如何从单独的线程中的dask.distributed运行进度条小部件？

Question

There is an example here , showing insertion of data for processing in a separate thread. I'm interested in reverse, when data are inserted manually and interactively, and background thread submits them to cluster for calculation.

Answer 1

This is already the case. When you call

client.submit(function, *args, **kwargs)

This serializes stuff immediately in the local thread (blocking), but then adds a callback to the Tornado IOLoop (running in a separate thread) to manage the actual communication to the scheduler. This all happens asynchronously to the main thread.