[regression] fix copy missing tensors for llama4

[xin3he]

Description

for google/gemma-3-4b-it: language_model.model.* ↔ model.language_model.*.
but for llama4, we need keep it.
So this change verifies both.

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

Fixes missing-tensor detection/copy logic to handle tensor-name prefix mismatches (reported as a Llama4 regression), and improves readability/determinism of layer-name summaries in logs.

Changes:

• Adds tensor-name aliasing so missing detection matches saved tensors even when prefixes differ (e.g., language_model.model.* ↔ model.language_model.*).
• Improves ignored-layer logging by compressing layer names before printing.
• Makes compress_layer_names() output deterministic by sorting the final parts list.

Reviewed changes

Copilot reviewed 2 out of 2 changed files in this pull request and generated 2 comments.

File	Description
auto_round/utils/missing_tensors.py	Enhances missing-tensor detection via name aliasing and improves config-update logging output.
auto_round/utils/common.py	Sorts compressed layer-name parts to make summaries deterministic.