Update configuration_llama_nsa.py

configuration_llama_nsa.py

@@ -185,6 +185,7 @@ class LlamaNSAConfig(PretrainedConfig):
 	block_size=64,
 	block_counts=16,
 	window_size=512,
+    inference_mode="sparse",
         **kwargs,
     ):
         self.vocab_size = vocab_size
@@ -214,6 +215,7 @@ class LlamaNSAConfig(PretrainedConfig):
         self.block_size = block_size # debug
         self.block_counts = block_counts
         self.window_size = window_size
+        self.inference_mode = inference_mode
 
         # Validate the correctness of rotary position embeddings parameters
         # BC: if there is a 'type' field, copy it it to 'rope_type'.
@@ -230,4 +232,4 @@ class LlamaNSAConfig(PretrainedConfig):
         )
 
 
-__all__ = ["LlamaNSAConfig"]
+__all__ = ["LlamaNSAConfig"]