func.func @random_uniform_int(%arg0: tensor<i32>, %arg1: tensor<i32>) -> tensor<1000xi32> {
    %0 = "tf.Const"() {value = dense<1000> : tensor<1xi32>} : () -> tensor<1xi32>
    // CHECK-NOT: tf.RandomUniformInt
    %1 = "tf.RandomUniformInt"(%0, %arg0, %arg1) {seed = 0 : i64, seed2 = 0 : i64} : (tensor<1xi32>, tensor<i32>, tensor<i32>) -> tensor<1000xi32>
    func.return %1 : tensor<1000xi32>
  }

  // CHECK-LABEL: multinomial

      /*error_tolerance=*/1.0);
}

TEST_F(ConvertTfQuantToMhloIntTest, UniformQuantizeAddToValidGraph) {
  constexpr absl::string_view kProgram = R"mlir(
func.func @main(%lhs: tensor<10x10xi32>, %rhs: tensor<10x10xi32>) -> tensor<10x10xi32> {
  %lhs_scale = "tf.Const"() { value = dense<0.518> : tensor<f32> } : () -> tensor<f32>
  %lhs_zp = "tf.Const"() { value = dense<42> : tensor<i32> } : () -> tensor<i32>

// CHECK:           [[VAL_5:%.*]] = call @func_1_GPU_FLOAT([[VAL_0]], [[VAL_3]]) {tac.device = "GPU", tac.inference_type = "FLOAT", tac.interface_name = "func_1"} : (tensor<100xf32>, tensor<100xf32>) -> tensor<100xf32>

}

// -----

func.func @pack_CPU(%arg0: tensor<100xf32>, %arg1: tensor<100xf32>) -> tensor<2x100xf32> attributes {tac.device = "CPU", tac.interface_name = "func_2"} {
  // CHECK: tac.cost = 1.000000e+02
  %0 = "tfl.pack"(%arg0, %arg1) {axis = 0 : i32, tac.device = "CPU", values_count = 2 : i32} : (tensor<100xf32>, tensor<100xf32>) -> tensor<2x100xf32>
  func.return %0 : tensor<2x100xf32>
}

  %3 = "tf.PreventGradient"(%1) : (tensor<10x10xi32>) -> tensor<10x10xi32>
  %4 = "tf.AddV2"(%2, %3) {device = ""} : (tensor<10x10xi32>, tensor<10x10xi32>) -> tensor<10x10xi32>
  func.return %4 : tensor<10x10xi32>

// CHECK-LABEL: preventGradient
// CHECK-NOT: %2 = "tf.PreventGradient"(%0) : (tensor<10x10xi32>) -> tensor<10x10xi32>

func.func @no_pushdown_multiple_outputs(%arg0: tensor<1000x2x!quant.uniform<i8:f32, 7.812500e-03>>) -> tensor<1000xf32> {
  %0 = "tfl.dequantize"(%arg0) : (tensor<1000x2x!quant.uniform<i8:f32, 7.812500e-03>>) -> tensor<1000x2xf32>
  %1:2 = "tfl.unpack"(%0) {axis = 1 : i32, num = 2 : i32} : (tensor<1000x2xf32>) -> (tensor<1000xf32>, tensor<1000xf32>)
  func.return %1#0 : tensor<1000xf32>

// CHECK-NEXT: tfl.dequantize
// CHECK-NEXT: tfl.unpack
}

func.func @main(%arg0: tensor<10xi32>, %arg1: tensor<10xi32>) -> tensor<10xi32>
attributes {tf.entry_function = {inputs = "input0,input1", outputs = "Add"}} {
  %graph = tf_executor.graph {
    %2:2 = tf_executor.island wraps "tf.Add"(%arg0, %arg1) {T = "tfdtype$DT_INT32", device = ""} : (tensor<10xi32>, tensor<10xi32>) -> tensor<10xi32> loc("Add")
    tf_executor.fetch %2 : tensor<10xi32>
  }
  func.return %graph : tensor<10xi32>
}

>, tensor<i8>, tensor<f64>, tensor<!tf_type.resource>)

# CHECK-SUBTYPE: func @main(%arg0: tensor<10xi32>, %arg1: tensor<!tf_type.variant<tensor<10xf32>>>, %arg2: tensor<10xi8>, %arg3: tensor<10xf64>, %arg4: tensor<!tf_type.resource<tensor<10xi32>>>) -> (tensor<10xi32>, tensor<!tf_type.variant<tensor<10xf32>>>, tensor<10xi8>, tensor<10xf64>, tensor<!tf_type.resource<tensor<10xi32>>>)...

      %2 = "tf.ReadVariableOp"(%arg15) : (tensor<*x!tf_type.resource<tensor<7x7x3x64xf32>>>) -> tensor<7x7x3x64xf32>
      %3 = "tf.ReadVariableOp"(%arg16) : (tensor<*x!tf_type.resource<tensor<1001xf32>>>) -> tensor<1001xf32>
      %4 = "tf.ReadVariableOp"(%arg17) : (tensor<*x!tf_type.resource<tensor<64x1001xf32>>>) -> tensor<64x1001xf32>
      %5 = "tf.ReadVariableOp"(%arg18) : (tensor<*x!tf_type.resource<tensor<f32>>>) -> tensor<f32>

# CHECK-LABEL: func @main
# CHECK-SAME:  ([[VAL_0:%.*]]: tensor<1x8x8x2xi32>) -> (tensor<1x8x8x2xi32>, tensor<i32>, tensor<10xi32>)
# CHECK-SAME:  control_outputs = ""
# CHECK-SAME:  inputs = "unranked"
# CHECK-SAME:  outputs = "unranked,static,static_10"
# CHECK-DAG:         [[VAL_2:%.*]] = arith.constant dense<0> : tensor<10xi32>
# CHECK-DAG:         [[VAL_1:%.*]] = arith.constant dense<0> : tensor<i32>