initial commit
Browse files- config.toml +30 -0
- pytorch_model.bin +3 -0
config.toml
ADDED
@@ -0,0 +1,30 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
[model]
|
2 |
+
path = "diarizen.models.eend.model_wavlm_conformer.Model"
|
3 |
+
|
4 |
+
[model.args]
|
5 |
+
wavlm_src = "wavlm_large_s80_md"
|
6 |
+
wavlm_layer_num = 25
|
7 |
+
wavlm_feat_dim = 1024
|
8 |
+
attention_in = 256
|
9 |
+
ffn_hidden = 1024
|
10 |
+
num_head = 4
|
11 |
+
num_layer = 4
|
12 |
+
dropout = 0.1
|
13 |
+
chunk_size = 8
|
14 |
+
use_posi = false
|
15 |
+
output_activate_function = false
|
16 |
+
selected_channel = 0
|
17 |
+
|
18 |
+
[inference.args]
|
19 |
+
seg_duration = 8
|
20 |
+
segmentation_step = 0.1
|
21 |
+
batch_size = 32
|
22 |
+
apply_median_filtering = true
|
23 |
+
|
24 |
+
[clustering.args]
|
25 |
+
method = "AgglomerativeClustering"
|
26 |
+
min_speakers = 2
|
27 |
+
max_speakers = 2
|
28 |
+
ahc_threshold = 0.6647095879538272
|
29 |
+
min_cluster_size = 16
|
30 |
+
min_frames_ratio = 0.1
|
pytorch_model.bin
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:9117101fefd61c74201edac55a47418f9137d6381fcc327db9cb9b780551a93d
|
3 |
+
size 278059174
|