[Refactor] Define MultiModalKwargsItems separate from MultiModalKwargs (#23053)

Signed-off-by: DarkLight1337 <tlleungac@connect.ust.hk>
This commit is contained in:
Cyrus Leung
2025-08-18 09:52:00 +00:00
committed by GitHub
parent 5c79b0d648
commit 27e8d1ea3e
77 changed files with 431 additions and 383 deletions
+3 -3
View File
@@ -41,7 +41,7 @@ from vllm.model_executor.layers.quantization import QuantizationConfig
from vllm.model_executor.layers.vocab_parallel_embedding import (
ParallelLMHead, VocabParallelEmbedding)
from vllm.model_executor.sampling_metadata import SamplingMetadata
from vllm.multimodal import MULTIMODAL_REGISTRY, MultiModalKwargs
from vllm.multimodal import MULTIMODAL_REGISTRY, MultiModalKwargsItems
from vllm.multimodal.inputs import (MultiModalDataDict, MultiModalFieldConfig,
MultiModalInputs, PlaceholderRange)
from vllm.multimodal.parse import ImageProcessorItems, MultiModalDataItems
@@ -237,7 +237,7 @@ class MultiModalProcessor(BaseMultiModalProcessor[MultiModalProcessingInfo]):
self,
mm_items: MultiModalDataItems,
hf_processor_mm_kwargs: Mapping[str, object],
out_mm_kwargs: MultiModalKwargs,
out_mm_kwargs: MultiModalKwargsItems,
):
"""
Given the original multi-modal items for this modality
@@ -372,7 +372,7 @@ class MultiModalProcessor(BaseMultiModalProcessor[MultiModalProcessingInfo]):
mm_tokens_per_modality["num_image_patches"]
) if "num_image_patches" in mm_tokens_per_modality else None
processed_data['num_image_patches'] = num_image_patches
mm_kwargs = MultiModalKwargs.from_hf_inputs(
mm_kwargs = MultiModalKwargsItems.from_hf_inputs(
processed_data,
self._get_mm_fields_config(processed_data, hf_processor_mm_kwargs,
num_image_patches),