diff --git a/vllm/renderers/base.py b/vllm/renderers/base.py index 2f10302c02..be5ecdf96f 100644 --- a/vllm/renderers/base.py +++ b/vllm/renderers/base.py @@ -105,7 +105,7 @@ class BaseRenderer(ABC, Generic[_T]): self._process_multimodal_async = make_async( self._process_multimodal, executor=self._mm_executor ) - if config.model_config.is_multimodal_model: + if mm_registry.supports_multimodal_inputs(config.model_config): mm_processor_cache = mm_registry.processor_cache_from_config(config) # Deep-copy the tokenizer so the multimodal processor gets its