Add a CP implementation variant with KV all-gather.

[xrennvidia]

Description

This is a CP implementation variant with KV all-gather. Currently, it can support:

• sliding window attention + causal + FlashAttention
• full window attention + causal + FlashAttention
• full window attention + causal + FusedAttention

Will add more functionality support later.

The KV all-gather communication is exposed, but the overheads should be small with GQA/MQA.

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)
• [x] 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:

• Add cp_comm_type in attention.py to allow users choose which CP implementation they want.
• Add unit tests for the new implementation variant.

Checklist:

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

[xrennvidia]

/te-ci

[xrennvidia]

/te-ci pytorch

[xrennvidia]

/te-ci pytorch

[xrennvidia]

/te-ci pytorch

[cyanguwa]

/te-ci pytorch