Error loading the _version_cpu.pyd file.

[DrMaxP]

I'm trying to get prior-depth-anything working on a Windows 11 machine. I have followed the setup guide and am attempting to run the sample - as detailed in this thread: SpatialVision/Prior-Depth-Anything#18

I'm hitting an issue where torch_cluster's __init__ routine fails to load the _version_cpu.pyd file from the environment - e.g. anaconda3\envs\priorda\Lib\site-packages\torch_cluster, giving me this error dialog:

This is with torch 2.2.2+cu121 and torch_cluster 1.6.3.

I've used various dependency walking tools to see if there is a missing dependency, but cannot find anything that looks wrong when I use those tools to load the pyd file. Why might it be failing in the python code? (Presumably that's using WinDll from ctypes).

Also, I have tried to import the cuda version of pytorch_cluster into the environment and I can see the code checks for pyd files with the _cpu or _cuda suffix in the init routine. Given I have CUDA installed - e.g. 12.1 - and an appropriate GPU (4090), why is it bringing in the CPU variant of torch_cluster and not the GPU/CUDA variant?

Apologies, my python coding is pretty rusty.