%32 = "tf.ConcatV2"(%29#0, %31, %14) {device = ""} : (tensor<2xi64>, tensor<1xi64>, tensor<i32>) -> tensor<3xi64>
  %33 = "tf.GatherV2"(%24#2, %29#1, %14) {batch_dims = 0 : i64, device = ""} : (tensor<?xi64>, tensor<?xi64>, tensor<i32>) -> tensor<?xi64>
  %34 = "tf.ConcatV2"(%33, %22, %14) {device = ""} : (tensor<?xi64>, tensor<1xi64>, tensor<i32>) -> tensor<?xi64>

  func.return %0#0 : tensor<8x8x8x8xf32>
}

// -----

// CHECK-LABEL: fusedBatchNormV3_noTraining_mixedPrecision

  %1 = "tfl.sub"(%0, %cst) {fused_activation_function = "NONE"} : (tensor<1x64x84x32xf32>, tensor<32xf32>) -> tensor<1x64x84x32xf32>
  func.return %1 : tensor<1x64x84x32xf32>

  // CHECK-DAG: %[[SHAPE:.*]] = arith.constant dense<[1, 64, 84, 32]> : tensor<4xi32>

//        start_indices = dense<0> : tensor<2xi64>,
//        strides = dense<1> : tensor<2xi64>} :
//        (tensor<4x6xf32>) -> tensor<4x1xf32>
// %1 = "mhlo.slice"(%input) {
//        limit_indices = dense<[4, 3]> : tensor<2xi64>,
//        start_indices = dense<[0, 1]> : tensor<2xi64>,
//        strides = dense<1> : tensor<2xi64>} :
//        (tensor<4x6xf32>) -> tensor<4x2xf32>