[Bugfix] NextStep cannot launch with offline example text_to_image.py

[fhfuih]

Purpose

The offline example text_to_image.py script specially caters to NextStep model by setting

omni_kwargs["model_class_name"] = "NextStep11Pipeline"
# ...
omni = Omni(**omni_kwargs)

But in this case, it skips the real setup in OmniDiffusion.

def __init__ # ...
    # ...
    elif model_type == "nextstep":
        if od_config.model_class_name is None:  # Not None any more!
            pipeline_class = "NextStep11Pipeline"  # This is not Set!

Thus, it results in the following error

Traceback (most recent call last):
  File "/usr/lib/python3.12/multiprocessing/process.py", line 314, in _bootstrap
    self.run()
  File "/usr/lib/python3.12/multiprocessing/process.py", line 108, in run
    self._target(*self._args, **self._kwargs)
  File "/usr/local/lib/python3.12/dist-packages/vllm_omni/entrypoints/omni_stage.py", line 889, in _stage_worker
    stage_engine = OmniDiffusion(
                   ^^^^^^^^^^^^^^
  File "/usr/local/lib/python3.12/dist-packages/vllm_omni/entrypoints/omni_diffusion.py", line 97, in __init__
    raise ValueError(f"Unknown model type: {model_type}, architectures: {architectures}")
ValueError: Unknown model type: nextstep, architectures: ['LlamaForCausalLM']

Note that this error is only about the offline inference mode (not AsyncOmniDuffusion).

Test Plan

Run the NextStep example snippet in the doc after this patch

Instead of showing the error above, it should run fine.

Test Result

Pass on my side. There is no CI test for NextStep for now.

---

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)

[Copilot]

Pull request overview

Fixes offline NextStep example startup by ensuring OmniDiffusion selects the NextStep pipeline class even when model_class_name is pre-set by the offline script.

Changes:

• Use od_config.model_class_name as the pipeline_class for model_type == "nextstep" when provided.
• Preserve the existing default behavior (NextStep11Pipeline) when model_class_name is not provided.

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

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

[wtomin]

LGTM.

[fhfuih]

Close as this file is deleted after #1908 . The offline entrypoint also uses async engine, which does not have this error any more