Fixes error in get losses functions

[canergen]

• Closes #xxxx (Replace xxxx with the GitHub issue number)
• Tests added and passed if fixing a bug or adding a new feature
• All code checks passed
• Added type annotations to new arguments/methods/functions
• Added an entry in the latest docs/release_notes/index.md file if fixing a bug or adding a new feature
• If the changes are patches for a version, I have added the on-merge: backport to x.x.x label

[codecov]

Codecov Report

Attention: 3 lines in your changes are missing coverage. Please review.

Comparison is base (2c63d44) 89.42% compared to head (37002f9) 88.90%.
Report is 1 commits behind head on main.

Files	Patch %	Lines
scvi/model/base/_log_likelihood.py	88.88%	3 Missing ⚠️

Additional details and impacted files

@@            Coverage Diff             @@
##             main    #2362      +/-   ##
==========================================
- Coverage   89.42%   88.90%   -0.52%     
==========================================
  Files         153      153              
  Lines       12564    12582      +18     
==========================================
- Hits        11235    11186      -49     
- Misses       1329     1396      +67     

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

[martinkim0]

Wait so does this mean that this method was not working properly before? Since the default is n_mc_samples_per_pass=1

[canergen]

Yes but we only use it for DE genes where it's called with n_samples_per_mc of 100 (?). ScANVI wasn't supporting importance weighting in DE beforehand.
ScANVI doesn't work with multiple samples (major work to fix this) and this might be an issue for multi-GPU support (see bug report). If you want to fix this, I have a fix for broadcast labels but then dropped it as also the encoder doesn't support it.

[martinkim0]

Probably best to also use torch operations here since compute_reconstruction_error doesn't use numpy. Let me know what you think

[canergen]

Fine with me. Can you puh those changes?

[martinkim0]

I'll get this in by today @canergen

[martinkim0]

#2429 needs to be addressed before this