// Returns a new buffer that is packed densely with 2 4-bit values in a byte.
// The packing format is low-bits-first, i.e. the lower nibble of a byte is
// filled first, followed by the upper nibble.
std::vector<uint8_t> PackInt4ValuesDensely(std::vector<uint8_t> src_buffer);

// Assumes `src_buffer` contains 2 4-bit elements packed in 8-bit.
// Returns a vector where each int8 element contains a int4 sign-extended value.

#include "tensorflow/compiler/mlir/lite/utils/low_bit_utils.h"

#include <cassert>
#include <cstdint>
#include <vector>

namespace tflite {

std::vector<uint8_t> PackInt4ValuesDensely(std::vector<uint8_t> src_buffer) {
  auto num_elements = src_buffer.size();
  auto packed_size = (num_elements + 1) / 2;
  std::vector<uint8_t> packed_buffer((num_elements + 1) / 2);

    std::vector<uint8_t> data;
    for (mlir::APInt v : attr.getValues<mlir::APInt>()) {
      data.emplace_back(static_cast<uint8_t>(*(v.getRawData())));
    }
    auto packed_buffer = tflite::PackInt4ValuesDensely(data);
    if (use_buffer_offset_) {
      buffer_data_map_[index] =
          std::string(packed_buffer.begin(), packed_buffer.end());
      return tflite::CreateBuffer(builder_, 0, 1, 1);
    } else {