%cst = "tfl.pseudo_const"() {value = dense<[[0], [0], [1], [1], [2], [2], [3], [3], [4], [4], [5], [5], [6], [6], [7], [7]]> : tensor<16x1xi32>} : () -> tensor<16x1xi32>
  %gather_nd_first = "tfl.gather_nd"(%arg0, %cst) : (tensor<8x1x8x1280xf32>, tensor<16x1xi32>) -> tensor<16x1x8x1280xf32>
  %transpose_first_perm = "tfl.pseudo_const"() {value = dense<[2, 1, 0, 3]> : tensor<4xi32>} : () -> tensor<4xi32>

  // CHECK:  %5 = "tfl.arg_max"(%arg0, %cst) : (tensor<1x32x1xf32>, tensor<1xi32>) -> tensor<1x1xi32>
  // CHECK:  return %4, %5 : tensor<1x1xf32>, tensor<1x1xi32>
}

// CHECK-LABEL:   func @convert_pytorch_argmax
func.func @convert_pytorch_argmax(%arg0: tensor<1x9xi32>) -> tensor<1xi32> {
  %0 = mhlo.constant dense<0> : tensor<i32>

  %0 = "tfl.pseudo_qconst"() {qtype = tensor<128x!quant.uniform<i32:f32, 0.7>>, value = dense<0> : tensor<128xi32>} : () -> tensor<128x!quant.uniform<i32:f32, 0.7>>

  // CHECK:      %[[SLICE1:.*]] = "tf.Slice"(%arg0, %[[CST5]], %[[CST1]]) : (tensor<3x8x4xi32>, tensor<3xi64>, tensor<3xi64>) -> tensor<1x8x4xi32>
  // CHECK:      %[[CONCAT:.*]] = "tf.ConcatV2"(%[[SLICE]], %[[SLICE1]], %[[CST2]]) : (tensor<2x8x4xi32>, tensor<1x8x4xi32>, tensor<i32>) -> tensor<3x8x4xi32>

    %3:2 = "tf.Split"(%0, %1) {device = ""} : (tensor<i32>, tensor<*xf32>) -> (tensor<*xf32>, tensor<*xf32>)
    %4 = tensor.cast %1 : tensor<*xf32> to tensor<?x?x?xf32>
    func.return %3#0, %3#1 : tensor<*xf32>, tensor<*xf32>
  }

  // CHECK-LABEL: infer_device_cluster
  func.func @infer_device_cluster(%arg0: tensor<1x8x2xi32>) -> (tensor<*xf32>, tensor<*xf32>) {

    %0 = "tf.Const"() {value = dense<1> : tensor<i64>} : () -> tensor<i64>
    %1 = "tf.Const"() {value = dense<0> : tensor<1x1xi32>} : () -> tensor<1x1xi32>
    %2 = "tf.Const"() {value = dense<[7, 7, 3, 64]> : tensor<4xi32>} : () -> tensor<4xi32>
    %3 = "tf.Const"() {value = dense<[[0, 0], [3, 3], [3, 3], [0, 0]]> : tensor<4x2xi32>} : () -> tensor<4x2xi32>

func.func @scatter_nd_add(%arg0: tensor<7xi64>, %arg1: tensor<1x1xi32>, %arg2: tensor<1xi64>) -> tensor<7xi64> {
  %0 = "tf.TensorScatterAdd"(%arg0, %arg1, %arg2) : (tensor<7xi64>, tensor<1x1xi32>, tensor<1xi64>) -> tensor<7xi64>
  func.return %0 : tensor<7xi64>

  // CHECK-LABEL: scatter_nd_add
  // CHECK:  %[[GATHER:.*]] = "tf.GatherNd"(%arg0, %arg1) <{bad_indices_policy = ""}> : (tensor<7xi64>, tensor<1x1xi32>) -> tensor<1xi64>

  // CHECK-DAG: %[[IOTA0:.*]] = "mhlo.iota"() <{iota_dimension = 0 : i64}> : () -> tensor<12x12xi32>
  // CHECK-DAG: %[[IOTA1:.*]] = "mhlo.iota"() <{iota_dimension = 1 : i64}> : () -> tensor<12x12xi32>
  // CHECK-DAG: %[[COMP:.*]] = mhlo.compare EQ, %[[IOTA0]], %[[IOTA1]], NOTYPE : (tensor<12x12xi32>, tensor<12x12xi32>) -> tensor<12x12xi1>
  // CHECK-DAG: %[[ZERO:.*]] = mhlo.constant dense<0.000000e+00> : tensor<f32>

  %4 = "tf.Const"() {value = dense<2.000000e+00> : tensor<f32>} : () -> tensor<f32>
  %5 = "tf.Const"() {value = dense<0> : tensor<i32>} : () -> tensor<i32>
  %6 = "tf.Const"() {value = dense<[[0], [16]]> : tensor<2x1xi32>} : () -> tensor<2x1xi32>
  %7 = "tf.Const"() {value = dense<[32, 1]> : tensor<2xi32>} : () -> tensor<2xi32>
  %8 = "tf.Const"() {value = dense<4> : tensor<i32>} : () -> tensor<i32>

  cf.br ^exit
^exit:
  func.return %1: tensor<8x16x32x64xf32>

  // CHECK: return %arg0
}

// CHECK-LABEL: testConcatCanonicalization
func.func @testConcatCanonicalization(%arg0: tensor<2x1xi32>, %arg1: tensor<2x1xi32>) -> tensor<2x2xi32> {
  // CHECK: %[[AXIS:.*]] = "tf.Const"
  %0 = "tf.Const"() { value = dense<1> : tensor<i32> } : () -> tensor<i32>