fix-1547-Bug-in-_partial_fit_and_predict_iterative-in-train_evaluator.py_

[eddiebergman]

Fixes the below bug and also provides the X_data with every call to self._loss, which was missing in some parts.

---

• Closes Bug in _partial_fit_and_predict_iterative in train_evaluator.py? #1547

[mfeurer]

I really like how this PR cuts down bloat. I have left a few minor comments. Could you also please add a unit test?

[eddiebergman]

I'm not really sure what to unit test here to be honest, could you be a bit more precise?

[mfeurer]

After thinking more about what this PR fixes (making the indexing of pandas dataframes and numpy arrays consistent) we don't need a unit test right now to ensure this - we don't have unit tests for pandas and therefore haven't previously caught this error, and this PR simplifies how we access pandas and we can see from the changelog that it is a clear improvement that right now does not require testing.

[mfeurer]

There are also unit tests failing wrt the changes (they are hidden among the other failiing unit tests, so you could maybe rebase on development where they are fixed?)?

[codecov]

Codecov Report

Merging #1550 (c4b2c2b) into development (9035f05) will increase coverage by 0.22%.
The diff coverage is 98.38%.

@@               Coverage Diff               @@
##           development    #1550      +/-   ##
===============================================
+ Coverage        84.43%   84.66%   +0.22%     
===============================================
  Files              157      157              
  Lines            11948    11976      +28     
  Branches          2063     2065       +2     
===============================================
+ Hits             10088    10139      +51     
+ Misses            1297     1286      -11     
+ Partials           563      551      -12     

[eddiebergman]

Seems the meta learning timeouts have struck again ... I really don't know why, going to rerun the tests