config.yaml

audio:
  chunk_size: 960000
  dim_f: 1024
  dim_t: 801 # don't work (use in model)
  hop_length: 441 # don't work (use in model)
  n_fft: 2048
  num_channels: 2
  sample_rate: 44100
  min_mean_abs: 0.0001

model:
  dim: 256
  depth: 12
  stereo: true
  num_stems: 1
  time_transformer_depth: 1
  freq_transformer_depth: 1
  linear_transformer_depth: 0
  freqs_per_bands: !!python/tuple
    - 2
    - 2
    - 2
    - 2
    - 2
    - 2
    - 2
    - 2
    - 2
    - 2
    - 2
    - 2
    - 2
    - 2
    - 2
    - 2
    - 2
    - 2
    - 2
    - 2
    - 2
    - 2
    - 2
    - 2
    - 4
    - 4
    - 4
    - 4
    - 4
    - 4
    - 4
    - 4
    - 4
    - 4
    - 4
    - 4
    - 12
    - 12
    - 12
    - 12
    - 12
    - 12
    - 12
    - 12
    - 24
    - 24
    - 24
    - 24
    - 24
    - 24
    - 24
    - 24
    - 48
    - 48
    - 48
    - 48
    - 48
    - 48
    - 48
    - 48
    - 128
    - 129
  dim_head: 64
  heads: 8
  attn_dropout: 0.0
  ff_dropout: 0.0
  flash_attn: true
  dim_freqs_in: 1025
  stft_n_fft: 2048
  stft_hop_length: 512
  stft_win_length: 2048
  stft_normalized: false
  mask_estimator_depth: 2
  multi_stft_resolution_loss_weight: 1.0
  multi_stft_resolutions_window_sizes: !!python/tuple
  - 4096
  - 2048
  - 1024
  - 512
  - 256
  multi_stft_hop_size: 147
  multi_stft_normalized: False
  mlp_expansion_factor: 4
  use_torch_checkpoint: True
  skip_connection: False


training:
  batch_size: 1
  gradient_accumulation_steps: 1
  grad_clip: 0
  instruments: ['vocals', 'instrument']
  lr: 1.0e-5
  patience: 5
  reduce_factor: 0.9
  target_instrument: instrument
  num_epochs: 1000
  num_steps: 1000
  q: 0.95
  coarse_loss_clip: true
  ema_momentum: 0.999
  optimizer: adam
  other_fix: false # it's needed for checking on multisong dataset if other is actually instrumental
  use_amp: true # enable or disable usage of mixed precision (float16) - usually it must be true


inference:
  batch_size: 2
  dim_t: 1876
  num_overlap: 4