[PyTorch] Fix multiple calls to saved_tensors in CP attention

[ksivaman]

Description

The torch autograd functions for attention with context parallelism incorrectly make multiple calls to ctx.saved_tensors in backward pass resulting in the following error:

Unpack is being triggered for a tensor that was already unpacked once. If you are calling ctx.saved_tensors in backward, make sure to do so only once. Otherwise please open an issue with details on your use case.

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

• Call ctx.saved_tensors exactly once before unpacking into separate variables

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

[ksivaman]

/te-ci pytorch L1

[xrennvidia]

LGTM, thanks.

[cyanguwa]

LGTM