Please add the allowedCudaVersions filter to templates for GPU pods. Many applications these days require at least CUDA 12.1 to function correctly and if users don't select the correct CUDA version(s) when deploying their pods, the applications don't work. Being able to add the allowedCudaVersions filter to a template will help solve this problem.
#Add allowedCudaVersions filter to templates
olive bison
olive bison
This is more of a priority now that A1111 uses torch 2.1.2 because xformers 0.0.23.post1 requires CUDA 12.1, so I need to be able to specify the CUDA version in ny template so that users don't get CUDA 11.8 or 12.0.
This was added to serverless under Advanced but it should be at a template level, not at an endpoint level.