207 lines
4.0 KiB
YAML
207 lines
4.0 KiB
YAML
|
# 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: UniASR
|
||
|
|
model_conf:
|
||
|
|
ctc_weight: 0.0
|
||
|
|
lsm_weight: 0.1
|
||
|
|
length_normalized_loss: true
|
||
|
|
predictor_weight: 1.0
|
||
|
|
decoder_attention_chunk_type: chunk
|
||
|
|
ctc_weight2: 0.0
|
||
|
|
predictor_weight2: 1.0
|
||
|
|
decoder_attention_chunk_type2: chunk
|
||
|
|
loss_weight_model1: 0.5
|
||
|
|
|
||
|
|
|
||
|
|
# encoder
|
||
|
|
encoder: SANMEncoderChunkOpt
|
||
|
|
encoder_conf:
|
||
|
|
output_size: 320
|
||
|
|
attention_heads: 4
|
||
|
|
linear_units: 1280
|
||
|
|
num_blocks: 35
|
||
|
|
dropout_rate: 0.1
|
||
|
|
positional_dropout_rate: 0.1
|
||
|
|
attention_dropout_rate: 0.1
|
||
|
|
input_layer: pe
|
||
|
|
pos_enc_class: SinusoidalPositionEncoder
|
||
|
|
normalize_before: true
|
||
|
|
kernel_size: 11
|
||
|
|
sanm_shfit: 0
|
||
|
|
selfattention_layer_type: sanm
|
||
|
|
chunk_size:
|
||
|
|
- 20
|
||
|
|
- 60
|
||
|
|
stride:
|
||
|
|
- 10
|
||
|
|
- 40
|
||
|
|
pad_left:
|
||
|
|
- 5
|
||
|
|
- 10
|
||
|
|
encoder_att_look_back_factor:
|
||
|
|
- 0
|
||
|
|
- 0
|
||
|
|
decoder_att_look_back_factor:
|
||
|
|
- 0
|
||
|
|
- 0
|
||
|
|
|
||
|
|
# decoder
|
||
|
|
decoder: FsmnDecoderSCAMAOpt
|
||
|
|
decoder_conf:
|
||
|
|
attention_dim: 256
|
||
|
|
attention_heads: 4
|
||
|
|
linear_units: 1024
|
||
|
|
num_blocks: 12
|
||
|
|
dropout_rate: 0.1
|
||
|
|
positional_dropout_rate: 0.1
|
||
|
|
self_attention_dropout_rate: 0.1
|
||
|
|
src_attention_dropout_rate: 0.1
|
||
|
|
att_layer_num: 6
|
||
|
|
kernel_size: 11
|
||
|
|
concat_embeds: true
|
||
|
|
|
||
|
|
# predictor
|
||
|
|
predictor: CifPredictorV2
|
||
|
|
predictor_conf:
|
||
|
|
idim: 320
|
||
|
|
threshold: 1.0
|
||
|
|
l_order: 1
|
||
|
|
r_order: 1
|
||
|
|
|
||
|
|
|
||
|
|
# encoder2
|
||
|
|
encoder2: SANMEncoderChunkOpt
|
||
|
|
encoder2_conf:
|
||
|
|
output_size: 320
|
||
|
|
attention_heads: 4
|
||
|
|
linear_units: 1280
|
||
|
|
num_blocks: 20
|
||
|
|
dropout_rate: 0.1
|
||
|
|
positional_dropout_rate: 0.1
|
||
|
|
attention_dropout_rate: 0.1
|
||
|
|
input_layer: pe
|
||
|
|
pos_enc_class: SinusoidalPositionEncoder
|
||
|
|
normalize_before: true
|
||
|
|
kernel_size: 21
|
||
|
|
sanm_shfit: 0
|
||
|
|
selfattention_layer_type: sanm
|
||
|
|
chunk_size:
|
||
|
|
- 45
|
||
|
|
- 70
|
||
|
|
stride:
|
||
|
|
- 35
|
||
|
|
- 50
|
||
|
|
pad_left:
|
||
|
|
- 5
|
||
|
|
- 10
|
||
|
|
encoder_att_look_back_factor:
|
||
|
|
- 0
|
||
|
|
- 0
|
||
|
|
decoder_att_look_back_factor:
|
||
|
|
- 0
|
||
|
|
- 0
|
||
|
|
|
||
|
|
# decoder
|
||
|
|
decoder2: FsmnDecoderSCAMAOpt
|
||
|
|
decoder2_conf:
|
||
|
|
attention_dim: 320
|
||
|
|
attention_heads: 4
|
||
|
|
linear_units: 1280
|
||
|
|
num_blocks: 12
|
||
|
|
dropout_rate: 0.1
|
||
|
|
positional_dropout_rate: 0.1
|
||
|
|
self_attention_dropout_rate: 0.1
|
||
|
|
src_attention_dropout_rate: 0.1
|
||
|
|
att_layer_num: 6
|
||
|
|
kernel_size: 11
|
||
|
|
concat_embeds: true
|
||
|
|
|
||
|
|
predictor2: CifPredictorV2
|
||
|
|
predictor2_conf:
|
||
|
|
idim: 320
|
||
|
|
threshold: 1.0
|
||
|
|
l_order: 1
|
||
|
|
r_order: 1
|
||
|
|
|
||
|
|
stride_conv: stride_conv1d
|
||
|
|
stride_conv_conf:
|
||
|
|
kernel_size: 2
|
||
|
|
stride: 2
|
||
|
|
pad:
|
||
|
|
- 0
|
||
|
|
- 1
|
||
|
|
|
||
|
|
# frontend related
|
||
|
|
frontend: WavFrontend
|
||
|
|
frontend_conf:
|
||
|
|
fs: 16000
|
||
|
|
window: hamming
|
||
|
|
n_mels: 80
|
||
|
|
frame_length: 25
|
||
|
|
frame_shift: 10
|
||
|
|
lfr_m: 7
|
||
|
|
lfr_n: 6
|
||
|
|
dither: 0.0
|
||
|
|
|
||
|
|
specaug: SpecAugLFR
|
||
|
|
specaug_conf:
|
||
|
|
apply_time_warp: false
|
||
|
|
time_warp_window: 5
|
||
|
|
time_warp_mode: bicubic
|
||
|
|
apply_freq_mask: true
|
||
|
|
freq_mask_width_range:
|
||
|
|
- 0
|
||
|
|
- 30
|
||
|
|
lfr_rate: 6
|
||
|
|
num_freq_mask: 1
|
||
|
|
apply_time_mask: true
|
||
|
|
time_mask_width_range:
|
||
|
|
- 0
|
||
|
|
- 12
|
||
|
|
num_time_mask: 1
|
||
|
|
|
||
|
|
train_conf:
|
||
|
|
accum_grad: 1
|
||
|
|
grad_clip: 5
|
||
|
|
max_epoch: 150
|
||
|
|
keep_nbest_models: 10
|
||
|
|
avg_nbest_model: 10
|
||
|
|
log_interval: 50
|
||
|
|
|
||
|
|
optim: adam
|
||
|
|
optim_conf:
|
||
|
|
lr: 0.0001
|
||
|
|
scheduler: warmuplr
|
||
|
|
scheduler_conf:
|
||
|
|
warmup_steps: 30000
|
||
|
|
|
||
|
|
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: 0
|
||
|
|
|
||
|
|
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
|