v1/bs_mega_53stem_back-vocal_mvsep_config.yaml

audio:
  chunk_size: 441000
  dim_f: 1024
  dim_t: 801
  hop_length: 441
  min_mean_abs: 0.0
  n_fft: 2048
  num_channels: 2
  sample_rate: 44100
inference:
  batch_size: 1
  chunk_size: 882000
  dim_t: 1101
  normalize: false
  num_overlap: 2
model:
  attn_dropout: 0.0
  depth: 12
  dim: 256
  dim_freqs_in: 1025
  dim_head: 64
  ff_dropout: 0.0
  flash_attn: true
  freq_transformer_depth: 1
  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
  heads: 8
  linear_transformer_depth: 0
  mask_estimator_depth: 2
  mlp_expansion_factor: 2
  multi_stft_hop_size: 147
  multi_stft_normalized: false
  multi_stft_resolution_loss_weight: 1.0
  multi_stft_resolutions_window_sizes: !!python/tuple
    - 4096
    - 2048
    - 1024
    - 512
    - 256
  num_stems: 1
  skip_connection: false
  stereo: true
  stft_hop_length: 512
  stft_n_fft: 2048
  stft_normalized: false
  stft_win_length: 2048
  time_transformer_depth: 1
  use_torch_checkpoint: false
training:
  batch_size: 1
  instruments:
    - back-vocal
    - other
  target_instrument: back-vocal
  use_amp: true