Cleanup(Cleanup<G>&& src)  // NOLINT
      : released_(src.is_released()), f_(src.release()) {}

  // Assignment to a Cleanup object behaves like destroying it
  // and making a new one in its place, analogous to unique_ptr
  // semantics.
  Cleanup& operator=(Cleanup&& src) {  // NOLINT
    if (!released_) f_();
    released_ = src.released_;
    f_ = src.release();
    return *this;
  }

  ~Cleanup() {

developers to experiment with novel optimizations and training algorithms. TensorFlow supports a variety of applications, with a focus on training and inference on deep neural networks. Several Google services use TensorFlow in production, we have released it as an open-source project, and it has become widely used for machine learning research. In this paper, we describe the TensorFlow dataflow model and demonstrate the compelling performance that TensorFlow achieves for several real-world applications....

    if (forward_input) {
      // Since the input was forwarded, we released the input handle right away
      // and hence expect the input to be forwarded to the return tensor.
      EXPECT_EQ(orig_ptr, TF_TensorData(t));
    } else {
      // In async mode we expect forwarding to work without releasing the input
      // handle since by the time the kernel is executed we have released the
      // handle in the client code.

  return TensorHandlePtr(TFE_NewCustomDeviceTensorHandle(
      context, parallel_device_name.c_str(), t_released->dtype(), t_released,
      handle_methods, status));
}

// For TFE_CustomDevice::copy_tensor_to_device in the parallel device
// registration.
//
// Since this function is used to satisfy the TFE_CustomDevice C API,

)

compile_pip_requirements_3_9(
    name = "requirements_3_9_release",
    extra_args = [
        "--allow-unsafe",
        "-P keras-nightly",
        "-P tb-nightly",
    ],
    requirements_in = "requirements.in",
    requirements_txt = "requirements_lock_3_9.txt",
)

compile_pip_requirements_3_10(
    name = "requirements_3_10_release",
    extra_args = [
        "--allow-unsafe",
        "-P keras-nightly",

# limitations under the License.
# ==============================================================================
source ci/official/envs/linux_x86
TFCI_BAZEL_COMMON_ARGS="--repo_env=TF_PYTHON_VERSION=$TFCI_PYTHON_VERSION --config release_gpu_linux"
TFCI_BAZEL_TARGET_SELECTING_CONFIG_PREFIX=linux_cuda
TFCI_BUILD_PIP_PACKAGE_ARGS="--repo_env=WHEEL_NAME=tensorflow"
TFCI_DOCKER_ARGS="--gpus all"
TFCI_LIB_SUFFIX="-gpu-linux-x86_64"

// Wraps a pointer to a context implementation.
//
// WARNING: Since the underlying object could be ref-counted a user of this
// interface cannot destruct the underlying context object. Instead, call
// TFE_DeleteContext who calls Release() on the context pointer and deletes
// the TFE_Context structure.
typedef struct TFE_Context TFE_Context;

namespace tensorflow {

DEFINE_CONVERSION_FUNCTIONS(tensorflow::ImmediateExecutionContext, TFE_Context);

      parent_op_(parent_op),
      tape_(tape),
      registry_(registry) {
  // TODO(b/172003047): Consider making AbstractOperation RefCounted.
  // parent_op_->Ref();
}
void TapeOperation::Release() {
  // TODO(srbs): Change to Unref().
  delete this;
}
TapeOperation::~TapeOperation() {
  // TODO(b/172003047): Consider making AbstractOperation RefCounted.
  // parent_op->Unref();
}

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

  TFE_DeleteContextOptions(opts);
  TFE_DeleteContext(ctx);
  TF_DeleteStatus(status);

  w0.release();
  w1.release();
}

TEST(CAPI, ShareVariableAcrossContextsAfterUpdateContextWorksWithTimeout) {
  tensorflow::ServerDef server_def_0 = GetServerDef(3);

1.  It must be a bug/performance issue or a feature request or a build issue or
    a documentation issue (for small doc fixes please send a PR instead).
1.  Make sure the Issue Template is filled out.
1.  The issue should be related to the repo it is created in.

**Here's why we have this policy:** We want to focus on the work that benefits
the whole community, e.g., fixing bugs and adding features. Individual support