DelayedΒΆ
Sometimes problems don’t fit into one of the collections like dask.array or
dask.dataframe. In these cases, users can parallelize custom algorithms
using the simpler dask.delayed interface. This allows one to create graphs
directly with a light annotation of normal python code.
>>> x = dask.delayed(inc)(1)
>>> y = dask.delayed(inc)(2)
>>> z = dask.delayed(add)(x, y)
>>> z.compute()
7
>>> z.vizualize()