}

// -----

// Tests that the output of the tf.XlaCallModule op has been replaced by
// a quantized type, and the corresponding quantfork.qcast ops that turned
// the float output to a quantized type is removed.

// CHECK-LABEL: quantize_simple_xla_call_module_no_operand
func.func private @quantize_simple_xla_call_module_no_operand() -> tensor<1x3xf32> {

  %2 = "tf.XlaCallModule"(%1#0, %0) <{Sout = [#tf_type.shape<1x3>], dim_args_spec = [], disabled_checks = [], has_token_input_output = false, module = "", platforms = [], version = 5 : i64}> {_entry_function = @composite_dot_general_fn_1, _original_entry_function = "c...

      %0 = "quantfork.stats"(%arg0) {layerStats = dense<[6.00000000e-6, 9.00000000e-1]> : tensor<2xf32>} : (tensor<1x2xf32>) -> tensor<1x2xf32>

// CHECK: %[[XLA_CALL_MODULE:.+]] = "tf.XlaCallModule"(%[[CUSTOM_AGGREGATOR_0]], %[[CST]])
// CHECK-SAME: _entry_function = @composite_dot_general_fn_1, _original_entry_function = "composite_dot_general_fn_1"

  func.func private @quantize_dot_general_per_tensor(%arg0: tensor<1x2xf32>) -> tensor<1x3xf32> attributes {tf._original_func_name = "main_0"} {
    %0 = stablehlo.constant dense<3.000000e-01> : tensor<2x3xf32>

    %0 = "quantfork.qcast"(%cst) : (tensor<2x3xf32>) -> tensor<2x3x!quant.uniform<i8:f32, 6.000000e-03>>
    %1 = "quantfork.dcast"(%0) : (tensor<2x3x!quant.uniform<i8:f32, 6.000000e-03>>) -> tensor<2x3xf32>

SmallVector<Value> AppendToVector(ArrayRef<Value> arguments, Value append);

// Checks if the `Method` attatched to the given `tf.XlaCallModule` op has
// `WeightOnlyPtq`.
bool HasWeightOnlyPtqMethod(TF::XlaCallModuleOp xla_call_module_op);

// Checks if an op is a `tf.XlaCallModule` op, contains 'conv' or 'dot_general'
// in its name and has `Method` with `WeightOnlyPtq`.
bool IsWeightOnlyQuantizableOp(const Operation& op);

      %cst_1 = arith.constant dense<2.0> : tensor<3xf32>
      %0 = "tf.XlaCallModule"(%arg0, %cst_0, %cst_1) <{Sout = [#tf_type.shape<1x4x4x3>], module = "", version = 9 : i64}> {_entry_function = @composite_fn_1, _original_entry_function = "composite_fn_1", _tfl_quant_trait = "fully_quantizable"} : (tensor<1x4x4x3xf32>, tensor<3x1x1x3xf32>, tensor<3xf32>) -> tensor<1x4x4x3xf32>

    %0 = "tf.XlaCallModule"(%output, %cst) <{Sout = [#tf_type.shape<10x1x3>], dim_args_spec = [], disabled_checks = [], function_list = [], has_token_input_output = false, module = "", platforms = ["CPU"], version = 9 : i64}> {_entry_function = @composite_dot_general_with_relu_fn_1,...