use logger.error directly

Author: rshaw@neuralmagic.com

vllm/engine/llm_engine.py

@@ -616,6 +616,7 @@ def _add_processed_request(
             decoder_inputs = processed_inputs
             encoder_inputs = None
 
+        print(f"{decoder_inputs=}")
         seq = Sequence(seq_id, decoder_inputs, block_size, eos_token_id,
                        lora_request, prompt_adapter_request)
 

vllm/entrypoints/openai/serving_completion.py

@@ -105,13 +105,16 @@ async def create_completion(
 
             tokenizer = await self.engine_client.get_tokenizer(lora_request)
 
+            print(f"{request.prompt=}")
             request_prompts, engine_prompts = await self._preprocess_completion(
                 request,
                 tokenizer,
                 request.prompt,
                 truncate_prompt_tokens=request.truncate_prompt_tokens,
                 add_special_tokens=request.add_special_tokens,
             )
+            print(f"{request_prompts=}")
+            print(f"{engine_prompts=}")
         except ValueError as e:
             logger.exception("Error in preprocessing prompt inputs")
             return self.create_error_response(str(e))

vllm/transformers_utils/tokenizer_group/tokenizer_group.py

@@ -57,7 +57,9 @@ def encode(self,
                request_id: Optional[str] = None,
                lora_request: Optional[LoRARequest] = None) -> List[int]:
         tokenizer = self.get_lora_tokenizer(lora_request)
+        print(f"{prompt=}")
         ret = tokenizer.encode(prompt)
+        print(f"{ret=}")
         self._raise_if_input_too_long(ret, lora_request)
         return ret
 

vllm/v1/engine/async_llm.py

@@ -18,7 +18,7 @@
 from vllm.transformers_utils.tokenizer import AnyTokenizer
 from vllm.transformers_utils.tokenizer_group import init_tokenizer_from_configs
 from vllm.usage.usage_lib import UsageContext
-from vllm.utils import get_exception_traceback, kill_process_tree
+from vllm.utils import kill_process_tree
 from vllm.v1.engine.core_client import EngineCoreClient
 from vllm.v1.engine.detokenizer import Detokenizer
 from vllm.v1.engine.processor import Processor
@@ -293,9 +293,8 @@ async def _run_output_handler(self):
                 # 4) Abort any requests that finished due to stop strings.
                 await self.engine_core.abort_requests_async(reqs_to_abort)
 
-        except Exception:
-            traceback = get_exception_traceback()
-            logger.error("EngineCore hit an exception: %s", traceback)
+        except Exception as e:
+            logger.error(e)
             kill_process_tree(os.getpid())
 
     async def abort(self, request_id: str) -> None: