Upload phase1-base/config.yaml with huggingface_hub
Browse files- phase1-base/config.yaml +418 -0
phase1-base/config.yaml
ADDED
|
@@ -0,0 +1,418 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
accumulate_grad_batches: 1
|
| 2 |
+
add_word_pos: true
|
| 3 |
+
amp: true
|
| 4 |
+
ast_context_model: MIT/ast-finetuned-audioset-10-10-0.4593
|
| 5 |
+
ast_context_scale: 0.25
|
| 6 |
+
ast_context_valid_prefix: false
|
| 7 |
+
audio_num_mel_bins: 100
|
| 8 |
+
audio_sample_rate: 24000
|
| 9 |
+
base_config:
|
| 10 |
+
- ./agent-1_nonlinear_phone_teacher_5k.yaml
|
| 11 |
+
binarizer_cls: data_gen.tts.base_binarizer.BaseBinarizer
|
| 12 |
+
binary_data_dir: /mnt/DATA/ResearchDrive/audiopaint/data/binary/libritts_bigvgan
|
| 13 |
+
cgpc_gt_view: utt
|
| 14 |
+
cgpc_tau: 0.1
|
| 15 |
+
clip_grad_norm: 1
|
| 16 |
+
clip_grad_value: 0
|
| 17 |
+
cond_dim: 0
|
| 18 |
+
dec_dilations:
|
| 19 |
+
- 1
|
| 20 |
+
- 1
|
| 21 |
+
- 1
|
| 22 |
+
- 1
|
| 23 |
+
dec_ffn_kernel_size: 9
|
| 24 |
+
dec_inp_add_noise: false
|
| 25 |
+
dec_kernel_size: 5
|
| 26 |
+
dec_layers: 4
|
| 27 |
+
dec_post_net_kernel: 3
|
| 28 |
+
decoder_rnn_dim: 0
|
| 29 |
+
decoder_type: conv
|
| 30 |
+
deletion_source_duration_enabled: false
|
| 31 |
+
deletion_source_probability: 0.0
|
| 32 |
+
detach_postflow_input: true
|
| 33 |
+
diff_decoder_type: wavenet
|
| 34 |
+
diff_loss_type: l1
|
| 35 |
+
dilation_cycle_length: 1
|
| 36 |
+
donor_splice_attempts: 24
|
| 37 |
+
donor_splice_exclude_conn_loss: true
|
| 38 |
+
donor_splice_match_prosody: true
|
| 39 |
+
donor_splice_neighbor_search: 128
|
| 40 |
+
donor_splice_operation: insertion
|
| 41 |
+
donor_splice_probability: 0.5
|
| 42 |
+
donor_splice_seam_top_k: 8
|
| 43 |
+
donor_splice_target_phone_max: 9
|
| 44 |
+
donor_splice_target_phone_min: 3
|
| 45 |
+
dropout: 0.0
|
| 46 |
+
ds_name: libritts
|
| 47 |
+
ds_workers: 2
|
| 48 |
+
dur_level: word
|
| 49 |
+
dur_predictor_kernel: 5
|
| 50 |
+
dur_predictor_layers: 3
|
| 51 |
+
enc_dec_norm: ln
|
| 52 |
+
enc_dilations:
|
| 53 |
+
- 1
|
| 54 |
+
- 1
|
| 55 |
+
- 1
|
| 56 |
+
- 1
|
| 57 |
+
enc_ffn_kernel_size: 5
|
| 58 |
+
enc_kernel_size: 5
|
| 59 |
+
enc_layers: 4
|
| 60 |
+
enc_post_net_kernel: 3
|
| 61 |
+
enc_pre_ln: true
|
| 62 |
+
enc_prenet: true
|
| 63 |
+
encoder_K: 8
|
| 64 |
+
encoder_type: conv
|
| 65 |
+
endless_ds: true
|
| 66 |
+
eval_max_batches: -1
|
| 67 |
+
f0_max: 600
|
| 68 |
+
f0_min: 80
|
| 69 |
+
ffn_hidden_size: 768
|
| 70 |
+
fft_size: 1024
|
| 71 |
+
fmax: null
|
| 72 |
+
fmin: 0
|
| 73 |
+
frames_multiple: 1
|
| 74 |
+
gen_dir_name: ''
|
| 75 |
+
gst_seed: 1234
|
| 76 |
+
gst_trainable: false
|
| 77 |
+
hidden_size: 192
|
| 78 |
+
hlac_frame_metric: norm
|
| 79 |
+
hlac_levels: fpw
|
| 80 |
+
hlac_w_frame: 1.0
|
| 81 |
+
hlac_w_phone: 1.0
|
| 82 |
+
hlac_w_word: 1.0
|
| 83 |
+
hop_size: 256
|
| 84 |
+
infer_mask_ratio: 0.3
|
| 85 |
+
keep_bins: 100
|
| 86 |
+
kl_min: 0.0
|
| 87 |
+
kl_start_steps: 10000
|
| 88 |
+
lambda_cgpc: 0.0
|
| 89 |
+
lambda_deletion_source_duration: 0.0
|
| 90 |
+
lambda_energy: 0.1
|
| 91 |
+
lambda_f0: 1.0
|
| 92 |
+
lambda_hlac: 0.0
|
| 93 |
+
lambda_kl: 1.0
|
| 94 |
+
lambda_mel_adv: 0.05
|
| 95 |
+
lambda_nonlinear_phone: 0.01
|
| 96 |
+
lambda_ph_dur: 0.1
|
| 97 |
+
lambda_phone_mel_contrast: 0.05
|
| 98 |
+
lambda_sent_dur: 0.0
|
| 99 |
+
lambda_uv: 1.0
|
| 100 |
+
lambda_word_dur: 1.0
|
| 101 |
+
latent_size: 16
|
| 102 |
+
layers_in_block: 2
|
| 103 |
+
load_ckpt: ''
|
| 104 |
+
load_ckpt_min_match: 0.9
|
| 105 |
+
loud_norm: false
|
| 106 |
+
lr: 2.0e-05
|
| 107 |
+
mask_type: continuous_
|
| 108 |
+
max_beta: 0.06
|
| 109 |
+
max_frames: 1548
|
| 110 |
+
max_input_tokens: 1550
|
| 111 |
+
max_sentences: 16
|
| 112 |
+
max_tokens: 40000
|
| 113 |
+
max_updates: 5000
|
| 114 |
+
max_valid_sentences: 1
|
| 115 |
+
max_valid_tokens: 60000
|
| 116 |
+
mel_disc_hidden_size: 128
|
| 117 |
+
mel_losses: l1:0.5|ssim:0.5
|
| 118 |
+
mel_type: bigvgan
|
| 119 |
+
mel_vmax: 1.5
|
| 120 |
+
mel_vmin: -6
|
| 121 |
+
min_frames: 0
|
| 122 |
+
noise_scale: 0.8
|
| 123 |
+
nonlinear_phone_teacher: assets/agent-1_nonlinear_phone_teacher.pt
|
| 124 |
+
num_ckpt_keep: 12
|
| 125 |
+
num_heads: 2
|
| 126 |
+
num_sanity_val_steps: 5
|
| 127 |
+
num_spk: 1261
|
| 128 |
+
num_valid_plots: 10
|
| 129 |
+
optimizer_adam_beta1: 0.9
|
| 130 |
+
optimizer_adam_beta2: 0.98
|
| 131 |
+
out_wav_norm: false
|
| 132 |
+
phone_mel_prototypes: assets/agent-1_phone_mel_prototypes.npz
|
| 133 |
+
phone_mel_temperature: 0.1
|
| 134 |
+
pitch_extractor: parselmouth
|
| 135 |
+
pitch_type: frame
|
| 136 |
+
posterior_start_steps: 0
|
| 137 |
+
predictor_dropout: 0.2
|
| 138 |
+
predictor_grad: 0.1
|
| 139 |
+
predictor_hidden: -1
|
| 140 |
+
predictor_kernel: 5
|
| 141 |
+
predictor_layers: 5
|
| 142 |
+
print_nan_grads: false
|
| 143 |
+
prior_flow_hidden: 64
|
| 144 |
+
prior_flow_kernel_size: 3
|
| 145 |
+
prior_flow_n_blocks: 4
|
| 146 |
+
processed_data_dir: /mnt/DATA/ResearchDrive/audiopaint/data/processed/libritts
|
| 147 |
+
profile_infer: false
|
| 148 |
+
raw_data_dir: data/raw/libritts
|
| 149 |
+
residual_channels: 256
|
| 150 |
+
residual_layers: 20
|
| 151 |
+
resume_from_checkpoint: 0
|
| 152 |
+
rope_half: true
|
| 153 |
+
rope_reset_at_nt: true
|
| 154 |
+
save_best: false
|
| 155 |
+
save_codes:
|
| 156 |
+
- tasks
|
| 157 |
+
- modules
|
| 158 |
+
save_gt: true
|
| 159 |
+
schedule_type: vpsde
|
| 160 |
+
scheduler: warmup
|
| 161 |
+
seed: 1234
|
| 162 |
+
share_wn_layers: 4
|
| 163 |
+
sigmoid_scale: false
|
| 164 |
+
sort_by_len: true
|
| 165 |
+
spec_max:
|
| 166 |
+
- 1.0604
|
| 167 |
+
- 1.4571
|
| 168 |
+
- 1.2852
|
| 169 |
+
- 1.3722
|
| 170 |
+
- 1.1957
|
| 171 |
+
- 1.6533
|
| 172 |
+
- 1.5299
|
| 173 |
+
- 1.1385
|
| 174 |
+
- 1.2986
|
| 175 |
+
- 1.3136
|
| 176 |
+
- 1.5607
|
| 177 |
+
- 1.3109
|
| 178 |
+
- 1.7511
|
| 179 |
+
- 1.7297
|
| 180 |
+
- 1.5513
|
| 181 |
+
- 1.6957
|
| 182 |
+
- 1.3587
|
| 183 |
+
- 1.8329
|
| 184 |
+
- 1.4253
|
| 185 |
+
- 1.2918
|
| 186 |
+
- 1.4762
|
| 187 |
+
- 1.3411
|
| 188 |
+
- 1.4748
|
| 189 |
+
- 1.3285
|
| 190 |
+
- 1.4044
|
| 191 |
+
- 1.1602
|
| 192 |
+
- 1.4003
|
| 193 |
+
- 1.2423
|
| 194 |
+
- 1.2876
|
| 195 |
+
- 1.0298
|
| 196 |
+
- 1.2514
|
| 197 |
+
- 1.1693
|
| 198 |
+
- 1.2147
|
| 199 |
+
- 1.1453
|
| 200 |
+
- 1.055
|
| 201 |
+
- 1.2303
|
| 202 |
+
- 1.4054
|
| 203 |
+
- 1.0691
|
| 204 |
+
- 1.1672
|
| 205 |
+
- 1.21
|
| 206 |
+
- 0.6927
|
| 207 |
+
- 0.6037
|
| 208 |
+
- 0.9786
|
| 209 |
+
- 0.361
|
| 210 |
+
- 0.7867
|
| 211 |
+
- 0.3277
|
| 212 |
+
- 0.4398
|
| 213 |
+
- 0.5989
|
| 214 |
+
- 0.4387
|
| 215 |
+
- 0.3292
|
| 216 |
+
- 0.1331
|
| 217 |
+
- 0.2449
|
| 218 |
+
- 0.2704
|
| 219 |
+
- -0.1298
|
| 220 |
+
- 0.0984
|
| 221 |
+
- -0.1073
|
| 222 |
+
- -0.1773
|
| 223 |
+
- -0.1167
|
| 224 |
+
- -0.1131
|
| 225 |
+
- -0.1729
|
| 226 |
+
- -0.3771
|
| 227 |
+
- -0.3133
|
| 228 |
+
- -0.5095
|
| 229 |
+
- -0.6342
|
| 230 |
+
- -0.1225
|
| 231 |
+
- -0.0924
|
| 232 |
+
- -0.2996
|
| 233 |
+
- -0.1631
|
| 234 |
+
- -0.4548
|
| 235 |
+
- -0.6299
|
| 236 |
+
- -0.4956
|
| 237 |
+
- -0.2327
|
| 238 |
+
- -0.2719
|
| 239 |
+
- -0.8649
|
| 240 |
+
- -0.4402
|
| 241 |
+
- -0.6376
|
| 242 |
+
- -0.7074
|
| 243 |
+
- -0.552
|
| 244 |
+
- -0.7659
|
| 245 |
+
- -0.7273
|
| 246 |
+
- -0.8326
|
| 247 |
+
- -0.7609
|
| 248 |
+
- -0.9281
|
| 249 |
+
- -0.6764
|
| 250 |
+
- -0.3611
|
| 251 |
+
- -1.0908
|
| 252 |
+
- -1.167
|
| 253 |
+
- -0.9377
|
| 254 |
+
- -0.699
|
| 255 |
+
- -0.7402
|
| 256 |
+
- -1.1614
|
| 257 |
+
- -1.1782
|
| 258 |
+
- -0.6824
|
| 259 |
+
- -0.8936
|
| 260 |
+
- -0.7902
|
| 261 |
+
- -1.1805
|
| 262 |
+
- -1.3148
|
| 263 |
+
- -1.3482
|
| 264 |
+
- -1.61
|
| 265 |
+
- -2.2738
|
| 266 |
+
spec_min:
|
| 267 |
+
- -11.5129
|
| 268 |
+
- -11.5129
|
| 269 |
+
- -11.5129
|
| 270 |
+
- -11.5129
|
| 271 |
+
- -11.5129
|
| 272 |
+
- -11.5129
|
| 273 |
+
- -11.5129
|
| 274 |
+
- -11.5129
|
| 275 |
+
- -11.5129
|
| 276 |
+
- -11.5129
|
| 277 |
+
- -11.5129
|
| 278 |
+
- -11.5129
|
| 279 |
+
- -11.5129
|
| 280 |
+
- -11.5129
|
| 281 |
+
- -11.5129
|
| 282 |
+
- -11.5129
|
| 283 |
+
- -11.5129
|
| 284 |
+
- -11.5129
|
| 285 |
+
- -11.5129
|
| 286 |
+
- -11.5129
|
| 287 |
+
- -11.5129
|
| 288 |
+
- -11.5129
|
| 289 |
+
- -11.5129
|
| 290 |
+
- -11.5129
|
| 291 |
+
- -11.5129
|
| 292 |
+
- -11.5129
|
| 293 |
+
- -11.5129
|
| 294 |
+
- -11.5129
|
| 295 |
+
- -11.5129
|
| 296 |
+
- -11.5129
|
| 297 |
+
- -11.5129
|
| 298 |
+
- -11.5129
|
| 299 |
+
- -11.5129
|
| 300 |
+
- -11.5129
|
| 301 |
+
- -11.5129
|
| 302 |
+
- -11.5129
|
| 303 |
+
- -11.5129
|
| 304 |
+
- -11.5129
|
| 305 |
+
- -11.5129
|
| 306 |
+
- -11.5129
|
| 307 |
+
- -11.5129
|
| 308 |
+
- -11.5129
|
| 309 |
+
- -11.5129
|
| 310 |
+
- -11.5129
|
| 311 |
+
- -11.5129
|
| 312 |
+
- -11.5129
|
| 313 |
+
- -11.5129
|
| 314 |
+
- -11.5129
|
| 315 |
+
- -11.5129
|
| 316 |
+
- -11.5129
|
| 317 |
+
- -11.5129
|
| 318 |
+
- -11.5129
|
| 319 |
+
- -11.5129
|
| 320 |
+
- -11.5129
|
| 321 |
+
- -11.5129
|
| 322 |
+
- -11.5129
|
| 323 |
+
- -11.5129
|
| 324 |
+
- -11.5129
|
| 325 |
+
- -11.5129
|
| 326 |
+
- -11.5129
|
| 327 |
+
- -11.5129
|
| 328 |
+
- -11.5129
|
| 329 |
+
- -11.5129
|
| 330 |
+
- -11.5129
|
| 331 |
+
- -11.5129
|
| 332 |
+
- -11.5129
|
| 333 |
+
- -11.5129
|
| 334 |
+
- -11.5129
|
| 335 |
+
- -11.5129
|
| 336 |
+
- -11.5129
|
| 337 |
+
- -11.5129
|
| 338 |
+
- -11.5129
|
| 339 |
+
- -11.5129
|
| 340 |
+
- -11.5129
|
| 341 |
+
- -11.5129
|
| 342 |
+
- -11.5129
|
| 343 |
+
- -11.5129
|
| 344 |
+
- -11.5129
|
| 345 |
+
- -11.5129
|
| 346 |
+
- -11.5129
|
| 347 |
+
- -11.5129
|
| 348 |
+
- -11.5129
|
| 349 |
+
- -11.5129
|
| 350 |
+
- -11.5129
|
| 351 |
+
- -11.5129
|
| 352 |
+
- -11.5129
|
| 353 |
+
- -11.5129
|
| 354 |
+
- -11.5129
|
| 355 |
+
- -11.5129
|
| 356 |
+
- -11.5129
|
| 357 |
+
- -11.5129
|
| 358 |
+
- -11.5129
|
| 359 |
+
- -11.5129
|
| 360 |
+
- -11.5129
|
| 361 |
+
- -11.5129
|
| 362 |
+
- -11.5129
|
| 363 |
+
- -11.5129
|
| 364 |
+
- -11.5129
|
| 365 |
+
- -11.5129
|
| 366 |
+
- -11.5129
|
| 367 |
+
task_cls: tasks.speech_editing.spec_denoiser.SpeechDenoiserTask
|
| 368 |
+
tb_log_interval: 100
|
| 369 |
+
test_ids: []
|
| 370 |
+
test_set_name: test
|
| 371 |
+
text_encoder_postnet: true
|
| 372 |
+
timescale: 1
|
| 373 |
+
timesteps: 8
|
| 374 |
+
train_set_name: train
|
| 375 |
+
train_sets: ''
|
| 376 |
+
training_mask_ratio: 0.15
|
| 377 |
+
trunk_adaptation: audio_lora
|
| 378 |
+
trunk_donor: Qwen/Qwen2.5-0.5B
|
| 379 |
+
trunk_freeze_tok_emb: true
|
| 380 |
+
trunk_grad_ckpt: true
|
| 381 |
+
trunk_inject: add
|
| 382 |
+
trunk_layer_pick: ends
|
| 383 |
+
trunk_layers: 14
|
| 384 |
+
trunk_lora_alpha: 16.0
|
| 385 |
+
trunk_lora_rank: 8
|
| 386 |
+
trunk_max_text: 48
|
| 387 |
+
trunk_pretrained_lr_scale: 0.05
|
| 388 |
+
trunk_use_ph: true
|
| 389 |
+
trunk_use_text: true
|
| 390 |
+
trunk_warm_start: true
|
| 391 |
+
two_stage: true
|
| 392 |
+
use_ast_context: false
|
| 393 |
+
use_cond_proj: false
|
| 394 |
+
use_conn_loss: true
|
| 395 |
+
use_dualffn_trunk: true
|
| 396 |
+
use_energy_embed: false
|
| 397 |
+
use_gt_dur: false
|
| 398 |
+
use_gt_f0: false
|
| 399 |
+
use_latent_cond: false
|
| 400 |
+
use_perceptual_loss: true
|
| 401 |
+
use_pitch_embed: true
|
| 402 |
+
use_post_flow: true
|
| 403 |
+
use_prior_flow: true
|
| 404 |
+
use_spk_embed: true
|
| 405 |
+
use_spk_id: false
|
| 406 |
+
use_txt_cond: true
|
| 407 |
+
use_uv: true
|
| 408 |
+
val_check_interval: 500
|
| 409 |
+
valid_infer_interval: 2000
|
| 410 |
+
valid_monitor_key: val_loss
|
| 411 |
+
valid_monitor_mode: min
|
| 412 |
+
valid_set_name: valid
|
| 413 |
+
vocoder: BigVGAN
|
| 414 |
+
vocoder_ckpt: pretrained/hifigan_hifitts
|
| 415 |
+
warmup_updates: 1
|
| 416 |
+
weight_decay: 0
|
| 417 |
+
win_size: 1024
|
| 418 |
+
work_dir: /mnt/DATA/ResearchDrive/audiopaint//checkpoints/agent-2_matched_horizon_no_deletion_5k
|