Yesterday's trained model no good today?

PL 0.13.1 The model was trained yesterday; after browser restart (and ctrl+F5 after previews first failed to appear) we end up with this from Data Settings. The PL instance has not been restarted since yesterday (yet!)
.
Any ideas?

My suspicion - PC sleep/wake: problem with previews went away after restarting the PL server - the PC had slept overnight. (Do I recall some bug that related to a resume from sleep/hibernation issue?)

image

Hmm, maybe the PL server (specifically the kernel) freezes after the PC goes to sleep/wakes up.
Looking at the error logs which are sent to us, it seems like the kernel didn’t respond to any of your frontend requests (they all failed).

There was a bug previously related to training a long time and PL automatically logging you out, and ages ago we had issues with the kernel freezing if training too long, but I don’t think we have dug into the effects of the computer going to sleep and waking up before.

:+1:
It’s only 1 datapoint so far, but at least anyone looking here might see that a server restart might be all that is required.