[Bugfix] Fixing max token error message for openai compatible server (v…

…llm-project#4016)

vllm/entrypoints/openai/serving_engine.py

@@ -206,6 +206,12 @@ def _validate_prompt_and_tokenize(
         token_num = len(input_ids)
 
         if request.max_tokens is None:
+            if token_num >= self.max_model_len:
+                raise ValueError(
+                    f"This model's maximum context length is "
+                    f"{self.max_model_len} tokens. However, you requested "
+                    f"{token_num} tokens in the messages, "
+                    f"Please reduce the length of the messages.", )
             request.max_tokens = self.max_model_len - token_num
 
         if token_num + request.max_tokens > self.max_model_len: