config: conf/tuning/prodiff_gst_xvector_base.yaml
print_config: false
log_level: INFO
dry_run: false
iterator_type: sequence
output_dir: exp/tts_prodiff_gst_xvector_base_raw_phn_none
ngpu: 1
seed: 0
num_workers: 6
num_att_plot: 3
num_valid_artifacts: 5
dist_backend: nccl
dist_init_method: env://
dist_world_size: 2
dist_rank: 0
local_rank: 0
dist_master_addr: localhost
dist_master_port: 32945
dist_launcher: null
multiprocessing_distributed: true
unused_parameters: false
sharded_ddp: false
growth_interval: 0
min_grad_scale: -1
cudnn_enabled: true
cudnn_benchmark: false
cudnn_deterministic: true
collect_stats: false
write_collected_feats: false
max_epoch: 600
patience: null
val_scheduler_criterion:
- valid
- loss
early_stopping_criterion:
- valid
- loss
- min
best_model_criterion:
-   - valid
    - loss
    - min
-   - train
    - loss
    - min
keep_nbest_models: 5
nbest_averaging_interval: 0
grad_clip: 1.0
grad_clip_type: 2.0
grad_noise: false
accum_grad: 1
no_forward_run: false
resume: true
train_dtype: float32
use_amp: false
log_interval: null
use_matplotlib: true
use_tensorboard: true
detect_anomaly: false
pretrain_path: null
init_param: []
ignore_init_mismatch: false
freeze_param: []
num_iters_per_epoch: 250
batch_size: 20
valid_batch_size: null
valid_num_batches: null
batch_bins: 6000000
valid_batch_bins: null
train_shape_file:
- exp/tts_stats_raw_phn_none/train/text_shape.phn
- exp/tts_stats_raw_phn_none/train/speech_shape
valid_shape_file:
- exp/tts_stats_raw_phn_none/valid/text_shape.phn
- exp/tts_stats_raw_phn_none/valid/speech_shape
batch_type: numel
valid_batch_type: null
fold_length:
- 150
- 240000
sort_in_batch: descending
sort_batch: descending
multiple_iterator: false
chunk_length: 500
chunk_shift_ratio: 0.5
num_cache_chunks: 1024
train_data_path_and_name_and_type:
-   - dump/raw/tr_no_dev_phn/text
    - text
    - text
-   - data/tr_no_dev_phn/durations
    - durations
    - text_int
-   - dump/raw/tr_no_dev_phn/wav.scp
    - speech
    - sound
-   - exp/tts_stats_raw_phn_none/train/collect_feats/pitch.scp
    - pitch
    - npy
-   - exp/tts_stats_raw_phn_none/train/collect_feats/energy.scp
    - energy
    - npy
-   - xvector/tr_no_dev_phn/xvector.scp
    - spembs
    - kaldi_ark
valid_data_path_and_name_and_type:
-   - dump/raw/dev_phn/text
    - text
    - text
-   - data/dev_phn/durations
    - durations
    - text_int
-   - dump/raw/dev_phn/wav.scp
    - speech
    - sound
-   - exp/tts_stats_raw_phn_none/valid/collect_feats/pitch.scp
    - pitch
    - npy
-   - exp/tts_stats_raw_phn_none/valid/collect_feats/energy.scp
    - energy
    - npy
-   -xvector/dev_phn/xvector.scp
    - spembs
    - kaldi_ark
allow_variable_data_keys: false
max_cache_size: 0.0
max_cache_fd: 32
valid_max_cache_size: null
optim: adamw
optim_conf:
    lr: 1.0
    betas:
    - 0.9
    - 0.98
scheduler: noamlr
scheduler_conf:
    model_size: 384
    warmup_steps: 2000
token_list:
- <blank>
- <unk>
- o
- a
- u
- i
- e
- k
- r
- t
- n
- 、
- N
- s
- sh
- d
- m
- g
- b
- w
- cl
- j
- ch
- sil
- h
- y
- p
- ts
- z
- f
- ky
- U
- ny
- gy
- ry
- I
- hy
- my
- by
- py
- v
- <sos/eos>
odim: null
model_conf:
    requires_word_duration: false
use_preprocessor: true
token_type: phn
bpemodel: null
non_linguistic_symbols: null
cleaner: null
g2p: null
feats_extract: fbank
feats_extract_conf:
    n_fft: 2048
    hop_length: 300
    win_length: 1200
    fs: 24000
    fmin: 80
    fmax: 7600
    n_mels: 80
normalize: global_mvn
normalize_conf:
    stats_file: stats/feats_stats.npz
tts: prodiff
tts_conf:
    adim: 384
    aheads: 2
    elayers: 4
    eunits: 1536
    positionwise_layer_type: conv1d-linear
    positionwise_conv_kernel_size: 9
    use_masking: true
    use_scaled_pos_enc: true
    encoder_normalize_before: true
    reduction_factor: 1
    init_type: xavier_uniform
    init_enc_alpha: 1.0
    transformer_enc_dropout_rate: 0.05
    transformer_enc_positional_dropout_rate: 0.05
    transformer_enc_attn_dropout_rate: 0.05
    duration_predictor_layers: 2
    duration_predictor_chans: 512
    duration_predictor_kernel_size: 3
    pitch_predictor_layers: 2
    pitch_predictor_chans: 512
    pitch_predictor_kernel_size: 3
    pitch_predictor_dropout: 0.5
    pitch_embed_kernel_size: 1
    pitch_embed_dropout: 0.0
    stop_gradient_from_pitch_predictor: true
    energy_predictor_layers: 2
    energy_predictor_chans: 512
    energy_predictor_kernel_size: 3
    energy_predictor_dropout: 0.5
    energy_embed_kernel_size: 1
    energy_embed_dropout: 0.0
    stop_gradient_from_energy_predictor: false
    spks: -1
    spk_embed_dim: 192
    denoiser_layers: 20
    denoiser_channels: 256
    diffusion_steps: 4
    diffusion_timescale: 1
    diffusion_beta: 40.0
    diffusion_scheduler: vpsde
    diffusion_cycle_ln: 1
    use_gst: true
    gst_heads: 8
    gst_tokens: 256
pitch_extract: dio
pitch_extract_conf:
    fs: 24000
    n_fft: 2048
    hop_length: 300
    f0max: 400
    f0min: 80
    reduction_factor: 1
pitch_normalize: global_mvn
pitch_normalize_conf:
    stats_file: stats/pitch_stats.npz
energy_extract: energy
energy_extract_conf:
    fs: 24000
    n_fft: 2048
    hop_length: 300
    win_length: 1200
    reduction_factor: 1
energy_normalize: global_mvn
energy_normalize_conf:
    stats_file: stats/energy_stats.npz
required:
- output_dir
- token_list
version: '202207'
distributed: true