config.yaml

config: conf/soundstream2d_dwconv_mag_phase_16k_n32_600k_step_rmseg_use_power.yaml
print_config: false
log_level: INFO
dry_run: false
iterator_type: sequence
output_dir: exp/soundstream2d_dwconv_mag_phase_16k_n32_600k_step_rmseg_use_power_raw_en_libritts
ngpu: 2
seed: 0
num_workers: 8
num_att_plot: 0
dist_backend: nccl
dist_init_method: env://
dist_world_size: null
dist_rank: null
local_rank: 0
dist_master_addr: null
dist_master_port: null
dist_launcher: null
multiprocessing_distributed: true
unused_parameters: true
sharded_ddp: false
cudnn_enabled: true
cudnn_benchmark: false
cudnn_deterministic: false
collect_stats: false
write_collected_feats: false
max_epoch: 60
max_update: 9223372036854775807
patience: null
val_scheduler_criterion:
- valid
- loss
early_stopping_criterion:
- valid
- loss
- min
best_model_criterion:
-   - valid
    - generator_multi_spectral_recon_loss
    - min
keep_nbest_models: 60
nbest_averaging_interval: 0
grad_clip: -1
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: 50
use_tensorboard: true
use_wandb: false
wandb_project: null
wandb_id: null
wandb_entity: null
wandb_name: null
wandb_model_log_interval: -1
detect_anomaly: false
pretrain_path: null
init_param: []
ignore_init_mismatch: true
freeze_param: []
num_iters_per_epoch: 10000
batch_size: 32
valid_batch_size: null
batch_bins: 2000000
valid_batch_bins: null
drop_last: true
train_shape_file:
- exp/tokenizer_states_16k/train/speech_shape
valid_shape_file:
- exp/tokenizer_states_16k/dev/speech_shape
batch_type: unsorted
valid_batch_type: null
speech_length_min: -1
speech_length_max: -1
fold_length:
- 512
- 150
sort_in_batch: descending
sort_batch: descending
multiple_iterator: false
chunk_length: 500
chunk_shift_ratio: 0.5
num_cache_chunks: 1024
dataset_type: small
dataset_conf: {}
train_data_file: null
valid_data_file: null
train_data_path_and_name_and_type:
-   - dump/raw_16k/train/wav.scp.pai
    - speech
    - kaldi_ark
valid_data_path_and_name_and_type:
-   - dump/raw_16k/dev/wav.scp.pai
    - speech
    - kaldi_ark
allow_variable_data_keys: false
max_cache_size: 0.0
max_cache_fd: 32
valid_max_cache_size: null
optim: adam
optim_conf:
    lr: 0.0003
    betas:
    - 0.5
    - 0.9
scheduler: null
scheduler_conf:
    step_size: 8
    gamma: 0.1
optim2: adam
optim2_conf:
    lr: 0.0003
    betas:
    - 0.5
    - 0.9
scheduler2: null
scheduler2_conf:
    step_size: 8
    gamma: 0.1
simple_ddp: false
num_worker_count: 1
generator_first: false
input_size: 3
cmvn_file: null
disc_grad_clip: -1
disc_grad_clip_type: 2.0
gen_train_interval: 1
disc_train_interval: 1
stat_flops: false
use_preprocessor: true
speech_volume_normalize: null
speech_rms_normalize: false
speech_max_length: 40800
sampling_rate: 16000
valid_max_length: 40800
frontend: null
frontend_conf: {}
normalize: null
normalize_conf: {}
encoder: encodec_seanet_encoder_2d
encoder_conf:
    ratios:
    -   - 4
        - 1
    -   - 4
        - 1
    -   - 4
        - 2
    -   - 4
        - 1
    norm: time_group_norm
    norm_params:
        num_groups: 1
    causal: false
    n_residual_layers: 3
    conv_group_ratio: 1
    dilation_base: 3
    kernel_size: 3
    last_kernel_size: 3
    seq_model: none
quantizer: costume_quantizer
quantizer_conf:
    codebook_size: 1024
    num_quantizers: 32
    ema_decay: 0.99
    kmeans_init: true
    sampling_rate: 16000
    quantize_dropout: true
    rand_num_quant:
    - 1
    - 2
    - 4
    - 8
    - 16
    - 32
    use_ddp: true
    encoder_hop_length: 320
decoder: encodec_seanet_decoder_2d
decoder_conf:
    ratios:
    -   - 4
        - 1
    -   - 4
        - 1
    -   - 4
        - 2
    -   - 4
        - 1
    norm: time_group_norm
    norm_params:
        num_groups: 1
    causal: false
    channels: 3
    dilation_base: 3
    n_residual_layers: 3
    kernel_size: 3
    last_kernel_size: 3
    tr_conv_group_ratio: 1
    conv_group_ratio: 1
    seq_model: none
model: freq_codec
model_conf:
    odim: 128
    multi_spectral_window_powers_of_two:
    - 5
    - 6
    - 7
    - 8
    - 9
    - 10
    target_sample_hz: 16000
    audio_normalize: true
    segment_dur: null
    overlap_ratio: null
    use_power_spec_loss: true
    codec_domain:
    - mag_phase
    - mag_phase
discriminator: multiple_disc
discriminator_conf:
    input_size: 1
    disc_conf_list:
    -   filters: 32
        name: encodec_multi_scale_stft_discriminator
distributed: true
version: 0.2.0