MLX
speech
leowangxyz commited on
Commit
0064cdd
·
verified ·
1 Parent(s): 2a2bed1

Upload 2 files

Browse files
Files changed (2) hide show
  1. config.json +74 -0
  2. weights.safetensors +3 -0
config.json ADDED
@@ -0,0 +1,74 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "extractor_mode": "default",
3
+ "encoder_layers": 12,
4
+ "encoder_embed_dim": 768,
5
+ "encoder_ffn_embed_dim": 3072,
6
+ "encoder_attention_heads": 12,
7
+ "activation_fn": "gelu",
8
+ "dropout": 0.1,
9
+ "attention_dropout": 0.1,
10
+ "activation_dropout": 0.0,
11
+ "encoder_layerdrop": 0.05,
12
+ "dropout_input": 0.1,
13
+ "dropout_features": 0.1,
14
+ "layer_norm_first": false,
15
+ "conv_feature_layers": "[(512,10,5)] + [(512,3,2)] * 4 + [(512,2,2)] * 2",
16
+ "conv_bias": false,
17
+ "feature_grad_mult": 0.1,
18
+ "mask_length": 10,
19
+ "mask_prob": 0.8,
20
+ "mask_selection": "static",
21
+ "mask_other": 0.0,
22
+ "no_mask_overlap": false,
23
+ "mask_min_space": 1,
24
+ "mask_channel_length": 10,
25
+ "mask_channel_prob": 0.0,
26
+ "mask_channel_selection": "static",
27
+ "mask_channel_other": 0.0,
28
+ "no_mask_channel_overlap": false,
29
+ "mask_channel_min_space": 1,
30
+ "conv_pos": 128,
31
+ "conv_pos_groups": 16,
32
+ "relative_position_embedding": true,
33
+ "num_buckets": 320,
34
+ "max_distance": 800,
35
+ "gru_rel_pos": true,
36
+ "normalize": false,
37
+ "conv_feature_layers_list": [
38
+ [
39
+ 512,
40
+ 10,
41
+ 5
42
+ ],
43
+ [
44
+ 512,
45
+ 3,
46
+ 2
47
+ ],
48
+ [
49
+ 512,
50
+ 3,
51
+ 2
52
+ ],
53
+ [
54
+ 512,
55
+ 3,
56
+ 2
57
+ ],
58
+ [
59
+ 512,
60
+ 3,
61
+ 2
62
+ ],
63
+ [
64
+ 512,
65
+ 2,
66
+ 2
67
+ ],
68
+ [
69
+ 512,
70
+ 2,
71
+ 2
72
+ ]
73
+ ]
74
+ }
weights.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:5031c85444aac1dff58bd0c40b92105ed9cbcc3cabe5b7f679134ecadb8615d8
3
+ size 377554512