base

dreaming-panda · dreaming-panda · commit ce3c81010e61 · 2024-12-16T12:34:25.000-05:00
diff --git a/evaluations/RULER/pred/attnserver_dist.py b/evaluations/RULER/pred/attnserver_dist.py
@@ -907,7 +907,7 @@ def __init__(self,
         num_local_tokens :int = 64,
         generation_buffer :int = 256,
         max_length: int = 8192,
-        dense_layers: list[int] = list(range(32)),
+        dense_layers: list[int] = list(range(80)),
         device :str = 'cuda:0',
         dtype = torch.bfloat16) -> None:
         
