)
        if bias_fn is not None:
          out = nn_ops.bias_add(out, self.bias)
        if has_batch_norm:
          # Fusing is supported for non-training case.
          out, _, _, _, _, _ = nn_ops.fused_batch_norm_v3(
              out, scale, offset, mean, variance, is_training=False
          )
        if activation_fn is not None:
          out = activation_fn(out)
        return {'output': out}

        )
        if has_bias:
          out = nn_ops.bias_add(out, self.bias)
        if has_batch_norm:
          # Fusing is supported for non-training case.
          out, _, _, _, _, _ = nn_ops.fused_batch_norm_v3(
              out, scale, offset, mean, variance, is_training=False
          )
        if activation_fn is not None:
          out = activation_fn(out)
        return {'output': out}

    // Match
    auto fused_batch_norm_op =
        dyn_cast_or_null<::mlir::TF::FusedBatchNormV3Op>(fused_batch_norm);
    root = fused_batch_norm_op;
    x = fused_batch_norm_op.getODSOperands(0);
    scale = fused_batch_norm_op.getODSOperands(1);
    offset = fused_batch_norm_op.getODSOperands(2);
    mean = fused_batch_norm_op.getODSOperands(3);
    variance = fused_batch_norm_op.getODSOperands(4);

          # activation in this test.
          if has_batch_norm:
            # Fusing is supported for non-training case.
            out, _, _, _, _, _ = nn_ops.fused_batch_norm_v3(
                out, scale, offset, mean, variance, is_training=False
            )
          out = activation_fn(out)
        out_min = array_ops.constant([-0.18, -0.32], dtype=dtypes.float32)