return axis < 0 ? axis + rank : axis;
}

/// Returns a `ConvertOp` that casts the elements to a i64 type while retaining
/// the shape of the input value.
static ConvertOp CastValueToI64(Location loc, Value value,
                                PatternRewriter *rewriter) {
  return rewriter->create<ConvertOp>(loc, value, rewriter->getIntegerType(64));
}

        &result_types);

    // Create a new if op with new operands and updated result types.
    auto converted = rewriter.create<TF::IfOp>(
        op.getLoc(), result_types, adaptor.getOperands(), op->getAttrs());
    converted->removeAttr("T");
    (void)UpdateFunctionTypesForIfOp(rewriter, converted, adaptor.getOperands(),
                                     tensor_list_args, resized_tensor_lists,

            for (expr in this@convertVarargsExpression) {
                val converted = expr.convertConstantExpression(builder) ?: continue

                if ((expr is FirSpreadArgumentExpression || expr is FirNamedArgumentExpression) && converted is KaArrayAnnotationValue) {
                    addAll(converted.values)
                } else {
                    add(converted)
                }

      T defaultType = (T) getFirstTypeParameter(type.getType());
      return defaultType;
    }
    if (type.getRawType() == Converter.class) {
      TypeToken<?> convertFromType = type.resolveType(Converter.class.getTypeParameters()[0]);
      TypeToken<?> convertToType = type.resolveType(Converter.class.getTypeParameters()[1]);
      @SuppressWarnings("unchecked") // returns default for both F and T

      T defaultType = (T) getFirstTypeParameter(type.getType());
      return defaultType;
    }
    if (type.getRawType() == Converter.class) {
      TypeToken<?> convertFromType = type.resolveType(Converter.class.getTypeParameters()[0]);
      TypeToken<?> convertToType = type.resolveType(Converter.class.getTypeParameters()[1]);
      @SuppressWarnings("unchecked") // returns default for both F and T

using CustomOpMap = std::unordered_map<std::string, CustomOpInfo>;

// Applies dynamic range quantization for the given model wehre the input_model
// type is flatbuffer but is converted to MLIR during quantization process and
// then converted back to flatbuffer for return. Note that this is part of
// reaching feature parity with the old quantizer for dynamic range
// quantization, specifically for

    }

    /**
     * Converts the given object into a {@code JavaVersion}.
     *
     * @param value An object whose toString() value is to be converted. May be null.
     * @return The version, or null if the provided value is null.
     * @throws IllegalArgumentException when the provided value cannot be converted.
     */

        val children = projects.groupBy { it.path.parent }
        val converted = mutableMapOf<Path, BuildStructureOperationProject>()
        for (project in projects) {
            convertProject(converted, project, rootProjectName, children)
        }
        return converted.getValue(Path.ROOT)
    }

    private
    fun convertProject(
        converted: MutableMap<Path, BuildStructureOperationProject>,

    const absl::flat_hash_set<std::string> &denylisted_nodes = {},
    bool enable_variable_quantization = false,
    bool disable_per_channel_for_dense_layers = false,
    const std::optional<const tensorflow::converter::DebugOptions>
        &debug_options = std::nullopt);

}  // namespace lite
}  // namespace mlir

            fqid == 'validation:property-validation:unsupported-notation'
            contextualLabel == 'Property \'input\' has unsupported value \'task \':dependencyTask\'\''
            details == "Type 'DefaultTask' cannot be converted to a $targetType"
            solutions == [
                'Use a String or CharSequence path, for example \'src/main/java\' or \'/usr/include\'',