[Fix] fix config and code errors of swin/hrformer

[ly015]

Motivation

Modification

• Fix misplaced paramwise_cfg in configs
• Fix a bug in swin backbone initialization

BC-breaking (Optional)

Use cases (Optional)

Checklist

Before PR:

• I have read and followed the workflow indicated in the CONTRIBUTING.md to create this PR.
• Pre-commit or linting tools indicated in CONTRIBUTING.md are used to fix the potential lint issues.
• Bug fixes are covered by unit tests, the case that causes the bug should be added in the unit tests.
• New functionalities are covered by complete unit tests. If not, please add more unit tests to ensure correctness.
• The documentation has been modified accordingly, including docstring or example tutorials.

After PR:

• CLA has been signed and all committers have signed the CLA in this PR.

[codecov]

Codecov Report

Base: 82.20% // Head: 82.16% // Decreases project coverage by -0.04% ⚠️

Coverage data is based on head (e4669fa) compared to base (a2b2a31).
Patch coverage: 0.00% of modified lines in pull request are covered.

❗ Current head e4669fa differs from pull request most recent head 4fa4866. Consider uploading reports for the commit 4fa4866 to get more accurate results

Additional details and impacted files

@@             Coverage Diff             @@
##           dev-1.x    #1995      +/-   ##
===========================================
- Coverage    82.20%   82.16%   -0.04%     
===========================================
  Files          214      214              
  Lines        12736    12734       -2     
  Branches      2154     2154              
===========================================
- Hits         10469    10463       -6     
- Misses        1806     1808       +2     
- Partials       461      463       +2     

Flag	Coverage Δ
unittests	82.16% <0.00%> (-0.04%)	⬇️

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

Impacted Files	Coverage Δ
mmpose/models/backbones/swin.py	84.32% <0.00%> (+0.25%)	⬆️
mmpose/codecs/associative_embedding.py	92.50% <0.00%> (-3.75%)	⬇️
mmpose/datasets/transforms/common_transforms.py	85.67% <0.00%> (+0.28%)	⬆️

Help us with your feedback. Take ten seconds to tell us how you rate us. Have a feature suggestion? Share it here.

☔ View full report at Codecov.
📢 Do you have feedback about the report comment? Let us know in this issue.

[Ben-Louis]

There are still some discrepancies between the Swin model and the checkpoint from the previous version.

The following error is reported when initializing the model defined by td-hm_swin-t-p4-w7_8xb32-210e_coco-256x192:

unexpected key in source state_dict: backbone.norm0.weight, backbone.norm0.bias, backbone.norm1.weight, backbone.norm1.bias, backbone.norm2.weight, backbone.norm2.bias

Is this behaving as expected?

[Ben-Louis]

There are still some discrepancies between the Swin model and the checkpoint from the previous version.

The following error is reported when initializing the model defined by td-hm_swin-t-p4-w7_8xb32-210e_coco-256x192:

unexpected key in source state_dict: backbone.norm0.weight, backbone.norm0.bias, backbone.norm1.weight, backbone.norm1.bias, backbone.norm2.weight, backbone.norm2.bias

Is this behaving as expected?

I get it. Only the normalization layers that are necessary for the out_indices are constructed.