liusongxiang / ppg-vc

PPG-Based Voice Conversion
Apache License 2.0
328 stars 72 forks source link

Help with "any to many voice conversion with location relative seq2seq modeling" paper #35

Open neginvahidi33 opened 11 months ago

neginvahidi33 commented 11 months ago

Dear Dr. Songxiang Liu,

I am trying to use your code for the one-shot VC and train linglf02mel seq2seq methods. However, I am getting a runtime error in the solver.exec() function. I think the error is caused by the fact that the if self.step > sekf.max_step == 1 loop does not terminate, and the code continues to run.

I think the error may be caused by incorrect data in the config file for the train fid lists. I would like to ask for your guidance on the following:

  1. Is the train fid list file a Python file or a text file that contains the paths to the vctk data? Are the vctk data audio or text?

  2. Is this the same for the dev and eval fid lists?

  3. Are the vctk_ppg_dir and libri_ppg_dir directories the output of the compute_ctc_att_bnf file?

  4. Are the vctk_f0_dir and libri_f0_dir directories the output of the compute_f0 file?

  5. Are the vctk_wav_dir and libri_wav_dir directories also audio datasets?

  6. Are the libri_spk_dvec_dir and libri_spk_devc_dir directories the output of the compute_spk_devcs file?

I would be very grateful if you could provide me with any guidance that you can. I will never forget your help.

Thank you, Negin Vahidi