Svein Ove Aas

I'm seeing the same behaviour on Linux. (RTX 4090; NixOS) saya :: ~/AI/instant-ngp ‹master*› » nix-shell --run './instant-ngp data/nerf/fox/' 148 ↵ 00:38:53 SUCCESS Initialized CUDA. Active GPU is #0: NVIDIA...

Is this still a problem on 3.11, or is upgrading Python a potential solution?

On the flip side, when I'm generating just a small fraction of an image, I almost certainly want it to be 512x512. A button for that would also be nice.

This is probably due to me going through the class and marking lots of things private. I'll add some kind of actual API before release, okay? ...and probably un-private it.

My argument, of course, is that treating it as an input would be more useful. Ideally I suppose it should be possible to choose.

I have about 270 GB of various checkpoint files, hardlinked into various UIs and branched repositories. The idea of converting all of them is... hoo boy. Deleting the originals isn't...

It's likely caused by the 0.15.x branch of Fabric. They fixed a bug that was causing a lot of client code to be loaded on the server, which of course...

That's a good point. Probably the most robust way would be to try each method in turn, swallowing exceptions if they fail; the _most_ robust way to see if xdg-open...

xdg-open might exist, but not work, in which case it'll exit with a nonzero return code. You'll also need to check for that. I'd see if I write a patch...

Some observations: - The problem persists between restarts of the NVML-using program. - If there is only a single compute process running, and I restart it, then the problem temporarily...