[Chore] Fix Bagel model import compatibility

[yuanheng-zhao]

PLEASE FILL IN THE PR DESCRIPTION HERE ENSURING ALL CHECKLIST ITEMS (AT THE BOTTOM) HAVE BEEN CONSIDERED.

Purpose

vLLM version has been updated to 0.19.0 but there's an import in Bagel left to be updated.

vLLM 0.19.0 https://github.com/vllm-project/vllm/blob/v0.19.0/vllm/multimodal/parse.py#L22 imports those classes from vllm.inputs

from vllm.inputs import ModalityData, MultiModalDataDict, MultiModalUUIDDict

To prevent export warning and keep future import clean, this PR updates the import path to utilize upstream vllm.inputs

cc @princepride

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. Please provide the test scripts & test commands. Please state the reasons if your codes don't require additional test scripts. For test file guidelines, please check the test style doc
• The test results. Please paste the results comparison before and after, or the e2e results.
• (Optional) The necessary documentation update, such as updating supported_models.md and examples for a new model. Please run mkdocs serve to sync the documentation editions to ./docs.
• (Optional) Release notes update. If your change is user-facing, please update the release notes draft.

BEFORE SUBMITTING, PLEASE READ https://github.com/vllm-project/vllm-omni/blob/main/CONTRIBUTING.md (anything written below this line will be removed by GitHub Actions)

[gcanlin]

Why could this test in CI pass? https://buildkite.com/vllm/vllm-omni/builds/5865/steps/canvas?sid=019d5b22-e3c6-467e-ab95-8ff38f44b359&tab=output

[gcanlin]

But the change LGTM.

[yuanheng-zhao]

Why could this test in CI pass? https://buildkite.com/vllm/vllm-omni/builds/5865/steps/canvas?sid=019d5b22-e3c6-467e-ab95-8ff38f44b359&tab=output

@gcanlin Oh I just re-checked. It's an import warning rather than an actual ImportError (ImportError is from my another WIP branch, not relatable to this one).

The CI didn't trigger ImportError because vLLM has a backward compatibility in https://github.com/vllm-project/vllm/blob/v0.19.0/vllm/multimodal/parse.py#L22 which imports those classes from vllm.inputs

from vllm.inputs import ModalityData, MultiModalDataDict, MultiModalUUIDDict

And the current one will only report a warning like

"ModalityData" is not exported from module "vllm.multimodal.parse"

This PR should be a chore import cleanup. Updated PR title and desc.

[gcanlin]

@yuanheng-zhao Thanks for investigating!