Fhrozen commited on
Commit
4281e90
1 Parent(s): 42e7292
dump/xvector/dev_phn/spk_xvector.ark ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:1fc0db3de80665cb43f7a04bbeb39fa4dc3c98aae959503417cfe4732b6539c6
3
+ size 39503
dump/xvector/dev_phn/spk_xvector.scp ADDED
@@ -0,0 +1,50 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ jvs001 dump/xvector/dev_phn/spk_xvector.ark:7
2
+ jvs002 dump/xvector/dev_phn/spk_xvector.ark:797
3
+ jvs003 dump/xvector/dev_phn/spk_xvector.ark:1587
4
+ jvs004 dump/xvector/dev_phn/spk_xvector.ark:2377
5
+ jvs005 dump/xvector/dev_phn/spk_xvector.ark:3167
6
+ jvs007 dump/xvector/dev_phn/spk_xvector.ark:3957
7
+ jvs008 dump/xvector/dev_phn/spk_xvector.ark:4747
8
+ jvs009 dump/xvector/dev_phn/spk_xvector.ark:5537
9
+ jvs010 dump/xvector/dev_phn/spk_xvector.ark:6327
10
+ jvs011 dump/xvector/dev_phn/spk_xvector.ark:7117
11
+ jvs012 dump/xvector/dev_phn/spk_xvector.ark:7907
12
+ jvs013 dump/xvector/dev_phn/spk_xvector.ark:8697
13
+ jvs014 dump/xvector/dev_phn/spk_xvector.ark:9487
14
+ jvs015 dump/xvector/dev_phn/spk_xvector.ark:10277
15
+ jvs016 dump/xvector/dev_phn/spk_xvector.ark:11067
16
+ jvs017 dump/xvector/dev_phn/spk_xvector.ark:11857
17
+ jvs018 dump/xvector/dev_phn/spk_xvector.ark:12647
18
+ jvs019 dump/xvector/dev_phn/spk_xvector.ark:13437
19
+ jvs020 dump/xvector/dev_phn/spk_xvector.ark:14227
20
+ jvs021 dump/xvector/dev_phn/spk_xvector.ark:15017
21
+ jvs022 dump/xvector/dev_phn/spk_xvector.ark:15807
22
+ jvs023 dump/xvector/dev_phn/spk_xvector.ark:16597
23
+ jvs024 dump/xvector/dev_phn/spk_xvector.ark:17387
24
+ jvs025 dump/xvector/dev_phn/spk_xvector.ark:18177
25
+ jvs026 dump/xvector/dev_phn/spk_xvector.ark:18967
26
+ jvs027 dump/xvector/dev_phn/spk_xvector.ark:19757
27
+ jvs029 dump/xvector/dev_phn/spk_xvector.ark:20547
28
+ jvs030 dump/xvector/dev_phn/spk_xvector.ark:21337
29
+ jvs031 dump/xvector/dev_phn/spk_xvector.ark:22127
30
+ jvs032 dump/xvector/dev_phn/spk_xvector.ark:22917
31
+ jvs033 dump/xvector/dev_phn/spk_xvector.ark:23707
32
+ jvs034 dump/xvector/dev_phn/spk_xvector.ark:24497
33
+ jvs035 dump/xvector/dev_phn/spk_xvector.ark:25287
34
+ jvs036 dump/xvector/dev_phn/spk_xvector.ark:26077
35
+ jvs037 dump/xvector/dev_phn/spk_xvector.ark:26867
36
+ jvs038 dump/xvector/dev_phn/spk_xvector.ark:27657
37
+ jvs039 dump/xvector/dev_phn/spk_xvector.ark:28447
38
+ jvs040 dump/xvector/dev_phn/spk_xvector.ark:29237
39
+ jvs041 dump/xvector/dev_phn/spk_xvector.ark:30027
40
+ jvs042 dump/xvector/dev_phn/spk_xvector.ark:30817
41
+ jvs043 dump/xvector/dev_phn/spk_xvector.ark:31607
42
+ jvs044 dump/xvector/dev_phn/spk_xvector.ark:32397
43
+ jvs045 dump/xvector/dev_phn/spk_xvector.ark:33187
44
+ jvs046 dump/xvector/dev_phn/spk_xvector.ark:33977
45
+ jvs047 dump/xvector/dev_phn/spk_xvector.ark:34767
46
+ jvs048 dump/xvector/dev_phn/spk_xvector.ark:35557
47
+ jvs049 dump/xvector/dev_phn/spk_xvector.ark:36347
48
+ jvs050 dump/xvector/dev_phn/spk_xvector.ark:37137
49
+ jvs051 dump/xvector/dev_phn/spk_xvector.ark:37927
50
+ tsukuyomi dump/xvector/dev_phn/spk_xvector.ark:38720
dump/xvector/eval1_phn/spk_xvector.ark ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:aef57ebfc1e59ddb8d92e62499b2f842609ff706ed7c451b858212dd52a0d080
3
+ size 39503
dump/xvector/eval1_phn/spk_xvector.scp ADDED
@@ -0,0 +1,50 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ jvs052 dump/xvector/eval1_phn/spk_xvector.ark:7
2
+ jvs053 dump/xvector/eval1_phn/spk_xvector.ark:797
3
+ jvs054 dump/xvector/eval1_phn/spk_xvector.ark:1587
4
+ jvs055 dump/xvector/eval1_phn/spk_xvector.ark:2377
5
+ jvs056 dump/xvector/eval1_phn/spk_xvector.ark:3167
6
+ jvs057 dump/xvector/eval1_phn/spk_xvector.ark:3957
7
+ jvs058 dump/xvector/eval1_phn/spk_xvector.ark:4747
8
+ jvs059 dump/xvector/eval1_phn/spk_xvector.ark:5537
9
+ jvs060 dump/xvector/eval1_phn/spk_xvector.ark:6327
10
+ jvs061 dump/xvector/eval1_phn/spk_xvector.ark:7117
11
+ jvs062 dump/xvector/eval1_phn/spk_xvector.ark:7907
12
+ jvs063 dump/xvector/eval1_phn/spk_xvector.ark:8697
13
+ jvs064 dump/xvector/eval1_phn/spk_xvector.ark:9487
14
+ jvs065 dump/xvector/eval1_phn/spk_xvector.ark:10277
15
+ jvs066 dump/xvector/eval1_phn/spk_xvector.ark:11067
16
+ jvs067 dump/xvector/eval1_phn/spk_xvector.ark:11857
17
+ jvs068 dump/xvector/eval1_phn/spk_xvector.ark:12647
18
+ jvs069 dump/xvector/eval1_phn/spk_xvector.ark:13437
19
+ jvs070 dump/xvector/eval1_phn/spk_xvector.ark:14227
20
+ jvs071 dump/xvector/eval1_phn/spk_xvector.ark:15017
21
+ jvs072 dump/xvector/eval1_phn/spk_xvector.ark:15807
22
+ jvs073 dump/xvector/eval1_phn/spk_xvector.ark:16597
23
+ jvs074 dump/xvector/eval1_phn/spk_xvector.ark:17387
24
+ jvs075 dump/xvector/eval1_phn/spk_xvector.ark:18177
25
+ jvs076 dump/xvector/eval1_phn/spk_xvector.ark:18967
26
+ jvs077 dump/xvector/eval1_phn/spk_xvector.ark:19757
27
+ jvs078 dump/xvector/eval1_phn/spk_xvector.ark:20547
28
+ jvs079 dump/xvector/eval1_phn/spk_xvector.ark:21337
29
+ jvs080 dump/xvector/eval1_phn/spk_xvector.ark:22127
30
+ jvs081 dump/xvector/eval1_phn/spk_xvector.ark:22917
31
+ jvs082 dump/xvector/eval1_phn/spk_xvector.ark:23707
32
+ jvs083 dump/xvector/eval1_phn/spk_xvector.ark:24497
33
+ jvs084 dump/xvector/eval1_phn/spk_xvector.ark:25287
34
+ jvs085 dump/xvector/eval1_phn/spk_xvector.ark:26077
35
+ jvs086 dump/xvector/eval1_phn/spk_xvector.ark:26867
36
+ jvs087 dump/xvector/eval1_phn/spk_xvector.ark:27657
37
+ jvs088 dump/xvector/eval1_phn/spk_xvector.ark:28447
38
+ jvs089 dump/xvector/eval1_phn/spk_xvector.ark:29237
39
+ jvs090 dump/xvector/eval1_phn/spk_xvector.ark:30027
40
+ jvs091 dump/xvector/eval1_phn/spk_xvector.ark:30817
41
+ jvs092 dump/xvector/eval1_phn/spk_xvector.ark:31607
42
+ jvs093 dump/xvector/eval1_phn/spk_xvector.ark:32397
43
+ jvs094 dump/xvector/eval1_phn/spk_xvector.ark:33187
44
+ jvs095 dump/xvector/eval1_phn/spk_xvector.ark:33977
45
+ jvs096 dump/xvector/eval1_phn/spk_xvector.ark:34767
46
+ jvs097 dump/xvector/eval1_phn/spk_xvector.ark:35557
47
+ jvs098 dump/xvector/eval1_phn/spk_xvector.ark:36347
48
+ jvs099 dump/xvector/eval1_phn/spk_xvector.ark:37137
49
+ jvs100 dump/xvector/eval1_phn/spk_xvector.ark:37927
50
+ tsukuyomi dump/xvector/eval1_phn/spk_xvector.ark:38720
dump/xvector/tr_no_dev_phn/spk_xvector.ark ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:f6236d94e2fc43a808f5536e5307040865506bc1201cae001486da31c9d17083
3
+ size 78213
dump/xvector/tr_no_dev_phn/spk_xvector.scp ADDED
@@ -0,0 +1,99 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ jvs001 dump/xvector/tr_no_dev_phn/spk_xvector.ark:7
2
+ jvs002 dump/xvector/tr_no_dev_phn/spk_xvector.ark:797
3
+ jvs003 dump/xvector/tr_no_dev_phn/spk_xvector.ark:1587
4
+ jvs004 dump/xvector/tr_no_dev_phn/spk_xvector.ark:2377
5
+ jvs005 dump/xvector/tr_no_dev_phn/spk_xvector.ark:3167
6
+ jvs007 dump/xvector/tr_no_dev_phn/spk_xvector.ark:3957
7
+ jvs008 dump/xvector/tr_no_dev_phn/spk_xvector.ark:4747
8
+ jvs009 dump/xvector/tr_no_dev_phn/spk_xvector.ark:5537
9
+ jvs010 dump/xvector/tr_no_dev_phn/spk_xvector.ark:6327
10
+ jvs011 dump/xvector/tr_no_dev_phn/spk_xvector.ark:7117
11
+ jvs012 dump/xvector/tr_no_dev_phn/spk_xvector.ark:7907
12
+ jvs013 dump/xvector/tr_no_dev_phn/spk_xvector.ark:8697
13
+ jvs014 dump/xvector/tr_no_dev_phn/spk_xvector.ark:9487
14
+ jvs015 dump/xvector/tr_no_dev_phn/spk_xvector.ark:10277
15
+ jvs016 dump/xvector/tr_no_dev_phn/spk_xvector.ark:11067
16
+ jvs017 dump/xvector/tr_no_dev_phn/spk_xvector.ark:11857
17
+ jvs018 dump/xvector/tr_no_dev_phn/spk_xvector.ark:12647
18
+ jvs019 dump/xvector/tr_no_dev_phn/spk_xvector.ark:13437
19
+ jvs020 dump/xvector/tr_no_dev_phn/spk_xvector.ark:14227
20
+ jvs021 dump/xvector/tr_no_dev_phn/spk_xvector.ark:15017
21
+ jvs022 dump/xvector/tr_no_dev_phn/spk_xvector.ark:15807
22
+ jvs023 dump/xvector/tr_no_dev_phn/spk_xvector.ark:16597
23
+ jvs024 dump/xvector/tr_no_dev_phn/spk_xvector.ark:17387
24
+ jvs025 dump/xvector/tr_no_dev_phn/spk_xvector.ark:18177
25
+ jvs026 dump/xvector/tr_no_dev_phn/spk_xvector.ark:18967
26
+ jvs027 dump/xvector/tr_no_dev_phn/spk_xvector.ark:19757
27
+ jvs029 dump/xvector/tr_no_dev_phn/spk_xvector.ark:20547
28
+ jvs030 dump/xvector/tr_no_dev_phn/spk_xvector.ark:21337
29
+ jvs031 dump/xvector/tr_no_dev_phn/spk_xvector.ark:22127
30
+ jvs032 dump/xvector/tr_no_dev_phn/spk_xvector.ark:22917
31
+ jvs033 dump/xvector/tr_no_dev_phn/spk_xvector.ark:23707
32
+ jvs034 dump/xvector/tr_no_dev_phn/spk_xvector.ark:24497
33
+ jvs035 dump/xvector/tr_no_dev_phn/spk_xvector.ark:25287
34
+ jvs036 dump/xvector/tr_no_dev_phn/spk_xvector.ark:26077
35
+ jvs037 dump/xvector/tr_no_dev_phn/spk_xvector.ark:26867
36
+ jvs038 dump/xvector/tr_no_dev_phn/spk_xvector.ark:27657
37
+ jvs039 dump/xvector/tr_no_dev_phn/spk_xvector.ark:28447
38
+ jvs040 dump/xvector/tr_no_dev_phn/spk_xvector.ark:29237
39
+ jvs041 dump/xvector/tr_no_dev_phn/spk_xvector.ark:30027
40
+ jvs042 dump/xvector/tr_no_dev_phn/spk_xvector.ark:30817
41
+ jvs043 dump/xvector/tr_no_dev_phn/spk_xvector.ark:31607
42
+ jvs044 dump/xvector/tr_no_dev_phn/spk_xvector.ark:32397
43
+ jvs045 dump/xvector/tr_no_dev_phn/spk_xvector.ark:33187
44
+ jvs046 dump/xvector/tr_no_dev_phn/spk_xvector.ark:33977
45
+ jvs047 dump/xvector/tr_no_dev_phn/spk_xvector.ark:34767
46
+ jvs048 dump/xvector/tr_no_dev_phn/spk_xvector.ark:35557
47
+ jvs049 dump/xvector/tr_no_dev_phn/spk_xvector.ark:36347
48
+ jvs050 dump/xvector/tr_no_dev_phn/spk_xvector.ark:37137
49
+ jvs051 dump/xvector/tr_no_dev_phn/spk_xvector.ark:37927
50
+ jvs052 dump/xvector/tr_no_dev_phn/spk_xvector.ark:38717
51
+ jvs053 dump/xvector/tr_no_dev_phn/spk_xvector.ark:39507
52
+ jvs054 dump/xvector/tr_no_dev_phn/spk_xvector.ark:40297
53
+ jvs055 dump/xvector/tr_no_dev_phn/spk_xvector.ark:41087
54
+ jvs056 dump/xvector/tr_no_dev_phn/spk_xvector.ark:41877
55
+ jvs057 dump/xvector/tr_no_dev_phn/spk_xvector.ark:42667
56
+ jvs058 dump/xvector/tr_no_dev_phn/spk_xvector.ark:43457
57
+ jvs059 dump/xvector/tr_no_dev_phn/spk_xvector.ark:44247
58
+ jvs060 dump/xvector/tr_no_dev_phn/spk_xvector.ark:45037
59
+ jvs061 dump/xvector/tr_no_dev_phn/spk_xvector.ark:45827
60
+ jvs062 dump/xvector/tr_no_dev_phn/spk_xvector.ark:46617
61
+ jvs063 dump/xvector/tr_no_dev_phn/spk_xvector.ark:47407
62
+ jvs064 dump/xvector/tr_no_dev_phn/spk_xvector.ark:48197
63
+ jvs065 dump/xvector/tr_no_dev_phn/spk_xvector.ark:48987
64
+ jvs066 dump/xvector/tr_no_dev_phn/spk_xvector.ark:49777
65
+ jvs067 dump/xvector/tr_no_dev_phn/spk_xvector.ark:50567
66
+ jvs068 dump/xvector/tr_no_dev_phn/spk_xvector.ark:51357
67
+ jvs069 dump/xvector/tr_no_dev_phn/spk_xvector.ark:52147
68
+ jvs070 dump/xvector/tr_no_dev_phn/spk_xvector.ark:52937
69
+ jvs071 dump/xvector/tr_no_dev_phn/spk_xvector.ark:53727
70
+ jvs072 dump/xvector/tr_no_dev_phn/spk_xvector.ark:54517
71
+ jvs073 dump/xvector/tr_no_dev_phn/spk_xvector.ark:55307
72
+ jvs074 dump/xvector/tr_no_dev_phn/spk_xvector.ark:56097
73
+ jvs075 dump/xvector/tr_no_dev_phn/spk_xvector.ark:56887
74
+ jvs076 dump/xvector/tr_no_dev_phn/spk_xvector.ark:57677
75
+ jvs077 dump/xvector/tr_no_dev_phn/spk_xvector.ark:58467
76
+ jvs078 dump/xvector/tr_no_dev_phn/spk_xvector.ark:59257
77
+ jvs079 dump/xvector/tr_no_dev_phn/spk_xvector.ark:60047
78
+ jvs080 dump/xvector/tr_no_dev_phn/spk_xvector.ark:60837
79
+ jvs081 dump/xvector/tr_no_dev_phn/spk_xvector.ark:61627
80
+ jvs082 dump/xvector/tr_no_dev_phn/spk_xvector.ark:62417
81
+ jvs083 dump/xvector/tr_no_dev_phn/spk_xvector.ark:63207
82
+ jvs084 dump/xvector/tr_no_dev_phn/spk_xvector.ark:63997
83
+ jvs085 dump/xvector/tr_no_dev_phn/spk_xvector.ark:64787
84
+ jvs086 dump/xvector/tr_no_dev_phn/spk_xvector.ark:65577
85
+ jvs087 dump/xvector/tr_no_dev_phn/spk_xvector.ark:66367
86
+ jvs088 dump/xvector/tr_no_dev_phn/spk_xvector.ark:67157
87
+ jvs089 dump/xvector/tr_no_dev_phn/spk_xvector.ark:67947
88
+ jvs090 dump/xvector/tr_no_dev_phn/spk_xvector.ark:68737
89
+ jvs091 dump/xvector/tr_no_dev_phn/spk_xvector.ark:69527
90
+ jvs092 dump/xvector/tr_no_dev_phn/spk_xvector.ark:70317
91
+ jvs093 dump/xvector/tr_no_dev_phn/spk_xvector.ark:71107
92
+ jvs094 dump/xvector/tr_no_dev_phn/spk_xvector.ark:71897
93
+ jvs095 dump/xvector/tr_no_dev_phn/spk_xvector.ark:72687
94
+ jvs096 dump/xvector/tr_no_dev_phn/spk_xvector.ark:73477
95
+ jvs097 dump/xvector/tr_no_dev_phn/spk_xvector.ark:74267
96
+ jvs098 dump/xvector/tr_no_dev_phn/spk_xvector.ark:75057
97
+ jvs099 dump/xvector/tr_no_dev_phn/spk_xvector.ark:75847
98
+ jvs100 dump/xvector/tr_no_dev_phn/spk_xvector.ark:76637
99
+ tsukuyomi dump/xvector/tr_no_dev_phn/spk_xvector.ark:77430
exp/tts_prodiff_gst_xvector_base_raw_phn_none/config.yaml ADDED
@@ -0,0 +1,274 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ config: conf/tuning/prodiff_gst_xvector_base.yaml
2
+ print_config: false
3
+ log_level: INFO
4
+ dry_run: false
5
+ iterator_type: sequence
6
+ output_dir: exp/tts_prodiff_gst_xvector_base_raw_phn_none
7
+ ngpu: 1
8
+ seed: 0
9
+ num_workers: 6
10
+ num_att_plot: 3
11
+ num_valid_artifacts: 5
12
+ dist_backend: nccl
13
+ dist_init_method: env://
14
+ dist_world_size: 2
15
+ dist_rank: 0
16
+ local_rank: 0
17
+ dist_master_addr: localhost
18
+ dist_master_port: 32945
19
+ dist_launcher: null
20
+ multiprocessing_distributed: true
21
+ unused_parameters: false
22
+ sharded_ddp: false
23
+ growth_interval: 0
24
+ min_grad_scale: -1
25
+ cudnn_enabled: true
26
+ cudnn_benchmark: false
27
+ cudnn_deterministic: true
28
+ collect_stats: false
29
+ write_collected_feats: false
30
+ max_epoch: 600
31
+ patience: null
32
+ val_scheduler_criterion:
33
+ - valid
34
+ - loss
35
+ early_stopping_criterion:
36
+ - valid
37
+ - loss
38
+ - min
39
+ best_model_criterion:
40
+ - - valid
41
+ - loss
42
+ - min
43
+ - - train
44
+ - loss
45
+ - min
46
+ keep_nbest_models: 5
47
+ nbest_averaging_interval: 0
48
+ grad_clip: 1.0
49
+ grad_clip_type: 2.0
50
+ grad_noise: false
51
+ accum_grad: 1
52
+ no_forward_run: false
53
+ resume: true
54
+ train_dtype: float32
55
+ use_amp: false
56
+ log_interval: null
57
+ use_matplotlib: true
58
+ use_tensorboard: true
59
+ detect_anomaly: false
60
+ pretrain_path: null
61
+ init_param: []
62
+ ignore_init_mismatch: false
63
+ freeze_param: []
64
+ num_iters_per_epoch: 250
65
+ batch_size: 20
66
+ valid_batch_size: null
67
+ valid_num_batches: null
68
+ batch_bins: 6000000
69
+ valid_batch_bins: null
70
+ train_shape_file:
71
+ - exp/tts_stats_raw_phn_none/train/text_shape.phn
72
+ - exp/tts_stats_raw_phn_none/train/speech_shape
73
+ valid_shape_file:
74
+ - exp/tts_stats_raw_phn_none/valid/text_shape.phn
75
+ - exp/tts_stats_raw_phn_none/valid/speech_shape
76
+ batch_type: numel
77
+ valid_batch_type: null
78
+ fold_length:
79
+ - 150
80
+ - 240000
81
+ sort_in_batch: descending
82
+ sort_batch: descending
83
+ multiple_iterator: false
84
+ chunk_length: 500
85
+ chunk_shift_ratio: 0.5
86
+ num_cache_chunks: 1024
87
+ train_data_path_and_name_and_type:
88
+ - - dump/raw/tr_no_dev_phn/text
89
+ - text
90
+ - text
91
+ - - data/tr_no_dev_phn/durations
92
+ - durations
93
+ - text_int
94
+ - - dump/raw/tr_no_dev_phn/wav.scp
95
+ - speech
96
+ - sound
97
+ - - exp/tts_stats_raw_phn_none/train/collect_feats/pitch.scp
98
+ - pitch
99
+ - npy
100
+ - - exp/tts_stats_raw_phn_none/train/collect_feats/energy.scp
101
+ - energy
102
+ - npy
103
+ - - dump/xvector/tr_no_dev_phn/xvector.scp
104
+ - spembs
105
+ - kaldi_ark
106
+ valid_data_path_and_name_and_type:
107
+ - - dump/raw/dev_phn/text
108
+ - text
109
+ - text
110
+ - - data/dev_phn/durations
111
+ - durations
112
+ - text_int
113
+ - - dump/raw/dev_phn/wav.scp
114
+ - speech
115
+ - sound
116
+ - - exp/tts_stats_raw_phn_none/valid/collect_feats/pitch.scp
117
+ - pitch
118
+ - npy
119
+ - - exp/tts_stats_raw_phn_none/valid/collect_feats/energy.scp
120
+ - energy
121
+ - npy
122
+ - - dump/xvector/dev_phn/xvector.scp
123
+ - spembs
124
+ - kaldi_ark
125
+ allow_variable_data_keys: false
126
+ max_cache_size: 0.0
127
+ max_cache_fd: 32
128
+ valid_max_cache_size: null
129
+ optim: adamw
130
+ optim_conf:
131
+ lr: 1.0
132
+ betas:
133
+ - 0.9
134
+ - 0.98
135
+ scheduler: noamlr
136
+ scheduler_conf:
137
+ model_size: 384
138
+ warmup_steps: 2000
139
+ token_list:
140
+ - <blank>
141
+ - <unk>
142
+ - o
143
+ - a
144
+ - u
145
+ - i
146
+ - e
147
+ - k
148
+ - r
149
+ - t
150
+ - n
151
+ - 、
152
+ - N
153
+ - s
154
+ - sh
155
+ - d
156
+ - m
157
+ - g
158
+ - b
159
+ - w
160
+ - cl
161
+ - j
162
+ - ch
163
+ - sil
164
+ - h
165
+ - y
166
+ - p
167
+ - ts
168
+ - z
169
+ - f
170
+ - ky
171
+ - U
172
+ - ny
173
+ - gy
174
+ - ry
175
+ - I
176
+ - hy
177
+ - my
178
+ - by
179
+ - py
180
+ - v
181
+ - <sos/eos>
182
+ odim: null
183
+ model_conf:
184
+ requires_word_duration: false
185
+ use_preprocessor: true
186
+ token_type: phn
187
+ bpemodel: null
188
+ non_linguistic_symbols: null
189
+ cleaner: null
190
+ g2p: null
191
+ feats_extract: fbank
192
+ feats_extract_conf:
193
+ n_fft: 2048
194
+ hop_length: 300
195
+ win_length: 1200
196
+ fs: 24000
197
+ fmin: 80
198
+ fmax: 7600
199
+ n_mels: 80
200
+ normalize: global_mvn
201
+ normalize_conf:
202
+ stats_file: exp/tts_stats_raw_phn_none/train/feats_stats.npz
203
+ tts: prodiff
204
+ tts_conf:
205
+ adim: 384
206
+ aheads: 2
207
+ elayers: 4
208
+ eunits: 1536
209
+ positionwise_layer_type: conv1d-linear
210
+ positionwise_conv_kernel_size: 9
211
+ use_masking: true
212
+ use_scaled_pos_enc: true
213
+ encoder_normalize_before: true
214
+ reduction_factor: 1
215
+ init_type: xavier_uniform
216
+ init_enc_alpha: 1.0
217
+ transformer_enc_dropout_rate: 0.05
218
+ transformer_enc_positional_dropout_rate: 0.05
219
+ transformer_enc_attn_dropout_rate: 0.05
220
+ duration_predictor_layers: 2
221
+ duration_predictor_chans: 512
222
+ duration_predictor_kernel_size: 3
223
+ pitch_predictor_layers: 2
224
+ pitch_predictor_chans: 512
225
+ pitch_predictor_kernel_size: 3
226
+ pitch_predictor_dropout: 0.5
227
+ pitch_embed_kernel_size: 1
228
+ pitch_embed_dropout: 0.0
229
+ stop_gradient_from_pitch_predictor: true
230
+ energy_predictor_layers: 2
231
+ energy_predictor_chans: 512
232
+ energy_predictor_kernel_size: 3
233
+ energy_predictor_dropout: 0.5
234
+ energy_embed_kernel_size: 1
235
+ energy_embed_dropout: 0.0
236
+ stop_gradient_from_energy_predictor: false
237
+ spks: -1
238
+ spk_embed_dim: 192
239
+ denoiser_layers: 20
240
+ denoiser_channels: 256
241
+ diffusion_steps: 4
242
+ diffusion_timescale: 1
243
+ diffusion_beta: 40.0
244
+ diffusion_scheduler: vpsde
245
+ diffusion_cycle_ln: 1
246
+ use_gst: true
247
+ gst_heads: 8
248
+ gst_tokens: 256
249
+ pitch_extract: dio
250
+ pitch_extract_conf:
251
+ fs: 24000
252
+ n_fft: 2048
253
+ hop_length: 300
254
+ f0max: 400
255
+ f0min: 80
256
+ reduction_factor: 1
257
+ pitch_normalize: global_mvn
258
+ pitch_normalize_conf:
259
+ stats_file: exp/tts_stats_raw_phn_none/train/pitch_stats.npz
260
+ energy_extract: energy
261
+ energy_extract_conf:
262
+ fs: 24000
263
+ n_fft: 2048
264
+ hop_length: 300
265
+ win_length: 1200
266
+ reduction_factor: 1
267
+ energy_normalize: global_mvn
268
+ energy_normalize_conf:
269
+ stats_file: exp/tts_stats_raw_phn_none/train/energy_stats.npz
270
+ required:
271
+ - output_dir
272
+ - token_list
273
+ version: '202207'
274
+ distributed: true
exp/tts_prodiff_gst_xvector_base_raw_phn_none/train.loss.ave_5best.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:d5a5d356a632ac39f6c952a733149878dbd56133557ed7dd26c3c8b12eb1bcb5
3
+ size 189585833
exp/tts_stats_raw_phn_none/train/energy_stats.npz ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:db0b080a4e6ad27382c91e47ba42d12b6d891aaa1fcab74235b51afd98889f73
3
+ size 770
exp/tts_stats_raw_phn_none/train/feats_stats.npz ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:233ee6b1bf4140bcd26dd84a2c7c1183705ffd9713395e49c7542e9294835648
3
+ size 1402
exp/tts_stats_raw_phn_none/train/pitch_stats.npz ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:03a17c6d25426835cc1a99e3b3e7dfc71fc2d594b0d6c5634f36ff4a0a718fa2
3
+ size 770
meta.yaml ADDED
@@ -0,0 +1,8 @@
 
 
 
 
 
 
 
 
 
1
+ espnet: '202207'
2
+ files:
3
+ model_file: exp/tts_prodiff_gst_xvector_base_raw_phn_none/train.loss.ave_5best.pth
4
+ python: "3.9.12 (main, Jun 1 2022, 11:38:51) \n[GCC 7.5.0]"
5
+ timestamp: 1673972233.479128
6
+ torch: 1.11.0
7
+ yaml_files:
8
+ train_config: exp/tts_prodiff_gst_xvector_base_raw_phn_none/config.yaml