pcunwa commited on
Commit
4e47d66
·
verified ·
1 Parent(s): d90408b

Upload 4 files

Browse files
Xe/bs_leap_xe_inst.ckpt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:33ee9415f491d257fa7a79f6e92a80d647bc71ed622bfef9a137b6e4250307d3
3
+ size 267798769
Xe/bs_leap_xe_voc.ckpt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:b739c1d2d87a81cd3dd3844ed9ad0bd678708c7a0a761a03a1aaff9af79a096d
3
+ size 267796851
Xe/leap_xe_config_inst.yaml ADDED
@@ -0,0 +1,156 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ audio:
2
+ chunk_size: 881559
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: 2
8
+ sample_rate: 44100
9
+ min_mean_abs: 0.
10
+
11
+ model:
12
+ dim: 256
13
+ depth: 16
14
+ stereo: true
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
+ - 2
27
+ - 2
28
+ - 2
29
+ - 2
30
+ - 2
31
+ - 2
32
+ - 2
33
+ - 2
34
+ - 2
35
+ - 2
36
+ - 2
37
+ - 2
38
+ - 2
39
+ - 2
40
+ - 2
41
+ - 2
42
+ - 2
43
+ - 2
44
+ - 4
45
+ - 4
46
+ - 4
47
+ - 4
48
+ - 4
49
+ - 4
50
+ - 4
51
+ - 4
52
+ - 4
53
+ - 4
54
+ - 4
55
+ - 4
56
+ - 4
57
+ - 4
58
+ - 4
59
+ - 4
60
+ - 4
61
+ - 4
62
+ - 4
63
+ - 4
64
+ - 4
65
+ - 4
66
+ - 4
67
+ - 4
68
+ - 4
69
+ - 4
70
+ - 4
71
+ - 4
72
+ - 4
73
+ - 4
74
+ - 4
75
+ - 4
76
+ - 4
77
+ - 4
78
+ - 4
79
+ - 4
80
+ - 12
81
+ - 12
82
+ - 12
83
+ - 12
84
+ - 12
85
+ - 12
86
+ - 12
87
+ - 12
88
+ - 12
89
+ - 12
90
+ - 12
91
+ - 12
92
+ - 12
93
+ - 12
94
+ - 12
95
+ - 12
96
+ - 24
97
+ - 24
98
+ - 24
99
+ - 24
100
+ - 24
101
+ - 24
102
+ - 24
103
+ - 24
104
+ - 48
105
+ - 48
106
+ - 48
107
+ - 48
108
+ - 128
109
+ - 129
110
+ dim_head: 64
111
+ heads: 8
112
+ attn_dropout: 0.0
113
+ ff_dropout: 0.0
114
+ flash_attn: true
115
+ dim_freqs_in: 1025
116
+ stft_n_fft: 2048
117
+ stft_hop_length: 512
118
+ stft_win_length: 2048
119
+ stft_normalized: false
120
+ mask_estimator_depth: 2
121
+ multi_stft_resolution_loss_weight: 1.0
122
+ multi_stft_resolutions_window_sizes: !!python/tuple
123
+ - 4096
124
+ - 2048
125
+ - 1024
126
+ - 512
127
+ - 256
128
+ multi_stft_hop_size: 147
129
+ multi_stft_normalized: False
130
+ mlp_expansion_factor: 4
131
+ use_torch_checkpoint: False
132
+ skip_connection: False
133
+
134
+
135
+ training:
136
+ batch_size: 1
137
+ gradient_accumulation_steps: 1
138
+ grad_clip: 0
139
+ instruments: ['vocals', 'other']
140
+ lr: 1.0e-04
141
+ patience: 5
142
+ reduce_factor: 0.9
143
+ target_instrument: other
144
+ num_epochs: 1000
145
+ num_steps: 1000
146
+ q: 0.95
147
+ coarse_loss_clip: true
148
+ ema_momentum: 0.999
149
+ optimizer: adamw
150
+ other_fix: false # it's needed for checking on multisong dataset if other is actually instrumental
151
+ use_amp: true # enable or disable usage of mixed precision (float16) - usually it must be true
152
+
153
+ inference:
154
+ batch_size: 2
155
+ dim_t: 1876
156
+ num_overlap: 2
Xe/leap_xe_config_voc.yaml ADDED
@@ -0,0 +1,156 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ audio:
2
+ chunk_size: 881559
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: 2
8
+ sample_rate: 44100
9
+ min_mean_abs: 0.
10
+
11
+ model:
12
+ dim: 256
13
+ depth: 16
14
+ stereo: true
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
+ - 2
27
+ - 2
28
+ - 2
29
+ - 2
30
+ - 2
31
+ - 2
32
+ - 2
33
+ - 2
34
+ - 2
35
+ - 2
36
+ - 2
37
+ - 2
38
+ - 2
39
+ - 2
40
+ - 2
41
+ - 2
42
+ - 2
43
+ - 2
44
+ - 4
45
+ - 4
46
+ - 4
47
+ - 4
48
+ - 4
49
+ - 4
50
+ - 4
51
+ - 4
52
+ - 4
53
+ - 4
54
+ - 4
55
+ - 4
56
+ - 4
57
+ - 4
58
+ - 4
59
+ - 4
60
+ - 4
61
+ - 4
62
+ - 4
63
+ - 4
64
+ - 4
65
+ - 4
66
+ - 4
67
+ - 4
68
+ - 4
69
+ - 4
70
+ - 4
71
+ - 4
72
+ - 4
73
+ - 4
74
+ - 4
75
+ - 4
76
+ - 4
77
+ - 4
78
+ - 4
79
+ - 4
80
+ - 12
81
+ - 12
82
+ - 12
83
+ - 12
84
+ - 12
85
+ - 12
86
+ - 12
87
+ - 12
88
+ - 12
89
+ - 12
90
+ - 12
91
+ - 12
92
+ - 12
93
+ - 12
94
+ - 12
95
+ - 12
96
+ - 24
97
+ - 24
98
+ - 24
99
+ - 24
100
+ - 24
101
+ - 24
102
+ - 24
103
+ - 24
104
+ - 48
105
+ - 48
106
+ - 48
107
+ - 48
108
+ - 128
109
+ - 129
110
+ dim_head: 64
111
+ heads: 8
112
+ attn_dropout: 0.0
113
+ ff_dropout: 0.0
114
+ flash_attn: true
115
+ dim_freqs_in: 1025
116
+ stft_n_fft: 2048
117
+ stft_hop_length: 512
118
+ stft_win_length: 2048
119
+ stft_normalized: false
120
+ mask_estimator_depth: 2
121
+ multi_stft_resolution_loss_weight: 1.0
122
+ multi_stft_resolutions_window_sizes: !!python/tuple
123
+ - 4096
124
+ - 2048
125
+ - 1024
126
+ - 512
127
+ - 256
128
+ multi_stft_hop_size: 147
129
+ multi_stft_normalized: False
130
+ mlp_expansion_factor: 4
131
+ use_torch_checkpoint: False
132
+ skip_connection: False
133
+
134
+
135
+ training:
136
+ batch_size: 1
137
+ gradient_accumulation_steps: 1
138
+ grad_clip: 0
139
+ instruments: ['vocals', 'other']
140
+ lr: 1.0e-04
141
+ patience: 5
142
+ reduce_factor: 0.9
143
+ target_instrument: vocals
144
+ num_epochs: 1000
145
+ num_steps: 1000
146
+ q: 0.95
147
+ coarse_loss_clip: true
148
+ ema_momentum: 0.999
149
+ optimizer: adamw
150
+ other_fix: false # it's needed for checking on multisong dataset if other is actually instrumental
151
+ use_amp: true # enable or disable usage of mixed precision (float16) - usually it must be true
152
+
153
+ inference:
154
+ batch_size: 2
155
+ dim_t: 1876
156
+ num_overlap: 2