How can I speed up the model ?

[allenz-me]

I download the model and deploy it in my server.

In my tests, it takes several seconds(usually 1 - 3 s) to get the result for a sentence of a medium length, even with a gpu on.

How can I speed up the model ?