Upload config.json with huggingface_hub
Browse files- config.json +7 -7
config.json
CHANGED
|
@@ -3,12 +3,12 @@
|
|
| 3 |
"architectures": [
|
| 4 |
"DizelLM"
|
| 5 |
],
|
| 6 |
-
"vocab_size":
|
| 7 |
-
"context_length":
|
| 8 |
-
"d_model":
|
| 9 |
-
"n_layers":
|
| 10 |
-
"n_heads":
|
| 11 |
-
"kv_heads":
|
| 12 |
"head_dim": 64,
|
| 13 |
"intermediate_size": 3072,
|
| 14 |
"norm_type": "rmsnorm",
|
|
@@ -16,6 +16,6 @@
|
|
| 16 |
"rope_theta": 10000.0,
|
| 17 |
"bias": false,
|
| 18 |
"weight_tying": true,
|
| 19 |
-
"dropout": 0.
|
| 20 |
"format_version": "1.5"
|
| 21 |
}
|
|
|
|
| 3 |
"architectures": [
|
| 4 |
"DizelLM"
|
| 5 |
],
|
| 6 |
+
"vocab_size": 100000,
|
| 7 |
+
"context_length": 2048,
|
| 8 |
+
"d_model": 1024,
|
| 9 |
+
"n_layers": 28,
|
| 10 |
+
"n_heads": 16,
|
| 11 |
+
"kv_heads": 8,
|
| 12 |
"head_dim": 64,
|
| 13 |
"intermediate_size": 3072,
|
| 14 |
"norm_type": "rmsnorm",
|
|
|
|
| 16 |
"rope_theta": 10000.0,
|
| 17 |
"bias": false,
|
| 18 |
"weight_tying": true,
|
| 19 |
+
"dropout": 0.0,
|
| 20 |
"format_version": "1.5"
|
| 21 |
}
|