DataoceanAI's picture
Upload folder using huggingface_hub
3c37b1b verified
accum_grad: 4
cmvn: global_cmvn
cmvn_conf:
cmvn_file: data/train/global_cmvn
is_json_cmvn: true
context_module: cppn
context_module_conf:
attention_heads: 4
dropout_rate: 0.1
embedding_size: 768
encoder_layers: 2
ctc: ctc
ctc_conf:
ctc_blank_id: 0
dataset: asr
dataset_conf:
batch_conf:
batch_size: 64
batch_type: static
context_conf:
batch_num_context: 200
len_max: 5
len_min: 2
mode: train
utt_num_context: 3
ctc_label: true
cycle: 100
fbank_conf:
dither: 0.1
frame_length: 25
frame_shift: 10
num_mel_bins: 80
filter_conf:
max_length: 3000
min_length: 0
token_max_length: 200
token_min_length: 1
no_time_idx: 3
remove_punctuation: true
remove_timestamp: true
resample_conf:
resample_rate: 16000
shuffle: true
shuffle_conf:
shuffle_size: 1024
sort: true
sort_conf:
sort_size: 1024
spec_aug: true
spec_aug_conf:
max_f: 10
max_t: 50
num_f_mask: 2
num_t_mask: 2
spec_trim: true
spec_trim_conf:
max_t: 30
speed_perturb: false
time_apply_prob: 0.0
decoder: transformer
decoder_conf:
attention_heads: 12
dropout_rate: 0.1
linear_units: 3072
num_blocks: 12
positional_dropout_rate: 0.1
self_attention_dropout_rate: 0.1
src_attention_dropout_rate: 0.1
use_sdpa: true
dtype: fp32
encoder: e_branchformer
encoder_conf:
activation_type: swish
attention_dropout_rate: 0.1
attention_heads: 12
causal: true
cgmlp_conv_kernel: 31
cgmlp_linear_units: 3072
dropout_rate: 0.1
gate_activation: identity
input_layer: conv2d
linear_units: 3072
merge_conv_kernel: 31
num_blocks: 12
output_size: 768
pos_enc_layer_type: rel_pos
positional_dropout_rate: 0.1
selfattention_layer_type: rel_selfattn
use_dynamic_chunk: true
use_dynamic_left_chunk: false
use_linear_after_conv: false
use_sdpa: true
grad_clip: 5
input_dim: 80
log_interval: 20
max_epoch: 100
model: asr_model
model_conf:
ctc_weight: 0.3
length_normalized_loss: false
lsm_weight: 0.1
model_dir: exp/dolphin_ebf_small_streaming_bias_share_ctc_sample
optim: adam
optim_conf:
lr: 0.0005
output_dim: 18173
save_interval: 200
save_states: model_only
scheduler: warmuplr
scheduler_conf:
warmup_steps: 2048
stats_dialect: true
tokenizer: char
tokenizer_conf:
special_tokens:
<asr>: 4
<blank>: 0
<eos>: 3
<sos>: 2
<unk>: 1
split_with_space: false
symbol_table_path: data/dict/units.txt
train_engine: torch_ddp
use_amp: false
vocab_size: 18173