[VLM][Core] Support profiling with multiple multi-modal inputs per prompt (#7126)

2024-08-15 01:55:42 +08:00
parent 70b746efcf
commit 3f674a49b5
38 changed files with 572 additions and 216 deletions
--- a/vllm/model_executor/model_loader/loader.py
+++ b/vllm/model_executor/model_loader/loader.py
@@ -133,7 +133,7 @@ def _get_model_initialization_kwargs(

    if supports_multimodal(model_class):
        if multimodal_config is None:
-            raise ValueError("Provide vision related configurations "
+            raise ValueError("Provide multi-modal related configurations "
                             "through LLM entrypoint or engine arguments.")

        extra_kwargs["multimodal_config"] = multimodal_config