Jyhan003 commited on
Commit
973ad07
·
1 Parent(s): 1cd51cc

make compatible with inference pipeline

Browse files
Files changed (2) hide show
  1. config.toml +16 -1
  2. pytorch_model.bin +2 -2
config.toml CHANGED
@@ -1,6 +1,7 @@
1
  [model]
2
  path = "diarizen.models.eend.model_wavlm_conformer.Model"
3
  [model.args]
 
4
  wavlm_layer_num = 13
5
  wavlm_feat_dim = 768
6
  attention_in = 256
@@ -12,4 +13,18 @@ chunk_size = 8
12
  use_posi = false
13
  output_activate_function = false
14
  selected_channel = 0
15
- max_speakers_per_chunk = 4
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
  [model]
2
  path = "diarizen.models.eend.model_wavlm_conformer.Model"
3
  [model.args]
4
+ wavlm_src = "wavlm_base"
5
  wavlm_layer_num = 13
6
  wavlm_feat_dim = 768
7
  attention_in = 256
 
13
  use_posi = false
14
  output_activate_function = false
15
  selected_channel = 0
16
+
17
+ [inference.args]
18
+ seg_duration = 8
19
+ segmentation_step = 0.1
20
+ batch_size = 32
21
+ apply_median_filtering = false
22
+
23
+ [clustering.args]
24
+ method = "AgglomerativeClustering"
25
+ merge_delta = 0.5
26
+ merge_max_length = 50
27
+ min_speakers = 2
28
+ max_speakers = 8
29
+ ahc_threshold = 0.7
30
+ min_cluster_size = 30
pytorch_model.bin CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:367e204af207aa02a1c9038ee922e9cdb462c6dfa5b3eb10147c197d64d87378
3
- size 402860182
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:9c4c4ee09ed5e5ab0982fe732e44268079fbad8adb3d69acfc4517c6448974e9
3
+ size 402866710