intermediate_size: 4096
Browse files
scripts/pretrain-core-model.yaml
CHANGED
|
@@ -10,15 +10,15 @@ model_config:
|
|
| 10 |
vocab_size: 32064
|
| 11 |
padded_vocab_size: 32064
|
| 12 |
n_layer: 32
|
| 13 |
-
n_head:
|
| 14 |
-
n_embd:
|
| 15 |
-
n_query_groups:
|
| 16 |
rotary_percentage: 1.0
|
| 17 |
parallel_residual: False
|
| 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:
|
|
|
|
| 10 |
vocab_size: 32064
|
| 11 |
padded_vocab_size: 32064
|
| 12 |
n_layer: 32
|
| 13 |
+
n_head: 4
|
| 14 |
+
n_embd: 512
|
| 15 |
+
n_query_groups: 4
|
| 16 |
rotary_percentage: 1.0
|
| 17 |
parallel_residual: False
|
| 18 |
bias: False
|
| 19 |
norm_class_name: "RMSNorm"
|
| 20 |
mlp_class_name: "LLaMAMLP"
|
| 21 |
+
intermediate_size: 4096
|
| 22 |
norm_eps: 1e-5
|
| 23 |
rope_base: 500000
|
| 24 |
rope_adjustments:
|