[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:
@@ -17,7 +17,7 @@ from transformers import PretrainedConfig
|
||||
|
||||
from vllm.model_executor.layers.quantization import QuantizationConfig
|
||||
from vllm.multimodal import MULTIMODAL_REGISTRY
|
||||
from vllm.multimodal.inputs import MultiModalKwargsItems
|
||||
from vllm.multimodal.inputs import MultiModalKwargsItems, MultiModalUUIDDict
|
||||
from vllm.multimodal.parse import (ImageEmbeddingItems, ImageProcessorItems,
|
||||
MultiModalDataItems)
|
||||
from vllm.multimodal.processing import (MultiModalProcessingInfo,
|
||||
@@ -479,7 +479,7 @@ class H2OVLMultiModalProcessor(
|
||||
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]:
|
||||
# The processor logic is different for len(images) <= 1 vs > 1
|
||||
# Since the processing cache assumes that the processor output is
|
||||
@@ -491,7 +491,7 @@ class H2OVLMultiModalProcessor(
|
||||
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,
|
||||
)
|
||||
|
||||
return super()._cached_apply_hf_processor(
|
||||
@@ -499,7 +499,7 @@ class H2OVLMultiModalProcessor(
|
||||
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,
|
||||
)
|
||||
|
||||
|
||||
|
||||
Reference in New Issue
Block a user