[Model][VLM] Add Qwen2.5-Omni model support (thinker only) (#15130)

Signed-off-by: fyabc <suyang.fy@alibaba-inc.com>
Signed-off-by: Roger Wang <ywang@roblox.com>
Co-authored-by: Roger Wang <136131678+ywang96@users.noreply.github.com>
Co-authored-by: Roger Wang <ywang@roblox.com>
Co-authored-by: Xiong Wang <wangxiongts@163.com>

vllm/transformers_utils/processor.py

@@ -111,6 +111,55 @@ def cached_processor_from_config(
     )
 
 
+def get_feature_extractor(
+    processor_name: str,
+    *args: Any,
+    trust_remote_code: bool = False,
+    **kwargs: Any,
+):
+    """Load an audio feature extractor for the given model name 
+    via HuggingFace."""
+    # don't put this import at the top level
+    # it will call torch.cuda.device_count()
+    from transformers import AutoFeatureExtractor
+    from transformers.feature_extraction_utils import FeatureExtractionMixin
+    try:
+        feature_extractor = AutoFeatureExtractor.from_pretrained(
+            processor_name,
+            *args,
+            trust_remote_code=trust_remote_code,
+            **kwargs)
+    except ValueError as e:
+        # If the error pertains to the processor class not existing or not
+        # currently being imported, suggest using the --trust-remote-code flag.
+        # Unlike AutoTokenizer, AutoImageProcessor does not separate such errors
+        if not trust_remote_code:
+            err_msg = (
+                "Failed to load the feature extractor. If the feature "
+                "extractor is a custom extractor not yet available in the "
+                "HuggingFace transformers library, consider setting "
+                "`trust_remote_code=True` in LLM or using the "
+                "`--trust-remote-code` flag in the CLI.")
+            raise RuntimeError(err_msg) from e
+        else:
+            raise e
+    return cast(FeatureExtractionMixin, feature_extractor)
+
+
+cached_get_feature_extractor = lru_cache(get_feature_extractor)
+
+
+def cached_feature_extractor_from_config(
+    model_config: "ModelConfig",
+    **kwargs: Any,
+):
+    return cached_get_feature_extractor(
+        model_config.model,
+        trust_remote_code=model_config.trust_remote_code,
+        **_merge_mm_kwargs(model_config, **kwargs),
+    )
+
+
 def get_image_processor(
     processor_name: str,
     *args: Any,