From 789dccecedc141014eab81f5fcd37bc832c50170 Mon Sep 17 00:00:00 2001
From: yinfan98 <32722923+yinfan98@users.noreply.github.com>
Date: Wed, 31 Jan 2024 19:23:41 +0800
Subject: [PATCH] fix and change all logger using get_logger (#1064)

* fix logger

* Merge the main branch for logger

* Fix some unchanged logging to logger
---
 benchmark/profile_hf_generation.py            |  3 +-
 lmdeploy/legacy/pytorch/adapters/__init__.py  |  6 ++--
 lmdeploy/legacy/pytorch/adapters/base.py      |  5 +--
 lmdeploy/legacy/pytorch/adapters/internlm.py  |  5 +--
 lmdeploy/legacy/pytorch/adapters/llama2.py    |  5 +--
 lmdeploy/legacy/pytorch/chat.py               | 26 ++++++++-------
 lmdeploy/legacy/pytorch/decode.py             |  5 +--
 lmdeploy/legacy/pytorch/model.py              |  5 +--
 lmdeploy/legacy/pytorch/session.py            |  5 +--
 lmdeploy/legacy/pytorch/utils.py              |  6 ++--
 .../pytorch/modeling/modeling_baichuan.py     |  4 +--
 .../pytorch/modeling/modeling_internlm.py     |  5 +--
 .../pytorch/modeling/modeling_internlm2.py    |  5 +--
 lmdeploy/pytorch/modeling/modeling_llama.py   |  5 +--
 lmdeploy/pytorch/models/falcon.py             |  3 --
 lmdeploy/serve/qos_engine/inner_group_schd.py |  5 +--
 lmdeploy/serve/qos_engine/qos_engine.py       |  4 +--
 lmdeploy/serve/turbomind/chatbot.py           | 21 ++++--------
 lmdeploy/turbomind/turbomind.py               | 24 ++++++--------
 lmdeploy/turbomind/utils.py                   |  5 +--
 lmdeploy/utils.py                             | 32 ++++++-------------
 21 files changed, 87 insertions(+), 97 deletions(-)

diff --git a/benchmark/profile_hf_generation.py b/benchmark/profile_hf_generation.py
index 1c6704bc4..6e981a191 100644
--- a/benchmark/profile_hf_generation.py
+++ b/benchmark/profile_hf_generation.py
@@ -75,8 +75,9 @@
 from transformers import AutoModelForCausalLM, GenerationConfig
 
 from lmdeploy.pytorch.accel import LoadNoInit
+from lmdeploy.utils import get_logger
 
-logger = logging.getLogger(__file__)
+logger = get_logger(__file__)
 logger.setLevel(logging.DEBUG)
 info = logger.info
 warning = logger.warning
diff --git a/lmdeploy/legacy/pytorch/adapters/__init__.py b/lmdeploy/legacy/pytorch/adapters/__init__.py
index f87c0546d..f4e93b7e5 100644
--- a/lmdeploy/legacy/pytorch/adapters/__init__.py
+++ b/lmdeploy/legacy/pytorch/adapters/__init__.py
@@ -1,14 +1,14 @@
 # Copyright (c) OpenMMLab. All rights reserved.
 
-import logging
-
 import torch.nn as nn
 
+from lmdeploy.utils import get_logger
+
 from .base import BasicAdapter, BasicAdapterFast
 from .internlm import InternLMAdapter
 from .llama2 import Llama2Adapter
 
-logger = logging.getLogger(__name__)
+logger = get_logger(__name__)
 
 
 def _get_default_adapter(tokenizer):
diff --git a/lmdeploy/legacy/pytorch/adapters/base.py b/lmdeploy/legacy/pytorch/adapters/base.py
index e958a0f37..e3eee2e9f 100644
--- a/lmdeploy/legacy/pytorch/adapters/base.py
+++ b/lmdeploy/legacy/pytorch/adapters/base.py
@@ -1,13 +1,14 @@
 # Copyright (c) OpenMMLab. All rights reserved.
 """Basic adapter suitable for general HuggingFace models."""
 
-import logging
 import re
 
 from transformers import (PreTrainedTokenizer, PreTrainedTokenizerBase,
                           PreTrainedTokenizerFast)
 
-logger = logging.getLogger(__name__)
+from lmdeploy.utils import get_logger
+
+logger = get_logger(__name__)
 
 
 class BaseAdapter:
diff --git a/lmdeploy/legacy/pytorch/adapters/internlm.py b/lmdeploy/legacy/pytorch/adapters/internlm.py
index 89e6f9f03..fc785c3b7 100644
--- a/lmdeploy/legacy/pytorch/adapters/internlm.py
+++ b/lmdeploy/legacy/pytorch/adapters/internlm.py
@@ -1,14 +1,15 @@
 # Copyright (c) OpenMMLab. All rights reserved.
-import logging
 import re
 
 import torch
 from transformers import (PreTrainedTokenizerFast, StoppingCriteria,
                           StoppingCriteriaList)
 
+from lmdeploy.utils import get_logger
+
 from .base import BaseAdapter
 
-logger = logging.getLogger(__name__)
+logger = get_logger(__name__)
 
 
 class InternLMStoppingCriteria(StoppingCriteria):
diff --git a/lmdeploy/legacy/pytorch/adapters/llama2.py b/lmdeploy/legacy/pytorch/adapters/llama2.py
index d0dcb6d25..5e46a1d6d 100644
--- a/lmdeploy/legacy/pytorch/adapters/llama2.py
+++ b/lmdeploy/legacy/pytorch/adapters/llama2.py
@@ -1,12 +1,13 @@
 # Copyright (c) OpenMMLab. All rights reserved.
-import logging
 import re
 
 from transformers import PreTrainedTokenizerFast
 
+from lmdeploy.utils import get_logger
+
 from .base import BasicAdapterFast
 
-logger = logging.getLogger(__name__)
+logger = get_logger(__name__)
 
 B_INST, E_INST = '[INST]', '[/INST]'
 B_SYS, E_SYS = '<<SYS>>\n', '\n<</SYS>>\n\n'
diff --git a/lmdeploy/legacy/pytorch/chat.py b/lmdeploy/legacy/pytorch/chat.py
index 9a167cc52..12c8d4df4 100644
--- a/lmdeploy/legacy/pytorch/chat.py
+++ b/lmdeploy/legacy/pytorch/chat.py
@@ -54,14 +54,14 @@
 import torch
 from transformers import GenerationConfig, PreTrainedModel
 
+from lmdeploy.utils import get_logger
+
 from .adapters import init_adapter
 from .dist import get_local_rank, get_rank, get_world_size
 from .model import accel_model, init_model
 from .session import BasicSessionManagerWithHistory
 from .utils import BasicStreamer, TerminalIO, control
 
-logger = logging.getLogger(__name__)
-
 
 def set_logging(log_file: str, debug: bool):
     torch.set_printoptions(linewidth=120)
@@ -69,15 +69,19 @@ def set_logging(log_file: str, debug: bool):
     log_file = log_file or 'chat.log'
     if r := get_rank() != 0:
         log_file = log_file + f'.{r}'
-    logging.basicConfig(level=level,
-                        format=('%(filename)s: '
-                                '%(levelname)s: '
-                                '%(funcName)s(): '
-                                '%(lineno)d:\t'
-                                '%(message)s'),
-                        filename=log_file,
-                        filemode='w')
+    format = '%(filename)s:   \
+              %(levelname)s:  \
+              %(funcName)s(): \
+              %(lineno)d:\t   \
+              %(message)s'
+
+    logger = get_logger(__name__,
+                        log_file=log_file,
+                        log_level=level,
+                        file_mode='w',
+                        log_formatter=format)
     print(f'Worker {get_rank()} logging to {log_file}')
+    return logger
 
 
 def main(
@@ -120,7 +124,7 @@ def main(
             based on `LlamaforCausalLM` class, this argument is required.
             Currently, only "llama1" is acceptable for llama1 models.
     """  # noqa: E501
-    set_logging(log_file, debug)
+    logger = set_logging(log_file, debug)
 
     # workers should sync in sampling
     torch.manual_seed(seed)
diff --git a/lmdeploy/legacy/pytorch/decode.py b/lmdeploy/legacy/pytorch/decode.py
index 594c7807b..25e296240 100644
--- a/lmdeploy/legacy/pytorch/decode.py
+++ b/lmdeploy/legacy/pytorch/decode.py
@@ -1,6 +1,5 @@
 # Copyright (c) OpenMMLab. All rights reserved.
 import argparse
-import logging
 import queue
 import warnings
 from typing import List, Optional
@@ -12,6 +11,8 @@
 from transformers import (AutoTokenizer, PreTrainedModel,
                           PreTrainedTokenizerBase)
 
+from lmdeploy.utils import get_logger
+
 from .model import accel_model, init_model
 
 
@@ -371,7 +372,7 @@ def __del__(self):
     test_path = args.test_path
     prompts = args.prompts
 
-    logger = logging.getLogger(__name__)
+    logger = get_logger(__name__)
     # logging.basicConfig(level=logging.DEBUG)
 
     # Use test file preferentially
diff --git a/lmdeploy/legacy/pytorch/model.py b/lmdeploy/legacy/pytorch/model.py
index 4a0476ff5..1982adf48 100644
--- a/lmdeploy/legacy/pytorch/model.py
+++ b/lmdeploy/legacy/pytorch/model.py
@@ -1,5 +1,4 @@
 # Copyright (c) OpenMMLab. All rights reserved.
-import logging
 import time
 import warnings
 from typing import Optional
@@ -7,9 +6,11 @@
 import torch
 from transformers import AutoModelForCausalLM, AutoTokenizer
 
+from lmdeploy.utils import get_logger
+
 from .dist import get_local_rank
 
-logger = logging.getLogger(__name__)
+logger = get_logger(__name__)
 
 
 class LoadWoInit:
diff --git a/lmdeploy/legacy/pytorch/session.py b/lmdeploy/legacy/pytorch/session.py
index 4689b258b..3ccad0946 100644
--- a/lmdeploy/legacy/pytorch/session.py
+++ b/lmdeploy/legacy/pytorch/session.py
@@ -1,10 +1,11 @@
 # Copyright (c) OpenMMLab. All rights reserved.
-import logging
 
 import torch
 from transformers.generation.utils import ModelOutput
 
-logger = logging.getLogger(__name__)
+from lmdeploy.utils import get_logger
+
+logger = get_logger(__name__)
 
 
 class BasicSessionManager:
diff --git a/lmdeploy/legacy/pytorch/utils.py b/lmdeploy/legacy/pytorch/utils.py
index 8f88e3279..90a645956 100644
--- a/lmdeploy/legacy/pytorch/utils.py
+++ b/lmdeploy/legacy/pytorch/utils.py
@@ -1,9 +1,9 @@
 # Copyright (c) OpenMMLab. All rights reserved.
 
-import logging
-
 from transformers.generation.streamers import BaseStreamer
 
+from lmdeploy.utils import get_logger
+
 from .dist import get_rank, master_only, master_only_and_broadcast_general
 
 try:
@@ -11,7 +11,7 @@
 except ImportError:  # readline not available
     pass
 
-logger = logging.getLogger(__name__)
+logger = get_logger(__name__)
 
 
 class TerminalIO:
diff --git a/lmdeploy/pytorch/modeling/modeling_baichuan.py b/lmdeploy/pytorch/modeling/modeling_baichuan.py
index 0292cf41e..e5fb0360b 100644
--- a/lmdeploy/pytorch/modeling/modeling_baichuan.py
+++ b/lmdeploy/pytorch/modeling/modeling_baichuan.py
@@ -27,13 +27,13 @@
 from transformers.activations import ACT2FN
 from transformers.modeling_outputs import (BaseModelOutputWithPast,
                                            CausalLMOutputWithPast)
-from transformers.utils import logging
 
 from lmdeploy.pytorch.modeling.convert_to_qmodules import convert_to_qmodules
+from lmdeploy.utils import get_logger
 
 from .configuration_baichuan import BaiChuanConfig
 
-logger = logging.get_logger(__name__)
+logger = get_logger(__name__)
 
 
 # Copied from transformers.models.bart.modeling_bart._make_causal_mask
diff --git a/lmdeploy/pytorch/modeling/modeling_internlm.py b/lmdeploy/pytorch/modeling/modeling_internlm.py
index 64c982bff..da7a167d2 100644
--- a/lmdeploy/pytorch/modeling/modeling_internlm.py
+++ b/lmdeploy/pytorch/modeling/modeling_internlm.py
@@ -33,14 +33,15 @@
                                            SequenceClassifierOutputWithPast)
 from transformers.modeling_utils import PreTrainedModel
 from transformers.utils import (add_start_docstrings,
-                                add_start_docstrings_to_model_forward, logging,
+                                add_start_docstrings_to_model_forward,
                                 replace_return_docstrings)
 
 from lmdeploy.pytorch.modeling.convert_to_qmodules import convert_to_qmodules
+from lmdeploy.utils import get_logger
 
 from .configuration_internlm import InternLMConfig
 
-logger = logging.get_logger(__name__)
+logger = get_logger(__name__)
 
 _CONFIG_FOR_DOC = 'InternLMConfig'
 
diff --git a/lmdeploy/pytorch/modeling/modeling_internlm2.py b/lmdeploy/pytorch/modeling/modeling_internlm2.py
index a5240a6ad..a29c37b2f 100644
--- a/lmdeploy/pytorch/modeling/modeling_internlm2.py
+++ b/lmdeploy/pytorch/modeling/modeling_internlm2.py
@@ -34,10 +34,11 @@
                                            SequenceClassifierOutputWithPast)
 from transformers.modeling_utils import PreTrainedModel
 from transformers.utils import (add_start_docstrings,
-                                add_start_docstrings_to_model_forward, logging,
+                                add_start_docstrings_to_model_forward,
                                 replace_return_docstrings)
 
 from lmdeploy.pytorch.modeling.convert_to_qmodules import convert_to_qmodules
+from lmdeploy.utils import get_logger
 
 try:
     from transformers.generation.streamers import BaseStreamer
@@ -46,7 +47,7 @@
 
 from .configuration_internlm import InternLMConfig as InternLM2Config
 
-logger = logging.get_logger(__name__)
+logger = get_logger(__name__)
 
 _CONFIG_FOR_DOC = 'InternLM2Config'
 
diff --git a/lmdeploy/pytorch/modeling/modeling_llama.py b/lmdeploy/pytorch/modeling/modeling_llama.py
index 68e0e5a8a..efe66123a 100644
--- a/lmdeploy/pytorch/modeling/modeling_llama.py
+++ b/lmdeploy/pytorch/modeling/modeling_llama.py
@@ -32,12 +32,13 @@
 from transformers.modeling_utils import PreTrainedModel
 from transformers.models.llama.configuration_llama import LlamaConfig
 from transformers.utils import (add_start_docstrings,
-                                add_start_docstrings_to_model_forward, logging,
+                                add_start_docstrings_to_model_forward,
                                 replace_return_docstrings)
 
 from lmdeploy.pytorch.modeling.convert_to_qmodules import convert_to_qmodules
+from lmdeploy.utils import get_logger
 
-logger = logging.get_logger(__name__)
+logger = get_logger(__name__)
 
 _CONFIG_FOR_DOC = 'LlamaConfig'
 
diff --git a/lmdeploy/pytorch/models/falcon.py b/lmdeploy/pytorch/models/falcon.py
index 9d18d9eaa..b708c0dbb 100644
--- a/lmdeploy/pytorch/models/falcon.py
+++ b/lmdeploy/pytorch/models/falcon.py
@@ -3,7 +3,6 @@
 # https://huggingface.co/tiiuae/falcon-7b-instruct
 # https://github.com/huggingface/transformers/blob/v4.33-release/src/transformers/models/falcon/modeling_falcon.py  # noqa
 
-import logging
 from typing import Optional, Tuple, Union
 
 import torch
@@ -20,8 +19,6 @@
 from ..kernels import (alibi_paged_attention_fwd, fill_kv_cache,
                        paged_attention_fwd)
 
-logger = logging.getLogger()
-
 
 # rotary pos emb helpers
 # (torch.jit.script does not seem to support staticmethod...)
diff --git a/lmdeploy/serve/qos_engine/inner_group_schd.py b/lmdeploy/serve/qos_engine/inner_group_schd.py
index 73a57e1ef..bc537c034 100644
--- a/lmdeploy/serve/qos_engine/inner_group_schd.py
+++ b/lmdeploy/serve/qos_engine/inner_group_schd.py
@@ -1,8 +1,9 @@
 # Copyright (c) OpenMMLab. All rights reserved.
 import collections
-import logging
 
-logger = logging.getLogger(__name__)
+from lmdeploy.utils import get_logger
+
+logger = get_logger(__name__)
 
 
 class UserRequestQueue:
diff --git a/lmdeploy/serve/qos_engine/qos_engine.py b/lmdeploy/serve/qos_engine/qos_engine.py
index ca5762de5..da2e65012 100644
--- a/lmdeploy/serve/qos_engine/qos_engine.py
+++ b/lmdeploy/serve/qos_engine/qos_engine.py
@@ -1,7 +1,6 @@
 # Copyright (c) OpenMMLab. All rights reserved.
 import asyncio
 import json
-import logging
 import threading
 import time
 from typing import List
@@ -11,8 +10,9 @@
                                             GenerateRequestQos)
 from lmdeploy.serve.qos_engine.inner_group_schd import UserRequestQueue
 from lmdeploy.serve.qos_engine.usage_stats import UsageStats
+from lmdeploy.utils import get_logger
 
-logger = logging.getLogger(__name__)
+logger = get_logger(__name__)
 
 
 class QosConfig:
diff --git a/lmdeploy/serve/turbomind/chatbot.py b/lmdeploy/serve/turbomind/chatbot.py
index f0fc9e2e9..7d350c616 100644
--- a/lmdeploy/serve/turbomind/chatbot.py
+++ b/lmdeploy/serve/turbomind/chatbot.py
@@ -18,7 +18,7 @@
 from lmdeploy.model import MODELS
 from lmdeploy.serve.turbomind.utils import (Postprocessor, Preprocessor,
                                             prepare_tensor)
-from lmdeploy.utils import filter_suffix
+from lmdeploy.utils import filter_suffix, get_logger
 
 
 @dataclass
@@ -51,13 +51,6 @@ def stream_callback(que, result, error):
         que.put(result.get_response(as_json=True))
 
 
-def get_logger(log_file=None, log_level=logging.INFO):
-    """Return the logger."""
-    from lmdeploy.utils import get_logger
-    logger = get_logger('service.ft', log_file=log_file, log_level=log_level)
-    return logger
-
-
 class Chatbot:
     """Chatbot for LLaMA series models with turbomind as inference engine.
 
@@ -130,7 +123,7 @@ def stream_infer(self,
         assert isinstance(session_id, int), \
             f'INT session id is required, but got {type(session_id)}'
 
-        logger = get_logger(log_level=self.log_level)
+        logger = get_logger('service.ft', log_level=self.log_level)
         logger.info(f'session {session_id}, request_id {request_id}, '
                     f'request_output_len {request_output_len}')
 
@@ -180,7 +173,7 @@ def end(self, session_id: int, *args, **kwargs):
         assert isinstance(session_id, int), \
             f'INT session id is required, but got {type(session_id)}'
 
-        logger = get_logger(log_level=self.log_level)
+        logger = get_logger('service.ft', log_level=self.log_level)
         logger.info(f'end session: {session_id}')
 
         if self._session is None:
@@ -218,7 +211,7 @@ def cancel(self, session_id: int, *args, **kwargs):
         """
         assert isinstance(session_id, int), \
             f'INT session id is required, but got {type(session_id)}'
-        logger = get_logger(log_level=self.log_level)
+        logger = get_logger('service.ft', log_level=self.log_level)
         logger.info(f'cancel session: {session_id}')
 
         if self._session is None:
@@ -267,7 +260,7 @@ def resume(self, session_id: int, *args, **kwargs):
         assert isinstance(session_id, int), \
             f'INT session id is required, but got {type(session_id)}'
 
-        logger = get_logger(log_level=self.log_level)
+        logger = get_logger('service.ft', log_level=self.log_level)
         logger.info(f'resume session: {session_id}')
 
         if self._session is None:
@@ -320,7 +313,7 @@ def infer(self,
         assert isinstance(session_id, int), \
             f'INT session id is required, but got {type(session_id)}'
 
-        logger = get_logger(log_level=self.log_level)
+        logger = get_logger('service.ft', log_level=self.log_level)
         logger.info(f'session {session_id}, request_id {request_id}, '
                     f'request_output_len {request_output_len}')
 
@@ -434,7 +427,7 @@ def _stream_infer(self,
         Yields:
             tuple: status, text, generated token number
         """
-        logger = get_logger(log_level=self.log_level)
+        logger = get_logger('service.ft', log_level=self.log_level)
         logger.info(f'session {session.session_id}, '
                     f'request id {session.request_id}, '
                     f'request_output_len {request_output_len}, '
diff --git a/lmdeploy/turbomind/turbomind.py b/lmdeploy/turbomind/turbomind.py
index c1dec981f..6af9764e2 100644
--- a/lmdeploy/turbomind/turbomind.py
+++ b/lmdeploy/turbomind/turbomind.py
@@ -35,7 +35,7 @@
 sys.path.append(osp.join(lmdeploy_dir, 'lib'))
 import _turbomind as _tm  # noqa: E402
 
-logger = logging.getLogger(__name__)
+logger = get_logger(__name__)
 
 
 def _construct_stop_or_bad_words(words: List[int] = None):
@@ -72,9 +72,8 @@ def _update_engine_config(config: TurbomindEngineConfig, **kwargs):
     for k, v in kwargs.items():
         if v and hasattr(config, k):
             setattr(config, k, v)
-            get_logger('turbomind').warning(
-                f'kwargs {k} is deprecated to initialize model, '
-                'use TurbomindEngineConfig instead.')
+            logger.warning(f'kwargs {k} is deprecated to initialize model, '
+                           'use TurbomindEngineConfig instead.')
     if config.model_name is not None:
         logger.warning('model_name is deprecated in TurbomindEngineConfig '
                        'and has no effect')
@@ -92,8 +91,7 @@ def _update_tm_config(dst: TurbomindModelConfig, src: TurbomindEngineConfig):
 
 
 def _compare_individual_gpu_memory(tp: int):
-    logging.basicConfig(level=logging.INFO)
-
+    logger.setLevel(level=logging.INFO)
     try:
         total_mem = []
         free_mem = []
@@ -108,12 +106,12 @@ def _compare_individual_gpu_memory(tp: int):
         all_free_equal = all(free == free_mem[0] for free in free_mem)
 
         if not all_total_equal or not all_free_equal:
-            logging.warning(
+            logger.warning(
                 f'Memory discrepancy detected: Total Memory={total_mem} MB, \
 Free Memory={free_mem} MB')
 
     except Exception as e:
-        logging.error(f'An exception occurred: {e}')
+        logger.error(f'An exception occurred: {e}')
 
 
 @contextmanager
@@ -170,7 +168,7 @@ def _catch_args(**kwargs):
 
             args = _catch_args(**kwargs, model_format=model_format, tp=tp)
             if len(args) > 0:
-                get_logger('turbomind').warning(
+                logger.warning(
                     f'loading from workspace, ignore args {args} '
                     'please use TurbomindEngineConfig or modify config.ini')
 
@@ -360,8 +358,7 @@ def _from_workspace(self, model_path: str,
         # check whether input tp is valid
         if cfg.tensor_para_size != 1 and \
                 self.gpu_count != cfg.tensor_para_size:
-            get_logger('turbomind').info(
-                f'found tp={cfg.tensor_para_size} in config.ini.')
+            logger.info(f'found tp={cfg.tensor_para_size} in config.ini.')
             self.gpu_count = cfg.tensor_para_size
 
         # update cfg
@@ -537,9 +534,8 @@ def _update_generation_config(self, config: EngineGenerationConfig,
             config.max_new_tokens = kwargs['request_output_len']
             deprecated_kwargs.append('request_output_len')
         for k in deprecated_kwargs:
-            get_logger('turbomind').warning(
-                f'kwargs {k} is deprecated for inference, '
-                'use GenerationConfig instead.')
+            logger.warning(f'kwargs {k} is deprecated for inference, '
+                           'use GenerationConfig instead.')
         return config
 
     def end(self, session_id: int):
diff --git a/lmdeploy/turbomind/utils.py b/lmdeploy/turbomind/utils.py
index c8ec34cf2..7fa6513d2 100644
--- a/lmdeploy/turbomind/utils.py
+++ b/lmdeploy/turbomind/utils.py
@@ -1,13 +1,14 @@
 # Copyright (c) OpenMMLab. All rights reserved.
 import dataclasses
 import json
-import logging
 import os
 
 from huggingface_hub import hf_hub_download
 from transformers.utils import ExplicitEnum
 
-logger = logging.getLogger(__name__)
+from lmdeploy.utils import get_logger
+
+logger = get_logger(__name__)
 
 
 class ModelSource(ExplicitEnum):
diff --git a/lmdeploy/utils.py b/lmdeploy/utils.py
index 39e4c5062..d4a861c95 100644
--- a/lmdeploy/utils.py
+++ b/lmdeploy/utils.py
@@ -1,14 +1,18 @@
 # Copyright (c) OpenMMLab. All rights reserved.
 import logging
+from logging import Logger
 from typing import List, Optional
 
 logger_initialized = {}
 
 
-def get_logger(name: str,
-               log_file: Optional[str] = None,
-               log_level: int = logging.INFO,
-               file_mode: str = 'w'):
+def get_logger(
+    name: Optional[str] = None,
+    log_file: Optional[str] = None,
+    log_level: int = logging.INFO,
+    file_mode: str = 'w',
+    log_formatter: str = '%(asctime)s - %(name)s - %(levelname)s - %(message)s'
+) -> Logger:
     """Initialize and get a logger by name.
 
     If the logger has not been initialized, this method will initialize the
@@ -22,25 +26,10 @@ def get_logger(name: str,
         log_level (int): The logger level.
         file_mode (str): The file mode used in opening log file.
             Defaults to 'w'.
+        log_formatter (str): The logger output format.
     Returns:
         logging.Logger: The expected logger.
     """
-    # use logger in mmengine if exists.
-    try:
-        from mmengine.logging import MMLogger
-        if MMLogger.check_instance_created(name):
-            logger = MMLogger.get_instance(name)
-        else:
-            logger = MMLogger.get_instance(name,
-                                           logger_name=name,
-                                           log_file=log_file,
-                                           log_level=log_level,
-                                           file_mode=file_mode)
-        return logger
-
-    except Exception:
-        pass
-
     logger = logging.getLogger(name)
     if name in logger_initialized:
         return logger
@@ -66,8 +55,7 @@ def get_logger(name: str,
         file_handler = logging.FileHandler(log_file, file_mode)
         handlers.append(file_handler)
 
-    formatter = logging.Formatter(
-        '%(asctime)s - %(name)s - %(levelname)s - %(message)s')
+    formatter = logging.Formatter(log_formatter)
     for handler in handlers:
         handler.setFormatter(formatter)
         handler.setLevel(log_level)