Open PetrochukM opened 6 years ago
Hi @PetrochukM, Thanks for your interest but there's quite a lot of training still to be done on both the spectrogram network and the wavenet before any audio samples will be ready and I only have one gpu. Even then, I'm not likely to train it for the million iterations that google did. My goal is to verify that all of the bugs are hammered out and then supply checkpoints that will allow other interested people to carry on training the network if they so desire.
(referencing the comic) Same.
Thanks for letting me know where you're at. Once you've got it figured out, I'm happy to train it for millions of iterations and pass the weights on.
Do you mind posting some samples?
Thanks!