apiserver_impersonation_attempts_total{mode, decision}
  apiserver_impersonation_attempts_duration_seconds{mode, decision}
  apiserver_impersonation_authorization_attempts_total{mode, decision}

// CheckPreconditionFn returns true if precondition check failed.
type CheckPreconditionFn func(o ObjectInfo) bool

// EvalMetadataFn validates input objInfo and GetObjectInfo error and returns an updated metadata and replication decision if any
type EvalMetadataFn func(o *ObjectInfo, gerr error) (ReplicateDecision, error)

// EvalRetentionBypassFn validates input objInfo and GetObjectInfo error and returns an error if retention bypass is not allowed.

    // is not available in GWT yet.  To keep the code simple and code size more,
    // we make a copy here, instead of creating a view of it.
    //
    // TODO: revisit if it's unbearably slow or when GWT supports
    // TreeMap.navigatbleKeySet().
    return ImmutableSortedSet.copyOf(comparator, sortedDelegate.keySet());
  }

  @Override

    feature.
*   Full new features (e.g., a new op implementing a cutting-edge algorithm)
    typically will live in
    [tensorflow/addons](https://github.com/tensorflow/addons) to get some
    airtime before a decision is made regarding whether they are to be migrated
    to the core.
*   As every PR requires several CPU/GPU hours of CI testing, we discourage
    submitting PRs to fix one typo, one warning,etc. We recommend fixing the

   * one from the pool. We must take the pooled connection because by taking it from the pool, we've
   * fully acquired it.
   *
   * This test yields threads to force the decision of plan1 to be deliberate and not lucky. In
   * particular, we set up this sequence of events:
   *
   *  1. take plan 0
   *  3. plan 0 connects
   *  4. finish taking plan 1
   *

     * b) this iteration converges to floor(sqrt(x)). In fact, the number of correct digits doubles
     * with each iteration, so this algorithm takes O(log(digits)) iterations.
     *
     * We start out with a double-precision approximation, which may be higher or lower than the
     * true value. Therefore, we perform at least one Newton iteration to get a guess that's

  // and not for correctness. The only downside of keeping this 1 seems to be
  // that the gradient accumulation is unbounded and we will never
  // aggressively aggregate accumulated gradients to recover memory.
  // Revisit and fix.
  return 1;
}

// Consumes references to the tensors in the gradient_tensors list and returns
// a tensor with the result.
AbstractTensorHandle* TapeVSpace::AggregateGradients(

     *
     * Historical note: At one point, we'd declared this method as accepting and returning nullable
     * values. For details on that, see earlier revisions of this file.
     */
    return convert(a);
  }

  /**
   * <i>May</i> return {@code true} if {@code object} is a {@code Converter} that behaves
   * identically to this converter.
   *

notice that is included in or attached to the work (an example is provided in the Appendix below). "Derivative Works" shall mean any work, whether in Source or Object form, that is based on (or derived from) the Work and for which the editorial revisions, annotations, elaborations, or other modifications represent, as a whole, an original work of authorship. For the purposes of this License, Derivative Works shall not include works that remain separable from, or merely link (or bind by name) to the...

     *
     * Historical note: At one point, we'd declared this method as accepting and returning nullable
     * values. For details on that, see earlier revisions of this file.
     */
    return convert(a);
  }

  /**
   * <i>May</i> return {@code true} if {@code object} is a {@code Converter} that behaves
   * identically to this converter.
   *