[Question] about --spec in model.sh

Your Question

I am wondering if --spec will work for megatron bridge?
For example, there is the {--spec "slime_plugins.models.qwen3_5" "get_qwen3_5_spec"} in scripts/models/qwen3.5-35B-A3B.sh. But I couldn't find where --spec should take effect in the Megatron Bridge. In the Megatron Bridge, qwen35_vl_provider.py calls block_spec = get_gpt_decoder_block_spec, not the externally passed --spec. Furthermore, I added some print statements in the get_qwen3_5_spec method, but these print statements had no effect when using the Megatron Bridge. So, do I understand that --spec doesn't work in the Megatron Bridge context?

What I've Tried

no

Environment (if relevant)

• slime version:
• Python version:
• PyTorch version:
• CUDA/ROCm version:
• GPU type and count:
• OS:

Additional Context

No response

Pre-submission Checklist

• I have read the CONTRIBUTING.md and understand the collaboration scope.
• I have read the documentation and FAQ and my question is not answered there.
• I have searched for existing issues and my question has not been asked before.

Start with scripts/models/qwen3.5-35B-A3B.sh and trace how --spec is passed into the Megatron Bridge. Compare that path with qwen35_vl_provider.py, especially its call to get_gpt_decoder_block_spec. Done means clarifying whether the external spec is used in this context and documenting the observed behavior.