Evaluator
Periodic in-loop evaluator.
Each evaluator owns one metric source (a Benchmark, a custom probe,
etc.) and carries its own run_every cadence. This lets a single
training run mix a fast val benchmark (run_every: 50) with a
heavier test benchmark (run_every: 500).
Attributes
attributenamestrShort name; used to prefix the metric keys (val/\<name>/\<key>).
attributerun_everyintPer-evaluator step cadence. 0 → disabled (never fires).
Positive → fire when (step + 1) % run_every == 0.
Functions
funcevaluate(self, sampler, *, model_path=None, step=None) -> dict[str, float]paramselfparamsamplerSamplingClientparammodel_pathstr | None= Noneparamstepint | None= NoneReturns
dict[str, float]