测试环境目录: 10.182.17.2:/data1/tlw/MinerU
export MINERU_MODEL_SOURCE=modelscope
uvicorn pdf_parser_with_image_process_async:app --host 0.0.0.0 --port 8018 --workers 1 --timeout-keep-alive 1800
export MINERU_MODEL_SOURCE=modelscope
CUDA_VISIBLE_DEVICES=1 MINERU_FORMULA_ENABLE=false mineru-api --host 0.0.0.0 --port 8007 --enable-vlm-preload true --gpu-memory-utilization 0.35
生产环境目录: 10.51.0.1:/data/MinerU
Nginx 配置文件: /etc/nginx/conf.d/mineru_api_pool.conf
export MINERU_MODEL_SOURCE=modelscope
uvicorn pdf_parser_with_image_process_async:app --host 0.0.0.0 --port 8018 --workers 1 --timeout-keep-alive 1800
解析模型(4个节点,端口8010~8013,NGINX端口8007):
CUDA_VISIBLE_DEVICES=4 mineru-api --host 0.0.0.0 --port 8010 --enable-vlm-preload true
`MinerU_1、MinerU_2、MinerU_3,方法同上