[Bugfix] Fix basic models tests hanging due to mm processor creation (vllm-project#22571)

Signed-off-by: Isotr0py <[email protected]>
Signed-off-by: Xiao Yu <[email protected]>

Author: Isotr0py

vllm/multimodal/registry.py

@@ -138,8 +138,8 @@ def supports_multimodal_inputs(self, model_config: "ModelConfig") -> bool:
         if not model_config.is_multimodal_model:
             return False
 
-        processor = self.create_processor(model_config, disable_cache=False)
-        supported_modalities = processor.info.get_supported_mm_limits()
+        info = self._create_processing_info(model_config, tokenizer=None)
+        supported_modalities = info.get_supported_mm_limits()
 
         mm_config = model_config.get_multimodal_config()
 
@@ -278,6 +278,26 @@ def _get_model_cls(self, model_config: "ModelConfig"):
         model_cls, _ = get_model_architecture(model_config)
         return model_cls
 
+    def _create_processing_ctx(
+        self,
+        model_config: "ModelConfig",
+        tokenizer: Optional[AnyTokenizer] = None,
+    ) -> InputProcessingContext:
+        if tokenizer is None and not model_config.skip_tokenizer_init:
+            tokenizer = cached_tokenizer_from_config(model_config)
+        return InputProcessingContext(model_config, tokenizer)
+
+    def _create_processing_info(
+        self,
+        model_config: "ModelConfig",
+        *,
+        tokenizer: Optional[AnyTokenizer] = None,
+    ) -> BaseProcessingInfo:
+        model_cls = self._get_model_cls(model_config)
+        factories = self._processor_factories[model_cls]
+        ctx = self._create_processing_ctx(model_config, tokenizer)
+        return factories.info(ctx)
+
     def create_processor(
         self,
         model_config: "ModelConfig",
@@ -291,15 +311,13 @@ def create_processor(
         if not model_config.is_multimodal_model:
             raise ValueError(f"{model_config.model} is not a multimodal model")
 
-        if tokenizer is None and not model_config.skip_tokenizer_init:
-            tokenizer = cached_tokenizer_from_config(model_config)
         if disable_cache is None:
             disable_cache = not model_config.enable_mm_processor_cache
 
         model_cls = self._get_model_cls(model_config)
         factories = self._processor_factories[model_cls]
 
-        ctx = InputProcessingContext(model_config, tokenizer)
+        ctx = self._create_processing_ctx(model_config, tokenizer)
         cache = None if disable_cache else self._get_processor_cache(
             model_config)