optuna.integration.TensorFlowPruningHook

class optuna.integration.TensorFlowPruningHook(trial, estimator, metric, run_every_steps)[source]

TensorFlow SessionRunHook to prune unpromising trials.

See the example if you want to add a pruning hook to TensorFlow’s estimator.

Parameters
  • trial (Trial) – A Trial corresponding to the current evaluation of the objective function.

  • estimator (tf.estimator.Estimator) – An estimator which you will use.

  • metric (str) – An evaluation metric for pruning, e.g., accuracy and loss.

  • run_every_steps (int) – An interval to watch the summary file.

Methods

after_run(run_context, run_values)

before_run(run_context)

begin()