TuneResult

class nvmolkit.autotune.TuneResult(

best_config: ~typing.Any,

best_throughput: float,

best_params: dict[str,

~typing.Any] = <factory>,

calibration_size: int = 0,

n_trials_run: int = 0,

study: ~typing.Any = None,

)

Result returned by every tune_* wrapper.

best_config

The configuration object (e.g. HardwareOptions or SubstructSearchConfig) corresponding to the best trial.

Type:

Any

best_throughput

Throughput in items per second of the best trial.

Type:

float

best_params

Raw parameter dictionary of the best optuna trial.

Type:

dict[str, Any]

calibration_size

Number of indices used in each trial after any adaptive shrink.

Type:

int

n_trials_run

Number of optuna trials that completed successfully.

Type:

int

study

The underlying optuna.Study object, or None if optuna is unavailable.

Type:

Any