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

Signed-off-by: DarkLight1337 <tlleungac@connect.ust.hk>
This commit is contained in:
Cyrus Leung
2025-03-14 00:59:56 -07:00
committed by GitHub
parent 09269b3127
commit 601bd3268e
27 changed files with 121 additions and 141 deletions
@@ -16,8 +16,7 @@ from vllm.model_executor.layers.sampler import SamplerOutput, get_sampler
from vllm.model_executor.models.clip import CLIPVisionModel
from vllm.model_executor.sampling_metadata import SamplingMetadata
from vllm.multimodal import MULTIMODAL_REGISTRY
from vllm.multimodal.inputs import (MultiModalFieldConfig, MultiModalKwargs,
NestedTensors)
from vllm.multimodal.inputs import MultiModalFieldConfig, MultiModalKwargs
from vllm.multimodal.parse import (ImageSize, MultiModalDataItems,
VideoEmbeddingItems, VideoProcessorItems)
from vllm.multimodal.processing import (BaseMultiModalProcessor,
@@ -27,7 +26,7 @@ from vllm.multimodal.profiling import BaseDummyInputsBuilder, ProcessorInputs
from vllm.sequence import IntermediateTensors
from vllm.utils import is_list_of
from .interfaces import SupportsMultiModal, SupportsPP
from .interfaces import MultiModalEmbeddings, SupportsMultiModal, SupportsPP
from .llava import init_vision_tower_for_llava
from .siglip import SiglipVisionModel
from .utils import (AutoWeightsLoader, init_vllm_registered_model,
@@ -421,8 +420,7 @@ class LlavaNextVideoForConditionalGeneration(nn.Module, SupportsMultiModal,
f"Unsupported type of video input {type(video_pixels)}")
def get_multimodal_embeddings(
self, **kwargs
) -> Union[list[torch.Tensor], torch.Tensor, tuple[torch.Tensor, ...]]:
self, **kwargs: object) -> Optional[MultiModalEmbeddings]:
video_input = self._parse_and_validate_video_input(**kwargs)
if video_input is None:
return None
@@ -432,7 +430,7 @@ class LlavaNextVideoForConditionalGeneration(nn.Module, SupportsMultiModal,
def get_input_embeddings(
self,
input_ids: torch.Tensor,
multimodal_embeddings: Optional[NestedTensors] = None,
multimodal_embeddings: Optional[MultiModalEmbeddings] = None,
) -> torch.Tensor:
inputs_embeds = self.language_model.get_input_embeddings(input_ids)
if multimodal_embeddings is not None: