Upload Wav2Vec2BertForMultilevelCTC
Browse files- config.json +98 -0
- model.safetensors +3 -0
config.json
ADDED
|
@@ -0,0 +1,98 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"activation_dropout": 0.0,
|
| 3 |
+
"adapter_act": "relu",
|
| 4 |
+
"adapter_kernel_size": 3,
|
| 5 |
+
"adapter_stride": 2,
|
| 6 |
+
"add_adapter": true,
|
| 7 |
+
"apply_spec_augment": true,
|
| 8 |
+
"architectures": [
|
| 9 |
+
"Wav2Vec2BertForMultilevelCTC"
|
| 10 |
+
],
|
| 11 |
+
"attention_dropout": 0.0,
|
| 12 |
+
"bos_token_id": 1,
|
| 13 |
+
"classifier_proj_size": 768,
|
| 14 |
+
"conformer_conv_dropout": 0.1,
|
| 15 |
+
"conv_depthwise_kernel_size": 31,
|
| 16 |
+
"ctc_loss_reduction": "mean",
|
| 17 |
+
"ctc_zero_infinity": false,
|
| 18 |
+
"eos_token_id": 2,
|
| 19 |
+
"feat_proj_dropout": 0.0,
|
| 20 |
+
"feature_projection_input_dim": 160,
|
| 21 |
+
"final_dropout": 0.1,
|
| 22 |
+
"hidden_act": "swish",
|
| 23 |
+
"hidden_dropout": 0.0,
|
| 24 |
+
"hidden_size": 1024,
|
| 25 |
+
"initializer_range": 0.02,
|
| 26 |
+
"intermediate_size": 4096,
|
| 27 |
+
"layer_norm_eps": 1e-05,
|
| 28 |
+
"layerdrop": 0.0,
|
| 29 |
+
"left_max_position_embeddings": 64,
|
| 30 |
+
"level_to_loss_weight": {
|
| 31 |
+
"ghonna": 0.05975,
|
| 32 |
+
"hams_or_jahr": 0.05975,
|
| 33 |
+
"istitala": 0.05975,
|
| 34 |
+
"itbaq": 0.05975,
|
| 35 |
+
"phonemes": 0.4,
|
| 36 |
+
"qalqla": 0.05975,
|
| 37 |
+
"safeer": 0.05975,
|
| 38 |
+
"shidda_or_rakhawa": 0.062,
|
| 39 |
+
"tafashie": 0.05975,
|
| 40 |
+
"tafkheem_or_taqeeq": 0.06,
|
| 41 |
+
"tikraar": 0.05975
|
| 42 |
+
},
|
| 43 |
+
"level_to_vocab_size": {
|
| 44 |
+
"ghonna": 3,
|
| 45 |
+
"hams_or_jahr": 3,
|
| 46 |
+
"istitala": 3,
|
| 47 |
+
"itbaq": 3,
|
| 48 |
+
"phonemes": 43,
|
| 49 |
+
"qalqla": 3,
|
| 50 |
+
"safeer": 3,
|
| 51 |
+
"shidda_or_rakhawa": 4,
|
| 52 |
+
"tafashie": 3,
|
| 53 |
+
"tafkheem_or_taqeeq": 3,
|
| 54 |
+
"tikraar": 3
|
| 55 |
+
},
|
| 56 |
+
"mask_feature_length": 10,
|
| 57 |
+
"mask_feature_min_masks": 0,
|
| 58 |
+
"mask_feature_prob": 0.0,
|
| 59 |
+
"mask_time_length": 10,
|
| 60 |
+
"mask_time_min_masks": 2,
|
| 61 |
+
"mask_time_prob": 0.0,
|
| 62 |
+
"max_source_positions": 5000,
|
| 63 |
+
"model_type": "multi_level_ctc",
|
| 64 |
+
"num_adapter_layers": 1,
|
| 65 |
+
"num_attention_heads": 16,
|
| 66 |
+
"num_hidden_layers": 24,
|
| 67 |
+
"output_hidden_size": 1024,
|
| 68 |
+
"pad_token_id": 0,
|
| 69 |
+
"position_embeddings_type": "relative_key",
|
| 70 |
+
"right_max_position_embeddings": 8,
|
| 71 |
+
"rotary_embedding_base": 10000,
|
| 72 |
+
"tdnn_dilation": [
|
| 73 |
+
1,
|
| 74 |
+
2,
|
| 75 |
+
3,
|
| 76 |
+
1,
|
| 77 |
+
1
|
| 78 |
+
],
|
| 79 |
+
"tdnn_dim": [
|
| 80 |
+
512,
|
| 81 |
+
512,
|
| 82 |
+
512,
|
| 83 |
+
512,
|
| 84 |
+
1500
|
| 85 |
+
],
|
| 86 |
+
"tdnn_kernel": [
|
| 87 |
+
5,
|
| 88 |
+
3,
|
| 89 |
+
3,
|
| 90 |
+
1,
|
| 91 |
+
1
|
| 92 |
+
],
|
| 93 |
+
"torch_dtype": "float32",
|
| 94 |
+
"transformers_version": "4.55.0",
|
| 95 |
+
"use_intermediate_ffn_before_adapter": false,
|
| 96 |
+
"use_weighted_layer_sum": false,
|
| 97 |
+
"xvector_output_dim": 512
|
| 98 |
+
}
|
model.safetensors
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:3102d98eff32afe653e487c6ea8232a1527af5f2bdca3cf686653e9faec19ecc
|
| 3 |
+
size 2423119912
|