Upload pyannote segmentation model converted for MLX Swift (80 cos+sin filters, 4 LSTM layers)
Browse files- config.json +45 -0
- model.safetensors +3 -0
config.json
ADDED
|
@@ -0,0 +1,45 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"model_type": "pyannote-segmentation",
|
| 3 |
+
"sample_rate": 16000,
|
| 4 |
+
"sincnet": {
|
| 5 |
+
"n_filters": [
|
| 6 |
+
80,
|
| 7 |
+
60,
|
| 8 |
+
60
|
| 9 |
+
],
|
| 10 |
+
"kernel_sizes": [
|
| 11 |
+
251,
|
| 12 |
+
5,
|
| 13 |
+
5
|
| 14 |
+
],
|
| 15 |
+
"strides": [
|
| 16 |
+
10,
|
| 17 |
+
1,
|
| 18 |
+
1
|
| 19 |
+
],
|
| 20 |
+
"pool_sizes": [
|
| 21 |
+
3,
|
| 22 |
+
3,
|
| 23 |
+
3
|
| 24 |
+
]
|
| 25 |
+
},
|
| 26 |
+
"lstm": {
|
| 27 |
+
"hidden_size": 128,
|
| 28 |
+
"num_layers": 4,
|
| 29 |
+
"bidirectional": true
|
| 30 |
+
},
|
| 31 |
+
"linear": {
|
| 32 |
+
"hidden_size": 128,
|
| 33 |
+
"num_layers": 2
|
| 34 |
+
},
|
| 35 |
+
"num_classes": 7,
|
| 36 |
+
"max_speakers": 3,
|
| 37 |
+
"powerset_max_classes": 2,
|
| 38 |
+
"num_frames_per_chunk": 589,
|
| 39 |
+
"chunk_duration": 10.0,
|
| 40 |
+
"chunk_step_ratio": 0.1,
|
| 41 |
+
"warm_up": [
|
| 42 |
+
0.0,
|
| 43 |
+
0.0
|
| 44 |
+
]
|
| 45 |
+
}
|
model.safetensors
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:d1630fa2c22f47e4c89034f8d5e3aff99884f55347d48ce70dd306328b4421f5
|
| 3 |
+
size 5960404
|