Hi I was able to implement the gstkaldi decoder successfullly. I have train a chain model on my own data and modified the yaml according to the sample_chinese_nnet3.yaml. However, I found that the decode result was slightly differernt from the online decoder in kaldi onlie2bin. Some words get wrong while the online2-wav-nnet3-latgent-faster was correct. Is there anything wrong with my config or any other reason?
Hi I was able to implement the gstkaldi decoder successfullly. I have train a chain model on my own data and modified the yaml according to the sample_chinese_nnet3.yaml. However, I found that the decode result was slightly differernt from the online decoder in kaldi onlie2bin. Some words get wrong while the online2-wav-nnet3-latgent-faster was correct. Is there anything wrong with my config or any other reason?