Mirco commited on
Commit
5dec5bb
1 Parent(s): 5c9ce57

upload model

Browse files
Files changed (4) hide show
  1. hyperparams.yaml +77 -0
  2. mean_var_norm.ckpt +0 -0
  3. model.ckpt +0 -0
  4. normalizer.ckpt +0 -0
hyperparams.yaml ADDED
@@ -0,0 +1,77 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ # ############################################################################
2
+ # Model: Small CRDNN model for Voice Activity Detection
3
+ # Author: Mirco Ravanelli, 2021
4
+ # ############################################################################
5
+
6
+ # Feature parameters
7
+ sample_rate: 16000
8
+ time_resolution: 0.01 # in seconds (e.g,, 0.01 = 10 ms)
9
+ n_fft: 400
10
+ n_mels: 40
11
+
12
+ # Model parameters
13
+ activation: !name:torch.nn.LeakyReLU
14
+ dropout: 0.15
15
+ cnn_blocks: 2
16
+ cnn_channels: (16, 32)
17
+ cnn_kernelsize: (3, 3)
18
+ rnn_layers: 2
19
+ rnn_neurons: 32
20
+ rnn_bidirectional: True
21
+ dnn_blocks: 1
22
+ dnn_neurons: 16
23
+ output_neurons: 1
24
+ device: 'cpu' # set 'cuda:0' for gpu
25
+
26
+ # Feature/Model objects
27
+ compute_features: !new:speechbrain.lobes.features.Fbank
28
+ sample_rate: !ref <sample_rate>
29
+ n_fft: !ref <n_fft>
30
+ n_mels: !ref <n_mels>
31
+ hop_length: !ref <time_resolution> * 1000 # in ms
32
+
33
+ mean_var_norm: !new:speechbrain.processing.features.InputNormalization
34
+ norm_type: sentence
35
+
36
+ cnn: !new:speechbrain.nnet.containers.Sequential
37
+ input_shape: [null, null, !ref <n_mels>]
38
+ norm1: !name:speechbrain.nnet.normalization.LayerNorm
39
+ cnn1: !name:speechbrain.lobes.models.CRDNN.CNN_Block
40
+ channels: 16
41
+ kernel_size: (3, 3)
42
+ cnn2: !name:speechbrain.lobes.models.CRDNN.CNN_Block
43
+ channels: 32
44
+ kernel_size: (3, 3)
45
+
46
+ rnn: !new:speechbrain.nnet.RNN.GRU
47
+ input_shape: [null, null, 320]
48
+ hidden_size: !ref <rnn_neurons>
49
+ num_layers: !ref <rnn_layers>
50
+ bidirectional: True
51
+
52
+ dnn: !new:speechbrain.nnet.containers.Sequential
53
+ input_shape: [null, null, !ref <rnn_neurons> * 2]
54
+ dnn1: !name:speechbrain.lobes.models.CRDNN.DNN_Block
55
+ neurons: !ref <dnn_neurons>
56
+ dnn2: !name:speechbrain.lobes.models.CRDNN.DNN_Block
57
+ neurons: !ref <dnn_neurons>
58
+ lin: !name:speechbrain.nnet.linear.Linear
59
+ n_neurons: !ref <output_neurons>
60
+ bias: False
61
+
62
+
63
+ model: !new:torch.nn.ModuleList
64
+ - [!ref <cnn>, !ref <rnn>, !ref <dnn>]
65
+
66
+ modules:
67
+ compute_features: !ref <compute_features>
68
+ model: !ref <model>
69
+ cnn: !ref <cnn>
70
+ rnn: !ref <rnn>
71
+ dnn: !ref <dnn>
72
+ mean_var_norm: !ref <mean_var_norm>
73
+
74
+ pretrainer: !new:speechbrain.utils.parameter_transfer.Pretrainer
75
+ loadables:
76
+ model: !ref <model>
77
+ mean_var_norm: !ref <mean_var_norm>
mean_var_norm.ckpt ADDED
Binary file (1.06 kB). View file
 
model.ckpt ADDED
Binary file (453 kB). View file
 
normalizer.ckpt ADDED
Binary file (1.06 kB). View file