// MIN-MAX-CHECK-NEXT:  [[lhs:%.*]], {{.*}}, {{.*}}, {{.*}} = "tf.CustomAggregator"(%arg0) <{calibration_method = 1 : i32, id = "composite_conv2d_with_relu6_fn_arg_0_calibration_method_1",...

func.func @pseudo_const() -> tensor<i32> {
  // CHECK: %[[CST:.*]] = arith.constant dense<1> : tensor<i32>
  // CHECK: return %[[CST]]
  %0 = "tfl.pseudo_const"() {value = dense<1> : tensor<i32>} : () -> tensor<i32>
  func.return %0 : tensor<i32>
}


// CHECK-LABEL: @range_int
func.func @range_int() -> tensor<?xi32> {
  %cst = arith.constant dense<0> : tensor<i32>
  %cst_1 = arith.constant dense<4> : tensor<i32>

// CHECK:           %[[VAL_6:.*]] = "tf.Minimum"(%[[VAL_4]], %[[VAL_5]]) : (tensor<i32>, tensor<i32>) -> tensor<i32>
// CHECK:           %[[VAL_7:.*]] = "tf.Maximum"(%[[VAL_6]], %[[VAL_3]]) : (tensor<i32>, tensor<i32>) -> tensor<i32>
// CHECK:           %[[VAL_8:.*]] = "tf.Cast"(%[[VAL_2]]) <{Truncate = false}> : (tensor<i32>) -> tensor<i32>
// CHECK:           %[[VAL_9:.*]] = "tf.Const"() <{value = dense<1> : tensor<i32>}> : () -> tensor<i32>

  // CHECK-DAG: %0 = mhlo.constant dense<1> : tensor<i32>
  // CHECK: %1 = "mhlo.broadcast_in_dim"(%0) <{broadcast_dimensions = dense<> : tensor<0xi64>}> : (tensor<i32>) -> tensor<1x750xi32>
  // CHECK: return %1 : tensor<1x750xi32>
}

// CHECK-LABEL: @splat_scalar_no_change
func.func @splat_scalar_no_change() -> (tensor<f32>, tensor<i32>) {
  // CHECK-NOT: mhlo.broadcast_in_dim

    // CHECK-NEXT: %4 = mhlo.convert %3 : tensor<i32>
    // CHECK-NEXT: %5 = "mhlo.slice"(%1) <{limit_indices = dense<2> : tensor<1xi64>, start_indices = dense<1> : tensor<1xi64>, strides = dense<1> : tensor<1xi64>}> : (tensor<4xi32>) -> tensor<1xi32>
    // CHECK-NEXT: %6 = mhlo.reshape %5 : (tensor<1xi32>) -> tensor<i32>
    // CHECK-NEXT: %7 = mhlo.convert %6 : tensor<i32>

  %conv = "tfl.conv_2d"(%0, %w, %b) {
    dilation_h_factor = 1 : i32, dilation_w_factor = 1 : i32, fused_activation_function = "RELU",
    padding = "SAME", stride_h = 1 : i32, stride_w = 1 : i32
  } : (tensor<1x5x5x2xf32>, tensor<3x1x1x2xf32>, tensor<3xf32>) -> tensor<1x5x5x3xf32>
  %conv2 = "tfl.conv_2d"(%0, %w, %b2) {
    dilation_h_factor = 1 : i32, dilation_w_factor = 1 : i32, fused_activation_function = "RELU",

  // CHECK-SAME: tfl.schema_version = 3 : i32

  // CHECK:          %{{.*}} = "tfl.pseudo_const"() <{value = dense<{{\[\[1, 2\], \[3, 4\], \[5, 6\]\]}}> : tensor<3x2xi32>}>
  // CHECK-NEXT:     [[SUB:%.*]] = tfl.sub %{{.*}}, %{{.*}} {fused_activation_function = "RELU6"} : tensor<3x2xi32>
  // CHECK-NEXT:     [[SCALAR:%.*]] = "tfl.pseudo_const"() <{value = dense<10> : tensor<i32>}> : () -> tensor<i32>

    %cst_3 = "tf.Const"() {value = dense<-1> : tensor<i32>} : () -> tensor<i32>
    %cst_4 = "tf.Const"() {value = dense<0.00117647066> : tensor<f32>} : () -> tensor<f32>
    %cst_5 = "tf.Const"() {value = dense<-43> : tensor<i32>} : () -> tensor<i32>
    %cst_6 = "tf.Const"() {value = dense<0.00156862743> : tensor<f32>} : () -> tensor<f32>

}

func.func @argmin(%arg0: tensor<3xi32>, %arg1: tensor<i32>) -> tensor<i32> {
  %0 = "tf.ArgMin"(%arg0, %arg1) : (tensor<3xi32>, tensor<i32>) -> tensor<i32>
  func.return %0 : tensor<i32>

// CHECK-LABEL: argmin
// CHECK:  "tfl.arg_min"(%arg0, %arg1) : (tensor<3xi32>, tensor<i32>) -> tensor<i32>
}

func.func @sigmoid(%arg0: tensor<?x88xf32>) -> tensor<?x88xf32> {

  }];

  let arguments = (ins
    TFL_TensorOf<[F32, I16, I32, I64, I8, UI8, QI8, QUI8, QI16]>:$value,
    TFL_1DTensorOf<[I32], [I32]>:$size_splits,
    TFL_0DTensorOf<[I32], [I32]>:$split_dim,
    ConfinedAttr<I32Attr, [IntPositive]>:$num_splits
  );

  let results = (outs
    TFL_VariadicTensorOf<[F32, I16, I32, I64, I8, UI8, QI8, QUI8, QI16]>:$outputs
  );

  let hasVerifier = 1;