Closed qq12cvhj closed 4 years ago
Hi @qq12cvhj , It should be fixed in latest commit. However, you should use this repo for learning purpose only, since it has bugs as noted in README and it only supported 2-stems model.
You should use [original repo] (https://github.com/deezer/spleeter) for production usage. Otherwise, you might try similar projects like demux or open-unmix.
Good job.I saved the model converted by pytorch,just to run it on mobile platforms,however tensorflow do not work on these platforms.So I choose your edition.Ok,I just use it in research work,not a productive env.I'll try it.
Here in the example file the L<512,so it is right.but what if the L > 512?There will be exceptions be thrown. mask = mask.permute([0, 3, 2, 1])[:, :, :L, 0].unsqueeze(-1)