|
|
|
|
# https://yaml.org/type/float.html
|
|
|
|
|
data:
|
|
|
|
|
train_manifest: data/manifest.tiny
|
|
|
|
|
dev_manifest: data/manifest.tiny
|
|
|
|
|
test_manifest: data/manifest.tiny
|
|
|
|
|
min_input_len: 0.0
|
|
|
|
|
max_input_len: 30.0
|
|
|
|
|
min_output_len: 0.0
|
|
|
|
|
max_output_len: 400.0
|
|
|
|
|
min_output_input_ratio: 0.05
|
|
|
|
|
max_output_input_ratio: 10.0
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
collator:
|
|
|
|
|
mean_std_filepath: data/mean_std.json
|
|
|
|
|
unit_type: char
|
|
|
|
|
vocab_filepath: data/lang_char/vocab.txt
|
|
|
|
|
augmentation_config: conf/augmentation.json
|
|
|
|
|
random_seed: 0
|
|
|
|
|
spm_model_prefix:
|
|
|
|
|
spectrum_type: linear
|
|
|
|
|
feat_dim:
|
|
|
|
|
delta_delta: False
|
|
|
|
|
stride_ms: 10.0
|
|
|
|
|
window_ms: 20.0
|
|
|
|
|
n_fft: None
|
|
|
|
|
max_freq: None
|
|
|
|
|
target_sample_rate: 16000
|
|
|
|
|
use_dB_normalization: True
|
|
|
|
|
target_dB: -20
|
|
|
|
|
dither: 1.0
|
|
|
|
|
keep_transcription_text: False
|
|
|
|
|
sortagrad: True
|
|
|
|
|
shuffle_method: batch_shuffle
|
|
|
|
|
num_workers: 0
|
|
|
|
|
batch_size: 4
|
|
|
|
|
|
|
|
|
|
model:
|
|
|
|
|
num_conv_layers: 2
|
|
|
|
|
num_rnn_layers: 4
|
|
|
|
|
rnn_layer_size: 2048
|
|
|
|
|
rnn_direction: forward
|
|
|
|
|
num_fc_layers: 2
|
|
|
|
|
fc_layers_size_list: 512, 256
|
|
|
|
|
use_gru: True
|
|
|
|
|
blank_id: 0
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
training:
|
|
|
|
|
n_epoch: 5
|
|
|
|
|
accum_grad: 1
|
|
|
|
|
lr: 1e-5
|
|
|
|
|
lr_decay: 1.0
|
|
|
|
|
weight_decay: 1e-06
|
|
|
|
|
global_grad_clip: 5.0
|
|
|
|
|
log_interval: 1
|
|
|
|
|
checkpoint:
|
|
|
|
|
kbest_n: 3
|
|
|
|
|
latest_n: 2
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
decoding:
|
|
|
|
|
batch_size: 128
|
|
|
|
|
error_rate_type: wer
|
|
|
|
|
decoding_method: ctc_beam_search
|
|
|
|
|
lang_model_path: data/lm/common_crawl_00.prune01111.trie.klm
|
|
|
|
|
alpha: 2.5
|
|
|
|
|
beta: 0.3
|
|
|
|
|
beam_size: 500
|
|
|
|
|
cutoff_prob: 1.0
|
|
|
|
|
cutoff_top_n: 40
|
|
|
|
|
num_proc_bsearch: 8
|
