[yaoyu-33]

Summary

• Accept and forward the MCore name constructor kwarg in Step35DecoderLayer
• Add focused unit coverage for the MTP-layer path forwarding name to TransformerLayer
• Audited custom Bridge TransformerLayer/decoder-layer subclasses for the same constructor-contract issue; no additional clear MTP decoder-layer cases found

Testing

• python3 AST check confirmed Step35DecoderLayer.__init__ now includes name
• uv tool run ruff check src/megatron/bridge/models/stepfun/step35_provider.py tests/unit_tests/models/stepfun/test_step35_provider.py
• uv tool run ruff format --check src/megatron/bridge/models/stepfun/step35_provider.py tests/unit_tests/models/stepfun/test_step35_provider.py
• python3 -m py_compile src/megatron/bridge/models/stepfun/step35_provider.py tests/unit_tests/models/stepfun/test_step35_provider.py
• git diff --check
• uv run --active --no-sync pre-commit run --files src/megatron/bridge/models/stepfun/step35_provider.py tests/unit_tests/models/stepfun/test_step35_provider.py

Focused pytest and full pre-commit were attempted with uv run, but local environment resolution is blocked because nvidia-resiliency-ext==0.6.0 has no compatible wheel for this workstation platform (manylinux_2_31_x86_64; available wheels target manylinux_2_39).

[copy-pr-bot]

This pull request requires additional validation before any workflows can run on NVIDIA's runners.

Pull request vetters can view their responsibilities here.

Contributors can view more details about this message here.

[claude]

Light Review

LGTM — clean, minimal fix. The name kwarg is accepted and forwarded to super().__init__() correctly, and the test coverage directly verifies the forwarding path via the recorder pattern.

No issues found.

Suggested test cases

No perf tests impacted.

[yaoyu-33]

/ok to test 8383660