What causes "Doesn't support floats yet" in stanfordnlp/CoreNLP?

Use the double-precision constructors, e.g. new QNMinimizer(Function) or new QNMinimizer()

stanfordnlp/CoreNLP · error · java.lang.UnsupportedOperationException

Doesn't support floats yet

Error message

Doesn't support floats yet

What it means

QNMinimizer has a constructor taking a FloatFunction that is intentionally unimplemented: it always throws UnsupportedOperationException("Doesn't support floats yet"). L-BFGS in this library is implemented for double precision only, so float-based monitoring/evaluation is not available.

Solutions

Use the double-precision constructors, e.g. new QNMinimizer(Function) or new QNMinimizer()
Convert your objective to implement the double-based Function interface
If float monitoring is needed, wrap it and pass a double-based Function monitor instead

Example fix

// before
QNMinimizer minimizer = new QNMinimizer((FloatFunction) objective);
// after
QNMinimizer minimizer = new QNMinimizer((Function) objective);

Defensive patterns

Strategy: validation

Validate before calling

if (monitor instanceof FloatFunction) throw new IllegalArgumentException("QNMinimizer only supports double-based Function monitors");
QNMinimizer m = new QNMinimizer((Function) monitor);

Type guard

boolean usableWithQN = !(monitor instanceof FloatFunction); // must be double-based Function

Try / catch

try { m = new QNMinimizer(monitor); } catch (UnsupportedOperationException e) { if (e.getMessage().contains("Doesn't support floats")) { m = new QNMinimizer(toDoubleFunction(monitor)); } else throw e; }

Prevention

Implement objectives against the double-based Function interface only
Avoid generic factory code that may pick the FloatFunction constructor
Check constructor overloads before wrapping float-based code

When it happens

Trigger: Invoking new QNMinimizer(FloatFunction) — any call to that constructor.

Common situations: Code written against an API that advertises float support; porting code from another optimizer library that uses floats; generic code that instantiates QNMinimizer with either Function or FloatFunction.

Related errors

AI-assisted analysis of stanfordnlp/CoreNLP@1b7edd19c4 (2026-09-10). Data as JSON: /api/errors/b5a7f45f0ee7af52. Report an issue: GitHub.

Appendix: source

Thrown at src/edu/stanford/nlp/optimization/QNMinimizer.java:186

  public QNMinimizer(Function monitor) {
    this.monitor = monitor;
  }

  public QNMinimizer(Function monitor, int m) {
    this(monitor, m, false);
  }

  public QNMinimizer(Function monitor, int m, boolean useRobustOptions) {
    this.monitor = monitor;
    mem = m;
    if (useRobustOptions) {
      this.setRobustOptions();
    }
  }

  public QNMinimizer(FloatFunction monitor) {
    throw new UnsupportedOperationException("Doesn't support floats yet");
  }

  public void setOldOptions() {
    useAveImprovement = true;
    useRelativeNorm = false;
    useNumericalZero = false;
    lsOpt = eLineSearch.BACKTRACK;
    scaleOpt = eScaling.SCALAR;
  }

  public final void setRobustOptions() {
    useAveImprovement = true;
    useRelativeNorm = true;
    useNumericalZero = true;
    lsOpt = eLineSearch.MINPACK;
    scaleOpt = eScaling.DIAGONAL;
  }

View on GitHub (pinned to 1b7edd19c4)