@tduncan / Nvidia-Optimization-Stack-VLLM Mirror · read-only / commits / b1340f9d55
Fork 0

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 16, 00:45 UTC

[V1] Bugfix: Validate Model Input Length (#12600)

SUMMARY: * avoid crashing the engine when we get an input longer than max_model_len FIX #12567(*link existing issues this PR will resolve*)

Robert Shaw committed Jan 31, 2025 at 21:32 UTC b1340f9d55cd36a92aff713213e95f354a1bd1b4
1 file changed +5
vllm/v1/engine/processor.py
+5
@@ -206,6 +206,11 @@ class Processor:
206 if prompt_ids is None or len(prompt_ids) == 0:
207 raise ValueError("Prompt cannot be empty")
208
209 + if len(prompt_ids) >= self.model_config.max_model_len:
210 + raise ValueError(
211 + f"Prompt length of {len(prompt_ids)} is longer than the "
212 + f"maximum model length of {self.model_config.max_model_len}.")
213 +
214 if self.model_config.is_multimodal_model:
215 max_prompt_len = self.model_config.max_model_len
216