Hugging Face's logo

ylacombe
/
w2v-bert-2.0

Feature Extraction
Transformers
Safetensors
wav2vec2-bert
Inference Endpoints
Model card Files Community
13

Upload Wav2Vec2BERTForPreTraining

#1
by ylacombe HF staff - opened
base: refs/heads/main
←
from: refs/pr/1
Discussion Files changed
+10
-46
Files changed (2) hide show
  1. config.json +8 -44
  2. model.safetensors +2 -2
config.json CHANGED
@@ -5,53 +5,24 @@
5
  "add_adapter": false,
6
  "apply_spec_augment": false,
7
  "architectures": [
8
- "Wav2Vec2ConformerForPreTraining"
9
  ],
10
  "attention_dropout": 0.0,
11
  "bos_token_id": 1,
12
  "classifier_proj_size": 768,
13
  "codevector_dim": 768,
14
- "conformer_conv_dropout": 0.0,
15
  "contrastive_logits_temperature": 0.1,
16
- "conv_bias": false,
17
  "conv_depthwise_kernel_size": 31,
18
- "conv_dim": [
19
- 512,
20
- 512,
21
- 512,
22
- 512,
23
- 512,
24
- 512,
25
- 160
26
- ],
27
- "conv_kernel": [
28
- 10,
29
- 3,
30
- 3,
31
- 3,
32
- 3,
33
- 2,
34
- 2
35
- ],
36
- "conv_stride": [
37
- 5,
38
- 2,
39
- 2,
40
- 2,
41
- 2,
42
- 2,
43
- 2
44
- ],
45
  "ctc_loss_reduction": "sum",
46
  "ctc_zero_infinity": false,
47
  "diversity_loss_weight": 0.1,
48
  "eos_token_id": 2,
49
- "feat_extract_activation": "gelu",
50
- "feat_extract_norm": "group",
51
  "feat_proj_dropout": 0.0,
52
  "feat_quantizer_dropout": 0.0,
 
53
  "final_dropout": 0.1,
54
- "hidden_act": "swish",
55
  "hidden_dropout": 0.0,
56
  "hidden_size": 1024,
57
  "initializer_range": 0.02,
@@ -64,17 +35,13 @@
64
  "mask_feature_prob": 0.0,
65
  "mask_time_length": 10,
66
  "mask_time_min_masks": 2,
67
- "mask_time_prob": 0.00,
68
  "max_source_positions": 5000,
69
- "model_type": "wav2vec2-conformer",
70
- "non_causal_depth_wise_conv": false,
71
- "num_adapter_layers": 0,
72
  "num_attention_heads": 16,
73
  "num_codevector_groups": 2,
74
  "num_codevectors_per_group": 320,
75
- "num_conv_pos_embedding_groups": 16,
76
- "num_conv_pos_embeddings": 128,
77
- "num_feat_extract_layers": 7,
78
  "num_hidden_layers": 24,
79
  "num_negatives": 100,
80
  "output_hidden_size": 1024,
@@ -83,9 +50,6 @@
83
  "proj_codevector_dim": 768,
84
  "right_max_position_embeddings": 8,
85
  "rotary_embedding_base": 10000,
86
- "skip_encoder_layer_norm": true,
87
- "skip_feature_encoder": true,
88
- "skip_pos_conv_embed": true,
89
  "tdnn_dilation": [
90
  1,
91
  2,
@@ -109,7 +73,7 @@
109
  ],
110
  "torch_dtype": "float32",
111
  "transformers_version": "4.37.0.dev0",
112
- "use_adapter_with_attention": false,
113
  "use_weighted_layer_sum": false,
114
  "vocab_size": null,
115
  "xvector_output_dim": 512
 
5
  "add_adapter": false,
6
  "apply_spec_augment": false,
7
  "architectures": [
8
+ "Wav2Vec2BERTForPreTraining"
9
  ],
10
  "attention_dropout": 0.0,
11
  "bos_token_id": 1,
12
  "classifier_proj_size": 768,
13
  "codevector_dim": 768,
14
+ "conformer_conv_dropout": 0.1,
15
  "contrastive_logits_temperature": 0.1,
 
16
  "conv_depthwise_kernel_size": 31,
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
17
  "ctc_loss_reduction": "sum",
18
  "ctc_zero_infinity": false,
19
  "diversity_loss_weight": 0.1,
20
  "eos_token_id": 2,
 
 
21
  "feat_proj_dropout": 0.0,
22
  "feat_quantizer_dropout": 0.0,
23
+ "feature_projection_input_dim": 160,
24
  "final_dropout": 0.1,
25
+ "hidden_act": "gelu",
26
  "hidden_dropout": 0.0,
27
  "hidden_size": 1024,
28
  "initializer_range": 0.02,
 
35
  "mask_feature_prob": 0.0,
36
  "mask_time_length": 10,
37
  "mask_time_min_masks": 2,
38
+ "mask_time_prob": 0.05,
39
  "max_source_positions": 5000,
40
+ "model_type": "wav2vec2-bert",
41
+ "num_adapter_layers": 1,
 
42
  "num_attention_heads": 16,
43
  "num_codevector_groups": 2,
44
  "num_codevectors_per_group": 320,
 
 
 
45
  "num_hidden_layers": 24,
46
  "num_negatives": 100,
47
  "output_hidden_size": 1024,
 
50
  "proj_codevector_dim": 768,
51
  "right_max_position_embeddings": 8,
52
  "rotary_embedding_base": 10000,
 
 
 
53
  "tdnn_dilation": [
54
  1,
55
  2,
 
73
  ],
74
  "torch_dtype": "float32",
75
  "transformers_version": "4.37.0.dev0",
76
+ "use_intermediate_ffn_before_adapter": false,
77
  "use_weighted_layer_sum": false,
78
  "vocab_size": null,
79
  "xvector_output_dim": 512
model.safetensors CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:5f09e925ebc936861e5d1f7320aff6e7bdd7f059c3becc3cffe79e1163eb09d7
3
- size 2328985440
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:4ce9f7bd65da973a161b8c1a276f8721eddd6c9add60970d4426065a44176e34
3
+ size 2328981576