chineseocr的gpu版本的docker镜像已经实现， 采用接口方式
GandalfLiu opened this issue · comments
docker pull ld1995docker/chineseocr:latest
直接拉取镜像运行 docker run --gpus all -d -p 8080:8080 chineseocr:latest 最后在通过接口方式实现识别 localhost:8080/ocr 就是接口， 部署采用gunicorn
字段
import requests
files = {"file": ("file_name", open(path, "rb"), "image/png")}
requests.post(url, files=files)
配置
+-----------------------------------------------------------------------------+
| NVIDIA-SMI 418.67 Driver Version: 418.67 CUDA Version: 10.1 |
|-------------------------------+----------------------+----------------------+
| GPU Name Persistence-M| Bus-Id Disp.A | Volatile Uncorr. ECC |
| Fan Temp Perf Pwr:Usage/Cap| Memory-Usage | GPU-Util Compute M. |
|===============================+======================+======================|
| 0 Tesla P4 On | 00000000:00:08.0 Off | 0 |
| N/A 41C P0 23W / 75W | 7447MiB / 7611MiB | 0% Default |
+-------------------------------+----------------------+----------------------+
+-----------------------------------------------------------------------------+
| Processes: GPU Memory |
| GPU PID Type Process name Usage |
|=============================================================================|
| 0 15054 C /root/miniconda3/bin/python 7437MiB |
+-----------------------------------------------------------------------------+
机器要求：安装好了nvidia驱动， 可以不用安装cuda， 安装的docker 版本大于等于19.03 ，安装nvidia-docker github地址： https://github.com/NVIDIA/nvidia-docker
贴一下Dockerfile
FROM nvidia/cuda:9.0-cudnn7-devel
MAINTAINER xxx
LABEL version="1.0"
EXPOSE 8080
WORKDIR /ocr
ADD . /ocr
RUN apt-get update
RUN apt-get install libglib2.0 libsm6 libxrender1 libxext-dev gcc -y
# 下载miniconda并且将Miniconda3-latest-Linux-x86_64.sh放在chineseocr文件中使用3.6的python环境
RUN cd /ocr/chineseocr && sh -c '/bin/echo -e "\nyes\n\nyes" | sh Miniconda3-latest-Linux-x86_64.sh'
RUN echo -e "\ny" && /root/miniconda3/bin/conda config --add channels https://mirrors.tuna.tsinghua.edu.cn/anaconda/pkgs/free/
RUN /root/miniconda3/bin/conda config --add channels https://mirrors.tuna.tsinghua.edu.cn/anaconda/pkgs/main/
RUN /root/miniconda3/bin/conda config --set show_channel_urls yes
RUN /root/miniconda3/bin/conda install python=3.6
RUN echo -e "\ny" && /root/miniconda3/bin/conda install -c conda-forge mahotas
RUN echo -e "\ny" && /root/miniconda3/bin/pip install flask jieba easydict opencv-contrib-python==4.0.0.21 Cython h5py pandas requests bs4 gunicorn matplotlib lxml -i https://mirrors.aliyun.com/pypi/simple
RUN /root/miniconda3/bin/pip install web.py==0.40.dev0 -i https://mirrors.aliyun.com/pypi/simple
RUN /root/miniconda3/bin/pip install -U pillow -i https://mirrors.aliyun.com/pypi/simple
RUN /root/miniconda3/bin/pip install keras==2.1.5 tensorflow==1.8 tensorflow-gpu==1.8 -i https://mirrors.aliyun.com/pypi/simple
RUN cd /ocr/pack && /root/miniconda3/bin/pip install * -i https://mirrors.aliyun.com/pypi/simple
RUN cd /ocr/pack && rm *
RUN cd /root && mkdir logs && cd logs && touch access.log && touch error.log
CMD ["/root/miniconda3/bin/gunicorn", "--chdir", "/ocr/chineseocr", "example:app", "-c", "/ocr/chineseocr/gconfig.py"]
ENV LANG C.UTF-8 LC_ALL=C.UTF-8
近期会升级到tf2.0