Open Fluqz opened 3 years ago
Hey dude, thanks for your reply. Im a bit late, but did you manage to use your audio files? I sure am interested in the files, if we find a way to share 40gb :D
Unfortunately not. I only had access to a newer machine which didn't support python 2 or TF1.5, so I had to tweak the nsynth training pipeline to make it run on python 3. I did manage to make it run on python 3 with the later TF versions but the training performance was really poor - would have run for a month on a A100 machine! This is probably a matter of GPU drivers as the same tweaked pipeline trained 3x as fast on Colab GPU (with Pro subscription).
Hey, I built a NSynth 1-2 years ago and always had the problem that id like to upload more or different samples, but I have a 15 year old windows computer, which will probably take weeks to process the audio. I also dont have linux installed. The NSynth Sound Maker is the only project I found with different audio files, but I think its not publicly available.
Id really appreciate if anyone would share his audio files!
Thanks in advance :)