ray.data.TaskPoolStrategy#

class ray.data.TaskPoolStrategy(size: int | None = None)[source]#

Bases: ComputeStrategy

Specify the task-based compute strategy for a Dataset transform.

TaskPoolStrategy executes dataset transformations using Ray tasks that are scheduled through a pool. Provide size to cap the number of concurrent tasks; leave it unset to allow Ray Data to scale the task count automatically.

Methods

__init__

Construct TaskPoolStrategy for a Dataset transform.