[Renderer] Move MM Hash parsing into Renderer (#34711)

Signed-off-by: DarkLight1337 <tlleungac@connect.ust.hk>
This commit is contained in:
Cyrus Leung
2026-02-17 19:18:55 -08:00
committed by GitHub
parent df3f537a66
commit a0d8d944e2
16 changed files with 254 additions and 217 deletions
+5 -4
View File
@@ -16,11 +16,12 @@ from transformers import PretrainedConfig
from vllm.model_executor.layers.quantization import QuantizationConfig
from vllm.multimodal import MULTIMODAL_REGISTRY
from vllm.multimodal.inputs import MultiModalKwargsItems, MultiModalUUIDDict
from vllm.multimodal.inputs import MultiModalKwargsItems
from vllm.multimodal.parse import (
ImageEmbeddingItems,
ImageProcessorItems,
MultiModalDataItems,
MultiModalUUIDItems,
)
from vllm.multimodal.processing.processor import (
MultiModalProcessingInfo,
@@ -491,9 +492,9 @@ class H2OVLMultiModalProcessor(BaseInternVLMultiModalProcessor[H2OVLProcessingIn
self,
prompt: str | list[int],
mm_data_items: MultiModalDataItems,
mm_uuid_items: MultiModalUUIDItems | None,
hf_processor_mm_kwargs: Mapping[str, object],
tokenization_kwargs: Mapping[str, object],
mm_uuids: MultiModalUUIDDict | None = 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
@@ -503,17 +504,17 @@ class H2OVLMultiModalProcessor(BaseInternVLMultiModalProcessor[H2OVLProcessingIn
return self._apply_hf_processor(
prompt=prompt,
mm_data_items=mm_data_items,
mm_uuid_items=mm_uuid_items,
hf_processor_mm_kwargs=hf_processor_mm_kwargs,
tokenization_kwargs=tokenization_kwargs,
mm_uuids=mm_uuids,
)
return super()._cached_apply_hf_processor(
prompt=prompt,
mm_data_items=mm_data_items,
mm_uuid_items=mm_uuid_items,
hf_processor_mm_kwargs=hf_processor_mm_kwargs,
tokenization_kwargs=tokenization_kwargs,
mm_uuids=mm_uuids,
)