feat: added small config
Browse files- small_config.json +34 -0
small_config.json
ADDED
|
@@ -0,0 +1,34 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"_name_or_path": "jinaai/jina-bert-implementation",
|
| 3 |
+
"model_max_length": 8192,
|
| 4 |
+
"architectures": [
|
| 5 |
+
"JinaBertForMaskedLM"
|
| 6 |
+
],
|
| 7 |
+
"auto_map": {
|
| 8 |
+
"AutoConfig": "jinaai/jina-bert-flash-implementation--configuration_bert.JinaBertConfig",
|
| 9 |
+
"AutoModel": "jinaai/jina-bert-flash-implementation--modeling_bert.BertModel",
|
| 10 |
+
"AutoModelForPreTraining": "jinaai/jina-bert-flash-implementation--modeling_bert.BertForPreTraining",
|
| 11 |
+
"AutoModelForMaskedLM": "jinaai/jina-bert-flash-implementation--modeling_bert.BertForPreTraining"
|
| 12 |
+
},
|
| 13 |
+
"vocab_size": 30528,
|
| 14 |
+
"hidden_size": 512,
|
| 15 |
+
"num_hidden_layers": 4,
|
| 16 |
+
"num_attention_heads": 8,
|
| 17 |
+
"intermediate_size": 2048,
|
| 18 |
+
"hidden_act": "gelu",
|
| 19 |
+
"hidden_dropout_prob": 0.1,
|
| 20 |
+
"attention_probs_dropout_prob": 0.1,
|
| 21 |
+
"type_vocab_size": 0,
|
| 22 |
+
"initializer_range": 0.02,
|
| 23 |
+
"layer_norm_eps": 1e-12,
|
| 24 |
+
"pad_token_id": 0,
|
| 25 |
+
"dense_seq_output": true,
|
| 26 |
+
"fused_mlp": false,
|
| 27 |
+
"mlp_checkpoint_lvl": 0,
|
| 28 |
+
"last_layer_subset": false,
|
| 29 |
+
"fused_dropout_add_ln": false,
|
| 30 |
+
"fused_bias_fc": false,
|
| 31 |
+
"pad_vocab_size_multiple": 1,
|
| 32 |
+
"num_tasks": 6,
|
| 33 |
+
"use_flash_attn": true
|
| 34 |
+
}
|