Upload folder using huggingface_hub
Browse files- 33.pt +3 -0
- 33.yaml +5 -0
- 7layer_20ms_33pt/ce_layer.pkl +3 -0
- 7layer_20ms_33pt/phn_center.npy +3 -0
- global_cmvn +1 -0
- train.yaml +107 -0
33.pt
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:0313e002bcecd7f24843300c9dcd6d5947d74d7cade0d12dfacaa2900977371b
|
| 3 |
+
size 135793931
|
33.yaml
ADDED
|
@@ -0,0 +1,5 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
cv_acc: 85.00080549078187
|
| 2 |
+
cv_loss: 0.6253250154177071
|
| 3 |
+
epoch: 33
|
| 4 |
+
lr: 3.864118600201355e-05
|
| 5 |
+
step: 1380057
|
7layer_20ms_33pt/ce_layer.pkl
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:41023de2f9277d0bb060aa3ad9a7f1aa07c61389985cce7dfb1067f0e6fd9b18
|
| 3 |
+
size 618037
|
7layer_20ms_33pt/phn_center.npy
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:0071d744a1172384cc9bc8fa3557ed7afe19feac1a32f16057cce447c131b67d
|
| 3 |
+
size 1230976
|
global_cmvn
ADDED
|
@@ -0,0 +1 @@
|
|
|
|
|
|
|
| 1 |
+
{"mean_stat": [75731376.0, 79514568.0, 87436296.0, 93233848.0, 97993608.0, 101069072.0, 102676224.0, 102601904.0, 102923536.0, 103032728.0, 104305008.0, 105199160.0, 106596456.0, 107461952.0, 107554288.0, 107720816.0, 107366488.0, 106614232.0, 106952936.0, 105242032.0, 103862080.0, 104817976.0, 103161792.0, 103698384.0, 102924024.0, 103407968.0, 102651056.0, 103237560.0, 102954800.0, 103020944.0, 103365968.0, 103501520.0, 103830472.0, 104461696.0, 105064072.0, 105814760.0, 106496224.0, 106974272.0, 107450032.0, 107183944.0, 107763728.0, 107420184.0, 107946312.0, 107919640.0, 108350736.0, 109008840.0, 109742152.0, 110260528.0, 110637304.0, 111002704.0, 110753840.0, 110596424.0, 110549024.0, 110660512.0, 110617936.0, 110718648.0, 110843784.0, 110688584.0, 110269328.0, 109632112.0, 109150904.0, 108340872.0, 107745632.0, 107056168.0, 106316872.0, 105559336.0, 104837960.0, 104368056.0, 104056392.0, 103780832.0, 103566048.0, 103452096.0, 103311136.0, 103106912.0, 102698656.0, 102027688.0, 100617560.0, 97610760.0, 92844152.0, 86920152.0], "var_stat": [912440832.0, 1013694080.0, 1220854400.0, 1375290752.0, 1511352832.0, 1604532096.0, 1656307200.0, 1658182656.0, 1666277504.0, 1666270848.0, 1703693824.0, 1732099968.0, 1775637376.0, 1802882432.0, 1806045696.0, 1810400512.0, 1798349056.0, 1773962240.0, 1783695616.0, 1730289024.0, 1687749632.0, 1714445696.0, 1662247680.0, 1676040576.0, 1651099520.0, 1664215808.0, 1640219904.0, 1657539840.0, 1648972032.0, 1650200576.0, 1660111104.0, 1663428352.0, 1673046528.0, 1691747200.0, 1709491072.0, 1732275712.0, 1753430272.0, 1767821696.0, 1782058880.0, 1773574528.0, 1790768640.0, 1779407744.0, 1795281664.0, 1794298496.0, 1808020864.0, 1829600000.0, 1854564992.0, 1872319616.0, 1885018112.0, 1896773888.0, 1888105216.0, 1882481152.0, 1880364416.0, 1884025472.0, 1883216512.0, 1887112064.0, 1891335680.0, 1886171392.0, 1872332544.0, 1851304704.0, 1835424896.0, 1809687040.0, 1790931840.0, 1769499776.0, 1746419840.0, 1722945280.0, 1700880896.0, 1686990592.0, 1678278144.0, 1670231936.0, 1663805824.0, 1660343680.0, 1656274560.0, 1650625536.0, 1638248320.0, 1617632256.0, 1576110336.0, 1489716864.0, 1356544256.0, 1198415744.0], "frame_num": 6834022}
|
train.yaml
ADDED
|
@@ -0,0 +1,107 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
accum_grad: 2
|
| 2 |
+
cmvn_file: exp/stream_wenet_giga/global_cmvn
|
| 3 |
+
dataset_conf:
|
| 4 |
+
batch_conf:
|
| 5 |
+
batch_size: 8
|
| 6 |
+
batch_type: static
|
| 7 |
+
cv_emb_conf:
|
| 8 |
+
emb_type: ivector
|
| 9 |
+
feat_type: fbank
|
| 10 |
+
h5_path: /apdcephfs/share_1316500/nenali/corpus/M2Met_beam/dev/data/embedding/ivector/python/ivector_python_dev_fbank.h5
|
| 11 |
+
perb_h5_path: /apdcephfs/share_1316500/nenali/corpus/M2Met_beam/dev/data/embedding/ivector/python/ivector_python_dev_fbank_perb.h5
|
| 12 |
+
fbank_conf:
|
| 13 |
+
dither: 0.1
|
| 14 |
+
frame_length: 25
|
| 15 |
+
frame_shift: 10
|
| 16 |
+
num_mel_bins: 80
|
| 17 |
+
filter_conf:
|
| 18 |
+
max_length: 1500
|
| 19 |
+
min_length: 10
|
| 20 |
+
token_max_length: 1500
|
| 21 |
+
token_min_length: 10
|
| 22 |
+
resample_conf:
|
| 23 |
+
resample_rate: 16000
|
| 24 |
+
shuffle: true
|
| 25 |
+
shuffle_conf:
|
| 26 |
+
shuffle_size: 1500
|
| 27 |
+
sort: true
|
| 28 |
+
sort_conf:
|
| 29 |
+
sort_size: 500
|
| 30 |
+
spec_aug: true
|
| 31 |
+
spec_aug_conf:
|
| 32 |
+
max_f: 30
|
| 33 |
+
max_t: 40
|
| 34 |
+
num_f_mask: 2
|
| 35 |
+
num_t_mask: 2
|
| 36 |
+
speed_perturb: false
|
| 37 |
+
test_emb_conf:
|
| 38 |
+
emb_type: ivector
|
| 39 |
+
feat_type: fbank
|
| 40 |
+
h5_path: /apdcephfs/share_1316500/nenali/corpus/M2Met_beam/test/data/embedding/ivector/python/ivector_python_test_fbank.h5
|
| 41 |
+
train_emb_conf:
|
| 42 |
+
emb_type: ivector
|
| 43 |
+
feat_type: fbank
|
| 44 |
+
h5_path: /apdcephfs/share_1316500/nenali/corpus/M2Met_beam/train/data/embedding/ivector/python/ivector_python_train_fbank.h5
|
| 45 |
+
perb_h5_path: /apdcephfs/share_1316500/nenali/corpus/M2Met_beam/train/data/embedding/ivector/python/ivector_python_train_fbank_perb.h5
|
| 46 |
+
use_emb: false
|
| 47 |
+
use_seg: false
|
| 48 |
+
wav_aug: true
|
| 49 |
+
wav_aug_conf:
|
| 50 |
+
musan_path: /apdcephfs/share_1149801/speech_user/nenali/private_nenali/corpus/wav_h5/musan_new.h5
|
| 51 |
+
rir_path: /apdcephfs/share_1149801/speech_user/nenali/private_nenali/corpus/wav_h5/rir_new.h5
|
| 52 |
+
decoder: transformer
|
| 53 |
+
decoder_conf:
|
| 54 |
+
attention_heads: 4
|
| 55 |
+
dropout_rate: 0.1
|
| 56 |
+
linear_units: 2048
|
| 57 |
+
num_blocks: 6
|
| 58 |
+
positional_dropout_rate: 0.1
|
| 59 |
+
self_attention_dropout_rate: 0.0
|
| 60 |
+
src_attention_dropout_rate: 0.0
|
| 61 |
+
encoder: conformer
|
| 62 |
+
encoder_conf:
|
| 63 |
+
activation_type: swish
|
| 64 |
+
attention_dropout_rate: 0.0
|
| 65 |
+
attention_heads: 4
|
| 66 |
+
causal: true
|
| 67 |
+
cnn_module_kernel: 15
|
| 68 |
+
cnn_module_norm: layer_norm
|
| 69 |
+
dropout_rate: 0.1
|
| 70 |
+
emb_dim: 100
|
| 71 |
+
input_layer: conv2d
|
| 72 |
+
linear_units: 2048
|
| 73 |
+
normalize_before: true
|
| 74 |
+
num_blocks: 7
|
| 75 |
+
output_size: 256
|
| 76 |
+
pos_enc_layer_type: rel_pos
|
| 77 |
+
positional_dropout_rate: 0.1
|
| 78 |
+
selfattention_layer_type: rel_selfattn
|
| 79 |
+
use_cnn_module: true
|
| 80 |
+
use_dynamic_chunk: true
|
| 81 |
+
use_emb: false
|
| 82 |
+
grad_clip: 5
|
| 83 |
+
input_dim: 80
|
| 84 |
+
is_json_cmvn: true
|
| 85 |
+
log_interval: 400
|
| 86 |
+
max_epoch: 100
|
| 87 |
+
model_conf:
|
| 88 |
+
ctc_weight: 0.0
|
| 89 |
+
length_normalized_loss: false
|
| 90 |
+
loss_margin: 0.2
|
| 91 |
+
lsm_weight: 0.1
|
| 92 |
+
ppg_loss: softmax
|
| 93 |
+
ppg_weight: 1.0
|
| 94 |
+
sv_conf:
|
| 95 |
+
pooling_type: stats
|
| 96 |
+
spk_dim: 128
|
| 97 |
+
spk_num: 339
|
| 98 |
+
sv_loss: softmax
|
| 99 |
+
sv_weight: 0.5
|
| 100 |
+
use_sv: false
|
| 101 |
+
optim: adam
|
| 102 |
+
optim_conf:
|
| 103 |
+
lr: 0.001
|
| 104 |
+
output_dim: 600
|
| 105 |
+
scheduler: warmuplr
|
| 106 |
+
scheduler_conf:
|
| 107 |
+
warmup_steps: 2000
|