Gramvaani importer file is not working

DeepSpeech
#21

okay also tell me after making custome language model i can change beam width lm alpha and lm beta anytime or i need to generate binary and trie file everytime i change those 3 hyperparameters? Asking cause i am not seeing any change in inference after changing value of hyperparameters.
i have included words like “three cheers chocolate” iin language model still it is giving result as “three cold” it is taking cold from “cold coffee”.

2)What are the values limit of lm apha,lm beta and beam width?
As i have around 200 words so should i decrease beam width which is currently 500?

#22

You can change the hyperparameters any time, they’re not tied to the files. There’s no explicit limits to the values. Beam width occurs at the timestep level, not word level, it’s unrelated to the size of your vocabulary, but you should definitely experiment with different values to see what works best for your use case.