[Model] Supplement to PR 24862: Pass param prefix to LLMHead

[whx-sjtu]

Purpose

This PR is a supplement to PR #24862, passing some prefix parameters that were previously missed in that PR to LLMHead

Test Plan

No extra test needed.

Test Result

All current ci tests should pass.

Essential Elements of an Effective PR Description Checklist

• The purpose of the PR, such as "Fix some issue (link existing issues this PR will resolve)".
• The test plan, such as providing test command.
• The test results, such as pasting the results comparison before and after, or e2e results
• (Optional) The necessary documentation update, such as updating supported_models.md and examples for a new model.
• (Optional) Release notes update. If your change is user facing, please update the release notes draft in the Google Doc.

[gemini-code-assist]

Code Review

This pull request aims to pass the prefix parameter to ParallelLMHead instances across various models, which is a crucial step for ensuring correct quantization. While the intent is correct, the implementation introduces inconsistencies in several models where the provided prefix does not match the module's attribute name. This could lead to problems with quantization configurations that rely on accurate module paths. Furthermore, in models that utilize a ModuleList for multiple LM heads, the same prefix is incorrectly applied to all heads, which would prevent them from being quantized differently if needed. I've added specific comments and suggestions to address these inconsistencies.

[mergify]

This pull request has merge conflicts that must be resolved before it can be
merged. Please rebase the PR, @whx-sjtu.

https://docs.github.com/en/pull-requests/collaborating-with-pull-requests/working-with-forks/syncing-a-fork