[Misc] Clean up type annotation for SupportsMultiModal (#14794)

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

vllm/model_executor/models/llava_next.py

@@ -16,12 +16,12 @@ from vllm.config import VllmConfig
 from vllm.model_executor.layers.sampler import SamplerOutput, get_sampler
 from vllm.model_executor.sampling_metadata import SamplingMetadata
 from vllm.multimodal import MULTIMODAL_REGISTRY
-from vllm.multimodal.inputs import MultiModalFieldConfig, NestedTensors
+from vllm.multimodal.inputs import MultiModalFieldConfig
 from vllm.multimodal.parse import ImageSize
 from vllm.sequence import IntermediateTensors
 
 from .clip import CLIPVisionModel
-from .interfaces import SupportsMultiModal, SupportsPP
+from .interfaces import MultiModalEmbeddings, SupportsMultiModal, SupportsPP
 from .llava import (BaseLlavaMultiModalProcessor, BaseLlavaProcessingInfo,
                     LlavaDummyInputsBuilder, LlavaLikeConfig,
                     LlavaMultiModalProjector, init_vision_tower_for_llava)
@@ -480,8 +480,7 @@ class LlavaNextForConditionalGeneration(nn.Module, SupportsMultiModal,
         ]
 
     def get_multimodal_embeddings(
-        self, **kwargs
-    ) -> Union[list[torch.Tensor], torch.Tensor, tuple[torch.Tensor, ...]]:
+            self, **kwargs: object) -> Optional[MultiModalEmbeddings]:
         image_input = self._parse_and_validate_image_input(**kwargs)
         if image_input is None:
             return None
@@ -491,7 +490,7 @@ class LlavaNextForConditionalGeneration(nn.Module, SupportsMultiModal,
     def get_input_embeddings(
         self,
         input_ids: torch.Tensor,
-        multimodal_embeddings: Optional[NestedTensors] = None,
+        multimodal_embeddings: Optional[MultiModalEmbeddings] = None,
     ) -> torch.Tensor:
 
         if multimodal_embeddings is None: