Hansss

Results 7 issues of Hansss

I think there's a bug in preprocess.py

https://github.com/ming024/FastSpeech2/blob/d4e79eb52e8b01d24703b2dfc0385544092958f3/preprocessor/preprocessor.py#L77-L90 On line 77, If there is no tg_path, the previous pitch, energy, and frame information will be included in the statistics. If there is no tg_path from the beginning,...

How about new parameter 'mex_mel_len' to replace 'max_seq_len'?

Hello, this is my first time to be a TTS model. I understand the max_seq_len parameter as the maximum phoneme length. because when phoneme data larger than max_seq_len is given...

영어 처리 과정에서 오류가 있습니다. [english.py]

https://github.com/Kyubyong/g2pK/blob/3bb9d5afc5159220d5d16492aca7a58f121b6073/g2pk/english.py#L20 정규식 내에 따옴표 ' 때문에 영어 처리가 제대로 이루어지지 않습니다. 위를 그대로 사용할 경우 **Input** : the shawshank redemption'이다. 언뜻 생각하면 'escape'를 썼을 법한데 'redemption'을 썼다. redemption의 사전적 의미는...

Fix: remove redefining duration_discriminator_type variable

refer to https://github.com/p0p4k/vits2_pytorch/issues/84#issue-2212649665

why 'duration_discriminator_type' is redefined?

https://github.com/p0p4k/vits2_pytorch/blob/1f4f3790568180f8dec4419d5cad5d0877b034bb/train_ms.py#L172-L188 Due to this line, the `net_dur_disc` maintain None

Stage 2 Training Fails with NaN Loss on Single GPU Due to Inconsistent Checkpoint Keys

### Description When trying to start stage 2 training after completing stage 1 using a **single** A100 80GB GPU with My Korean dataset, I encountered an issue where `g_loss` becomes...

feat: Improve model checkpoint loading

### Summary This PR fixes the checkpoint loading issue in the second stage of training when using a **single GPU.** The second stage adds a **'module.'** prefix to all parameter...