Upload config.json
Browse files- config.json +28 -0
config.json
ADDED
|
@@ -0,0 +1,28 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"architectures": [
|
| 3 |
+
"LeanAlbertForPretraining", "LeanAlbertForTokenClassification", "LeanAlbertForSequenceClassification"
|
| 4 |
+
],
|
| 5 |
+
"model_type": "lean_albert",
|
| 6 |
+
"num_hidden_layers": 32,
|
| 7 |
+
"num_hidden_groups": 32,
|
| 8 |
+
"num_inner_groups": 1,
|
| 9 |
+
"share_large_matrices": true,
|
| 10 |
+
"adapter_dim": 32,
|
| 11 |
+
"hidden_size": 2560,
|
| 12 |
+
"intermediate_size": 10240,
|
| 13 |
+
"embedding_size": 256,
|
| 14 |
+
"num_attention_heads": 64,
|
| 15 |
+
"vocab_size": 999,
|
| 16 |
+
"hidden_act": "gelu_new",
|
| 17 |
+
"hidden_act_gated": true,
|
| 18 |
+
"sandwich_norm": true,
|
| 19 |
+
"position_embedding_type": "rotary",
|
| 20 |
+
"hidden_dropout_prob": 0,
|
| 21 |
+
"classifier_dropout_prob": 0.1,
|
| 22 |
+
"attention_probs_dropout_prob": 0,
|
| 23 |
+
"layer_norm_eps": 1e-12,
|
| 24 |
+
"type_vocab_size": 2,
|
| 25 |
+
"pad_token_id": 0,
|
| 26 |
+
"bos_token_id": 2,
|
| 27 |
+
"eos_token_id": 3
|
| 28 |
+
}
|