Now I am using pytorch1.0 in win10 with cuda9.0. I’d like to use tensorboard to visualize the training progress and check the network structure.
However, I find pytorch1.3 have the official support for tensorboard.
But pytorch1.3 need cuda9.2.
What should I do?
If you have the driver version >= 398.26, then you could install pytorch 1.3. cuda 9.2 here doesn’t refer to the cuda version at the user side, but it means the version of cuda used during build.
God, I use search tool in win10, and I find the nvidia-smi!
Here the driver version is 432, I’ll try to install pytorch1.3 in my PC.
Thanks a lot
It works.
I install pytorch1.3.1 with cuda10 and cudnn7.5 in the GTX1080
And we need to install the tensorboard with version higher 1.14.0