Fix LXMERT with DataParallel (#7471)

src/transformers/modeling_lxmert.py

@@ -958,7 +958,7 @@ class LxmertModel(LxmertPreTrainedModel):
         # positions we want to attend and -10000.0 for masked positions.
         # Since we are adding it to the raw scores before the softmax, this is
         # effectively the same as removing these entirely.
-        extended_attention_mask = extended_attention_mask.to(dtype=next(self.parameters()).dtype)
+        extended_attention_mask = extended_attention_mask.to(dtype=self.dtype)
         extended_attention_mask = (1.0 - extended_attention_mask) * -10000.0
 
         # Process the visual attention mask