Tensorflowtts: fastspeech2 data

[fly2mun]

hello.
i try to make the fastspeech2 model used by Ljspeech data set.
but i can not train any more.

1. try to ljspeech dataset.
   tensorflow-tts-preprocess
   tensorflow-tts-normalize
2. I can train tacotron2 by ljspeech.

But fastspeech2 train fail.

Traceback (most recent call last):
File "examples/fastspeech2/train_fastspeech2.py", line 412, in
main()
File "examples/fastspeech2/train_fastspeech2.py", line 325, in main
mel_length_threshold=mel_length_threshold,
File "./examples/fastspeech2/fastspeech2_dataset.py", line 101, in __init__
), f"Number of charactor, mel, duration, f0 and energy files are different"
AssertionError: Number of charactor, mel, duration, f0 and energy files are different

please help me.

[dathudeptrai]

you need extract duration before training FastSpeech2, you can use tacotron2 or MFA to do that.

[fly2mun]

I see, Thank you for your cooperation.