Transformer-TTS
Transformer-TTS copied to clipboard
soobinseo
A Pytorch Implementation of "Neural Speech Synthesis with Transformer Network"
Many thanks for your great work. I have a question about the Post-Net after the Mel Linear layer, I find in the inference stage, you use the mel_pred as output...
Can you kindly also add training curve plots for postnet? Need to know what was the loss during its convergence. Thanks in advance!
I tried to train using the LJSpeech dataset (i followed your instruction on how to) and when i try to synthesize, there is no sound or anything?
Here's the output after training postnet to 66000 steps and transformer to 272000 It's on a custom dataset that didn't have very good attention on tacotron2 but did have an...
Hi @soobinseo , when i run the file: python prepare_data.py . I am facing the issue: AttributeError: 'DataFrame' object has no attribute 'ix' What should i do? Best regards, PeterPham
Hello . I try to use your model to finish melspec to melspec conversion task. I try to add a guided attention like this : dal_loss = (hp.lamda_attn_dal * dal_g...
Dear soobinseo, Thanks for your wonderful work. I have a one question about your mel-spectrogram normalization method, after I look carefully at your work. From util.py, # normalize mel =...
Dear Soobinseo: Thank you for your wonderful work. When I run train_transformer.py after finished the first 3 steps you instructed in README, I got a RuntimeError: ``` (transformer-tts) xxx@zjlab:~/TTS/Transformer-TTS$ python...
Hello, I see you use 1,0000 epochs for training the model. How many days did you used? 10000 epochs seems really huge. Many thanks!
Hi, thanks for your contribution, Just had some questions, 1. Could you tell me where did you get this idea of query concatenation in mha?? Is it discussed somewhere?? paper??...
