const int64_t dim = GetTensorSpatialDimIndex(num_dims, data_format, i);
      spatial_dim_indices.push_back(dim);

      const int64_t dilation = get_int(dilations[dim]);
      rhs_dilations.push_back(dilation);
      const int64_t stride = get_int(strides[dim]);
      window_strides.push_back(stride);

      Value pad_low, pad_high;

        verify_size(window_strides.size(), "window strides") &&
        verify_size(padding.size(), "padding entries") &&
        verify_size(lhs_dilation.size(), "lhs dilation factors") &&
        verify_size(rhs_dilation.size(), "rhs dilation factors")))
    return std::nullopt;

  xla::Window window;
  for (size_t i = 0; i < window_dimensions.size(); i++) {
    auto dim = window.add_dimensions();

    Arg<TF_I32OrI64Tensor, [{the padding to apply at the start and end of each input dimensions}]>:$padding,
    Arg<TF_I32OrI64Tensor, [{dilation to apply between input elements}]>:$lhs_dilation,
    Arg<TF_I32OrI64Tensor, [{dilation to apply between kernel elements}]>:$rhs_dilation,
    Arg<TF_I32OrI64Tensor, [{number of feature groups for grouped convolution.}]>:$feature_group_count,

        *   (since v2.7) `tf.math.bincount`
        *   (since v2.7) `tf.nn.depthwise_conv2d` backprop to `filter` when not
            using cuDNN convolution
        *   (since v2.7) `tf.nn.dilation2d` gradient
        *   (since v2.7) `tf.nn.max_pool_with_argmax` gradient
        *   (since v2.7) `tf.raw_ops.DebugNumericSummary` and
            `tf.raw_ops.DebugNumericSummaryV2`