Mirror vocoder/config.json from cvssp/audioldm2@c8e7e189
Browse files
encoders/cvssp/audioldm2/vocoder/config.json
ADDED
|
@@ -0,0 +1,50 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"architectures": [
|
| 3 |
+
"SpeechT5HifiGan"
|
| 4 |
+
],
|
| 5 |
+
"initializer_range": 0.01,
|
| 6 |
+
"leaky_relu_slope": 0.1,
|
| 7 |
+
"model_in_dim": 64,
|
| 8 |
+
"model_type": "hifigan",
|
| 9 |
+
"normalize_before": false,
|
| 10 |
+
"resblock_dilation_sizes": [
|
| 11 |
+
[
|
| 12 |
+
1,
|
| 13 |
+
3,
|
| 14 |
+
5
|
| 15 |
+
],
|
| 16 |
+
[
|
| 17 |
+
1,
|
| 18 |
+
3,
|
| 19 |
+
5
|
| 20 |
+
],
|
| 21 |
+
[
|
| 22 |
+
1,
|
| 23 |
+
3,
|
| 24 |
+
5
|
| 25 |
+
]
|
| 26 |
+
],
|
| 27 |
+
"resblock_kernel_sizes": [
|
| 28 |
+
3,
|
| 29 |
+
7,
|
| 30 |
+
11
|
| 31 |
+
],
|
| 32 |
+
"sampling_rate": 16000,
|
| 33 |
+
"torch_dtype": "float32",
|
| 34 |
+
"transformers_version": "4.32.0.dev0",
|
| 35 |
+
"upsample_initial_channel": 1024,
|
| 36 |
+
"upsample_kernel_sizes": [
|
| 37 |
+
16,
|
| 38 |
+
16,
|
| 39 |
+
8,
|
| 40 |
+
4,
|
| 41 |
+
4
|
| 42 |
+
],
|
| 43 |
+
"upsample_rates": [
|
| 44 |
+
5,
|
| 45 |
+
4,
|
| 46 |
+
2,
|
| 47 |
+
2,
|
| 48 |
+
2
|
| 49 |
+
]
|
| 50 |
+
}
|