[Model] Support nested structures for TensorSchema

[DarkLight1337]

Purpose

This helps simplify the implementation of HyperCLOVAX-SEED-Vision and enable merge_by_field_config=True for that model.

Other changes:

• Improve the name of the field in Phi3VImagePixelInputs
• Fix invalid import for GLM4.1V caused by [Multi Modal] Configurable MM Profiling #25631

cc @bbeckca

Test Plan

Test Result

---

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 introduces support for nested tensor structures within TensorSchema, a significant enhancement for managing complex multimodal inputs. The refactoring in hyperclovax_vision.py to adopt this new schema notably improves code clarity and maintainability. I have identified two critical issues in the _call_hf_processor method where video input data structures are incorrectly constructed, which would result in runtime errors. My review provides specific suggestions to rectify these problems. The other modifications, including the core logic updates in tensor_schema.py and the corresponding test adjustments, appear to be well-implemented.

[chatgpt-codex-connector]

💡 Codex Review

Here are some automated review suggestions for this pull request.

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

• Open a pull request for review
• Mark a draft as ready
• Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

[DarkLight1337]

/gemini review

[gemini-code-assist]

Code Review

This pull request introduces support for nested structures in TensorSchema, which simplifies the implementation for models like HyperCLOVAX-SEED-Vision. The changes include refactoring hyperclovax_vision.py to use the new TensorSchema capabilities, renaming a field in Phi3VImagePixelInputs for clarity, and fixing a minor bug in glm4_1v.py. The refactoring in hyperclovax_vision.py is substantial and improves code clarity and maintainability. I have one suggestion to further improve the maintainability of the parsing logic for multimodal inputs.

[Isotr0py]

LGTM!

[bbeckca]

Looks good. Thanks for improving @DarkLight1337!