Pytorch（GPU加速）框架:Ubuntu18.04+Anaconda3+NVIDIA430+cuda10+cuDNN+pytorch

在安装anaconda3时,磁盘空间不足,无奈给ubuntu扩容,插上安装盘后进入试用版ubuntu,打开gparted扩容,参考博客:

https://blog.csdn.net/weixin_40920349/article/details/80964825

https://blog.csdn.net/jx1605/article/details/80624599

 

Pytorch（GPU加速）框架安装过程参考 https://blog.csdn.net/wuzhiwuweisun/article/details/82753403

注意:

1. 显卡驱动nividia必须和cuda版本匹配:

电脑自带显卡驱动nividia390,安装cuda10.0之后发现显卡驱动nividia390和cuda10.0版本不匹配,参考博客 https://blog.csdn.net/zhw864680355/article/details/90411288 找到各对应版本,因为不想降低gcc版本而选择安装和cuda10.0匹配的nividia430,参考博客 https://blog.csdn.net/u014682691/article/details/80605201

 

2. gcc g++版本必须和cuda版本匹配:

https://www.cnblogs.com/tiandsp/p/9441494.html

 

3.pytorch 直接conda命令安装会失败,用pip轮子安装可以

https://pytorch.org/get-started/previous-versions/