[PyTorch] Debug checkpointing with te.Sequential

[timmoon10]

Description

TE 2.0 changed how the fusible ops handled FP8 state, i.e. using quantizers rather than fp8_meta dicts, but the get_extra_state/set_extra_state functions were not updated. This PR updates these functions so they are similar to the module get_extra_state/set_extra_state functions:

TransformerEngine/transformer_engine/pytorch/module/base.py

Line 556 in be055eb

def get_extra_state(self) -> torch.Tensor:

TransformerEngine/transformer_engine/pytorch/module/base.py

Line 616 in be055eb

def set_extra_state(self, state: torch.Tensor) -> None:

I've also added a unit test.

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 refactoring

Changes

• Update checkpointing functions in fusible ops with changes from TE 2.0
• Add tests for checkpointing fusible ops

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

[timmoon10]

/te-ci pytorch