patrickvonplaten
commited on
Commit
•
9323626
1
Parent(s):
083e923
update model for pretraining compatible
Browse files- config.json +15 -3
- pytorch_model.bin +2 -2
config.json
CHANGED
@@ -2,10 +2,12 @@
|
|
2 |
"activation_dropout": 0.0,
|
3 |
"apply_spec_augment": true,
|
4 |
"architectures": [
|
5 |
-
"
|
6 |
],
|
7 |
"attention_dropout": 0.1,
|
8 |
"bos_token_id": 1,
|
|
|
|
|
9 |
"conv_bias": false,
|
10 |
"conv_dim": [
|
11 |
512,
|
@@ -34,19 +36,23 @@
|
|
34 |
2,
|
35 |
2
|
36 |
],
|
|
|
|
|
|
|
37 |
"do_stable_layer_norm": false,
|
38 |
"eos_token_id": 2,
|
39 |
"feat_extract_activation": "gelu",
|
40 |
"feat_extract_norm": "group",
|
41 |
"feat_proj_dropout": 0.1,
|
|
|
42 |
"final_dropout": 0.0,
|
43 |
"freeze_feat_extract_train": true,
|
|
|
44 |
"hidden_act": "gelu",
|
45 |
"hidden_dropout": 0.1,
|
46 |
"hidden_size": 768,
|
47 |
"initializer_range": 0.02,
|
48 |
"intermediate_size": 3072,
|
49 |
-
"gradient_checkpointing": true,
|
50 |
"layer_norm_eps": 1e-05,
|
51 |
"layerdrop": 0.05,
|
52 |
"mask_channel_length": 10,
|
@@ -54,6 +60,8 @@
|
|
54 |
"mask_channel_other": 0.0,
|
55 |
"mask_channel_prob": 0.0,
|
56 |
"mask_channel_selection": "static",
|
|
|
|
|
57 |
"mask_time_length": 10,
|
58 |
"mask_time_min_space": 1,
|
59 |
"mask_time_other": 0.0,
|
@@ -63,11 +71,15 @@
|
|
63 |
"no_mask_channel_overlap": false,
|
64 |
"no_mask_time_overlap": false,
|
65 |
"num_attention_heads": 12,
|
|
|
|
|
66 |
"num_conv_pos_embedding_groups": 16,
|
67 |
"num_conv_pos_embeddings": 128,
|
68 |
"num_feat_extract_layers": 7,
|
69 |
"num_hidden_layers": 12,
|
|
|
70 |
"pad_token_id": 0,
|
71 |
-
"
|
|
|
72 |
"vocab_size": 32
|
73 |
}
|
|
|
2 |
"activation_dropout": 0.0,
|
3 |
"apply_spec_augment": true,
|
4 |
"architectures": [
|
5 |
+
"Wav2Vec2ForPreTraining"
|
6 |
],
|
7 |
"attention_dropout": 0.1,
|
8 |
"bos_token_id": 1,
|
9 |
+
"codevector_dim": 256,
|
10 |
+
"contrastive_logits_temperature": 0.1,
|
11 |
"conv_bias": false,
|
12 |
"conv_dim": [
|
13 |
512,
|
|
|
36 |
2,
|
37 |
2
|
38 |
],
|
39 |
+
"ctc_loss_reduction": "sum",
|
40 |
+
"ctc_zero_infinity": false,
|
41 |
+
"diversity_loss_weight": 0.1,
|
42 |
"do_stable_layer_norm": false,
|
43 |
"eos_token_id": 2,
|
44 |
"feat_extract_activation": "gelu",
|
45 |
"feat_extract_norm": "group",
|
46 |
"feat_proj_dropout": 0.1,
|
47 |
+
"feat_quantizer_dropout": 0.0,
|
48 |
"final_dropout": 0.0,
|
49 |
"freeze_feat_extract_train": true,
|
50 |
+
"gradient_checkpointing": true,
|
51 |
"hidden_act": "gelu",
|
52 |
"hidden_dropout": 0.1,
|
53 |
"hidden_size": 768,
|
54 |
"initializer_range": 0.02,
|
55 |
"intermediate_size": 3072,
|
|
|
56 |
"layer_norm_eps": 1e-05,
|
57 |
"layerdrop": 0.05,
|
58 |
"mask_channel_length": 10,
|
|
|
60 |
"mask_channel_other": 0.0,
|
61 |
"mask_channel_prob": 0.0,
|
62 |
"mask_channel_selection": "static",
|
63 |
+
"mask_feature_length": 10,
|
64 |
+
"mask_feature_prob": 0.0,
|
65 |
"mask_time_length": 10,
|
66 |
"mask_time_min_space": 1,
|
67 |
"mask_time_other": 0.0,
|
|
|
71 |
"no_mask_channel_overlap": false,
|
72 |
"no_mask_time_overlap": false,
|
73 |
"num_attention_heads": 12,
|
74 |
+
"num_codevector_groups": 2,
|
75 |
+
"num_codevectors_per_group": 320,
|
76 |
"num_conv_pos_embedding_groups": 16,
|
77 |
"num_conv_pos_embeddings": 128,
|
78 |
"num_feat_extract_layers": 7,
|
79 |
"num_hidden_layers": 12,
|
80 |
+
"num_negatives": 100,
|
81 |
"pad_token_id": 0,
|
82 |
+
"proj_codevector_dim": 256,
|
83 |
+
"transformers_version": "4.7.0.dev0",
|
84 |
"vocab_size": 32
|
85 |
}
|
pytorch_model.bin
CHANGED
@@ -1,3 +1,3 @@
|
|
1 |
version https://git-lfs.github.com/spec/v1
|
2 |
-
oid sha256:
|
3 |
-
size
|
|
|
1 |
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:3249fe98bfc62fcbc26067f724716a6ec49d12c4728a2af1df659013905dff21
|
3 |
+
size 380267417
|