A high-throughput and memory-efficient inference and serving engine for LLMs
Mirrored from https://github.com/tduncan-lum/Nvidia-Optimization-Stack---VLLM.git
Synced Nov 12, 06:00 UTC
[Misc] Throw error on unintended access to scheduler_config.max_model_len (#29771)
Signed-off-by: Wei Wei <wwei6@meta.com>
Wei Wei committed
Dec 1, 2025 at 18:58 UTC
fc95521ba59f4800c8afa7fcdcebc3cf08dd197c
1 file changed
+5
vllm/config/scheduler.py
+5
@@ -274,3 +274,8 @@ class SchedulerConfig:
274
)
275
276
return self
277
+
278
+ def __getattribute__(self, name: str) -> Any:
279
+ if name == "max_model_len" or name == "is_encoder_decoder":
280
+ raise AttributeError(f"{name} is an init-only parameter. ")
281
+ return object.__getattribute__(self, name)