Embedding Policy Results

This policy looks very promising

I tried to train my model, but it needs some serious processing power to even finish one epoch. Is this expected? I am using the same hyperparams as in your paper.