Skip to content

[Bugfix] Fix request batching regression on Ming-flash-omni ImageGen #4837

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
Gaohan123 merged 3 commits into from
Jul 3, 2026
Merged

[Bugfix] Fix request batching regression on Ming-flash-omni ImageGen #4837

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
3 commits
Select commit Hold shift + click to select a range
a14c43a
fix MingImagePipeline diffusion output type
yuanheng-zhao Jul 2, 2026
52b538d
fix orchestrator branch
yuanheng-zhao Jul 2, 2026
78cc4dc
Merge branch 'main' into fix/request-batching-ming-imagegen
Gaohan123 Jul 3, 2026
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
12 changes: 3 additions & 9 deletions vllm_omni/diffusion/models/ming_flash_omni/pipeline_ming_imagegen.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -446,7 +446,7 @@ def forward(self, req: DiffusionRequestBatch) -> DiffusionOutput:
None if ref_latent is None else tuple(ref_latent.shape),
)
try:
outputs = super().forward(
outputs: DiffusionOutput = super().forward(
z_req,
prompt=None,
height=height,
Expand All @@ -462,13 +462,7 @@ def forward(self, req: DiffusionRequestBatch) -> DiffusionOutput:
finally:
set_forward_context_ref_latent(None)

output = outputs[0]
if hasattr(output, "output") and output.output is not None:
raw = output.output
elif hasattr(output, "images"):
raw = output.images
else:
raw = output
raw = outputs.output
if not isinstance(raw, torch.Tensor):
raise RuntimeError(f"ZImagePipeline returned non-tensor output: {type(raw).__name__}")
if logger.isEnabledFor(logging.DEBUG):
Expand All @@ -478,7 +472,7 @@ def forward(self, req: DiffusionRequestBatch) -> DiffusionOutput:
raw.float().min().item(),
raw.float().max().item(),
)
return DiffusionOutput(output=raw)
return outputs


# ----------------------------------------------------------------------
Expand Down
2 changes: 1 addition & 1 deletion vllm_omni/engine/orchestrator.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -1287,7 +1287,7 @@ async def _forward_to_next_stage(
[req_id, *self._cfg_tracker.cleanup_parent(req_id)],
)
return
if already_submitted and len(diffusion_prompt) == 1:
if len(diffusion_prompt) == 1:

@yJader yJader Jul 2, 2026

Copy link
Copy Markdown
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

already_submitted is only used here to distinguish the orchestrator submission path (submit_update vs submit_initial). For diffusion, both paths eventually enter the same diffusion client API and then DiffusionEngine.step/step_streaming -> add_request, so DiffusionEngine does not attach different semantics to initial vs update submissions.

Since diffusion StagePool rejects list prompts in both submit_initial and submit_update, singleton-list normalization should not depend on already_submitted. Applying it consistently makes the diffusion payload contract clearer and avoids the initial path still passing a list-shaped payload.

diffusion_prompt = diffusion_prompt[0]
else:
diffusion_prompt = req_state.prompt
Expand Down
Loading