loyi commited on
Commit
30321b3
·
1 Parent(s): b1d194f

add sena(姬野星奏) model

Browse files
sena441/sena441/clean_sena.ckpt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:ecca3a4d938bc08322a35d3c2a6188319fcee44fc972325996c9debfcc25dd06
3
+ size 227118939
sena441/sena441/config.yaml ADDED
@@ -0,0 +1,445 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ K_step: 1000
2
+ accumulate_grad_batches: 1
3
+ audio_num_mel_bins: 128
4
+ audio_sample_rate: 44100
5
+ binarization_args:
6
+ shuffle: false
7
+ with_align: true
8
+ with_f0: true
9
+ with_hubert: true
10
+ with_spk_embed: false
11
+ with_wav: false
12
+ binarizer_cls: preprocessing.SVCpre.SVCBinarizer
13
+ binary_data_dir: data/binary/sena
14
+ check_val_every_n_epoch: 10
15
+ choose_test_manually: true
16
+ clip_grad_norm: 1
17
+ config_path: training/config_nsf_sena.yaml
18
+ content_cond_steps: []
19
+ cwt_add_f0_loss: false
20
+ cwt_hidden_size: 128
21
+ cwt_layers: 2
22
+ cwt_loss: l1
23
+ cwt_std_scale: 0.8
24
+ datasets:
25
+ - opencpop
26
+ debug: false
27
+ dec_ffn_kernel_size: 9
28
+ dec_layers: 4
29
+ decay_steps: 10000
30
+ decoder_type: fft
31
+ dict_dir: ''
32
+ diff_decoder_type: wavenet
33
+ diff_loss_type: l2
34
+ dilation_cycle_length: 4
35
+ dropout: 0.1
36
+ ds_workers: 4
37
+ dur_enc_hidden_stride_kernel:
38
+ - 0,2,3
39
+ - 0,2,3
40
+ - 0,1,3
41
+ dur_loss: mse
42
+ dur_predictor_kernel: 3
43
+ dur_predictor_layers: 5
44
+ enc_ffn_kernel_size: 9
45
+ enc_layers: 4
46
+ encoder_K: 8
47
+ encoder_type: fft
48
+ endless_ds: false
49
+ f0_bin: 256
50
+ f0_max: 1100.0
51
+ f0_min: 40.0
52
+ ffn_act: gelu
53
+ ffn_padding: SAME
54
+ fft_size: 2048
55
+ fmax: 16000
56
+ fmin: 40
57
+ fs2_ckpt: ''
58
+ gaussian_start: true
59
+ gen_dir_name: ''
60
+ gen_tgt_spk_id: -1
61
+ hidden_size: 256
62
+ hop_size: 512
63
+ hubert_gpu: true
64
+ hubert_path: ckpt/hubert/hubert_soft.pt
65
+ infer: false
66
+ keep_bins: 128
67
+ lambda_commit: 0.25
68
+ lambda_energy: 0.0
69
+ lambda_f0: 1.0
70
+ lambda_ph_dur: 0.3
71
+ lambda_sent_dur: 1.0
72
+ lambda_uv: 1.0
73
+ lambda_word_dur: 1.0
74
+ load_ckpt: ''
75
+ log_interval: 100
76
+ loud_norm: false
77
+ lr: 0.0008
78
+ max_beta: 0.02
79
+ max_epochs: 3000
80
+ max_eval_sentences: 1
81
+ max_eval_tokens: 60000
82
+ max_frames: 42000
83
+ max_input_tokens: 60000
84
+ max_sentences: 128
85
+ max_tokens: 42000
86
+ max_updates: 1000000
87
+ mel_loss: ssim:0.5|l1:0.5
88
+ mel_vmax: 1.5
89
+ mel_vmin: -6.0
90
+ min_level_db: -120
91
+ no_fs2: true
92
+ norm_type: gn
93
+ num_ckpt_keep: 10
94
+ num_heads: 2
95
+ num_sanity_val_steps: 1
96
+ num_spk: 1
97
+ num_test_samples: 0
98
+ num_valid_plots: 10
99
+ optimizer_adam_beta1: 0.9
100
+ optimizer_adam_beta2: 0.98
101
+ out_wav_norm: false
102
+ pe_ckpt: ckpt/0102_xiaoma_pe/model_ckpt_steps_60000.ckpt
103
+ pe_enable: false
104
+ perform_enhance: true
105
+ pitch_ar: false
106
+ pitch_enc_hidden_stride_kernel:
107
+ - 0,2,5
108
+ - 0,2,5
109
+ - 0,2,5
110
+ pitch_extractor: parselmouth
111
+ pitch_loss: l2
112
+ pitch_norm: log
113
+ pitch_type: frame
114
+ pndm_speedup: 10
115
+ pre_align_args:
116
+ allow_no_txt: false
117
+ denoise: false
118
+ forced_align: mfa
119
+ txt_processor: zh_g2pM
120
+ use_sox: true
121
+ use_tone: false
122
+ pre_align_cls: data_gen.singing.pre_align.SingingPreAlign
123
+ predictor_dropout: 0.5
124
+ predictor_grad: 0.1
125
+ predictor_hidden: -1
126
+ predictor_kernel: 5
127
+ predictor_layers: 5
128
+ prenet_dropout: 0.5
129
+ prenet_hidden_size: 256
130
+ pretrain_fs_ckpt: ''
131
+ processed_data_dir: xxx
132
+ profile_infer: false
133
+ raw_data_dir: data/raw/sena
134
+ ref_norm_layer: bn
135
+ rel_pos: true
136
+ reset_phone_dict: true
137
+ residual_channels: 512
138
+ residual_layers: 20
139
+ save_best: false
140
+ save_ckpt: true
141
+ save_codes:
142
+ - configs
143
+ - modules
144
+ - src
145
+ - utils
146
+ save_f0: true
147
+ save_gt: false
148
+ schedule_type: linear
149
+ seed: 1234
150
+ sort_by_len: true
151
+ speaker_id: sena
152
+ spec_max:
153
+ - 0.17097608745098114
154
+ - 0.3397301733493805
155
+ - 0.6082707047462463
156
+ - 0.7658080458641052
157
+ - 0.9264223575592041
158
+ - 1.1378915309906006
159
+ - 1.2659214735031128
160
+ - 1.231523036956787
161
+ - 1.2456761598587036
162
+ - 1.2685513496398926
163
+ - 1.1583133935928345
164
+ - 1.1537851095199585
165
+ - 1.2560111284255981
166
+ - 1.1849987506866455
167
+ - 1.1610959768295288
168
+ - 1.2605981826782227
169
+ - 1.226485013961792
170
+ - 1.1268006563186646
171
+ - 1.137515664100647
172
+ - 1.177683711051941
173
+ - 1.170212745666504
174
+ - 1.1891556978225708
175
+ - 1.204882025718689
176
+ - 1.151033878326416
177
+ - 1.1880069971084595
178
+ - 1.2375954389572144
179
+ - 1.1038519144058228
180
+ - 1.1212972402572632
181
+ - 1.197728157043457
182
+ - 1.1437523365020752
183
+ - 1.0745612382888794
184
+ - 1.1244035959243774
185
+ - 1.1003185510635376
186
+ - 1.106966257095337
187
+ - 1.1127562522888184
188
+ - 1.0868451595306396
189
+ - 1.0899468660354614
190
+ - 1.0303136110305786
191
+ - 1.04606294631958
192
+ - 1.03463876247406
193
+ - 1.016716480255127
194
+ - 1.0034751892089844
195
+ - 1.0667635202407837
196
+ - 1.0440107583999634
197
+ - 0.9983909130096436
198
+ - 0.9837781190872192
199
+ - 0.9780927300453186
200
+ - 0.9547947645187378
201
+ - 1.0833947658538818
202
+ - 1.0186176300048828
203
+ - 1.0262666940689087
204
+ - 0.9818507432937622
205
+ - 0.9809918403625488
206
+ - 0.9516422748565674
207
+ - 0.9858482480049133
208
+ - 0.8438718914985657
209
+ - 0.9086287617683411
210
+ - 0.8336443901062012
211
+ - 0.8408522009849548
212
+ - 0.7966897487640381
213
+ - 0.7060108184814453
214
+ - 0.6853682398796082
215
+ - 0.6755683422088623
216
+ - 0.6235848665237427
217
+ - 0.6333395838737488
218
+ - 0.6069809198379517
219
+ - 0.5742683410644531
220
+ - 0.665706217288971
221
+ - 0.6048570275306702
222
+ - 0.7021833062171936
223
+ - 0.7196664810180664
224
+ - 0.5961004495620728
225
+ - 0.5049697756767273
226
+ - 0.49245569109916687
227
+ - 0.5433599352836609
228
+ - 0.6448594331741333
229
+ - 0.731346607208252
230
+ - 0.524942934513092
231
+ - 0.44391942024230957
232
+ - 0.5200350880622864
233
+ - 0.4035450220108032
234
+ - 0.4356535077095032
235
+ - 0.49859124422073364
236
+ - 0.48476505279541016
237
+ - 0.5568605661392212
238
+ - 0.45708560943603516
239
+ - 0.4374412000179291
240
+ - 0.45580312609672546
241
+ - 0.4294637441635132
242
+ - 0.3111153542995453
243
+ - 0.2814539670944214
244
+ - 0.2886189818382263
245
+ - 0.2661425471305847
246
+ - 0.20193226635456085
247
+ - 0.38692978024482727
248
+ - 0.3419096767902374
249
+ - 0.48901742696762085
250
+ - 0.4254615008831024
251
+ - 0.38061267137527466
252
+ - 0.35519418120384216
253
+ - 0.35050907731056213
254
+ - 0.5192219018936157
255
+ - 0.4214952886104584
256
+ - 0.4718898832798004
257
+ - 0.428811639547348
258
+ - 0.5504823327064514
259
+ - 0.5378233194351196
260
+ - 0.49835407733917236
261
+ - 0.40051230788230896
262
+ - 0.34427186846733093
263
+ - 0.46247372031211853
264
+ - 0.3569100499153137
265
+ - 0.4144101142883301
266
+ - 0.4045094847679138
267
+ - 0.45712941884994507
268
+ - 0.43897101283073425
269
+ - 0.47350218892097473
270
+ - 0.5039759874343872
271
+ - 0.4046401381492615
272
+ - 0.354009747505188
273
+ - 0.3337729573249817
274
+ - 0.3514827489852905
275
+ - 0.19709141552448273
276
+ - 0.09294259548187256
277
+ - -0.09443746507167816
278
+ - 0.13250915706157684
279
+ - -0.014945651404559612
280
+ - -0.06271237879991531
281
+ spec_min:
282
+ - -4.999994277954102
283
+ - -4.999994277954102
284
+ - -4.999994277954102
285
+ - -4.999994277954102
286
+ - -4.999994277954102
287
+ - -4.999994277954102
288
+ - -4.999994277954102
289
+ - -4.999994277954102
290
+ - -4.999994277954102
291
+ - -4.999994277954102
292
+ - -4.999994277954102
293
+ - -4.999994277954102
294
+ - -4.999994277954102
295
+ - -4.999994277954102
296
+ - -4.999994277954102
297
+ - -4.999994277954102
298
+ - -4.999994277954102
299
+ - -4.999994277954102
300
+ - -4.999994277954102
301
+ - -4.999994277954102
302
+ - -4.999994277954102
303
+ - -4.999994277954102
304
+ - -4.999994277954102
305
+ - -4.999994277954102
306
+ - -4.999994277954102
307
+ - -4.999994277954102
308
+ - -4.999994277954102
309
+ - -4.999994277954102
310
+ - -4.999994277954102
311
+ - -4.999994277954102
312
+ - -4.999994277954102
313
+ - -4.999994277954102
314
+ - -4.999994277954102
315
+ - -4.999994277954102
316
+ - -4.999994277954102
317
+ - -4.999994277954102
318
+ - -4.999994277954102
319
+ - -4.999994277954102
320
+ - -4.999994277954102
321
+ - -4.999994277954102
322
+ - -4.999994277954102
323
+ - -4.999994277954102
324
+ - -4.999994277954102
325
+ - -4.999994277954102
326
+ - -4.999994277954102
327
+ - -4.999994277954102
328
+ - -4.999994277954102
329
+ - -4.999994277954102
330
+ - -4.999994277954102
331
+ - -4.999994277954102
332
+ - -4.999994277954102
333
+ - -4.999994277954102
334
+ - -4.999994277954102
335
+ - -4.999994277954102
336
+ - -4.999994277954102
337
+ - -4.999994277954102
338
+ - -4.999994277954102
339
+ - -4.999994277954102
340
+ - -4.999994277954102
341
+ - -4.999994277954102
342
+ - -4.999994277954102
343
+ - -4.999994277954102
344
+ - -4.999994277954102
345
+ - -4.999994277954102
346
+ - -4.999994277954102
347
+ - -4.999994277954102
348
+ - -4.999994277954102
349
+ - -4.999994277954102
350
+ - -4.999994277954102
351
+ - -4.999994277954102
352
+ - -4.999994277954102
353
+ - -4.999994277954102
354
+ - -4.999994277954102
355
+ - -4.999994277954102
356
+ - -4.999994277954102
357
+ - -4.999994277954102
358
+ - -4.999994277954102
359
+ - -4.999994277954102
360
+ - -4.999994277954102
361
+ - -4.999994277954102
362
+ - -4.999994277954102
363
+ - -4.999994277954102
364
+ - -4.999994277954102
365
+ - -4.999994277954102
366
+ - -4.999994277954102
367
+ - -4.999994277954102
368
+ - -4.999994277954102
369
+ - -4.999994277954102
370
+ - -4.999994277954102
371
+ - -4.999994277954102
372
+ - -4.999994277954102
373
+ - -4.999994277954102
374
+ - -4.999994277954102
375
+ - -4.999994277954102
376
+ - -4.999994277954102
377
+ - -4.999994277954102
378
+ - -4.999994277954102
379
+ - -4.999994277954102
380
+ - -4.999994277954102
381
+ - -4.999994277954102
382
+ - -4.999994277954102
383
+ - -4.999994277954102
384
+ - -4.999994277954102
385
+ - -4.999994277954102
386
+ - -4.999994277954102
387
+ - -4.999994277954102
388
+ - -4.999994277954102
389
+ - -4.999994277954102
390
+ - -4.999994277954102
391
+ - -4.999994277954102
392
+ - -4.999994277954102
393
+ - -4.999994277954102
394
+ - -4.999994277954102
395
+ - -4.999994277954102
396
+ - -4.999994277954102
397
+ - -4.999994277954102
398
+ - -4.999994277954102
399
+ - -4.999994277954102
400
+ - -4.999994277954102
401
+ - -4.999994277954102
402
+ - -4.999994277954102
403
+ - -4.999994277954102
404
+ - -4.999994277954102
405
+ - -4.999994277954102
406
+ - -4.999994277954102
407
+ - -4.999994277954102
408
+ - -4.999994277954102
409
+ - -4.999994277954102
410
+ spk_cond_steps: []
411
+ stop_token_weight: 5.0
412
+ task_cls: training.task.SVC_task.SVCTask
413
+ test_ids: []
414
+ test_input_dir: ''
415
+ test_num: 0
416
+ test_prefixes:
417
+ - data\raw\sena\test
418
+ test_set_name: test
419
+ timesteps: 1000
420
+ train_set_name: train
421
+ use_crepe: true
422
+ use_denoise: false
423
+ use_energy_embed: false
424
+ use_gt_dur: false
425
+ use_gt_f0: false
426
+ use_midi: false
427
+ use_nsf: true
428
+ use_pitch_embed: true
429
+ use_pos_embed: true
430
+ use_spk_embed: false
431
+ use_spk_id: false
432
+ use_split_spk_id: false
433
+ use_uv: false
434
+ use_var_enc: false
435
+ use_vec: false
436
+ val_check_interval: 1000
437
+ valid_num: 0
438
+ valid_set_name: valid
439
+ vocoder: network.vocoders.nsf_hifigan.NsfHifiGAN
440
+ vocoder_ckpt: ckpt/nsf_hifigan/model
441
+ warmup_updates: 2000
442
+ wav2spec_eps: 1e-6
443
+ weight_decay: 0
444
+ win_size: 2048
445
+ work_dir: ckpt/sena