```

### 3. Failure after conversion
If the conversion is successful, but the generated model is wrong, then state what is wrong:

- Model produces wrong results and/or has lesser accuracy.
- Model produces correct results, but it is slower than expected.

### 4. (optional) RNN conversion support
If converting TF RNN to TFLite fused RNN ops, please prefix [RNN] in the title.

 * Stopwatch} requires object allocation and additional method calls, which can reduce the accuracy
 * of the elapsed times reported. {@code Stopwatch} is still suitable for logging and metrics where
 * reasonably accurate values are sufficient. If the uncommon case that you need to maximize
 * accuracy, use {@code System.nanoTime()} directly instead.
 *
 * <p>Basic usage:
 *
 * <pre>{@code

        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);

}

// To collect RX stats during "mc support perf net"
// RXSample holds the RX bytes for the duration between
// the last peer to connect and the first peer to disconnect.
// This is to improve the RX throughput accuracy.
type netPerfRX struct {
	RX                uint64    // RX bytes
	lastToConnect     time.Time // time at which last peer to connect to us
	firstToDisconnect time.Time // time at which the first peer disconnects from us

        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);

pkg maps, func EqualFunc[$0 interface{ ~map[$2]$3 }, $1 interface{ ~map[$2]$4 }, $2 comparable, $3 interface{}, $4 interface{}]($0, $1, func($3, $4) bool) bool #57436
pkg math/big, method (*Int) Float64() (float64, Accuracy) #56984
pkg net/http, method (*ProtocolError) Is(error) bool #41198
pkg net/http, method (*ResponseController) EnableFullDuplex() error #57786
pkg net/http, var ErrSchemeMismatch error #44855

     * 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.