meshed-memory-transformer
meshed-memory-transformer copied to clipboard
The model runs quite slowly
Hi @marcellacornia,
I have tried to build a demo using M2 transformer because I found that the model worked very well. Unfortunately, when I tried to make inference using device = CPU, it took about 30 seconds for 64 character length sequence, I guess the beam-search algorithm is the reason why the inference consumed lots of time. Do you have any idea that I could enhance the performance of this implementation?
How long does it take to train the network?