Error when run generate trie

DeepSpeech
#1

On MacOS, I have created all the files necessary for training a specific Vietnamese model and now i am stuck at trie generation.

I give these command:

/Users/tringuyen/Documents/DeepSpeech/DeepSpeech/generate_trie /
/Users/tringuyen/Documents/DeepSpeech/mymodels/alphabet.txt /
/Users/tringuyen/Documents/DeepSpeech/mymodels/mylm.binary /
/Users/tringuyen/Documents/DeepSpeech/mymodels

I get this error:

ERROR: VectorFst::Write: Write failed: <unspecifed>

Can anyone please help me?

#2

Hello and welcome! Have you taken a look at this tutorial? There are more explicit directions for interacting with KenLM and creation of a custom LM.
Tutorial

1 Like
#3

or this previously reported issue regarding the alphabet.txt

#4

Thank you @tuttlebr , I have fixed the error with the last param (the path to store the trie after generate).

#5

Sorry @tuttlebr, can you explain or give me some docs about the concept and the usage of 2 files ‘lm.binary’ & ‘trie’, what is the mission of this 2 files. Thank you and I hope that you can explain it for me.

#6

These files are used in the decoding process to score characters, words, and sequences of words in order to decide which transcription is more likely to be the true transcription of the audio.

1 Like
#7

For technical documentation: from the creator of KenLM https://kheafield.com/code/kenlm/

For theoretical documentation: more information about chain probabilities and language modeling

1 Like
#8

Thank guys for helping me to resolve my problem.