data, or by relaxing `Input` shape assumptions (note that you can pass
        shapes with `None` entries for axes that are meant to be dynamic). You
        can also disable the input checking entirely by setting
        `model.input_spec = None`.
    *   Several changes have been made to
        `tf.keras.mixed_precision.experimental`. Note that it is now recommended

      run: |
        function map() {
          # The "digest" that allows us to pull an image is not the digest as
          # returned by the API, but a sha256sum of the entire chunk of image
          # metadata. gcr.io helpfully includes it in the header of the response
          # as docker-content-digest: sha256:[digest]. Note we use egrep to
          # match exactly sha256:<hash> because curl may include a ^M symbol at

      // added since the last TF_SessionRun() call.
      for (auto id = session->last_num_graph_nodes; id < num_nodes; ++id) {
        Node* const node = graph.FindNodeId(id);
        if (node != nullptr && node->IsOp()) {
          NodeDef* const node_def = graph_def.add_node();
          *node_def = node->def();
        }
      }
      *graph_def.mutable_library() = graph.flib_def().ToProto();

-		  : memset (s, c, n)))
-#  endif
-# endif
-
-/* GCC < 3.0 optimizes memset(s, 0, n) but not bzero(s, n).
-   The optimization is broken before EGCS 1.1.
-   GCC 3.0+ has __builtin_bzero as well, but at least till GCC 3.4
-   if it decides to call the library function, it calls memset
-   and not bzero.  */
-# if __GNUC_PREREQ (2, 91)
-#  define __bzero(s, n) __builtin_memset (s, '\0', n)
-# endif
-

# See the License for the specific language governing permissions and
# limitations under the License.
# ==============================================================================
# Note: this file gets sourced in utilities/setup.sh, which has "set -u"
# (error on undefined variables). This ensures that (a) every TFCI variable
# has an explicit default value, and (b) no script can accidentally use a

  tfrun python3 tensorflow/tools/ci_build/update_version.py --nightly
fi
source ci/official/utilities/get_versions.sh

# Note on gsutil commands:
# "gsutil cp" always "copies into". It cannot act on the contents of a directory
# and it does not seem possible to e.g. copy "gs://foo/bar" as anything other than
# "/path/bar". This script uses "gsutil rsync" instead, which acts on directory

build:monolithic --define framework_shared_object=false
build:monolithic --define tsl_protobuf_header_only=false
build:monolithic --experimental_link_static_libraries_once=false  # b/229868128

# Please note that MKL on MacOS is still not supported.
# If you would like to use a local MKL instead of downloading, please set the
# environment variable "TF_MKL_ROOT" every time before build.
build:mkl --define=build_with_mkl=true --define=enable_mkl=true

// `shape_inference::InferenceContext` constructor. Note the following:
//   - The inputs of the `op` are not used for shape inference. So, it is
//     OK to not have the inputs properly set in `op`. See `input_tensors`
//     if you want shape inference to consider the input tensors of the
//     op for shape inference.
//   - The types need not be set in `input_shapes` as it is not used.

    kiwisolver macholib matplotlib nose numpy packaging pandas pickleshare pip \
    prompt-toolkit protobuf psutil pyasn1 pyasn1-modules pycparser Pygments \
    pyparsing pyreadline python-dateutil pytz pywin32 requests rsa setuptools \
    simplegeneric six Tempita traitlets uritemplate urllib3 virtualenv wcwidth \
    wheel win-unicode-console;

# Hardcoding Android license since I did not find any solution on accepting it

# with clang. It does not imply that CUDA support has been enabled.
alias(
    name = "is_cuda_compiler_clang",
    actual = if_oss(
        "@local_config_cuda//:is_cuda_compiler_clang",
        "@local_config_cuda//cuda:TRUE",
    ),
)

# Config setting that is satisfied when CUDA device code should be compiled
# with nvcc. It does not imply that CUDA support has been enabled.
alias(