corrected change to 64
Browse files- myolmoe/config.json +2 -3
myolmoe/config.json
CHANGED
|
@@ -15,7 +15,7 @@
|
|
| 15 |
"norm_topk_prob": false,
|
| 16 |
"num_attention_heads": 16,
|
| 17 |
"num_experts": 64,
|
| 18 |
-
"num_small_experts":
|
| 19 |
"small_expert_intermediate_size": 0,
|
| 20 |
"num_experts_per_tok": 2,
|
| 21 |
"num_hidden_layers": 16,
|
|
@@ -34,6 +34,5 @@
|
|
| 34 |
"small_expert_intermediate_ratio": 16,
|
| 35 |
"small_expert_count": 64,
|
| 36 |
"small_expert_sparsity_coef": 0.1,
|
| 37 |
-
"max_small_expert_count": 64
|
| 38 |
-
"num_small_experts": 64
|
| 39 |
}
|
|
|
|
| 15 |
"norm_topk_prob": false,
|
| 16 |
"num_attention_heads": 16,
|
| 17 |
"num_experts": 64,
|
| 18 |
+
"num_small_experts": 64,
|
| 19 |
"small_expert_intermediate_size": 0,
|
| 20 |
"num_experts_per_tok": 2,
|
| 21 |
"num_hidden_layers": 16,
|
|
|
|
| 34 |
"small_expert_intermediate_ratio": 16,
|
| 35 |
"small_expert_count": 64,
|
| 36 |
"small_expert_sparsity_coef": 0.1,
|
| 37 |
+
"max_small_expert_count": 64
|
|
|
|
| 38 |
}
|