[Frontend] Update the warning log when using VLLM_ALLOW_LONG_MAX_MODEL_LEN (#20904)

Signed-off-by: wang.yuqi <[email protected]>
Signed-off-by: Harry Mellor <[email protected]>
Co-authored-by: Harry Mellor <[email protected]>

Author: noooop

vllm/config/__init__.py

@@ -3021,16 +3021,20 @@ def _get_and_verify_max_len(
                 f"User-specified max_model_len ({max_model_len}) is greater "
                 f"than the derived max_model_len ({max_len_key}="
                 f"{derived_max_model_len} or model_max_length="
-                f"{model_max_length} in model's config.json). This may lead "
-                "to incorrect model outputs or CUDA errors.")
+                f"{model_max_length} in model's config.json).")
+            warning = (
+                "VLLM_ALLOW_LONG_MAX_MODEL_LEN must be used with extreme "
+                "caution. If the model uses relative position encoding (RoPE), "
+                "positions exceeding derived_max_model_len lead to nan. If the "
+                "model uses absolute position encoding, positions exceeding "
+                "derived_max_model_len will cause a CUDA array out-of-bounds "
+                "error.")
             if envs.VLLM_ALLOW_LONG_MAX_MODEL_LEN:
-                logger.warning(
-                    "%s Make sure the value is correct and within the "
-                    "model context size.", msg)
+                logger.warning_once("%s %s", msg, warning)
             else:
                 raise ValueError(
                     f"{msg} To allow overriding this maximum, set "
-                    "the env var VLLM_ALLOW_LONG_MAX_MODEL_LEN=1")
+                    f"the env var VLLM_ALLOW_LONG_MAX_MODEL_LEN=1. {warning}")
     return int(max_model_len)