Executors¶

A Flow supports multiple backend executors which configure how work is executed in the workflow.

Docker¶

By default, flows are configured to use the Docker executor which launches each Layer registered to a Flow in its own docker container. The launched containers can be configured with the layer’s Container configuration.

from laminar import Flow
from laminar.configurations import executors, layers

class DockerFlow(Flow):
    ...

@DockerFlow.register(container=layers.Container(cpu=1, memory=1500))
class A(Layer):
    ...

flow = DockerFlow(executor=executors.Docker())

Thread¶

The Thread executor executes layers directly in the main Python process. This is very useful for testing.

from laminar import Flow
from laminar.configurations import executors

class ThreadFlow(Flow):
    ...

flow = ThreadFlow(executor=executors.Thread())

AWS.Batch¶

Warning

AWS.Batch is experimental.

The `AWS.Batch`` executor executes layers on the AWS Batch compute service.

from laminar import Flow
from laminar.configurations import executors

class BatchFlow(Flow):
    ...

flow = BatchFlow(executor=executors.AWS.Batch(job_queue_arn=..., job_role_arn=...))