Upload 2 files
Browse files- ns2_38_args.json +271 -0
- ns2_4_args.json +271 -0
ns2_38_args.json
ADDED
|
@@ -0,0 +1,271 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"base_config": "egs/tts/NaturalSpeech2/exp_config_base.json",
|
| 3 |
+
"dataset": [
|
| 4 |
+
"libritts",
|
| 5 |
+
],
|
| 6 |
+
"exp_name": "ns2_wenet_16_3.8_new",
|
| 7 |
+
"log_dir": "ckpts/tts",
|
| 8 |
+
"model": {
|
| 9 |
+
"diffusion": {
|
| 10 |
+
"beta_max": 20,
|
| 11 |
+
"beta_min": 0.05,
|
| 12 |
+
"diffusion_type": "diffusion",
|
| 13 |
+
"noise_factor": 1.0,
|
| 14 |
+
"ode_solver": "euler",
|
| 15 |
+
"sigma": 1.0,
|
| 16 |
+
"wavenet": {
|
| 17 |
+
"attn_head": 8,
|
| 18 |
+
"cross_attn_per_layer": 3,
|
| 19 |
+
"dilation_cycle": 2,
|
| 20 |
+
"drop_out": 0.2,
|
| 21 |
+
"hidden_size": 512,
|
| 22 |
+
"input_size": 128,
|
| 23 |
+
"num_layers": 40,
|
| 24 |
+
"out_size": 128,
|
| 25 |
+
},
|
| 26 |
+
},
|
| 27 |
+
"inference_step": 500,
|
| 28 |
+
"latent_dim": 128,
|
| 29 |
+
"prior_encoder": {
|
| 30 |
+
"duration_predictor": {
|
| 31 |
+
"attn_head": 8,
|
| 32 |
+
"conv_layers": 30,
|
| 33 |
+
"cross_attn_per_layer": 3,
|
| 34 |
+
"drop_out": 0.5,
|
| 35 |
+
"filter_size": 512,
|
| 36 |
+
"input_size": 512,
|
| 37 |
+
"kernel_size": 3,
|
| 38 |
+
},
|
| 39 |
+
"encoder": {
|
| 40 |
+
"conv_filter_size": 2048,
|
| 41 |
+
"conv_kernel_size": 9,
|
| 42 |
+
"encoder_dropout": 0.2,
|
| 43 |
+
"encoder_head": 8,
|
| 44 |
+
"encoder_hidden": 512,
|
| 45 |
+
"encoder_layer": 6,
|
| 46 |
+
"use_cln": true,
|
| 47 |
+
},
|
| 48 |
+
"pitch_bins_num": 512,
|
| 49 |
+
"pitch_max": 1100,
|
| 50 |
+
"pitch_min": 50,
|
| 51 |
+
"pitch_predictor": {
|
| 52 |
+
"attn_head": 8,
|
| 53 |
+
"conv_layers": 30,
|
| 54 |
+
"cross_attn_per_layer": 3,
|
| 55 |
+
"drop_out": 0.5,
|
| 56 |
+
"filter_size": 512,
|
| 57 |
+
"input_size": 512,
|
| 58 |
+
"kernel_size": 5,
|
| 59 |
+
},
|
| 60 |
+
"vocab_size": 100,
|
| 61 |
+
},
|
| 62 |
+
"prompt_encoder": {
|
| 63 |
+
"conv_filter_size": 2048,
|
| 64 |
+
"conv_kernel_size": 9,
|
| 65 |
+
"encoder_dropout": 0.2,
|
| 66 |
+
"encoder_head": 8,
|
| 67 |
+
"encoder_hidden": 512,
|
| 68 |
+
"encoder_layer": 6,
|
| 69 |
+
"use_cln": false,
|
| 70 |
+
},
|
| 71 |
+
"query_emb": {
|
| 72 |
+
"head_num": 8,
|
| 73 |
+
"hidden_size": 512,
|
| 74 |
+
"query_token_num": 32,
|
| 75 |
+
},
|
| 76 |
+
},
|
| 77 |
+
"model_type": "NaturalSpeech2",
|
| 78 |
+
"preprocess": {
|
| 79 |
+
"align_mel_duration": false,
|
| 80 |
+
"audio_dir": "audios",
|
| 81 |
+
"bits": 8,
|
| 82 |
+
"clip_mode": "start",
|
| 83 |
+
"code_dir": "code",
|
| 84 |
+
"contentvec_dir": "contentvec",
|
| 85 |
+
"data_augment": false,
|
| 86 |
+
"dur_dir": "durs",
|
| 87 |
+
"duration_dir": "duration",
|
| 88 |
+
"emo2id": "emo2id.json",
|
| 89 |
+
"energy_dir": "energys",
|
| 90 |
+
"energy_extract_mode": "from_mel",
|
| 91 |
+
"energy_norm": false,
|
| 92 |
+
"energy_remove_outlier": false,
|
| 93 |
+
"extract_acoustic_token": false,
|
| 94 |
+
"extract_amplitude_phase": false,
|
| 95 |
+
"extract_audio": false,
|
| 96 |
+
"extract_contentvec_feature": false,
|
| 97 |
+
"extract_duration": false,
|
| 98 |
+
"extract_energy": false,
|
| 99 |
+
"extract_label": false,
|
| 100 |
+
"extract_linear_spec": false,
|
| 101 |
+
"extract_mcep": false,
|
| 102 |
+
"extract_mel": false,
|
| 103 |
+
"extract_mert_feature": false,
|
| 104 |
+
"extract_phone": false,
|
| 105 |
+
"extract_pitch": false,
|
| 106 |
+
"extract_uv": false,
|
| 107 |
+
"extract_wenet_feature": false,
|
| 108 |
+
"extract_whisper_feature": false,
|
| 109 |
+
"file_lst": "file.lst",
|
| 110 |
+
"fmax": 12000,
|
| 111 |
+
"fmin": 0,
|
| 112 |
+
"hop_size": 120,
|
| 113 |
+
"imaginary_dir": "imaginarys",
|
| 114 |
+
"lab_dir": "labs",
|
| 115 |
+
"label_dir": "labels",
|
| 116 |
+
"lexicon_path": "./text/lexicon/librispeech-lexicon.txt",
|
| 117 |
+
"linear_dir": "linears",
|
| 118 |
+
"log_amplitude_dir": "log_amplitudes",
|
| 119 |
+
"mcep_dir": "mcep",
|
| 120 |
+
"mel_dir": "mels",
|
| 121 |
+
"mel_extract_mode": "",
|
| 122 |
+
"mel_min_max_norm": false,
|
| 123 |
+
"melspec_dir": "mel",
|
| 124 |
+
"metadata_dir": "metadata",
|
| 125 |
+
"min_level_db": -115,
|
| 126 |
+
"n_fft": 1024,
|
| 127 |
+
"n_mel": 80,
|
| 128 |
+
"num_silent_frames": 8,
|
| 129 |
+
"phase_dir": "phases",
|
| 130 |
+
"phone_dir": "phones",
|
| 131 |
+
"phone_energy_dir": "phone_energys",
|
| 132 |
+
"phone_extractor": "espeak",
|
| 133 |
+
"phone_pitch_dir": "phone_pitches",
|
| 134 |
+
"phone_seq_file": "phone_seq_file",
|
| 135 |
+
"pitch_dir": "pitch",
|
| 136 |
+
"pitch_extractor": "parselmouth",
|
| 137 |
+
"pitch_norm": false,
|
| 138 |
+
"pitch_remove_outlier": false,
|
| 139 |
+
"processed_dir": "data",
|
| 140 |
+
"raw_data": "raw_data",
|
| 141 |
+
"read_metadata": true,
|
| 142 |
+
"real_dir": "reals",
|
| 143 |
+
"ref_level_db": 20,
|
| 144 |
+
"sample_rate": 24000,
|
| 145 |
+
"spk2id": "spk2id.json",
|
| 146 |
+
"symbols_dict": "symbols.dict",
|
| 147 |
+
"train_file": "train.json",
|
| 148 |
+
"trim_fft_size": 512,
|
| 149 |
+
"trim_hop_size": 128,
|
| 150 |
+
"trim_silence": false,
|
| 151 |
+
"trim_top_db": 30,
|
| 152 |
+
"trimmed_wav_dir": "trimmed_wavs",
|
| 153 |
+
"use_amplitude_phase": false,
|
| 154 |
+
"use_audio": false,
|
| 155 |
+
"use_code": true,
|
| 156 |
+
"use_cross_reference": true,
|
| 157 |
+
"use_dur": false,
|
| 158 |
+
"use_duration": true,
|
| 159 |
+
"use_emoid": false,
|
| 160 |
+
"use_frame_duration": false,
|
| 161 |
+
"use_frame_energy": false,
|
| 162 |
+
"use_frame_pitch": false,
|
| 163 |
+
"use_lab": false,
|
| 164 |
+
"use_label": false,
|
| 165 |
+
"use_len": true,
|
| 166 |
+
"use_linear": false,
|
| 167 |
+
"use_log_scale_energy": false,
|
| 168 |
+
"use_log_scale_pitch": false,
|
| 169 |
+
"use_mel": false,
|
| 170 |
+
"use_min_max_norm_mel": false,
|
| 171 |
+
"use_one_hot": false,
|
| 172 |
+
"use_phn_seq": false,
|
| 173 |
+
"use_phone": true,
|
| 174 |
+
"use_phone_duration": false,
|
| 175 |
+
"use_phone_energy": false,
|
| 176 |
+
"use_phone_pitch": false,
|
| 177 |
+
"use_pitch": true,
|
| 178 |
+
"use_spkid": true,
|
| 179 |
+
"use_text": false,
|
| 180 |
+
"use_uv": false,
|
| 181 |
+
"use_wav": false,
|
| 182 |
+
"use_wenet": false,
|
| 183 |
+
"utt2emo": "utt2emo",
|
| 184 |
+
"utt2spk": "utt2spk",
|
| 185 |
+
"uv_dir": "uvs",
|
| 186 |
+
"valid_file": "test.json",
|
| 187 |
+
"wav_dir": "wavs",
|
| 188 |
+
"wenet_dir": "wenet",
|
| 189 |
+
"win_size": 480,
|
| 190 |
+
},
|
| 191 |
+
"supported_model_type": [
|
| 192 |
+
"GANVocoder",
|
| 193 |
+
"Fastspeech2",
|
| 194 |
+
"DiffSVC",
|
| 195 |
+
"Transformer",
|
| 196 |
+
"EDM",
|
| 197 |
+
"CD",
|
| 198 |
+
],
|
| 199 |
+
"task_type": "",
|
| 200 |
+
"train": {
|
| 201 |
+
"adam": {
|
| 202 |
+
"lr": 0.0001,
|
| 203 |
+
},
|
| 204 |
+
"adamw": {
|
| 205 |
+
"lr": 0.0004,
|
| 206 |
+
},
|
| 207 |
+
"batch_size": 12,
|
| 208 |
+
"dataloader": {
|
| 209 |
+
"num_worker": 16,
|
| 210 |
+
"pin_memory": true,
|
| 211 |
+
},
|
| 212 |
+
"ddp": true,
|
| 213 |
+
"diff_ce_loss_lambda": 0.5,
|
| 214 |
+
"diff_noise_loss_lambda": 1.0,
|
| 215 |
+
"epochs": 5000,
|
| 216 |
+
"gradient_accumulation_step": 1,
|
| 217 |
+
"keep_checkpoint_max": 100,
|
| 218 |
+
"keep_last": [
|
| 219 |
+
1000,
|
| 220 |
+
],
|
| 221 |
+
"lr_scheduler": "cosine",
|
| 222 |
+
"lr_warmup_steps": 5000,
|
| 223 |
+
"max_epoch": 5000,
|
| 224 |
+
"max_sentences": 32,
|
| 225 |
+
"max_steps": 1000000,
|
| 226 |
+
"max_tokens": 7500,
|
| 227 |
+
"multi_speaker_training": false,
|
| 228 |
+
"num_train_steps": 800000,
|
| 229 |
+
"optimizer": "AdamW",
|
| 230 |
+
"random_seed": 114,
|
| 231 |
+
"reducelronplateau": {
|
| 232 |
+
"factor": 0.8,
|
| 233 |
+
"min_lr": 0.0001,
|
| 234 |
+
"patience": 10,
|
| 235 |
+
},
|
| 236 |
+
"run_eval": [
|
| 237 |
+
true,
|
| 238 |
+
],
|
| 239 |
+
"sampler": {
|
| 240 |
+
"drop_last": true,
|
| 241 |
+
"holistic_shuffle": true,
|
| 242 |
+
},
|
| 243 |
+
"save_checkpoint_stride": [
|
| 244 |
+
1,
|
| 245 |
+
],
|
| 246 |
+
"save_checkpoints_steps": 2000,
|
| 247 |
+
"save_summary_steps": 500,
|
| 248 |
+
"scheduler": "ReduceLROnPlateau",
|
| 249 |
+
"total_training_steps": 800000,
|
| 250 |
+
"tracker": [
|
| 251 |
+
"tensorboard",
|
| 252 |
+
],
|
| 253 |
+
"train_feature_dirs": [
|
| 254 |
+
"/path/labels_with_dur_75",
|
| 255 |
+
"/path/mels_16k_75",
|
| 256 |
+
"/path/mos38_normed_encodec_16",
|
| 257 |
+
"/path/norm_wavs.scp",
|
| 258 |
+
],
|
| 259 |
+
"train_fileid_list_path": "/path/train_pure_3.8.txt",
|
| 260 |
+
"use_dynamic_batchsize": false,
|
| 261 |
+
"valid_feature_dirs": [
|
| 262 |
+
"/path/labels_with_dur_75",
|
| 263 |
+
"/path/mels_16k_75",
|
| 264 |
+
"/path/mos38_normed_encodec_16",
|
| 265 |
+
"/path/norm_wavs.scp",
|
| 266 |
+
],
|
| 267 |
+
"valid_fileid_list_path": "/path/test.txt",
|
| 268 |
+
"valid_interval": 2000,
|
| 269 |
+
},
|
| 270 |
+
"use_custom_dataset": false,
|
| 271 |
+
}
|
ns2_4_args.json
ADDED
|
@@ -0,0 +1,271 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"base_config": "egs/tts/NaturalSpeech2/exp_config_base.json",
|
| 3 |
+
"dataset": [
|
| 4 |
+
"libritts",
|
| 5 |
+
],
|
| 6 |
+
"exp_name": "ns2_wenet_16_mos4_finetune",
|
| 7 |
+
"log_dir": "ckpts/tts",
|
| 8 |
+
"model": {
|
| 9 |
+
"diffusion": {
|
| 10 |
+
"beta_max": 20,
|
| 11 |
+
"beta_min": 0.05,
|
| 12 |
+
"diffusion_type": "diffusion",
|
| 13 |
+
"noise_factor": 1.0,
|
| 14 |
+
"ode_solver": "euler",
|
| 15 |
+
"sigma": 1.0,
|
| 16 |
+
"wavenet": {
|
| 17 |
+
"attn_head": 8,
|
| 18 |
+
"cross_attn_per_layer": 3,
|
| 19 |
+
"dilation_cycle": 2,
|
| 20 |
+
"drop_out": 0.2,
|
| 21 |
+
"hidden_size": 512,
|
| 22 |
+
"input_size": 128,
|
| 23 |
+
"num_layers": 40,
|
| 24 |
+
"out_size": 128,
|
| 25 |
+
},
|
| 26 |
+
},
|
| 27 |
+
"inference_step": 500,
|
| 28 |
+
"latent_dim": 128,
|
| 29 |
+
"prior_encoder": {
|
| 30 |
+
"duration_predictor": {
|
| 31 |
+
"attn_head": 8,
|
| 32 |
+
"conv_layers": 30,
|
| 33 |
+
"cross_attn_per_layer": 3,
|
| 34 |
+
"drop_out": 0.5,
|
| 35 |
+
"filter_size": 512,
|
| 36 |
+
"input_size": 512,
|
| 37 |
+
"kernel_size": 3,
|
| 38 |
+
},
|
| 39 |
+
"encoder": {
|
| 40 |
+
"conv_filter_size": 2048,
|
| 41 |
+
"conv_kernel_size": 9,
|
| 42 |
+
"encoder_dropout": 0.2,
|
| 43 |
+
"encoder_head": 8,
|
| 44 |
+
"encoder_hidden": 512,
|
| 45 |
+
"encoder_layer": 6,
|
| 46 |
+
"use_cln": true,
|
| 47 |
+
},
|
| 48 |
+
"pitch_bins_num": 512,
|
| 49 |
+
"pitch_max": 1100,
|
| 50 |
+
"pitch_min": 50,
|
| 51 |
+
"pitch_predictor": {
|
| 52 |
+
"attn_head": 8,
|
| 53 |
+
"conv_layers": 30,
|
| 54 |
+
"cross_attn_per_layer": 3,
|
| 55 |
+
"drop_out": 0.5,
|
| 56 |
+
"filter_size": 512,
|
| 57 |
+
"input_size": 512,
|
| 58 |
+
"kernel_size": 5,
|
| 59 |
+
},
|
| 60 |
+
"vocab_size": 100,
|
| 61 |
+
},
|
| 62 |
+
"prompt_encoder": {
|
| 63 |
+
"conv_filter_size": 2048,
|
| 64 |
+
"conv_kernel_size": 9,
|
| 65 |
+
"encoder_dropout": 0.2,
|
| 66 |
+
"encoder_head": 8,
|
| 67 |
+
"encoder_hidden": 512,
|
| 68 |
+
"encoder_layer": 6,
|
| 69 |
+
"use_cln": false,
|
| 70 |
+
},
|
| 71 |
+
"query_emb": {
|
| 72 |
+
"head_num": 8,
|
| 73 |
+
"hidden_size": 512,
|
| 74 |
+
"query_token_num": 32,
|
| 75 |
+
},
|
| 76 |
+
},
|
| 77 |
+
"model_type": "NaturalSpeech2",
|
| 78 |
+
"preprocess": {
|
| 79 |
+
"align_mel_duration": false,
|
| 80 |
+
"audio_dir": "audios",
|
| 81 |
+
"bits": 8,
|
| 82 |
+
"clip_mode": "start",
|
| 83 |
+
"code_dir": "code",
|
| 84 |
+
"contentvec_dir": "contentvec",
|
| 85 |
+
"data_augment": false,
|
| 86 |
+
"dur_dir": "durs",
|
| 87 |
+
"duration_dir": "duration",
|
| 88 |
+
"emo2id": "emo2id.json",
|
| 89 |
+
"energy_dir": "energys",
|
| 90 |
+
"energy_extract_mode": "from_mel",
|
| 91 |
+
"energy_norm": false,
|
| 92 |
+
"energy_remove_outlier": false,
|
| 93 |
+
"extract_acoustic_token": false,
|
| 94 |
+
"extract_amplitude_phase": false,
|
| 95 |
+
"extract_audio": false,
|
| 96 |
+
"extract_contentvec_feature": false,
|
| 97 |
+
"extract_duration": false,
|
| 98 |
+
"extract_energy": false,
|
| 99 |
+
"extract_label": false,
|
| 100 |
+
"extract_linear_spec": false,
|
| 101 |
+
"extract_mcep": false,
|
| 102 |
+
"extract_mel": false,
|
| 103 |
+
"extract_mert_feature": false,
|
| 104 |
+
"extract_phone": false,
|
| 105 |
+
"extract_pitch": false,
|
| 106 |
+
"extract_uv": false,
|
| 107 |
+
"extract_wenet_feature": false,
|
| 108 |
+
"extract_whisper_feature": false,
|
| 109 |
+
"file_lst": "file.lst",
|
| 110 |
+
"fmax": 12000,
|
| 111 |
+
"fmin": 0,
|
| 112 |
+
"hop_size": 120,
|
| 113 |
+
"imaginary_dir": "imaginarys",
|
| 114 |
+
"lab_dir": "labs",
|
| 115 |
+
"label_dir": "labels",
|
| 116 |
+
"lexicon_path": "./text/lexicon/librispeech-lexicon.txt",
|
| 117 |
+
"linear_dir": "linears",
|
| 118 |
+
"log_amplitude_dir": "log_amplitudes",
|
| 119 |
+
"mcep_dir": "mcep",
|
| 120 |
+
"mel_dir": "mels",
|
| 121 |
+
"mel_extract_mode": "",
|
| 122 |
+
"mel_min_max_norm": false,
|
| 123 |
+
"melspec_dir": "mel",
|
| 124 |
+
"metadata_dir": "metadata",
|
| 125 |
+
"min_level_db": -115,
|
| 126 |
+
"n_fft": 1024,
|
| 127 |
+
"n_mel": 80,
|
| 128 |
+
"num_silent_frames": 8,
|
| 129 |
+
"phase_dir": "phases",
|
| 130 |
+
"phone_dir": "phones",
|
| 131 |
+
"phone_energy_dir": "phone_energys",
|
| 132 |
+
"phone_extractor": "espeak",
|
| 133 |
+
"phone_pitch_dir": "phone_pitches",
|
| 134 |
+
"phone_seq_file": "phone_seq_file",
|
| 135 |
+
"pitch_dir": "pitch",
|
| 136 |
+
"pitch_extractor": "parselmouth",
|
| 137 |
+
"pitch_norm": false,
|
| 138 |
+
"pitch_remove_outlier": false,
|
| 139 |
+
"processed_dir": "data",
|
| 140 |
+
"raw_data": "raw_data",
|
| 141 |
+
"read_metadata": true,
|
| 142 |
+
"real_dir": "reals",
|
| 143 |
+
"ref_level_db": 20,
|
| 144 |
+
"sample_rate": 24000,
|
| 145 |
+
"spk2id": "spk2id.json",
|
| 146 |
+
"symbols_dict": "symbols.dict",
|
| 147 |
+
"train_file": "train.json",
|
| 148 |
+
"trim_fft_size": 512,
|
| 149 |
+
"trim_hop_size": 128,
|
| 150 |
+
"trim_silence": false,
|
| 151 |
+
"trim_top_db": 30,
|
| 152 |
+
"trimmed_wav_dir": "trimmed_wavs",
|
| 153 |
+
"use_amplitude_phase": false,
|
| 154 |
+
"use_audio": false,
|
| 155 |
+
"use_code": true,
|
| 156 |
+
"use_cross_reference": true,
|
| 157 |
+
"use_dur": false,
|
| 158 |
+
"use_duration": true,
|
| 159 |
+
"use_emoid": false,
|
| 160 |
+
"use_frame_duration": false,
|
| 161 |
+
"use_frame_energy": false,
|
| 162 |
+
"use_frame_pitch": false,
|
| 163 |
+
"use_lab": false,
|
| 164 |
+
"use_label": false,
|
| 165 |
+
"use_len": true,
|
| 166 |
+
"use_linear": false,
|
| 167 |
+
"use_log_scale_energy": false,
|
| 168 |
+
"use_log_scale_pitch": false,
|
| 169 |
+
"use_mel": false,
|
| 170 |
+
"use_min_max_norm_mel": false,
|
| 171 |
+
"use_one_hot": false,
|
| 172 |
+
"use_phn_seq": false,
|
| 173 |
+
"use_phone": true,
|
| 174 |
+
"use_phone_duration": false,
|
| 175 |
+
"use_phone_energy": false,
|
| 176 |
+
"use_phone_pitch": false,
|
| 177 |
+
"use_pitch": true,
|
| 178 |
+
"use_spkid": true,
|
| 179 |
+
"use_text": false,
|
| 180 |
+
"use_uv": false,
|
| 181 |
+
"use_wav": false,
|
| 182 |
+
"use_wenet": false,
|
| 183 |
+
"utt2emo": "utt2emo",
|
| 184 |
+
"utt2spk": "utt2spk",
|
| 185 |
+
"uv_dir": "uvs",
|
| 186 |
+
"valid_file": "test.json",
|
| 187 |
+
"wav_dir": "wavs",
|
| 188 |
+
"wenet_dir": "wenet",
|
| 189 |
+
"win_size": 480,
|
| 190 |
+
},
|
| 191 |
+
"supported_model_type": [
|
| 192 |
+
"GANVocoder",
|
| 193 |
+
"Fastspeech2",
|
| 194 |
+
"DiffSVC",
|
| 195 |
+
"Transformer",
|
| 196 |
+
"EDM",
|
| 197 |
+
"CD",
|
| 198 |
+
],
|
| 199 |
+
"task_type": "",
|
| 200 |
+
"train": {
|
| 201 |
+
"adam": {
|
| 202 |
+
"lr": 0.0001,
|
| 203 |
+
},
|
| 204 |
+
"adamw": {
|
| 205 |
+
"lr": 0.0004,
|
| 206 |
+
},
|
| 207 |
+
"batch_size": 12,
|
| 208 |
+
"dataloader": {
|
| 209 |
+
"num_worker": 16,
|
| 210 |
+
"pin_memory": true,
|
| 211 |
+
},
|
| 212 |
+
"ddp": true,
|
| 213 |
+
"diff_ce_loss_lambda": 0.5,
|
| 214 |
+
"diff_noise_loss_lambda": 1.0,
|
| 215 |
+
"epochs": 5000,
|
| 216 |
+
"gradient_accumulation_step": 1,
|
| 217 |
+
"keep_checkpoint_max": 100,
|
| 218 |
+
"keep_last": [
|
| 219 |
+
1000,
|
| 220 |
+
],
|
| 221 |
+
"lr_scheduler": "cosine",
|
| 222 |
+
"lr_warmup_steps": 5000,
|
| 223 |
+
"max_epoch": 5000,
|
| 224 |
+
"max_sentences": 32,
|
| 225 |
+
"max_steps": 1000000,
|
| 226 |
+
"max_tokens": 7500,
|
| 227 |
+
"multi_speaker_training": false,
|
| 228 |
+
"num_train_steps": 800000,
|
| 229 |
+
"optimizer": "AdamW",
|
| 230 |
+
"random_seed": 114,
|
| 231 |
+
"reducelronplateau": {
|
| 232 |
+
"factor": 0.8,
|
| 233 |
+
"min_lr": 0.0001,
|
| 234 |
+
"patience": 10,
|
| 235 |
+
},
|
| 236 |
+
"run_eval": [
|
| 237 |
+
true,
|
| 238 |
+
],
|
| 239 |
+
"sampler": {
|
| 240 |
+
"drop_last": true,
|
| 241 |
+
"holistic_shuffle": true,
|
| 242 |
+
},
|
| 243 |
+
"save_checkpoint_stride": [
|
| 244 |
+
1,
|
| 245 |
+
],
|
| 246 |
+
"save_checkpoints_steps": 2000,
|
| 247 |
+
"save_summary_steps": 500,
|
| 248 |
+
"scheduler": "ReduceLROnPlateau",
|
| 249 |
+
"total_training_steps": 800000,
|
| 250 |
+
"tracker": [
|
| 251 |
+
"tensorboard",
|
| 252 |
+
],
|
| 253 |
+
"train_feature_dirs": [
|
| 254 |
+
"/path/labels_with_dur_75",
|
| 255 |
+
"/path/mels_16k_75",
|
| 256 |
+
"/path/mos38_normed_encodec_16",
|
| 257 |
+
"/path/norm_wavs.scp",
|
| 258 |
+
],
|
| 259 |
+
"train_fileid_list_path": "/path/train_pure_4.txt",
|
| 260 |
+
"use_dynamic_batchsize": false,
|
| 261 |
+
"valid_feature_dirs": [
|
| 262 |
+
"/path/labels_with_dur_75",
|
| 263 |
+
"/path/mels_16k_75",
|
| 264 |
+
"/path/mos38_normed_encodec_16",
|
| 265 |
+
"/path/norm_wavs.scp",
|
| 266 |
+
],
|
| 267 |
+
"valid_fileid_list_path": "/path/test.txt",
|
| 268 |
+
"valid_interval": 2000,
|
| 269 |
+
},
|
| 270 |
+
"use_custom_dataset": false,
|
| 271 |
+
}
|