config.yaml

global:
  sr:                   44100     # Hz
  segment:              11.85     # seconds
  shift:                2         # seconds
  sample_rate:          44100     # Hz
  n_workers:            16
  prefetch_factor:      6


inference_models:
  kick:     'pretrained_larsnet_models/kick/pretrained_kick_unet.pth'
  snare:    'pretrained_larsnet_models/snare/pretrained_snare_unet.pth'
  toms:     'pretrained_larsnet_models/toms/pretrained_toms_unet.pth'
  hihat:    'pretrained_larsnet_models/hihat/pretrained_hihat_unet.pth'
  cymbals:  'pretrained_larsnet_models/cymbals/pretrained_cymbals_unet.pth'


data_augmentation:
  augmentation_prob:          0.5
  kit_swap_augment_prob:      0.5
  doubling_augment_prob:      0.3
  pitch_shift_augment_prob:   0.3
  saturation_augment_prob:    0.3
  channel_swap_augment_prob:  0.5
  remix_augment_prob:         0.3


kick:
  F:                    2048
  T:                    512
  batch_size:           24
  learning_rate:        1e-4
  epochs:               22
  training_mode:        'stft'
  model_id:             'default_kick_unet'


snare:
  F:                    2048
  T:                    512
  batch_size:           24
  learning_rate:        1e-4
  epochs:               22
  training_mode:        'stft'
  model_id:             'default_snare_unet'


toms:
  F:                    2048
  T:                    512
  batch_size:           24
  learning_rate:        1e-4
  epochs:               22
  training_mode:        'stft'
  model_id:             'default_toms_unet'


hihat:
  F:                    2048
  T:                    512
  batch_size:           24
  learning_rate:        1e-4
  epochs:               22
  training_mode:        'stft'
  model_id:             'default_hihat_unet'


cymbals:
  F:                    2048
  T:                    512
  batch_size:           24
  learning_rate:        1e-4
  epochs:               22
  training_mode:        'stft'
  model_id:             'default_cymbals_unet'