Overview#

Welcome to theror project, which proposes a more customizable and flexible way to createpipelines programmatically in Python. You can define simple forward propagating pipelines where theinput of the first stage is computed on and then passed to the next stage until you reach a terminalstage. Which makes it easier to have a good separation of concern when constructing complex inferencelogic in ML projects.

Key Features#

The current version of ror isv0.1.0 and supports the following features:

• 🏗Forward Pipelines: Supports simple forward passing pipelines with a separation of concern for each pipeline stage.

• 🚀Robust ML Pipelines Faster: Parallelize the work of an entire pipeline by developing stages independently which can be linked at the end.

• 📚Readable ML Code: Move away from massive ML Python files to using logically constructed code in the form of stages.

Benefits#

• High-Readability

• Better Maintainability

• Enabling Better Collaboration

Future Additions#

• Conditional Stages: Based on the computation of some stage, the output can be directed to different sub-pipelines.

• Caching InitStage: Load your model/data once and cache such that it is not loaded for each pipeline inference.

Documentation Overview#