#include "tensorflow/core/framework/kernel_def.pb.h"
#include "tensorflow/core/framework/log_memory.h"
#include "tensorflow/core/framework/node_def_util.h"
#include "tensorflow/core/framework/op_kernel.h"
#include "tensorflow/core/framework/partial_tensor_shape.h"
#include "tensorflow/core/framework/shape_inference.h"
#include "tensorflow/core/framework/tensor.h"
#include "tensorflow/core/framework/tensor.pb.h"  // NOLINT

          "multiplier");
    }

    // Slicing with dynamic offsets (helper method advised)
    auto create_slice = [&](mlir::Value tensor, int depth_idx, int channel_idx,
                            bool is_kernel = false) -> mlir::Value {
      std::vector<int64_t> tensor_shape =
          mlir::cast<ShapedType>(tensor.getType()).getShape().vec();

      // Calculate offsets based on depth_idx, channel_idx and tensor_shape

    y_shape = [v if v is not None else n for v in shapes[1]]

    class MatmulModel(module.Module):

      def __init__(self, bias: Optional[core.Tensor]):
        self._bias = bias
        self._kernel = np.random.uniform(size=y_shape).astype('f4')
        self._min = (-0.8, -0.8, -0.9)
        self._max = (0.9, 0.9, 1.0)

      @def_function.function(
          input_signature=[