Upload model
Browse files- config.json +3 -5
- pytorch_model.bin +1 -1
config.json
CHANGED
|
@@ -1,11 +1,9 @@
|
|
| 1 |
{
|
|
|
|
| 2 |
"architectures": [
|
| 3 |
-
"
|
| 4 |
],
|
| 5 |
"attention_bias": false,
|
| 6 |
-
"auto_map": {
|
| 7 |
-
"AutoConfig": "modelling_sparrow.SparrowConfig"
|
| 8 |
-
},
|
| 9 |
"dropout": 0.0,
|
| 10 |
"flash_attn": true,
|
| 11 |
"hidden_dim": 512,
|
|
@@ -14,7 +12,7 @@
|
|
| 14 |
"max_seq_len": 512,
|
| 15 |
"mlp_bias": false,
|
| 16 |
"model_type": "sparrow",
|
| 17 |
-
"norm_eps": 1e-
|
| 18 |
"num_attention_heads": 16,
|
| 19 |
"num_hidden_layers": 8,
|
| 20 |
"num_key_value_heads": 16,
|
|
|
|
| 1 |
{
|
| 2 |
+
"_name_or_path": "/data/sparrow/model/pretrain/",
|
| 3 |
"architectures": [
|
| 4 |
+
"SparrowModelForCausalLM"
|
| 5 |
],
|
| 6 |
"attention_bias": false,
|
|
|
|
|
|
|
|
|
|
| 7 |
"dropout": 0.0,
|
| 8 |
"flash_attn": true,
|
| 9 |
"hidden_dim": 512,
|
|
|
|
| 12 |
"max_seq_len": 512,
|
| 13 |
"mlp_bias": false,
|
| 14 |
"model_type": "sparrow",
|
| 15 |
+
"norm_eps": "1e-5",
|
| 16 |
"num_attention_heads": 16,
|
| 17 |
"num_hidden_layers": 8,
|
| 18 |
"num_key_value_heads": 16,
|
pytorch_model.bin
CHANGED
|
@@ -1,3 +1,3 @@
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:
|
| 3 |
size 204011452
|
|
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:3f24f36c610a73c1b64408eef0ca529c4a87e8461cf17446e276eb2ee26b8ebc
|
| 3 |
size 204011452
|