Open h770347 opened 2 years ago
There are other models at https://kaldi-asr.org/models.html but just spent an hour going through some of them. Their file structures do not look like the example yaml file (the Tedlium RNNLM) or when they do have the same types of files (mdl, mat, txt, HCLC) like for example the Gigaspeech XL model, the worker log gives an error Expected token "<DIMENSION>", got instead "<Nnet3>"
that seems to require yaml/Kaldi knowledge, or perhaps a newer version of kaldi-gstreamer-server than this image was built on. If anyone is able to use this docker image as of 2023 for English speech recognition, any hints or pointers are appreciated.
wget https://phon.ioc.ee/~tanela/tedlium_nnet_ms_sp_online.tgz converted 'https://phon.ioc.ee/~tanela/tedlium_nnet_ms_sp_online.tgz' (ANSI_X3.4-1968) -> 'https://phon.ioc.ee/~tanela/tedlium_nnet_ms_sp_online.tgz' (UTF-8) --2022-01-05 19:37:12-- https://phon.ioc.ee/~tanela/tedlium_nnet_ms_sp_online.tgz Resolving phon.ioc.ee (phon.ioc.ee)... 193.40.251.126 Connecting to phon.ioc.ee (phon.ioc.ee)|193.40.251.126|:443... failed: Connection refused.