Commit
•
aae5600
1
Parent(s):
322a106
Upload config.json with huggingface_hub
Browse files- config.json +28 -0
config.json
ADDED
|
@@ -0,0 +1,28 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"model_type": "mamba",
|
| 3 |
+
"d_model": 768,
|
| 4 |
+
"d_intermediate": 0,
|
| 5 |
+
"n_layer": 12,
|
| 6 |
+
"vocab_size": 1880,
|
| 7 |
+
"ssm_cfg": {
|
| 8 |
+
"layer": "Mamba2"
|
| 9 |
+
},
|
| 10 |
+
"attn_layer_idx": [],
|
| 11 |
+
"attn_cfg": {},
|
| 12 |
+
"rms_norm": true,
|
| 13 |
+
"residual_in_fp32": true,
|
| 14 |
+
"fused_add_norm": true,
|
| 15 |
+
"pad_vocab_size_multiple": 16,
|
| 16 |
+
"tie_embeddings": true,
|
| 17 |
+
"dropout_rate": 0.1,
|
| 18 |
+
"eos_token_id": 2,
|
| 19 |
+
"bos_token_id": 1,
|
| 20 |
+
"pad_token_id": 3,
|
| 21 |
+
"num_labels": 1,
|
| 22 |
+
"summary_type": "cls_index",
|
| 23 |
+
"summary_use_proj": true,
|
| 24 |
+
"summary_activation": "tanh",
|
| 25 |
+
"summary_proj_to_labels": true,
|
| 26 |
+
"summary_first_dropout": 0.4,
|
| 27 |
+
"summary_hidden_size": 128
|
| 28 |
+
}
|