2026.5.10：百度的baidu/Qianfan-OCR进行推理，包括识别图片、pdf等

百度的baidu/Qianfan-OCR进行推理，包括识别图片、pdf等

需要的环境:
Ubuntu-26.04
GPU
uv

1.配置pyproject.toml

[project] name = "Qianfan-OCR-Inference" version = "0.1.0" description = "Qianfan-OCR-Inference" readme = "README.md" requires-python = ">=3.13,<3.14" dependencies = [ "accelerate>=1.13.0", "huggingface-hub>=1.14.0", "torch", "torchaudio", "torchvision", "vllm>=0.20.1", ] [tool.uv.sources] torch = [ { index = "pytorch-cu128", marker = "sys_platform == 'linux' or sys_platform == 'win32'" }, ] torchvision = [ { index = "pytorch-cu128", marker = "sys_platform == 'linux' or sys_platform == 'win32'" }, ]