[Feature] Support mmcls with NPU backend.

[wangjiangben-hw]

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

Added ascending device support in mmcv.

Modification

• The main modification points are as follows:
  1. In mmcls/apis/train.py, We added an NPU device in the DDP scenario, and the mixing accuracy is enabled by default when using the NPU
  2. In ``mmcls/utils/distribution.py```, We have added DP and DDP classes for NPU
  3. In ``mmcls/datasets/samplers/distributed_sampler.py```, We added device judgment to make the broadcast within the sync_random_seed function execute correctly on the npu

BC-breaking (Optional)

A new mmcv containing the NPU background needed.

Use cases (Optional)

We tested resnet50_8xb32_in1k.py with Top-1 (%): 76.4

Checklist

Before PR:

• Pre-commit or other linting tools are used to fix the potential lint issues.
• Bug fixes are fully covered by unit tests, the case that causes the bug should be added in the unit tests.
• The modification is covered by complete unit tests. If not, please add more unit test to ensure the correctness.
• The documentation has been modified accordingly, like docstring or example tutorials.

After PR:

• If the modification has potential influence on downstream or other related projects, this PR should be tested with those projects, like MMDet or MMSeg.
• CLA has been signed and all committers have signed the CLA in this PR.

[CLAassistant]

All committers have signed the CLA.

[codecov]

Codecov Report

Base: 86.09% // Head: 85.98% // Decreases project coverage by -0.11% ⚠️

Coverage data is based on head (905ea52) compared to base (517bd3d).
Patch coverage: 13.33% of modified lines in pull request are covered.

Additional details and impacted files

@@            Coverage Diff             @@
##           master    #1072      +/-   ##
==========================================
- Coverage   86.09%   85.98%   -0.12%     
==========================================
  Files         140      142       +2     
  Lines        9676     9906     +230     
  Branches     1677     1715      +38     
==========================================
+ Hits         8331     8518     +187     
- Misses       1092     1124      +32     
- Partials      253      264      +11     

Flag	Coverage Δ
unittests	85.93% <13.33%> (-0.12%)	⬇️

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

Impacted Files	Coverage Δ
mmcls/apis/train.py	15.00% <0.00%> (-0.39%)	⬇️
mmcls/utils/distribution.py	7.40% <0.00%> (-3.12%)	⬇️
mmcls/datasets/samplers/distributed_sampler.py	85.71% <100.00%> (+0.52%)	⬆️
mmcls/models/backbones/repvgg.py	96.15% <0.00%> (-0.69%)	⬇️
mmcls/apis/test.py	23.27% <0.00%> (-0.66%)	⬇️
mmcls/models/backbones/van.py	96.83% <0.00%> (ø)
mmcls/models/utils/__init__.py	100.00% <0.00%> (ø)
mmcls/models/utils/attention.py	96.91% <0.00%> (ø)
mmcls/models/backbones/__init__.py	100.00% <0.00%> (ø)
... and 5 more

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.

[mzr1996]

To be compatible with the old version of MMCV, please avoid importing new API from MMCV.