julius
julius copied to clipboard
julius-speech
Training recipe for DNN-HMMs compatible with Julius
Hello
It is possible to publish a training recipe for generation of DNN-HMMs compatible with Julius? I am specifically interested in steps leading from triphone GMM-HMMs to DNN-HMMs which are a part of dictation toolkit. I am using HTK for training DNN-HMMs but it contains a very general recipe which I am not sure how compatible is with Julius acoustic model requirements.
Kind Regards Leslaw
I spent a lot of time reading recent scientific publications about Julius and DNN which were published by Kyoto University scientists. I have a good understanding on how DNN Julius implementation is working. Also recently HTK has published a beta version with support for DNN training. Is it possible to extend Julius to accept new versions of HTK models? Or at least could you provide the code or URL from which the Japanese training toolkit DNN models were based on? I am specifically interested how you converted baseline HTK model to Kaldi DNN model and then how you converted Kaldi model to the one that is currently supported by DNN Julius.
Hello,
I am interested in this thing too. I think a poorly documented SW easily become "no one want to use" thing. I chose Julius because Voxforge.org has a good tutorial. But I am thinking about Kaldi.
Bye Estal
Julius does not support ReLU yet, so you can't convert Kaldi DNN.
pingfengluo [email protected] ezt írta (időpont: 2019. júl. 13., Szo 7:08):
I am interested in this thing that Julius support Kaldi model too. @Akinobu https://github.com/Akinobu Lee
— You are receiving this because you commented. Reply to this email directly, view it on GitHub https://github.com/julius-speech/julius/issues/46?email_source=notifications&email_token=AHNHTWUGWCMQCKHV7TMJOOLP7FPMPA5CNFSM4CQNHIPKYY3PNVWWK3TUL52HS4DFVREXG43VMVBW63LNMVXHJKTDN5WW2ZLOORPWSZGODZ3JWHQ#issuecomment-511089438, or mute the thread https://github.com/notifications/unsubscribe-auth/AHNHTWVXLKKVV2OUW42TEDLP7FPMPANCNFSM4CQNHIPA .