alphacep / vosk-api

Offline speech recognition API for Android, iOS, Raspberry Pi and servers with Python, Java, C# and Node
Apache License 2.0
7.74k stars 1.09k forks source link

about some model questions #908

Closed Veango closed 2 years ago

Veango commented 2 years ago

hello,i notice that code if model has hclg.fst, it doesn't use decode.fst decoder = new kaldi::SingleUtteranceNnet3Decoder(model->nnet3_decodingconfig, model_->transmodel, model_->decodableinfo, model_->hclgfst ? model_->hclgfst : decodefst, featurepipeline); my question is 1:if the model has hclg.fst, it certainly hasn't hclr.fst and gr.fst fisles? 2: if the model just have hclg.fst, it doesn't support set the grammar?

nshmyrev commented 2 years ago

if the model has hclg.fst, it certainly hasn't hclr.fst and gr.fst fisles?

Yes, although it is a good idea to have all of them probably.

2: if the model just have hclg.fst, it doesn't support set the grammar?

Yes

Veango commented 2 years ago

thank you so much