cinnamon-generic 0.2.3

[Generic Package] A simple high-level framework for research

Cinnamon Generic

The generic package offers several Component and related Configuration for machine-learning.

Additionally, it provides the first set of commands: high-level APIs for speeding up cinnamon registration.

Documentation

Check the online documentation of cinnamon-generic for more information.

Check also the central cinnamon documentation to overview all cinnamon packages!

Components and Configurations

The generic package defines Component and Configuration for

• Hyper-parameter calibration
• Callbacks
• Data loaders
• Data splitters (e.g., train, validation and test)
• File manager
• Framework helper (e.g., for deterministic runs)
• Metrics
• Model API
• Pipeline API (i.e., a special nested Component that executes children sequentially)
• Data processors
• Model validation routines: train and test, cross-validation

Commands

Commands are syntactic sugar for several cinnamon operations.

For instance, the setup_registry() performs (i) custom module loading; (ii) registration DAG evaluation and (iii) registration. Thus, it is just necessary to invoke this command at the beginning of each runnable script to use cinnamon.

Other commands are more specific to the generic package:

• Running a generic Component.
• Running a Routine component in training and inference modes.
• Running multiple Component in sequential fashion.
• Running multiple Routine components in sequential fashion.
• Running a Calibrator component.

Install

pip

  pip install cinnamon-generic

git

  git clone https://github.com/federicoruggeri/cinnamon_generic

Contribute

Want to contribute with new Component and Configuration?

Feel free to submit a merge request!

Cinnamon is meant to be a community project :)

Contact

Don't hesitate to contact:

• Federico Ruggeri @ [e-mail]

for questions/doubts/issues!