PaddleHub-GPU镜像制作

接前文，本文介绍制作PaddleOCR/PaddleHub的GPU版本Docker镜像，以便在GPU环境下高效运行。

环境依赖

🔔  镜像制作过程在下述环境中进行的，酌情参考。

基础环境

• ubuntu 22.04
• gcc/g++ 11.04
• Python 3.10
• paddlepaddle-gpu==2.6.1.post120
• paddleocr 2.7.5
• opencv-python 4.9.0.80
• paddlehub 2.4.0

GPU环境

• cuda 12.2
• Driver API Version: 12.4
• Runtime API Version: 11.8

环境变量

必须的步骤

export CUDA_VISIBLE_DEVICES=0

export LD_LIBRARY_PATH=/usr/local/cuda-12.2/lib64:$LD_LIBRARY_PATH
export PATH=/usr/local/cuda-12.2/bin:$PATH
export CUDA_HOME=/usr/local/cuda-12.2

基础镜像

拉取官方镜像并启动容器：

docker run -it --gpus all --name=paddlegpu -d paddlepaddle-gpu:2.6.1

说明：

如果无法拉取这个镜像（或嫌弃该镜像内容太多），也可以使用前文中生成的镜像 paddlehub:base ，然后逐步处理。

安装paddlepaddle-gpu

安装 CUDA12.0 包含 cuDNN 动态链接库的 PaddlePaddle

python3 -m pip install paddlepaddle-gpu==2.6.1.post120 -f https://www.paddlepaddle.org.cn/whl/linux/cudnnin/stable.html

验证

python3.10 -c "import paddle; paddle.utils.run_check()"

cuda和cuDNN安装完成后，验证正常如下：

root@b40610924ca2:/usr/local/cuda-12.2/targets/x86_64-linux/lib# python3.10 -c "import paddle; paddle.utils.run_check()"
Running verify PaddlePaddle program ... 
I0511 02:00:54.641423   814 program_interpreter.cc:212] New Executor is Running.
W0511 02:00:54.643276   814 gpu_resources.cc:119] Please NOTE: device: 0, GPU Compute Capability: 8.0, Driver API Version: 12.4, Runtime API Version: 11.8
W0511 02:00:54.643945   814 gpu_resources.cc:164] device: 0, cuDNN Version: 8.9.
I0511 02:00:55.097067   814 interpreter_util.cc:624] Standalone Executor is Used.
PaddlePaddle works well on 1 GPU.
PaddlePaddle is installed successfully! Let's start deep learning with PaddlePaddle now.

 可能出现的问题

 执行检测指令后出现异常响应：

root@e93431fbf153:/PaddleOCR# python3.10 -c "import paddle; paddle.utils.run_check()"
/usr/local/lib/python3.10/dist-packages/paddle/base/framework.py:688: UserWarning: You are using GPU version Paddle, but your CUDA device is not set properly. CPU device will be used by default.
  warnings.warn(
Running verify PaddlePaddle program ... 
[2024-05-10 13:17:24,497] [ WARNING] install_check.py:60 - You are using GPU version PaddlePaddle, but there is no GPU detected on your machine. Maybe CUDA devices is not set properly.
 Original Error is 
I0510 13:17:24.505074  4123 program_interpreter.cc:212] New Executor is Running.
I0510 13:17:24.539232  4123 interpreter_util.cc:624] Standalone Executor is Used.
PaddlePaddle works well on 1 CPU.
PaddlePaddle is installed successfully! Let's start deep learning with PaddlePaddle now.

解决

安装 CUDA 和 cuDNN。

• CUDA

如果宿主机安装了CUDA，可直接拷贝到容器内（宿主机上CUDA安装过程不演示）：

docker cp /usr/local/cuda paddlegpu:/usr/local

# 下面的步骤是因为有软连接。如果没有，则省略
# docker cp /usr/local/cuda-12.2 paddlegpu:/usr/local

##添加环境变量
export PATH=/usr/local/cuda/bin:$PATH
export LD_LIBRARY_PATH=/usr/local/cuda/lib64:$LD_LIBRARY_PATH

容器内查看CUDA安装结果：

root@b40610924ca2:/# nvcc --version
nvcc: NVIDIA (R) Cuda compiler driver
Copyright (c) 2005-2023 NVIDIA Corporation
Built on Tue_Aug_15_22:02:13_PDT_2023
Cuda compilation tools, release 12.2, V12.2.140
Build cuda_12.2.r12.2/compiler.33191640_0

说明：

如果基础镜像中包含了旧版本的CUDA，可直接删除。然后拷贝新版本使用。

• 安装cuDNN

容器内执行下面这个指令，确保执行成功：

python3 -m pip install paddlepaddle-gpu==2.6.1.post120 -f https://www.paddlepaddle.org.cn/whl/linux/cudnnin/stable.html

安装PoddleOCR

在容器内执行：

wget https://github.com/PaddlePaddle/PaddleOCR/archive/refs/tags/v2.7.5.tar.gz && \
    tar -zxvf v2.7.5.tar.gz && \
    mv PaddleOCR-2.7.5 PaddleOCR \
    && rm -rf v2.7.5.tar.gz \
    && cd PaddleOCR \
    && pip3.10 install -r requirements.txt -i https://mirror.baidu.com/pypi/simple \
    && mkdir -p /PaddleOCR/inference/ \
    && wget https://paddleocr.bj.bcebos.com/dygraph_v2.0/ch/ch_ppocr_mobile_v2.0_det_infer.tar && \
    tar xf ch_ppocr_mobile_v2.0_det_infer.tar -C /PaddleOCR/inference/ && \
    wget https://paddleocr.bj.bcebos.com/dygraph_v2.0/ch/ch_ppocr_mobile_v2.0_cls_infer.tar && \
    tar xf ch_ppocr_mobile_v2.0_cls_infer.tar -C /PaddleOCR/inference/ && \
    wget https://paddleocr.bj.bcebos.com/dygraph_v2.0/ch/ch_ppocr_mobile_v2.0_rec_infer.tar && \
    tar xf ch_ppocr_mobile_v2.0_rec_infer.tar -C /PaddleOCR/inference/ \
    && rm -rf ch_ppocr_mobile_v2.0_det_infer.tar \
    && rm -rf ch_ppocr_mobile_v2.0_cls_infer.tar \
    && rm -rf ch_ppocr_mobile_v2.0_rec_infer.tar

安装paddlehub

pip3.10 install --upgrade -i https://mirror.baidu.com/pypi/simple paddlehub

## 以下包 是执行hub serving install 时需要安装的包
pip3.10 install -i https://mirror.baidu.com/pypi/simple shapely
pip3.10 install -i https://mirror.baidu.com/pypi/simple pyclipper
pip3.10 install -i https://mirror.baidu.com/pypi/simple scikit-image
pip3.10 install -i https://mirror.baidu.com/pypi/simple imgaug
pip3.10 install -i https://mirror.baidu.com/pypi/simple lmdb

安装模块

#表格识别
hub install deploy/hubserving/structure_table

#文字识别
hub install deploy/hubserving/ocr_system

#版面识别
hub install deploy/hubserving/structure_layout

不报错，说明安装成功了。

说明：

其他模块按需安装。

其他问题

ImportError: libGL.so.1

File "/usr/local/lib/python3.10/dist-packages/paddlehub/vision/utils.py", line 18, in <module>
    import cv2
ImportError: libGL.so.1: cannot open shared object file: No such file or directory

需要安装如下包（容器内）：

apt-get update
apt-get install libgl1-mesa-glx -y
apt-get install libglib2.0-0 -y

降级protobuf

报错：

TypeError: Descriptors cannot be created directly.
If this call came from a _pb2.py file, your generated code is out of date and must be regenerated with protoc >= 3.19.0.
If you cannot immediately regenerate your protos, some other possible workarounds are:
 1. Downgrade the protobuf package to 3.20.x or lower.
…………

解决方案：降级

pip3.10 install -i https://mirror.baidu.com/pypi/simple protobuf==3.20.2

缺少模型文件

类似如下报错：

    raise ValueError(
ValueError: not find model.pdmodel or inference.pdmodel in ./inference/ch_PP-OCRv3_det_infer/

少哪个模型就下载、解压到inference目录下即可。

同时，修改params.py文件：

cd /PaddleOCR
vi deploy/hubserving/structure_table/params.py

def read_params():
    cfg = pp_ocr_read_params()

    # params for table structure model
    cfg.table_max_len = 488
    #cfg.table_model_dir = './inference/en_ppocr_mobile_v2.0_table_structure_infer/'
    cfg.table_model_dir = './inference/ch_ppstructure_mobile_v2.0_SLANet_infer/'
    #cfg.table_char_dict_path = './ppocr/utils/dict/table_structure_dict.txt'
    cfg.table_char_dict_path = './ppocr/utils/dict/table_structure_dict_ch.txt'
    cfg.show_log = False
    return cfg

版本问题

启动gpu版本的服务后，客户端调用时报错ÿ