Add split vocoder component (audio_vae unchanged, still usable by sglang's combined loader)
Browse files- model_index.json +4 -0
- vocoder/config.json +44 -0
- vocoder/diffusion_pytorch_model.safetensors +3 -0
model_index.json
CHANGED
|
@@ -43,5 +43,9 @@
|
|
| 43 |
"audio_vae": [
|
| 44 |
"diffusers",
|
| 45 |
"MMAudioVAE"
|
|
|
|
|
|
|
|
|
|
|
|
|
| 46 |
]
|
| 47 |
}
|
|
|
|
| 43 |
"audio_vae": [
|
| 44 |
"diffusers",
|
| 45 |
"MMAudioVAE"
|
| 46 |
+
],
|
| 47 |
+
"vocoder": [
|
| 48 |
+
"diffusers",
|
| 49 |
+
"MMAudioVocoder"
|
| 50 |
]
|
| 51 |
}
|
vocoder/config.json
ADDED
|
@@ -0,0 +1,44 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"_class_name": "MMAudioVocoder",
|
| 3 |
+
"_diffusers_version": "0.40.0.dev0",
|
| 4 |
+
"num_mels": 128,
|
| 5 |
+
"resblock_dilation_sizes": [
|
| 6 |
+
[
|
| 7 |
+
1,
|
| 8 |
+
3,
|
| 9 |
+
5
|
| 10 |
+
],
|
| 11 |
+
[
|
| 12 |
+
1,
|
| 13 |
+
3,
|
| 14 |
+
5
|
| 15 |
+
],
|
| 16 |
+
[
|
| 17 |
+
1,
|
| 18 |
+
3,
|
| 19 |
+
5
|
| 20 |
+
]
|
| 21 |
+
],
|
| 22 |
+
"resblock_kernel_sizes": [
|
| 23 |
+
3,
|
| 24 |
+
7,
|
| 25 |
+
11
|
| 26 |
+
],
|
| 27 |
+
"upsample_initial_channel": 1536,
|
| 28 |
+
"upsample_kernel_sizes": [
|
| 29 |
+
16,
|
| 30 |
+
8,
|
| 31 |
+
4,
|
| 32 |
+
4,
|
| 33 |
+
4,
|
| 34 |
+
4
|
| 35 |
+
],
|
| 36 |
+
"upsample_rates": [
|
| 37 |
+
8,
|
| 38 |
+
4,
|
| 39 |
+
2,
|
| 40 |
+
2,
|
| 41 |
+
2,
|
| 42 |
+
2
|
| 43 |
+
]
|
| 44 |
+
}
|
vocoder/diffusion_pytorch_model.safetensors
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:abf0536824bdce8dcf8ee10b815770111f4b5d05d0122e1bab4b5719324edca7
|
| 3 |
+
size 488692656
|