std::vector<int64_t> source_tensor_ids = MakeTensorIDList(sources);
  tensorflow::gtl::FlatSet<int64_t> sources_set(source_tensor_ids.begin(),
                                                source_tensor_ids.end());
  std::unordered_map<int64_t, TapeTensor> sources_that_are_targets;
  for (int i = 0; i < target_tensor_ids.size(); ++i) {
    int64_t target_id = target_tensor_ids[i];

            I pushed these containers:
            
            - `gcr.io/tensorflow-sigs/build:${{ github.event.number }}-python3.12`
            - `gcr.io/tensorflow-sigs/build:${{ github.event.number }}-python3.11`
            - `gcr.io/tensorflow-sigs/build:${{ github.event.number }}-python3.10`
            - `gcr.io/tensorflow-sigs/build:${{ github.event.number }}-python3.9`
            

  const int last_node_id = graph->graph.num_node_ids();
  tensorflow::ImportGraphDefResults results;
  status->status = tensorflow::ImportGraphDef(opts->opts, def, &graph->graph,
                                              &graph->refiner, &results);
  if (!status->status.ok()) return;

  // Add new nodes to name_map
  for (int i = last_node_id; i < graph->graph.num_node_ids(); ++i) {
    auto* node = graph->graph.FindNodeId(i);

    "creators": [
        {
            "name": "TensorFlow Developers"
        }
    ],
    "access_right": "open",
    "notes": "Specific TensorFlow versions can be found in the \"Versions\" list on the right side of this page.<br>See the full list of authors <a href=\"https://github.com/tensorflow/tensorflow/graphs/contributors\">on GitHub</a>."

### 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
running models or processes in the same system. GPUs can also have
implementation bugs that might allow attackers to leave malicious code running

///   invariants (i.e., guarantee the preconditions before calling the operation
///   and check the postconditions after the operation returns). If plugins need
///   additional invariants, they should be checked on the plugin side and the
///   `status` out variable should be updated accordingly (e.g., to include
///   plugin version information that relates to the condition change).
///

* `tf.tpu.experimental.embedding.TPUEmbeddingV2`
    * Add `compute_sparse_core_stats` for sparse core users to profile the 
      data with this API to get the `max_ids` and `max_unique_ids`. These
      numbers will be needed to configure the sparse core embedding mid level 
      api.
    * Remove the `preprocess_features` method since that's no longer needed.

## Thanks to our Contributors

  }
  if (result.size() != source_tensor_ids.size()) {
    return errors::Internal("Expected result Span to be of size ",
                            source_tensor_ids.size(), " found ", result.size(),
                            " in call to Tape::ComputeGradient.");
  }
  std::unordered_set<int64_t> used_gradient_ids(source_tensor_ids.size());
  for (int i = 0; i < source_tensor_ids.size(); i++) {

# This flag is set to true in Bazel 1.0 and newer versions. We tried to migrate
# Tensorflow to the default, however test coverage wasn't enough to catch the
# errors.
# There is ongoing work on Bazel team's side to provide support for transitive
# shared libraries. As part of migrating to transitive shared libraries, we
# hope to provide a better mechanism for control over symbol exporting, and
# then tackle this issue again.
#

# Builds the following Docker images for Linux ARM64. See the accompanying
# Dockerfile for more details:
# - gcr.io/tensorflow-sigs/build-arm64:jax-latest-multi-python
# - gcr.io/tensorflow-sigs/build-arm64:tf-latest-multi-python

set -exo pipefail

function is_continuous_or_release() {
  [[ "$KOKORO_JOB_TYPE" == "CONTINUOUS_INTEGRATION" ]] || [[ "$KOKORO_JOB_TYPE" == "RELEASE" ]]