greg@FX11:~$ docker pull nvcr.io/nvidia/tensorflow:17.10
FATA[0000] Post http:///var/run/docker.sock/v1.18/images/create?fromImage=nvcr.io%2Fnvidia%2Ftensorflow%3A17.10: dial unix /var/run/docker.sock: no such file or directory. Are you trying to connect to a TLS-enabled daemon without TLS?
You will not be able to run the NGC containers on a Windows or MacOS system. NGC GPU enabled containers require nvidia-docker to run and are only supported on Linux systems.
Incidentally, it would appear from your error message that you do not have docker configured correctly on your Win 10 system and that is causing your error on docker pull. Rather than try to fix your Windows docker installation, I suggest you move to Linux to work with NGC.
I have tensorflow w/ GPU running native on an AWS instance, Linux 17 (my work computer) and Windows 10 (my home computer). So I was curious if I could run the gpu optimized version under docker under WSL.
Right now the Tensorflow-cpu only version runs on WSL, but without GPU support and without docker. I was just trying to figure out if it’d be easier to maintain w/ docker. It sounds like not all the system calls that are needed for GPU support are implemented.
Anyways, for my home computer and by your comments, it sounds like it’s better to run Tensorflow with GPU support natively rather than using the NGC docker images.