Torch-ecg

----------------------------------------------------------------------------------------

Added
~~~~~

- Add metrics computation class ``WaveDelineationMetrics`` for evaluating the
performance of ECG wave delineation models.
- Add methods for computing the metrics to the output classes (in the module
``torch_ecg.components.outputs``).
- Add script ``push2pypi.sh`` for pushing the package to PyPI.
- Add attribute ``global_pool_size`` to the configuration of the classification
models (``torch_ecg.models.ECG_CRNN``).

Changed
~~~~~~~

- ``flake8`` check ignore list is updated.
- ``README`` is updated.

Removed
~~~~~~~

- Usage of ``easydict`` is removed. Now we use ``torch_ecg.cfg.CFG`` for
configuration.

Fixed
~~~~~

- Computation of the metric of `mean_error` for ECG wave delineation is corrected.
- Fix bugs in ``SpaceToDepth`` layer (``torch_ecg.models.resnet``).

----------------------------------------------------------------------------------------

Changed
~~~~~~~

- Some out-of-date ``sample-data`` files are updated, unnecessary files
are removed.
- Passing a path that does not exist to a database reader class now raises
no error, but a warning is issued instead.
- Include ``isort`` and ``flake8`` in the code formatting and linting steps.
Code are reformatted and linted.

----------------------------------------------------------------------------------------

Changed
~~~~~~~

- Docstrings are cleaned up.
- Unit tests are updated.

----------------------------------------------------------------------------------------

Added
~~~~~

- Add ``BibLookup`` class for looking up BibTeX entries from DOIs
of papers related to datasets and models.
- Add ``RPeaksDetectionMetrics`` class to the ``torch_ecg.components.metrics``
module for evaluating the performance of R-peaks detection models.
- Add CI workflow for running tests via GitHub Actions.

Changed
~~~~~~~

- The loading methods (``load_data``, ``load_ann``, etc.) of the database
reader classes are enhanced to accept ``int`` type record name argument
(``rec``), which redirects to the record with the corresponding index
in the ``all_records`` attribute of the database reader class.

----------------------------------------------------------------------------------------

Added
~~~~~

- Add decorator ``add_docstring`` for adding/modifying docstrings of functions
and classes.
- Add method ``append`` for the ``BaseOutput`` class.
- Add several metrics computation functions in ``torch_ecg/utils/utils_metrics.py``:
- ``confusion_matrix``
- ``ovr_confusion_matrix``
- ``auc``
- ``accuracy``
- ``f_measure``
- ``QRS_score``
- Add top-level module ``torch_ecg.components``.
- Add classes for metrics computation to the ``torch_ecg.components.metrics`` module.

Changed
~~~~~~~

- ``Dataset`` classes and corresponding config classes are added to the
``__init__.py`` file of the ``torch_ecg.databases.dataset`` module
so that they can be imported directly from the module.
- Logger classes, output classes, and trainer classes are moved to the new
module ``torch_ecg.components``.
- Callbacks in ``BaseTrainer`` are enhanced, allowing empty monitor, and allowing
non-positive number of checkpoints to be saved (i.e., no checkpoint is saved).

----------------------------------------------------------------------------------------

Fixed
~~~~~

- Bugs in extracting compressed files in the ``http_get`` function
of the ``utils.download`` module.

Security
~~~~~~~~