BugFix: qwen pp cannot send bool when using fuse loss

Author: Jason233333

paddlenlp/transformers/qwen2/modeling.py

@@ -1486,7 +1486,8 @@ def __init__(self, config: Qwen2Config, embedding_weights=None, transpose_y=Fals
     def forward(self, hidden_states, tensor_parallel_output=None, batch_size=None):
         # add this for fused_head_and_loss_fn
         if self.config.use_fused_head_and_loss_fn:
-            return hidden_states, self.weight, None, self.transpose_y
+            # return hidden_states, self.weight, None, self.transpose_y
+            return hidden_states, self.weight, None, None
 
         if self.config.sequence_parallel:
             hidden_states = GatherOp.apply(hidden_states)