// 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>
  %0 = "tfl.softmax"(%arg0) {beta = 1.000000e+00 : f32} : (tensor<128x16xf32>) -> tensor<128x16xf32>

  %group_assignment = "tf.Const"() { value = dense<[[0],[1]]> : tensor<2x1xi32> } : () -> tensor<2x1xi32>
  // CHECK: "mhlo.all_reduce"
  // CHECK{LITERAL}: replica_groups = dense<[[0], [1]]> : tensor<2x1xi64>
  // CHECK-NOT: channel_handle
  // CHECK: mhlo.add
  %0 = "tf.XlaAllReduce"(%input, %group_assignment) {reduce_op = "Add", mode = "CrossReplica"} : (tensor<f32>, tensor<2x1xi32>) -> tensor<f32>
  func.return %0 : tensor<f32>
}

// -----

    %11 = stablehlo.convert %2 : (tensor<1x1x1xi8>) -> tensor<1x1x1xi32>
    %12 = stablehlo.broadcast_in_dim %11, dims = [0, 1, 2] : (tensor<1x1x1xi32>) -> tensor<8x16x16xi32>
    %13 = stablehlo.subtract %10, %12 : tensor<8x16x16xi32>  // q1 - z1
    %14 = stablehlo.convert %9 : (tensor<8x16x4xi8>) -> tensor<8x16x4xi32>
    %15 = stablehlo.convert %4 : (tensor<1x1x1xi8>) -> tensor<1x1x1xi32>

// -----

func.func @constant_tensor_invalid_3(%arg0: vector<1xi32>) -> tensor<1x1xi32> {
    // expected-error@+1 {{input and output should have same shape and element type}}
  %0 = "tfr.constant_tensor"(%arg0) : (vector<1xi32>) -> tensor<1x1xi32>
  func.return %0 : tensor<1x1xi32>
}

// -----

func.func @constant_tensor_invalid_4(%arg0: i32) -> tensor<1x1xi32> {
    // expected-error@+1 {{input can not be converted to an output tensor}}

  %cst = "tf.Const"() {device = "", value = dense<2> : tensor<i32>} : () -> tensor<i32>
  // idx 1 should be a compile time constant
  %0 = "tf.ExpandDims"(%arg0, %cst) {device = ""} : (tensor<?x2xi32>, tensor<i32>) -> tensor<?x2x1xi32>
  %1 = "tf.AddV2"(%cst, %cst) {device = ""} : (tensor<i32>, tensor<i32>) -> tensor<i32>

  return %0 : tensor<?x2x1xi32>
}

  %1 = "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>>

  %cst = arith.constant dense<[[1], [2]]> : tensor<2x1xi32>

  // CHECK: %[[CST:.*]] = arith.constant dense<2> : tensor<1xi32>
  // CHECK: return %[[CST]]
  %0 = "tfl.rank"(%cst) : (tensor<2x1xi32>) -> tensor<1xi32>
  func.return %0 : tensor<1xi32>
}

// CHECK-LABEL: @rank_input_known_rank
func.func @rank_input_known_rank(%arg0 : tensor<2x1xi32>) -> tensor<1xi32> {

  // 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>>

    %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>