custom_task_tutorial.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "id": "b4552ef1-39b6-4b32-bb8d-9df481c9d5e9",
   "metadata": {
    "pycharm": {
     "name": "#%% md\n"
    }
   },
   "source": [
    "# Building a Custom Task\n",
    "\n",
    "This notebook walks through the general workflow of building a custom task using python classes. We'll also demonstrate how to deploy your custom task to DataRobot and integrate it into an existing blueprint!\n",
    "\n",
    "## Note\n",
    "The later sections of this tutorial require that you have access to Cloud DataRobot (app.datarobot.com or app.eu.datarobot.com)\n",
    "\n",
    "## Agenda\n",
    "In this tutorial, we'll learn:\n",
    "1. How to create a custom task using simple python classes\n",
    "2. How to test your python class\n",
    "3. How to use the drum cli tools to test out your custom task \n",
    "4. How to use the DataRobot API to deploy your custom task to the DataRobot cloud for use in projects\n",
    "5. How to insert a custom task on the DataRobot cloud into a blueprint\n",
    "\n",
    "## Setup and Requirements\n",
    "\n",
    "1. Ensure you have the DataRobot datarobot-user-models (DRUM) downloaded: https://github.com/datarobot/datarobot-user-models\n",
    "2. Navigate to the drum repo, create a new virtual environment, and pip install -r requirements.txt from that repo\n",
    "3. Follow the installation guide here for the blueprint workshop: https://blueprint-workshop.datarobot.com/setup.html \n",
    "\n",
    "Note: all filepaths are relative to the top level directory of the datarobot-user-models above. If you are in the correct directory you should see folders \n",
    "like task_templates, public_dropin_environments, tests, etc. \n",
    "\n",
    "\n",
    "## Configuring Models and Environments\n",
    "For more information on how to properly configure custom tasks and environments, read the README of our [DataRobot User Models repository](https://github.com/datarobot/datarobot-user-models)."
   ]
  },
  {
   "cell_type": "markdown",
   "id": "d838b9ec-24b8-4307-b5e7-042ae8f0ba5d",
   "metadata": {
    "pycharm": {
     "name": "#%% md\n"
    }
   },
   "source": [
    "# Building a custom task\n",
    "\n",
    "First, we need to import a few things"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 196,
   "id": "2e2f39ff-8382-494d-a085-e9d69515c84d",
   "metadata": {
    "pycharm": {
     "name": "#%%\n"
    }
   },
   "outputs": [],
   "source": [
    "from pathlib import Path\n",
    "import pandas as pd\n",
    "import tensorflow as tf\n",
    "import pickle\n",
    "\n",
    "import keras.models\n",
    "from tensorflow.keras.models import Sequential\n",
    "from tensorflow.keras.layers import Dense\n",
    "from tensorflow.keras.callbacks import EarlyStopping\n",
    "from tensorflow.keras.wrappers.scikit_learn import KerasRegressor\n",
    "\n",
    "from sklearn.pipeline import Pipeline\n"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "26180fa4-85f4-470c-ad85-ea83ac0fabbe",
   "metadata": {
    "pycharm": {
     "name": "#%% md\n"
    }
   },
   "source": [
    "Now let's build a neural network! First we'll lay out the code, then we'll walk through it"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 197,
   "id": "19999210-29c2-47a4-81a0-6625f173cd1a",
   "metadata": {
    "pycharm": {
     "name": "#%%\n"
    }
   },
   "outputs": [],
   "source": [
    "from datarobot_drum.custom_task_interfaces import RegressionEstimatorInterface\n",
    "\n",
    "class CustomTask(RegressionEstimatorInterface):\n",
    "    \n",
    "    def fit(self, X, y, row_weights=None, **kwargs):\n",
    "        tf.random.set_seed(1234)\n",
    "        input_dim, output_dim = len(X.columns), 1\n",
    "\n",
    "        model = Sequential(\n",
    "            [\n",
    "                Dense(\n",
    "                    input_dim, activation=\"relu\", input_dim=input_dim, kernel_initializer=\"normal\"\n",
    "                ),\n",
    "                Dense(input_dim // 2, activation=\"relu\", kernel_initializer=\"normal\"),\n",
    "                Dense(output_dim, kernel_initializer=\"normal\"),\n",
    "            ]\n",
    "        )\n",
    "        model.compile(loss=\"mse\", optimizer=\"adam\", metrics=[\"mae\", \"mse\"])\n",
    "\n",
    "        callback = EarlyStopping(monitor=\"loss\", patience=3)\n",
    "        model.fit(\n",
    "            X, y, epochs=20, batch_size=8, validation_split=0.33, verbose=1, callbacks=[callback]\n",
    "        )\n",
    "\n",
    "        # Attach the model to our object for future use\n",
    "        self.estimator = model\n",
    "        return self\n",
    "\n",
    "    \n",
    "    def save(self, artifact_directory):\n",
    "\n",
    "        # If your estimator is not pickle-able, you can serialize it using its native method,\n",
    "        # i.e. in this case for keras we use model.save, and then set the estimator to none\n",
    "        keras.models.save_model(self.estimator, Path(artifact_directory) / \"model.h5\")\n",
    "\n",
    "        # Helper method to handle serializing, via pickle, the CustomTask class\n",
    "        self.save_task(artifact_directory, exclude=[\"estimator\"])\n",
    "\n",
    "        return self\n",
    "\n",
    "\n",
    "    @classmethod\n",
    "    def load(cls, artifact_directory):\n",
    "        \n",
    "        # Helper method to load the serialized CustomTask class\n",
    "        custom_task = cls.load_task(artifact_directory)\n",
    "        custom_task.estimator = keras.models.load_model(Path(artifact_directory) / \"model.h5\")\n",
    "\n",
    "        return custom_task\n",
    "    \n",
    "\n",
    "    def predict(self, X, **kwargs):\n",
    "        # Note how the regression estimator only outputs one column, so no explicit column names are needed\n",
    "        return pd.DataFrame(data=self.estimator.predict(X))\n",
    "\n"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "7aa9b5af-854b-4a8f-b174-6c58b7fa005e",
   "metadata": {
    "pycharm": {
     "name": "#%% md\n"
    }
   },
   "source": [
    "There's a lot above, but the key idea is that we have 4 hooks: fit, save, load, and predict. DataRobot will use these hooks automatically to run our custom task. As you can probably guess, these hooks run in a specific order: first we train (fit) a model, then we serialize it (see the section [below]), then we load (i.e. deserialize) it again, and finally we make predictions. \n",
    "\n",
    "One thing to note is that the above CustomTask is simply a python class, which means we can also add helper methods or have functions / classes in a helper file that we import. The more complex your CustomTask, the more it probably makes sense to import a separate helper file to keep things simple. See [here] for an example of directly using helper methods in the class and [here] for using a separate helper file. "
   ]
  },
  {
   "cell_type": "markdown",
   "id": "177d7dfa-e1d9-4ac7-bbd8-0f31e99db09b",
   "metadata": {
    "pycharm": {
     "name": "#%% md\n"
    }
   },
   "source": [
    "## Training our model with Fit"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "0016bfbc-dc73-4f2e-a936-d6db6e4ccef7",
   "metadata": {
    "pycharm": {
     "name": "#%% md\n"
    }
   },
   "source": [
    "Now let's actually use the class above. Since this is an ordinary python class, all we need to do is build an object and we can test it out to ensure our methods work! First, let's grab a dataset and then separate out the target column"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 198,
   "id": "e8723510-6c27-4ece-8bfc-0e28921fef1f",
   "metadata": {
    "pycharm": {
     "name": "#%%\n"
    }
   },
   "outputs": [],
   "source": [
    "df = pd.read_csv(\"tests/testdata/juniors_3_year_stats_regression.csv\")\n",
    "\n",
    "y = df['Grade 2014']\n",
    "X = df.drop(labels=['Grade 2014'], axis=1)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "14e4c8c8-110d-464d-834b-c6560d230625",
   "metadata": {
    "pycharm": {
     "name": "#%% md\n"
    }
   },
   "source": [
    "Now let's train our model!"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 199,
   "id": "1b4a1f57-c912-4ced-9bbe-fa7ab1cbc2c0",
   "metadata": {
    "pycharm": {
     "name": "#%%\n"
    }
   },
   "outputs": [],
   "source": [
    "task = CustomTask()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 200,
   "id": "82989943-c38f-403f-bc08-27b8370d59b7",
   "metadata": {
    "pycharm": {
     "name": "#%%\n"
    }
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 1/20\n",
      "124/124 [==============================] - 1s 2ms/step - loss: 243.6316 - mae: 11.2293 - mse: 243.6316 - val_loss: 127.8057 - val_mae: 7.8969 - val_mse: 127.8057\n",
      "Epoch 2/20\n",
      "124/124 [==============================] - 0s 2ms/step - loss: 119.6218 - mae: 7.3784 - mse: 119.6218 - val_loss: 126.4541 - val_mae: 7.1894 - val_mse: 126.4541\n",
      "Epoch 3/20\n",
      "124/124 [==============================] - 0s 2ms/step - loss: 119.6898 - mae: 7.4064 - mse: 119.6898 - val_loss: 131.0404 - val_mae: 6.7115 - val_mse: 131.0404\n",
      "Epoch 4/20\n",
      "124/124 [==============================] - 0s 1ms/step - loss: 118.2549 - mae: 7.2968 - mse: 118.2549 - val_loss: 126.3128 - val_mae: 8.0251 - val_mse: 126.3128\n",
      "Epoch 5/20\n",
      "124/124 [==============================] - 0s 1ms/step - loss: 118.4696 - mae: 7.3713 - mse: 118.4696 - val_loss: 124.3396 - val_mae: 7.1751 - val_mse: 124.3396\n",
      "Epoch 6/20\n",
      "124/124 [==============================] - 0s 1ms/step - loss: 119.1182 - mae: 7.2953 - mse: 119.1182 - val_loss: 129.7029 - val_mae: 6.6129 - val_mse: 129.7029\n",
      "Epoch 7/20\n",
      "124/124 [==============================] - 0s 1ms/step - loss: 119.7277 - mae: 7.4162 - mse: 119.7277 - val_loss: 123.8151 - val_mae: 7.6164 - val_mse: 123.8151\n"
     ]
    }
   ],
   "source": [
    "task = task.fit(X,y)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "f14e29fc-b5a3-4448-a9c5-f521f8c8244d",
   "metadata": {
    "pycharm": {
     "name": "#%% md\n"
    }
   },
   "source": [
    "## Saving and Loading our Custom Task"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "9a889669-2be2-403d-985b-f62c4a61a9ca",
   "metadata": {
    "pycharm": {
     "name": "#%% md\n"
    }
   },
   "source": [
    "You may be wondering why we need to save our model only to immediately load it again to make predictions. The reason is that model training and prediction, also known as inference or scoring, are distinct tasks that may have very different resource requirements. As an extreme example, one of DataRobot's proprietary models is a genetic algorithm known as Eureqa. Training this model can take some time, as it iterates through mathematical transformations. However the output of this model is a simple mathematical equation, which can run almost instantly on very modest computational resources. So during training we want to allocate a high amount of compute, but while the model is making predictions, e.g. it is deployed and waiting to receive new data, we want a far lower amount of compute allocated.  \n",
    "\n",
    "The way we achieve this is by using Docker containers, which we can think of as extremely lightweight virtual machines. This allows our training container to have significantly different computational resources allocated than our prediction container. But since the training and prediction steps are in separate containers, we need a way to move trained models and other useful artifacts, e.g. class labels, between them. The solution is to write out the artifacts to disk, i.e. serialize them. So our save method at the end of training will write out the model to a shared file storage location and then our load method at the beginning of making predictions will read the artifacts into memory, i.e. deserialize them. \n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 201,
   "id": "ea43b8ec-80f9-4b99-8eaf-f778aba66225",
   "metadata": {
    "pycharm": {
     "name": "#%%\n"
    }
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "<__main__.CustomTask at 0x18730a810>"
      ]
     },
     "execution_count": 201,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "task.save(\".\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 202,
   "id": "d5419aaa-ab0e-4bd9-bd54-918af84a35d6",
   "metadata": {
    "pycharm": {
     "name": "#%%\n"
    }
   },
   "outputs": [],
   "source": [
    "task = task.load(\".\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 203,
   "id": "0ffeebdf-403e-4efb-a40c-4fc92af07c34",
   "metadata": {
    "pycharm": {
     "name": "#%%\n"
    }
   },
   "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>0</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>27.046207</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>25.635019</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>26.916660</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>25.292889</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>28.178623</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>...</th>\n",
       "      <td>...</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1472</th>\n",
       "      <td>24.894941</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1473</th>\n",
       "      <td>26.109985</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1474</th>\n",
       "      <td>31.427782</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1475</th>\n",
       "      <td>23.208033</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1476</th>\n",
       "      <td>31.082373</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "<p>1477 rows × 1 columns</p>\n",
       "</div>"
      ],
      "text/plain": [
       "              0\n",
       "0     27.046207\n",
       "1     25.635019\n",
       "2     26.916660\n",
       "3     25.292889\n",
       "4     28.178623\n",
       "...         ...\n",
       "1472  24.894941\n",
       "1473  26.109985\n",
       "1474  31.427782\n",
       "1475  23.208033\n",
       "1476  31.082373\n",
       "\n",
       "[1477 rows x 1 columns]"
      ]
     },
     "execution_count": 203,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "task.predict(X)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "d12611cf-00f2-420c-b237-6ce09111cf45",
   "metadata": {
    "pycharm": {
     "name": "#%% md\n"
    }
   },
   "source": [
    "Let's take a look more deeply at the save method:"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 204,
   "id": "69f5d9ff-a732-4b4b-982b-3b7c98b543b0",
   "metadata": {
    "pycharm": {
     "name": "#%%\n"
    }
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "\u001B[0;31mSignature:\u001B[0m \u001B[0mtask\u001B[0m\u001B[0;34m.\u001B[0m\u001B[0msave\u001B[0m\u001B[0;34m(\u001B[0m\u001B[0martifact_directory\u001B[0m\u001B[0;34m)\u001B[0m\u001B[0;34m\u001B[0m\u001B[0;34m\u001B[0m\u001B[0m\n",
       "\u001B[0;31mDocstring:\u001B[0m\n",
       "Serializes the object and stores it in `artifact_directory`\n",
       "\n",
       "Parameters\n",
       "----------\n",
       "artifact_directory: str\n",
       "    Path to the directory to save the serialized artifact(s) to.\n",
       "\n",
       "Returns\n",
       "-------\n",
       "self\n",
       "\u001B[0;31mSource:\u001B[0m   \n",
       "    \u001B[0;32mdef\u001B[0m \u001B[0msave\u001B[0m\u001B[0;34m(\u001B[0m\u001B[0mself\u001B[0m\u001B[0;34m,\u001B[0m \u001B[0martifact_directory\u001B[0m\u001B[0;34m)\u001B[0m\u001B[0;34m:\u001B[0m\u001B[0;34m\u001B[0m\n",
       "\u001B[0;34m\u001B[0m\u001B[0;34m\u001B[0m\n",
       "\u001B[0;34m\u001B[0m        \u001B[0;31m# If your estimator is not pickle-able, you can serialize it using its native method,\u001B[0m\u001B[0;34m\u001B[0m\n",
       "\u001B[0;34m\u001B[0m        \u001B[0;31m# i.e. in this case for keras we use model.save, and then set the estimator to none\u001B[0m\u001B[0;34m\u001B[0m\n",
       "\u001B[0;34m\u001B[0m        \u001B[0mkeras\u001B[0m\u001B[0;34m.\u001B[0m\u001B[0mmodels\u001B[0m\u001B[0;34m.\u001B[0m\u001B[0msave_model\u001B[0m\u001B[0;34m(\u001B[0m\u001B[0mself\u001B[0m\u001B[0;34m.\u001B[0m\u001B[0mestimator\u001B[0m\u001B[0;34m,\u001B[0m \u001B[0mPath\u001B[0m\u001B[0;34m(\u001B[0m\u001B[0martifact_directory\u001B[0m\u001B[0;34m)\u001B[0m \u001B[0;34m/\u001B[0m \u001B[0;34m\"model.h5\"\u001B[0m\u001B[0;34m)\u001B[0m\u001B[0;34m\u001B[0m\n",
       "\u001B[0;34m\u001B[0m\u001B[0;34m\u001B[0m\n",
       "\u001B[0;34m\u001B[0m        \u001B[0;31m# Helper method to handle serializing, via pickle, the CustomTask class\u001B[0m\u001B[0;34m\u001B[0m\n",
       "\u001B[0;34m\u001B[0m        \u001B[0mself\u001B[0m\u001B[0;34m.\u001B[0m\u001B[0msave_task\u001B[0m\u001B[0;34m(\u001B[0m\u001B[0martifact_directory\u001B[0m\u001B[0;34m,\u001B[0m \u001B[0mexclude\u001B[0m\u001B[0;34m=\u001B[0m\u001B[0;34m[\u001B[0m\u001B[0;34m\"estimator\"\u001B[0m\u001B[0;34m]\u001B[0m\u001B[0;34m)\u001B[0m\u001B[0;34m\u001B[0m\n",
       "\u001B[0;34m\u001B[0m\u001B[0;34m\u001B[0m\n",
       "\u001B[0;34m\u001B[0m        \u001B[0;32mreturn\u001B[0m \u001B[0mself\u001B[0m\u001B[0;34m\u001B[0m\u001B[0;34m\u001B[0m\u001B[0m\n",
       "\u001B[0;31mFile:\u001B[0m      /var/folders/zd/m7w2dbld2v9dlsdrtdrb1ms40000gq/T/ipykernel_58586/2228518297.py\n",
       "\u001B[0;31mType:\u001B[0m      method\n"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "??task.save"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "14df8a48-7dae-40ee-b007-16fcaa28bd38",
   "metadata": {
    "pycharm": {
     "name": "#%% md\n"
    }
   },
   "source": [
    "As we can see, we use two distinct functions to save our model. First, we use the keras function save_model to save our self.estimator, i.e. the trained model from the fit function. Then we use a built in helper function, save_task. Let's look at save task quickly:\n",
    "\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 24,
   "id": "50e28c39-8f44-4ffb-a86c-6d74fe9008a9",
   "metadata": {
    "pycharm": {
     "name": "#%%\n"
    }
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "\u001B[0;31mSignature:\u001B[0m \u001B[0mtask\u001B[0m\u001B[0;34m.\u001B[0m\u001B[0msave_task\u001B[0m\u001B[0;34m(\u001B[0m\u001B[0martifact_directory\u001B[0m\u001B[0;34m,\u001B[0m \u001B[0mexclude\u001B[0m\u001B[0;34m=\u001B[0m\u001B[0;32mNone\u001B[0m\u001B[0;34m)\u001B[0m\u001B[0;34m\u001B[0m\u001B[0;34m\u001B[0m\u001B[0m\n",
       "\u001B[0;31mSource:\u001B[0m   \n",
       "    \u001B[0;32mdef\u001B[0m \u001B[0msave_task\u001B[0m\u001B[0;34m(\u001B[0m\u001B[0mself\u001B[0m\u001B[0;34m,\u001B[0m \u001B[0martifact_directory\u001B[0m\u001B[0;34m,\u001B[0m \u001B[0mexclude\u001B[0m\u001B[0;34m=\u001B[0m\u001B[0;32mNone\u001B[0m\u001B[0;34m)\u001B[0m\u001B[0;34m:\u001B[0m\u001B[0;34m\u001B[0m\n",
       "\u001B[0;34m\u001B[0m        \u001B[0;34m\"\"\"\u001B[0m\n",
       "\u001B[0;34m        Helper function that abstracts away pickling the CustomTask object. It also can\u001B[0m\n",
       "\u001B[0;34m        automatically set previously serialized variables to None, e.g. when using keras you likely\u001B[0m\n",
       "\u001B[0;34m        want to serialize self.estimator using model.save() or keras.models.save_model() and then\u001B[0m\n",
       "\u001B[0;34m        pass in exclude='estimator'\u001B[0m\n",
       "\u001B[0;34m\u001B[0m\n",
       "\u001B[0;34m        Parameters\u001B[0m\n",
       "\u001B[0;34m        ----------\u001B[0m\n",
       "\u001B[0;34m        artifact_directory: str\u001B[0m\n",
       "\u001B[0;34m            Path to the directory to save the serialized artifact(s) to.\u001B[0m\n",
       "\u001B[0;34m        exclude: List[str]\u001B[0m\n",
       "\u001B[0;34m            Variables on the CustomTask object we want to exclude from serialization by setting to None\u001B[0m\n",
       "\u001B[0;34m\u001B[0m\n",
       "\u001B[0;34m        Returns\u001B[0m\n",
       "\u001B[0;34m        -------\u001B[0m\n",
       "\u001B[0;34m        None\u001B[0m\n",
       "\u001B[0;34m        \"\"\"\u001B[0m\u001B[0;34m\u001B[0m\n",
       "\u001B[0;34m\u001B[0m        \u001B[0;31m# If any custom task variables are excluded in the pickle, temporarily store them here, set them to None, then\u001B[0m\u001B[0;34m\u001B[0m\n",
       "\u001B[0;34m\u001B[0m        \u001B[0;31m# restore them back onto the class after serialization\u001B[0m\u001B[0;34m\u001B[0m\n",
       "\u001B[0;34m\u001B[0m        \u001B[0mvariables_to_restore\u001B[0m \u001B[0;34m=\u001B[0m \u001B[0;34m{\u001B[0m\u001B[0;34m}\u001B[0m\u001B[0;34m\u001B[0m\n",
       "\u001B[0;34m\u001B[0m\u001B[0;34m\u001B[0m\n",
       "\u001B[0;34m\u001B[0m        \u001B[0;32mif\u001B[0m \u001B[0mexclude\u001B[0m\u001B[0;34m:\u001B[0m\u001B[0;34m\u001B[0m\n",
       "\u001B[0;34m\u001B[0m            \u001B[0;32mfor\u001B[0m \u001B[0mcustom_task_variable\u001B[0m \u001B[0;32min\u001B[0m \u001B[0mexclude\u001B[0m\u001B[0;34m:\u001B[0m\u001B[0;34m\u001B[0m\n",
       "\u001B[0;34m\u001B[0m                \u001B[0;32mtry\u001B[0m\u001B[0;34m:\u001B[0m\u001B[0;34m\u001B[0m\n",
       "\u001B[0;34m\u001B[0m                    \u001B[0;31m# Ensure the variable actually exists in the custom task\u001B[0m\u001B[0;34m\u001B[0m\n",
       "\u001B[0;34m\u001B[0m                    \u001B[0mvariables_to_restore\u001B[0m\u001B[0;34m[\u001B[0m\u001B[0mcustom_task_variable\u001B[0m\u001B[0;34m]\u001B[0m \u001B[0;34m=\u001B[0m \u001B[0mgetattr\u001B[0m\u001B[0;34m(\u001B[0m\u001B[0mself\u001B[0m\u001B[0;34m,\u001B[0m \u001B[0mcustom_task_variable\u001B[0m\u001B[0;34m)\u001B[0m\u001B[0;34m\u001B[0m\n",
       "\u001B[0;34m\u001B[0m                \u001B[0;32mexcept\u001B[0m \u001B[0mAttributeError\u001B[0m\u001B[0;34m:\u001B[0m\u001B[0;34m\u001B[0m\n",
       "\u001B[0;34m\u001B[0m                    \u001B[0;32mraise\u001B[0m \u001B[0mDrumCommonException\u001B[0m\u001B[0;34m(\u001B[0m\u001B[0;34m\u001B[0m\n",
       "\u001B[0;34m\u001B[0m                        \u001B[0;34mf\"The object named {custom_task_variable} passed in exclude= was not found\"\u001B[0m\u001B[0;34m\u001B[0m\n",
       "\u001B[0;34m\u001B[0m                    \u001B[0;34m)\u001B[0m\u001B[0;34m\u001B[0m\n",
       "\u001B[0;34m\u001B[0m\u001B[0;34m\u001B[0m\n",
       "\u001B[0;34m\u001B[0m                \u001B[0;31m# Set it to None so it does not get serialized\u001B[0m\u001B[0;34m\u001B[0m\n",
       "\u001B[0;34m\u001B[0m                \u001B[0msetattr\u001B[0m\u001B[0;34m(\u001B[0m\u001B[0mself\u001B[0m\u001B[0;34m,\u001B[0m \u001B[0mcustom_task_variable\u001B[0m\u001B[0;34m,\u001B[0m \u001B[0;32mNone\u001B[0m\u001B[0;34m)\u001B[0m\u001B[0;34m\u001B[0m\n",
       "\u001B[0;34m\u001B[0m        \u001B[0;32mwith\u001B[0m \u001B[0mopen\u001B[0m\u001B[0;34m(\u001B[0m\u001B[0;34m\u001B[0m\n",
       "\u001B[0;34m\u001B[0m            \u001B[0mos\u001B[0m\u001B[0;34m.\u001B[0m\u001B[0mpath\u001B[0m\u001B[0;34m.\u001B[0m\u001B[0mjoin\u001B[0m\u001B[0;34m(\u001B[0m\u001B[0martifact_directory\u001B[0m\u001B[0;34m,\u001B[0m \u001B[0mSerializable\u001B[0m\u001B[0;34m.\u001B[0m\u001B[0mdefault_artifact_filename\u001B[0m\u001B[0;34m)\u001B[0m\u001B[0;34m,\u001B[0m \u001B[0;34m\"wb\"\u001B[0m\u001B[0;34m\u001B[0m\n",
       "\u001B[0;34m\u001B[0m        \u001B[0;34m)\u001B[0m \u001B[0;32mas\u001B[0m \u001B[0mfp\u001B[0m\u001B[0;34m:\u001B[0m\u001B[0;34m\u001B[0m\n",
       "\u001B[0;34m\u001B[0m            \u001B[0mpickle\u001B[0m\u001B[0;34m.\u001B[0m\u001B[0mdump\u001B[0m\u001B[0;34m(\u001B[0m\u001B[0mself\u001B[0m\u001B[0;34m,\u001B[0m \u001B[0mfp\u001B[0m\u001B[0;34m)\u001B[0m\u001B[0;34m\u001B[0m\n",
       "\u001B[0;34m\u001B[0m\u001B[0;34m\u001B[0m\n",
       "\u001B[0;34m\u001B[0m        \u001B[0;32mfor\u001B[0m \u001B[0mcustom_task_variable\u001B[0m\u001B[0;34m,\u001B[0m \u001B[0mvalue\u001B[0m \u001B[0;32min\u001B[0m \u001B[0mvariables_to_restore\u001B[0m\u001B[0;34m.\u001B[0m\u001B[0mitems\u001B[0m\u001B[0;34m(\u001B[0m\u001B[0;34m)\u001B[0m\u001B[0;34m:\u001B[0m\u001B[0;34m\u001B[0m\n",
       "\u001B[0;34m\u001B[0m            \u001B[0msetattr\u001B[0m\u001B[0;34m(\u001B[0m\u001B[0mself\u001B[0m\u001B[0;34m,\u001B[0m \u001B[0mcustom_task_variable\u001B[0m\u001B[0;34m,\u001B[0m \u001B[0mvalue\u001B[0m\u001B[0;34m)\u001B[0m\u001B[0;34m\u001B[0m\u001B[0;34m\u001B[0m\u001B[0m\n",
       "\u001B[0;31mFile:\u001B[0m      ~/.virtualenvs/drum_1.6.6/lib/python3.7/site-packages/datarobot_drum/custom_task_interfaces/custom_task_interface.py\n",
       "\u001B[0;31mType:\u001B[0m      method\n"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "??task.save_task"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "7d003172-6a15-4a36-8f89-f39da5200894",
   "metadata": {
    "pycharm": {
     "name": "#%% md\n"
    }
   },
   "source": [
    "Don't worry about understanding every line above. The key point is that we set everything passed in the exclude parameter to None, then we use a the standard python library pickle to serialize the CustomTask object. The reason we do this is flexibility. There are a wide array of python ML frameworks, e.g. keras / tensorflow, pytorch, xgboost, etc. Many of these frameworks, particularly those around neural networks, have their own serialization functions that handle all the complexities around storing weights, archiectures, etc. \n",
    "\n",
    "So the recommended pattern is to save your estimator using your framework's serialization function, e.g. keras.models.save_model above, and then use the helper function save_task we provide to save the rest of your CustomTask object. \n",
    "\n",
    "If we look at the load method, we see that we simply reverse the order. First we use the helper function load_task to load our CustomTask object using pickle, then we load our estimator into self.estimator using the keras function load_model:"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 25,
   "id": "4b8a6eb6-8bb5-4cbc-b0c0-659b3069c107",
   "metadata": {
    "pycharm": {
     "name": "#%%\n"
    }
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "\u001B[0;31mSignature:\u001B[0m \u001B[0mtask\u001B[0m\u001B[0;34m.\u001B[0m\u001B[0mload\u001B[0m\u001B[0;34m(\u001B[0m\u001B[0martifact_directory\u001B[0m\u001B[0;34m)\u001B[0m\u001B[0;34m\u001B[0m\u001B[0;34m\u001B[0m\u001B[0m\n",
       "\u001B[0;31mDocstring:\u001B[0m\n",
       "Deserializes the object stored within `artifact_directory`.\n",
       "\n",
       "Parameters\n",
       "----------\n",
       "artifact_directory: str\n",
       "    Path to the directory to save the serialized artifact(s) to.\n",
       "\n",
       "Returns\n",
       "-------\n",
       "cls\n",
       "    The deserialized object\n",
       "\u001B[0;31mSource:\u001B[0m   \n",
       "    \u001B[0;34m@\u001B[0m\u001B[0mclassmethod\u001B[0m\u001B[0;34m\u001B[0m\n",
       "\u001B[0;34m\u001B[0m    \u001B[0;32mdef\u001B[0m \u001B[0mload\u001B[0m\u001B[0;34m(\u001B[0m\u001B[0mcls\u001B[0m\u001B[0;34m,\u001B[0m \u001B[0martifact_directory\u001B[0m\u001B[0;34m)\u001B[0m\u001B[0;34m:\u001B[0m\u001B[0;34m\u001B[0m\n",
       "\u001B[0;34m\u001B[0m        \u001B[0;34m\u001B[0m\n",
       "\u001B[0;34m\u001B[0m        \u001B[0;31m# Helper method to load the serialized CustomTask class\u001B[0m\u001B[0;34m\u001B[0m\n",
       "\u001B[0;34m\u001B[0m        \u001B[0mcustom_task\u001B[0m \u001B[0;34m=\u001B[0m \u001B[0mcls\u001B[0m\u001B[0;34m.\u001B[0m\u001B[0mload_task\u001B[0m\u001B[0;34m(\u001B[0m\u001B[0martifact_directory\u001B[0m\u001B[0;34m)\u001B[0m\u001B[0;34m\u001B[0m\n",
       "\u001B[0;34m\u001B[0m        \u001B[0mcustom_task\u001B[0m\u001B[0;34m.\u001B[0m\u001B[0mestimator\u001B[0m \u001B[0;34m=\u001B[0m \u001B[0mkeras\u001B[0m\u001B[0;34m.\u001B[0m\u001B[0mmodels\u001B[0m\u001B[0;34m.\u001B[0m\u001B[0mload_model\u001B[0m\u001B[0;34m(\u001B[0m\u001B[0mPath\u001B[0m\u001B[0;34m(\u001B[0m\u001B[0martifact_directory\u001B[0m\u001B[0;34m)\u001B[0m \u001B[0;34m/\u001B[0m \u001B[0;34m\"model.h5\"\u001B[0m\u001B[0;34m)\u001B[0m\u001B[0;34m\u001B[0m\n",
       "\u001B[0;34m\u001B[0m\u001B[0;34m\u001B[0m\n",
       "\u001B[0;34m\u001B[0m        \u001B[0;32mreturn\u001B[0m \u001B[0mcustom_task\u001B[0m\u001B[0;34m\u001B[0m\u001B[0;34m\u001B[0m\u001B[0m\n",
       "\u001B[0;31mFile:\u001B[0m      /var/folders/zd/m7w2dbld2v9dlsdrtdrb1ms40000gq/T/ipykernel_58586/2228518297.py\n",
       "\u001B[0;31mType:\u001B[0m      method\n"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "??task.load"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "7bb8e275-f42a-42a4-b814-b70f1247223b",
   "metadata": {
    "pycharm": {
     "name": "#%% md\n"
    }
   },
   "source": [
    "It's important to understand that some python ML frameworks, notably Sklearn, use pickle for their serialization as well. This means we don't need to write our own save / load functions in our CustomTask, as the default functions will simply pickle everyting including the model. The below example is from the template [here] and shows how this looks for a simple sklearn model. You can see that the whole CustomTask is around 10 lines of code. Pretty neat, huh?"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 27,
   "id": "961306ce-2a76-4f6a-b229-eb468507726d",
   "metadata": {
    "pycharm": {
     "name": "#%%\n"
    }
   },
   "outputs": [],
   "source": [
    "class CustomTask(RegressionEstimatorInterface):\n",
    "    def fit(self, X, y, row_weights=None, **kwargs):\n",
    "        \n",
    "        self.estimator = Ridge()\n",
    "        self.estimator.fit(X, y)\n",
    "\n",
    "        return self\n",
    "\n",
    "    def predict(self, X, **kwargs):\n",
    "\n",
    "        return pd.DataFrame(data=self.estimator.predict(X))\n",
    "\n"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "ff725e50-b847-471e-9d47-34f7dd1cdc72",
   "metadata": {
    "pycharm": {
     "name": "#%% md\n"
    }
   },
   "source": [
    "## Making Predictions with the Correct Output"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "98816191-ba64-433d-a11d-a3d5752bc405",
   "metadata": {
    "pycharm": {
     "name": "#%% md\n"
    }
   },
   "source": [
    "A CustomTask currently has to output a pandas dataframe where the rows are the samples and the columns the predictions. \n",
    "\n",
    "As you may have noticed, all of examples so far have been regressors, i.e. outputting a single, numeric prediction. So our rows are just the number of samples and we have a single column (which doesn't need a specific name). We can see that our CustomTasks above inherit from the RegressionEstimatorInterface, which enforces this behavior.\n",
    "\n",
    "We can use the exact same behavior when we are creating an anomaly detection CustomTask, because the output is again a single numeric column representing how anomalous each sample is. There is a corresponding AnomalyEstimatorInterface.\n"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "3e6c045f-98f3-4a0b-8d5a-c5566b426c52",
   "metadata": {
    "pycharm": {
     "name": "#%% md\n"
    }
   },
   "source": [
    "Things are a little trickier though when we need to create a binary or multiclass estimator. That's because we'll need to align the columns to the classes. Keep in mind that our CustomTask will run inside a DataRobot blueprint, which will be given a list of classes in the target. Let's take a look at an example binary estimator:"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 30,
   "id": "f98e0b28-99ed-4f50-9cec-9f174c5b9300",
   "metadata": {
    "pycharm": {
     "name": "#%%\n"
    }
   },
   "outputs": [],
   "source": [
    "from datarobot_drum.custom_task_interfaces import BinaryEstimatorInterface\n",
    "\n",
    "class CustomTask(BinaryEstimatorInterface):\n",
    "    def fit(self, X, y, row_weights=None, **kwargs):\n",
    "        \n",
    "        self.estimator = DecisionTreeClassifier()\n",
    "        self.estimator.fit(X, y)\n",
    "\n",
    "        return self\n",
    "\n",
    "    def predict_proba(self, X, **kwargs):\n",
    "\n",
    "        # Note that binary estimators require two columns in the output, the positive and negative class labels\n",
    "        # So we need to pass in the class names derived from the estimator as column names OR\n",
    "        # we can use the class labels from DataRobot stored in\n",
    "        # kwargs['positive_class_label'] and kwargs['negative_class_label']\n",
    "        return pd.DataFrame(data=self.estimator.predict_proba(X), columns=self.estimator.classes_)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "72145ce3-a5a3-4a79-a560-a6bd2dd06d66",
   "metadata": {
    "pycharm": {
     "name": "#%% md\n"
    }
   },
   "source": [
    "The first thing to notice is that because we're outputting probabilities, we define a predict_proba instead of a predict function. The second thing to notice is that we have to provide the column names of our dataframe, and they have to align to the classes of our dataset. If you look at the fit function, you'll notice we directly pass in the target column y. This will have our target labels and these will be passed to our model as we train it, i.e. self.estimator.fit()\n"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "87859af7-9def-4eac-b5e4-c85746cd330f",
   "metadata": {
    "pycharm": {
     "name": "#%% md\n"
    }
   },
   "source": [
    "For binary classification, DataRobot requires there to be 2 columns: the positive class prediction and the negative class prediction \n",
    "(which is the inverse). Obviously, these two numbers should sum up to 1.0. \n",
    "\n",
    "For frameworks that output 2 classes, like sklearn, we can simply use the classes stored by the sklearn model itself, i.e. self.estimator.classes_\n",
    "\n",
    "Some frameworks, such as pytorch, instead only output one column (typically the positive class probability). In those cases we have to derive the negative class column, as seen below:"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 32,
   "id": "87d006f7-71d1-46ef-ba53-55c5fdea8824",
   "metadata": {
    "pycharm": {
     "name": "#%%\n"
    }
   },
   "outputs": [],
   "source": [
    "    def predict_proba(self, X, **kwargs):\n",
    "        \"\"\"Since pytorch only outputs a single probability, i.e. the probability of the positive class,\n",
    "         we use the class labels passed in kwargs to label the columns\"\"\"\n",
    "        data_tensor = torch.from_numpy(X.values).type(torch.FloatTensor)\n",
    "        predictions = self.estimator(data_tensor).cpu().data.numpy()\n",
    "        \n",
    "        predictions = pd.DataFrame(predictions, columns=[kwargs[\"positive_class_label\"]])\n",
    "\n",
    "        # The negative class probability is just the inverse of what the model predicts above\n",
    "        predictions[kwargs[\"negative_class_label\"]] = (\n",
    "            1 - predictions[kwargs[\"positive_class_label\"]]\n",
    "        )\n",
    "        return predictions\n"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "2528c5b1-5bbd-4c02-878b-11d9c9f1cee2",
   "metadata": {
    "pycharm": {
     "name": "#%% md\n"
    }
   },
   "source": [
    "Multiclass is slightly more challenging. Here we'll need to output the probability of each class as a separate column. If our framework stores. the classes, like many sklearn models, we can use the same exact same approach as above with binary classification. If the framework doesn't store the classes, e.g. pytorch, then we'll need to store the classes during the fit step on the self object so it can be used as the columns (Note: the save & load methods are excluded below to focus in on the unique aspects of multiclass):"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 35,
   "id": "a9a46774-4c5d-46d6-995e-2cf423ebc690",
   "metadata": {
    "pycharm": {
     "name": "#%%\n"
    }
   },
   "outputs": [],
   "source": [
    "from datarobot_drum.custom_task_interfaces import MulticlassEstimatorInterface\n",
    "\n",
    "\n",
    "class CustomTask(MulticlassEstimatorInterface):\n",
    "    def fit(self, X, y, row_weights=None, **kwargs):\n",
    "        \"\"\"Note how we encode the class labels and store them on self to be used in the predict hook\"\"\"\n",
    "        self.lb = LabelEncoder().fit(y)\n",
    "        y = self.lb.transform(y)\n",
    "\n",
    "        # For reproducible results\n",
    "        torch.manual_seed(0)\n",
    "\n",
    "        self.estimator, optimizer, criterion = build_classifier(X, len(self.lb.classes_))\n",
    "        train_classifier(X, y, self.estimator, optimizer, criterion)\n",
    "        \n",
    "\n",
    "    def predict_proba(self, X, **kwargs):\n",
    "        \"\"\"Note how the column names come from the encoded class labels in the fit hook above\"\"\"\n",
    "        data_tensor = torch.from_numpy(X.values).type(torch.FloatTensor)\n",
    "        predictions = self.estimator(data_tensor).cpu().data.numpy()\n",
    "\n",
    "        # Note that multiclass estimators require one column per class in the output\n",
    "        # So we need to pass in the class names derived from the estimator as column names.\n",
    "        return pd.DataFrame(data=predictions, columns=self.lb.classes_)\n"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "ff386820-c309-4d99-9243-2e51272817fa",
   "metadata": {
    "pycharm": {
     "name": "#%% md\n"
    }
   },
   "source": [
    "## Transformers vs. Estimators"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "fcc84954-0660-46e4-9263-8ca4b4aca0b5",
   "metadata": {
    "pycharm": {
     "name": "#%% md\n"
    }
   },
   "source": [
    "So far we've focused on Estimators, which output a prediction. We can also create transforms, which manipulate the data and pass it along to (eventually) an estimator. Note: this final estimator can either be a built in DataRobot task or a CustomTask.\n",
    "\n",
    "The key difference between estimators and transforms is that instead of a predict function we have a transform function. The transform function also returns a dataframe, but instead of predictions it returns the transformed data. Note that while you can certainly create or remove columns, the transform function must output the same number of rows, e.g. you can't create new synthetic data.\n",
    "\n",
    "It's important to keep in mind that the fit and transform functions also run in separate containers. So any artifacts need to be saved, either by using the default pickling functions that run automatically, or by defining your own save and load functions. \n",
    "\n",
    "Below is a simple example where you fill in missing values using median imputation. Note that because we are simply storing the imputation function as part of self, we can use the built in save and load functions so we don't have to define our own:"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 37,
   "id": "45a3492b-d926-439b-8653-e56622cd28a3",
   "metadata": {
    "pycharm": {
     "name": "#%%\n"
    }
   },
   "outputs": [],
   "source": [
    "from datarobot_drum.custom_task_interfaces import TransformerInterface\n",
    "\n",
    "\n",
    "class CustomTask(TransformerInterface):\n",
    "    def fit(self, X, y, **kwargs):\n",
    "\n",
    "        # Any information derived from the training data (i.e. median values for each column) should be stored to self.\n",
    "        # Then, in the transform hook below, we use this information to transform any data that passes through this task\n",
    "        self.medians = X.median(axis=0, numeric_only=True, skipna=True).to_dict()\n",
    "        return self\n",
    "\n",
    "    def transform(self, X, **kwargs):\n",
    "        return X.fillna(self.medians)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "c540cc1a-0c9f-42d2-afa0-10e672f4ee29",
   "metadata": {
    "pycharm": {
     "name": "#%% md\n"
    }
   },
   "source": [
    "One important thing to note is that we provide the target column as part of the fit function above so you can create target encoders or other related transforms. Since your CustomTask transform will run in a DataRobot blueprint, DataRobot will automatically handle setting up the defined partitioning strategy, e.g. 5 fold cross validation, and only run the fit function on the training folds. But it's always possible to introduce target leakage / overfitting with any encoding strategy, so please take care!\n"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "45714758-be06-4e9a-9bd2-1fce3c82a55e",
   "metadata": {
    "pycharm": {
     "name": "#%% md\n"
    }
   },
   "source": [
    "## Understanding the CustomTask Interface (Optional)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "d1a1495a-db5e-4a14-90cf-8a09d26cf2f6",
   "metadata": {
    "pycharm": {
     "name": "#%% md\n"
    }
   },
   "source": [
    "In the above sections, we mentioned that DataRobot automatically serialize your CustomTask as a pickle file unless you provided your own save / load functions. This is because we defined a series of interfaces that inherit from each other and have default implementations. Let's walk through them quickly.\n",
    "\n",
    "The top level interface is the target type interfaces, e.g. TransformerInterface, MulticlassEstimatorInterface, etc. These provide the relevant predict functions. Each of these inherits from a CustomTaskInterface, which provides the default fit method. \n",
    "\n",
    "Finally, the top level function is. the Serializable interface. You can see that by default, the save and load functions simply call the save_task and load_task helper functions. This is to provide a level of abstraction. When you define your own CustomTask save or load function, you can use the save_task and load_task functions to avoid having to write your own code to pickle the CustomTask object. "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 47,
   "id": "238fedae-7ea1-41f8-be82-2d4f76663c3f",
   "metadata": {
    "pycharm": {
     "name": "#%%\n"
    }
   },
   "outputs": [],
   "source": [
    "class Serializable(object):\n",
    "    default_artifact_filename = \"drum_artifact.pkl\"\n",
    "\n",
    "    def save(self, artifact_directory):\n",
    "\n",
    "        self.save_task(artifact_directory)\n",
    "\n",
    "        # For use in easy chaining, e.g. CustomTask().fit().save().load()\n",
    "        return self\n",
    "\n",
    "    def save_task(self, artifact_directory, exclude=None):\n",
    "        \"\"\"\n",
    "        Helper function that abstracts away pickling the CustomTask object. It also can\n",
    "        automatically set previously serialized variables to None, e.g. when using keras you likely\n",
    "        want to serialize self.estimator using model.save() or keras.models.save_model() and then\n",
    "        pass in exclude='estimator'\n",
    "\n",
    "        Parameters\n",
    "        ----------\n",
    "        artifact_directory: str\n",
    "            Path to the directory to save the serialized artifact(s) to.\n",
    "        exclude: List[str]\n",
    "            Variables on the CustomTask object we want to exclude from serialization by setting to None\n",
    "\n",
    "        Returns\n",
    "        -------\n",
    "        None\n",
    "        \"\"\"\n",
    "        # If any custom task variables are excluded in the pickle, temporarily store them here, set them to None, then\n",
    "        # restore them back onto the class after serialization\n",
    "        variables_to_restore = {}\n",
    "\n",
    "        if exclude:\n",
    "            for custom_task_variable in exclude:\n",
    "                try:\n",
    "                    # Ensure the variable actually exists in the custom task\n",
    "                    variables_to_restore[custom_task_variable] = getattr(self, custom_task_variable)\n",
    "                except AttributeError:\n",
    "                    raise DrumCommonException(\n",
    "                        f\"The object named {custom_task_variable} passed in exclude= was not found\"\n",
    "                    )\n",
    "\n",
    "                # Set it to None so it does not get serialized\n",
    "                setattr(self, custom_task_variable, None)\n",
    "        with open(\n",
    "            os.path.join(artifact_directory, Serializable.default_artifact_filename), \"wb\"\n",
    "        ) as fp:\n",
    "            pickle.dump(self, fp)\n",
    "\n",
    "        for custom_task_variable, value in variables_to_restore.items():\n",
    "            setattr(self, custom_task_variable, value)\n",
    "\n",
    "    @classmethod\n",
    "    def load(cls, artifact_directory):\n",
    "\n",
    "        return cls.load_task(artifact_directory)\n",
    "\n",
    "    @classmethod\n",
    "    def load_task(cls, artifact_directory):\n",
    "        \"\"\"\n",
    "        Helper method to abstract deserializing the pickle object stored within `artifact_directory` and\n",
    "        returning the custom task. Any variables that were excluded in `save_task` must be manually loaded\n",
    "        proceeding this function.\n",
    "\n",
    "        Parameters\n",
    "        ----------\n",
    "        artifact_directory: str\n",
    "            Path to the directory to save the serialized artifact(s) to.\n",
    "\n",
    "        Returns\n",
    "        -------\n",
    "        cls\n",
    "            The deserialized object\n",
    "        \"\"\"\n",
    "        with open(\n",
    "            os.path.join(artifact_directory, Serializable.default_artifact_filename), \"rb\"\n",
    "        ) as fp:\n",
    "            deserialized_object = pickle.load(fp)\n",
    "\n",
    "        if not isinstance(deserialized_object, cls):\n",
    "            raise DrumCommonException(\n",
    "                \"load_task method must return a {} class\".format(cls.__name__)\n",
    "            )\n",
    "        return deserialized_object"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "726af462-9065-4c1e-a248-bbc4eb997326",
   "metadata": {
    "pycharm": {
     "name": "#%% md\n"
    }
   },
   "source": [
    "# Testing the Custom Task "
   ]
  },
  {
   "cell_type": "markdown",
   "id": "8d984337-fe11-470d-97cc-6e9c2c406c63",
   "metadata": {
    "pycharm": {
     "name": "#%% md\n"
    }
   },
   "source": [
    "Because our CustomTask is simply a python class, we can utilize standard python testing libraries / frameworks. \n",
    "\n",
    "E.g. we can verify with a test dataset that our CustomTask matches the expected output"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 184,
   "id": "6ffdf3e7-928f-4f14-9258-631339228d1a",
   "metadata": {
    "pycharm": {
     "name": "#%%\n"
    }
   },
   "outputs": [],
   "source": [
    "# Note: for this example we're simply checking the outputs against itself. But in reality this would be a list of previous predictions\n",
    "previous_predictions = task.predict(X).values\n",
    "assert all(task.predict(X).values == previous_predictions)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "78c16fef-b069-47f2-8c87-544b54a52b4a",
   "metadata": {
    "pycharm": {
     "name": "#%% md\n"
    }
   },
   "source": [
    "We can also make assertions about our output to ensure our predictions are in a sensible range"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 189,
   "id": "520123e9-7651-4ae6-ac69-5ab57deea65e",
   "metadata": {
    "pycharm": {
     "name": "#%%\n"
    }
   },
   "outputs": [],
   "source": [
    "maximum_prediction = 50\n",
    "assert task.predict(X).values.max() < maximum_prediction"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "544830e3-ee18-4f32-aa0e-f3ace55a799e",
   "metadata": {
    "pycharm": {
     "name": "#%% md\n"
    }
   },
   "source": [
    "These tests can easily be incorporated into a CI/CD pipeline and run automatically!"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "4014a38c-5dbc-4fc9-a605-abb31de2e673",
   "metadata": {
    "pycharm": {
     "name": "#%% md\n"
    }
   },
   "source": [
    "# How to upload a Custom Task via the web application"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "142c7bd9-5dfa-4b72-9973-ef8aa33dc70d",
   "metadata": {
    "pycharm": {
     "name": "#%% md\n"
    }
   },
   "source": [
    "The easiest way to upload your custom task is to place all of your code into a file called custom.py. For practice, you can copy the code in the CustomTask above (the Keras regression estimator near the top) and place it in a custom.py. Note that [here] is a example of this exact Keras Regressor Custom Task with all the necessary files if you want to inspect it. \n",
    "\n",
    "The next step is to define a model-metadata.yaml. This tells DataRobot a bit about your custom task, e.g. if it's a binary or regression estimator, and also provides guardrails about what type of data it accepts as input and output. E.g. our Keras estimator above only handles numeric data and can't handle missing values. \n",
    "\n",
    "Finally, you may need to create a requirements.txt file. As mentioned above, your CustomTask is run in a Docker Container. DataRobot by default provides a series of drop in environments for each language / framework (see [here]). In the case of the Keras regression CustomTask above, we don't need to import anything outside what that environment contains. You can see an example of a CustomTask using a requirements.txt file [here]"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "d6b62112-57d4-4ce0-a07d-d0597d67779d",
   "metadata": {
    "pycharm": {
     "name": "#%% md\n"
    }
   },
   "source": [
    "Once you have all the files ready, you can follow instructions [here] to upload your CustomTask through the Model Registry -> Tasks page. Then you can follow instructions [here] to use your custom task in a DataRobot blueprint"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "1a6aac02-0d78-4945-9806-d580877785f5",
   "metadata": {
    "pycharm": {
     "name": "#%% md\n"
    }
   },
   "source": [
    "# How to upload a Custom Task through the API"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "39afa616-6e7c-4d4b-bc50-fedcd828e75e",
   "metadata": {
    "pycharm": {
     "name": "#%% md\n"
    }
   },
   "source": [
    "DataRobot supports a robust python, R, and REST API that you use to programatically control data prep, modeling, and deployments. The goal is to let you create an end to end model pipeline without ever having to leave your favorite IDE or notebook (although please at least look at our web application, it's pretty!) \n",
    "\n",
    "In the next few sections we'll show how to quickly create a project to use as an example. We'll then run a few of DataRobot's default blueprints so we can add our CustomTask to them. Then we'll upload the CustomTask we created above, visualize the existing blueprint, and finally add our custom blueprint to the task. \n",
    "\n",
    "Please keep in mind the focus here is on CustomTasks and not a general walkthrough of the API. For that, please see the excellent tutorials [here]"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "a668c8b8-ae0f-40e0-bee5-e8b4fecdccd0",
   "metadata": {
    "pycharm": {
     "name": "#%% md\n"
    }
   },
   "source": [
    "## Create and run a project"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "e19f12be-f7a7-4145-b244-b087d493e2f4",
   "metadata": {
    "pycharm": {
     "name": "#%% md\n"
    }
   },
   "source": [
    "We'll be using the docs here if you'd like more information: https://datarobot-public-api-client.readthedocs-hosted.com/en/v2.27.1/reference/modeling/project.html#create-a-project"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 54,
   "id": "e949628d-78cb-45b2-a4dd-cd4a1253bb03",
   "metadata": {
    "pycharm": {
     "name": "#%%\n"
    }
   },
   "outputs": [],
   "source": [
    "import datarobot as dr"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 56,
   "id": "efbd8a80-571f-4f39-a009-5fbf2378ffd6",
   "metadata": {
    "pycharm": {
     "name": "#%%\n"
    }
   },
   "outputs": [],
   "source": [
    "# Note, we can also use dr.Dataset.create_from_in_memory_data to create our dataset directly from the pandas dataframe above\n",
    "dataset = dr.Dataset.create_from_file(file_path=\"tests/testdata/juniors_3_year_stats_regression.csv\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 57,
   "id": "3d0fb99a-2f86-46ba-9e4c-8b912fa87028",
   "metadata": {
    "pycharm": {
     "name": "#%%\n"
    }
   },
   "outputs": [],
   "source": [
    "# Note: we could also have just used dr.Project.create to directly create the project. \n",
    "project = dr.Project.create_from_dataset(dataset.id, project_name=\"Keras Regression Custom Task Tutorial\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 59,
   "id": "f89a6a87-e64e-42cf-b777-27afd22b657b",
   "metadata": {
    "pycharm": {
     "name": "#%%\n"
    }
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "Project(Keras Regression Custom Task Tutorial)"
      ]
     },
     "execution_count": 59,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# This will start the modeling, note that it will run asynchronously \n",
    "target_name = \"Grade 2014\"\n",
    "project.analyze_and_model(target=target_name,\n",
    "                   worker_count=-1, # Max workers\n",
    "                   mode=dr.AUTOPILOT_MODE.QUICK\n",
    "                  )\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 60,
   "id": "91ec0a77-de76-43fe-ac7c-ff27174bddcd",
   "metadata": {
    "pycharm": {
     "name": "#%%\n"
    }
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "In progress: 2, queued: 10 (waited: 0s)\n",
      "In progress: 2, queued: 10 (waited: 0s)\n",
      "In progress: 2, queued: 10 (waited: 1s)\n",
      "In progress: 2, queued: 10 (waited: 1s)\n",
      "In progress: 2, queued: 10 (waited: 2s)\n",
      "In progress: 2, queued: 10 (waited: 4s)\n",
      "In progress: 1, queued: 9 (waited: 7s)\n",
      "In progress: 4, queued: 6 (waited: 14s)\n",
      "In progress: 4, queued: 1 (waited: 27s)\n",
      "In progress: 0, queued: 0 (waited: 47s)\n",
      "In progress: 4, queued: 1 (waited: 67s)\n",
      "In progress: 1, queued: 0 (waited: 87s)\n",
      "In progress: 0, queued: 0 (waited: 108s)\n",
      "In progress: 1, queued: 0 (waited: 128s)\n",
      "In progress: 1, queued: 0 (waited: 148s)\n",
      "In progress: 1, queued: 0 (waited: 168s)\n",
      "In progress: 1, queued: 0 (waited: 188s)\n",
      "In progress: 0, queued: 0 (waited: 208s)\n",
      "In progress: 0, queued: 0 (waited: 229s)\n"
     ]
    }
   ],
   "source": [
    "# This will let us know when autopilot is finished\n",
    "project.wait_for_autopilot()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 62,
   "id": "8c21316f-7835-4ab1-acf9-08727373cfcb",
   "metadata": {
    "pycharm": {
     "name": "#%%\n"
    }
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "[Model('Keras Slim Residual Neural Network Regressor using Training Schedule (1 Layer: 64 Units)'), Model('Keras Slim Residual Neural Network Regressor using Training Schedule (1 Layer: 64 Units)'), Model('Ridge Regressor')]\n"
     ]
    }
   ],
   "source": [
    "# We can now list the top 3 models that ran\n",
    "models = project.get_models()\n",
    "print(models[:3])"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "e45f560c-4b02-460c-a8da-16d7b571438a",
   "metadata": {
    "pycharm": {
     "name": "#%% md\n"
    }
   },
   "source": [
    "## Upload our Custom Task"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "2be6c143-5425-48c0-a428-3eef225613f5",
   "metadata": {
    "pycharm": {
     "name": "#%% md\n"
    }
   },
   "source": [
    "Now that we have a project with models to work with, let's upload our CustomTask! \n",
    "We'll be using the docs here: https://datarobot-public-api-client.readthedocs-hosted.com/en/v2.27.1/reference/modeling/spec/custom_task.html"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "afc84db6-a310-4cf5-be96-fa6e55c911df",
   "metadata": {
    "pycharm": {
     "name": "#%% md\n"
    }
   },
   "source": [
    "The first thing we'll need to do is ensure we have the appropriate execution environment for our CustomTask. You can find a list of these in the github [here]."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 65,
   "id": "31bc4380-db44-438a-8d0a-6ce52c0e7e2e",
   "metadata": {
    "pycharm": {
     "name": "#%%\n"
    }
   },
   "outputs": [],
   "source": [
    "execution_environment = dr.ExecutionEnvironment.create(\n",
    "    name=\"Python3 Keras Environment\",\n",
    "    description=\"This environment contains Python3 keras library.\",\n",
    ")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 66,
   "id": "b3d99e06-a85f-44d7-9276-83fcc1e5114c",
   "metadata": {
    "pycharm": {
     "name": "#%%\n"
    }
   },
   "outputs": [],
   "source": [
    "# Note this may take a few minutes to build the Docker image\n",
    "execution_environment_version = dr.ExecutionEnvironmentVersion.create(execution_environment.id, \n",
    "                                                                      \"public_dropin_environments/python3_keras\", \n",
    "                                                                      \"First Version\")"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "bdf7bd20-d7f4-4b1d-b724-be325b38434a",
   "metadata": {
    "pycharm": {
     "name": "#%% md\n"
    }
   },
   "source": [
    "Now we can create a CustomTask and a CustomTaskVersion. We use versions because it's likely will iterate on our CustomTask, \n",
    "and DataRobot by default will use the default version"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 68,
   "id": "4dfedc99-8a50-433b-b5bb-bcdf558db7ef",
   "metadata": {
    "pycharm": {
     "name": "#%%\n"
    }
   },
   "outputs": [],
   "source": [
    "from datarobot.enums import CUSTOM_TASK_TARGET_TYPE\n",
    "\n",
    "keras_regression = dr.CustomTask.create(\n",
    "    name=\"keras custom task regressor\",\n",
    "    target_type=CUSTOM_TASK_TARGET_TYPE.REGRESSION\n",
    ")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 69,
   "id": "d5582354-d697-4073-9a45-d253a4e316ab",
   "metadata": {
    "pycharm": {
     "name": "#%%\n"
    }
   },
   "outputs": [],
   "source": [
    "custom_task_folder = \"tests/fixtures/custom_task_interface_keras_regression\"\n",
    "\n",
    "task_version = dr.CustomTaskVersion.create_clean(\n",
    "    custom_task_id=keras_regression.id,\n",
    "    base_environment_id=execution_environment.id,\n",
    "    folder_path=custom_task_folder,\n",
    ")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 72,
   "id": "36dfd987-1f34-437a-82dc-42845b55c689",
   "metadata": {
    "pycharm": {
     "name": "#%%\n"
    }
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "CustomTaskVersion('v1.0')"
      ]
     },
     "execution_count": 72,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "keras_regression.refresh() # We have to perform a GET request to update\n",
    "keras_regression.latest_version"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "8a78a09f-5816-4113-8647-f4b7980faab9",
   "metadata": {
    "pycharm": {
     "name": "#%% md\n"
    }
   },
   "source": [
    "## Visualize an existing Blueprint"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "912043dc-cc61-4cf8-8b0c-1585b58ad54a",
   "metadata": {
    "pycharm": {
     "name": "#%% md\n"
    }
   },
   "source": [
    "Now we'll use the blueprint workshop API: https://blueprint-workshop.datarobot.com/getting_started.html\n",
    "\n",
    "Please make sure you've installed the appropriate dependencies, notably graphviz (see https://blueprint-workshop.datarobot.com/setup.html) \n",
    "\n",
    "The key point to remember is that DataRobot's Blueprints (BPs) are Directed Acyclic Graphs (DAGs). This means that data flows from left to right. Typically the data is split by data types and fed to different paths or branches that apply preprocessing. Our dataset only has numeric data, so that's all we see. "
   ]
  },
  {
   "cell_type": "markdown",
   "id": "d3cbd998-c743-49d4-91de-a2b2094015eb",
   "metadata": {
    "pycharm": {
     "name": "#%% md\n"
    }
   },
   "source": [
    "Let's visualize an exisitng blueprint that was created by DataRobot's AutoML"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 78,
   "id": "0e386c52-9986-429e-9db7-e2aeec801eaf",
   "metadata": {
    "pycharm": {
     "name": "#%%\n"
    }
   },
   "outputs": [],
   "source": [
    "from datarobot_bp_workshop import Workshop, Visualize\n",
    "w = Workshop()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 191,
   "id": "3afea1c9-3aa6-42c0-b663-2526aec4f8eb",
   "metadata": {
    "pycharm": {
     "name": "#%%\n"
    }
   },
   "outputs": [],
   "source": [
    "# One important note is that models and blueprints are separate. A blueprint trained on a dataset becomes a model\n",
    "\n",
    "# This will default to getting the top model \n",
    "models = project.get_models()\n",
    "top_model = models[5] # This is for convenience, you can choose any model you'd like\n",
    "blueprint = dr.Blueprint.get(project.id, top_model.blueprint_id)\n"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "5e4111f6-d8e4-4a11-b033-768270d17a3a",
   "metadata": {
    "pycharm": {
     "name": "#%% md\n"
    }
   },
   "source": [
    "Note: if you're comparing to the DataRobot Web Application UI, make sure to click \"Copy and Edit\" otherwise you might see a different visual.\n",
    "This is because the blueprint shown on the leaderboard automatically removes from the visualization any pathways that aren't used. In this case the leaderboard UI will show only the Numeric Variables path because our dataset only has numerics. But the same blueprint run on Numerics + Text data would show both pathways."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 173,
   "id": "fc06c683-ce82-4f30-aeca-8cf6e12fc48c",
   "metadata": {
    "pycharm": {
     "name": "#%%\n"
    }
   },
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<IPython.core.display.Image object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "Visualize.show_dr_blueprint(blueprint) "
   ]
  },
  {
   "cell_type": "markdown",
   "id": "e955f8b5-722f-4ec0-916c-a245b5cbb042",
   "metadata": {
    "pycharm": {
     "name": "#%% md\n"
    }
   },
   "source": [
    "We can inspect any blueprint to understand what tasks it has. This will give us information if we want to create a brand new blueprint using our CustomTask or insert our CustomTask into an existing blueprint and then modify it further with DataRobot's tasks"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 130,
   "id": "991d9baa-7005-41f3-b71b-d256eb040766",
   "metadata": {
    "pycharm": {
     "name": "#%%\n"
    }
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "w = Workshop(user_blueprint_id='6202e28f1255e7118a50d079')\n",
      "\n",
      "pdm3 = w.Tasks.PDM3(w.TaskInputs.CAT)\n",
      "pdm3.set_task_parameters(cm=50000, sc=10)\n",
      "\n",
      "ndc = w.Tasks.NDC(w.TaskInputs.NUM)\n",
      "\n",
      "rdt5 = w.Tasks.RDT5(ndc)\n",
      "\n",
      "ptm3 = w.Tasks.PTM3(w.TaskInputs.TXT)\n",
      "ptm3.set_task_parameters(d2=0.8, mxf=20000, d1=1, n='l1')\n",
      "\n",
      "ndc_1 = w.Tasks.NDC(w.TaskInputs.DATE)\n",
      "\n",
      "rst = w.Tasks.RST(ndc_1)\n",
      "rst.set_task_parameters(spt=1)\n",
      "\n",
      "kerasr = w.Tasks.KERASR(rdt5, rst, pdm3, ptm3)\n",
      "kerasr.set_task_parameters(epochs=8, hidden_dropout=0.05, hidden_units='list(64)', learning_rate=0.03, use_training_schedule=1)\n",
      "\n",
      "kerasr_blueprint = w.BlueprintGraph(kerasr, name='Keras Slim Residual Neural Network Regressor using Training Schedule (1 Layer: 64 Units)')\n"
     ]
    }
   ],
   "source": [
    "blueprint_graph = w.clone(top_model.blueprint_id, project.id)\n",
    "source_code = blueprint_graph.to_source_code(to_stdout=True)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "72c2d748-2e49-4591-b0b0-e326fdaa4a82",
   "metadata": {
    "pycharm": {
     "name": "#%% md\n"
    }
   },
   "source": [
    "You can also easily list all the tasks, see https://blueprint-workshop.datarobot.com/examples/walkthrough/Walkthrough.html#Help-with-Tasks for more details"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "089a4083-3a3c-40dc-a4fc-a0619e4f2499",
   "metadata": {
    "pycharm": {
     "name": "#%% md\n"
    }
   },
   "source": [
    "## Retrieve Custom Task"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 155,
   "id": "14580243-7f64-49c3-99a0-13ac6fc9b238",
   "metadata": {
    "pycharm": {
     "name": "#%%\n"
    }
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "keras custom task regressor: [CUSTOMR_6202d89c1255e7118a50d059] \n",
       "  - (No description)"
      ]
     },
     "execution_count": 155,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "w.search_tasks(\"keras custom task regressor\") # the name of our task above"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "d43845c9-da52-4c44-9332-ef1b2839ee8c",
   "metadata": {
    "pycharm": {
     "name": "#%% md\n"
    }
   },
   "source": [
    "## Add our CustomTask"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "0b58fcef-ffd1-43d4-be97-24b69fab4919",
   "metadata": {
    "pycharm": {
     "name": "#%% md\n"
    }
   },
   "source": [
    "Now that we have our top blueprint, let's replace the Keras Neural Network estimator with our own CustomTask! We can grab the source code from the blueprint above, but we'll replace the existing keras estimator with our custom task\n",
    "\n"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 192,
   "id": "5e530da0-0ff5-4924-b037-ea4672945370",
   "metadata": {
    "pycharm": {
     "name": "#%%\n"
    }
   },
   "outputs": [],
   "source": [
    "# Copied from above\n",
    "w = Workshop(user_blueprint_id='6202e28f1255e7118a50d079')\n",
    "\n",
    "pdm3 = w.Tasks.PDM3(w.TaskInputs.CAT)\n",
    "pdm3.set_task_parameters(cm=50000, sc=10)\n",
    "\n",
    "ndc = w.Tasks.NDC(w.TaskInputs.NUM)\n",
    "\n",
    "rdt5 = w.Tasks.RDT5(ndc)\n",
    "\n",
    "ptm3 = w.Tasks.PTM3(w.TaskInputs.TXT)\n",
    "ptm3.set_task_parameters(d2=0.8, mxf=20000, d1=1, n='l1')\n",
    "\n",
    "ndc_1 = w.Tasks.NDC(w.TaskInputs.DATE)\n",
    "\n",
    "rst = w.Tasks.RST(ndc_1)\n",
    "rst.set_task_parameters(spt=1)\n",
    "\n",
    "# This is the only difference, we replace the KERASR with our Custom Task\n",
    "kerasr = w.CustomTasks.CUSTOMR_6202d89c1255e7118a50d059(rdt5, rst, pdm3, ptm3)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 168,
   "id": "b3d075a1-ef02-4347-b6eb-d7204b012c86",
   "metadata": {
    "pycharm": {
     "name": "#%%\n"
    }
   },
   "outputs": [],
   "source": [
    "keras_custom_task_blueprint = w.BlueprintGraph(kerasr, name='Keras Custom Task')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 169,
   "id": "84608d81-e736-457f-bfb7-8eec1ec8c5a5",
   "metadata": {
    "pycharm": {
     "name": "#%%\n"
    }
   },
   "outputs": [
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<IPython.core.display.Image object>"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    }
   ],
   "source": [
    "keras_custom_task_blueprint.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 171,
   "id": "a24ec610-8543-421f-ac46-f20e4ef67fa2",
   "metadata": {
    "pycharm": {
     "name": "#%%\n"
    }
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Training requested! Blueprint Id: e30d8f9ec2290d95f65b97de4c78307e\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "Name: 'Keras Custom Task'\n",
       "\n",
       "Input Data: Categorical | Numeric | Text | Date\n",
       "Tasks: One-Hot Encoding | Numeric Data Cleansing | Smooth Ridit Transform | Matrix of word-grams occurrences | Standardize | keras custom task regressor"
      ]
     },
     "execution_count": 171,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# Note this is asynchronous\n",
    "keras_custom_task_blueprint.train(project_id=project.id)"
   ]
  },
  {
   "cell_type": "markdown",
   "id": "8d551510-cb1e-4f07-b41a-790d7fd70777",
   "metadata": {
    "pycharm": {
     "name": "#%% md\n"
    }
   },
   "source": [
    "Congratulations! You've now successfully created a CustomTask, uploaded it to DataRobot, and inserted it into a Blueprint all within a Jupyter Notebook!"
   ]
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "drum_1.6.6",
   "language": "python",
   "name": "drum_1.6.6"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.7.12"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 5
}