_convert_: all
accum_grad: 1
adapter: lora
adapter_conf: {}
allow_multi_rates: false
allow_variable_data_keys: false
aux_ctc_tasks: []
batch_bins: 1000000
batch_size: 20
batch_type: folded
best_model_criterion:
- - valid/acc
  - 1
  - max
bpemodel: ./data/bpe_30/bpe.model
category_sample_size: 10
category_upsampling_factor: 0.5
chunk_default_fs: null
chunk_discard_short_samples: true
chunk_excluded_key_prefixes: []
chunk_length: 500
chunk_max_abs_length: null
chunk_shift_ratio: 0.5
cleaner: null
collect_stats: false
create_dataset:
  archive_path: ./../../egs2/mini_an4/asr1/downloads.tar.gz
  dataset_dir: ./data/mini_an4
  func: src.create_dataset.create_dataset
create_graph_in_tensorboard: false
ctc_conf:
  brctc_group_strategy: end
  brctc_risk_factor: 0.0
  brctc_risk_strategy: exp
  ctc_type: builtin
  dropout_rate: 0.0
  ignore_nan_grad: null
  reduce: true
  zero_infinity: true
cudnn_benchmark: false
cudnn_deterministic: true
cudnn_enabled: true
data_aug_effects:
- - 0.1
  - contrast
  - enhancement_amount: 75.0
- - 0.1
  - highpass
  - Q: 0.707
    cutoff_freq: 5000
- - 0.1
  - equalization
  - Q: 0.707
    center_freq: 1000
    gain: 0
- - 0.1
  - - - 0.3
      - speed_perturb
      - factor: 0.9
    - - 0.3
      - speed_perturb
      - factor: 1.1
    - - 0.3
      - speed_perturb
      - factor: 1.3
data_aug_num:
- 1
- 4
data_aug_prob: 1.0
data_dir: ./data
dataloader:
  collate_fn:
    _target_: espnet2.train.collate_fn.CommonCollateFn
    int_pad_value: -1
  train:
    iter_factory:
      _target_: espnet2.iterators.sequence_iter_factory.SequenceIterFactory
      batches:
        batch_bins: 200000
        batch_size: 2
        shape_files:
        - ./exp/stats/train/feats_shape
        type: sorted
      collate_fn:
        _target_: espnet2.train.collate_fn.CommonCollateFn
        int_pad_value: -1
      num_workers: 0
      shuffle: true
    multiple_iterator: false
    num_shards: 1
  valid:
    iter_factory:
      _target_: espnet2.iterators.sequence_iter_factory.SequenceIterFactory
      batches:
        batch_bins: 200000
        batch_size: 2
        shape_files:
        - ./exp/stats/valid/feats_shape
        type: sorted
      collate_fn:
        _target_: espnet2.train.collate_fn.CommonCollateFn
        int_pad_value: -1
      shuffle: false
    multiple_iterator: false
    num_shards: 1
dataset:
  _target_: espnet3.components.data.data_organizer.DataOrganizer
  train:
  - dataset:
      _target_: src.dataset.MiniAN4Dataset
      manifest_path: ./data/mini_an4/manifest/train_nodev.tsv
    name: train_nodev
  valid:
  - dataset:
      _target_: src.dataset.MiniAN4Dataset
      manifest_path: ./data/mini_an4/manifest/train_dev.tsv
    name: train_dev
dataset_dir: ./data/mini_an4
dataset_scaling_factor: 1.2
dataset_upsampling_factor: 0.5
ddp_comm_hook: null
decode_dir: ./exp/train_asr_rnn_data_aug_debug/decode
decoder: rnn
decoder_conf:
  hidden_size: 2
deepspeed_config: null
detect_anomaly: false
dist_backend: nccl
dist_init_method: env://
dist_launcher: null
dist_master_addr: null
dist_master_port: null
dist_rank: null
dist_world_size: null
drop_last_iter: false
dry_run: false
early_stopping_criterion:
- valid
- loss
- min
encoder: vgg_rnn
encoder_conf:
  hidden_size: 2
  num_layers: 1
  output_size: 2
exclude_weight_decay: false
exclude_weight_decay_conf: {}
exp_dir: ./exp/train_asr_rnn_data_aug_debug
exp_tag: train_asr_rnn_data_aug_debug
fold_length: []
freeze_param: []
frontend: default
frontend_conf:
  hop_length: 160
  n_fft: 512
  win_length: 400
fs: 16000
g2p: null
grad_clip: 5.0
grad_clip_type: 2.0
grad_noise: false
gradient_as_bucket_view: true
ignore_init_mismatch: false
init: null
init_param: []
input_size: null
iterator_type: sequence
joint_net_conf: {}
keep_nbest_models:
- 10
local_rank: null
log_interval: null
log_level: INFO
max_batch_size: null
max_cache_fd: 32
max_cache_size: 0.0
max_epoch: 40
min_batch_size: 1
model: espnet
model_conf:
  ctc_weight: 0.3
  length_normalized_loss: false
  lsm_weight: 0.1
multi_task_dataset: false
multiple_iterator: false
multiprocessing_distributed: false
nbest_averaging_interval: 0
no_forward_run: false
noise_apply_prob: 1.0
noise_db_range: '13_15'
noise_scp: null
non_linguistic_symbols: null
normalize: utterance_mvn
normalize_conf:
  eps: 1.0e-20
  norm_means: true
  norm_vars: false
num_att_plot: 3
num_cache_chunks: 1024
num_device: 1
num_iters_per_epoch: null
num_nodes: 1
num_workers: 1
optim:
  _target_: torch.optim.Adam
  lr: 0.001
  weight_decay: 0.0
optim_conf:
  capturable: false
  differentiable: false
  eps: 1.0e-06
  foreach: null
  lr: 1.0
  maximize: false
  rho: 0.9
  weight_decay: 0
output_dir: null
parallel:
  env: local
  n_workers: 1
  options: {}
patience: null
postencoder: null
postencoder_conf: {}
preencoder: null
preencoder_conf: {}
preprocessor: default
preprocessor_conf:
  audio_pad_value: 0.0
  data_aug_effects: null
  data_aug_num:
  - 1
  - 1
  data_aug_prob: 0.0
  delimiter: null
  force_single_channel: false
  fs: 0
  min_sample_size: -1
  nonsplit_symbol: null
  space_symbol: <space>
  speech_name: speech
  text_name: text
  unk_symbol: <unk>
  whisper_language: null
  whisper_task: null
pretrain_path: null
recipe_dir: .
resume: false
rir_apply_prob: 1.0
rir_scp: null
save_strategy: all
scheduler:
  _target_: torch.optim.lr_scheduler.ReduceLROnPlateau
  factor: 0.5
  mode: min
  patience: 1
scheduler_conf: {}
seed: 0
sharded_ddp: false
short_noise_thres: 0.5
shuffle_within_batch: false
sort_batch: descending
sort_in_batch: descending
specaug: null
specaug_conf: {}
speech_volume_normalize: null
stats_dir: ./exp/stats
task: espnet3.systems.asr.task.ASRTask
token_list: ./data/bpe_30/tokens.txt
token_type: bpe
tokenizer:
  character_coverage: 1.0
  model_type: bpe
  save_path: ./data/bpe_30
  text_builder:
    func: src.tokenizer.gather_training_text
    manifest_path: ./data/mini_an4/manifest/train_nodev.tsv
  vocab_size: 30
train: true
train_data_path_and_name_and_type: []
train_dtype: float32
train_shape_file: []
trainer:
  accumulate_grad_batches: 1
  check_val_every_n_epoch: 1
  devices: 1
  gradient_clip_val: 1.0
  limit_train_batches: 1
  limit_val_batches: 1
  log_every_n_steps: 1
  max_epochs: 1
  num_nodes: 1
  precision: 32
  reload_dataloaders_every_n_epochs: 1
  use_distributed_sampler: false
unused_parameters: false
upsampling_factor: 0.5
use_adapter: false
use_amp: false
use_deepspeed: false
use_lang_prompt: false
use_matplotlib: true
use_nlp_prompt: false
use_preprocessor: true
use_tensorboard: true
use_tf32: false
use_wandb: false
val_scheduler_criterion: valid/loss
valid_batch_bins: null
valid_batch_size: null
valid_batch_type: null
valid_data_path_and_name_and_type: []
valid_iterator_type: null
valid_max_cache_size: null
valid_shape_file: []
vocab_size: 30
wandb_entity: null
wandb_id: null
wandb_model_log_interval: -1
wandb_name: null
wandb_project: null
write_collected_feats: false