Pydgn

Added
- We can now pass arguments to `pydgn.training.callback.metric.Metric` objects to create arbitrary behavior


Changed
- Added possibility to skip unfinished experiments when calling `pydgn.evaluation.util.retrieve_experiments`. See documentation for more information.

Changed
- Replaced torch.save with atomic_save of PyDGN to handle cases where metrics data storage corrupted the file

Fixed

- Minor fix in `filter_experiments`

Fixed

- Experiment retrieval routines not working as expected when loading checkpoint of a specific configuration rather than the best one.
- Added function `instantiate_data_provider_from_config` that was somehow missing (see tutorials)

Changed

- Dropping tests on MacOS 13, as they are causing too many problems since recent changes.

Added

- You can now store metrics trend across epochs using `Plotter`. Just pass the argument `store_on_disk=True` in the configuration file of the experiment.

Fixed

- TOML project to comply with latest releases of `macos` and `coverage` package
- `pydgn-train` and `pydgn-dataset` not being found in version `1.5.4`