############################################
#           Network Architecture           #
############################################
cmvn_file_type: "json"
# encoder related
encoder: conformer
encoder_conf:
    output_size: 512    # dimension of attention
    attention_heads: 8
    linear_units: 2048  # the number of units of position-wise feed forward
    num_blocks: 12      # the number of encoder blocks
    dropout_rate: 0.1
    positional_dropout_rate: 0.1
    attention_dropout_rate: 0.0
    input_layer: conv2d # encoder input type, you can chose conv2d, conv2d6 and conv2d8
    normalize_before: True
    use_cnn_module: True
    cnn_module_kernel: 15
    cnn_module_norm: layer_norm
    activation_type: swish
    pos_enc_layer_type: rel_pos
    selfattention_layer_type: rel_selfattn

# decoder related
decoder: transformer
decoder_conf:
    attention_heads: 8
    linear_units: 2048
    num_blocks: 6
    dropout_rate: 0.1
    positional_dropout_rate: 0.1
    self_attention_dropout_rate: 0.0
    src_attention_dropout_rate: 0.0

# hybrid CTC/attention
model_conf:
    ctc_weight: 0.3
    lsm_weight: 0.1     # label smoothing option
    length_normalized_loss: false

# https://yaml.org/type/float.html
###########################################
#                   Data                  #
###########################################
train_manifest: data/train_l/data.list
dev_manifest: data/dev/data.list
test_manifest: data/test_meeting/data.list

###########################################
#              Dataloader                 #
###########################################
use_stream_data: True
unit_type: 'char'
vocab_filepath: data/lang_char/vocab.txt 
cmvn_file: data/mean_std.json
spm_model_prefix: ''
feat_dim: 80
stride_ms: 10.0
window_ms: 25.0
dither: 0.1
sortagrad: 0 # Feed samples from shortest to longest ; -1: enabled for all epochs, 0: disabled, other: enabled for 'other' epochs 
batch_size: 32
minlen_in: 10
maxlen_in: 1200  # if input length(number of frames) > maxlen-in, data is automatically removed
minlen_out: 0
maxlen_out: 150  # if output length(number of tokens) > maxlen-out, data is automatically removed
resample_rate: 16000
shuffle_size: 1500
sort_size: 1000
num_workers: 8
prefetch_factor: 10
dist_sampler: True
num_encs: 1
augment_conf:
  max_w: 80
  w_inplace: True
  w_mode: "PIL"
  max_f: 30
  num_f_mask: 2
  f_inplace: True
  f_replace_with_zero: False
  max_t: 40
  num_t_mask: 2
  t_inplace: True
  t_replace_with_zero: False


###########################################
#                 Training                #
###########################################
n_epoch: 30 
accum_grad: 32
global_grad_clip: 5.0
log_interval: 100
checkpoint:
  kbest_n: 50
  latest_n: 5
optim: adam
optim_conf:
  lr: 0.001
  weight_decay: 1.0e-6
scheduler: warmuplr     
scheduler_conf:
  warmup_steps: 5000
  lr_decay: 1.0