"tfl.custom"(%0, %2) {custom_code = "FlexAssignVariableOp", custom_option = #tfl<const_bytes : "0x1041737369676E5661726961626C654F70003B121041737369676E5661726961626C654F701A001A002A0B0A056474797065120230012A140A0E76616C69646174655F736861706512022800320000024F3E1414042801">} : (tensor<!tf_type.resource<tensor<1x10xf32>>>, tensor<1x10xf32>) -> ()

// -----

// CHECK-LABEL: float_matmul
func.func @float_matmul(
  %arg0: tensor<1x10xf32>, %arg1: tensor<10x10xf32>) -> (tensor<*xf32>, tensor<*xf32>, tensor<*xf32>) {
  %cst = "tf.Const"() {value = dense<0.000000e+00> : tensor<10xf32>} : () -> tensor<10xf32>
  %0 = "tf.MatMul"(%arg0, %arg1) {
    transpose_a = false, transpose_b = false
  } : (tensor<1x10xf32>, tensor<10x10xf32>) -> tensor<*xf32>

// NOFALLBACK-LABEL: @xla_svd
func.func @xla_svd(%arg0: tensor<1x1xf32>) -> (tensor<1xf32>, tensor<1x1xf32>, tensor<1x1xf32>) {
  // NOFALLBACK: XlaSvd
  %s, %u, %v = "tf.XlaSvd"(%arg0) {max_iter = 1, epsilon = 1.0E-09 : f32, precision_config = ""} : (tensor<1x1xf32>) -> (tensor<1xf32>, tensor<1x1xf32>, tensor<1x1xf32>)
  func.return %s, %u, %v : tensor<1xf32>, tensor<1x1xf32>, tensor<1x1xf32>
}

      %3 = stablehlo.reshape %2 : (tensor<1x3xf32>) -> tensor<3x1xf32>
      %4 = "quantfork.qcast"(%3) {volatile} : (tensor<3x1xf32>) -> tensor<3x1x!quant.uniform<i8:f32, 0.13170163023705575:-1>>
      %5 = "quantfork.dcast"(%4) : (tensor<3x1x!quant.uniform<i8:f32, 0.13170163023705575:-1>>) -> tensor<3x1xf32>
      return %5 : tensor<3x1xf32>
    }
  }
)mlir";

  %3 = "tfl.pack"(%1, %2) {axis = 0 : i32, per_device_costs = {CPU = 2.0 : f32, GPU = -1.0 : f32}, values_count = 2 : i32, tac.device = "CPU"} : (tensor<1xf32>, tensor<1xf32>) -> tensor<2x1xf32>
  func.return %3 : tensor<2x1xf32>
})";
  const std::string kExpectedFB = CreateRuntimeMetadata();
  mlir::DialectRegistry registry;
  registry.insert<mlir::TFL::TensorFlowLiteDialect, mlir::arith::ArithDialect,

  // CHECK: return %[[v1]] : tensor<2x?x?x1xf32>
}

func.func @einsum_broadcast(%arg0: tensor<3x4x5xf32>, %arg1: tensor<5x6xf32>) -> tensor<3x4x6xf32> {
  %0 = "tf.Einsum"(%arg0, %arg1) {T = "tfdtype$DT_FLOAT", equation = "ijk,km->ijm"}: (tensor<3x4x5xf32>, tensor<5x6xf32>) -> tensor<3x4x6xf32>
  func.return %0 : tensor<3x4x6xf32>

    %1:2 = "tf.IteratorGetNext"(%arg4) {device = "/job:localhost/replica:0/task:0/device:CPU:0"} : (tensor<*x!tf_type.resource>) -> (tensor<2x224x224x3xf32>, tensor<2x1xf32>)

  "tf.SomeOp2"() {device = "/device:CPU:0", _output_shapes = ["tfshape$"], f.Tin = [f32], f._read_only_resource_inputs = []} : () -> ()
  func.return
}

// CHECK-LABEL: func @basic
func.func @basic(
    %arg0: tensor<3x1xf32>,
    %arg1: tensor<!tf_type.resource<tensor<1x3xf32>>>) -> (tensor<3x3xf32>) {

    func.return %1#0 : tensor<i32>
  }
}

// -----

// CHECK: module attributes {tfl._legalize_tfl_variables = true}
module {
  func.func @main(%arg0 : tensor<!tf_type.resource<tensor<4096xf32>>>,
      %arg1 : tensor<*x!tf_type.variant>) {
    %cst_0 = arith.constant dense<2> : tensor<i64>
    %cst_1 = arith.constant dense<0> : tensor<i32>
    %0 = "tf.RepeatDataset"(%arg1, %cst_0) {device = "",

        unknown_rank: true
      }
    }
  }
  attr {
    key: "dtype"
    value {
      type: DT_BOOL
    }
  }
}

# CHECK-DAG: tf.IteratorGetNext{{.+}}-> (tensor<1x8xbf16>, tensor<2x?x16xf32>, tensor<*xf64>)
# CHECK-DAG: tf.IteratorGetNextSync{{.+}}-> (tensor<*xi16>, tensor<3x24xi32>, tensor<?x4x32xi64>)