%1 = "tf.XlaCallModule"(%0, %cst_0, %cst_1) <{Sout = [#tf_type.shape<1x4x4x3>], module = "", version = 9 : i64}> {_entry_function = @composite_fn_2, _original_entry_function = "composite_fn_2", _tfl_quant_trait = "fully_quantizable"} : (tensor<1x4x4x3xf32>, tensor<3x1x1x3xf32>, tensor<3xf32>) -> tensor<1x4x4x3xf32>
      return %1 : tensor<1x4x4x3xf32>

  func.return %17 : tensor<1x2x3xf32>

  // CHECK: %[[NONE:.*]] = "tfl.no_value"() <{value}> : () -> none
  // CHECK: %[[DQ_1:.*]] = "tfl.dequantize"({{.*}}) : (tensor<1x1xf16>) -> tensor<1x1xf32>
  // CHECK: %[[DQ_2:.*]] = "tfl.dequantize"({{.*}}) : (tensor<1x1xf16>) -> tensor<1x1xf32>
  // CHECK: %[[DQ_3:.*]] = "tfl.dequantize"({{.*}}) : (tensor<1x1xf16>) -> tensor<1x1xf32>

func.func @convert_dot_general_dynamic_batch_dim(%arg0: tensor<2x?x2x3xf32>, %arg1: tensor<2x?x4x3xf32>) -> tensor<2x?x2x4xf32> {
%0 = "mhlo.dot_general"(%arg0, %arg1) {
  dot_dimension_numbers = #mhlo.dot<
    lhs_batching_dimensions = [0, 1],
    rhs_batching_dimensions = [0, 1],
    lhs_contracting_dimensions = [3],
    rhs_contracting_dimensions = [3]
  >} : (tensor<2x?x2x3xf32>, tensor<2x?x4x3xf32>) -> tensor<2x?x2x4xf32>

func.func @batchmatmulv2_basic(%arg0: tensor<1x4x2xf32>, %arg1: tensor<3x2x4xf32>) -> tensor<3x4x4xf32> {
// CHECK-LABEL:   func @batchmatmulv2_basic
// CHECK-SAME:        ([[LHS:%.*]]: tensor<1x4x2xf32>, [[RHS:%.*]]: tensor<3x2x4xf32>) -> tensor<3x4x4xf32>
// CHECK:           [[LHSSHAPE:%.*]] = shape.shape_of [[LHS]] : tensor<1x4x2xf32>
// CHECK:           [[RHSSHAPE:%.*]] = shape.shape_of [[RHS]] : tensor<3x2x4xf32>

    // CHECK-NEXT: %[[CMP:.*]] = mhlo.compare GT, %[[INP]], %[[ZERO]], NOTYPE : (tensor<1x4x4x3xf32>, tensor<1x4x4x3xf32>) -> tensor<1x4x4x3xi1>
    // CHECK-NEXT: %[[RES:.*]] = mhlo.select %[[CMP]], %[[INP]], %[[LEAKY]] : tensor<1x4x4x3xi1>, tensor<1x4x4x3xf32>
    // CHECK-NEXT: return %[[RES]] : tensor<1x4x4x3xf32>
    %0 = "tf.LeakyRelu"(%arg0) {alpha = 2.000000e-01 : f32, device = ""} : (tensor<1x4x4x3xf32>) -> tensor<1x4x4x3xf32>

  constexpr absl::string_view
      kModuleXlaCallModuleOpWithDefaultQuantizationMethod = R"mlir(
    func.func @xla_call_module_default_quantization_method(%arg0: tensor<1x1x3xf32>, %arg1: tensor<3x4xf32>) -> tensor<1x1x4xf32> {

    %cst_1 = "tf.Const"() <{value = dense<[[-0.795477629, 0.581315517, 0.921566545], [0.138622552, 0.463866323, 0.95474267], [-0.143770888, -0.796835303, 0.899996876], [0.0989735424, -0.483384758, -7.277030e-01]]> : tensor<4x3xf32>}> {device = ""} : () -> tensor<4x3xf32>

    %0 = stablehlo.constant dense<1.000000e+00> : tensor<4x3xf32>
    %1 = "quantfork.qcast"(%0) {volatile} : (tensor<4x3xf32>) -> tensor<4x3x!quant.uniform<i8<-127:127>:f32:1, {5.000000e-03, 5.000000e-03, 5.000000e-03}>>
    %2 = "quantfork.dcast"(%1) : (tensor<4x3x!quant.uniform<i8<-127:127>:f32:1, {5.000000e-03, 5.000000e-03, 5.000000e-03}>>) -> tensor<4x3xf32>

  "tf_device.cluster_func"(%1) {func = @cluster_func, use_spmd_for_xla_partitioning = true, num_cores_per_replica = 8 : i64} : (tensor<4x4x4xf32>) -> ()
  func.return
}

func.func @cluster_func(%arg0: tensor<4x4x4xf32>) {
  func.return
}

// -----

// Tests partitioned inputs/outputs with no sharding (via XLA SPMD) defaults to
// replicate sharding ("").

// CHECK-LABEL: func @partitioned_input_output

  %0 = stablehlo.constant dense<1.0> : tensor<4x3xf32>
  %1 = stablehlo.dot_general %arg0, %0, contracting_dims = [1] x [0], precision = [DEFAULT, DEFAULT] : (tensor<1x4xf32>, tensor<4x3xf32>) -> tensor<1x3xf32>
  return %1 : tensor<1x3xf32>
}
// CHECK: @main(%[[ARG_0:.+]]: tensor<1x4xf32>) -> tensor<1x3xf32>
// CHECK-DAG: %[[CST:.+]] = stablehlo.constant dense<1.000000e+00> : tensor<4x3xf32>