2.7.6版本按照如下配置启动，还是无法切换到cpu

[walker202004-netizen]

export MINERU_DEVICE_MODE=cpu
export CUDA_VISIBLE_DEVICES=""
mineru-api --host 0.0.0.0 --port 8000

2026-03-11 01:47:45.392 | WARNING | mineru.utils.pdf_page_id:get_end_page_id:8 - end_page_id is out of range, use images length
2026-03-11 01:47:46.038 | INFO | mineru.utils.engine_utils:get_vlm_engine:32 - Using vllm-async-engine as the inference engine for VLM.
2026-03-11 01:47:47,362 - modelscope - INFO - Intra-cloud acceleration enabled for downloading from OpenDataLab/MinerU2.5-2509-1.2B
Downloading Model from https://www.modelscope.cn to directory: /root/.cache/modelscope/hub/models/OpenDataLab/MinerU2.5-2509-1.2B
2026-03-11 01:47:47,910 - modelscope - INFO - Target directory already exists, skipping creation.
INFO 03-11 01:47:48 [init.py:220] No platform detected, vLLM is running on UnspecifiedPlatform
WARNING 03-11 01:47:49 [_custom_ops.py:20] Failed to import from vllm._C with ImportError('libcuda.so.1: cannot open shared object file: No such file or directory')
2026-03-11 01:47:49.675 | INFO | mineru.backend.vlm.utils:enable_custom_logits_processors:31 - CUDA not available, disabling custom_logits_processors

[dosubot[bot]]

Hi @walker202004-netizen! I'm Dosu and I’m helping the MinerU team.

设置 MINERU_DEVICE_MODE=cpu 是不够的，因为默认的 vllm-async-engine 后端不支持纯 CPU 运行。你需要使用 --backend pipeline 参数：

MINERU_DEVICE_MODE=cpu mineru-api --host 0.0.0.0 --port 8000 --backend pipeline

MINERU_DEVICE_MODE 环境变量只对 pipeline 和 vlm-transformers 后端生效，而 vLLM 相关后端必须依赖 GPU。支持 CPU 的后端包括：pipeline、vlm-transformers 和 http-client。

如果需要使用 VLM 能力但没有本地 GPU，也可以考虑轻量级部署方式，通过 vlm-http-client 连接远程 GPU 服务器。

To reply, just mention @dosu.

---

^{How did I do? Good | Irrelevant | Incorrect | Verbose | Hallucination | Report 🐛 | Other}

   

[walker202004-netizen]

换了一个报错信息：

MINERU_DEVICE_MODE=cpu mineru-api --host 0.0.0.0 --port 8000 --backend pipeline
Start MinerU FastAPI Service: http://0.0.0.0:8000
API documentation: http://0.0.0.0:8000/docs
INFO: Started server process [12810]
INFO: Waiting for application startup.
INFO: Application startup complete.
INFO: Uvicorn running on http://0.0.0.0:8000 (Press CTRL+C to quit)
2026-03-11 01:59:09.854 | ERROR | mineru.cli.fast_api:parse_pdf:417 - mineru.cli.common.aio_do_parse() got multiple values for keyword argument 'backend'
Traceback (most recent call last):

[dosubot[bot]]

这是一个已知问题，--backend 参数在命令行和 API 内部配置中重复传递导致冲突。

解决方法是不要在启动命令中指定 --backend，而是在 API 请求时传递：

启动服务：

MINERU_DEVICE_MODE=cpu mineru-api --host 0.0.0.0 --port 8000

调用 API 时指定 backend：

curl -X POST http://localhost:8000/file_parse \
  -F "files=@/path/to/file.pdf" \
  -F "backend=pipeline"

这样将基础设施配置（host/port）和请求参数（backend）分开就可以避免冲突了。

To reply, just mention @dosu.

---

^{How did I do? Good | Irrelevant | Incorrect | Verbose | Hallucination | Report 🐛 | Other}