251105 수 작업일지

1. mlops 시연 영상 구축(llm 모델 - gpt-oss-20b 모델 활용)

- 모델 / 서빙 / 스트림릿 화면 제작 ( 스트림릿 화면은 사용자를 위해)

2. mlops 템플릿 만들기(현대차 배터리 dpast 모델 - 전처리 / 모델 / 학습 ) 

- mlflow에 best_epoch 을 저장하는 것에 시간 소요

3. 남은 과업

- gpt-oss-20b 연동오류(토큰이 안맞다는데 관련해서 토큰 발급 블로그 만들것)

 

_manager.py:472} INFO - [base] (APIServer pid=21)     return ModelConfig(
[2025-11-05T13:35:25.865+0900] {pod_manager.py:472} INFO - [base] (APIServer pid=21)            ^^^^^^^^^^^^
[2025-11-05T13:35:25.865+0900] {pod_manager.py:472} INFO - [base] (APIServer pid=21)   File "/opt/conda/lib/python3.11/site-packages/pydantic/_internal/_dataclasses.py", line 123, in __init__
[2025-11-05T13:35:25.865+0900] {pod_manager.py:472} INFO - [base] (APIServer pid=21)     s.__pydantic_validator__.validate_python(ArgsKwargs(args, kwargs), self_instance=s)
[2025-11-05T13:35:25.865+0900] {pod_manager.py:472} INFO - [base] (APIServer pid=21)   File "/opt/conda/lib/python3.11/site-packages/vllm/config/__init__.py", line 603, in __post_init__
[2025-11-05T13:35:25.865+0900] {pod_manager.py:472} INFO - [base] (APIServer pid=21)     self.model, self.tokenizer = maybe_override_with_speculators_target_model(  # noqa: E501
[2025-11-05T13:35:25.865+0900] {pod_manager.py:472} INFO - [base] (APIServer pid=21)                                  ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
[2025-11-05T13:35:25.865+0900] {pod_manager.py:472} INFO - [base] (APIServer pid=21)   File "/opt/conda/lib/python3.11/site-packages/vllm/transformers_utils/config.py", line 487, in maybe_override_with_speculators_target_model
[2025-11-05T13:35:25.866+0900] {pod_manager.py:472} INFO - [base] (APIServer pid=21)     config_dict, _ = PretrainedConfig.get_config_dict(
[2025-11-05T13:35:25.866+0900] {pod_manager.py:472} INFO - [base] (APIServer pid=21)                      ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
[2025-11-05T13:35:25.866+0900] {pod_manager.py:472} INFO - [base] (APIServer pid=21)   File "/opt/conda/lib/python3.11/site-packages/transformers/configuration_utils.py", line 662, in get_config_dict
[2025-11-05T13:35:25.866+0900] {pod_manager.py:472} INFO - [base] (APIServer pid=21)     config_dict, kwargs = cls._get_config_dict(pretrained_model_name_or_path, **kwargs)
[2025-11-05T13:35:25.866+0900] {pod_manager.py:472} INFO - [base] (APIServer pid=21)                           ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
[2025-11-05T13:35:25.866+0900] {pod_manager.py:472} INFO - [base] (APIServer pid=21)   File "/opt/conda/lib/python3.11/site-packages/transformers/configuration_utils.py", line 721, in _get_config_dict
[2025-11-05T13:35:25.866+0900] {pod_manager.py:472} INFO - [base] (APIServer pid=21)     resolved_config_file = cached_file(
[2025-11-05T13:35:25.866+0900] {pod_manager.py:472} INFO - [base] (APIServer pid=21)                            ^^^^^^^^^^^^
[2025-11-05T13:35:25.866+0900] {pod_manager.py:472} INFO - [base] (APIServer pid=21)   File "/opt/conda/lib/python3.11/site-packages/transformers/utils/hub.py", line 321, in cached_file
[2025-11-05T13:35:25.866+0900] {pod_manager.py:472} INFO - [base] (APIServer pid=21)     file = cached_files(path_or_repo_id=path_or_repo_id, filenames=[filename], **kwargs)
[2025-11-05T13:35:25.867+0900] {pod_manager.py:472} INFO - [base] (APIServer pid=21)            ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
[2025-11-05T13:35:25.867+0900] {pod_manager.py:472} INFO - [base] (APIServer pid=21)   File "/opt/conda/lib/python3.11/site-packages/transformers/utils/hub.py", line 542, in cached_files
[2025-11-05T13:35:25.867+0900] {pod_manager.py:472} INFO - [base] (APIServer pid=21)     raise OSError(
[2025-11-05T13:35:25.867+0900] {pod_manager.py:472} INFO - [base] (APIServer pid=21) OSError: You are trying to access a gated repo.
[2025-11-05T13:35:25.867+0900] {pod_manager.py:472} INFO - [base] (APIServer pid=21) Make sure to have access to it at https://huggingface.co/google/gemma-2-2b-it.
[2025-11-05T13:35:25.867+0900] {pod_manager.py:472} INFO - [base] (APIServer pid=21) 401 Client Error. (Request ID: Root=1-690ad40d-6065067156b6c5ac666168e4;1c1791cc-fe12-495c-80a4-fcd647aed6a4)
[2025-11-05T13:35:25.867+0900] {pod_manager.py:472} INFO - [base] (APIServer pid=21)
[2025-11-05T13:35:25.867+0900] {pod_manager.py:472} INFO - [base] (APIServer pid=21) Cannot access gated repo for url https://huggingface.co/google/gemma-2-2b-it/resolve/main/config.json.
[2025-11-05T13:35:26.861+0900] {pod_manager.py:490} INFO - [base] (APIServer pid=21) Access to model google/gemma-2-2b-it is restricted. You must have access to it and be authenticated to access it. Please log in.

 

에러 요약

 

• OSError: You are trying to access a gated repo … google/gemma-2-2b-it … 401 Client Error
• 의미: Hugging Face 게이트드(승인 필요) 모델에 접근 중인데, vLLM 프로세스에서 인증 토큰을 인식 못 함.

 

---

 

핵심 원인

 

1. 모델이 게이트드 → 계정에서 접근 승인이 필요
2. Pod/컨테이너/프로세스 환경에 HF 토큰 미주입 또는 잘못된 변수명
3. 실행 유저 불일치(토큰은 A유저 홈, 프로세스는 B유저)
4. 네트워크/프록시로 huggingface.co 접근 실패

 

 

1) 접근 권한 확인

 

• 모델 페이지에서 Access Granted인지 확인(이미 승인되었다면 OK)

 

아직 해결 못함