td>162.000000</td>\n",
       "      <td>3.880000</td>\n",
       "      <td>5.080000</td>\n",
       "      <td>0.660000</td>\n",
       "      <td>3.580000</td>\n",
       "      <td>13.000000</td>\n",
       "      <td>1.710000</td>\n",
       "      <td>4.000000</td>\n",
       "      <td>1680.000000</td>\n",
       "      <td>2.000000</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "          alcohol  malic_acid         ash  alcalinity_of_ash   magnesium  \\\n",
       "count  178.000000  178.000000  178.000000         178.000000  178.000000   \n",
       "mean    13.000618    2.336348    2.366517          19.494944   99.741573   \n",
       "std      0.811827    1.117146    0.274344           3.339564   14.282484   \n",
       "min     11.030000    0.740000    1.360000          10.600000   70.000000   \n",
       "25%     12.362500    1.602500    2.210000          17.200000   88.000000   \n",
       "50%     13.050000    1.865000    2.360000          19.500000   98.000000   \n",
       "75%     13.677500    3.082500    2.557500          21.500000  107.000000   \n",
       "max     14.830000    5.800000    3.230000          30.000000  162.000000   \n",
       "\n",
       "       total_phenols  flavanoids  nonflavanoid_phenols  proanthocyanins  \\\n",
       "count     178.000000  178.000000            178.000000       178.000000   \n",
       "mean        2.295112    2.029270              0.361854         1.590899   \n",
       "std         0.625851    0.998859              0.124453         0.572359   \n",
       "min         0.980000    0.340000              0.130000         0.410000   \n",
       "25%         1.742500    1.205000              0.270000         1.250000   \n",
       "50%         2.355000    2.135000              0.340000         1.555000   \n",
       "75%         2.800000    2.875000              0.437500         1.950000   \n",
       "max         3.880000    5.080000              0.660000         3.580000   \n",
       "\n",
       "       color_intensity         hue  od280/od315_of_diluted_wines      proline  \\\n",
       "count       178.000000  178.000000                    178.000000   178.000000   \n",
       "mean          5.058090    0.957449                      2.611685   746.893258   \n",
       "std           2.318286    0.228572                      0.709990   314.907474   \n",
       "min           1.280000    0.480000                      1.270000   278.000000   \n",
       "25%           3.220000    0.782500                      1.937500   500.500000   \n",
       "50%           4.690000    0.965000                      2.780000   673.500000   \n",
       "75%           6.200000    1.120000                      3.170000   985.000000   \n",
       "max          13.000000    1.710000                      4.000000  1680.000000   \n",
       "\n",
       "           target  \n",
       "count  178.000000  \n",
       "mean     0.938202  \n",
       "std      0.775035  \n",
       "min      0.000000  \n",
       "25%      0.000000  \n",
       "50%      1.000000  \n",
       "75%      2.000000  \n",
       "max      2.000000  "
      ]
     },
     "execution_count": 5,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "data.describe()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "<matplotlib.axes._subplots.AxesSubplot at 0x7fbfd733f250>"
      ]
     },
     "execution_count": 6,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "data.target.value_counts().plot.bar()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "#data.pivot(columns='target').alcohol.plot(alpha=0.6,kind = 'hist',bins=20)\n",
    "ax = data.pivot(columns='target').alcohol.plot(alpha=0.6,kind = 'hist',bins=20)\n",
    "ax.figure.savefig('EDA_example.png')\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "<matplotlib.axes._subplots.AxesSubplot at 0x7fbfd991dd90>"
      ]
     },
     "execution_count": 8,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "data.pivot(columns='target').color_intensity.plot(alpha=0.6,kind = 'hist', bins=20)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "INFO:root:Split into train set shape (133, 13) and test set shape (45, 13)\n"
     ]
    }
   ],
   "source": [
    "X = data.drop(columns=['target'])\n",
    "y = data['target']\n",
    "\n",
    "X_train, X_test, y_train, y_test = train_test_split(X, y, test_size = 0.25, random_state = 100)\n",
    "logger.info(f'Split into train set shape {X_train.shape} and test set shape {X_test.shape}')\n"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Model Training with MLFlow tracking "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {},
   "outputs": [],
   "source": [
    "train_set = lgb.Dataset(X_train, label=y_train)\n",
    "test_set  = lgb.Dataset(X_test, label = y_test)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 29,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[1]\ttest's multi_logloss: 1.09567\n",
      "[2]\ttest's multi_logloss: 1.03913\n",
      "[3]\ttest's multi_logloss: 0.986517\n",
      "[4]\ttest's multi_logloss: 0.938852\n",
      "[5]\ttest's multi_logloss: 0.89676\n",
      "[6]\ttest's multi_logloss: 0.858452\n",
      "[7]\ttest's multi_logloss: 0.821099\n",
      "[8]\ttest's multi_logloss: 0.788379\n",
      "[9]\ttest's multi_logloss: 0.757815\n",
      "[10]\ttest's multi_logloss: 0.729938\n",
      "[11]\ttest's multi_logloss: 0.703839\n",
      "[12]\ttest's multi_logloss: 0.680795\n",
      "[13]\ttest's multi_logloss: 0.657841\n",
      "[14]\ttest's multi_logloss: 0.636529\n",
      "[15]\ttest's multi_logloss: 0.617145\n",
      "[16]\ttest's multi_logloss: 0.599812\n",
      "[17]\ttest's multi_logloss: 0.582328\n",
      "[18]\ttest's multi_logloss: 0.566805\n",
      "[19]\ttest's multi_logloss: 0.55166\n",
      "[20]\ttest's multi_logloss: 0.538621\n",
      "[21]\ttest's multi_logloss: 0.525541\n",
      "[22]\ttest's multi_logloss: 0.513225\n",
      "[23]\ttest's multi_logloss: 0.500235\n",
      "[24]\ttest's multi_logloss: 0.49071\n",
      "[25]\ttest's multi_logloss: 0.478019\n",
      "[26]\ttest's multi_logloss: 0.462843\n",
      "[27]\ttest's multi_logloss: 0.455223\n",
      "[28]\ttest's multi_logloss: 0.44146\n",
      "[29]\ttest's multi_logloss: 0.432254\n",
      "[30]\ttest's multi_logloss: 0.421047\n",
      "[31]\ttest's multi_logloss: 0.415256\n",
      "[32]\ttest's multi_logloss: 0.407785\n",
      "[33]\ttest's multi_logloss: 0.397045\n",
      "[34]\ttest's multi_logloss: 0.391692\n",
      "[35]\ttest's multi_logloss: 0.381042\n",
      "[36]\ttest's multi_logloss: 0.375156\n",
      "[37]\ttest's multi_logloss: 0.366049\n",
      "[38]\ttest's multi_logloss: 0.362195\n",
      "[39]\ttest's multi_logloss: 0.358208\n",
      "[40]\ttest's multi_logloss: 0.352642\n",
      "[41]\ttest's multi_logloss: 0.348858\n",
      "[42]\ttest's multi_logloss: 0.341586\n",
      "[43]\ttest's multi_logloss: 0.333657\n",
      "[44]\ttest's multi_logloss: 0.330651\n",
      "[45]\ttest's multi_logloss: 0.321556\n",
      "[46]\ttest's multi_logloss: 0.319284\n",
      "[47]\ttest's multi_logloss: 0.313898\n",
      "[48]\ttest's multi_logloss: 0.305753\n",
      "[49]\ttest's multi_logloss: 0.305932\n",
      "[50]\ttest's multi_logloss: 0.299722\n",
      "[51]\ttest's multi_logloss: 0.29794\n",
      "[52]\ttest's multi_logloss: 0.292378\n",
      "[53]\ttest's multi_logloss: 0.292039\n",
      "[54]\ttest's multi_logloss: 0.285667\n",
      "[55]\ttest's multi_logloss: 0.279823\n",
      "[56]\ttest's multi_logloss: 0.275348\n",
      "[57]\ttest's multi_logloss: 0.269572\n",
      "[58]\ttest's multi_logloss: 0.26982\n",
      "[59]\ttest's multi_logloss: 0.264582\n",
      "[60]\ttest's multi_logloss: 0.263925\n",
      "[61]\ttest's multi_logloss: 0.259535\n",
      "[62]\ttest's multi_logloss: 0.253162\n",
      "[63]\ttest's multi_logloss: 0.247948\n",
      "[64]\ttest's multi_logloss: 0.24385\n",
      "[65]\ttest's multi_logloss: 0.24272\n",
      "[66]\ttest's multi_logloss: 0.236983\n",
      "[67]\ttest's multi_logloss: 0.23376\n",
      "[68]\ttest's multi_logloss: 0.233192\n",
      "[69]\ttest's multi_logloss: 0.231918\n",
      "[70]\ttest's multi_logloss: 0.228831\n",
      "[71]\ttest's multi_logloss: 0.224762\n",
      "[72]\ttest's multi_logloss: 0.221375\n",
      "[73]\ttest's multi_logloss: 0.219359\n",
      "[74]\ttest's multi_logloss: 0.221096\n",
      "[75]\ttest's multi_logloss: 0.215758\n",
      "[76]\ttest's multi_logloss: 0.216601\n",
      "[77]\ttest's multi_logloss: 0.213923\n",
      "[78]\ttest's multi_logloss: 0.210556\n",
      "[79]\ttest's multi_logloss: 0.208907\n",
      "[80]\ttest's multi_logloss: 0.208798\n",
      "[81]\ttest's multi_logloss: 0.205317\n",
      "[82]\ttest's multi_logloss: 0.207773\n",
      "[83]\ttest's multi_logloss: 0.204117\n",
      "[84]\ttest's multi_logloss: 0.204468\n",
      "[85]\ttest's multi_logloss: 0.201084\n",
      "[86]\ttest's multi_logloss: 0.1992\n",
      "[87]\ttest's multi_logloss: 0.200818\n",
      "[88]\ttest's multi_logloss: 0.196651\n",
      "[89]\ttest's multi_logloss: 0.196785\n",
      "[90]\ttest's multi_logloss: 0.193743\n",
      "[91]\ttest's multi_logloss: 0.195193\n",
      "[92]\ttest's multi_logloss: 0.192296\n",
      "[93]\ttest's multi_logloss: 0.193571\n",
      "[94]\ttest's multi_logloss: 0.191923\n",
      "[95]\ttest's multi_logloss: 0.191903\n",
      "[96]\ttest's multi_logloss: 0.189219\n",
      "[97]\ttest's multi_logloss: 0.186119\n",
      "[98]\ttest's multi_logloss: 0.186432\n",
      "[99]\ttest's multi_logloss: 0.187195\n",
      "[100]\ttest's multi_logloss: 0.184508\n",
      "              precision    recall  f1-score   support\n",
      "\n",
      "         0.0       0.91      1.00      0.95        10\n",
      "         1.0       0.80      0.94      0.86        17\n",
      "         2.0       1.00      0.78      0.88        18\n",
      "\n",
      "    accuracy                           0.89        45\n",
      "   macro avg       0.90      0.91      0.90        45\n",
      "weighted avg       0.90      0.89      0.89        45\n",
      "\n"
     ]
    },
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "INFO:root:f1_score macro: 0.8974152724152723\n",
      "INFO:root:precision_score macro: 0.903030303030303\n",
      "INFO:root:recall_score macro: 0.9063180827886711\n",
      "INFO:root:f1_score weighted: 0.8883669383669384\n",
      "INFO:root:precision_score weighted: 0.9042424242424242\n",
      "INFO:root:recall_score weighted: 0.8888888888888888\n"
     ]
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 864x504 with 1 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "text/plain": [
       "<Figure size 864x504 with 0 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 576x504 with 2 Axes>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "    mf.set_experiment('MLflow expriement example')\n",
    "\n",
    "    \n",
    "    with mlflow.start_run(): \n",
    "        params={}\n",
    "        params['learning_rate']=0.06\n",
    "        params['boosting_type']='gbdt' #GradientBoostingDecisionTree\n",
    "        params['objective']='multiclass' #Multi-class target feature\n",
    "        params['metric']='multi_logloss' #metric for multi-class\n",
    "        params['max_depth']=2\n",
    "        params['num_class']=3\n",
    "        params['random_seed']=100\n",
    "        \n",
    "             \n",
    "        clf = lgb.train(params, train_set, num_boost_round=100, valid_sets=[test_set], valid_names=['test'])\n",
    "        preds = clf.predict(X_test)\n",
    "        predictions = []\n",
    "\n",
    "        for x in preds:\n",
    "            predictions.append(np.argmax(x))\n",
    "        from sklearn.metrics import accuracy_score, f1_score, precision_score, recall_score, classification_report, confusion_matrix\n",
    "\n",
    "        print(classification_report(y_test, predictions))\n",
    "        \n",
    "        # log artifact: classification_report\n",
    "        plt.rc('figure', figsize=(12, 7))\n",
    "        plt.text(0.01, 0.05, str(classification_report(y_test, predictions)), {'fontsize': 10}, fontproperties = 'monospace') \n",
    "        plt.axis('off')\n",
    "        plt.tight_layout()\n",
    "        summary_fn = 'clf_summary.png'\n",
    "        plt.savefig(summary_fn)\n",
    "        mlflow.log_artifact(summary_fn) # logging to mlflow\n",
    "        plt.close()\n",
    "        \n",
    "        \n",
    "        \n",
    "        f1_macro=f1_score(y_test, predictions, average=\"macro\")\n",
    "        precision_macro=precision_score(y_test, predictions, average=\"macro\")\n",
    "        recall_macro=recall_score(y_test, predictions, average=\"macro\") \n",
    "        f1_weighted=f1_score(y_test, predictions, average=\"weighted\")\n",
    "        precision_weighted=precision_score(y_test, predictions, average=\"weighted\")\n",
    "        recall_weighted=recall_score(y_test, predictions, average=\"weighted\")\n",
    "        \n",
    "        logger.info(f'f1_score macro: {f1_macro}')\n",
    "        logger.info(f'precision_score macro: {precision_macro}')\n",
    "        logger.info(f'recall_score macro: {recall_macro}')\n",
    "        logger.info(f'f1_score weighted: {f1_weighted}')\n",
    "        logger.info(f'precision_score weighted: {precision_weighted}')\n",
    "        logger.info(f'recall_score weighted: {recall_weighted}')\n",
    "        \n",
    "        #log the ML model, params, metrics and figures with MLflow \n",
    "        mf.log_param(\"learning_rate\", params['learning_rate'])\n",
    "        mf.log_param(\"max_depth\", params['max_depth'])\n",
    "        mf.log_param(\"boosting_type\",  params['boosting_type'])\n",
    "\n",
    "\n",
    "        mf.log_metric('F1_macro', f1_macro)\n",
    "        mf.log_metric('Precision_macro', precision_macro)\n",
    "        mf.log_metric('Recall_macro', recall_macro)\n",
    "        mf.log_metric('F1_score weighted', f1_weighted)\n",
    "        mf.log_metric('Precision_score weighted', precision_weighted)\n",
    "        mf.log_metric('Recall_score weighted', recall_weighted)\n",
    "        \n",
    "        mf.log_artifact('data.csv')\n",
    "        mf.lightgbm.log_model(clf, 'Model')\n",
    "\n",
    "        \n",
    "        mf.log_artifact('EDA_example.png')  \n",
    "        \n",
    "        plotImp(clf,X_train,'LightGBM Features importants plot')   \n",
    "        confusion_matrix_test(y_test, predictions)\n",
    "        mf.end_run()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  },
  {
   "cell_type": "code",
   "execution_count": 30,
   "metadata": {},
   "outputs": [],
   "source": [
    "mf.end_run()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 31,
   "metadata": {},
   "outputs": [],
   "source": [
    "from git import Repo\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 23,
   "metadata": {},
   "outputs": [],
   "source": [
    "def auto_commit(repo, file_path, home_dir=''):\n",
    "    if repo.git.diff(\"--cached\"):\n",
    "        print('Hiding Staged Changes')\n",
    "        repo.git.commit(\"-m temporary\")\n",
    "        repo.git.add(file_path)\n",
    "        repo.git.commit(\"-m AUTOML\")\n",
    "        stash_result = repo.git.stash('save')\n",
    "        stash_created = False\n",
    "        if stash_result != \"No local changes to save\":\n",
    "            stash_created = True\n",
    "\n",
    "        # switch commits\n",
    "        recent_commits = [commit.hexsha for commit in list(repo.iter_commits('master', max_count=3))]\n",
    "        repo.git.reset(\"--hard\",recent_commits[2])\n",
    "        repo.git.cherry_pick(recent_commits[0])\n",
    "        repo.git.cherry_pick(recent_commits[1])\n",
    "\n",
    "        # Revert the storage of the stage\n",
    "        repo.git.reset(\"HEAD^\")  # rollback the \"temporary\" commit\n",
    "        repo.git.add('.')  # add everything from that commit back to staging\n",
    "        if stash_created:\n",
    "            repo.git.stash(\"pop\")\n",
    "\n",
    "    # Call git directly because otherwise the filters are not applied\n",
    "    else:\n",
    "        repo.git.add(file_path)\n",
    "        try:\n",
    "            repo.git.commit(\"-m AUTOML commit\")\n",
    "        except:\n",
    "            # Nothing to commit because file was untouched\n",
    "            pass\n",
    "    \n",
    "    last_commit = list(repo.iter_commits('master', max_count=2))[:1][0]\n",
    "    return last_commit.hexsha"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 237,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "'/Users/gershonc/workspace/gc/mlflow_showcase'"
      ]
     },
     "execution_count": 237,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "pwd"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 25,
   "metadata": {},
   "outputs": [],
   "source": [
    "from git import Repo\n",
    "\n",
    "repo_path = '/Users/gershonc/workspace/gc/mlflow_showcase'\n",
    "repo = Repo(repo_path)\n",
    "\n",
    "#mlflow.start_run(source_version=auto_commit(repo, sys.argv[0]):"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 26,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "<git.repo.base.Repo '/Users/gershonc/workspace/gc/mlflow_showcase/.git'>"
      ]
     },
     "execution_count": 26,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "repo"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 27,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "<git.Commit \"5179b53a84aaaa6bc421daddbd5e2a7616f6f9f1\">"
      ]
     },
     "execution_count": 27,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "last_commit = list(repo.iter_commits(repo.active_branch))[:1][0]\n",
    "last_commit"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 28,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "<git.Commit \"7fa36d44d9506185d724ddd1bed2ae64cfc3bd6c\">"
      ]
     },
     "execution_count": 28,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "repo.git.add('*')\n",
    "repo.git.commit(\"-m Auto MLflow commit\")\n",
    "#repo.git.push()\n",
    "last_commit = list(repo.iter_commits(repo.active_branch))[:1][0]\n",
    "last_commit"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  }
 ],
 "metadata": {
  "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": "3.7.6"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 4
}