Tweak fit_data_scitype fallback to resolve issue with class weights

ablaom commented 7 months ago

Resolves #170.

We need an Any type where we had AbstractDict{...}, which doesn't make sense, because currently dictionaries have Unknown scitype, unless they are keyed on integer (multisets).

ablaom commented 7 months ago

To do:

[x] Check MLJ integration tests locally

codecov[bot] commented 7 months ago

Codecov Report

All modified and coverable lines are covered by tests :white_check_mark:

Project coverage is 88.53%. Comparing base (3e36cc7) to head (12f588e).

Additional details and impacted files

```diff @@ Coverage Diff @@ ## dev #195 +/- ## ======================================= Coverage 88.53% 88.53% ======================================= Files 8 8 Lines 375 375 ======================================= Hits 332 332 Misses 43 43 ```

:umbrella: View full report in Codecov by Sentry.
:loudspeaker: Have feedback on the report? Share it here.

JuliaAI / MLJModelInterface.jl

Tweak fit_data_scitype fallback to resolve issue with class weights #195

Codecov Report