Allow Gemma3 to take image embeddings (#28483)

Signed-off-by: tingtinggithub <streamttt@gmail.com>

vllm/v1/engine/processor.py

@@ -14,6 +14,7 @@ from vllm.lora.request import LoRARequest
 from vllm.multimodal import MULTIMODAL_REGISTRY, MultiModalRegistry
 from vllm.multimodal.cache import processor_cache_from_config
 from vllm.multimodal.inputs import MultiModalFeatureSpec, MultiModalUUIDDict
+from vllm.multimodal.parse import MultiModalDataParser
 from vllm.multimodal.processing import EncDecMultiModalProcessor
 from vllm.multimodal.utils import argsort_mm_positions
 from vllm.pooling_params import PoolingParams
@@ -340,7 +341,12 @@ class Processor:
 
         mm_uuids: dict[str, list[str | None] | str] = {}
         for modality, data in mm_data.items():
-            n = len(data) if isinstance(data, list) else 1
+            # Hash each item for embedding inputs.
+            n = (
+                len(data)
+                if isinstance(data, list) or MultiModalDataParser.is_embeddings(data)
+                else 1
+            )
             mm_uuids[modality] = [f"{request_id}-{modality}-{i}" for i in range(n)]
         return mm_uuids