%outputs_0, %control_1 = tf_executor.island(%control) wraps "tf.Const"() {value = dense<0.000000e+00> : tensor<16xf32>} : () -> tensor<16xf32>
      %outputs_2, %control_3 = tf_executor.island(%control_1) wraps "tf.Const"() {value = dense<0.000000e+00> : tensor<5x5x32x16xf32>} : () -> tensor<5x5x32x16xf32>

  func.return %0 : tensor<256x32x32x16xf32>
}

func.func @testConv2DDynamicShape(tensor<?x32x32x3xf32>, tensor<16x3x3x3xf32>, tensor<16xf32>) -> tensor<?x32x32x16xf32> {
^bb0(%arg0: tensor<?x32x32x3xf32>, %arg1: tensor<16x3x3x3xf32>, %arg2: tensor<16xf32>):
  // CHECK: _arithmetic_count = -1 : i64

// PerChannel-DAG: %[[zp1:.*]] = "tf.Const"() <{value = dense<{{[0-9]+}}> : tensor<3xi32>}> : () -> tensor<3xi32>
// PerChannel-DAG: %[[zp2:.*]] = "tf.Const"() <{value = dense<{{[0-9]+}}> : tensor<6xi32>}> : () -> tensor<6xi32>
// PerChannel: %[[out_1:.*]] = "tf.PartitionedCall"(%arg0, %[[q_w1]], %[[scale1]], %[[zp1]]) <{config = "", config_proto = "", executor_type = "",

func.func @padSliceTo4D(%arg0: tensor<4x384x32xf32>) -> tensor<1x384x32xf32> {
  %0 = "tfl.pseudo_const"() {value = dense<0> : tensor<3xi32>} : () -> tensor<3xi32>
  %1 = "tfl.pseudo_const"() {value = dense<[1, 384, 32]> : tensor<3xi32>} : () -> tensor<3xi32>
  %2 = "tfl.slice"(%arg0, %0, %1) : (tensor<4x384x32xf32>, tensor<3xi32>, tensor<3xi32>) -> tensor<1x384x32xf32>
  func.return %2 : tensor<1x384x32xf32>
}

  %0 = "tfr.cast"(%arg0) : (tensor<*xf32>) -> !tfr.tensor
  %1 = "tfr.cast"(%0) : (!tfr.tensor) -> tensor<*xi32>
  %2 = "tfr.cast"(%0) : (!tfr.tensor) -> tensor<2xi32>
  func.return %1, %2 : tensor<*xi32>, tensor<2xi32>

// CHECK: %[[tf_cast_unranked:.*]] = "tf.Cast"(%arg0) <{Truncate = false}> : (tensor<*xf32>) -> tensor<*xi32>

// PerTensor-LABEL: QuantizeGatherWeightOnly
func.func @QuantizeGatherWeightOnly(%arg0: tensor<3xi32>) -> tensor<3x3x3x3xf32> {
  %w = arith.constant dense<1.270000e+02> : tensor<64x3x3x3xf32>
  %emb = "tfl.gather"(%w, %arg0) {axis = 0 : i32, batch_dims = 0 : i32} : (tensor<64x3x3x3xf32>, tensor<3xi32>) -> tensor<3x3x3x3xf32>

  %min = arith.constant dense<0.0> : tensor<16xf32>
  %max = arith.constant dense<255.0> : tensor<16xf32>
  %mini = "tf.Identity"(%min) : (tensor<16xf32>) -> tensor<16xf32>
  %maxi = "tf.Identity"(%max) : (tensor<16xf32>) -> tensor<16xf32>
  %fq = "tf.FakeQuantWithMinMaxVarsPerChannel"(%in, %mini, %maxi) {num_bits = 5, narrow_range = false} : (tensor<3x3x3x16xf32>, tensor<16xf32>, tensor<16xf32>) -> tensor<3x3x3x16xf32>

  %min = arith.constant dense<0.0> : tensor<16xf32>
  %max = arith.constant dense<15.0> : tensor<16xf32>
  %mini = "tf.Identity"(%min) : (tensor<16xf32>) -> tensor<16xf32>
  %maxi = "tf.Identity"(%max) : (tensor<16xf32>) -> tensor<16xf32>
  %fq = "tf.FakeQuantWithMinMaxVarsPerChannel"(%in, %mini, %maxi) {num_bits = 3, narrow_range = false} : (tensor<3x3x3x16xf32>, tensor<16xf32>, tensor<16xf32>) -> tensor<3x3x3x16xf32>

// CHECK:       %10 = "tfl.reduce_prod"(%9#2, %cst_9) <{keep_dims = true}> {tac.device = "CPU", tac.inference_type = "FLOAT"} : (tensor<2xi32>, tensor<1xi32>) -> tensor<1xi32>

  %2 = "tf.AddN"(%arg0, %1, %1) : (tensor<2xi32>, tensor<2xi32>, tensor<2xi32>) -> tensor<2xi32>
  %3 = "tf.AddN"(%1, %arg0, %1) : (tensor<2xi32>, tensor<2xi32> , tensor<2xi32>) -> tensor<2xi32>
  %4 = "tf.AddN"(%1, %1) : (tensor<2xi32>, tensor<2xi32>) -> tensor<2xi32>
  %5 = "tf.AddN"(%arg0, %1, %0) : (tensor<2xi32>, tensor<2xi32>, tensor<2xi32>) -> tensor<2xi32>
  func.return %2, %3, %4, %5: tensor<2xi32>, tensor<2xi32>, tensor<2xi32>, tensor<2xi32>
}