fix google/gemma-3-4b-it

[xin3he]

Description

The MTP missing tensor detection logic cannot handle gemma since the tensor name is using language_model.layers instead of model.language_model.layers.

Type of Change

• Bug fix
• New feature
• Documentation update
• Performance improvement
• Code refactoring
• Other (please specify):

Related Issues

Fixes or relates to #

Checklist Before Submitting

• My code has been tested locally.
• Documentation has been updated as needed.
• New or updated tests are included where applicable.

[Copilot]

Pull request overview

This PR aims to fix missing-tensor (MTP) detection for Gemma models where tensor paths use language_model.layers.* instead of model.language_model.layers.*, so tensors aren’t incorrectly treated as missing and copied.

Changes:

• Extend _is_truly_missing() to treat certain parent-layer paths as equivalent when only a leading prefix differs.

---

You can also share your feedback on Copilot code review. Take the survey.

[wenhuach21]

better add an ut to test mtp layers in the future.

[xin3he]

better add an ut to test mtp layers in the future.

We had many UTs in test_missing_tensors.py, it's just hard to handle all corner cases.