hi are there any docs on how to use the

@checkpoint

decorator when using

@parallel

steps? The docs I found here say TODO https://github.com/outerbounds/metaflow-checkpoint-examples/blob/master/documentation/checkpoint_deco/checkpoint_usage.md#saving--loa[…]rallel-steps

@hallowed-glass-14538 might have stuff handy

so the checkpoint decorator gives more explicit controls over checkpointing but that means something has to call the api (

current.checkpoint.save

). If you are fully using Ray then you don't even need the decorator since you can let ray fully control the checkpoint storage/loading. Example :

from metaflow.metaflow_config import DATASTORE_SYSROOT_S3
from metaflow import current
import os

path_to_checkpoints = os.path.join(
    DATASTORE_SYSROOT_S3,
    "mf.ray_checkpoints",
    current.flow_name,
    current.run_id,
    current.step_name,
    current.parallel.control_task_id,
)
from ray.train import RunConfig
RunConfig(
    storage_path = path_to_checkpoints
    # ...
)

This way every execution will store checkpoints in a different location. You can set the value of

path_to_checkpoints

to

self

to make it a data artifact.

oh nice thanks for explaining this