Hugging Face's logo

sparanoid
/
milky-green-diff-svc

Audio-to-Audio
Chinese
English
Japanese
Model card Files Community
sparanoid commited on
Commit
1ad6484
•
1 Parent(s): b32fb43

Upload config.yaml

Browse files
Files changed (1) hide show
  1. config.yaml +349 -0
config.yaml ADDED
@@ -0,0 +1,349 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ K_step: 1000
2
+ accumulate_grad_batches: 1
3
+ audio_num_mel_bins: 80
4
+ audio_sample_rate: 24000
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/mqnl-covers
14
+ check_val_every_n_epoch: 10
15
+ choose_test_manually: false
16
+ clip_grad_norm: 1
17
+ config_path: training/config.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: 40000
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: 50.0
52
+ ffn_act: gelu
53
+ ffn_padding: SAME
54
+ fft_size: 512
55
+ fmax: 12000
56
+ fmin: 30
57
+ fs2_ckpt: ''
58
+ gaussian_start: true
59
+ gen_dir_name: ''
60
+ gen_tgt_spk_id: -1
61
+ hidden_size: 256
62
+ hop_size: 128
63
+ hubert_gpu: true
64
+ hubert_path: checkpoints/hubert/hubert_soft.pt
65
+ infer: false
66
+ keep_bins: 80
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: 0.0004
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: 88
85
+ max_tokens: 128000
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: checkpoints/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/mqnl-covers
134
+ ref_norm_layer: bn
135
+ rel_pos: true
136
+ reset_phone_dict: true
137
+ residual_channels: 256
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: mqnl-covers
152
+ spec_max:
153
+ - 0.11258526891469955
154
+ - 0.029963258653879166
155
+ - 0.25167587399482727
156
+ - 0.3379969000816345
157
+ - 0.4209271967411041
158
+ - 0.4627160131931305
159
+ - 0.41560280323028564
160
+ - 0.3913577198982239
161
+ - 0.34750282764434814
162
+ - 0.29936230182647705
163
+ - 0.32207849621772766
164
+ - 0.3448757231235504
165
+ - 0.3113778233528137
166
+ - 0.35317060351371765
167
+ - 0.4166089594364166
168
+ - 0.3650610148906708
169
+ - 0.2879890203475952
170
+ - 0.21680839359760284
171
+ - 0.2509251832962036
172
+ - 0.25092849135398865
173
+ - 0.24402980506420135
174
+ - 0.28403371572494507
175
+ - 0.32604289054870605
176
+ - 0.3334726393222809
177
+ - 0.38054946064949036
178
+ - 0.41315314173698425
179
+ - 0.2945516109466553
180
+ - 0.350572794675827
181
+ - 0.4011584222316742
182
+ - 0.1728031039237976
183
+ - 0.17655976116657257
184
+ - 0.1981496810913086
185
+ - 0.11094453185796738
186
+ - 0.23602479696273804
187
+ - 0.06451099365949631
188
+ - 0.24015556275844574
189
+ - 0.01039820071309805
190
+ - 0.038468487560749054
191
+ - -0.03273174166679382
192
+ - 0.015290342271327972
193
+ - -0.0906442329287529
194
+ - -0.19544707238674164
195
+ - -0.03575506806373596
196
+ - -0.10758063197135925
197
+ - -0.12621892988681793
198
+ - -0.2831754982471466
199
+ - -0.2826099097728729
200
+ - -0.08007119596004486
201
+ - 0.09162873029708862
202
+ - -0.01513915229588747
203
+ - -0.131241112947464
204
+ - -0.2449709177017212
205
+ - -0.3528900146484375
206
+ - -0.2924903929233551
207
+ - -0.15330247581005096
208
+ - -0.1724396049976349
209
+ - -0.19285175204277039
210
+ - -0.11945571005344391
211
+ - -0.25330424308776855
212
+ - -0.27141112089157104
213
+ - -0.24878466129302979
214
+ - -0.18733099102973938
215
+ - -0.20597964525222778
216
+ - -0.23352883756160736
217
+ - -0.17950132489204407
218
+ - -0.24911648035049438
219
+ - -0.5155487656593323
220
+ - -0.49901074171066284
221
+ - -0.6347367763519287
222
+ - -0.7135335206985474
223
+ - -0.7645825743675232
224
+ - -0.8687030673027039
225
+ - -0.7830072641372681
226
+ - -0.7961702346801758
227
+ - -0.731216549873352
228
+ - -0.840649425983429
229
+ - -0.7183194756507874
230
+ - -0.663295328617096
231
+ - -1.1195011138916016
232
+ - -2.001723051071167
233
+ spec_min:
234
+ - -6.000000476837158
235
+ - -6.000000476837158
236
+ - -6.000000476837158
237
+ - -6.000000476837158
238
+ - -6.000000476837158
239
+ - -6.000000476837158
240
+ - -6.000000476837158
241
+ - -6.000000476837158
242
+ - -6.000000476837158
243
+ - -6.000000476837158
244
+ - -6.000000476837158
245
+ - -6.000000476837158
246
+ - -6.000000476837158
247
+ - -6.000000476837158
248
+ - -6.000000476837158
249
+ - -6.000000476837158
250
+ - -6.000000476837158
251
+ - -6.000000476837158
252
+ - -6.000000476837158
253
+ - -6.000000476837158
254
+ - -6.000000476837158
255
+ - -6.000000476837158
256
+ - -6.000000476837158
257
+ - -6.000000476837158
258
+ - -6.000000476837158
259
+ - -6.000000476837158
260
+ - -6.000000476837158
261
+ - -6.000000476837158
262
+ - -6.000000476837158
263
+ - -6.000000476837158
264
+ - -6.000000476837158
265
+ - -6.000000476837158
266
+ - -6.000000476837158
267
+ - -6.000000476837158
268
+ - -6.000000476837158
269
+ - -6.000000476837158
270
+ - -6.000000476837158
271
+ - -6.000000476837158
272
+ - -6.000000476837158
273
+ - -6.000000476837158
274
+ - -6.000000476837158
275
+ - -6.000000476837158
276
+ - -6.000000476837158
277
+ - -6.000000476837158
278
+ - -6.000000476837158
279
+ - -6.000000476837158
280
+ - -6.000000476837158
281
+ - -6.000000476837158
282
+ - -6.000000476837158
283
+ - -6.000000476837158
284
+ - -6.000000476837158
285
+ - -6.000000476837158
286
+ - -6.000000476837158
287
+ - -6.000000476837158
288
+ - -6.000000476837158
289
+ - -6.000000476837158
290
+ - -6.000000476837158
291
+ - -6.000000476837158
292
+ - -6.000000476837158
293
+ - -6.000000476837158
294
+ - -6.000000476837158
295
+ - -6.000000476837158
296
+ - -6.000000476837158
297
+ - -6.000000476837158
298
+ - -6.000000476837158
299
+ - -6.000000476837158
300
+ - -6.000000476837158
301
+ - -6.000000476837158
302
+ - -6.000000476837158
303
+ - -6.000000476837158
304
+ - -6.000000476837158
305
+ - -6.000000476837158
306
+ - -6.000000476837158
307
+ - -6.000000476837158
308
+ - -6.000000476837158
309
+ - -6.000000476837158
310
+ - -6.000000476837158
311
+ - -6.000000476837158
312
+ - -6.000000476837158
313
+ - -6.000000476837158
314
+ spk_cond_steps: []
315
+ stop_token_weight: 5.0
316
+ task_cls: training.task.SVC_task.SVCTask
317
+ test_ids: []
318
+ test_input_dir: ''
319
+ test_num: 0
320
+ test_prefixes:
321
+ - test
322
+ test_set_name: test
323
+ timesteps: 1000
324
+ train_set_name: train
325
+ use_crepe: true
326
+ use_denoise: false
327
+ use_energy_embed: false
328
+ use_gt_dur: false
329
+ use_gt_f0: false
330
+ use_midi: false
331
+ use_nsf: true
332
+ use_pitch_embed: true
333
+ use_pos_embed: true
334
+ use_spk_embed: false
335
+ use_spk_id: false
336
+ use_split_spk_id: false
337
+ use_uv: false
338
+ use_var_enc: false
339
+ use_vec: false
340
+ val_check_interval: 2000
341
+ valid_num: 0
342
+ valid_set_name: valid
343
+ vocoder: network.vocoders.hifigan.HifiGAN
344
+ vocoder_ckpt: checkpoints/0109_hifigan_bigpopcs_hop128
345
+ warmup_updates: 2000
346
+ wav2spec_eps: 1e-6
347
+ weight_decay: 0
348
+ win_size: 512
349
+ work_dir: checkpoints/mqnl-covers