namespace quant {
namespace {
bool GetBooleanSpecs(const std::string& bool_val) {
  bool result;
  std::stringstream iss(bool_val);
  iss >> std::boolalpha >> result;
  return result;
}
}  // namespace

void ParseCustomOpSpecs(const absl::string_view node_names,
                        const CustomOpUpdateOptions& update_option,
                        CustomOpMap& custom_op_map) {
  if (node_names.empty()) return;

    quant_specs.nodes_blocklist = absl::flat_hash_set<std::string>(
        nodes_blocklist_flag_.begin(), nodes_blocklist_flag_.end());
  }

  if (!enable_custom_op_weight_only_.empty()) {
    ParseCustomOpSpecs(enable_custom_op_weight_only_,
                       quant::CustomOpUpdateOptions::kWeightOnly,
                       quant_specs.custom_map);
  }
  if (enable_float16_quantization_) {

  // to be eliminated and fused into quantized kernels.
  QDQConversionMode qdq_conversion_mode = QDQConversionMode::kQDQNone;
};

// Parses the command line flag strings to the CustomOpMap specification.
void ParseCustomOpSpecs(absl::string_view node_names,
                        const CustomOpUpdateOptions& update_option,
                        CustomOpMap& custom_op_map);

};

#include "tensorflow/compiler/mlir/lite/transforms/generated_post_quantize.inc"

void PostQuantizePass::runOnOperation() {
  if (!enable_custom_op_no_side_effect_.empty()) {
    ParseCustomOpSpecs(enable_custom_op_no_side_effect_,
                       quant::CustomOpUpdateOptions::kNoSideEffect,
                       custom_op_map_);
  }

  RewritePatternSet patterns(&getContext());

      !enable_dynamic_range_per_channel_quantization_for_dense_layers_;
  quant_specs_.minimum_elements_for_weights = min_elements_for_weights_;

  if (!enable_custom_op_quantization_.empty()) {
    ParseCustomOpSpecs(enable_custom_op_quantization_,
                       quant::CustomOpUpdateOptions::kInputIndices,
                       quant_specs_.custom_map);
  }

  ConvertTFLQuantOpsToMlirQuantOps(func);