Implement a base model for LFR

[fernandoGubiMarques]

Describe your changes

This branch contains the implementation of a base model for the Learning from Random Projectors pretext task for self-supervised learning.

This model requires two torch.nn.ModuleList. For convenience, a RepeatedModuleList module is also implemented to instantiate multiple modules of the same class.

Additionally, it contains an MLP class (unrelated to LFR) to facilitate the creation of future simple prediction and projection heads.

Issue ticket number and link (If apply)

Checklist before requesting a review

• I have performed a self-review of my code.
• I have added tests to this funtionality.
• All tests are passing.
• I have updated the documentation as needed.
• I have followed the project's coding guidelines.

[GabrielBG0]

Some changes are needed before merging.

You should also relocate your nets to the specific use case it is intended to have, e.g. image, time series etc...
Take a look in the new model organization.

[GabrielBG0]

Define types for __init__ parameters

[GabrielBG0]

@fernandoGubiMarques Forgot to include in the review, but if you could add the new models to the models/README.md would be nice too

[fernandoGubiMarques]

Neither LFR or MLP have a specific use, so I'm not sure where you want me to put them. LearningFromRandomnessModel is similar in nature with SimpleSupervisedModel, which is also outside any submodules. At any rate, the typing mismatches have been resolved.

[GabrielBG0]

LGTM

[GabrielBG0]

Code looks good but i would like @otavioon 's opinion on where to put the models

[otavioon]

Please move minerva/models/nets/lfr.py to minerva/models/ssl/lfr.py.

[otavioon]

LGTM