feat: implement `config`, `registry`, and `pipeline` - Githubissues

OmicsML / dance

DANCE: a deep learning library and benchmark platform for single-cell analysis

https://pydance.readthedocs.io

BSD 2-Clause "Simplified" License

348 stars 34 forks source link

feat: implement `config`, `registry`, and `pipeline` #390

Closed RemyLau closed 8 months ago

RemyLau commented 8 months ago

This PR contributes toward AutoML capabilities with pipeline tuning and hyperparameter tuning for each function/transformation.

The following modules are implemented:

pipeline object for handling complex (multi-level) transformation/data/method pipelines.
config for handling pipeline configs.
registry for registering functions/transformations to be used in pipeline tuning.

Future TODO notes:

[x] Register transforms (preprocessors) and make sure they are registered correctly; make necessary reorganizations (#393)
[x] Pipeline tuning (#397)