In order to broadcast 2d arrays and create a 3d array, i use this simple code.
import numpy as np
import dask.array as da
x = da.random.uniform(size=(200, 100000), chunks=(200,100))
ind = da.random.binomial(1, 0.5, (200, 200, 1)).astype(np.bool)
x = da.where(ind, x, 0)
x = x.persist() #the crash happens here
print(x)`
I expect x to be a (200,200,100000) array with 0 where ind is False. However, its cause a sudden increase of the RAM usage and my pc freeze and it becomes unusable. I tried several chunk size but with the same result. setup: os: linux mint 18 cpu: i5 4440 ram: 8gb dask version: 1.2.0
thank you for your help