make 500: InternalServerError more informative (#20610)

Signed-off-by: Masataro Asai <guicho2.71828@gmail.com>

vllm/v1/engine/async_llm.py

@@ -461,9 +461,9 @@ class AsyncLLM(EngineClient):
             raise
 
         # Request validation error.
-        except ValueError:
+        except ValueError as e:
             if self.log_requests:
-                logger.info("Request %s failed (bad request).", request_id)
+                logger.info("Request %s failed (bad request): %s.", request_id, e)
             raise
 
         # Unexpected error in the generate() task (possibly recoverable).
@@ -471,7 +471,15 @@ class AsyncLLM(EngineClient):
             if q is not None:
                 await self.abort(q.request_id, internal=True)
             if self.log_requests:
-                logger.info("Request %s failed.", request_id)
+                try:
+                    s = f"{e.__class__.__name__}: {e}"
+                except Exception as e2:
+                    s = (
+                        f"{e.__class__.__name__}: "
+                        + "error during printing an exception of class"
+                        + e2.__class__.__name__
+                    )
+                logger.info("Request %s failed due to %s.", request_id, s)
             raise EngineGenerateError() from e
 
     def _run_output_handler(self):