[BugFix] Normalize diffusion parallel CLI overrides into stage parallel_config

[xiaohajiayou]

Purpose

Diffusion stages in omni multi-stage configs consume parallel settings from engine_args.parallel_config, but CLI overrides were being applied as flat top-level engine_args fields.

This could produce a resolved stage config where:

• diffusion parallel_config still kept the old values
• CLI overrides appeared only at the top level
• LLM and diffusion stages did not handle the same override fields consistently

This change fixes that at stage-config materialization time:

• for diffusion stages, fields defined by DiffusionParallelConfig are normalized into engine_args.parallel_config
• those fields are no longer duplicated at the top level for diffusion stages
• for LLM stages, the existing top-level override behavior is preserved

Test Plan

• Add a unit test covering diffusion stages with an existing parallel_config, verifying that CLI overrides replace the nested values
• Add a unit test covering diffusion stages without an existing parallel_config, verifying that the nested config is created from CLI overrides
• Add a unit test covering LLM stages, verifying that shared parallel fields remain top-level and do not create parallel_config
• cmd

pytest -q tests/test_config_factory.py -k "diffusion_parallel_overrides or llm_parallel_overrides or deploy_override_fields_include_deploy_schema_fields"
pytest -q tests/test_arg_utils.py -k "nullify_stage_engine_defaults_resets_inherited_defaults or non_override_flags_keep_real_defaults_after_nullify"

Test Result

-- Docs: https://docs.pytest.org/en/stable/how-to/capture-warnings.html
--- Running Summary
4 passed, 104 deselected, 18 warnings in 1.57s



-- Docs: https://docs.pytest.org/en/stable/how-to/capture-warnings.html
--- Running Summary
2 passed, 22 deselected, 18 warnings in 1.63s
(vllm-omni) root@autodl-container-wpn6tqcvq6-04abd7f3:~/vllm-omni# 

---

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)

[chatgpt-codex-connector]

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: 3ecb6edb3d

ℹ️ About Codex in GitHub

Codex has been enabled to automatically 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 👍.

When you sign up for Codex through ChatGPT, Codex can also answer questions or update the PR, like "@codex address that feedback".

[chatgpt-codex-connector]

Preserve plain dict parallel_config values

When a diffusion stage already has engine_args.parallel_config as a normal dict (the common result of parsed YAML/registry deploy config, and also what the new tests construct), this unconditional to_dict() call goes through the OmegaConf wrapper, which only accepts OmegaConf containers and raises for plain Python dicts. That means any diffusion deploy YAML that includes a nested parallel_config now fails during StageConfig.to_omegaconf() before overrides can be applied; handle mappings/dataclass instances directly or only call to_dict() for OmegaConf objects.

Useful? React with 👍 / 👎.

[chatgpt-codex-connector]

Do not treat parser defaults as diffusion overrides

This moves every non-None DiffusionParallelConfig field from runtime_overrides into the nested config, but the CLI nullifier only covers deploy_override_field_names() and does not include several diffusion-only fields; for example vllm_omni/entrypoints/cli/serve.py still defaults --cfg-parallel-size and --vae-patch-parallel-size to 1 and --ulysses-mode to "strict". Starting serve without typing those flags will therefore overwrite any deploy YAML parallel_config values for those fields with parser defaults, reintroducing the deploy-precedence bug this change is trying to avoid.

Useful? React with 👍 / 👎.

[xiaohajiayou]

Good catch. This was a real issue.

I previously did not account for the omni deploy-YAML path where one of the stages can be a diffusion stage, so these diffusion-only parallel fields also need to be included in the deploy override whitelist.

This is now addressed by adding the diffusion parallel_config override fields to StageDeployConfig, so they are included in deploy_override_field_names() and go through the existing nullifier flow. As a result, untyped parser defaults for these diffusion-only fields no longer override deploy YAML values, while explicit CLI values still do.

Also added/updated tests around the deploy-override field set and nullifier behavior.

[xiaohajiayou]

This pr is closed and now handled by #3483.