Closed yaotingwangofficial closed 4 months ago
@yaotingwangofficial That might need some extra optimization on the engineering side. Since this is only a research-based project, so I do not conduct thorough optimization on the inference speed. You are most welcome to contribute on that if you want of course.
Thanks!
Thanks for your cool work, I tried some samples and the performance is really good and love the token-saving. But seems the decoding process is too low compared to others, is there any possible method to accelerate the decoding into <1s for decoding around 10s speech? Many thanks!