// CHECK-LABEL: func private @composite_depthwise_conv2d_fn(
// CHECK-SAME:                                             %arg0: tensor<1x3x4x3xf32>,
// CHECK-SAME:                                             %arg1: tensor<2x3x3x2xf32>)

// CHECK-LABEL: func private @composite_depthwise_conv2d_fn_0(
// CHECK-SAME:                                             %arg0: tensor<1x3x4x3xf32>,
// CHECK-SAME:                                             %arg1: tensor<2x3x1x6xf32>)

// CHECK-LABEL: func @main
// CHECK: tf.AddV2
// CHECK-SAME: {__op_key = 0 : i32}

// CHECK: tf.AddV2
// CHECK-SAME: {__op_key = 1 : i32}

// CHECK: tf.AddV2
// CHECK-SAME: {__op_key = 2 : i32}

// CHECK: tf.AddV2
// CHECK-SAME: {__op_key = 3 : i32}

// CHECK: tf.Sub
// CHECK-SAME: {__op_key = 4 : i32}

// CHECK: tf.Sub
// CHECK-SAME: {__op_key = 5 : i32}

// CHECK: tf.Sub

# it is being invoked.
# CHECK: module
# CHECK-SAME: tf.versions
# CHECK-SAME: bad_consumers
# CHECK-SAME: min_consumer
# CHECK-SAME: producer

# CHECK: "tf_saved_model.global_tensor"() <{is_mutable, sym_name = "[[VAR:[a-zA-Z_0-9]+]]", type = tensor<1x3xf32>, value = {{.*}} : tensor<1x3xf32>}> : () -> ()

# CHECK:      func {{@[a-zA-Z_0-9]+}}(
# CHECK-SAME:   [[ARG0:%.*]]: tensor<3x1xf32> {tf_saved_model.index_path = ["x"]},

// CHECK:       ENTRY %main.{{[0-9]+}}
// CHECK-SAME:  (arg_tuple.{{[0-9]+}}: (f32[128,10], f32[10,1024], f32[128,1024])) -> (f32[128,10], f32[10,1024], f32[128,1024]) {
// CHECK:         ROOT %tuple.{{[0-9]+}}
// CHECK-SAME:    sharding={
// CHECK-SAME:    {devices=[1,2]0,1}
// CHECK-SAME:    {maximal device=0}
// CHECK-SAME:    {replicated}

  #
  # CHECK:      func {{@[a-zA-Z_0-9]+}}(
  # CHECK-SAME:   %arg0: tensor<f32> {tf._user_specified_name = "x", tf_saved_model.index_path = [0]},
  # CHECK-SAME:   %arg1: tensor<!tf_type.resource<{{.*}}>>
  # CHECK-SAME: ) -> (
  # CHECK-SAME:   tensor<f32> {tf_saved_model.index_path = [0]},
  # CHECK-SAME:   tensor<f32> {tf_saved_model.index_path = [1]})
  # CHECK-SAME: attributes{{.*}}tf_saved_model.exported_names = ["callee"]

  // CHECK: [[PI_0:%.*]] = "tf.TPUPartitionedInputV2"([[ARG0]], [[ARG1]])
  // CHECK-SAME: _XlaSharding = ""
  // CHECK-SAME: partition_dims = []
  // CHECK: [[PI_1:%.*]] = "tf.TPUPartitionedInputV2"([[ARG2]], [[ARG3]])
  // CHECK-SAME: _XlaSharding = ""
  // CHECK-SAME: partition_dims = []
  // CHECK: [[RI:%.*]] = "tf.TPUReplicatedInput"([[PI_0]], [[PI_1]])

          strides = [1, 1, 2, 2]
        } : (tensor<?x3x230x230xf32>, tensor<7x7x3x64xf32>) -> tensor<?x64x112x112xf32>

  // CHECK: %[[CONV0:[0-9]*]] = "tf.Conv2D"
  // CHECK-SAME: %[[PAD]]
  // CHECK-SAME: data_format = "NHWC"
  // CHECK-SAME: strides = [1, 2, 2, 1]

  %6, %batch_mean, %batch_variance, %reserved_1, %reserved_2, %reserved_3 =
       "tf.FusedBatchNormV3"(%5, %arg1, %arg1, %arg1, %arg1)
       {

    func.return %4 : tensor<100x50xf32>
  }
  // CHECK: "tf.VarHandleOp"
  // CHECK-SAME: shared_name = "var1"
  // CHECK-SAME: _is_initialized = true
  // CHECK: "tf.VarHandleOp"
  // CHECK-SAME: shared_name = "var2"
  // CHECK-SAME: _is_initialized = true
  // CHECK: "tf.VarHandleOp"
  // CHECK-SAME: shared_name = "var3"
  // CHECK-SAME: _is_initialized = false

  // INVALID-NOT: _is_initialized

}

# CHECK: tf.HashTableV2
# CHECK-SAME: shared_name = "hash_table_node"

# CHECK: "tf.Variable"
# CHECK-SAME: shared_name = "variable_node"
# CHECK: "tf.VariableV2"
# CHECK-SAME: shared_name = "variable_v2_node"

# CHECK: "tf.BatchFunction"
# CHECK-SAME: shared_name = "batch_node"

# CHECK: func private @create_resource
# CHECK: tf.HashTableV2
# CHECK-SAME: shared_name = "hash_table_node@create_resource"

       : (tensor<1x3x4x3xbf16>, tensor<2x3x3x2xbf16>) -> tensor<1x3x2x2xbf16> loc("Model/conv2d")
  %2 = "tf.Cast"(%1) {Truncate = false} : (tensor<1x3x2x2xbf16>) -> tensor<1x3x2x2xf32>
  %3 = "tf.IdentityN"(%2) {device = ""} : (tensor<1x3x2x2xf32>) -> tensor<1x3x2x2xf32>
  return %3 : tensor<1x3x2x2xf32>
// CHECK: tf.Conv2D
// CHECK-SAME: loc("Model/conv2d")
}