Update modeling_gptbert.py

modeling_gptbert.py

@@ -394,7 +394,7 @@ class SelfAttention(nn.Module):
             query=query,
             key=key,
             value=value,
-            attn_mask=attention_mask,
+            attn_mask=attention_mask if not self.is_causal else None,
             dropout_p=self.config.attention_dropout if self.training else 0.0,
             is_causal=self.is_causal
         )