fix a bug for return attn_weights

unicore/modules/multihead_attention.py

@@ -92,24 +92,30 @@ class SelfMultiheadAttention(nn.Module):
             )
             attn_weights = attn_weights.view(bsz * self.num_heads, tgt_len, src_len)
 
-        attn_probs = softmax_dropout(
-            attn_weights, self.dropout, self.training, bias=attn_bias
-        )
+        if not return_attn:
+            attn = softmax_dropout(
+                attn_weights, self.dropout, self.training, bias=attn_bias,
+            )
+        else:
+            attn_weights += attn_bias
+            attn = softmax_dropout(
+                attn_weights, self.dropout, self.training,
+            )
 
-        attn = torch.bmm(attn_probs, v)
-        assert list(attn.size()) == [bsz * self.num_heads, tgt_len, self.head_dim]
+        o = torch.bmm(attn, v)
+        assert list(o.size()) == [bsz * self.num_heads, tgt_len, self.head_dim]
 
-        attn = (
-            attn.view(bsz, self.num_heads, tgt_len, self.head_dim)
+        o = (
+            o.view(bsz, self.num_heads, tgt_len, self.head_dim)
             .transpose(1, 2)
             .contiguous()
             .view(bsz, tgt_len, embed_dim)
         )
-        attn = self.out_proj(attn)
+        o = self.out_proj(o)
         if not return_attn:
-            return attn
+            return o
         else:
-            return attn, attn_weights, attn_probs
+            return o, attn_weights, attn
 
 
 class CrossMultiheadAttention(nn.Module):
@@ -201,16 +207,16 @@ class CrossMultiheadAttention(nn.Module):
             )
             attn_weights = attn_weights.view(bsz * self.num_heads, tgt_len, src_len)
 
-        attn_probs = softmax_dropout(attn_weights, self.dropout, self.training, bias=attn_bias)
+        attn = softmax_dropout(attn_weights, self.dropout, self.training, bias=attn_bias)
 
-        attn = torch.bmm(attn_probs, v)
-        assert list(attn.size()) == [bsz * self.num_heads, tgt_len, self.head_dim]
+        o = torch.bmm(attn, v)
+        assert list(o.size()) == [bsz * self.num_heads, tgt_len, self.head_dim]
 
-        attn = (
-            attn.view(bsz, self.num_heads, tgt_len, self.head_dim)
+        o = (
+            o.view(bsz, self.num_heads, tgt_len, self.head_dim)
             .transpose(1, 2)
             .contiguous()
             .view(bsz, tgt_len, embed_dim)
         )
-        attn = self.out_proj(attn)
-        return attn
+        o = self.out_proj(o)
+        return o