Hugging Face's logo

Lakoc
/
zipformer2_medium

Transformers
wav2vec2-zipformer
Model card Files
xet
Community
Lakoc commited on
Commit
c1647e0
·
1 Parent(s): c8596e9

Upload config

Browse files
Files changed (1) hide show
  1. config.json +178 -0
config.json ADDED
@@ -0,0 +1,178 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "activation_dropout": 0.1,
3
+ "adapter_attn_dim": null,
4
+ "adapter_kernel_size": 3,
5
+ "adapter_stride": 2,
6
+ "add_adapter": false,
7
+ "apply_spec_augment": true,
8
+ "attention_dropout": 0.1,
9
+ "bos_token_id": 1,
10
+ "causal": false,
11
+ "chunk_size": [
12
+ 16,
13
+ 32,
14
+ 64,
15
+ -1
16
+ ],
17
+ "classifier_proj_size": 256,
18
+ "cnn_module_kernel": [
19
+ 31,
20
+ 31,
21
+ 15,
22
+ 15,
23
+ 15,
24
+ 31
25
+ ],
26
+ "codevector_dim": 256,
27
+ "contrastive_logits_temperature": 0.1,
28
+ "conv_bias": false,
29
+ "conv_dim": [
30
+ 512,
31
+ 512,
32
+ 512,
33
+ 512,
34
+ 512,
35
+ 512,
36
+ 512
37
+ ],
38
+ "conv_kernel": [
39
+ 10,
40
+ 3,
41
+ 3,
42
+ 3,
43
+ 3,
44
+ 2,
45
+ 2
46
+ ],
47
+ "conv_stride": [
48
+ 5,
49
+ 2,
50
+ 2,
51
+ 2,
52
+ 2,
53
+ 2,
54
+ 2
55
+ ],
56
+ "ctc_loss_reduction": "sum",
57
+ "ctc_zero_infinity": false,
58
+ "diversity_loss_weight": 0.1,
59
+ "do_stable_layer_norm": false,
60
+ "downsampling_factor": [
61
+ 1,
62
+ 2,
63
+ 4,
64
+ 8,
65
+ 4,
66
+ 2
67
+ ],
68
+ "dropout": 0.1,
69
+ "encoder_dim": [
70
+ 192,
71
+ 256,
72
+ 384,
73
+ 512,
74
+ 384,
75
+ 192
76
+ ],
77
+ "encoder_unmasked_dim": [
78
+ 192,
79
+ 192,
80
+ 256,
81
+ 256,
82
+ 256,
83
+ 192
84
+ ],
85
+ "eos_token_id": 2,
86
+ "feat_extract_activation": "gelu",
87
+ "feat_extract_norm": "group",
88
+ "feat_proj_dropout": 0.0,
89
+ "feat_quantizer_dropout": 0.0,
90
+ "feedforward_dim": [
91
+ 512,
92
+ 768,
93
+ 1024,
94
+ 1536,
95
+ 1024,
96
+ 512
97
+ ],
98
+ "final_dropout": 0.1,
99
+ "hidden_act": "gelu",
100
+ "hidden_dropout": 0.1,
101
+ "hidden_size": 192,
102
+ "initializer_range": 0.02,
103
+ "intermediate_size": 3072,
104
+ "layer_norm_eps": 1e-05,
105
+ "layerdrop": 0.1,
106
+ "left_context_frames": [
107
+ 64,
108
+ 128,
109
+ 256,
110
+ -1
111
+ ],
112
+ "mask_feature_length": 10,
113
+ "mask_feature_min_masks": 0,
114
+ "mask_feature_prob": 0.0,
115
+ "mask_time_length": 10,
116
+ "mask_time_min_masks": 2,
117
+ "mask_time_prob": 0.05,
118
+ "model_type": "wav2vec2-zipformer",
119
+ "num_adapter_layers": 3,
120
+ "num_attention_heads": 12,
121
+ "num_codevector_groups": 2,
122
+ "num_codevectors_per_group": 320,
123
+ "num_conv_pos_embedding_groups": 16,
124
+ "num_conv_pos_embeddings": 128,
125
+ "num_encoder_layers": [
126
+ 2,
127
+ 2,
128
+ 3,
129
+ 4,
130
+ 3,
131
+ 2
132
+ ],
133
+ "num_feat_extract_layers": 7,
134
+ "num_heads": [
135
+ 4,
136
+ 4,
137
+ 4,
138
+ 8,
139
+ 4,
140
+ 4
141
+ ],
142
+ "num_hidden_layers": 12,
143
+ "num_negatives": 100,
144
+ "output_downsampling_factor": 2,
145
+ "output_hidden_size": 768,
146
+ "pad_token_id": 0,
147
+ "pos_dim": 48,
148
+ "pos_head_dim": 4,
149
+ "proj_codevector_dim": 256,
150
+ "query_head_dim": 32,
151
+ "tdnn_dilation": [
152
+ 1,
153
+ 2,
154
+ 3,
155
+ 1,
156
+ 1
157
+ ],
158
+ "tdnn_dim": [
159
+ 512,
160
+ 512,
161
+ 512,
162
+ 512,
163
+ 1500
164
+ ],
165
+ "tdnn_kernel": [
166
+ 5,
167
+ 3,
168
+ 3,
169
+ 1,
170
+ 1
171
+ ],
172
+ "transformers_version": "4.34.1",
173
+ "use_weighted_layer_sum": false,
174
+ "value_head_dim": 12,
175
+ "vocab_size": 32,
176
+ "warmup_batches": 32000.0,
177
+ "xvector_output_dim": 512
178
+ }