this.header = new AvlNode<>();
    successor(header, header);
    this.rootReference = new Reference<>();
  }

  /** A function which can be summed across a subtree. */
  private enum Aggregate {
    SIZE {
      @Override
      int nodeAggregate(AvlNode<?> node) {
        return node.elemCount;
      }

      @Override
      long treeAggregate(@Nullable AvlNode<?> root) {

        Keras training loops like `fit`/`evaluate`, the unreduced vector loss is
        passed to the optimizer but the reported loss will be a scalar value.
    *   `SUM`: Scalar sum of weighted losses. 4. `SUM_OVER_BATCH_SIZE`: Scalar
        `SUM` divided by number of elements in losses. This reduction type is
        not supported when used with `tf.distribute.Strategy` outside of
        built-in training loops like `tf.keras` `compile`/`fit`.

          queueIterator.remove();
        }
      }
      assertThat(remaining).isEqualTo(0);
      assertIntact(queue);
      assertThat(queue).containsExactlyElementsIn(elements);
    }
  }

  private enum Element {
    ONE,
    TWO,
    THREE,
    FOUR,
    FIVE;
  }

  public void testRandomAddsAndRemoves_duplicateElements() {
    Random random = new Random(0);

          protected void verify(List<Integer> elements) {
            assertEquals(new HashSet<>(elements), multimap.get("foo"));
          }
        };

    tester.test();
  }

  private enum Color {
    BLUE,
    RED,
    YELLOW,
    GREEN
  }

  private abstract static class CountingSupplier<E> implements Supplier<E>, Serializable {
    int count;

    abstract E getImpl();

    this.header = new AvlNode<>();
    successor(header, header);
    this.rootReference = new Reference<>();
  }

  /** A function which can be summed across a subtree. */
  private enum Aggregate {
    SIZE {
      @Override
      int nodeAggregate(AvlNode<?> node) {
        return node.elemCount;
      }

      @Override
      long treeAggregate(@Nullable AvlNode<?> root) {

      /*
       * The result has the same type argument we started with. We just couldn't express EnumSet<E>
       * or ImmutableEnumSet<E> along the way because our own <E> isn't <E extends Enum<E>>.
       *
       * We are also performing a safe covariant cast to change <? extends E> to <E>.
       */
      @SuppressWarnings("unchecked")
      ImmutableSet<E> result = (ImmutableSet<E>) untypedResult;

   *
   * <p>If the mapped elements contain duplicates (according to {@code comparator}), the first
   * occurrence in encounter order appears in the resulting multiset, with count equal to the sum of
   * the outputs of {@code countFunction.applyAsInt(t)} for each {@code t} mapped to that element.
   *
   * @since 22.0
   */
  public static <T extends @Nullable Object, E>

   * Returns a predicate that returns the result of {@link File#isFile} on input files.
   *
   * @since 15.0
   */
  public static Predicate<File> isFile() {
    return FilePredicate.IS_FILE;
  }

  private enum FilePredicate implements Predicate<File> {
    IS_DIRECTORY {
      @Override
      public boolean apply(File file) {
        return file.isDirectory();
      }

      @Override

   * Returns a predicate that returns the result of {@link File#isFile} on input files.
   *
   * @since 15.0
   */
  public static Predicate<File> isFile() {
    return FilePredicate.IS_FILE;
  }

  private enum FilePredicate implements Predicate<File> {
    IS_DIRECTORY {
      @Override
      public boolean apply(File file) {
        return file.isDirectory();
      }

      @Override

// called after a successful TF_NewWhile() call.
TF_CAPI_EXPORT extern void TF_AbortWhile(const TF_WhileParams* params);

// Adds operations to compute the partial derivatives of sum of `y`s w.r.t `x`s,
// i.e., d(y_1 + y_2 + ...)/dx_1, d(y_1 + y_2 + ...)/dx_2...
//
// `dx` are used as initial gradients (which represent the symbolic partial
// derivatives of some loss function `L` w.r.t. `y`).