ubuntu16.04安装TensorRT

TensorRT是NVIDIA推出的一款深度学习推理优化器，可以方便深度学习模型在低成本的单个GPU机器或者一些嵌入式平台上的部署，这里介绍TensorRT在ubuntu16.04下的安装。

TensorRT下载地址：https://developer.nvidia.com/nvidia-tensorrt-5x-download 下载需要登录NVIDIA帐号

可以点击下载自己cuda对应的TensorRT版本，比如TensorRT-5.0.2.6.Ubuntu-16.04.4.x86_64-gnu.cuda-9.0.cudnn7.3.tar.gz

需要注意的是TensorRT的版本需要和cuda以及cudnn一致，TensorRT-5.0.2.6.对应cuda9.0和cudnn7.3。解压Tensorrt下载文件，并添加环境变量（这里的$HOME为TensrRT-5.0.26所在目录）

vim ~/.bashrc
export LD_LIBRARY_PATH=$HOME/TensorRT-5.0.2.6/lib:$LD_LIBRARY_PATH 
export CUDA_INSTALL_DIR=/usr/local/cuda-9.0 
export CUDNN_INSTALL_DIR=/usr/local/cuda-9.0
source ~/.bashrc

然后进入解压目录，安装TensorRT

#install TensorRT
cd $HOME/tensorrt-5.0.2.6/python
sudo pip3 install tensorrt-5.0.2.6-py2.py3-none-any.whl
#install UFF
cd ../uff
sudo pip3 install uff-0.5.5-py2.py3-none-any.whl
#install graphsurgeon
cd ../graphsurgeon
sudo pip3 install graphsurgeon-0.3.2-py2.py3-none-any.whl

测试TensorRT是否安装成功，进入Python编辑器导入tensorrt和uff模块

import tensorrt
import uff
print(tensorrt.__version__)
print(uff.__version__)

常见报错问题

ImportError: libnvinfer.so.5: cannot open shared object file: Nosuch file or directory
ImportError: libnvonnxparser.so.0: cannot open shared object file:No such file or directory
ImportError: libnvonnxparser_runtime.so.0: cannot open shared objectfile: No such file or directory
ImportError: libnvparsers.so.5: cannot open shared object file: Nosuch file or directory
ImportError: libnvinfer_plugin.so.5: cannot open shared object file:No such file or directory

解决办法 如下

sudo cp $HOME/TensorRT-5.0.2.6/targets/x86_64-linux-gnu/lib/libnvinfer.so.5 /usr/lib/
sudo cp $HOME/TensorRT-5.0.2.6/targets/x86_64-linux-gnu/lib/libnvonnxparser.so.0 /usr/lib/
sudo cp $HOME/TensorRT-5.0.2.6/targets/x86_64-linux-gnu/lib/libnvonnxparser_runtime.so.0 /usr/lib/
sudo cp $HOME/TensorRT-5.0.2.6/targets/x86_64-linux-gnu/lib/libnvparsers.so.5 /usr/lib/
sudo cp $HOME/TensorRT-5.0.2.6/targets/x86_64-linux-gnu/lib/ libnvinfer_plugin.so.5/usr/lib/

另外通过命令：which convert-to-uff   可以输出uff的安装路径

运行sample测试下TensorRT是否安装成功

cd TensorRT-5.1.2.2/samples/sampleMNIST 
make clean 
make 
cd /TensorRT-XXX/bin
./sample_mnist

显示如下结果，说明TensorRT已正常安装

[I] Output:
0: 
1: 
2: 
3: 
4: 
5: 
6: 
7: 
8: **********
9: 

&&&& PASSED TensorRT.sample_mnist # ./sample_mnist