Releases: microsoft/MLOpsPython
Releases · microsoft/MLOpsPython
MLOps with Azure ML
Enable deploy of model registered by previous build (#207) Allows you to skip the first two stages of the pipeline and deploy a model registered by a previous build by setting the variable MODEL_BUILD_ID to the previous build's id when queuing the pipeline. This speeds up the process of fixing issues with the scoring or deployment configuration.
MLOps with Azure ML
123544 fix hard coded subscription reference (#205)
MLOps with Azure ML
123536 Remove 60 minute job timeout #203
MLOps with Azure ML
123095 Use correct parameter names in web app deploy (#202)
MLOps with Azure ML
123023 Clarify that web app must exist for the deploy to work (#201)
MLOps with Azure ML
Add basic descriptions to YAML pipelines (#200) * Add basic descriptions to YAML pipelines * Updated comments to address review feedback * Add missing feedback fix
MLOps with Azure ML
fix model input features (#198) Co-authored-by: Eugene Fedorenko <[email protected]>
MLOps with Azure ML
Getting started refactoring (#197) * State explicitly the flow order (Template as it is with Diabetes-> Bootstrap with your project code) * Fix “multistage pipeline structure” explanation * Clarify the R approach. It has only the model training. There is no evaluation/registration * Fix the confusion with AzureResourceConnection and WORKSPACE_SVC_CONNECTION * Explain using of a Docker image in the pipeline * Link to bring-your-own-code (Bryan’s) article * Fix broken links (e.g. diabetes_regression-ci-build-train.yml) * Provide ML Service connection screenshot * Explain explicit "diabetes" names. (e.g. The repo contains a sample “diabetes regression” project so here and there all names contain "diabetes") * Clarify the folder structure (Common folders (e.g. .pipelines, ml_service) vs Project folders (e.g. diabetes_regression))
MLOps with Azure ML
122630 run unit tests in any case (#199)
3.0.0
Microsoft ❤ MLOps release
There have been 33 PRs to master since the 2.2.0 release (sorry, we're not keeping up with semver). At this point there are several major changes to the repository since then which make it hard to simply upgrade and plenty of significant improvements to warrant 3.0. A big push was made to make it a better experience to bring your own ML project code based on usage from customers using this template for their production implementations.
Features:
- Add Build UI link as a tag so Azure ML pipeline runs can link back to Azure DevOps build which triggered the run #132 @jotaylo
- Added experimentation folder and Juypter notebooks for diabetes ridge regression model #145 @bjcmit
- Change folder structure to support multiple models and intellisense (code --> diabetes_regression) #147
@sbaidachni - Discover scoring model & add swagger endpoint to scoring container #149 @algattik
- Add toggle for evaluation step #152 @sudivate
- Always run Azure DevOps pipeline evaluation step to avoid confusion with cancellation #154 @jotaylo
- Add .amlignore example of usage #155 @sbaidachni
- Remove sys.path imports everywhere to support IntelliSense #162 @sbaidachni
- Add code coverage #150 @algattik
- Consistent manage environments in conda YML files #158, #168 @algattik
- Remove the pkl extension requirement for model name #164 @sbaidachni
- Support config file for code reference, switch alpha parameter from pipeline to config #169 @sbaidachni
- Fail when there is an error with R code #173 @algattik
- Add bootstrap code to more easily bring your own code #186 @sudivate
- Bootstrap code renames all instances of "diabetes" #189 @dtzar
- Bootstrap supports mac and linux machines #190 @eedorenko
- Modified to use dataset and pipelinedata #193 @GraemeMalcolm
Fixes:
- Doc: Clarifications in getting started #131 @jotaylo
- Build_uri assignment bug #133 @starlord-daniel
- extending use of SOURCES_DIR_TRAIN variable #143 & fix #146 @jotaylo
- pipeline with r name reference #167 @algattik
- sklearn joblib warning #174 @algattik
- Doc: link to point to not remove yml files #176 @timpoth
- ACI deployment #178 @sudivate
- Doc: links & typos #175 @eedorenko
- Add required azureml-dataprep package to deps #182 @eedorenko
- Rearrange & fix unit tests #184 @bjcmit
- Doc: clarifying comments #187 @eedorenko