Fixes problems with weighted cross-entropy

[franchuterivera]

address #248

[codecov]

Codecov Report

Merging #263 (0d1e50a) into development (0771dce) will decrease coverage by 51.94%.
The diff coverage is 0.00%.

❗ Current head 0d1e50a differs from pull request most recent head 4f9b37e. Consider uploading reports for the commit 4f9b37e to get more accurate results

@@               Coverage Diff                @@
##           development     #263       +/-   ##
================================================
- Coverage        81.18%   29.23%   -51.95%     
================================================
  Files              150      150               
  Lines             8566     8620       +54     
  Branches          1333     1323       -10     
================================================
- Hits              6954     2520     -4434     
- Misses            1131     6100     +4969     
+ Partials           481        0      -481     

Impacted Files	Coverage Δ
autoPyTorch/utils/implementations.py	33.33% <0.00%> (-42.28%)	⬇️
autoPyTorch/optimizer/utils.py	0.00% <0.00%> (-100.00%)	⬇️
autoPyTorch/api/tabular_regression.py	0.00% <0.00%> (-96.88%)	⬇️
...omponents/setup/network_backbone/ResNetBackbone.py	19.62% <0.00%> (-80.38%)	⬇️
autoPyTorch/api/base_task.py	7.24% <0.00%> (-76.91%)	⬇️
autoPyTorch/ensemble/ensemble_selection.py	18.94% <0.00%> (-76.85%)	⬇️
autoPyTorch/data/tabular_target_validator.py	16.49% <0.00%> (-75.26%)	⬇️
autoPyTorch/data/tabular_feature_validator.py	12.50% <0.00%> (-73.90%)	⬇️
...orch/pipeline/components/training/metrics/utils.py	13.75% <0.00%> (-73.75%)	⬇️
autoPyTorch/optimizer/smbo.py	6.72% <0.00%> (-73.28%)	⬇️
... and 130 more

---

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update 76bdde7...4f9b37e. Read the comment docs.

[ravinkohli]

Hey, this PR looks great. I think this can be merged. Thanks for the added tests to ensure we don't mess this up going forward.

[nabenabe0928]

Thanks for the PR.
I put a comment regarding the validity of the calculation.
Please check the current implementation.