CheckpointsConfig

toloka.client.quality_control.QualityControl.CheckpointsConfig | Source code

CheckpointsConfig(
self,
*,
real_settings: Optional[Settings] = None,
golden_settings: Optional[Settings] = None,
training_settings: Optional[Settings] = None
)

Random check majority opinion.

Only some tasks are issued with a high overlap (for example, "5") and are being tested. Other tasks are issued with the overlap set in the pool settings (for example, "1") and remain without verification. Spot check saves money and speeds up pool execution.

You can reduce the frequency of checks over time.

Example settings: in the first 25 tasks completed by the Toloker in the pool, issue every fifth task with an overlap "5" to check the answers. In subsequent tasks issue each 25 task with an overlap "5".

Parameters Description

ParametersTypeDescription
real_settingsOptional[Settings]

Checkpoints settings for main tasks.

golden_settingsOptional[Settings]

Checkpoints settings for golden tasks.

training_settingsOptional[Settings]

Checkpoints settings for train tasks.

Toloka-Kit
OverviewGetting OAuth tokenQuick start
Recipes
Reference
toloka.client
toloka.async_client
toloka.autoquality [autoquality]
toloka.metrics
toloka.streaming