// CHECK:  %cst = arith.constant dense<2> : tensor<1xi32>
  // CHECK:  %4 = "tfl.reduce_max"(%arg0, %cst) <{keep_dims = false}> : (tensor<4x32x256xf32>, tensor<1xi32>) -> tensor<4x32xf32>
  // CHECK:  %5 = "tfl.arg_max"(%arg0, %cst) : (tensor<4x32x256xf32>, tensor<1xi32>) -> tensor<4x32xi32>
  // CHECK:  return %4, %5 : tensor<4x32xf32>, tensor<4x32xi32>

}

// CHECK-LABEL: identity_with_const
func.func @identity_with_const() -> tensor<*xf32> {
  %cst = "tf.Const"() {device = "", value = dense<[2.167590e-01, 2.89403105]> : tensor<2xf32>} : () -> tensor<2xf32>
  %cst_1 = "tf.Const"() {device = "", value = dense<1.000000e-03> : tensor<f32>} : () -> tensor<f32>
  %0 = "tf.Identity"(%cst) {device = ""} : (tensor<2xf32>) -> tensor<*xf32>

    %cst = "tf.Const"() <{value = dense<[0, 1]> : tensor<2xi32>}> {device = ""} : () -> tensor<2xi32>
    %cst_0 = "tf.Const"() <{value = dense<1.000000e+01> : tensor<f32>}> {device = ""} : () -> tensor<f32>
    %0 = "tf.Sum"(%arg0, %cst) <{keep_dims = false}> {device = ""} : (tensor<1x4xf32>, tensor<2xi32>) -> tensor<f32>

// CHECK-NEXT:    %cst = "tf.Const"() <{value = dense<1> : tensor<i32>}> : () -> tensor<i32>
// CHECK-NEXT:    %0 = "tf_mlrt.tf_await"(%arg1) : (!mlrt.future) -> tensor<i32>
// CHECK-NEXT:    %1 = "tf.AddV2"(%0, %cst) : (tensor<i32>, tensor<i32>) -> tensor<i32>
// CHECK-NEXT:    "tf_mlrt.tf_promise"(%arg2, %1) : (!mlrt.promise, tensor<i32>) -> ()

  func.return %1, %2 : tensor<2x1xf32>, tensor<2x1xf32>

// CHECK:  %cst = arith.constant
// CHECK:  %[[FQ:.*]] = "tf.FakeQuantWithMinMaxVars"(%arg0, %arg1, %arg2)
// CHECK:  %[[T1:.*]] = "tf.Transpose"(%[[FQ]], %cst)
// CHECK:  %[[T2:.*]] = "tf.Transpose"(%[[FQ]], %cst)
}

// CHECK-LABEL: fakeQuantFollowedByReshape

}

// -----

module {
  func.func @quantize_sharded_weights_with_xladot(%arg0: tensor<?x?x?x?xf32>) -> tensor<?x?x?x?xbf16> {
    %cst = "tf.Const"() {device = "", value = dense<1.000000e+01> : tensor<512x512xf32>} : () -> tensor<512x512xf32>

  func.return %1, %2 : tensor<2x1xf32>, tensor<2x1xf32>

// CHECK:  %cst = arith.constant
// CHECK:  %[[FQ:.*]] = "tf.FakeQuantWithMinMaxVars"(%arg0, %arg1, %arg2)
// CHECK:  %[[T1:.*]] = "tf.Transpose"(%[[FQ]], %cst)
// CHECK:  %[[T2:.*]] = "tf.Transpose"(%[[FQ]], %cst)
}

// CHECK-LABEL: fakeQuantFollowedByReshape

// CHECK-LABEL: HandleReturnedDequantizeWithAnotherUse
func.func @HandleReturnedDequantizeWithAnotherUse(%arg0: tensor<128x16xf32>) -> (tensor<128x16xf32>, tensor<128xi32>) {
// CHECK-NEXT:  %[[cst:.*]] = arith.constant dense<1> : tensor<i32>
  %cst = arith.constant dense<1> : tensor<i32>
// CHECK-NEXT:  %[[softmax:.*]] = "tfl.softmax"(%arg0) <{beta = 1.000000e+00 : f32}> : (tensor<128x16xf32>) -> tensor<128x16xf32>

// Variable is mutated, nothing should be removed.
module {
  func.func @f() {
    %cst = "tf.Const"() { value = dense<1.0> : tensor<0xf32> } : () -> tensor<0xf32>
    %handle = "tf.VarHandleOp"() {container="", shared_name="var1", device = "/job:worker/replica:0/task:1/device:CPU:0"} : () -> tensor<!tf_type.resource<tensor<0xf32>>>
    "tf.AssignVariableOp"(%handle, %cst) : (tensor<!tf_type.resource<tensor<0xf32>>>, tensor<0xf32>) -> ()

// RUN: tf-quant-opt %s -split-input-file -quant-insert-quantized-functions -quant-quantize-composite-functions | FileCheck %s

module {
  func.func @conv(%arg0: tensor<1x2x2x3xf32>) -> (tensor<*xf32>, tensor<*xf32>) {