scorevision: push artifact
Browse files- original/config.json +18 -0
original/config.json
ADDED
|
@@ -0,0 +1,18 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"num_hidden_layers": 24,
|
| 3 |
+
"num_experts": 32,
|
| 4 |
+
"experts_per_token": 4,
|
| 5 |
+
"vocab_size": 201088,
|
| 6 |
+
"hidden_size": 2880,
|
| 7 |
+
"intermediate_size": 2880,
|
| 8 |
+
"swiglu_limit": 7.0,
|
| 9 |
+
"head_dim": 64,
|
| 10 |
+
"num_attention_heads": 64,
|
| 11 |
+
"num_key_value_heads": 8,
|
| 12 |
+
"sliding_window": 128,
|
| 13 |
+
"initial_context_length": 4096,
|
| 14 |
+
"rope_theta": 150000,
|
| 15 |
+
"rope_scaling_factor": 32.0,
|
| 16 |
+
"rope_ntk_alpha": 1,
|
| 17 |
+
"rope_ntk_beta": 32
|
| 18 |
+
}
|