[Fix] Fix npu optim

[luomaoling]

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

Adaptation torch_npu cases where optim is not included.

Modification

mmcv/runner/optimizer/builder.py.

BC-breaking (Optional)

Does the modification introduce changes that break the backward-compatibility of the downstream repositories?
None.

Use cases (Optional)

python tools/train.py configs/resnet/resnet50_8xb32_in1k.py --cfg-options optimizer.type=NpuFusedSGD.

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.