Hi,
Could you tell me please if calling da.to_zarr with compute=True computes all previous operations (e.g. transpose, convolve)?
Best regards,
Aliaksei
Hi,
Could you tell me please if calling da.to_zarr with compute=True computes all previous operations (e.g. transpose, convolve)?
Best regards,
Aliaksei
@aliaksei-chareshneu Welcome to Discourse!
Could you tell me please if calling da.to_zarr with compute=True computes all previous operations (e.g. transpose, convolved)?
That’s correct, and compute=True
is also the default for to_zarr
. If you specify compute=False
, it will return a Delayed
object which you then need to compute
to execute all previous operations including the to_zarr
.
Example:
import dask.array as da
from dask.distributed import Client
client = Client()
client
s = da.ones((10, 1), chunks=(1, 1))
s1 = s + s.T
da.to_zarr(s1, url="my_zarr_files_computed") # see the dashboard, also files will be present in the directory
f = da.to_zarr(s1, url="my_zarr_files_not_yet_computed", compute=False) # no files in the directory
f.compute() # files will appear