Hugging Face's logo

lainlives
/
Singer-Preprocessors

NeMo
Model card Files
xet
 Community
lainlives commited on
Commit
52007f2
·
verified ·
1 Parent(s): 28ea6c8

Upload 4 files

Browse files
Files changed (4) hide show
  1. mdx23c/MDX23C-8KFFT-InstVoc_HQ_2.ckpt +3 -0
  2. mdx23c/MDX23C-De-Reverb-aufr33-jarredou.ckpt +3 -0
  3. mdx23c/mdx23c_dereverb_aufr33_jarredou_config.yaml +114 -0
  4. mdx23c/mdx23c_instvoc_hq2_config.yaml +44 -0
mdx23c/MDX23C-8KFFT-InstVoc_HQ_2.ckpt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:7d960d8e40a458120412c1bd807e013d2dbca7b959cc9da2bbcb0eb203d1daea
3
+ size 448093770
mdx23c/MDX23C-De-Reverb-aufr33-jarredou.ckpt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:eae2471b707758d74db38ac1b1d5800e12f57c4e9d1ebbb2faf004b8e086e914
3
+ size 448098867
mdx23c/mdx23c_dereverb_aufr33_jarredou_config.yaml ADDED
@@ -0,0 +1,114 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ audio:
2
+ chunk_size: 261120
3
+ dim_f: 4096
4
+ dim_t: 256
5
+ hop_length: 1024
6
+ n_fft: 8192
7
+ num_channels: 2
8
+ sample_rate: 44100
9
+ min_mean_abs: 0.001
10
+ model:
11
+ act: gelu
12
+ bottleneck_factor: 4
13
+ growth: 128
14
+ norm: InstanceNorm
15
+ num_blocks_per_scale: 2
16
+ num_channels: 128
17
+ num_scales: 5
18
+ num_subbands: 4
19
+ scale:
20
+ - 2
21
+ - 2
22
+ training:
23
+ batch_size: 2
24
+ gradient_accumulation_steps: 1
25
+ grad_clip: 0
26
+ instruments:
27
+ - dry
28
+ - other
29
+ lr: 1.0e-06
30
+ patience: 4
31
+ reduce_factor: 0.93
32
+ target_instrument: null
33
+ num_epochs: 40
34
+ num_steps: 1000
35
+ q: 0.95
36
+ coarse_loss_clip: true
37
+ ema_momentum: 0.999
38
+ optimizer: adamw
39
+ read_metadata_procs: 8
40
+ other_fix: false
41
+ use_amp: true
42
+ augmentations:
43
+ enable: false
44
+ loudness: true
45
+ loudness_min: 0.5
46
+ loudness_max: 1.5
47
+ mixup: true
48
+ mixup_probs: !!python/tuple
49
+ - 0.2
50
+ - 0.02
51
+ mixup_loudness_min: 0.5
52
+ mixup_loudness_max: 1.5
53
+ all:
54
+ channel_shuffle: 0.5
55
+ random_inverse: 0.05
56
+ random_polarity: 0.5
57
+ pedalboard_chorus: 0.001
58
+ pedalboard_chorus_rate_hz_min: 1.0
59
+ pedalboard_chorus_rate_hz_max: 7.0
60
+ pedalboard_chorus_depth_min: 0.25
61
+ pedalboard_chorus_depth_max: 0.95
62
+ pedalboard_chorus_centre_delay_ms_min: 3
63
+ pedalboard_chorus_centre_delay_ms_max: 10
64
+ pedalboard_chorus_feedback_min: 0.0
65
+ pedalboard_chorus_feedback_max: 0.01
66
+ pedalboard_chorus_mix_min: 0.1
67
+ pedalboard_chorus_mix_max: 0.9
68
+ pedalboard_phazer: 0.001
69
+ pedalboard_phazer_rate_hz_min: 1.0
70
+ pedalboard_phazer_rate_hz_max: 10.0
71
+ pedalboard_phazer_depth_min: 0.25
72
+ pedalboard_phazer_depth_max: 0.95
73
+ pedalboard_phazer_centre_frequency_hz_min: 200
74
+ pedalboard_phazer_centre_frequency_hz_max: 12000
75
+ pedalboard_phazer_feedback_min: 0.0
76
+ pedalboard_phazer_feedback_max: 0.5
77
+ pedalboard_phazer_mix_min: 0.1
78
+ pedalboard_phazer_mix_max: 0.9
79
+ pedalboard_pitch_shift: 0.01
80
+ pedalboard_pitch_shift_semitones_min: -7
81
+ pedalboard_pitch_shift_semitones_max: 7
82
+ pedalboard_resample: 0.001
83
+ pedalboard_resample_target_sample_rate_min: 4000
84
+ pedalboard_resample_target_sample_rate_max: 44100
85
+ mp3_compression_min_bitrate: 32
86
+ mp3_compression_max_bitrate: 320
87
+ mp3_compression_backend: lameenc
88
+ dry:
89
+ pedalboard_distortion: 0.001
90
+ pedalboard_distortion_drive_db_min: 1.0
91
+ pedalboard_distortion_drive_db_max: 25.0
92
+ tanh_distortion: 0.05
93
+ tanh_distortion_min: 0.1
94
+ tanh_distortion_max: 0.7
95
+ pedalboard_bitcrash: 0.005
96
+ pedalboard_bitcrash_bit_depth_min: 4
97
+ pedalboard_bitcrash_bit_depth_max: 16
98
+ seven_band_parametric_eq: 0.24
99
+ seven_band_parametric_eq_min_gain_db: -9
100
+ seven_band_parametric_eq_max_gain_db: 9
101
+ gaussian_noise: 0.005
102
+ gaussian_noise_min_amplitude: 0.001
103
+ gaussian_noise_max_amplitude: 0.01
104
+ time_stretch: 0.01
105
+ time_stretch_min_rate: 0.8
106
+ time_stretch_max_rate: 1.25
107
+ other:
108
+ seven_band_parametric_eq: 0.24
109
+ seven_band_parametric_eq_min_gain_db: -9
110
+ seven_band_parametric_eq_max_gain_db: 9
111
+ inference:
112
+ batch_size: 1
113
+ dim_t: 256
114
+ num_overlap: 2
mdx23c/mdx23c_instvoc_hq2_config.yaml ADDED
@@ -0,0 +1,44 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ audio:
2
+ chunk_size: 261120
3
+ dim_f: 4096
4
+ dim_t: 256
5
+ hop_length: 1024
6
+ n_fft: 8192
7
+ num_channels: 2
8
+ sample_rate: 44100
9
+ min_mean_abs: 0.001
10
+ model:
11
+ act: gelu
12
+ bottleneck_factor: 4
13
+ growth: 128
14
+ norm: InstanceNorm
15
+ num_blocks_per_scale: 2
16
+ num_channels: 128
17
+ num_scales: 5
18
+ num_subbands: 4
19
+ scale:
20
+ - 2
21
+ - 2
22
+ training:
23
+ batch_size: 6
24
+ grad_clip: 0
25
+ instruments:
26
+ - Vocals
27
+ - Instrumental
28
+ lr: 1.0e-05
29
+ patience: 2
30
+ reduce_factor: 0.95
31
+ target_instrument: null
32
+ num_epochs: 1000
33
+ num_steps: 1000
34
+ augmentation: 1
35
+ augmentation_type: simple1
36
+ augmentation_mix: true
37
+ q: 0.95
38
+ coarse_loss_clip: true
39
+ ema_momentum: 0.999
40
+ use_amp: true
41
+ inference:
42
+ batch_size: 1
43
+ dim_t: 256
44
+ num_overlap: 2