Skip to content
Hayonix

API reference · Schema

FineTuneMetricPoint

One point per step, in step order: plot step against any metric

  • objectstringrequired

    Always `fine_tune_metric_point`: the type of this object

    One offine_tune_metric_point

  • stepintegerrequired

    Optimizer step the values were recorded at

  • epochnumber | nullrequired

    Epoch at this step, fractional; null when not reported

  • train_lossnumber | nullrequired

    Training loss; null when not reported at this step

  • eval_lossnumber | nullrequired

    Loss on the validation file; null at steps without an evaluation

  • learning_ratenumber | nullrequired

    Learning rate at this step; null when not reported

  • grad_normnumber | nullrequired

    Gradient norm before clipping; null when not reported

  • valuesobjectrequired

    Every metric at this step (preference training adds reward metrics)

  • recorded_atstring (date-time)required

    Time recorded (ISO 8601, UTC)

Example
{
  "object": "fine_tune_metric_point",
  "step": 1,
  "epoch": 1.5,
  "train_loss": 1.5,
  "eval_loss": 1.5,
  "learning_rate": 1.5,
  "grad_norm": 1.5,
  "values": {},
  "recorded_at": "2026-01-01T00:00:00.000Z"
}

Used by