databricks-mlflow-mlops-azureml's Introduction

Databricks-MLflow-MLOps-AzureML

Introduction

Using databricks mlflow model registry with azureml to build out an mlops pipeline

MLflow Model Registry Web-hook for MLOps

MLflow Model Registry is a central repository to store and query models for production use and consumable across different teams. Users can publish and version models, discover new models and model versions, request and approve stage transitions, pull information about models from production systems, and download model artifacts.

Through the web-hook APIs you can automate devops pipelines

Getting Started

Dependencies

Instructions

Create test web-hook

json='{"model_name": "$model_name",    "events": ["TRANSITION_REQUEST_CREATED"],    "description": "Teams notifications",   "status": "TEST_MODE",    "http_url_spec": {      "url": "$webhook_url",      "secret": "anyRandomString"}}'
url='https://$region.azuredatabricks.net/api/2.0/mlflow/registry-webhooks/create'


result=`curl -X POST -H "Accept: application/json" -H "Authorization: Bearer ${token}" --data "$json" $url`
echo result: $result

List web-hook

json='{"model_name": "$model_name"}'
url='https://$region.azuredatabricks.net/api/2.0/mlflow/registry-webhooks/list'
result=`curl -X GET -H "Accept: application/json" -H "Authorization: Bearer ${token}" --data "$json" $url`

echo result: $result

Test web-hook

json='{"id":"$id"}'
url='https://$region.azuredatabricks.net/api/2.0/mlflow/registry-webhooks/test'
result=`curl -X POST -H "Accept: application/json" -H "Authorization: Bearer ${token}" --data "$json" $url`

echo result: $result

Activate the web-hook now


json='{"id":"$id", "status": "ACTIVE"}'
url='https://$region.azuredatabricks.net/api/2.0/mlflow/registry-webhooks/update'
result=`curl -X PATCH -H "Accept: application/json" -H "Authorization: Bearer ${token}" --data "$json" $url`

echo result: $result