*/
    public ScaleAndIndex index(int index) {
      return new ScaleAndIndex(scale, index);
    }

    /**
     * Specifies multiple quantile indexes to be calculated, each index being the k in the kth
     * q-quantile.
     *
     * @param indexes the quantile indexes, each of which must be in the inclusive range [0, q] for

    }
  },
  ;

  /**
   * Calculates a single quantile. Equivalent to {@code
   * Quantiles.scale(scale).index(index).computeInPlace(dataset)}.
   */
  abstract double singleQuantile(int index, int scale, double[] dataset);

  /**
   * Calculates multiple quantiles. Equivalent to {@code
   * Quantiles.scale(scale).indexes(indexes).computeInPlace(dataset)}.
   */
  abstract Map<Integer, Double> multipleQuantiles(

  public void testScale_zero() {
    assertThrows(IllegalArgumentException.class, () -> Quantiles.scale(0));
  }

  public void testScale_negative() {
    assertThrows(IllegalArgumentException.class, () -> Quantiles.scale(-4));
  }

  public void testScale_index_negative() {
    Quantiles.Scale intermediate = Quantiles.scale(10);
    assertThrows(IllegalArgumentException.class, () -> intermediate.index(-1));
  }

        `inference_output_type` for full integer quantized models. This allows
        users to modify the model input and output type to integer types
        (`tf.int8`, `tf.uint8`) instead of defaulting to float type
        (`tf.float32`).
*   NNAPI
    *   Adds NNAPI Delegation support for requantization use cases by converting
        the operation into a dequantize-quantize pair.

ti
ci
vi
lo
la
li
le
gli
ne
il
un
uno
una
ma
ed
se
perché
anche
come
dov
dove
che
chi
cui
non
più
quale
quanto
quanti
quanta
quante
quello
quelli
quella
quelle
questo
questi
questa
queste
si
tutto
tutti
a
c
e
i
l
o
ho
hai
ha
abbiamo
avete
hanno

      Map<Integer, Double> quantiles = algorithm.multipleQuantiles(indexes, 100, dataset.clone());
      assertWithMessage("Wrong keys from " + algorithm).that(quantiles.keySet()).isEqualTo(indexes);
      for (int i : indexes) {
        assertWithMessage("Mismatch between %s and %s at %s", algorithm, REFERENCE_ALGORITHM, i)
            .that(quantiles.get(i))
            .isWithin(ALLOWED_ERROR)

      Map<Integer, Double> quantiles = algorithm.multipleQuantiles(indexes, 100, dataset.clone());
      assertWithMessage("Wrong keys from " + algorithm).that(quantiles.keySet()).isEqualTo(indexes);
      for (int i : indexes) {
        assertWithMessage("Mismatch between %s and %s at %s", algorithm, REFERENCE_ALGORITHM, i)
            .that(quantiles.get(i))
            .isWithin(ALLOWED_ERROR)

    * Si on compare Uvicorn, il faut le comparer à d'autre applications de serveurs comme Daphne, Hypercorn, uWSGI, etc.
* **Starlette** :

import com.google.common.collect.ImmutableSet;
import com.google.common.collect.Range;
import java.util.Random;

/** Benchmarks some algorithms providing the same functionality as {@link Quantiles}. */
public class QuantilesBenchmark {

  private static final ContiguousSet<Integer> ALL_DECILE_INDEXES =
      ContiguousSet.create(Range.closed(0, 10), DiscreteDomain.integers());

import com.google.common.collect.ImmutableSet;
import com.google.common.collect.Range;
import java.util.Random;

/** Benchmarks some algorithms providing the same functionality as {@link Quantiles}. */
public class QuantilesBenchmark {

  private static final ContiguousSet<Integer> ALL_DECILE_INDEXES =
      ContiguousSet.create(Range.closed(0, 10), DiscreteDomain.integers());