Hi,
I just got started with OOD, could you please help me with integrating Tensorboard, I am on RHEL8.
Is tensorboard just as a package enough (via pip install) or should I install tensorboard as a module in my server?
Hi and welcome! You can install through pip if it’s global so that others can use it, though I’d suggest it as a module so that you have better control over versioning.
One issue with tensorboard though is it provides no authentication. So once you boot it up, it’s accessible to everyone and everything that can find it without any sort of security.
We’ve had these updates staged for a long time, but I just got them merged into our example tensorboard application below. Note that we do not use this in production, it was just an example that we hacked together so many years ago.
But within those hacks, we were able to provide some security through a network namespace and an authentication proxy provided by Stanford, though this requires subuids enabled on your system.
There could be easier ways to lock tensorboard down, I’m just not sure of them.