1 H2O AI Cloud: Quickstart.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# End-to-end workflow in the H2O AI Cloud\n",
    "\n",
    "Get started building and deploying machine learning models in the H2O AI Cloud.\n",
    "\n",
    "* Create an AI engine for building models\n",
    "* Use AutoML to build a machine learning model\n",
    "* Deploy the model to production"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Prerequistes\n",
    "\n",
    "Update the `h2o_ai_cloud.py` file with the connection parameters for your H2O AI Cloud environemnt:\n",
    "1. Login to your H2O AI Cloud environment\n",
    "1. Click your username or avatar in the H2O AI Cloud navigation bar\n",
    "1. Navigate to `CLI & API Access`\n",
    "1. Use the variables from the `Accessing H2O AI Cloud APIs` section to populate the parameters"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Create an AI engine\n",
    "Create a Driverless AI engine for access to automated machine learning to build models for us on our data. \n",
    "\n",
    "See the `2 AI Engines` tutorial for more details on how to use and interact with **Steam** for creating and managing your AI Engines."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [],
   "source": [
    "from h2o_ai_cloud import steam_client\n",
    "from h2osteam.clients import DriverlessClient"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Visit https://cloud.h2o.ai/auth/get-platform-token to get your platform token\n",
      "Enter your platform token: ········\n"
     ]
    }
   ],
   "source": [
    "steam = steam_client()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "['visitor-driverless-kubernetes']\n"
     ]
    }
   ],
   "source": [
    "# Profiles determin the resourcs that an AI Engine will have access to - see the Steam tutorial for more information\n",
    "dai_profiles = [profile[\"name\"] for profile in steam.get_profiles() if profile[\"profile_type\"] == \"driverless_kubernetes\"]\n",
    "\n",
    "print(dai_profiles)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "Creating a new machine can take 2-20 minutes depending on your environment and configuration settings. Please reachout to your admin as needed to reduce start times. "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Driverless AI instance is submitted, please wait...\n",
      "Driverless AI instance is running\n"
     ]
    }
   ],
   "source": [
    "dai_machine = DriverlessClient(steam).launch_instance(\n",
    "    name=\"my-automl\",\n",
    "    version=steam.get_driverless_engines()[-1][\"version\"],  # newest version of the AutoML software\n",
    "    profile_name=dai_profiles[0], # profiles will have different names in different environments\n",
    "    max_uptime_h=1\n",
    ")\n",
    "dai = dai_machine.connect()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Build a model"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Import data"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Complete 100.00% - [4/4] Computed stats for column Churn?\n"
     ]
    }
   ],
   "source": [
    "telco_churn = dai.datasets.create(\n",
    "    data=\"https://h2o-internal-release.s3-us-west-2.amazonaws.com/data/Splunk/churn.csv\",  \n",
    "    data_source=\"s3\", \n",
    "    name=\"Telco_Churn\",\n",
    "    force=True\n",
    ")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "b34aade4-fcc5-11ec-bbbf-8af6970414ab | Telco_Churn\n",
      "\n",
      "Columns: ['State', 'Account Length', 'Area Code', 'Phone', \"Int'l Plan\", 'VMail Plan', 'VMail Message', 'Day Mins', 'Day Calls', 'Day Charge', 'Eve Mins', 'Eve Calls', 'Eve Charge', 'Night Mins', 'Night Calls', 'Night Charge', 'Intl Mins', 'Intl Calls', 'Intl Charge', 'CustServ Calls', 'Churn?']\n",
      "\n",
      "Shape: (3333, 21)\n"
     ]
    }
   ],
   "source": [
    "print(telco_churn.key, \"|\", telco_churn.name)\n",
    "print(\"\\nColumns:\", telco_churn.columns)\n",
    "print('\\nShape:', telco_churn.shape)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Run an AutoML experiment\n",
    "\n",
    "See the `3 AutoML` tutorial for more details on how to use and interact with Driverless AI for AutoML. "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Experiment launched at: https://steam.cloud.h2o.ai:443/proxy/driverless/3287/#/experiment?key=bc7539d4-fcc5-11ec-bbbf-8af6970414ab\n",
      "Complete 100.00% - Status: Complete                                             \n"
     ]
    }
   ],
   "source": [
    "default_baseline = dai.experiments.create(\n",
    "    name='Default Baseline', \n",
    "    train_dataset=telco_churn, \n",
    "    target_column=\"Churn?\", \n",
    "    task=\"classification\",\n",
    "    accuracy=1, time=1, interpretability=6  # a quick AutoML experiment to see a baseline    \n",
    ")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 8,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Status: Complete\n",
      "Experiment: Default Baseline (bc7539d4-fcc5-11ec-bbbf-8af6970414ab)\n",
      "  Version: 1.10.3, 2022-07-06 00:53\n",
      "  Settings: 1/1/6, seed=6190721, GPUs enabled\n",
      "  Train data: Telco_Churn (3333, 21)\n",
      "  Validation data: N/A\n",
      "  Test data: N/A\n",
      "  Target column: Churn? (binary, 14.491% target class)\n",
      "System specs: Docker/Linux, 54 GB, 32 CPU cores, 1/1 GPU\n",
      "  Max memory usage: 0.763 GB, 0 GB GPU\n",
      "Recipe: AutoDL (5 iterations, 2 individuals)\n",
      "  Validation scheme: stratified, 1 internal holdout\n",
      "  Feature engineering: 28 features scored (19 selected)\n",
      "Timing: MOJO latency 0.0399 millis (214.2kB), Python latency 84.1064 millis (141.1kB)\n",
      "  Data preparation: 15.50 secs\n",
      "  Shift/Leakage detection: 1.18 secs\n",
      "  Model and feature tuning: 17.01 secs (6 models trained)\n",
      "  Feature evolution: 1.30 secs (0 of 3 model trained)\n",
      "  Final pipeline training: 26.99 secs (7 models trained)\n",
      "  Python / MOJO scorer building: 36.72 secs / 15.54 secs\n",
      "Validation score: LOGLOSS = 0.4137853 (constant preds of -1.775)\n",
      "Validation score: LOGLOSS = 0.3111744 +/- 0.0127315 (baseline)\n",
      "Validation score: LOGLOSS = 0.1618368 +/- 0.01040678 (final pipeline)\n",
      "Test score:       LOGLOSS = N/A\n"
     ]
    }
   ],
   "source": [
    "default_baseline.summary()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Deploy the model"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Create a project\n",
    "We willl create a project for our use case to easily share our work with others and deploy the models"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 9,
   "metadata": {},
   "outputs": [],
   "source": [
    "churn_project = dai.projects.create(\n",
    "    name=\"Telco Churn Predictions\", \n",
    "    description=\"Which of our customers is likely to cancel their contract?\",\n",
    "    force=\"True\"\n",
    ")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 10,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "<driverlessai._projects.Project at 0x110df63a0>"
      ]
     },
     "execution_count": 10,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "churn_project.link_experiment(default_baseline)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Connect to MLOps\n",
    "See the `5 Model Deployment` tutorial for more details on how to use and interact with MLOps for managing models and deploymnets. "
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {},
   "outputs": [],
   "source": [
    "from h2o_ai_cloud import mlops_client"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Visit https://cloud.h2o.ai/auth/get-platform-token to get your platform token\n",
      "Enter your platform token: ········\n"
     ]
    }
   ],
   "source": [
    "mlops = mlops_client()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Create a deployment"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "5ba7c218-8065-4e11-a819-c3bcbb991e1f DEV\n",
      "9ad0c266-5913-4616-9057-9daa3188252f PROD\n"
     ]
    }
   ],
   "source": [
    "my_project_deployment_environments = mlops.storage.deployment_environment.list_deployment_environments(body={\n",
    "    'project_id': churn_project.key, \n",
    "}).deployment_environment\n",
    "\n",
    "for de in my_project_deployment_environments:\n",
    "    print(de.id, de.display_name)\n",
    "    if de.display_name == \"DEV\":\n",
    "        deployment_env = de.id"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 14,
   "metadata": {},
   "outputs": [],
   "source": [
    "churn_model_deployed = mlops.storage.deployment_environment.deploy({\n",
    "    'deployment_environment_id': de.id,\n",
    "    'experiment_id': default_baseline.key,\n",
    "    'type': 'SINGLE_MODEL'\n",
    "})"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### View information about our deployment"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 15,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "{'deployment_id': '',\n",
       " 'grafana_endpoint': '',\n",
       " 'last_modified_time': None,\n",
       " 'message': '',\n",
       " 'scorer': None,\n",
       " 'state': 'DEPLOYMENT_STATE_UNSPECIFIED'}"
      ]
     },
     "execution_count": 15,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "mlops.deployer.deployment_status.get_deployment_status({\n",
    "    'deployment_id': churn_model_deployed.deployment.id\n",
    "}).deployment_status"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Wait for deployment to become healthy"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 16,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "{'deployment_id': '7b695155-de7e-4b27-b219-48d150ab6930',\n",
       " 'grafana_endpoint': 'https://mlops-monitoring.cloud.h2o.ai/d/7b695155-de7e-4b27-b219-48d150ab6930',\n",
       " 'last_modified_time': datetime.datetime(2022, 7, 6, 0, 57, 29, tzinfo=tzutc()),\n",
       " 'message': 'N/A',\n",
       " 'scorer': {'capabilities': {'method': 'GET',\n",
       "                             'url': 'https://model.cloud.h2o.ai/7b695155-de7e-4b27-b219-48d150ab6930/model/capabilities'},\n",
       "            'sample_request': {'method': 'GET',\n",
       "                               'url': 'https://model.cloud.h2o.ai/7b695155-de7e-4b27-b219-48d150ab6930/model/sample_request'},\n",
       "            'score': {'method': 'POST',\n",
       "                      'url': 'https://model.cloud.h2o.ai/7b695155-de7e-4b27-b219-48d150ab6930/model/score'},\n",
       "            'type': 'MOJO2_REST_SCORER'},\n",
       " 'state': 'HEALTHY'}"
      ]
     },
     "execution_count": 16,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "import time\n",
    "\n",
    "deployment_status = mlops.deployer.deployment_status.get_deployment_status({\n",
    "        'deployment_id': churn_model_deployed.deployment.id\n",
    "    }).deployment_status.state\n",
    "\n",
    "while deployment_status != \"HEALTHY\":\n",
    "    time.sleep(5)\n",
    "    \n",
    "    deployment_status = mlops.deployer.deployment_status.get_deployment_status({\n",
    "        'deployment_id': churn_model_deployed.deployment.id\n",
    "    }).deployment_status.state\n",
    "    \n",
    "mlops.deployer.deployment_status.get_deployment_status({\n",
    "    'deployment_id': churn_model_deployed.deployment.id\n",
    "}).deployment_status"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 17,
   "metadata": {},
   "outputs": [],
   "source": [
    "sample_request = mlops.deployer.deployment_status.get_deployment_status({\n",
    "    'deployment_id': churn_model_deployed.deployment.id\n",
    "}).deployment_status.scorer.sample_request.url\n",
    "\n",
    "\n",
    "deployment_url = mlops.deployer.deployment_status.get_deployment_status({\n",
    "    'deployment_id': churn_model_deployed.deployment.id\n",
    "}).deployment_status.scorer.score.url"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Make a prediction with new data"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 18,
   "metadata": {},
   "outputs": [],
   "source": [
    "import requests\n",
    "import json"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "{'fields': ['State',\n",
       "  'Account Length',\n",
       "  'Area Code',\n",
       "  \"Int'l Plan\",\n",
       "  'VMail Plan',\n",
       "  'VMail Message',\n",
       "  'Day Mins',\n",
       "  'Day Calls',\n",
       "  'Day Charge',\n",
       "  'Eve Mins',\n",
       "  'Eve Calls',\n",
       "  'Eve Charge',\n",
       "  'Night Mins',\n",
       "  'Night Calls',\n",
       "  'Night Charge',\n",
       "  'Intl Mins',\n",
       "  'Intl Calls',\n",
       "  'Intl Charge',\n",
       "  'CustServ Calls'],\n",
       " 'rows': [['text',\n",
       "   '0',\n",
       "   '0',\n",
       "   'text',\n",
       "   'text',\n",
       "   '0',\n",
       "   '0',\n",
       "   '0',\n",
       "   '0',\n",
       "   '0',\n",
       "   '0',\n",
       "   '0',\n",
       "   '0',\n",
       "   '0',\n",
       "   '0',\n",
       "   '0',\n",
       "   '0',\n",
       "   '0',\n",
       "   '0']]}"
      ]
     },
     "execution_count": 19,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "sample_request_as_text = requests.get(sample_request).text\n",
    "sample_request = json.loads(sample_request_as_text)\n",
    "sample_request"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 20,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "{'fields': ['Churn?.False.', 'Churn?.True.'],\n",
       " 'id': 'bc7539d4-fcc5-11ec-bbbf-8af6970414ab',\n",
       " 'score': [['0.96486217', '0.035137847']]}"
      ]
     },
     "execution_count": 20,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "new_predictions = requests.post(\n",
    "    url=deployment_url,\n",
    "    json={\n",
    "        'fields': sample_request[\"fields\"],\n",
    "        'rows': sample_request[\"rows\"]\n",
    "    }\n",
    ")\n",
    "\n",
    "predictions = json.loads(new_predictions.text)\n",
    "predictions"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 21,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Churn Percent: 3.5%\n"
     ]
    }
   ],
   "source": [
    "print(f\"Churn Percent: {round(float(predictions['score'][0][1]) * 100, 1)}%\")"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Clean up\n",
    "Delete our AutoML machine - we can import the project to any new Driverless AI machine, and the model will remain deployed to MLOps"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 22,
   "metadata": {},
   "outputs": [],
   "source": [
    "dai_machine.terminate()"
   ]
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 3 (ipykernel)",
   "language": "python",
   "name": "python3"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.8.12"
  },
  "toc": {
   "base_numbering": 1,
   "nav_menu": {},
   "number_sections": true,
   "sideBar": true,
   "skip_h1_title": true,
   "title_cell": "Table of Contents",
   "title_sidebar": "Contents",
   "toc_cell": false,
   "toc_position": {},
   "toc_section_display": true,
   "toc_window_display": true
  }
 },
 "nbformat": 4,
 "nbformat_minor": 5
}