Bug compare models

[paulsonak]

• running predictions on classification models using the balancing weight transformers requires weights in the MinimalDataset representing the data to predict on

• using get_multitask_perf_from_files_new returned NaN metrics for single task models if they were mixed in with multitask models. Now, they both return the proper metrics.

((I am submitting this to be merged into bug_transformer_fitting but I can change it, I think it's independent I just need all of these bugs fixed in the same branch.))

[codecov]

Codecov Report

Attention: Patch coverage is 6.25000% with 45 lines in your changes missing coverage. Please review.

Files with missing lines	Patch %	Lines
atomsci/ddm/pipeline/compare_models.py	0.00%	43 Missing ⚠️
atomsci/ddm/pipeline/model_datasets.py	60.00%	2 Missing ⚠️

@@                     Coverage Diff                     @@
##           bug_transformer_fitting     #387      +/-   ##
===========================================================
+ Coverage                    28.53%   30.98%   +2.45%     
===========================================================
  Files                           47       47              
  Lines                        12919    13795     +876     
===========================================================
+ Hits                          3687     4275     +588     
- Misses                        9232     9520     +288     

Files with missing lines	Coverage Δ
atomsci/ddm/pipeline/model_datasets.py	63.55% <60.00%> (+2.68%)	⬆️
atomsci/ddm/pipeline/compare_models.py	5.58% <0.00%> (-0.07%)	⬇️

... and 3 files with indirect coverage changes

[paulsonak]

This belongs with issue #382

[stewarthe6]

Is there a test for this?