ming024/FastSpeech2

Minor bug in loading vocoders

Mahyar24 opened this issue · 0 comments

These three lines don't load the models into the desired device and should be replaced:

ckpt = torch.load(ckpt_path)

ckpt = torch.load("hifigan/generator_LJSpeech.pth.tar")

ckpt = torch.load("hifigan/generator_universal.pth.tar")

with something like this: torch.load(ckpt_path, map_location=device).