func.return %0 : tensor<1x112x112x32xf32>
}

// -----

func.func @testAvgPool(%arg0: tensor<256x32x32x3xf32>) -> tensor<256x30x30x16xf32> {
  // CHECK: tac.device = "GPU", tac.inference_type = "FLOAT"

  %6 = "tfl.quantize"(%5) {qtype = tensor<1x112x112x32x!quant.uniform<u8:f32, 0.023528476789885875>>} : (tensor<1x112x112x32xf32>) -> tensor<1x112x112x32x!quant.uniform<u8:f32, 0.023528476789885875>>
  func.return %6 : tensor<1x112x112x32x!quant.uniform<u8:f32, 0.023528476789885875>>

  func.return %3 : tensor<1x112x112x32xf32>

  func.return %3 : tensor<1x112x112x32xf32>

// RUN: tf-opt %s -tf-optimize | FileCheck %s

// CHECK-LABEL: @fuseMulIntoConv2d
func.func @fuseMulIntoConv2d(%arg0: tensor<1x112x112x3xf32>) -> tensor<1x28x23x2xf32> {
  %cst0 = arith.constant dense<[[[[1.0, 2.0], [3.0, 4.0], [5.0, 6.0]], [[7.0, 8.0], [9.0, 10.0], [11.0, 12.0]], [[13.0, 14.0], [15.0, 16.0], [17.0, 18.0]]]]> : tensor<1x3x3x2xf32>
  %cst2 = arith.constant dense<[1.0, 2.0]> : tensor<2xf32>

  %6 = "tfl.quantize"(%5) {qtype = tensor<1x112x112x32x!quant.uniform<u8:f32, 0.023528476789885875>>} : (tensor<1x112x112x32xf32>) -> tensor<1x112x112x32x!quant.uniform<u8:f32, 0.023528476789885875>>
  func.return %6 : tensor<1x112x112x32x!quant.uniform<u8:f32, 0.023528476789885875>>

     %space_to_depth = "tf.SpaceToDepth"(%input) {block_size = 2, ...}:
        (tensor<2x224x224x3xf32>) -> tensor<2x112x112x12xf32>
     %device_launch = "tf_device.launch_func"(%space_to_depth,...) {func = @_func,...)
     return ...
   }
   func @_func(%input: tensor<2x112x112x12xf32>,
              %filter: tensor<7x7x3x64xf32>) {
     %filter_transform = "tf.Pad/tf.Transpose/tf.Reshape"(%filter):

  func.return %2 : tensor<1x112x112x32xf32>
}

// -----

// CHECK-LABEL: @test_fullyconnected_replace_float
func.func @test_fullyconnected_replace_float(%arg0: tensor<4x256x6x6xf32>) -> tensor<4x256x36xf32> {

    // CHECK-DAG: %[[SPACETODEPTH0:.*]] = "tf.SpaceToDepth"([[INPUT:.*]]) <{block_size = 2 : i64, data_format = "NHWC"}> : (tensor<2x224x224x3xf32>) -> tensor<2x112x112x12xf32>
    %2 = "tf.AddV2"(%arg2, %arg3) {device = ""} : (tensor<i32>, tensor<i32>) -> tensor<i32>
    %3 = "tf.ReadVariableOp"(%arg6) : (tensor<!tf_type.resource<tensor<7x7x3x64xf32>>>) -> tensor<7x7x3x64xf32>

        %space_to_depth = "tf.SpaceToDepth"(%input) {block_size = 2, ...}: (tensor<2x224x224x3xf32>) -> tensor<2x112x112x12xf32>
        %device_launch = "tf_device.cluster_func"(%space_to_depth,...) {func = @_func,...)
        return ...
      }
      func @_func(%input: tensor<2x112x112x12xf32>, %filter: tensor<7x7x3x64xf32>) {