mmdeploy框架环境配置教程

• 1 通过 MIM 安装 MMCV
  略

• 2 安装 MMDeploy 和 推理引擎

• 2.1 Linux-x86_64, CPU, ONNX Runtime 1.8.1

# 安装 MMDeploy TensorRT 自定义算子库和推理 SDK
wget https://github.com/open-mmlab/mmdeploy/releases/download/v0.9.0/mmdeploy-0.9.0-linux-x86_64-cuda11.1-tensorrt8.2.3.0.tar.gz
tar -zxvf mmdeploy-0.9.0-linux-x86_64-cuda11.1-tensorrt8.2.3.0.tar.gz
cd mmdeploy-0.9.0-linux-x86_64-cuda11.1-tensorrt8.2.3.0
pip install dist/mmdeploy-0.9.0-py3-none-linux_x86_64.whl
pip install sdk/python/mmdeploy_python-0.9.0-cp38-none-linux_x86_64.whl
cd ..
# 安装推理引擎 TensorRT
# !!! 从 NVIDIA 官网下载 TensorRT-8.2.3.0 CUDA 11.x 安装包并解压到当前目录
pip install TensorRT-8.2.3.0/python/tensorrt-8.2.3.0-cp38-none-linux_x86_64.whl
pip install pycuda
export TENSORRT_DIR=$(pwd)/TensorRT-8.2.3.0
export LD_LIBRARY_PATH=${TENSORRT_DIR}/lib:$LD_LIBRARY_PATH
# !!! 从 NVIDIA 官网下载 cuDNN 8.2.1 CUDA 11.x 安装包并解压到当前目录
export CUDNN_DIR=$(pwd)/cuda
export LD_LIBRARY_PATH=$CUDNN_DIR/lib64:$LD_LIBRARY_PATH

• 2.2 Linux-x86_64, CUDA 11.x, TensorRT 8.2.3.0

# 安装 MMDeploy TensorRT 自定义算子库和推理 SDK
wget https://github.com/open-mmlab/mmdeploy/releases/download/v0.9.0/mmdeploy-0.9.0-linux-x86_64-cuda11.1-tensorrt8.2.3.0.tar.gz
tar -zxvf mmdeploy-0.9.0-linux-x86_64-cuda11.1-tensorrt8.2.3.0.tar.gz
cd mmdeploy-0.9.0-linux-x86_64-cuda11.1-tensorrt8.2.3.0
pip install dist/mmdeploy-0.9.0-py3-none-linux_x86_64.whl
pip install sdk/python/mmdeploy_python-0.9.0-cp38-none-linux_x86_64.whl
cd ..
# 安装推理引擎 TensorRT
# !!! 从 NVIDIA 官网下载 TensorRT-8.2.3.0 CUDA 11.x 安装包并解压到当前目录
pip install TensorRT-8.2.3.0/python/tensorrt-8.2.3.0-cp38-none-linux_x86_64.whl
pip install pycuda
export TENSORRT_DIR=$(pwd)/TensorRT-8.2.3.0
export LD_LIBRARY_PATH=${TENSORRT_DIR}/lib:$LD_LIBRARY_PATH
# !!! 从 NVIDIA 官网下载 cuDNN 8.2.1 CUDA 11.x 安装包并解压到当前目录
export CUDNN_DIR=$(pwd)/cuda
export LD_LIBRARY_PATH=$CUDNN_DIR/lib64:$LD_LIBRARY_PATH

问题点1： 安装pip install pycuda报错

解决办法：

1、下载pycuda源码
wget https://files.pythonhosted.org/packages/5a/56/4682a5118a234d15aa1c8768a528aac4858c7b04d2674e18d586d3dfda04/pycuda-2021.1.tar.gz
2、解压tar -xvzf pycuda-2021.1.tar.gz
3、cd pycuda-2021.1
4、vim setup.py
5、在下图位置添加conf[“CXXFLAGS”].append(’-std=c++11’)
6、运行python3 setup.py install

export PATH=/usr/local/cuda-10.2/bin:/usr/local/cuda/bin:$PATH
source ~/.bashrc

参考：
Could not build wheels for pycuda which use PEP 517 and cannot be install
src/cpp/cuda.hpp:14:10: fatal error: cuda.h: No such file or directory