{
          "shape": [1, 4],
          "name": "input_activation_state18",
          "is_variable": true,
          "quantization": {
            "min": [-0.8],
            "max": [1.6]
          }
        },
        {
          "shape": [1, 2],
          "name": "input_cell_state19",
          "is_variable": true,
          "quantization": {
            "min": [-2.0],
            "max": [4.0]

// CHECK-LABEL: main
func.func @main() -> tensor<3x2xi32> {
  // CHECK: "tfl.pseudo_const"() <{value = dense<0> : tensor<3x2xi32>}> {tfl.is_variable} : () -> tensor<3x2xi32>
  %0 = "tfl.pseudo_const"() {value = dense<0> : tensor<3x2xi32>, tfl.is_variable} : () -> tensor<3x2xi32> loc("variable")
  func.return %0 : tensor<3x2xi32>

// RUN: flatbuffer_translate -mlir-to-tflite-flatbuffer %s -o - | flatbuffer_to_string - | FileCheck %s

func.func @main() -> tensor<3x2xi32> {
  %0 = "tfl.pseudo_const" () {value = dense<0> : tensor<3x2xi32>, tfl.is_variable} : () -> tensor<3x2xi32> loc("variable")
  func.return %0 : tensor<3x2xi32>
}

// CHECK:      {
// CHECK-NEXT:     version: 3,
// CHECK-NEXT:     operator_codes: [ ],
// CHECK-NEXT:     subgraphs: [ {
// CHECK-NEXT:       tensors: [ {

// CHECK-DAG: %[[RES0:.*]] = "tfl.pseudo_const"() <{value = dense<{{.*}}> : tensor<1x4xf32>}> {tfl.is_variable} : () -> tensor<1x4xf32>
// CHECK-DAG: %[[RES1:.*]] = "tfl.pseudo_const"() <{value = dense<{{.*}}> : tensor<1x4xf32>}> {tfl.is_variable} : () -> tensor<1x4xf32>

// CHECK-EMPTY:
// CHECK-NEXT:       },
// CHECK-NEXT:       is_variable: true,
// CHECK-NEXT:       has_rank: true
// CHECK-NEXT:     }, {
// CHECK-NEXT:       shape: [ 1, 4 ],
// CHECK-NEXT:       name: "Const1",
// CHECK-NEXT:       quantization: {
// CHECK-EMPTY:
// CHECK-NEXT:       },
// CHECK-NEXT:       is_variable: true,
// CHECK-NEXT:       has_rank: true
// CHECK-NEXT:     }, {

// CHECK-EMPTY:
// CHECK-NEXT:       },
// CHECK-NEXT:       is_variable: true,
// CHECK-NEXT:       has_rank: true
// CHECK-NEXT:     }, {
// CHECK-NEXT:       shape: [ 1, 4 ],
// CHECK-NEXT:       name: "Const1",
// CHECK-NEXT:       quantization: {
// CHECK-EMPTY:
// CHECK-NEXT:       },
// CHECK-NEXT:       is_variable: true,
// CHECK-NEXT:       has_rank: true
// CHECK-NEXT:     }, {

// CHECK-EMPTY:
// CHECK-NEXT:       },
// CHECK-NEXT:       is_variable: true,
// CHECK-NEXT:       has_rank: true
// CHECK-NEXT:     }, {
// CHECK-NEXT:       shape: [ 4, 4 ],
// CHECK-NEXT:       name: "Const1",
// CHECK-NEXT:       quantization: {
// CHECK-EMPTY:
// CHECK-NEXT:       },
// CHECK-NEXT:       is_variable: true,
// CHECK-NEXT:       has_rank: true
// CHECK-NEXT:     }, {

            return LogicalResult::failure();
          }

          if (is_variable) {
            input.emitOpError() << "unexpected multiple TPUReplicatedInputOp "
                                << "for single argument";
            return LogicalResult::failure();
          }
          is_variable = true;
        } else {
          is_non_variable = true;
        }
      }

  // marked as a stateful. If so, set the tensor's is_variable as true
  // This is v1 ref variable semantics in the TFLite runtime.
  bool is_variable = false;
  for (auto& use : value.getUses()) {
    is_variable = IsStatefulOperand(use.getOwner(), use.getOperandNumber());
    if (is_variable) {
      break;
    }
  }

                                  const std::vector<uint8_t>& buffer,
                                  bool is_variable, OpBuilder builder,
                                  Location loc, bool use_stablehlo_constant) {
  if (tensor.sparsity != nullptr) {
    return BuildSparseConstOp(tensor, buffer, builder, loc);
  }

  if (is_variable) {
    return BuildVariableOp(tensor, builder, loc);
  }