Update config.json
Browse files- config.json +10 -0
config.json
CHANGED
|
@@ -4,11 +4,21 @@
|
|
| 4 |
],
|
| 5 |
"model_type": "mixture_of_recursions",
|
| 6 |
"vocab_size": 31985,
|
|
|
|
| 7 |
"hidden_size": 384,
|
|
|
|
| 8 |
"num_hidden_layers": 12,
|
|
|
|
| 9 |
"num_attention_heads": 6,
|
|
|
|
| 10 |
"intermediate_size": 2048,
|
|
|
|
| 11 |
"max_position_embeddings": 128,
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 12 |
"torch_dtype": "float32",
|
| 13 |
"transformers_version": "4.57.0",
|
| 14 |
"auto_map": {
|
|
|
|
| 4 |
],
|
| 5 |
"model_type": "mixture_of_recursions",
|
| 6 |
"vocab_size": 31985,
|
| 7 |
+
"d_model": 384,
|
| 8 |
"hidden_size": 384,
|
| 9 |
+
"n_layers": 12,
|
| 10 |
"num_hidden_layers": 12,
|
| 11 |
+
"n_heads": 6,
|
| 12 |
"num_attention_heads": 6,
|
| 13 |
+
"dim_feedforward": 2048,
|
| 14 |
"intermediate_size": 2048,
|
| 15 |
+
"max_seq_len": 128,
|
| 16 |
"max_position_embeddings": 128,
|
| 17 |
+
"max_steps": 4,
|
| 18 |
+
"dropout": 0.1,
|
| 19 |
+
"router_type": "adaptive",
|
| 20 |
+
"padding_idx": 0,
|
| 21 |
+
"pos_encoding": "learned",
|
| 22 |
"torch_dtype": "float32",
|
| 23 |
"transformers_version": "4.57.0",
|
| 24 |
"auto_map": {
|