Upload 5 files
Browse files- README.md +3 -3
- cnceleb_resnet34_LM.onnx +3 -0
- config.yaml +97 -0
- model_5.pt +3 -0
- pytorch_model.bin +3 -0
README.md
CHANGED
@@ -1,3 +1,3 @@
|
|
1 |
-
---
|
2 |
-
license:
|
3 |
-
---
|
|
|
1 |
+
---
|
2 |
+
license: apache-2.0
|
3 |
+
---
|
cnceleb_resnet34_LM.onnx
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:e7584940aeac8d5512d875e58ce6c09ba4ddad65d8128e1dac0d93aadd087ebb
|
3 |
+
size 26530309
|
config.yaml
ADDED
@@ -0,0 +1,97 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
model:
|
2 |
+
_target_: pyannote.audio.models.embedding.WeSpeakerResNet34
|
3 |
+
sample_rate: 16000
|
4 |
+
num_channels: 1
|
5 |
+
num_mel_bins: 80
|
6 |
+
frame_length: 25
|
7 |
+
frame_shift: 10
|
8 |
+
dither: 0.0
|
9 |
+
window_type: hamming
|
10 |
+
use_energy: false
|
11 |
+
|
12 |
+
checkpoint: exp/cnceleb_v2/resnet/wespeaker/renset34_new_margin-LM/models/model_0.pt
|
13 |
+
data_type: shard
|
14 |
+
dataloader_args:
|
15 |
+
batch_size: 64
|
16 |
+
drop_last: true
|
17 |
+
num_workers: 16
|
18 |
+
pin_memory: false
|
19 |
+
prefetch_factor: 8
|
20 |
+
dataset_args:
|
21 |
+
aug_prob: 0.6
|
22 |
+
fbank_args:
|
23 |
+
dither: 1.0
|
24 |
+
frame_length: 25
|
25 |
+
frame_shift: 10
|
26 |
+
num_mel_bins: 80
|
27 |
+
num_frms: 600
|
28 |
+
shuffle: true
|
29 |
+
shuffle_args:
|
30 |
+
shuffle_size: 2500
|
31 |
+
spec_aug: false
|
32 |
+
spec_aug_args:
|
33 |
+
max_f: 8
|
34 |
+
max_t: 10
|
35 |
+
num_f_mask: 1
|
36 |
+
num_t_mask: 1
|
37 |
+
prob: 0.6
|
38 |
+
speed_perturb: false
|
39 |
+
do_lm: true
|
40 |
+
exp_dir: exp/cnceleb_v2/resnet/wespeaker/renset34_new_margin-LM
|
41 |
+
gpus:
|
42 |
+
- 0
|
43 |
+
- 1
|
44 |
+
- 2
|
45 |
+
- 3
|
46 |
+
- 4
|
47 |
+
- 5
|
48 |
+
- 6
|
49 |
+
- 7
|
50 |
+
log_batch_interval: 100
|
51 |
+
loss: CrossEntropyLoss
|
52 |
+
loss_args: {}
|
53 |
+
margin_scheduler: MarginScheduler
|
54 |
+
margin_update:
|
55 |
+
epoch_iter: 992
|
56 |
+
final_margin: 0.5
|
57 |
+
fix_start_epoch: 1
|
58 |
+
increase_start_epoch: 1
|
59 |
+
increase_type: exp
|
60 |
+
initial_margin: 0.5
|
61 |
+
update_margin: true
|
62 |
+
model: ResNet34
|
63 |
+
model_args:
|
64 |
+
embed_dim: 256
|
65 |
+
feat_dim: 80
|
66 |
+
pooling_func: TSTP
|
67 |
+
two_emb_layer: false
|
68 |
+
model_init: null
|
69 |
+
noise_data: data/cnceleb_wespeaker/musan/lmdb
|
70 |
+
num_avg: 1
|
71 |
+
num_epochs: 5
|
72 |
+
optimizer: SGD
|
73 |
+
optimizer_args:
|
74 |
+
lr: 0.0001
|
75 |
+
momentum: 0.9
|
76 |
+
nesterov: true
|
77 |
+
weight_decay: 0.0001
|
78 |
+
projection_args:
|
79 |
+
easy_margin: false
|
80 |
+
embed_dim: 256
|
81 |
+
num_class: 8379
|
82 |
+
project_type: arc_margin
|
83 |
+
scale: 32.0
|
84 |
+
reverb_data: data/cnceleb_wespeaker/rirs/lmdb
|
85 |
+
save_epoch_interval: 1
|
86 |
+
scheduler: ExponentialDecrease
|
87 |
+
scheduler_args:
|
88 |
+
epoch_iter: 992
|
89 |
+
final_lr: 2.5e-05
|
90 |
+
initial_lr: 0.0001
|
91 |
+
num_epochs: 5
|
92 |
+
scale_ratio: 8.0
|
93 |
+
warm_from_zero: true
|
94 |
+
warm_up_epoch: 1
|
95 |
+
seed: 42
|
96 |
+
train_data: data/cnceleb_wespeaker/cnceleb_train/shard.list
|
97 |
+
train_label: data/cnceleb_wespeaker/cnceleb_train/utt2spk
|
model_5.pt
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:4fed1fbf1bb1e772f794e54f7dd8c5635485a0ccce799ecb429a9232d22596f7
|
3 |
+
size 35222731
|
pytorch_model.bin
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:4fed1fbf1bb1e772f794e54f7dd8c5635485a0ccce799ecb429a9232d22596f7
|
3 |
+
size 35222731
|