nerfstudio

nerfstudio copied to clipboard

I am using the ns-train command on custom data & edited nerf algorithm. But I keep getting an "ImportError".

Steps to reproduce the behavior:

1. Created a new conda environment. The new environment is different than another conda environment that has been working perfectly with nerfstudio.
2. Copied existing nerf algorithm and tweaked it a bit in a new .py file located in the models folder
3. Added custom nerf algorithm name in method_configs to call it using the command line
4. Used command ns-train custom_name --data ./Video_apt/ --viewer.websocket-port 7008 (used same command on original conda environment)

It is supposed to run a copy of nerfacto. I tried using nerfacto and the edited nerfacto but they keep raising the same "ImportError" error about "cannot import name 'VanillaDataManagerConfig' from 'nerfstudio.data.datamanagers'" in this conda environment. But when I run nerfacto in the original conda environment, it works.

I have tried changing different variables in the method_configs file to see if any new code is messing up the project. I have also tried digging deep into the ns-train, train, base_datamanagers, and method_configs file to locate the problem, to no avail.

instead of using ns-train, could you try using python scripts/train.py?

Can you try the latest github main version. It has some fixes regarding circular imports that may fix this issue.

Thanks, sorry for the late replies but neither option worked unfortunately.

Adding my partner @SanakovDiar who can test the data to this issue thread.

Hi, any updates?

Hi, any updates?

Hello, yes - unfortunately neither option worked. The error message remained the same.

Can you try pulling the latest github version and then post the error you are seeing.

Closing due to inactivity. Feel free to reopen if it is still and issue.