ZeroDevisionError when using 0<dask_distributed_local_core_fraction<1 on machine with one CPU

[observingClouds]

While writing a CI test, I realize that the following fails when using a dask_distributed_local_core_fraction other than 0 or 1 if the machine has only one core (like the standard GitHub CI compute).

mllam-data-prep/mllam_data_prep/__main__.py

Lines 56 to 64 in a283eff

	n_system_cores = os.cpu_count()
	# compute the number of cores to use
	n_local_cores = int(args.dask_distributed_local_core_fraction * n_system_cores)
	# get the total system memory
	total_memory = psutil.virtual_memory().total
	# compute the memory per worker
	memory_per_worker = (
	total_memory / n_local_cores * args.dask_distributed_local_memory_fraction
	)

Even though this case will hardly ever be an issue for our computations, I suggest to replace int() with numpy.ceil() or raise a specific warning. Currently one just gets a ZeroDevisionError.

[leifdenby]

replace int() with numpy.ceil() or raise a specific warning

Good catch. ceil makes sense to me, good idea