Fix erroranalysis conversion exception with object detection task for complex pred Y and true Y data and numpy>=1.26.0 by specifying object type

[imatiach-msft]

Description

Fix erroranalysis conversion exception with object detection task for complex pred Y and true Y data and numpy>=1.26.0 by specifying object type.

With numpy>=1.26.0 erroranalysis package throws exception:

...
  File "/azureml-envs/responsibleai-vision/lib/python3.9/site-packages/responsibleai/managers/error_analysis_manager.py", line 337, in compute
    report = self._analyzer.create_error_report(
  File "/azureml-envs/responsibleai-vision/lib/python3.9/site-packages/erroranalysis/analyzer/error_analyzer.py", line 423, in create_error_report
    tree = self.compute_error_tree(self.feature_names,
  File "/azureml-envs/responsibleai-vision/lib/python3.9/site-packages/erroranalysis/analyzer/error_analyzer.py", line 344, in compute_error_tree
    return _compute_error_tree(self,
  File "/azureml-envs/responsibleai-vision/lib/python3.9/site-packages/erroranalysis/_internal/surrogate_error_tree.py", line 266, in compute_error_tree
    return compute_error_tree_on_dataset(
  File "/azureml-envs/responsibleai-vision/lib/python3.9/site-packages/erroranalysis/_internal/surrogate_error_tree.py", line 169, in compute_error_tree_on_dataset
    booster, dataset_indexed_df, cat_info = get_surrogate_booster_local(
  File "/azureml-envs/responsibleai-vision/lib/python3.9/site-packages/erroranalysis/_internal/surrogate_error_tree.py", line 345, in get_surrogate_booster_local
    pred_y = np.array(pred_y)
ValueError: setting an array element with a sequence. The requested array has an inhomogeneous shape after 1 dimensions. The detected shape was (10,) + inhomogeneous part.

Based on stackoverflow suggestion:
https://stackoverflow.com/questions/67183501/setting-an-array-element-with-a-sequence-requested-array-has-an-inhomogeneous-sh

The fix is to use numpy object type when converting list to numpy array for complex data which is used in tasks like object detection. A test was also added to reproduce the exception locally and validate that the issue is resolved after the type cast was added for newer numpy>=1.26.0.

Checklist

• I have added screenshots above for all UI changes.
• I have added e2e tests for all UI changes.
• Documentation was updated if it was needed.

[codecov-commenter]

Codecov Report

Attention: Patch coverage is 71.42857% with 4 lines in your changes missing coverage. Please review.

Project coverage is 87.84%. Comparing base (dfa37be) to head (1aef947).

Files with missing lines	Patch %	Lines
...is/erroranalysis/_internal/surrogate_error_tree.py	55.55%	4 Missing ⚠️

Additional details and impacted files

@@            Coverage Diff             @@
##             main    #2594      +/-   ##
==========================================
- Coverage   87.89%   87.84%   -0.05%     
==========================================
  Files         113      113              
  Lines        6648     6657       +9     
==========================================
+ Hits         5843     5848       +5     
- Misses        805      809       +4     

Flag	Coverage Δ
unittests	87.84% <71.42%> (-0.05%)	⬇️

Flags with carried forward coverage won't be shown. Click here to find out more.

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