This commit is contained in:
Awni Hannun 2025-01-12 11:02:51 -08:00
parent 0e908bddff
commit 89c49d6e23

View File

@ -158,7 +158,7 @@ class CohereModel(nn.Module):
if cache is None: if cache is None:
cache = [None] * len(self.layers) cache = [None] * len(self.layers)
if mask is None: if mask is None:
j = self.args.sliding_window_pattern j = self.args.sliding_window_pattern
mask = create_attention_mask(h, cache[j - 1 : j]) mask = create_attention_mask(h, cache[j - 1 : j])