fix function not having return value in dist case (#76186)

liufengwei0103 · web-flow · commit a8a926991055 · 2025-11-04T14:19:42.000+08:00
diff --git a/python/paddle/nn/functional/flash_attention.py b/python/paddle/nn/functional/flash_attention.py
@@ -1453,6 +1453,7 @@ def scaled_dot_product_attention(
             dropout_p,
             is_causal,
         )
+        return out
 
     if attn_mask is None:
         # downgraded to ordinary flash attention implementation

Original file line number	Diff line number	Diff line change
`@@ -1453,6 +1453,7 @@ def scaled_dot_product_attention(`
`1453`	`1453`	`dropout_p,`
`1454`	`1454`	`is_causal,`
`1455`	`1455`	`)`
	`1456`	`+ return out`
`1456`	`1457`
`1457`	`1458`	`if attn_mask is None:`
`1458`	`1459`	`# downgraded to ordinary flash attention implementation`