Pytorch - very slow on: example Translation with a Sequence to Sequence Network and Attention

my model was really slow with pytorch so I tried one of your examples and saw it is slow as well (seq2seq.

It is still running. Has been running for at least an hour and still hasn’t finished…

I’m running only on CPU
I’m working on Ubuntu 16.04.3.
64 cores.
Python 3.6.3 :: Anaconda custom (64-bit).

tried using MKL:
installed via conda
then ran: export CMAKE_PREFIX_PATH=/…/anaconda3
didn’t work. I’m I doing it wrong?

Any ideas?