- code_block: |
        <pre class = "prettyprint">
        // Syntactically similar to LLVM:
        func @testFunction(%arg0: i32) {
          %x = call @thingToCall(%arg0) : (i32) -> i32
          br ^bb1
        ^bb1:
          %y = arith.addi %x, %x : i32
          return %y : i32
        }
        </pre>

  - classname: devsite-landing-row-cards
    items:

func.func @tensorlistReserveConstantUnknownElementShapeDim(%arg0: tensor<i32>, %arg1: tensor<i32>) -> tensor<?x7xf32> {
  %cst = arith.constant dense<[-1, 7]> : tensor<2xi32>
  %0 = "tf.TensorListReserve"(%cst, %arg0) : (tensor<2xi32>, tensor<i32>) -> tensor<!tf_type.variant<tensor<?x7xf32>>>
  %1 = "tf.TensorListGetItem"(%0, %arg1, %cst) : (tensor<!tf_type.variant<tensor<?x7xf32>>>, tensor<i32>, tensor<2xi32>) -> tensor<?x7xf32>
  func.return %1 : tensor<?x7xf32>

// RUN: flatbuffer_translate -mlir-to-tflite-flatbuffer %s | flatbuffer_to_string - | FileCheck %s

func.func @main(%arg0: tensor<i32>, %arg1: tensor<i32>) -> tensor<!tf_type.variant<tensor<*xi32>>> {
  %0 = "tfl.custom"(%arg0, %arg1) {custom_code = "TensorListReserve", custom_option = #tfl<const_bytes : "0x02">} : (tensor<i32>, tensor<i32>) -> tensor<!tf_type.variant<tensor<*xi32>>>
  func.return %0 : tensor<!tf_type.variant<tensor<*xi32>>>
}

  %zps = "tf.Const"() { value = dense<3> : tensor<i32> } : () -> tensor<i32>

  // expected-error@+1 {{'tf.UniformQuantize' op is illegal as it is a UQ op or contains uq/qint types}}
  %0 = "tf.UniformQuantize"(%arg0, %scales, %zps) {
    quantization_axis = -1 : i64, quantization_min_val = -128 : i64, quantization_max_val = 127 : i64
  } : (tensor<1xf32>, tensor<f32>, tensor<i32>) -> tensor<1x!tf_type.qint8>

// CHECK: %[[CUSTOM_AGGREGATOR_0:.+]], {{.*}}, {{.*}}, {{.*}} = "tf.CustomAggregator"(%[[ARG_0]]) <{calibration_method = 1 : i32, id = "composite_dot_general_fn_1_arg_0_calibration_method_1", max_percentile = 0.000000e+00 : f32, min_percentile = 0.000000e+00 : f32, num_bins = 0 : i32}> : (tensor<1x4xf32>) -> (tensor<1x4xf32>, tensor<f32>, tensor<f32>, tensor<0xi64>)

  // CHECK-NEXT:   buffer: 3
  // CHECK-NEXT:   } ]
  // CHECK-NEXT:   signature_defs: [ ]
  // CHECK-NEXT: }

  %0 = "tfl.average_pool_2d"(%arg0) {filter_height = 3 : i32, filter_width = 6 : i32, fused_activation_function = "NONE", padding = "VALID", stride_h = 3 : i32, stride_w = 1 : i32} : (tensor<1x6x6x16xf32>) -> tensor<1x1x1x16xf32> loc("avgpool")
  func.return %0 : tensor<1x1x1x16xf32>

  func.func @convert_cluster_func(%arg0: tensor<i32>) -> () {
    %2 = "tf_device.parallel_execute"() ({

      %3 = "tf_device.cluster_func"(%arg0) {device = "/job:localhost/replica:0/task:0/device:GPU:0", func = @func} : (tensor<i32>) -> tensor<i32>

      tf_device.return %3 : tensor<i32>

    }) : () -> tensor<i32>
    return
  }
  func.func @func(%arg0: tensor<i32>) -> tensor<i32> {
    return %arg0 : tensor<i32>
  }
}
)";

    ```mlir
      func @send_to_host(%arg0: tensor<i32>) {
        "tf.XlaSendToHost"(%arg0) {key = "send_key"} : (tensor<i32>) -> ()
        return
      }
    ```

    This might be legalized like this:

    ```mlir
      func @send_to_host(%arg0: tensor<i32>) {
        %0 = "mhlo.create_token"() : () -> !mhlo.token
        %1 = "mhlo.send"(%arg0, %0) {...}} : (tensor<i32>, !mhlo.token) -> !mhlo.token
        return
      }

    value {
      shape {
        unknown_rank: true
      }
    }
  }
}
versions {
  producer: 216
}

# CHECK-NO-SHAPE: func @main(%arg0: tensor<i32>, %arg1: tensor<!tf_type.resource<tensor<i32>>>) -> (tensor<i32>, tensor<!tf_type.resource<tensor<i32>>>)

// This file test the selective quantiation feature in TF Quantizer. In the test
// config, the op named "test_opt_out" will not be quantized.

module attributes {tf.versions = {bad_consumers = [], min_consumer = 12 : i32, producer = 1269 : i32}} {
  func.func @conv2d_unmatching_unit(%arg0: tensor<1x3x4x3xf32>) -> (tensor<1x3x2x2xf32>) {
    %cst = "tf.Const"() {device = "", value = dense<1.0> : tensor<2x3x3x2xf32>} : () -> tensor<2x3x3x2xf32>