fix: Adding a flag of whether to run Importance Sampling in RESOLVI Differential Experssion

[ori-kron-wis]

No description provided.

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 75.52%. Comparing base (14daf47) to head (61e56c8).

❗ There is a different number of reports uploaded between BASE (14daf47) and HEAD (61e56c8). Click for more details.

HEAD has 47 uploads less than BASE

Flag	BASE (14daf47)	HEAD (61e56c8)
	50	3

Additional details and impacted files

@@             Coverage Diff             @@
##             main    #3679       +/-   ##
===========================================
- Coverage   86.62%   75.52%   -11.10%     
===========================================
  Files         226      226               
  Lines       21859    21859               
===========================================
- Hits        18936    16510     -2426     
- Misses       2923     5349     +2426     

Files with missing lines	Coverage Δ
src/scvi/external/resolvi/_model.py	90.24% <ø> (-3.05%)	⬇️
src/scvi/external/resolvi/_utils.py	69.23% <ø> (-17.16%)	⬇️
src/scvi/train/__init__.py	94.44% <ø> (ø)

... and 59 files with indirect coverage changes

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.

[canergen]

There is already a flag: weights. It's just the value is ignored. Otherwise, thanks for proposing the fix :).

[ori-kron-wis]

There is already a flag: weights. It's just that the value is ignored. Otherwise, thanks for proposing the fix :).

No No. weights, is still used and is doing a weighted choice of indexes at the very end.

It says it's supposed to not run IS if selected as "uniform", but IS will be run regardless. This is what you meant.

Up to you if to approve and merge this PR, or suggest a better fix, as I think the original intention was different from what is implemented in get_normalized_expression_importance right now.

In any case the flag of IS + num of samples / Monte Carlo steps should be open for user, as it can really take long time for them.