Feature/moscot not

[MUCDK]

[codecov[bot]]

Codecov Report

Merging #567 (2db4d93) into main (cb53435) will increase coverage by 0.33%. Report is 2 commits behind head on main. The diff coverage is 79.31%.

:exclamation: Current head 2db4d93 differs from pull request most recent head 4566eb9. Consider uploading reports for the commit 4566eb9 to get more accurate results

Additional details and impacted files

@@            Coverage Diff             @@
##             main     #567      +/-   ##
==========================================
+ Coverage   77.45%   77.78%   +0.33%     
==========================================
  Files          36       41       +5     
  Lines        3589     4840    +1251     
  Branches      633      824     +191     
==========================================
+ Hits         2780     3765     +985     
- Misses        542      732     +190     
- Partials      267      343      +76     

Files	Coverage Δ
src/moscot/backends/ott/nets/_nets.py	100.00% <100.00%> (ø)
src/moscot/base/problems/birth_death.py	90.08% <100.00%> (+0.08%)	:arrow_up:
src/moscot/base/problems/compound_problem.py	73.75% <100.00%> (ø)
src/moscot/base/problems/manager.py	78.33% <100.00%> (ø)
src/moscot/utils/subset_policy.py	65.11% <100.00%> (+0.49%)	:arrow_up:
src/moscot/backends/utils.py	79.31% <90.90%> (+1.04%)	:arrow_up:
src/moscot/base/solver.py	87.20% <75.00%> (-1.82%)	:arrow_down:
src/moscot/problems/time/_neural.py	93.75% <93.75%> (ø)
src/moscot/backends/ott/_jax_data.py	95.00% <95.00%> (ø)
src/moscot/backends/ott/nets/_icnn.py	95.74% <95.74%> (ø)
... and 10 more

[jannisborn]

Very nice effort! Is there any timeline on when this PR will be merged? Or, if you are looking for help, could you describe what is missing for this PR to be merged?

[MUCDK]

Very nice effort! Is there any timeline on when this PR will be merged? Or, if you are looking for help, could you describe what is missing for this PR to be merged?

Hi @jannisborn ,

Thanks for your interest in this and your offer for help! As we are also planning to include more recent and performant neural OT estimators, we currently work on their implementations on the OTT-JAX side, and this way get a better idea of how to structure the different neural OT models into one module. As we would like to build moscot upon OTT-JAX as much as possible, we would like to wait until we have a better idea about the implementation on the OTT-JAX side.

Regarding the time line: We will work very intensely on the OTT-JAX implementation next week, and hence - hopefully - the week after, we can focus on the moscot implementation.

If you would like to contribute, or test the implementation and give usability feedback, we would highly appreciate this!

[hbenedek]

Hi @MUCDK!

I am working with @jannisborn and saw that you started to incorporate this PR to ott-jax (and GENOT and more). I would be more than happy to help out with testing/ giving feedback or contributing!

[MUCDK]

Hi @hbenedek , Great to hear you're interested in collaborating! Indeed, I would highly appreciate your help. I think/hope we'll be done with the implementation on the ott-jax side by the end of this week. Still, please feel free to play around with the current implementation and provide feedback. This way, you'll get to know the API, and we can implement the models (i.e. OT-FM, GENOT, and eventually ICNN-OT (all in their conditional format, too) into moscot next week. Therefore, it'd be great if you are familiar with these models, as this will make the workflow easier!

Let's chat about details of the ott-jax -> moscot implementation in the beginning of next week!

[hbenedek]

Thanks @MUCDK! I will try to familiarize myself with the API and the models.