adapt npu FA. (#8171)

PaddlePaddle · Mar 25, 2024 · a0457d1 · a0457d1
1 parent a5d87f5
commit a0457d1
Showing 1 changed file with 1 addition and 0 deletions.
diff --git a/paddlenlp/transformers/llama/modeling.py b/paddlenlp/transformers/llama/modeling.py
@@ -239,6 +239,7 @@ def scaled_dot_product_attention(
                     attention_mask is None,
                     True,
                     False,
+                    False,
                 )[0]
             else:
                 attn_output = F.scaled_dot_product_attention(