[VLM] Avoid unnecessary dummy multimodal data during processing (#16416)

Signed-off-by: DarkLight1337 <tlleungac@connect.ust.hk>

vllm/model_executor/models/pixtral.py

@@ -32,13 +32,14 @@ from vllm.model_executor.layers.sampler import SamplerOutput, get_sampler
 from vllm.model_executor.model_loader.weight_utils import default_weight_loader
 from vllm.model_executor.sampling_metadata import SamplingMetadata
 from vllm.multimodal import MULTIMODAL_REGISTRY, MultiModalKwargs
-from vllm.multimodal.inputs import MultiModalFieldConfig, NestedTensors
+from vllm.multimodal.inputs import (MultiModalDataDict, MultiModalFieldConfig,
+                                    NestedTensors)
 from vllm.multimodal.parse import (ImageProcessorItems, ImageSize,
                                    MultiModalDataItems)
 from vllm.multimodal.processing import (BaseMultiModalProcessor,
                                         BaseProcessingInfo, PromptReplacement,
                                         PromptUpdate, PromptUpdateDetails)
-from vllm.multimodal.profiling import BaseDummyInputsBuilder, ProcessorInputs
+from vllm.multimodal.profiling import BaseDummyInputsBuilder
 from vllm.sequence import IntermediateTensors
 from vllm.transformers_utils.tokenizer import (MistralTokenizer,
                                                cached_tokenizer_from_config)
@@ -203,28 +204,26 @@ class PixtralProcessingInfo(BaseProcessingInfo):
 
 class PixtralDummyInputsBuilder(BaseDummyInputsBuilder[PixtralProcessingInfo]):
 
-    def get_dummy_processor_inputs(
+    def get_dummy_text(self, mm_counts: Mapping[str, int]) -> str:
+        return ""
+
+    def get_dummy_mm_data(
         self,
         seq_len: int,
         mm_counts: Mapping[str, int],
-    ) -> ProcessorInputs:
+    ) -> MultiModalDataDict:
         num_images = mm_counts.get("image", 0)
 
         target_width, target_height = \
             self.info.get_image_size_with_most_features()
 
-        mm_data = {
+        return {
             "image":
             self._get_dummy_images(width=target_width,
                                    height=target_height,
                                    num_images=num_images)
         }
 
-        return ProcessorInputs(
-            prompt_text="",
-            mm_data=mm_data,
-        )
-
 
 class PixtralMultiModalProcessor(BaseMultiModalProcessor[PixtralProcessingInfo]
                                  ):