// merging same input values if they're identical. These optimizations are
// desirable but not for those TFLite ops which have variable tensors as inputs.
// Yes, they have identical input values, but those identical values are
// "stateful", their values can change during invocations.
//
// A typical example is unidirectional_sequence_lstm have two variable tensor

    def "configuration inputs are tracked during task graph serialization"() {
        given:
        def configurationCache = newConfigurationCacheFixture()

        and: "a project that has a provider having undeclared configuration inputs, which is linked to a task"
        buildFile("""
            abstract class MyTask extends DefaultTask {
                @Input

    if (op.getAdjX()) std::swap(equation[3], equation[4]);
    if (op.getAdjY()) std::swap(equation[6 + 3], equation[6 + 4]);

    rewriter.replaceOpWithNewOp<TF::EinsumOp>(
        op, op.getType(),
        /*inputs=*/ValueRange({input_lhs, input_rhs}),
        /*equation=*/equation);

    return success();
  }
};

#define GEN_PASS_DEF_BATCHMATMULTOEINSUMPASS

    %handle_0 = "tf.VarHandleOp"() {container="c", shared_name="a"} : () -> tensor<!tf_type.resource<tensor<1x10xf32>>>
    %handle_1 = "tf.VarHandleOp"() {container="c", shared_name="b"} : () -> tensor<!tf_type.resource<tensor<1x10xi64>>>
    %0 = "tf.ReadVariableOp"(%handle_0) : (tensor<!tf_type.resource<tensor<1x10xf32>>>) -> tensor<1x10xf32>

// RUN: tf-tfrt-opt -tf-to-tfrt %s | FileCheck %s --dump-input=fail --dump-input-filter=all
// RUN: tf-tfrt-opt -pass-pipeline='builtin.module(tf-to-tfrt{target-tpurt=true tpu-use-core-selector=false})' %s | FileCheck %s --dump-input=fail --dump-input-filter=all

// CHECK-LABEL: func @_tfrt_fallback_init
// CHECK-SAME: {{.*}} !tfrt.chain
// CHECK: tfrt_fallback_async.createop(%arg0) key(0) device("/device:CPU:0") "tf.ParseExampleV2"()

  std::vector<tensorflow::Tensor> inputs;

  // Adds inputs to the TF operation.
  for (const ElementsAttr& operand : operands) {
    tensorflow::Tensor tensor;
    RETURN_FAILURE_IF_ERROR(tensorflow::ConvertToTensor(operand, &tensor));
    inputs.push_back(std::move(tensor));
  }

  std::vector<tensorflow::TensorValue> input_values;
  for (tensorflow::Tensor& tensor : inputs) {
    input_values.emplace_back();

#define GEN_PASS_DEF_ANNOTATEPARAMETERREPLICATIONPASS
#include "tensorflow/compiler/mlir/tensorflow/transforms/tf_device_passes.h.inc"

// Analyzes the inputs to ClusterFuncOps in the module, and annotates their
// invoked functions whether each input has the same data across replicas.
struct AnnotateParameterReplicationPass
    : public impl::AnnotateParameterReplicationPassBase<
          AnnotateParameterReplicationPass> {

      }))
    return failure();

  SmallVector<ElementsAttr, 4> inputs;
  inputs.reserve(operands.size());
  for (auto input : operands) {
    inputs.push_back(mlir::cast<ElementsAttr>(input));
  }

  SmallVector<Attribute> constants;
  LogicalResult status = EvaluateOperation(inst, inputs, constants);
  results.assign(constants.begin(), constants.end());
  return status;
}

        val gradleUserHomeDir: File,
        val jvm: String,
        val startParameterProperties: Map<String, Any?>,
        /**
         * Whether the undeclared inputs accessed while serializing the task graph will be
         * excluded from input tracking. This is a temporary opt-out flag after a change
         * was made in that behavior.
         */
        val ignoreInputsInConfigurationCacheTaskGraphWriting: Boolean,

)

var Decode cbor.DecMode = func() cbor.DecMode {
	decode, err := cbor.DecOptions{
		// Maps with duplicate keys are well-formed but invalid according to the CBOR spec
		// and never acceptable. Unlike the JSON serializer, inputs containing duplicate map
		// keys are rejected outright and not surfaced as a strict decoding error.
		DupMapKey: cbor.DupMapKeyEnforcedAPF,

		// For JSON parity, decoding an RFC3339 string into time.Time needs to be accepted