Mineru Docker镜像

Mineru Docker镜像

打包了几个镜像并封装了api服务供调用。
镜像打包完成后可在离线环境下使用。

英伟达GPU版本

Dockerfile

FROM python:3.10-bookworm
LABEL authors="chunf"
RUN rm -f /etc/apt/sources.list.d/debian.sources
# sources.list为清华大学debian12镜像源
COPY ./sources.list /etc/apt/sources.list
RUN apt-get update
# 安装libreoffice和相关字体
RUN apt-get install libreoffice -y
RUN apt-get install libreoffice-l10n-zh-cn fonts-wqy-zenhei fonts-noto-cjk -y
# Fonts.zip为本人windows电脑所有字体文件
COPY ./Fonts.zip /tmp/Fonts.zip
RUN unzip -n -d /usr/share/fonts /tmp/Fonts.zip && fc-cache -fv && rm -f /tmp/Fonts.zip
# opencv等库需要
RUN apt-get install libgl1-mesa-glx -y

# 安装python相关包
RUN pip config set global.index-url https://mirrors.aliyun.com/pypi/simple
RUN pip install torch==2.3.1 torchvision==0.18.1 torchaudio==2.3.1 --index-url https://download.pytorch.org/whl/cu118
RUN pip install -U "magic-pdf[full]" -i https://mirrors.tuna.tsinghua.edu.cn/pypi/web/simple --extra-index-url https://wheels.myhloli.com &&  \
    pip install fastapi uvicorn python-multipart xlrd modelscope
RUN pip uninstall paddlepaddle -y && \
    pip install paddlepaddle-gpu==3.0.0b1 -i https://www.paddlepaddle.org.cn/packages/stable/cu118/

# 识别过程中的配置文件和模型文件，请自行下载
COPY ./magic-pdf-cuda.json /root/magic-pdf.json
COPY ./modelscope /root/.cache/modelscope
COPY ./.paddleocr /root/.paddleocr

• 模型和相关配置文件也可以使用挂载的方式，不必打包到镜像。

docker-compose

version: "3"
services:
  mineru:
    image: hcf/mineru-gpu
    container_name: mineru-gpu
    command:
      - python
      - /app/main.py
    volumes:
      - ./app:/app
      - ./output:/output
      - ./magic-pdf-cuda.json:/root/magic-pdf.json
    environment:
      - MAGIC_FILE_OUTPUT_DIR=/output
    ports:
      - "8000:8000"
    deploy:
      resources:
        reservations:
          devices:
            - driver: nvidia
              count: 1
              capabilities: [ gpu ]

• app目录为FastAPI服务源代码。

CPU版本

Dockerfile

FROM python:3.10-bookworm
LABEL authors="chunf"
RUN rm -f /etc/apt/sources.list.d/debian.sources
COPY ./sources.list /etc/apt/sources.list
RUN apt-get update
RUN apt-get install libreoffice -y
RUN apt-get install libreoffice-l10n-zh-cn fonts-wqy-zenhei fonts-noto-cjk -y
COPY ./Fonts.zip /tmp/Fonts.zip
RUN unzip -n -d /usr/share/fonts /tmp/Fonts.zip && fc-cache -fv && rm -f /tmp/Fonts.zip
RUN apt-get install libgl1-mesa-glx -y
RUN pip config set global.index-url https://mirrors.aliyun.com/pypi/simple
RUN pip install torch==2.3.1 torchvision==0.18.1 torchaudio==2.3.1 --index-url https://download.pytorch.org/whl/cpu
RUN pip install -U "magic-pdf[full]" --extra-index-url https://wheels.myhloli.com
RUN pip install fastapi uvicorn python-multipart xlrd modelscope

COPY ./magic-pdf-cpu.json /root/magic-pdf.json
COPY ./modelscope /root/.cache/modelscope
COPY ./.paddleocr /root/.paddleocr

docker-compose

version: "3"
services:
  mineru:
    image: hcf/mineru
    container_name: mineru-cpu
    command:
      - python
      - /app/main.py
    volumes:
      - ./app:/app
      - ./output:/output
      - ./magic-pdf-cpu.json:/root/magic-pdf.json
    environment:
      - MAGIC_FILE_OUTPUT_DIR=/output
    ports:
      - "8000:8000"

DCU版本

FROM image.sourcefind.cn:5000/dcu/admin/base/pytorch:2.3.0-ubuntu22.04-dtk24.04.3-py3.10
LABEL authors="HuangChunfeng"

RUN apt-get update
RUN apt-get install libreoffice -y
RUN apt-get install libreoffice-l10n-zh-cn fonts-wqy-zenhei fonts-noto-cjk -y
RUN apt-get install libgl1-mesa-glx -y

# 新增其他字体
COPY ./Fonts.zip /tmp/Fonts.zip
RUN unzip -n -d /usr/share/fonts /tmp/Fonts.zip && fc-cache -f -v && rm -rf /tmp/Fonts.zip


RUN pip install -U "magic-pdf[full]" --index-url https://mirrors.tuna.tsinghua.edu.cn/pypi/web/simple --extra-index-url https://wheels.myhloli.com
RUN pip install fastapi uvicorn python-multipart modelscope xlrd
RUN pip uninstall paddlepaddle -y

COPY ./paddlepaddle-2.6.1+das.opt1.dtk24043-cp310-cp310-manylinux_2_28_x86_64.whl /tmp/paddlepaddle-2.6.1+das.opt1.dtk24043-cp310-cp310-manylinux_2_28_x86_64.whl
RUN pip install /tmp/paddlepaddle-2.6.1+das.opt1.dtk24043-cp310-cp310-manylinux_2_28_x86_64.whl && rm -f /tmp/paddlepaddle-2.6.1+das.opt1.dtk24043-cp310-cp310-manylinux_2_28_x86_64.whl

COPY ./magic-pdf-cuda.json /root/magic-pdf.json
COPY ./modelscope /root/.cache/modelscope
COPY ./.paddleocr /root/.paddleocr

GPU版本离线测试

windows docker-desktop测试正常。