Fix "Add online serving to Stable Audio Diffusion and introduce v1/audio/generate endpoint" (#1794)

Signed-off-by: Ekagra Ranjan <3116519+ekagra-ranjan@users.noreply.github.com>
Signed-off-by: Yueqian Lin <linyueqian@outlook.com>
Co-authored-by: Copilot <175728472+Copilot@users.noreply.github.com>
Co-authored-by: Yueqian Lin <70319226+linyueqian@users.noreply.github.com>
Co-authored-by: Yueqian Lin <linyueqian@outlook.com>

vllm_omni/diffusion/models/stable_audio/pipeline_stable_audio.py

@@ -360,7 +360,6 @@ class StableAudioPipeline(nn.Module, SupportAudioOutput, DiffusionPipelineProfil
         negative_prompt: str | list[str] | None = None,
         audio_end_in_s: float | None = None,
         audio_start_in_s: float = 0.0,
-        num_inference_steps: int = 100,
         guidance_scale: float = 7.0,
         num_waveforms_per_prompt: int = 1,
         generator: torch.Generator | list[torch.Generator] | None = None,
@@ -378,7 +377,6 @@ class StableAudioPipeline(nn.Module, SupportAudioOutput, DiffusionPipelineProfil
             negative_prompt: Negative prompt for CFG
             audio_end_in_s: Audio end time in seconds (max ~47s for stable-audio-open-1.0)
             audio_start_in_s: Audio start time in seconds
-            num_inference_steps: Number of denoising steps
             guidance_scale: CFG scale
             num_waveforms_per_prompt: Number of audio outputs per prompt
             generator: Random generator for reproducibility
@@ -398,8 +396,12 @@ class StableAudioPipeline(nn.Module, SupportAudioOutput, DiffusionPipelineProfil
             negative_prompt = None
         elif req.prompts:
             negative_prompt = ["" if isinstance(p, str) else (p.get("negative_prompt") or "") for p in req.prompts]
-
-        num_inference_steps = req.sampling_params.num_inference_steps or num_inference_steps
+        num_inference_steps = req.sampling_params.num_inference_steps
+        if num_inference_steps is None:
+            num_inference_steps = 100  # Default steps
+        num_inference_steps = req.sampling_params.num_inference_steps
+        if num_inference_steps is None:
+            num_inference_steps = 50  # Default steps
         if req.sampling_params.guidance_scale_provided:
             guidance_scale = req.sampling_params.guidance_scale