Upload config.yaml with huggingface_hub
Browse files- config.yaml +97 -0
config.yaml
ADDED
@@ -0,0 +1,97 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
encoder: SenseVoiceEncoderSmall
|
2 |
+
encoder_conf:
|
3 |
+
output_size: 512
|
4 |
+
attention_heads: 4
|
5 |
+
linear_units: 2048
|
6 |
+
num_blocks: 50
|
7 |
+
tp_blocks: 20
|
8 |
+
dropout_rate: 0.1
|
9 |
+
positional_dropout_rate: 0.1
|
10 |
+
attention_dropout_rate: 0.1
|
11 |
+
input_layer: pe
|
12 |
+
pos_enc_class: SinusoidalPositionEncoder
|
13 |
+
normalize_before: true
|
14 |
+
kernel_size: 11
|
15 |
+
sanm_shfit: 0
|
16 |
+
selfattention_layer_type: sanm
|
17 |
+
|
18 |
+
|
19 |
+
model: SenseVoiceSmall
|
20 |
+
model_conf:
|
21 |
+
length_normalized_loss: true
|
22 |
+
sos: 1
|
23 |
+
eos: 2
|
24 |
+
ignore_id: -1
|
25 |
+
|
26 |
+
tokenizer: SentencepiecesTokenizer
|
27 |
+
tokenizer_conf:
|
28 |
+
bpemodel: null
|
29 |
+
unk_symbol: <unk>
|
30 |
+
split_with_space: true
|
31 |
+
|
32 |
+
frontend: WavFrontend
|
33 |
+
frontend_conf:
|
34 |
+
fs: 16000
|
35 |
+
window: hamming
|
36 |
+
n_mels: 80
|
37 |
+
frame_length: 25
|
38 |
+
frame_shift: 10
|
39 |
+
lfr_m: 7
|
40 |
+
lfr_n: 6
|
41 |
+
cmvn_file: null
|
42 |
+
|
43 |
+
|
44 |
+
dataset: SenseVoiceCTCDataset
|
45 |
+
dataset_conf:
|
46 |
+
index_ds: IndexDSJsonl
|
47 |
+
batch_sampler: EspnetStyleBatchSampler
|
48 |
+
data_split_num: 32
|
49 |
+
batch_type: token
|
50 |
+
batch_size: 14000
|
51 |
+
max_token_length: 2000
|
52 |
+
min_token_length: 60
|
53 |
+
max_source_length: 2000
|
54 |
+
min_source_length: 60
|
55 |
+
max_target_length: 200
|
56 |
+
min_target_length: 0
|
57 |
+
shuffle: true
|
58 |
+
num_workers: 4
|
59 |
+
sos: ${model_conf.sos}
|
60 |
+
eos: ${model_conf.eos}
|
61 |
+
IndexDSJsonl: IndexDSJsonl
|
62 |
+
retry: 20
|
63 |
+
|
64 |
+
train_conf:
|
65 |
+
accum_grad: 1
|
66 |
+
grad_clip: 5
|
67 |
+
max_epoch: 20
|
68 |
+
keep_nbest_models: 10
|
69 |
+
avg_nbest_model: 10
|
70 |
+
log_interval: 100
|
71 |
+
resume: true
|
72 |
+
validate_interval: 10000
|
73 |
+
save_checkpoint_interval: 10000
|
74 |
+
|
75 |
+
optim: adamw
|
76 |
+
optim_conf:
|
77 |
+
lr: 0.00002
|
78 |
+
scheduler: warmuplr
|
79 |
+
scheduler_conf:
|
80 |
+
warmup_steps: 25000
|
81 |
+
|
82 |
+
specaug: SpecAugLFR
|
83 |
+
specaug_conf:
|
84 |
+
apply_time_warp: false
|
85 |
+
time_warp_window: 5
|
86 |
+
time_warp_mode: bicubic
|
87 |
+
apply_freq_mask: true
|
88 |
+
freq_mask_width_range:
|
89 |
+
- 0
|
90 |
+
- 30
|
91 |
+
lfr_rate: 6
|
92 |
+
num_freq_mask: 1
|
93 |
+
apply_time_mask: true
|
94 |
+
time_mask_width_range:
|
95 |
+
- 0
|
96 |
+
- 12
|
97 |
+
num_time_mask: 1
|