[Algorithm] Update Stable KLD (NeurIPS'21)

[yangxue0827]

Thanks for your contribution and we appreciate it a lot. The following instructions would make your pull request more healthy and more easily get feedback. If you do not understand some items, don't worry, just make the pull request and seek help from maintainers.

Motivation

Solve the instability issue (NAN) of KLD training.

Modification

• Use gaussian_dist_loss intead of gaussian_dist_loss_v1
• Preliminary analysis is due to lack of truncation processing (torch.clamp).
• Update README.md
• Add new configs with the key word stable.

BC-breaking (Optional)

Does the modification introduce changes that break the back-compatibility of the downstream repos?
If so, please describe how it breaks the compatibility and how the downstream projects should modify their code to keep compatibility with this PR.

Use cases (Optional)

If this PR introduces a new feature, it is better to list some use cases here, and update the documentation.

Checklist

1. Pre-commit or other linting tools are used to fix the potential lint issues.
2. The modification is covered by complete unit tests. If not, please add more unit test to ensure the correctness.
3. The documentation has been modified accordingly, like docstring or example tutorials.

[codecov-commenter]

Codecov Report

Merging #183 (7ce3def) into dev (44b0aec) will increase coverage by 0.57%.
The diff coverage is n/a.

@@            Coverage Diff             @@
##              dev     #183      +/-   ##
==========================================
+ Coverage   28.30%   28.88%   +0.57%     
==========================================
  Files         108      109       +1     
  Lines        6967     7074     +107     
  Branches     1048     1064      +16     
==========================================
+ Hits         1972     2043      +71     
- Misses       4927     4961      +34     
- Partials       68       70       +2     

Flag	Coverage Δ
unittests	28.86% <ø> (+0.57%)	⬆️

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

Impacted Files	Coverage Δ
mmrotate/models/detectors/oriented_rcnn.py	41.17% <0.00%> (-42.16%)	⬇️
mmrotate/core/visualization/image.py	68.57% <0.00%> (-7.43%)	⬇️
...ate/models/roi_heads/oriented_standard_roi_head.py	16.66% <0.00%> (-0.32%)	⬇️
mmrotate/models/roi_heads/roi_trans_roi_head.py	11.88% <0.00%> (-0.26%)	⬇️
mmrotate/datasets/sar.py	100.00% <0.00%> (ø)
mmrotate/datasets/dota.py	15.78% <0.00%> (ø)
mmrotate/core/visualization/__init__.py	100.00% <0.00%> (ø)
mmrotate/models/losses/gaussian_dist_loss.py	84.12% <0.00%> (ø)
mmrotate/core/visualization/palette.py	93.54% <0.00%> (ø)
mmrotate/models/roi_heads/gv_ratio_roi_head.py	17.30% <0.00%> (+0.64%)	⬆️
... and 2 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 44b0aec...7ce3def. Read the comment docs.

[liuyanyi]

LGTM

[liuyanyi]

LGTM

[zytx121]

LGTM