Translate to Turkish (Türkçe).

Language code: tr.

### Core principle

Don't translate word-by-word. Rewrite naturally in Turkish as if writing the doc from scratch. Preserve meaning, but prioritize fluency over literal accuracy.

### Grammar and tone

- Use instructional Turkish, consistent with existing Turkish docs.
- Use imperative/guide language (e.g. "açalım", "gidin", "kopyalayalım", "bir bakalım").

        assertTrue(StrictMath.pow(2.0, log2 + 1) > d);
      }
    }
  }

  @GwtIncompatible // DoubleMath.log2(double, RoundingMode)
  public void testRoundLog2Half() {
    // We don't expect perfect rounding accuracy.
    for (int exp : asList(-1022, -50, -1, 0, 1, 2, 3, 4, 100, 1022, 1023)) {
      for (RoundingMode mode : asList(HALF_EVEN, HALF_UP, HALF_DOWN)) {
        double x = Math.scalb(Math.sqrt(2) + 0.001, exp);

        assertTrue(StrictMath.pow(2.0, log2 + 1) > d);
      }
    }
  }

  @GwtIncompatible // DoubleMath.log2(double, RoundingMode)
  public void testRoundLog2Half() {
    // We don't expect perfect rounding accuracy.
    for (int exp : asList(-1022, -50, -1, 0, 1, 2, 3, 4, 100, 1022, 1023)) {
      for (RoundingMode mode : asList(HALF_EVEN, HALF_UP, HALF_DOWN)) {
        double x = Math.scalb(Math.sqrt(2) + 0.001, exp);

     * removed in one segment while checking another, in which case the table was never actually
     * empty at any point. (The sum ensures accuracy up through at least 1<<31 per-segment
     * modifications before recheck.)  Method containsValue() uses similar constructions for
     * stability checks.
     */
    long sum = 0L;

     * removed in one segment while checking another, in which case the table was never actually
     * empty at any point. (The sum ensures accuracy up through at least 1<<31 per-segment
     * modifications before recheck.) Method containsValue() uses similar constructions for
     * stability checks.
     */
    long sum = 0L;

     * removed in one segment while checking another, in which case the table was never actually
     * empty at any point. (The sum ensures accuracy up through at least 1<<31 per-segment
     * modifications before recheck.) Method containsValue() uses similar constructions for
     * stability checks.
     */
    long sum = 0L;

    *   `model.add_loss(symbolic_tensor)` should work in ambient eager.
    *   Update metric name to always reflect what the user has given in compile.
        Affects following cases
    *   When name is given as 'accuracy'/'crossentropy'
    *   When an aliased function name is used eg. 'mse'
    *   Removing the `weighted` prefix from weighted metric names.
    *   Allow non-Tensors through v2 losses.

Do not make a parameter `@Nullable` just to check and throw a `NullPointerException` when it is `null`. 

## Consequences

* Classes are forced to have consistent nullability annotations.
* IDE warnings become more accurate, reducing warnings fatigue.
* Public API nullability annotations become closer to reality.
* Java compilation of Gradle code suffers from small performance penalty (up to 10%).

  }

  public void testServiceStartupTimes_selfStartingServices() {
    // This tests to ensure that:
    // 1. service times are accurate when the service is started by the manager
    // 2. service times are recorded when the service is not started by the manager (but they may
    // not be accurate).
    Service b =
        new NoOpDelayedService(353) {
          @Override
          protected void doStart() {

  }

  public void testServiceStartupTimes_selfStartingServices() {
    // This tests to ensure that:
    // 1. service times are accurate when the service is started by the manager
    // 2. service times are recorded when the service is not started by the manager (but they may
    // not be accurate).
    Service b =
        new NoOpDelayedService(353) {
          @Override
          protected void doStart() {