Closed wada-s closed 5 years ago
If I use your script as a vocoder, consider the following to specify the number of frames to be included in one batch. (In this case 16 frames are specified)
trn_loader = DataLoader(dataset, collate_fn=lambda batch: env.collate(0, 16, 0, batch), batch_size=batch_size, num_workers=2, shuffle=True, pin_memory=True)
Is this because you use 960 samples for learning like a paper?
Hi ! Thanks for sharing this repo!
I use your script as a vocoder implementation. I'm reading it but I don't know what seq_len is used for. Usually, I think that it is used as the number of samples put in one batch.