### Typical Pull Request Workflow -

**1. New PR**

- As a contributor, you submit a New PR on GitHub.
- We inspect every incoming PR and add certain labels to the PR such as `size:`,
  `comp:` etc.  At this stage we check if the PR is valid and meets certain
  quality requirements. For example, we check if the CLA is signed, PR has

// Fetches the current number of inputs attached to `op`.
//
// Does not use the operation's definition to determine how many inputs should
// be attached. It is intended for use with TFE_OpGetFlatInput to inspect an
// already-finalized operation.
//
// Note that TFE_OpGetFlatInputCount and TFE_OpGetFlatInput operate on a flat
// sequence of inputs, unlike TFE_OpGetInputLength (for getting the length of a

                                      TF_Status* status) -> int64_t {
    EXPECT_EQ(n, block_size);
    EXPECT_EQ(offset % block_size, 0);
    EXPECT_EQ(calls.find(offset), calls.end()) << "at offset " << offset;
    calls.insert(offset);
    memset(buffer, 'x', n);
    TF_SetStatus(status, TF_OK, "");
    return n;
  };
  const uint32 block_count = 256;
  tf_gcs_filesystem::RamFileBlockCache cache(

  auto cluster = server_def.mutable_cluster();
  auto client_job = cluster->add_job();
  client_job->set_name("localhost");
  int client_port = tensorflow::testing::PickUnusedPortOrDie();
  client_job->mutable_tasks()->insert(
      {0, strings::StrCat("localhost:", client_port)});
  server_def.set_job_name("localhost");
  auto serialized = server_def.SerializeAsString();

// APIs for generically dealing with op attributes (e.g. when forwarding them
// through custom device implementations).
//
// TODO(allenl): Currently these are black boxes, but we should have some way to
// inspect values. This would let people e.g. copy over most attributes and then
// modify some based on their values.

// A reference to an op's name -> attribute mapping
typedef struct TFE_OpAttrs TFE_OpAttrs;

  auto iter = registry_.find(op_name);
  if (iter != registry_.end()) {
    const string error_msg = "Gradient already exists for op: " + op_name + ".";
    return errors::AlreadyExists(error_msg);
  }
  registry_.insert({op_name, gradient_function_factory});
  return absl::OkStatus();
}
Status GradientRegistry::Lookup(
    const ForwardOperation& op,
    std::unique_ptr<GradientFunction>* gradient_function) const {

  tensorflow::JobDef* job_def = cluster_def->add_job();
  job_def->set_name(job_name);
  for (int i = 0; i < num_tasks; i++) {
    int port = tensorflow::testing::PickUnusedPortOrDie();
    job_def->mutable_tasks()->insert(
        {i, tensorflow::strings::StrCat("localhost:", port)});
    LOG(INFO) << "Picked test port: " << port << " for job: " << job_name
              << ", task: " << i;
  }
  return server_def;
}

  void DeleteTrace(int64_t tensor_id);

  // Consumes the internal state of the tape (so cannot be called more than
  // once) and produces the gradient of the target tensors with respect to the
  // source tensors. The output gradients are used if not empty and not
  // null. The result is populated with one tensor per target element.
  // When running backward functions, builds zeros-like tensors for

                           TF_DataType value) {
  auto iter = builder->attr_names.insert(attr_name).first;
  builder->Set(*iter, static_cast<tensorflow::DataType>(value));
}

void TF_AttrBuilderSetTypeList(TF_AttrBuilder* builder, const char* attr_name,
                               const TF_DataType* values, int num_values) {
  auto iter = builder->attr_names.insert(attr_name).first;

  tensorflow::ClusterDef* cluster_def = server_def.mutable_cluster();
  tensorflow::JobDef* job_def = cluster_def->mutable_job(0);
  int port = tensorflow::testing::PickUnusedPortOrDie();
  job_def->mutable_tasks()->insert(
      {2, tensorflow::strings::StrCat("localhost:", port)});
  server_def.set_task_index(0);
  string serialized_update = server_def.SerializeAsString();
  TFE_ContextUpdateServerDef(ctx, 0, serialized_update.data(),