Upload config.yaml
Browse files- Woosh-AE/config.yaml +27 -0
Woosh-AE/config.yaml
ADDED
|
@@ -0,0 +1,27 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
exclude_from_checkpoint: true
|
| 2 |
+
trainable: true
|
| 3 |
+
istft_head:
|
| 4 |
+
_target_: woosh.module.model.vocos.ISTFTCircleHead
|
| 5 |
+
_partial_: true
|
| 6 |
+
padding: center
|
| 7 |
+
conv_pad: same
|
| 8 |
+
conv_kernel: 1
|
| 9 |
+
softclip: softplus
|
| 10 |
+
ztransform:
|
| 11 |
+
_target_: woosh.module.model.vocos.ZeroDropoutTransform
|
| 12 |
+
_partial_: true
|
| 13 |
+
p: 1.0
|
| 14 |
+
_target_: woosh.module.model.VocosAutoEncoder
|
| 15 |
+
channels: 1
|
| 16 |
+
z_dim: 128
|
| 17 |
+
d_model: 2048
|
| 18 |
+
intermediate_dim: 3072
|
| 19 |
+
n_fft: 960
|
| 20 |
+
hop_length: 480
|
| 21 |
+
num_layers: 8
|
| 22 |
+
enc_num_layers: 8
|
| 23 |
+
input_layer_norm: true
|
| 24 |
+
final_layer_norm: true
|
| 25 |
+
stft_normalized: false
|
| 26 |
+
spec_embed: stft-complex
|
| 27 |
+
sample_rate: 48000
|