[Core] Simplify and unify mm uuid handling & auto-generated mm hash overrides processing. (#24271)

Signed-off-by: Chenheli Hua <huachenheli@outlook.com>
This commit is contained in:
Chenheli Hua
2025-09-09 21:36:09 -07:00
committed by GitHub
parent 0efdb5c3ba
commit 009d689b0c
13 changed files with 110 additions and 127 deletions
+4 -3
View File
@@ -31,7 +31,8 @@ from vllm.model_executor.models.whisper import WhisperEncoder
from vllm.model_executor.sampling_metadata import SamplingMetadata
from vllm.multimodal import MULTIMODAL_REGISTRY
from vllm.multimodal.inputs import (MultiModalDataDict, MultiModalFieldConfig,
MultiModalKwargsItems, NestedTensors)
MultiModalKwargsItems, MultiModalUUIDDict,
NestedTensors)
from vllm.multimodal.parse import (AudioProcessorItems, MultiModalDataItems,
MultiModalDataParser)
from vllm.multimodal.processing import (BaseMultiModalProcessor,
@@ -290,14 +291,14 @@ class VoxtralMultiModalProcessor(BaseMultiModalProcessor[VoxtralProcessingInfo]
mm_data_items: MultiModalDataItems,
hf_processor_mm_kwargs: Mapping[str, object],
tokenization_kwargs: Mapping[str, object],
mm_hash_overrides: Optional[dict[str, list[str]]] = None,
mm_uuids: Optional[MultiModalUUIDDict] = None,
) -> tuple[list[int], MultiModalProcessingInfo, bool]:
prompt_ids, mm_info, _ = super()._cached_apply_hf_processor(
prompt=prompt,
mm_data_items=mm_data_items,
hf_processor_mm_kwargs=hf_processor_mm_kwargs,
tokenization_kwargs=tokenization_kwargs,
mm_hash_overrides=mm_hash_overrides,
mm_uuids=mm_uuids,
)
# NOTE: The tokens are already inserted by the chat template