MLOpsPython

Features:

• Multi-stage pipeline and scoring observability #122 @algattik
  • Replaces separate build + release pipelines with a single multi-stage pipeline with stages for Build, Train, ACI, AKS and Webapp deployment.
  • ACI, AKS and Webapp deployment are optionally enabled (through Variables).
  • Added example of custom traces (logging) and correlating traces with requests with Kusto queries.
  • Added example of distributed tracing (client passes a trace header, logged by the scoring service).
• Canary deployment sample #125 @eedorenko
  • Deploys two scoring images to a K8s cluster with Istio service mesh
  • Routes the traffic to the old/new models according to Canary weights
  • Routes the traffic to the old/new models analyzing the request header (A/B testing)

Fixes:

• Enabling continuous deployment for AML model error fix #91 @algattik
• No model associated with the scoring image fix #120 @algattik
• Azure ML Model deploy to ACI error fix #127 @eedorenko

• Generalize model tag finder, tag model experiment_name #103 @dtzar
• Bump all the pip requirements to latest and resolve install errors #104 @dtzar
• Cleanup getting started doc #112 @dtzar
• Add some code comments #114 @dtzar
• Simplify build with Azure CLI, various bug fixes #121 @algattik
  • Use Azure CLI tasks to remove need for variables SP_APP_ID, SP_APP_SECRET, SUBSCRIPTION_ID, TENANT_ID
  • Updated getting_started.md to point out variable path to artifact (closes #117)
  • Made random hyperparameter (ridge regression alpha) a pipeline parameter, setting its value in the DevOps pipeline (closes #107)
  • Changed unit test to test actual training code (partially solves #74)
  • Fixed: attach_compute should set exit code after exception (closes #113)
  • Fixed scoring endpoint HTTP behavior (closes #110)
  • Fixed PipelineParameters format in call to Azure ML Extension (closes #118)
  • Fixed Model build environment for Azure Web App for containers (closes #119)

There have been quite a few fixes and enhancements since the 2.0.0 release and want to summarize the changes here sorted by PR merge date.

Features:

1. Add support for scoring image using Web App for containers #54 @sudivate
2. Add support for Training R model code on ML Compute & Databricks #77 @eedorenko
3. Upgrade build train CI to use new multi-stage pipelines #90 @dtzar
4. Switch CI to use new Azure ML Agentless Task & improve/overhaul model evaluation, registration #101 @dtzar

Updates:

1. Register to use newer sdk method #65 @jpe316
2. Use Azure ML SDK more consistently #67 @tarockey
3. badge to proper AzDO repo & better pipeline filters #81 @dtzar
4. Enable App Insight by default on scoring image #86 @alla15747
5. Newest Conda dependencies for scoring images and project #93 @dtzar
6. Cleaner / Less variables including using AzDO variable template #99 @dariuszparys
7. Tag models with experiment name, filter models in evaluation via experiment name #103 @dtzar

Doc Enhancements:

1. Update to flow getting started #63 @omartin2010
2. Better getting started clarity #82 @Smarker

Fixes:

1. Authentication error fix #51 @eedorenko
2. Doc: Code description spelling fix #59 @tarockey
3. Doc: permissions fix #61 @sudivate
4. Doc: Readme typo fix #62 @anand372
5. Doc: Fix typo #71 @Seddryck