[Core][Frontend] Support Passing Multimodal Processor Kwargs (#8657)

Signed-off-by: Alex-Brooks <Alex.Brooks@ibm.com>
2024-09-23 01:44:48 -06:00
parent d23679eb99
commit 9b8c8ba119
16 changed files with 590 additions and 117 deletions
--- a/vllm/entrypoints/llm.py
+++ b/vllm/entrypoints/llm.py
@@ -134,6 +134,7 @@ class LLM:
        max_seq_len_to_capture: int = 8192,
        disable_custom_all_reduce: bool = False,
        disable_async_output_proc: bool = False,
+        mm_processor_kwargs: Optional[Dict[str, Any]] = None,
        **kwargs,
    ) -> None:
        '''
@@ -174,6 +175,7 @@ class LLM:
            max_seq_len_to_capture=max_seq_len_to_capture,
            disable_custom_all_reduce=disable_custom_all_reduce,
            disable_async_output_proc=disable_async_output_proc,
+            mm_processor_kwargs=mm_processor_kwargs,
            **kwargs,
        )
        self.llm_engine = LLMEngine.from_engine_args(