```mlir
      %0 = "tf.Const"() {value = dense<[[42.0]]> : tensor<1x1xf32>} : () -> tensor<1x1xf32>
      %1 = "tf.Const"() {device = "", value = dense<[[42.0]]> : tensor<1x1xf32>} : () -> tensor<1x1xf32>
      %2 = "tf.Const"() {device = "baz", value = dense<[[42.0]]> : tensor<1x1xf32>} : () -> tensor<1x1xf32>
    ```

    then running this pass with 'default-device=foobar', we get:

    ```mlir

}

func.func @einsum_reducesum(%arg0: tensor<2x5x7xf32>, %arg1: tensor<5x2xf32>) -> tensor<5x7xf32> {
  %0 = "tf.Einsum"(%arg0, %arg1) {T = "tfdtype$DT_FLOAT", equation = "lbh,bl->bh"}: (tensor<2x5x7xf32>, tensor<5x2xf32>) -> tensor<5x7xf32>
  func.return %0 : tensor<5x7xf32>
  // CHECK-LABEL: einsum_reducesum
  // CHECK-DAG: %[[cst:.*]] = arith.constant dense<[1, 2, 0]> : tensor<3xi32>

  func.return %0 : tensor<10x10xf32>
  // CHECK-LABEL: strided_slice_with_constant_attributes
  // CHECK-DAG: [[BEGIN:%cst.*]] = arith.constant dense<[-1, 0, 0]> : tensor<3xi32>
  // CHECK-DAG: [[END:%cst.*]] = arith.constant dense<[0, 10, 10]> : tensor<3xi32>

  %max_size = "tf.Const"() {value = dense<10> : tensor<i32>} : () -> tensor<i32>
  // CHECK-NOT: tf.EmptyTensorList
  // CHECK: tf.BroadcastTo
  // CHECK-SAME: tensor<10x16xf32>
  %tl0 = "tf.EmptyTensorList"(%element_shape, %max_size) : (tensor<1xi32>, tensor<i32>) -> tensor<!tf_type.variant<tensor<16xf32>>>
  // CHECK-NOT: tf.TensorListReserve
  // CHECK: tf.BroadcastTo

// CHECK:   [[ZERO:%.*]] = arith.constant dense<0> : tensor<i32>
// CHECK:   [[EXP_ITEM:%.*]] = "tf.ExpandDims"([[ITEM]], [[ZERO]]) {{.*}} -> tensor<1x10xf32>
// CHECK:   [[RESULT:%.*]] = "tf.Concat"(%cst, [[INPUT]], [[EXP_ITEM]]) : {{.*}} -> tensor<?x10xf32>
// CHECK:   return [[RESULT]] : tensor<?x10xf32>
}

// -----

// CHECK:           }) : (tensor<5x10xf32>, tensor<3x1xi32>, tensor<10x3xf32>) -> tensor<5x10xf32>
// CHECK:           return %[[VAL_3]] : tensor<5x10xf32>
// CHECK:         }
func.func @convert_scatter_update_with_non_trailing_update_window_dims(
  %arg0: tensor<5x10xf32>,
  %arg1: tensor<3x1xi32>,
  %arg2: tensor<10x3xf32>) -> tensor<5x10xf32>
{
  %0 = "mhlo.scatter"(%arg0, %arg1, %arg2) ({

func.func @testDynamicBatchMatMulToV2(%arg0: tensor<2x3x5xf32>, %arg1: tensor<?x5x7xf32>) -> tensor<2x3x7xf32> {
  // CHECK: "tf.BatchMatMul"(%arg0, %arg1) <{adj_x = false, adj_y = false}> {device = "/job:localhost/replica:0/task:0/device:GPU:0"}
  %0 = "tf.BatchMatMul"(%arg0, %arg1) {adj_x = false, adj_y = false, device = "/job:localhost/replica:0/task:0/device:GPU:0"} : (tensor<2x3x5xf32>, tensor<?x5x7xf32>) -> tensor<2x3x7xf32>
  func.return %0: tensor<2x3x7xf32>
}

mlir_module = '''python
func @main(%arg0 : tensor<10xf32>, %arg1 : tensor<10xf32>) -> tensor<10x10xf32> {
   %add = "magic.op"(%arg0, %arg1) : (tensor<10xf32>, tensor<10xf32>) -> tensor<10x10xf32>
   return %ret : tensor<10x10xf32>
}
'''

@tf.function
def foo(x, y):
  return mlir_passthrough_op([x, y], mlir_module, Toutputs=[tf.float32])