Return data from the backend fit method

[mauicv]

What is this:

this fixes: #813. Given we fit differently in different cases the data returned is different in case.

• SVM(..., optimization='sgd'): Returns dictionary containing converged and n_iter where n_iter is the number of iterations until convergence.
• SVM(..., optimization='bgd'): Returns dictionary containing converged, n_iter and lower_bound where n_iter is as above and lower_bound is the loss achieved during training.
• GMM(..., backend='pytorch'): converged, lower_bound and n_epochs - because the GMM PyTorch backend performs gradient descent over n epochs this value is the number of epochs performed.
• GMM(..., backend='sklearn'): converged, lower_bound and n_iter.

[jklaise]

LGTM but would add the info on what's returned in the public docstrings of fit. Also seems fit_logs isn't used?

Is the plan also to return info from the existing ODs? As a minimum we should open a ticket and investigate what those returns might be.

[codecov]

Codecov Report

Merging #835 (05ed926) into master (2a0c090) will increase coverage by 0.01%.
The diff coverage is 100.00%.

Additional details and impacted files

@@            Coverage Diff             @@
##           master     #835      +/-   ##
==========================================
+ Coverage   81.86%   81.87%   +0.01%     
==========================================
  Files         159      159              
  Lines       10317    10323       +6     
==========================================
+ Hits         8446     8452       +6     
  Misses       1871     1871              

Impacted Files	Coverage Δ
alibi_detect/od/pytorch/gmm.py	93.75% <ø> (ø)
alibi_detect/od/pytorch/svm.py	99.16% <ø> (ø)
alibi_detect/od/_gmm.py	97.43% <100.00%> (ø)
alibi_detect/od/_svm.py	97.50% <100.00%> (ø)
alibi_detect/od/pytorch/base.py	97.43% <100.00%> (+0.21%)	⬆️

[mauicv]

LGTM but would add the info on what's returned in the public docstrings of fit. Also seems fit_logs isn't used?

Is the plan also to return info from the existing ODs? As a minimum we should open a ticket and investigate what those returns might be.

Agree w.r.t. the docstrings, good catch! I'll add that update. I'll open a ticket for the old ODs. They could definitely return something but it's a bit of a bigger job. ticket here: #837