BurstAttention and Ulyless all2all support for long sequence training.

[MayDomine]

BurstAttention and Ulyless all2all support for long sequence training.

Issue Reference

N/A

Description

1. Add BurstAttention as distributed ring_flash attention implementation.
2. Add all2all communication ops for all2all attention (Same with DeepSpeed Ulyless).
3. Modify training example code to support sequence parallel(SP)
4. Add SP communicator and communication stream
5. All test related to above change

Type of Change

• 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)
• This change requires a documentation update

How Has This Been Tested?

Please describe the tests that you ran to verify your changes. Provide instructions so we can reproduce.

Checklist

• I have read the CONTRIBUTING document.
• My code follows the code style of this project.
• My change requires a change to the documentation.
• I have updated the documentation accordingly.
• I have added tests to cover my changes.
• All new and existing tests passed.

Additional Information

Any additional information, configuration, or data that might be necessary for the review.