[Enhancement] Update RAFT model

[MeowZheng]

1. As fix the bug in RAFT decoder [Fix] Fix RAFT wrong refine iter after evaluation #62, the model has been retrained.
2. Use SyncBN to improve the performance.
3. Release model fine-tuned on KITTI2015

[codecov]

Codecov Report

Merging #68 (b97f667) into master (af18201) will decrease coverage by 0.10%.
The diff coverage is n/a.

@@            Coverage Diff             @@
##           master      #68      +/-   ##
==========================================
- Coverage   67.91%   67.81%   -0.11%     
==========================================
  Files          84       84              
  Lines        4828     4828              
  Branches      689      689              
==========================================
- Hits         3279     3274       -5     
- Misses       1444     1450       +6     
+ Partials      105      104       -1     

Flag	Coverage Δ
unittests	67.81% <ø> (-0.11%)	⬇️

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

Impacted Files	Coverage Δ
mmflow/datasets/pipelines/transforms.py	79.33% <0.00%> (-1.17%)	⬇️
mmflow/models/decoders/gma_decoder.py	97.95% <0.00%> (+2.04%)	⬆️

---

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 af18201...b97f667. Read the comment docs.