eesen
eesen copied to clipboard
Is it possible to use warp-ctc instead?
The benchmark in https://github.com/baidu-research/warp-ctc is much faster than eesen's implementation.
We are studying the validity of the comparisons they presented, specifically the linear correlation between time and N for Eesen.
Parallelization is still an issue for eesen, which needs to be addressed if it is be used massively.
X.
在 2016/1/18 23:37, Yajie Miao 写道:
We are studying the validity of the comparisons they presented, specifically the linear correlation between time and N for Eesen.
— Reply to this email directly or view it on GitHub https://github.com/srvk/eesen/issues/28#issuecomment-172563775.
Any news/results about it?
We've looked at it, and it does not seem to be that straightforward. Baidu's code is only CTC (no LSTM) so one would need to put a few more things in place before it'll be usable. We also had difficulty understanding the published comparisons. Yajie, I know you're traveling - maybe you could comment some more?