The human-centric platform for production ML & AI

Outerbounds

how can we write a test for a flow that uses @batch decorator without setting s3 as the default METAFLOW_DEFAULT_DATASTORE in config_test.json?

you can do `python myflow.py --datastore=local run`

or if it's easier, `METAFLOW_DATASTORE=local python myflow.py run`

oh, I missed the `@batch` part. Batch requires S3 since it needs to move data around. For testing purposes you can setup a separate `METAFLOW_PROFILE` with a different bucket so test data stays separate

you can do `python myflow.py --metadata=local run`  to keep metadata separate for testing