Skip to main content

limen.inference

Replay selected experiment rounds into validated models and run them bar-by-bar at inference time.

Canonical docs​

What this package owns​

Owns the reconstruction-and-inference path: replaying a selected permutation from a completed experiment, validating it against the original log, and wrapping it as a live Sensor. Does not own experiment search execution (owned by limen.experiment) or the reference model implementations it reuses (owned by limen.sfd.reference_architecture).

Key entry points​

Entry pointUse caseNotes
TrainerReplay selected permutations from a completed YAML experimentRe-runs prepare_data()/run_model() with the original split and compares metrics to the log
SensorInference wrapper around a trained YAML modelPredicts one bar at a time for live use
BarPredictionResult of a single-bar predictionCarries prediction, probability, and a reason
ReconstructionErrorRaised when metric validation deviates beyond toleranceSignals a permutation could not be faithfully rebuilt

Adjacent modules​

  • limen.experiment produces the experiment logs and artifacts Trainer reconstructs. It does not re-export inference names; import them from limen.inference or top-level limen.
  • limen.sfd.reference_architecture.ReferenceModel is the model type a Sensor wraps.
  • limen.yaml.compiler.CompiledSFD supplies the compiled manifest used during reconstruction.

Quick orientation​

inference/
├── trainer.py   # Trainer: reconstruct and validate a permutation
├── sensor.py    # Sensor + BarPrediction: live bar-by-bar prediction
└── errors.py    # ReconstructionError

Things to know​

  • Trainer validates by re-running the manifest and comparing metrics to the original log; a deviation past tolerance raises ReconstructionError rather than returning a mismatched model.
  • Metric comparison uses a float tolerance (with a wider tolerance for stochastic models), and a fixed set of bookkeeping columns (id, execution_time, …) is skipped during comparison.
  • BarPrediction.reason explains a None prediction ('warm-up', 'inside-training-window', 'null-features', 'sensor-error') — a null prediction is expected during warm-up, not an error.