Been using the hub.docker.com public nvidia/cuda containers for quite some time now without fault, but last night I think a rogue push might have broken the development containers:
❯ docker run -ti nvidia/cuda:10.2-devel-ubuntu18.04 /bin/bash
root@e5b73e77f905:/# nvcc
bash: nvcc: command not found
root@e5b73e77f905:/# gcc
bash: gcc: command not found
root@e5b73e77f905:/# exit
❯ docker run -ti nvidia/cuda:11.0-devel-ubuntu18.04-rc /bin/bash
root@78d084c08d90:/# nvcc
bash: nvcc: command not found
root@78d084c08d90:/# gcc
bash: gcc: command not found
root@78d084c08d90:/#
My guess is this wasn’t intended. Any idea when they’ll be back up? I certainly didn’t expect the 10.2-devel container to change suddenly; it’s been pretty stable for a while.