[Misc] Remove unnecessary decode call (vllm-project#12833)

Signed-off-by: Louis Ulmer <[email protected]>

Author: DarkLight1337

vllm/inputs/preprocess.py

@@ -260,8 +260,6 @@ def _process_multimodal(
         mm_processor = self.mm_registry.create_processor(
             self.model_config, tokenizer)
 
-        if isinstance(prompt, list):
-            prompt = tokenizer.decode(prompt)
         if mm_processor_kwargs is None:
             mm_processor_kwargs = {}