[PyTorch] Bugfix for wgrad bulk overlap conflict when dgrad overlap is reduce-scatter

[denera]

Description

When Userbuffers config dictionary sets overlap method to ring-exchange or pipeline for any *_dgrad layer, that layer's *_wgrad overlap needs to be disabled in order for ub_overlap_rs_dgrad=True option for related TE modules to function correctly.

This PR fixes a bug where the "*_wgrad" overlap was persisting in the Userbuffer configuration and the corresponding UB object was being initialized even when it was not needed.

• @fanshiqing for viz.

Type of change

• [ ] Documentation change (change only to the documentation, either a fix or a new content)
• [x] 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:

• *_wgrad overlap is now removed from methods["bulk"] list when the same layer's *_dgrad overlap has its method set to either ring-exchange or pipeline.
• add_ub(name, **ub_cfg) is now only called if name is in the original user-provided ub_cfg. This avoids creating UB objects with default configs that may conflict with the user's intended TP overlap use.

Checklist:

• [x] I have read and followed the contributing guidelines
• [x] The functionality is complete
• [ ] 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

[timmoon10]

/te-ci pytorch L0 L1

[denera]

/te-ci pytorch L0 L1