3090 error during training in w Cuda 11.6

[vade]

Describe the bug Hi

I am trying to run through the install and setup. I was able to install nerfstudio on my 3x 3090 box with driver 510.85.02 and coda 11.6 (and have functioning PyTorch installed in other conda envs).

Im unable to run

ns-train nerfacto --data data/nerfstudio/poster :

NVIDIA GeForce RTX 3090 with CUDA capability sm_86 is not compatible with the current PyTorch installation. The current PyTorch install supports CUDA capabilities sm_37 sm_50 sm_60 sm_70. If you want to use the NVIDIA GeForce RTX 3090 GPU with PyTorch, please check the instructions at https://pytorch.org/get-started/locally/

If I update to the latest released pytorch via git, I then get:

NameError: name 'tcnn' is not defined

What is the expected manner to install with an upgraded coda runtime / torch?

Ive tried editing the nerfstudio.egg_info/requirements.txt but must be missing some dependency layer and it keeps reseting torch to 1.12.1

Thanks for any guidance.