grokadamw.GrokAdamW
Browse files
scripts/pretrain-core-model.yaml
CHANGED
|
@@ -18,7 +18,7 @@ model_config:
|
|
| 18 |
bias: False
|
| 19 |
norm_class_name: "RMSNorm"
|
| 20 |
mlp_class_name: "LLaMAMLP"
|
| 21 |
-
intermediate_size:
|
| 22 |
norm_eps: 1e-5
|
| 23 |
rope_base: 500000
|
| 24 |
rope_adjustments:
|
|
|
|
| 18 |
bias: False
|
| 19 |
norm_class_name: "RMSNorm"
|
| 20 |
mlp_class_name: "LLaMAMLP"
|
| 21 |
+
intermediate_size: 2688 # n_embd * 5.25
|
| 22 |
norm_eps: 1e-5
|
| 23 |
rope_base: 500000
|
| 24 |
rope_adjustments:
|