How to train with utf-8 flag

DeepSpeech
#1

I Just Want to know about How to use utf-8 flag.

In documentation, utf-8 flag ignore alphabet. but it isn’t.

So I make alphabet.txt File.

for example,
00
01
02
…
ff

or

\x00
\x01
…
\xff

but it is not work

So, How do I set the alphabet file to use the UTF-8 flag?

#2

You don’t.

You are obviously wrong here.

This flag is here for exploring usage on some specific languages. You need to read the code and understand what it does before using it. It’s 100% likely you don’t have to use that flag in your case.

#3

I am going to make a Korean STT model.
When using the utf-8 flag, I mistook it for an alphabet.txt problem because errors occur without the alphabet.txt path.
But today I realized the real problem and solved it.
The problem was identified in the process of passing python str to C code in mvs_ctcdecoder, and it seems to be solved by encoding the str to utf-8.