Is MKL a must-have to run torch.fft() on a non-GPU environment?

I try the official conda release of PyTorch (1.8.0) on Mac with M1 chip. Everything runs smoothly except the torch.fft()module that complains about MKL.

I was just wondering if there is some way to run torch.fft() without MKL or MKL
is a prerequisite for it.