Fix the CustomNonLinearClsHead when the batch_size is set to 1

[sungmanc]

Summary

This PR introduces a new trick to avoid an error that happens in the bs=1 case at the BN1d included at the CustomNonLinearClsHead. For now, just applying to the loss != IBLoss, need to consider further cases.

How to test

otx train src/otx/algorithms/classification/configs/mobilenet_v3_large_1_cls_incr/template.yaml --train-data-roots tests/assets/classification_dataset --val-data-roots tests/assets/classification_dataset params --learning_parameters.batch_size 1

Checklist

• I have added unit tests to cover my changes.​
• I have added integration tests to cover my changes.​
• I have added e2e tests for validation.
• I have added the description of my changes into CHANGELOG in my target branch (e.g., CHANGELOG in develop).​
• I have updated the documentation in my target branch accordingly (e.g., documentation in develop).
• I have linked related issues.

License

• I submit my code changes under the same Apache License that covers the project.
  Feel free to contact the maintainers if that's a concern.
• I have updated the license header for each file (see an example below).

# Copyright (C) 2023 Intel Corporation
# SPDX-License-Identifier: Apache-2.0

[codecov]

Codecov Report

Attention: 2 lines in your changes are missing coverage. Please review.

Comparison is base (6d44fc0) 81.51% compared to head (a8dedb9) 81.50%.

Additional details and impacted files

@@             Coverage Diff             @@
##           develop    #2569      +/-   ##
===========================================
- Coverage    81.51%   81.50%   -0.01%     
===========================================
  Files          516      516              
  Lines        38255    38259       +4     
===========================================
  Hits         31184    31184              
- Misses        7071     7075       +4     

Flag	Coverage Δ
py310	81.50% <50.00%> (-0.01%)	⬇️
py38	81.48% <50.00%> (-0.01%)	⬇️
py39	81.48% <50.00%> (-0.01%)	⬇️

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

Files	Coverage Δ
...ion/adapters/mmcls/models/heads/custom_cls_head.py	90.62% <50.00%> (-2.71%)	⬇️

... and 1 file with indirect coverage changes

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

[eunwoosh]

It just a question. It seems that error is raised due to dimension problem, is it right?
if it's right, how about changing dimension of tensor instead of concatenating same tensor?

[sungmanc]

It just a question. It seems that error is raised due to dimension problem, is it right? if it's right, how about changing dimension of tensor instead of concatenating same tensor?

Right, however, if you mean just changing the shape of the tensor (i.e. [1, 960] --> [2, 480]. it could affect to the output since we used BatchNorm1d

[sungmanc]

Discard this PR and will made PR for the release branch

[eunwoosh]

Right, however, if you mean just changing the shape of the tensor (i.e. [1, 960] --> [2, 480]. it could affect to the output since we used BatchNorm1d

ah I thought that [960] needs to be changed to [1, 960], but it's already [1, 960]. I think I just misunderstood it.