You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
On a customer benchmark I am measuring read_csv is significantly slower with the default client rather than with the configured one.
read_csv: 13.72040319442749 # Client() # worker threads and processes take place
read_csv: 3.934723377227783 # Client(n_workers=16, threads_per_worker=1) # pure multiprocessing
We can set DaskThreadsPerWorker to 1 to get pure multiprocessing in Dask.
The text was updated successfully, but these errors were encountered:
On a customer benchmark I am measuring
read_csv
is significantly slower with the default client rather than with the configured one.We can set
DaskThreadsPerWorker
to 1 to get pure multiprocessing in Dask.The text was updated successfully, but these errors were encountered: