alphacep / vosk-api

Offline speech recognition API for Android, iOS, Raspberry Pi and servers with Python, Java, C# and Node
Apache License 2.0
7.38k stars 1.04k forks source link

training with vosk problem #1424

Open Mary123447 opened 11 months ago

Mary123447 commented 11 months ago

Hello, I am trying to perform a training of a Vosk model. I ran the code in https://github.com/alphacep/vosk-api/blob/master/python/example/colab/vosk-training.ipynb and after several attempts I was able to run a training for a number of epochs and get the output model. I got a folder containing several files but I noticed that, compared to the pre-defined models, the two files HCLr.fst and Gr.fst are missing. I then tried to run the code https://github.com/alphacep/vosk-api/blob/master/python/example/colab/vosk-adaptation.ipynb initially trying to reproduce the vosk-model-small-en-us-0.15-compile-colab.tar.gz with the data obtained from the previous training but the compile-graph.sh show these errors: Screenshot 2023-08-04 193231 Later I tried the code with the default folder but also in this case there are problems with converting HCLG.fst to HCLr.fst and to Gr.fst, can anyone help me to solve these problems?