Tags

1.3.0

- Smarter Container Building: The system now creates a unique fingerprint for Docker containers based on their contents and reuses existing containers when nothing has changed, avoiding unnecessary rebuilds and saving time.

- Enhanced Notebook Execution: Instead of having separate "train" and "score" jobs, there's now a unified system that automatically detects which Jupyter notebooks have been modified and runs them. Users can also manually specify which notebook to run using special commands in their commit messages like [run my_notebook.ipynb] or skip execution entirely with [skip run].

- Improved Release Process: The release system now only creates releases for production tags (excluding development, release candidate, alpha, and beta versions) and includes an automatic release feature that triggers when production containers are built.

- Cleaner Code Structure: The configuration has been reorganized with better variable names (like changing "TRAIN_RUNNER" to "COMMIT_RUNNER") and improved error handling throughout the pipeline.

dev-1.3.0

1.2.2

Fix gpu training loggins

1.2.1

Cleanup papermill job logging

1.2.0

Promote to production via tags

1.1.0

Ability to use multiple containers and have a separate production/deployment branch

1.0.3

Release 1.0.3 of components repository gitlab-data/ds-component-pipeline

1.0.2

Update README; add parameters

1.0.1

1.0.0