for ms-swift

modeling_hyper_qwen2.py

@@ -464,8 +464,7 @@ class HyperQwen2SdpaAttention(HyperQwen2Attention):
                 curr_query_layer = query_states[bi:bi+1]
                 curr_key_layer = key_states[bi:bi+1]
                 curr_value_layer = value_states[bi:bi+1]
-                full_mask = causal_mask
-                is_causal = True if causal_mask is None and q_len > 1 else False
+                is_causal = True if q_len > 1 else False
                 if is_causal:
                     full_mask = None
                 else: