Hugging Face's logo

mimba
/
matcha-plt-m2-scratch

Model card Files
xet
 Community
mimba commited on
Commit
ad010d4
·
verified ·
1 Parent(s): 5a041fb

epoch 023 | val_loss=2.1854

Browse files
Files changed (6) hide show
  1. .gitattributes +2 -0
  2. checkpoints/epoch_023.ckpt +3 -0
  3. inference/epoch_023/metrics.json +328 -0
  4. inference/epoch_023/t1_spk_m2.wav +3 -0
  5. inference/epoch_023/t2_spk_m2.wav +3 -0
  6. metrics.json +330 -2
.gitattributes CHANGED
@@ -77,3 +77,5 @@ inference/epoch_021/t1_spk_m2.wav filter=lfs diff=lfs merge=lfs -text
77
  inference/epoch_021/t2_spk_m2.wav filter=lfs diff=lfs merge=lfs -text
78
  inference/epoch_022/t1_spk_m2.wav filter=lfs diff=lfs merge=lfs -text
79
  inference/epoch_022/t2_spk_m2.wav filter=lfs diff=lfs merge=lfs -text
 
 
 
77
  inference/epoch_021/t2_spk_m2.wav filter=lfs diff=lfs merge=lfs -text
78
  inference/epoch_022/t1_spk_m2.wav filter=lfs diff=lfs merge=lfs -text
79
  inference/epoch_022/t2_spk_m2.wav filter=lfs diff=lfs merge=lfs -text
80
+ inference/epoch_023/t1_spk_m2.wav filter=lfs diff=lfs merge=lfs -text
81
+ inference/epoch_023/t2_spk_m2.wav filter=lfs diff=lfs merge=lfs -text
checkpoints/epoch_023.ckpt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:77e97f5b90ab8618679ce48d521aa151efad35fe57c26d612f84cec42c8ce41b
3
+ size 218850584
inference/epoch_023/metrics.json ADDED
@@ -0,0 +1,328 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "step": 24885.0,
3
+ "sub_loss/train_dur_loss": 0.3276459276676178,
4
+ "sub_loss/train_dur_loss_step": 0.34054887294769287,
5
+ "sub_loss/train_prior_loss": 0.9636818766593933,
6
+ "sub_loss/train_prior_loss_step": 0.9643590450286865,
7
+ "sub_loss/train_diff_loss": 0.8940833210945129,
8
+ "sub_loss/train_diff_loss_step": 0.9285227656364441,
9
+ "loss/train": 2.185410737991333,
10
+ "loss/train_step": 2.2334306240081787,
11
+ "grad_norm/grad_2.0_norm/encoder.emb.weight": 0.010179021395742893,
12
+ "grad_norm/grad_2.0_norm/encoder.prenet.conv_layers.0.weight": 0.00047572783660143614,
13
+ "grad_norm/grad_2.0_norm/encoder.prenet.conv_layers.0.bias": 1.6653610146022402e-05,
14
+ "grad_norm/grad_2.0_norm/encoder.prenet.conv_layers.1.weight": 0.0005618097493425012,
15
+ "grad_norm/grad_2.0_norm/encoder.prenet.conv_layers.1.bias": 1.8629147234605625e-05,
16
+ "grad_norm/grad_2.0_norm/encoder.prenet.conv_layers.2.weight": 0.0007703870069235563,
17
+ "grad_norm/grad_2.0_norm/encoder.prenet.conv_layers.2.bias": 2.832677637343295e-05,
18
+ "grad_norm/grad_2.0_norm/encoder.prenet.norm_layers.0.gamma": 2.503781252016779e-05,
19
+ "grad_norm/grad_2.0_norm/encoder.prenet.norm_layers.0.beta": 2.5653840566519648e-05,
20
+ "grad_norm/grad_2.0_norm/encoder.prenet.norm_layers.1.gamma": 2.4070699510048144e-05,
21
+ "grad_norm/grad_2.0_norm/encoder.prenet.norm_layers.1.beta": 3.3490050554974005e-05,
22
+ "grad_norm/grad_2.0_norm/encoder.prenet.norm_layers.2.gamma": 5.208807851886377e-05,
23
+ "grad_norm/grad_2.0_norm/encoder.prenet.norm_layers.2.beta": 0.00017012649914249778,
24
+ "grad_norm/grad_2.0_norm/encoder.prenet.proj.weight": 0.0024400739930570126,
25
+ "grad_norm/grad_2.0_norm/encoder.prenet.proj.bias": 0.0008538054535165429,
26
+ "grad_norm/grad_2.0_norm/encoder.encoder.attn_layers.0.conv_q.weight": 0.0020137473475188017,
27
+ "grad_norm/grad_2.0_norm/encoder.encoder.attn_layers.0.conv_q.bias": 0.00017171133367810398,
28
+ "grad_norm/grad_2.0_norm/encoder.encoder.attn_layers.0.conv_k.weight": 0.0028118372429162264,
29
+ "grad_norm/grad_2.0_norm/encoder.encoder.attn_layers.0.conv_k.bias": 0.00010522449883865193,
30
+ "grad_norm/grad_2.0_norm/encoder.encoder.attn_layers.0.conv_v.weight": 0.002319223713129759,
31
+ "grad_norm/grad_2.0_norm/encoder.encoder.attn_layers.0.conv_v.bias": 0.00043287305743433535,
32
+ "grad_norm/grad_2.0_norm/encoder.encoder.attn_layers.0.conv_o.weight": 0.002975519048050046,
33
+ "grad_norm/grad_2.0_norm/encoder.encoder.attn_layers.0.conv_o.bias": 0.0007131283637136221,
34
+ "grad_norm/grad_2.0_norm/encoder.encoder.attn_layers.1.conv_q.weight": 0.002366220112890005,
35
+ "grad_norm/grad_2.0_norm/encoder.encoder.attn_layers.1.conv_q.bias": 0.00017606209439691156,
36
+ "grad_norm/grad_2.0_norm/encoder.encoder.attn_layers.1.conv_k.weight": 0.0025677415542304516,
37
+ "grad_norm/grad_2.0_norm/encoder.encoder.attn_layers.1.conv_k.bias": 0.00014030555030331016,
38
+ "grad_norm/grad_2.0_norm/encoder.encoder.attn_layers.1.conv_v.weight": 0.0030527263879776,
39
+ "grad_norm/grad_2.0_norm/encoder.encoder.attn_layers.1.conv_v.bias": 0.00041086156852543354,
40
+ "grad_norm/grad_2.0_norm/encoder.encoder.attn_layers.1.conv_o.weight": 0.003840845078229904,
41
+ "grad_norm/grad_2.0_norm/encoder.encoder.attn_layers.1.conv_o.bias": 0.0006546062068082392,
42
+ "grad_norm/grad_2.0_norm/encoder.encoder.attn_layers.2.conv_q.weight": 0.002656548051163554,
43
+ "grad_norm/grad_2.0_norm/encoder.encoder.attn_layers.2.conv_q.bias": 0.0002256667212350294,
44
+ "grad_norm/grad_2.0_norm/encoder.encoder.attn_layers.2.conv_k.weight": 0.0027526067569851875,
45
+ "grad_norm/grad_2.0_norm/encoder.encoder.attn_layers.2.conv_k.bias": 0.0001322456810157746,
46
+ "grad_norm/grad_2.0_norm/encoder.encoder.attn_layers.2.conv_v.weight": 0.004119835328310728,
47
+ "grad_norm/grad_2.0_norm/encoder.encoder.attn_layers.2.conv_v.bias": 0.00043101023766212165,
48
+ "grad_norm/grad_2.0_norm/encoder.encoder.attn_layers.2.conv_o.weight": 0.003995231352746487,
49
+ "grad_norm/grad_2.0_norm/encoder.encoder.attn_layers.2.conv_o.bias": 0.0006926574860699475,
50
+ "grad_norm/grad_2.0_norm/encoder.encoder.attn_layers.3.conv_q.weight": 0.0038935565389692783,
51
+ "grad_norm/grad_2.0_norm/encoder.encoder.attn_layers.3.conv_q.bias": 0.0002771899744402617,
52
+ "grad_norm/grad_2.0_norm/encoder.encoder.attn_layers.3.conv_k.weight": 0.005292035173624754,
53
+ "grad_norm/grad_2.0_norm/encoder.encoder.attn_layers.3.conv_k.bias": 0.00021608290262520313,
54
+ "grad_norm/grad_2.0_norm/encoder.encoder.attn_layers.3.conv_v.weight": 0.003980815410614014,
55
+ "grad_norm/grad_2.0_norm/encoder.encoder.attn_layers.3.conv_v.bias": 0.00042934296652674675,
56
+ "grad_norm/grad_2.0_norm/encoder.encoder.attn_layers.3.conv_o.weight": 0.003352918429300189,
57
+ "grad_norm/grad_2.0_norm/encoder.encoder.attn_layers.3.conv_o.bias": 0.0007153137703426182,
58
+ "grad_norm/grad_2.0_norm/encoder.encoder.attn_layers.4.conv_q.weight": 0.004747373051941395,
59
+ "grad_norm/grad_2.0_norm/encoder.encoder.attn_layers.4.conv_q.bias": 0.00028694901266135275,
60
+ "grad_norm/grad_2.0_norm/encoder.encoder.attn_layers.4.conv_k.weight": 0.005936896428465843,
61
+ "grad_norm/grad_2.0_norm/encoder.encoder.attn_layers.4.conv_k.bias": 0.0002524383016861975,
62
+ "grad_norm/grad_2.0_norm/encoder.encoder.attn_layers.4.conv_v.weight": 0.004644670989364386,
63
+ "grad_norm/grad_2.0_norm/encoder.encoder.attn_layers.4.conv_v.bias": 0.00049459288129583,
64
+ "grad_norm/grad_2.0_norm/encoder.encoder.attn_layers.4.conv_o.weight": 0.003199503291398287,
65
+ "grad_norm/grad_2.0_norm/encoder.encoder.attn_layers.4.conv_o.bias": 0.0007712735095992684,
66
+ "grad_norm/grad_2.0_norm/encoder.encoder.attn_layers.5.conv_q.weight": 0.0032443455420434475,
67
+ "grad_norm/grad_2.0_norm/encoder.encoder.attn_layers.5.conv_q.bias": 0.0002529523626435548,
68
+ "grad_norm/grad_2.0_norm/encoder.encoder.attn_layers.5.conv_k.weight": 0.0036140424199402332,
69
+ "grad_norm/grad_2.0_norm/encoder.encoder.attn_layers.5.conv_k.bias": 5.029261956224218e-05,
70
+ "grad_norm/grad_2.0_norm/encoder.encoder.attn_layers.5.conv_v.weight": 0.0076249465346336365,
71
+ "grad_norm/grad_2.0_norm/encoder.encoder.attn_layers.5.conv_v.bias": 0.0006648091948591173,
72
+ "grad_norm/grad_2.0_norm/encoder.encoder.attn_layers.5.conv_o.weight": 0.004593569785356522,
73
+ "grad_norm/grad_2.0_norm/encoder.encoder.attn_layers.5.conv_o.bias": 0.0011054789647459984,
74
+ "grad_norm/grad_2.0_norm/encoder.encoder.norm_layers_1.0.gamma": 0.0007502355147153139,
75
+ "grad_norm/grad_2.0_norm/encoder.encoder.norm_layers_1.0.beta": 0.0007287278422154486,
76
+ "grad_norm/grad_2.0_norm/encoder.encoder.norm_layers_1.1.gamma": 0.0007848882814869285,
77
+ "grad_norm/grad_2.0_norm/encoder.encoder.norm_layers_1.1.beta": 0.0006988961831666529,
78
+ "grad_norm/grad_2.0_norm/encoder.encoder.norm_layers_1.2.gamma": 0.0008045095601119101,
79
+ "grad_norm/grad_2.0_norm/encoder.encoder.norm_layers_1.2.beta": 0.0007149989251047373,
80
+ "grad_norm/grad_2.0_norm/encoder.encoder.norm_layers_1.3.gamma": 0.0008514974033460021,
81
+ "grad_norm/grad_2.0_norm/encoder.encoder.norm_layers_1.3.beta": 0.0007861618069000542,
82
+ "grad_norm/grad_2.0_norm/encoder.encoder.norm_layers_1.4.gamma": 0.0008179837604984641,
83
+ "grad_norm/grad_2.0_norm/encoder.encoder.norm_layers_1.4.beta": 0.0009000417194329202,
84
+ "grad_norm/grad_2.0_norm/encoder.encoder.norm_layers_1.5.gamma": 0.0011050268076360226,
85
+ "grad_norm/grad_2.0_norm/encoder.encoder.norm_layers_1.5.beta": 0.0014269533567130566,
86
+ "grad_norm/grad_2.0_norm/encoder.encoder.ffn_layers.0.conv_1.weight": 0.006145291030406952,
87
+ "grad_norm/grad_2.0_norm/encoder.encoder.ffn_layers.0.conv_1.bias": 0.0002593389945104718,
88
+ "grad_norm/grad_2.0_norm/encoder.encoder.ffn_layers.0.conv_2.weight": 0.008098908700048923,
89
+ "grad_norm/grad_2.0_norm/encoder.encoder.ffn_layers.0.conv_2.bias": 0.0006639388157054782,
90
+ "grad_norm/grad_2.0_norm/encoder.encoder.ffn_layers.1.conv_1.weight": 0.007951445877552032,
91
+ "grad_norm/grad_2.0_norm/encoder.encoder.ffn_layers.1.conv_1.bias": 0.0003412013757042587,
92
+ "grad_norm/grad_2.0_norm/encoder.encoder.ffn_layers.1.conv_2.weight": 0.00750324223190546,
93
+ "grad_norm/grad_2.0_norm/encoder.encoder.ffn_layers.1.conv_2.bias": 0.0006179813644848764,
94
+ "grad_norm/grad_2.0_norm/encoder.encoder.ffn_layers.2.conv_1.weight": 0.008253040723502636,
95
+ "grad_norm/grad_2.0_norm/encoder.encoder.ffn_layers.2.conv_1.bias": 0.0003635578032117337,
96
+ "grad_norm/grad_2.0_norm/encoder.encoder.ffn_layers.2.conv_2.weight": 0.008296418003737926,
97
+ "grad_norm/grad_2.0_norm/encoder.encoder.ffn_layers.2.conv_2.bias": 0.0006220432696864009,
98
+ "grad_norm/grad_2.0_norm/encoder.encoder.ffn_layers.3.conv_1.weight": 0.008848525583744049,
99
+ "grad_norm/grad_2.0_norm/encoder.encoder.ffn_layers.3.conv_1.bias": 0.00038747471990063787,
100
+ "grad_norm/grad_2.0_norm/encoder.encoder.ffn_layers.3.conv_2.weight": 0.007914482615888119,
101
+ "grad_norm/grad_2.0_norm/encoder.encoder.ffn_layers.3.conv_2.bias": 0.0006922960164956748,
102
+ "grad_norm/grad_2.0_norm/encoder.encoder.ffn_layers.4.conv_1.weight": 0.01127774827182293,
103
+ "grad_norm/grad_2.0_norm/encoder.encoder.ffn_layers.4.conv_1.bias": 0.00047135885688476264,
104
+ "grad_norm/grad_2.0_norm/encoder.encoder.ffn_layers.4.conv_2.weight": 0.006439025979489088,
105
+ "grad_norm/grad_2.0_norm/encoder.encoder.ffn_layers.4.conv_2.bias": 0.0008436174248345196,
106
+ "grad_norm/grad_2.0_norm/encoder.encoder.ffn_layers.5.conv_1.weight": 0.023061564192175865,
107
+ "grad_norm/grad_2.0_norm/encoder.encoder.ffn_layers.5.conv_1.bias": 0.0008471981273032725,
108
+ "grad_norm/grad_2.0_norm/encoder.encoder.ffn_layers.5.conv_2.weight": 0.0068195536732673645,
109
+ "grad_norm/grad_2.0_norm/encoder.encoder.ffn_layers.5.conv_2.bias": 0.0015057931886985898,
110
+ "grad_norm/grad_2.0_norm/encoder.encoder.norm_layers_2.0.gamma": 0.0007468069088645279,
111
+ "grad_norm/grad_2.0_norm/encoder.encoder.norm_layers_2.0.beta": 0.0007797700236551464,
112
+ "grad_norm/grad_2.0_norm/encoder.encoder.norm_layers_2.1.gamma": 0.0008247159421443939,
113
+ "grad_norm/grad_2.0_norm/encoder.encoder.norm_layers_2.1.beta": 0.0007752308738417923,
114
+ "grad_norm/grad_2.0_norm/encoder.encoder.norm_layers_2.2.gamma": 0.0009003481245599687,
115
+ "grad_norm/grad_2.0_norm/encoder.encoder.norm_layers_2.2.beta": 0.0008182940073311329,
116
+ "grad_norm/grad_2.0_norm/encoder.encoder.norm_layers_2.3.gamma": 0.0008685484062880278,
117
+ "grad_norm/grad_2.0_norm/encoder.encoder.norm_layers_2.3.beta": 0.0009516056743450463,
118
+ "grad_norm/grad_2.0_norm/encoder.encoder.norm_layers_2.4.gamma": 0.0011581492144614458,
119
+ "grad_norm/grad_2.0_norm/encoder.encoder.norm_layers_2.4.beta": 0.0011406283592805266,
120
+ "grad_norm/grad_2.0_norm/encoder.encoder.norm_layers_2.5.gamma": 0.00243854196742177,
121
+ "grad_norm/grad_2.0_norm/encoder.encoder.norm_layers_2.5.beta": 0.0023418134078383446,
122
+ "grad_norm/grad_2.0_norm/encoder.proj_m.weight": 0.09363279491662979,
123
+ "grad_norm/grad_2.0_norm/encoder.proj_m.bias": 0.008155187591910362,
124
+ "grad_norm/grad_2.0_norm/encoder.proj_w.conv_1.weight": 0.14825886487960815,
125
+ "grad_norm/grad_2.0_norm/encoder.proj_w.conv_1.bias": 0.006943495012819767,
126
+ "grad_norm/grad_2.0_norm/encoder.proj_w.norm_1.gamma": 0.002364248037338257,
127
+ "grad_norm/grad_2.0_norm/encoder.proj_w.norm_1.beta": 0.0026497221551835537,
128
+ "grad_norm/grad_2.0_norm/encoder.proj_w.conv_2.weight": 0.07042104005813599,
129
+ "grad_norm/grad_2.0_norm/encoder.proj_w.conv_2.bias": 0.0023474604822695255,
130
+ "grad_norm/grad_2.0_norm/encoder.proj_w.norm_2.gamma": 0.0011766505194827914,
131
+ "grad_norm/grad_2.0_norm/encoder.proj_w.norm_2.beta": 0.004002234432846308,
132
+ "grad_norm/grad_2.0_norm/encoder.proj_w.proj.weight": 0.17083770036697388,
133
+ "grad_norm/grad_2.0_norm/encoder.proj_w.proj.bias": 0.009552001953125,
134
+ "grad_norm/grad_2.0_norm/decoder.estimator.time_mlp.linear_1.weight": 0.0033950291108340025,
135
+ "grad_norm/grad_2.0_norm/decoder.estimator.time_mlp.linear_1.bias": 0.0004075357865076512,
136
+ "grad_norm/grad_2.0_norm/decoder.estimator.time_mlp.linear_2.weight": 0.006900752894580364,
137
+ "grad_norm/grad_2.0_norm/decoder.estimator.time_mlp.linear_2.bias": 0.000763674033805728,
138
+ "grad_norm/grad_2.0_norm/decoder.estimator.down_blocks.0.0.mlp.1.weight": 0.01261177472770214,
139
+ "grad_norm/grad_2.0_norm/decoder.estimator.down_blocks.0.0.mlp.1.bias": 0.0013403829652816057,
140
+ "grad_norm/grad_2.0_norm/decoder.estimator.down_blocks.0.0.block1.block.0.weight": 0.01363199483603239,
141
+ "grad_norm/grad_2.0_norm/decoder.estimator.down_blocks.0.0.block1.block.0.bias": 0.0006541350740008056,
142
+ "grad_norm/grad_2.0_norm/decoder.estimator.down_blocks.0.0.block1.block.1.weight": 0.0013086594408378005,
143
+ "grad_norm/grad_2.0_norm/decoder.estimator.down_blocks.0.0.block1.block.1.bias": 0.0010349287185817957,
144
+ "grad_norm/grad_2.0_norm/decoder.estimator.down_blocks.0.0.block2.block.0.weight": 0.03309207409620285,
145
+ "grad_norm/grad_2.0_norm/decoder.estimator.down_blocks.0.0.block2.block.0.bias": 0.0010355969425290823,
146
+ "grad_norm/grad_2.0_norm/decoder.estimator.down_blocks.0.0.block2.block.1.weight": 0.00514656538143754,
147
+ "grad_norm/grad_2.0_norm/decoder.estimator.down_blocks.0.0.block2.block.1.bias": 0.00493122823536396,
148
+ "grad_norm/grad_2.0_norm/decoder.estimator.down_blocks.0.0.res_conv.weight": 0.11264612525701523,
149
+ "grad_norm/grad_2.0_norm/decoder.estimator.down_blocks.0.0.res_conv.bias": 0.8733417987823486,
150
+ "grad_norm/grad_2.0_norm/decoder.estimator.down_blocks.0.1.0.norm1.weight": 0.004071284085512161,
151
+ "grad_norm/grad_2.0_norm/decoder.estimator.down_blocks.0.1.0.norm1.bias": 0.004936905112117529,
152
+ "grad_norm/grad_2.0_norm/decoder.estimator.down_blocks.0.1.0.attn1.to_q.weight": 0.004860785789787769,
153
+ "grad_norm/grad_2.0_norm/decoder.estimator.down_blocks.0.1.0.attn1.to_k.weight": 0.002252493752166629,
154
+ "grad_norm/grad_2.0_norm/decoder.estimator.down_blocks.0.1.0.attn1.to_v.weight": 0.05356993526220322,
155
+ "grad_norm/grad_2.0_norm/decoder.estimator.down_blocks.0.1.0.attn1.to_out.0.weight": 0.002795935608446598,
156
+ "grad_norm/grad_2.0_norm/decoder.estimator.down_blocks.0.1.0.attn1.to_out.0.bias": 0.009306102991104126,
157
+ "grad_norm/grad_2.0_norm/decoder.estimator.down_blocks.0.1.0.norm3.weight": 0.004220674745738506,
158
+ "grad_norm/grad_2.0_norm/decoder.estimator.down_blocks.0.1.0.norm3.bias": 0.006370976101607084,
159
+ "grad_norm/grad_2.0_norm/decoder.estimator.down_blocks.0.1.0.ff.net.0.alpha": 0.0003204054373782128,
160
+ "grad_norm/grad_2.0_norm/decoder.estimator.down_blocks.0.1.0.ff.net.0.beta": 0.0002711393462959677,
161
+ "grad_norm/grad_2.0_norm/decoder.estimator.down_blocks.0.1.0.ff.net.0.proj.weight": 0.03016696125268936,
162
+ "grad_norm/grad_2.0_norm/decoder.estimator.down_blocks.0.1.0.ff.net.0.proj.bias": 0.0021761737298220396,
163
+ "grad_norm/grad_2.0_norm/decoder.estimator.down_blocks.0.1.0.ff.net.2.weight": 0.022060168907046318,
164
+ "grad_norm/grad_2.0_norm/decoder.estimator.down_blocks.0.1.0.ff.net.2.bias": 0.003218117170035839,
165
+ "grad_norm/grad_2.0_norm/decoder.estimator.down_blocks.0.2.conv.weight": 0.035109687596559525,
166
+ "grad_norm/grad_2.0_norm/decoder.estimator.down_blocks.0.2.conv.bias": 0.0010453894501551986,
167
+ "grad_norm/grad_2.0_norm/decoder.estimator.down_blocks.1.0.mlp.1.weight": 0.00982450507581234,
168
+ "grad_norm/grad_2.0_norm/decoder.estimator.down_blocks.1.0.mlp.1.bias": 0.0009652215521782637,
169
+ "grad_norm/grad_2.0_norm/decoder.estimator.down_blocks.1.0.block1.block.0.weight": 0.007580431178212166,
170
+ "grad_norm/grad_2.0_norm/decoder.estimator.down_blocks.1.0.block1.block.0.bias": 0.0001582414988661185,
171
+ "grad_norm/grad_2.0_norm/decoder.estimator.down_blocks.1.0.block1.block.1.weight": 0.0007583931437693536,
172
+ "grad_norm/grad_2.0_norm/decoder.estimator.down_blocks.1.0.block1.block.1.bias": 0.0007107752608135343,
173
+ "grad_norm/grad_2.0_norm/decoder.estimator.down_blocks.1.0.block2.block.0.weight": 0.0070860073901712894,
174
+ "grad_norm/grad_2.0_norm/decoder.estimator.down_blocks.1.0.block2.block.0.bias": 0.0005084277945570648,
175
+ "grad_norm/grad_2.0_norm/decoder.estimator.down_blocks.1.0.block2.block.1.weight": 0.0006098756566643715,
176
+ "grad_norm/grad_2.0_norm/decoder.estimator.down_blocks.1.0.block2.block.1.bias": 0.0005726104718632996,
177
+ "grad_norm/grad_2.0_norm/decoder.estimator.down_blocks.1.0.res_conv.weight": 0.024373618885874748,
178
+ "grad_norm/grad_2.0_norm/decoder.estimator.down_blocks.1.0.res_conv.bias": 0.09879472851753235,
179
+ "grad_norm/grad_2.0_norm/decoder.estimator.down_blocks.1.1.0.norm1.weight": 0.0007012482965365052,
180
+ "grad_norm/grad_2.0_norm/decoder.estimator.down_blocks.1.1.0.norm1.bias": 0.00143358763307333,
181
+ "grad_norm/grad_2.0_norm/decoder.estimator.down_blocks.1.1.0.attn1.to_q.weight": 0.0019879851024597883,
182
+ "grad_norm/grad_2.0_norm/decoder.estimator.down_blocks.1.1.0.attn1.to_k.weight": 0.002155584516003728,
183
+ "grad_norm/grad_2.0_norm/decoder.estimator.down_blocks.1.1.0.attn1.to_v.weight": 0.008918555453419685,
184
+ "grad_norm/grad_2.0_norm/decoder.estimator.down_blocks.1.1.0.attn1.to_out.0.weight": 0.0011629601940512657,
185
+ "grad_norm/grad_2.0_norm/decoder.estimator.down_blocks.1.1.0.attn1.to_out.0.bias": 0.0008767198887653649,
186
+ "grad_norm/grad_2.0_norm/decoder.estimator.down_blocks.1.1.0.norm3.weight": 0.0010608871234580874,
187
+ "grad_norm/grad_2.0_norm/decoder.estimator.down_blocks.1.1.0.norm3.bias": 0.0012027847114950418,
188
+ "grad_norm/grad_2.0_norm/decoder.estimator.down_blocks.1.1.0.ff.net.0.alpha": 0.0004097867349628359,
189
+ "grad_norm/grad_2.0_norm/decoder.estimator.down_blocks.1.1.0.ff.net.0.beta": 0.00040464234189130366,
190
+ "grad_norm/grad_2.0_norm/decoder.estimator.down_blocks.1.1.0.ff.net.0.proj.weight": 0.009615248069167137,
191
+ "grad_norm/grad_2.0_norm/decoder.estimator.down_blocks.1.1.0.ff.net.0.proj.bias": 0.0007450803532265127,
192
+ "grad_norm/grad_2.0_norm/decoder.estimator.down_blocks.1.1.0.ff.net.2.weight": 0.009156358428299427,
193
+ "grad_norm/grad_2.0_norm/decoder.estimator.down_blocks.1.1.0.ff.net.2.bias": 0.0008174045942723751,
194
+ "grad_norm/grad_2.0_norm/decoder.estimator.down_blocks.1.2.weight": 0.03646900877356529,
195
+ "grad_norm/grad_2.0_norm/decoder.estimator.down_blocks.1.2.bias": 0.0004777516005560756,
196
+ "grad_norm/grad_2.0_norm/decoder.estimator.mid_blocks.0.0.mlp.1.weight": 0.007459959480911493,
197
+ "grad_norm/grad_2.0_norm/decoder.estimator.mid_blocks.0.0.mlp.1.bias": 0.000627974106464535,
198
+ "grad_norm/grad_2.0_norm/decoder.estimator.mid_blocks.0.0.block1.block.0.weight": 0.006041461601853371,
199
+ "grad_norm/grad_2.0_norm/decoder.estimator.mid_blocks.0.0.block1.block.0.bias": 5.324760059011169e-05,
200
+ "grad_norm/grad_2.0_norm/decoder.estimator.mid_blocks.0.0.block1.block.1.weight": 0.0005003157421015203,
201
+ "grad_norm/grad_2.0_norm/decoder.estimator.mid_blocks.0.0.block1.block.1.bias": 0.00042986817425116897,
202
+ "grad_norm/grad_2.0_norm/decoder.estimator.mid_blocks.0.0.block2.block.0.weight": 0.006311975885182619,
203
+ "grad_norm/grad_2.0_norm/decoder.estimator.mid_blocks.0.0.block2.block.0.bias": 0.000369485147530213,
204
+ "grad_norm/grad_2.0_norm/decoder.estimator.mid_blocks.0.0.block2.block.1.weight": 0.00034039438469335437,
205
+ "grad_norm/grad_2.0_norm/decoder.estimator.mid_blocks.0.0.block2.block.1.bias": 0.0003005445178132504,
206
+ "grad_norm/grad_2.0_norm/decoder.estimator.mid_blocks.0.0.res_conv.weight": 0.02558288909494877,
207
+ "grad_norm/grad_2.0_norm/decoder.estimator.mid_blocks.0.0.res_conv.bias": 0.0469089038670063,
208
+ "grad_norm/grad_2.0_norm/decoder.estimator.mid_blocks.0.1.0.norm1.weight": 0.0006509429658763111,
209
+ "grad_norm/grad_2.0_norm/decoder.estimator.mid_blocks.0.1.0.norm1.bias": 0.0008774683228693902,
210
+ "grad_norm/grad_2.0_norm/decoder.estimator.mid_blocks.0.1.0.attn1.to_q.weight": 0.0028540780767798424,
211
+ "grad_norm/grad_2.0_norm/decoder.estimator.mid_blocks.0.1.0.attn1.to_k.weight": 0.0016212090849876404,
212
+ "grad_norm/grad_2.0_norm/decoder.estimator.mid_blocks.0.1.0.attn1.to_v.weight": 0.00638106744736433,
213
+ "grad_norm/grad_2.0_norm/decoder.estimator.mid_blocks.0.1.0.attn1.to_out.0.weight": 0.0010032415157184005,
214
+ "grad_norm/grad_2.0_norm/decoder.estimator.mid_blocks.0.1.0.attn1.to_out.0.bias": 0.00048305917880497873,
215
+ "grad_norm/grad_2.0_norm/decoder.estimator.mid_blocks.0.1.0.norm3.weight": 0.0007569156587123871,
216
+ "grad_norm/grad_2.0_norm/decoder.estimator.mid_blocks.0.1.0.norm3.bias": 0.0008760644705034792,
217
+ "grad_norm/grad_2.0_norm/decoder.estimator.mid_blocks.0.1.0.ff.net.0.alpha": 0.0002850043820217252,
218
+ "grad_norm/grad_2.0_norm/decoder.estimator.mid_blocks.0.1.0.ff.net.0.beta": 0.00030988469370640814,
219
+ "grad_norm/grad_2.0_norm/decoder.estimator.mid_blocks.0.1.0.ff.net.0.proj.weight": 0.007048940751701593,
220
+ "grad_norm/grad_2.0_norm/decoder.estimator.mid_blocks.0.1.0.ff.net.0.proj.bias": 0.0005304201622493565,
221
+ "grad_norm/grad_2.0_norm/decoder.estimator.mid_blocks.0.1.0.ff.net.2.weight": 0.008041159249842167,
222
+ "grad_norm/grad_2.0_norm/decoder.estimator.mid_blocks.0.1.0.ff.net.2.bias": 0.0004873759171459824,
223
+ "grad_norm/grad_2.0_norm/decoder.estimator.mid_blocks.1.0.mlp.1.weight": 0.006650703493505716,
224
+ "grad_norm/grad_2.0_norm/decoder.estimator.mid_blocks.1.0.mlp.1.bias": 0.0006425073370337486,
225
+ "grad_norm/grad_2.0_norm/decoder.estimator.mid_blocks.1.0.block1.block.0.weight": 0.006960422731935978,
226
+ "grad_norm/grad_2.0_norm/decoder.estimator.mid_blocks.1.0.block1.block.0.bias": 4.733038440463133e-05,
227
+ "grad_norm/grad_2.0_norm/decoder.estimator.mid_blocks.1.0.block1.block.1.weight": 0.0004921250510960817,
228
+ "grad_norm/grad_2.0_norm/decoder.estimator.mid_blocks.1.0.block1.block.1.bias": 0.0004338321159593761,
229
+ "grad_norm/grad_2.0_norm/decoder.estimator.mid_blocks.1.0.block2.block.0.weight": 0.005244262516498566,
230
+ "grad_norm/grad_2.0_norm/decoder.estimator.mid_blocks.1.0.block2.block.0.bias": 0.0003615597961470485,
231
+ "grad_norm/grad_2.0_norm/decoder.estimator.mid_blocks.1.0.block2.block.1.weight": 0.0002653438423294574,
232
+ "grad_norm/grad_2.0_norm/decoder.estimator.mid_blocks.1.0.block2.block.1.bias": 0.0002546966716181487,
233
+ "grad_norm/grad_2.0_norm/decoder.estimator.mid_blocks.1.0.res_conv.weight": 0.03545884042978287,
234
+ "grad_norm/grad_2.0_norm/decoder.estimator.mid_blocks.1.0.res_conv.bias": 0.031797852367162704,
235
+ "grad_norm/grad_2.0_norm/decoder.estimator.mid_blocks.1.1.0.norm1.weight": 0.0006199093186296523,
236
+ "grad_norm/grad_2.0_norm/decoder.estimator.mid_blocks.1.1.0.norm1.bias": 0.0008437394862994552,
237
+ "grad_norm/grad_2.0_norm/decoder.estimator.mid_blocks.1.1.0.attn1.to_q.weight": 0.0020717547740787268,
238
+ "grad_norm/grad_2.0_norm/decoder.estimator.mid_blocks.1.1.0.attn1.to_k.weight": 0.0020844489336013794,
239
+ "grad_norm/grad_2.0_norm/decoder.estimator.mid_blocks.1.1.0.attn1.to_v.weight": 0.006132850889116526,
240
+ "grad_norm/grad_2.0_norm/decoder.estimator.mid_blocks.1.1.0.attn1.to_out.0.weight": 0.0007109093712642789,
241
+ "grad_norm/grad_2.0_norm/decoder.estimator.mid_blocks.1.1.0.attn1.to_out.0.bias": 0.00044272647937759757,
242
+ "grad_norm/grad_2.0_norm/decoder.estimator.mid_blocks.1.1.0.norm3.weight": 0.0007580879027955234,
243
+ "grad_norm/grad_2.0_norm/decoder.estimator.mid_blocks.1.1.0.norm3.bias": 0.0008920146501623094,
244
+ "grad_norm/grad_2.0_norm/decoder.estimator.mid_blocks.1.1.0.ff.net.0.alpha": 0.00031814113026484847,
245
+ "grad_norm/grad_2.0_norm/decoder.estimator.mid_blocks.1.1.0.ff.net.0.beta": 0.0003361067792866379,
246
+ "grad_norm/grad_2.0_norm/decoder.estimator.mid_blocks.1.1.0.ff.net.0.proj.weight": 0.007014248985797167,
247
+ "grad_norm/grad_2.0_norm/decoder.estimator.mid_blocks.1.1.0.ff.net.0.proj.bias": 0.000548130483366549,
248
+ "grad_norm/grad_2.0_norm/decoder.estimator.mid_blocks.1.1.0.ff.net.2.weight": 0.00817958079278469,
249
+ "grad_norm/grad_2.0_norm/decoder.estimator.mid_blocks.1.1.0.ff.net.2.bias": 0.00045939270057715476,
250
+ "grad_norm/grad_2.0_norm/decoder.estimator.up_blocks.0.0.mlp.1.weight": 0.010318427346646786,
251
+ "grad_norm/grad_2.0_norm/decoder.estimator.up_blocks.0.0.mlp.1.bias": 0.0010112684685736895,
252
+ "grad_norm/grad_2.0_norm/decoder.estimator.up_blocks.0.0.block1.block.0.weight": 0.014999396167695522,
253
+ "grad_norm/grad_2.0_norm/decoder.estimator.up_blocks.0.0.block1.block.0.bias": 8.258102752733976e-05,
254
+ "grad_norm/grad_2.0_norm/decoder.estimator.up_blocks.0.0.block1.block.1.weight": 0.0007028156542219222,
255
+ "grad_norm/grad_2.0_norm/decoder.estimator.up_blocks.0.0.block1.block.1.bias": 0.000596357393078506,
256
+ "grad_norm/grad_2.0_norm/decoder.estimator.up_blocks.0.0.block2.block.0.weight": 0.0061602480709552765,
257
+ "grad_norm/grad_2.0_norm/decoder.estimator.up_blocks.0.0.block2.block.0.bias": 0.0005165530019439757,
258
+ "grad_norm/grad_2.0_norm/decoder.estimator.up_blocks.0.0.block2.block.1.weight": 0.00038624528679065406,
259
+ "grad_norm/grad_2.0_norm/decoder.estimator.up_blocks.0.0.block2.block.1.bias": 0.00032755982829257846,
260
+ "grad_norm/grad_2.0_norm/decoder.estimator.up_blocks.0.0.res_conv.weight": 0.05575540289282799,
261
+ "grad_norm/grad_2.0_norm/decoder.estimator.up_blocks.0.0.res_conv.bias": 0.08856959640979767,
262
+ "grad_norm/grad_2.0_norm/decoder.estimator.up_blocks.0.1.0.norm1.weight": 0.000856694474350661,
263
+ "grad_norm/grad_2.0_norm/decoder.estimator.up_blocks.0.1.0.norm1.bias": 0.001748208305798471,
264
+ "grad_norm/grad_2.0_norm/decoder.estimator.up_blocks.0.1.0.attn1.to_q.weight": 0.0028679671231657267,
265
+ "grad_norm/grad_2.0_norm/decoder.estimator.up_blocks.0.1.0.attn1.to_k.weight": 0.0022758597042411566,
266
+ "grad_norm/grad_2.0_norm/decoder.estimator.up_blocks.0.1.0.attn1.to_v.weight": 0.010127125307917595,
267
+ "grad_norm/grad_2.0_norm/decoder.estimator.up_blocks.0.1.0.attn1.to_out.0.weight": 0.0009156019659712911,
268
+ "grad_norm/grad_2.0_norm/decoder.estimator.up_blocks.0.1.0.attn1.to_out.0.bias": 0.0006679215002804995,
269
+ "grad_norm/grad_2.0_norm/decoder.estimator.up_blocks.0.1.0.norm3.weight": 0.0010506961261853576,
270
+ "grad_norm/grad_2.0_norm/decoder.estimator.up_blocks.0.1.0.norm3.bias": 0.00132385129109025,
271
+ "grad_norm/grad_2.0_norm/decoder.estimator.up_blocks.0.1.0.ff.net.0.alpha": 0.0004299134307075292,
272
+ "grad_norm/grad_2.0_norm/decoder.estimator.up_blocks.0.1.0.ff.net.0.beta": 0.00042983077582903206,
273
+ "grad_norm/grad_2.0_norm/decoder.estimator.up_blocks.0.1.0.ff.net.0.proj.weight": 0.00906638614833355,
274
+ "grad_norm/grad_2.0_norm/decoder.estimator.up_blocks.0.1.0.ff.net.0.proj.bias": 0.0007534258766099811,
275
+ "grad_norm/grad_2.0_norm/decoder.estimator.up_blocks.0.1.0.ff.net.2.weight": 0.008831251412630081,
276
+ "grad_norm/grad_2.0_norm/decoder.estimator.up_blocks.0.1.0.ff.net.2.bias": 0.0007010450935922563,
277
+ "grad_norm/grad_2.0_norm/decoder.estimator.up_blocks.0.2.conv.weight": 0.09383069723844528,
278
+ "grad_norm/grad_2.0_norm/decoder.estimator.up_blocks.0.2.conv.bias": 0.0016627543373033404,
279
+ "grad_norm/grad_2.0_norm/decoder.estimator.up_blocks.1.0.mlp.1.weight": 0.028144413605332375,
280
+ "grad_norm/grad_2.0_norm/decoder.estimator.up_blocks.1.0.mlp.1.bias": 0.003071507904678583,
281
+ "grad_norm/grad_2.0_norm/decoder.estimator.up_blocks.1.0.block1.block.0.weight": 0.030205750837922096,
282
+ "grad_norm/grad_2.0_norm/decoder.estimator.up_blocks.1.0.block1.block.0.bias": 0.0005511984345503151,
283
+ "grad_norm/grad_2.0_norm/decoder.estimator.up_blocks.1.0.block1.block.1.weight": 0.002031352138146758,
284
+ "grad_norm/grad_2.0_norm/decoder.estimator.up_blocks.1.0.block1.block.1.bias": 0.0021938334684818983,
285
+ "grad_norm/grad_2.0_norm/decoder.estimator.up_blocks.1.0.block2.block.0.weight": 0.025340333580970764,
286
+ "grad_norm/grad_2.0_norm/decoder.estimator.up_blocks.1.0.block2.block.0.bias": 0.0018560552271082997,
287
+ "grad_norm/grad_2.0_norm/decoder.estimator.up_blocks.1.0.block2.block.1.weight": 0.002064154017716646,
288
+ "grad_norm/grad_2.0_norm/decoder.estimator.up_blocks.1.0.block2.block.1.bias": 0.001897836453281343,
289
+ "grad_norm/grad_2.0_norm/decoder.estimator.up_blocks.1.0.res_conv.weight": 0.09192795306444168,
290
+ "grad_norm/grad_2.0_norm/decoder.estimator.up_blocks.1.0.res_conv.bias": 0.025354549288749695,
291
+ "grad_norm/grad_2.0_norm/decoder.estimator.up_blocks.1.1.0.norm1.weight": 0.004005036316812038,
292
+ "grad_norm/grad_2.0_norm/decoder.estimator.up_blocks.1.1.0.norm1.bias": 0.0044862860813736916,
293
+ "grad_norm/grad_2.0_norm/decoder.estimator.up_blocks.1.1.0.attn1.to_q.weight": 0.003407246433198452,
294
+ "grad_norm/grad_2.0_norm/decoder.estimator.up_blocks.1.1.0.attn1.to_k.weight": 0.001814698800444603,
295
+ "grad_norm/grad_2.0_norm/decoder.estimator.up_blocks.1.1.0.attn1.to_v.weight": 0.04642708599567413,
296
+ "grad_norm/grad_2.0_norm/decoder.estimator.up_blocks.1.1.0.attn1.to_out.0.weight": 0.001497728400863707,
297
+ "grad_norm/grad_2.0_norm/decoder.estimator.up_blocks.1.1.0.attn1.to_out.0.bias": 0.00255810865201056,
298
+ "grad_norm/grad_2.0_norm/decoder.estimator.up_blocks.1.1.0.norm3.weight": 0.0026136802043765783,
299
+ "grad_norm/grad_2.0_norm/decoder.estimator.up_blocks.1.1.0.norm3.bias": 0.003123651025816798,
300
+ "grad_norm/grad_2.0_norm/decoder.estimator.up_blocks.1.1.0.ff.net.0.alpha": 0.0011864164844155312,
301
+ "grad_norm/grad_2.0_norm/decoder.estimator.up_blocks.1.1.0.ff.net.0.beta": 0.0009428896591998637,
302
+ "grad_norm/grad_2.0_norm/decoder.estimator.up_blocks.1.1.0.ff.net.0.proj.weight": 0.023969616740942,
303
+ "grad_norm/grad_2.0_norm/decoder.estimator.up_blocks.1.1.0.ff.net.0.proj.bias": 0.0018177471356466413,
304
+ "grad_norm/grad_2.0_norm/decoder.estimator.up_blocks.1.1.0.ff.net.2.weight": 0.014123436063528061,
305
+ "grad_norm/grad_2.0_norm/decoder.estimator.up_blocks.1.1.0.ff.net.2.bias": 0.002645216416567564,
306
+ "grad_norm/grad_2.0_norm/decoder.estimator.up_blocks.1.2.weight": 0.18219496309757233,
307
+ "grad_norm/grad_2.0_norm/decoder.estimator.up_blocks.1.2.bias": 0.0013607360888272524,
308
+ "grad_norm/grad_2.0_norm/decoder.estimator.final_block.block.0.weight": 0.1666623204946518,
309
+ "grad_norm/grad_2.0_norm/decoder.estimator.final_block.block.0.bias": 0.0008949644980020821,
310
+ "grad_norm/grad_2.0_norm/decoder.estimator.final_block.block.1.weight": 0.01216113194823265,
311
+ "grad_norm/grad_2.0_norm/decoder.estimator.final_block.block.1.bias": 0.011881562881171703,
312
+ "grad_norm/grad_2.0_norm/decoder.estimator.final_proj.weight": 0.11852734535932541,
313
+ "grad_norm/grad_2.0_norm/decoder.estimator.final_proj.bias": 0.006633310113102198,
314
+ "grad_norm/grad_2.0_norm_total": 0.9888164401054382,
315
+ "sub_loss/val_dur_loss": 0.32394927740097046,
316
+ "sub_loss/val_dur_loss_epoch": 0.32394927740097046,
317
+ "sub_loss/val_prior_loss": 0.96378493309021,
318
+ "sub_loss/val_prior_loss_epoch": 0.96378493309021,
319
+ "sub_loss/val_diff_loss": 0.8976922035217285,
320
+ "sub_loss/val_diff_loss_epoch": 0.8976922035217285,
321
+ "loss/val": 2.1854262351989746,
322
+ "loss/val_epoch": 2.1854262351989746,
323
+ "sub_loss/train_dur_loss_epoch": 0.3276459276676178,
324
+ "sub_loss/train_prior_loss_epoch": 0.9636818766593933,
325
+ "sub_loss/train_diff_loss_epoch": 0.8940833210945129,
326
+ "loss/train_epoch": 2.185410737991333,
327
+ "epoch": 23
328
+ }
inference/epoch_023/t1_spk_m2.wav ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:901f1351954ab754601c812c06c42bcd59cf1f53d82bae3f5c8275ee97f57ac9
3
+ size 176172
inference/epoch_023/t2_spk_m2.wav ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:f19c6286c1c4eb586e3994b5fab984d907e9032dbe778dc358b8b53348c0df59
3
+ size 142380
metrics.json CHANGED
@@ -1,6 +1,6 @@
1
  {
2
- "last_epoch": 22,
3
- "last_val_loss": 2.185152530670166,
4
  "history": [
5
  {
6
  "step": 1081.0,
@@ -7217,6 +7217,334 @@
7217
  "sub_loss/train_diff_loss_epoch": 0.8967382311820984,
7218
  "loss/train_epoch": 2.1882729530334473,
7219
  "epoch": 22
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
7220
  }
7221
  ]
7222
  }
 
1
  {
2
+ "last_epoch": 23,
3
+ "last_val_loss": 2.1854262351989746,
4
  "history": [
5
  {
6
  "step": 1081.0,
 
7217
  "sub_loss/train_diff_loss_epoch": 0.8967382311820984,
7218
  "loss/train_epoch": 2.1882729530334473,
7219
  "epoch": 22
7220
+ },
7221
+ {
7222
+ "step": 24885.0,
7223
+ "sub_loss/train_dur_loss": 0.3276459276676178,
7224
+ "sub_loss/train_dur_loss_step": 0.34054887294769287,
7225
+ "sub_loss/train_prior_loss": 0.9636818766593933,
7226
+ "sub_loss/train_prior_loss_step": 0.9643590450286865,
7227
+ "sub_loss/train_diff_loss": 0.8940833210945129,
7228
+ "sub_loss/train_diff_loss_step": 0.9285227656364441,
7229
+ "loss/train": 2.185410737991333,
7230
+ "loss/train_step": 2.2334306240081787,
7231
+ "grad_norm/grad_2.0_norm/encoder.emb.weight": 0.010179021395742893,
7232
+ "grad_norm/grad_2.0_norm/encoder.prenet.conv_layers.0.weight": 0.00047572783660143614,
7233
+ "grad_norm/grad_2.0_norm/encoder.prenet.conv_layers.0.bias": 1.6653610146022402e-05,
7234
+ "grad_norm/grad_2.0_norm/encoder.prenet.conv_layers.1.weight": 0.0005618097493425012,
7235
+ "grad_norm/grad_2.0_norm/encoder.prenet.conv_layers.1.bias": 1.8629147234605625e-05,
7236
+ "grad_norm/grad_2.0_norm/encoder.prenet.conv_layers.2.weight": 0.0007703870069235563,
7237
+ "grad_norm/grad_2.0_norm/encoder.prenet.conv_layers.2.bias": 2.832677637343295e-05,
7238
+ "grad_norm/grad_2.0_norm/encoder.prenet.norm_layers.0.gamma": 2.503781252016779e-05,
7239
+ "grad_norm/grad_2.0_norm/encoder.prenet.norm_layers.0.beta": 2.5653840566519648e-05,
7240
+ "grad_norm/grad_2.0_norm/encoder.prenet.norm_layers.1.gamma": 2.4070699510048144e-05,
7241
+ "grad_norm/grad_2.0_norm/encoder.prenet.norm_layers.1.beta": 3.3490050554974005e-05,
7242
+ "grad_norm/grad_2.0_norm/encoder.prenet.norm_layers.2.gamma": 5.208807851886377e-05,
7243
+ "grad_norm/grad_2.0_norm/encoder.prenet.norm_layers.2.beta": 0.00017012649914249778,
7244
+ "grad_norm/grad_2.0_norm/encoder.prenet.proj.weight": 0.0024400739930570126,
7245
+ "grad_norm/grad_2.0_norm/encoder.prenet.proj.bias": 0.0008538054535165429,
7246
+ "grad_norm/grad_2.0_norm/encoder.encoder.attn_layers.0.conv_q.weight": 0.0020137473475188017,
7247
+ "grad_norm/grad_2.0_norm/encoder.encoder.attn_layers.0.conv_q.bias": 0.00017171133367810398,
7248
+ "grad_norm/grad_2.0_norm/encoder.encoder.attn_layers.0.conv_k.weight": 0.0028118372429162264,
7249
+ "grad_norm/grad_2.0_norm/encoder.encoder.attn_layers.0.conv_k.bias": 0.00010522449883865193,
7250
+ "grad_norm/grad_2.0_norm/encoder.encoder.attn_layers.0.conv_v.weight": 0.002319223713129759,
7251
+ "grad_norm/grad_2.0_norm/encoder.encoder.attn_layers.0.conv_v.bias": 0.00043287305743433535,
7252
+ "grad_norm/grad_2.0_norm/encoder.encoder.attn_layers.0.conv_o.weight": 0.002975519048050046,
7253
+ "grad_norm/grad_2.0_norm/encoder.encoder.attn_layers.0.conv_o.bias": 0.0007131283637136221,
7254
+ "grad_norm/grad_2.0_norm/encoder.encoder.attn_layers.1.conv_q.weight": 0.002366220112890005,
7255
+ "grad_norm/grad_2.0_norm/encoder.encoder.attn_layers.1.conv_q.bias": 0.00017606209439691156,
7256
+ "grad_norm/grad_2.0_norm/encoder.encoder.attn_layers.1.conv_k.weight": 0.0025677415542304516,
7257
+ "grad_norm/grad_2.0_norm/encoder.encoder.attn_layers.1.conv_k.bias": 0.00014030555030331016,
7258
+ "grad_norm/grad_2.0_norm/encoder.encoder.attn_layers.1.conv_v.weight": 0.0030527263879776,
7259
+ "grad_norm/grad_2.0_norm/encoder.encoder.attn_layers.1.conv_v.bias": 0.00041086156852543354,
7260
+ "grad_norm/grad_2.0_norm/encoder.encoder.attn_layers.1.conv_o.weight": 0.003840845078229904,
7261
+ "grad_norm/grad_2.0_norm/encoder.encoder.attn_layers.1.conv_o.bias": 0.0006546062068082392,
7262
+ "grad_norm/grad_2.0_norm/encoder.encoder.attn_layers.2.conv_q.weight": 0.002656548051163554,
7263
+ "grad_norm/grad_2.0_norm/encoder.encoder.attn_layers.2.conv_q.bias": 0.0002256667212350294,
7264
+ "grad_norm/grad_2.0_norm/encoder.encoder.attn_layers.2.conv_k.weight": 0.0027526067569851875,
7265
+ "grad_norm/grad_2.0_norm/encoder.encoder.attn_layers.2.conv_k.bias": 0.0001322456810157746,
7266
+ "grad_norm/grad_2.0_norm/encoder.encoder.attn_layers.2.conv_v.weight": 0.004119835328310728,
7267
+ "grad_norm/grad_2.0_norm/encoder.encoder.attn_layers.2.conv_v.bias": 0.00043101023766212165,
7268
+ "grad_norm/grad_2.0_norm/encoder.encoder.attn_layers.2.conv_o.weight": 0.003995231352746487,
7269
+ "grad_norm/grad_2.0_norm/encoder.encoder.attn_layers.2.conv_o.bias": 0.0006926574860699475,
7270
+ "grad_norm/grad_2.0_norm/encoder.encoder.attn_layers.3.conv_q.weight": 0.0038935565389692783,
7271
+ "grad_norm/grad_2.0_norm/encoder.encoder.attn_layers.3.conv_q.bias": 0.0002771899744402617,
7272
+ "grad_norm/grad_2.0_norm/encoder.encoder.attn_layers.3.conv_k.weight": 0.005292035173624754,
7273
+ "grad_norm/grad_2.0_norm/encoder.encoder.attn_layers.3.conv_k.bias": 0.00021608290262520313,
7274
+ "grad_norm/grad_2.0_norm/encoder.encoder.attn_layers.3.conv_v.weight": 0.003980815410614014,
7275
+ "grad_norm/grad_2.0_norm/encoder.encoder.attn_layers.3.conv_v.bias": 0.00042934296652674675,
7276
+ "grad_norm/grad_2.0_norm/encoder.encoder.attn_layers.3.conv_o.weight": 0.003352918429300189,
7277
+ "grad_norm/grad_2.0_norm/encoder.encoder.attn_layers.3.conv_o.bias": 0.0007153137703426182,
7278
+ "grad_norm/grad_2.0_norm/encoder.encoder.attn_layers.4.conv_q.weight": 0.004747373051941395,
7279
+ "grad_norm/grad_2.0_norm/encoder.encoder.attn_layers.4.conv_q.bias": 0.00028694901266135275,
7280
+ "grad_norm/grad_2.0_norm/encoder.encoder.attn_layers.4.conv_k.weight": 0.005936896428465843,
7281
+ "grad_norm/grad_2.0_norm/encoder.encoder.attn_layers.4.conv_k.bias": 0.0002524383016861975,
7282
+ "grad_norm/grad_2.0_norm/encoder.encoder.attn_layers.4.conv_v.weight": 0.004644670989364386,
7283
+ "grad_norm/grad_2.0_norm/encoder.encoder.attn_layers.4.conv_v.bias": 0.00049459288129583,
7284
+ "grad_norm/grad_2.0_norm/encoder.encoder.attn_layers.4.conv_o.weight": 0.003199503291398287,
7285
+ "grad_norm/grad_2.0_norm/encoder.encoder.attn_layers.4.conv_o.bias": 0.0007712735095992684,
7286
+ "grad_norm/grad_2.0_norm/encoder.encoder.attn_layers.5.conv_q.weight": 0.0032443455420434475,
7287
+ "grad_norm/grad_2.0_norm/encoder.encoder.attn_layers.5.conv_q.bias": 0.0002529523626435548,
7288
+ "grad_norm/grad_2.0_norm/encoder.encoder.attn_layers.5.conv_k.weight": 0.0036140424199402332,
7289
+ "grad_norm/grad_2.0_norm/encoder.encoder.attn_layers.5.conv_k.bias": 5.029261956224218e-05,
7290
+ "grad_norm/grad_2.0_norm/encoder.encoder.attn_layers.5.conv_v.weight": 0.0076249465346336365,
7291
+ "grad_norm/grad_2.0_norm/encoder.encoder.attn_layers.5.conv_v.bias": 0.0006648091948591173,
7292
+ "grad_norm/grad_2.0_norm/encoder.encoder.attn_layers.5.conv_o.weight": 0.004593569785356522,
7293
+ "grad_norm/grad_2.0_norm/encoder.encoder.attn_layers.5.conv_o.bias": 0.0011054789647459984,
7294
+ "grad_norm/grad_2.0_norm/encoder.encoder.norm_layers_1.0.gamma": 0.0007502355147153139,
7295
+ "grad_norm/grad_2.0_norm/encoder.encoder.norm_layers_1.0.beta": 0.0007287278422154486,
7296
+ "grad_norm/grad_2.0_norm/encoder.encoder.norm_layers_1.1.gamma": 0.0007848882814869285,
7297
+ "grad_norm/grad_2.0_norm/encoder.encoder.norm_layers_1.1.beta": 0.0006988961831666529,
7298
+ "grad_norm/grad_2.0_norm/encoder.encoder.norm_layers_1.2.gamma": 0.0008045095601119101,
7299
+ "grad_norm/grad_2.0_norm/encoder.encoder.norm_layers_1.2.beta": 0.0007149989251047373,
7300
+ "grad_norm/grad_2.0_norm/encoder.encoder.norm_layers_1.3.gamma": 0.0008514974033460021,
7301
+ "grad_norm/grad_2.0_norm/encoder.encoder.norm_layers_1.3.beta": 0.0007861618069000542,
7302
+ "grad_norm/grad_2.0_norm/encoder.encoder.norm_layers_1.4.gamma": 0.0008179837604984641,
7303
+ "grad_norm/grad_2.0_norm/encoder.encoder.norm_layers_1.4.beta": 0.0009000417194329202,
7304
+ "grad_norm/grad_2.0_norm/encoder.encoder.norm_layers_1.5.gamma": 0.0011050268076360226,
7305
+ "grad_norm/grad_2.0_norm/encoder.encoder.norm_layers_1.5.beta": 0.0014269533567130566,
7306
+ "grad_norm/grad_2.0_norm/encoder.encoder.ffn_layers.0.conv_1.weight": 0.006145291030406952,
7307
+ "grad_norm/grad_2.0_norm/encoder.encoder.ffn_layers.0.conv_1.bias": 0.0002593389945104718,
7308
+ "grad_norm/grad_2.0_norm/encoder.encoder.ffn_layers.0.conv_2.weight": 0.008098908700048923,
7309
+ "grad_norm/grad_2.0_norm/encoder.encoder.ffn_layers.0.conv_2.bias": 0.0006639388157054782,
7310
+ "grad_norm/grad_2.0_norm/encoder.encoder.ffn_layers.1.conv_1.weight": 0.007951445877552032,
7311
+ "grad_norm/grad_2.0_norm/encoder.encoder.ffn_layers.1.conv_1.bias": 0.0003412013757042587,
7312
+ "grad_norm/grad_2.0_norm/encoder.encoder.ffn_layers.1.conv_2.weight": 0.00750324223190546,
7313
+ "grad_norm/grad_2.0_norm/encoder.encoder.ffn_layers.1.conv_2.bias": 0.0006179813644848764,
7314
+ "grad_norm/grad_2.0_norm/encoder.encoder.ffn_layers.2.conv_1.weight": 0.008253040723502636,
7315
+ "grad_norm/grad_2.0_norm/encoder.encoder.ffn_layers.2.conv_1.bias": 0.0003635578032117337,
7316
+ "grad_norm/grad_2.0_norm/encoder.encoder.ffn_layers.2.conv_2.weight": 0.008296418003737926,
7317
+ "grad_norm/grad_2.0_norm/encoder.encoder.ffn_layers.2.conv_2.bias": 0.0006220432696864009,
7318
+ "grad_norm/grad_2.0_norm/encoder.encoder.ffn_layers.3.conv_1.weight": 0.008848525583744049,
7319
+ "grad_norm/grad_2.0_norm/encoder.encoder.ffn_layers.3.conv_1.bias": 0.00038747471990063787,
7320
+ "grad_norm/grad_2.0_norm/encoder.encoder.ffn_layers.3.conv_2.weight": 0.007914482615888119,
7321
+ "grad_norm/grad_2.0_norm/encoder.encoder.ffn_layers.3.conv_2.bias": 0.0006922960164956748,
7322
+ "grad_norm/grad_2.0_norm/encoder.encoder.ffn_layers.4.conv_1.weight": 0.01127774827182293,
7323
+ "grad_norm/grad_2.0_norm/encoder.encoder.ffn_layers.4.conv_1.bias": 0.00047135885688476264,
7324
+ "grad_norm/grad_2.0_norm/encoder.encoder.ffn_layers.4.conv_2.weight": 0.006439025979489088,
7325
+ "grad_norm/grad_2.0_norm/encoder.encoder.ffn_layers.4.conv_2.bias": 0.0008436174248345196,
7326
+ "grad_norm/grad_2.0_norm/encoder.encoder.ffn_layers.5.conv_1.weight": 0.023061564192175865,
7327
+ "grad_norm/grad_2.0_norm/encoder.encoder.ffn_layers.5.conv_1.bias": 0.0008471981273032725,
7328
+ "grad_norm/grad_2.0_norm/encoder.encoder.ffn_layers.5.conv_2.weight": 0.0068195536732673645,
7329
+ "grad_norm/grad_2.0_norm/encoder.encoder.ffn_layers.5.conv_2.bias": 0.0015057931886985898,
7330
+ "grad_norm/grad_2.0_norm/encoder.encoder.norm_layers_2.0.gamma": 0.0007468069088645279,
7331
+ "grad_norm/grad_2.0_norm/encoder.encoder.norm_layers_2.0.beta": 0.0007797700236551464,
7332
+ "grad_norm/grad_2.0_norm/encoder.encoder.norm_layers_2.1.gamma": 0.0008247159421443939,
7333
+ "grad_norm/grad_2.0_norm/encoder.encoder.norm_layers_2.1.beta": 0.0007752308738417923,
7334
+ "grad_norm/grad_2.0_norm/encoder.encoder.norm_layers_2.2.gamma": 0.0009003481245599687,
7335
+ "grad_norm/grad_2.0_norm/encoder.encoder.norm_layers_2.2.beta": 0.0008182940073311329,
7336
+ "grad_norm/grad_2.0_norm/encoder.encoder.norm_layers_2.3.gamma": 0.0008685484062880278,
7337
+ "grad_norm/grad_2.0_norm/encoder.encoder.norm_layers_2.3.beta": 0.0009516056743450463,
7338
+ "grad_norm/grad_2.0_norm/encoder.encoder.norm_layers_2.4.gamma": 0.0011581492144614458,
7339
+ "grad_norm/grad_2.0_norm/encoder.encoder.norm_layers_2.4.beta": 0.0011406283592805266,
7340
+ "grad_norm/grad_2.0_norm/encoder.encoder.norm_layers_2.5.gamma": 0.00243854196742177,
7341
+ "grad_norm/grad_2.0_norm/encoder.encoder.norm_layers_2.5.beta": 0.0023418134078383446,
7342
+ "grad_norm/grad_2.0_norm/encoder.proj_m.weight": 0.09363279491662979,
7343
+ "grad_norm/grad_2.0_norm/encoder.proj_m.bias": 0.008155187591910362,
7344
+ "grad_norm/grad_2.0_norm/encoder.proj_w.conv_1.weight": 0.14825886487960815,
7345
+ "grad_norm/grad_2.0_norm/encoder.proj_w.conv_1.bias": 0.006943495012819767,
7346
+ "grad_norm/grad_2.0_norm/encoder.proj_w.norm_1.gamma": 0.002364248037338257,
7347
+ "grad_norm/grad_2.0_norm/encoder.proj_w.norm_1.beta": 0.0026497221551835537,
7348
+ "grad_norm/grad_2.0_norm/encoder.proj_w.conv_2.weight": 0.07042104005813599,
7349
+ "grad_norm/grad_2.0_norm/encoder.proj_w.conv_2.bias": 0.0023474604822695255,
7350
+ "grad_norm/grad_2.0_norm/encoder.proj_w.norm_2.gamma": 0.0011766505194827914,
7351
+ "grad_norm/grad_2.0_norm/encoder.proj_w.norm_2.beta": 0.004002234432846308,
7352
+ "grad_norm/grad_2.0_norm/encoder.proj_w.proj.weight": 0.17083770036697388,
7353
+ "grad_norm/grad_2.0_norm/encoder.proj_w.proj.bias": 0.009552001953125,
7354
+ "grad_norm/grad_2.0_norm/decoder.estimator.time_mlp.linear_1.weight": 0.0033950291108340025,
7355
+ "grad_norm/grad_2.0_norm/decoder.estimator.time_mlp.linear_1.bias": 0.0004075357865076512,
7356
+ "grad_norm/grad_2.0_norm/decoder.estimator.time_mlp.linear_2.weight": 0.006900752894580364,
7357
+ "grad_norm/grad_2.0_norm/decoder.estimator.time_mlp.linear_2.bias": 0.000763674033805728,
7358
+ "grad_norm/grad_2.0_norm/decoder.estimator.down_blocks.0.0.mlp.1.weight": 0.01261177472770214,
7359
+ "grad_norm/grad_2.0_norm/decoder.estimator.down_blocks.0.0.mlp.1.bias": 0.0013403829652816057,
7360
+ "grad_norm/grad_2.0_norm/decoder.estimator.down_blocks.0.0.block1.block.0.weight": 0.01363199483603239,
7361
+ "grad_norm/grad_2.0_norm/decoder.estimator.down_blocks.0.0.block1.block.0.bias": 0.0006541350740008056,
7362
+ "grad_norm/grad_2.0_norm/decoder.estimator.down_blocks.0.0.block1.block.1.weight": 0.0013086594408378005,
7363
+ "grad_norm/grad_2.0_norm/decoder.estimator.down_blocks.0.0.block1.block.1.bias": 0.0010349287185817957,
7364
+ "grad_norm/grad_2.0_norm/decoder.estimator.down_blocks.0.0.block2.block.0.weight": 0.03309207409620285,
7365
+ "grad_norm/grad_2.0_norm/decoder.estimator.down_blocks.0.0.block2.block.0.bias": 0.0010355969425290823,
7366
+ "grad_norm/grad_2.0_norm/decoder.estimator.down_blocks.0.0.block2.block.1.weight": 0.00514656538143754,
7367
+ "grad_norm/grad_2.0_norm/decoder.estimator.down_blocks.0.0.block2.block.1.bias": 0.00493122823536396,
7368
+ "grad_norm/grad_2.0_norm/decoder.estimator.down_blocks.0.0.res_conv.weight": 0.11264612525701523,
7369
+ "grad_norm/grad_2.0_norm/decoder.estimator.down_blocks.0.0.res_conv.bias": 0.8733417987823486,
7370
+ "grad_norm/grad_2.0_norm/decoder.estimator.down_blocks.0.1.0.norm1.weight": 0.004071284085512161,
7371
+ "grad_norm/grad_2.0_norm/decoder.estimator.down_blocks.0.1.0.norm1.bias": 0.004936905112117529,
7372
+ "grad_norm/grad_2.0_norm/decoder.estimator.down_blocks.0.1.0.attn1.to_q.weight": 0.004860785789787769,
7373
+ "grad_norm/grad_2.0_norm/decoder.estimator.down_blocks.0.1.0.attn1.to_k.weight": 0.002252493752166629,
7374
+ "grad_norm/grad_2.0_norm/decoder.estimator.down_blocks.0.1.0.attn1.to_v.weight": 0.05356993526220322,
7375
+ "grad_norm/grad_2.0_norm/decoder.estimator.down_blocks.0.1.0.attn1.to_out.0.weight": 0.002795935608446598,
7376
+ "grad_norm/grad_2.0_norm/decoder.estimator.down_blocks.0.1.0.attn1.to_out.0.bias": 0.009306102991104126,
7377
+ "grad_norm/grad_2.0_norm/decoder.estimator.down_blocks.0.1.0.norm3.weight": 0.004220674745738506,
7378
+ "grad_norm/grad_2.0_norm/decoder.estimator.down_blocks.0.1.0.norm3.bias": 0.006370976101607084,
7379
+ "grad_norm/grad_2.0_norm/decoder.estimator.down_blocks.0.1.0.ff.net.0.alpha": 0.0003204054373782128,
7380
+ "grad_norm/grad_2.0_norm/decoder.estimator.down_blocks.0.1.0.ff.net.0.beta": 0.0002711393462959677,
7381
+ "grad_norm/grad_2.0_norm/decoder.estimator.down_blocks.0.1.0.ff.net.0.proj.weight": 0.03016696125268936,
7382
+ "grad_norm/grad_2.0_norm/decoder.estimator.down_blocks.0.1.0.ff.net.0.proj.bias": 0.0021761737298220396,
7383
+ "grad_norm/grad_2.0_norm/decoder.estimator.down_blocks.0.1.0.ff.net.2.weight": 0.022060168907046318,
7384
+ "grad_norm/grad_2.0_norm/decoder.estimator.down_blocks.0.1.0.ff.net.2.bias": 0.003218117170035839,
7385
+ "grad_norm/grad_2.0_norm/decoder.estimator.down_blocks.0.2.conv.weight": 0.035109687596559525,
7386
+ "grad_norm/grad_2.0_norm/decoder.estimator.down_blocks.0.2.conv.bias": 0.0010453894501551986,
7387
+ "grad_norm/grad_2.0_norm/decoder.estimator.down_blocks.1.0.mlp.1.weight": 0.00982450507581234,
7388
+ "grad_norm/grad_2.0_norm/decoder.estimator.down_blocks.1.0.mlp.1.bias": 0.0009652215521782637,
7389
+ "grad_norm/grad_2.0_norm/decoder.estimator.down_blocks.1.0.block1.block.0.weight": 0.007580431178212166,
7390
+ "grad_norm/grad_2.0_norm/decoder.estimator.down_blocks.1.0.block1.block.0.bias": 0.0001582414988661185,
7391
+ "grad_norm/grad_2.0_norm/decoder.estimator.down_blocks.1.0.block1.block.1.weight": 0.0007583931437693536,
7392
+ "grad_norm/grad_2.0_norm/decoder.estimator.down_blocks.1.0.block1.block.1.bias": 0.0007107752608135343,
7393
+ "grad_norm/grad_2.0_norm/decoder.estimator.down_blocks.1.0.block2.block.0.weight": 0.0070860073901712894,
7394
+ "grad_norm/grad_2.0_norm/decoder.estimator.down_blocks.1.0.block2.block.0.bias": 0.0005084277945570648,
7395
+ "grad_norm/grad_2.0_norm/decoder.estimator.down_blocks.1.0.block2.block.1.weight": 0.0006098756566643715,
7396
+ "grad_norm/grad_2.0_norm/decoder.estimator.down_blocks.1.0.block2.block.1.bias": 0.0005726104718632996,
7397
+ "grad_norm/grad_2.0_norm/decoder.estimator.down_blocks.1.0.res_conv.weight": 0.024373618885874748,
7398
+ "grad_norm/grad_2.0_norm/decoder.estimator.down_blocks.1.0.res_conv.bias": 0.09879472851753235,
7399
+ "grad_norm/grad_2.0_norm/decoder.estimator.down_blocks.1.1.0.norm1.weight": 0.0007012482965365052,
7400
+ "grad_norm/grad_2.0_norm/decoder.estimator.down_blocks.1.1.0.norm1.bias": 0.00143358763307333,
7401
+ "grad_norm/grad_2.0_norm/decoder.estimator.down_blocks.1.1.0.attn1.to_q.weight": 0.0019879851024597883,
7402
+ "grad_norm/grad_2.0_norm/decoder.estimator.down_blocks.1.1.0.attn1.to_k.weight": 0.002155584516003728,
7403
+ "grad_norm/grad_2.0_norm/decoder.estimator.down_blocks.1.1.0.attn1.to_v.weight": 0.008918555453419685,
7404
+ "grad_norm/grad_2.0_norm/decoder.estimator.down_blocks.1.1.0.attn1.to_out.0.weight": 0.0011629601940512657,
7405
+ "grad_norm/grad_2.0_norm/decoder.estimator.down_blocks.1.1.0.attn1.to_out.0.bias": 0.0008767198887653649,
7406
+ "grad_norm/grad_2.0_norm/decoder.estimator.down_blocks.1.1.0.norm3.weight": 0.0010608871234580874,
7407
+ "grad_norm/grad_2.0_norm/decoder.estimator.down_blocks.1.1.0.norm3.bias": 0.0012027847114950418,
7408
+ "grad_norm/grad_2.0_norm/decoder.estimator.down_blocks.1.1.0.ff.net.0.alpha": 0.0004097867349628359,
7409
+ "grad_norm/grad_2.0_norm/decoder.estimator.down_blocks.1.1.0.ff.net.0.beta": 0.00040464234189130366,
7410
+ "grad_norm/grad_2.0_norm/decoder.estimator.down_blocks.1.1.0.ff.net.0.proj.weight": 0.009615248069167137,
7411
+ "grad_norm/grad_2.0_norm/decoder.estimator.down_blocks.1.1.0.ff.net.0.proj.bias": 0.0007450803532265127,
7412
+ "grad_norm/grad_2.0_norm/decoder.estimator.down_blocks.1.1.0.ff.net.2.weight": 0.009156358428299427,
7413
+ "grad_norm/grad_2.0_norm/decoder.estimator.down_blocks.1.1.0.ff.net.2.bias": 0.0008174045942723751,
7414
+ "grad_norm/grad_2.0_norm/decoder.estimator.down_blocks.1.2.weight": 0.03646900877356529,
7415
+ "grad_norm/grad_2.0_norm/decoder.estimator.down_blocks.1.2.bias": 0.0004777516005560756,
7416
+ "grad_norm/grad_2.0_norm/decoder.estimator.mid_blocks.0.0.mlp.1.weight": 0.007459959480911493,
7417
+ "grad_norm/grad_2.0_norm/decoder.estimator.mid_blocks.0.0.mlp.1.bias": 0.000627974106464535,
7418
+ "grad_norm/grad_2.0_norm/decoder.estimator.mid_blocks.0.0.block1.block.0.weight": 0.006041461601853371,
7419
+ "grad_norm/grad_2.0_norm/decoder.estimator.mid_blocks.0.0.block1.block.0.bias": 5.324760059011169e-05,
7420
+ "grad_norm/grad_2.0_norm/decoder.estimator.mid_blocks.0.0.block1.block.1.weight": 0.0005003157421015203,
7421
+ "grad_norm/grad_2.0_norm/decoder.estimator.mid_blocks.0.0.block1.block.1.bias": 0.00042986817425116897,
7422
+ "grad_norm/grad_2.0_norm/decoder.estimator.mid_blocks.0.0.block2.block.0.weight": 0.006311975885182619,
7423
+ "grad_norm/grad_2.0_norm/decoder.estimator.mid_blocks.0.0.block2.block.0.bias": 0.000369485147530213,
7424
+ "grad_norm/grad_2.0_norm/decoder.estimator.mid_blocks.0.0.block2.block.1.weight": 0.00034039438469335437,
7425
+ "grad_norm/grad_2.0_norm/decoder.estimator.mid_blocks.0.0.block2.block.1.bias": 0.0003005445178132504,
7426
+ "grad_norm/grad_2.0_norm/decoder.estimator.mid_blocks.0.0.res_conv.weight": 0.02558288909494877,
7427
+ "grad_norm/grad_2.0_norm/decoder.estimator.mid_blocks.0.0.res_conv.bias": 0.0469089038670063,
7428
+ "grad_norm/grad_2.0_norm/decoder.estimator.mid_blocks.0.1.0.norm1.weight": 0.0006509429658763111,
7429
+ "grad_norm/grad_2.0_norm/decoder.estimator.mid_blocks.0.1.0.norm1.bias": 0.0008774683228693902,
7430
+ "grad_norm/grad_2.0_norm/decoder.estimator.mid_blocks.0.1.0.attn1.to_q.weight": 0.0028540780767798424,
7431
+ "grad_norm/grad_2.0_norm/decoder.estimator.mid_blocks.0.1.0.attn1.to_k.weight": 0.0016212090849876404,
7432
+ "grad_norm/grad_2.0_norm/decoder.estimator.mid_blocks.0.1.0.attn1.to_v.weight": 0.00638106744736433,
7433
+ "grad_norm/grad_2.0_norm/decoder.estimator.mid_blocks.0.1.0.attn1.to_out.0.weight": 0.0010032415157184005,
7434
+ "grad_norm/grad_2.0_norm/decoder.estimator.mid_blocks.0.1.0.attn1.to_out.0.bias": 0.00048305917880497873,
7435
+ "grad_norm/grad_2.0_norm/decoder.estimator.mid_blocks.0.1.0.norm3.weight": 0.0007569156587123871,
7436
+ "grad_norm/grad_2.0_norm/decoder.estimator.mid_blocks.0.1.0.norm3.bias": 0.0008760644705034792,
7437
+ "grad_norm/grad_2.0_norm/decoder.estimator.mid_blocks.0.1.0.ff.net.0.alpha": 0.0002850043820217252,
7438
+ "grad_norm/grad_2.0_norm/decoder.estimator.mid_blocks.0.1.0.ff.net.0.beta": 0.00030988469370640814,
7439
+ "grad_norm/grad_2.0_norm/decoder.estimator.mid_blocks.0.1.0.ff.net.0.proj.weight": 0.007048940751701593,
7440
+ "grad_norm/grad_2.0_norm/decoder.estimator.mid_blocks.0.1.0.ff.net.0.proj.bias": 0.0005304201622493565,
7441
+ "grad_norm/grad_2.0_norm/decoder.estimator.mid_blocks.0.1.0.ff.net.2.weight": 0.008041159249842167,
7442
+ "grad_norm/grad_2.0_norm/decoder.estimator.mid_blocks.0.1.0.ff.net.2.bias": 0.0004873759171459824,
7443
+ "grad_norm/grad_2.0_norm/decoder.estimator.mid_blocks.1.0.mlp.1.weight": 0.006650703493505716,
7444
+ "grad_norm/grad_2.0_norm/decoder.estimator.mid_blocks.1.0.mlp.1.bias": 0.0006425073370337486,
7445
+ "grad_norm/grad_2.0_norm/decoder.estimator.mid_blocks.1.0.block1.block.0.weight": 0.006960422731935978,
7446
+ "grad_norm/grad_2.0_norm/decoder.estimator.mid_blocks.1.0.block1.block.0.bias": 4.733038440463133e-05,
7447
+ "grad_norm/grad_2.0_norm/decoder.estimator.mid_blocks.1.0.block1.block.1.weight": 0.0004921250510960817,
7448
+ "grad_norm/grad_2.0_norm/decoder.estimator.mid_blocks.1.0.block1.block.1.bias": 0.0004338321159593761,
7449
+ "grad_norm/grad_2.0_norm/decoder.estimator.mid_blocks.1.0.block2.block.0.weight": 0.005244262516498566,
7450
+ "grad_norm/grad_2.0_norm/decoder.estimator.mid_blocks.1.0.block2.block.0.bias": 0.0003615597961470485,
7451
+ "grad_norm/grad_2.0_norm/decoder.estimator.mid_blocks.1.0.block2.block.1.weight": 0.0002653438423294574,
7452
+ "grad_norm/grad_2.0_norm/decoder.estimator.mid_blocks.1.0.block2.block.1.bias": 0.0002546966716181487,
7453
+ "grad_norm/grad_2.0_norm/decoder.estimator.mid_blocks.1.0.res_conv.weight": 0.03545884042978287,
7454
+ "grad_norm/grad_2.0_norm/decoder.estimator.mid_blocks.1.0.res_conv.bias": 0.031797852367162704,
7455
+ "grad_norm/grad_2.0_norm/decoder.estimator.mid_blocks.1.1.0.norm1.weight": 0.0006199093186296523,
7456
+ "grad_norm/grad_2.0_norm/decoder.estimator.mid_blocks.1.1.0.norm1.bias": 0.0008437394862994552,
7457
+ "grad_norm/grad_2.0_norm/decoder.estimator.mid_blocks.1.1.0.attn1.to_q.weight": 0.0020717547740787268,
7458
+ "grad_norm/grad_2.0_norm/decoder.estimator.mid_blocks.1.1.0.attn1.to_k.weight": 0.0020844489336013794,
7459
+ "grad_norm/grad_2.0_norm/decoder.estimator.mid_blocks.1.1.0.attn1.to_v.weight": 0.006132850889116526,
7460
+ "grad_norm/grad_2.0_norm/decoder.estimator.mid_blocks.1.1.0.attn1.to_out.0.weight": 0.0007109093712642789,
7461
+ "grad_norm/grad_2.0_norm/decoder.estimator.mid_blocks.1.1.0.attn1.to_out.0.bias": 0.00044272647937759757,
7462
+ "grad_norm/grad_2.0_norm/decoder.estimator.mid_blocks.1.1.0.norm3.weight": 0.0007580879027955234,
7463
+ "grad_norm/grad_2.0_norm/decoder.estimator.mid_blocks.1.1.0.norm3.bias": 0.0008920146501623094,
7464
+ "grad_norm/grad_2.0_norm/decoder.estimator.mid_blocks.1.1.0.ff.net.0.alpha": 0.00031814113026484847,
7465
+ "grad_norm/grad_2.0_norm/decoder.estimator.mid_blocks.1.1.0.ff.net.0.beta": 0.0003361067792866379,
7466
+ "grad_norm/grad_2.0_norm/decoder.estimator.mid_blocks.1.1.0.ff.net.0.proj.weight": 0.007014248985797167,
7467
+ "grad_norm/grad_2.0_norm/decoder.estimator.mid_blocks.1.1.0.ff.net.0.proj.bias": 0.000548130483366549,
7468
+ "grad_norm/grad_2.0_norm/decoder.estimator.mid_blocks.1.1.0.ff.net.2.weight": 0.00817958079278469,
7469
+ "grad_norm/grad_2.0_norm/decoder.estimator.mid_blocks.1.1.0.ff.net.2.bias": 0.00045939270057715476,
7470
+ "grad_norm/grad_2.0_norm/decoder.estimator.up_blocks.0.0.mlp.1.weight": 0.010318427346646786,
7471
+ "grad_norm/grad_2.0_norm/decoder.estimator.up_blocks.0.0.mlp.1.bias": 0.0010112684685736895,
7472
+ "grad_norm/grad_2.0_norm/decoder.estimator.up_blocks.0.0.block1.block.0.weight": 0.014999396167695522,
7473
+ "grad_norm/grad_2.0_norm/decoder.estimator.up_blocks.0.0.block1.block.0.bias": 8.258102752733976e-05,
7474
+ "grad_norm/grad_2.0_norm/decoder.estimator.up_blocks.0.0.block1.block.1.weight": 0.0007028156542219222,
7475
+ "grad_norm/grad_2.0_norm/decoder.estimator.up_blocks.0.0.block1.block.1.bias": 0.000596357393078506,
7476
+ "grad_norm/grad_2.0_norm/decoder.estimator.up_blocks.0.0.block2.block.0.weight": 0.0061602480709552765,
7477
+ "grad_norm/grad_2.0_norm/decoder.estimator.up_blocks.0.0.block2.block.0.bias": 0.0005165530019439757,
7478
+ "grad_norm/grad_2.0_norm/decoder.estimator.up_blocks.0.0.block2.block.1.weight": 0.00038624528679065406,
7479
+ "grad_norm/grad_2.0_norm/decoder.estimator.up_blocks.0.0.block2.block.1.bias": 0.00032755982829257846,
7480
+ "grad_norm/grad_2.0_norm/decoder.estimator.up_blocks.0.0.res_conv.weight": 0.05575540289282799,
7481
+ "grad_norm/grad_2.0_norm/decoder.estimator.up_blocks.0.0.res_conv.bias": 0.08856959640979767,
7482
+ "grad_norm/grad_2.0_norm/decoder.estimator.up_blocks.0.1.0.norm1.weight": 0.000856694474350661,
7483
+ "grad_norm/grad_2.0_norm/decoder.estimator.up_blocks.0.1.0.norm1.bias": 0.001748208305798471,
7484
+ "grad_norm/grad_2.0_norm/decoder.estimator.up_blocks.0.1.0.attn1.to_q.weight": 0.0028679671231657267,
7485
+ "grad_norm/grad_2.0_norm/decoder.estimator.up_blocks.0.1.0.attn1.to_k.weight": 0.0022758597042411566,
7486
+ "grad_norm/grad_2.0_norm/decoder.estimator.up_blocks.0.1.0.attn1.to_v.weight": 0.010127125307917595,
7487
+ "grad_norm/grad_2.0_norm/decoder.estimator.up_blocks.0.1.0.attn1.to_out.0.weight": 0.0009156019659712911,
7488
+ "grad_norm/grad_2.0_norm/decoder.estimator.up_blocks.0.1.0.attn1.to_out.0.bias": 0.0006679215002804995,
7489
+ "grad_norm/grad_2.0_norm/decoder.estimator.up_blocks.0.1.0.norm3.weight": 0.0010506961261853576,
7490
+ "grad_norm/grad_2.0_norm/decoder.estimator.up_blocks.0.1.0.norm3.bias": 0.00132385129109025,
7491
+ "grad_norm/grad_2.0_norm/decoder.estimator.up_blocks.0.1.0.ff.net.0.alpha": 0.0004299134307075292,
7492
+ "grad_norm/grad_2.0_norm/decoder.estimator.up_blocks.0.1.0.ff.net.0.beta": 0.00042983077582903206,
7493
+ "grad_norm/grad_2.0_norm/decoder.estimator.up_blocks.0.1.0.ff.net.0.proj.weight": 0.00906638614833355,
7494
+ "grad_norm/grad_2.0_norm/decoder.estimator.up_blocks.0.1.0.ff.net.0.proj.bias": 0.0007534258766099811,
7495
+ "grad_norm/grad_2.0_norm/decoder.estimator.up_blocks.0.1.0.ff.net.2.weight": 0.008831251412630081,
7496
+ "grad_norm/grad_2.0_norm/decoder.estimator.up_blocks.0.1.0.ff.net.2.bias": 0.0007010450935922563,
7497
+ "grad_norm/grad_2.0_norm/decoder.estimator.up_blocks.0.2.conv.weight": 0.09383069723844528,
7498
+ "grad_norm/grad_2.0_norm/decoder.estimator.up_blocks.0.2.conv.bias": 0.0016627543373033404,
7499
+ "grad_norm/grad_2.0_norm/decoder.estimator.up_blocks.1.0.mlp.1.weight": 0.028144413605332375,
7500
+ "grad_norm/grad_2.0_norm/decoder.estimator.up_blocks.1.0.mlp.1.bias": 0.003071507904678583,
7501
+ "grad_norm/grad_2.0_norm/decoder.estimator.up_blocks.1.0.block1.block.0.weight": 0.030205750837922096,
7502
+ "grad_norm/grad_2.0_norm/decoder.estimator.up_blocks.1.0.block1.block.0.bias": 0.0005511984345503151,
7503
+ "grad_norm/grad_2.0_norm/decoder.estimator.up_blocks.1.0.block1.block.1.weight": 0.002031352138146758,
7504
+ "grad_norm/grad_2.0_norm/decoder.estimator.up_blocks.1.0.block1.block.1.bias": 0.0021938334684818983,
7505
+ "grad_norm/grad_2.0_norm/decoder.estimator.up_blocks.1.0.block2.block.0.weight": 0.025340333580970764,
7506
+ "grad_norm/grad_2.0_norm/decoder.estimator.up_blocks.1.0.block2.block.0.bias": 0.0018560552271082997,
7507
+ "grad_norm/grad_2.0_norm/decoder.estimator.up_blocks.1.0.block2.block.1.weight": 0.002064154017716646,
7508
+ "grad_norm/grad_2.0_norm/decoder.estimator.up_blocks.1.0.block2.block.1.bias": 0.001897836453281343,
7509
+ "grad_norm/grad_2.0_norm/decoder.estimator.up_blocks.1.0.res_conv.weight": 0.09192795306444168,
7510
+ "grad_norm/grad_2.0_norm/decoder.estimator.up_blocks.1.0.res_conv.bias": 0.025354549288749695,
7511
+ "grad_norm/grad_2.0_norm/decoder.estimator.up_blocks.1.1.0.norm1.weight": 0.004005036316812038,
7512
+ "grad_norm/grad_2.0_norm/decoder.estimator.up_blocks.1.1.0.norm1.bias": 0.0044862860813736916,
7513
+ "grad_norm/grad_2.0_norm/decoder.estimator.up_blocks.1.1.0.attn1.to_q.weight": 0.003407246433198452,
7514
+ "grad_norm/grad_2.0_norm/decoder.estimator.up_blocks.1.1.0.attn1.to_k.weight": 0.001814698800444603,
7515
+ "grad_norm/grad_2.0_norm/decoder.estimator.up_blocks.1.1.0.attn1.to_v.weight": 0.04642708599567413,
7516
+ "grad_norm/grad_2.0_norm/decoder.estimator.up_blocks.1.1.0.attn1.to_out.0.weight": 0.001497728400863707,
7517
+ "grad_norm/grad_2.0_norm/decoder.estimator.up_blocks.1.1.0.attn1.to_out.0.bias": 0.00255810865201056,
7518
+ "grad_norm/grad_2.0_norm/decoder.estimator.up_blocks.1.1.0.norm3.weight": 0.0026136802043765783,
7519
+ "grad_norm/grad_2.0_norm/decoder.estimator.up_blocks.1.1.0.norm3.bias": 0.003123651025816798,
7520
+ "grad_norm/grad_2.0_norm/decoder.estimator.up_blocks.1.1.0.ff.net.0.alpha": 0.0011864164844155312,
7521
+ "grad_norm/grad_2.0_norm/decoder.estimator.up_blocks.1.1.0.ff.net.0.beta": 0.0009428896591998637,
7522
+ "grad_norm/grad_2.0_norm/decoder.estimator.up_blocks.1.1.0.ff.net.0.proj.weight": 0.023969616740942,
7523
+ "grad_norm/grad_2.0_norm/decoder.estimator.up_blocks.1.1.0.ff.net.0.proj.bias": 0.0018177471356466413,
7524
+ "grad_norm/grad_2.0_norm/decoder.estimator.up_blocks.1.1.0.ff.net.2.weight": 0.014123436063528061,
7525
+ "grad_norm/grad_2.0_norm/decoder.estimator.up_blocks.1.1.0.ff.net.2.bias": 0.002645216416567564,
7526
+ "grad_norm/grad_2.0_norm/decoder.estimator.up_blocks.1.2.weight": 0.18219496309757233,
7527
+ "grad_norm/grad_2.0_norm/decoder.estimator.up_blocks.1.2.bias": 0.0013607360888272524,
7528
+ "grad_norm/grad_2.0_norm/decoder.estimator.final_block.block.0.weight": 0.1666623204946518,
7529
+ "grad_norm/grad_2.0_norm/decoder.estimator.final_block.block.0.bias": 0.0008949644980020821,
7530
+ "grad_norm/grad_2.0_norm/decoder.estimator.final_block.block.1.weight": 0.01216113194823265,
7531
+ "grad_norm/grad_2.0_norm/decoder.estimator.final_block.block.1.bias": 0.011881562881171703,
7532
+ "grad_norm/grad_2.0_norm/decoder.estimator.final_proj.weight": 0.11852734535932541,
7533
+ "grad_norm/grad_2.0_norm/decoder.estimator.final_proj.bias": 0.006633310113102198,
7534
+ "grad_norm/grad_2.0_norm_total": 0.9888164401054382,
7535
+ "sub_loss/val_dur_loss": 0.32394927740097046,
7536
+ "sub_loss/val_dur_loss_epoch": 0.32394927740097046,
7537
+ "sub_loss/val_prior_loss": 0.96378493309021,
7538
+ "sub_loss/val_prior_loss_epoch": 0.96378493309021,
7539
+ "sub_loss/val_diff_loss": 0.8976922035217285,
7540
+ "sub_loss/val_diff_loss_epoch": 0.8976922035217285,
7541
+ "loss/val": 2.1854262351989746,
7542
+ "loss/val_epoch": 2.1854262351989746,
7543
+ "sub_loss/train_dur_loss_epoch": 0.3276459276676178,
7544
+ "sub_loss/train_prior_loss_epoch": 0.9636818766593933,
7545
+ "sub_loss/train_diff_loss_epoch": 0.8940833210945129,
7546
+ "loss/train_epoch": 2.185410737991333,
7547
+ "epoch": 23
7548
  }
7549
  ]
7550
  }