Given the documentation stating that
I would expect matmul to be implemented for complex tensors, however when I try to execute the following:
a = torch.tensor([[1.4 + 3j, 2 + 5j], [1.4 + 3j, 2 + 5j]], dtype=torch.cfloat) a @ a
RuntimeError: _th_addmm_out not supported on CPUType for ComplexFloat.
This also happens when using
torch.mm instead of the short operator.
Am I doing something wrong?
Edit: The error occurs on both CPU and GPU.