root
commited on
Commit
•
5473fae
1
Parent(s):
67922e4
upload config.json to resolve from_pretrained error
Browse files- config.json +22 -0
config.json
ADDED
|
@@ -0,0 +1,22 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"activation_function": "swiglu",
|
| 3 |
+
"bos_token_id": 13,
|
| 4 |
+
"eos_token_id": 13,
|
| 5 |
+
"hidden_act": "silu",
|
| 6 |
+
"hidden_size": 2304,
|
| 7 |
+
"initializer_range": 0.02,
|
| 8 |
+
"intermediate_size": 6144,
|
| 9 |
+
"max_position_embeddings": 16384,
|
| 10 |
+
"num_attention_heads": 18,
|
| 11 |
+
"num_hidden_layers": 28,
|
| 12 |
+
"num_key_value_heads": 18,
|
| 13 |
+
"pad_token_id": 0,
|
| 14 |
+
"pretraining_tp": 1,
|
| 15 |
+
"rms_norm_eps": 1e-06,
|
| 16 |
+
"tie_word_embeddings": false,
|
| 17 |
+
"transformers_version": "4.31.0",
|
| 18 |
+
"use_cache": true,
|
| 19 |
+
"vocab_size": 96000,
|
| 20 |
+
"rope_scaling": {"type":"linear", "factor": 4.0 }
|
| 21 |
+
}
|
| 22 |
+
|