training accuracy reported as 0.0 with CUB hierarchical label for the classification

[yunchu]

Describe the bug
Training accuracy reported as 0.0 with CUB hierarchical label for the classification
here is the part of regression_config.json for this testing

Steps to Reproduce

1. run classification regression TC

$ CI_DATA_ROOT=/home/yunchule/data/ci_datasets/ tox -vvv -e tests-cls-py310-pt1 -- tests/regression/classification/test_classification.py::TestRegressionHierarchicalLabelClassification::test_otx_train[Custom_Image_Classification_EfficientNet-V2-S]

2. check the log

>       assert test_result["passed"] is True, test_result["log"]
E       AssertionError: [EfficientNet-V2-S] Performance: (0.0) < Criteria: (0.5418), threshold: 0.1.
E       assert False is True

Environment:

• OS: Ubuntu 20.04
• Framework version: Pytorch
• Python version: 3.10
• OpenVINO version: 2023.0
• CUDA/cuDNN version: 11.7
• GPU model and memory: 3090/24G

[wonjuleee]

The problem was because of the hierarchical classification dataset generation from the original CUB data. Since CUB categories are composed of proper noun of birds, it is not appropriate to split the category names into hierarchies.

[yunchu]

Thanks for your update @wonjuleee
I will close this issue.

[goodsong81]

Found root cause, and fixed in #2626