Package jdistlib.inference

Class SamplingOptions

java.lang.Object
jdistlib.inference.SamplingOptions
public final class SamplingOptions extends Object
Immutable common MCMC warmup, retention, adaptation, and safety options.

  • Method Details

    • builder

      public static SamplingOptions.Builder builder()

    • toBuilder

      public SamplingOptions.Builder toBuilder()
      Starts a builder that preserves every option, including streaming callbacks.

    • warmupIterations

      public int warmupIterations()

    • sampleIterations

      public int sampleIterations()

    • thinning

      public int thinning()

    • stepSize

      public double stepSize()

    • targetAcceptance

      public double targetAcceptance()

    • leapfrogSteps

      public int leapfrogSteps()

    • maximumTreeDepth

      public int maximumTreeDepth()

    • maximumEnergyError

      public double maximumEnergyError()

    • sliceWidth

      public double sliceWidth()

    • maximumSliceSteps

      public int maximumSliceSteps()

    • adaptStepSize

      public boolean adaptStepSize()

    • adaptMassMatrix

      public boolean adaptMassMatrix()

    • denseMassMatrix

      public boolean denseMassMatrix()

    • allowFiniteDifferences

      public boolean allowFiniteDifferences()

    • warmupSchedule

      public WarmupSchedule warmupSchedule()

    • metricConfiguration

      public MetricConfiguration metricConfiguration()

    • integrationTime

      public double integrationTime()
      Static-HMC integration time; NaN retains the legacy fixed leapfrog count.

    • stepSizeJitter

      public double stepSizeJitter()

    • storeDraws

      public boolean storeDraws()

    • computeBackend

      public Compute computeBackend()
      Compute policy for accelerator-aware numerical targets and operations.

    • nutsBackend

      public ComputeNuts nutsBackend()
      NUTS-specific accelerator policy; tree construction always remains on CPU.

    • cancelled

      public boolean cancelled()