For that you would need to persist the results yourself - there is no such feature in Prefect to automatically infer what you want to persist unless you return it and checkpoint the result.
Do you happen to use AWS S3? If so, you could persist this data in a single command using awswrangler
- I love this library!
https://aws-data-wrangler.readthedocs.io/en/stable/
Alternatively, you could manually call .write()
method on results within your task:
from prefect import Flow, task
from prefect.engine.results import S3Result
@task
def my_task():
s3_result = S3Result(bucket='bucket_of_models')
s3_result.write(your_dask_dataframe)
More on that:
Also, you can check: