tune.iterative

tune.iterative.asha

class ASHAJudge(schedule, always_checkpoint=False, study_early_stop=None, trial_early_stop=None, monitor=None)[source]

Bases: tune.concepts.flow.judge.TrialJudge

Parameters

schedule (List[Tuple[float, int]]) –
always_checkpoint (bool) –
study_early_stop (Optional[Callable[[List[Any], List[tune.iterative.asha.RungHeap]], bool]]) –
trial_early_stop (Optional[Callable[[tune.concepts.flow.report.TrialReport, List[tune.concepts.flow.report.TrialReport], List[tune.iterative.asha.RungHeap]], bool]]) –
monitor (Optional[tune.concepts.flow.judge.Monitor]) –

property always_checkpoint: bool

can_accept(trial)[source]

Parameters: trial (tune.concepts.flow.trial.Trial) –
Return type: bool

get_budget(trial, rung)[source]

Parameters

trial (tune.concepts.flow.trial.Trial) –
rung (int) –

Return type

float

judge(report)[source]

Parameters: report (tune.concepts.flow.report.TrialReport) –
Return type: tune.concepts.flow.judge.TrialDecision

property schedule: List[Tuple[float, int]]

class RungHeap(n)[source]

Bases: object

Parameters: n (int) –

property best: float

property bests: List[float]

property capacity: int

property full: bool

push(report)[source]

Parameters: report (tune.concepts.flow.report.TrialReport) –
Return type: bool

values()[source]

Return type: Iterable[tune.concepts.flow.report.TrialReport]

tune.iterative.objective

class IterativeObjectiveFunc[source]

Bases: object

copy()[source]

Return type: tune.iterative.objective.IterativeObjectiveFunc

property current_trial: tune.concepts.flow.trial.Trial

finalize()[source]

Return type: None

generate_sort_metric(value)[source]

Parameters: value (float) –
Return type: float

initialize()[source]

Return type: None

load_checkpoint(fs)[source]

Parameters: fs (fs.base.FS) –
Return type: None

run(trial, judge, checkpoint_basedir_fs)[source]

Parameters

trial (tune.concepts.flow.trial.Trial) –
judge (tune.concepts.flow.judge.TrialJudge) –
checkpoint_basedir_fs (fs.base.FS) –

Return type

None

run_single_iteration()[source]

Return type: tune.concepts.flow.report.TrialReport

run_single_rung(budget)[source]

Parameters: budget (float) –
Return type: tune.concepts.flow.report.TrialReport

property rung: int

save_checkpoint(fs)[source]

Parameters: fs (fs.base.FS) –
Return type: None

validate_iterative_objective(func, trial, budgets, validator, continuous=False, checkpoint_path='', monitor=None)[source]

Parameters

func (tune.iterative.objective.IterativeObjectiveFunc) –
trial (tune.concepts.flow.trial.Trial) –
budgets (List[float]) –
validator (Callable[[List[tune.concepts.flow.report.TrialReport]], None]) –
continuous (bool) –
checkpoint_path (str) –
monitor (Optional[tune.concepts.flow.judge.Monitor]) –

Return type

None

tune.iterative.sha

tune.iterative.study

class IterativeStudy(objective, checkpoint_path)[source]

Bases: object

Parameters

objective (tune.iterative.objective.IterativeObjectiveFunc) –
checkpoint_path (str) –

optimize(dataset, judge)[source]

Parameters

dataset (tune.concepts.dataset.TuneDataset) –
judge (tune.concepts.flow.judge.TrialJudge) –

Return type

tune.concepts.dataset.StudyResult