torch-mlir
torch-mlir copied to clipboard
Force double precision in aten.mean.dim
Accuracy for aten.mean.dim was off for large inputs for the same reason as aten.var.correction (see #1100)
Looks like Refbackend fails to run the test on GHA ? Should we skip it by default ?
Is the error not reproducible locally?
Closing as this is fixed by #1605