[Bug][Hunyuanimage 3.0] fix different AR encode behavior between online and offline (#3500)

Signed-off-by: dengyunyang <584797741@qq.com>

vllm_omni/entrypoints/openai/serving_chat.py

@@ -2129,6 +2129,7 @@ class OmniOpenAIServingChat(OpenAIServingChat, AudioMixin):
         extra_body: dict[str, Any],
         reference_images: list[Image.Image],
         gen_params: OmniDiffusionSamplingParams,
+        tokenizer: Any = None,
     ) -> tuple[OmniTextPrompt, list[Any]]:
         """Build the shared multistage generation prompt and stage params."""
         stage_configs = getattr(engine, "stage_configs", None) or []
@@ -2159,16 +2160,26 @@ class OmniOpenAIServingChat(OpenAIServingChat, AudioMixin):
             else:
                 engine_prompt_data = {"image": reference_images}
 
+        prompt_token_ids: list[int] | None = None
         if bot_task:
-            from vllm_omni.diffusion.models.hunyuan_image3.prompt_utils import build_prompt
+            from vllm_omni.diffusion.models.hunyuan_image3.prompt_utils import (
+                build_prompt,
+                build_prompt_tokens,
+            )
+
+            if tokenizer is not None:
+                prompt_token_ids = build_prompt_tokens(prompt, tokenizer, task=bot_task)
+            else:
+                prompt = build_prompt(prompt, task=bot_task)
 
-            prompt = build_prompt(prompt, task=bot_task)
             if reference_images and len(reference_images) == 1:
                 engine_prompt_data = {"image": reference_images[0]}
                 modalities = ["image"]
 
         engine_prompt: OmniTextPrompt = {"prompt": prompt}
         engine_prompt["modalities"] = modalities
+        if prompt_token_ids is not None:
+            engine_prompt["prompt_token_ids"] = prompt_token_ids
         if negative_prompt is not None:
             engine_prompt["negative_prompt"] = negative_prompt
 
@@ -2337,12 +2348,20 @@ class OmniOpenAIServingChat(OpenAIServingChat, AudioMixin):
             diffusion_engine = cast(AsyncOmni, engine)
             stage_configs = getattr(diffusion_engine, "stage_configs", None) or []
             if len(stage_configs) > 1:
+                tokenizer = None
+                get_tok = getattr(diffusion_engine, "get_tokenizer", None)
+                if get_tok is not None:
+                    try:
+                        tokenizer = await get_tok()
+                    except Exception as exc:
+                        logger.warning("get_tokenizer failed: %s", exc)
                 engine_prompt, sampling_params_list = self._build_multistage_generation_inputs(
                     engine=diffusion_engine,
                     prompt=prompt,
                     extra_body=extra_body,
                     reference_images=pil_images,
                     gen_params=gen_params,
+                    tokenizer=tokenizer,
                 )
             else:
                 engine_prompt = gen_prompt