{ "cells": [ { "attachments": {}, "cell_type": "markdown", "metadata": { "tags": [] }, "source": [ "## Prediction\n", "### Deploy ML model using KServe inference service" ] }, { "cell_type": "code", "execution_count": null, "metadata": {}, "outputs": [], "source": [ "import yaml\n", "import json\n", "\n", "def deploy_kserve_model(json_file_path):\n", " try:\n", " # Run kubectl apply command using subprocess\n", " subprocess.run(['kubectl', 'apply', '-f', json_file_path], check=True)\n", " print(\"Deployment successful!\")\n", " except subprocess.CalledProcessError as e:\n", " print(f\"Deployment failed. Error: {e}\")\n", "\n", "###################################################################################\n", "## update kserve-mlflow-deploy \n", "with open('kserve/kserve-mlflow-deploy.yaml', 'r') as file:\n", " yaml_data = yaml.safe_load(file)\n", "\n", "## set model name - tutorial default logs model under experiment name\n", "model_name = 'bike-sharing-exp'\n", "with open('best-model-uri.txt','r') as f:\n", " storage_uri = f.read()\n", "# Convert YAML to JSON\n", "json_data = json.dumps(yaml_data)\n", "data = json.loads(json_data)\n", "\n", "data[\"metadata\"][\"namespace\"] = os.environ['NOTEBOOK_NAMESPACE']\n", "data[\"metadata\"][\"name\"] = model_name\n", "data['spec']['predictor']['sklearn']['storageUri'] = f\"{storage_uri}/model.pkl\"\n", "\n", "# Save the JSON data to a file\n", "with open('kserve/kserve-mlflow-deploy.json', 'w') as file:\n", " file.write(json.dumps(data))\n", "\n", "# Specify the path to your JSON file\n", "json_file_path = 'kserve/kserve-mlflow-deploy.json'\n", "\n", "# Call the function to deploy the Kubernetes resource to deploy ml model \n", "deploy_kserve_model(json_file_path)" ] }, { "cell_type": "markdown", "metadata": {}, "source": [ "### Deployment of kserve will take few minutes to comes to ready state, so wait for few minutes before actually executing the below cell." ] }, { "cell_type": "code", "execution_count": null, "metadata": {}, "outputs": [], "source": [ "import os\n", "import requests\n", "\n", "DOMAIN_NAME = \"svc.cluster.local\" # change this to your domain for external access \n", "NAMESPACE = os.environ['NOTEBOOK_NAMESPACE']\n", "DEPLOYMENT_NAME = model_name\n", "MODEL_NAME = model_name\n", "if DOMAIN_NAME == \"svc.cluster.local\":\n", " PROTO = \"http\"\n", " SVC = f'{DEPLOYMENT_NAME}.{NAMESPACE}.{DOMAIN_NAME}'\n", " VERIFY = False\n", "else:\n", " PROTO = \"https\"\n", " SVC = f'{DEPLOYMENT_NAME}.{NAMESPACE}.serving.{DOMAIN_NAME}'\n", " VERIFY = os.environ['EZUA_DOMAIN_CA_CERT_PATH']\n", "URL = f\"{PROTO}://{SVC}/v2/models/{MODEL_NAME}/infer\"\n", "\n", "print(URL)\n", "\n", "names = ['season', 'year', 'month', 'hour_of_day', 'is_holiday', 'weekday', 'is_workingday', \n", " 'weather_situation', 'temperature', 'feels_like_temperature', 'humidity', 'windspeed']\n", "\n", "input_data = [\n", " [1, 2, 1, 0, 0, 6, 0, 1, 0.24, 0.2879, 0.81, 0.0000],\n", " [1, 5, 1, 0, 0, 6, 1, 1, 0.24, 0.2879, 0.81, 0.0000]\n", "]\n", "\n", "inputs = {\n", " \"inputs\": [\n", " {\n", " \"name\": \"ndarray\",\n", " \"shape\": [2, 12],\n", " \"datatype\": \"FP32\",\n", " \"data\": input_data\n", " }\n", " ]\n", "}\n", "\n", "%update_token\n", "\n", "headers = {\"Authorization\": f\"Bearer {os.environ['AUTH_TOKEN']}\"}\n", "\n", "response = requests.post(URL, json=inputs, headers=headers, verify=VERIFY)\n", "\n", "print(response.reason)\n", "\n", "output = response.json()['outputs'][0]['data']\n", "\n", "print(\"Rendted Bikes Per Hours:\\n\")\n", "for item, out in zip(input_data, output):\n", " input_dict = dict(zip(names,item))\n", " print(f\"Input Data: {input_dict} \\n\\nBike Per Hour: {out}\\n\")" ] }, { "cell_type": "markdown", "metadata": { "collapsed": false, "jupyter": { "outputs_hidden": false } }, "source": [ "### Accessing KServe API from an External Network\n", "\n", "#### Initializing Configuration Variables\n", "```bash\n", "DOMAIN=\n", "USERNAME=\n", "PASSWORD=\n", "KC_ADDR=\n", "```\n", "\n", "\n", "#### Acquiring Access and Refresh Tokens for Authentication\n", "\n", "```bash\n", "response_json=$(curl -k --data \"username=$USERNAME&password=$PASSWORD&grant_type=password&client_id=ua-grant\" \"https://$KC_ADDR/realms/UA/protocol/openid-connect/token\")\n", "\n", "REFRESH_TOKEN=$(echo \"$response_json\" | jq -r '.refresh_token')\n", "ACCESS_TOKEN=$(echo \"$response_json\" | jq -r '.access_token')\n", "```\n", "\n", "#### Configuring Variables for Model Inference\n", "\n", "```bash\n", "SVC=\n", "MODEL_NAME=\n", "NAMESPACE=\n", "```\n", "\n", "#### Executing Inference Request (Example: Bike Sharing Data)\n", "\n", "```bash\n", "curl --request POST \"https://$SVC.$NAMESPACE.$DOMAIN/v2/models/$MODEL_NAME/infer\" -k -H \"Authorization: Bearer $ACCESS_TOKEN\" --header 'Content-Type: application/json' --data-raw '{\"inputs\":[{\"name\":\"ndarray\",\"datatype\":\"FP32\",\"shape\":[2,12],\"data\":[[1, 2, 1, 0, 0, 6, 0, 1, 0.24, 0.2879, 0.81, 0.0000],[1, 5, 1, 0, 0, 6, 1, 1, 0.24, 0.2879, 0.81, 0.0000]]}]}'\n", "```\n", "\n", "#### Renewing Access Token Using Refresh Token (When Necessary)\n", "\n", "```bash\n", "response_json=$(curl -k --data \"grant_type=refresh_token&client_id=ua-grant&refresh_token=$REFRESH_TOKEN\" \"https://$KC_ADDR/realms/UA/protocol/openid-connect/token\")\n", "ACCESS_TOKEN=$(echo \"$response_json\" | jq -r '.access_token')\n", "```" ] } ], "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": "3.8.5" } }, "nbformat": 4, "nbformat_minor": 4 }