drop me: adjust comment

Author: NeoLegends

i6_models/parts/conformer/mhsa_rel_pos.py

@@ -219,7 +219,7 @@ def forward(self, input_tensor: torch.Tensor, sequence_mask: torch.Tensor) -> to
         print("q_with_bias_u", q_with_bias_u.transpose(-3, -2).shape)
         print("k", k.transpose(-3, -2).shape)
         print("v", v.transpose(-3, -2).shape)
-        print("attn_bd_mask_scaled", attn_bd_mask_scaled.transpose(-3, -2).shape)
+        print("attn_bd_mask_scaled", attn_bd_mask_scaled.shape)
         attn_output = F.scaled_dot_product_attention(
             q_with_bias_u.transpose(-3, -2),  # [B, #heads, T, F']
             k.transpose(-3, -2),  # [B, #heads, T', F']