func.func @testSpaceToDepthF32(%arg0: tensor<1x2x2x1xf32>) -> tensor<1x1x1x4xf32> {
  // CHECK: %[[ARG:.*]]: tensor<1x2x2x1xf32>
  // CHECK: "tfl.space_to_depth"(%[[ARG]]) <{block_size = 2 : i32}> : (tensor<1x2x2x1xf32>) -> tensor<1x1x1x4xf32>
  %0 = "tfl.space_to_depth"(%arg0) {block_size = 2: i32} : (tensor<1x2x2x1xf32>) -> tensor<1x1x1x4xf32>
  func.return %0 : tensor<1x1x1x4xf32>
}

// -----

// CHECK:           %[[VAL_8:.*]] = "tfl.reshape"(%[[VAL_7]], %[[VAL_3]]) {tac.device = "GPU", tac.inference_type = "FLOAT"} : (tensor<1x1x1x2xf32>, tensor<1xi32>) -> tensor<2xf32>
// CHECK:           %[[VAL_9:.*]] = "tfl.reshape"(%[[VAL_8]], %[[VAL_4]]) {tac.device = "GPU", tac.inference_type = "FLOAT"} : (tensor<2xf32>, tensor<2xi32>) -> tensor<2x1xf32>

    %cst = "tf.Const"() {value = dense<3.00000000e-1> : tensor<2x3x3x2xf32>} : () -> tensor<2x3x3x2xf32>
    %cst_0 = "tf.Const"() {value = dense<4.00000000e-1> : tensor<1x1x1x2xf32>} : () -> tensor<1x1x1x2xf32>
    %0 = "quantfork.stats"(%arg0) {layerStats = dense<[6.00000000e-6, 9.00000000e-1]> : tensor<2xf32>} : (tensor<1x3x4x3xf32>) -> tensor<1x3x4x3xf32>
    %1 = "tf.XlaCallModule"(%0, %cst, %cst_0) {

  func.return %1 : tensor<1x1x1x16xf32>

// CHECK: %[[avgp:.*]] = "tfl.average_pool_2d"(%arg0)
// CHECK: %[[dq:.*]] = "tfl.dequantize"(%[[avgp]]) : (tensor<1x1x1x16x!quant.uniform<u8:f32, 7.812500e-03:128>>) -> tensor<1x1x1x16xf32>
// CHECK: return %[[dq]] : tensor<1x1x1x16xf32>
}

// CHECK-LABEL: QuantizeReshape2D

// CustomOpNotWeightOnly-LABEL: QuantizeCustomOp
func.func @QuantizeCustomOp(%arg0: tensor<1x1x1x1xf32>) -> tensor<*xf32> attributes {tf.entry_function = {inputs = "input", outputs = "custom_op"}} {
  %0 = "quantfork.stats"(%arg0) {layerStats = dense<[0.000000e+00, 2.550000e+02]> : tensor<2xf32>} : (tensor<1x1x1x1xf32>) -> tensor<1x1x1x1xf32>
  %w = arith.constant dense<127.0> : tensor<1024x1x1x1xf32>

func.func @attribute_cast(%arg0: tensor<1x4x4x1xf32>) -> tensor<1x2x2x1xf32> {
  %0 = "tfr.cast"(%arg0) : (tensor<1x4x4x1xf32>) -> !tfr.tensor
  %stride_i32 = arith.constant 2 : i32
  %1 = tfr.call @tf__my_max_pool(%0, %stride_i32, %stride_i32) : (!tfr.tensor, i32, i32) -> !tfr.tensor
  %2 = "tfr.cast"(%1) : (!tfr.tensor) -> tensor<1x2x2x1xf32>
  func.return %2 : tensor<1x2x2x1xf32>
// CHECK: tf__max_pool
}

  func.func @simple_folding(%arg0: tensor<1x1x1x1xi32>, %arg1: tensor<1x1x1x1xf32>) -> tensor<?x?x?x?xf32> {
    // CHECK: %[[SHAPE:.*]] = "tf.Shape"
    // CHECK: %[[CONV:.*]] = "tf.Conv2DBackpropInput"(%[[SHAPE]]
    // CHECK-SAME: (tensor<4xi32>, tensor<1x1x1x1xf32>, tensor<1x1x1x1xf32>) -> tensor<1x1x1x1xf32>
    // CHECK: return %[[CONV]] : tensor<1x1x1x1xf32>
    %0 = "tf.Shape"(%arg0) : (tensor<1x1x1x1xi32>) -> tensor<4xi32>

  %0 = "tf.Less"(%arg0, %arg1) : (tensor<8x7x6x5x?x3x2x1xf32>, tensor<?x3x2x1xf32>) -> tensor<8x7x6x5x?x3x2x1xi1>
  %1 = "tf.SelectV2"(%0, %arg0, %arg1) : (tensor<8x7x6x5x?x3x2x1xi1>, tensor<8x7x6x5x?x3x2x1xf32>, tensor<?x3x2x1xf32>) -> tensor<8x7x6x5x?x3x2x1xf32>
  func.return %1 : tensor<8x7x6x5x?x3x2x1xf32>

func.func @InvalidFuseTileAlreadyBroadcastAlongTileDim(%arg0: tensor<1x1x1x1xf32>) -> tensor<1x6x8x1xf32> {
  %cst_1 = arith.constant dense<[1, 6, 8, 1]> : tensor<4xi32>
  %cst_2 = arith.constant dense<[1, 1, 1, 46]> : tensor<4xi32>
  %cst_20 = arith.constant dense<4.600000e+01> : tensor<f32>
  %0 = "tfl.tile"(%arg0, %cst_1) : (tensor<1x1x1x1xf32>, tensor<4xi32>) -> tensor<1x6x8x1xf32>

    } : (tensor<1x6x6x16xf32>) -> tensor<1x1x1x16xf32>
  func.return %1 : tensor<1x1x1x16xf32>

// CHECK: %0 = "tfl.dequantize"(%arg0)
// CHECK: %1 = "tfl.average_pool_2d"(%0)
// CHECK: %2 = "tfl.quantize"(%1)
// CHECK: %3 = "tfl.dequantize"(%2)
// CHECK: return %3 : tensor<1x1x1x16xf32>
}

// CHECK-LABEL: QuantizeMaximum