[Misc] Automatically resolve HF processor init kwargs (#22005)

Signed-off-by: DarkLight1337 <tlleungac@connect.ust.hk>
2025-08-01 13:44:10 +08:00
parent ad57f23f6a
commit 82de9b9d46
40 changed files with 334 additions and 727 deletions
--- a/vllm/model_executor/models/mllama4.py
+++ b/vllm/model_executor/models/mllama4.py
@@ -533,7 +533,7 @@ class Mllama4ProcessingInfo(BaseProcessingInfo):

    def get_hf_processor(self, **kwargs: object) -> Llama4Processor:
        return self.ctx.get_hf_processor(Llama4Processor,
-                                         use_fast=True,
+                                         use_fast=kwargs.pop("use_fast", True),
                                         **kwargs)

    def get_supported_mm_limits(self) -> Mapping[str, Optional[int]]: