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

  const char* tensor_data[] = {&tensor1[0], &tensor2[0]};
  const size_t tensor_size[] = {TF_ARRAYSIZE(tensor1), TF_ARRAYSIZE(tensor2)};
  const int64_t* tensor_dims[] = {&dims1[0], &dims2[0]};
  const size_t tensor_ndims[] = {ndims1, ndims2};
  for (int i = 0; i < 2; ++i) {
    TF_Tensor* v = values[i];

      });
}
bool Tape::ShouldRecord(
    absl::Span<const AbstractTensorHandle* const> tensors) const {
  std::vector<int64_t> tensor_ids(tensors.size());
  std::vector<tensorflow::DataType> tensor_dtypes(tensors.size());
  for (int i = 0; i < tensors.size(); i++) {
    tensor_ids[i] = ToId(tensors[i]);
    tensor_dtypes[i] = tensors[i]->DataType();
  }
  return GradientTape::ShouldRecord(tensor_ids, tensor_dtypes);
}

                       const string& op_name = "");
  // Returns whether any tensor in a list of tensors is being watched and has
  // a trainable dtype.
  bool ShouldRecord(
      absl::Span<const AbstractTensorHandle* const> tensors) const;
  // Unwatches this tensor on the tape. Mainly used for cleanup when deleting
  // eager tensors.
  void DeleteTrace(const AbstractTensorHandle*);

    // Input tensors
    const std::vector<std::pair<string, Tensor>>& input_pairs,
    // Output tensors
    const std::vector<string>& output_tensor_names, TF_Tensor** c_outputs,
    // Target nodes
    const std::vector<string>& target_oper_names, TF_Buffer* run_metadata,
    TF_Status* status) {
  const int noutputs = output_tensor_names.size();
  std::vector<Tensor> outputs(noutputs);
  Status result;

                                             TF_Status* status,
                                             void* device_info);

  // Method to copy a tensor from the custom device to a target device.
  TFE_TensorHandle* (*copy_tensor_from_device)(TFE_Context* context,
                                               TFE_TensorHandle* tensor,

TF_AbstractTensor* TF_OutputListGet(TF_OutputList* o, int i);
// Append a tensor at the end of the output list, growing its size by one.
void TF_OutputListPushBack(TF_OutputList* o, TF_AbstractTensor* tensor,
                           TF_Status*);

// TF_ExecuteOperation will, if in eager mode, execute, if in graph mode, maybe
// capture some inputs and then add a node in the graph. The output tensors are
// returned through the provided TF_OutputList.

  TFE_DeleteContextOptions(opts);

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

  /* Want to test simple MatMul example with abstract tensors:
    [[1,2],   *  [[5,6],   =   [[19,22],
     [3,4]]       [7,8]]        [43,50]]
  */

  // Build 1st Matrix.
  int64_t dims[] = {2, 2};  // Matrices will be 2 x 2

TF_Tensor* BoolTensor(int32_t v);

// Create a tensor with values of type TF_INT8 provided by `values`.
TF_Tensor* Int8Tensor(const int64_t* dims, int num_dims, const char* values);

// Create a tensor with values of type TF_INT32 provided by `values`.
TF_Tensor* Int32Tensor(const int64_t* dims, int num_dims,
                       const int32_t* values);

// Create 1 dimensional tensor with values from `values`

  }
};

// An abstract operation describes an operation by its type, name, and
// attributes. It can be "executed" by the context with some input tensors.
// It is allowed to reusing the same abstract operation for multiple execution
// on a given context, with the same or different input tensors.
class TracingOperation : public AbstractOperation {
 protected:
  explicit TracingOperation(AbstractOperationKind kind)

        "TFE_TensorHandleDeviceMemorySize may not be called on a ",
        handle->TypeString(), " tensor handle.");
    return 0;
  }
  const tensorflow::Tensor* tensor;
  status->status = handle->Tensor(&tensor);
  if (!status->status.ok()) {
    return 0;
  }
  return tensor->TotalBytes();
}

TFE_Op* TFE_NewOp(TFE_Context* ctx, const char* op_or_function_name,