Predicting using MLJ interface is slow

[pat-alt]

@pasq-cat I think we should indeed just go with the direct MLJ interface #121 but for now I would still merge this cause it addresses the compute times.

[codecov]

Codecov Report

All modified and coverable lines are covered by tests ✅

Project coverage is 96.48%. Comparing base (78c846e) to head (9873c55).
Report is 5 commits behind head on main.

Additional details and impacted files

@@            Coverage Diff             @@
##             main     #123      +/-   ##
==========================================
- Coverage   96.65%   96.48%   -0.17%     
==========================================
  Files          22       22              
  Lines         658      655       -3     
==========================================
- Hits          636      632       -4     
- Misses         22       23       +1     

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

[pasq-cat]

nvm this message , i was wrong

[pat-alt]

shoudln't the same be done in function MLJFlux.predict(model::LaplaceRegression, fitresult, Xnew)

?

Yes, I've done that

[pasq-cat]

shoudln't the same be done in function MLJFlux.predict(model::LaplaceRegression, fitresult, Xnew)
?

Yes, I've done that

btw i was sure that laplace accepted vectors as input. has anything been changed or i was wrong the whole time? the documentation says abstractarray but i distinctly remember this requirement....

[pat-alt]

shoudln't the same be done in function MLJFlux.predict(model::LaplaceRegression, fitresult, Xnew)
?

Yes, I've done that

btw i was sure that laplace accepted vectors as input. has anything been changed or i was wrong the whole time? the documentation says abstractarray but i distinctly remember this requirement....

hmm not sure where you say this https://juliatrustworthyai.github.io/LaplaceRedux.jl/stable/reference/#LaplaceRedux.predict-Tuple{LaplaceRedux.AbstractLaplace,%20AbstractArray}

[pasq-cat]

shoudln't the same be done in function MLJFlux.predict(model::LaplaceRegression, fitresult, Xnew)
?

Yes, I've done that

btw i was sure that laplace accepted vectors as input. has anything been changed or i was wrong the whole time? the documentation says abstractarray but i distinctly remember this requirement....

hmm not sure where you say this https://juliatrustworthyai.github.io/LaplaceRedux.jl/stable/reference/#LaplaceRedux.predict-Tuple{LaplaceRedux.AbstractLaplace,%20AbstractArray}

i know i know, must have been something that i have misunderstood in the first days, mah