[Feature] Support engine 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 mmcv/device/, We've added our NPU devices. Since the NPU does not support the DataParallel mode, DP and DDP have been customized, the Scatter related methods have been rewritten, and the use case of MLU has been referenced, thanks.
  2. In mmcv/runner/dist_utils.py, We have added NPU-related distributed initialization methods.
  3. In mmcv/runner/hooks/optimizer.py, We've added the amp method on npu.
  4. In mmcv/utils/, We've added NPU devices.

BC-breaking (Optional)

None

Use cases (Optional)

• test case
  • tests/test_device/test_device_utils.py
  • tests/test_device/test_functions.py
  • tests/test_device/test_npu/test_npu_parallel.py

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 test to ensure the correctness.
• The documentation has been modified accordingly, including docstring or example tutorials.

After PR:

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

[CLAassistant]

All committers have signed the CLA.

[grimoire]

Why do we remove HIP_DIFF?

[wangjiangben-hw]

https://github.com/open-mmlab/mmcv/blob/master/setup.py#L283

This might be a comparison error...

[grimoire]

Please update tests/test_ops/test_focal_loss.py.

[wangjiangben-hw]

Please update tests/test_ops/test_focal_loss.py.

done..

[grimoire]

LGTM