Skip to content

Open-source speech recognition toolkit for training, inference, streaming ASR, VAD, punctuation, speaker diarization pipelines, and OpenAI-compatible/MCP serving.

2.6 KB
Raw Blame
# This is an example that demonstrates how to configure a model file.
# You can modify the configuration according to your own requirements.

# to print the register_table:
# from funasr.register import tables
# tables.print()

# network architecture
model: Branchformer
model_conf:
    ctc_weight: 0.3
    lsm_weight: 0.1     # label smoothing option
    length_normalized_loss: false

# encoder
encoder: BranchformerEncoder
encoder_conf:
    output_size: 256
    use_attn: true
    attention_heads: 4
    attention_layer_type: rel_selfattn
    pos_enc_layer_type: rel_pos
    rel_pos_type: latest
    use_cgmlp: true
    cgmlp_linear_units: 2048
    cgmlp_conv_kernel: 31
    use_linear_after_conv: false
    gate_activation: identity
    merge_method: concat
    cgmlp_weight: 0.5               # used only if merge_method is "fixed_ave"
    attn_branch_drop_rate: 0.0      # used only if merge_method is "learned_ave"
    num_blocks: 24
    dropout_rate: 0.1
    positional_dropout_rate: 0.1
    attention_dropout_rate: 0.1
    input_layer: conv2d
    stochastic_depth_rate: 0.0

# decoder
decoder: TransformerDecoder
decoder_conf:
    attention_heads: 4
    linear_units: 2048
    num_blocks: 6
    dropout_rate: 0.1
    positional_dropout_rate: 0.1
    self_attention_dropout_rate: 0.
    src_attention_dropout_rate: 0.


# frontend related
frontend: WavFrontend
frontend_conf:
    fs: 16000
    window: hamming
    n_mels: 80
    frame_length: 25
    frame_shift: 10
    dither: 0.0
    lfr_m: 1
    lfr_n: 1

specaug: SpecAug
specaug_conf:
    apply_time_warp: true
    time_warp_window: 5
    time_warp_mode: bicubic
    apply_freq_mask: true
    freq_mask_width_range:
    - 0
    - 30
    num_freq_mask: 2
    apply_time_mask: true
    time_mask_width_range:
    - 0
    - 40
    num_time_mask: 2

train_conf:
  accum_grad: 1
  grad_clip: 5
  max_epoch: 150
  keep_nbest_models: 10
  log_interval: 50

optim: adam
optim_conf:
   lr: 0.001
   weight_decay: 0.000001
scheduler: warmuplr
scheduler_conf:
   warmup_steps: 35000

dataset: AudioDataset
dataset_conf:
    index_ds: IndexDSJsonl
    batch_sampler: BatchSampler
    batch_type: example # example or length
    batch_size: 1 # if batch_type is example, batch_size is the numbers of samples; if length, batch_size is source_token_len+target_token_len;
    max_token_length: 2048 # filter samples if source_token_len+target_token_len > max_token_length,
    buffer_size: 500
    shuffle: True
    num_workers: 4

tokenizer: CharTokenizer
tokenizer_conf:
  unk_symbol: <unk>
  split_with_space: true


ctc_conf:
    dropout_rate: 0.0
    ctc_type: builtin
    reduce: true
    ignore_nan_grad: true
normalize: null