huseinzol05
commited on
Commit
•
2e64721
1
Parent(s):
9b1df44
Upload 2 files
Browse files- model_config.yaml +115 -0
- model_weights.ckpt +3 -0
model_config.yaml
ADDED
@@ -0,0 +1,115 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
train_ds:
|
2 |
+
manifest_filepath: /ws/manifests/raid/combined/train_manifest.json
|
3 |
+
sample_rate: 16000
|
4 |
+
labels: null
|
5 |
+
batch_size: 64
|
6 |
+
shuffle: true
|
7 |
+
time_length: 8
|
8 |
+
augmentor:
|
9 |
+
noise:
|
10 |
+
manifest_path: /ws/manifests/raid/musan/musan_music_noise_manifest_dur8.json
|
11 |
+
prob: 0.2
|
12 |
+
min_snr_db: 5
|
13 |
+
max_snr_db: 15
|
14 |
+
num_workers: 4
|
15 |
+
validation_ds:
|
16 |
+
manifest_filepath: /ws/manifests/raid/voxceleb/small_manifest.json
|
17 |
+
sample_rate: 16000
|
18 |
+
labels: null
|
19 |
+
batch_size: 64
|
20 |
+
shuffle: false
|
21 |
+
time_length: 8
|
22 |
+
num_workers: 1
|
23 |
+
preprocessor:
|
24 |
+
_target_: nemo.collections.asr.modules.AudioToMelSpectrogramPreprocessor
|
25 |
+
normalize: per_feature
|
26 |
+
window_size: 0.02
|
27 |
+
sample_rate: 16000
|
28 |
+
window_stride: 0.01
|
29 |
+
window: hann
|
30 |
+
features: 64
|
31 |
+
n_fft: 512
|
32 |
+
frame_splicing: 1
|
33 |
+
dither: 1.0e-05
|
34 |
+
stft_conv: false
|
35 |
+
encoder:
|
36 |
+
_target_: nemo.collections.asr.modules.ConvASREncoder
|
37 |
+
feat_in: 64
|
38 |
+
activation: relu
|
39 |
+
conv_mask: true
|
40 |
+
jasper:
|
41 |
+
- filters: 512
|
42 |
+
repeat: 1
|
43 |
+
kernel:
|
44 |
+
- 3
|
45 |
+
stride:
|
46 |
+
- 1
|
47 |
+
dilation:
|
48 |
+
- 1
|
49 |
+
dropout: 0.5
|
50 |
+
residual: true
|
51 |
+
separable: true
|
52 |
+
- filters: 512
|
53 |
+
repeat: 2
|
54 |
+
kernel:
|
55 |
+
- 7
|
56 |
+
stride:
|
57 |
+
- 1
|
58 |
+
dilation:
|
59 |
+
- 1
|
60 |
+
dropout: 0.5
|
61 |
+
residual: true
|
62 |
+
separable: true
|
63 |
+
- filters: 512
|
64 |
+
repeat: 2
|
65 |
+
kernel:
|
66 |
+
- 11
|
67 |
+
stride:
|
68 |
+
- 1
|
69 |
+
dilation:
|
70 |
+
- 1
|
71 |
+
dropout: 0.5
|
72 |
+
residual: true
|
73 |
+
separable: true
|
74 |
+
- filters: 512
|
75 |
+
repeat: 2
|
76 |
+
kernel:
|
77 |
+
- 15
|
78 |
+
stride:
|
79 |
+
- 1
|
80 |
+
dilation:
|
81 |
+
- 1
|
82 |
+
dropout: 0.5
|
83 |
+
residual: true
|
84 |
+
separable: true
|
85 |
+
- filters: 1500
|
86 |
+
repeat: 1
|
87 |
+
kernel:
|
88 |
+
- 1
|
89 |
+
stride:
|
90 |
+
- 1
|
91 |
+
dilation:
|
92 |
+
- 1
|
93 |
+
dropout: 0.0
|
94 |
+
residual: false
|
95 |
+
separable: true
|
96 |
+
decoder:
|
97 |
+
_target_: nemo.collections.asr.modules.SpeakerDecoder
|
98 |
+
feat_in: 1500
|
99 |
+
num_classes: 7205
|
100 |
+
pool_mode: xvector
|
101 |
+
emb_sizes: 256
|
102 |
+
angular: true
|
103 |
+
loss:
|
104 |
+
scale: 30
|
105 |
+
margin: 0.2
|
106 |
+
optim:
|
107 |
+
name: sgd
|
108 |
+
lr: 0.006
|
109 |
+
weight_decay: 0.001
|
110 |
+
sched:
|
111 |
+
name: CosineAnnealing
|
112 |
+
warmup_ratio: 0.1
|
113 |
+
min_lr: 0.0001
|
114 |
+
momentum: 0.9
|
115 |
+
target: nemo.collections.asr.models.label_models.EncDecSpeakerLabelModel
|
model_weights.ckpt
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:5a112bf4194dd5b92d47e2a27f8161d7572fc4a505ced74ef1b1d1c046b733d8
|
3 |
+
size 23577441
|