Skip to content

Latest commit

 

History

History
55 lines (29 loc) · 1.98 KB

37-mlops.md

File metadata and controls

55 lines (29 loc) · 1.98 KB
Raw

HOWTO - Quickstart: MLOps with Azure Machine Learning and Databricks

Prerequisites

HOWTO - install AzureML SDK v1+v2 and ESML accelerator library

HOWTO - Supported use cases & Accelerated use cases

Context and pipeline outputs: MLOps in relation to DataOps

1) Configure & Run ESML template Notebooks

  1. Configure the lake_settings
  1. Run the 3 notebooks of your choice to genereate the Azure ML Pipelines:
  • Example: If you want to work with Databricks and pyspark, for batch deployment: 1 + 2b + 3b
  • Example: If you want to use Azure ML compute and AutoML, for online deployment AND/OR batch deployment: 1 + 2a + 3b and/or 3c

Output:

  • 2 Azure Machine Learning pipelines for: Training and Inference
  • 1 Online endpoint

2) CI/CD (Python): Configure the ESML MLOps template, to use your project, model, pipelines

  1. Configure the inline python parameters in the file 21-train_in_2_gold_train_pipeline.py

Parameters to configure in 21-train_in_2_gold_train_pipeline.py at line 49, 50

advanced_mode = False # ADVANCED MODE (DatabricksSteps also) + Manual ML (or AutoML if defined in Databricks notebook)
use_automl = True # SIMPLE MODE + AutoMLStep (if True)Manual ML Step (if False)

3) Import & Configre the GHA/ADO ESML Azure Devops pipeline

  1. Import Pipeline from from template - Template location
  • Point the pipeline to your project and models branch, such as "project001_M11_dev_branch"
  1. Configure the Variables in Azure Devops / Github Actions