### Added

* Add `attach_handler` method to `PhlowerTrainer` to add an extra handler at training process.
* Add `lazy_load` parameter to `TrainingSetting` to load data lazily.
* When `time_series_length` is -1, `PhlowerGroupModule` determines the time series length automatically from the input data.
* Add `NaNStoppingHandler` to stop training when loss becomes NaN.
* Add sliding window method for time series data in training process.
* Add distributed data parallel (DDP) training.
* Add `evaluate_context_manager` parameter to `TrainerSetting` to choose context manager during evaluation.
* Add `LayerNorm` module in `phlower.nn`
* Add `aggregation_method` parameter to `LossCalculator` to choose aggregation method of losses. (sum or mean)
* Add `cg` mode in iteration solver of `PhlowerGroupModule`.
* Add `PhlowerPresetGroupModule` as a preset group.

### Changed
* Time series tensor is splitted into each time step when forwarding with `time_series_length` in `PhlowerGroupModule`.
* Display details of losses at training process.
* Default inference mode is changed to `torch.inference_mode` from `torch.no_grad`.

### Fixed
* Fix `restart` method not to load recursively previous restarted checkpoints.
* Fix to set empty tcp port for DDP

Fixed


Fix PhlowerTensor with physical dimension to handle torch.stack.


* Changed
  * Decompose input members when to apply reverse transform after prediction
  * Change default value of bias parameter in setting class of GCN. (False -> True)
  * Change default value of bias parameter in setting class of coefficient_network in IsoGCN. (False -> True)

* Fixed

  * Fix PhlowerTensor with physical dimension to handle torch.stack.

Merge branch 'develop' into 'main'

Develop update to main

See merge request ricos/machine_learning/phlower!113

0.2.0 Release


Update
* Add ML layers (DeepSets, Dirichlet, ..)
* Improve interface to define ML model

Release initial features


* Add PhlowerTensor which handles physical dimension
* High Level API for scaling, training and prediction
* Yaml file to define model structure

Register PyPI