Skip to main content

Running non-blocking dask computations in the background

superduperdb offers the possiblity to run all long running blocking jobs in the background via dask. Read about the dask project here.

To configure this feature, configure:

from superduperdb import CFG

CFG.mode = 'production'

When this is so-configured the following functions push their computations to the dask cluster:

  • db.add
  • db.insert
  • db.update
  • Model.predict

When dask is configured, these functions returns either a superduperdb.job.Job object, or an iterable thereof.

job = m.predict(     # a `superduper.job.ComponentJob` object
) # watch the `stdout` of the `Job`