Overview

The examples folder contains a complete machine-learning pipeline built with cinnamon. It is a good starting point to see how components and configurations work together in a real project.

The pipeline covers:

• A DataLoader that downloads and parses the IMDB dataset.

• Two Processor components — one for tf-idf features, one for label encoding.

• An SVCModel wrapping scikit-learn’s SVC classifier.

• An SVCBenchmark that wires everything together and evaluates the model.

All components are registered in the Registry and can be swapped independently.

Project layout

examples/
    components/
        data_loader.py      ← IMDBLoader
        processor.py        ← TfIdfProcessor, LabelProcessor
        model.py            ← SVCModel
        benchmark.py        ← SVCBenchmark
    configurations/
        data_loader.py      ← IMDBLoaderConfig
        processor.py        ← TfIdfProcessorConfig, LabelProcessor registration
        model.py            ← SVCModelConfig
        benchmark.py        ← SVCBenchmarkConfig
    demos/
        demo_data_loader.py ← run IMDBLoader standalone
        demo_benchmark.py   ← run the full pipeline

Installation

Install cinnamon and the example dependencies:

pip install "cinnamon-core[examples]"

Running the demos

Run the full benchmark pipeline, from the repository root:

python -m examples.demos.demo_benchmark

Run only the data loader:

python -m examples.demos.demo_data_loader

Run them as modules, from the root rather than from inside examples. The components are bound by their full import path, examples.components.benchmark.SVCBenchmark, so the directory that has to be importable is the one holding examples. Running python demos/demo_benchmark.py from inside examples fails with ModuleNotFoundError: No module named 'examples' at the moment the first component is built.

Both scripts call Registry.build() to discover and register all configurations under the configurations/ folder before building the components.

Using these components in your own project

You can reuse the example components and configurations from your own project by pointing Registry.build() to the examples folder as an external directory:

from pathlib import Path
from cinnamon.registry import Registry

Registry.build(
    directory=Path('.'),                           # your project
    external_directories=[Path('path/to/examples')]
)

Once built, you can register configurations that reference the example keys:

from cinnamon.configuration import Configuration, Param
from cinnamon.registry import RegistrationKey, register_method

class CustomBenchmarkConfig(Configuration):
    data_loader: RegistrationKey = Param(
        RegistrationKey(name='data_loader', tags={'imdb'}, namespace='examples')
    )
    model: RegistrationKey = Param(
        RegistrationKey(name='model', tags={'svc'}, namespace='examples')
    )

    @classmethod
    @register_method(name='benchmark', tags={'custom'}, namespace='my_project',
                     component='components.CustomBenchmark', run_method='run')
    def default(cls) -> 'CustomBenchmarkConfig':
        return super().default()

Note

Make sure to install the example dependencies before importing example components, otherwise the import will fail during Registry.build().