const auto& shape = reader->GetVariableToShapeMap().at(name);
  int rank = shape.dims();
  if (num_dims != rank) {
    status->status = InvalidArgument("Expected rank is ", num_dims,
                                     " but actual rank is ", rank);
    return;
  }
  for (int i = 0; i < num_dims; i++) {
    dims[i] = shape.dim_size(i);
  }
}

          echo "TF_VERSION=${major_version}.${minor_version}" >> "$GITHUB_OUTPUT"
          # Also get the current date to do cache busting. Assumes one day
          # is an ok range for rebuilds
          echo "DATE=$(date +'%Y-%m-%d')" >> "$GITHUB_OUTPUT"
        id: tf-version
      -
        name: Build and push
        id: docker_build

    if (r > 0) {
      dst += r;
      offset += static_cast<uint64_t>(r);
      n -= r;  // safe as 0 < r <= n so n will never underflow
      read += r;
    } else if (r == 0) {
      TF_SetStatus(status, TF_OUT_OF_RANGE, "Read fewer bytes than requested");
      break;
    } else if (errno == EINTR || errno == EAGAIN) {
      // Retry
    } else {
      TF_SetStatusFromIOError(status, errno, posix_file->filename);
      break;
    }

      looping.
  """
  default = environ_cp.get(var_name) or var_default
  full_query = '%s [Default is %s]: ' % (
      ask_for_var,
      default,
  )

  for _ in range(n_ask_attempts):
    val = get_from_env_or_user_or_default(environ_cp, var_name, full_query,
                                          default)
    if check_success(val):
      break
    if not suppress_default_error:

                                   " dimensions which is over the limit of ",
                                   TensorShape::MaxDimensions(), ".");
  }
  TensorShapeProto proto;
  if (num_dims < 0) {
    proto.set_unknown_rank(true);
  } else {
    for (int d = 0; d < num_dims; ++d) {
      proto.add_dim()->set_size(dims[d]);
    }
  }

  forward_op_->attrs.Set(attr_name, proto);

// required except where indicated.
typedef struct TFE_CustomDeviceTensorHandleMethods {
  int version = TFE_CUSTOM_DEVICE_VERSION;

  // Computes the rank of the tensor handle.
  //
  // Shapes are specified via callbacks because retrieving the shape of a tensor
  // is a blocking operation for async eager; custom devices should avoid

        `DynamicUpdateSlice`.
    *   Enabled a new MLIR-based dynamic range quantization backend by default
        *   The new backend is used for post-training int8 dynamic range
            quantization and post-training float16 quantization.
        *   Set `experimental_new_dynamic_range_quantizer` in
            tf.lite.TFLiteConverter to False to disable this change

          &(gcs_file->buffer[0]), status);
      gcs_file->buffer_end_is_past_eof =
          (TF_GetCode(status) == TF_OUT_OF_RANGE);
      if (read_fill_buffer >= 0) gcs_file->buffer.resize(read_fill_buffer);
      if (TF_GetCode(status) != TF_OK &&
          TF_GetCode(status) != TF_OUT_OF_RANGE) {
        // Empty the buffer to avoid caching bad reads.
        gcs_file->buffer.resize(0);
        return -1;
      }

to the network can execute arbitrary code with the privileges of the user
running the `tf.train.Server`.

## Untrusted inputs during training and prediction

TensorFlow supports a wide range of input data formats. For example it can
process images, audio, videos, and text. There are several modules specialized
in taking those formats, modifying them, and/or converting them to intermediate

  TF_Operation* feed2 = Placeholder(func_graph_, s_, "feed2");
  TF_Operation* add = Add(feed1, feed2, func_graph_, s_);
  DefineT(-1, {}, {{feed1, 0}, {feed2, 2}}, {{add, 0}}, {}, true);
  EXPECT_EQ(TF_OUT_OF_RANGE, TF_GetCode(s_));
  EXPECT_EQ(string("Node 'feed2' (type: 'Placeholder', num of outputs: 1) does "
                   "not have output 2\n\tEncountered while processing "
                   "input 1 into function 'MyFunc'"),