Pytorch install and cudnn issue

1

I just upgraded cudnn to 7.0.5 from 6.0.21 for cuda 8.0 on my Titan X GPU (Ubuntu 14.04). When I install pytorch in using the command ‘conda install pytorch torchvision -c pytorch’, I still get packages for earlier cudnn version 6.0.21. Can someone please tell me what I am doing wrong?

Some diagnostic commands are below

(kme-torch) watts@Maximus:~/Downloads$ cat /usr/local/cuda/include/cudnn.h | grep CUDNN_MAJOR -A 2
#define CUDNN_MAJOR 7
#define CUDNN_MINOR 0
#define CUDNN_PATCHLEVEL 5

#define CUDNN_VERSION (CUDNN_MAJOR * 1000 + CUDNN_MINOR * 100 + CUDNN_PATCHLEVEL)

#include “driver_types.h”
(kme-torch) watts@Maximus:~/Downloads$

(kme-torch) watts@Maximus:~/Downloads$ conda install pytorch torchvision -c pytorch
Fetching package metadata …
Solving package specifications: .

All requested packages already installed.

packages in environment at /home/watts/anaconda2/envs/kme-torch:

pytorch 0.1.12 py35cuda8.0cudnn6.0_1
torchvision 0.1.8 py35_0
(kme-torch) watts@Maximus:~/Downloads$

2

Hi,

You need to upgrade your conda to get the latest version of pytorch (and cudnn).