Scaling Up¶

Container¶

Not all layers in a Flow need to use the same resources. Some layers might be memory focused and require many gigabytes of memory, and others might be compute focused and require many CPUs. laminar provides a layer Container configuration that can modify the settings of the container the Layer is being run in.

from laminar import Flow, Layer
from laminar.configurations.layers import Container

class ContainerFlow(Flow):
    ...

@ContainerFlow.register(container=Container(cpu=4, memory=2000, workdir="/app"))
class Task(Layer):
    ...

A Container configuration can be shared across multiple layers.

from laminar import Flow, Layer
from laminar.configurations.layers import Container

class ContainerFlow(Flow):
    ...

container = Container(cpu=4, memory=2000, workdir="/app")

@ContainerFlow.register(container=container)
class First(Layer):
    ...

@ContainerFlow.register(container=container)
class Second(Layer):
    ...