1
1
{
2
- "model" : " CDVAEGAN " ,
2
+ "model" : " CDVAECLSGAN " ,
3
3
"model_module" : " model.cdvae-cls-gan-mcc" ,
4
- "trainer" : " CDVAEGANTrainer " ,
5
- "trainer_module" : " trainer.cdvae-cls-gan-mcc " ,
6
- "stat_dir " : " /mnt/md1/datasets/ vcc2018/world/etc-new " ,
7
- "spklist" : " /mnt/md1/datasets/ vcc2018/world/etc/speakers.tsv " ,
4
+ "trainer" : " CDVAECLSGANTrainer " ,
5
+ "trainer_module" : " trainer.cdvae-cls-gan" ,
6
+ "stats " : " ./data/ vcc2018/stats/stats.h5 " ,
7
+ "spklist" : " ./data/ vcc2018/conf/spk.list " ,
8
8
9
9
"z_dim" : 16 ,
10
10
"y_dim" : 12 ,
58
58
}
59
59
},
60
60
"training" : {
61
- "train_file_pattern" : [" /mnt/md1/datasets/ vcc2018/world/ bin-dynamic /VAD/tr /*/[12]00[0-6]?.bin" ,
62
- " /mnt/md1/datasets/ vcc2018/world/ bin-dynamic /VAD/tr /*/[12]0070.bin" ],
63
- "valid_file_pattern" : [" /mnt/md1/datasets/ vcc2018/world/ bin-dynamic /VAD/tr /*/[12]007[1-9].bin" ,
64
- " /mnt/md1/datasets/ vcc2018/world/ bin-dynamic /VAD/tr /*/[12]008[0-1].bin" ],
61
+ "train_file_pattern" : [" ./data/ vcc2018/bin/VAD/*/[12]00[0-6]?.bin" ,
62
+ " ./data/ vcc2018/bin/VAD/*/[12]0070.bin" ],
63
+ "valid_file_pattern" : [" ./data/ vcc2018/bin/VAD/*/[12]007[1-9].bin" ,
64
+ " ./data/ vcc2018/bin/VAD/*/[12]008[0-1].bin" ],
65
65
"batch_size" : 16 ,
66
66
"crop_length" : 128 ,
67
67
"lr" : 1e-4 ,
68
68
"beta1" : 0.5 ,
69
69
"beta2" : 0.999 ,
70
70
"nIterD" : 5 ,
71
- "vae_iter" : 100000 ,
72
- "cls_iter" : 100000 ,
73
- "gan_iter" : 300000 ,
74
- "max_iter" : 500000 ,
71
+ "vae_iter" : 100 ,
72
+ "cls_iter" : 100 ,
73
+ "gan_iter" : 300 ,
74
+ "max_iter" : 500 ,
75
75
"save_freq" : 10000 ,
76
76
"summary_freq" : 1000 ,
77
- "log_freq" : 1000 ,
77
+ "log_freq" : 10 ,
78
78
"gamma" : 100 ,
79
79
"lambda" : 1000 ,
80
80
"gp_weight" : 10
81
81
},
82
82
"conversion" : {
83
- "input" : " mcc" ,
84
- "output" : " mcc" ,
85
- "test_file_pattern" : " /mnt/md1/datasets/vcc2018/world/bin-dynamic/no_VAD/ev/{}/*.bin"
83
+ "test_file_pattern" : " ./data/vcc2018/bin/noVAD/{}/3*.bin"
86
84
},
87
85
"feat_param" :{
88
86
"fs" : 22050 ,
89
87
"shiftms" : 5 ,
90
88
"fftl" : 1024 ,
91
89
"mcep_alpha" : 0.455 ,
92
- "sp_dim" : 513 ,
93
- "mcc_dim" : 34 ,
94
- "feat_dim" : 2710 ,
90
+ "mcep_dim" : 34 ,
91
+ "feat_dim" : 1064 ,
95
92
"dim" :{
96
93
"sp" : 513 ,
97
- "mcc " : 34 ,
98
- "feat" : 2710
94
+ "mcep " : 34 ,
95
+ "feat" : 1064
99
96
}
100
97
},
101
- "normalizer_files " :{
98
+ "normalizer " :{
102
99
"sp" : {
103
- "type" : " minmax" ,
104
- "dim" : null ,
105
- "max" : " sp_max.npf" ,
106
- "min" : " sp_min.npf"
100
+ "type" : [" minmax" ]
107
101
},
108
102
"mcc" : {
109
- "type" : " minmax" ,
110
- "dim" : null ,
111
- "max" : " mcc_max.npf" ,
112
- "min" : " mcc_min.npf"
103
+ "type" : [" minmax" ]
113
104
}
114
105
}
115
- }
106
+ }
0 commit comments