%2 = "quantfork.stats"(%1) {layerStats = dense<[5.00000000e-6, 7.00000000e-1]> : tensor<2xf32>} : (tensor<2x2x3xf32>) -> tensor<2x2x3xf32>
    return %2 : tensor<2x2x3xf32>
  }
// CHECK: func.func private @quantize_dot_general_batch_per_tensor_quantized_fn(%[[ARG_0:.+]]: tensor<2x2x2xf32>) -> tensor<2x2x3xf32> attributes {tf._original_func_name = "main_0"}

  %0 = "tfl.var_handle"() {container = "", shared_name = "states"} : () -> tensor<!tf_type.resource<tensor<1x2x3xf32>>>
  %1 = "tfl.read_variable"(%0) : (tensor<!tf_type.resource<tensor<1x2x3xf32>>>) -> tensor<1x2x3xf32>
  %2 = "tfl.quantize"(%1) {qtype = tensor<1x2x3x!quant.uniform<u8<1:255>:f32, 0.047244094488188976:128>>} : (tensor<1x2x3xf32>) -> tensor<1x2x3x!quant.uniform<u8<1:255>:f32, 0.047244094488188976:128>>

func.func @leaky_relu_grad(%arg0: tensor<1x4x4xf32>, %arg1: tensor<1x4x4xf32>) -> tensor<1x4x4xf32> attributes {tf.entry_function = {}} {
    // CHECK-NEXT: %[[ALPHA:.*]] = "chlo.constant_like"(%arg1) <{value = 2.000000e-01 : f32}> : (tensor<1x4x4xf32>) -> tensor<1x4x4xf32>
    // CHECK-NEXT: %[[ZERO:.*]] = "chlo.constant_like"(%arg1) <{value = 0.000000e+00 : f32}> : (tensor<1x4x4xf32>) -> tensor<1x4x4xf32>

}

func.func @ConvertIdentityGatherNdOp(%arg0: tensor<4x3xf32>) -> tensor<4x3xf32> {
  %cst = arith.constant dense<[[0], [1], [2], [3]]> : tensor<4x1xi32>
  %0 = "tfl.gather_nd"(%arg0, %cst) : (tensor<4x3xf32>, tensor<4x1xi32>) -> tensor<4x3xf32>
  func.return %0 : tensor<4x3xf32>

// CHECK-LABEL: ConvertIdentityGatherNdOp
// CHECK-SAME: (%[[ARG:.*]]: tensor<4x3xf32>) -> tensor<4x3xf32>
// CHECK-NEXT: return %[[ARG]] : tensor<4x3xf32>
}

}

// -----

// CHECK-LABEL: topk_2
func.func @topk_2(%arg0: tensor<3x4x8xf32>) -> (tensor<3x4x2xf32>, tensor<3x4x2xi32>) {
  %0 = arith.constant dense<2> : tensor<i32>
  %1:2 = "tfl.topk_v2"(%arg0, %0) : (tensor<3x4x8xf32>, tensor<i32>) -> (tensor<3x4x2xf32>, tensor<3x4x2xi32>)
  func.return %1#0, %1#1: tensor<3x4x2xf32>, tensor<3x4x2xi32>
}

// -----

// CHECK-LABEL: topk_d

// CHECK:           }) : (tensor<5x4x3x7xf32>, tensor<2x2xi32>, tensor<2x5x3xf32>) -> tensor<5x4x3x7xf32>
// CHECK:           return %[[VAL_3]] : tensor<5x4x3x7xf32>
// CHECK:         }
func.func @convert_scatter_update_to_non_trailing_operand_dimensions(
  %arg0: tensor<5x4x3x7xf32>,
  %arg1: tensor<2x2xi32>,
  %arg2: tensor<2x5x3xf32>) -> tensor<5x4x3x7xf32>
{
  %0 = "mhlo.scatter"(%arg0, %arg1, %arg2) ({

  func.func @batchmatmulv2(%arg0: tensor<1x4x2xf32>, %arg1: tensor<3x2x4xf32>) -> tensor<3x4x4xf32> {
    // CHECK: mhlo.reduce
    // CHECK: mhlo.dot_general
    // CHECK: mhlo.transpose
    %0 = "tf.BatchMatMulV2"(%arg0, %arg1) {T = f32, adj_x = false, adj_y = false, grad_x = false, grad_y = false, device = ""} : (tensor<1x4x2xf32>, tensor<3x2x4xf32>) -> tensor<3x4x4xf32>
    func.return %0 : tensor<3x4x4xf32>
  }

func.func @partitioned_input_rank_mismatch(%arg0: tensor<!tf_type.resource<tensor<1x4x4xf32>>>) {
  // expected-error @+1 {{rank}}
  %0 = "tf.TPUPartitionedInputV2"(%arg0) {_XlaSharding = "\08\03\1A\05\04\01\01\01\02\22\08\00\01\02\03\04\05\06\070\01", partition_dims = [4, 1, 1, 2], is_packed = true} : (tensor<!tf_type.resource<tensor<1x4x4xf32>>>) -> tensor<!tf_type.resource<tensor<4x4x4xf32>>>

func.func private @dense_image_warp_invalid_input_shape(%arg0: tensor<2x4x4xf32>, %arg1: tensor<2x4x4x2xf32>) -> tensor<2x4x4x1xf32> attributes {tf._implements = "addons:DenseImageWarp"}

// expected-warning @+1 {{Flow should be a 4D float tensor}}
func.func private @dense_image_warp_invalid_flow_shape(%arg0: tensor<2x4x4x1xf32>, %arg1: tensor<2x4x4xf32>) -> tensor<2x4x4x1xf32> attributes {tf._implements = "addons:DenseImageWarp"}

// %0:3 = "tf.SplitV"(%input, %split_sizes, %split_dim) :
//                   (tensor<4x6xf32>, tensor<3xi32>, tensor<i32>) ->
//                   (tensor<4x1xf32>, tensor<4x2xf32>, tensor<4x3xf32>)
//
// We will generate slices following slices:
// %0 = "mhlo.slice"(%input) {
//        limit_indices = dense<[4, 1]> : tensor<2xi64>,
//        start_indices = dense<0> : tensor<2xi64>,