cemsubakan
commited on
Update hyperparams.yaml
Browse files- hyperparams.yaml +10 -10
hyperparams.yaml
CHANGED
@@ -13,11 +13,11 @@ kernel_size: 16
|
|
13 |
kernel_stride: 8
|
14 |
|
15 |
# Specifying the network
|
16 |
-
Encoder:
|
17 |
kernel_size: 16
|
18 |
out_channels: 128
|
19 |
|
20 |
-
intra_mdl:
|
21 |
num_layers: 8
|
22 |
d_model: 128
|
23 |
nhead: 8
|
@@ -28,7 +28,7 @@ intra_mdl: &id001 !new:speechbrain.lobes.models.resepformer.SBTransformerBlock_w
|
|
28 |
use_norm: true
|
29 |
use_skip: true
|
30 |
|
31 |
-
mem_mdl:
|
32 |
num_layers: 8
|
33 |
d_model: 128
|
34 |
nhead: 8
|
@@ -39,7 +39,7 @@ mem_mdl: &id002 !new:speechbrain.lobes.models.resepformer.SBTransformerBlock_wno
|
|
39 |
use_norm: true
|
40 |
use_skip: true
|
41 |
|
42 |
-
MaskNet:
|
43 |
input_dim: 128
|
44 |
num_spk: 2
|
45 |
causal: false
|
@@ -47,10 +47,10 @@ MaskNet: &id005 !new:speechbrain.lobes.models.resepformer.ResourceEfficientSepar
|
|
47 |
segment_size: 150
|
48 |
layer: 2
|
49 |
mem_type: av
|
50 |
-
seg_model:
|
51 |
-
mem_model:
|
52 |
|
53 |
-
Decoder:
|
54 |
in_channels: 128
|
55 |
out_channels: 1
|
56 |
kernel_size: 16
|
@@ -58,9 +58,9 @@ Decoder: &id004 !new:speechbrain.lobes.models.dual_path.Decoder
|
|
58 |
bias: false
|
59 |
|
60 |
modules:
|
61 |
-
encoder:
|
62 |
-
decoder:
|
63 |
-
masknet:
|
64 |
|
65 |
pretrainer: !new:speechbrain.utils.parameter_transfer.Pretrainer
|
66 |
loadables:
|
|
|
13 |
kernel_stride: 8
|
14 |
|
15 |
# Specifying the network
|
16 |
+
Encoder: !new:speechbrain.lobes.models.dual_path.Encoder
|
17 |
kernel_size: 16
|
18 |
out_channels: 128
|
19 |
|
20 |
+
intra_mdl: !new:speechbrain.lobes.models.resepformer.SBTransformerBlock_wnormandskip
|
21 |
num_layers: 8
|
22 |
d_model: 128
|
23 |
nhead: 8
|
|
|
28 |
use_norm: true
|
29 |
use_skip: true
|
30 |
|
31 |
+
mem_mdl: !new:speechbrain.lobes.models.resepformer.SBTransformerBlock_wnormandskip
|
32 |
num_layers: 8
|
33 |
d_model: 128
|
34 |
nhead: 8
|
|
|
39 |
use_norm: true
|
40 |
use_skip: true
|
41 |
|
42 |
+
MaskNet: !new:speechbrain.lobes.models.resepformer.ResourceEfficientSeparator
|
43 |
input_dim: 128
|
44 |
num_spk: 2
|
45 |
causal: false
|
|
|
47 |
segment_size: 150
|
48 |
layer: 2
|
49 |
mem_type: av
|
50 |
+
seg_model: !ref <intra_mdl>
|
51 |
+
mem_model: !ref <mem_mdl>
|
52 |
|
53 |
+
Decoder: !new:speechbrain.lobes.models.dual_path.Decoder
|
54 |
in_channels: 128
|
55 |
out_channels: 1
|
56 |
kernel_size: 16
|
|
|
58 |
bias: false
|
59 |
|
60 |
modules:
|
61 |
+
encoder: !ref <Encoder>
|
62 |
+
decoder: !ref <Decoder>
|
63 |
+
masknet: !ref <MaskNet>
|
64 |
|
65 |
pretrainer: !new:speechbrain.utils.parameter_transfer.Pretrainer
|
66 |
loadables:
|