search

p0p4k / vits2_pytorch

unofficial vits2-TTS implementation in pytorch
https://arxiv.org/abs/2307.16430
MIT License
477 stars 85 forks source link

how to know whether the model is fitting #66

Closed creed2415 closed 10 months ago

creed2415 commented 11 months ago

based on these metrics（[loss_disc, loss_gen, loss_fm, loss_mel, loss_dur, loss_kl,global_step, lr]） in the log

p0p4k commented 11 months ago

I would look at loss_mel, loss_dur, loss_kl. loss_mel is reconstructed wav's mel and real mel difference. loss_dur is aligner and dur_pred difference. loss_kl is the distance between p(z_mel | mel) and p(z_mel | text)