装完cuda,可以用nvcc -V 来测试是否安装成功,
前提是要改变环境变量,长远点看,要添加一个lib,一个bin。
#cuda bin settingpycuda安装
官网教程:
https://wiki.tiker.net/PyCuda/Installation/Linux/Ubuntu
下载:
https://pypi.python.org/pypi/pycuda
------------------------------------------------
二、cuDNN 安装(多版本)
https://blog.csdn.net/xll_bit/article/details/78373576
下载地址:(我选择的是cuda-8.0对应cudnn-7.1)
https://developer.nvidia.com/rdp/cudnn-download#a-collapse705-8
将tgz文件解压到指定目录:
tar zxvf test.tgz -C 指定目录
修改环境变量.bashrc中最后一行加入:
export LD_LIBRARY_PATH=/your/path/to/cudnn/lib64:$LD_LIBRARY_PATH
source .bashrc
cudnn.h 要放在
cd cuda/include
sudo cp *.h /usr/local/cuda/include/
因为所有版本的这个文件都是一样的,所以不用单独写路径到 .bashrc
参考链接:
caffe 多个cudnn:
https://blog.csdn.net/zhaishengfu/article/details/52333674
最正确的姿势安装cudnn,网上大多数教程都太坑了:
https://blog.csdn.net/Lucifer_zzq/article/details/76675239
-------------------------------------------
三、安装TensorRT
下载地址:
https://developer.nvidia.com/nvidia-tensorrt-download
安装手册:
http://developer.download.nvidia.com/compute/machine-learning/tensorrt/secure/3.0/ga/TensorRT-Installation-Guide-3.0.4.pdf?SzMTX6bbvuAaT9j_gu0mVYmhuVqCdCHaZJ4xgojL_sYWmXelQ31tpvUYnpszWi-sU5dgXT0tPX3bRD5Crmy3MTxOK9tMZbNTlDU5QjsGr_X91LYDMSVV00W6HcpYaNcdBZVhxpPNOIwzz0ovb8ZzQFfPwl4Kxx6DDeDltyNpeR_6ttDIb0Tybx9nZGYFXEQfNyFUqA
安装步骤:
--------------------------------
四、
---------------------------------
五、几个官网教程
cuda & cublas
https://www.clear.rice.edu/comp422/resources/cuda/html/cublas/index.html#introduction
tensorrt guide
https://docs.nvidia.com/deeplearning/sdk/tensorrt-developer-guide/index.html#overview