How can I make the results of a numpy random replicable using Dask or joblib Parallel if I can’t with Dask?
The get_random function is more complex and manipulates pandas.Dataframe objects.
`rng = numpy.random.RandomState(123)
seeds = rng.randint(0, 2**31 – 1, size=10)
def get_random(seed):
in_rng = numpy.random.RandomState(seed)
return in_rng.randint(0, 100)
delayed_results = []
for seed in seeds:
result = dask.delayed(getRandom)(seed)
delayed_results.append(result)
results = dask.compute(*delayed_results)`