Closed younes-mirinezhad closed 1 month ago
I started training without validation and the model was trained.
Hi, may I ask how you were able to run the train process, I am having the same problem now?
@bdthanh can you try rerunning your mfa train
command with --clean
appended and see if that works?
@bdthanh can you try rerunning your
mfa train
command with--clean
appended and see if that works?
I solved it alr, thanks for your attention. It turns out my command was wrong at corpus path
Hi,
I need to use the MFA to align my voices and texts, so I have started to train the MFA for my target language.
First, I prepare a folder (speech_corpus) that contains .wav (voice) and .lab (transcript) files.
After that, I prepare a pronunciation dictionary for my target language (pronunciation_dictionary.txt)
Then, I install my working environment
As I see, my dataset has just one speaker so I use
--single_speaker
in all commands.Then I start to validate my data
mfa validate --single_speaker ./speech_corpus pronunciation_dictionary.txt
This is my validation output:As you can see after Iteration 40 of 40:
Generating alignments...
,Accumulating transition stats...
,Collecting phone and word alignments from monophone_ali lattices...
,Collecting training data...
doesn't finish. Whay?After that, I validate my pronunciation dictionary
mfa validate_dictionary --single_speaker pronunciation_dictionary.txt
And this is my outputs:Now, I want to train the model and I run this command
mfa train --single_speaker speech_corpus/ pronunciation_dictionary.txt newModel.zip
And this is my error:What is the problem and what should I do?