Upload 23 files

configs/config_dnr_bandit_bsrnn_multi_mus64.yaml

@@ -0,0 +1,78 @@
+name: "MultiMaskMultiSourceBandSplitRNN"
+audio:
+  chunk_size: 264600
+  num_channels: 2
+  sample_rate: 44100
+  min_mean_abs: 0.001
+
+model:
+  in_channel: 1
+  stems: ['speech', 'music', 'effects']
+  band_specs: "musical"
+  n_bands: 64
+  fs: 44100
+  require_no_overlap: false
+  require_no_gap: true
+  normalize_channel_independently: false
+  treat_channel_as_feature: true
+  n_sqm_modules: 8
+  emb_dim: 128
+  rnn_dim: 256
+  bidirectional: true
+  rnn_type: "GRU"
+  mlp_dim: 512
+  hidden_activation: "Tanh"
+  hidden_activation_kwargs: null
+  complex_mask: true
+  n_fft: 2048
+  win_length: 2048
+  hop_length: 512
+  window_fn: "hann_window"
+  wkwargs: null
+  power: null
+  center: true
+  normalized: true
+  pad_mode: "constant"
+  onesided: true
+
+training:
+  batch_size: 4
+  gradient_accumulation_steps: 4
+  grad_clip: 0
+  instruments:
+  - speech
+  - music
+  - effects
+  lr: 9.0e-05
+  patience: 2
+  reduce_factor: 0.95
+  target_instrument: null
+  num_epochs: 1000
+  num_steps: 1000
+  q: 0.95
+  coarse_loss_clip: true
+  ema_momentum: 0.999
+  optimizer: adam
+  other_fix: true # 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
+
+augmentations:
+  enable: true # enable or disable all augmentations (to fast disable if needed)
+  loudness: true # randomly change loudness of each stem on the range (loudness_min; loudness_max)
+  loudness_min: 0.5
+  loudness_max: 1.5
+  mixup: true # mix several stems of same type with some probability (only works for dataset types: 1, 2, 3)
+  mixup_probs: !!python/tuple # 2 additional stems of the same type (1st with prob 0.2, 2nd with prob 0.02)
+    - 0.2
+    - 0.02
+  mixup_loudness_min: 0.5
+  mixup_loudness_max: 1.5
+  all:
+    channel_shuffle: 0.5 # Set 0 or lower to disable
+    random_inverse: 0.1 # inverse track (better lower probability)
+    random_polarity: 0.5 # polarity change (multiply waveform to -1)
+
+inference:
+  batch_size: 1
+  dim_t: 256
+  num_overlap: 4

configs/config_htdemucs_6stems.yaml

@@ -0,0 +1,127 @@
+audio:
+  chunk_size: 485100 # samplerate * segment
+  min_mean_abs: 0.001
+  hop_length: 1024
+
+training:
+  batch_size: 8
+  gradient_accumulation_steps: 1
+  grad_clip: 0
+  segment: 11
+  shift: 1
+  samplerate: 44100
+  channels: 2
+  normalize: true
+  instruments: ['drums', 'bass', 'other', 'vocals', 'guitar', 'piano']
+  target_instrument: null
+  num_epochs: 1000
+  num_steps: 1000
+  optimizer: adam
+  lr: 9.0e-05
+  patience: 2
+  reduce_factor: 0.95
+  q: 0.95
+  coarse_loss_clip: true
+  ema_momentum: 0.999
+  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
+
+augmentations:
+  enable: true # enable or disable all augmentations (to fast disable if needed)
+  loudness: true # randomly change loudness of each stem on the range (loudness_min; loudness_max)
+  loudness_min: 0.5
+  loudness_max: 1.5
+  mixup: true # mix several stems of same type with some probability (only works for dataset types: 1, 2, 3)
+  mixup_probs: [0.2, 0.02]
+  mixup_loudness_min: 0.5
+  mixup_loudness_max: 1.5
+  all:
+    channel_shuffle: 0.5 # Set 0 or lower to disable
+    random_inverse: 0.1 # inverse track (better lower probability)
+    random_polarity: 0.5 # polarity change (multiply waveform to -1)
+
+inference:
+  num_overlap: 4
+  batch_size: 8
+
+model: htdemucs
+
+htdemucs:  # see demucs/htdemucs.py for a detailed description
+  # Channels
+  channels: 48
+  channels_time:
+  growth: 2
+  # STFT
+  num_subbands: 1
+  nfft: 4096
+  wiener_iters: 0
+  end_iters: 0
+  wiener_residual: false
+  cac: true
+  # Main structure
+  depth: 4
+  rewrite: true
+  # Frequency Branch
+  multi_freqs: []
+  multi_freqs_depth: 3
+  freq_emb: 0.2
+  emb_scale: 10
+  emb_smooth: true
+  # Convolutions
+  kernel_size: 8
+  stride: 4
+  time_stride: 2
+  context: 1
+  context_enc: 0
+  # normalization
+  norm_starts: 4
+  norm_groups: 4
+  # DConv residual branch
+  dconv_mode: 3
+  dconv_depth: 2
+  dconv_comp: 8
+  dconv_init: 1e-3
+  # Before the Transformer
+  bottom_channels: 0
+  # CrossTransformer
+  # ------ Common to all
+  # Regular parameters
+  t_layers: 5
+  t_hidden_scale: 4.0
+  t_heads: 8
+  t_dropout: 0.0
+  t_layer_scale: True
+  t_gelu: True
+  # ------------- Positional Embedding
+  t_emb: sin
+  t_max_positions: 10000 # for the scaled embedding
+  t_max_period: 10000.0
+  t_weight_pos_embed: 1.0
+  t_cape_mean_normalize: True
+  t_cape_augment: True
+  t_cape_glob_loc_scale: [5000.0, 1.0, 1.4]
+  t_sin_random_shift: 0
+  # ------------- norm before a transformer encoder
+  t_norm_in: True
+  t_norm_in_group: False
+  # ------------- norm inside the encoder
+  t_group_norm: False
+  t_norm_first: True
+  t_norm_out: True
+  # ------------- optim
+  t_weight_decay: 0.0
+  t_lr:
+  # ------------- sparsity
+  t_sparse_self_attn: False
+  t_sparse_cross_attn: False
+  t_mask_type: diag
+  t_mask_random_seed: 42
+  t_sparse_attn_window: 400
+  t_global_window: 100
+  t_sparsity: 0.95
+  t_auto_sparsity: False
+  # Cross Encoder First (False)
+  t_cross_first: False
+  # Weight init
+  rescale: 0.1
+

configs/config_musdb18_bs_roformer.yaml

@@ -0,0 +1,134 @@
+audio:
+  chunk_size: 131584
+  dim_f: 1024
+  dim_t: 256
+  hop_length: 512
+  n_fft: 2048
+  num_channels: 2
+  sample_rate: 44100
+  min_mean_abs: 0.001
+
+model:
+  dim: 192
+  depth: 6
+  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: 0.1
+  ff_dropout: 0.1
+  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
+
+training:
+  batch_size: 10
+  gradient_accumulation_steps: 1
+  grad_clip: 0
+  instruments:
+  - vocals
+  - bass
+  - drums
+  - other
+  lr: 5.0e-05
+  patience: 2
+  reduce_factor: 0.95
+  target_instrument: vocals
+  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
+
+augmentations:
+  enable: true # enable or disable all augmentations (to fast disable if needed)
+  loudness: true # randomly change loudness of each stem on the range (loudness_min; loudness_max)
+  loudness_min: 0.5
+  loudness_max: 1.5
+
+inference:
+  batch_size: 1
+  dim_t: 256
+  num_overlap: 4

configs/config_musdb18_demucs3_mmi.yaml

@@ -0,0 +1,72 @@
+audio:
+  chunk_size: 485100 # samplerate * segment
+  min_mean_abs: 0.000
+  hop_length: 1024
+
+training:
+  batch_size: 8
+  gradient_accumulation_steps: 1
+  grad_clip: 0
+  segment: 11
+  shift: 1
+  samplerate: 44100
+  channels: 2
+  normalize: true
+  instruments: ['drums', 'bass', 'other', 'vocals']
+  target_instrument: null
+  num_epochs: 1000
+  num_steps: 1000
+  optimizer: adam
+  lr: 9.0e-05
+  patience: 2
+  reduce_factor: 0.95
+  q: 0.95
+  coarse_loss_clip: true
+  ema_momentum: 0.999
+  other_fix: false # it's needed for checking on multisong dataset if other is actually instrumental
+  use_amp: false # enable or disable usage of mixed precision (float16) - usually it must be true
+
+augmentations:
+  enable: true # enable or disable all augmentations (to fast disable if needed)
+  loudness: true # randomly change loudness of each stem on the range (loudness_min; loudness_max)
+  loudness_min: 0.5
+  loudness_max: 1.5
+
+inference:
+  num_overlap: 4
+  batch_size: 8
+
+model: hdemucs
+
+hdemucs:  # see demucs/hdemucs.py for a detailed description
+  channels: 48
+  channels_time: null
+  growth: 2
+  nfft: 4096
+  wiener_iters: 0
+  end_iters: 0
+  wiener_residual: False
+  cac: True
+  depth: 6
+  rewrite: True
+  hybrid: True
+  hybrid_old: False
+  multi_freqs: []
+  multi_freqs_depth: 3
+  freq_emb: 0.2
+  emb_scale: 10
+  emb_smooth: True
+  kernel_size: 8
+  stride: 4
+  time_stride: 2
+  context: 1
+  context_enc: 0
+  norm_starts: 4
+  norm_groups: 4
+  dconv_mode: 1
+  dconv_depth: 2
+  dconv_comp: 4
+  dconv_attn: 4
+  dconv_lstm: 4
+  dconv_init: 0.001
+  rescale: 0.1

configs/config_musdb18_htdemucs.yaml

@@ -0,0 +1,119 @@
+audio:
+  chunk_size: 485100 # samplerate * segment
+  min_mean_abs: 0.001
+  hop_length: 1024
+
+training:
+  batch_size: 8
+  gradient_accumulation_steps: 1
+  grad_clip: 0
+  segment: 11
+  shift: 1
+  samplerate: 44100
+  channels: 2
+  normalize: true
+  instruments: ['drums', 'bass', 'other', 'vocals']
+  target_instrument: null
+  num_epochs: 1000
+  num_steps: 1000
+  optimizer: adam
+  lr: 9.0e-05
+  patience: 2
+  reduce_factor: 0.95
+  q: 0.95
+  coarse_loss_clip: true
+  ema_momentum: 0.999
+  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
+
+augmentations:
+  enable: true # enable or disable all augmentations (to fast disable if needed)
+  loudness: true # randomly change loudness of each stem on the range (loudness_min; loudness_max)
+  loudness_min: 0.5
+  loudness_max: 1.5
+
+inference:
+  num_overlap: 4
+  batch_size: 8
+
+model: htdemucs
+
+htdemucs:  # see demucs/htdemucs.py for a detailed description
+  # Channels
+  channels: 48
+  channels_time:
+  growth: 2
+  # STFT
+  num_subbands: 1
+  nfft: 4096
+  wiener_iters: 0
+  end_iters: 0
+  wiener_residual: false
+  cac: true
+  # Main structure
+  depth: 4
+  rewrite: true
+  # Frequency Branch
+  multi_freqs: []
+  multi_freqs_depth: 3
+  freq_emb: 0.2
+  emb_scale: 10
+  emb_smooth: true
+  # Convolutions
+  kernel_size: 8
+  stride: 4
+  time_stride: 2
+  context: 1
+  context_enc: 0
+  # normalization
+  norm_starts: 4
+  norm_groups: 4
+  # DConv residual branch
+  dconv_mode: 3
+  dconv_depth: 2
+  dconv_comp: 8
+  dconv_init: 1e-3
+  # Before the Transformer
+  bottom_channels: 512
+  # CrossTransformer
+  # ------ Common to all
+  # Regular parameters
+  t_layers: 5
+  t_hidden_scale: 4.0
+  t_heads: 8
+  t_dropout: 0.0
+  t_layer_scale: True
+  t_gelu: True
+  # ------------- Positional Embedding
+  t_emb: sin
+  t_max_positions: 10000 # for the scaled embedding
+  t_max_period: 10000.0
+  t_weight_pos_embed: 1.0
+  t_cape_mean_normalize: True
+  t_cape_augment: True
+  t_cape_glob_loc_scale: [5000.0, 1.0, 1.4]
+  t_sin_random_shift: 0
+  # ------------- norm before a transformer encoder
+  t_norm_in: True
+  t_norm_in_group: False
+  # ------------- norm inside the encoder
+  t_group_norm: False
+  t_norm_first: True
+  t_norm_out: True
+  # ------------- optim
+  t_weight_decay: 0.0
+  t_lr:
+  # ------------- sparsity
+  t_sparse_self_attn: False
+  t_sparse_cross_attn: False
+  t_mask_type: diag
+  t_mask_random_seed: 42
+  t_sparse_attn_window: 400
+  t_global_window: 100
+  t_sparsity: 0.95
+  t_auto_sparsity: False
+  # Cross Encoder First (False)
+  t_cross_first: False
+  # Weight init
+  rescale: 0.1
+

configs/config_musdb18_mdx23c.yaml

@@ -0,0 +1,182 @@
+audio:
+  chunk_size: 261120
+  dim_f: 4096
+  dim_t: 256
+  hop_length: 1024
+  n_fft: 8192
+  num_channels: 2
+  sample_rate: 44100
+  min_mean_abs: 0.001
+
+model:
+  act: gelu
+  bottleneck_factor: 4
+  growth: 128
+  norm: InstanceNorm
+  num_blocks_per_scale: 2
+  num_channels: 128
+  num_scales: 5
+  num_subbands: 4
+  scale:
+  - 2
+  - 2
+
+training:
+  batch_size: 6
+  gradient_accumulation_steps: 1
+  grad_clip: 0
+  instruments:
+  - vocals
+  - bass
+  - drums
+  - other
+  lr: 9.0e-05
+  patience: 2
+  reduce_factor: 0.95
+  target_instrument: null
+  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
+
+augmentations:
+  enable: true # enable or disable all augmentations (to fast disable if needed)
+  loudness: true # randomly change loudness of each stem on the range (loudness_min; loudness_max)
+  loudness_min: 0.5
+  loudness_max: 1.5
+  mixup: true # mix several stems of same type with some probability (only works for dataset types: 1, 2, 3)
+  mixup_probs: !!python/tuple # 2 additional stems of the same type (1st with prob 0.2, 2nd with prob 0.02)
+    - 0.2
+    - 0.02
+  mixup_loudness_min: 0.5
+  mixup_loudness_max: 1.5
+
+  # apply mp3 compression to mixture only (emulate downloading mp3 from internet)
+  mp3_compression_on_mixture: 0.01
+  mp3_compression_on_mixture_bitrate_min: 32
+  mp3_compression_on_mixture_bitrate_max: 320
+  mp3_compression_on_mixture_backend: "lameenc"
+
+  all:
+    channel_shuffle: 0.5 # Set 0 or lower to disable
+    random_inverse: 0.1 # inverse track (better lower probability)
+    random_polarity: 0.5 # polarity change (multiply waveform to -1)
+    mp3_compression: 0.01
+    mp3_compression_min_bitrate: 32
+    mp3_compression_max_bitrate: 320
+    mp3_compression_backend: "lameenc"
+
+    # pedalboard reverb block
+    pedalboard_reverb: 0.01
+    pedalboard_reverb_room_size_min: 0.1
+    pedalboard_reverb_room_size_max: 0.9
+    pedalboard_reverb_damping_min: 0.1
+    pedalboard_reverb_damping_max: 0.9
+    pedalboard_reverb_wet_level_min: 0.1
+    pedalboard_reverb_wet_level_max: 0.9
+    pedalboard_reverb_dry_level_min: 0.1
+    pedalboard_reverb_dry_level_max: 0.9
+    pedalboard_reverb_width_min: 0.9
+    pedalboard_reverb_width_max: 1.0
+
+    # pedalboard chorus block
+    pedalboard_chorus: 0.01
+    pedalboard_chorus_rate_hz_min: 1.0
+    pedalboard_chorus_rate_hz_max: 7.0
+    pedalboard_chorus_depth_min: 0.25
+    pedalboard_chorus_depth_max: 0.95
+    pedalboard_chorus_centre_delay_ms_min: 3
+    pedalboard_chorus_centre_delay_ms_max: 10
+    pedalboard_chorus_feedback_min: 0.0
+    pedalboard_chorus_feedback_max: 0.5
+    pedalboard_chorus_mix_min: 0.1
+    pedalboard_chorus_mix_max: 0.9
+
+    # pedalboard phazer block
+    pedalboard_phazer: 0.01
+    pedalboard_phazer_rate_hz_min: 1.0
+    pedalboard_phazer_rate_hz_max: 10.0
+    pedalboard_phazer_depth_min: 0.25
+    pedalboard_phazer_depth_max: 0.95
+    pedalboard_phazer_centre_frequency_hz_min: 200
+    pedalboard_phazer_centre_frequency_hz_max: 12000
+    pedalboard_phazer_feedback_min: 0.0
+    pedalboard_phazer_feedback_max: 0.5
+    pedalboard_phazer_mix_min: 0.1
+    pedalboard_phazer_mix_max: 0.9
+
+    # pedalboard distortion block
+    pedalboard_distortion: 0.01
+    pedalboard_distortion_drive_db_min: 1.0
+    pedalboard_distortion_drive_db_max: 25.0
+
+    # pedalboard pitch shift block
+    pedalboard_pitch_shift: 0.01
+    pedalboard_pitch_shift_semitones_min: -7
+    pedalboard_pitch_shift_semitones_max: 7
+
+    # pedalboard resample block
+    pedalboard_resample: 0.01
+    pedalboard_resample_target_sample_rate_min: 4000
+    pedalboard_resample_target_sample_rate_max: 44100
+
+    # pedalboard bitcrash block
+    pedalboard_bitcrash: 0.01
+    pedalboard_bitcrash_bit_depth_min: 4
+    pedalboard_bitcrash_bit_depth_max: 16
+
+    # pedalboard mp3 compressor block
+    pedalboard_mp3_compressor: 0.01
+    pedalboard_mp3_compressor_pedalboard_mp3_compressor_min: 0
+    pedalboard_mp3_compressor_pedalboard_mp3_compressor_max: 9.999
+
+  vocals:
+    pitch_shift: 0.1
+    pitch_shift_min_semitones: -5
+    pitch_shift_max_semitones: 5
+    seven_band_parametric_eq: 0.25
+    seven_band_parametric_eq_min_gain_db: -9
+    seven_band_parametric_eq_max_gain_db: 9
+    tanh_distortion: 0.1
+    tanh_distortion_min: 0.1
+    tanh_distortion_max: 0.7
+  bass:
+    pitch_shift: 0.1
+    pitch_shift_min_semitones: -2
+    pitch_shift_max_semitones: 2
+    seven_band_parametric_eq: 0.25
+    seven_band_parametric_eq_min_gain_db: -3
+    seven_band_parametric_eq_max_gain_db: 6
+    tanh_distortion: 0.2
+    tanh_distortion_min: 0.1
+    tanh_distortion_max: 0.5
+  drums:
+    pitch_shift: 0.33
+    pitch_shift_min_semitones: -5
+    pitch_shift_max_semitones: 5
+    seven_band_parametric_eq: 0.25
+    seven_band_parametric_eq_min_gain_db: -9
+    seven_band_parametric_eq_max_gain_db: 9
+    tanh_distortion: 0.33
+    tanh_distortion_min: 0.1
+    tanh_distortion_max: 0.6
+  other:
+    pitch_shift: 0.1
+    pitch_shift_min_semitones: -4
+    pitch_shift_max_semitones: 4
+    gaussian_noise: 0.1
+    gaussian_noise_min_amplitude: 0.001
+    gaussian_noise_max_amplitude: 0.015
+    time_stretch: 0.01
+    time_stretch_min_rate: 0.8
+    time_stretch_max_rate: 1.25
+
+
+inference:
+  batch_size: 1
+  dim_t: 256
+  num_overlap: 4

configs/config_musdb18_mel_band_roformer.yaml

@@ -0,0 +1,73 @@
+audio:
+  chunk_size: 131584
+  dim_f: 1024
+  dim_t: 256
+  hop_length: 512
+  n_fft: 2048
+  num_channels: 2
+  sample_rate: 44100
+  min_mean_abs: 0.001
+
+model:
+  dim: 192
+  depth: 8
+  stereo: true
+  num_stems: 1
+  time_transformer_depth: 1
+  freq_transformer_depth: 1
+  linear_transformer_depth: 0
+  num_bands: 60
+  dim_head: 64
+  heads: 8
+  attn_dropout: 0.1
+  ff_dropout: 0.1
+  flash_attn: True
+  dim_freqs_in: 1025
+  sample_rate: 44100 # needed for mel filter bank from librosa
+  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
+
+training:
+  batch_size: 7
+  gradient_accumulation_steps: 1
+  grad_clip: 0
+  instruments:
+    - vocals
+    - bass
+    - drums
+    - other
+  lr: 5.0e-05
+  patience: 2
+  reduce_factor: 0.95
+  target_instrument: vocals
+  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
+
+augmentations:
+  enable: true # enable or disable all augmentations (to fast disable if needed)
+  loudness: true # randomly change loudness of each stem on the range (loudness_min; loudness_max)
+  loudness_min: 0.5
+  loudness_max: 1.5
+
+inference:
+  batch_size: 1
+  dim_t: 256
+  num_overlap: 4

configs/config_musdb18_scnet.yaml

@@ -0,0 +1,64 @@
+audio:
+  chunk_size: 264600
+  num_channels: 2
+  sample_rate: 44100
+  min_mean_abs: 0.001
+
+model:
+  dims: [4, 32, 64, 128]
+  bandsplit_ratios: [.175, .392, .433]
+  downsample_strides: [1, 4, 16]
+  n_conv_modules: [3, 2, 1]
+  n_rnn_layers: 6
+  rnn_hidden_dim: 128
+  n_sources: 4
+
+  n_fft: 4096
+  hop_length: 1024
+  win_length: 4096
+  stft_normalized: false
+
+  use_mamba: true
+  d_state: 16
+  d_conv: 4
+  d_expand: 2
+
+training:
+  batch_size: 10
+  gradient_accumulation_steps: 1
+  grad_clip: 0
+  instruments:
+    - vocals
+    - bass
+    - drums
+    - other
+  lr: 5.0e-04
+  patience: 2
+  reduce_factor: 0.95
+  target_instrument: null
+  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
+
+augmentations:
+  enable: true # enable or disable all augmentations (to fast disable if needed)
+  loudness: true # randomly change loudness of each stem on the range (loudness_min; loudness_max)
+  loudness_min: 0.5
+  loudness_max: 1.5
+  mixup: true # mix several stems of same type with some probability (only works for dataset types: 1, 2, 3)
+  mixup_probs:
+    !!python/tuple # 2 additional stems of the same type (1st with prob 0.2, 2nd with prob 0.02)
+    - 0.2
+    - 0.02
+  mixup_loudness_min: 0.5
+  mixup_loudness_max: 1.5
+
+inference:
+  batch_size: 1
+  dim_t: 256
+  num_overlap: 4

configs/config_musdb18_segm_models.yaml

@@ -0,0 +1,92 @@
+audio:
+  chunk_size: 261632
+  dim_f: 4096
+  dim_t: 512
+  hop_length: 512
+  n_fft: 8192
+  num_channels: 2
+  sample_rate: 44100
+  min_mean_abs: 0.001
+
+model:
+  encoder_name: tu-maxvit_large_tf_512 # look here for possibilities: https://github.com/qubvel/segmentation_models.pytorch#encoders-
+  decoder_type: unet # unet, fpn
+  act: gelu
+  num_channels: 128
+  num_subbands: 8
+
+training:
+  batch_size: 7
+  gradient_accumulation_steps: 1
+  grad_clip: 0
+  instruments:
+  - vocals
+  - bass
+  - drums
+  - other
+  lr: 5.0e-05
+  patience: 2
+  reduce_factor: 0.95
+  target_instrument: null
+  num_epochs: 1000
+  num_steps: 2000
+  q: 0.95
+  coarse_loss_clip: true
+  ema_momentum: 0.999
+  optimizer: adamw
+  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
+
+augmentations:
+  enable: true # enable or disable all augmentations (to fast disable if needed)
+  loudness: true # randomly change loudness of each stem on the range (loudness_min; loudness_max)
+  loudness_min: 0.5
+  loudness_max: 1.5
+  mixup: true # mix several stems of same type with some probability (only works for dataset types: 1, 2, 3)
+  mixup_probs: !!python/tuple # 2 additional stems of the same type (1st with prob 0.2, 2nd with prob 0.02)
+    - 0.2
+    - 0.02
+  mixup_loudness_min: 0.5
+  mixup_loudness_max: 1.5
+
+  # apply mp3 compression to mixture only (emulate downloading mp3 from internet)
+  mp3_compression_on_mixture: 0.01
+  mp3_compression_on_mixture_bitrate_min: 32
+  mp3_compression_on_mixture_bitrate_max: 320
+  mp3_compression_on_mixture_backend: "lameenc"
+
+  all:
+    channel_shuffle: 0.5 # Set 0 or lower to disable
+    random_inverse: 0.1 # inverse track (better lower probability)
+    random_polarity: 0.5 # polarity change (multiply waveform to -1)
+    mp3_compression: 0.01
+    mp3_compression_min_bitrate: 32
+    mp3_compression_max_bitrate: 320
+    mp3_compression_backend: "lameenc"
+
+  vocals:
+    pitch_shift: 0.1
+    pitch_shift_min_semitones: -5
+    pitch_shift_max_semitones: 5
+    seven_band_parametric_eq: 0.25
+    seven_band_parametric_eq_min_gain_db: -9
+    seven_band_parametric_eq_max_gain_db: 9
+    tanh_distortion: 0.1
+    tanh_distortion_min: 0.1
+    tanh_distortion_max: 0.7
+  other:
+    pitch_shift: 0.1
+    pitch_shift_min_semitones: -4
+    pitch_shift_max_semitones: 4
+    gaussian_noise: 0.1
+    gaussian_noise_min_amplitude: 0.001
+    gaussian_noise_max_amplitude: 0.015
+    time_stretch: 0.01
+    time_stretch_min_rate: 0.8
+    time_stretch_max_rate: 1.25
+
+
+inference:
+  batch_size: 1
+  dim_t: 512
+  num_overlap: 4

configs/config_musdb18_torchseg.yaml

@@ -0,0 +1,92 @@
+audio:
+  chunk_size: 261632
+  dim_f: 4096
+  dim_t: 512
+  hop_length: 512
+  n_fft: 8192
+  num_channels: 2
+  sample_rate: 44100
+  min_mean_abs: 0.001
+
+model:
+  encoder_name: maxvit_tiny_tf_512 # look with torchseg.list_encoders(). Currently 858 available
+  decoder_type: unet # unet, fpn
+  act: gelu
+  num_channels: 128
+  num_subbands: 8
+
+training:
+  batch_size: 18
+  gradient_accumulation_steps: 1
+  grad_clip: 0
+  instruments:
+  - vocals
+  - bass
+  - drums
+  - other
+  lr: 5.0e-05
+  patience: 2
+  reduce_factor: 0.95
+  target_instrument: null
+  num_epochs: 1000
+  num_steps: 2000
+  q: 0.95
+  coarse_loss_clip: true
+  ema_momentum: 0.999
+  optimizer: adamw
+  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
+
+augmentations:
+  enable: true # enable or disable all augmentations (to fast disable if needed)
+  loudness: true # randomly change loudness of each stem on the range (loudness_min; loudness_max)
+  loudness_min: 0.5
+  loudness_max: 1.5
+  mixup: true # mix several stems of same type with some probability (only works for dataset types: 1, 2, 3)
+  mixup_probs: !!python/tuple # 2 additional stems of the same type (1st with prob 0.2, 2nd with prob 0.02)
+    - 0.2
+    - 0.02
+  mixup_loudness_min: 0.5
+  mixup_loudness_max: 1.5
+
+  # apply mp3 compression to mixture only (emulate downloading mp3 from internet)
+  mp3_compression_on_mixture: 0.01
+  mp3_compression_on_mixture_bitrate_min: 32
+  mp3_compression_on_mixture_bitrate_max: 320
+  mp3_compression_on_mixture_backend: "lameenc"
+
+  all:
+    channel_shuffle: 0.5 # Set 0 or lower to disable
+    random_inverse: 0.1 # inverse track (better lower probability)
+    random_polarity: 0.5 # polarity change (multiply waveform to -1)
+    mp3_compression: 0.01
+    mp3_compression_min_bitrate: 32
+    mp3_compression_max_bitrate: 320
+    mp3_compression_backend: "lameenc"
+
+  vocals:
+    pitch_shift: 0.1
+    pitch_shift_min_semitones: -5
+    pitch_shift_max_semitones: 5
+    seven_band_parametric_eq: 0.25
+    seven_band_parametric_eq_min_gain_db: -9
+    seven_band_parametric_eq_max_gain_db: 9
+    tanh_distortion: 0.1
+    tanh_distortion_min: 0.1
+    tanh_distortion_max: 0.7
+  other:
+    pitch_shift: 0.1
+    pitch_shift_min_semitones: -4
+    pitch_shift_max_semitones: 4
+    gaussian_noise: 0.1
+    gaussian_noise_min_amplitude: 0.001
+    gaussian_noise_max_amplitude: 0.015
+    time_stretch: 0.01
+    time_stretch_min_rate: 0.8
+    time_stretch_max_rate: 1.25
+
+
+inference:
+  batch_size: 1
+  dim_t: 512
+  num_overlap: 4

configs/config_vocals_bandit_bsrnn_multi_mus64.yaml

@@ -0,0 +1,73 @@
+name: "MultiMaskMultiSourceBandSplitRNN"
+audio:
+  chunk_size: 264600
+  num_channels: 2
+  sample_rate: 44100
+  min_mean_abs: 0.001
+
+model:
+  in_channel: 1
+  stems: ['vocals', 'other']
+  band_specs: "musical"
+  n_bands: 64
+  fs: 44100
+  require_no_overlap: false
+  require_no_gap: true
+  normalize_channel_independently: false
+  treat_channel_as_feature: true
+  n_sqm_modules: 8
+  emb_dim: 128
+  rnn_dim: 256
+  bidirectional: true
+  rnn_type: "GRU"
+  mlp_dim: 512
+  hidden_activation: "Tanh"
+  hidden_activation_kwargs: null
+  complex_mask: true
+  n_fft: 2048
+  win_length: 2048
+  hop_length: 512
+  window_fn: "hann_window"
+  wkwargs: null
+  power: null
+  center: true
+  normalized: true
+  pad_mode: "constant"
+  onesided: true
+
+training:
+  batch_size: 4
+  gradient_accumulation_steps: 4
+  grad_clip: 0
+  instruments:
+  - vocals
+  - other
+  lr: 9.0e-05
+  patience: 2
+  reduce_factor: 0.95
+  target_instrument: null
+  num_epochs: 1000
+  num_steps: 1000
+  q: 0.95
+  coarse_loss_clip: true
+  ema_momentum: 0.999
+  optimizer: adam
+  other_fix: true # 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
+
+augmentations:
+  enable: true # enable or disable all augmentations (to fast disable if needed)
+  loudness: true # randomly change loudness of each stem on the range (loudness_min; loudness_max)
+  loudness_min: 0.5
+  loudness_max: 1.5
+  mixup: true # mix several stems of same type with some probability (only works for dataset types: 1, 2, 3)
+  mixup_probs: !!python/tuple # 2 additional stems of the same type (1st with prob 0.2, 2nd with prob 0.02)
+    - 0.2
+    - 0.02
+  mixup_loudness_min: 0.5
+  mixup_loudness_max: 1.5
+
+inference:
+  batch_size: 1
+  dim_t: 256
+  num_overlap: 4

configs/config_vocals_bs_roformer.yaml

@@ -0,0 +1,138 @@
+audio:
+  chunk_size: 131584
+  dim_f: 1024
+  dim_t: 256
+  hop_length: 512
+  n_fft: 2048
+  num_channels: 2
+  sample_rate: 44100
+  min_mean_abs: 0.001
+
+model:
+  dim: 192
+  depth: 6
+  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: 0.1
+  ff_dropout: 0.1
+  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
+
+training:
+  batch_size: 10
+  gradient_accumulation_steps: 1
+  grad_clip: 0
+  instruments:
+  - vocals
+  - other
+  lr: 5.0e-05
+  patience: 2
+  reduce_factor: 0.95
+  target_instrument: vocals
+  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
+
+augmentations:
+  enable: true # enable or disable all augmentations (to fast disable if needed)
+  loudness: true # randomly change loudness of each stem on the range (loudness_min; loudness_max)
+  loudness_min: 0.5
+  loudness_max: 1.5
+  mixup: true # mix several stems of same type with some probability (only works for dataset types: 1, 2, 3)
+  mixup_probs: !!python/tuple # 2 additional stems of the same type (1st with prob 0.2, 2nd with prob 0.02)
+    - 0.2
+    - 0.02
+  mixup_loudness_min: 0.5
+  mixup_loudness_max: 1.5
+
+inference:
+  batch_size: 1
+  dim_t: 256
+  num_overlap: 4

configs/config_vocals_htdemucs.yaml

@@ -0,0 +1,123 @@
+audio:
+  chunk_size: 485100 # samplerate * segment
+  min_mean_abs: 0.001
+  hop_length: 1024
+
+training:
+  batch_size: 10
+  gradient_accumulation_steps: 1
+  grad_clip: 0
+  segment: 11
+  shift: 1
+  samplerate: 44100
+  channels: 2
+  normalize: true
+  instruments: ['vocals', 'other']
+  target_instrument: null
+  num_epochs: 1000
+  num_steps: 1000
+  optimizer: adam
+  lr: 9.0e-05
+  patience: 2
+  reduce_factor: 0.95
+  q: 0.95
+  coarse_loss_clip: true
+  ema_momentum: 0.999
+  other_fix: true # 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
+
+augmentations:
+  enable: true # enable or disable all augmentations (to fast disable if needed)
+  loudness: true # randomly change loudness of each stem on the range (loudness_min; loudness_max)
+  loudness_min: 0.5
+  loudness_max: 1.5
+  mixup: true # mix several stems of same type with some probability (only works for dataset types: 1, 2, 3)
+  mixup_probs: [0.2, 0.02]
+  mixup_loudness_min: 0.5
+  mixup_loudness_max: 1.5
+
+inference:
+  num_overlap: 2
+  batch_size: 8
+
+model: htdemucs
+
+htdemucs:  # see demucs/htdemucs.py for a detailed description
+  # Channels
+  channels: 48
+  channels_time:
+  growth: 2
+  # STFT
+  num_subbands: 1
+  nfft: 4096
+  wiener_iters: 0
+  end_iters: 0
+  wiener_residual: false
+  cac: true
+  # Main structure
+  depth: 4
+  rewrite: true
+  # Frequency Branch
+  multi_freqs: []
+  multi_freqs_depth: 3
+  freq_emb: 0.2
+  emb_scale: 10
+  emb_smooth: true
+  # Convolutions
+  kernel_size: 8
+  stride: 4
+  time_stride: 2
+  context: 1
+  context_enc: 0
+  # normalization
+  norm_starts: 4
+  norm_groups: 4
+  # DConv residual branch
+  dconv_mode: 3
+  dconv_depth: 2
+  dconv_comp: 8
+  dconv_init: 1e-3
+  # Before the Transformer
+  bottom_channels: 512
+  # CrossTransformer
+  # ------ Common to all
+  # Regular parameters
+  t_layers: 5
+  t_hidden_scale: 4.0
+  t_heads: 8
+  t_dropout: 0.0
+  t_layer_scale: True
+  t_gelu: True
+  # ------------- Positional Embedding
+  t_emb: sin
+  t_max_positions: 10000 # for the scaled embedding
+  t_max_period: 10000.0
+  t_weight_pos_embed: 1.0
+  t_cape_mean_normalize: True
+  t_cape_augment: True
+  t_cape_glob_loc_scale: [5000.0, 1.0, 1.4]
+  t_sin_random_shift: 0
+  # ------------- norm before a transformer encoder
+  t_norm_in: True
+  t_norm_in_group: False
+  # ------------- norm inside the encoder
+  t_group_norm: False
+  t_norm_first: True
+  t_norm_out: True
+  # ------------- optim
+  t_weight_decay: 0.0
+  t_lr:
+  # ------------- sparsity
+  t_sparse_self_attn: False
+  t_sparse_cross_attn: False
+  t_mask_type: diag
+  t_mask_random_seed: 42
+  t_sparse_attn_window: 400
+  t_global_window: 100
+  t_sparsity: 0.95
+  t_auto_sparsity: False
+  # Cross Encoder First (False)
+  t_cross_first: False
+  # Weight init
+  rescale: 0.1
+

configs/config_vocals_mdx23c.yaml

@@ -0,0 +1,95 @@
+audio:
+  chunk_size: 261120
+  dim_f: 4096
+  dim_t: 256
+  hop_length: 1024
+  n_fft: 8192
+  num_channels: 2
+  sample_rate: 44100
+  min_mean_abs: 0.001
+
+model:
+  act: gelu
+  bottleneck_factor: 4
+  growth: 128
+  norm: InstanceNorm
+  num_blocks_per_scale: 2
+  num_channels: 128
+  num_scales: 5
+  num_subbands: 4
+  scale:
+  - 2
+  - 2
+
+training:
+  batch_size: 6
+  gradient_accumulation_steps: 1
+  grad_clip: 0
+  instruments:
+  - vocals
+  - other
+  lr: 9.0e-05
+  patience: 2
+  reduce_factor: 0.95
+  target_instrument: null
+  num_epochs: 1000
+  num_steps: 1000
+  q: 0.95
+  coarse_loss_clip: true
+  ema_momentum: 0.999
+  optimizer: adam
+  other_fix: true # 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
+
+augmentations:
+  enable: true # enable or disable all augmentations (to fast disable if needed)
+  loudness: true # randomly change loudness of each stem on the range (loudness_min; loudness_max)
+  loudness_min: 0.5
+  loudness_max: 1.5
+  mixup: true # mix several stems of same type with some probability (only works for dataset types: 1, 2, 3)
+  mixup_probs: !!python/tuple # 2 additional stems of the same type (1st with prob 0.2, 2nd with prob 0.02)
+    - 0.2
+    - 0.02
+  mixup_loudness_min: 0.5
+  mixup_loudness_max: 1.5
+
+  # apply mp3 compression to mixture only (emulate downloading mp3 from internet)
+  mp3_compression_on_mixture: 0.01
+  mp3_compression_on_mixture_bitrate_min: 32
+  mp3_compression_on_mixture_bitrate_max: 320
+  mp3_compression_on_mixture_backend: "lameenc"
+
+  all:
+    channel_shuffle: 0.5 # Set 0 or lower to disable
+    random_inverse: 0.1 # inverse track (better lower probability)
+    random_polarity: 0.5 # polarity change (multiply waveform to -1)
+    mp3_compression: 0.01
+    mp3_compression_min_bitrate: 32
+    mp3_compression_max_bitrate: 320
+    mp3_compression_backend: "lameenc"
+
+  vocals:
+    pitch_shift: 0.1
+    pitch_shift_min_semitones: -5
+    pitch_shift_max_semitones: 5
+    seven_band_parametric_eq: 0.25
+    seven_band_parametric_eq_min_gain_db: -9
+    seven_band_parametric_eq_max_gain_db: 9
+    tanh_distortion: 0.1
+    tanh_distortion_min: 0.1
+    tanh_distortion_max: 0.7
+  other:
+    pitch_shift: 0.1
+    pitch_shift_min_semitones: -4
+    pitch_shift_max_semitones: 4
+    gaussian_noise: 0.1
+    gaussian_noise_min_amplitude: 0.001
+    gaussian_noise_max_amplitude: 0.015
+    time_stretch: 0.01
+    time_stretch_min_rate: 0.8
+    time_stretch_max_rate: 1.25
+
+inference:
+  batch_size: 1
+  dim_t: 256
+  num_overlap: 4

configs/config_vocals_mel_band_roformer.yaml

@@ -0,0 +1,77 @@
+audio:
+  chunk_size: 131584
+  dim_f: 1024
+  dim_t: 256
+  hop_length: 512
+  n_fft: 2048
+  num_channels: 2
+  sample_rate: 44100
+  min_mean_abs: 0.001
+
+model:
+  dim: 192
+  depth: 8
+  stereo: true
+  num_stems: 1
+  time_transformer_depth: 1
+  freq_transformer_depth: 1
+  linear_transformer_depth: 0
+  num_bands: 60
+  dim_head: 64
+  heads: 8
+  attn_dropout: 0.1
+  ff_dropout: 0.1
+  flash_attn: True
+  dim_freqs_in: 1025
+  sample_rate: 44100  # needed for mel filter bank from librosa
+  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
+
+training:
+  batch_size: 7
+  gradient_accumulation_steps: 1
+  grad_clip: 0
+  instruments:
+  - vocals
+  - other
+  lr: 5.0e-05
+  patience: 2
+  reduce_factor: 0.95
+  target_instrument: vocals
+  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
+
+augmentations:
+  enable: true # enable or disable all augmentations (to fast disable if needed)
+  loudness: true # randomly change loudness of each stem on the range (loudness_min; loudness_max)
+  loudness_min: 0.5
+  loudness_max: 1.5
+  mixup: true # mix several stems of same type with some probability (only works for dataset types: 1, 2, 3)
+  mixup_probs: !!python/tuple # 2 additional stems of the same type (1st with prob 0.2, 2nd with prob 0.02)
+    - 0.2
+    - 0.02
+  mixup_loudness_min: 0.5
+  mixup_loudness_max: 1.5
+
+inference:
+  batch_size: 1
+  dim_t: 256
+  num_overlap: 4

configs/config_vocals_scnet.yaml

@@ -0,0 +1,71 @@
+audio:
+  chunk_size: 264600
+  num_channels: 2
+  sample_rate: 44100
+  min_mean_abs: 0.000
+
+model:
+  sources: ['vocals', 'other']
+  audio_channels: 2
+  # dims: [4, 32, 64, 128] # small version
+  dims: [4, 64, 128, 256]
+  nfft: 4096
+  hop_size: 1024
+  win_size: 4096
+  normalized: True
+  band_configs: {
+    'low': { 'SR': .175, 'stride': 1, 'kernel': 3 },
+    'mid': { 'SR': .392, 'stride': 4, 'kernel': 4 },
+    'high': { 'SR': .433, 'stride': 16, 'kernel': 16 }
+  }
+  conv_depths: [3, 2, 1]
+  compress: 4
+  conv_kernel: 3
+  # Dual-path RNN
+  num_dplayer: 6
+  expand: 1
+  # mamba
+  use_mamba: False
+  mamba_config: {
+    'd_stat': 16,
+    'd_conv': 4,
+    'd_expand': 2
+  }
+
+training:
+  batch_size: 4
+  gradient_accumulation_steps: 2
+  grad_clip: 0
+  instruments:
+    - vocals
+    - other
+  lr: 5.0e-04
+  patience: 2
+  reduce_factor: 0.95
+  target_instrument: null
+  num_epochs: 1000
+  num_steps: 1000
+  q: 0.95
+  coarse_loss_clip: true
+  ema_momentum: 0.999
+  optimizer: adam
+  other_fix: true # 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
+
+augmentations:
+  enable: true # enable or disable all augmentations (to fast disable if needed)
+  loudness: true # randomly change loudness of each stem on the range (loudness_min; loudness_max)
+  loudness_min: 0.5
+  loudness_max: 1.5
+  mixup: true # mix several stems of same type with some probability (only works for dataset types: 1, 2, 3)
+  mixup_probs:
+    !!python/tuple # 2 additional stems of the same type (1st with prob 0.2, 2nd with prob 0.02)
+    - 0.2
+    - 0.02
+  mixup_loudness_min: 0.5
+  mixup_loudness_max: 1.5
+
+inference:
+  batch_size: 8
+  dim_t: 256
+  num_overlap: 4

configs/config_vocals_scnet_unofficial.yaml

@@ -0,0 +1,62 @@
+audio:
+  chunk_size: 264600
+  num_channels: 2
+  sample_rate: 44100
+  min_mean_abs: 0.000
+
+model:
+  dims: [4, 32, 64, 128]
+  bandsplit_ratios: [.175, .392, .433]
+  downsample_strides: [1, 4, 16]
+  n_conv_modules: [3, 2, 1]
+  n_rnn_layers: 6
+  rnn_hidden_dim: 128
+  n_sources: 2
+
+  n_fft: 4096
+  hop_length: 1024
+  win_length: 4096
+  stft_normalized: false
+
+  use_mamba: false
+  d_state: 16
+  d_conv: 4
+  d_expand: 2
+
+training:
+  batch_size: 10
+  gradient_accumulation_steps: 2
+  grad_clip: 0
+  instruments:
+    - vocals
+    - other
+  lr: 5.0e-04
+  patience: 2
+  reduce_factor: 0.95
+  target_instrument: null
+  num_epochs: 1000
+  num_steps: 1000
+  q: 0.95
+  coarse_loss_clip: true
+  ema_momentum: 0.999
+  optimizer: adam
+  other_fix: true # 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
+
+augmentations:
+  enable: true # enable or disable all augmentations (to fast disable if needed)
+  loudness: true # randomly change loudness of each stem on the range (loudness_min; loudness_max)
+  loudness_min: 0.5
+  loudness_max: 1.5
+  mixup: true # mix several stems of same type with some probability (only works for dataset types: 1, 2, 3)
+  mixup_probs:
+    !!python/tuple # 2 additional stems of the same type (1st with prob 0.2, 2nd with prob 0.02)
+    - 0.2
+    - 0.02
+  mixup_loudness_min: 0.5
+  mixup_loudness_max: 1.5
+
+inference:
+  batch_size: 8
+  dim_t: 256
+  num_overlap: 4

configs/config_vocals_segm_models.yaml

@@ -0,0 +1,78 @@
+audio:
+  chunk_size: 261632
+  dim_f: 4096
+  dim_t: 512
+  hop_length: 512
+  n_fft: 8192
+  num_channels: 2
+  sample_rate: 44100
+  min_mean_abs: 0.001
+
+model:
+  encoder_name: tu-maxvit_large_tf_512 # look here for possibilities: https://github.com/qubvel/segmentation_models.pytorch#encoders-
+  decoder_type: unet # unet, fpn
+  act: gelu
+  num_channels: 128
+  num_subbands: 8
+
+loss_multistft:
+  fft_sizes:
+  - 1024
+  - 2048
+  - 4096
+  hop_sizes:
+  - 512
+  - 1024
+  - 2048
+  win_lengths:
+  - 1024
+  - 2048
+  - 4096
+  window: "hann_window"
+  scale: "mel"
+  n_bins: 128
+  sample_rate: 44100
+  perceptual_weighting: true
+  w_sc: 1.0
+  w_log_mag: 1.0
+  w_lin_mag: 0.0
+  w_phs: 0.0
+  mag_distance: "L1"
+
+
+training:
+  batch_size: 8
+  gradient_accumulation_steps: 1
+  grad_clip: 0
+  instruments:
+  - vocals
+  - other
+  lr: 5.0e-05
+  patience: 2
+  reduce_factor: 0.95
+  target_instrument: null
+  num_epochs: 1000
+  num_steps: 2000
+  q: 0.95
+  coarse_loss_clip: true
+  ema_momentum: 0.999
+  optimizer: adamw
+  other_fix: true # 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
+
+augmentations:
+  enable: true # enable or disable all augmentations (to fast disable if needed)
+  loudness: true # randomly change loudness of each stem on the range (loudness_min; loudness_max)
+  loudness_min: 0.5
+  loudness_max: 1.5
+  mixup: true # mix several stems of same type with some probability (only works for dataset types: 1, 2, 3)
+  mixup_probs: !!python/tuple # 2 additional stems of the same type (1st with prob 0.2, 2nd with prob 0.02)
+    - 0.2
+    - 0.02
+  mixup_loudness_min: 0.5
+  mixup_loudness_max: 1.5
+
+inference:
+  batch_size: 1
+  dim_t: 512
+  num_overlap: 4

configs/config_vocals_swin_upernet.yaml

@@ -0,0 +1,50 @@
+audio:
+  chunk_size: 261632
+  dim_f: 4096
+  dim_t: 512
+  hop_length: 512
+  n_fft: 8192
+  num_channels: 2
+  sample_rate: 44100
+  min_mean_abs: 0.001
+
+model:
+  act: gelu
+  num_channels: 16
+  num_subbands: 8
+
+training:
+  batch_size: 14
+  gradient_accumulation_steps: 4
+  grad_clip: 0
+  instruments:
+  - vocals
+  - other
+  lr: 3.0e-05
+  patience: 2
+  reduce_factor: 0.95
+  target_instrument: null
+  num_epochs: 1000
+  num_steps: 1000
+  q: 0.95
+  coarse_loss_clip: true
+  ema_momentum: 0.999
+  optimizer: adamw
+  other_fix: true # it's needed for checking on multisong dataset if other is actually instrumental
+
+augmentations:
+  enable: true # enable or disable all augmentations (to fast disable if needed)
+  loudness: true # randomly change loudness of each stem on the range (loudness_min; loudness_max)
+  loudness_min: 0.5
+  loudness_max: 1.5
+  mixup: true # mix several stems of same type with some probability (only works for dataset types: 1, 2, 3)
+  mixup_probs: !!python/tuple # 2 additional stems of the same type (1st with prob 0.2, 2nd with prob 0.02)
+    - 0.2
+    - 0.02
+  mixup_loudness_min: 0.5
+  mixup_loudness_max: 1.5
+
+inference:
+  batch_size: 1
+  dim_t: 512
+  num_overlap: 4

configs/config_vocals_torchseg.yaml

@@ -0,0 +1,58 @@
+audio:
+  chunk_size: 261632
+  dim_f: 4096
+  dim_t: 512
+  hop_length: 512
+  n_fft: 8192
+  num_channels: 2
+  sample_rate: 44100
+  min_mean_abs: 0.000
+
+model:
+  encoder_name: maxvit_tiny_tf_512 # look with torchseg.list_encoders(). Currently 858 available
+  decoder_type: unet # unet, fpn
+  act: gelu
+  num_channels: 128
+  num_subbands: 8
+
+training:
+  batch_size: 18
+  gradient_accumulation_steps: 1
+  grad_clip: 1.0
+  instruments:
+  - vocals
+  - other
+  lr: 1.0e-04
+  patience: 2
+  reduce_factor: 0.95
+  target_instrument: null
+  num_epochs: 1000
+  num_steps: 1000
+  q: 0.95
+  coarse_loss_clip: true
+  ema_momentum: 0.999
+  optimizer: adam
+  other_fix: true # 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
+
+augmentations:
+  enable: false # enable or disable all augmentations (to fast disable if needed)
+  loudness: true # randomly change loudness of each stem on the range (loudness_min; loudness_max)
+  loudness_min: 0.5
+  loudness_max: 1.5
+  mixup: true # mix several stems of same type with some probability (only works for dataset types: 1, 2, 3)
+  mixup_probs: !!python/tuple # 2 additional stems of the same type (1st with prob 0.2, 2nd with prob 0.02)
+    - 0.2
+    - 0.02
+  mixup_loudness_min: 0.5
+  mixup_loudness_max: 1.5
+
+  all:
+    channel_shuffle: 0.5 # Set 0 or lower to disable
+    random_inverse: 0.1 # inverse track (better lower probability)
+    random_polarity: 0.5 # polarity change (multiply waveform to -1)
+
+inference:
+  batch_size: 8
+  dim_t: 512
+  num_overlap: 2

configs/viperx/model_bs_roformer_ep_317_sdr_12.9755.yaml

@@ -0,0 +1,126 @@
+audio:
+  chunk_size: 352800
+  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: 0.000
+
+model:
+  dim: 512
+  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: 0.1
+  ff_dropout: 0.1
+  flash_attn: true
+  dim_freqs_in: 1025
+  stft_n_fft: 2048
+  stft_hop_length: 441
+  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
+
+training:
+  batch_size: 2
+  gradient_accumulation_steps: 1
+  grad_clip: 0
+  instruments:
+  - vocals
+  - other
+  lr: 1.0e-05
+  patience: 2
+  reduce_factor: 0.95
+  target_instrument: vocals
+  num_epochs: 1000
+  num_steps: 1000
+  q: 0.95
+  coarse_loss_clip: true
+  ema_momentum: 0.999
+  optimizer: adam
+  other_fix: true # 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: 4
+  dim_t: 801
+  num_overlap: 2

configs/viperx/model_bs_roformer_ep_937_sdr_10.5309.yaml

@@ -0,0 +1,138 @@
+audio:
+  chunk_size: 131584
+  dim_f: 1024
+  dim_t: 256
+  hop_length: 512
+  n_fft: 2048
+  num_channels: 2
+  sample_rate: 44100
+  min_mean_abs: 0.001
+
+model:
+  dim: 384
+  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: 0.1
+  ff_dropout: 0.1
+  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
+
+training:
+  batch_size: 4
+  gradient_accumulation_steps: 1
+  grad_clip: 0
+  instruments:
+  - vocals
+  - other
+  lr: 5.0e-05
+  patience: 2
+  reduce_factor: 0.95
+  target_instrument: other
+  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
+
+augmentations:
+  enable: true # enable or disable all augmentations (to fast disable if needed)
+  loudness: true # randomly change loudness of each stem on the range (loudness_min; loudness_max)
+  loudness_min: 0.5
+  loudness_max: 1.5
+  mixup: true # mix several stems of same type with some probability (only works for dataset types: 1, 2, 3)
+  mixup_probs: !!python/tuple # 2 additional stems of the same type (1st with prob 0.2, 2nd with prob 0.02)
+    - 0.2
+    - 0.02
+  mixup_loudness_min: 0.5
+  mixup_loudness_max: 1.5
+
+inference:
+  batch_size: 8
+  dim_t: 512
+  num_overlap: 2

configs/viperx/model_mel_band_roformer_ep_3005_sdr_11.4360.yaml

@@ -0,0 +1,65 @@
+audio:
+  chunk_size: 352800
+  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: 0.000
+
+model:
+  dim: 384
+  depth: 12
+  stereo: true
+  num_stems: 1
+  time_transformer_depth: 1
+  freq_transformer_depth: 1
+  linear_transformer_depth: 0
+  num_bands: 60
+  dim_head: 64
+  heads: 8
+  attn_dropout: 0.1
+  ff_dropout: 0.1
+  flash_attn: True
+  dim_freqs_in: 1025
+  sample_rate: 44100  # needed for mel filter bank from librosa
+  stft_n_fft: 2048
+  stft_hop_length: 441
+  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
+
+training:
+  batch_size: 1
+  gradient_accumulation_steps: 8
+  grad_clip: 0
+  instruments:
+  - vocals
+  - other
+  lr: 4.0e-05
+  patience: 2
+  reduce_factor: 0.95
+  target_instrument: vocals
+  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: 4
+  dim_t: 801
+  num_overlap: 2