auto dtype = input->DataType();
  if (DataTypeIsFloating(BaseType(dtype)) ||
      DataTypeIsInteger(BaseType(dtype))) {
    return tensorflow::ops::Identity(ctx, input, output, name);
  } else if (!DataTypeIsComplex(BaseType(dtype)) &&
             BaseType(dtype) != DT_VARIANT) {
    return errors::InvalidArgument(
        "Expected numeric or variant tensor, got dtype ", dtype);
  }

// 1. does not cause an error
// 2. preserves dtype and shape.
TEST_P(SavedVariableLoadingTest, LoadSavedVariableSuccessful) {
  auto& test_params = GetParam();
  DataType dtype = std::get<0>(test_params);
  TensorShape shape(std::get<1>(test_params));

  SavedVariable saved_variable;
  saved_variable.set_dtype(dtype);
  shape.AsProto(saved_variable.mutable_shape());

  std::unique_ptr<Variable> var;

  if (tensor.dtype() != expected_tensor.dtype()) {
    if (listener->IsInterested()) {
      *listener << "\nexpected tensor of type "
                << DataType_Name(expected_tensor.dtype())
                << " but found one of type " << DataType_Name(tensor.dtype());
      return false;
    }
  }

  switch (tensor.dtype()) {
    case DT_HALF:

        auto operand_or_result = value_shaped_type.getElementType();
        std::string dtype;

        TypeSwitch<Type>(operand_or_result)
            .Case<IntegerType>([&](Type) {
              dtype =
                  absl::StrCat("i", operand_or_result.getIntOrFloatBitWidth());
            })
            .Case<FloatType>([&](Type) {
              dtype =
                  absl::StrCat("f", operand_or_result.getIntOrFloatBitWidth());

                                           DataType dtype, TensorShape shape,
                                           const char* raw_device_name,
                                           ImmediateTensorHandlePtr* handle) {
  ImmediateOpPtr varhandle_op(ctx->CreateOperation());

  TF_RETURN_IF_ERROR(varhandle_op->Reset("VarHandleOp", raw_device_name));
  TF_RETURN_IF_ERROR(varhandle_op->SetAttrType("dtype", dtype));

}

TEST(OStreamTest, TestTensorSpec) {
  std::stringstream stream;
  TensorSpec tensor_spec;
  tensor_spec.shape = tensorflow::PartialTensorShape({2});
  tensor_spec.dtype = tensorflow::DT_FLOAT;
  stream << tensor_spec;
  ASSERT_EQ(stream.str(), "TensorSpec(shape = [2], dtype = 1)");
}

}  // namespace impl
}  // namespace libtf

  mutex_lock ml(*variable->mu());
  OP_REQUIRES(
      context,
      !variable->is_initialized || variable->tensor()->dtype() == dtype_,
      errors::InvalidArgument(
          "Trying to assign variable with wrong dtype. Expected ",
          DataTypeString(variable->tensor()->dtype()), " got ",
          DataTypeString(dtype_)));
  variable->is_initialized = true;
  *variable->tensor() = value;
}

          DataType dtype = n->output_type(0);
          AddNodeAttr("dtype", dtype, &const_def);
          TensorProto value;
          value.set_dtype(dtype);
          value.mutable_tensor_shape()->add_dim()->set_size(
              shape_proto.dim_size());
          for (const auto& dim : shape_proto.dim()) {
            if (dtype == DT_INT32) {
              value.add_int_val(dim.size());

// Functor that compares the arg values of two different signatures. Returns
// true when the args are not equal.
struct SignatureNotEqual {
  bool operator()(const Tensor& arg, const Tensor& other) {
    return arg.dtype() != other.dtype() || arg.shape() != other.shape() ||
           arg.tensor_data() != other.tensor_data();
  }
  bool operator()(const TensorTypeAndShape& arg,
                  const TensorTypeAndShape& other) {

TensorSpec::TensorSpec(PartialTensorShape shape, DataType dtype)
    : shape_(std::move(shape)), dtype_(dtype) {}

TensorSpec::TensorSpec(const TensorSpecProto& proto)
    : shape_(proto.shape()), dtype_(proto.dtype()) {}

const PartialTensorShape& TensorSpec::shape() const { return shape_; }

DataType TensorSpec::dtype() const { return dtype_; }