Thomas Müller

I believe docker images corresponding to newer CUDA versions should work out of the box. If someone has the time to try one, confirm it working and make a pull...

Looks good to me. In the pull request, you're asking for someone to test -- I don't have a suitable setup to do this at the moment, so I will...

`scale: 1.0` should do the trick. That's the parameter that's 0.33 by default.

In that case, tweaking `scale` is probably not the answer and I'd recommend instead to downscale the vertices of the output mesh. Instant NGP makes use of several overlaid grids...

This sounds to me like your `transforms.json` have an inherently different scale across datasets (which carries through to the mesh). This is unavoidable if you use e.g. COLMAP to estimate...

There's no feature in tiny-cuda-nn that makes this deterministic, sorry. The reason that the backward pass through the hash encoding is non-deterministic in the first place is that colliding training...

Unfortunately, this seems like a case of compatibility hell to me -- ~I don't think there's a way to compile for compute capability 89 _and_ have a PyTorch version with...

Hi, I suspect the higher quality is a direct consequence of the SDF not being defined everywhere. This makes it easier for the model to fit it accurately near the...

Hi Dal, INGP already renders a depth buffer alongside the "main" rendering (as determined by `ERenderMode`). There's no Python binding for this, but you could for example modify the existing...

Hi, that is indeed unexpected. Which operating system are you on? If Windows, does the public binary release have DLSS working? Have you tried deleting the build folder and re-building...