standardize AbstractFloats

[tiemvanderdeure]

Changes the float type of the fitresult of Standardizer to AbstractFloat. This means standardizers will no longer force floats to Float64.

I also removed the second definition of fitresult_given_feature - I think this must have been an oversight.

[ablaom]

Thanks for this. It seems that codecov is hanging. I've been having this issue elsewhere. Can you try updating the codecov GH action to version @v3. You can do that here:

MLJModels.jl/.github/workflows/ci.yml

Line 45 in 0de3aec

- uses: codecov/codecov-action@v1

Don't use @v4 as that requires us to add a security token.

[codecov]

Codecov Report

All modified and coverable lines are covered by tests ✅

Project coverage is 77.08%. Comparing base (5dc7eb2) to head (1757802).
Report is 5 commits behind head on dev.

❗ Current head 1757802 differs from pull request most recent head 66303c4. Consider uploading reports for the commit 66303c4 to get more accurate results

Additional details and impacted files

@@            Coverage Diff             @@
##              dev     #555      +/-   ##
==========================================
+ Coverage   76.92%   77.08%   +0.16%     
==========================================
  Files          16       16              
  Lines        1170     1126      -44     
==========================================
- Hits          900      868      -32     
+ Misses        270      258      -12     

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

[ablaom]

Could you please also fix this line:

MLJModels.jl/src/builtins/Transformers.jl

Line 506 in 1757802

std(v) > eps(Float64) ||

More correct is eps(T), right?

[ablaom]

Good to have this, thanks. 🙏🏾

[tiemvanderdeure]

MLJModels.jl/src/builtins/Transformers.jl

Line 506 in 1757802

std(v) > eps(Float64) ||

More correct is eps(T), right?

eps(T) would fail if T is an integer, but eps(std(v)) should always work.