Spaces:
Configuration error
Configuration error
Commit
·
e9bd2c2
1
Parent(s):
e22d040
Upload 2 files
Browse files
checkpoints/0109_hifigan_bigpopcs_hop128/config.yaml
ADDED
|
@@ -0,0 +1,241 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
accumulate_grad_batches: 1
|
| 2 |
+
adam_b1: 0.8
|
| 3 |
+
adam_b2: 0.99
|
| 4 |
+
amp: false
|
| 5 |
+
audio_num_mel_bins: 80
|
| 6 |
+
audio_sample_rate: 24000
|
| 7 |
+
aux_context_window: 0
|
| 8 |
+
#base_config:
|
| 9 |
+
#- egs/egs_bases/singing/pwg.yaml
|
| 10 |
+
#- egs/egs_bases/tts/vocoder/hifigan.yaml
|
| 11 |
+
binarization_args:
|
| 12 |
+
reset_phone_dict: true
|
| 13 |
+
reset_word_dict: true
|
| 14 |
+
shuffle: false
|
| 15 |
+
trim_eos_bos: false
|
| 16 |
+
trim_sil: false
|
| 17 |
+
with_align: false
|
| 18 |
+
with_f0: true
|
| 19 |
+
with_f0cwt: false
|
| 20 |
+
with_linear: false
|
| 21 |
+
with_spk_embed: false
|
| 22 |
+
with_spk_id: true
|
| 23 |
+
with_txt: false
|
| 24 |
+
with_wav: true
|
| 25 |
+
with_word: false
|
| 26 |
+
binarizer_cls: data_gen.tts.singing.binarize.SingingBinarizer
|
| 27 |
+
binary_data_dir: data/binary/big_popcs_24k_hop128
|
| 28 |
+
check_val_every_n_epoch: 10
|
| 29 |
+
clip_grad_norm: 1
|
| 30 |
+
clip_grad_value: 0
|
| 31 |
+
datasets: []
|
| 32 |
+
debug: false
|
| 33 |
+
dec_ffn_kernel_size: 9
|
| 34 |
+
dec_layers: 4
|
| 35 |
+
dict_dir: ''
|
| 36 |
+
disc_start_steps: 40000
|
| 37 |
+
discriminator_grad_norm: 1
|
| 38 |
+
discriminator_optimizer_params:
|
| 39 |
+
eps: 1.0e-06
|
| 40 |
+
lr: 0.0002
|
| 41 |
+
weight_decay: 0.0
|
| 42 |
+
discriminator_params:
|
| 43 |
+
bias: true
|
| 44 |
+
conv_channels: 64
|
| 45 |
+
in_channels: 1
|
| 46 |
+
kernel_size: 3
|
| 47 |
+
layers: 10
|
| 48 |
+
nonlinear_activation: LeakyReLU
|
| 49 |
+
nonlinear_activation_params:
|
| 50 |
+
negative_slope: 0.2
|
| 51 |
+
out_channels: 1
|
| 52 |
+
use_weight_norm: true
|
| 53 |
+
discriminator_scheduler_params:
|
| 54 |
+
gamma: 0.999
|
| 55 |
+
step_size: 600
|
| 56 |
+
dropout: 0.1
|
| 57 |
+
ds_workers: 1
|
| 58 |
+
enc_ffn_kernel_size: 9
|
| 59 |
+
enc_layers: 4
|
| 60 |
+
endless_ds: true
|
| 61 |
+
ffn_act: gelu
|
| 62 |
+
ffn_padding: SAME
|
| 63 |
+
fft_size: 512
|
| 64 |
+
fmax: 12000
|
| 65 |
+
fmin: 30
|
| 66 |
+
frames_multiple: 1
|
| 67 |
+
gen_dir_name: ''
|
| 68 |
+
generator_grad_norm: 10
|
| 69 |
+
generator_optimizer_params:
|
| 70 |
+
eps: 1.0e-06
|
| 71 |
+
lr: 0.0002
|
| 72 |
+
weight_decay: 0.0
|
| 73 |
+
generator_params:
|
| 74 |
+
aux_channels: 80
|
| 75 |
+
dropout: 0.0
|
| 76 |
+
gate_channels: 128
|
| 77 |
+
in_channels: 1
|
| 78 |
+
kernel_size: 3
|
| 79 |
+
layers: 30
|
| 80 |
+
out_channels: 1
|
| 81 |
+
residual_channels: 64
|
| 82 |
+
skip_channels: 64
|
| 83 |
+
stacks: 3
|
| 84 |
+
upsample_net: ConvInUpsampleNetwork
|
| 85 |
+
upsample_params:
|
| 86 |
+
upsample_scales:
|
| 87 |
+
- 2
|
| 88 |
+
- 4
|
| 89 |
+
- 4
|
| 90 |
+
- 4
|
| 91 |
+
use_nsf: false
|
| 92 |
+
use_pitch_embed: true
|
| 93 |
+
use_weight_norm: true
|
| 94 |
+
generator_scheduler_params:
|
| 95 |
+
gamma: 0.999
|
| 96 |
+
step_size: 600
|
| 97 |
+
griffin_lim_iters: 60
|
| 98 |
+
hidden_size: 256
|
| 99 |
+
hop_size: 128
|
| 100 |
+
infer: false
|
| 101 |
+
lambda_adv: 1.0
|
| 102 |
+
lambda_cdisc: 4.0
|
| 103 |
+
lambda_energy: 0.0
|
| 104 |
+
lambda_f0: 0.0
|
| 105 |
+
lambda_mel: 5.0
|
| 106 |
+
lambda_mel_adv: 1.0
|
| 107 |
+
lambda_ph_dur: 0.0
|
| 108 |
+
lambda_sent_dur: 0.0
|
| 109 |
+
lambda_uv: 0.0
|
| 110 |
+
lambda_word_dur: 0.0
|
| 111 |
+
load_ckpt: ''
|
| 112 |
+
loud_norm: false
|
| 113 |
+
lr: 2.0
|
| 114 |
+
max_epochs: 1000
|
| 115 |
+
max_frames: 2400
|
| 116 |
+
max_input_tokens: 1550
|
| 117 |
+
max_samples: 8192
|
| 118 |
+
max_sentences: 20
|
| 119 |
+
max_tokens: 24000
|
| 120 |
+
max_updates: 3000000
|
| 121 |
+
max_valid_sentences: 1
|
| 122 |
+
max_valid_tokens: 60000
|
| 123 |
+
mel_loss: ssim:0.5|l1:0.5
|
| 124 |
+
mel_vmax: 1.5
|
| 125 |
+
mel_vmin: -6
|
| 126 |
+
min_frames: 0
|
| 127 |
+
min_level_db: -120
|
| 128 |
+
num_ckpt_keep: 3
|
| 129 |
+
num_heads: 2
|
| 130 |
+
num_mels: 80
|
| 131 |
+
num_sanity_val_steps: 5
|
| 132 |
+
num_spk: 100
|
| 133 |
+
num_test_samples: 0
|
| 134 |
+
num_valid_plots: 10
|
| 135 |
+
optimizer_adam_beta1: 0.9
|
| 136 |
+
optimizer_adam_beta2: 0.98
|
| 137 |
+
out_wav_norm: false
|
| 138 |
+
pitch_extractor: parselmouth
|
| 139 |
+
pitch_type: frame
|
| 140 |
+
pre_align_args:
|
| 141 |
+
allow_no_txt: false
|
| 142 |
+
denoise: false
|
| 143 |
+
sox_resample: true
|
| 144 |
+
sox_to_wav: false
|
| 145 |
+
trim_sil: false
|
| 146 |
+
txt_processor: zh
|
| 147 |
+
use_tone: false
|
| 148 |
+
pre_align_cls: data_gen.tts.singing.pre_align.SingingPreAlign
|
| 149 |
+
predictor_grad: 0.0
|
| 150 |
+
print_nan_grads: false
|
| 151 |
+
processed_data_dir: ''
|
| 152 |
+
profile_infer: false
|
| 153 |
+
raw_data_dir: ''
|
| 154 |
+
ref_level_db: 20
|
| 155 |
+
rename_tmux: true
|
| 156 |
+
rerun_gen: true
|
| 157 |
+
resblock: '1'
|
| 158 |
+
resblock_dilation_sizes:
|
| 159 |
+
- - 1
|
| 160 |
+
- 3
|
| 161 |
+
- 5
|
| 162 |
+
- - 1
|
| 163 |
+
- 3
|
| 164 |
+
- 5
|
| 165 |
+
- - 1
|
| 166 |
+
- 3
|
| 167 |
+
- 5
|
| 168 |
+
resblock_kernel_sizes:
|
| 169 |
+
- 3
|
| 170 |
+
- 7
|
| 171 |
+
- 11
|
| 172 |
+
resume_from_checkpoint: 0
|
| 173 |
+
save_best: true
|
| 174 |
+
save_codes: []
|
| 175 |
+
save_f0: true
|
| 176 |
+
save_gt: true
|
| 177 |
+
scheduler: rsqrt
|
| 178 |
+
seed: 1234
|
| 179 |
+
sort_by_len: true
|
| 180 |
+
stft_loss_params:
|
| 181 |
+
fft_sizes:
|
| 182 |
+
- 1024
|
| 183 |
+
- 2048
|
| 184 |
+
- 512
|
| 185 |
+
hop_sizes:
|
| 186 |
+
- 120
|
| 187 |
+
- 240
|
| 188 |
+
- 50
|
| 189 |
+
win_lengths:
|
| 190 |
+
- 600
|
| 191 |
+
- 1200
|
| 192 |
+
- 240
|
| 193 |
+
window: hann_window
|
| 194 |
+
task_cls: tasks.vocoder.hifigan.HifiGanTask
|
| 195 |
+
tb_log_interval: 100
|
| 196 |
+
test_ids: []
|
| 197 |
+
test_input_dir: ''
|
| 198 |
+
test_num: 50
|
| 199 |
+
test_prefixes: []
|
| 200 |
+
test_set_name: test
|
| 201 |
+
train_set_name: train
|
| 202 |
+
train_sets: ''
|
| 203 |
+
upsample_initial_channel: 512
|
| 204 |
+
upsample_kernel_sizes:
|
| 205 |
+
- 16
|
| 206 |
+
- 16
|
| 207 |
+
- 4
|
| 208 |
+
- 4
|
| 209 |
+
upsample_rates:
|
| 210 |
+
- 8
|
| 211 |
+
- 4
|
| 212 |
+
- 2
|
| 213 |
+
- 2
|
| 214 |
+
use_cdisc: false
|
| 215 |
+
use_cond_disc: false
|
| 216 |
+
use_fm_loss: false
|
| 217 |
+
use_gt_dur: true
|
| 218 |
+
use_gt_f0: true
|
| 219 |
+
use_mel_loss: true
|
| 220 |
+
use_ms_stft: false
|
| 221 |
+
use_pitch_embed: true
|
| 222 |
+
use_ref_enc: true
|
| 223 |
+
use_spec_disc: false
|
| 224 |
+
use_spk_embed: false
|
| 225 |
+
use_spk_id: false
|
| 226 |
+
use_split_spk_id: false
|
| 227 |
+
val_check_interval: 2000
|
| 228 |
+
valid_infer_interval: 10000
|
| 229 |
+
valid_monitor_key: val_loss
|
| 230 |
+
valid_monitor_mode: min
|
| 231 |
+
valid_set_name: valid
|
| 232 |
+
vocoder: pwg
|
| 233 |
+
vocoder_ckpt: ''
|
| 234 |
+
vocoder_denoise_c: 0.0
|
| 235 |
+
warmup_updates: 8000
|
| 236 |
+
weight_decay: 0
|
| 237 |
+
win_length: null
|
| 238 |
+
win_size: 512
|
| 239 |
+
window: hann
|
| 240 |
+
word_size: 3000
|
| 241 |
+
work_dir: checkpoints/0109_hifigan_bigpopcs_hop128
|
checkpoints/0109_hifigan_bigpopcs_hop128/model_ckpt_steps_1512000.ckpt
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:1cb68f3ce0c46ba0a8b6d49718f1fffdf5bd7bcab769a986fd2fd129835cc1d1
|
| 3 |
+
size 55827436
|