facebookresearch/demucs

apply_model run time

mathpopo opened this issue · 5 comments

i try to use separate(htdemuc),but i meet : i put 100ms / 1s ,apply_model run time is same,how can i do to see different?

What system are you using? What is your CPU and GPU?

ubuntu 20.04, GPU, other issue:all model point:
stereo,44100?
i put a mono 16000, but ->torch audio load : stereo,44100

padded_mix = mix.padded(valid_length), should here again enlange the length, end set segment "1", just 44100
so:

  1. 16k mono->stereo,44100
  2. length->padded

because segment,so every epoch is same time

What? Are you TRAINING demucs?

@CarlGao4 no, now i just test , debug all struct, understand how toadd log