from the `envs` directory that is filled with configuration options:

-   Nightly jobs (Run nightly on the `nightly` branch)
    -   Uses `wheel.sh`, `libtensorflow.sh`, `code_check_full.sh`
-   Continuous jobs (Run on every GitHub commit)
    -   Uses `pycpp.sh`
-   Presubmit jobs (Run on every GitHub PR)
    -   Uses `pycpp.sh`, `code_check_changed_files.sh`

// is for performance optimization by reusing an exiting unused op rather than
// creating a new op every time. If `raw_device_name` is `NULL` or empty, it
// does not set the device name. If it's not `NULL`, then it attempts to parse
// and set the device name. It's effectively `TFE_OpSetDevice`, but it is faster
// than separately calling it because if the existing op has the same

    gtl::ArraySlice<AbstractTensorHandle*> output_gradients,
    absl::Span<AbstractTensorHandle*> result) const {
  if (gradient_function == nullptr) {
    return errors::InvalidArgument(
        "Provided null gradient_function for '", op_type, "'.\n",
        "If the intent is to treat this op as non-differentiable consider "
        "using RegisterNotDifferentiable or "
        "NotDifferentiableGradientFunction.");
  }

if [[ "$TFCI_DOCKER_PULL_ENABLE" == 1 ]]; then
  # Simple retry logic for docker-pull errors. Sleeps for 15s if a pull fails.
  # Pulling an already-pulled container image will finish instantly, so
  # repeating the command costs nothing.
  docker pull "$TFCI_DOCKER_IMAGE" || sleep 15
  docker pull "$TFCI_DOCKER_IMAGE" || sleep 15
  docker pull "$TFCI_DOCKER_IMAGE"
fi 

    *   `tf.data.Dataset.zip` now supports Python-style zipping, i.e. `Dataset.zip(a, b, c)`.
    * `tf.data.Dataset.shuffle` now supports `tf.data.UNKNOWN_CARDINALITY` When doing a "full shuffle" using  `dataset = dataset.shuffle(dataset.cardinality())`. But remember, a "full shuffle" will load the full dataset into memory so that it can be shuffled, so make sure to only use this with small datasets or datasets of small objects (like filenames).

*   `tf.math`

    core::RefCountPtr<AbstractFunction> scoped_func;
    // Returning null tensors from a tf.function is not supported, so we keep
    // track of indices in the model's outputs are nullptr in this set.
    // The FunctionDef only outputs the non-null tensors. We later pad the
    // function op outputs to have nullptrs at the `null_indices`.
    absl::flat_hash_set<int> null_indices;
    {
      AbstractContextPtr func_ctx(BuildFunction(fn_name));

  # The TFCI Mac VM image seems to have uncommitted local changes to the Pyenv
  # repository so we have to discard them and reset the working directory before
  # we can pull in the latest changes.
  cd /Users/kbuilder/.pyenv/ && git reset --hard HEAD && git pull && cd -
fi

# "TFCI_MACOS_PYENV_INSTALL_ENABLE" controls whether to use Pyenv to install
# the Python version set in "TFCI_PYTHON_VERSION" and use it as default.

config_setting(
    name = "lite_protos_legacy",
    define_values = {"TENSORFLOW_PROTOS": "lite"},
    visibility = ["//visibility:public"],
)

config_setting(
    name = "full_protos",
    define_values = {"TENSORFLOW_PROTOS": "full"},
    visibility = ["//visibility:public"],
)

selects.config_setting_group(
    name = "lite_protos",
    match_any = [":lite_protos_legacy"],

#include "tensorflow/c/tf_datatype.h"
#include "tensorflow/c/tf_status.h"
#include "tensorflow/c/tf_status_helper.h"
#include "xla/tsl/c/tsl_status_internal.h"
#include "tensorflow/core/framework/full_type.pb.h"
#include "tensorflow/core/framework/shape_inference.h"
#include "tensorflow/core/framework/tensor_shape.h"
#include "tensorflow/core/framework/types.pb.h"
#include "tensorflow/core/lib/llvm_rtti/llvm_rtti.h"

    status->status =
        tensorflow::errors::InvalidArgument("Unparseable AttrValue proto");
    return;
  }
  if (op == nullptr) {
    status->status = tensorflow::errors::InvalidArgument(
        "Got a null or uninitialized `op` argument");
    return;
  }
  tensorflow::EagerOperation* operation =
      OperationFromInterface(tensorflow::unwrap(const_cast<TFE_Op*>(op)));