[Fix] permute fusion number of forward output and backward input is not match

[RuibinCheung]

Description

Fix permute fusion number of forward output and backward input is not match when inp.numel() is zero.

This is a nvidia's upstream code bug. And it was fixed on PR NVIDIA/TransformerEngine#1468. But this PR involves a feature which relates to PTX. It can't cherry-pick it directly now.

It will return two tensor when expert of MoE dispatched token is 0 but backward function only accepts one tensor. It will raise a runtime error because the number of parameters and arguments is not matched.

Type of change

• Documentation change (change only to the documentation, either a fix or a new content)
• Bug fix (non-breaking change which fixes an issue)
• New feature (non-breaking change which adds functionality)
• Breaking change (fix or feature that would cause existing functionality to not work as expected)
• Infra/Build change
• Code refractor

Changes

Please list the changes introduced in this PR:

• Change A
• Change B

Checklist:

• I have read and followed the contributing guidelines
• The functionality is complete
• I have commented my code, particularly in hard-to-understand areas
• I have made corresponding changes to the documentation
• My changes generate no new warnings
• I have added tests that prove my fix is effective or that my feature works
• New and existing unit tests pass locally with my changes

[RuibinCheung]

@wangye805 PTAL

[wangye805]

LGTM