Imho, the stronghold of cuda is slowly eroding. Inference can run without it, and could so for years via ONNX. Now we are starting to see more back-ends becomi

Imho, the stronghold of cuda is slowly eroding. Inference can run without it, and could so for years via ONNX. Now we are starting to see more back-ends becoming available. see https://github.com/openxla https://github.com/openxla