accumulate_grad_batches: 1
add_word_pos: true
amp: false
audio_num_mel_bins: 80
audio_sample_rate: 22050
base_config:
- ./ps_flow.yaml
binarization_args:
  min_sil_duration: 0.1
  shuffle: false
  test_range:
  - 0
  - 523
  train_range:
  - 871
  - -1
  trim_eos_bos: false
  valid_range:
  - 523
  - 871
  with_align: true
  with_f0: true
  with_f0cwt: false
  with_linear: false
  with_spk_embed: false
  with_wav: false
binarizer_cls: data_gen.tts.base_binarizer.BaseBinarizer
binary_data_dir: data/binary/ljspeech
check_val_every_n_epoch: 10
clip_grad_norm: 1
clip_grad_value: 0
conv_use_pos: false
debug: false
dec_dilations:
- 1
- 1
- 1
- 1
dec_ffn_kernel_size: 9
dec_inp_add_noise: false
dec_kernel_size: 5
dec_layers: 4
dec_post_net_kernel: 3
decoder_rnn_dim: 0
decoder_type: conv
detach_postflow_input: true
dropout: 0.0
ds_workers: 2
dur_level: word
dur_predictor_kernel: 5
dur_predictor_layers: 3
enc_dec_norm: ln
enc_dilations:
- 1
- 1
- 1
- 1
enc_ffn_kernel_size: 5
enc_kernel_size: 5
enc_layers: 4
enc_post_net_kernel: 3
enc_pre_ln: false
enc_prenet: true
encoder_K: 8
encoder_type: rel_fft
endless_ds: true
eval_max_batches: -1
f0_max: 600
f0_min: 80
ffn_act: gelu
ffn_hidden_size: 768
fft_size: 1024
fmax: 7600
fmin: 80
frames_multiple: 4
fvae_dec_n_layers: 4
fvae_decoder_type: wn
fvae_enc_dec_hidden: 192
fvae_enc_n_layers: 8
fvae_encoder_type: wn
fvae_kernel_size: 5
fvae_noise_scale: 1.0
fvae_strides: 4
gen_dir_name: ''
griffin_lim_iters: 30
hidden_size: 192
hop_size: 256
infer: false
infer_post_glow: true
kl_min: 0.0
kl_start_steps: 10000
lambda_commit: 0.25
lambda_energy: 0.1
lambda_f0: 1.0
lambda_kl: 1.0
lambda_ph_dur: 0.1
lambda_sent_dur: 0.0
lambda_uv: 1.0
lambda_word_dur: 1.0
latent_size: 16
layers_in_block: 2
load_ckpt: ''
loud_norm: false
lr: 0.0002
max_epochs: 1000
max_frames: 1548
max_input_tokens: 1550
max_sentences: 64
max_tokens: 40000
max_updates: 480000
max_valid_sentences: 1
max_valid_tokens: 60000
mel_losses: l1:0.5|ssim:0.5
mel_vmax: 1.5
mel_vmin: -6
min_frames: 0
noise_scale: 0.8
num_ckpt_keep: 3
num_heads: 2
num_sanity_val_steps: 5
num_spk: 1
num_valid_plots: 10
optimizer_adam_beta1: 0.9
optimizer_adam_beta2: 0.98
out_wav_norm: false
pitch_extractor: parselmouth
pitch_key: pitch
pitch_type: frame
post_flow_lr: 0.001
post_glow_hidden: 192
post_glow_kernel_size: 3
post_glow_n_block_layers: 3
post_glow_n_blocks: 12
post_glow_training_start: 160000
post_share_cond_layers: false
posterior_start_steps: 0
predictor_dropout: 0.2
predictor_grad: 0.1
predictor_hidden: -1
predictor_kernel: 5
predictor_layers: 2
preprocess_args:
  add_eos_bos: true
  mfa_group_shuffle: false
  mfa_offset: 0.02
  nsample_per_mfa_group: 1000
  reset_phone_dict: true
  reset_word_dict: true
  save_sil_mask: true
  txt_processor: en
  use_mfa: true
  vad_max_silence_length: 12
  wav_processors: []
  with_phsep: true
preprocess_cls: egs.datasets.audio.lj.preprocess.LJPreprocess
print_nan_grads: false
prior_flow_hidden: 64
prior_flow_kernel_size: 3
prior_flow_n_blocks: 4
processed_data_dir: data/processed/ljspeech
profile_infer: false
raw_data_dir: data/raw/LJSpeech-1.1
ref_norm_layer: bn
rename_tmux: true
resume_from_checkpoint: 0
save_best: false
save_codes:
- tasks
- modules
- egs
save_f0: false
save_gt: true
scheduler: warmup
seed: 1234
share_wn_layers: 4
sigmoid_scale: false
sort_by_len: true
task_cls: tasks.tts.ps_flow.PortaSpeechFlowTask
tb_log_interval: 100
test_ids:
- 0
- 1
- 2
- 3
- 4
- 5
- 6
- 7
- 8
- 9
- 10
- 11
- 12
- 13
- 14
- 15
- 16
- 17
- 18
- 19
- 68
- 70
- 74
- 87
- 110
- 172
- 190
- 215
- 231
- 294
- 316
- 324
- 402
- 422
- 485
- 500
- 505
- 508
- 509
- 519
test_input_yaml: ''
test_num: 100
test_set_name: test
text_encoder_postnet: false
train_set_name: train
train_sets: ''
two_stage: true
use_cond_proj: false
use_fvae: true
use_gt_dur: false
use_gt_f0: false
use_latent_cond: false
use_pitch_embed: false
use_pos_embed: true
use_post_flow: true
use_prior_flow: true
use_spk_embed: false
use_spk_id: false
use_txt_cond: true
use_uv: true
use_word_encoder: false
use_word_input: false
val_check_interval: 2000
valid_infer_interval: 10000
valid_monitor_key: val_loss
valid_monitor_mode: min
valid_set_name: valid
vocoder: HifiGAN
vocoder_ckpt: checkpoints/hifi_lj
warmup_updates: 8000
weight_decay: 0
win_size: 1024
word_dict_size: 10000
word_enc_layers: 4
word_encoder_type: rel_fft
work_dir: checkpoints/ps_normal_exp