Enforce valid max_num_batched_tokens when disable_chunked_mm_input=True (#16447)

Signed-off-by: mgoin <mgoin64@gmail.com>
2025-04-11 02:09:52 -06:00
parent f7030df3be
commit aa3b3d76e0
3 changed files with 18 additions and 1 deletions
--- a/vllm/engine/arg_utils.py
+++ b/vllm/engine/arg_utils.py
@@ -1030,7 +1030,7 @@ class EngineArgs:
            action=StoreBoolean,
            default=EngineArgs.disable_chunked_mm_input,
            nargs="?",
-            const="False",
+            const="True",
            help="Disable multimodal input chunking attention for V1. "
            "If set to true and chunked prefill is enabled, we do not want to"
            " partially schedule a multimodal item. This ensures that if a "