jeff n
01/05/2021, 7:21 PMJim Crist-Harif
01/05/2021, 7:28 PMflow.run()
yourself, the results of all tasks will be stored on the output. For small results this is fine, but for larger things that could be problematic. You might write intermediates to disk and pass a filepath around if needed. This design was intentional, many users calling flow.run
wish to inspect the output of all tasks in flow.run
.
For flows run via cloud/server (registered via flow.register()
) intermediate results won't be kept around in memory, so once an intermediate result is no longer needed in memory it will be dropped.jeff n
01/05/2021, 7:48 PM