Upload config.yaml
Browse files- 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/aqua-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/aqua-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: aqua-covers
|
| 152 |
+
spec_max:
|
| 153 |
+
- 0.30652669072151184
|
| 154 |
+
- 0.08383000642061234
|
| 155 |
+
- 0.1723981499671936
|
| 156 |
+
- 0.3240015506744385
|
| 157 |
+
- 0.39965277910232544
|
| 158 |
+
- 0.44550102949142456
|
| 159 |
+
- 0.38052627444267273
|
| 160 |
+
- 0.32977527379989624
|
| 161 |
+
- 0.3479712903499603
|
| 162 |
+
- 0.34630587697029114
|
| 163 |
+
- 0.3231307864189148
|
| 164 |
+
- 0.38071632385253906
|
| 165 |
+
- 0.4098763167858124
|
| 166 |
+
- 0.4250231385231018
|
| 167 |
+
- 0.47273561358451843
|
| 168 |
+
- 0.4468965232372284
|
| 169 |
+
- 0.4200228452682495
|
| 170 |
+
- 0.3831767439842224
|
| 171 |
+
- 0.37592482566833496
|
| 172 |
+
- 0.3783853352069855
|
| 173 |
+
- 0.4049440622329712
|
| 174 |
+
- 0.37970447540283203
|
| 175 |
+
- 0.40022099018096924
|
| 176 |
+
- 0.41112345457077026
|
| 177 |
+
- 0.46689096093177795
|
| 178 |
+
- 0.4474470615386963
|
| 179 |
+
- 0.39477235078811646
|
| 180 |
+
- 0.36646541953086853
|
| 181 |
+
- 0.3191830515861511
|
| 182 |
+
- 0.33559781312942505
|
| 183 |
+
- 0.3327852189540863
|
| 184 |
+
- 0.34980878233909607
|
| 185 |
+
- 0.3240199685096741
|
| 186 |
+
- 0.3037862181663513
|
| 187 |
+
- 0.29198917746543884
|
| 188 |
+
- 0.36840683221817017
|
| 189 |
+
- 0.2427365928888321
|
| 190 |
+
- 0.2464374154806137
|
| 191 |
+
- 0.29265183210372925
|
| 192 |
+
- 0.2961692810058594
|
| 193 |
+
- 0.1467614769935608
|
| 194 |
+
- 0.0934368446469307
|
| 195 |
+
- 0.06043912097811699
|
| 196 |
+
- 0.05212782323360443
|
| 197 |
+
- 0.058152664452791214
|
| 198 |
+
- 0.10727789998054504
|
| 199 |
+
- 0.037161730229854584
|
| 200 |
+
- 0.01805332489311695
|
| 201 |
+
- 0.014705522917211056
|
| 202 |
+
- -0.005861940328031778
|
| 203 |
+
- -0.00127512332983315
|
| 204 |
+
- -0.10097749531269073
|
| 205 |
+
- -0.09864751249551773
|
| 206 |
+
- -0.0720948874950409
|
| 207 |
+
- -0.12240472435951233
|
| 208 |
+
- -0.22892576456069946
|
| 209 |
+
- -0.19455470144748688
|
| 210 |
+
- -0.2074085921049118
|
| 211 |
+
- -0.16607750952243805
|
| 212 |
+
- -0.3007032871246338
|
| 213 |
+
- -0.22413626313209534
|
| 214 |
+
- -0.36319100856781006
|
| 215 |
+
- -0.2630736529827118
|
| 216 |
+
- -0.4355252683162689
|
| 217 |
+
- -0.4011249542236328
|
| 218 |
+
- -0.4347180128097534
|
| 219 |
+
- -0.36945652961730957
|
| 220 |
+
- -0.39407187700271606
|
| 221 |
+
- -0.47727906703948975
|
| 222 |
+
- -0.45892977714538574
|
| 223 |
+
- -0.34440308809280396
|
| 224 |
+
- -0.4578149616718292
|
| 225 |
+
- -0.5542540550231934
|
| 226 |
+
- -0.2935250997543335
|
| 227 |
+
- -0.3627168536186218
|
| 228 |
+
- -0.5405836701393127
|
| 229 |
+
- -0.6675882935523987
|
| 230 |
+
- -0.7014155983924866
|
| 231 |
+
- -0.9406138062477112
|
| 232 |
+
- -1.8107411861419678
|
| 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 |
+
- -5.8976850509643555
|
| 289 |
+
- -5.985969066619873
|
| 290 |
+
- -5.844642639160156
|
| 291 |
+
- -5.920450687408447
|
| 292 |
+
- -5.921976566314697
|
| 293 |
+
- -5.874476432800293
|
| 294 |
+
- -5.934454441070557
|
| 295 |
+
- -5.875627040863037
|
| 296 |
+
- -5.845851898193359
|
| 297 |
+
- -5.956720352172852
|
| 298 |
+
- -5.952122211456299
|
| 299 |
+
- -5.828412055969238
|
| 300 |
+
- -5.834786891937256
|
| 301 |
+
- -5.867711544036865
|
| 302 |
+
- -5.819331645965576
|
| 303 |
+
- -5.861200332641602
|
| 304 |
+
- -5.8309478759765625
|
| 305 |
+
- -5.837533473968506
|
| 306 |
+
- -5.745879173278809
|
| 307 |
+
- -5.765843868255615
|
| 308 |
+
- -5.7412519454956055
|
| 309 |
+
- -5.756224632263184
|
| 310 |
+
- -5.7770514488220215
|
| 311 |
+
- -5.83022928237915
|
| 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/aqua-covers
|