mitudesk commited on
Commit
738692f
1 Parent(s): 21719ef

Upload 2 files

Browse files
Files changed (2) hide show
  1. models/uma/config.yaml +349 -0
  2. models/uma/teio.ckpt +3 -0
models/uma/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/teio
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: 21000
83
+ max_input_tokens: 60000
84
+ max_sentences: 16
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/teio
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: teio
152
+ spec_max:
153
+ - -1.5451143980026245
154
+ - -1.5177826881408691
155
+ - -1.1807013750076294
156
+ - -0.6732071042060852
157
+ - -0.47006210684776306
158
+ - -0.271837055683136
159
+ - -0.27174147963523865
160
+ - -0.3395537734031677
161
+ - -0.2529868483543396
162
+ - -0.22453370690345764
163
+ - -0.24767500162124634
164
+ - -0.22861438989639282
165
+ - -0.28668588399887085
166
+ - -0.335957795381546
167
+ - -0.3118636906147003
168
+ - -0.34530898928642273
169
+ - -0.35274678468704224
170
+ - -0.4730182886123657
171
+ - -0.45395755767822266
172
+ - -0.4338522255420685
173
+ - -0.41395917534828186
174
+ - -0.29468369483947754
175
+ - -0.16852207481861115
176
+ - -0.3900046646595001
177
+ - -0.6241626739501953
178
+ - -0.5899035930633545
179
+ - -0.6534764170646667
180
+ - -0.6667397022247314
181
+ - -0.6992383003234863
182
+ - -0.7867978811264038
183
+ - -0.8457596302032471
184
+ - -0.43350857496261597
185
+ - -0.629216730594635
186
+ - -0.9135912656784058
187
+ - -0.9230040311813354
188
+ - -0.6756577491760254
189
+ - -0.8399246335029602
190
+ - -0.8495144248008728
191
+ - -0.781493067741394
192
+ - -1.0347247123718262
193
+ - -1.0051935911178589
194
+ - -1.1246198415756226
195
+ - -1.021154522895813
196
+ - -0.851677417755127
197
+ - -0.8443652987480164
198
+ - -0.9016147255897522
199
+ - -0.7618780732154846
200
+ - -1.0490750074386597
201
+ - -1.2046996355056763
202
+ - -1.2022035121917725
203
+ - -0.9753153324127197
204
+ - -1.2503044605255127
205
+ - -1.0664823055267334
206
+ - -1.1236635446548462
207
+ - -1.2223032712936401
208
+ - -1.0116488933563232
209
+ - -1.2263423204421997
210
+ - -1.2552075386047363
211
+ - -1.3846945762634277
212
+ - -1.2681812047958374
213
+ - -1.3416036367416382
214
+ - -1.264938235282898
215
+ - -1.2763726711273193
216
+ - -1.4651004076004028
217
+ - -1.4880361557006836
218
+ - -1.5735552310943604
219
+ - -1.4097294807434082
220
+ - -1.468385100364685
221
+ - -1.3768259286880493
222
+ - -1.3312186002731323
223
+ - -1.3547866344451904
224
+ - -1.4387739896774292
225
+ - -1.1861546039581299
226
+ - -1.1709729433059692
227
+ - -1.1812609434127808
228
+ - -1.1489264965057373
229
+ - -1.5605546236038208
230
+ - -2.2702553272247314
231
+ - -4.064557075500488
232
+ - -5.809507846832275
233
+ spec_min:
234
+ - -5.740882873535156
235
+ - -6.0
236
+ - -6.0
237
+ - -6.0
238
+ - -6.0
239
+ - -6.0
240
+ - -6.0
241
+ - -6.0
242
+ - -5.959110260009766
243
+ - -6.0
244
+ - -6.0
245
+ - -6.0
246
+ - -6.0
247
+ - -6.0
248
+ - -6.0
249
+ - -6.0
250
+ - -6.0
251
+ - -6.0
252
+ - -6.0
253
+ - -6.0
254
+ - -6.0
255
+ - -6.0
256
+ - -6.0
257
+ - -6.0
258
+ - -6.0
259
+ - -6.0
260
+ - -6.0
261
+ - -6.0
262
+ - -6.0
263
+ - -6.0
264
+ - -6.0
265
+ - -5.999546527862549
266
+ - -6.0
267
+ - -5.995517730712891
268
+ - -6.0
269
+ - -6.0
270
+ - -6.0
271
+ - -6.0
272
+ - -6.0
273
+ - -6.0
274
+ - -6.0
275
+ - -5.960205078125
276
+ - -5.93423318862915
277
+ - -6.0
278
+ - -5.933608531951904
279
+ - -6.0
280
+ - -6.0
281
+ - -6.0
282
+ - -5.953958511352539
283
+ - -5.908934593200684
284
+ - -5.911312580108643
285
+ - -5.882552623748779
286
+ - -5.932425498962402
287
+ - -5.91495943069458
288
+ - -5.826524257659912
289
+ - -5.777952671051025
290
+ - -5.775007724761963
291
+ - -5.849961280822754
292
+ - -5.7793660163879395
293
+ - -5.781087875366211
294
+ - -5.818603992462158
295
+ - -5.765895366668701
296
+ - -5.834509372711182
297
+ - -5.817623615264893
298
+ - -5.855445384979248
299
+ - -5.844409465789795
300
+ - -5.760529518127441
301
+ - -5.713063716888428
302
+ - -5.74588680267334
303
+ - -5.855954647064209
304
+ - -5.874588489532471
305
+ - -5.81571626663208
306
+ - -5.849369049072266
307
+ - -5.963766574859619
308
+ - -5.8541646003723145
309
+ - -5.922942161560059
310
+ - -6.0
311
+ - -6.0
312
+ - -6.0
313
+ - -6.0
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/teio
models/uma/teio.ckpt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:162c23e83526cbc5675d37edf43e71ca54f85ac020e7af36e802e8f0f8b65b1a
3
+ size 188902673