values = Reshape(scope, values, flat_values_shape);
  }

  indices = Reshape(scope, indices, indices_size);
  Output params_grad =
      UnsortedSegmentSum(scope, values, indices, gather_dim_size);

  if (batch_dims != 0) {
    // Put back the batch dimensions.
    params_grad = Reshape(scope, params_grad, params_shape);
  }
  return params_grad;
}

    "BUCKETIZE",
    "RANDOM_UNIFORM",
    "MULTINOMIAL",
    "GELU",
    "DYNAMIC_UPDATE_SLICE",
    "RELU_0_TO_1",
    "UNSORTED_SEGMENT_PROD",
    "UNSORTED_SEGMENT_MAX",
    "UNSORTED_SEGMENT_SUM",
    "ATAN2",
    "UNSORTED_SEGMENT_MIN",
    "SIGN",
    "BITCAST",
    "BITWISE_XOR",
    "RIGHT_SHIFT",
    "STABLEHLO_LOGISTIC",
    "STABLEHLO_ADD",
    "STABLEHLO_DIVIDE",

      "UniformRequantize",
      "Unique",
      "UniqueV2",
      "UpperBound",
      "UnsortedSegmentMax",
      "UnsortedSegmentMin",
      "UnsortedSegmentProd",
      "UnsortedSegmentSum",
      "VarIsInitializedOp",
      "VariableShape",
      "Where",
      "While",
      "XlaAllReduce",
      "XlaBroadcastHelper",
      "XlaCallModule",
      "XlaConcatND",

func.func @unsortedSegmentReduction(%data: tensor<7x10x8xf32>, %segment_ids: tensor<7x10x8x1xi32>, %num_segments: tensor<i32>) {
  // expected-error @+1 {{requires segment ids rank to be less than or equal to data's rank}}
  %0 = "tf.UnsortedSegmentSum"(%data, %segment_ids, %num_segments) : (tensor<7x10x8xf32>, tensor<7x10x8x1xi32>, tensor<i32>) -> (tensor<?x8xf32>)
  func.return
}

// -----