ray.tune.Tuner.fit#

Tuner.fit() ResultGrid[source]#

Executes hyperparameter tuning job as configured and returns result.

Failure handling: For the kind of exception that happens during the execution of a trial, one may inspect it together with stacktrace through the returned result grid. See ResultGrid for reference. Each trial may fail up to a certain number. This is configured by RunConfig.FailureConfig.max_failures.

Exception that happens beyond trials will be thrown by this method as well. In such cases, there will be instruction like the following printed out at the end of console output to inform users on how to resume.

Please use Tuner.restore to resume.

import os
from ray.tune import Tuner

trainable = ...

tuner = Tuner.restore(
    os.path.expanduser("~/ray_results/tuner_resume"),
    trainable=trainable
)
tuner.fit()
Raises:
  • RayTaskError – If user-provided trainable raises an exception

  • TuneError – General Ray Tune error.