func.return %4 : tensor<!tf_type.variant>
}
// CHECK-LABEL: __inference_Dataset_flat_map_lambda_190
func.func private @__inference_Dataset_flat_map_lambda_190(%arg0: tensor<i64> {tf._user_specified_name = "args_0"}) -> tensor<!tf_type.variant> attributes {tf._original_func_name = "__inference_Dataset_flat_map_lambda_19", tf._tf_data_function = true, tf.signature.is_stateful} {

// The signature of the function is not changed.
// CHECK: func @cast_bf16_add_v2_to_fp32(%[[ARG_0:.*]]: tensor<2xbf16>, %[[ARG_1:.*]]: tensor<2xbf16>) -> tensor<2xf32>

// bfloat16 operands are cast to f32 operands.
// CHECK-DAG: %[[CAST_0:.*]] = "tf.Cast"(%[[ARG_0]]) <{Truncate = false}> : (tensor<2xbf16>) -> tensor<2xf32>
// CHECK-DAG: %[[CAST_1:.*]] = "tf.Cast"(%[[ARG_1]]) <{Truncate = false}> : (tensor<2xbf16>) -> tensor<2xf32>

  selector:
    k8s-app: kube-dns
  template:
    metadata:
      labels:
        k8s-app: kube-dns
    spec:
      containers:
      - name: carbon-relay
        image: banno/carbon-relay
        args:
        -

    }
    tf_executor.fetch
  }
  func.return
}
// CHECK-LABEL: empty_island_multiple_data_results
// CHECK-SAME: (%[[ARG_0:.*]]: tensor<*xf32>, %[[ARG_1:.*]]: tensor<*xi32>)
// CHECK: %[[IDENTITY_N:.*]]:2 = "tf.IdentityN"
// CHECK-SAME: (%[[ARG_0]], %[[ARG_1]])
// CHECK: tf_executor.yield %[[IDENTITY_N]]#0, %[[IDENTITY_N]]#1

        prometheus.io/path: /stats/prometheus
        prometheus.io/port: "15020"
        prometheus.io/scrape: "true"
        proxy.istio.io/overrides: '{"initContainers":[{"name":"istio-proxy","command":["envoy"],"args":["-c","my-config.yaml"],"resources":{}}]}'

tf_device.replicate([%0#0, %1#0] as %arg13: tensor<2x224x224x3xf32>, [%0#1, %1#1] as %arg14: tensor<2x1xf32>, %arg6 as %arg15: tensor<*x!tf_type.resource<tensor<7x7x3x64xf32>>>, %arg8 as %arg16: tensor<*x!tf_type.resource<tensor<1001xf32>>>, %arg7 as %arg17: tensor<*x!tf_type.resource<tensor<64x1001xf32>>>, %arg9 as %arg18: tensor<*x!tf_type.resource<tensor<f32>>>, %arg10 as %arg19: tensor<*x!tf_type.resource<tensor<f32>>>, %arg11 as %arg20: tensor<*x!tf_type.resource<tensor<f32>>>, %arg12 as %arg21:...

  %arg6: tensor<2048x640xf32>,
  %arg7: tensor<2048x640xf32>,
  %arg8: tensor<2048x640xf32>,
  %arg9: tensor<2048xf32>,
  %arg10: tensor<2048xf32>,
  %arg11: tensor<2048xf32>,
  %arg12: tensor<2048xf32>,
  %arg13: tensor<640x2048xf32>,
  %arg14: tensor<640xf32>,
  %arg15: tensor<2048xf32>,
  %arg16: tensor<2048xf32>,
  %arg17: tensor<2048xf32>,
  %arg18: tensor<2048xf32>,
  %arg19: tensor<1x640xf32>,

    key: "T"
    value {
      type: DT_FLOAT
    }
  }
  attr {
    key: "Tshape"
    value {
      type: DT_INT32
    }
  }
}

# MLIR-LABEL: func @main
# MLIR-SAME:  (%[[ARG_0:[a-z0-9]+]]: tensor<1x1x1x256x!quant.uniform<i8:f32, 0.21632751372549019:27>>) -> tensor<1x6x31x!quant.uniform<i8:f32, 0.09363494573854933:22>>
# MLIR-SAME:  control_outputs = ""
# MLIR-SAME:  inputs = "input"

        given:
        sourceFile << """
            #define ARGS_3 ("hello.h")
            #define FUNC_NAME() F
            #define F(X) X
            #define CONCAT_(X, Y) X ## Y
            #define CONCAT(X, Y) CONCAT_(X, Y)
            #define HEADER(X, Y, Z) CONCAT(X ## Y, ARGS_## Z)
            #include HEADER(FUNC_NAME, (), 3)
        """

        expect:

      func.return
    }
    ```

    with the following reduction function:

    ```mlir
    func.func private @__reduce_func_1(%arg0: tensor<i64> {tf._user_specified_name = "args_0"},
      %arg1: tensor<32xf32> {tf._user_specified_name = "args_1"}) -> (tensor<i64>)
      attributes {tf._tf_data_function = true, tf.signature.is_stateful} {
        %0 = "tf.JustPretend"(%arg1) : (tensor<32xf32>) -> (tensor<i64>)