Hello, quick question, on my cluster I can load either CUDA libraries of versions 11.8 or 12.6. Would sleap-nn work with CUDA 12.6, provided I add the setting to pyproject.toml to pull in the corresponding Torch build? I'm not very familiar with the entire Torch+CUDA framework and whether there could be differences in functionality between 12.6 and 12.8 that SLEAP may depend on. So, while it seems to build successfully, I'm not sure if I should be concerned about odd failures down the line due to forward incompatibility.

Also, less importantly, how big of a difference is it to use 11 vs 12? Is it worth it to try to use 12, or are the differences minor enough that I shouldn't worry and should rather just match the version fully and go with 11.8?

Thanks!