Various Improvements - Githubissues

Main changes:

evaluate.py and evaluate_all.py ready to evaluate respectively one model and all the models (see #22)
create_job_sub.py:
- now proposes submission file to evaluate and predict;
- results get saved in the same folder (see #16);
evaluate_all.pbs to evaluate all the models present (see #17)
discretization.py: improvement of the first discretization ; proposition of a new absolute discretization (see #24)
tests have been developed for various aspects of the project
training pipeline improved (see #25)
models.py: a new model, ProtNet has been proposed
correction and improvement generally:
- in ExamplesIterator: now it does not return empty batchs (the length returned was wrong)
- little correction in ModelInspector: only show subfolder of results that are present
some more settings and comment

jjerphan / CS5242Project