[Fix] Fix AMP in Ascend and support using NPUJITCompile environment

[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

1. Support automatic mixed-precision training with pytorch version 1.10 or above.
2. Support manually set up the use of dynamic shapes and static shapes.

Modification

1. mmengine/runner/amp.py. complements to NPU autocast.
2. mmengine/device/utils.py. supplemented with a switch to turn on static shape，dynamic shapes are used by default.

BC-breaking (Optional)

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

Use cases (Optional)

Open the static shape：export NPUJITCompile=True

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. If the modification has potential influence on downstream projects, this PR should be tested with downstream projects, like MMDet or MMCls.
4. The documentation has been modified accordingly, like docstring or example tutorials.

[codecov]

Codecov Report

❗ No coverage uploaded for pull request base (main@0d25625). Click here to learn what that means.
Patch has no changes to coverable lines.

❗ Current head 2df4355 differs from pull request most recent head 4c5e458. Consider uploading reports for the commit 4c5e458 to get more accurate results

Additional details and impacted files

@@           Coverage Diff           @@
##             main     #994   +/-   ##
=======================================
  Coverage        ?   76.57%           
=======================================
  Files           ?      139           
  Lines           ?    11037           
  Branches        ?     2207           
=======================================
  Hits            ?     8452           
  Misses          ?     2219           
  Partials        ?      366           

Flag	Coverage Δ
unittests	76.57% <0.00%> (?)

Flags with carried forward coverage won't be shown. Click here to find out 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 in Codecov by Sentry.
📢 Do you have feedback about the report comment? Let us know in this issue.

[wangjiangben-hw]

/lgtm

[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

Supplement the support of some functions on the NPU.

Modification

1.mmengine/runner/amp.py. complements to NPU autocast. 2.mmengine/device/utils.py. supplemented with a switch to turn on static shape.