Prune the tree: move general functions to base class NNCV

[luigibonati]

Description

Moved fit function to NeuralNetworkCV class. In addition, added default functions for prepare_dataset, evaluate_dataset and train_epoch

List of changes

• [X] Moved fit in NNCV
• [X] Added prepare_dataset function (both for base class and children where needed)
• [X] Move training parameters (e.g. loss_type for tica) outside fit
• [X] add custom_train_epoch to change the workflow
• [x] add tests for custom_train
• [x] add custom_eval function
• [x] change log info (e.g. train loss) to dictionary
• [x] changed .to(device) into set_device

[luigibonati]

• I did some refactoring here, moving a lot of stuff from cv classes to base NNCV (@EnricoTrizio see for instance the new TDA file).
• The only breaking change is for DeepTICA, related to the parameters of the loss function (loss_type and n_eig) that are removed from the fit function and moved to a specific set_loss_function method
• @EnricoTrizio I also changed how the training log is printed, so in principle if you want to print something else (e.g. contribution to tda loss) you could simply add new keys and values to the self.logs dictionary --> an example is in DeepTICA where I add a new key in the constructor and set it in the evaluate_dataset function

[lgtm-com[bot]]

This pull request fixes 1 alert when merging 82202c5fbc5081231a35a2319300e61cc7e9a96c into adf73aff972d89af47632d0adf2f3739ae96e36b - view on LGTM.com

fixed alerts:

• 1 for Unused import

Heads-up: LGTM.com's PR analysis will be disabled on the 5th of December, and LGTM.com will be shut down ⏻ completely on the 16th of December 2022. Please enable GitHub code scanning, which uses the same CodeQL engine :gear: that powers LGTM.com. For more information, please check out our post on the GitHub blog.