AutoMLOps 1.2.3

[1.2.3] - 2023-09-29

Added

• Support for Github Actions integration
• Added workload identity federation integration

Changed

• Updated documentation
• .git and relevant files are now placed inside the AutoMLOps/ base directory

Fixed

• Set default cloud functions registry to the artifact-registry that is created as part of the provision operation

AutoMLOps 1.2.2

[1.2.2] - 2023-09-25

Added

Changed

• Commented out change from 1.2.1 to set cloud functions default registry to artifact-registry; there appears to be an auth bug with this setting
• Updated documentation

Fixed

• Removed legacy ml.googleapis.com API that is deprecated as of 9/25/2023

AutoMLOps 1.2.1

[1.2.1] - 2023-09-20

Added

• Support for Vertex AI experiments
• Added in an include statement for the build trigger (only runs on changes to AutoMLOps/**)

Changed

• Set default cloud functions registry to artifact-registry
• Changed recommended roles and permissions for deploying with use_ci=False.

Fixed

• Fixed issues with use_ci=False provision directory not being created properly
• Fixed issues with use_ci=False terraform config

AutoMLOps 1.2.0

[1.2.0] - 2023-09-19

Added

Major version updates:

• Code is now broken down into 5 main operations: generate, provision, deprovision, deploy, and launchAll
• Uses jinja2 templates for storing code templates and writing to files
• Additional package dependencies using Jinja2 templates and deploying with precheck function
• Provides additional warnings and checks for giving IAM permissions visibility to the user
• Creates a .gitignore by default
• Support for cloud-functions in addition to cloud-run for the submission service
• Added 2 new generated folders: provision/ and services/
• Added naming_prefix parameter to allow for differentiating between AutoMLOps pipelines in the same project
• Significant generalization in terms of tooling (now allows for specifying provisioning_frameworks, deployment_frameworks, etc.)
• Renamed backend src code folder to google_cloud_automlops to avoid naming confusion
• Added enum and config files, which is different than previous approach of class inheritance

Changed

• Updated README.md and documentation
• Templatized code has now been pulled out and placed into jinja2 templates.
• Gitops workflow placed into separate folder and file, will only version AutoMLOps/ directory instead of the whole cwd.
• Reworked deployment workflow and build configuration (submission service and cloud scheduler are now created as part of the provision step).
• Update notebook examples
• Changed wording and branding descriptions
• Significant updates to unit tests

Fixed

• Bugs related to provisioning with terraform

AutoMLOps 1.1.4

[1.1.4] - 2023-07-25

Added

• Writes .gitkeep to scripts/pipeline_spec directory by default
• Generates a readme.md into generated AutoMLOps codebase now by default

Changed

• Two newlines after functions (linting)
• Parameter mapping (list -> JsonArray, map -> JsonObject)
• Updated documentation: added examples section into main readme, changed package version deps in examples notebooks

Fixed

• Migration issues with pyyaml 5.4.1 since release of cython>3.0; Fixed by updating pyyaml version to 6.0.1.

AutoMLOps 1.1.3

[1.1.3] - 2023-07-07

Added

• Added a BQML retail notebook example

Changed

• Updated the git workflow to check that the remote is pointing to the correct project id.

Fixed

• Pinned all kfp versions to kfp<2.0.0. to address the recent migration to kfp2+.

AutoMLOps 1.1.2

Added

• Added in optional parameter for specifying a base_image.
• Created an example notebook that walks the user through a transfer learning example using a GPU.
• Added in a clear_cache function which deletes all files within the tmpfiles directory.

Changed

• Updated readme and implementation guide.
• Long-term change: .tmpfiles/ subdirectory to .AutoMLOps-cache/
• Long-term change: removed use_kfp_spec from parameter lists and switched to determining this at run-time.

Fixed

• Removed redundant code for cloudbuild config generation.
• Updated constants.py file to remove constants no longer being used.
• Fixed custom imports mismatch between kfp spec and custom automlops spec.
• Verified custom_training_jobs_specs works as intended.

AutoMLOps 1.1.1

Added

• Refactored backend modules to be structured based on frameworks (e.g. kfp, tfx, etc.) and deployments (e.g. cloudbuild, github actions, etc.)
• Added some unit tests for the utils.py module.

Changed

• Moved unit tests to /tests directory.

AutoMLOps 1.1.0

Added

• New interface for defining AutoMLOps components and Pipelines. Removed the need to call Jupyter cell decorators and replaced them with Python function based decorators instead.
• New feature updates allow for running AutoMLOps outside of Jupyter Notebooks.
• Examples for running AutoMLOps outside of a notebook, as well as an example inferencing pipeline.
• Faster build jobs

Changed

• Updated readme and implementation guide.
• Better logging
• Better handling of requirements.txt (remove dups, infer from pipreqs, option to input explicit versions, sorted order)
• Better versioning of models

Fixed

• Bug that can change current working directory on failure with run_local=True.

AutoMLOps 1.0.5

Official release on PyPI.

Changed

• Cleaned up wheel and egg files from repo.
• Removed dist/ and build/ directories.