Hugging Face's logo

Lakoc
/
zipformer2_base

Transformers
wav2vec2-zipformer
Model card Files
xet
 Community
Lakoc commited on
Commit
ce50b41
·
1 Parent(s): 10b30bc

Upload config

Browse files
Files changed (1) hide show
  1. config.json +182 -0
config.json ADDED
@@ -0,0 +1,182 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "_name_or_path": "Lakoc/zipformer2_medium",
3
+ "activation_dropout": 0.1,
4
+ "adapter_attn_dim": null,
5
+ "adapter_kernel_size": 3,
6
+ "adapter_stride": 2,
7
+ "add_adapter": false,
8
+ "apply_spec_augment": true,
9
+ "attention_dropout": 0.1,
10
+ "bos_token_id": 1,
11
+ "causal": false,
12
+ "chunk_size": [
13
+ 16,
14
+ 32,
15
+ 64,
16
+ -1
17
+ ],
18
+ "classifier_proj_size": 256,
19
+ "cnn_module_kernel": [
20
+ 31,
21
+ 31,
22
+ 15,
23
+ 15,
24
+ 15,
25
+ 31
26
+ ],
27
+ "codevector_dim": 256,
28
+ "contrastive_logits_temperature": 0.1,
29
+ "conv_bias": false,
30
+ "conv_depthwise_kernel_size": 31,
31
+ "conv_dim": [
32
+ 512,
33
+ 512,
34
+ 512,
35
+ 512,
36
+ 512,
37
+ 512,
38
+ 512
39
+ ],
40
+ "conv_kernel": [
41
+ 10,
42
+ 3,
43
+ 3,
44
+ 3,
45
+ 3,
46
+ 2,
47
+ 2
48
+ ],
49
+ "conv_stride": [
50
+ 5,
51
+ 2,
52
+ 2,
53
+ 2,
54
+ 2,
55
+ 2,
56
+ 2
57
+ ],
58
+ "ctc_loss_reduction": "sum",
59
+ "ctc_zero_infinity": false,
60
+ "diversity_loss_weight": 0.1,
61
+ "do_stable_layer_norm": true,
62
+ "downsampling_factor": [
63
+ 1,
64
+ 2,
65
+ 4,
66
+ 8,
67
+ 4,
68
+ 2
69
+ ],
70
+ "dropout": 0.1,
71
+ "encoder_dim": [
72
+ 192,
73
+ 256,
74
+ 512,
75
+ 768,
76
+ 512,
77
+ 256
78
+ ],
79
+ "encoder_unmasked_dim": [
80
+ 192,
81
+ 192,
82
+ 384,
83
+ 512,
84
+ 384,
85
+ 256
86
+ ],
87
+ "eos_token_id": 2,
88
+ "feat_extract_activation": "gelu",
89
+ "feat_extract_norm": "layer",
90
+ "feat_proj_dropout": 0.0,
91
+ "feat_quantizer_dropout": 0.0,
92
+ "feedforward_dim": [
93
+ 384,
94
+ 512,
95
+ 1024,
96
+ 1536,
97
+ 1024,
98
+ 512
99
+ ],
100
+ "final_dropout": 0.1,
101
+ "hidden_act": "gelu",
102
+ "hidden_dropout": 0.1,
103
+ "hidden_size": 192,
104
+ "initializer_range": 0.02,
105
+ "intermediate_size": 2048,
106
+ "layer_norm_eps": 1e-05,
107
+ "layerdrop": 0.0,
108
+ "left_context_frames": [
109
+ 64,
110
+ 128,
111
+ 256,
112
+ -1
113
+ ],
114
+ "mask_feature_length": 10,
115
+ "mask_feature_min_masks": 0,
116
+ "mask_feature_prob": 0.0,
117
+ "mask_time_length": 10,
118
+ "mask_time_min_masks": 2,
119
+ "mask_time_prob": 0.05,
120
+ "max_source_positions": 5000,
121
+ "model_type": "wav2vec2-zipformer",
122
+ "num_adapter_layers": 3,
123
+ "num_attention_heads": 8,
124
+ "num_codevector_groups": 2,
125
+ "num_codevectors_per_group": 320,
126
+ "num_conv_pos_embedding_groups": 16,
127
+ "num_conv_pos_embeddings": 128,
128
+ "num_encoder_layers": [
129
+ 2,
130
+ 2,
131
+ 3,
132
+ 4,
133
+ 3,
134
+ 2
135
+ ],
136
+ "num_feat_extract_layers": 7,
137
+ "num_heads": [
138
+ 4,
139
+ 4,
140
+ 4,
141
+ 8,
142
+ 4,
143
+ 4
144
+ ],
145
+ "num_hidden_layers": 12,
146
+ "num_mel_bins": 84,
147
+ "num_negatives": 100,
148
+ "output_downsampling_factor": 2,
149
+ "output_hidden_size": 512,
150
+ "pad_token_id": 0,
151
+ "pos_dim": 48,
152
+ "pos_head_dim": 4,
153
+ "proj_codevector_dim": 256,
154
+ "query_head_dim": 32,
155
+ "tdnn_dilation": [
156
+ 1,
157
+ 2,
158
+ 3,
159
+ 1,
160
+ 1
161
+ ],
162
+ "tdnn_dim": [
163
+ 512,
164
+ 512,
165
+ 512,
166
+ 512,
167
+ 1500
168
+ ],
169
+ "tdnn_kernel": [
170
+ 5,
171
+ 3,
172
+ 3,
173
+ 1,
174
+ 1
175
+ ],
176
+ "transformers_version": "4.34.1",
177
+ "use_weighted_layer_sum": false,
178
+ "value_head_dim": 12,
179
+ "vocab_size": null,
180
+ "warmup_batches": 32000.0,
181
+ "xvector_output_dim": 512
182
+ }