Update configuration_deepseek_v32.py
Browse files
configuration_deepseek_v32.py
CHANGED
|
@@ -175,7 +175,7 @@ class DeepseekV32Config(PretrainedConfig):
|
|
| 175 |
self.hidden_size = hidden_size
|
| 176 |
self.intermediate_size = intermediate_size
|
| 177 |
self.num_hidden_layers = num_hidden_layers
|
| 178 |
-
self.
|
| 179 |
|
| 180 |
self.moe_intermediate_size = moe_intermediate_size
|
| 181 |
self.num_attention_heads = num_attention_heads
|
|
|
|
| 175 |
self.hidden_size = hidden_size
|
| 176 |
self.intermediate_size = intermediate_size
|
| 177 |
self.num_hidden_layers = num_hidden_layers
|
| 178 |
+
self.first_k_dense_replace = first_k_dense_replace
|
| 179 |
|
| 180 |
self.moe_intermediate_size = moe_intermediate_size
|
| 181 |
self.num_attention_heads = num_attention_heads
|