A high-throughput and memory-efficient inference and serving engine for LLMs
Mirrored from https://github.com/tduncan-lum/Nvidia-Optimization-Stack---VLLM.git
Synced Nov 15, 22:15 UTC
[Bugfix] Add `mm_processor_kwargs` to chat-related protocols (#13644)
Roger Wang committed
Feb 20, 2025 at 22:04 UTC
a30c093502f0a671969ab561aacc80bc430f8ed6
1 file changed
+8
vllm/entrypoints/openai/protocol.py
+8
@@ -974,6 +974,10 @@ class EmbeddingChatRequest(OpenAIBaseModel):
974
description=("Additional kwargs to pass to the template renderer. "
975
"Will be accessible by the chat template."),
976
)
977
+ mm_processor_kwargs: Optional[Dict[str, Any]] = Field(
978
+ default=None,
979
+ description=("Additional kwargs to pass to the HF processor."),
980
+ )
981
priority: int = Field(
982
default=0,
983
description=(
@@ -1394,6 +1398,10 @@ class TokenizeChatRequest(OpenAIBaseModel):
1398
description=("Additional kwargs to pass to the template renderer. "
1399
"Will be accessible by the chat template."),
1400
)
1401
+ mm_processor_kwargs: Optional[Dict[str, Any]] = Field(
1402
+ default=None,
1403
+ description=("Additional kwargs to pass to the HF processor."),
1404
+ )
1405
1406
@model_validator(mode="before")
1407
@classmethod