ways to detect malicious models/graphs/checkpoints, so the recommended way to
mitigate the risk in this scenario is to sandbox the model execution.

### Hardware attacks

Physical GPUs or TPUs can also be the target of attacks. [Published
research](https://scholar.google.com/scholar?q=gpu+side+channel) shows that it
might be possible to use side channel attacks on the GPU to leak data from other

        "cpu": "mips64",
    },
    visibility = ["//visibility:public"],
)

# TODO(jakeharmon8): Remove in favor of TSL version
config_setting(
    name = "windows",
    # Internal builds query the target OS.
    constraint_values = if_google(
        ["//third_party/bazel_platforms/os:windows"],
        [],
    ),
    # OSS builds query the CPU type.
    values = if_oss(
        {"cpu": "x64_windows"},

      ImmediateExecutionTensorHandle* handle,
      const tensorflow::string& target_device_name,
      ImmediateExecutionTensorHandle** result) override {
    TF_Status status;
    handle->Ref();
    TFE_TensorHandle* result_handle = device_.copy_tensor_from_device(
        context_, tensorflow::wrap(handle), target_device_name.c_str(), &status,
        info_);
    handle->Unref();

*   <IF RELEASE CONTAINS MULTIPLE FEATURES FROM SAME AREA, GROUP THEM TOGETHER>

*   Add `is_cpu_target_available`, which indicates whether or not TensorFlow was
    built with support for a given CPU target. This can be useful for skipping
    target-specific tests if a target is not supported.

*   `tf.data`
    * Support `data.experimental.distribued_save`. `distribued_save` uses
      tf.data service

  | sort -u
}

# Verify that, given a build target and a license-list generator target, all of
# the dependencies of that target which include a license notice file are then
# included when generating that license. Necessary because the license targets
# in TensorFlow are manually enumerated rather than generated automatically.
do_external_licenses_check(){
  BUILD_TARGET="$1"
  LICENSES_TARGET="$2"

  in_grads.reserve(input_tensors.size());
  for (int target_index = 0; target_index < input_tensors.size();
       ++target_index) {
    const auto current_grad =
        accumulated_gradients_.find(input_tensors[target_index].GetID());
    if (current_grad == accumulated_gradients_.end()) {
      if (IsDtypeTrainable(input_tensors[target_index].GetDType())) {

  // If target is a directory return TF_FAILED_PRECONDITION.
  // Target might be missing, so don't error in that case.
  struct stat st;
  if (stat(dst, &st) != 0) {
    if (errno != ENOENT) {
      TF_SetStatusFromIOError(status, errno, dst);
      return;
    }
  } else if (S_ISDIR(st.st_mode)) {
    TF_SetStatus(status, TF_FAILED_PRECONDITION, "target path is a directory");
    return;
  }

source ci/official/envs/linux_x86
TFCI_BAZEL_COMMON_ARGS="--repo_env=TF_PYTHON_VERSION=$TFCI_PYTHON_VERSION --config release_cpu_linux --config=tpu"
TFCI_BAZEL_TARGET_SELECTING_CONFIG_PREFIX=linux_tpu
TFCI_BUILD_PIP_PACKAGE_ARGS="--repo_env=WHEEL_NAME=tensorflow_tpu"
TFCI_LIB_SUFFIX="-tpu-linux-x86_64"
TFCI_WHL_BAZEL_TEST_ENABLE=0
TFCI_WHL_IMPORT_TEST_ENABLE=0
TFCI_WHL_SIZE_LIMIT=580M

# limitations under the License.
# ==============================================================================
TFCI_BAZEL_COMMON_ARGS="--repo_env=TF_PYTHON_VERSION=$TFCI_PYTHON_VERSION --config release_macos_arm64"
TFCI_BAZEL_TARGET_SELECTING_CONFIG_PREFIX=macos_arm64
TFCI_BUILD_PIP_PACKAGE_ARGS="--repo_env=WHEEL_NAME=tensorflow"
TFCI_INDEX_HTML_ENABLE=1
TFCI_LIB_SUFFIX="-cpu-darwin-arm64"
TFCI_MACOS_BAZEL_TEST_DIR_ENABLE=1

        "//tensorflow/c:tf_status",
        "//tensorflow/c/experimental/filesystem:filesystem_interface",
    ],
)

# Since building pip package and API tests require a filesystem, we provide a
# static registration target that they should link against.
cc_library(
    name = "posix_filesystem_static",
    srcs = ["posix_filesystem_static.cc"],
    visibility = ["//visibility:public"],
    deps = [
        ":posix_filesystem_impl",