%1 = "tf.Transpose"(%arg0, %0) : (tensor<1x4x4x8xf32>, tensor<4xi32>) -> tensor<1x8x4x4xf32>
  %2 = "tf.Transpose"(%1, %0) : (tensor<1x8x4x4xf32>, tensor<4xi32>) -> tensor<1x4x8x4xf32>
  %3 = "tf.Transpose"(%arg1, %0) : (tensor<1x4x4x8xf32>, tensor<4xi32>) -> tensor<1x8x4x4xf32>
  %4 = "tf.Transpose"(%3, %0) : (tensor<1x8x4x4xf32>, tensor<4xi32>) -> tensor<1x4x8x4xf32>

  %group_assignment = "tf.Const"() { value = dense<[[0, 1]]> : tensor<1x2xi32> } : () -> tensor<1x2xi32>
  %instance_key = "tf.Const"() { value = dense<3> : tensor<i32> } : () -> tensor<i32>
  %group_size, %group_key = "tf.CollectiveAssignGroupV2"(%group_assignment, %rank, %key_base) {} : (tensor<1x2xi32>, tensor<i32>, tensor<i32>) -> (tensor<i32>, tensor<i32>)
  // CHECK-NOT: "tf.CollectiveAssignGroupV2"

  // CHECK: _arithmetic_count = 3000 : i64
  %0 = "tfl.fully_connected"(%arg0, %arg1, %arg2) {fused_activation_function = "NONE", keep_num_dims = false, weights_format = "DEFAULT"} : (tensor<1x37xf32>, tensor<40x37xf32>, tensor<40xf32>) -> tensor<1x40xf32>
  func.return %0 : tensor<1x40xf32>
}

        }) : () -> tensor<f32>
        "tf_device.cluster"() ({
          %group = "tf.Const"() <{value = dense<[[0, 1]]> : tensor<1x2xi32>}> : () -> tensor<1x2xi32>
          %arg1_reduced = "tf.XlaAllReduce"(%arg1_id, %group) <{mode = "CrossReplica", reduce_op = "Add"}> : (tensor<f32>, tensor<1x2xi32>) -> tensor<f32>
          "tf.OpA"(%arg1_reduced) : (tensor<f32>) -> ()
          tf_device.return
        }) : () -> ()

  // CHECK-LABEL: @xla_svd
  func.func @xla_svd(%arg0: tensor<1x1xf32>) -> (tensor<1xf32>, tensor<1x1xf32>, tensor<1x1xf32>) {
    // CHECK-NOT: XlaSvd
    %s, %u, %v = "tf.XlaSvd"(%arg0) {max_iter = 1, epsilon = 1.0E-09 : f32, precision_config = ""} : (tensor<1x1xf32>) -> (tensor<1xf32>, tensor<1x1xf32>, tensor<1x1xf32>)
    func.return %s, %u, %v : tensor<1xf32>, tensor<1x1xf32>, tensor<1x1xf32>
  }

  func.func @identity(%arg0: f32) -> f32 {

}

// CHECK-LABEL: QuantizeConcat
func.func @QuantizeConcat(tensor<1x2xf32>, tensor<1x2xf32>) -> tensor<2x2x!quant.uniform<u8:f32, 1.000000e-01:128>> {
^bb0(%arg0: tensor<1x2xf32>, %arg1: tensor<1x2xf32>):
  %0 = "tfl.concatenation"(%arg0, %arg1) {axis = 0 : i32, fused_activation_function = "NONE"} : (tensor<1x2xf32>, tensor<1x2xf32>) -> tensor<2x2xf32>

  return %1#0, %1#1 : tensor<1x2xf32>, tensor<1x2xf32>
}
func.func @_func(%arg0: tensor<2x4xf32>, %arg1: tensor<4x2xf32>) -> tensor<2x2xf32> {
  %0 = "tf.MatMul"(%arg0, %arg1) {_XlaSharding = "\08\03\1A\02\02\01\22\02\00\01"} : (tensor<2x4xf32>, tensor<4x2xf32>) -> tensor<2x2xf32>
  return %0 : tensor<2x2xf32>
}

// -----

      %0 = stablehlo.dot_general %arg0, %arg1, contracting_dims = [1] x [0], precision = [DEFAULT, DEFAULT] : (tensor<1x1024xf32>, tensor<1024x3xf32>) -> tensor<1x3xf32>
      return %0 : tensor<1x3xf32>
    }
  }
)mlir";

TEST_F(LiftAsFunctionCallTest, LiftedFunctionSucceeds) {

  %8 = "tfl.concatenation"(%2, %0) {axis = -1 : i32, fused_activation_function = "NONE"} : (tensor<1x1xf32>, tensor<1x1xf32>) -> tensor<1x2xf32>
  %9 = "quantfork.stats"(%8) {layerStats = dense<[-0.488159984, 0.189515018]> : tensor<2xf32>} : (tensor<1x2xf32>) -> tensor<1x2xf32>
  %10 = "tfl.concatenation"(%9, %7) {axis = -1 : i32, fused_activation_function = "NONE"} : (tensor<1x2xf32>, tensor<1x2xf32>) -> tensor<1x4xf32>

}

func.func @main2(%arg0: tensor<2x4xf32>, %arg1: tensor<2x4xf32>) -> tensor<2x4xf32> {
  %0 = "tfl.quantize"(%arg0) {qtype = tensor<2x4x!quant.uniform<u8:f32, 0.49803921568627452>>} : (tensor<2x4xf32>) -> tensor<2x4x!quant.uniform<u8:f32, 0.49803921568627452>>
  %1 = "tfl.quantize"(%arg1) {qtype = tensor<2x4x!quant.uniform<u8:f32, 0.49803921568627452>>} : (tensor<2x4xf32>) -> tensor<2x4x!quant.uniform<u8:f32, 0.49803921568627452>>