test-mlir-cuda* CI jobs pass but are useless
The "test-mlir-cuda*" CI jobs pass, but since there are no runners with GPUs (and I also believe the docker image for mlir-cuda is missing some things), they simply fall back to generating code for CPU. That means that the test-mlir-cpu tests basically run twice, which is useless, moreover, the fact that jobs named "test-mlir-cuda" pass is misleading, since they're not testing CUDA at all like I said.
Is it possible to disable or skip these tests for the moment ?