[Misc] Rename MultiModalInputsV2 -> MultiModalInputs (#12244)

Signed-off-by: DarkLight1337 <[email protected]>

Author: DarkLight1337

docs/source/api/multimodal/inputs.md

@@ -43,7 +43,7 @@
 ```
 
 ```{eval-rst}
-.. autoclass:: vllm.multimodal.inputs.MultiModalInputsV2
+.. autoclass:: vllm.multimodal.inputs.MultiModalInputs
     :members:
     :show-inheritance:
 ```

vllm/inputs/data.py

@@ -9,7 +9,7 @@
 if TYPE_CHECKING:
     from vllm.multimodal import (MultiModalDataDict, MultiModalKwargs,
                                  MultiModalPlaceholderDict)
-    from vllm.multimodal.inputs import MultiModalInputsV2
+    from vllm.multimodal.inputs import MultiModalInputs
 
 
 class TextPrompt(TypedDict):
@@ -207,7 +207,7 @@ def token_inputs(
     return inputs
 
 
-DecoderOnlyInputs = Union[TokenInputs, "MultiModalInputsV2"]
+DecoderOnlyInputs = Union[TokenInputs, "MultiModalInputs"]
 """
 The inputs in :class:`~vllm.LLMEngine` before they are
 passed to the model executor.
@@ -222,14 +222,14 @@ class EncoderDecoderInputs(TypedDict):
 
     This specifies the required data for encoder-decoder models.
     """
-    encoder: Union[TokenInputs, "MultiModalInputsV2"]
+    encoder: Union[TokenInputs, "MultiModalInputs"]
     """The inputs for the encoder portion."""
 
-    decoder: Union[TokenInputs, "MultiModalInputsV2"]
+    decoder: Union[TokenInputs, "MultiModalInputs"]
     """The inputs for the decoder portion."""
 
 
-SingletonInputs = Union[TokenInputs, "MultiModalInputsV2"]
+SingletonInputs = Union[TokenInputs, "MultiModalInputs"]
 """
 A processed :class:`SingletonPrompt` which can be passed to
 :class:`vllm.sequence.Sequence`.
@@ -311,7 +311,7 @@ def multi_modal_hashes(self) -> List[str]:
             return inputs.get("multi_modal_hashes", [])
 
         if inputs["type"] == "multimodal":
-            # only the case when we use MultiModalInputsV2
+            # only the case when we use MultiModalInputs
             return inputs.get("mm_hashes", [])  # type: ignore[return-value]
 
         assert_never(inputs)  # type: ignore[arg-type]

vllm/inputs/preprocess.py

@@ -7,7 +7,7 @@
 from vllm.logger import init_logger
 from vllm.lora.request import LoRARequest
 from vllm.multimodal import MULTIMODAL_REGISTRY, MultiModalRegistry
-from vllm.multimodal.inputs import MultiModalDataDict, MultiModalInputsV2
+from vllm.multimodal.inputs import MultiModalDataDict, MultiModalInputs
 from vllm.prompt_adapter.request import PromptAdapterRequest
 from vllm.transformers_utils.tokenizer_group import BaseTokenizerGroup
 
@@ -247,7 +247,7 @@ def _process_multimodal(
         mm_data: MultiModalDataDict,
         mm_processor_kwargs: Optional[Mapping[str, object]],
         lora_request: Optional[LoRARequest],
-    ) -> MultiModalInputsV2:
+    ) -> MultiModalInputs:
         """
         Apply the model's multi-modal processor to a multi-modal prompt,
         returning the corresponding token IDs and metadata.
@@ -271,7 +271,7 @@ async def _process_multimodal_async(
         mm_data: MultiModalDataDict,
         mm_processor_kwargs: Optional[Mapping[str, object]],
         lora_request: Optional[LoRARequest],
-    ) -> MultiModalInputsV2:
+    ) -> MultiModalInputs:
         """Async version of :meth:`_process_multimodal`."""
         tokenizer_group = self.get_tokenizer_group()
         tokenizer = await tokenizer_group.get_lora_tokenizer_async(lora_request

vllm/model_executor/models/blip2.py

@@ -15,7 +15,7 @@
 from vllm.model_executor.sampling_metadata import SamplingMetadata
 from vllm.multimodal import MULTIMODAL_REGISTRY
 from vllm.multimodal.inputs import (MultiModalDataDict, MultiModalFieldConfig,
-                                    MultiModalInputsV2, MultiModalKwargs,
+                                    MultiModalInputs, MultiModalKwargs,
                                     NestedTensors, PlaceholderRange)
 from vllm.multimodal.parse import MultiModalDataItems
 from vllm.multimodal.processing import (BaseMultiModalProcessor,
@@ -490,7 +490,7 @@ def apply(
         prompt: Union[str, list[int]],
         mm_data: MultiModalDataDict,
         hf_processor_mm_kwargs: Mapping[str, object],
-    ) -> MultiModalInputsV2:
+    ) -> MultiModalInputs:
         result = super().apply(prompt, mm_data, hf_processor_mm_kwargs)
 
         # Only <image> tokens should be considered as placeholders,

vllm/model_executor/models/chameleon.py

@@ -29,7 +29,7 @@
 from vllm.model_executor.utils import set_weight_attrs
 from vllm.multimodal import MULTIMODAL_REGISTRY
 from vllm.multimodal.inputs import (MultiModalDataDict, MultiModalFieldConfig,
-                                    MultiModalInputsV2, MultiModalKwargs,
+                                    MultiModalInputs, MultiModalKwargs,
                                     NestedTensors, PlaceholderRange)
 from vllm.multimodal.parse import MultiModalDataItems
 from vllm.multimodal.processing import (BaseMultiModalProcessor,
@@ -159,7 +159,7 @@ def apply(
         prompt: Union[str, list[int]],
         mm_data: MultiModalDataDict,
         hf_processor_mm_kwargs: Mapping[str, object],
-    ) -> MultiModalInputsV2:
+    ) -> MultiModalInputs:
         result = super().apply(prompt, mm_data, hf_processor_mm_kwargs)
 
         # Only <image> tokens should be considered as placeholders,

vllm/model_executor/models/fuyu.py

@@ -31,7 +31,7 @@
 from vllm.model_executor.sampling_metadata import SamplingMetadata
 from vllm.multimodal import MULTIMODAL_REGISTRY
 from vllm.multimodal.inputs import (MultiModalDataDict, MultiModalFieldConfig,
-                                    MultiModalInputsV2, MultiModalKwargs,
+                                    MultiModalInputs, MultiModalKwargs,
                                     NestedTensors, PlaceholderRange)
 from vllm.multimodal.parse import (ImageProcessorItems, ImageSize,
                                    MultiModalDataItems)
@@ -232,7 +232,7 @@ def apply(
         prompt: Union[str, list[int]],
         mm_data: MultiModalDataDict,
         hf_processor_mm_kwargs: Mapping[str, object],
-    ) -> MultiModalInputsV2:
+    ) -> MultiModalInputs:
         result = super().apply(prompt, mm_data, hf_processor_mm_kwargs)
 
         # Only |SPEAKER| (image) tokens should be considered as placeholders,

vllm/model_executor/models/llava.py

@@ -24,7 +24,7 @@
 from vllm.model_executor.sampling_metadata import SamplingMetadata
 from vllm.multimodal import MULTIMODAL_REGISTRY
 from vllm.multimodal.inputs import (MultiModalDataDict, MultiModalFieldConfig,
-                                    MultiModalInputsV2, MultiModalKwargs,
+                                    MultiModalInputs, MultiModalKwargs,
                                     NestedTensors)
 from vllm.multimodal.parse import (ImageEmbeddingItems, ImageProcessorItems,
                                    ImageSize, MultiModalDataItems)
@@ -746,7 +746,7 @@ def apply(
         prompt: Union[str, list[int]],
         mm_data: MultiModalDataDict,
         hf_processor_mm_kwargs: Mapping[str, object],
-    ) -> MultiModalInputsV2:
+    ) -> MultiModalInputs:
         hf_config = self.info.get_hf_config()
         image_token_id = hf_config.image_token_index
 
@@ -805,7 +805,7 @@ def get_replacement_mantis(item_idx: int):
             for modality, placeholders in mm_placeholders.items()
         }
 
-        return MultiModalInputsV2(
+        return MultiModalInputs(
             type="multimodal",
             prompt=prompt,
             prompt_token_ids=prompt_ids,

vllm/model_executor/models/phi3v.py

@@ -31,7 +31,7 @@
 from vllm.model_executor.sampling_metadata import SamplingMetadata
 from vllm.multimodal import MULTIMODAL_REGISTRY
 from vllm.multimodal.inputs import (MultiModalDataDict, MultiModalFieldConfig,
-                                    MultiModalInputsV2, MultiModalKwargs,
+                                    MultiModalInputs, MultiModalKwargs,
                                     NestedTensors, PlaceholderRange)
 from vllm.multimodal.parse import (ImageEmbeddingItems, ImageProcessorItems,
                                    ImageSize, MultiModalDataItems)
@@ -484,7 +484,7 @@ def apply(
         prompt: Union[str, list[int]],
         mm_data: MultiModalDataDict,
         hf_processor_mm_kwargs: Mapping[str, object],
-    ) -> MultiModalInputsV2:
+    ) -> MultiModalInputs:
         result = super().apply(prompt, mm_data, hf_processor_mm_kwargs)
 
         # Only <|image|> tokens should be considered as placeholders,

vllm/model_executor/models/qwen2_audio.py

@@ -37,7 +37,7 @@
 from vllm.model_executor.sampling_metadata import SamplingMetadata
 from vllm.multimodal import MULTIMODAL_REGISTRY
 from vllm.multimodal.inputs import (MultiModalDataDict, MultiModalFieldConfig,
-                                    MultiModalInputsV2, MultiModalKwargs,
+                                    MultiModalInputs, MultiModalKwargs,
                                     NestedTensors, PlaceholderRange)
 from vllm.multimodal.parse import (AudioProcessorItems, MultiModalDataItems,
                                    MultiModalDataParser)
@@ -245,7 +245,7 @@ def apply(
         prompt: Union[str, list[int]],
         mm_data: MultiModalDataDict,
         hf_processor_mm_kwargs: Mapping[str, object],
-    ) -> MultiModalInputsV2:
+    ) -> MultiModalInputs:
         result = super().apply(prompt, mm_data, hf_processor_mm_kwargs)
 
         # Only <|AUDIO|> tokens should be considered as placeholders,

vllm/multimodal/inputs.py

@@ -491,7 +491,7 @@ def get_items(self, modality: str) -> Sequence[MultiModalKwargsItem]:
 """
 
 
-class MultiModalInputsV2(TypedDict):
+class MultiModalInputs(TypedDict):
     """
     Represents the outputs of
     :class:`vllm.multimodal.processing.BaseMultiModalProcessor`,