【bug】大模型微调bug：OSError: Failed to load tokenizer.| Lora

LLaMA-Factory 默认使用的是 Hugging Face Hub（huggingface.co）作为模型下载源。如果你在国内、或者使用的是 ModelScope 镜像（魔搭社区），就必须手动改配置或环境变量，让它从 modelscope.cn 拉取模型。

报错原文

root@dsw-1400022-546bdb8d54-27nsv:/mnt/workspace/LLaMA-Factory# llamafactory-cli train examples/train_lora/llama3_lora_sft.yaml
2025-10-16 13:00:45.897740: I tensorflow/core/util/port.cc:113] oneDNN custom operations are on. You may see slightly different numerical results due to floating-point round-off errors from different computation orders. To turn them off, set the environment variable `TF_ENABLE_ONEDNN_OPTS=0`.
2025-10-16 13:00:45.937885: I tensorflow/core/platform/cpu_feature_guard.cc:210] This TensorFlow binary is optimized to use available CPU instructions in performance-critical operations.
To enable the following instructions: AVX2 AVX512F AVX512_VNNI FMA, in other operations, rebuild TensorFlow with the appropriate compiler flags.
2025-10-16 13:00:46.853637: W tensorflow/compiler/tf2tensorrt/utils/py_utils.cc:38] TF-TRT Warning: Could not find TensorRT
[2025-10-16 13:00:49,181] [INFO] [real_accelerator.py:254:get_accelerator] Setting ds_accelerator to cuda (auto detect)
df: /root/.triton/autotune: 没有那个文件或目录
[2025-10-16 13:00:51,610] [INFO] [logging.py:107:log_dist] [Rank -1] [TorchCheckpointEngine] Initialized with serialization = False
[INFO|2025-10-16 13:00:52] llamafactory.hparams.parser:423 >> Process rank: 0, world size: 1, device: cuda:0, distributed training: False, compute dtype: torch.bfloat16
[INFO|tokenization_auto.py:898] 2025-10-16 13:01:02,405 >> Could not locate the tokenizer configuration file, will try to use the model config instead.
Traceback (most recent call last):File "/usr/local/lib/python3.11/site-packages/urllib3/connection.py", line 198, in _new_connsock = connection.create_connection(^^^^^^^^^^^^^^^^^^^^^^^^^^^^^File "/usr/local/lib/python3.11/site-packages/urllib3/util/connection.py", line 85, in create_connectionraise errFile "/usr/local/lib/python3.11/site-packages/urllib3/util/connection.py", line 73, in create_connectionsock.connect(sa)
OSError: [Errno 101] Network is unreachableThe above exception was the direct cause of the following exception:Traceback (most recent call last):File "/usr/local/lib/python3.11/site-packages/urllib3/connectionpool.py", line 787, in urlopenresponse = self._make_request(^^^^^^^^^^^^^^^^^^^File "/usr/local/lib/python3.11/site-packages/urllib3/connectionpool.py", line 488, in _make_requestraise new_eFile "/usr/local/lib/python3.11/site-packages/urllib3/connectionpool.py", line 464, in _make_requestself._validate_conn(conn)File "/usr/local/lib/python3.11/site-packages/urllib3/connectionpool.py", line 1093, in _validate_connconn.connect()File "/usr/local/lib/python3.11/site-packages/urllib3/connection.py", line 704, in connectself.sock = sock = self._new_conn()^^^^^^^^^^^^^^^^File "/usr/local/lib/python3.11/site-packages/urllib3/connection.py", line 213, in _new_connraise NewConnectionError(
urllib3.exceptions.NewConnectionError: <urllib3.connection.HTTPSConnection object at 0x7f63fc1bfcd0>: Failed to establish a new connection: [Errno 101] Network is unreachableThe above exception was the direct cause of the following exception:Traceback (most recent call last):File "/usr/local/lib/python3.11/site-packages/requests/adapters.py", line 667, in sendresp = conn.urlopen(^^^^^^^^^^^^^File "/usr/local/lib/python3.11/site-packages/urllib3/connectionpool.py", line 841, in urlopenretries = retries.increment(^^^^^^^^^^^^^^^^^^File "/usr/local/lib/python3.11/site-packages/urllib3/util/retry.py", line 519, in incrementraise MaxRetryError(_pool, url, reason) from reason  # type: ignore[arg-type]^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
urllib3.exceptions.MaxRetryError: HTTPSConnectionPool(host='huggingface.co', port=443): Max retries exceeded with url: /LLM-Research/Meta-Llama-3-8B-Instruct/resolve/main/config.json (Caused by NewConnectionError('<urllib3.connection.HTTPSConnection object at 0x7f63fc1bfcd0>: Failed to establish a new connection: [Errno 101] Network is unreachable'))During handling of the above exception, another exception occurred:Traceback (most recent call last):File "/usr/local/lib/python3.11/site-packages/huggingface_hub/file_download.py", line 1546, in _get_metadata_or_catch_errormetadata = get_hf_file_metadata(^^^^^^^^^^^^^^^^^^^^^File "/usr/local/lib/python3.11/site-packages/huggingface_hub/utils/_validators.py", line 114, in _inner_fnreturn fn(*args, **kwargs)^^^^^^^^^^^^^^^^^^^File "/usr/local/lib/python3.11/site-packages/huggingface_hub/file_download.py", line 1463, in get_hf_file_metadatar = _request_wrapper(^^^^^^^^^^^^^^^^^File "/usr/local/lib/python3.11/site-packages/huggingface_hub/file_download.py", line 286, in _request_wrapperresponse = _request_wrapper(^^^^^^^^^^^^^^^^^File "/usr/local/lib/python3.11/site-packages/huggingface_hub/file_download.py", line 309, in _request_wrapperresponse = http_backoff(method=method, url=url, **params, retry_on_exceptions=(), retry_on_status_codes=(429,))^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^File "/usr/local/lib/python3.11/site-packages/huggingface_hub/utils/_http.py", line 310, in http_backoffresponse = session.request(method=method, url=url, **kwargs)^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^File "/usr/local/lib/python3.11/site-packages/requests/sessions.py", line 589, in requestresp = self.send(prep, **send_kwargs)^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^File "/usr/local/lib/python3.11/site-packages/requests/sessions.py", line 703, in sendr = adapter.send(request, **kwargs)^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^File "/usr/local/lib/python3.11/site-packages/huggingface_hub/utils/_http.py", line 96, in sendreturn super().send(request, *args, **kwargs)^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^File "/usr/local/lib/python3.11/site-packages/requests/adapters.py", line 700, in sendraise ConnectionError(e, request=request)
requests.exceptions.ConnectionError: (MaxRetryError("HTTPSConnectionPool(host='huggingface.co', port=443): Max retries exceeded with url: /LLM-Research/Meta-Llama-3-8B-Instruct/resolve/main/config.json (Caused by NewConnectionError('<urllib3.connection.HTTPSConnection object at 0x7f63fc1bfcd0>: Failed to establish a new connection: [Errno 101] Network is unreachable'))"), '(Request ID: 173af757-1a47-42b9-9662-480a9cec02c5)')The above exception was the direct cause of the following exception:Traceback (most recent call last):File "/usr/local/lib/python3.11/site-packages/transformers/utils/hub.py", line 479, in cached_fileshf_hub_download(File "/usr/local/lib/python3.11/site-packages/huggingface_hub/utils/_validators.py", line 114, in _inner_fnreturn fn(*args, **kwargs)^^^^^^^^^^^^^^^^^^^File "/usr/local/lib/python3.11/site-packages/huggingface_hub/file_download.py", line 1010, in hf_hub_downloadreturn _hf_hub_download_to_cache_dir(^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^File "/usr/local/lib/python3.11/site-packages/huggingface_hub/file_download.py", line 1117, in _hf_hub_download_to_cache_dir_raise_on_head_call_error(head_call_error, force_download, local_files_only)File "/usr/local/lib/python3.11/site-packages/huggingface_hub/file_download.py", line 1661, in _raise_on_head_call_errorraise LocalEntryNotFoundError(
huggingface_hub.errors.LocalEntryNotFoundError: An error happened while trying to locate the file on the Hub and we cannot find the requested files in the local cache. Please check your connection and try again or make sure your Internet connection is on.The above exception was the direct cause of the following exception:Traceback (most recent call last):File "/mnt/workspace/LLaMA-Factory/src/llamafactory/model/loader.py", line 78, in load_tokenizertokenizer = AutoTokenizer.from_pretrained(^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^File "/usr/local/lib/python3.11/site-packages/transformers/models/auto/tokenization_auto.py", line 1069, in from_pretrainedconfig = AutoConfig.from_pretrained(^^^^^^^^^^^^^^^^^^^^^^^^^^^File "/usr/local/lib/python3.11/site-packages/transformers/models/auto/configuration_auto.py", line 1250, in from_pretrainedconfig_dict, unused_kwargs = PretrainedConfig.get_config_dict(pretrained_model_name_or_path, **kwargs)^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^File "/usr/local/lib/python3.11/site-packages/transformers/configuration_utils.py", line 649, in get_config_dictconfig_dict, kwargs = cls._get_config_dict(pretrained_model_name_or_path, **kwargs)^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^File "/usr/local/lib/python3.11/site-packages/transformers/configuration_utils.py", line 708, in _get_config_dictresolved_config_file = cached_file(^^^^^^^^^^^^File "/usr/local/lib/python3.11/site-packages/transformers/utils/hub.py", line 321, in cached_filefile = cached_files(path_or_repo_id=path_or_repo_id, filenames=[filename], **kwargs)^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^File "/usr/local/lib/python3.11/site-packages/transformers/utils/hub.py", line 553, in cached_filesraise OSError(
OSError: We couldn't connect to 'https://huggingface.co' to load the files, and couldn't find them in the cached files.
Check your internet connection or see how to run the library in offline mode at 'https://huggingface.co/docs/transformers/installation#offline-mode'.The above exception was the direct cause of the following exception:Traceback (most recent call last):File "/usr/local/bin/llamafactory-cli", line 8, in <module>sys.exit(main())^^^^^^File "/mnt/workspace/LLaMA-Factory/src/llamafactory/cli.py", line 24, in mainlauncher.launch()File "/mnt/workspace/LLaMA-Factory/src/llamafactory/launcher.py", line 152, in launchrun_exp()File "/mnt/workspace/LLaMA-Factory/src/llamafactory/train/tuner.py", line 110, in run_exp_training_function(config={"args": args, "callbacks": callbacks})File "/mnt/workspace/LLaMA-Factory/src/llamafactory/train/tuner.py", line 72, in _training_functionrun_sft(model_args, data_args, training_args, finetuning_args, generating_args, callbacks)File "/mnt/workspace/LLaMA-Factory/src/llamafactory/train/sft/workflow.py", line 48, in run_sfttokenizer_module = load_tokenizer(model_args)^^^^^^^^^^^^^^^^^^^^^^^^^^File "/mnt/workspace/LLaMA-Factory/src/llamafactory/model/loader.py", line 93, in load_tokenizerraise OSError("Failed to load tokenizer.") from e
OSError: Failed to load tokenizer.

bug分析

我的模型下载源默认是在Hubgging Face Hub。但是我是在国内，用的是ModelScope社区镜像，魔塔社区，拉取模型的时候吗，下载外网镜像会失败。

解决方案

将镜像源切换到国内的魔塔社区。

在 llama3_lora_sft.yaml 里直接指定 ModelScope 模型路径：

model_name_or_path: LLM-Research/Meta-Llama-3-8B-Instruct

设置环境变量：

export USE_MODELSCOPE_HUB=1
export MODELSCOPE_CACHE=/mnt/workspace/.cache/modelscope

开始训练：

llamafactory-cli train examples/train_lora/llama3_lora_sft.yaml

训练成功如图所示：