Ilya Basharov

1) what if I will solve `torch-scatter` library problem ? 2) > many changes break backwards compatibility yeah, because `nerfstudio` does not fully support gaussians. It is more focused on...

> Hi @Ilyabasharov > > I'm doing some studies and I'm trying to test the ZipNerf in NerfStudio. I pulled the branch, did the installation again, updated the cli, but...

@jingyibo123 hello, yes, I set default implementation to `torch`. But it is working also with `tcnn` which is much faster

@kerrj Did you test it?  I've run nerfacto-huge and attached visualization at the late stages of training when the gradient BARF is no longer working

> Interesting, does this scene diverge without the scheduling for you as well? nono, all good without gradient BARF

> > @kerrj Did you test it?  I've run nerfacto-huge and attached visualization at the late stages of training when the gradient BARF is no longer working > >...

> Also how many scheduling steps did you use here? first 8k steps (figure taken at 40k)

> @Ilyabasharov you might want to try a lower number of scheduling steps to be less than the number of proposal annealing steps (which is 5k for nerfacto huge). If...

@Mxbonn Fully agree! #2242

 @AdamRashid96 New bug: the indexes of the current image samples and the indexes of the cached depth maps may not match at different launches. Therefore, either fix the train and...