[💡SUG] Automatic dataset and dataloader save path

[damicoedoardo]

For my understanding there are two different level of dataset preprocessing,
Level 1 filtering and preprocessing operation: you can save a dataset after this step calling dataset.save()
Level 2 dataset train/(val)/test split: you can save the different dataloaders with save_split_dataloaders()

In the first case you can specify the dir where you want to save the dataset
In the second case you can not specify the dir where you want to save, instead the dataloaders are saved under config['checkpoint_dir']

Is it possible to unify the behaviour of the two saving processes?

I would propose to use the data_path config parameter to store the three different level of preprocessed data that can be stored under three different dir:

1. f'{data_path}/atomic': all the atomic files which are now under dataset
2. f'{data_path}/preprocessed': All the dataset which have been preprocessed (level 1 mentioned above)
   e) f'{data_path}/{model_name}/splitted': All the dataloaders for an algorithm or category of algorithms (level 2 mentioned above)

I see the proposed structure more intuitive and easy to use.

Thanks

[2017pxy]

@damicoedoardo Thanks for your advice, we will consider it!