Great catch! Our cloud machine encountered a cuda error (the GPU fell off PCIe) - had to restart it. It's back to normal now.

All the more reason to have a managed version of services :)