Hugging Face's logo

anvuew
/
dereverb_room

Model card Files
xet
 Community
anvuew commited on
Commit
97e7a63
·
1 Parent(s): 3304b37

add model

Browse files
Files changed (2) hide show
  1. dereverb_room_anvuew.yaml +132 -0
  2. dereverb_room_anvuew_sdr_13.7432.ckpt +3 -0
dereverb_room_anvuew.yaml ADDED
@@ -0,0 +1,132 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ audio:
2
+ chunk_size: 384000
3
+ dim_f: 1024
4
+ dim_t: 801 # don't work (use in model)
5
+ hop_length: 441 # don't work (use in model)
6
+ n_fft: 2048
7
+ num_channels: 1
8
+ sample_rate: 44100
9
+ min_mean_abs: 0.000
10
+
11
+ model:
12
+ dim: 128
13
+ depth: 16
14
+ stereo: false
15
+ num_stems: 1
16
+ time_transformer_depth: 1
17
+ freq_transformer_depth: 1
18
+ linear_transformer_depth: 0
19
+ freqs_per_bands: !!python/tuple
20
+ - 2
21
+ - 2
22
+ - 2
23
+ - 2
24
+ - 2
25
+ - 2
26
+ - 3
27
+ - 3
28
+ - 3
29
+ - 3
30
+ - 3
31
+ - 4
32
+ - 4
33
+ - 4
34
+ - 4
35
+ - 4
36
+ - 5
37
+ - 5
38
+ - 5
39
+ - 5
40
+ - 6
41
+ - 6
42
+ - 6
43
+ - 6
44
+ - 7
45
+ - 7
46
+ - 7
47
+ - 8
48
+ - 8
49
+ - 8
50
+ - 9
51
+ - 9
52
+ - 10
53
+ - 10
54
+ - 11
55
+ - 12
56
+ - 13
57
+ - 14
58
+ - 15
59
+ - 16
60
+ - 17
61
+ - 18
62
+ - 19
63
+ - 20
64
+ - 21
65
+ - 22
66
+ - 23
67
+ - 24
68
+ - 25
69
+ - 27
70
+ - 29
71
+ - 31
72
+ - 33
73
+ - 35
74
+ - 37
75
+ - 39
76
+ - 41
77
+ - 43
78
+ - 45
79
+ - 48
80
+ - 52
81
+ - 57
82
+ - 64
83
+ dim_head: 16
84
+ heads: 8
85
+ attn_dropout: 0.0
86
+ ff_dropout: 0.0
87
+ flash_attn: true
88
+ dim_freqs_in: 1025
89
+ stft_n_fft: 2048
90
+ stft_hop_length: 512
91
+ stft_win_length: 2048
92
+ stft_normalized: False
93
+ mask_estimator_depth: 3
94
+ multi_stft_resolution_loss_weight: 1.0
95
+ multi_stft_resolutions_window_sizes: !!python/tuple
96
+ - 4096
97
+ - 2048
98
+ - 1024
99
+ - 512
100
+ - 256
101
+ multi_stft_hop_size: 147
102
+ multi_stft_normalized: False
103
+ mlp_expansion_factor: 4
104
+ use_torch_checkpoint: True
105
+ skip_connection: False
106
+
107
+
108
+ training:
109
+ batch_size: 4
110
+ gradient_accumulation_steps: 1
111
+ grad_clip: 1000.0
112
+ instruments: ['noreverb', 'reverb']
113
+ lr: 5.0e-5
114
+ patience: 5
115
+ reduce_factor: 0.75
116
+ target_instrument: noreverb
117
+ num_epochs: 1000
118
+ num_steps: 1000
119
+ q: 0.95
120
+ coarse_loss_clip: true
121
+ ema_momentum: 0.999
122
+ optimizer: adam
123
+
124
+ other_fix: False # it's needed for checking on multisong dataset if other is actually instrumental
125
+ use_amp: true # enable or disable usage of mixed precision (float16) - usually it must be true
126
+
127
+
128
+
129
+ inference:
130
+ batch_size: 1
131
+ dim_t: 871
132
+ num_overlap: 1
dereverb_room_anvuew_sdr_13.7432.ckpt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:2edec521f09e26341c1923dc82c8c52dbc86478b42b9999f679535743c970cb3
3
+ size 118128452