RoFormer (models)

models/audio-separator-models/.gitattributes

@@ -0,0 +1,35 @@
+*.7z filter=lfs diff=lfs merge=lfs -text
+*.arrow filter=lfs diff=lfs merge=lfs -text
+*.bin filter=lfs diff=lfs merge=lfs -text
+*.bz2 filter=lfs diff=lfs merge=lfs -text
+*.ckpt filter=lfs diff=lfs merge=lfs -text
+*.ftz filter=lfs diff=lfs merge=lfs -text
+*.gz filter=lfs diff=lfs merge=lfs -text
+*.h5 filter=lfs diff=lfs merge=lfs -text
+*.joblib filter=lfs diff=lfs merge=lfs -text
+*.lfs.* filter=lfs diff=lfs merge=lfs -text
+*.mlmodel filter=lfs diff=lfs merge=lfs -text
+*.model filter=lfs diff=lfs merge=lfs -text
+*.msgpack filter=lfs diff=lfs merge=lfs -text
+*.npy filter=lfs diff=lfs merge=lfs -text
+*.npz filter=lfs diff=lfs merge=lfs -text
+*.onnx filter=lfs diff=lfs merge=lfs -text
+*.ot filter=lfs diff=lfs merge=lfs -text
+*.parquet filter=lfs diff=lfs merge=lfs -text
+*.pb filter=lfs diff=lfs merge=lfs -text
+*.pickle filter=lfs diff=lfs merge=lfs -text
+*.pkl filter=lfs diff=lfs merge=lfs -text
+*.pt filter=lfs diff=lfs merge=lfs -text
+*.pth filter=lfs diff=lfs merge=lfs -text
+*.rar filter=lfs diff=lfs merge=lfs -text
+*.safetensors filter=lfs diff=lfs merge=lfs -text
+saved_model/**/* filter=lfs diff=lfs merge=lfs -text
+*.tar.* filter=lfs diff=lfs merge=lfs -text
+*.tar filter=lfs diff=lfs merge=lfs -text
+*.tflite filter=lfs diff=lfs merge=lfs -text
+*.tgz filter=lfs diff=lfs merge=lfs -text
+*.wasm filter=lfs diff=lfs merge=lfs -text
+*.xz filter=lfs diff=lfs merge=lfs -text
+*.zip filter=lfs diff=lfs merge=lfs -text
+*.zst filter=lfs diff=lfs merge=lfs -text
+*tfevents* filter=lfs diff=lfs merge=lfs -text

models/audio-separator-models/README.md

@@ -0,0 +1,3 @@
+---
+license: mit
+---

models/audio-separator-models/roformers/MelBandRoformerBigSYHFTV1.ckpt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:e2327e3e81f19e67c307f8c830c54267c09ecb0e9c6ad2b40a80c310899c955f
+size 1479738496

models/audio-separator-models/roformers/MelBandRoformerSYHFT.ckpt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:f319dfcde4396ea3106658f457f5eb0bc577e113491f61ae8bab216fe84b0c0c
+size 913096702

models/audio-separator-models/roformers/MelBandRoformerSYHFTV2.5.ckpt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:916e3a2c1e63b1457bcad823b98ca705e4933deffd2a5ab3a370e10f68bf47e2
+size 913090472

models/audio-separator-models/roformers/MelBandRoformerSYHFTV2.ckpt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:2e99f8efa5315300c197295592bd7e56c21c1d77e1884c904b5128c54a2a4632
+size 913095346

models/audio-separator-models/roformers/MelBandRoformerSYHFTV3Epsilon.ckpt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:4c886092e4aae13aa089263a0d54d483643f58c16ec221aed37268e2c1031397
+size 913090472

models/audio-separator-models/roformers/aspiration_mel_band_roformer_less_aggr_sdr_18.1201.ckpt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:83bfe991cec4fbadde9f30d1f79cd5293ad0b1f936256be327bba5cbb4883374
+size 835982664

models/audio-separator-models/roformers/aspiration_mel_band_roformer_sdr_18.9845.ckpt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:9e791258c866c6c8da66052693d8cc3b64f1f42c01e052dbdc570cd278380cc5
+size 835983746

models/audio-separator-models/roformers/bs_roformer_instrumental_resurrection_unwa.ckpt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:16311025a5133ae6411760ccfe9e3e66b31a01d9d8bec0a03fa7ec4bedac7a15
+size 204483033

models/audio-separator-models/roformers/bs_roformer_male_female_by_aufr33_sdr_7.2889.ckpt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:3cf11736d1b42a11ae55d8299316585921477dd2a671b24b663660846ca9861b
+size 527119779

models/audio-separator-models/roformers/bs_roformer_vocals_gabox.ckpt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:18d58efe5e949e70fab11b875329af6d06ef11ccc29574bfe943fb57cc827f38
+size 639254584

models/audio-separator-models/roformers/bs_roformer_vocals_resurrection_unwa.ckpt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:9dbfe5cb572e4ed32a15ec727d7bd06c8d7aba97509e6fda5bc008bb1e0b2dd5
+size 204510749

models/audio-separator-models/roformers/bs_roformer_vocals_revive_unwa.ckpt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:f1d7e4bfdfef07c6b2bc1d65283a7d03c3c38f8c7dbc8d729b785f93c8b8699a
+size 639326600

models/audio-separator-models/roformers/bs_roformer_vocals_revive_v2_unwa.ckpt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:58098850c882a7472dad39f99fb8040ce6eaafe671cfe9881d89aea276bbb5f5
+size 639326600

models/audio-separator-models/roformers/bs_roformer_vocals_revive_v3e_unwa.ckpt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:1b0751b9a15c591407c3b77f08eb4ad3005e42e96051f3f2b39760f1130c467b
+size 639326600

models/audio-separator-models/roformers/config_aspiration_mel_band_roformer.yaml

@@ -0,0 +1,76 @@
+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: 256
+  depth: 8
+  stereo: true
+  num_stems: 2
+  time_transformer_depth: 1
+  freq_transformer_depth: 1
+  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:
+  - aspiration
+  - other
+  lr: 4.0e-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: false # 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: 4
+  dim_t: 801
+  num_overlap: 2

models/audio-separator-models/roformers/config_bs_roformer_instrumental_resurrection_unwa.yaml

@@ -0,0 +1,135 @@
+audio:
+  chunk_size: 749259
+  dim_f: 1024
+  dim_t: 1700 # 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: 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.
+  ff_dropout: 0.
+  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']
+  patience: 3
+  reduce_factor: 0.95
+  target_instrument: other
+  num_epochs: 1000
+  num_steps: 1000
+  augmentation: false # enable augmentations by audiomentations and pedalboard
+  augmentation_type: simple1
+  use_mp3_compress: false # Deprecated
+  augmentation_mix: true # Mix several stems of the same type with some probability
+  augmentation_loudness: true # randomly change loudness of each stem
+  augmentation_loudness_type: 1 # Type 1 or 2
+  augmentation_loudness_min: 0.5
+  augmentation_loudness_max: 1.5
+  q: 0.95
+  coarse_loss_clip: true
+  ema_momentum: 0.999
+  # optimizer: prodigy
+  optimizer: adam
+  # lr: 1.0
+  lr: 1.0e-5
+  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: 1700
+  num_overlap: 2
+  normalize: false

models/audio-separator-models/roformers/config_bs_roformer_vocals_gabox.yaml

@@ -0,0 +1,133 @@
+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: 0.001
+
+model:
+  dim: 512
+  depth: 12
+  stereo: true
+  num_stems: 1
+  time_transformer_depth: 1
+  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
+  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: 16
+  gradient_accumulation_steps: 1
+  grad_clip: 0
+  instruments:
+  - Vocals
+  - Instrumental
+  lr: 5.0e-05
+  patience: 2
+  reduce_factor: 0.95
+  target_instrument: Vocals
+  num_epochs: 1000
+  num_steps: 1000
+  augmentation: false # enable augmentations by audiomentations and pedalboard
+  augmentation_type: simple1
+  use_mp3_compress: false # Deprecated
+  augmentation_mix: true # Mix several stems of the same type with some probability
+  augmentation_loudness: true # randomly change loudness of each stem
+  augmentation_loudness_type: 1 # Type 1 or 2
+  augmentation_loudness_min: 0.5
+  augmentation_loudness_max: 1.5
+  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: 1
+  dim_t: 801
+  num_overlap: 4

models/audio-separator-models/roformers/config_bs_roformer_vocals_resurrection_unwa.yaml

@@ -0,0 +1,135 @@
+audio:
+  chunk_size: 785920
+  dim_f: 1024
+  dim_t: 1536 # 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: 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.
+  ff_dropout: 0.
+  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']
+  patience: 3
+  reduce_factor: 0.95
+  target_instrument: vocals
+  num_epochs: 1000
+  num_steps: 1000
+  augmentation: false # enable augmentations by audiomentations and pedalboard
+  augmentation_type: simple1
+  use_mp3_compress: false # Deprecated
+  augmentation_mix: true # Mix several stems of the same type with some probability
+  augmentation_loudness: true # randomly change loudness of each stem
+  augmentation_loudness_type: 1 # Type 1 or 2
+  augmentation_loudness_min: 0.5
+  augmentation_loudness_max: 1.5
+  q: 0.95
+  coarse_loss_clip: true
+  ema_momentum: 0.999
+  # optimizer: prodigy
+  optimizer: adam
+  # lr: 1.0
+  lr: 1.0e-5
+  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: 1536
+  num_overlap: 2
+  normalize: false

models/audio-separator-models/roformers/config_bs_roformer_vocals_revive_unwa.yaml

@@ -0,0 +1,134 @@
+audio:
+  chunk_size: 485100 #352800 #485100
+  dim_f: 1024
+  dim_t: 1101
+  hop_length: 441
+  n_fft: 2048
+  num_channels: 2
+  sample_rate: 44100
+  min_mean_abs: 0.
+
+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.
+  ff_dropout: 0.
+  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: 1
+  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
+  augmentation: false # enable augmentations by audiomentations and pedalboard
+  augmentation_type: null
+  use_mp3_compress: false # Deprecated
+  augmentation_mix: false # Mix several stems of the same type with some probability
+  augmentation_loudness: false # randomly change loudness of each stem
+  augmentation_loudness_type: 1 # Type 1 or 2
+  augmentation_loudness_min: 0
+  augmentation_loudness_max: 0
+  q: 0.95
+  coarse_loss_clip: false
+  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: 2
+  dim_t: 1101
+  num_overlap: 2

models/audio-separator-models/roformers/config_chorus_male_female_bs_roformer.yaml

@@ -0,0 +1,125 @@
+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: 8
+  stereo: true
+  num_stems: 2
+  time_transformer_depth: 1
+  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
+  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: 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: 1
+  grad_clip: 0
+  instruments:
+  - male
+  - female
+  lr: 1.0e-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
+
+inference:
+  batch_size: 1
+  dim_t: 801
+  num_overlap: 2

models/audio-separator-models/roformers/config_dereverb-echo_mel_band_roformer.yaml

@@ -0,0 +1,76 @@
+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: 256
+  depth: 8
+  stereo: true
+  num_stems: 2
+  time_transformer_depth: 1
+  freq_transformer_depth: 1
+  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:
+  - dry
+  - No dry
+  lr: 4.0e-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: false # 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: 4
+  dim_t: 801
+  num_overlap: 4

models/audio-separator-models/roformers/config_dereverb-echo_mel_band_roformer_sdr_13.4843_v2.yaml

@@ -0,0 +1,64 @@
+audio:
+  chunk_size: 352800
+  dim_f: 1024
+  dim_t: 801
+  hop_length: 441
+  n_fft: 2048
+  num_channels: 2
+  sample_rate: 44100
+  min_mean_abs: 0.000
+
+model:
+  dim: 256
+  depth: 8
+  stereo: true
+  num_stems: 1
+  time_transformer_depth: 1
+  freq_transformer_depth: 1
+  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
+  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:
+  - dry
+  - No dry
+  lr: 1.0e-05
+  patience: 2
+  reduce_factor: 0.95
+  target_instrument: dry
+  num_epochs: 1000
+  num_steps: 1000
+  q: 0.95
+  coarse_loss_clip: true
+  ema_momentum: 0.999
+  optimizer: adam
+  other_fix: false
+  use_amp: true
+
+inference:
+  batch_size: 1
+  dim_t: 801
+  num_overlap: 4

models/audio-separator-models/roformers/config_dereverb_echo_mel_band_roformer_v2.yaml

@@ -0,0 +1,64 @@
+audio:
+  chunk_size: 352800
+  dim_f: 1024
+  dim_t: 801
+  hop_length: 441
+  n_fft: 2048
+  num_channels: 2
+  sample_rate: 44100
+  min_mean_abs: 0.000
+
+model:
+  dim: 256
+  depth: 8
+  stereo: true
+  num_stems: 1
+  time_transformer_depth: 1
+  freq_transformer_depth: 1
+  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
+  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:
+  - dry
+  - other
+  lr: 1.0e-05
+  patience: 2
+  reduce_factor: 0.95
+  target_instrument: dry
+  num_epochs: 1000
+  num_steps: 1000
+  q: 0.95
+  coarse_loss_clip: true
+  ema_momentum: 0.999
+  optimizer: adam
+  other_fix: false
+  use_amp: true
+
+inference:
+  batch_size: 1
+  dim_t: 801
+  num_overlap: 4

models/audio-separator-models/roformers/config_mel_band_roformer_bleed_suppressor_v1.yaml

@@ -0,0 +1,51 @@
+audio:
+  chunk_size: 485100
+  dim_f: 1024
+  dim_t: 801
+  hop_length: 441
+  n_fft: 2048
+  num_channels: 2
+  sample_rate: 44100
+  min_mean_abs: 0.000
+
+model:
+  dim: 384
+  depth: 6
+  stereo: true
+  num_stems: 1
+  time_transformer_depth: 1
+  freq_transformer_depth: 1
+  num_bands: 60
+  dim_head: 64
+  heads: 8
+  attn_dropout: 0
+  ff_dropout: 0
+  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:
+  instruments:
+  - Instrumental
+  - Bleed
+  target_instrument: Instrumental
+  use_amp: True
+
+inference:
+  batch_size: 1
+  dim_t: 801
+  num_overlap: 2

models/audio-separator-models/roformers/config_mel_band_roformer_instrumental_becruily.yaml

@@ -0,0 +1,72 @@
+audio:
+  chunk_size: 352800
+  dim_f: 1024
+  dim_t: 256
+  hop_length: 441
+  n_fft: 2048
+  num_channels: 2
+  sample_rate: 44100
+  min_mean_abs: 0.000
+
+model:
+  dim: 384
+  depth: 6
+  stereo: true
+  num_stems: 1
+  time_transformer_depth: 1
+  freq_transformer_depth: 1
+  num_bands: 60
+  dim_head: 64
+  heads: 8
+  attn_dropout: 0
+  ff_dropout: 0
+  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: 1
+  grad_clip: 0
+  instruments:
+  - Instrumental
+  - Vocals
+  lr: 0.0005
+  patience: 2
+  reduce_factor: 0.95
+  target_instrument: Instrumental
+  num_epochs: 1000
+  num_steps: 1000
+  augmentation: false # enable augmentations by audiomentations and pedalboard
+  augmentation_type: null
+  use_mp3_compress: false # Deprecated
+  augmentation_mix: false # Mix several stems of the same type with some probability
+  augmentation_loudness: false # randomly change loudness of each stem
+  augmentation_loudness_type: 1 # Type 1 or 2
+  augmentation_loudness_min: 0
+  augmentation_loudness_max: 0
+  q: 0.95
+  coarse_loss_clip: false
+  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
+
+inference:
+  batch_size: 1
+  dim_t: 1101
+  num_overlap: 2

models/audio-separator-models/roformers/config_mel_band_roformer_instrumental_gabox.yaml

@@ -0,0 +1,51 @@
+audio:
+  chunk_size: 485100
+  dim_f: 1024
+  dim_t: 1101
+  hop_length: 441
+  n_fft: 2048
+  num_channels: 2
+  sample_rate: 44100
+  min_mean_abs: 0.000
+
+model:
+  dim: 384
+  depth: 6
+  stereo: true
+  num_stems: 1
+  time_transformer_depth: 1
+  freq_transformer_depth: 1
+  num_bands: 60
+  dim_head: 64
+  heads: 8
+  attn_dropout: 0
+  ff_dropout: 0
+  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:
+  instruments:
+  - Instrumental
+  - Vocals
+  target_instrument: Instrumental
+  use_amp: True
+
+inference:
+  batch_size: 1
+  dim_t: 1101
+  num_overlap: 2

models/audio-separator-models/roformers/config_mel_band_roformer_karaoke_becruily.yaml

@@ -0,0 +1,72 @@
+audio:
+  chunk_size: 485100
+  dim_f: 1024
+  dim_t: 256
+  hop_length: 441
+  n_fft: 2048
+  num_channels: 2
+  sample_rate: 44100
+  min_mean_abs: 0.000
+
+model:
+  dim: 384
+  depth: 6
+  stereo: true
+  num_stems: 2
+  time_transformer_depth: 1
+  freq_transformer_depth: 1
+  num_bands: 60
+  dim_head: 64
+  heads: 8
+  attn_dropout: 0
+  ff_dropout: 0
+  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: 1
+  grad_clip: 0
+  instruments:
+  - Vocals
+  - Instrumental
+  lr: 0.0005
+  patience: 2
+  reduce_factor: 0.95
+  target_instrument: null
+  num_epochs: 1000
+  num_steps: 1000
+  augmentation: false # enable augmentations by audiomentations and pedalboard
+  augmentation_type:
+  use_mp3_compress: false # Deprecated
+  augmentation_mix: false # Mix several stems of the same type with some probability
+  augmentation_loudness: false # randomly change loudness of each stem
+  augmentation_loudness_type: 1 # Type 1 or 2
+  augmentation_loudness_min: 0
+  augmentation_loudness_max: 0
+  q: 0.95
+  coarse_loss_clip: false
+  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
+
+inference:
+  batch_size: 1
+  dim_t: 1101
+  num_overlap: 8

models/audio-separator-models/roformers/config_mel_band_roformer_karaoke_gabox.yaml

@@ -0,0 +1,72 @@
+audio:
+  chunk_size: 485100
+  dim_f: 1024
+  dim_t: 256
+  hop_length: 441
+  n_fft: 2048
+  num_channels: 2
+  sample_rate: 44100
+  min_mean_abs: 0.000
+
+model:
+  dim: 384
+  depth: 6
+  stereo: true
+  num_stems: 1
+  time_transformer_depth: 1
+  freq_transformer_depth: 1
+  num_bands: 60
+  dim_head: 64
+  heads: 8
+  attn_dropout: 0
+  ff_dropout: 0
+  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: true
+
+training:
+  batch_size: 1
+  gradient_accumulation_steps: 1
+  grad_clip: 0
+  instruments:
+    - Vocals
+    - Instrumental
+  lr: 0.0005
+  patience: 2
+  reduce_factor: 0.95
+  target_instrument: Vocals
+  num_epochs: 1000
+  num_steps: 1000
+  augmentation: false # enable augmentations by audiomentations and pedalboard
+  augmentation_type:
+  use_mp3_compress: false # Deprecated
+  augmentation_mix: false # Mix several stems of the same type with some probability
+  augmentation_loudness: false # randomly change loudness of each stem
+  augmentation_loudness_type: 1 # Type 1 or 2
+  augmentation_loudness_min: 0
+  augmentation_loudness_max: 0
+  q: 0.95
+  coarse_loss_clip: false
+  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
+
+inference:
+  batch_size: 1
+  dim_t: 1101
+  num_overlap: 8

models/audio-separator-models/roformers/config_mel_band_roformer_kim_ft_unwa.yaml

@@ -0,0 +1,72 @@
+audio:
+  chunk_size: 485100
+  dim_f: 1024
+  dim_t: 801
+  hop_length: 441
+  n_fft: 2048
+  num_channels: 2
+  sample_rate: 44100
+  min_mean_abs: 0.000
+
+model:
+  dim: 384
+  depth: 6
+  stereo: true
+  num_stems: 1
+  time_transformer_depth: 1
+  freq_transformer_depth: 1
+  num_bands: 60
+  dim_head: 64
+  heads: 8
+  attn_dropout: 0
+  ff_dropout: 0
+  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: 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
+  augmentation: false # enable augmentations by audiomentations and pedalboard
+  augmentation_type: null
+  use_mp3_compress: false # Deprecated
+  augmentation_mix: false # Mix several stems of the same type with some probability
+  augmentation_loudness: false # randomly change loudness of each stem
+  augmentation_loudness_type: 1 # Type 1 or 2
+  augmentation_loudness_min: 0
+  augmentation_loudness_max: 0
+  q: 0.95
+  coarse_loss_clip: false
+  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: 1
+  dim_t: 801
+  num_overlap: 8

models/audio-separator-models/roformers/config_mel_band_roformer_vocal_fullness_aname.yaml

@@ -0,0 +1,54 @@
+audio:
+  chunk_size: 661500
+  dim_f: 1024
+  dim_t: 256
+  hop_length: 441
+  n_fft: 2048
+  num_channels: 2
+  sample_rate: 44100
+  min_mean_abs: 0.001
+
+model:
+  dim: 384
+  depth: 6
+  stereo: true
+  num_stems: 1
+  time_transformer_depth: 1
+  freq_transformer_depth: 1
+  num_bands: 60
+  dim_head: 64
+  heads: 8
+  attn_dropout: 0
+  ff_dropout: 0
+  flash_attn: True
+  dim_freqs_in: 1025
+  sample_rate: 44100
+  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: 1
+  grad_clip: 0
+  instruments:
+  - vocals
+  - other
+  target_instrument: vocals
+  use_amp: true
+
+inference:
+  batch_size: 4
+  dim_t: 1101
+  num_overlap: 4

models/audio-separator-models/roformers/config_mel_band_roformer_vocals_becruily.yaml

@@ -0,0 +1,72 @@
+audio:
+  chunk_size: 352800
+  dim_f: 1024
+  dim_t: 256
+  hop_length: 441
+  n_fft: 2048
+  num_channels: 2
+  sample_rate: 44100
+  min_mean_abs: 0.000
+
+model:
+  dim: 384
+  depth: 6
+  stereo: true
+  num_stems: 1
+  time_transformer_depth: 1
+  freq_transformer_depth: 1
+  num_bands: 60
+  dim_head: 64
+  heads: 8
+  attn_dropout: 0
+  ff_dropout: 0
+  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: 1
+  grad_clip: 0
+  instruments:
+  - vocals
+  - other
+  lr: 0.0005
+  patience: 2
+  reduce_factor: 0.95
+  target_instrument: vocals
+  num_epochs: 1000
+  num_steps: 1000
+  augmentation: false # enable augmentations by audiomentations and pedalboard
+  augmentation_type: null
+  use_mp3_compress: false # Deprecated
+  augmentation_mix: false # Mix several stems of the same type with some probability
+  augmentation_loudness: false # randomly change loudness of each stem
+  augmentation_loudness_type: 1 # Type 1 or 2
+  augmentation_loudness_min: 0
+  augmentation_loudness_max: 0
+  q: 0.95
+  coarse_loss_clip: false
+  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
+
+inference:
+  batch_size: 1
+  dim_t: 1101
+  num_overlap: 2

models/audio-separator-models/roformers/config_mel_band_roformer_vocals_gabox.yaml

@@ -0,0 +1,51 @@
+audio:
+  chunk_size: 352800
+  dim_f: 1024
+  dim_t: 256
+  hop_length: 441
+  n_fft: 2048
+  num_channels: 2
+  sample_rate: 44100
+  min_mean_abs: 0.001
+
+model:
+  dim: 384
+  depth: 6
+  stereo: true
+  num_stems: 1
+  time_transformer_depth: 1
+  freq_transformer_depth: 1
+  num_bands: 60
+  dim_head: 64
+  heads: 8
+  attn_dropout: 0
+  ff_dropout: 0
+  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:
+  instruments:
+  - Vocals
+  - Instrumental
+  target_instrument: Vocals
+
+inference:
+  batch_size: 1
+  dim_t: 1101
+  num_overlap: 1
+  chunk_size: 352800

models/audio-separator-models/roformers/config_melband_roformer_big_beta5e.yaml

@@ -0,0 +1,51 @@
+audio:
+  chunk_size: 485100
+  dim_f: 1024
+  dim_t: 801
+  hop_length: 441
+  n_fft: 2048
+  num_channels: 2
+  sample_rate: 44100
+  min_mean_abs: 0.000
+
+model:
+  dim: 384
+  depth: 6
+  stereo: true
+  num_stems: 1
+  time_transformer_depth: 1
+  freq_transformer_depth: 1
+  num_bands: 60
+  dim_head: 64
+  heads: 8
+  attn_dropout: 0
+  ff_dropout: 0
+  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: 3
+  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:
+  instruments:
+  - vocals
+  - other
+  target_instrument: vocals
+  use_amp: True
+
+inference:
+  batch_size: 1
+  dim_t: 801
+  num_overlap: 2

models/audio-separator-models/roformers/config_melbandroformer_big_beta4.yaml

@@ -0,0 +1,51 @@
+audio:
+  chunk_size: 485100
+  dim_f: 1024
+  dim_t: 1101
+  hop_length: 441
+  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
+  num_bands: 60
+  dim_head: 64
+  heads: 8
+  attn_dropout: 0
+  ff_dropout: 0
+  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: 3
+  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:
+  instruments:
+  - vocals
+  - other
+  target_instrument: vocals
+  use_amp: True
+
+inference:
+  batch_size: 1
+  dim_t: 1101
+  num_overlap: 2

models/audio-separator-models/roformers/config_melbandroformer_big_beta6.yaml

@@ -0,0 +1,72 @@
+audio:
+  chunk_size: 529200
+  dim_f: 1024
+  dim_t: 256
+  hop_length: 441
+  n_fft: 2048
+  num_channels: 2
+  sample_rate: 44100
+  min_mean_abs: 0.000
+
+model:
+  dim: 512
+  depth: 6
+  stereo: true
+  num_stems: 1
+  time_transformer_depth: 1
+  freq_transformer_depth: 1
+  num_bands: 60
+  dim_head: 64
+  heads: 8
+  attn_dropout: 0
+  ff_dropout: 0
+  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: 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
+  augmentation: false # enable augmentations by audiomentations and pedalboard
+  augmentation_type: null
+  use_mp3_compress: false # Deprecated
+  augmentation_mix: false # Mix several stems of the same type with some probability
+  augmentation_loudness: false # randomly change loudness of each stem
+  augmentation_loudness_type: 1 # Type 1 or 2
+  augmentation_loudness_min: 0
+  augmentation_loudness_max: 0
+  q: 0.95
+  coarse_loss_clip: false
+  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: 2
+  dim_t: 1201
+  num_overlap: 2

models/audio-separator-models/roformers/config_melbandroformer_big_beta6x.yaml

@@ -0,0 +1,72 @@
+audio:
+  chunk_size: 529200
+  dim_f: 1024
+  dim_t: 256
+  hop_length: 441
+  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
+  num_bands: 60
+  dim_head: 64
+  heads: 8
+  attn_dropout: 0
+  ff_dropout: 0
+  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: 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
+  augmentation: false # enable augmentations by audiomentations and pedalboard
+  augmentation_type: null
+  use_mp3_compress: false # Deprecated
+  augmentation_mix: false # Mix several stems of the same type with some probability
+  augmentation_loudness: false # randomly change loudness of each stem
+  augmentation_loudness_type: 1 # Type 1 or 2
+  augmentation_loudness_min: 0
+  augmentation_loudness_max: 0
+  q: 0.95
+  coarse_loss_clip: false
+  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: 2
+  dim_t: 1201
+  num_overlap: 2

models/audio-separator-models/roformers/config_melbandroformer_inst.yaml

@@ -0,0 +1,51 @@
+audio:
+  chunk_size: 485100
+  dim_f: 1024
+  dim_t: 1101
+  hop_length: 441
+  n_fft: 2048
+  num_channels: 2
+  sample_rate: 44100
+  min_mean_abs: 0.000
+
+model:
+  dim: 384
+  depth: 6
+  stereo: true
+  num_stems: 1
+  time_transformer_depth: 1
+  freq_transformer_depth: 1
+  num_bands: 60
+  dim_head: 64
+  heads: 8
+  attn_dropout: 0
+  ff_dropout: 0
+  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:
+  instruments:
+  - other
+  - vocals
+  target_instrument: other
+  use_amp: True
+
+inference:
+  batch_size: 1
+  dim_t: 1101
+  num_overlap: 2

models/audio-separator-models/roformers/config_melbandroformer_inst_v2.yaml

@@ -0,0 +1,51 @@
+audio:
+  chunk_size: 485100
+  dim_f: 1024
+  dim_t: 1101
+  hop_length: 441
+  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
+  num_bands: 60
+  dim_head: 64
+  heads: 8
+  attn_dropout: 0
+  ff_dropout: 0
+  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: 3
+  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:
+  instruments:
+  - Instrumental
+  - Vocals
+  target_instrument: Instrumental
+  use_amp: True
+
+inference:
+  batch_size: 1
+  dim_t: 1101
+  num_overlap: 2

models/audio-separator-models/roformers/config_melbandroformer_instvoc_duality.yaml

@@ -0,0 +1,51 @@
+audio:
+  chunk_size: 485100
+  dim_f: 1024
+  dim_t: 256
+  hop_length: 441
+  n_fft: 2048
+  num_channels: 2
+  sample_rate: 44100
+  min_mean_abs: 0.000
+
+model:
+  dim: 384
+  depth: 6
+  stereo: true
+  num_stems: 2
+  time_transformer_depth: 1
+  freq_transformer_depth: 1
+  num_bands: 60
+  dim_head: 64
+  heads: 8
+  attn_dropout: 0
+  ff_dropout: 0
+  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:
+  instruments:
+  - Vocals
+  - Instrumental
+  target_instrument: null
+  use_amp: True
+
+inference:
+  batch_size: 1
+  dim_t: 1101
+  num_overlap: 2

models/audio-separator-models/roformers/config_vocals_mel_band_roformer_big_v1_ft.yaml

@@ -0,0 +1,51 @@
+audio:
+  chunk_size: 485100
+  dim_f: 1024
+  dim_t: 801
+  hop_length: 441
+  n_fft: 2048
+  num_channels: 2
+  sample_rate: 44100
+  min_mean_abs: 0.000
+
+model:
+  dim: 384
+  depth: 6
+  stereo: true
+  num_stems: 1
+  time_transformer_depth: 1
+  freq_transformer_depth: 1
+  num_bands: 60
+  dim_head: 64
+  heads: 8
+  attn_dropout: 0
+  ff_dropout: 0
+  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: 3
+  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:
+  instruments:
+  - vocals
+  - other
+  target_instrument: vocals
+  use_amp: True
+
+inference:
+  batch_size: 1
+  dim_t: 801
+  num_overlap: 2

models/audio-separator-models/roformers/config_vocals_mel_band_roformer_ft.yaml

@@ -0,0 +1,72 @@
+audio:
+  chunk_size: 352800
+  dim_f: 1024
+  dim_t: 256
+  hop_length: 441
+  n_fft: 2048
+  num_channels: 2
+  sample_rate: 44100
+  min_mean_abs: 0.000
+
+model:
+  dim: 384
+  depth: 6
+  stereo: true
+  num_stems: 1
+  time_transformer_depth: 1
+  freq_transformer_depth: 1
+  num_bands: 60
+  dim_head: 64
+  heads: 8
+  attn_dropout: 0
+  ff_dropout: 0
+  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: 1.0e-04
+  patience: 2
+  reduce_factor: 0.95
+  target_instrument: vocals
+  num_epochs: 1000
+  num_steps: 100
+  augmentation: true # enable augmentations by audiomentations and pedalboard
+  augmentation_type: null
+  use_mp3_compress: false # Deprecated
+  augmentation_mix: true # Mix several stems of the same type with some probability
+  augmentation_loudness: true # randomly change loudness of each stem
+  augmentation_loudness_type: 1 # Type 1 or 2
+  augmentation_loudness_min: 0
+  augmentation_loudness_max: 0
+  q: 0.95
+  coarse_loss_clip: false
+  ema_momentum: 0.999
+  optimizer: adamw8bit
+  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: 256
+  num_overlap: 2

models/audio-separator-models/roformers/denoise_mel_band_roformer_aufr33_aggr_sdr_27.9768.ckpt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:a25e3b233722cd81e2de7b8e798a3fef29d4b9799ccacda60b0dc958a1e2a5bb
+size 913097300

models/audio-separator-models/roformers/denoise_mel_band_roformer_aufr33_aggr_sdr_27.9768_config.yaml

@@ -0,0 +1,71 @@
+audio:
+  chunk_size: 352800
+  dim_f: 1024
+  dim_t: 801
+  hop_length: 441
+  n_fft: 2048
+  num_channels: 2
+  sample_rate: 44100
+  min_mean_abs: 000
+
+model:
+  dim: 384
+  depth: 6
+  stereo: true
+  num_stems: 1
+  time_transformer_depth: 1
+  freq_transformer_depth: 1
+  num_bands: 60
+  dim_head: 64
+  heads: 8
+  attn_dropout: 0
+  ff_dropout: 0
+  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: 2
+  gradient_accumulation_steps: 1
+  grad_clip: 0
+  instruments:
+  - dry
+  - other
+  lr: 1.0e-05
+  patience: 8
+  reduce_factor: 0.95
+  target_instrument: dry
+  num_epochs: 1000
+  num_steps: 4032
+  augmentation: false # enable augmentations by audiomentations and pedalboard
+  augmentation_type: null
+  use_mp3_compress: false # Deprecated
+  augmentation_mix: false # Mix several stems of the same type with some probability
+  augmentation_loudness: false # randomly change loudness of each stem
+  augmentation_loudness_type: 1 # Type 1 or 2
+  augmentation_loudness_min: 0
+  augmentation_loudness_max: 0
+  q: 0.95
+  coarse_loss_clip: false
+  ema_momentum: 0.999
+  optimizer: adam
+  other_fix: false # it's needed for checking on multisong dataset if other is actually instrumental
+    
+inference:
+  batch_size: 2
+  dim_t: 801
+  num_overlap: 4

models/audio-separator-models/roformers/denoise_mel_band_roformer_aufr33_sdr_27.9959.ckpt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:7c1c39191edc34e942ca7f2346ce6b6c0e1208a5f76349ffce6f696bd12910de
+size 913097300

models/audio-separator-models/roformers/denoise_mel_band_roformer_aufr33_sdr_27.9959_config.yaml

@@ -0,0 +1,71 @@
+audio:
+  chunk_size: 352800
+  dim_f: 1024
+  dim_t: 801
+  hop_length: 441
+  n_fft: 2048
+  num_channels: 2
+  sample_rate: 44100
+  min_mean_abs: 000
+
+model:
+  dim: 384
+  depth: 6
+  stereo: true
+  num_stems: 1
+  time_transformer_depth: 1
+  freq_transformer_depth: 1
+  num_bands: 60
+  dim_head: 64
+  heads: 8
+  attn_dropout: 0
+  ff_dropout: 0
+  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: 2
+  gradient_accumulation_steps: 1
+  grad_clip: 0
+  instruments:
+  - dry
+  - other
+  lr: 1.0e-05
+  patience: 8
+  reduce_factor: 0.95
+  target_instrument: dry
+  num_epochs: 1000
+  num_steps: 4032
+  augmentation: false # enable augmentations by audiomentations and pedalboard
+  augmentation_type: null
+  use_mp3_compress: false # Deprecated
+  augmentation_mix: false # Mix several stems of the same type with some probability
+  augmentation_loudness: false # randomly change loudness of each stem
+  augmentation_loudness_type: 1 # Type 1 or 2
+  augmentation_loudness_min: 0
+  augmentation_loudness_max: 0
+  q: 0.95
+  coarse_loss_clip: false
+  ema_momentum: 0.999
+  optimizer: adam
+  other_fix: false # it's needed for checking on multisong dataset if other is actually instrumental
+    
+inference:
+  batch_size: 2
+  dim_t: 801
+  num_overlap: 4

models/audio-separator-models/roformers/dereverb-echo_mel_band_roformer_sdr_10.0169.ckpt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:cd2b737a394cfb80cd48cc9fcbaf89f5f4062f6b93066c2911617a06d8b7860a
+size 835997896

models/audio-separator-models/roformers/dereverb-echo_mel_band_roformer_sdr_13.4843_v2.ckpt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:396432f5af25992fe82d0286634bd879027c073721db6ab10199e75459708b9f
+size 455862568

models/audio-separator-models/roformers/dereverb_big_mbr_ep_362.ckpt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:0506455e74ffc02bbec700df9863ae243597034003815f1418227c6dee33b6ea
+size 455864012