eonu / sequentia / a71b38e4-620d-4b65-a7f9-45b107a52cb8

83%

master: 83%

Pull #242

circleci

refactor: full `scikit-learn` compatibility + general refactor (#241)

- Use `scikit-learn` ≥1.4 metadata routing for full API compatibility
- Switch from `setup.py` to `pyproject.toml`
- Switch to `poetry` for dependency management
- Use `tox` for documentation building, unit tests and linting
- Add pre-commit hooks
- Add coverage checks via `coveralls`
- Drop support for Python <3.11
- Support Python 3.11 and 3.12
- Add issue and pull request templates
- Add license information to `.py` files
- Update repository documentation
- Use `cliff` for `CHANGELOG.md` generation
- Update Sphinx documenation
- Switch to `numpy` docstring format (via `napoleon`)
- Move library code from `lib/` to `sequentia/`
- Move unit tests from `lib/test` to `tests/`
- Add `sequentia.enums` module for configuration values.
- Add `sequentia.version` sub-package for versioning information.
- Add `sequentia._internal` sub-package for internal logic.
- Remove `sequentia.pipeline` module.
- Make topologies internal
- Remove `pydantic` validation classes for model inputs
- Remove `pydantic` validation for `numpy` arrays
- Use key-word only arguments more widely (e.g. for `lengths`)
- Use `pydantic.validate_call` for validating constructors and other functions
- Use type hinting
- Use `ABCMeta` metaclass for base classes (e.g. `ClassifierMixin`/`RegressorMixin` and `BaseHMM`)
- Use `abc` decorators for abstract methods
- Avoid unnecessarily redefining methods on subclasses for only documentation purposes
- Add `predict_log_proba` to classifiers
- Move majority of validation logic to `sequentia._internal._validation` module
- Prefix all HMM parameter setter functions with `set_state`
- Rename `set_transitions` to set_state_transition_probs`
- Rename `set_state_emissions` to `set_state_emission_probs`
- Rename `set_state_covariances` to `set_state_covars`
- Use static methods to define default HMM parameters
- Handle KNN querying for single traini... (continued)

1168 of 1195 new or added lines in 22 files covered. (97.74%)

2187 of 2413 relevant lines covered (90.63%)

1.73 hits per line