Closed nackjaylor closed 1 year ago
This seems to have fixed itself... I shall close this issue, but it appears that the eval code may have some robustness issues (which is strange saying considering it has worked fine for months with no problem... very odd).
Hi team,
It seems some updates to either JAX or Linux in some way shape or form is playing havoc with the render code provided in MultiNeRF. As it stands, I am able to train, but can no longer render/eval from NeRFs using this codebase.
Is anyone else facing similar issues? I've been able to replicate across 2 different machines and from code running in a container.
Any chance you could take a look into the inference code and see what might be the probable cause?
Cheers,
Jack