for machine learning. It has a comprehensive, flexible ecosystem of
[tools](https://www.tensorflow.org/resources/tools),
[libraries](https://www.tensorflow.org/resources/libraries-extensions), and
[community](https://www.tensorflow.org/community) resources that lets
researchers push the state-of-the-art in ML and developers easily build and
deploy ML-powered applications.

                                                        TF_Status* status);

// Some TF ops need a step container to be set to limit the lifetime of some
// resources (mostly TensorArray and Stack, used in while loop gradients in
// graph mode). Calling this on a context tells it to start a step.
TF_CAPI_EXPORT extern void TFE_ContextStartStep(TFE_Context* ctx);

  std::unordered_map<int64_t, TapeTensor> sources_that_are_targets;
  for (int i = 0; i < target_tensor_ids.size(); ++i) {
    int64_t target_id = target_tensor_ids[i];
    if (sources_set.find(target_id) != sources_set.end()) {
      auto tensor = targets[i];
      sources_that_are_targets.insert(
          std::make_pair(target_id, TapeTensor(tensor)));
    }
  }

        "//tensorflow/core:test_main",
        "//tensorflow/core/kernels:array",
        "//tensorflow/core/kernels:control_flow_ops",
        "//tensorflow/core/kernels:math",
        "//tensorflow/core/platform:resource_loader",
    ],
)

tf_cc_test(
    name = "c_api_experimental_test",
    size = "medium",
    srcs = ["c_api_experimental_test.cc"],
    data = [
        "testdata/tf_record",

                        : const_cast<TF_Tensor**>(input_tensors.data()),
                    /*input_tensors_as_shapes*/ nullptr,
                    /*input_resource_shapes_and_types*/ nullptr, &output_shapes,
                    /*output_resource_shapes_and_types*/ nullptr, status_);
    CHECK_EQ(TF_OK, TF_GetCode(status_)) << TF_Message(status_);
    CHECK_EQ(output_shapes->num_items, 1);

  virtual void ClearCachesAndThreadExecutors() = 0;

  // Initialize the step resource container for a training step. This is used
  // in current TF runtime. For tfrt, it is used by fallback op handler.
  virtual void StartStep() = 0;
  // Destroy the step resource container for a training step.
  virtual void EndStep() = 0;

  std::unordered_map<int64_t, TapeTensor> sources_that_are_targets;
  for (int grad_index = 0, end = grad.size(); grad_index < end; ++grad_index) {
    Gradient* grad_tensor = grad[grad_index];
    if (grad_tensor != nullptr) {
      int64_t tensor_id = vspace_.TensorId(grad_tensor);
      targets.push_back(tensor_id);
      if (sources_set.find(tensor_id) != sources_set.end()) {
        sources_that_are_targets.emplace(

// `output_resource_shapes_and_types`. The caller is responsible for freeing the
// memory in these buffers by calling `TF_DeleteShapeAndTypeList`.
TF_CAPI_EXPORT extern void TFE_InferShapes(
    TFE_Op* op, TF_ShapeAndTypeList* input_shapes, TF_Tensor** input_tensors,
    TF_ShapeAndTypeList* input_tensor_as_shapes,
    TF_ShapeAndTypeList** input_resource_shapes_and_types,

                     TF_ShapeAndTypeList* input_tensors_as_shapes,
                     TF_ShapeAndTypeList** input_resource_shapes_and_types,
                     TF_ShapeAndTypeList** output_shapes,
                     TF_ShapeAndTypeList*** output_resource_shapes_and_types,
                     TF_Status* status) {
  using tensorflow::NodeDef;
  using tensorflow::OpRegistrationData;

  ASSERT_EQ(TF_OK, TF_GetCode(status)) << TF_Message(status);

  // The device of var_handle0 is local device which is the same before and
  // after cluster update. Remove resource with valid device should succeed.
  TFE_Op* op = TFE_NewOp(ctx, "DestroyResourceOp", status);
  ASSERT_EQ(TF_OK, TF_GetCode(status)) << TF_Message(status);
  TFE_OpAddInput(op, var_handle0, status);