[VLM][Bugfix] Enable specifying prompt target via index (#14038)

vllm/model_executor/models/blip2.py

@@ -19,8 +19,8 @@ from vllm.multimodal.inputs import (MultiModalFieldConfig, MultiModalKwargs,
                                     NestedTensors)
 from vllm.multimodal.parse import MultiModalDataItems
 from vllm.multimodal.processing import (BaseMultiModalProcessor,
-                                        BaseProcessingInfo, PromptInsertion,
-                                        PromptUpdate)
+                                        BaseProcessingInfo, PromptIndexTargets,
+                                        PromptInsertion, PromptUpdate)
 from vllm.multimodal.profiling import BaseDummyInputsBuilder, ProcessorInputs
 from vllm.sequence import IntermediateTensors
 
@@ -490,7 +490,7 @@ class Blip2MultiModalProcessor(BaseMultiModalProcessor[Blip2ProcessingInfo]):
         return [
             PromptInsertion(
                 modality="image",
-                target="",
+                target=PromptIndexTargets.start(),
                 insertion=image_tokens,
             )
         ]